Citation
A multi-level study of vulnerability of Mongolian pastoralists to natural hazards and its consequences on individual and household well-being

Material Information

Title:
A multi-level study of vulnerability of Mongolian pastoralists to natural hazards and its consequences on individual and household well-being
Alternate Title:
A multilevel study of vulnerability of Mongolian pastoralists to natural hazards and its consequences on individual and household well-being
Alternate Title:
A multi level study of vulnerability of Mongolian pastoralists to natural hazards and its consequences on individual and household well-being
Uncontrolled:
Multilevel study of vulnerability of Mongolian pastoralists to natural hazards and its consequences on individual and household well-being
Creator:
Chuluundorj, Oyuntsetseg ( author )
Publication Date:
Language:
English
Physical Description:
1 online resource (xii, 255 leaves) : forms ;

Subjects

Subjects / Keywords:
Herders -- Economic conditions -- Mongolia ( lcsh )
Pastoral systems -- Mongolia ( lcsh )
Natural disasters -- Mongolia ( lcsh )
Adaptability (Psychology) -- Mongolia ( lcsh )
Adaptability (Psychology) ( fast )
Economic history ( fast )
Herders -- Economic conditions ( fast )
Natural disasters ( fast )
Pastoral systems ( fast )
Economic conditions -- Mongolia ( lcsh )
Mongolia ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Abstract:
"The transition to the free-market economy following the break-up of a socialist block in early 1990s totally transformed the pastoralist husbandry in Mongolia from state-supported collective farms to independent, subsistence-based herding households. The old socialist system provided a buffer against frequent cases of drought and winter storms. Plus, many negative social and economic consequences, including but not limited to livestock theft, market failure, increasing poverty and inequality, conflict over pasture and water sources has significantly reduced the coping capabilities of herders to resist to natural stress. This study employs a multi-level research design to explore the adaptive and coping strategies employed by rural herders in Mongolia to natural hazard events and the effectiveness of these strategies on their well-being measured in terms of economic and health status. The results of the study indicate that climate stress is not a strong predictor of lower socioeconomic well-being and poorer health outcomes. The role of factors such as gender, age, education and household size that make individuals and households more vulnerable to suffering from natural hazards and their consequences and of adaptive strategies to buffer the effects of natural disasters such as social capital and better herd management skills are crucial in the final outcomes" - Abstract.
Thesis:
Thesis (Ph.D.)--University of Colorado at Denver and Health Sciences Center, 2006.
Bibliography:
Includes bibliographical references (leaves 242-255).
Statement of Responsibility:
Oyuntsesteg [i.e. Oyuntsetseg] Chuluundorj.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
991872481 ( OCLC )
ocn991872481

Downloads

This item has the following downloads:


Full Text
A MULTI-LEVEL STUDY OF VULNERABILITY OF MONGOLIAN PASTORALISTS TO NATURAL HAZARDS AND ITS CONSEQUENCES ON INDIVIDUAL AND HOUSEHOLD WELL-BEING
by
Oyuntsesteg Chuluundorj
B.S., Mongolian National Medical University, 1996 M.A., University of Colorado at Denver, 2001
A thesis submitted to the
University of Colorado at Denver and Health Sciences Center in partial fulfillment of the requirements for the degree of Doctor of Philosophy Health and Behavioral Sciences
2006


This thesis for the Doctor of Philosophy degree by
Oyuntsetseg Chuluundorj has been approved by


Chuluundoij, Oyuntsetseg (Ph.D., Health and Behavioral Sciences)
A Multi-Level Study of Vulnerability of Mongolian Pastoralists to Natural Hazards and Its Consequences on Individual and Household Well-Being
Thesis directed by Professor Craig R. Janes
ABSTRACT
The transition to the ffee-market economy following the break-up of a socialist block in early 1990s totally transformed the pastoralist husbandry in Mongolia from state-supported collective farms to independent, subsistence-based herding households. The old socialist system provided a buffer against frequent cases of drought and winter storms. Plus, many negative social and economic consequences, including but not limited to livestock theft, market failure, increasing poverty and inequality, conflict over pasture and water sources has significantly reduced the coping capabilities of herders to resist to natural stress.
This study employs a multi-level research design to explore the adaptive and coping strategies employed by rural herders in Mongolia to natural hazard events and the effectiveness of these strategies on their well-being measured in terms of economic and health status. The results of the study indicate that climate stress is not a strong predictor of lower socioeconomic well-being and poorer health outcomes. The role of factors such as gender, age, education and household size that make individuals and households more vulnerable to suffering from


natural hazards and their consequences and of adaptive strategies to buffer the effects of natural disasters such as social capital and better herd management skills are crucial in the final outcomes.
This abstract accurately represents the content of the candidates thesis.
I recommend its publication.
Signed


ACKNOWLEDGEMENT
I would like to acknowledge Craig Janes, the chair of my thesis committee, and Kitty Corbett, Susan Dreisbach, Deborah Thomas and David Tracer, the members of my committee, for the outstanding support and guidance through this process.
My gratitude also goes to Sunmin Lee and Steve Sain for statistical advice.
My thanks due to my friends, Oyungerel Nanzad and Delgermaa Tsagaankhuu for much needed help in the data collection, and Solongo Altangerel for providing me with iron supplements.
I would like to thank the Public Entity Risk Institute for financial support that made the fieldtrips possible.


TABLE OF CONTENTS
Figures............................................. ix
Tables.............................................. x
CHAPTER
1. INTRODUCTION....................................... 1
2. BACKGROUND........................................... 6
Demographic and Socioeconomic Characteristics Mongolia.......................................... 6
Ecology and Climate.............................. 11
Natural Hazards.................................. 14
Pastoralism in Mongolia.......................... 16
Adaptation....................................... 21
Traditional Adaptive Strategies .......... 22
Non-Traditional Adaptive Strategies....... 26
Natural Hazards and Health....................... 31
3. THEORY.............................................. 35
Vulnerability Paradigm........................... 35
Natural Disaster.......................... 35
Vulnerability............................. 37
Theory of Political Ecology...................... 43
vi


Social Capital Theory.............................. 46
Definition of Social Capital................. 46
Social Capital and Economy................... 50
Social Capital and Health.................... 53
Gender..........................:.............. 55
Theoretical Framework.............................. 57
4. METHODOLOGY.......................................... 60
Research Plan ..................................... 61
A Spatial Ecological Study at the County Level. 61
Sampling..................................... 64
Variables.................................... 64
Analysis................................... 67
A Cross-Sectional Study at the Household Level. 68
Sampling..................................... 69
Interview Sites.............................. 73
Interview Setting............................ 77
Subject Payment.............................. 77
Household Questionnaire...................... 77
Hematological and Anthropometric Data Collection................................... 80
Variables.................................... 83
Analysis..................................... 92
5. RESULTS............................................... 94
Findings from County-Level Spatial Data Analysis... 94
vii


Descriptive Statistics................... 94
Conditional Autoregressive Models........ 96
Findings from Data Analysis of Household
Interviews..................................... 102
Descriptive Statistics.................. 102
Regional Differences.................... 115
Natural Disaster........................ 128
Socioeconomic Status.................... 129
Health Status.;......................... 165
Disaster and Health..................... 201
6. DUSCUSSION AND CONCLUSIONS....................... 203
Significance................................... 215
Limitations.................................... 217
APPENDIX................................................. 219
A. HOUSEHOLD QUESTIONNAIRE......................... 220
GLOSSARY................................................. 241
REFERENCES................................................. 242
viii


LIST OF FIGURES
Figure
2.1 Administrative Map of Mongolia.............................. 7
2.2 Contribution of Agriculture to GDP.......................... 8
2.3 Population Distribution..................................... 9
2.4 Ecological Zones........................................... 12
2.5 Livestock by Species, 1989-2005 ........................... 20
3.1 Theoretical Framework...................................... 59
4.1 Counties Selected for Interviews........................... 71
4.2 Interview Sites in Bayankhutag County...................... 73
4.3 Interview Sites in Olziit County........................... 74
4.4 Interview Sites in Khovd County............................ 75
4.5 Interview Sites in Bayan-Ondor County...................... 76
IX


LIST OF TABLES
Table
2.1 Dzud years and animal losses.................................. 15
2.2 Summary of adaptive strategies................................ 31
3.1 Definitions of vulnerability.................................. 37
4.1 Independent variables in the spatial ecological study......... 65
4.2 Dependent variables in the spatial ecological study........... 67
4.3 Sampling frame for household interviews....................... 70
4.4 Anemia cut-off levels......................................... 81
4.5 Sociodemographic variables.................................... 83
4.6 Socioeconomic variables....................................... 84
4.7 Social capital variables..................................... 85
4.8 Natural disaster variables.................................... 88
4.9 Herd management variables..................................... 89
4.10 Health variables .. .......................................... 90
5.1 Independent variables in the spatial ecological study......... 95
5.2 Dependent variables in the spatial ecological study........... 95
5.3 Crude mortality rate per 1000 persons......................... 97
5.4 Maternal mortality rate per 100,000 women of reproductive
age........................................................... 98
x


5.5 Under five mortality rate per 1000 livebirths................. 99
5.6 Crude morbidity rate per 1000 persons........................ 100
5.7 Demographic characteristics.................................. 102
5.8 Socioeconomic status......................................... 103
5.9 Social network............................................... 105
5.10 Trust and solidarity ........................................ 106
5.11 Collective action and cooperation............................ 107
5.12 Information and communication................................ 107
5.13 Social cohesion and inclusion................................ 108
5.14 Empowerment and political action............................. 109
5.15 Herd management.............................................. 110
5.16 Natural disasters............................................ Ill
5.17 Dzud........................................................ 112
5.18 Drought...................................................... 113
5.19 Continuous outcome variables................................. 114
5.20 Binary outcome variables..................................... 114
5.21 One-way ANOVA results: regional differences.................. 116
5.22 Post-hoc test results: regional differences.................. 120
5.23 Fixed and random part results: number of livestock in SFU ... 132
5.24 Fixed and random part results: number of milk livestock
in SFU....................................................... 138
5.25 Fixed and random part results: yearly household income in
thousand tugrics............................................. 142
5.26 Fixed and random part results: number of household items ... 146
xi


5.27 Fixed and random part results: number of transportation
items..............................'........................ 150
5.28 Fixed and random part results: large expenses in thousand
tugrics......................................................... 154
5.29 Fixed and random part results: monthly food expenses per
person in thousand tugrics...................................... 158
5.30 Fixed and random part results: meat (kg) consumed per
person per year................................................. 162
5.31 Fixed and random part results: hemoglobin in g/dl............. 167
5.32 Fixed and random part results: presence of anemia............. 173
5.33 Fixed and random part results: Body Mass Index................ 178
5.34 Fixed and random part results: overweight..................... 183
5.35 Fixed and random part results: triceps skinfold in mm..... 188
5.36 Fixed and random part results: abdominal skinfold in mm .... 193
5.37 Fixed and random part results: sick........................... 198
5.38 Logistic regression with dzud affects health
as the dependent variable....................................... 201
5.39 Logistic regression with drought affects health
as the dependent variable....................................... 202
xii


CHAPTER 1
INTRODUCTION
Beginning in 1990, Mongolia, a former client state of what was then the Soviet Union, undertook sweeping economic reforms. One important consequence of these reforms was the transformation of the rural, primarily pastoral, economy. Former state collective farms were dismantled and herding households were thrown into a highly insecure subsistence mode of production, and as a consequence, have become increasingly vulnerable to local fluctuations in rainfall, winter disasters, and the availability and quality of forage. Although Mongolian pastoralists have practiced, and continue to practice, a wide variety of adaptive strategies to manage risk, the economic transition has led to privatization and state disinvestment in the.rural infrastructure to such a degree that the institutional structures which had previously distributed risk among households by managing access to resources, providing access to essential commodities, and marketing animal products, have largely disappeared. Given the lack of formal social institutions which would protect production and subsistence under uncertain environmental and market circumstances, it is unlikely that traditional adaptive strategies alone are
1


sufficient to reduce risk. Some researchers suggest that various informal social arrangements that act as adaptive safety nets are emerging in some regions of Mongolia, yet the characteristics and success of such strategies are not well understood (Humphrey & Sneath, 1999).
In the absence of state support, households attempt to manage environmental risk in a number of ways. They diversify sources of income within the household (e.g., combining wage labor with herding activities). They undertake, with new intensity, the adaptive strategies many nomadic peoples use in unpredictable environments (e.g., management of herd size and mix, highly mobile foraging, creation of social bonds of reciprocity, absentee-herd owning). Some family members may move away for entire seasons to work in towns or cities. In some households, the elderly are sent to live in county and provincial centers so that they can support grandchildren sent to attend school and provide links to town-based services. The success or failure of these strategies has a direct effect on livelihoods of pastoralist herders.
Some policymakers argue that intensifying livestock production and reducing environmental risk by introducing a sedentary form of herding is the only solution to enhance the livelihood of Mongolian pastoralists (Hoohdoi, 2002), but the long-term environmental consequences of a settled herding can be devastating. In the case of Inner Mongolia, researchers found that reducing the mobility of herders by introducing technological innovation increased productivity per unit of land and labor in the short run, but eventually led to environmental degradation (Bates & Lees, 1996).
2


The objective of this research is to understand how environmental challenges interact with political and socioeconomic circumstances and the adaptive strategies employed by herders to affect the well-being of rural pastoral households. The research employs a combination of geographic, social, economic and anthropological methodologies (Borgerhoff-Mulder & Sellen, 1994). The main outcome variable, well-being, is measured multi -dimensionally at the county, household, and individual levels, and includes both socioeconomic status and general health outcome.
To meet the overall objective of the research, the following specific aims were developed:
Aim #1. To identify vulnerability to natural hazards at the county level by assessing the exposure to hazards, sensitivity (a potential of being affected by a climate stress) and adaptive/coping strategies using climate and county socioeconomic and demographic data.
Aim #2. To explore the relationships between vulnerability and health outcomes at the county level.
Aim #3. To study the impact of vulnerability to drought and dzud (a Mongolian term for winter disasters) on the economic well-being of rural households.
Aim #4. To investigate the relationship of vulnerability to drought and dzud and the health status of individuals within households.
The study hypothesis can be phrased as follows: high levels of vulnerability which is a composite measure of sensitivity, household adaptive
3


strategies, and exposure to natural hazards will be associated with low-levels of household well-being (economic and health).
The results of this study will help to identify factors that are most important in disaster mitigation among pastoralist households in Mongolia. It will assist in the implementation of programs to secure pastoral livelihoods. Ideally, these programs will enhance the general health status of pastoralist households. The use of different levels of analysis (county and household) in the research will give a more comprehensive picture of vulnerability in Mongolia and a better understanding of the processes that shape vulnerability at multiple spatial scales.
The dissertation is divided into seven chapters. Chapter 2 provides background information on Mongolias demographic and socioeconomic situation during the transition period, ecological and climate peculiarities, pastoralism and adaptive strategies to overcome constraints of a physical environment, impact of natural disasters on health, and application of GIS and spatial statistics in health research.
Chapter 3 develops the theoretical principles that guided the research, including the vulnerability paradigm and theories of political ecology, social capital and gender.
Chapter 4 describes the methodology of the research. Predictor and outcome variables used for each level of the research, the sampling frame, study sites and settings, and data collection procedures are described in this section.
4


Chapter 5 presents the results of the data analysis: spatial data analysis at the county level and multilevel data analysis at the household and individual levels.
The discussion of results, conclusions and final recommendations follow in Chapter 6.
5


CHAPTER 2
BACKGROUND
Demographic and Socioeconomic Characteristics of Mongolia
Mongolia is located in Central Asia between Russia and China. The territory is 1.567 million square kilometers, with a population of only 2.7 million people, which results in one of the worlds lowest population densities1.5 persons per square kilometer. Fifty two percent of the population are urban-dwellers, the majority of which live in the capital of Ulaanbaatar. Of the 48% who are rural residents, the majority are nomadic pastoralists.
The country is divided into 18 rural provinces and four cities. Each province is divided into 15-16 counties on average, and each county is divided into small townships with an average population of 500-600 persons (Figure 2.1).
6


Figure 2.1. Administrative Map of Mongolia
After seventy years of state-controlled, centrally-planned economy, in 1990 Mongolia undertook political and economic reforms in transition to a market-economy. These reforms consisted of the following elements: price liberalization, removal of restrictions on international trade and foreign investment, privatization of state-owned enterprises, and a marked reduction in governments involvement in the economy. By most reports, these reforms resulted in widespread social chaos and economic collapse (Griffin, 2001).
Aid to Mongolia from the former Soviet Union and other socialist countries was immediately eliminated, contributing to a sudden 30% drop in GDP. Technical assistance was withdrawn, the Council for Mutual Economic Assistance that had regulated trade between countries in the former Soviet
7


block collapsed, and Mongolia lost import and export alliances. Industrial production, which had up to 1989 been a growing sector of the economy, declined sharply, resulting in widespread unemployment and poverty. As a result, the value of industrial production as a percentage of GDP declined from 35% to 20% between 1990 and 2000 (UNDP, 2001). The contribution of agriculture to GDP, mainly-livestock husbandry, increased accordingly as seen in Figure 2.2.
Figure 2.2. Contribution of Agriculture to GDP (Griffin, 2001)
The shock of poverty and high levels of unemployment in urban areas led to widespread migration from the cities to the countryside as individuals sought to benefit from livestock privatization. Griffin called this process ruralization. As the result, between 1989 and 1998 the proportion of the population living in urban areas declined by 13%, and the rural population grew by 17% (Griffin, 2001). Mongolia became once again a predominantly rural
8


country. Since 1998, however, due to a string of severe winters which may have led to the failure of many inexperienced herders, along with a stabilization of the urban economy, population has begun to return to the towns and cities (see Figure 2.3). The western provinces experienced the largest levels of emigration from 1995 to 2000, eighty eight thousand and six hundred persons, and eastern provinces the smallest, twenty nine thousand and nine hundred persons (Bum, 2003).
Figure 2.3. Population Distribution (NSOM, 2004).
Liberal economic reform radically transformed the structure and organization of the rural economy. The results have been a rapid dismantling of the old collective institutions-negdel, privatization of all livestock and assets, disinvestment in social services, and poor access to markets (Femandez-Gimenez, 1998; UNDP, 2001). The collapse of the negdels and the privatization of herds have eliminated the formal institutions that buffered
9


economic and ecologic risk and protected the environment by regulating access to and maintaining land, pasturage and water resources; providing veterinary services; and establishing reserves of food, hay, warm clothes and medical supplies for use in emergencies (Baas, Batjargal, & Swift, 2001; Griffin, 2001; UNDP, 2001).
Indicators of social and health services also show significant deterioration. Maternal mortality rates have been consistently high over the last ten years (about 200 deaths/100,000 livebirths compared to 120 deaths/100,000 livebirths in 1989). The majority of deaths occur in rural counties and these women were most likely to be herdswomen, those with less education, or ones from poor families with little social support (Chuluundorj, 2001; Janes & Chuluundorj, 2004). The incidences of infectious diseases such as tuberculosis, sexually transmitted infections, infectious hepatitis (esp. hepatitis B virus and hepatitis C virus), and brucellosis have grown disproportionately in the poor segment of the population in both rural and urban areas (Government of Mongolia & UNDP, 2004). There has also been a rapid increase of non-communicable diseases such as cancer, cardiovascular diseases and injuries (Foggin, Farkas, Shirev-Adiya, & Chinbat, 1997).
Although the above mentioned socioeconomic changes occurred everywhere in the country, different regions and provinces faced unique challenges depending on environmental conditions, local economic resources, access to domestic and foreign (e.g., Russia, China) markets, and infrastructure development. Consequently, the economic rewards of pastoralism vary greatly
10


from one region to another (Humphrey & Sneath, 1999). Concomitantly, disparities also exist in the health status of a population in different geographic regions: e.g. the maternal mortality rate is consistently highest in the economically disadvantaged western provinces (Demberelsuren & Dorjpurev, 2000). However, little is understood about the impact of ecological differences between different regions of the country and their impact on a local populations socioeconomic well-being and health. This is one question addressed in this study.
Ecology and Climate
There are three different ecological zones within the territory of Mongolia: the forest zone with high annual standing vegetation (3500 to 4000 kg/ha), the grass steppe or grassland zone with a moderate annual standing vegetation (1500 to 3000 kg/ha), and the desert zone with low annual standing vegetation (375 to 1500 kg/ha). Grassland is the transitional ecosystem between forest and desert ecosystems: it occurs where the precipitation is too low to maintain forests, but high enough to prevent desertification (Government of Mongolia, 2001; Moran, 2000). A map of ecological zones is shown in Figure 2.4.
11


Figure 2.4. Ecological Zones
Climatic conditions fluctuate greatly from one ecological zone to another. The precipitation varies among ecological zones and is confined mainly to summer: 100 mm/year in southern Gobi areas and up to 500 mm/year in northern mountainous forest regions. The majority of the country receives on average 350 mm of precipitation per year (Government of Mongolia, 2001). Summer rainfall has declined between 1970-1990 in the Gobi, and the number of heavy rainfall events has fallen significantly (Lai, Harasawa, & Muriyarso, 2001). Droughts have been more frequent in the 20th century, and may be linked to the climatic consequences of global warming.
12


The land is covered by snow for seven-to-eight months per year, which results in a growing season of only four months. Temperatures can easily reach -40C in mountainous region during winter. The annual mean surface temperature in Mongolia has increased by 0.7 degrees Celsius over the past 50 years (Lai et al., 2001).The difference between night and day temperatures is high, frequently reaching 20-30C. The combination of low precipitation and high temperature fluctuation make agricultural production risky: only one percent of total land in Mongolia is classified as arable (UNDP, 2002).
Given these ecological and climatic constraints, pastoralism is the most efficient means of production in Mongolia (Moran, 2000). Because the amount of energy that can be extracted via animal herding from a unit of grassland is small, pastoralists require high mobility and access to a large territory of grazing land in order to reduce the risk from droughts and unstable weather conditions. These demands vary across the country depending on local and regional ecological and climatic conditions (Bates & Lees, 1996).
In the future, global warming is likely to cause ecological and climate changes: shift the boundaries of ecological zones, reduce land productivity and biodiversity, and raise the occurrence of pests and diseases in these areas. Increased temperatures with the combination of low precipitation is likely to cause severe water shortage and reduce land productivity by 40-90% (Lai et al., 2001). This may eventually place severe constraints on pasture availability and negatively affect animal breeding and wool/cashmere productivity. Continuing land degradation could also have profoundly negative impacts on the well-
13


being of a large number of rural people whose livelihoods are highly dependent on environmental resources (Corvalan et ah, 2005).
Natural Hazards
Mongolian pastoralists face several environmental hazards (source of danger that may potentially harm individuals or groups): winter disaster or dzud, drought, wild fire, predation, damage caused to vegetation by large rodents, and animal diseases (Baas et ah, 2001; Haddow & Bullock, 2006; Pelling, 2003). Among these, droughts and dzuds are the greatest threat to subsistence (Templer, Swift, & Payne, 1993).
A dzud can be defined in many ways, meteorologists define a dzud as snow cover of more than 25 cms, a sudden prolonged snow storm, or prolonged extreme cold. Mongolian herders typically distinguish between several kinds of dzuds. A white dzud occurs when deep snow cover prevents animals from grazing. A storm dzud occurs when severe weather conditions drive animals downwind, and they get lost and freeze to death. Freezing dzuds occur when temperatures fall sharply and animals cannot maintain their body temperature (Morinaga, Tian, & Shinoda, 2003; Swift, 1999). Regardless of the specifics of definitions, any of these events is a real threat to livestock loss. A sequence of summer drought and winter dzud is the most devastating because animals who do not gain sufficient weight during the summer easily perish even in mild winter conditions.
14


Drought is a period of several months or even years of abnormal dryness due to below-average rainfall that causes a pronounced decrease in forage yield relative to what is expected in an average year (Rothauge, 1998, p. 1).
Drought is a slow-onset hazard and can affect large areas. In fact, the dzuds with the greatest animal loss were all preceded by summer droughts: the dzuds of 1944-1945,1967-1968, and 2001-2002 (Table 2.1).
Table 2.1. Dzud years and animal losses (Hoohdoi, 2002).
Year Affected Places Animal loss in thousand heads Animal loss as % of total livestock
1944-1945 9 provinces, 65% of total land 8638.0 35.5
1954-1955 No information available 1887.7 8.2
1956-1957 No information available 1008.0 4.1
1967-1968 13 provinces, 80% of total land 3800.0 17.0
1976-1977 15 provinces, 90% of total land 1453.9 6.1
1993 spring 3 provinces, 30 soums 689.5 2.7
1996-1997 11 provinces, 69 soums 700.0 2.4
1999-2000 13 provinces, 158 soums, 70% of land 2614.0 10.0
2000-2001 17 provinces, 98 soums 4800.0 18.2
2001-2002 20 provinces 3400.0 9.5
15


Some actions have been taken during dzuds by the Mongolian government to help households in dzud-affected areas. These include the distribution of fodder, vegetable seeds to encourage agriculture, dietary diversification and greater food supplies, and support of the health care system with equipment and drugs (Norovlin et ah, 2003). These efforts seem insufficient to reduce herd loss and prevent adverse health outcomes in the pastoralists. One study conducted in Mongolia revealed that the prevalence of growth stunting was significantly greater among children aged 6-23 months in dzud-affected areas than in unaffected areas: 38.3% versus 26.0%, p=0.04 (Norovlin et al., 2003).
The effect of such climatic stress on the socioeconomic conditions of pastoralists is tremendous. Loss of livestock directly translates to loss of financial security, marginalization and impoverishment. Importantly, the loss of livestock may also lead to loss of social ties and the support that flows through these channels. Pastoral herding requires a high reciprocity and families with little or no means to engage in mutually beneficial relationships may be left out. Families may also move to more urban areas to make a living and become detached from their kin- and place-based social networks. These families comprise a large percentage of the urban poor.
The essential pastoral strategy is probably neither maximization nor optimization but risk aversion, an attempt to decrease uncertainty by anticipation (Galaty & Johnson, 1990). As the quote states, avoiding, anticipating and coping with these negative climatic stresses or their
16


consequences is the primary strategy of pastoralists. Understanding the importance of these strategies and developing policies to enhance the capacity of pastoralists gain more importance: the occurrence of extreme climate events is likely to increase in the future and bring threats to pastoralist livelihood more than ever (Corvalan et al., 2005). Specific adaptive and coping strategies will be delineated later, but before that an introduction to pastoral livelihood system is necessary.
Pastoralism in Mongolia
Changes that occurred in the pastoralist system in 20th century in Mongolia are quite remarkable. In the early 20th century, livestock was the property of a landed aristocracy: local lords and monasteries. The majority of the population were poor and subject to exploitation by the elite. The coordination of access to pasture and seasonal migration was done effectively by the aristocracy to avoid overgrazing and secure water sources.
Beginning in the 1940s, the process of appropriation of private animals by the state began. The process was completed in early 1960s: the state firmly established its control over all resources, including all animals, through organizing collectives and state farms. The state organized the livestock production system down to the household level and controlled the use of natural pastures with a goal to achieve a maximum use of natural resources without overusing them. Herders became salaried employees of the state and the state
17


provided food items at subsidized prices, health and social services free of charge, free education to children, animal shelter, veterinary services, hay, and extra labor when needed.
Politico-economic processes occurring during the transition that began in 1990 brought new challenges into livestock production in Mongolia. Livestock was given back to people, creating small household-level production units. The privatization of livestock resulted in an increase to the pastoralist population: many urban dwellers moved to the countryside to benefit from receiving livestock shares. In 2000, 191,526 households (about 40% of total population) made their living primarily from pastoralism compared to 74,710 in 1990 (Government of Mongolia & UNDP, 2004). The importance of livestock production to Mongolias economy increased over this time period: it accounted for 83% of total agricultural production in 2003 (Meams, 2004).
Although livestock has been privatized, land is still common property. Having common land is a necessary condition when the productivity of pasture is highly variable. It allows herders access to large territories to reduce risks of inclement weather (Gilles & Gefii, 1990). However, the emergence of many small-scale private herd owners has decreased rotational access to pastures.
This is due to increased competition over the best pastures in the absence of supra-household regulatory mechanisms. Without institutions to govern shared access to better pastures, overgrazing and poor access to good pasture and water source have become a real problem. Poorer households or households with fewer or weaker social ties are cut off from best resources (UNDP, 2002).
18


As more and more people have become involved in pastoral production, several trends have been identified:
1. The number of total livestock has increased in the last decade (Figures 2.5), but the majority of herding households have relatively few animals. The average number of livestock per household declined from 346 in 1990 to 158 in 2000. The percentage of households with fewer than 100 animals was 58% in 1999, 63% in 2000, 67% in 2001, and 69% in 2002. In 2002, 88% of all herding households had fewer than 200 animals (Ravsal,
2003). A recent study by the Ministry of Finance and Economy on herd restocking strategies has concluded that the number of livestock for a reasonable living for an average family of 4-5 people would range from 200 to 300 animals (Meams, 2004). Thus, there has been an increase in poor herding households over the last decade. At the same time there is an emerging group of wealthy herders in the country who owned more than 2000 animals (Government of Mongolia & UNDP, 2004).
2. There is a change in the structure of herds. The total number of sheep remained constant throughout the decade, but the number of goats increased by over 200%. The percentage of goats in the national livestock population increased from 20% in 1990 to 42% in 2003, whereas the percentage of other livestock declined by 1-16% (Figure 2.5). The increase in the number of goats is likely due to the market value of cashmere, the most valuable of the animal products traded by rural herders.
19


16,000,000 -|
14.000. 000
12.000. 000
10,000,000
8,000,000
6,000,000
4.000. 000
2.000. 000
Figure 2.5. Livestock by Species, 1989-2005 (NSOM)
The declining numbers of camels, horses and cattle in the total national livestock population is attributed by herders and livestock experts to the use of mechanized transportation rather than camels, the slaughter of camels due to the increased price of camel meat (180-200$ per camel), an increased price for horse hides (20$), theft of horses in far pastures, and a loss of horses and cattle to dzud and drought. Maintaining the right ratio of different species in a herd is extremely important to effective pasture use, and experienced herders attribute pasture degradation to too many goats (Grayson & Baatarjav, 2004).
3. Disappearance of economic/market services in rural areas contributed to market failure in the countryside and overgrazing of pastures close to cities and province centers. Herders move to be closer to cities and towns where they have opportunities to sell their products, resulting in higher livestock-to-pasture ratios in these regions and causing pasture degradation.
20


Thus, economic liberalization has enabled some individuals and households to increase their profits from livestock significantly, yet this has produced dramatic increases in inequality between the richest and the poorest segments of the rural population as some are more able than others to respond to increased market opportunity and climate stress, and significant pressure on natural recourses. Households with a few animals, young and inexperienced herders, single-headed households, households with insufficient labor (high dependency ratios), households who entered herding production to benefit from privatization and are new to herding are particularly at higher risk of impoverishment and social marginalization (Cooper, 1993).
Adaptation
The chief problems presented by grassland ecosystems center around the exploitation of water and pasture, the composition and size of herds, the establishment of workable relationships between pastoralists and agriculturalists, and the achievement of a balance between the human and the animal population under conditions of great climatic uncertainty (Moran, 2000, p. 220).
The many adaptive strategies used by Mongolian herders focus on avoiding hazardous natural events, maximizing the use of natural resources, increasing their income, and securing their livelihoods. The adaptive strategies
21


delineated below are divided into two groups: traditional and non-traditional or emerging. The traditional adaptive strategies have been practiced for centuries and are common among pastoralist populations in the world. The non-traditional adaptive strategies have emerged recently in Mongolia as a response to socioeconomic and political changes that have occurred over the last 15 years.
Traditional Adaptive Strategies
Low Livestock Density/Low Settlement Density. Low livestock and human densities are crucial for successful herding. Having a larger area to graze during different seasons ensures availability of good pasture throughout the year. Mongolia has one of the lowest population densities in the world-1.5 persons per square kilometers. Seventy nine percent of Mongolias total land (1567 million square kms) is under pasture (Fratkin, 1997). However, since privatization herding households have moved to central provinces where the market opportunities are better, resulting in higher population and livestock densities in those areas.
High Mobility. Low livestock and human density is the condition for high mobility. Pastoralists move every season. In fact, there are several different types of movements practiced by pastoralists. These include seasonal migration moving between ecological zones to benefit from different stages of
22


vegetation cycle; middle range move between two pasturelands; and short range camp site transfer movement within one pasture (Schareika, 2003).
The high mobility of herding households has decreased with the privatization of livestock mostly due to unavailability of transportation means, lack of labor and unavailability of good quality pastures (Femandez-Gimenez, 1998; Foster, 2003). Decline in herding mobility leads to an increasing conflict over pasture usage and water sources. Increased pressures on pasture eventually will result in overgrazing and pasture degradation. Intensive grazing of 2-3 periods on the same area in a season may lead to a decline in pasture yield of up to 72% the following year (Tserendash & Erdenebaatar, 1993). According to a recent study, 1.4 % of total pasture land in Mongolia is degraded very seriously (50% or more), 20.7%seriously (30% to 49%), 50.8%mildly (20% to 29%), 25.4%minimally (10-19%) and 1.7%not degraded (less than 10%). The problem of pasture degradation is especially serious in central provinces where the livestock and human density is much higher (Marriott & Erdene-Ochir, 2004).
The importance of transportation in mobility is enormous. Wealthier families have been able to acquire motorized transportation which has enabled them to transfer their livestock in case of natural disasters; gives them access to markets for purchasing food and products and selling animal products; and allows them to become involved in alternative economic activities to supplement income from pastoral production.
23


Large Number of Livestock. It is the primary goal of pastoralists to increase the number of their livestock. The Mongolian climate is defined as a non-equilibrium system: it is characterized by a great fluctuation of livestock numbers as a result of unpredictable and uncontrollable natural hazards that periodically decimate herds. Livestock numbers increase rapidly for some period of time after a disaster, then declines sharply as a result of a natural disaster. Having a larger herd means better security in times of disasters. Wealthier households may lose more animals, but they will also have more animals left to rebuild a herd. But poor families are likely to be left with number of livestock below the minimum subsistence level and may not be able to recover from such losses in the longer term. In addition, livestock defines wealth and wealthier households are able to procure access to better pastures and water sources, especially if wells need to be dug and maintained, and enjoy greater leisure time and overall higher prestige (Begzsuren, Ellis, Ojima, Coughenour, & Chuluun, 2004; Goldstein & Beall, 2002; Mulder & Sellen, 1994).
Increased Number of Species in a Herd and Breed Selection. Multispecies herds have many advantages: better use of pasture due to different foraging habits of different species, different species provide diversified products for consumption and sale, different mortality of different species ensures that some animals are left after natural disasters (cattle and horses
24


perish first, goats are the last to die in dzud). Purchase of highly adaptive breeds may also reduce the loss during disasters.
Household Splitting. In times of droughts or winter disasters families split their herd for some period of time to benefit from better pastures and favorable climatic conditions. This is an option for families who have enough people to split and live separately for a few months (Meams, 2004).
Large Family Size. Pastoralists tend to have larger families compared to urban dwellers. Having a large family guarantees labor availability, more social ties and more resources flowing through them. In Mongolia, the total fertility rate was 3.66 in rural areas compared to 2.46 in urban areas according to the Reproductive Health Survey of 1998 (NSOM & UNFPA, 1999).
Large Social Network. Having a large social network is an important asset to the herding household. Households cooperate with others on many of their difficult tasks, including moving, sheering, combing, cutting hay, and fetching water and fire wood. Hospitality and reciprocity facilitate mobility and help herders gather needed information. Creating alliances across different ecological zones is also crucial to secure better pastures in times of droughts and dzuds (Galaty & Johnson, 1990). Waller and Sobania (1994) noted that in Africa asociality was equal to poverty. Having a large social network may not be as important to wealthier households as it is to poorer ones. Cooper (1993)
25


recognized the importance of expanding social ties to sustain the labor and consumption requirements for the poor. Wealthier households can purchase labor or services whenever they need them, but the poor rely on others generosity for these services.
Large Percentage of Female Animals. Female animals are the foundation for building assets for pastoralists. In addition, they provide milk that is a necessary part of the diet for herders (Chen, 1991).
Non-Traditional Adaptive Strategies
Increased Cash Income. One of the changes that occurred with the introduction of the market economy was an increased value of cash for rural herders. Sixty .two percent of income of herding households comes from the sale of wool and cashmere, and the rest from a sale of livestock (Government of Mongolia & UNDP, 2004). Converting surplus livestock products to cash means access to health care and other services when needed, ability to purchase transportation to enhance mobility, better supply of cereals and other food items that are not produced by herders, and good education for their children. Substitution of hand-made products by manufactured products further deepens the dependency on markets and cash. Market opportunities to sell livestock are not the same across the country. Herders closer to cities, borders and trade
26


centers benefit from easier access to markets and higher prices. Herders in remote locations must often sell their products to middlemen at lower prices.
Livelihood Diversification. Pastoralists engage in various activities besides livestock husbandry to increase their income and support their subsistence. Some of these activities are hunting, begging, farming of vegetables, gathering wild foods, sewing, knitting, and animal theft. Theft of larger animals such as camels, horses and cattle is common because these animals graze farther from the camp and are not usually tended by herders on their pastures. Diversification may not be a risk-averting strategy for the poor: mainly they do it for survival. But middle-wealth and rich families do it to minimize risk or accumulate more wealth (Little, Smith, Cellarius, Coppock, & Barrett, 2001).
Cooperation through Informal Camps and Neighborhood Groupings (Khot AilL In the absence of institutional support, herding households form small groupings to share labor. They each take turns to tend animals on pastures, milk animals, process dairy, or go to markets to sell or trade. If a household has to go to the province center to seek health care, other households can take care of their children and animals. Often these groupings are based on kinship relationships, but friendship may also provide a basis for affiliation.
27


Livestock Insurance. There are only two companies that offer livestock insurance at the current moment: Mongol Daatgal company and one other state-owned company. Livestock loss is covered at 100% but the premiums are set at six percent (of the cash value of livestock) which is expensive for pastoralists. There is no regional variability in the premiums. But the current insurance was not marketed aggressively and herders knowledge how insurance works is very rudimentary. Other forms of livestock insurance are being discussed, including weather-based and mortality-based insurances. The feasibility of these schemes remain to be tested but these are likely to offer more benefits than traditional livestock insurance (Skees & Enkh-Amgalan, 2002).
Involvement in Wage Labor. Often families with many adult members send one or two persons to province centers or cities to secure permanent jobs. Having a job means cash income. It also creates rural-urban linkages to promote exchange of goods and services.
Patron-Client Relationships. These are linkages between better-off and poor households, with little or no reciprocity. This type of relationship is profitable to wealthier families: they benefit from extra labor the client can provide. In return, the patron grants a small salary and free dairy products and meat to the client. It may be helpful to poorer families who otherwise would have moved to urban areas where their livelihood may be even worse.
However, this patron-client relationship does not offer any opportunities to
28


clients to build their own livestock herds and secure their livelihoods. They remain highly dependent on the willingness and ability of their patrons to support them.
Access to Credit. Having access to any of type of formal (banks, in-kind through restocking) and informal (relatives, friends, kiosks, pawnbrokers, money lenders including cashmere traders) credit is important, especially for poorer households who do not have or have very little cash income. Households who have lost all of their livestock to natural disasters may receive animals from their relatives and friends. If there is an urgent need to borrow money to cover, for example, medical expenses, having a reliable source for loans may make all the difference.
Animal restocking programs have been introduced in five provinces (Dundgobi, Bayankhongor, Zavkhan, Ovorkhangai and Uvs) as an experiment. Households are given cash in the amount of one million tugrics (an equivalent of less than 1000 USD) for five years (with an interest of six percent paid starting the third year) and they purchase animals from local herders. Criteria for receiving such grants include: loss of an entire herd or being left with a few animals not sufficient for subsistence; possession of a permanent winter shelter; a multi-generational herder, no criminal history, and agreement of all household members to the conditions of the award. Households new to herding or young and inexperienced herders are not able to benefit from this program (Hoohdoi, 2002).
29


Reducing Consumption. Other strategies, such as reducing consumption or switching to cheaper inferior food, have also been observed (Chen, 1991; Fratkin & Roth, 1996; Little, 2002). Siurua and Swift (2002) have found that reducing consumption during disasters was a common approach among pastoralists in affected areas.
Migration to Urban Areas. Population migration from cities to rural areas from 1990 to 1995 was reversed in the second half of the 1990s. Families who lost all of their livestock have moved to province centers or cities to seek employment. According to the statistics from the National Statistical Office of Mongolia, one third of the total population in Ulaanbaatar is migrants from rural areas (UNDP, 2003). This is the most extreme form of adaptation, which often worsens socioeconomic conditions of households and the health of individuals.
Absentee-FIerding. Some pastoralist households herd animals of other households (absentee-herders) who live in urban areas. It can be useful to both sides: the herding household may receive services from the absentee-owners in return for tending their animals. For absentee-herders, having their own herd on pastures means availability of dairy and meat year around (Femandez-Gimenez, 1999).
In summary, Mongolian pastoralists engage in various activities to buffer their economic loss during disasters and secure their subsistence. Little is
30


known about the relative importance of these strategies on households level of well-being. This is the major question addressed in this thesis.
Table 2.2. Summary of adaptive strategies
Traditional
Non-Traditional
Low livestock density/Low settlement
density
High mobility
Increased number of species in a herd and breed selection Large number of livestock
Household splitting
Large family size
Large social network
Large percentage of female animals
Increased cash income
Livelihood diversification
Cooperation through informal camps and neighborhood groupings Livestock insurance
Involvement in wage labor
Patron-client relationships
Access to credits
Reducing consumption
Migration to urban areas
Absentee-herding
Natural Hazards and Health
The number of natural hazards, such as floods, droughts, wildfires, winter storms and extreme heat, is likely to increase in the future as the result of an increased accumulation of greenhouse gases in the lower atmosphere and stratospheric ozone depletion, and these natural events pose a serious threat to the human livelihood (Burton, Kates, & White, 1993; Haddow & Bullock,
31


2006). The mechanisms through which these natural hazards affect health vary depending on nature of the events.
Flood is the most widespread and the most costly natural hazard. The direct damages include drowning and injury of humans, food insecurity due to the devastation of animals, crops and natural resources, and disruption of roads to transport people to safety supplies. The indirect damage includes outbreaks of waterborne infections arising from the contamination of drinking water by sewage and vector-borne infections such as malaria because of increased reproduction of mosquitoes in stagnant water, and spread of harmful chemicals such as pesticides (Blaikie, Cannon, Davis, & Wisner, 1994; Hewitt, 1997).
Reduced vegetation growth and death of livestock from dehydration during droughts endangers the food security of populations. Reduced flow of water in streams, lakes and wells are likely to put limitations on water usage and worsen sanitary conditions. Poor water quality, changes in its salinity, and accumulation of toxic substances may eventually have a major impact on the health of human populations (Hewitt, 1997).
The impact of dzuds on human health can also be direct and indirect. The direct effect includes hypothermia, frostbite, and injuries due to motor vehicle collisions or collapses of structures (Hewitt, 1997). The indirect effects such as poor nutrition due to high mortality of livestock and inaccessibility of roads may impact health slowly.
In any natural hazard a variety of problems may arise as the result of displacement in case of any natural hazard such as housing in unsafe places,
32


unsafe water and sanitary conditions, poor mental health, and poor nutrition (Hewitt, 1997; Kovats & Bouma, 2002). The significance of these natural events is growing: the number of people killed, injured and thrown to poverty due to natural hazards is swelling (Kovats, Menne, McMichael, Corvalan, & Bertollini, 2000).
The relationship between hazards and certain health outcomes are not as straightforward. It is attributed to a multiple causality of diseases, diversity of diseases, prolonged effects of hazards, and great variability in the manifestation of natural events (Kovats et al., 2000).
In addition, hazards do not affect everyone in the same manner. Individuals, households, regional and/or national capacities to avoid, reduce or cope with environmental stress or its consequences buffer negative effects of hazardous events on human health and well-being. Those who have the fewest-resources to avoid natural events or their negative impacts are at most risk of becoming impoverished and sick. The loss of human life and loss in property are the highest in developing nations (Lai et al., 2001).
The health of children is especially sensitive to natural hazards. They are a biologically vulnerable group and suffer from those diseases likely to be caused or exacerbated by reduced income and/or reduced public medical care. Nutritional deficiency, which can be caused by a deficiency in vitamin, mineral, and the energy/protein content of food, perhaps the most commonly used health outcome in studies exploring the relationships between major environmental
33


events and their health consequences (Kormondy & Brown, 1998; Musgrove, 1987).
In summary, Chapter 2 described main socioeconomic and sociodemographic processes that have occurred in Mongolia during the transition period, changes in pastoral livelihoods as the consequence of these processes, strategies employed by Mongolian pastoralists to overcome ecological, political and socioeconomic challenges, and the effectiveness of these strategies to buffer negative environmental impacts. The following chapter introduces the theoretical framework and main theories employed in this study.
34


CHAPTER 3
THEORY
Vulnerability Paradigm
Natural Disaster
The impacts of natural hazards are likely to be felt more severely in developing countries than in developed countries, irrespective of the magnitude of climate change, because of the poor resource and infrastructure bases found in such countries. In countries where weather-sensitive production systems such as crop/vegetable growing and livestock husbandry contribute significantly to the national economy, natural hazards are major threats to the livelihoods of people, particularly of those who are poor. In Mongolia, where livestock husbandry contributes 20 percent to the GDP and is the main source of subsistence for rural Mongolians (constituting 40 percent of the total population); the impacts of natural events such as winter storms and droughts are dramatic and potentially devastating (NSOM, 2004).
35


Winter storms and droughts are not new phenomena to experienced Mongolian pastoralists. But their impact on human livelihoods and health has changed dramatically during the transition from a state-oriented to a market-oriented economy that began in the early 1990s. Under the socialist system, livestock was a state property and herders were hired employees of the state. The latter provided salary, health and social services. During the privatization of the early 1990s, all livestock was divided among herding households. Each household received approximately 50-60 animals on which they had to rely on for subsistence. A summer drought and a winter storm became a real threat: one winter storm could swipe away all animals and leave families with nothing.
This experience of rural herders in Mongolia is a classic example of how natural disasters are bom. Blaikie et al. (1994) defined a natural disaster as the sum of a hazard or natural event and vulnerability. A disaster occurs when a significant number of vulnerable people experience a hazard and suffer severe damage and/or disruption of their livelihood system in such a way that recovery is unlikely without external aid. The disruptions may occur on multiple scales, from individuals health and mental distress to local and/or national socioeconomic downfall and failures in the global political economy (Pelling, 2003). Natural events do not cause natural disasters. They are a necessary part of natural disasters, but they most often act as triggers of underlying social inequality and social forces, or in other words, a disaster is a consequence of vulnerability (Pelling, 2003). Vulnerability is a social process.
36


Vulnerability
The term vulnerability has its origins in food security literature, and has been applied in the last decade or so to research on climate change and its effects (Vincent, 2004). The concept is very much in its initial stage of formulation and there is no all-agreed definition of vulnerability. However, the main elements in the definition of vulnerability have evolved as seen in Table 3.1.
Table 3.1. Definitions of vulnerability (Weichselgartner, 2001)
Authors Definition
Gabor and Griffith (1980) The threat (to hazardous materials) to which people are exposed (including chemical agents and the ecological situation of the communities and their level of emergency preparedness)
Susman et al. (1983) The degree to which different classes of society are differentially at risk
Smith (1992) Human sensitivity to environmental hazards represents a combination of physical exposure and human vulnerabilitythe breadth of social and economic tolerance available a the same time
Cutter (1993) The likelihood that an individual or group will be exposed to and adversely affected by a hazard
Watts and Bohle. (1993) Vulnerability is best defined as an aggregate measure of human welfare that integrates environmental, social, economic and political exposure to a range of potential harmful perturbations. Vulnerability is a multilayered and multidimensional social space defined by the determinate, political, economic and institutional capabilities of people in specific places at specific times.
37


Table 3.1. Definitions of vulnerability (Cont.)
Authors Definition
Blaikie et The characteristics of a person or group in terms of their
al. (1994) capacity to anticipate, cope with, resist and recover from the
impact of a natural hazard
IPCC The degree to which a system is susceptible to, or unable to
(2001) cope with, adverse effects of climate change, including climate variability and extremes. Vulnerability is a function of the character, magnitude, and rate of climate variation to which a system is exposed, its sensitivity, and its adaptive capacity.
The development of theories of natural disasters has its roots in the early 1960s with Burton and Kates' work. They described natural disasters as environmental events that can cause harm to humans (Pelling, 2003). Vulnerability has been treated as an increased exposure to natural hazards: living in zones where a hazard is likely to occur. This construct is known as biophysical vulnerability. The frequency, intensity, duration of natural events and the extent of damage were the main focus of studies that use the definition of biophysical vulnerability. In this literature, reducing an exposure to a hazard by introducing technological innovations becomes the key to preventing or mitigating disasters. OBrien named the construct end-point vulnerability (Adger, Brooks, Bentham, Agnew, & Eriksen, 2004; Gilbert, 1995; O'Brien, Eriksen, Schjolden, & Nygaard, 2004).
White and Haas (1975) postulated that people continue to live in regions
which they know to be hazardous hoping that the benefits of occupying the area
will outweigh potential losses. Individuals were seen as active agents who can
38


make choices about potential risks of disaster. But the environment was still seen as the major force that shapes the outcome of a disaster.
This environmental deterministic view dominated the field until 1983, when Hewitt published his seminal work, "Interpretations of calamity: from the viewpoint of human ecology" in which he described natural disasters as an outcome of the society-nature interaction (Pelling, 2003). Hewitt argued that human responses to natural hazards are not dependent on the natural events, but are shaped by the social order, institutions and historical circumstances present in the society (Tobin & Montz, 1997).
Following Hewitts work, researchers started to use vulnerability in a different sense, and people's capacities to avoid, resist and recover from hazards became an essential part of vulnerability. In this literature, vulnerability is seen as a state that exists within a system before it encounters a hazard event and is exacerbated by it (Brooks, 2003). All social and economic processes leading to marginalization, poverty and inequality determine the outcome of a hazard event. The way to decrease vulnerability involves changing structures and institutions that govern human lives and consequently social and economic processes. This type of vulnerability is known as social vulnerability or vulnerability as a start-point (Gilbert, 1995). Some authors define biophysical vulnerability as a function of hazard and social vulnerability (O'Brien, Eriksen, Schjolden, & Nygaard, 2003; Vincent, 2004).
Thus, it is the interaction between the structure and the individual(s) that ultimately defines vulnerability (Pelling, 2003). Adaptive processes that occur
39


at larger scalenation or worldare necessary to make resources available at the smaller scalelocal or individual. Yet it is an individuals action(s) that define in some ways the mechanisms to employ these resources for their own well-being (Hewitt, 1995).
The disaster pressure and release model defined by Blaikie et al. (1994) summarizes the stages of vulnerability in a very useful way. Root causes such as limited access to power, structures and resources, and ideologies of political and economic systems create dynamic pressures on individuals. These pressures can be lack of local institutions, training, appropriate skills, local investments and markets, press freedom and ethnical standards, and macrolevel forces such as rapid urbanization and population growth, arms expenditures, debt repayment schedules, deforestation and land degradation. All these dynamic pressures can create unsafe living conditions for humans: dangerous locations and infrastructure, poverty and livelihood insecurity, vulnerable groups, and insufficient disaster preparedness. And the presence of unsafe conditions in the presence of hazard events results in disaster.
There is a broad array of terms used to describe vulnerability. Resilience is another term used in contrast to vulnerability. It is defined as the ability of an actor to cope with or adapt to hazard stress and return to the previous stable condition without incurring any long-term negative consequences (Lai et al., 2001; Pelling, 2003).
Sensitivity, susceptibility, resistance, capacity, and potentiality are some of the constructs used to describe vulnerability. But the following are the
40


components of vulnerability that are most commonly used in the scholarly literature:
1. Exposure the degree of climate stress. Exposure can be both shortterm and long-term. Frequency, duration and intensity are some of the features of exposure.
2. Sensitivity the degree to which the system is affected by climate stress directly or indirectly and negatively or positively (White et al., 2001). Some authors use resistance instead of sensitivity to describe the individuals or groups capacity to withstand the impact of a hazard (Pelling, 2003). Methods to reduce sensitivity can be a change of economic, social and political circumstances. The number of possible hazards is large and the manifestation of a particular hazard varies. Consequently, sensitivity should not be viewed as general and applicable to all hazardous events. Sensitivity is a hazard-specific phenomenon and an acknowledgement of its unique features will help to.define successful policies to prevent or mitigate disasters.
3. Adaptation or coping an adjustment of a system to climatic stress, potential damages, or consequences (O'Brien, Sygna, & Haugen, 2004). The function of adaptation is to reduce social vulnerability and promote sustainable development by making changes in ecological, social, and economic systems (Smit & Pilosofa, 2001). Adaptation depends greatly on the adaptive capacity or adaptability of an affected system, region, or community to cope with the impacts and risks of climate change. Adaptability includes technological options, availability and access to resources, human and social capital, decision-
41


making process and structure of critical institutions (Brooks, 2003; O'Brien et al., 2004). As in the case of sensitivity, adaptation is hazard-specific; it varies depending on exposure (Leatherman & Thomas, 2001).
In summary, there are several important features of vulnerability:
Vulnerability is variable (contextual): it varies across geographical space and social groups.
Vulnerability is scale-dependent: vulnerability at the local level may not be the same as vulnerability at the regional and/or national levels.
Vulnerability is dynamic: it varies depending on changes in social structure and forces over time (Hewitt, 1995).
The vulnerability paradigm is appropriate for studying the risks Mongolian pastoralists face today and the impact of natural hazards on their livelihoods. As mentioned earlier in this chapter, severe climatic events were always a threat to nomads occupying Mongolian grasslands for centuries. Political and socioeconomic changes that have occurred in the country since the early 1990s significantly challenged their lives. Insecurity, poverty, and lack of state support became a reality of everyday life. OBrien and Leichenko (2000) called this situation a double exposure: a population confronted by consequences of both climate and social changes.
As described in the previous chapter, various adaptive strategies have surfaced to cope with severe climatic hardships. Understanding the impact of these adaptive strategies in reducing vulnerability to hazards and their outcomes
42


is important in securing livelihoods of pastoralists in Mongolia. The results of this work might be usefully applied in other contexts.
Theory of Political Ecology
Many vulnerability studies draw on theories of political economy and political ecology in explaining the factors that lead to vulnerability, and on social capital as a means of claiming access to resources and pursuing coping mechanisms (O'Brien et ah, 2004; Olmos, 2001; Pelling, 2003). The physical and biological environment we live in today is politicized: environmental problems cannot be understood in isolation from their political and economic contexts within which they are created and/or exacerbated (Cutter, 1996). Both political economy and political ecology explore the role of power relations in human uses of the environment. Exposure, sensitivity and adaptation/coping are all defined by individuals' access to resources and assets. The access is rooted in global political and socioeconomic structures (Brooks, 2003; Pelling, 2003).
But political ecology offers more to the objectives of this research. Political ecology as opposed to political economy, "seeks to understand the complex relations between nature and society through a careful analysis of what one might call the forms of access and control over resources and their implications for environmental health and sustainable livelihoods (Watts, 2000).
43


Leatherman and Thomas (2001) point to two important things that need to be incorporated into research on human-environmental interaction: social relations and environmental and social contexts. Relations of power or access to resources through social relations are key elements that define livelihoods of individuals and households, their exposure to different exposures and coping abilities. On the other hand, the environmental context within which these relations of power take place shape local context, the latter is important in social relations (O'Brien et al., 2004).
In summary, the theory of political ecology emphasizes the following premises:
The local environment is shaped by the processes occurring at the higher levels. It is important to understand how the manifestations of globalization affect adaptive capacity within localities.
Human-environment interaction is shaped by these local circumstances.
Humans are active agents trying to alter their local environment and manage environmental risks (Moran, 2000).
In Mongolia, structural adjustment programs implemented by the International Monetary Fund (IMF) and World Bank introduced price' liberalization, privatization of state enterprises, and removal of restrictions on international trade in early 1990s (Janes & Chuluundorj, 2004). Aid from the former socialist countries, mainly the Soviet Union, had ended, leaving industries without electricity and agricultural machines without fuel.
44


Unemployment, poverty, hunger, increase in crime and alcohol abuse were the major consequences of these policies.
For rural herding households, these changes at the macro level resulted in a lack of support from the state that was available in the past in the form of human resource, transportation, hay, fodder, and health care and livestock replacement in case of natural disasters and loss. This dramatically increased their vulnerability to summer droughts and winter storms. Survival in the short term became the primary concern of pastoralists, and for many, exhausted all of their resources.
In response to these macro and household level changes, securing access to resources through social ties and improving knowledge and skills to accumulate assets has become more important than ever. Those households who had social ties to both rural and urban areas, within the local government, and to health care and social services had better opportunities for trade of livestock products for basic necessities, for receiving appropriate care when needed, and for sending children to school. The importance of these resources increases when environmental stress is greater. Those who have larger families and many friends and relatives can move to more distant regions where the environment is more favorable. Those who have connections to the local government may secure better winter camps and benefit from occasional aid in food items, warm clothing, and hay and fodder at discounted prices. Therefore, political and economic changes in the country were reflected in changes to the
45


livelihoods of Mongolians, and new forms of strategies to adapt to circumstances and secure access to resources emerged as a result.
Social Capital Theory
The theory of social capital is of great importance in this research. As discussed earlier in the dissertation, mobilization of a variety of social resources has emerged as an important adaptive strategy employed by Mongolian pastoralists to prevent or overcome negative impacts of natural disasters.
Definition of Social Capital
During the last decade or so, social capital has become one of the most popular concepts in the social sciences. Originating in the work of a French sociologist Emile Durkheim in the 19th century, its application has expanded to multiple fields, including health-related disciplines. Social capital is often seen as the third form of capital, the first two being financial and human. Portes (1998) defines economic capital as people's bank accounts, human capital as inside their heads, and social capital as the structure of their relationships.
The origin of social capital dates to the 1980s when Bourdieu gave the first systematic contemporary analysis of social capital. He defined social capital as: .. .the aggregate of the actual or potential resources which are linked
46


to possession of a durable network of more or less institutionalized relationships of mutual acquaintance or recognition (Bourdieu, 1985, p. 248).
Following Bourdieu, Coleman, in his seminal work, defined social capital through its functions:
Social capital is defined by its function. It is ... a variety of entities with two elements in common: they all consist of some aspect of social structures, and they facilitate certain actions of actors whether persons or corporate actors within the structure (Coleman, 1988, p. 98).
Both Bourdieu and Coleman mention social network or structure as the foundation of social capital. Bourdieu put more emphasis on formal networks, whereas Colemans ideas of social capital emphasize informal networks.
Putnam added the importance of trust and norms to social networks. He stated, .. .social capital.. .refers to features of social organization, such as trust, norms, and networks that can improve the efficiency of society... (Putnam, Leonardi, & Nanetti, 1993).
In recent years more and more researchers define social capital as consisting of those resources available through social networks (Lin, 2001; Portes, 1998). Social networks, in this case, are defined as, the web of social relationships that surround an individual and the characteristics of those ties (Berkman & Glass, 2000, p. 145).
47


Clearly, the definition of social capital varies, but researchers tend to separate group-level structural social capital that emphasizes resources available through social relations and structures, such as interpersonal trust, norms and values, civic engagement, rule of law, and governance; and individual-level social support that flows in through social networks. This variation in a definition leads to methodological differences in measuring social capital (Grootaert & Van Bastelaer, 2002; Kawachi & Berkman, 2000).
Capturing all dimensions of social capital in one research project may 'be a daunting task, though choosing a specific dimension and designing a research to capture only this dimension can produce meaningful results. Societies are unique in their social relationships, thus in their social capital. A dimension of social capital that is important in one society may not be relevant in another. This is certainly true for rural pastoralist communities in Mongolia. Formal organizations are not typically important in their everyday lives, but local norms and values, and trust can have significant influence in their lives. Yet access to formal organizations credits, insurance, etc. may be important during times of stress. In addition, informal networks and resources flowing through these networks may be very important, especially for the poor. These micro- and meso-level dimensions of social capital are the main focus of this research.
48


From a methodological perspective, researchers define three types of social capital:
bonding social capital: ties to people who are similar in terms of their demographic characteristics;
bridging social capital: ties to people who do not share same demographic characteristics; and
linking social capital: ties to people in positions of authority (Grootaert, Narayan, Jones, & Woolcock, 2004).
All three functional types of social capital are important for socioeconomic well-being and health of individuals and groups. One type might offer benefits that are not available though other types. Linking social capital may offer more opportunities to access social services, whereas bonding social capital facilitates the adoption of healthy behaviors, and bridging social capital facilitates information exchange.
It is worthwhile mentioning that social capital may also have a negative impact on those both within and outside a community:
adverse effects on outsiders: excluding outsiders and maintaining inequalities between groups (Waldinger, 1995); and
adverse effects on insiders: restricting individual privacy and autonomy, making it hard to get out of "bad" groups such as drug dealers, gangs, and reducing the inflow of new ideas (Productivity Commission of Australia, 2004; Bourgois, 1995; Portes, 1998).
49


These negative effects of social capital may severely affect households who by virtue of escaping harsh climatic conditions are forced to move to different counties: help in new places may not be available. For this reason unless they face severe hazards, households do not tend to move to places where they do not have any social ties and support.
Social Capital and Economy
Links between social capital and development have been examined in a range of contexts. Higher levels of social capital appears to be beneficial to economic development, effective political institutions, and reduction of political problems (Fukuyama, 1995; Putnam, 1995). The following are examples of research conducted in both developing and developed countries that have explored the relationship between social capital and socioeconomic development:
Guiso et al. (2000) found that in Italy, in areas with high levels of social trust, households invest less in cash and more in stock, use more checks, have higher access to formal credits. Firms also benefit from higher social trust: they have more access to credits and are more likely to have multiple shareholders (Guiso, Sapienza, & Zingales, 2000).
Social capital measures (e.g. family network) were associated with higher secondary school graduation rate, college enrollment,
50


socioeconomic status, and avoidance of criminal activities among children of teenage mothers in the U.S. (Furstenberg & Hughes, 1995).
Higher social capital is shown to have a positive impact on watershed conservation and in cooperative development activities in Rajasthan, India (Krishna & Uphoff, 1999).
More trust, reciprocity, and sharing in neighborhoods of Dhaka, Bangladesh predicted a likelihood of a neighborhood to have a voluntary solid waste management system (Pargal, Huq, & Gilligan, 1999).
Social capital measured by the number of memberships in associations, diversity of memberships, number of meetings, and cash and time contribution to associations are positively related to asset accumulation and access to credits in Indonesia (Grootaert, 2000).
The role of social capital in economic well-being of households may be even greater in developing countries where almost all of the transactions between individuals are performed based on individual trust and trust in informal institutions. Formal institutions (e.g. courts) that typically regulate transactions may not function properly or may be too expensive (Durlauf & Fafchamps, 2004).
Researchers have identified the following mechanisms through which social capital facilitates economic development:
Social capital facilitates transactions among individuals or groups.
Common rules and trust allow people to interact efficiently.
51


Social capital makes information and knowledge exchange more efficient. Larger networks ensure a greater more flow of information, whether it is about hazards, market prices, social services, or governmental aid.
Participation in social networks and the development of trust makes collective action easier. Developmental programs often have to rely on collective action.
High levels of social control puts pressure on individuals in a network and forces them to engage in positive behaviors that are mutually beneficial (Grootaert & Van Bastelaer, 2002).
Social capital increases access to social services (Productivity Commission of Australia, 2004; Kawachi, Kennedy, & Glass, 1999).
These mechanisms do not relate only to economic development and household economic welfare. They are relevant to the association of social capital to health as well. In addition, socioeconomic status may become an intermediary link in the relationship of social capital to health.
The importance of social capital and support has become greater for Mongolian herders since privatization. Livestock husbandry is a difficult job that requires adequate human resources, access to better pasture and water especially in times of unfavorable environmental conditions, reliable information flow about weather, market, and government policy, and assurance of resources available through connections with local government. All of these can be secured only through expanding their social ties and establishing a flow
52


of mutual support through these networks. Reciprocity and hospitality are two features that are common among pastoralist societies. These features facilitate broader larger social ties and support that are crucial to successful herding.
Social Capital and Health
The effect of structural social capital and individual social support on health is well documented. Studies conducted across different populations show that people who have greater social capital and/or social support live longer (Christensen, Wiebe, Smith, & Turner, 1994; Giles, Glonek, Luszcz, & Andrews, 2005; Kawachi et al., 1999; Kawachi, Kennedy, Lochner, & Prothrow-Stith, 1997), are less susceptible to non-communicable diseases such as cardiovascular diseases, cancer and arthritis (Eng, Rimm, Fitzmaurice, & Kawachi, 2002; Eriksen, 1994; Kinney et al., 2003; Rosengren, Wilhelmsen, & Orth-Gomer, 2004; Vogt, Mullooly, Ernst, Pope, & Hollis, 1992; Weinberger, Tierney, Booher, & Hiner, 1990), are more likely to survive myocardial infarction (Berkman, Leo-Summers, & Horwitz, 1992; Farmer & Meyer, 1996; Kawachi et al., 1996; Orth-Gomer, Rosengren, & Wilhelmsen, 1993; Seeman, 1996), are less likely to suffer from mental illnesses (Bal, Crombez, Van Oost, & Debourdeaudhuij, 2003; Bassuk, Glass, & Berkman, 1999; Michael, Berkman, Colditz, & Kawachi, 2001; Penninx et al., 1997), are less susceptible to infectious diseases (Holtgrave & Crosby, 2003; Theorell et al., 1995), and are more likely to engage in health-promoting behaviors (Treiber et al., 1991).
53


The mechanisms through which social capital influences health is the same as in case of economic development: reducing transaction costs, disseminating information and knowledge, promoting healthy behaviors, increasing access to health services, plus psychological benefits (buffer of stress, increased self-esteem, stability and control over environment affects the neuroendocrine and immune systems, and influence the overall physical health) (Cohen & Syme, 1985).
In a study of maternal mortality in Mongolia, social support available through the family network was an important determinant of maternal deaths (Janes & Chuluundorj, 2004). A woman could not receive a medical care on time because no one was available to take care of her children while she was gone, or no one was able to lend her money to cover necessary expenses. Pregnant women do not stop doing heavy household chores such as fetching water and firewood, milking animals and processing dairy products. All of these involve lifting of a heavy staff, which can facilitate placental abruption and hemorrhage. Often women who died in pregnancy or childbirth did not have anyone nearby who could help while their husbands were gone herding their animals or selling their livestock products.
The concept of social capital has been increasingly used by researchers in health-related fields but there are several important limitations to such studies:
Most measures are measures of outcomes, not of social capital itself.
They measure only the quantity of social capital, not quality.
54


Tendency to use a single indicator of social capital, which fails to capture the multi-dimensionality of the phenomenon.
A failure to recognize that social capital will vary by network type and social scale.
A failure to differentiate the effects of social capital from other forces such as political institutions (Productivity Commission of Australia, 2004).
Since Colemans work, researchers increasingly acknowledge that social capital is a group/community attribute. Problems may rise when social capital measurements are taken at the individual level. Yet, Brehm and Rahn (1997) argue that it is individuals who ultimately build relationships and trust, and who participate in various community activities. Thus, it is also important to measure social capital at an individual level in addition to group-level measurements. In developing countries, where the quality of data collected at the community levels is often poor, individual-level variables may be the only valid option.
Gender
Gender is an important covariate in health related research not only due to the biological differences between female and male bodies, but also due to the social differences that may put one gender at larger risk of certain illnesses. Gender theory attributes such differences to the varying roles that females and
55


males perform in their household livelihood systems. Chen (1991) argues that women's activities are more multidimensional than men's, some of the activities are a sole responsibility of women, and women's role in creating and keeping networks are greater than men's. However, viewing labor division as the only factor in health inequalities among men and women carries serious flaws. It does not acknowledge social structural explanations of womens vulnerability such as power, access to resources, discrimination against women, and domestic violence.
Environmental challenges such as land degradation, drought or flood just to name a few, limit the abilities of women to handle their household responsibilities and increase their work burdens, negatively affecting their health status. The literature reveals a pattern of gender differentiation throughout the disaster process. The differences are largely attributed to childcare responsibilities, poverty, social networks, traditional roles, discrimination, and other issues related to gender stratification. More women die during natural disasters compared to men because often they are confined to homes and do not escape on time. Women experience more emotional problems during and after disasters compared to men. It is often on womens shoulders to rebuild their and their childrens lives because men tend to work outside of their homes (Fothergill, 1996).
The role of women in the pastoral economy is expanding in Mongolia. Traditionally, men were responsible for herding and women were responsible for dairy processing and other household duties. With an increase of herd size
56


and mix of animals, it became necessary for women to engage in other non-traditional activities to help the family, including a sale or trade of livestock products and wage labor for cash income. They are now important contributors to the households economy and this may increase womens risk for sickness. Demand for labor during disasters may result in delay in seeking care for health problems. Whether or not Mongolian males and females experience disasters differently and have varying health outcomes has not yet been studied.
Theoretical Framework
All of the theories and constructs used in the study are summarized in the theoretical framework presented schematically in Figure 3.1.
Vulnerability consists of three components: natural hazards, sensitivity factors and adaptive capacities of individuals and households. All three components are shaped by global processes. Global warming due to the increased industrialization and higher emission of greenhouse gases increase the number of drought and dzud events. Political processes that occurred in the countries of the former Soviet block have brought significant changes in national and local economies and peoples livelihoods. Structural adjustment programs taking place in developing countries increase the number of poor people and widen the gap between the rich and the poor. Thus, political ecology affects the processes that occur in the physical environment and social forces that define the vulnerability to natural events.
57


The outcome of a natural hazard, measured by socioeconomic loss and health conditions, is dependent on sensitivity and adaptation of individuals, households or regions. Lower sensitivity and higher adaptive capacity buffer negative impacts of hazard events. Very young or very old people, women, the poor, and those who do not have adequate skills are most sensitive to natural hazards. Individuals and households with greater social capital, better herd management skills and more accumulation of resources are less likely to suffer from short- and long-term consequences of natural hazards.
Sensitivity and adaptation are also related. Enhancing adaptive capacity will result in lower sensitivity and lower sensitivity also increases adaptive capacity. Better socioeconomic status and health conditions of individuals and households will in turn lower the sensitivity to natural hazards and increase the adaptive capacities. Thus, all elements of vulnerability and the outcome are interrelated with each other, forming a circle in the theoretical framework shown below.
58


J
GLOBAL PROCESSES
Global warming
Structural adjustment programs
Political processes
SENSITIVITY
Age: young/elderly
. Gender: femaleheaded households
High population growth/dependency ratio
Poor
Inadequate skills: education, occupation
V U L N E R A B I L I T
Y
adaptation!}
Social capital/social network and support
Good herd management skills
Accumulation of resources
WELL-BEING
Socioeconomic Status
Health
Figure 3.1. Theoretical Framework
59


CHAPTER 4
METHODOLOGY
The objective of this research is to understand how environmental challenges interact with political and socioeconomic circumstances and the adaptive strategies employed by herders to affect the well-being of rural pastoral households. Of particular importance is the identification of household access to social resources and/or emerging or incipient cooperative institutions that spread risk among numbers of participating households.
To meet the overall objective of the research, the following specific aims were developed:
Aim #1. To identify vulnerability to natural hazards at the county level by assessing the exposure to hazards, sensitivity (a potential of being affected by a climate stress) and adaptive/coping strategies using climate and county socioeconomic and demographic data.
Aim #2. To explore the relationships between vulnerability and health outcomes at the county level.
60


Aim #3. To study the impact of vulnerability to drought and dzud (a Mongolian term for winter disasters) on the economic well-being of rural households.
Aim #4. To investigate the relationship of vulnerability to drought and dzud and the health status of individuals within households.
A multi-level research design is employed with the main outcome variable, well-being, measured multi-dimensionally at the county, household, and individual levels, and includes both socioeconomic status and general health outcome.
Research Plan
Identification of sensitivity factors and adaptive strategies in the context of diverse environmental risks requires a two-stage research design: a spatial ecological study at the county/community level, and a cross-sectional study of. 120 households sampled from four counties in Mongolia. The details of each element of the study design will be described in the following sections.
A Spatial Ecological Study at the County Level
The spatial ecological study involved assessing relationships among and between climate, socioeconomic factors, and health in all of Mongolias rural counties.
61


Spatial statistical methods are commonly used to assess rates across geographic space, adjust relationships for noise (e.g. spatial autocorrelation -areas closer to each other tend to have similar values), identify disease outbreaks, and evaluate the impact of specific exposures.
Problems of analyzing spatially referenced data using nonspatial methods pose several problems, the most important one of which is spatial autocorrelation. The first law of geography states that Everything is related to everything else, but near things are more related than far things (Tobler cited in Waller & Gotway, 2004, p. 3). Thus, using nonspatial statistics violates the assumption of independent observations which may result in serious flaws. Spatial statistical methods consider this spatial dependency of observations and allow analysis of differentiating trends/relationships from spatial autocorrelation.
The availability of mapping tools, including Geographical Information Systems (GIS), greatly influenced the development of medical geography. The use of spatial statistics and GIS in health sciences has mainly focused on creating demographic, economic and lifestyle profiles of communities and their relationship to environmental hazards (Bellander et al., 2001; Dunn, Woodhouse, Bhopal, & Acquilla, 1995; Moran & Butler, 2001; Morrow, 1999; Pine & Diaz, 2000; Speer, Semenza, Kurosaki, & Anton-Culver, 2002). Studies have found an association between exposure to environmental hazards and cancer (Entwisle, Rindfuss, Walsh, Evans, & Curran, 1997; Hjalmas, Kulldorff, Gustafson, & Nagarwalla, 1996; Krautheim & Aldrich, 1997), permitted
62


surveillance and monitoring of vector-borne or infectious diseases (Boone et al., 2000; Emch, 1998; Fost, 1990; Kistemann, Munzinger, & Dangendorf, 2002; Marfin et al., 2001; Tanser & le Sueur, 2002); quantifying environmental levels of toxic substances and their health effects (Kohli, Noorlind-Brage, & Lofman, 2000; Margai, 2001); measuring access to health services (Brabyn & Skelly, 2002; Eyles, 1990; McLafferty, 2003; Parker & Campbell, 1998; Rosero-Bixby, 2004); planning service areas (Bhana, 1998; Foley, 2002; Kofie & Moller-Jensen, 2001; Perry & Gesler, 2000); and predicting pedestrian injuries (Durkin, McElroy, Guan, Bigelow, & Brazelton, 2005; LaScala, Johnson, & Gruenewald, 2001; Lightstone, Dhillon, Peek-Asa, & Kraus, 2001). Lately, spatial statistics and GIS methodologies are increasingly used in social epidemiology to study the distribution of the social and behavioral determinants of health outcomes (Entwisle et al., 1997).
There are several important concerns in using spatial statistics such as spatial autocorrelation mentioned above, including the modifiable area unit problem (changing spatial units of analysis may result in different relationships), the ecological fallacy (relationship at aggregate level may not apply to individual level), and non-uniformity of space and edge effects (the spatial units of analysis are drawn arbitrarily and events on the edges of a unit are forced to relate to the center) (Waller & Gotway, 2004). Careful design and analysis may significantly reduce these problems and enhance the validity of results.
63


Sampling
250 rural counties (out of 302 total) in Mongolia are included in the spatial analysis of climatic, socioeconomic, and health conditions in Mongolia.
Variables
The analysis examines three categories of independent variables, all measured in 2003 at the county level. The descriptions of independent variables are given in Table 4.1.
Monthly temperature and precipitation data were obtained from 108 weather stations across Mongolia for the period 1993-2003. Universal kriging was performed using GIS to predict the monthly averages for each county. Several indicators, including coefficient of variability, average standardized anomaly, aridity index, yearly average, and yearly range, were created from predicted monthly data and regressed towards the number of livestock and livestock growth rate for 1993-2003. The best predictors of the number of livestock and livestock growth rate were mean yearly precipitation and temperature ranges, therefore used in further analyses.
Sociodemographic characteristics and livestock indicators were collected at the end of 2003. The data were not available for the previous years. Thus, the data only allow us one to look at the more immediate consequences of climate stress.
64


Table 4.1. Independent variables in the spatial ecological study
Variable Categories Independent Variables Explanation of Variables Data Source
Mean yearly The difference between the Institute of
precipitation maximum and minimum Meteorology
cn 00 O u +-> C/3 c5 B Mean yearly The difference between the
u temperature range 1993-2003 maximum and minimum monthly average temperatures is calculated for each year and averaged. A larger range means higher temperatures in the summer and low temperatures in the winter.
Dependency Ratio of the population National
C/3 O ratio under 18 and over 65 years Statistical
CA of age to the population Office of
In 1 between 19-64 years of age Mongolia
b es O T3 Percent of Percentage of people whose
O .S people with income is under absolute
income below poverty line in the total
2 > to O 11 _o 'o o 00 poverty line population. The poverty line is set by the government for each region and can be either a minimum wage per person or minimum number of livestock per person.
65


Table 4.1. Independent variables in the spatial ecological study (Cont.)
Variable Categories Independent Variables Explanation of Variables Data Source
1 o &£ .a ab-g *2 Unemployment rate Percentage of people who are not employed in the total population of 18-64 years of age. National Statistical Office of Mongolia
2 fc B -Q >> s 8? 2 y O w o e oo S C/D Percent of people receiving pensions and allowances Percentage of people who receives pensions and allowances in the total population
Livestock indicators (indicators of adaptive strategies) Livestock density in SFU Livestock per person in SFU Ratio of the number of livestock in SFU to total land in hectares Ratio of the number of livestock in SFU to total population National Statistical Office of Mongolia
The health-related dependent variables of the first stage of analysis are shown in Table 4.2. The crude morbidity rate reflects the total number of hospital visits, including visits to outpatient clinics.
66


Table 4.2. Dependent variables in the spatial ecological study
Variable Categories Dependent Variables Explanation of Variable Data Source
Health Crude Number of hospital- National
morbidity rate recorded visits/Total population* 1000 Statistical Office of Mongolia,
Crude mortality Number of deaths/Total Ministry of
rate population *1000 Health,
Under five Number of deaths of World
mortality rate children <5 years of age/ Health
Maternal mortality rate Total livebirths *1000 Number of maternal deaths/ Number of women of reproductive age *100000 Organization
Analysis
A conditional autoregressive (CAR) model was used to analyze the effects of the independent variables on health indicators. CAR models include spatial parameters to measure spatial autocorrelation by comparing values in a county to the values in neighboring counties. In other words, the means and variances defined by the CAR model are conditional, thus dependent on the values of neighbors. This permits separating the relationship between independent and dependent variables from the relationship that occurs due to spatial autocorrelation. If spatial autocorrelation is not adjusted in the analysis,
67


parameter estimates and standard errors may overestimate existing relationships.
The common problems that rates pose for data analysis are abnormal distributions, dependence of the variances of rates on the mean rate itself, and the dependence of variance on population size. These are serious problems that may lead to wrong conclusions if not fixed appropriately. The Freeman-Tukey. square root transformation was thus used for outcome variables to make their distributions normal. This transformation also removes a connection between the rate and the variance of the rate. Subsequently, a weighted analysis is performed to remedy the dependence of the variance on population size (heteroscedasticity) (Waller & Gotway, 2004). Nonspatial and spatial parameters of the analyses are presented in Chapter 5.
A Cross-Sectional Study at the Household Level
In the second stage of the research, 120 household-level structured interviews of households in four counties,- stratified by vulnerability and health indicators, were conducted. The main objective of this study was to explore the impact of ecological factors on individuals and households, factors that determine the extent of such effect(s), and the outcome of this interaction expressed in individuals and households levels of economic status and health. These interviews permit analysis of household vulnerability and the long-term outcome of natural disasters, and provide important information on how the
68


coping and adaptive strategies employed by rural herding households mitigate the consequences of adverse environmental events. The complex interaction among environmental constraints (same as in the spatial analysis stage of the study), adaptive and coping capacities of households and their socioeconomic well-being and health status can only be assessed in details via interviews. The results of analyses at this stage will compliment the findings from the first stage of the research.
Sampling
Analysis of the county-level data were used to identify four counties that differed on two dimensions: vulnerability (combination of climatologic stress, socioeconomic factors, and adaptive strategies identifiable from county-level data), and level of household well-being (as reflected by health indicators). Counties were ranked for each independent and dependent variable shown in Tables 4.1 and 4.2. Mean ranks for each category of variables (climate stress, sociodemographic, socioeconomic and health) were calculated for counties.
Counties were sorted into quartiles by ranks on health status and three categories of vulnerability statuses. Counties falling in the lower and/or upper quartiles on each category were included in the sampling frame, generating a 2x2 table. There were five counties in overall that fit the criteria for the sampling fame. One county in each cell was selected to be most representative
69


of the ecological diversity of Mongolia. The following 2x2 table describes the sampling frame and indicates the counties chosen for household interviews.
Table 4.3. Sampling frame for household interviews
Socioeconomic and Environmental Vulnerability
HIGH
LOW
POOR Olziit (desert) Khovd (steppe)
GOOD
Bayankhutag
(steppe)
Bayan-Ondor
(desert)
It is important to remember that this scheme is based on county level data; the indicators are only indirect reflections of what happens to individual households. The role of the aggregate data analysis is to develop a sampling frame that captures the widest possible range of household experiences in the context of diverse environmental stresses. The scheme serves to maximize variability, and, thereby, generate a sample with a greater probability of highlighting household-level coping and adaptive strategies, and intrahousehold sociodemographic factors that affect household well-being.
The following four counties were selected using a 2x2 sampling frame: Bayankhutag county in Khentii province, Olziit county in Dundgovi province, Bayan-Ondor county in Bayankhongor province, and Khovd county in Khovd
70


province (Figure 4.1). These counties represent two different ecological zones: steppe (Bayankhutag and Khovd) and desert (Olziit and Bayan-Ondor).
Figure 4.1. Counties Selected for Interviews
Thirty interviews in each county were conducted in the summer of 2005. This provided an acceptable level of power for identifying variables comprising household vulnerability and linking these to individual health outcomes (based on identifying variability in health outcomes of between 2-5% in the population, with a sensitivity of +/- 2%). A variety of power calculations, using the stat-calc function in Epilnfo showed the sample size to be adequate to testing the main elements of the study hypothesis. This said, the sample is by epidemiologic standards relatively small. This should be kept in mind when evaluating tests of significance.
71


County officials and health care providers were contacted at the onset to locate households for sampling. A transect sampling method was used to select households within counties: households along the road, river or valley were counted and a systematic sampling method was used to select from households. Every second to every fifth household was approached for an interview. A second and/or third attempt was made if adult members of a household were not present.
It is important to note here that the households that had suffered significant losses of animals, and were thus the most vulnerable, were likely to have been forced to migrate to urban areas. The resulting household sample may thus not represent the most vulnerable of the population.
In Olziit county a systematic sampling strategy was not feasible: many households had left the county territory because of a severe drought, and every household we could locate was interviewed. In Bayan-Ondor county, households moved to the north and to the territory of a neighboring county due to the drought as well. Thus interviews took place in the northern part of the county and also in the territory of a neighboring Erdene county of Gobi-Altai province. Locations of each interview site were recorded using the Global Positioning System. If there was more than one household in the camping group, the first household with a head or spouse present was approached for an interview.
72


Interview Sites
Bayankhutag County. The territory of Bayankhutag county (Figure 4.2) is divided into three townships. Township #2 was selected randomly to conduct household interviews. At the time of the study the majority of households had moved closer to the Kherlen river on the northern border of the county due to a severe drought. The county center is located approximately 10 kms south of the province centerOndorkhaan. The population in Bayankhutag county are predominantly Khalkh Mongols, but a small number of Uriankhai and Buriat ethnic minority groups were also present. The population in 2003 was 2200. Patron-client relationships were well established in this county. Wealthier families with many livestock hired poor families to help in herding and related activities. In turn, patrons provided clients with meat, dairy and a monthly salary averaging approximately 40000 tugrics (equivalent of 35 USD).
Figure 4.2. Interview Sites in Bayankhutag County
73


Olziit County. Olziit is the largest and southernmost county in Dundgovi province (Figure 4.3). The county is notable for having the largest number of camels in Mongolia. With a population of 2900 in 2003, it is the third most populated county in Dundgovi province. Music, folklore songs, games and airag (fermented horse milk) are known to be popular among people in Dundgovi province. They are by reputation a socially active, friendly and fun-loving people. Because of extremely dry conditions and poor vegetation, herding households move very often: it is not uncommon for families to move 2-3 times per week. Households keep great distances from each other, and it was often difficult to locate single gers (yurts) among the sand dunes.
Figure 4.3. Interview Sites in Olziit County
Khovd County. Khovd county is located on the eastern slopes of the Altai mountain ranges and to the west of Khovd province center (Figure 4.4).
74


The closest households were located within five kms of the province center. There are only two townships in this county: the 1st township where all residents are engaged in fanning, and the 2nd township where livestock husbandry is the dominant mode of production. Interviews took place in the 2nd township.
The total population was 5202 in 2003. Khovd county is the only county in Khovd province where the majority of its (96%) population are Khazakhs. Khazakhs are the largest of the ethnic minorities in Mongolia. They speak a distinct language (though many also speak Khalkh Mongol), and, in contrasts to most Mongolians, are Muslim. The relationship between Khazakh and Mongol residents seems to be very good, although some disputes about animal theft between them takes place occasionally. Khazakh people are known for their great hospitality, close family ties, great sense of humor, and beautiful crafts.
Figure 4.4. Interview Sites in Khovd County
75


Bayan-Ondor County. Bayan-Ondor is the southernmost county of Bayankhongor province: it borders China, a fact that opens great trading opportunities for local herders (Figure 4.5). Perhaps as a consequence, Bayankhongor province is notable for having the largest number of goats in the country cashmere is a particularly important product for cross-border trade. There are three townships in this county, but because of a severe drought households from all townships had moved to the north, and many into an adjacent county.
Figure 4.5. Interview Sites in Bayan-Ondor County
76


Interview Setting
Interviews took place in a ger. Often there were people from other households present during the interview. After presenting an introduction to the study, a consent form was read and signed by a head of the household and/or spouse, and anthropometric measurements and hemoglobin tests of capillary blood were taken from each member of the household. Once all measurements were taken interviews were done with the head of the household and/or spouse. Each interview together with measurements took on average from 1.5 to 2 hours. Only two households refused to participate in the study. These were both located in Khovd county. The heads of these two households were not present and their spouses did not speak Mongolian.
Subject Payment
An equivalent of 5 USD was given as an incentive to each household participating in the study. Iron supplements were given to those who had anemia.
Household Questionnaire
A modified version of a social capital questionnaire, the Social Capital Integrated Questionnaire (SC-IQ) developed by a team of the World Bank, was
77


used in the research. This instrument had been tested for validity and reliability in developing country settings such as Kygyzstan and Nepal (Grootaert et al., 2004), though not in nomadic populations. Modifications were thus made to this questionnaire to fit it to the unique social characteristics of a highly mobile, pastoral population. Questions about herding practices, natural disasters, economic and health statuses were also added.
The questionnaire had the following sections:
1. Demographic information on household members: gender, year of birth, marital status, occupation, years of education, years of residency in the county, months present in the household within the last year.
2. Information on the khot ail (cooperative group): gender, age, and relationship, types of support given to and received from a khot ail.
3. Information about other immediate family members (parents, siblings, children and significant others): gender, age, relationship, occupation, education, place of residency, frequency of meetings, support given and/or received.
4. Social capital:
a. Social network: number of close friends and relatives, their residency, frequency of meetings, closeness to a county governor, likelihood to get help in different situations, access to loans and restocking programs.
b. Trust: likelihood to get help, trust in people, friends, relatives, local government and media/press.
78


c. Social cooperation: engagement in social formal and informal activities, likelihood to cooperate in various situations.
d. Information and communication: types of information sources, number of visits to county and province centers, distance to the nearest phone.
e. Social cohesion: wealth differentiation, number of visitors and households visited, number of games played and food shared with others, likelihood to help new people, and area safety.
f. Empowerment: impact on neighborhood, likelihood of a local government to listen to people, number of petitions of government, number of people who voted in the last election
5. Herding practices: number of moves, their duration and distances, purchase of livestock insurance, winter shelter and veterinary services, conflict over pasture and water sources, amount of hay and mineral lick prepared for winter, problems of overgrazing.
6. Experience of natural disasters in the past 15 years: livestock loss, aid from friends/relatives, government, disaster warning and preparedness, and accessibility of roads.
7. Health status during the last six months for each household member: presence of illness, access to health care, cost of care, availability of cash for, health care.
8. Economic status: livestock inventory, yearly income from livestock products and from pensions, wages and allowances, supplementary
79


income sources, large expenses within a year, and an inventory of household items such as TV, radio, battery, satellite dish, and transportation.
A complete questionnaire can be viewed in the Appendix.
Hematological and Anthropometric Data Collection
Anthropometric measures are known to be a relatively cheap method for early detection of a nutritional deficiency (Bailey & Ferro-Luzzi, 1995). Body Mass Index (BMI) is commonly used in assessing individuals nutritional status. It is an indirect indicator of body fatness. The formula for BMI is weight in kilograms divided by height in meters squared. Skinfold thicknesses also provide an estimation of general fatness. Of the many mineral deficiencies, iron deficiency anemia is a common problem in developed countries, where it is found as high as in 67% of children and 33% of women of reproductive age (Kormondy & Brown, 1998).
Hemoglobin. The blood hemoglobin concentrations of all household members, except children under six months of age, were tested in the field using a portable hemoglobinmeter manufactured by HemoCue Ltd, UK. A sample of capillary blood was obtained from the ring finger or middle finger of the left hand using a microlance, and immediately analyzed in the hemoglobinmeter. The sensitivity of Hemocue is 82.4% and specificity is
80


94.2% (Paddle, 2002; Sari et al., 2001). Hemocue of capillary blood has a 85% correlation with a gold standard: direct cyanmethaemoglobin (Sari et al., 2001). The measured hemoglobin was used to determine whether an individual was anemic using UNICEF/UNU/WHO/MI criteria. The following table shows the hemoglobin thresholds for anemia used in this study.
Table 4.4. Anemia cut-off levels
Category Hemoglobin in g/dl
Children under five years of age 11.0 !
Children 5-11 years of age 11.5
Children 12-14 years of age 12.0
Women non-pregnant 12.0
Women pregnant 11.0
Men 13.0
All children diagnosed with anemia were given ferrous sulfate syrup and adults were given ferrous sulfate tablets. Directions for use were given and a printed copy of the directions in Mongolian was provided to an adult female member of the household.
Body Weight. Body weight was measured in all subjects over two years of age using standard scale (precision of 100 g, made in Russia) which was periodically checked using a five kg standard weight. Subjects were weighed in
81


bare feet with a minimum of clothing. No corrections were made in the analyses for clothing. Body weight for children under two years of age was measured using a battery-operated digital weight scale (precision lOOg, Seca).
A child was placed in the middle of the scale with light clothing and the reading from the digital scale was recorded. The measurement was repeated if the baby was moving and would not lie still.
Body Length and Height. Length was measured in children under two years of age with a plastic measuring mat (5 mm precision, Seca). For children over two years of age and adults, height is measured using a portable stadiometer (1 mm precision, Seca) with a movable bar.
Skinfold Thickness. Abdominal and triceps skinfolds were measured using skinfold calipers. An abdominal skinfold was measured 1 cm to the right of umbilicus in a horizontal fold while the person was standing. A triceps skinfold was measured with the right arm hanging loosely, 1 cm posterior from the middle point between the lateral projection of the acromion process of the scapula and the inferior border of the olecranon process of the ulna while the elbow was flexed to 90 (Lohman, Roche, & Martorell, 1988). The correlation between the abdominal and triceps skinfolds was .90.
82


Variables
Complete lists of all variables entered into the analyses are given in the following tables. Independent variables used in the analysis are presented in Tables 4.5 4.9. Variables in Table 4.6 (socioeconomic indicators) are also used as outcome variables in some analyses. Descriptions of health outcome variables are given in Table 4.10.
Table 4.5. Household sociodemographic variables
Variables Explanation of Variables
Household size Number of permanent household members
Mean education years Mean education years of household members >18 years of age
Mean occupational rank Each adult household member is given a score from 1 to 3 based on occupation and a mean for each household is calculated: 1 for herder, farmer, worker or salesperson (does not require training) 2 for agent, administrative worker, carpenter, construction worker, cook, driver, mechanic, miner, sewer and typist (requires 1-2 years of training) 3 for engineer, feldsher/nurse, financial advisor, forecaster, teacher, policeman, veterinarian, and college student (requires specialized training of at least 3 years)
Percent of members in workforce Percentage of household members in workforce (18-64 years of age)
83


Table 4.6. Household socioeconomic variables
Variables Explanation of Variables
Number of livestock in SFU Sheep Forage Unit (1 kg forage per day): 1 sheep = 1 SFU 1 goat = 1 SFU 1 cattle = 6 SFU 1 horse = 7 SFU 1 camel = 7 SFU
Number of milk livestock in SFU Calculated the same way as above
Yearly household income in thousand tugrics Sum of income from the sale of cashmere, wool, meat, dairy, and hides, wages, pensions and allowances in one year
Number of household items Sum of the household possessions such as TV, radio, power generator, and extra ger
Number of transportation items Sum of the households transportation means such as motorbike, sedan, and truck
Large expenses in thousand tugrics Sum of the households large expenses within one year
Monthly food expenses per person in thousand tugrics A total amount of cash spent on food in the last month divided by number of people living in that household during this period.
Meat (kg) consumed per person per year Total amount of meat consumed by the household over the year is divided by the number of a permanent household members
84


Table 4.7. Household social capital variables
Variables Explanation of Variables
Number of households in khot ail Number of households defined by the interviewee as khot ail
Number of other family members Number of immediate family members (parents, children, siblings) and significant others not living in the household
Number of other family members in the same county Number of immediate family members (parents, children, siblings) and significant others living in the same county but not in the same household
Number of close relatives With whom the household members talk about private matters
Number of close friends With whom the household members talk about private matters
Asked local (county) governor for help Binary variable coded 1 for as Yes and 0 No
Number of people turned for help Number of people who turned for help within 1 month
Access to loans and restocking programs Binary variable coded 1 for as Yes and 0 No
Trust people Answers to the question How trustful are people nowadays? are scored as 1 not trustful, 2 little, 3 -somewhat, 4 very much
Trust relatives Answers to the question How much do you trust your relatives? are scored as 1 not trustful, 2 little, 3 -somewhat, 4 very much
Trust friends Answers to the question How much do you trust your friends are scored as 1 not at all, 2 little, 3 -somewhat, 4 very much
85


Table 4.7. Household social capital variables (Cont.)
Variables Explanation of Variables
Trust local government Answers to the question How much do you trust your county government? are scored as 1 not at all, 2 little, 3 somewhat, 4 very much
Trust press and media Answers to the question How much do you trust press and media are scored as 1 not at all, 2 -little, 3 somewhat, 4 very much
Number of communal activities participated Number of communal activities such as township or county meetings, fundraising events within the last year
Number of common issues discussed Number of informal meetings with people to discuss issues within the last year
Likelihood to cooperate on building a well Is scored as 1 very unlikely, 2 somewhat unlikely, 3 somewhat likely, 4 very likely
Likelihood to help in case of illness and disaster Is scored as 1 very unlikely, 2 somewhat unlikely, 3 somewhat likely, 4 very likely
Number of information sources about government activities Reported number
Number of information sources about market prices Reported number
Distance to the nearest phone Reported distance in kilometers
Wealth difference Is scored as 1 different to a great extent, 2 -different to a small extent, 3 not different
86


Table 4.7. Household social capital variables (Cont.)
Variables Explanation of Variables
Frequency of having food/drinks Frequency of having food/drink with others in special occasions such as birthday, funeral arrival of special guests within the last month
Frequency of playing games Frequency of playing games with others within the last month
Number of visitors in your ger The average number of visitors in a single day was asked and multiplied by 30 (one person may be counted multiple times)
Number of other gers anyone from your household visited The average number of other households any of the household members visited in a single day, multiplied by 30 (one household may be counted multiple times)
Likelihood to help newcomers Is scored as 1 very unlikely, 2 somewhat unlikely, 3 somewhat likely, 4 very likely
Area safety Is scored as 1- very unsafe, 2 somewhat unsafe, 3 -somewhat safe, 4 very safe
Your impact on neighborhood Is scored as 1 no impact, 2 a small impact, 3 a big impact
Likelihood of local government listening to people Is scored as 1 not likely, 2 to a small extent, 3 -to a great extent
87


Table 4.8. Natural disaster variables
Variables Explanation of Variables
Animal loss in SFU Animal deaths for each species are converted to SFU units and summed
Number of dzud years Since 1990
Number of drought years Since 1990
Number of dzuds and droughts combined Since 1990
88


Full Text

PAGE 1

A MULTI-LEVEL STUDY OF VULNERABILITY OF MONGOLIAN PASTORALISTS TO NATURAL HAZARDS AND ITS CONSEQUENCES ON INDIVIDUAL AND HOUSEHOLD WELL-BEING by Oyuntsesteg Chuluundorj B.S., Mongolian National Medical University, 1996 M.A., University of Colorado at Denver, 2001 A thesis submitted to the University of Colorado at Denver and Health Sciences Center in partial fulfillment of the requirements for the degree of Doctor of Philosophy Health and Behavioral Sciences 2006

PAGE 2

This thesis for the Doctor of Philosophy degree by Oyuntsetseg ChuluundOlj has been approved by Susan DreIsbach Deborah S. Thomas Date

PAGE 3

ChuluundOlj, Oyuntsetseg (Ph.D., Health and Behavioral Sciences) A Multi-Level Study of Vulnerability of Mongolian Pastoralists to Natural Hazards and Its Consequences on Individual and Household Well-Being Thesis directed by Professor Craig Janes ABSTRACT The transition to the free-market economy following the break-up of a socialist block in early 1990s totally transformed the pastoralist husbandry in Mongolia from state-supported collective farms to independent, subsistence-based herding households. The old socialist system provided a buffer against frequent cases of drought and winter storms. Plus, many negative social and economic consequences, including but not limited to livestock theft, market failure, increasing poverty and inequality, conflict over pasture and water sources has significantly reduced the coping capabilities of herders to resist to natural stress. c This study employs a multi-level research design to explore the adaptive and coping strategies employed by rural herders in Mongolia to natural hazard events and the effectiveness of these strategies on their well-being measured in terms of economic and health status. The results of the study indicate that climate stress is not a strong predictor oflower socioeconomic weU;'being and poorer health outcomes. The role of factors such as gender, age, education and household size that make individuals and households more vulnerable to suffering from

PAGE 4

natural hazards and their consequences and of adaptive strategies to buffer the effects of natural disasters such as social capital and better herd management skills are crucial in the final outcomes. This abstract accurately represents the content of the candidate's thesis. I recommend its publication. aigR. s

PAGE 5

ACKNOWLEDGEMENT I would like to acknowledge Craig Janes, the chair of my thesis committee, and Kitty Corbett, Susan Dreisbach, Deborah Thomas and David Tracer, the members of my committee, for the outstanding support and guidance through this process. My gratitude also goes to Sunmin Lee and Steve Sain for statistical advice. My thanks due to my friends, Oyungerel Nanzad and Delgermaa Tsagaankhuu for much needed help in the data collection, and Solongo Altangerel for providing me with iron supplements. I would like to thank the Public Entity Risk Institute for financial support that made the fieldtrips possible.

PAGE 6

TABLE OF CONTENTS Figures. ............. ................. .......... ..... ............. ..... IX Tables ................................................................. CHAPTER INTRQDUCTION....... ...................................... 1 2. BACKGROUND............................................... 6 Demographic and Socioeconomic Characteristics Mongolia .................................................... 6 Ecology and Climate ....................................... 11 Natural Hazards ............................................. 14 Pastoralism in Mongolia. . . . . . . . ... 16 Adaptation. . . . . . . . . . . . .... 21 Traditional Adaptive Strategies . . . ... 22 NonTraditional Adaptive Strategies........... 26 Natural Hazards and Health.................... ........ ... 31 3. THEORy..................................................... .... 35 Vulnerability Paradigm........ ......................... ... 35 Natural Disaster................................ .... 35 Vulnerability....................................... 37 Theory of Political Ecology............................... 43 VI

PAGE 7

Social Capital Theory ....................................... 46 Definition of Social Capital ...................... 46 Social Capital and Economy................. .... 50 Social Capital and Health..... .................. 53 Gender ................. .. .. .......... : .. .................. 55 Theoretical Framework..... ........ .................... ... 57 4. METHODOLOGY .............................................. 60 Research Plan ............................................... 61 A Spatial Ecological Study at the County Level....... 61 Sampling. . . . . . . . . . . ... 64 Variables ............................................ 64 Analysis .............................................. 67 A Cross-Sectional Study at the Household Level.... 68 Sampling . . .. . .. . . . . . .. 69 Interview Sites .................................... 73 Interview Setting. . . . . . . . .. 77 Subject Payment.................................. 77 Household Questionnaire ........................ 77 Hematological and Anthropometric Data Collection .......................................... 80 Variables ........................................... 83 Analysis ............................................. 92 5. RESULTS ........................................................ 94 Findings from County-Level Spatial Data Analysis... 94 Vll

PAGE 8

Descriptive Statistics .............................. 94 Conditional Autoregressive Models ............ 96 Findings from Data Analysis of Household Interviews ................................................... 102 Descriptive Statistics .............................. 102 Regional Differences .............................. 115 Natural Disaster.. ..................... ......... ... 128 Socioeconomic Status..... ..... ......... ........ 129 Health Status ....................................... 165 Disaster and Health. . . . . . . . .. 201 6. DUSCUSSION AND CONCLUSIONS.................... 203 Significance ................................................. 215 Limitations .................................................. 217 APPENDIX .................................................................... 219 HOUSEHOLD QUESTIONNAIRE.... ............ .... 220 GLOSSARy.............................................. ............. ....... 241 REFERENCES ............................................................... 242 Vlll

PAGE 9

LIST OF FIGURES Figure 2.1 Administrative Map of Mongolia ................. ... ........ ..... 7 2.2 Contribution of Agriculture to GDP .............................. 8 2.3 Population Distribution. . . . . . . . . . . . 9 2.4 Ecological Zones....... ... ...... .................................... 12 2.5 Livestock by Species, 1989-2005 ................................. 20 3.1 Theoretical Framework.. ........ ............................... .... 59 4.1 Counties Selected forInterviews .................................. 71 4.2 Interview Sites in Bayankhutag County .......................... 73 4.3 Interview Sites in Olziit County................................... 74 4.4 Interview Sites in Khovd County ................................. 75 4.5 Interview Sites in Bayan-Ondor County......................... 76 ix

PAGE 10

LIST OF TABLES Table years and animal losses ..................................... 15 2.2 Summary of adaptive strategies .................... .............. 31 3.1 Definitions of vulnerability ........................................ 37 4.1 Independent variables in the spatial ecological study........ ... 65 4.2 Dependent variables in the spatial ecological study.... ......... 67 4.3 Sampling frame for household interviews ....................... 70 4.4 Anemia cut-offlevels ................................... ........ .... 81 4.5 Sociodemographic variables....... .... ........ ............ .... ... 83 4.6 Socioeconomic variables................................ ........... 84 4.7 Social capital variables............................................. 85 4.8 Natural disaster variables.......................................... 88 4.9 Herd management variables ....................................... 89 4.10 Health variables .. :................................................ ... 90 5.1 Independent variables in the spatial ecological study........ ... 95 5.2 Dependent variables in the spatial ecological study.... ........ 95 5.3 Crude mortality rate per 1000 persons........................... 97 5.4 Maternal mortality rate 100,000 women of reproductive age ..................................................................... 98

PAGE 11

5.5 Under five mortality rate per 1000 livebirths ................. ... 99 5.6 Crude morbidity rate per 1000 persons....................... .... 100 5.7 Demographic characteristics.............. ......................... 102 5.8 Socioeconomic status............................................ ... 103 5.9 Social network....................................................... 105 5.10 Trust and solidarity ........................... ...... .................. 106 5.11 Collective action and cooperation................................. 107 5.12 Information and communication ................. ........... ...... 107 5.13 Social cohesion and inclusion..................................... 108 5.14 Empowerment and political action ................................ 109 5.15 Herd management. . . . . . . . . . . . .... 110 5 .16 Natural disasters ..................................................... 111 5.17 ................................................................ 112 5.18 Drought. . . . . . . . . . . . . . . . 113 5.19 Continuous outcome variables ........................ ............ 114 5.20 Binary outcome variables.......................................... 114 5.21 One-way ANOVA results: regional differences................ 116 5.22 Post-hoc test results: regional differences.................... .... 120 5.23 Fixed and random part results: number oflivestock in SFU ... 132 5.24 Fixed and random part results: number of milk livestock in SFU ................................................................. 138 5.25 Fixed and random part results: yearly household income in thousand tugrics ...................................................... 142 5.26 Fixed and random part results: number of household items ... 146 Xl

PAGE 12

5.27 Fixed and random part results: number of transportation items ................................... :.............................. 150 5.28 Fixed and random part results: large expenses in thousand tugrics .............................................................. ... 154 5.29 Fixed and random part results: monthly food expenses per person in thousand tugrics .......................................... 158 5.30 Fixed and random part results: meat (kg) consumed per person per year ........................................................................ 162 5.31 Fixed and random part results: hemoglobin in g/dl........... ... 167 5.32 Fixed and random part results: presence of anemia ............. 173 5.33 Fixed and random part results: Body Mass Index ............... 178 5.34 Fixed and random part results: overweight ........................ .... 183 5.35 Fixed and random part results: triceps skinfold in mm ..... .... 188 5.36 Fixed and random part results: abdomina:l skinfold in mm .... 193 5.37 Fixed and random part results: sick..... ....................... .... 198 5.38 Logistic regression with affects health" as the dependent variable .......................................... 5.39 Logistic regression with "drought affects health" as the dependent variable. . . . . . . . . . ..... 202 XlI

PAGE 13

CHAPTER 1 INTRODUCTION in 1990, Mongolia, a former client state of what was then the Soviet Union, undertook sweeping economic reforms. One important consequence of these reforms was the transformation of the rural, primarily pastoral, economy. Former state collective farms were dismantled and herding households were thrown into a highly insecure subsistence mode of production, and as a consequence, have become increasingly vulnerable to local fluctuations in rainfall, winter disasters, and the availability and quality of forage. Although Mongolian pastoralists have practiced, and continue to practice, a wide variety of adaptive strategies to manage risk, the economic transition has led to privatization and state disinvestment in the.rurql infrastructure to such a degree that the institutional structures which had previously distributed risk among households by managing access to resources, providing access to essential commodities, and marketing animal products, have largely disappeared. Given the lack of formal social institutions which would protect production and subsistence under uncertain environmental and market circumstances, it is unlikely that traditional adaptive strategies alone are

PAGE 14

sufficient to reduce risk. Some researchers suggest that various informal social arrangements that act as adaptive "safety nets" are emerging in some regions of Mongolia, yet the characteristics and success of such strategies are not well understood (Humphrey Sneath, 1999). In the absence of state support, households attempt to manage environmental risk in a number of ways. They diversify sources of income within the household (e.g., combining wage labor with herding activities). They undertake, with new intensity, the adaptive strategies many nomadic peoples use in unpredictable environments (e.g., management of herd size and mix, highly mobile foraging, creation of social bonds of reciprocity, absentee-herd owning). Some family members may move away for entire seasons to work in towns or cities. In some households, the elderly are sent to live in county and provincial centers so that they can support grandchildren sent to attend school and provide links to town-based services. The success or failure of these strategies has a direct effect on livelihoods of pastoralist herders. Some policymakers argue that intensifying livestock production and reducing environmental risk by introducing a sedentary form of herding is the only solution to enhance the livelihood of Mongolian pastoralists (Hoohdoi, 2002), but the long-tern1 environmental consequences of a settled herding can be devastating. In the case of Inner Mongolia, researchers found that reducing the mobility of herders by introducing technological innovation increased productivity per unit of land and labor in the short run, but eventually led to environmental degradation (Bates Lees, 1996). 2

PAGE 15

The objective of this research is to understand how environmental challenges interact with political and socioeconomic circumstances and the adaptive strategies employed by herders to affect the well-being of rural pastoral households. The research employs a combination of geographic, social, economic and anthropological methodologies (Borgerhoff-Mulder & Sellen 1?94). The main outcome variable, well-being, is measured multi dimensionally at the county, household, and individual levels, and includes both socioeconomic status and general health outcome. To meet the overall objective of the research, the following specific were developed: #1. To identify vulnerability to natural hazards at the county level by assessing the exposure to hazards, sensitivity (a potential of being affected by a climate stress) and adaptive/coping strategies using "climate and county socioeconomic and demographic data. #2. To explore the relationships between vulnerability and health outcomes at the county level. #3. To study the impact of vulnerability to drought and (a Mongolian term for winter disasters) on the economic well-being of rural households. #4. To investigate the relationship of vulnerability to drought and dzud and the health status of individuals within households. The study hypothesis can be phrased as follows: high levels of vulnerability -which is a composite measure of sensitivity household adaptive 3

PAGE 16

strategies, and exposure to natural hazards will be associated with low-levels of household well-being (economic and health). The results of this study will help to identify factors that are most important in disaster mitigation among pastoralist households in Mongolia. will assist in the implementation of programs to secure pastoral livelihoods. Ideally, these programs will enhance the general health status of past ora list households. The use of different levels of analysis (county and household) in the research will give a more comprehensive picture of vulnerability in Mongolia and a better understanding of the processes that shape vulnerability at multiple spatial scales. The dissertation is divided into seven chapters. Chapter 2 provides background information on Mongolia's demographic and socioeconomic situation during the transition period, ecological and climate peculiarities, pastoralism and adaptive strategies to overcome constraints of a physical environment, impact of natural disasters on health, and application of GIS and spatial statistics in health research. Chapter 3 develops the theoretical principles that guided the research, including the vulnerability paradigm and theories of political ecology, social capital and gender. Chapter 4 describes the methodology of the research. Predictor and outcome variables used for each level of the research, the sampling frame, study sites and settings, and data collection procedures are described in this section. 4

PAGE 17

Chapter 5 presents the results of the data analysis: spatial data analysis at the county level and multilevel data analysis at the household and individual levels. The discussion of results, conclusions and final recommendations follow in Chapter 6.

PAGE 18

CHAPTER 2 BACKGROUND Demographic and Socioeconomic Characteristics of Mongolia Mongolia is located in Central Asia between Russia and China. The territory is 1.567 million square kilometers, with a population of only 2.7 million people, which results in one of the wofld's lowest population densities-l.5 persons per square kilometer. Fifty two percent of the population are urban-dwellers, the majority which live in the capital ofUlaanbaatar. Of the 48% who are rural residents, the majority are nomadic pastoralists. The country is divided into 18 rural provinces and four cities. Each province is divided into 15-16 counties on average, and each county is divided into small townships with an average population of 500-600 persons (Figure 2.1). 6

PAGE 19

Figure 2.1. Administrative Map of Mongolia After seventy years of state-controlled, centrally-planned economy, in 1990 Mongolia undertook political and reforms in transition to a market-economy. These reforms consisted of the following elements: price liberalization, removal of restrictions on international trade and foreign investment, privatization of state-owned enterprises, and a marked reduction in government's involvement in the economy By most reports, these reforms resulted in widespread social chaos and economic collapse (Griffin, 2001). Aid to Mongolia from the former Soviet Union and other socialist countries was immediately eliminated, contributing to a sudden 30% drop in GDP. Technical assistance was withdrawn, the Council for Mutual Economic Assistance that had regulated trade between countries in the former Soviet 7

PAGE 20

block collapsed, and Mongolia lost import and export alliances. Industrial production, which had up to 1989 been a growing sector of the economy, declined sharply, resulting in widespread unemployment and poverty. As a result, the value of industrial production as a percentage of GDP declined from 35% to 20% between 1990 and 2000 (UNDP, 2001). The contribution of agriculture to GDP, mainly-livestock husbandry, increased accordingly as seen in Figure 2.2. 20 .. o 1 0) It) O) .... co 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) ..... .... .... .... .... Figure 2.2. Contribution of Agriculture to GDP (Griffin, 2001) The shock of poverty and high levels of unemployment urban areas led to widespread migration from the cities to the countryside as individuals sought to benefit from livestock privatization. Griffin called this process "ruralization." As the result, between 1989 and 1998 the proportion of the population living in urban areas declined by l3%, and the rural population grew by 17% (Griffin, 2001). Mongolia became once again a predominantly rural

PAGE 21

country. Since 1998, however, due to a string of severe winters which may have led to the failure of many inexperienced herders, along with a stabilization of the urban economy, population has begun to return to the towns and cities (see Figure 2.3). The western provinces experienced the largest levels of emigration from 1995 to 2000, eighty eight thousand and six hundred persons, and eastern provinces the smallest, twenty nine thousand and nine hundred persons (Bum 2003). 70 60 50 40 30 20 10 o v 0 0 0 0 0 0 0 0 N N N N Figure 2.3. Population Distribution (NSOM, 2004). Liberal economic reform radically transformed the structure and organization of the rural economy. The results have been a rapid dismantling of the old collective institutions-"negdel", privatization of all livestock and assets, disinvestment in social services, and poor access to markets (FernandezGimenez, 1998; UNDP, 2001). The collapse of the and the privatization of herds have eliminated the formal institutions that buffered

PAGE 22

economic and ecologic risk and protected the environment by regulating access to and maintaining land, pasturage and water resources; providing veterinary services; and establishing reserves of food, hay, warm clothes and medical supplies for use in emergencies (Baas, Batjargal, Swift, 2001; Griffin, 2001; UNDP,2001). Indicators of social and health services also show significant deterioration. Maternal mortality rates have been consistently high over the last ten years (about 200 deaths/lOO,OOO livebirths compared to 120 deaths/lOO,OOO livebirths in 1989). The majority of deaths occur in rural counties and these women were most likely to be herdswomen, those with less education, or ones from poor families with little social support (Chuluundorj, 2001; Janes Chuluundorj, 2004). The incidences of infectious diseases such as tuberculosis, sexually transmitted infections, infectious hepatitis (esp. hepatitis B virus and hepatitis C virus), and brucellosis have grown disproportionately in the poor segment of the population in both rural and urban areas (Government of Mongolia UNDP, 2004). There has also been a rapid increase of non communicable diseases such as cancer, cardiovascular diseases and injuries (Foggin, Farkas, Shirev-Adiya, Chinbat, 1997). Although the above mentioned socioeconomic changes occurred everywhere in the country, different regions and provinces faced unique challenges depending on environmental conditions, local economic resources, access to domestic and foreign (e.g., Russia, China) markets, and infrastructure development. Consequently, the economic rewards of pastoralism vary greatly 10

PAGE 23

from one region to another (Humphrey Sneath, 1999). Concomitantly, disparities also exist in the health status of a population in different geographic regions: e.g. the maternal mortality rate is consistently highest in the economically disadvantaged western provinces (Demberelsuren Dorjpurev, 2000) However, little is understood about the impact of ecological differences between different regions of the country and their impact on a local population's socioeconomic well-being and health. This is one question addressed in this study Ecology and Climate There are three different ecological zones within the territory of Mongolia: the forest zone with high annual standing vegetation (3500 to 4000 kg/ha), the grass steppe or grassland zone with a moderate annual standing vegetation (1500 to 3000 kg/ha), and the desert zone with low annual standing vegetation (375 to 1500 kg/ha). Grassland is the transitional ecosystem between forest qnd desert ecosystems: it occurs where the precipitation is too low to maintain forests, but high enough to prevent desertification (Government of Mongolia, 2001; Moran, 2000). A map of ecological zones is shown in Figure 2.4. 11

PAGE 24

Desert Steppe Forest Figure 2.4. Ecological Zones Climatic conditions fluctuate greatly from cine ecological zone to another. The precipitation varies among ecological zones and is confined mainly to summer: 100 mm/year in southern Gobi areas and up to 500 mmlyear in northern mountainous forest regions. The majority of the country receives on average 350 mm of precipitation per year (Government of Mongolia, 2001). Summer rainfall has declined between 1970-1990 in the Gobi, and the number of heavy rainfall events has fallen significantly (Lal, Harasawa, Muriyarso, 2001). Droughts have been more frequent in the 20th century, and may be linked to the climatic consequences of global warming. 12

PAGE 25

The land is covered by snow for seven-to-eight months per year, which results in a growing season of only four months. Temperatures can easily reach -40C in mountainous region during winter. The annual mean surface temperature in Mongolia has increased by 0.7 degrees Celsius over the past 50 years (Lal et aI., 2001).The difference between night and day temperatures is high, frequently reaching 20-30C. The combination oflow precipitation and high temperature fluctuation make agricultural production risky: only one percent of total land in Mongolia is classified as arable (UNDP, 2002). Given these ecological and climatic constraints, pastoralism is the most efficient means of production in Mongolia (Moran, 2000). Because the amount of energy that can be extracted via animal herding from a unit of grassland is small, pastoralists require high mobility and access to a large territory of grazing land in order to reduce the risk from droughts and unstable weather conditions. These demands vary acr
PAGE 26

being of a large number of rural people whose livelihoods are highly dependent on environmental resources (Corvalan et aI., 2005). Mongolian pastoralists face several environmental hazards (source of danger that may potentially harm individuals or groups): winter disaster or drought, wild fire, predation, damage caused to vegetation by large rodents, and animal diseases (Baas et aI., 2001; fIaddow Bullock, 2006; Pelling, 2003). Among these, droughts and are the greatest threat to subsistence (Templer, Swift, Payne, 1993). A can be defined in many ways, meteorologists define a as snow cover of more than 25 cms, a sudden prolonged snow storm, or prolonged extreme cold. Mongolian herders typically distinguish between several kinds of A "white dzud" occurs when deep snow cover prevents animals from grazing. A "storm dzud" occurs when severe weather conditions drive animals downwind, and they get lost and freeze to death. "Freezing dzuds" occur when temperatures fall sharply and animals cannot maintain their body temperature (Morinaga, Tian, Shinoda, 2003; Swift, 1999). Regardless of the specifics of definitions, any of these events is a real threat to livestock loss. A sequence of summer drought and winter is the most devastating because animals who do not gain sufficient weight during the summer easily perish even in mild winter conditions. 14

PAGE 27

Drought is a period of several months or even years of abnormal dryness due to below-average rainfall that causes a pronounced decrease in forage yield relative to what is expected in an average year (Rothauge, 1998, p. 1). Drought is a slow-onset hazard and can affect large areas. In fact, the with the greatest animal loss were all preceded by summer droughts: the of 1944-1945, 1967-1968 and 2001-2002 (Table 2.1). Table 2.1. years and animal losses (Hoohdoi, 2002). Animal loss Animal loss Year Affected Places in thousand as % of total heads livestock 1944-1945 9 provinces, 65% of total land 8638.0 35.5 1954-1955 No information available 1887.7 8.2 1956-1957 No information available 1008.0 4.1 1967-1968 13 provinces, 80% of total land 3800.0 17.0 1976-1977 15 provinces, 90% of total land 1453.9 6.1 1993 spring 3 provinces, 30 soums 689.5 2.7 1996-1997 11 provinces, 69 soums 700.0 2.4 1999-2000 13 provinces, 158 soums, 2614.0 10.0 70% of land 2000-2001 17 provinces, 98 soums 4800.0 18.2 2001-2002 20 provinces 3400.0 9.5 15

PAGE 28

Some actions have been taken during by the Mongolian government to help households in dzud-affected areas. These include the distribution of fodder, vegetable seeds to encourage agriculture, dietary diversification and greater food supplies, and support of the health care system with equipment and drugs (Norovlin et aI., 2003). These efforts seem insufficient to reduce herd loss and prevent adverse health outcomes in the pastoralists. One study conducted in Mongolia revealed that the prevalence of growth stunting was significantly greater among children aged 6-23 months in dzud-affected areas than in unaffected areas: 38.3% versus 26.0%, p=0.04 (Norovlin et aI., 2003). The effect of such climatic stress on the socioeconomic conditions of pastoralists is tremendous. Loss oflivestock directly translates to loss of financial security, marginalization and impoverishment. Importantly, the loss of livestock may also lead to loss of social ties and the support that flows through these channels. Pastoral herding requires a high reciprocity and families with little or no means to engage in mutually beneficial relationships may be left out. Families may also move to more urban areas to make a living and become detached from their kinand place-based social networks. These families comprise a large percentage of the urban poor. "The essential pastoral strategy is probably neither maximization nor optimization but risk aversion, an attempt to decrease uncertainty by anticipation" (Galaty Johnson, 1990). As the quote states, avoiding, anticipating and coping with these negative climatic stresses or their 16

PAGE 29

consequences is the primary strategy of pastoralists. Understanding the importance of these strategies and developing policies to enhance the capacity of pastoralists gain more importance: the occurrence of extreme climate events is likely to increase in the future and bring threats to pastoralist livelihood more than ever (Corvalan et al., 2005). Specific adaptive and coping strategies will be delineated later, but before that an .introduction to pastoral livelihood system is necessary. Pastoralism in Mongolia Changes that occurred in the pastoralist system in 20th century in Mongolia are quite remarkable. In the early 20th century, livestock was the property of a landed aristocracy: local lords and monasteries. The majority of the population were poor and subject to exploitation by the elite. The coordination of access to pasture and seasonal migration was done effectively by the aristocracy to avoid overgrazing and secure water sources. Beginning in the 1940s, the process of appropriation of private animals by the state began. The process was completed in early 1960s: the state firmly established its control over all resources, including all animals, through organizing collectives and state farms. The state organized the livestock production system down to the household level and controlled the use of natural pastures with a goal to achieve a maximum use of natural resources without overusing them. Herders became salaried employees of the state and the state 17

PAGE 30

provided food items at subsidized prices, health and social services free of charge, free education to children, animal shelter, veterinary services, hay, and extra labor when needed. Politico-economic processes occurring during the transition that began in 1990 brought new challenges into livestock production in Mongolia. Livestock was given back to people, creating small household-level units. The privatization of livestock resulted in an increase to the pastoralist population: many urban dwellers moved to the countryside to benefit from receiving livestock shares. In 2000, 191,526 households (about 40% of total population) made their living primarily from pastoralism compared to 74,710 in 1990 (Government of Mongolia UNDP, 2004). The importance oflivestock production to Mongolia's economy increased over this time period: it accounted for 83% of total agricultural production in 2003 (Mearns, 2004). Although livestock has been privatized, land is still common property. Having common land is a necessary condition when the productivity of pasture is highly variable. allows herders access to large territories to reduce risks of inclement weather (Gilles Gefu, 1990). However, the emergence of many small-scale private herd owners has decreased rotational access to pastures. This is due to increased competition over the best pastures in the absence of supra-household regulatory mechanisms. Without institutions to govern shared access to better pastures, overgrazing and poor access to good pasture and water source have become a real problem. Poorer households or households with fewer or weaker social ties are cut off from best resources (UNDP, 2002). 18

PAGE 31

As more and more people have become involved in pastoral production, several trends have been identified: 1. The number of total livestock has increased in the last decade (Figures 2.5), but the majority of herding households have relatively few animals. The average number oflivestock per household declined from 346 in 1990 to 158 in 2000. The percentage of households with fewer than 100 animals was 58% in 1999, 63% in 2000,67% in 2001, and 69% in 2002. In 2002, 88% of all herding households had fewer than 200 animals (Ravsal, 2003). A recent study by the Ministry of Finance and Economy on herd restocking strategies has concluded that the number of livestock for a reasonable living for an average family of 4-5 people would range from 200 to 300 animals (Mearns, 2004). Thus, there has been an increase in poor herding households over the last decade. At the same time there is an emerging group of wealthy herders in the country who owned more than 2000 animals (Government of Mongolia UNDP, 2004). 2. There is a change in the structure of herds. The total number of sheep remained constant throughout the decade, but the number of goats increased by over 200%. The percentage of goats in the national livestock population increased from 20% in 1990 to 42% in 2003, whereas the percentage of other livestock declined by 1-16% (Figure 2.5). The increase in the number of goats is likely due to the market value of cashmere, the most valuable of the animal products traded by rural herders. 19

PAGE 32

+-_______ "I Q) 10 r.... Q) 10 co Q) Q) Q) Q) Q) Q) Q) Q) Q) Q) Q) C'\I C'\I C'\I Figure 2.5. Livestock by Species, 1989-2005 (NSOM) The declining numbers of camels, horses and cattle in the total national livestock population is attributed by herders and livestock experts to the use of mechanized transportation rather than camels, the slaughter of camels due to the increased price of camel meat (180-200$ per camel), an increased price for horse hides (20$), theft of horses in far pastures, and a loss of horses and cattle to dzud and drought. Maintaining the right ratio of different species in a herd is extremely important to effective pasture use, and experienced herders attribute pasture degradation to too many goats (Grayson Baatarjav, 2004). 3 Disappearance of economic/market services in rural areas contributed to market failure in the countryside and overgrazing of pastures close to cities and province centers. Herders move to be closer to cities and towns where they have opportunities to sell their products resulting in higher livestock-to-pasture ratios in these regions and causing pasture degradation. 20

PAGE 33

Thus, economic liberalization has enabled some individuals and households to increase their profits from livestock significantly, yet this has produced dramatic increases in inequality between the richest and the poorest segments of the rural population as some are more able than others to respond to increased market opportunity and climate stress, and significant pressure on natural recourses. Households with a few animals, young and inexperienced herders, single-headed households, households with insufficient labor (high dependency ratios), households who entered herding production to benefit from privatization and are new to herding are particularly at higher risk of impoverishment and social marginalization (Cooper, 1993). Adaptation The chief problems presented by grassland ecosystems center around the exploitation of water and pasture, the composition and size of herds, the establishment of workable relationships between pastoralists and agriculturalists, and the achievement of a balance between the human and the animal population under conditions of great climatic uncertainty (Moran, 2000, p. 220). The many adaptive strategies used by Mongolian herders focus on avoiding hazardous natural events, maximizing the use of natural resources, increasing their income, and securing their livelihoods. The adaptive strategies 21

PAGE 34

delineated below are divided into two groups: traditional and non-traditional or emerging. The traditional adaptive strategies have been practiced for centuries and are common among pastoralist populations in the world. The non traditional adaptive strategies have emerged recently in Mongolia as a response to socioeconomic and political changes that have occurred over the last 15 years. Traditional Adaptive Strategies Low Livestock DensitvlLow Settlement Density. Low livestock and human densities are crucial for successful herding. Having a larger area to graze during different seasons ensures availability of good pasture throughout the year. Mongolia has one of the lowest population densities in the world-1.5 persons per square kilometers. Seventy nine percent of Mongolia's total land (1567 million square kms) is under pasture (Fratkin, 1997). However, since privatization herding households have moved to central provinces where the market opportunities are better, resulting in higher popUlation and livestock densities in those areas. High Mobility. Low livestock and human density is the condition for high mobility. Pastoralists move every season. In fact, there are several different types of movements practiced by pastoralists. These include seasonal migration --moving between ecological zones to benefit from different stages of 22

PAGE 35

vegetation cycle; middle range move -between two pasturelands; and short range camp site transfer --movement within one pasture (Schareika, 2003). The high mobility of herding households has decreased with the privatization of lIvestock mostly due to unavailability of transportation means, lack of labor and unavailability of good quality pastures (Fernandez-Gimenez, 1998; Foster, 2003). Decline in herding mobility leads to an increasing conflict over pasture usage and water sources. Increased pressures on pasture eventually will result in overgrazing and pasture degradation. Intensive grazing of2-3 periods on the same area in a season may lead to a decline in pasture yield of up to 72% the following year (Tserendash Erdenebaatar, 1993). According to a recent study, 1.4 % of total pasture land in Mongolia is degraded very seriously (50% or more), 20.7%--seriously (30% to 49%), 50.8%--mildly (20% to 29%), 25.4%--minimally (10-19%) and 1.7%--not degraded (less than 10%). The problem of pasture degr:adation is especially serious in central provinces where the livestock and human density is much higher (Marriott Erdene-Ochir, 2004). The importance of transportation in mobility is enonnous. Wealthier families have been able to acquire motorized transportation which has enabled them to transfer, their livestock in case of natural disasters; gives them access to markets for purchasing food and products and selling animal products; and allows them to become involved in alternative economic activities to supplement income from pastoral production. 23

PAGE 36

Large Number of Livestock. is the primary goal ofpastoralists to increase the number of their livestock. The Mongolian climate is defined as a non-equilibrium system: it is characterized by a great fluctuation oflivestock numbers as a result of unpredictable and uncontrollable natural hazards that periodically decimate herds. Livestock numbers increase rapidly for some period of time after a disaster, then declines sharply as a result of a natural disaster. Having a larger herd means better security in times of disasters. Wealthier households may lose more animals, but they will also have more animals left to rebuild a herd. But poor families are likely to be left with number of livestock below the minimum subsistence level and may not be able to recover from such losses in the longer term. In addition, livestock defines wealth and wealthier households are able to procure access to better pastures and water sources, especially if wells need to be dug and maintained, and enjoy greater leisure time and overall higher prestige (Begzsuren, Ellis, Ojima, Coughenour, Chuluun, 2004; Goldstein Beall, 2002; Mulder Sellen, 1994). Increased Number of Species in a Herd and Breed Selection. Multi. species herds have many advantages: better use of pasture due to different foraging habits of different species, different species provide diversified products for consumption and sale, different mortality of different species ensures that some animals are left after natural disasters (cattle and horses 24

PAGE 37

perish first, goats are the last to die in Purchase of highly adaptive breeds may also reduce the loss during disasters. Household Splitting. In times of droughts or winter disasters families split their herd for some period of time to benefit from better pastures and favorable climatic conditions. This is an option for families who have enough people to split and live separately for a few months (Mearns, 2004). Large Family Size. Pastoralists tend to have larger families compared to urban dwellers. Having a large family guarantees labor availability, more social ties and more resources flowing through them. In Mongolia, the total fertility rate was 3.66 in rural areas compared to 2.46 in urban areas according to the Reproductive Health Survey of 1998 (NSOM UNFPA, 1999). Large Social Network. Having a large social network is an important asset to the herding household. Households cooperate with others on many of their difficult tasks, including moving, sheering, combing, c"!ltting hay, and fetching water and fire wood. Hospitality and reciprocity facilitate mobility and help herders gather needed information. Creating alliances across different ecological zones is also crucial to secure better pastures in times of droughts and (Galaty Johnson, 1990). Waller and Sobania (1994) noted that in Africa "asociality" was equal to poverty. Having a large social network may not be as important to wealthier households as it is to poorer ones. Cooper (1993) 25

PAGE 38

recognized the importance of expanding social ties to sustain the labor and consumption requirements for the poor. Wealthier households can purchase labor or services whenever they need them, but the poor rely on others' generosity for these services. Large Percentage of Female Animals. Female animals are the foundation for building assets for pastoralists. addition, they provide milk that is a necessary part of the diet for herders (Chen, 1991). NonTraditional Adaptive Strategies Increased Cash Income. One of the changes that occurred with the introduction of the market economy was an increased value of cash for rural herders. Sixty two percent of income of herding households comes from the sale of wool and cashmere, and the rest from a sale oflivestock (Government of Mongolia UNDP, 2004). Converting surplus livestock products to cash means access to health care and other services when needed, ability to purchase transportation to enhance mobility, better supply of cereals and other food items that are not produced by herders, and good education for their children. Substitution of hand-made products by manufactured products further deepens the dependency on markets and cash. Market opportunities to sell livestock are not the same across the country. Herders closer to cities, borders and trade 26

PAGE 39

centers benefit from easier access to markets and higher prices. Herders in remote locations must often sell their products to middlemen at lower prices. Livelihood Diversification. Pastoralists engage in various activities besides livestock husbandry to increase their income and support their subsistence. Some of these activities are hunting, begging, farming of vegetables, gathering wild foods, sewing, knitting, and animal theft. Theft of larger animals such as camels, horses and cattle is common because these animals graze farther from the camp and are not usually tended by herders on their pastures. Diversification may not be a risk-averting strategy for the poor: mainly they do it for survival. But middle-wealth and rich families do it to minimize risk or accumulate more wealth (Little, Smith, Cellarjus, Coppock, Bartett 2001). Cooperation through Informal Camps and Neighborhood Groupings (Khot Ail). In the absence of institutional support, herding households form small groupings to share labor. They each take turns to tend animals on pastures, milk animals, process dairy, or go to markets to sell or trade. If a household has to go to the province center to seek health care, other households can take care of their children and animals. Often these groupings are based on kinship relationships, but friendship may also provide a basis for affiliation. 27

PAGE 40

Livestock Insurance. There are only two companies that offer livestock insurance at the current moment: Mongol Daatgal company and one other state owned company. Livestock loss is covered at 100% but the premiums are set at six percent (of the cash value of livestock) which is expensive for pastoralists. There is no regional variability in the premiums. But the current insurance was not marketed aggressively and herders' knowledge how insurance works is very rudimentary. Other forms of livestock insurance are being discussed, including weather-based and mortality-based insurances. The feasibility of these schel11es remain to be tested but these are likely to offer more benefits than traditional livestock insurance (Skees Enkh-Amgalan, 2002). Involvement in Wage Labor. Often families with many adult members send one or two persons to province centers or cities to secure permanent jobs. Having a job means cash income. also creates rural-urban linkages to promote exchange of goods and services. Patron-Client Relationships. These are linkages between better-off and poor households, with little or no reciprocity. This type of relationship is profitable to wealthier families: they benefit from extra labor the client can provide. In return, the patron grants a small salary and free dairy products and meat to the client. may be helpful to poorer families who otherwise would have moved to urban areas where their livelihood may be even worse. However, this patron-client relationship does not offer any opportunities to 28

PAGE 41

clients to build their own livestock herds and secure their livelihoods. They remain highly dependent on the willingness and ability of their patrons to support them. Access to Credit. Having access to any of type of formal (banks, in-kind through restocking) and informal (relatives, friends, kiosks, pawnbrokers, money lenders including cashmere traders) credit is important, especially for poorer households who do not have or have very little cash income. Households who have lost all of their livestock to natural disasters may receive animals from their relatives and friends. If there is an urgent need to borrow money to cover, for example, medical expenses, having a reliable source for loans may make all the difference. Animal restocking programs have been introduced in five provinces (Dundgobi, Bayankhongor, Zavkhan, Ovorkhangai and Uvs) as an experiment. Households are given cash in the amount of one million tugrics (an equivalent of less than 1000 USD) for five years (with an interest of six percent paid starting the third year) and they purchase animals from local herders. Criteria for receiving such grants include: loss of an entire herd or being left with a few animals not sufficient for subsistence; possession of a permanent winter shelter; a multi-generational herder, no criminal history, and agreement of all household members to the conditions of the award. Households new to herding or young and inexperienced herders are not able to benefit from this program (Hoohdoi, 2002). 29

PAGE 42

Reducing Consumption. Other strategies, such as reducing consumption or switching to cheaper inferior food, have also been observed (Chen, 1991; Fratkin Roth, 1996; Little, 2002). Siurua and Swift (2002) have found that reducing consumption during disasters was a common approach among pastoralists in affected areas. Migration to Urban Areas. Population migration from cities to rural areas from 1990 to 1995 was reversed in the second half of the 1990s. Families who lost all of their livestock have moved to province centers or cities to seek employment. According to the statistics from the National Statistical Office of Mongolia, 'one third of the total population in Ulaanbaatar is migrants .from rural areas (UNDP, 2003). This is the most extreme form of adaptation, which often worsens socioeconomic conditions of households and the health of individuals. Absentee-Herding. Some pastoralist households herd animals of other households (absentee-herders) who live in urban areas. It can be useful to both sides: the herding household may receive services from the absentee-owners in return for tending their animals. For absentee:-herders, having their own herd on pastures means availability of dairy and meat year around (Fernandez-Gimenez, 1999). In summary, Mongolian pastoralists engage in various .activities to buffer their economic loss during disasters and secure their subsistence. Little is 30

PAGE 43

known about the relative importance of these strategies on households' level of well-being. This is the major question addressed in this thesis. Table 2.2. Summary of adaptive strategies Traditional Low livestock density/Low settlement density High mobility Increased number of species in a herd and breed selection Large number of livestock Household splitting Large family size Large social network Large percentage of female animals NonTraditional Increased cash income Livelihood diversification Cooperation through informal camps and neighborhood groupings Livestock insurance Involvement in wage labor Patron-client relationships Access to credits Reducing consumption Migration to urban areas Absenteeherding Natural Hazards and Health The number of natural hazards, such as floods, droughts, wildfires, winter storms and extreme heat, is likely to increase in the future as the result of an increased accumulation of greenhouse gases in the lower atmosphere and stratospheric ozone depletion, and these natural events pose a serious threat to the human livelihood (Burton, Kates, White, 1993; Haddow Bullock, 31

PAGE 44

2006). The mechanisms through which these natural hazards affect health vary depending on nature of the events. Flood is the most widespread and the most costly natural hazard. The direct damages include drowning and injury of humans, food insecurity due to the devastation of animals crops and natural resources, and disruption of roads to transport people to safety supplies. The indirect damage includes outbreaks of waterborne infections arising from the contamination of drinking water by sewage and vector-borne infections such as malaria because of increased reproduction of mosquitoes in stagnant water, and spread of harmful chemicals such as pesticides (Blaikie, Cannon, Davis, Wisner, 1994; Hewitt, 1997). Reduced vegetation growth and death of livestock from dehydration during droughts endangers the food security of populations. Reduced flow of water in streams, lakes and wells are likely to put limitations on water usage and worsen sanitary conditions. Poor water quality, changes in its salinity, and accumulation of toxic substances may eventually have a major impact on the health of human populations (Hewitt, 1997). The impact of on human health can also be direct and indirect. The direct effect includes hypothermia, frostbite, and injuries due to motor vehicle collisions or collapses of structures (Hewitt, 1997). The indirect effects such as poor nutrition due to high mortality of livestock and inaccessibility of roads may impact health slowly. In any natural hazard a variety of problems may arise as the result of displacement in case of any natural hazard such as housing in unsafe places, 32

PAGE 45

unsafe water and sanitary conditions, poor mental health, and poor nutrition (Hewitt, 1997; Kovats Bouma, 2002). The significance of these natural events is growing: the number of people killed, injured and thrown to poverty due to natural hazards is swelling (Kovats, Menne, McMichael, Corvalan, Bertollini, 2000). The relationship between hazards and certain health outcomes are not as straightforward. is attributed to a multiple causality of diseases, diversity of diseases, prolonged effects of hazards, and great variability in the manifestation of natural events (Kovats et aI., 2000). In addition, hazards do not affect everyone in the same manner. Individuals, households, regional and/or national capacities to avoid, reduce or cope with environmental stress or its consequences buffer negative effects of hazardous events on human health and well-being. Those who have the fewest. resources to avoid natural events or their negative impacts are at most risk of becoming impoverished and sick. The loss of human life and loss in property are the highest in developing nations (Lal et aI., 2001). The health of children is especially sensitive to natural hazards. They are a biologically vulnerable group and suffer from those diseases likely to be caused or exacerbated by reduced income and/or reduced public medical care. Nutritional deficiency, which can be caused by a deficiency in vitamin, mineral, and the energy/protein content of food, perhaps the most commonly used health outcome in studies exploring the relationships between major environmental 33

PAGE 46

events and their health consequences (Konnondy & Brown, 1998; Musgrove, 1987). In summary, Chapter 2 described main socioeconomic and sociodemographic processes that have occurred in Mongolia during the transition period, changes in pastoral livelihoods as the consequence of these processes, strategies employed by Mongolian pastoralists to overcome ecological, political and socioeconomic challenges, and the effectiveness of these strategies to buffer negative environmental impacts. The following chapter introduces the theoretical framework and main theories employed in this study. 34

PAGE 47

CHAPTER 3 THEORY Vulnerability Paradigm Natural Disaster The impacts of natural hazards are likely to be felt more severely in developing countries than in developed countries, irrespective of the magnitude of climate change, because of the poor resource and infrastructure bases found in such countries. In countries where weather-sensitive production systems such as growing and livestock husbandry contribute significantly to the national economy, natural hazards are major threats to the livelihoods of people, particularly of those who are poor. In Mongolia, where livestock husbandry contributes 20 percent to the GDP and is the main source of subsistence for rural Mongolians (constituting 40 percent of the total population); the impacts of natural events such as winter storms and droughts are dramatic and potentially devastating (NSOM, 2004). 35

PAGE 48

Winter storms and droughts are not new phenomena to experienced Mongolian pastoralists. But their impact on human livelihoods and health has changed dramatically during the transition from a state-oriented to a market oriented economy that began in the early 1990s. Under the socialist system, livestock was a state property and herders were hired employees ofthe state. The latter provided salary, health and social services. During the privatization of the early 1990s, all livestock was divided among herding households. Each household received approximately 50-60 animals on which they had to rely on for subsistence. A summer drought and a winter stOlm became a real threat: one winter storm could swipe away all animals and leave families with nothing. This experience of rural herders in Mongolia is a classic example of how natural disasters are born. Blaikie et al. (1994) defined a natural disaster as the sum of a hazard or natural event and vulnerability. disaster occurs "when a significant number of vulnerable people experience a hazard and suffer severe damage and/or disruption of their livelihood system in such a way that recovery is unlikely without external aid". The disruptions may occur on mUltiple scales, from individuals' health and mental distress to local andlor national socioeconomic downfall and failures in the global political economy (Pelling, 2003). Natural events do not cause natural disasters. They are a necessary part of natural disasters, but they most often act as triggers of underlying social inequality and social forces, or in other words, a disaster is a consequence of vulnerability (Pelling, 2003). Vulnerability is a social process. 36

PAGE 49

Vulnerability The term vulnerability has its origins in food security literature, and has been applied in the last decade or so to research on climate change and its effects (Vincent 2004). The concept is very much in its initial stage of formulation and there is no all-agreed definition of vulnerability. However, the main elements in the definition of vulnerability have evolved as seen in Table 3 .1. Table 3.1. Definitions of vulnerability (Weichselgartner, 2001) Authors Gabor and Griffith (1980) Susman et al. (1983) Smith (1992) Cutter (1993) Watts and Bohle. (1993) Definition The threat (to hazardous materials) to which people are exposed (including chemical agents and the ecological situation of the communities and their level of emergency preparedness) The degree to which different classes of society are differentially at risk Human sensitivity to environmental hazards represents a combination of physical exposure and human "vulnerability-the breadth of social and economic tolerance available a the same time The likelihood that an individual or group will be exposed to and adversely affected by a hazard Vulnerability is best defined as an aggregate measure of human welfare that integrates environmental, social, economic and political exposure to a range of potential harmful perturbations. Vulnerability is a multilayered and multidimensional social space defined by the determinate, political, economic and institutional capabilities of people in specific places at specific times. 37

PAGE 50

Table 3.1. Definitions of vulnerability (Cont.) Authors Blaikie et al. (1994) IPCC (2001) Definition The characteristics of a person or group in terms of their capacity to anticipate, cope with, resist and recover from the impact of a natural hazard The degree to which a system is susceptible to, or unable to cope with, adverse effects of climate change, including climate variability and extremes. Vulnerability is a function of the character, magnitude, and rate of climate variation to which a system is exposed, its sensitivity, and its adaptive capacity. The development of theories of natural disasters has its roots in the early 1960s with Burton and Kates' work. They described natural disasters as environmental events that can cause harm to humans (Pelling, 2003). Vulnerability has been treated as an increased exposure to natural hazards: living in zones where a hazard is likely to occur. This construct is known as biophysical vulnerability. The frequency, intensity, duration of natural events and the extent of damage were the main focus of studies that use the definition of biophysical vulnerability. In this literature, reducing an exposure to a hazard by introducing technological innovations becomes the key to preventing or mitigating disasters. O'Brien named the construct end-point vulnerability (Adger, Brooks, Bentham, Agnew, Eriksen, 2004; Gilbert, 1995; O'Brien, Eriksen, Schjolden, Nygaard, 2004). White and Haas (1975) postulated that people continue to live in regions which they know to be hazardous hoping that the benefits of oGcupying the area will outweigh potential losses. Individuals were seen as active agents who can 38

PAGE 51

make choices about potential risks of disaster. But the environment was still seen as the major force that shapes the outcome of a disaster. This environmental deterministic view dominated the field until 1983, when Hewitt published his seminal work, "Interpretations of calamity: from the viewpoint of human ecology" in which he described natural disasters as an outcome of the society-nature interaction (Pelling, 2003). Hewitt argued that human responses to natural hazards are not dependent on the natural events, but are shaped by the social order, institutions and historical circumstances present in the society (Tobin Montz, 1997). Following Hewitt's work, researchers started to use vulnerability in a different sense, and people's capacities to avoid, resist and recover from hazards became an essential part of vulnerability. In this literature, vulnerability is seen as a state that exists within a system before it encounters a hazard event and is exacerbated by it (Brooks, 2003). All social and economic processes leading to marginalization, poverty and inequality determine the outcome of a hazard event. The way to decrease vulnerability involves changing structures and institutions that govern human lives and consequently social and economic processes. This type of vulnerability is known as social vulnerability or vulnerability as a "start-point" (Gilbert, 1995). Some authors define biophysical vulnerability as a function of hazard and social vulnerability (O'Brien, Eriksen, Schjolden, Nygaard, 2003; Vincent, 2004). Thus, it is the interaction between the structure and the individual(s) that ultimately defines vulnerability (pelling, 2003). Adaptive processes that occur 39

PAGE 52

at larger scale-nation or world-are necessary to make resources available at the smaller scale-local or individual. Yet it is an individual's action( s) that define in some ways the mechanisms to employ these resources for their own well-being (Hewitt, 1995). The disaster pressure and release model defined by Blaikie et al. (1994) summarizes the stages of vulnerability in a very useful way. Root causes such as limited access to power, structures and resources, and ideologies of political and economic systems create dynamic pressures on individuals. These pressures can be lack of local institutions, training, appropriate skills, local investments and markets, press freedom and ethnical standards, and macro level forces such as rapid urbanization and population growth, arms expenditures, debt repayment schedules, deforestation and land degradation. All these dynamic pressures can create unsafe living conditions for humans: dangerous locations and infrastructure, poverty and livelihood insecurity, vulnerable groups, and insufficient disaster preparedness. And the presence of unsafe conditions in the presence of hazard events results in disaster. There is a broad array oftemls used to describe vulnerability. Resilience is another term used in contrast to vulnerability. is defined as the ability of an actor to cope with or adapt to hazard stress and return to the previous stable condition without incurring any long-term negative consequences (Lal et aI., 2001; Pelling, 2003). Sensitivity, susceptibility, resistance, capacity, and potentiality are some of the constructs used to describe vulnerability. But the following are the 40

PAGE 53

components of vulnerability that are most commonly used in the scholarly literature: 1. Exposure the degree of climate stress. Exposure can be both short term and long-term. Frequency, duration and intensity are some of the features of exposure. 2. Sensitivity the degree to which the system is affected by climate stress directly or indirectly and negatively or positively (White et aI., 2001) Some authors use "resistance" instead of "sensitivity to describe the individual's or group's capacity to withstand the impact of a hazard (Pelling, 2003). Methods to reduce sensitivity can be a change of economic, social and political circumstances. The number of possible hazards is large and the manifestation of a particular hazard varies. Consequently, sensitivity should not be viewed as general and applicable to all hazardous events. Sensitivity is a hazard-specific phenomenon and an acknowledgement of its unique features will help to define successful policies to prevent or mitigate disasters. 3. Adaptation or coping an adjustment of a system to climatic stress, potential damages, or consequences (O'Brien, Sygna, Haugen, 2004). The function of adaptation is to reduce social vulnerability and promote sustainable development by making changes in ecological, social, and economic systems (Smit Pilosofa, 2001). Adaptation depends greatly on the adaptive capacity or adaptability of an affected system, region, or community to cope with the impacts and risks of climate change. Adaptability includes technological options, availability and access to resources, human and social capital, decision-41

PAGE 54

making process and structure of critical institutions (Brooks, 2003; O'Brien et aI., 2004). As in the case of sensitivity, adaptation is hazard-specific; it varies depending on exposure (Leatherman Thomas, 2001). In summary, there are several important features of vulnerability: Vulnerability is variable (contextual): it varies across geographical space and social groups. Vulnerability is scale-dependent: vulnerability at the local level may not be the same as vulnerability at the regional and/or national levels. Vulnerability is dynamic: it varies depending on changes in social structure and forces over time (Hewitt, 1995). The vulnerability paradigm is appropriate for studying the risks Mongolian pastoralists face today and the impact of natural hazards on their livelihoods. As mentioned earlier in this chapter, severe climatic events were always a threat to nomads Mongolian grasslands for centuries. Political and socioeconomic changes that have occurred in the country since the early 1990s significantly challenged their lives. Insecurity, poverty, and lack of state support became a reality of everyday life. O'Brien and Leichenko (2000) called this situation a "double exposure": a population confronted by consequences of both climate and social changes. As described in the previous chapter, var'ious adaptive strategies have surfaced to cope with severe climatic hardships. Understanding the impact of these adaptive strategies in reducing vulnerability to hazards and their outcomes 42

PAGE 55

is important in securing livelihoods of pastoralists in Mongolia. The results of this work might be usefully applied in other contexts. Theory of Political Ecology Many vulnerability studies draw on theories of political economy and political ecology in explaining the factors that lead to vulnerability, and on social capital as a means of claiming access to resources and pursuing coping mechanisms (O'Brien et aL, 2004; Olmos, 2001; Pelling, 2003). The physical and biological environment we live in today is "politicized": environmental problems cannot be understood in isolation from their political and economic contexts within which they are created and/or exacerbated (Cutter, 1996). Both political economy and political ecology explore the role of power relations in human uses of the environment. Exposure, sensitivity and adaptation/coping are all defined by individuals' access to resources and assets. The access is rooted in global political and socioeconomic structures (Brooks, 2003; Pelling, 2003). But political ecology offers more to the objectives of this research. Political ecology as opposed to political economy, "seeks to understand the complex relations between nature and society through a careful analysis of what one might call the forms of access and control over resources and their implications for environmental health and sustainable livelihoods" (Watts, 2000). 43

PAGE 56

Leatherman and Thomas (2001) point to two important things that need to be incorporated into research on human-environmental interaction: social relations and environmental and social contexts. "Relations of power" or access to resources through social relations are key elements that define livelihoods of individuals and households, their exposure to different exposures and coping abilities. On the other hand, the environmental context within which these relations of power take place shape local context, the latter is important in social relations (O'Brien et aI., 2004). In summary, the theory of political ecology emphasizes the following premises: The local environment is shaped by the processes occurring at the higher levels. It is important to understand how the manifestations of globalization affect adaptive capacity within localities. Human-environment interaction is shaped by these local circumstances. Humans are active agents trying to alter their local environment and manage environmental risks (Moran, 2000). In Mongolia, structural adjustment programs implemented by the International Monetary Fund (IMF) and World Bank introduced price' liberalization, privatization of state enterprises, and removal of restrictions on international trade in early 1990s (Janes Chuluundorj, 2004). Aid from the former socialist countries, mainly the Soviet Union, had ended, leaving industries without electricity and agricultural machines without fuel. 44

PAGE 57

Unemployment, poverty, hunger, increase in crime and alcohol abuse were the major consequences of these policies. For rural herding households, these changes at the macro level resulted in a lack of support from the state that was available in the past in the form of human resource, transportation, hay, fodder, and health care and livestock replacement in case of natural disasters and loss. This dramatically increased their vulnerability to summer droughts and winter storms. Survival in the short term became the primary concern of pastoralists, and for many, exhausted all of their resources. In response to these macro and household level changes, securing access to resources through social ties and improving knowledge and skills to accumulate assets has become more important than ever. Those households who had social ties to both rural and urban areas, within the local government, and to health care and social services had better opportunities for trade of livestock products for basic necessities, for receiving appropriate care when needed, and for sending children to school. The importance of these resources increases when environmental stress is greater. Those who have larger families and many friends and relatives can move to more distant regions where the environment is more favorable. Those who have connections to the local government may secure better winter camps and benefit from occasional aid in food items, warm clothing, and hay and fodder at discounted prices. Therefore, political and economic changes in the country were reflected in changes to the 45

PAGE 58

livelihoods of Mongolians, and new forms of strategies to adapt to circumstances and secure access to resources emerged as a result. Social Capital Theory The theory of social capital is of great importance in this research. As discussed earlier in the dissertation, mobilization of a variety of social resources has emerged as an important adaptive strategy employed by Mongolian pastoralists to prevent or overcome negative impacts of natural disasters. Definition of Social Capital During the last decade or so, social capital has become one of the most popular concepts in the social sciences. Originating in the work of a French sociologist Emile Durkheim in the 19th century, its application has expanded to multiple fields, including health-related disciplines. Social capital is often seen as the'third form of capital, the first two being financial and human. Portes' (1998) defines economic capital as "people's bank accounts", human capital as "inside their heads", and social capital as "the structure of their relationships". The origin of social capital dates to the 1980s when Bourdieu gave the first systematic contemporary analysis of social capital. He defined social capital as: ... the aggregate of the actual or potential resources which are linked 46

PAGE 59

to possession of a durable network of more or less institutionalized relationships of mutual acquaintance or recognition" (Bourdieu, 1985, p. 248). Following Bourdieu, Coleman, in his seminal work, defined social capital through its functions: Social capital is defined by its function. is ... a variety of entities with two elements in common: they all consist of some aspect of social structures, and they facilitate certain actions of actors whether persons or corporate actors within the structure (Coleman, 1988, p. 98). Both Bourdieu and Coleman mention social network or structure as the foundation of social capital. Bourdieu put more emphasis on formal networks, whereas Coleman's ideas of social capital emphasize informal networks. Putnam added the importance of trust and norms to social networks. He stated, ... social capital. .. refers to features of social organization, such as trust, norms, and networks that can improve the efficiency of society ... (Putnam, Leonardi, Nanetti, 1993). In recent years more and more researchers define social capital as consisting of those resources available through social networks (Lin, 200 1; Portes, 1998). Social networks, in this case, are defined as, "the web of social relationships that surround an individual and the characteristics ofthose ties" (Berkman Glass, 2000, p. 145). 47

PAGE 60

Clearly, the definition of social capital varies, but researchers tend to separate group-level structural social capital that emphasizes resources available through social relations and structures, such as interpersonal trust, norms and values, civic engagement, rule of law, and governance; and individual-level social suppOli that flows in through social networks. This J variation in a definition leads to methodological differences in measuring social capital (Grootaert Van Bastelaer, 2002; Kawachi Berkman, 2000). Capturing all dimensions of social capital in one research project may be a daunting task, though choosing a specific dimension and designing a research to capture only this dimension can produce meaningful results. Societies are unique in their social relationships, thus in their social capital. A dimension of social capital that is important in one society may not be relevant another. This is certainly true for rural pastoralist communities in Mongolia. Formal organizations are not typically important in their everyday lives, but local norms and values, and trust can have significant influence in their lives. Yet access to formal organizations -: credits, insurance, etc. may be important during times of stress. In addition, informal networks and resources flowing through these networks may be very important, especially for the poor. These microand meso-level dimensions of social capital are the main focus of this research. 48

PAGE 61

From a methodological perspective, researchers define three types of social capital: bonding social capital: ties to people who are similar in terms of their demographic characteristics; bridging social capital: ties to people who do not share same demographic characteristics; and linking social capital: ties to people in positions of authority (Grootaert, Narayan, Jones, Woolcock, 2004). All three functional types of social capital are important for socioeconomic well-being and health of individuals and groups. One type might offer benefits that are not available though other types. Linking social capital may offer more opportunities to access social services, whereas bonding social capital facilitates the adoption of healthy behaviors, and bridging social capital facilitates information exchange. is worthwhile mentioning that social capital may also have a negative impact on those both within and outside a community: adverse effects on outsiders: excluding outsiders and maintaining inequalities between groups (Waldinger, 1995); and adverse effects on insiders: restricting individual privacy and autonomy, making it hard to get out of "bad" groups such as drug dealers, gangs, and reducing the inflow of new ideas (Productivity Commission of Australia, 2004; Bourgois, 1995; Portes, 1998). 49

PAGE 62

These negative effects of social capital may severely affect households who by virtue of escaping harsh climatic conditions are forced to move to different counties: help in new places may not be available. For this reason unless they face severe hazards, households do not tend to move to places where they do not have any social ties and support. Social Capital and Economy Links between social capital and development have been examined in a range of contexts. Higher levels of social capital appears to be beneficial to economic development, effective political institutions, and reduction of political problems (Fukuyama, 1995; Putnam, 1995). The following are examples of research conducted in both developing and developed countries that have explored the relationship between social capital and socioeconomic development: Guiso et al. (2000) found that in Italy, in areas with high levels of social trust, households invest less in cash and more in stock, use more checks, have higher access to formal credits. Firms also benefit from higher social trust: they have more access to credits and are more likely to have multiple shareholders (Guiso, Sapienza, Zingales, 2000). Social capital measures (e.g. family network) were associated with higher secondary school graduation rate, college enrollment, 50

PAGE 63

socioeconomic status, and avoidance of criminal activities among children of teenage mothers in the U.S. (Furstenberg Hughes, 1995). Higher social capital is shown to have a positive impact on watershed conservation and in cooperative development activities in Rajasthan, India (Krishna & Uphoff, 1999). More trust, reciprocity, and sharing in neighborhoods of Dhaka, Bangladesh predicted a likelihood of a neighborhood to have a voluntary solid waste management system (Pargal, Huq, Gilligan, 1999). Social capital measured by the number of memberships in associations, diversity of memberships, number of meetings, and cash and time contribution to associations are positively related to asset accumulation and access to credits in Indonesia (Grootaert, 2000). The role of social capital in economic well-being of households may be even greater in developing countries where almost all of the transactions between individuals are performed based on individual trust and trust in informal institutions. Formal institutions (e.g. courts) that typically regulate transactions may not function properly or may be too expensive (Durlauf Fafchamps, 2004). Researchers have identified the following mechanisms through which social capital facilitates economic development: Social capital facilitates transactions among individuals or groups. Common rules and trust allow people to interact efficiently. 51

PAGE 64

Social capital makes information and knowledge exchange more efficient. Larger networks ensure a greater more flow of information, whether it is about hazards, market prices, social services, or governmental aid. Participation in social networks and the development of trust makes collective action easier. Developmental programs often have to rely on collective action. High levels of social control puts pressure on individuals in a network and forces them to engage in positive behaviors that are mutually beneficial (Grootaert Van Bastelaer, 2002). Social capital increases access to social services (Productivity Commission of Australia, 2004; Kawachi, Kennedy, Glass, 1999). These mechanisms do not relate only to economic development and household economic welfare. They are relevant to the association of social capital to health as well. In addition, socioeconomic status may become an intermediary link in the relationship of social capital to health. The importance of social capital and support has become greater for Mongolian herders since privatization. Livestock husbandry is a difficult job that requires adequate human resources, access to better pasture and water especially in times of unfavorable environmental conditions, reliable information flow about weather, market, and government policy, and assurance of resources available through connections with local government. All of these can be secured only through expanding their social ties and establishing a flow 52

PAGE 65

of mutual support through these networks. Reciprocity and hospitality are two features that are common among pastoralist societies. These features facilitate broader larger social ties and support that are crucial to successful herding. Social Capital and Health The effect of structural social capital and individual social support on health is well documented. Studies conducted across different populations show that people who have greater social capital and/or social support live longer (Christensen, Wiebe, Smith, Turner, 1994; Giles, Glonek, Luszcz, Andrews, 2005; Kawachi et aI., 1999; Kawachi, Kennedy, Lochner, Prothrow-Stith, 1997), are less susceptible to non-communicable diseases such as cardiovascular diseases, cancer and arthritis (Eng, Rimm, Fitzmaurice, Kawachi, 2002; Eriksen, 1994; Kinney et aI., 2003; Rosengren, Wilhelmsen, Orth-Gomer, 2004; Vogt, Mullooly, Ernst, Pope, Hollis, 1992; Weinberger, Tierney, Booher, Hiner, 1990), are more likely to survive myocardial infarction (Berkman, Leo-Summers, Horwitz, 1992; Farmer Meyer, 1996; Kawachi et aI., 1996; Orth-Gomer, Rosengren, Wilhelmsen, 1993; Seeman, 1996), are less likely to suffer from mental illnesses (Bal, Crombez, Van Oost, Debourdeaudhuij, 2003; Bassuk, Glass, Berkman, 1999; Michael, Berkman, Colditz, Kawachi, 2001; Penninx et aI., 1997), are less susceptible to infectious diseases (Holtgrave Crosby, 2003; Theorell et aI., 1995), and are more likely to engage in health-promoting behaviors (Treiber et aI., 1991). 53

PAGE 66

The mechanisms through which social capital influences health is the same as in case of economic development: reducing transaction costs, disseminating information and knowledge, promoting healthy behaviors, increasing access to health services, plus psychological benefits (buffer of stress, increased self-esteem, stability and control over environment affects the neuroendocrine and immune systems, and influence the overall physical health) (Cohen Syme, 1985). In a study of maternal mortality in Mongolia, social support available through the family network was an important determinant of maternal deaths (Janes Chuluundorj, 2004). A woman could not receive a medical care on time because no one was available to take care of her children while she was gone, or no one was able to lend her money to cover necessary expenses. Pregnant women do not stop doing heavy household chores such as fetching water and firewood, milking animals and processing dairy products. All of these involve lifting of a heavy staff, which can facilitate placental abruption and hemorrhage. Often women who died in pregnancy or childbirth did not have anyone nearby who could help while their husbands were gone herding their animals or selling their livestock products. The concept of social capital has been increasingly used by researchers in health-related fields but there are several important limitations to such studies: Most measures are measures of outcomes, not of social capital itself. They measure only the quantity of social capital, not quality. 54

PAGE 67

Tendency to use a single indicator of social capital which fails to capture the multi-dimensionality of the phenomenon. A failure to recognize that social capital will vary by network type and social scale. A failure to differentiate the effects of social capital from other forces such as political institutions (Productivity Commission of Australia, 2004). Since Coleman's work, researchers increasingly acknowledge that social capital is a group/community attribute. Problems may rise when social capital measurements are taken at the individual level. Yet, Brehm and Rahn (1997) argue that it is individuals who ultimately build relationships and trust, and who participate in various community activities. Thus, it is also important to measure social capital at an individual level in addition to group-level measurements. In developing countries, where the quality of data collected at the community levels is often poor, individual-level variables may be the only valid option. Gender Gender is an important covariate in health related research not only due to the biological differences between female and male bodies, but also due to the social differences that may put one gender at larger risk of certain illnesses. Gender theory attributes such differences to the varying roles that females and 55

PAGE 68

males perform in their household livelihood systems. Chen (1991) argues that women's activities are more multidimensional than men's, some of the activities are a sole responsibility of women, and women's role in creating and keeping networks are greater than men's. However, viewing labor division as the only factor in health inequalities among men and women carries serious flaws. does not acknowledge social structural explanations of women's vulnerability such as power, access to resources, discrimination against women, and domestic violence. Environmental challenges such as land degradation, drought or flood just to name a few, limit the abilities of women to handle their household responsibilities and increase their work burdens, negatively affecting their health status. The literature reveals a pattern of gender differentiation throughout the disaster process. The differences are largely attributed to childcare responsibilities, poverty, social networks, traditional roles, discrimination, and other issues related to gender stratification. More women die during natural disasters compared to men because often they are confined to homes and do not escape on time. Women experience more emotional problems during and after disasters compared to men. is often on women's shoulders to rebuild their and their children's lives because men tend to work outside of their homes (Fothergill, 1996). The role of women in the pastoral economy is expanding in Mongolia. Traditionally, men were responsible for herding and women were responsible for dairy processing and other household duties. With an increase of herd size 56

PAGE 69

and mix of animals, it became necessary for women to engage in other non traditional activities to help the family, including a sale or trade of livestock products and wage labor for cash income. They are now important contributors to the household's economy and this may increase women's risk for sickness. Demand for labor during disasters may result in delay in seeking care for health problems. Whether or not Mongolian males and females experience disasters differently and have varying health outcomes has not yet been studied. Theoretical Framework All of the theories and constructs used in the study are summarized in the theoretical framework presented schematically in Figure 3.1. Vulnerability consists of three components: natural hazards, sensitivity factors and adaptive capacities of individuals and households. All three components are shaped by global processes. Global warming due to the increased industrialization and higher emission of greenhouse gases increase the number of drought and events. Political processes that occurred in the countries of the former Soviet block have brought significant changes in national and lo'eal economies and people's livelihoods. Structural adjustment programs taking place in developing countries increase the number of poor people and widen the gap between the rich and the poor. Thus, political ecology affects the processes that occur in the physical environment and social forces that define the vulnerability to natural events. 57

PAGE 70

The outcome of a natural hazard, measured by socioeconomic loss and health conditions, is dependent on sensitivity and adaptation of individuals, households or regions. Lower sensitivity and higher adaptive capacity buffer negative impacts of hazard events. Very young or very old people, women, the poor, and those who do not have adequate skills are most sensitive to natural hazards. Individuals and households with greater social capital, better herd management skills and more accumulation of resources are less likely to suffer from shortand long-tenn consequences of natural hazards. Sensitivity and adaptation are also related. Enhancing adaptive capacity will result in lower sensitivity and lower sensitivity also increases adaptive capacity. Better socioeconomic status and health conditions of individuals and households will in turn lower the sensitivity to natural hazards and increase the adaptive capacities. Thus, all elements of vulnerability and the outcome are interrelated with each other, fonning a circle in the theoretical framework shown below. 58

PAGE 71

Age: young/elderly Gender: female headed households High population ratio Poor Inadequate skills:, education, occupation Global warming Structural.adjustment programs '. Political processes Socioeconomic Status Health Social capital/social network and support Good herd management skills Accumulation of resources Figure 3.1. Theoretical Framework 59

PAGE 72

CHAPTER 4 METHODOLOGY The objective of this research is to understand how environmental challenges interact with political and socioeconomic circumstances and the adaptive strategies employed by herders to affect the well-being of rural pastoral households. Of particular importance is the identification of household access to social resources and/or emerging or incipient cooperative institutions that spread risk among numbers of participating households. To meet the overall objective of the research the following specific aims were developed: #1. To identify vulnerability to natural hazards at the county level by assessing the exposure to hazards, sensitivity (a potential of being affected by a climate stress) and adaptive/coping strategies using climate and county socioeconomic and demographic data. #2. To explore the relationships between vulnerability and health outcomes at the county level. 60

PAGE 73

#3. To study the impact of vulnerability to drought and (a Mongolian term for winter disasters) on the economic well-being of rural households. #4. To investigate the relationship of vulnerability to drought and dzud and the health status of individuals within households. A multi-level research design is employed with the main outcome variable, well-being, measured multi-dimensionally at the county, household, and individual levels, and includes both socioeconomic status and general health outcome. Research Plan Identification of sensitivity factors and adaptive strategies in the context of diverse environmental risks requires a two-stage research design: a spatial ecological study at the county/community level, and a cross-sectional study of. 120 households sampled from four counties in Mongolia. The details of each element of the study design will be described in the following sections. A Spatial Ecological Study at the County Level The spatial ecological study involved assessing relationships among and between climate, socioeconomic factors, and health in all of Mongolia's rural counties. 61

PAGE 74

Spatial statistical methods are commonly used to assess rates across geographic space, adjust relationships for noise (e.g. spatial autocorrelation areas closer to each other tend to have similar values), identify disease outbreaks, and evaluate the impact of specific exposures Problems of analyzing spatially referenced data using nonspatial methods pose several problems, the most important one of which is spatial autocorrelation. The first law of geography states that "Everything is related to everything else, but near things are more related than far things" (Tobler cited in Waller Gotway, 2004, p 3). Thus, using nonspatial statistics violates the assumption of independent observations which may result in serious flaws Spatial statistical methods consider this spatial dependency of observations and allow analysis of differentiating trendslrelationships from spatial autocorrelation. The availability of mapping tools, including greatly influenced the development of medical geography. The use of spatial statistics and GIS in health sciences has mainly focused on creating demographic, economic and lifestyle profiles of communities and their relationship to environmental hazards (Bellander et aI., 2001; Dunn, Woodhouse, Bhopal, Acquilla, 1995; Moran Butler, 2001; Morrow, 1999; Pine & Diaz, 2000; Speer, Semenza, Kurosaki, & Anton-Culver, 2002). Studies have found an association between exposure to environmental hazards and cancer (Entwisle, Rindfuss, Walsh, Evans, Curran, 1997; Hjalmas, Kulldorff, Gustafson, & Nagarwalla, 1996; Krautheim & Aldrich, permitted 62

PAGE 75

surveillance and monitoring of vector-borne or infectious diseases (Boone et aI., 2000; Emch, 1998; Fost, 1990; Kistemann, Munzinger, Dangendorf, 2002; Marfin et aI., 2001; Tanser Ie Sueur, 2002); quantifying environmental levels of toxic substances and their health effects (Kohli, Noorlind-Brage, Lofman, 2000; Margai, 2001); measuring access to health services (Brabyn Skelly, 2002; Eyles, 1990; McLafferty, 2003; Parker Campbell, 1998; Rosero-Bixby, 2004); planning service areas (Bhana, 1998; Foley, 2002; Kofie & Moller-Jensen, 2001; Perry & Gesler, 2000); and predicting pedestrian injuries (Durkin, McElroy, Guan, Bigelow, Brazelton, 2005; LaScala, Johnson, Gruenewald, 2001; Lightstone, Dhillon, Peek-Asa, Kraus, 2001). Lately, spatial statistics and GIS methodologies are increasingly used in social epidemiology to study the distribution of the social and behavioral determinants of health outcomes (Entwisle et aI., 1997). There are several important concerns in using spatial statistics such as spatial autocorrelation mentioned above, including the modifiable area unit problem (changing spatial units of analysis may result in different relationships), the ecological fallacy (relationship at aggregate level may not apply to individual level), and non-uniformity of space and edge effects (the spatial units of analysis are drawn arbitrarily and events on the edges of a unit are forced to relate to the center) (Waller Gotway, 2004). Careful design and analysis may significantly reduce these problems and enhance the validity of results. 63

PAGE 76

Sampling 250 rural counties (out of 302 total) in Mongolia are included in the spatial analysis of climatic, socioeconomic, and health conditions in Mongolia. Variables The analysis examines three categories of independent variables, all measured in 2003 at the county level. The descriptions of independent variables are given in Table 4.1. Monthly temperature and precipitation data were obtained from 108 weather stations across Mongolia for the period 1993-2003. Universal kriging was performed using GIS to predict the monthly averages for each county. Several indicators, including coefficient of variability, average standardized anomaly, aridity index, yearly average, and yearly range, were created from predicted monthly data and regressed towards the number of livestock and livestock growth rate for 1993-2003. The best predictors of the number of livestock and livestock growth rate were mean yearly precipitation and temperature ranges, therefore used in further analyses. Sociodemographic characteristics and livestock indicators were collected at the end of2003. The data were not available for the previous years. Thus, the data only allow us one to look at the more immediate consequences of climate stress. 64

PAGE 77

Table 4.1. Independent variables in the spatial ecological study Variable Independent Explanation of Variables Data Source Categories Variables Mean yearly The difference between the Institute of precipi tation maximum and minimum Meteorology range 1993-2003 monthly average and precipitation is calculated Hydrology for each year and averaged. A larger range is an indicator of more rainfall in rn the summer and less rn (1) .l:l snowfall in the winter. rn (1) ..... ro Mean yearly The difference between the ..... .temperature maximum and minimum range 1993-2003 monthly average temperatures is calculated for each year and averaged. A larger range means higher temperatures in the summer and low temperatures in the winter. Dependency Ratio of the population National rn ratio under 18 and over 65 years Statistical C,) ..... ..... of age to the population Office of rn ..... 5)--between 19-64 years of age Mongolia ..... rn C,) I-< ro 0 8 ..t:: Percent of Percentage of people whose C,)"O .S people with income is under absolute ..t:: :>. income below poverty line in the total <.t:: I-< poverty line population. The poverty line bJ) ..... o ..... is set by the government for (1) (1) each region and can be 0 ..... either a minimum wage per C,) 0 person or minimum number CZl of livestock per person. 65

PAGE 78

Table 4.1. Independent variables in the spatial ecological study (Cont.) Variable Independent Explanation of Variables Data Categories Variables Source Unemployment Percentage of people who National rate are not employed in the total Statistical Ul 1-0 ..... 0 population of 18-64 years of Office of Ul ...... 0.. (.) age Mongolia on .... I::::: o ..... S t> .c Percent of Percentage of people who 1-0 people receives pensions and o ..... G .t:: o (.) Ul receiving allowances in the total Ul pensions and population allowances Ul Livestock Ratio of the number of National 1-0 ..... o ...... ...... 0.. density in SFU livestock in SFU to total Statistical land in hectares Office of I::::: 4-< ..... ..... 0 on Mongolia (.) 1-0 Livestock per Ratio of the number of 001-0 .................. Ul Ul person in SFU livestock in SFU to total ..... population .S The health-related dependent variables of the first stage of analysis are shown in Table 4.2 The crude morbidity rate reflects the total number of hospital visits, including visits to outpatient clinics. 66 __ __ -L'-_______________ __ _____________________ ______

PAGE 79

Table 4.2. Dependent variables in the spatial ecological study Variable Dependent Explanation of Variable Data Source Categories Variables Health Crude Number of hospitalNational morbidity rate recorded visits/Total Statistical population 1000 Office of Crude mortality Number of deaths/Total Mongolia, Ministry of rate population 1000 Health, Under five Number of deaths of World mortality rate children <5 years of agel Health Totallivebirths 1000 Organization Maternal Number of maternal mortality rate deaths/ Number of women of reproductive age *100000 Analysis A conditional autoregressive (CAR) model was used to analyze the effects of the independent variables on health indicators. CAR models include spatial parameters to measure spatial autocorrelation by comparing values in a county to the values in neighboring counties. In other words, the means and variances defined by the CAR model are conditional, thus dependent on the values of neighbors. This permits separating the relationship between independent and dependent variables from the relationship that occurs due to spatial autocorrelation. spatial autocorrelation is not adjusted in the analysis, 67

PAGE 80

parameter estimates and standard errors may overestimate existing relationships. The common problems that rates pose for data analysis are abnormal distributions, dependence of the variances of rates on the mean rate itself, and the dependence of variance on population size. These are serious problems that may lead to wrong conclusions ifnot fixed appropriately. The Freeman-Tukey square root transformation was thus use d for outcome variables to make their distributions normal. This transformation also removes a connection between the rate and the variance of the rate. Subsequently, a weighted analysis is performed to remedy the dependence of the variance on population size (heteroscedasticity) (Waller Gotway, 2004). Nonspatial and spatial parameters of the analyses are presented in Chapter 5. A Cross-Sectional Study at the Household Level the second stage of the research, 120 household-level structured interviews of households in four counties,. stratified by vulnerability and health indicators, were conducted. The main objective of this study was to explore the impact of ecological factors on individuals and households, factors that determine the extent of such effect( s), and the outcome of this interaction expressed in individuals' and households' levels of economic status and health. These interviews permit analysis of household vulnerability and the long-term outcome of natural disasters, and provide important information on how the 68

PAGE 81

coping and adaptive strategies employed by rural herding households mitigate the consequences of adverse environmental events. The complex interaction among environmental constraints (same as in the spatial analysis stage of the study), adaptive and coping capacities of households and their socioeconomic well-being and health status can only be assessed in details via interviews. The results of analyses at this stage will compliment the findings from the first stage of the research. Sampling Analysis of the county-level data were used to identify four counties that differed on two dimensions: vulnerability (combination of climatologic stress, socioeconomic factors, and adaptive strategies identifiable from county level data), and level of household well-being (as reflected by health indicators). Counties were ranked for each independent and dependent variable shown in Tables 4.1 and 4.2. Mean ranks for each category of variables (climate stress, sociodemographic, socioeconomic and health) were calculated for counties. Counties were sorted into quartiles by ranks on health status and three categories of vulnerability statuses. Counties falling in the lower and/or upper quartiles on each category were included in the sampling frame, generating a 2x2 table. There were five counties in overall that fit the criteria for the sampling fame. One county in each cell was selected to be most representative 69

PAGE 82

of the ecological diversity of Mongolia. The following 2x2 table describes the sampling frame and indicates the counties chosen for household interviews. Table 4.3. Sampling frame for household interviews POOR ..... ro <1) ::c: GOOD Socioeconomic and Environmental Vulnerability HIGH LOW Olziit (desert) Khovd (steppe) Bayankhutag Bayan-Ondor (steppe) (desert) is important to remember that this scheme is based on county level data; the indicators are only indirect reflections of what happens to individual households. The role of the aggregate data analysis is to develop a sampling frame that captures the widest possible range of household experiences in the context of diverse environmental stresses. The scheme serves to maximize variability, and, thereby, generate a sample with a greater probability of highlighting household-level coping and adaptive strategies, and intrahousehold sociodemographic factors that affect household well-being. The following four counties were selected using a 2x2 sampling frame: Bayankhutag county in Khentii province, Olziit county in Dundgovi province, Bayan-Ondor county in Bayankhongor province, and Khovd county in Khovd 70

PAGE 83

province (Figure 4.1). These counties represent two different ecological zones: steppe (Bayankhutag and Khovd) and desert (Olziit and Bayan-Ondor). \:l-aimag 1a Figure 4.1. Counties Selected for Interviews Thirty interviews in each county were conducted in the summer of 2005. This provided an acceptable level of power for identifying variables comprising household vulnerability and linking these to individual health outcomes (based on identifying variability in health outcomes of between 2-5% in the population with a sensitivity of+/-2%). A variety of power calculations using the "stat-calc" function in EpiInfo showed the sample size to be adequate to testing the main elements of the study hypothesis. This said the sample is by epidemiologic standards relatively small. This should be kept in mind when evaluating tests of significance 71

PAGE 84

County officials and health care providers were contacted at the onset to locate households for sampling. A transect sampling method was used to select households within counties: households along the road, river or valley were counted and a systematic sampling method was used to select from households. Every second to every fifth household was approached for an interview. A second and/or third attempt was made if adult members of a household were not present. is important to note here that the households that had suffered significant losses of animals, and were thus the most vulnerable, were likely to have been forced to migrate to urban areas. The resulting household sample may thus not represent the most vulnerable of the population. In Olziit county a systematic sampling strategy was not feasible: many households had left the county territory because of a severe drought, and every household we could locate was interviewed. In Bayan-Ondor county, households moved to the north and to the territory of a neighboring county due to the drought as well. Thus interviews took place in the northern part of the county and also in the territory of a neighboring Erdene county of Gobi-Altai province. Locations of each interview site were recorded using the Global Positioning System. there was more than one household in the camping group, the first household with a head or spouse present was approached for an interview. 72

PAGE 85

Interview Sites Bayankhutag County. The territory of Bayankhutag county (Figure 4.2) is divided into three townships. Township #2 was selected randomly to conduct household interviews. At the time of the study the majority of households had moved closer to the Kherlen river on the northern border of the county due to a severe drought. The county center is located approximately 10 kms south of the province center--Ondorkhaan. The population in Bayankhutag county are predominantly Khalkh Mongols, but a small number ofUriankhai and Buriat ethnic minority groups were also present. The population in 2003 was 2200. Patron-client relationships were well established in this county. Wealthier families with many livestock hired poor families to help in herding and related activities. In tum, patrons provided clients with meat, dairy and a monthly salary averaging approximately 40000 tugrics (equivalent of35 USD). Figure 4.2. Interview Sites in Bayankhutag County 73

PAGE 86

Olziit County. Olziit is the largest and southernmost county in Dundgovi province (Figure 4.3). The county is notable for having the largest number of camels in Mongolia. With a population of2900 in 2003, it is the third most populated county in Dundgovi province. Music, folklore songs, games and airag (fermented horse milk) are known to be popular among people in Dundgovi province. They are by reputation a socially active, friendly and fun-loving people. Because of extremely dry conditions and poor vegetation, herding households move very often: it is not uncommon for families to move 2-3 times per week. Households keep great distances from each other, and it was often difficult to locate single "gers" (yurts) among the sand dunes. Figure 4.3. Interview Sites in Olziit County Khovd County. Khovd county is located on the eastern slopes of the Altai mountain ranges and to the west of Khovd province center (Figure 4.4). 74

PAGE 87

The closest households were located within five kms of the province center. There are only two townships in this county: the 1 st township where all residents are engaged in farming, and the 2nd township where livestock husbandry is the dominant mode of production. Interviews took place in the 2nd township. The total population was 5202 in 2003. Khovd county is the only county in Khovd province where the majority of its (96%) population are Khazakhs. Khazakhs are the largest of the ethnic minorities in Mongolia. They speak a distinct language (though many also speak Khalkh Mongol), and, in contrasts to most Mongolians, are Muslim. The relationship between Khazakh and Mongol residents seems to be very good, although some disputes about animal theft between them takes place occasionally. Khazakh people are known for their great hospitality, close family ties, great sense of humor, and beautiful crafts. Figure 4.4. Interview Sites in Khovd County 75 -

PAGE 88

Bayan-Ondor County. Bayan-Ondor is the southernmost county of Bayankhongor province: it borders China, a fact that opens great trading opportunities for local herders (Figure 4.5). Perhaps as a consequence, Bayankhongor province is notable for having the largest number of-goats in the country cashmere is a particularly important product for cross-border trade. There are three townships in this county, but because of a severe drought households from all townships had moved to the north, and many into an adjacent county. Figure 4.5. Interview Sites in Bayan-Ondor County 76

PAGE 89

Interview Setting Interviews took place in a Often there were people from other households present during the interview. After presenting an introduction to the study, a consent form was read and signed by a head ofthe household and/or spouse, and anthropometric measurements and hemoglobin tests of capillary blood were taken from each member of the household. Once all measurements were taken interviews were done with the head of the household and/or spouse. Each interview together with measurements took on average from 1.5 to 2 hours. Only two households refused to participate in the study. These were both located in Khovd county. The heads of these two households were not present and their spouses did not speak Mongolian. Subject Payment An equivalent of 5 USD was given as an incentive to each household participating in the study. Iron supplements were given to those who had anemia. Household Questionnaire A modified version of a social capital questionnaire, the Social Capital Integrated Questionnaire (SC-IQ) developed by a team of the World Bank, was 77

PAGE 90

used in the research. This instrument had been tested for validity and reliability in developing country settings such as Kygyzstan and Nepal (Grootaert et aI., 2004), though not in nomadic populations. Modifications were thus made to this questionnaire to fit it to the unique social characteristics of a highly mobile, pastoral population. Questions about herding practices, natural disasters, economic and health statuses were also added. The questionnaire had the following sections: 1. Demographic information on household members: gender, year of birth, marital status, occupation, years of education, years of residency in the county, months present in the household within the last year. 2. Information on the "khot ail" (cooperative group): gender, age, and relationship, types of support given to and received from a 3. Information about other immediate family members (parents, siblings, children and significant others): gender, age, relationship, occupation, education, place of residency, frequency of meetings, support given and/or received. 4. Social capital: a. Social network: number of close friends and relatives, their residency, frequency of meetings, closeness to a county governor, likelihood to get help in different situations, access to loans and restocking programs. b. Trust: likelihood to get help, trust in people, friends, relatives, local government and media/press. 78

PAGE 91

c. Social cooperation: engagement in social formal and informal activities, likelihood to cooperate in various situations. d. Information and communication: types of information sources, number of visits to county and province centers, distance to the nearest phone. e. Social cohesion: wealth differentiation, number of visitors and households visited, number of games played and food shared with others, likelihood to help new people, and area safety. f. Empowerment: impact on neighborhood, likelihood of a local government to listen to people, number of petitions of government, number of people who voted in the last election 5. Herding practices: number of moves, their duration and distances, purchase of livestock insurance, winter shelter and veterinary services, conflict over pasture and water sources, amount of hay and mineral lick prepared for winter, problems of overgrazing. 6. Experience of natural disasters in the past 15 years: livestock loss, aid from friendslrelatives, government, disaster warning and preparedness, and accessibility of roads. 7. Health status during the last six months for each household member: presence of illness,access to health care, cost of care, availability of cash for health care. 8. Economic status: livestock inventory, yearly income from livestock products and from pensions, wages and allowances, supplementary 79

PAGE 92

income sources, large expenses within a year, and an inventory of household items such as TV radio, battery, satellite dish, and transportation. A complete questionnaire can be viewed in the Appendix. Hematological and Anthropometric Data Collection Anthropometric measures are to be a relatively cheap method for early detection of a nutritional deficiency (Bailey Ferro-Luzzi, 1995). Body Mass Index (BMI) is commonly used in assessing individuals' nutritional status. is an indirect indicator of body fatness. The formula for BMI is weight in kilograms divided by height in meters squared. Skinfold thicknesses also provide an estimation of general fatness. Of the many mineral deficiencies iron deficiency anemia is a common problem in developed countries, where it is found as high as in 67% of children and 33% of women of reproductive age (Kormondy Brown, 1998). Hemoglobin. The blood hemoglobin concentrations of all household members, except children under six months of age were tested in the field using a portable hemoglobinmeter manufactured by HemoCue Ltd, UK. A sample of capillary blood was obtained from the ring finger or middle finger of the left hand using a microlance, and immediately analyzed in the hemoglobinmeter. The sensitivity ofHemocue is 82.4% and specificity is 80

PAGE 93

94.2% (Paddle, 2002; Sari et aI., 2001). Hemocue of capillary blood has a 85% correlation with a gold standard: direct cyanmethaemoglobin (Sari et aI., 2001). The measured hemoglobin was used to determine whether an individual was anemic using UNICEFIUNU/WHOIMI criteria. The following table shows the hemoglobin thresholds for anemia used in this study. Table 4.4. Anemia cut-off levels Category Children under five years of age Children 5-11 years of age Children 12-14 years of age Women non-pregnant Women pregnant Men Hemoglobin in g/dl 11.0 11.5 12.0 12.0 11.0 13.0 All children diagnosed with anemia were given ferrous sulfate syrup and adults were given ferrous sulfate tablets. Directions for use were given and a printed copy of the directions in Mongolian was provided to an adult female member of the household. Body Weight. Body weight was measured in all subjects over two years of age using standard scale (precision of 100 g, made in Russia) which was periodically checked using a five kg standard weight. Subjects were weighed in 81

PAGE 94

bare feet with a minimum of clothing. No corrections were made in the analyses for clothing. Body weight for children under two years of age was measured using a battery-operated digital weight scale (precision 100g, Seca). A child was placed in the middle of the scale with light clothing and the reading from the digital scale was recorded. The measurement was repeated if the baby was moving and would not lie still. Body Length and Height. Length was measured in children under two years of age with a plastic measuring mat (5 mm precision, Seca). For children over two years of age and adults, height is measured using a portable stadiometer mm precision, Seca) with a movable bar. Skinfold Thickness. Abdominal and triceps skinfolds were measured using skinfold calipers. An abdominal skinfold was measured 1 cm to the right of umbilicus in a horizontal fold while the person was standing. A triceps skinfold was measured with the right arm hanging loosely, 1 cm posterior from the middle point between the lateral projection of the acromion process of the scapula and the inferior border of the olecranon process of the ulna while the elbow was flexed to 90 (Lohman, Roche, Martorell, 1988). The correlation between the abdominal and triceps skinfolds was .90. 82

PAGE 95

Variables Complete lists of all variables entered into the analyses are given in the following tables. Independent variables used in the analysis are presented in Tables 4.5 4.9. Variables in Table 4.6 (socioeconomic indicators) are also used as outcome variables in some analyses. Descriptions of health outcome variables are given in Table 4.10. Table 4.5. Household sociodemographic variables Variables Explanation of Variables Household size Number of permanent household members Mean education Mean education years of household members> 18 years of years Mean occupational rank Percent of members in workforce age Each adult household member is given a score from 1 to 3 based on occupation and a mean for each household is calculated: 1 for herder, farmer, worker or salesperson (does not require training) 2 for agent, administrative worker, carpenter, construction worker, cook, driver, mechanic, miner, sewer and typist (requires 1-2 years oftraining) 3 for engineer, feldsher/nurse, financial advisor, forecaster, teacher, policeman, veterinarian, and college student (requires specialized training of at least 3 years) Percentage of household members in workforce (18-64 years of age) 83

PAGE 96

Table 4.6. Household socioeconomic variables Variables Number of livestock in SFU Number of milk livestock inSFU Yearly household income in thousand tugrics Number of household items Number of transportation items Large expenses in thousand tugrics Monthly food expenses per person in thousand tugrics Meat (kg) consumed per person per year Explanation of Variables Sheep Forage Unit kg forage per day): 1 sheep = 1 SFU 1 goat 1 SFU 1 cattle 6 SFU 1 horse = 7 SFU 1 camel = 7 SFU Calculated the same way as above Sum of income from the sale of cashmere, wool, meat, dairy, and hides, wages, pensions and allowances in one year Sum of the household possessions such as TV, radio, power generator, and extra "ger" Sum of the household's transportation means such as motorbike, sedan, and truck Sum of the household's large expenses within one year A total amount of cash spent on food in the last month divided by number of people living in that household during this period. Total amount of meat consumed by the household over the year is divided by the number of a permanent household members 84

PAGE 97

Table 4.7. Household social capital variables Variables Number of households in Number of other family members Number of other family members in the samecourtty Number of close relatives Number of close friends Asked local ( county) governor for help Explanation of Variables Number of households defined by the interviewee as Number of immediate family members (parents, children, siblings) and significant others not living in the household Number of immediate family members (parents, children, siblings) and significant others living in the same county but not in the same household With whom the household members talk about private matters With whom the household members talk about private matters Binary variable coded 1 for as "Yes" and 0 "No" Number of people Number of people who turned for help within 1 month turned for help Access to loans and Binary variable coded 1 for as "Yes" and 0 "No" restocking programs Trust people Answers to the question "How trustful are people nowadays?" are scored as 1 not trustful, 2 little, 3 somewhat, 4 very much Trust relatives Answers to the question "How much do you trust your relatives?" are scored as 1 not trustful, 2 little, 3 somewhat, 4 very much Trust friends Answers to the question "How much do you trust your friends" are scored as 1 not at all' 2 little, 3 somewhat, 4 very much 85

PAGE 98

Table 4.7. Household social capital variables (Cont.) Variables Trust local government Trust press and 'media Number of communal activities participated Number of common issues discussed Likelihood to cooperate on building a well Explanation of Variables Answers to the question "How much do you trust your county government?" are scored as 1 not at all, 2 little, 3 somewhat, 4 very much Answers to the question "How much do you trust press and media" are scored as 1 not at all 2 little, 3 somewhat, 4 very much Number of communal activities such as township or county meetings, fundraising events within the last year Number of informal meetings with people to discuss issues within the last year Is scored as 1 very unlikely, 2 somewhat unlikely, 3 somewhat likely, 4 very likely Likelihood to help in Is scored as 1 very unlikely, 2 somewhat case of illness and unlikely, 3 somewhat likely, 4 very likely disaster Number of information sources about government activities Number of information sources about market prices Distance to the nearest phone Wealth difference Reported number Reported number Reported distance in kilometers Is scored as 1 different to a great extent, 2 different to a small extent, 3 not different 86

PAGE 99

Table 4.7. Household social capital variables (Cont.) Variables Frequency of having foodl drinks Explanation of Variables Frequency of having foodldrink with others in special occasions such as birthday, funeral arrival of special guests within the last month Frequency of playing Frequency of playing games with others within the games last month Number of visitors in The average number of visitors in asingle day was your asked and multiplied by 30 (one person may be counted multiple times) Number of other anyone from your household visited Likelihood to help newcomers Area safety Your impact on neighborhood Likelihood of local government listening to people The average number of other households any of the household members visited in a single day, multiplied by 30 (one household may be counted multiple times) Is scored as 1 very unlikely, 2 somewhat unlikely, 3 somewhat likely, 4 very likely Is scored as 1-very unsafe, 2 somewhat unsafe, 3 somewhat safe, 4 very safe Is scored as 1 no impact, 2 -a small impact, 3 -a big impact Is scored as 1 not likely, 2 to a small extent, 3 to a great extent 87

PAGE 100

Table 4.8. Natural disaster variables Variables Explanation of Variables Animal loss in SFU Animal deaths for each species are converted to SFU units and summed Number of Since 1990 years Number of drought years Number of and droughts combined Since 1990 Since 1990 88

PAGE 101

Table 4.9. Herd management variables Variables Have livestock insurance Have permanent winter shelter Have conflict over pasture Have conflict over water Prepared hay (in the past year) Prepared mineral lick (in the past year) Had a vet exam (in the past year) Problem of serious overgrazing Had distant moves Explanation of Variables 1 if "Yes", 0 if "No" 1 if"Yes", 0 if "No" o if "Yes", 1 if"No" o if "Yes", 1 if"No" 1 if "Yes", 0 if "No" 1 if "Yes", 0 if "No" 1 if"Yes", 0 if"No" o if "Yes", 1 if "No" Z-scores were calculated within each ecological zone and households who moved more than 2SD from the zonal mean were counted as distant movers (coded as 1), others were not (coded as 0). This was done to account for regional differences in moving distances. 89

PAGE 102

Table 4.10. Individual health variables Variables Explanation of Variables Hemoglobin From capillary blood gldl References Anemic (binary variable) UNICEFIUNU/WHOIMI Body Mass Index Overweight (binary variable) hemoglobin cut-off level is used to create a binary variable anemic and 0 not anemic Body Mass Index is calculated for individuals over 2 years of age using the following formula: BMI weight in kg height2 in em Individuals over 20 years of age are classified as overweight if BMI2: 25; Individuals 2 20 years of age are classified as overweight if their BMI falls at 85th percentile or above of CDC graph chart; For children from 0 to 2 year of age, ovelweight is assessed according to the growth chart developed by the Ministry of Health in Mongolia. All coded as 1 if overweight or as 0 ifnot. 90 Center for Disease Control Ministry of Health in Mongolia

PAGE 103

Table 4.10. Health variables (Cont.) Variables Underweight (binary variable) Triceps skinfoldmm Abdominal skinfoldmm Sick (binary variable) Diagnosed (binary variable) Treated (binary variable Explanation of Variables Individuals over 20 years of age are classified as underweight ifBMI 18.5; Individuals 2 20 years of age are classified as underweight if their BMI falls below or at 5% percentile of CDC graph chart; F or children from 0 to 2 year of age, underweight is assessed according to the growth chart developed by the Ministry of Health in Mongolia. All coded as 1 if underweight or as 0 if not. Coded as 1 if a household member was sick during the past six months and as 0 if otherwise 'Coded as 1 if diagnosed by a medical professional and as 0 if not Coded as 1 if treated and as 0 if not 91 References Center for Disease Control Ministry of Health in Mongolia

PAGE 104

Analysis A multivariate regression technique is used for analyses of household socioeconomic status using ML WIN software. A multilevel modeling procedure (MLM) is used to analyze the relationships between independent and dependent variables. MLM is an extension of ordinary least squares regression analysis. With exposure variables measured at the individual and household levels, and outcome variables at the household (socioeconomic indicators) and individual (health outcomes) levels, ordinary least squares regression analysis will allow only two options: aggregating data to the household level or distributing data from household to individual level. These two methods may lead to either ecological or atomistic fallacies. addition, distribution of higher level data to a lower level reduces the variation in measures (subjects within a household will have similar values), and minimizes standard errors. This may lead to an overestimation of relationships as significant when they are not (Ferrer, Palmer, Burge, 2005). On the other hand, an aggregation of a lower level data to a higher level data in analyses does not take into account individual characteristics within the groups. MLM takes care of these problems because it calculates the variation between higher level units (households) and subjects within a higher level unit (individuals in a household) separately. other words, by using MLM it is feasible to test whether or not differences in outcome indicators among households is due to the individual characteristics of households (compositional 92

PAGE 105

effect) or due to differences of household-level predictors when individual ,characteristics are controlled (contextual effects) (Diez Roux, 2002; Leyland Groenewegen, 2003; Macintyre Ellaway, 2003; Subramanian, Jones, Duncan, 2003). 93

PAGE 106

CHAPTERS RESULTS The goal of this study is to explore the relationship between vulnerability measured by exposure to natural hazards, sensitivity and adaptive abilities, and socioeconomic well-being and health outcomes at household and county levels. Findings of data analysis at both county and household levels are described in this chapter. Findings from the County-Level Spatial Data Analysis Descriptive Statistics The means, standard deviations, minimum and maximum values for variables used in the spatial data analysis are given in Tables 5.1 and 5.2. The data on the percent of people with income below poverty line were not available for Uvs andHuvsgul counties, and on the percent of people receiving pensions and allowances for Dornogovi counties. 94

PAGE 107

Table 5.1. Independent variables in the spatial ecological study Variables Mean SD Min Max Mean yearly precipitation 302 66.52 18.45 38.00 106.70 range Mean yearly temperature 302 40.36 2.08 35.40 45.80 range Dependency ratio 302 0.72 0.14 0.05 2 .21 Percent of people with income 262 0.29 0.18 0 00 1.00 below poverty line Unemployment rate 302 0.02 0.02 0.00 0 .16 Percent people receiving 290 0.10 0.02 0 04 0.24 pensions and allowances Livestock density in SFU 302 1.00 0.65 0.06 3.50 Livestock per person in SFU 302 48.05 21.16 6.07 121.46 Table 5.2. Dependent variables in the spatial ecological study Variables Mean SD Min Max Crude mortality rate per 302 6.58 2.22 1.00 15. 00 1000 persons Maternal mortality rate 302 12.01 39.43 0.00 226.76 per 100000 women of reproductive age Under five mortality rate 302 35.59 27 82 0.00 128 .21 per 1000 livebirths Crude morbidity rate per 302 240.65 130.81 35.19 811.39 1000 per persons 95

PAGE 108

Conditional Autoregressive Models Two hundred fifty counties out of 302 were entered into the conditional autoregressive model. Fifty two counties were omitted from data analysis due to missing information. Four separate models were calculated for each dependent variable using the R statistical program. The summary statistics of each model are presented in Tables 5.3 5.6. In the analysis of crude mortality rate in Table 5.3, spatial dependence is not present (likelihood ratio test statistic of spatial autocorrelation is not statistically significant). The mean yearly precipitation range and dependency ratio are shown as significant predictors of crude mortality rate. Higher mean yearly precipitation which indicates lower climate stress (B=O.Oll, SE=0.004, p-value
PAGE 109

Table 5.3. Crude mortality rate per 1000 persons Parameter Intercept Mean yearly precipitation range Mean yearly temperature range Dependency ratio Percent of people with income below poverty line Unemployment rate Percent of people receiving pensions and allowances Livestock density in SFU Livestock per person in SFU Spatial Parameter Phi Likelihood ratio test statistic significant at significant at 01 Estimate 3.478 0.011 0.006 1.275 -0.003 -0.653 0.739 -0.092 -0.002 0.0003 -0.021 1.222 Std. Error z-value 1.497 2.323 0.004 2.962** 0.033 0.183 0.552 2.313* 0.003 -0.815 2.990 -0.219 2.305 0.321 0.099 -0.923 0.003 -0.777 In the analysis of maternal mortality rate in Table 5.4, all indicators except mean yearly precipitation range and number oflivestock per person in SFU were statistically insignificant. Higher mean yearly precipitation range, which indicates lower climate stress, was correlated with lower maternal mortality rate (B=-0.005, SE=0.002, p<0.05). Larger number oflivestock per person in SFU was related with higher maternal mortality rate (B=0.007, SE=0.002, p-value
PAGE 110

significant, indicating an absence of spatial dependence of maternal mortality rate (phi=..:0.042, likelihood ratio test statistic= 1.869, p-value=0.17). Table 5.4. Maternal mortality rate per 100,000 women of reproductive age Parameter Intercept Mean yearly precipitation range Mean yearly temperature range Dependency ratio Percent of people with income below poverty line Unemployment rate Percent of people receiving pensions and allowances Livestock density in SFU Livestock per person in SFU (JA2 Spatial Parameter Phi Likelihood ratio test statistic significant" at significant at 01 Estimate 0.712 -0.005 0.027 -0.301 -0.002 1.050 -1.801 -0.067 0.007 0.0003 -0.042 1.869 Std. Error z-value 0.830 0.857 0.002 -2.473* 0.018 1.462 0.308 -0.977 0.002 -0.747 1.709 0.615 1.307 -1.378 0.056 -1.187 0.002 4.279** In Table 5.5, the percent of people with income below poverty line and livestock per person in SFU were statistically significant predictors of under five mortality rate. Under five mortality increases as percent of people with income below poverty line increases (B=0.054, SE=0.019,p-value<0.0l) and 98

PAGE 111

decreases as livestock per person in SFU increases (B=-0.032, SE=0.014, p<0.05). There was a weak negative spatial dependence of under five mortality rate (phi=-0.039, likelihood ratio test statistic=2.203, p<0.05). Table 5.5. Under five mortality rate per 1000 livebirths Parameter Estimate Std. Error z-value Intercept 5.785 7.971 0.726 Mean yearly precipitation range 0.001 0.018 0.077 Mean yearly temperature range 0.172 0.175 0.987 Dependency ratio -0.007 2.967 -0.002 Percent of people with income below 0.054 0.019 2.818** poverty line Unemployment rate -12.080 13.561 -0.891 Percent of people receiving pensions -4.109 12.867 -0.319 and allowances Livestock density in SFU 0.250 0.513 0.487 Livestock per person in SFU -0.032 0.014 -2.222* 0.446 Spatial Parameter Phi -0.039 Likelihood ratio test statistic 2.203* significant at ** significant at The crude morbidity rate in Table 5.6 shows a weak (phi=0.004, likelihood ratio test statistic=5.404, p
PAGE 112

morbidity rate. A higher mean yearly precipitation range (B=0.0933, SE=0.031, p-value
PAGE 113

The results of CAR models indicate that climate stress is not a strong predictor of mortality and morbidity compared to sociodemographic and socioeconomic variables entered into the equations. Indicators of lower socioeconomic status such as a higher percentage of people with income below poverty line and people receiving pensions and allowances were associated with a greater under five mortality and crude morbidity rates. The dependency ratio was also positively related to crude mortality rate. A higher number of animals per unit of land, which is an of greater economic resources of the county, was predictive of a lower crude morbidity rate. The greater number oflivestock per person was associated with a lower under five mortality rate but with a higher maternal mortality rate The number oflivestock per person may not be a sensitive indicator of economic status: it does not provide information on income or wealth inequality between households and distribution of economic resources within households. The choice of variables used in these analyses was limited by the availability of data at the county level. Missing data, poor quality of data, and different standards used in the collection of data in different counties certainly pose a threat to the validity and generalizability of the results. is also important to acknowledge that this is an ecological study and individuals from poor households may have necessarily suffered from poor health or died in early ages or from pregnancyand childbirth-related complications. However, the second part of the research conducted at the household level was an attempt to minimize these limitations. A comparison of the results 101

PAGE 114

of models to the results of household level data analyses may give very useful insights on the possible application of county-level data in further research. Findings from Data Analysis of Household Interviews Descriptive Statistics Sociodemographic Indicators; The sociodemographic characteristics of the study sample are shown in Table 5.7. Table 5.7. Household demographic characteristics Variables Mean SD Min Max Household size 5 6 2.3 2.0 12.0 Mean education years 6.6 2.1 2.5 10.7 Mean occupational r a nk 1.2 0.4 1.0 3.0 Percent of members workforce 68. 2 24 7 0 0 100 0 In 78.3% (94 households) of cases heads of the households were married (11.7% not married, 9.2% widowed, and 0.8% divorced). Ten percent (12) of households were female-headed and 10.8% (13) single male headed The majority of female heads were widowers whereas the majority of 102

PAGE 115

single male heads were young adults who had not been married and lived with elderly parents. Almost 83% (99) of households were traditional herding families (multigenerational herding), the rest were households who had begun livestock herding during the transition period. Socioeconomic Indicators. Socioeconomic status was measured by a combination of the variables shown in Table 5.8. Table 5.8. Household socioeconomic status Variables Mean SD Min Max Number of livestock in SFU 453.9 562.2 0.0 4896.0 Number of milk livestock in SFU 56.1 57.3 0.0 371.0 Yearly household income in 1928.0 1475.0 103.0 8568.0 thousand tugrics Number of household items 2:6 1.2 0.0 6.0 Number of transportation items 1.1 0.8 0.0 3.0 Large expenses in thousand 1030.0 1079.0 30.0 5900.0 tugrics Monthly food expenses per 14.0 11.0 0.0 75.0 person in thousand tugrics Meat (kg) consumed per person 192.0 146.0 17.0 975.0 per year 103

PAGE 116

Interestingly, the share of milk livestock in a total herd was very small12%. The contribution of other activities to supplement household income were negligible. A few households were engaged in carpentering, bootmaking, or growing vegetables, but these were small-scale activities with a primary purpose of meeting a household's own needs rather than selling for a profit. The majority of people said that their livelihood had improved during the transition period (51 households 43.6%). Only a small proportion of households indicated that their livelihood had deteriorated (21 households 17.9%). For others it stayed about the same. The main reason given for improved livelihood was having one's own herd and being able to control the income that comes from livestock production. Social Capital Indicators. Social capital indicators are described in Tables 5.9 5.14. 104

PAGE 117

Table 5.9. Household social network Variables Mean SD Min Max Number of households in 1.1 1.0 0 4 Number of other family members 9.7 4.5 1 23 Number of other family members 5.4 4.0 0 20 in the same county Number of close relatives 5.4 4.9 0 30 Number of close friends 6.8 12.3 0 20 Number of people turned for help 1.4 2.6 0 16 The frequency of meeting with family members and friends was strongly associated with the number of family members and friends, thus frequency of meetings was not used in further analyses. The majority of study participants were native residents of their counties and knew their county governors very well or well. In such cases, a question of whether or not the household asked their county governor a favor was a better indicator of their relationship with the county governor. Seventy six point seven percent (92) of households had not asked their county governor for any favor. More than half of households had access to loans and livestock restocking programs (60.8%). Responses to situational questions asked during the interviews such as "Whom do you talk about health issues?" or "If you had suddenly to go away for a couple of days, do you have someone to take care of your children?" did 105

PAGE 118

not show any variation in responses. All households had at least someone to do these favors for them and these were mostly relatives. Table 5.10. Trust and solidarity Variables Mean SD Min Max Trust people 2.1 0.6 1 Trust relatives 3.4 0.7 2 Trust friends 3.2 0.7 1 Trust local government 2.6 0.7 1 Trust press and media 2.5 0.8 1 Trust in relatives and friends was much higher than trust in others (Table 5.10). Seventy-one percent of study households mentioned that the level of trust in their community had gotten worse during the transition period, as opposed to nine percent who thought it had gotten better. 106

PAGE 119

Table 5.11. Collective action and cooperation Variables Mean SD Min Max Number of communal activities 1.5 2.9 0 24 participated Number of common issues 0.6 l.3 0 10 discussed Likelihood to cooperate on 3.5 0.7 1 4 building a well Likelihood to help in case of illness and disaster 3.6 0.6 1 4 Fifty-five percent of households reported that with the transition period, the willingness of people to help each other had decreased, as opposed to 19.5% who thought it had increased. Table 5.12. Information and communication Variables Mean SD Min Max Number of information sources 2.3 0.9 0 5 about government activities Number of information sources 2.1 0.9 1 5 about market prices Distance (kms) to the nearest 25.2 14.8 0 80 phone 107

PAGE 120

The most important sources of information about government activities were radio (83%), TV (66%) and social networks (37%). Radio was also the most important information source for market prices (75%), followed by social networks (49%), TV (42%) and local shops and markets (30%). Newspapers and magazines have become a rare commodity in the countryside since the transition: there is no longer an infrastructure in rural areas for the distribution of print media. Radio has thus become the most important information source, especially for the poorer households who cannot afford to purchase a satellite dish, power generator and TV. Eighty percent of households listen to radio or TV every day. Compared to the pre-transition period, access to information has increased for the majority of households (57%), but has decreased for 21 % of households. Table 5.13. Social cohesion and iriclusion Variables Mean SD Min Max Wealth differences 1.5 0.7 1 3 Frequency of having food/drinks 3.4 7.0 0 30 Frequency of playing games 1.8 5.1 0 30 Number of visitors in your 92.3 111.8 2 900 Number of other anyone 37.7 57.1 0 450 from your household visited Likelihood to help newcomers 2.7 1.0 1 4 Area safety 1.8 0.7 1 3 108

PAGE 121

Table 5.14. Empowerment and political action Variables Mean SD Min Max Your impact on neighborhood Likelihood of local government listening to people 2.1 1.9 0.7 0.7 A number of questions were asked about social activities. These 3 3 questions measure the domain in social capital of social cohesion and inclusion. There were almost no cases of joint petitions to government officials and political leaders, so this question was omitted from the analyses. Also, almost all households voted in the last presidential election and it was not necessary to include this question. Herd Management Indicators. The average distance households moved within a year was 206 kms (SD=162.4), but the distance varied greatly within ecological zones. Fifty seven percent of households moved more times compared to the pre-transition period, as opposed to 15% of households who moved fewer times. Frequencies of responses to herd management variables are given in the next table. 109

PAGE 122

Table 5.15. Herd management Variables Response N Percent Yes 12 10.0 Have livestock insurance No 108 90.0 Have permanent winter Yes 81 67.5 shelter for animals No 39 32.5 Have conflict over Yes 87 72.5 pasture No 33 27.5 Yes 102 85.0 Have conflict over water No 18 15.0 Yes 105 87.5 Prepared hay No 15 12.5 Yes 99 82.5 Prepared mineral lick No 21 17.5 Yes 96 80.0 Had a vet exam No 24 20.0 Problem of serious Yes 21 17.5 overgrazing No 99 82.5 Yes 16 13.3 Had distant moves No 104 86.7 Natural Disaster Indicators. Animal loss and frequencies of droughts and are presented below. This information reflects the experience of natural hazards in the past 15 years. 110

PAGE 123

Table 5 .16. Natural disasters Variables Mean SD Min Max Animal loss in SFU 355.5 382.2 0 2090 Number of years 1.7 0.9 0 Number of drought years 1.5 1.1 0 Number of and droughts 0.8 0.8 0 3 combined Another set of questions was asked to assess the efforts taken to mitigate livestock loss during These include: availability of additional human resources, winter shelter for animals, and fodder from the State Emergency Fodder Fund and medical services; supplies of food, warm clothes and milk substitutes from the Mongolian Disaster Mitigation and Management Agency; purchase of hay and fodder from private companies; presence of an early warning; and migration during a The responses are shown in Table 5.17. 111

PAGE 124

Table 5.17. Aid! Activities Response N Percent Additional human Yes 74 36.1 resources No 131 63.9 Fodder supply from Yes 33 16.6 the state fund No 166 83.4 Purchased fodder Yes 122 62.9 No 72 37.1 Yes 56 27.2 Supply of food No 150 72.8 Supply of warm Yes 15 7.3 clothes No 190 92.7 Availability of Yes 65 68.4 medical services No 141 31.6 Supply of milk Yes 38 18.4 substitutes No 168 81.6 Yes 106 52.0 Moved during No 98 48.0 Received an early Yes 95 46.1 warnmg No 111 53.9 Roads inaccessible Yes 85 47.5 No 94 52.5 Had a well-kept Yes 140 68.3 winter shelter No 65 31.7 In 52.4 % of cases, droughts preceded The following Table 5.18 describes the conditions/activities present during droughts: 112

PAGE 125

Table 5.18. Drought Aid/Activities Response N Percent Split animals among Yes 23 12.5 different households No 161 87.5 Yes 51 27.7 Had reserve pasture No 133 72.3 Yes 55 30.2 Purchased grass No 127 69.8 Yes 28 15.2 Conflict over pasture No 156 84.8 Reduced food Yes 42 23.0 No 141 77.0 Shortage of dairy Yes 121 66.1 products No 62 33.9 Received food aid Yes 5 2.7 No 179 97.3 Received an early Yes 62 33.9 wammg No 121 66.1 Nineteen percent (35) of droughts were considered very serious and 35.3 % (65) as serious. On average, during dry summers households had to move nine times. Animal loss during droughts was not significant. Health Indicators. Continuous and biriary health variables measured at an individual level are presented in Tables 5.19 and 5.20. Some of these variables (underweight diagnosed and treated) did not yield meaningful results in multilevel data analysis due to a small sample size (141 children) or small variations in responses. 113

PAGE 126

Table 5.19. Continuous outcome variables Variables Mean SD Min Max Hemoglobin g/dl 13,5 2.1 5 19 BMI 21.7 5.4 10 37 Triceps skinfold (mm) 14.1 7.9 4 41 Abdominal skinfold (mm) 19.9 13.1 2 54 Table 5.20. Binary outcome variables Variables Presence N Percent Anemic Yes 91 24.1 No 287 75.9 Yes 107 29.2 Overweight No 259 70.8 Yes 22 6.0 Underweight No 344 94.0 Sick Yes 131 26.4 (in the past six months) No 365 73.6 Yes 112 87.5 Diagnosed No 16 12.5 Treated Yes 105 82.0 No 23 18.0 114

PAGE 127

Regional Differences Statistically significant differences in predictor variables are shown in Tables 5.21 and 5.22. Variables that did not result in statistically significant regional differences are not shown in these tables. There were no differences among counties on any of the health outcome variables. Pastoralists in Khovd county camped in larger groups, prepared more hay, and had the largest number of milk animals when compared to other counties. Herders in Bayan-Ondor county had larger families, resided farther away from the county center, visited each other more frequently, felt safer about their environment, and reported more years of drought and combined, and possessed a greater number of household items when compared to households in other counties. Households interviewed in Olziit county had the fewest numbers of families in their and moved far greater distances. Residents in Bayankhutag county showed more willingness to cooperate on building and reconstructing wells, expressed a greater likelihood to help each other in case of illness and disaster and help newcomers from other counties prepared more mineral lick, had the largest number of animals, but had the fewest number of household items, vet exams, and experienced the fewest 115

PAGE 128

Table 5.21. One-way ANOVA results: regional differences Parameters Sum of Squares df Mean Square F Sig. Number of households in Between Groups 14.3 3 4.8' 5.568 .001 Within Groups 99.5 116 .9 Total 113.9 119 Number of other family members Between Groups 618.3 3 206.1 13.716 .000 Within Groups 1742.9 116 15. 0 Total 2361.2 119 Trust friends Between Groups 4.5 3 1.5 3.515 .018 Within Groups 48.0 113 .4 Total 52.5 116 Likelihood to cooperate on building a well Between Groups 7.2 3 2.4 5.959 .001 Within Groups 46.1 114 .4 Total 53.3 117 Likelihood to help in case of illness and Between Groups 5.6 3 1.9 4.936 .003 disaster Within Groups 43.9 116 .4 Total 49.5 119 Distance to the nearest phone Between Groups 6621.4 3 2207.1 13.174 .000 Within Groups 19433.9 116 167.5 Total 26055.3 119 116

PAGE 129

Table 5.21. One-way ANOVA results: regional differences (Cont.) Parameters Sum of Squares df Mean Square F Sig. Number of visitors in your Between Groups 158913.5 3 52971.2 4.629 .004 Within Groups 1327411.5 116 11443 2 Total 1486325 0 119 Number of other anyone from your Between Groups 25245.4 3 8415.1 2.697 .049 household visited Within Groups 355731.8 114 3120.454 Total 380977.0 117 Likelihood to help newcomers Between Groups 17.9 3 6.0 7.482 000 Within Groups 89.6 112 8 Total 107.6 115 Area safety Between Groups 20.5 3 6.8 17.586 .000 Within Groups 45.0 116 .4 Total 65.5 119 Distance moved within a year Between Groups 340431.4 3 113477 1 4.704 .004 Within Groups 2798413 8 116 24124 3 Total 3138845.2 119 117

PAGE 130

Table 5.21. One-way ANOVA results: regional differences (Cont.) Parameters Sum of Squares df Mean Square F Sig. Hay prepared in kg Between Groups 1167510659.1 3 389170219.7 60.128 .000 Within Groups 744319017.2 115 6472339.3 Total 1911829676.3 118 Mineral lick prepared in kg Between Groups 36347799.0 3 12115933.0 10.698 .000 Within Groups 131373043.0 116 1132526.2 Total 167720842.0 119 Frequency of veterinarian exams Between Groups 5.4 3 1.8 3.947 .010 Within Groups 52.0 114 .5 Total 57.4 117 Number of years Between Groups 14.6 3 4.9 6.673 .000 Within Groups 84.6 116 .7 Total 99.2 119 Number of drought years Between Groups 18.4' 3 6.1 5.247 .002 Within Groups 135.6 116 1.2 Total 154.0 119 Number of droughts and combined Between Groups 7.6 3 2.5 4.156 .008 Within Groups 70.4 116 .6 Total 78.0 119 118

PAGE 131

""--------------------------Table 5.21. One-way ANOVA results: regional differences (Cont.) Parameters Sum of Squares df Mean Square F Sig. Number of livestock in SFU Between Groups 3391327.7 3 1130442.6 3 833 .012 Within Groups 34214872.9 116 294955.8 Total 37606200.6 119 Number of milk livestock in SFU Between Groups 80504.1 3 26834.7 10.0'17 .000 Within Groups 310767.5 116 2679.0 Total 391271.6 119 Number of household items Between Groups 17.7 3 5.9 4.511 .005 Within Groups 151.6 116 1.3 Total 169.3 119 Number of transportation items Between Groups 13.5 3 4.5 7.949 .000 Within Groups 65. 7 116 .6 Total 79.2 119 119

PAGE 132

Table 5.22. Post-hoc test results: regional differences (I) (J) Mean Std. Variables Difference Sig. soum code soum code (I-J} Error Number of Bayan-Ondor Olziit .433 .239 .273 households in Khovd -.467 .239 .213 Bayankhutag .300 .239 .594 Olzi1t Bayan-Ondor -.433 .239 .273 Khovd -.900(*) .239 .001 Bayankhutag -.133 .239 .944 Khovd Bayan-Ondor .467 .239 .213 Olziit .900(*) .239 .001 Bayankhutag .767(*) .239 .009 Bayankhutag Bayan-Ondor -.300 .239 .594 Olziit .133 .239 .944 Khovd -.767(*) .239 .009 Number of Bayan-Ondor Olziit 4.067(*) 1.001 .001 other family Khovd 3.333(*) 1.001 .006 members Bayankhutag 6.333(*) 1.001 .000 Olziit Bayan-Ondor -4.067(*) 1.001 .001 Khovd -.733 1.001 .884 Bayankhutag 2.267 1.001 .112 Khovd Bayan-Ondor -3.333(*) 1.001 .006 Olziit .733 1.001 .884 Bayankhutag 3.000(*) 1.001 .017 Bayankhutag Bayan-Ondor .-6.333(*) 1.001 .000 Olziit -2.267 1.001 .112 Khovd -3.000(*) 1.001 .017 Trust friends Bayan-Ondor Olziit .200 .168 .636 Khovd -.180 .170 .713 Bayankhutag .340 .171 .199 Olziit Bayan-Ondor -.200 .168 .636 Khovd -.380 .170 .119 Bayankhutag .140 .171 .845 *The mean difference is significant at the .05 level. 120

PAGE 133

Table 5.22. Post-hoc test results: regional differences (Cont.) (1) (J) Mean Std. Variables Difference Sig. soum code soum code (1-J) Error Trust Khovd Bayan-Ondor .180 .170 .713 friends Olziit .380 .170 .119 Bayankhutag .521(*) .173 .017 Bayankhutag Bayan-Ondor -.340 .171 .199 Olziit -.140 .171 .845 Khovd -.521(*) .173 .017 Likelihood Bayan-Ondor Olziit .200 .164 .616 to cooperate Khovd .167 .164 .741 on building Bayankhutag .167 .054 a well Olziit Bayan-Ondor -.200 .164 .616 Khovd -.033 .164 .997 Bayankhutag -.631(*) .167 .001 Khovd Bayan-Ondor -.167 .164 .741 Olziit .033 .164 .997 Bayankhutag -.598(*) .167 .003 Bayankhutag Bayan-Ondor .167 .054 Olziit .631(*) .167 .001 Khovd .598(*) .167 .003 Likelihood Bayan-Ondor Olziit .067 .159 .975 to help in Khovd case of .159 .062 illness and Bayankhutag -.200 .159 .590 disaster Olziit Bayan-Ondor -.067 .159 .975 Khqvd .333 .159 .159 Bayankhutag -.267 .159 .339 Khovd Bayan-Ondor .159 .062 Olziit -.333 .159 .159 Bayankhutag -.600(*) .159 .001 Bayankhutag Bayan-Ondor .200 .159 .590 Olziit .267 .159 .339 Khovd .600(*) .159 .001 *The mean difference is significant at the .05 level. 121

PAGE 134

Table 5.22. Post-hoc test results: regional differences (Cont.) (I) (1) Mean Std. Variables Difference Sig. soum code soum code {I-J} Error Distance to Bayan-Ondor Olziit 11.633(*) 3.342 .004 the nearest Khovd 19.467(*) 3.342 .000 phone Bayankhutag 16.567(*) 3.342 .000 Olziit Bayan-Ondor -11.633(*) 3.342 .004 Khovd 7.833 3.342 .094 Bayankhutag 4.933 3.342 .455 Khovd Bayan-Ondor -19.467(*) 3.342 .000 Olziit -7.833 3.342 .094 Bayankhutag -2.900 3.342 .821 Bayankhutag Bayan-Ondor -16.567(*) 3.342 .000 Olziit -4.933 3.342 .455 Khovd 2.900 3.342 .821 Number of Bayan-Ondor Olziit 64.033 27.620 .100 visitors in Khovd 87.000(*) 27.620 .011 your Bayankhutag 90.867(*) 27.620 .007 Olziit Bayan-Ondor -64.033 27.620 .100 Khovd 22.967 27.620 .839 Bayankhutag 26.833 27.620 .766 Khovd Bayan-Ondor -87.000(*) 27.620 .011 Olziit -22.967 27.620 .839 Bayankhutag 3.867 27.620 .999 Bayankhutag Bayan-Ondor -90.867(*) 27.620 .007 Olziit -26.833 27.620 .766 Khovd -3.867 27.620 .999 Number of Bayan-Ondor Olziit -17.133 14.423 .636 other Khovd .744 14.547 1.00 anyone from Bayankhutag 24.054 14.547 .353 your household Olziit Bayan-Ondor 17.133 14.423 .636 visited Khovd 17.877 .14.547 .610 Bayankhutag 41.187(*) 14.547 .028 *The mean difference is significant at the .. 05 level. 122

PAGE 135

Table 5.22. Post-hoc test results: regional differences (Cont.) (I) (1) Mean Std. Variables Difference Sig. soum code soum code (1-1} Error Number of Khovd Bayan-Ondor -.744 14.547 1.000 other Olziit -17.877 14.547 .610 anyone from Bayankhutag 23.310 14.670 .389 your household Bayankhutag Bayan-Ondor -24.054 14.547 .353 visited Olziit -41.187(*) 14.547 .028 Khovd -23.310 14.670 .389 Likelihood Bayan-Ondor Olziit .447 .237 .240 to help Khovd .583 .233 .065 newcomers Bayankhutag -.414 .235 .297 Olziit Bayan-Ondor -.447 .237 .240 Khovd .136 .235 .939 Bayankhutag -.861(*) .237 .002 Khovd Bayan-Ondor -.583 .233 .065 Olziit -.136 .235 .939 Bayankhutag -.997(*) .233 .000 Bayankhutag Bayan-Ondor .414 .235 .297 Olziit .861(*) .237 .002 Khovd .997(*) .233 .000 Area safety Bayan-Ondor Olziit .533(*) .161 .007 Khovd 1.033(*) .161 .000 Bayankhutag .967(*) .161 .000 Olziit Bayan-Ondor -.533(*) .161 .007 Khovd .500(*) .161 .012 Bayankhutag .433(*) .161 .040 Khovd Bayan-Ondor -1.033(*) .161 .000 Olziit -.500(*) .161 .012 Bayankhutag -.067 .161 .976 Bayankhutag Bayan-Ondor -.967(*) .l61 .000 Olziit -.433(*) .161 .040 Khovd .067 .161 .976 *The mean difference is significant at the .05 level. 123

PAGE 136

Table 5.22. Post-hoc test results: regional differences (Cont.) (J) Mean Std. Variables Difference Sig. soum code soum code (I-J) Error Distance Bayan-Ondor Olziit -79.267 40.103 .203 moved Khovd 11.967 40.103 .991 within a year Bayankhutag 70.167 40 103 .303 Olziit Bayan-Ondor 79.267 40.103 .203 Khovd 91.233 40.103 .110 Bayankhutag 149.433(*) 40.103 .002 Khovd Bayan-Ondor -11.967 40 .103 .991 Olziit -91.233 40.103 .110 Bayankhutag 58.200 40.103 .470 Bayankhutag Bayan-Ondor -70 167 40.103 .303 Olziit -149.433(*) 40.103 .002 Khovd -58.200 40.103 .470 Hay Bayan-Ondor Olziit -261.833 656.878 978 prepared Khovd -7476 833(*) 656.878 000 in kg Bayankhutag -585.247 662.517 .813 Olziit Bayan-Ondor 261.833 656.878 .978 Khovd -7215.000(*) 656.878 000 Bayankhutag -323.414 662.517 .962 Khovd Bayan-Ondor 7476.833(*) 656.878 000 Olziit 7215.000(*) 656.878 000 Bayankhutag 6891.586( ) 662.517 .000 Bayankhutag Bayan-Ondor 585.247 662.517 813 Olziit 323.414 662.517 962 Khovd -6891.586(*) 662 517 000 Mineral Bayan-Ondor Olziit 778.500(*) 274.776 027 lick Khovd 118.500 274.776 973 prepared Bayankhutag -771.633(*) 274.776 029 in kg Olziit Bayan-Ondor -778.500( ) 274.776 027 Khovd -660.000 274.776 082 Bayankhutag -1550.133(*) 274.776 .000 The mean difference is significant at the .05 level. 124

PAGE 137

Table 5.22. Post-hoc test results: regional differences (Cont.) (J) Mean Std. Variables Difference Sig. soum code soum code (I-J) Error Mineral Khovd Bayan-Ondor -118.500 274.776 .973 lick Olziit 660.000 274.776 .082 prepared Bayankhutag -890.133(*) 274.776 .008 in kg Bayankhutag Bayan-Ondor 771.633(*) 274.776 .029 Olziit 1550.133(*) 274.776 .000 Khovd 890.133(*) 274.776 .008 Frequency Bayan-Ondor Olziit -.003 .176 1.000 of vet Khovd .379 .177 .147 exams Bayankhutag .463(*) .176 .047 Olziit Bayan-Ondor .003 .176 1.000 Khovd .383 .176 .136 Bayankhutag .467(*) .174 .042 Khovd Bayan-Ondor -.379 .177 .147 Olziit -.383 .. 176 .136 Bayankhutag .0'84 .176 .964 Bayankhutag Bayan-Ondor -.463(*) .176 .047 Olziit -.467(*) .174 .042 Khovd -.084 .176 :964 Number of Bayan-Ondor Olziit .100 .221 .969 Khovd .100 .221 .969 Bayankhutag .867(*) .221 .001 Olziit Bayan-Ondor -.100 .221 .969 Khovd .000 .221 1.000 Bayankhutag .767(*) .221 .004 Khovd Bayan-:Ondor -.100 .221 .969 Olziit .000 .221 1.000 Bayankhutag .767(*) .221 .004 Bayankhutag Bayan-Ondor -.867(*) .221 .001 Olziit -.767(*) .221 .004 Khovd -.767(*) .221 .004 *The mean difference is significant at the .05 level. 125

PAGE 138

Table 5.22. Post-hoc test results: regional differences (Cont.) (I) (J) Mean Std. Variables Difference Sig. soum code soum code (1-1) Error Number of Bayan-Ondor Olziit .333 .279 .632 drought Khovd .800(*) .279 .025 years Bayankhutag 1.000(*) .279 .003 Olziit Bayan-Ondor -.333 .279 .632 Khovd .467 .279 .343 Bayankhutag .667 .279 .085 Khovd Bayan-Ondor -.800(*) .279 .025 Olziit -.467 .279 .343 Bayankhutag .200 .279 .890 Bayankhutag Bayan-Ondor -1.000(*) .279 .003 Olziit -.667 .279 .085 Khovd -.200 .279 .890 Number of Bayan-Ondor Olziit .533(*) .201 .045 and Khovd .333 .201 .351 droughts Bayankhutag .667(*) .201 .007 combined Olziit Bayan-Ondor -.533(*) .201 .045 Khovd -.200 .201 .753 Bayankhutag .133 .201 .911 Khovd Bayan-Ondor -.333 .201 .351 Olziit .200 .201 .753 Bayankhutag .333 .201 .351 Bayankhutag Bayan-Ondor -.667(*) .201 .007 Olziit -.133 .201 .911 Khovd -.333 .201 .351 Number of Bayan-Ondor Olziit -90.633 140.227 .917 livestock inSFU Khovd -129.467 140.227 .792 Bayankhutag -446.133(*) 140.227 .010 Olziit Bayan-Ondor 90.633 140.227 .917 Khovd -38.833 140.227 .993 Bayankhutag -355.500 140.227 .060 *The mean difference is significant at the .05 level. 126

PAGE 139

Table 5.22. Post-hoc test results: regional differences (Cont.) (J) Mean Std. Variables Difference Sig. soum code soum code (1-J) Error Number of Khovd Bayan-Ondor 129.467 140.227 .792 livestock Olziit 38.833 140.227 .993 in SFU Bayankhutag -316.667 140.227 .114 Bayankhutag Bayan-Ondor 446.133(*) 140.227 .010 Olziit 355.500 140.227 .060 Khovd 316.667 140.227 .114 Number of Bayan-Ondor Olziit 41.033(*) 13.364 .014 milk Khovd -27.333 13.364 .178 livestock in SFU Bayankhutag 25.067 13.364 .244 Olziit Bayan-Ondor -41.033(*) 13.364 .014 Khovd -68.367(*) 13.364 .000 Bayankhtitag -15.967 13.364 .631 Khovd Bayan-Ondor 27.333 13.364 .178 Olziit 68.367(*) 13.364 .000 Bayankhutag 52.400(*) 13.364 .001 Bayankhutag Bayan-Ondor -25.067 13.364 .244 Olziit 15.967 13.364 .631 Khovd -52.400(*) 13.364 .001 Number of Bayan-Ondor Olziit .867(*) .295 .021 household Khovd .633 .295 .145 items Bayankhutag 1.000(*) .295 .005 Olziit Bayan-Ondor -.867(*) .295 .021 Khovd -.233 .295 .859 Bayankhutag .133 .295 .969 Khovd Bayan-Ondor -.633 .295 .145 Olziit .233 .295 .859 Bayankhutag 367 .295 .602 Bayankhutag Bayan-Ondor -1.000(*) .295 .005 Olziit -.133 .295 .969 Khovd ,..367 .295 .602 *The mean difference is significant at the .05 level. 127

PAGE 140

Table 5.22. Post-hoc test results: regional differences (Cont.) (I) (J) Mean Std. Variables Difference Sig. soum code soum code (I-J) Error Number of Bayan-Ondor Olziit -.200 .194 .733 transportKhovd .233 .194 .627 tation items Bayankhutag .700(*) .194 .003 Olziit Bayan-Ondor .200 ;194 .733 Khovd .433 .194 .121 Bayankhutag .900(*) .194 .000 Khovd Bayan-Ondor -.233 .194 .627 Olziit -.433 .194 .121 Bayankhutag .467 .194 .082 Bayankhutag Bayan-Ondor -.700(*) .194 .003 Olziit -.900(*) .194 .000 Khovd -.467 .194 .082 *The mean difference is significant at the .05 level. Natural Disaster The number of number of droughts, number of and droughts combined, number of days roads were inaccessible, availability of help in the form of human resources, fodder, food, warm clothes, health care, milk substitute, purchase of fodder, distant moves, and presence of an early warning and winter shelter for animals were regressed against the number of animals lost in SFU. In the results, the number of days roads were inaccessible (B=2, SE=0.51, p
PAGE 141

loss. More days roads were closed were associated with greater animal loss. But more frequent droughts and were predictive of smaller animal loss. is possible that herders who live in droughtand areas are more experienced and can overcome the hazards with minimal losses. is also likely that the days roads were closed in an indicator of dzud severity. Drought per se did not result in livestock loss. Interestingly, government aid did not prevent livestock loss by any means. Socioeconomic Status Multivariate regression analyses were conducted using ML WIN statistical software to explore the relationship between various sociodemographic, social capital, herd management and natural disaster variables, and household socioeconomic status. The household socioeconomic status (an outcome) was represented by a number oflivestock in SFU, number of milk livestock in SFU, yearly household income in thousand tugrics, number of household items, number of transportation items, large expenses in thousand tugrics, monthly food expenses per person in thousand tugrics, and meat consumed per person per year. Each outcome variable was entered into the model separately. Number of Livestock in SFD. The average number of livestock in SFU was 452.86 (Table None of the models were significant except Model 4 129

PAGE 142

(p
PAGE 143

to herding have more years of education but may have fewer livestock due to the fact that they came to herding later and have relatively less experience. 131

PAGE 144

Table 5.23. Fixed and random part results: number oflivestock in SFU Parameters Fixed Constant Sociodemo graphic Model Model 2 Model 3 Model 4 Model 5 452.9 682.4: 753.2--I 493.8 5-2-3-.0--......... _______ 1-188.3 (141.1L! -237.7 (154.6) -69.2 (146.62 : -71.7 (147.9L_ Percent of members in workforce 1.1 (2.1) 1.7 (2.3) 1.9 (2.1) : 1.9 (2.1) -------1 Jl..umber ---o-------! -62.8 (57.9) -107.2 __ _1J3.1_(57.5_ Social capital Number of other family members 21.4 (19.1) 2.2 (18 Number of other family members in the same county -23.6 (20.6) -9.6 (19.2) -9.5 (20.3) Number of close friends-r-12.5 (5.0)* 9.2 (4.9) : 8.7 (5.0) _____________________ 1 _______ ________ 1 ___ ____ Asked local g.Qvernor f9r help -58.3 (129.5) -19.8 (121.1) -12.1_(1211)_ Number ofjJeople turned for help 10.4 (23.0) 22.8 (21.2) 24.2 (22.0) ----.--.. ----.-.-.. ---Access to loans/restocking ** significant at the .05 level significant at the .01 level 132

PAGE 145

Table 5.23. Fixed and random part results: number oflivestock in SFU (Cont.) Parameters Social capital Model 1 Model 2 Model 3 Model 4 Model 5 Trust friends -251.6 (103.0) -180.9 (92.2)* -177.2 (95.3) Trust local government I 5.3 (95.4) -19.4 (87.5) -27.8 (91.3) Trust press and media -39.'0 (77.2) 1.0 (71.1) : -4.8 (73.2) -------------------:----;--. -----1------.. ------r--.------------.. .. --. ---... --------------------:--------------------. --------actIVItIes -30.7 (27.9) -58.23 (26.8)* -54.9 (28.5) common issues -10.8 (40.0) -7.6 (36.4) -7.6 (37.0) 1--.. ---....... -------------.. ----.-----------------------.---------------. -----to cooperate on : -33.3 (87.5) -60.2 (81.8) -54.7 (84.2) a well to help in case of 211.2 (104.5)* 219.7 (98.5)* 230.0 (102.3)* Illness/dIsaster Number of information sources 1-------[--------1 --64.7 -60.8 (67.2) -62.3 (71.4) -5\.0 (77.3) _ -1-_ ___ _________ ______ _____ ..7.:1J ___ Wealth difference 33.9 (79.6) 16.0 (71.0) 15:7 (73.5) of having food/drinks [ 2.1 (9.6)--_ -3.9 (8.8) -4.5 (9.3) Frequency of playing_games 2.4 (11.0) 3.0 (9.7) 2.9 Number of visitors in your -0.9 (0.5) -1.8 (0.5)** -1.8 (0.5)** significant at the .05 level significant at the .01 level 133

PAGE 146

Table 5.23. Fixed and random part results: number oflivestock in SFU (Cont.) Parameters Natural disaster significant at the .05 level significant at the .01 level Modell Model2 134 Model 3 Model 4ModelS

PAGE 147

Table 5.23. Fixed and random part results: number oflivestock in SFU (Cont.) Parameters Model 1 Model 2; Model 3 Model 4 Model 5 Natural : Number of and droughts 23.06 (93.53) __ combineQ __ ____ _________ _--"--___ ___ ________ 1 ___________ _________ Random -------.--_---. -_._1561.1 153Q.6significant at the .05 level significant at the .01 level significant at the .001 level 135

PAGE 148

Number of milk livestock in SFD. Models 2 through 4 in Table 5.24 were' statistically significant (p
PAGE 149

statistically significant (B=3.3, SE=1.6, p
PAGE 150

Table 5.24. Fixed and random part results: number of milk livestock in SFU Parameters Model 1 Model 2 Model 3 Model 4 Model 5 Fixed ____ 56.1 31.0 -51.5 -79.5 -82.4 n ------._n u Social capital N b fh h ld 6 8 ___ -:l4 CiJ.4 -0.6 { 4.41_ -0.2 (1.4) 0.2 1.4 urn ero ouse 0 sm Number of other family members 1.7 { Number of other family members J 0.4 ( same county Number of close friends 1.7) 1.8 (1.5) 1.0 (1.5) 0.9 (0 --+ Number ofIJeople turned for help 3.3 ( .4 ) 0.1 (0.4) 0.2 (0.4) LQL -2.1 9.4 Access to --i 15.0 ( Trust people -9.7 ( 1.9 2_._ 3 .1.&,,(11 10.7 2 11.3 (9.2) 13.3 (9.3)_ 8 3) -1.4 (7.3) (7.22_ ----_ _____ ._. ____ ._ .. ______ _____ __ J ___ 7 __ -8.0 (7.2) -8.4 {7.2)_ Trust friends Trust local government Trust press and media significant at the .05 level significant at the .01 level 138 -10.4 2.2 { T 22.5 (6 7.92 -1.9 (6.8) -0.8 6 .9) .4)** 22.4 (5.5)** 23.9 (5.5)*

PAGE 151

Table 5.24. fixed and random part results: number of milk livestock in SFU (Cont.) Parameters Social capital Model 1 Model 2 Model 3 Number of communal activities 1.2 (2.3) Number ofcominon issues discussed 1.9 (3.3) Likelihood to cooperate on building a well 4.7 (7.2) LIkelIhood to help m case of Illness/dIsaster : 1.1 (8.6) Model 4 ModelS -0.6 )_1-_-0--,-. 6 (2.2 0.7 (2.8) 0.5 (2.8) 9.6 8.1 ----;-------.----------------------:--------_._-------------------------_._ ------sources about -2.9 (6.2) -4.1 (5.2) -0.1 (5.4) ofmformatlOn sources about market 5.2 (6.4) ___ : ___ (5.2!_ Frequency of having food/drinks 1.7 (0.8)* : 1.7 (0.7)* 2.0 (0.7)** -,------------____ ___ ___ ___ _______ _______ -1-_____ -----.--__ Number of visitors in your 0.07 (0.04LL9.01 (0.04) 0.01 (0.04) Number of other anyone from your family visited -0.1 (0.1) 1 Your on neighborhood -2.1 (5.6) -6.3 (5.1) 1_;7.3 (5.0) Likellihood oflocal government listening to 6.4 (7.4) 2.0 (6.5) 3.2 (6.6) peop e significant at the .05 level significant at the .01 level 139

PAGE 152

of milk --------.-.-----.----.----l---------,--.-----.;. -----.--... .-. -----------_.--. .. -. -.--.. -.---.--,----.-'----.-.---_----1 -----------------,-----. ------... Prepared hay ____ ___ 1 ___ ____ ___ i_L -15.1 (9.9L l {l0.0)_. __ ________ _ ----1--..-. --Number ofdro:ught years dzuds f------326M 28433 1517.0 r 952_5-(420.9) (367.1) (211.4) 40. ? t J_ 1132.7) *** .01 level

PAGE 153

Yearly household income in thousand tugrics. The mean yearly household income (Table 5.25) for households was 1928 thousand tugrics which in 2005 is equivalent to about US$ 1750. All of the models were statistically significant in predicting income except for Model 5, where natural disaster variables were added to previous models. Of the sociodemographic characteristics of households analyzed in Model 2, household size was the only significant predictor. Yearly income increased as households became larger (B=219.72, SE=57.53, p
PAGE 154

Table 5.25. Fixed and random part results: yearly household income in thousand tugrics Parameters I Modell Model 2 Model 3 Model 4 Model 5 Fixed Constant 1928 584 496 185 344 Sociodemo__ __ ._n ____________ n _____ n __ n ___ .:_ _, graphic' Mean education years 16 (62) : -29 (75) -103 (68) -119 (73) c-Mean .. (39 i) -Percent of members workforce 7 (5) 11 (6) 10 (6) 10 {) Social households ______ 14 7J_J_.:.161_( 152) capital of other family members I -23 :?JJizt+_ Number of other family members in the -33 (55) -15 (51) -23 (54) -------i 1. ____ L ....... Number of people turned for help -0.4 (61) 22 (56) 5 (58) --: to loans/ restockmKprograms 914 (344)** 830 809 (325)* Trust people L208 (268) I -99 (248) .----.. .-Trust local government -t 370 (254) 256 (233) 246 (242) Trust press and media 172 (20Srl---129 (189)1-142 (194)-significant at the .05 level significant at the .01 level 142

PAGE 155

Table 5.25. Fixed and random part results: yearly household income in thousand tugrics (Cont.) Parameters Social capital ; Model 1 Model 2 Model 3 Model 4 Model 5 Number of communal -t----__ .. ___ : :J30 -{71)-L --l19 (]6) issues (97) : 41 (98) I LIkelIhood to on bmldmg a well : 27 (233) T-71 (217) -65 (223) to helIU.n __ Number of sources about' -82 (200) -172 (178) -1l5 (189) government actIVItIes _. __ ._ .. _----_ ... _._ ...... __ _----_ .... _._-_._---_. ... --+---_ .. _--. _.----_._--.--------_._._ sources about : 3 (206) (181) to the nearest Qhone -5 -2 -4 12) Wealth difference 87 (212) 48 (189) Fre _quen.y of having food/drinks 43 33 (25) of playing_games 10 (29) 5 (26) 6 (26) __ .. ... ____ _._._ ... _l__ ... __ ___ . __ ..... .... -1._::. L._ anyone 1 0.1 (3) -2 (3) LIkelIhood to help newcomers -215 (171) [ 190 (155) -156 (163) Area ___ __ ___ 465 (231)* l 444 people significant at the .05 level significant at the .01 level 143

PAGE 156

Table 5.25. Fixed and random part results: yearly household income in thousand tugrics (Cont.) Parameters Model 1 Model 2 Model 3 Model 4 Model 5 Herd .... ..... J]]J __ 6 (5 .. management : Have permarient winter shelter -2 (337) -15 (338) Have conflict over pasture -563 (370) -592 (376) --1'" ..... . -+ ... ... Prepared mineral lick L-----+---_ __ l_ -122 (3931_. -89 (406) Had a vet exam ._1._ : 484 (340) .472 (348) __ Problem of serious overgrazing I 096 (381 )** 1014 (409)* Natural Animal loss in SFU 0.2 (0.3) disaster Number of years -149 (222) Number of drought years 116 (153) of and droughts combmed Rand-o-m---! 2157731 1901817 1584492 1177531 1163751 __ ._. .. __ .. __ _(220794) __ C!.6j085L _____ (162165) __ 2091 2076**[ 1763***[ 1732*** 1731 -2LL *** significant at the .05 level significant at the .01 level significant at the .001 level 144

PAGE 157

Number of household items. On average, households possessed 2.6 household items such as TVs, radios, satellite dishes, etc. (Table 5.26). Cash income of pastoralist households can greatly vary depending on a season and market fluctuations and may not be an accurate measure of socioeconomic status. The number of household items may be used as a proxy for socioeconomic status where self-reported income data may not be reliable. Model 2 (household sociodemographic variables) and Model 3 (social capital variables) showed statistically significant reductions of -2 logIikelihood (p-values were less than 0.05 and 0.001 respectively). In Model 2, the number of household items was positively associated with household size (B=0.17, SE=0.05, p
PAGE 158

Table 5.26. Fixed and random part results: number of household items Parameters Model 1 Model 2 Model 3 Model 4 Model 5 Fixed I -!; Constant : 2.58 1.12: -1.16 I -2.07 -_2_.2_4 __ Sociodem-o---+--f_!I-_ ___ .. ___________ ... .9.: .. _I graphic Mean education 0.0110.05) __ 0.01 0.01,.{0.05)_ Mean occupational rank ; -0.09 (0.29) -0.27 (0.29) -0.16 (0.30) r -0.18 (0.29) i-------. _____ ---l' Percent of members workforce 0.01 (0.00) 0.01 (0.00) 0.01 (0.00) 0.01 (0.00 Social capital : Number of households in -0.22 (0.11)* -0.26 (0.11)* 'b-0.23 (0.11)* Number of other family members : 0.10 (0.04)* I ,-... ------------+-----1-------+---__---'--: Number of other family members -0.04 (0.04) -0.04 (0.04) -0.07 (0.04) m t e same county! ; Number of close friends (0-:-01) [-9.01 (0.01) 0.01 (0.01) ____ ------1--_____ ___ __ _____ ____ ____ __ Q_"_91jO.01t_ : Asked local governor for help 0.44 (0.24)* 10.25)_1_ 0.27 J0.25)_ Number for help __ ---+_ 0.06 (0.04) 0.06 (0.04) 0.03 (0.04) Access to loans/restocking _________ 0.04 (0.24) -0.06 (0.24) -0.15 (0.24) .. ----------t---------Trust friends 0.26 (0.19) 0.33 (0.19) 0.43 (0.19)* significant at the .05 level significant at the .01 level 146

PAGE 159

Table 5.26. Fixed and random part results: number of household items (Cont.) Parameters Model 1 Model 2 Model 3 Model 4 ModelS Social capital : Trust local g'-'-o_ve_rn_m_e_nt ________ +----"-'--'o .-=--0-,-7--,-_--'-0-'-.0'--1----'(, 0 _.1_8'-) -t-_0_._04_-,Trustpress and media -0.05 (0.14) : -0.01 (0.15) 0.06(0.15) ______ _________ __ ___ ; Number of common issues discussed -0.09 -0 : 07 -0.07 (0.07)_ : Likelihood to cooperate on building a well c------------------_lt_-0.01 0.08 Likeliho Q d -to help in case 0.32 (0.20) 0.39 (0.20) : 0.27 (0.20) Number of information sources about -0.10 (0.14) \ -0.1_.1 (0.14) -0.03 (0.14) : government activIties ___ +----" __ ___ r -Numberofinformationsources about 0.28 (0.15) 0.35 (0.14)* 0.31 (0.14)* ; market prices : : Distance to the nearest Qhone -0.01 (0.01) -0.01 (0.01) -0.01 (0.01) !Wealthdifference 0.16(0.15) : 0.16(0.15) 0.20(0.15) y-0.00 (O.OO)--r-o.ooCO.Oo)Number of other anyone from your : family visited 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) !Likelihoodtohelpnewcomers -0.19(0.12) 1-0.21(0.12) -0.17(012) Area safety 0.33 (0.16)* \ 0.36 (0.17)* 0.45 (0. 16E_ I Your impact on neighborhood 0.11 (0.13) 0.06 (0.14) 0.08 (0 13) significant at the .05 level significant at the .01 level 147

PAGE 160

Table 5.26. Fixed and random part results: number of household items (Cont.) Parameters Modell! Model 2 Model 3 Model 4 Model 5 Social capital : Likelihood of local government -0.21 0 29 (0 17) -0.28 (0.17) [listening to people (0.17) -. -H-e-r-d---+i-H-a-ve--'livestock insurance (0.38) management -1------1-----------Have conflict over water---r---Q.44 (.28) __ I 0.15 (0.36) 0.07 (0.37) mineral lick ________ JQ.31L __ iliad a vet exam : __ -0.09 (0.26) -0.16 (0.26) : Problem of serious overgrazing 0.34 (0.30) 0.09 (0.31) : Had distant moves 0.54 (0.32) 0.56 (0.31) -------f----:. Animal loss in SFU : 0.00 (0.00) Natural disaster Number of and droughts 0.08 (0.19) combined Random significant at the level significant at the .01 level significant at the .001 level 148

PAGE 161

Number of transportation items. On average, a household owned one transportation item (a truck, a sedan, or a motorbike) (Table 5.27). Only Model 3 and Model 5 were statistically significant in predicting ownership of a means of transport. In Model 3, where social capital variables were entered into the model, the number of information sources about government activities was inversely correlated with the outcome (B=-0.21, SE=O.11, p
PAGE 162

Table 5.27. Fixed and random part results: number of transportation items Parameters Modell: Model 2 Model 3 Model 4 ModelS Fixed Constant Sociodemo graphic : 1.08 0.36 -1.23 I -1.68 -1.81 _ .. __ _____ __ ___ __ Mean education years (0.042.. ___ Q_-9.&i0.04.)_r-_0.05 (0.04) __ }1eap. .. _______ ---l-=0.09 _.-9..1 7 ____ -+-_Pe_ r ce n t -'-of. members in workforce 0.00 (0.00) 1 0.00 (0.00) 0.00 (O.OO)..-J 0.00 (0.00) ___ __ Social capital Number of other family members -0.01 (0.03) 0.02 (0.0]) -0.01 (0.03) Number of other family members in the I 0.02 (0.03) -0.01 (0.03) -0.01 (0.03) same county Number of close friends 0.00 (0.01) 0.00 (0.01) 0.00 (0.01) Number of close relatives -0.02 (0.02) -0.01 (0.02) 0.00 (0.02) --_ u __ (ol]j---Number for help 0.00 0.00 (0.03) -0.01 (9.03L ccess to 10ans/restockinKprcigrams 0.09 (0.18) 0.11 (0.17) Trust people 1-0.07(0)4) -0.12(0.14) -0.11(0.13) r -----Trust local government 0.17 (0.13) 0.17 (0.13) 0.19 (0.13) Trust press and media 0.02 (0.11) -0.05 (0.11) -0.01 (0.10) significant at the .05 level significant at the .01 level 150

PAGE 163

Table 5.27. Fixed and random part results: number of transportation items (Cont.) Parameters Social capital Model 1 Model 2 Model 3 Model 4 ModelS Number of communal activities participated -0.03 (0.04) -0.04 (0.04) -0.04 (0.04) Number of common issues discussed -----r-. -0-.-04------'(0-.0-6-" ) 0.04 (0.05) 0.02 (0.05) .: -j-__ __ Number o.fmformatlOn sources about : 0.18 (0.11} 0.20 (0.10)* 0.16 (0.10) Distance to the nearest phone 0.00 (0.01) 1) Wealth difference : : 0.10(0.11) 0.12(0.11) 0.10(0.10) Frequency of having food/drinks 0.02 (0.01) 0.02 (0.01)* 0.03 (0.01)** VISI e 0.00 ___ [ -1--Your impact on neighborhood -0.14 (0.09) -0.18 (0.10) -0.21 (0.09)* of local government listening to -0.03 (0.12) -0.08 (0.12) -0.08 (0.12) significant at the .05 level significant at the .01 level 151

PAGE 164

Table 5.27. Fixed and random part results: number of transportation items (eont.) Parameters Model 1 Model 2 Model 3 Model 4 ModelS Herd Have livestock insurance 0.09 (0.28) -0.05 (0.27) management ______ __1_ m m __ -_un n __ r------------------.----------____ r-_0.01 (0.21) -0.040 20) haY __ 0.42 (0.26) 0.45 Prepared mineral lick 1 0.15 (0.22) 0.19 (0.22) f-H;ad a "yet exam __ ._. __ r------------________ ___________ Problem of serious overgrazing -0.05 (0.23) 0.42 (0.22) Had distant moves -0.05 (0.22) Natural Animal loss SFU 0.00 (0 .00) disaster of years -0.05 (0.12) Number of drought years 0.28 (0.08)** ------. -----------r------------r--------------------.------Random. of and droughts combined -0.26 (0.13 ** significant at the .05 level significant at the .01 level significant at the .001 level 152

PAGE 165

Large expenses in thousand tugrics. The mean amount households spent on large purchases were 1,030 thousand tugrics which is an equivalent of about US$ 935 (Table 5.28). These expenses usually were college tuition, purchase of transportation, satellite dish or battery, and holiday preparations. All of the models except Model 5 showed a significant -2 loglikelihood reduction. From household sociodemographic variables in Model 2, larger households (B=94.26, SE=42.85, p<0.05) and larger number of household members in a workforce (B=10.30, SE=3.90, p
PAGE 166

Table 5.28. Fixed and random part results: large expenses in thousand tugrics Parameters Modell Model 2 Model 3 Model 4 Model 5 Fixed Constant 1030 -499 -55 -846 -1366 SociodemoHousehold size ." .. _._.91:.(?QL. ____ __ .. __ __ 59 .i5.,Q) ._._ ------.------------"._-------_._---. ----.-----------------.--graphic Mean education years 60 (462 6 (56) Mean rank ---82 (264) _..J.0 9 --_ ... ... _._---Percent of members in workforce 10 11 (5)* 6 (4) 6 (4) Social Number of households in ___ -:2. 1112)_ -146 (115) ._=21JV_?.1 ----------f----------_capital Number of other family members -2 (37) 9 (36) __ Number of other family members in the 48 (41) 34 (40) 8 (41) same county Number of close friends 2 (10) 4 (10) 5 {I02 Number of close relatives ___ .:..1.5._ 125 ___ t_ _____ .' ____ ___ __ ._' ___ _________ ._. ______ -____ ___ .0 '_-. __ -________ 0 __ -______ ___ "."._ ----Asked local governor for hell' 185 (251) 304 (257) 76 (248_) __ _____ 42 ___ (412 __ ** Access to loans/restockinKl'rograms Trust people 1----Trust relatives ._-----_. -----Trust friends Trust local government Trust press and media significant at the .05 level significant at the .011evel ----------_. _---... ---------.154 458 (2582 466 (244) 369 (246)_ -225 (201) -178 (194) -174 (190) _. _.163 '!UJ!12L -547 (205)* -412 (191)* -308 (191) -53 (182LF 43 (183)_. -15 (154) 77 (148) 166 (147)

PAGE 167

Table 5.28. Fixed and random part results: large expenses in thousand tugrics (Cont.) Parameters Soci a l capital Modell Model 2 Model 3 Model 4 Model 5 Number of communal activities particip a te d-i-____ -f--_ Number of common issues discussed -73 (80) -31 (76) -35 (74) abou t -58 (139) -23 (143) government activities Number of information sources about marketJ)rices 89 (154) 170 (141) 151 (138) Distance to the nearest phone I 3 (9) 7 (9) 0.8 (9) Wealth difference 174 (158) 153 (147) 231 (147) ____ _____ ____ Number of visitors in your 0.3 (1)_ -0.1 (1) O.UD __ Number of other anyone from your -2 (2) -3 (2.0) -4 (2)* family visited t.9 he!p ____________ J ____ _______________ __=i2 __ (1211 __ __ safety -26 91 (163) Your impact on neighborhood -95 (136) 17 (137) 43 (133) Likelihood of local government listening to people 45 (179) 162 (173) 152(174) significant at the .05 level significant at the .01 level 155

PAGE 168

Table 5.28. Fixed and random part results: large expenses in thousand tugrics (Cont.) Parameters Herd management Natural disaster : Model 1 Model 2 Model 3 Model 4 Model 5 : Animal loss in SFU 0 3 Number of years 160 (168) combined t.--)4403((1181 7 6 of and droughts ------'1'---_ _-------.,-._-_._-+-------+-----+-----+--------+, ---_._-Random ------'1---------------1-----+1 1155127 1055011 886572 718916 667130 _cr_2e ___ (136201) (123541) (100179) (93003) 2015.71 2004* I 1703*** 1681 ** 1674 significant at the .05 level significant at the .01 level significant at the .001 level 156

PAGE 169

Monthly food expenses per person in thousand tugrics. Monthly food expenses per person were 13,630 tugrics, an equivalent ofUS$ 12 (Table 5.29). Both sociodemographic (Model 2) and social capital (Model 3) variables had a significant effect on the monthly food expenditure (p<0.01 and p
PAGE 170

Table 5.29. Fixed and random part results: monthly food expenses per person in thousand tugrics Parameters Fixed Constant Sociodemo graphic Social capital Modell Model 2 Model 3 Model 4 Model 5 l 13 63 8 84 -2 62 -3 95 -2 22 Household size -0.52 (0.42) -0.97 (0.46)* -0.79 (0.48) -0.91 (0.50) -. -----"-----. --.--------------.----. --. .1. ________ ... -----.----____________ -0.16 (Q .. 52L -0.52 (0.52) -0.53 (0.55) Mean occuRational rank _.J. ) 1.29 (2.572 -0.69 (2.86t ___ 1.11 (2.97LJ_ 1.64 (2.96) Percent of members in workforce 0.13 (0.04)** 0.13 (0.04)** 0.12 (0.04)** QJ1J0.04)** of households in -+______ : 0.05 ___ 9-=-QQ_ D.:l?) of other family members 1--_ ___ 0.18 (0.35) 0.26 (0.36) 0.25 __ Number of other family members in -0.41 -0.58 (0.39) -0.56 (0.41) the same county of close friends 0.05 (0.09) tJ[1 0-(-0-.1-0--) 0 -.12 ( -0.1 -0)== Number ofeople turned for 0.55 (0.42) ___ 0.68 (9-=--43)_1 _-.9.50 {0.44) Access to 10ans/restockinKprograms : -0.60 (2.402 -0.97 (2.39) -1.85 (2.46) Trust people -3.94 (1.87)* -5.10 (1.89)** -5.55 (1.90)** -----0--. Trust local government -1.97(1.77) -2.52 (1.78) -2.61 (1.83) Trust press and media -0.07 (1043) 0.52 (1.44) 0.61 (1.47) ** significant at the .05 level significant at the .01 level 158

PAGE 171

Table 5.29. Fixed and random part results: monthly food expenses per person in thousand tugrics (Cont.) Parameters Model 1 Model 2 Model 3 Model 4 Model 5 Social Number of communal activities participated 0.37 (0.52) -0.16 (0.54) -0.07 (0.57) capital Number of common issues discussed : -1.08 (0.74) -0.80 (0.74) -0.62 (0.74) ----+-----: -------i-s sources about ---1 0.59 (1.39) 1.16 _government actlvItles Number o.finformation sources about r -0.03 (1.43) 0.22 (1.38) 0.02 (1.38) markeU2f1ces : Distance to the nearest phone (0.09) 0.10 0 09 ) : 0.09 (0.09) Wealth difference -0.81 (1.47) -0.83 (1.44) -0.68 (1.47) Frequency of having food/drinks -0.49 (0.18)** -0.63 (0.18)** 1 -0.66 (0.19)** Nun:ber other anyone from your -0.01 (0.02) -0.02 (0.02) -0.02 (0.02) j'amlly VISIted i -=.0.21jhJ9) ___ (J.2}1 __ Area safety 2.37 2.28 (1.62) I 2.75 (1.63) Your impact on neighborhood -3.24 -2 .16 (1.34) -1.98 (1.33) I I Likelihood oflocal government listening to -1.87 (1.66) -1.13 (1.69) -0.79 (1.74) people significant at the .05 level significant at the .01 level 159

PAGE 172

Table 5.29. Fixed and random part results: monthly food expenses per person in thousand tugrics (Cont.) Parameters Modell Model 2 Model 3 Model 4 Model 5 Herd Have livestock insurance 6.69 (3.80) 7.03 (3.83) management __ __ __ __ __ J_ --;Hi Prepared hay 1.55 (3.54) __ j).61i3 7 Natural disaster Animal loss in SFU I 0.00 (0.00) -1.04(1.68) Number of drought years! : ______ 6L_ and droughts combined 2.37 (1.87) Prepared ----!-0 26 (3 QQ2 0 17 (3 08) a vet ____ ___ ___ r-= _____ ; __ Problem of serious overgrazing -3.68 -5.30 (3.09)_ Had distant moves I 4.75 (3.11) 4.87 (3.10) Random ---I __ __ (14.67) (9.57) -2LL 908.47 893.01** 739.30*** 727.94 *** significant at the .05 level significant at the .01 level significant at the .001 level 160 66.72 (9.30) 724.96

PAGE 173

Meat (kg) consumed per person per year. On average, a person consumed 192 kg of meat per year (Table 5.30). That is almost 530 g of meat each day, including children. Meat consumption increases with a larger proportion of household members in workforce (B=1.63, SE=0.52, p<0.01) in Model 2. In Model 3, a greater number of other family members (B=1O.33, SE=4.57, p<0.05), higher trust in relatives (B=51.78, SE=22.18, p
PAGE 174

Table 5.30. Fixed and random part results: meat (kg) consumed per person per year Parameters Fixed Constant Sociodemo graphic : Model Model 2 Model 3 Model 4 Model 5 192.53 141.13 63.56 58.02 45.17 _______ m_. .. __ ; ___ n L __ education years : -0.34 ( 6.18) -1.09 (6.76) (6.43L_ -6.74 ( 6.88) rank I -10.44 (36.83) 32.75 m.01) ___ -:-Percent of members in workforce! 1.63-<0.52)** 1.86 (0.55)** 1.52 (0.52)** : 1.59 (0.53)** Number of households in: 13.65 (13.88) 12.80 (13.90) 12.30 (14.39) Social capital ; ---------4.46}** : 11.89 Numberofotherfamilymembersin! -4.94(4 94) -6.94(4.82) -7.21(5. 08) the same county Number of close friends 1.08 (1.20) 2.13 (1.23) 1.94 (1.24) _____ _ ________ _____ ___ Asked local governor for L-38.90 15.50 (30.43) 15.44 (31.05L Number of IJeople tl!rned for _____ J)...=(5.50L 2 } 3 (53O)i 11.11 (5.50)* Access to 10ans/restockinKprograms 29.57 QJ.03) 10.47 (29.57) 11.28 (30.73) Trust eople -6.12 (24.19) -14.20 (23.43) -11.16 (23.80) Trust relatives : 51. 78 (22.18)* 50.19 (21.66)* 51.95 (21.77)* -----22.92 ( 2i16) T 2U -S(2i86) Trust loc .al government -31.62 (22.88) -40.67 (21.99) -40.78 (22.86) Trust press and media -12.32 (18.50) -2.21 (17.86) -3.19 (18.32) significant at the .05 level significant at the .01 level 162

PAGE 175

Table 5.30. Fixed and random part results: meat (kg) consumed per person per year (Cont.) Parameters Social capital Model 1 Model 2 Model 3 Model 4 Model 5 Number of communal activities participated -5.21 (6.69) II -14.43 (6.72)* -14.45 (7.14)* Number of common issues discussed 451 (958) 412 (915) 300 (9.26) r-governmentactivities I (18.01) (16.87)* 1 (17.86)* Number of-information sources about 36.33 43.54 45.58 market prices (18.52)* (17.12)* Distance to the nearest phone LUQ) ______ 0.03 t109LJ 0.15 (1.12) __ Wealth difference 14.20 (19.09) 16.72 (17.83) 17.68 (18.40) ----11 --.-..... --.---. ------.f.-_________ 1 .. -(0:-14)--Number of other your j---household visited -0.13 (0.24) -0.42 -0.42 (0.25) Likelihood to helg}lewcomers ___ _:)0.39 Area : -12.23 (20.80) -18.20 (20.05) -21.09 (20.42) Your impact on neighborhood -48.79 -26.10 -25.87 Likelihood of local government listening to people (16.62) 22.00 (21.56) 28.41 (20.89) 22.76 (21:72) significant at the .05 level significant at the .01 level 163

PAGE 176

Table 5.30. Fixed and random part results: meat (kg) consumed per person per year (Cont.) Parameters 1 Model 1 1 Model 2 Model 3 Model 4 Model Herd Have livestock insurance 1 141.80 (47.12)** 138.35 (47 81)** management Have permanent winter shelter J -1.90 (31.84) -3.19 (31.97) ___ __ Have conflict over water ---. .1--------' ________ (34.92) ___ ____ ---t-----.. -.. : -6.63 : -8.43 Prepared mineral lick -63.19 (37.15) -59.50 (38.44) --1 Had distant moves 98.74 (38.47)* 100.86 (38.75)** Natural Animal loss 1. -0.03 (0 03) disaster Number of years.__ -1 11.60 (21.03 Random combmed --------L-------------l-------+-------------+--21127.21 18872.91 12867.84 10534.92 10422.76 (2727.51) (2436.48) (1793.09) (1468.01) (1452.38) 1535 .54-1522.o -6**-I-f266:94***-j--i----1245.23 ---*** significant at the .05 level significant at the .01 level significant at the .001 level 164

PAGE 177

Health Status Multilevel modeling was conducted with 496 individuals at level one and 120 households at level two to assess the relationships of predictor variables (age and gender of individuals, household sociodemographic, socioeconomic, social capital and natural disaster variables) and health outcomes. All predictor variables except gender and age were measured at the household level. The health outcome indicators were measured at the individual)evel. Multilevel modeling of outcome variables such as presence of underweight, diagnosed and treated have not converged to meaningful results. Hemoglobin in g/dl. Modell (Table 5.31) provides a mean hemoglobin level with no adjustment for any individualor household-level predictors. Individual predictors such as gender and age were entered into Model 2. The decrease of -2 loglikelihood in the Model 2 was statistically significant at p
PAGE 178

Model 4: a larger number of household items (B=O.32, SE=O.11, p
PAGE 179

Tri.ble .31. Fixed and random part results: hemoglobin in g/dl Parameters Modell: Model 2 Model 3 Model 4 Model Model 6 _F_i_x __ ed __ ___ Predictors Agel ce;'teredL ___ (O.OJ** (0.01)** -().04(O:O )*. I 0,06 ((l.0 1)** Household Predictors +_ size 0.1 (0.1) 0.1 (0.1) Mean education years I 0.2 (0.1)** 0.2 (0.1)* 0.2 (0.1)* 0.2 (0.1)** _ean occupational 1 _____ -1.0 (0.4)** f----0.7 __________________ m! 0.00 (0.01) 0.0 (0.01) 0.0 (0.0) .0 (0.0) SocioeconomIC ---------1 ------------------... ----!. ... significant at the level ** significant at the .01 level 167

PAGE 180

Table 5.31. Fixed and random part results: hemoglobin in g/dl (Cont.) Parameters Social capital Model 1 Model 2 Model 3 Model 4 : Model 5 Model 6 ____ _+__ I Number of;ther family members in the 0.1 (0.1) 0.1 (0.16) __ --------------------------------------------_ 1 ___ ------------__ Number of close friends t--0.0 (0.01) -0.0 (O.OL close relat __________ r -0.1 (0.1 ) Asked local governor for help I 0.1 (0.4) -0.0 {O.4L Number of turned for help -0.0 (0.1) 0.0 (0.1) ______ _ ______ Trust people -0. 7 (0.3)* -0.6 (0.3)* I Trust relatives 0.4 (0.3) 0.5 (0.3) friends ____ 0.6 (0.3) -0.5 (0.3 L Trust ______ ____ ________ ____ ___ ; ----L----______ Trust ress and media I : 0.0 (Q.2L__ 0.1 {O.2L_ of communal activities -------------I -0.14 (0.07)* -0.2 (0.1)* partlclpated -----------------------'1---1:!kelihoo(to on building a well r--------+-(0.3L Likelihood to help iri case of I I illness/disaster -0.0 (0.3) -0.1 (0.3) significant at the .05 level significant at the .01 level 168

PAGE 181

Table 5.31. Fixed and random part results: hemoglobin in (Cont.) Parameters : Modell: Model 2 Model 3 Model 4 ModelS Model 6 Social Number of information sources about ___________ ________ ____________ '-____________ ________________ _______ '-_ ____ ____ _______ __ capital Number o.f information sources about -0.2 (0.2) : -0.1 (0.2) market pnces -------1-----------\"---------; ---------t --i(o.oi) [ ---o.oO(Mi)--;-0.1 (0.2)--! 0.2 (0.2) Frequency : -----: : (0.6)--! (0.03)* _____________ ------------1--------------:----------1 ------------:-----------f-------------,--------------Likelihood to hel newcomers 0.0 (0.2) -0.0 (0.2) -:----R lIstenmg to people significant at the .05 level significant at the .01 level 169

PAGE 182

Table 5.31. Fixed and random part results: hemoglobin in g/dl (Cont.) Parameters Natural disaster significant at the .05 level significant at the .01 level significant at the .001 level : Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 170

PAGE 183

Presence of anemia. Hb levels in g/dl (Table were converted to a binary variable (anemic-1/not anemic-O) using WHO anemia cut-offlevels described in the methods section. A multilevel logistic regression analysis was performed to explore the independent variables that predicted the presence of anemia. In Model 2, being male was associated with lower risk of anemia (B=0.61, SE=O.2S, p
PAGE 184

relationship in the opposite direction (B=-1.24, SE=O.55, p
PAGE 185

Table 5.32. Fixed and random part results: presence of anemia Parameters I Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 -1.2 -o.iiO])*:I:o I'rediC1orsrg<>lce11lereilL uu. _-, _,
PAGE 186

Table 5.32. Fixed and random part results: presence of anemia (Cont.) Parameters 1 Model 1 Model 2 Model 3 Model 4 ModelS Model 6 Socio(centeredL; ______ ___ ; _____ economic Amount spend on ; 0.0 (0.0) 0.0 (0.0) 0.0 (O.OL 0.01 (0.0)* 0.1 (0.1)* 0.1 (0.1) n:!0nt!?:ly food per person __ Meat consumed per : person (centered} 1 _____ ---:-_______ -:--_ ___ -S-oc ia l Number of households in 0.0 (0.0) 0.0 (0.0) 0.0 (0.0) capital !:-___ -+______ -+__ ___ Number of other family -0.6 (0.3) -0.5 (0.4) other family -----r --1-members in the same 0.2 (0.1) 0.2 (0.1) -----1----0.1 (0.1) -0.2 (0.1) of close --:---:--------!-_O.O---,(Ml O.QiO.OL 0.0 (0.1} 0.1 Number of close :--:.--Asked local governor for (0 ) 0 (0 8) helL_ Number of people turned 0.0 (0.1) -0.0 (0.1) for help_ Access to .------0.4 (0.7) 0.5 (0.9) loans/restocking programs : significant at the level ** significant at the .01 level 174

PAGE 187

Table 5.32. Fixed and random part results: presence of anemia (Cont.) Parameters Social capital Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 175

PAGE 188

Table 5.32. Fixed and random part results: presence of anemia (Cont.) Parameters Social capital Natural disaster significant at the .05 level significant at the .01 level : Modell: Model 2 Model 3 : Model 4 176 ModelS Model 6

PAGE 189

Body Mass Index. The mean overall BMI was 21.71 (Table 5.33). In Model 2, gender and age were entered into the regression equation. As expected, males had a slightly lower BMI (B=-1.43, SE=O.38, p
PAGE 190

Table 5.33. Fixed and random part results: Body Mass Index Parameters Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 -.... --. ... : ... -.-... -. ...... ---:.--.. -().1.CO.-J)--oj (C).iTdemographic Mean education years 0.2 (0.1) : 0.1 (0.1) : 0.5 (0.2)** 0.4 (0.2)** Mean .-.. ---.-:-:... --... -0.5 (l.Ol. of members m 0.0 (0.0) : 0.0 (0.0) 0.0 (0.0) 0.0 (0.0) -.--.--1--.-. ---...... .--.-.. .. -... .... -... -.-... _n .. -.--. SocioNumber of livestock O.OO(OQL : --.9:Q.{O.OL_. economic Number of milk livestock : : 0.00(0.0) : -0.02 (0.01)* 1 -0.02 (0.01)* of transportation :1 ..; 0.7 (0.3)* :-0.9 (;4)* 0.7-;0.4; ems : -.----.--.. ----. ----. : i--------r ----i ----------.-----Meat consumed per person 0.0 (0.0) 0.0 (0.0) 0.0 (0.0) significant at the .05 level significant at the .01 level 178

PAGE 191

Table 5.33. Fixed and random part results: Body Mass Index (Cont.) Parameters Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Social _____ ____ capital _____________ --: ----------1--____ --L-______ : _________ __ .. ( ................. Number of close relatives-; -----1 (0.1)1--0.0 (0.1) -Askedlocal -------------r-----------. ---u-r--------;------Number turned for help _________ J_-______ ____ 0.2 Access to 10ans/restockinKprograms _ _____ 0.0 (0.7) [ 0.1 (0.7) Trust L : -0.4 (0.6) -0.5 (0.6) i ----------------1Likelihood in of ill disaster :-----------f-_ I 0.6 (0.6) 1 9 (0.6L_ Number sources about .: -0.7 (0.4) -0.7 (0.4) government actIvItIeS! significant at the .05 level significant at the .01 level 179

PAGE 192

Table 5.33. Fixed and random part results: Body Mass Index (Cont.) Parameters Social capital Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Numkber o finformation sources about 0.7 (0.4) 0.8 (0.5) mar et l)flces ------: l----:--+ ----1==1-. -____ ___ ____ ,-: your household vlSlted Likelihood to help ___ ____ _______ 0.4 (0.3) --____ l!t.2 L. ikeli.hood of local government 1 = I I 0.6 (0.4) 0.4 (0.4) e Na-tu-r-al----+-A-n-im-alloss in SFU ..J: 0:0(0.6) disaster l of years .. =P: --0-.6-->-(0-.5-<-)----1 -------combmed significant at the .05 level significant at the .01 level 180

PAGE 193

Table 5.33. Fixed and random part results: Body Mass Index (Cont.) Parameters Model 1 Model 2 Model 3 Model 4 ModelS Model 6 significant at the .05 level significant at the .01 level significant at the 0.01 level 181

PAGE 194

Overweight. A multilevel logistic regression analysis was conducted to define the factors associated with ovelweight (Table A binary variable was created following the guidelines developed by CDC and the Ministry of Health in Mongolia as described in the methods section. Overweight (inclusive of obesity) was coded as I and others as O. is important to note that in this analysis overweight was regarded as a favorable condition, indicating a better nutritional status. As in the analysis ofBMI, gender and age were statistically significant predictors of overweight. Males were two times less likely to be overweight compared to females (OR=0.S3, 9S%CI=0.31-0.90). The risk of overweight increased with age.(B=O.OS, SE=0.01, p
PAGE 195

Table 5.34. Fixed and random part results: overweight Parameters Model 1 Model 2 Model 3 Model 4 ModelS Model 6 Fixed L__ Constant I : -1.0 -0.9 -2.6 -3.0! -9.4 : -9.8 -.. --:---;---------1 ------_. ----------.. .... --------, ------------;-------,--------------IndIvIdual (male) ____ _______ -0.6 -0.7 (O.4E...... __ =lJ_ (O.4E..... ....... -+ ... (0.0 l)**' .o,
PAGE 196

Table 5.34. Fixed and random part results: overweight (Cont.) Parameters Socio. economIC Social capital Model 1 Model 2 Model 3 Model 4 : Model 5 Model 6 184

PAGE 197

Table 5.34. Fixed and random part results: overweight (Cont.) Parameters Social capital Model 1 Model 2 Model 3 4 ModelS Model 6 __ 0.2 (Q.51_ .... --j-the phone -0.0 (0.0) -_________ _____ _____ ____ _______ J ____________ Frequency of having food/drinks 0.0 (0.1) 0.1 (0.1) ------------------------------------------D.1(D.1)--j---O-.l( 0.1 "---l-... ...... m Your il1!pact on neighborhood __ =---___ -0.0 (O.3Ll_ -0.0 (0.3L \ 1 Likelihood of local government listening to : 0.2 (0.4) people V.L. : significant at the .05 level significant at the .01 level 185

PAGE 198

Table 5.34. Fixed and random part results: overweight (Cont.) significant at the .05 level significant at the .01 level 186

PAGE 199

Triceps skinfold in mm. The overall mean triceps skinfold of individuals was 14.1S mm (Table S.3S). Males had lower triceps skinfold compared to females (B=-6.40, SE=O.66, p
PAGE 200

Table 5.35. Fixed and random part results: triceps skinfold in mm Parameters Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 j--Ge;;;ier n .. -_. -_---f = Q,2.JQJiQJ>2 Jl.2J9.:<>2rSociodemoHousehold size iI.2 0.2 (0.2) 0.5 (0.3L 0.4 (0.3) graphic Mean education __ ___ 0.2 (0.2) 0.6 (0.3)* 0.6 ------------1 SocioNumber oflivestock 1 0.0 (0.0) 0.0 0.0 (0.0) economIC N b f lk 1 k -0.03 -0.06 -0.06 urn 0 mI .__ _(0.01)** (0.01)** (0.01)_** household income O .QJQ.Q) 0.0 iQ.QL_ 0.0 # of household items 0.3 (0.4) 0.6 (0.6L 11 trans:Q0rtation items 1.0 ( 0.5) __ !JL{0.8) 0.9 (0.8) ____ __ O.OiQ.O) __ 0.0 (0.0) Monthly food expenses --1 -0.0 (0.1 -0.1 (0.1) -0.1 (0.1) Meat consumed per person 0.0 (0.0) 0.0 (0.0) 0.0 (0.0) significant at the .05 level significant at the .01 level 188

PAGE 201

Table Fixed and random part results: triceps skinfold in mm (Cont.) Parameters Social capital Model Model 2 1 Model 3 1 Model 4 Model Model 6 Number of households in 1 -11-_00 .. 81 QO,--.62))--; Number of other family members 1 I -,----------------------------------------------------1-------------------------i----------tJ------------1------------,------------Number of others in the same countL ___ -L __ 0.1 (0.2) ____ ---l-------'--t----Asked hel p _________ 1 __ -1.2 (1.2) CU)_ Number of people turned for help, ___ ..Q.:QJQ:_?Lj __ __ f--------------------------------------__ n ______ [ ______________________ 1 ---.----------------Access to loans/restocking programs -0.3 (1.3) : __________ ____ __ ___ :lQ_(L1L_ Trust relatives 0.7 (1.0) I 0:6 (1.0) ,-'Irust ________ L_ ________ -0.1 (1.2L 0.3 (1.2) __ Trust local government 1 --HO.6 (0.9) : -0.4 (l.OL 8 (Q. ..r:._;_.1.7 Number ofeommunal ________________ _______________________-0.1 -0.1 (0.3) Number of common issues discussed ----0.1 (0.3) -0.1 (0.3L -0.9 (0.7) -0.9 (0.7)* significant at the level significant at the .01 level 189

PAGE 202

Table 5.35. Fixed and random part results: triceps skinfold in mm (Cont.) Parameters Model 1 Model 2 Model 3 Model 4 ModelS! Model 6 Social Number o finformation sources about 0.3 (0.8) : 0.8 (0.8) capital _______ ____ ..: _.-----.--_____ 1. ________ ___ ___________ ______ -=--___ : Number of visitors in your 0.0 (0.0) Number of anyone from your 0.0 (0 0) 1 0 0 (0.0) ------i-------+----0.3 (0.6) _________ ___ _______ _____ __ __ ____ __ ___ ______ __________ Your imJlact on neighborhood __ __ -0.3 -0 5 (0.6L_ Likelihood of local government 0.5 (0.8) 0.0 (0.8) to people -+_ in SFU ____ + _____ ---t-_._ -----1--0-.0-(-0.0)--Number of years 0.2 (0.9) Number of drought years 0.5 (0.6) Number of and droughts combined 0 8 (0 9) disaster Natural significant at the .05 level significant at the .01 level 190

PAGE 203

Table 5.35. Fixed and random part results: triceps skinfold in mm (Cont.) Parameters Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 ______________ 6 _0,0 (
PAGE 204

Abdominal skinfold in mm. The overall mean abdominal skinfold was 19.87 mm (Table 5.36). In Model 2, both male gender (B=-8.58, SE=l.OO, p
PAGE 205

Table 5.36. Fixed and random part results: abdominal skinfold in mm Parameters Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 .. --19.91 24.1 .. 19.5 .. 1 ... 18.9 .:. 10.0. 7.5 -Inclividual--==--=--I-':SX (1.0)" -:8XCI .0)** 0 -8.4 (1.0)*'-r:8.311.2)".1 :8.3 (1.2)*' rge.fcentefl)(jL_--n 0 0.4 (0.03)**0.4 (O.03j**_p.4 -<0.03)** 0.4 (0.03)** SociodemoHousehold size 0.2 (0.2) 0.4 (0.3) 0.8 (0.4) 0.8 graphic Mean education years 0.4 (0.32 _____ o.3 (0.3) I 0.6 (0.4) 0.6 (0.4) ----1---00 _______ ___ ______ ___ __ __ _________ ____ .. Socio0.0 (0.0) (Q.Ol_L_O.O (0.0)_ economiC J':-Iumber of milk live.stock __ ____ ____ __ .. _. _____ -0.02@..QlLI -0.04 -0.04 (0.02)* Yearly household income 0.0 (0.0) 0.0 (0.0) 0.0 (0.0) items 0.4 (0.5) 0.3 (0.9) 0.3 Monthly food .. -0.0 Meat consumed per person 0.0 (0.0) 0.0 (0.0) 0.0 (0.0) significant at the .05 level significant at the .01 level 193

PAGE 206

Table 5.36. Fixed and random part results: abdominal skinfold in mm (Cont.) Parameters Social capital Modell Model 2 Model 3 Model 4 Model 5 Model 6 194

PAGE 207

Table 5.36. Fixed and random part results: abdominal skinfold in mm (Cont.) Parameters Social capital Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Number of information sources about market 0.7 (1.2) 1.3 (1.3) pnces ,: -------}----------------------J?istance the_ nearest _____ ______ _-___ ---.. Jiumber of visitors in your 0.0 (0.0) 0.0 (0.0) Number of anyone from your 0.0 (0.0) 0.0 (0.0) household VIsIted ------=--people -N-a-t-ur-a-l -------+--Animalloss in SFU 1 0.0 (0.0) disaster Number of years 0.0 (1.4) Number of drought years 0.9 (1.0) -Number of and droughts combined 1.2 (1.3) significant at the .05 level significant at the .01 level 195

PAGE 208

Table 5.36. Fixed and random part results: abdominal skinfold in mm (Cont.) Parameters Model 1 Model 2 Model 3 Model 4 ModelS Model 6 Random significant at the .05 level significant at the .01 level significant at the .001 level 196

PAGE 209

Sick (in the past six months). A logistic regression analysis was conducted with a binary dependent variable with the categories sick (coded as 1) and not sick (coded as 0) (Table S.37). In Model 2, both gender and age were significant predictors of sickness: males were two times less likely to report sickness than females (OR=0.S7, 9S%CI=0.37-0.88 or B=-0.S7, SE=0.22, p
PAGE 210

Table 5.37. Fixed and random part results: sick Parameters Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Fixed !-_ ------,____ -. -1.0 Constant (0.1) -2.2 (0.3) -0.9 (0.7) -1.1 (0.9) ; -1.8 (2.9) : -3.3 (3.1) -... : I Sociodemosize _______ 1-_ _____ : ______ -0.1 (0.1)_.L_ -0.2 graphic -------+---------.----P b ercent mem ers -0.0 (0.0) 0.0 (0.0) -0.02 (0.01)* : -0.02 (0.01)* ---1----. hYearly household income 1__ 0.0 (O.QLJ 0.0 (O.OL..J_..Q:9 (0.0 ) # household items : : -0.1 (0.1) -0.5 (0.2)* -0.6 # transportation items l -0.1 (0.2) 0.2 (0.3) 0.3 (0.3) ___ I ____ :.......J 0.0 (0.0) L 0.0 (0.0) 1 ___ 0.0 0 ) food expenses 0.0 (0.0) 0.0 (0.0) 0.0 (0.1) Meat consumed 0.0 (0.0) 0.0 (0.0) 0.0 (0.0) significant at the .05 level significant at the .01 level 198

PAGE 211

Table 5.37. Fixed and random part results: sick (Cont.) Parameters Social capital Number of other family members in the Wme-"ounty __ Number of close friends Number of close relatives Asked local governor for help Number of turned for help Access to Trust peoQle Trust relatives .-_ Trust friends _. ..IIUst ls>cal __ ______ ._ Trust press and media Number of communal activities pi\!!!cipated ._ _. .. ________ ._._ ___ __ Number of common issues discussed Likelihood to cooperate on building a well Likelihood to help in case of \ illness/disaster significant at the .05 level significant at the .01 level Model 1 Model 2 Model 3 Model 4 ModelS Model 6 '-'_._-----.--..... ... 0.1 (OJ) 0.0 (0.1) _-_ _-_,._--------------.---.------.. 9.Q_(0.0) 0.04 '----1 0.0 (0: 1) -0.0 (0.1) -----0.1 (0.5) -0.5 (0.62 -0.0 (0.1) 0.0 (0.1) -----------------0.6 (0.5) {0.6) 0.2 (0.5) -0.0 (0.4) -0.5 (0.4) -0.5 __ 0.5 (0.5) ----. --_."----_._. _., _------0.4) 0.3 (0.4) 0.5 {0.42 0.1 (0.1) -0.0 (0.1) -----_ 0.2 (0.1) -. 1) ... ---0.2 (0.4) -0.0 (0.5) 0.6 (0.4) 0.6 (0.5) 199

PAGE 212

Table 5.37. Fixed and random part results: sick (Cont.) Parameters Social capital Model 1 Model 2 : Model 3 Model ModelS Model 6 # information sources about gov. activities: 0.2 (0.3) -0.4 (0.3) __ ______ : _________ ___ : ___ __ _______ ____ -QAJ9:iL f_pistance to the nearest phone -_---J 0.0 (0.0) Wealth difference -0.9 (0.3)** ; -0.8 (0.3)* -----------------yrequency of having fdod/drinks : 0.0 (0.0) 0.0 (0.0) Frequency of playing gaines : '0.1 (0.0) 0.1 (0.0) -----------Number of visitors in your I 0.0 (0.0) 0.0 (0.0) --T ---(--:-----1-----I Likelihood to help newcomers : -0.3 (0.3) -0.3 (0.3) Likelihood oflocal government listening to 06 (0 3) -0.7 (0.4) people r-__ of and droughts -0.2 (0.4) Random -1---T-------0.0 (0.0) 0.0 (0.0) -! 0.0 (0.0) 0.0 (O.O)! 0.0 (0.0) 0.0 (0.0) significant at the .05 level significant at the .01 level 200

PAGE 213

Disaster and Health Logistic regression analyses were perfonned at the household level to identify factors that indicate a poor health outcome due to disasters. The dependent variables were nominal variables affects health" and "drought affects health" with binary responses "yes" or "no". Close to 47% of study households said that affects their health and 49.2% said that drought affects health. None of the variables on natural disasters were statistically significant predictors of responses to questions whether or not affects health in logistic regression analyses (Tables 5.38 and 5.39). However, a higher number of and droughts combined was associated (marginally significant) with a greater probability of responding that affects health (B=0.57, SE=0.32, p=0.08). Table 5.38. Logistic regression with affects health" as the dependent variable Variables Odds Ratio Total dzud loss in SFU .001 .001 1.001 Number of dzud years -.092 .277 .912 Number of drought -.044 .206 .957 years Number of dzuds and .571 .321 1.770 droughts combined Constant -.512 .443 .. 599 201 .316 .741 .832 .076 .247

PAGE 214

Table 5.39. Logistic regression with "drought affects health" as the dependent variable Variables Odds Ratio Total dzud loss in SFU .000 .001 1.000 Number of dzud years .109 .273 1.115 Number of drought years .172 .202 1.188 Number of dzuds and -.134 .311 .875 droughts combined Constant -.440 .436 .644 .443 .690 .395 .667 .313 The results of hierarchical logistic regression analyses performed with sociodemographic, socioeconomic and social capital variables as predictors showed no association with outcome variables affects health" or "drought affects health". 202

PAGE 215

CHAPTER 6 DISCUSSION AND CONCLUSIONS Vulnerability to natural hazards is becoming a major issue in international arenas, especially in semi-arid regions where the impact of hazards is expected to be the most severe given the natural conditions and poor development in this area. Mongolia is one of these countries, but the dependence of the country's economy on livestock production makes the situation even more serious. Frequent droughts, land degradation, deforestation, and are central problems to Mongolian agriculture which mainly consist oflivestock husbandry. The presence of natural events in the absence of social vulnerability is not sufficient to create disasters. Poor infrastructure development, human and social capital, and access to services, marginalization and social inequality tend to amplify the social consequences of hazards. These factors increase sensitivitY to hazards and reduce adaptive capacities of human populations, posing a critical threat to individual and household well-being. The main hypothesis of the research was that high levels of vulnerability, which is a composite measure of sensitivity and household adaptive strategies in the presence of a natural hazard, will be associated with low-levels of household 203

PAGE 216

well-being (economic and health). To test this hypothesis, the following specific aims were developed. #1. To identify vulnerability to natural hazards at the county level by assessing the exposure to hazards, sensitivity (a potential of being affected by a climate stress) and adaptive/coping strategies using climate and county socioeconomic and demographic data. #2. To explore the relationships between vulnerability and health outcomes at the county level (using spatial data analyses). #3. To study the impact of vulnerability to drought and (a Mongolian term for winter disasters) on the economic well-being of rural households. #4. To investigate the relationship of vulnerability to drought and dzud and the health status of individuals within households (using multilevel modeling techniques). The findings from both county and household levels of analysis indicate that climate stress in not the main factor that defines the outcome of disasters: sociodemographic and socioeconomic characteristics that indicate sensitivity to hazards and social resources available to resist and recover from hazards are the most significant to household and individual well-being. At the county level, sensitivity to hazards, represented by socioeconomic and sociodemographic indicators such as dependency ratio, percent of people with income below poverty line, percent of people receiving pensions and allowances, shows a strong relationship with morbidity and/or mortality indicator(s). Higher

PAGE 217

levels of sensitivity were indicative of poorer health outcomes similarly to what has been hypothesized. In general, better adaptive strategies (livestock per person and livestock density) also were related to better health outcomes with one exception: a larger number oflivestock per person was predictive of higher maternal mortality which is contrary to what has been proposed. A possible explanation of this finding is that the number of livestock per person may not necessarily reflect an equal distribution of livestock. As mentioned earlier in the background section, inequality is increasing among rural pastoralists. Perhaps a larger number of livestock is related to social inequality: a number of households that own large assets and a large number of households have fewer assets. An indicator that the distribution oflivestock within a county may have been a better predictor than an average number of livestock per person. The relationship between climate stress and health does not follow the hypothesized direction: worse climatic conditions have shown a protective effect on population health. One explanation to this relationship might be a relatively greater importance of other factors such as sensitivity and adaptation in affecting health conditions. Another issue may be the quality of data that were used in the model. Ten years of monthly precipitation and temperature data were used in the analysis, but socioeconomic indicators were collected only for the last of these ten years. It was assumed that the largest effect of climate stress on general health status of popuiations is to be shown not immediately following the hazardous events, but over a longer term. In fact, one study has shown that there was no 205

PAGE 218

evident nutritional deficiency following droughts and because of the buffering effects of adaptive strategies employed by herders in Mongolia (Siurua Swift, 2002). Lastly, ten years of climate data may not be sufficient to use as an indicator of the overall climate stress. might be interesting to study if the climate data for 30 or more years shows the relationships in same directions. At the household level, there were several factors that had a significant impact on the socioeconomic status of pastoralists. First, a large household size and a large proportion of members in the workforce were associated with increased wealth of households. Having many children ensures greater availability of labor and more productivity, better access to resources through multiple channels of their members, and greater support in times of need. Second, access to resources through formal networks also emerges as an important asset for households: having ties to the county administration is an important source of information and resource flow and is related to better economic well-being of households. Aid and programs fromlofthe governm((nt and international organizations are distributed through the county administration and securing a good relationship with it may directly translate into acquiring these resources. Third, having many close relatives outside of the nuclear household was not necessarily associated with better socioeconomic status, but having many relatives who lived in different places and greater trust in relatives was associated with more wealth. may reflect the importance of urban-rural linkages in the accumulation of wealth. These weak ties may be important to gain access 206

PAGE 219

resources not available in the county, such as medical treatment and rare goods or commodities, to provide housing for school children, and to sell livestock products. Dense networks may be detrimental for the household's economic situation because they have to share resources on a daily basis. As has been discussed in the anthropological literature, in contexts of poverty and periodic resource scarcity, dense social networks are highly adaptive, resulting in a rapid circulation of goods and resources. These same networks, however, also tend to keep participants poor (Stack, 1979). Fourth, perceptions ofliving in a safer area were associated with more wealth. Livestock theft is a major concern for pastoralists. Perceptions of safety likely reflect low levels of livestock theft, with benefits for herd size and income. Livestock theft was present at some degree in all counties where the interviews were conducted. Fifth, greater levels of social cohesion and inclusion resulted in more wealth. Sharing food and engaging in recreational activities with other people indicates inclusion into the social network of neighbors and greater solidarity. Establishing and nurturing these social networks may payoff in the future, especially in times of stress. someone needs a lift to the market to sell cashmere and wool, the neighbor may give a ride. If someone lost most of the animals to a and wants to rebuild a herd, relatives and neighbors may give each a few animals from their herd. Greater likelihood to help in times of illness and disaster was associated with larger number of livestock, which supports the relationship between greater social cohesion and inclusion and more wealth. 207

PAGE 220

Finally, successful herd management strategies emerge as another important factor that contributes to the economic security of pastoralist households. Among these, livestock insurance was the strongest predictor of household's wealth. This association may not be necessarily due to the settlement after the loss of livestock to disasters for those who have livestock insurance. In the prevalence of livestock insurance purchased was very low (10%). What may be relevant here is that wealthier households are more likely to purchase livestock insurance to buffer their losses in times of More questions should be asked in the future research on the details how loss is compensated for those who have livestock insurance. The relationships between trust and indicators of socioeconomic status showed relationships in different directions: greater trust in relatives was associated with more wealth, but greater trust in friends was related to less wealth. The responses to the questions that asked about the likelihood to trust relatives, friends, local government, and press and media may not have been accurate since the interviews took place in a where other people were present including relatives and friends. Some people were suspicious when the question about the trust in local government was asked. Thus, the responses to these questions may have not been reliable and resulted in relationships in different directions. Perhaps asking questions about past experiences (e.g. if there was the case a friend or a relative borrowed money and did not return) may provide more useful information. 208

PAGE 221

The number of visitors and number of other visited may not have been good indicators. These questions only tell the quantity of visits and do not imply anything about the quality or usefulness of these encounters. Pastoralists visit each other very often to find out about possible whereabouts of lost animals or pasture conditions at different locations, and these visits may not necessarily involve a flow of resources. The effect of vulnerability on health varies from one health outcome to another. But in general, several conclusions can be drawn. First, gender does matter. Females were likely to be more anemic than males but they also tend to be more overweight compared to males. Women are likely to be report more illness episodes within the past six months. But there was no statistically significant difference on health seeking behavior among those who had been sick: the percentage of males and females that had been diagnosed and treated was similar. Second, age impacts susceptibility to certain illnesses. Children are more susceptible to anemia and nutritional deficiency compared to adults. Older people tend to be sicker than younger adults. Third, education appears to be very important. Higher levels of education were associated with a decreased risk of anemia and nutritional deficiency. NumerouS studies have found that higher education of adults in the household, especially of mothers, is associated with decreased mortality and morbidity independent of wealth (Lanata, 2001). is argued that educated parents are 209

PAGE 222

increasingly motivated to improve their livelihood and probably have a better knowledge and increased utilization of different resources available. Fourth, having a large percentage of household members in the workforce was related to better health outcomes. These are adults from age 18 to 64 who are least susceptible to illnesses since morbidity is the highest on the tails of the age distribution. Besides the biological explanation for this observation, more working people in a household means higher incomes and a better flow of resources, which in tum affects their susceptibility to illnesses and access to health care. Fifth, more wealth is associated with a lower risk of anemia, nutritional deficiency and illness. Socioeconomic position is one of the fundamental causes ofhealthldisease: it influences multiple health outcomes, including but not limited to anemia and nutritional status through a variety of mechanisms, such as access to resources and changes in lifestyle (Link Phelan, 2004). Besides material resources, wealthier households enjoy more leisure time which in tum can be spent on socializing with people and maintaining social bonds of reciprocity. The number of milk animals, one of the indicators ofa household's wealth, showed a relationship in the opposite direction: the larger the number of milk animals, the poorer the health outcomes. Milk animals are not slaughtered for consumption, and milk production may be reduced during dry summers to such a degree that milk is not available even for making milk tea. The percentage of milk animals within a herd was not very high--only 12%. Thus, having a large 210

PAGE 223

number of milk animals does not necessarily or directly translate into better food security during drought events. Another interesting finding appears in the relationship of money spent on food and anemia. More money spent on the purchase of food items resulted in greater risk of anemia. The possible explanation of this situation may be reduced consumption of meat when intake of cereals is increased (herders do not buy meat, and cash they are spending for food goes to the purchase of cereals, salt and tea). Sixth, the effects of social capital variables on health outcomes were rather complex and often contradictory. Having many trustworthy relatives who lived close emerged as a significant predictor of good health. Interestingly, it had the opposite relationship with a household socioeconomic status. Having a large network of relatives may require material resources to be shared, but it may become an important source of tangible or instrumental support. These close relatives can take care of children and animals and make the travel to the hospital possible or the relatives living in urban areas can send medicines that might otherwise be hard to obtain in the county. The absence of wealth difference was related to less illness reported in the past six months. Inequality may affect health through varying mechanisms some of which include reduced social capital, especially trust and cooperation, and psychological harm. Another variable reflecting social cohesion-frequency of playing games-was positively related with hemoglobin: higher the frequency of games

PAGE 224

was associated with more hemoglobin. Greater social cohesion may facilitate community actions that benefit health conditions. For example, in a cohesive community, if one household slaughters an animal for consumption, other households are going to receive small shares of meat. The others do the same,'and everybody enjoys fresh meat once in a while. This is especially important during the summertime, when animals are not slaughtered often and the basic diet of pastoralists consists of only dairy. Larger numbers of information sources about market prices was associated with better nutritional status. Having good information may increase the income from sale of livestock products and result in a better nutritional status. On the other hand, information sources about market prices may also become a source of knowledge about risks for illness or availability oftreatments. The variables related to trust showed relationships in different directions. was hypothesized that greater trust in people, relatives, friends, and local government would promote participation in community actions that enforces healthy behavioral norms, facilitate community actions that benefit health, and have direct impact. Press and media, besides broadcasting information on healthy lifestyle factors, can also facilitate community cooperation and benefit health through these collaborative actions. As the results ofthe analyses show, the findings were inconsistent with the hypothesized directions. As it was mentioned before, it may be related to the presence of other people when interviews were taking place. 212

PAGE 225

Perceived area safety was another variable that was negatively associated with BMI. was hypothesized that safer the area is, better the health outcomes are. Safer area may be related to more wealth, less psychological stress and greater social cohesion and cooperation, all of which can result in better health conditions. However, the relationship did not follow this direction. may be associated with other social capital variables being an intermediate link between perceived area safety and health outcome. is important to acknowledge that the interviews did not involve families who migrated to towns and cities after the loss of their animals. Lack of social support in the county may have been an important contributor to their migration, and this study did not include these households. At last, the combination of drought and emerges as more important in sickness of household members than drought or taken separately. Pastoralists consistently emphasized the devastation of preceded by droughts. On the other hand, larger number of years was associated with lower risk for anemia. is possible that weak animals are slaughtered for consumption before they die during winter. The findings that more droughts and were predictive of less livestock loss may suggest the same thing: as herders learn from previous episodes of hazardous events, they tend to slaughter weaker animals beforehand to minimize their economic loss. In fact, the government has encouraged slaughtering weak animals to reduce the economic loss due to dzuds. Siurua and Swift had found that herders slaughtered more animals for winter meat when a was likely to occur (Siurua Swift, 2002). 213

PAGE 226

Thus, the results of countyand household-level analyses show the same pattern: climate 'stress by itself is not a major predictor of human well-being due to many strategies pastoralists use to mitigate its impact. Climate stress may also be perceived differently by pastoralists depending on their knowledge, past experiences and resources available. Perceived risk is relevant because it shapes the actions of pastoralists to prevent possible losses from natural hazards. However the perception of hazard risks varies and this variation makes the comparisons difficult among different households. This may explain why the number of droughts, and the combination of the two showed relationships in different directions. More research is needed to understand how pastoralists perceive the environmental risks and how these perceptions shape their economic and herd management decisions. The results show that what is important here is the sensitivity to natural events and adaptive strategies to minimize their impact and return to normal life. Poor households, households with an inadequate labor force, households with no formal or informal ties to resources, and households with inadequate human capital (lacking education or skills) are at greater risk to suffer from consequences of summer droughts and winter disasters. These are the households that are more likely to loose their main asset--livestock to a migrate to urban areas seeking livelihood opportunities and end up in urban slums with little means for a better and healthier life. These are the households that are likely to have inadequate nutrition and get sick during or after disasters. 214

PAGE 227

This is where the disaster mitigation policies should focus. Developing programs aimed at enhancing regional and household adaptive capacities, including supporting social cooperation and cohesion, and promoting useful herding skills is a promising way to reduce economic losses from natural hazards and improve health outcomes in the longer run. Improving adaptive capacities will in tum lower the sensitivity to hazards, by reducing the poverty among pastoralists and enhancing human capital (e.g. education, skills that would help to diversify their income sources). The current relief programs during natural disasters are only limited to distributing a few basic goods after the disaster strikes, and, as the results of this study indicate, they do not have much effect in preventing animal losses or easing the longer term negative impacts of disasters on human well-being. Focusing disaster mitigation programs on increasing adaptive capacity and lowering human sensitivity to natural hazards can also be a useful strategy in the context of other types of hazard events elsewhere. Reducing social vulnerability to hazards should always be the center of any disaster mitigation policy because it is the most important factor that shapes the human experience with natural events. Significance There are several contributions this study made to the existing body of research in vulnerability to natural disasters in Mongolia. This was the first research that looked at multiple social factors that contributed to the negative 215

PAGE 228

outcomes of natural disasters on socioeconomic well-being and health of pastoralists in Mongolia. Several studies have been conducted in the past on short-term nutritional deficiencies after disasters, adaptive strategies used by herders to reduce the livestock loss, and sociodemographic factors of the households who suffered the most from natural disasters. But none of them looked at these factors simultaneously to assess their relative importance to health. The use of spatial analysis techniques and multilevel modeling helped to reduce the overestimation of relationships common in traditional non-spatial or single-level data analysis techniques used for geographical and multilayered data. These data analysis methods can be very useful in vulnerability research given the multidimensionality and complexity of the concept and should be encouraged. The study used social capital as one of the key resources. This is the first study that looks at potential application of social capital to Mongolians. The insights from this research will help to identify the key elements of social capital that have the most significance in pastoralist lifestyle. Identifying these elements will be important for the future development of programs to reduce social vulnerability of Mongolian pastoralists. The SC-IQ developed by the experts in the World Bank was a valuable tool to determine important dimensions of social capital in rural Mongolian context, especially in regards to socioeconomic well-being of past ora lists. As the results indicate, though, social capital is a multidimensional concept and the importance of different dimensions such as trust, social cohesion, social inclusion,

PAGE 229

social network and support may not be the same in different cultures and may not be always beneficial as it is commonly thought. Limitations Besides the strengths and significance of the research, there are some limitations that are important to acknowledge. The study did not include households who migrated to urban centers. There was no possibility of comparing how these households were different from the study sample. is very likely that these households suffered the most during and after droughts and and had to move to urban areas because there was nothing left to rely on for subsistence. The other limitation of the study is the use of secondary data at the county level which has been collected for different purposes. Availability of data constrained the use of indicators at this level of analysis. In addition, unavailability of .county data limited the possibility of analyzing short-term effects of climate stress. The quality of secondary data is always questionable and there is no guarantee that all county officials followed the same standard in collecting these data. The number of households used in the study was small. This may have resulted in a smaller variation in predictors and lower power to detect otherwise significant relationships. This is a'cross-sectional study. Although some attempts were made to gather data on preceding climate stress, all other variables were measures at one 217

PAGE 230

point in time. The nature of the research design makes it impossible to draw causal relationships. In future, it would be a nice addition to study migrant populations in urban areas. This type of research requires more time and resources, but given the potential increase of natural calamities and following rural-to-urban migration, understanding the difficulties the migrants face are crucial in developing programs to assist them. It is also important to facilitate the collection of data at the county level at national level statistical offices (e.g. National Statistical Office of Mongolia, Ministry of Health). The county information is collected in each province for each year but they are aggregated at provinces before reaching the national offices. This puts serious limits on the use of these data for research and policy purposes. As the results of this study suggest, county level information can be useful when a smaller level data is unavailable. Enhancing the accuracy and validity of this information will likely to increase its applicability in research. 218

PAGE 231

219

PAGE 232

N o 00 -...l w ......... Code Name Relationship to the household head Year of birth Sex Marital status Occupation Current job Years in school withdrawn from school, why Years of residency in this county Months present during the last 12 months .... 0 ::= ;s. 0 0 ::r: CZl ::r: 0 0 0 ""'C .... =-::= 0 CZl

PAGE 233

Individual Height Weight Skinfold mm Hb Given iron code em kg supplement Comments: 221

PAGE 234

In(ormation Oil Khot Ails # 1. 3. 4. 5. 6. # 2 2. 3. 4. 5. 6. # 3 2. 3. 4. 5. 6. 2. 3. 4. 5. 6.

PAGE 235

I. What are the main benefits of khot ails for your household? 2. 3. 4. 5. 2. What kind of support does your household give to other khot ails? 2. 3. 4. 5. 223

PAGE 236

Name Gender Age Relationship to Occupation the interviewee 2. 3. 4. 5. 6. 7 8. 1. Have you received any help during the last month? a. Yes b. No *Codes: 1. Send money 3. Provide housing for my children when they go to school 5. Help with transportation 7. Help with connections to important people 9. Help with information 11. Emotional support Employment Place of Frequency residency of meetings 2. Help with medical care getting medicines 4. Help with selling animal products 6. Help with extra labor when needed 8. Provide care when someone is sick Type of help 10. Send goods that are hard to obtain in countryside 12. Other, specIfy 224

PAGE 237

1. About how many close friends do you have whom you can talk about your private matters? 2. Where do the most of your friends live a. The same khot ail b. The same county c. The same province d. Cities e. Other 3. On average, how often do you talk to your friends? a. Everyday b. A few times a week c. Once a week d. Less than once a week e. Never 4. About how many relatives do you have whom you can talk about your private matters? 5. Where do most of your relatives live a. The same khot ail b. The same county c. The same province d. Cities e. Other 6. On average, how often do you talk to your relatives? a. Everyday b. A few times a week c. Once a week d. Less than once a week e. Never 225

PAGE 238

7. How well do you know your county governor? a. Not at all b. Little c. Somewhat d. Very well 8. Have you asked your county governor to help you in private matters? a. Yes No 9. Whom do you talk about health issues a. c. b. d. 10. Whom do you talk about financial issues a. c. b. 11. Whom do you talk about herding issues a. c. b. d. 12. Whom do you ask to borrow tools a. c. b. d. 13. Whom do you ask to get help in cutting hay a. c. 226 d.

PAGE 239

14. If you suddenly needed to borrow 10,000 tugrics for about a week, are there people beyond your family members to whom you could tum and who would be willing and able to provide this money? a. Definitely not b. Probably not c. Probably d. Definitely 15. Whom would you ask for this money a. c. b. d. 16. If you had suddenly go away for a couple of days, do you have someone to take care of your children? a. Definitely not b. Probably not c. Probably d. Definitely 17. Whom would you live your children with? a. c. b. d. 18. In the past month, how many people with a personal problem have turned to you for assistance? 19. Do you have an access to bank loans and restocking animals? a. Yes b. No 20. Compared to pre-transition period, how has people's willingness to help each other changed? a. Increased b. Decreased c .. Stayed the same 227

PAGE 240

1. Would people in this area help you if you need it? a. Definitely not b. Probably not c. Probably d. Definitely 2. Generally speaking, how trustful are people nowadays? a. Absolutely not trustful b. Not trustful c. Trustful d. Very trustful 3. How much do you trust your relatives? 4. How much do you trust your friends? 5. How much do you trust your county officials? 6. How much do you trust press and media? Do not trust at all 2 -Do not trust 3 -Trust Trust completel 7. How much has the level of trust in this neighborhood changed during the transition period? a. Gotten better the same b. Gotten worse c. Stayed 1. Since the N aadam of a last year did you or anyone in your household participate in any communal activities, in which people came together to do some work for the benefit of the neighborhood? a. Yes, No How many times? 228

PAGE 241

2. Since the Naadam oflast year did you or anyone in your household joined with others in your neighborhood to address a common issue? a. Yes, b. No How many times? 3. Ifthere is a necessity to build a new well for both human and livestock needs, how likely is it that people cooperate? a. Very unlikely b. Somewhat unlikely c. Somewhat likely d. Very likely 4. Suppose something unfortunate happened to someone in your neighborhood, such as serious illness or natural disaster. How likely is it that people in the neighborhood get together to help them? a. Very unlikely b. Somewhat unlikely c. Somewhat likely d. Very likely Information and Communication 1. What are the main sources of information about what the government is doing a. Relatives, friends and neighbors b. Local markets, shops c. Local newspaper d. National newspaper e. Radio Television g. Bhag leaders h. Soum officials 2. What are the most important sources of market inform ation on livestock products a. Relatives, friends and neighbors b. Local markets, shops c. Local newspaper d. National newspaper e. Radio 229

PAGE 242

Television g. Bhag leaders h. Soum officials 3. How many times have people in your household traveled to the county center in the last 6 months? 4. How many times have people in your household traveled to the province center in the last 6 months? 5. How often do you listen to radio? a. Everyday b. few times a week c. Once a week d. Less than once a week e. Never 6. How far does it take you to get to the nearest phone? 7. Compared to the pre-transition period, how has the access to information changed? a. Improved b. Deteriorated c. Stayed the same 1. How are people in your neighborhood different from each other in terms of number of livestock and wealth? a. To a great extent b. To a small extent c. Not different 2. If they are different, do any ofthese differences cause problems? a. Yes No What kind of problems? ______ 3. How many times in the past month have your household members got together with people to have food or drinks? __ 4. In the past months, how many times have your household members gotten together with people to play games? __ 5. In the past month, how many times have people visited your ger? __ 230

PAGE 243

6. In the past month, how many times have your household members visited people in their homes? __ 7. Ifa family moves into your area from a different county, how much people would be willing to help them? a. Not at all b. Little c. Somewhat d. Very much 8. In general, safe is this area from crime? a. Very unsafe Safe b. Little unsafe 1. In overall, how much impact do you think you have in making this neighborhood a safer place to live? a. No impact b. A small impact c. A large impact c. 2. To what extent does the soum government take into account concerns by you and your area when they make decisions relevant to your county? a. Not at all b. To a small extent c. To a great extent 3. In the past 12 months, how often have people in this neighborhood got together to jointly petition government officials or political leaders for something benefiting the neighborhood? __ 4. How many people in your household did vote in the last presidential election? 1. How many times have you moved since the Naadam of last year? __ 2. About how many kilometers does your household move in a year? __ 231

PAGE 244

3. Is this fewer, about the same, or more than before transition? a. Fewer b. More c. The same 4. How far do you move to your summer camp? km 5. How far do you move to your winter camp? km 6. Did you do "otor" this spring/summer? a. Yes b. No Forhowlong? ______ 7. Did you purchase livestock insurance this year? a. Yes b. No For what percentage of your livestock? ______ __ 8. Do you have a permanent winter shelter for your livestock? a. Yes No 9. Do you have a lease for this winter camp? a. Yes b. No 10. How fairly are the leases given out by the soum leaders? a. Fair b. Not fair 11. Has there been any conflict over pasture in this area? a. Yes b. No How did you solve it? __________ 12. Has there been a conflict over water source? a. Yes b. No How did you solve it? ____ 13. How much hay did you prepare last summer? __ kg 14. How much mineral lick did you prepare for last fall? __ kg 15. Who does tend the livestock to pastures? __________ 232

PAGE 245

16. Who does the milking? 17. Who does the dairy processing? 18. Where do your household and your neighbors sell their livestock products? 19. Are there any problems with selling your animal products? a. Yes b. No Whatarethey? __________________ ___ 20 When was the last time you had a veterinarian examine your livestock? 21. How serious is the problem of overgrazing in this area? a. Very serious b Somewhat serious c Not serious 22. What do you think needs to be done to prevent from or stop overgrazing? 1. Have you experienced any dzud(s) since the transition? a. Yes b. No 233

PAGE 246

Years 2. What was your loss? Camel Camel --Cattle Cattle --Horse Horse --Sheep Sheep Goat Goat 3.Did you get any extra Yes, from whom? Yes, from whom? manpower help from your relatives or friends? No No 4.Did you get fodder from the Yes, amount? Yes, amount? national fodder fund? No No S.Did you have to purchase Yes, how much? Yes, how much? fodder from private companies? No No 6.Did you receive any food Yes, amount? Yes, amount? aid? No No 7.Did you receive any warm Yes, amount? Yes, amount? clothes as an aid? No No 8.Did you receive necessary Yes, illness(es)? Yes, illness(es)? medical services? No No 9.Did you receive milk Yes, amount? Yes, amount? substitutes for younger animals? No No 10. Did you have to move Yes, how far? Yes, how far? __ during the dzud? No No 11. Have you received any Yes, from whom? Yes, from whom? warning about the possibility ofdzud? No No 12. How did you respond to a.Fix shelters a.Fix shelters this warning? b.Prepare hay b.Prepare hay c.Focus on herd c. Focus on herd management management d.Prepare food for d.Prepare food for winter winter e.Prepare wood e.Prepare wood f. Other f. Other 13. Were the roads to your Yes Yes camp accessible for a No, for how long? No, for how long? whole winter? 14. Did you have a well-kept Yes Yes winter shelter for animals? No No 234

PAGE 247

15. Do you personally know anyone who lost all of their animals to any of these dzud(s)? a. Yes. b. No What has happened to them? ________ 16. Did you or anyone in your neighborhood receive a livestock through restocking programs? a. Yes No How many animals? 17. Have you experiences any droughts since the transition? a. Yes b. No 235

PAGE 248

Years 18. How serious was the a. Very serious a. Very serious drought? b. Serious b. Serious c. Not serious c. Not serious 19. How often did you have to move to get better pastures times times during a drought? 20. Did you have to split your livestock temporarily Yes Yes among different No No households? 21. Did you have a reserve Yes Yes pasture? No No 22. Did you have to purchase Yes Yes _grass? No No 23. Was there a conflict over Yes Yes pasture and water source? No No 24. Did you have to do Yes, what animals? Yes, what animals? selective herding? No No 25. Have you experienced a Yes, for how long? Yes, for how long? shortage of dairy products? __ days/months days/months No No 26. Did you have to reduce Yes Yes your household food No No consumption? 27. Did you get any aid in food Yes, from whom? Yes, from whom? supply? No No 28. Did you have any animal Camel Camel ----loss? Cattle Cattle --Horse Horse -Sheep Sheep Goat Goat 29. Did you receive an early Yes, from whom? Yes, from whom? warning about a drought? No No 30. How did you respond to this warning? 236

PAGE 249

Name/ N Code 0 0 0 0 Has she/he been sick in the last 6 months? 0 (II ::r' 0 (II {J) ...:> ...:> Has the illness been 0.. 0 0.. ::r' ....... {J) diagnosed by a health professional? yes, by whom and ::r' ::r' (II DS? no. why not? P' (II l;l P' N (II l;l v.> (II -...J ....... () (II ....... 0.. (II Has the illness been treated? yes, what kind of 0 0 ::r' 0 ::r' cr-cr-0 {J) (II ::r' (II 0 ::r' 0 ....-0 0 ....treatment? no, why? Costs ::r' (II ::r' P' (II P' ....... ::r' -{J) ::r' ....... {J) P' ....... 2 P' {J) 2 ...:> ...:> Did you have to borrow money to for treatment? Has the illness been cured?

PAGE 250

Household Socioeconomic Status Livestock Inventory Camel Number of livestock received by privatization 2. Number oflivestock as of today 3. Number of milk animals Number of livestock purchased in 2005 5. Number of livestock sold/traded in 2005 6. Number of animals given to others as gifts in 2005, including your close relatives 7. Number of animals lost to theft, disease, and predation in 2005 8. Number of animals consumed since last August Income 1. Income from a livestock: a. Did you sell any cashmere this s rin ? b. Did you sell any wool this spring? c. Did you sell any livestock since last Au st? b. Did you sell any meat since last fall? c. Did you sell any hides and skins since last fall? d. Did you sell any dairy since spring? Cattle Horse Yes, what was the income? Sheep Goat No 1. What is your household monthly income from pensions and allowances? __ tugrics 2. What is your household monthly income from wages? __ tugrics 238

PAGE 251

3. Do you do any of the following to supplement your income or help your livelihood? a. Growin vegetables and cro s b. Huntin c. Collectin wild food d. e. Yes, whose No responsibility is this? 9. What are the priority needs for your household at this moment (things need to be purchased as a priority)? a. b c. 10. What would be your household's largest expense in the nearest future? 11. Do you own any of the following Yes No things? a. TV b. Radio c. Power generator d. Motorbike e. Car/Jeep f. Truck g Extra "ger" 12. What were the costs associated with any of the following in the past year a. Health care b. Holidays c. Weddings d. Tuition e. Other 239

PAGE 252

13. Amount of money spent on food for each month _____ tugrics 14. How has the livelihood of your household changed compared to the pre transition period? a. Improved b. Deteriorated 15. What were the reasons of such change? a. c. 240 b c. Stayed the same

PAGE 253

Annual Standing Crop grass in one square hectares of land is cut, dried and weighed. Ecological zone -a classification based on vegetation and precipitation. Feldsher trained mid-level medical professional. Ger traditional dwelling, known as "yurt". Gross Domestic Product (GDP) the total output of goods and services for final use within the domestic territory of a given country, regardless of the allocation to domestic and foreign claims. Khot ail-a camping group of households. Land degradation the loss of primary production, often through soil erosion but also through changes in vegetation and through processes such as salinization and shifting sand (Corvalan et aI., 2005). Sheep Forage Unit (SFU) -a measure that makes possible to express the size of herds containing diverse species in a common unit. is based on the food requirements of the different species relative to sheep (Neupert, 1999). 1 camel or 1 horse equals to 7 sheep, 1 C'ow to 6 sheep, and 1 goat to 1 sheep (Grayson Baatarjav, 2004). Total Fertility Rate the measure of a number of children a woman is likely to have in her lifetime Tugrics Mongolian currency, approximately 1100 tugrics are equivalent to 1 USD. 241

PAGE 254

REFERENCES Adger, N., Brooks, N., Bentham, G., Agnew, M., &Eriksen, S. (2004). (technical report). Norwich: Tyndal Centre for Climate Change Research. Baas, S., Batjargal, E., Swift, J. (2001, Oct. 12,2003). Paper presented at the Asia-Pacific conference in early warning, prevention, preparedness and management of disasters in food and agriculture, Chiangmai, Thailand. Bailey, K. V., Ferro-Luzzi, A. (1995). Use of body mass index of adults in assessing individual and community nutritional status. 73(5),673-680. Bal, S., Crombez, G., Van Oost, P., Debourdeaudhuij, I. (2003). The role of social support in well-being and coping with self-reported stressful events in adolescents. 27, 1377-1395. Bassuk, S. S., Glass, T. A., Berkman, L. F. (1999). Social disengagement and incident cognitive decline in community-dwelling elderly persons. 131 (3), 165-173. Bates, D. G., Lees, S. H. (1996). Chapter II. Pastoralism. In D. G. Bates S. H. Lees (Eds.), (pp. 153-157). New York: Plenum Press. Begzsuren, S., Ellis, J. E., Ojima, D. S., Coughenour, M. B., Chuluun, (2004). Livestock responses to droughts and severe winter weather in the Gobi Three Beauty National Park, Mongolia. 59, 785-796. Bellander, T., Berglind, N., Gustavvson, P., Jonson, T., Nyberg, F., Pershagen, G. (2001). Using geographical information systems to asses individual historical exposure to air pollution from traffic and house heating in Stockholm. 242

PAGE 255

Berkman, L. F., Glass, T. (2000). Social integration, social networks, social support, and health. L. F. Berkman Kawachi (Eds.), (pp. 137-173). New York: Oxford University Press. Berkman, L. F., Leo-Summers, L., Horwitz, (1992). Emotional support and survival after myocardial infarction. 1003-1009. Bhana, A. (1998). The use ofthe geographical information system (GIS) to determine potential access and allocation of public mental health resources in Kwazulu-Natal. 28(4),222-234. Blaikie, P., Cannon, T., Davis, I., Wisner, B. (1994). New York: Routledge. Boone,1. D., McGwire, Otteson, E. W., DeBaca, S., Kuhn, E. A., Villard, P., et al. (2000). Remote sensing and Geographic Information Systems: charting Sin Nombre Virus infections in Deer Mice. 6(3),248-258. Borgerhoff-Mulder, M., Sellen, D. W. (1994). Pastoralist decision-making: behavioral ecological perspective. Fratkin, A. Galvin & E. Roth (Eds.), (pp. 205-229). Boulder: LYlll1,e Rienners Publishers. Bourdieu, P. (1985). The forms of capital. J. G. Richardson (Ed.), (pp. 241-258). New York: Greenwood. Bourgois, P. (1995). New York: Cambridge University Press. Brabyn, L., Skelly, C. (2002). Modeling population access to New Zealand public hospitals. 1(3), 1-9. Brehm, J., Rahn, W. (1997). Individual-level evidence for the causes and consequences of social capital. 41(3),999-1023. Brooks, N. (2003). (working paper No. 38). NOlwich: Tyndall Centre for Climate Change Research and Centre for Social and Economic Research on the Global Environment. Bum, N. (2003). (research report). Ulaanbaatar: UNIFEM UNDP. 243

PAGE 256

Burton, 1., Kates, W., White, G. F. (1993) (second ed.). New York: The Guilford Press. Chen, M. A. (1991). India: Sage Publications. Christensen, A. l, Wiebe, J. S., Smith, T. W., Turner, C. W. (1994). Predictors of survival among hemodialysis patients: effect of perceived family support. 13(6),521-525. Chuluundorj, O. (2001). Unpublished Master's, University of Colorado at Denver, Denver. Cohen, S., Syme, S. L. (1985). Issues in the study and application of social support. S. Cohen S. L. Syme (Eds.), (pp. 3-23). Orlando: Academic Press, Inc. Coleman, (1988). Social capital in the creation of human capital. 94s, 95-120. Cooper, L. (1993). Patterns of mutual assistance in the Mongolian pastoral economy. 33, 153-162. Corvalan, C., Hales, S., McMichael, A., Butler, C., Campbell-Lendrum, D., Confalonieri, U., et al. (2005). Geneva: World Health Organization. Cutter, S. (1996). Vulnerability to environmental hazards. Demberelsuren, Dorjpurev, T. (2000). Ulaanbaatar: MMOHSW, UNICEF, and The Maternal and Children's Research Center. Diez Roux, A. V. (2002). A glossary for multilevel analysis. 56,588-594. Dunn, C. E., Woodhouse, R. S., Bhopal, R. S., Acquilla, S. D. (1995). Asthma and factory emissions in northern England: addressing public health concern by combining geographical and epidemiological methods. 49(4),395-400. Durkin, M., McElroy, l, Guan, H., Bigelow, W., Brazelton, T. (2005). Geostatistical analysis of traffic injury in Wisconsin: Impact on case fatality of distance to level IIII trauma care. 244

PAGE 257

Durlauf, S. N., Fafchamps, M. (2004). (working paper). Cambridge: National Bureau of Economic Research. Emch, M. (1998). Paper presented at the Geographical Information Systems in Public Health, San Diego. Eng, P. M., Rimm, E. B., Fitzmaurice, G., Kawachi, 1. (2002). Social ties and change in social ties in relation to subsequent total and cause-specific mortality and coronary heart disease incidence in men. 155(8), 700-709. Entwisle, B., Rindfuss, R R, Walsh, S. J., Evans, T. P., Curran, S. R. (1997). Geographical Information Systems, spatial network analysis, and contraceptive choice. 34(2), 171-187. Eriksen, W. (1994). The role of social support in the pathogenesis of coronary heart disease: a literature review. 11,201-209. Eyles, J. (1990). How significant are the spatial configurations of health care systems? Farmer,1. P., Meyer, P. S. (1996). Higher levels of social support predict greater survival following acute myocardial infarction. 22(2), 59-66. Fernandez-Gimenez, M. (1998). Paper presented at the "Crossing Boundaries," the seventh annual conference of the International Association for the Study of Common Property, Vancouver, BC, Canada. Fernandez-Gimenez, M. (1999). Reconsidering the role of absentee her owners: A view from Mongolia. 27(1), 1-27. Ferrer, R L., Palmer, R., Burge, S. (2005). The family contribution to health status: A population-level estimate. 3(2), 102108. Foggin, P. M., Farkas, 0., Shirev-Adiya, S., Chinbat, B. (1997). Health status and risk factors of seminomadic pastoralists in Mongolia: A geographical approach. 44(11), 1623-1647. Foley, R (2002). Assessing the applicability of GIS in a health and social care setting: Planning services for informal carers in East Sussex, England. 55, 79-96. 245

PAGE 258

Fost, D. (1990). Using maps to tackle AIDS. 12(4),22. Foster, S. (2003). Ottawa: International Development Research Center. Fothergill, A. (1996). Gender, risk and disaster. 14(1),33-56. Fratkin, E. (1997). Pastoralism: Governance and development issues. 26,235-261. Fratkin, E., Roth, E. (1996). Who survives drought? Measuring winners and losers among the Ariaal Rendille pastoralists in Kenya. In D. G. Bates S. H. Lees (Eds.), New York: Plenum Press. Fukuyama, F. (1995). New York: The Free Press. Furstenberg, F., Hughes, M. (1995). Social capital and successful development among at-risk youth. 57, 580-592. Galaty, J. G., Johnson, D. L. (1990). Introduction: Pastoral systems in global perspective. In J. G. Galaty D. L. Johnson (Eds.), (pp. p. 1-32). New York: The Guilford Press. Gilbert, C. (1995). Studying disaster: a review of the main conceptual tools. 3(3),231-240. Giles, L. Glonek, G. F. Y., Luszcz, M A., Andrews, G. R. (2005). Effect of social networks on 10 year survival in very old Australians: The Australian longitudinal study of aging. 59, 574-579. Gilles, J. L., Gefu, (1990). Chapter 3. Nomads, ranchers, and the state: The sociocultural aspects of pastoralism. In G. Galaty D. L. Johnson (Eds.), (pp. 103-118). New York: The Guilford Press. Goldstein, M. C., Beall, C. M. (2002). Changing pattern of Tibetan nomadic pastoralism. In W. R. Leonard M. Crawford (Eds.), (first ed., pp. 131-150). Cambridge: Cambridge University Press. Government of Mongolia. (2001). Ulaanbaatar: Government of Mongolia Asian Development Bank. 246

PAGE 259

Government of Mongolia, UNDP. (2004). Ulaanbaatar: UNDP. Grayson, R., Baatarjav, M. (2004). (research report). Ulaanbaatar: Open Society Forum. Griffin, (2001). Ulaanbaatar: UNDP. Grootaert, C. (2000). Washington, D.C.: The World Bank. Grootaert, C., Narayan, D., Jones, V. N., & Woolcock, M. (2004). (working paper #18). Washington, D.C.: The World Bank. Grootaert, C., Van Bastelaer, T. (2002). Washington, D.C.: The World Bank. Guiso, L., Sapienza, P., Zingales, L. (2000). Cambridge: National Bureau of Economic Research. Haddow, G D., Bullock, J. A. (2006). (second ed.). Oxford: Elsevier Butterworth-Henemann. Hewitt, (1995). Excluded perspectives in the social construction of disaster. 13(3),317-339. Hewitt, (1997). Essex, England: Addison Wesley Longman Limited. Hjalmas, U;, Kulldorff, G., Gustafson, G., Nagarwalla, N. (1996). Childhood leukaemia in Sweden: Using GIS and a spatial scan statistic for cluster detection. 15(7/9), 707-715. Holtgrave, D. R., Crosby, R. A. (2003). Social capital, poverty, and income inequality as predictors of gonorrhoea, syphilis, chlamydia and AIDS case rates in the United States. 79, 62-64. Hoohdoi, C. (2002). Ulaanbaatar: T U Printing. Humphrey, Sneath, D. (1999). Durham, NC: Duke University Press. 247

PAGE 260

Janes, C. R., Chuluundorj, O. (2004). Free markets and dead mothers: The social ecology of maternal mortality in post-socialist Mongolia. 18(2),230-257 Kawachi, I., Berkman, F. (2000). Social cohesion, social capital, and health. In F. Berkman I. Kawachi (Eds.), (pp. 174-190). New York: Oxford University Press. Kawachi, I., Colditz, G. A., Ascherio, A., Rimm, E. B., Giovannucci, E., Stamfer, M. J., et al. (1996). A prospective study of social networks in relation to total mortality and cardiovascular disease in men in the USA. Kawachi, I., Kennedy, B. P., Glass, R. (1999). Social capital and self-rated health: A contextual analysis. 1187-1193. Kawachi, I., Kennedy, B. P., Lochner, K., Prothrow-Stith, D. (1997). Social capital, income inequality and mortality. 87(9), 1491-1498. Kinney, A. Y., Bloor, E., Dudley, W. N., Millikan, R. C., Marshall, E., Martin, C., et al. (2003). Roles ofreligous involvement and social support in the risk of colon cancer among Blacks and Whites. 158(11), 1097-1107. Kistemann, T., Munzinger, A., Dangendorf, F. (2002). Spatial patterns of tuberculosis incidence in Cologne (Germany). 55, 7-19. Kotie, R. Y., Moller-Jensen, (2001). Towards a framework for delineating sub-districts for primary health care administration in rural Ghana: A case study using GIS. 55,26-33. Kohli, S., Noorlind-Brage, H., Lofman, O. (2000). Childhood leukaemia in areas with different radon level: A spatial and temporal analysis using GIS. 54, 822-826. Kormondy, E. J Brown, D. E. (1998). New Jersey: Prentice Hall. Kovats, R. S., Bouma, M. (2002). Chapter 6. Retrospective studies: Analogue approaches to describing climate variability and health. In P. Martens A. J. McMichael (Eds.), (pp. 144-171). Cambridge: Cambridge University Press. 248

PAGE 261

Kovats, S., Menne, B., McMichael, A J., Corvalan, C., Bertollini, R. (2000). Geneva: World Health Organization. Krautheim, K. R., Aldrich, T. E. (1997). Geographic information system (GIS) studies of cancer around NPL sites. 13(2-3),357-362. Krishna, A., Uphoff, N. (1999). (working paper). Washington, D.C.: The World Bank. Lal, M., Harasawa, H., Muriyarso, D. (2001). (report). Cambridge: Intergovernmental Panel on Climate Change. Lanata, C. F. (2001). Children's health in developing countries: Issues of coping, child neglect and marginalization. In D. A. Leon G. Walt (Eds.), (pp. 137158). Oxford: Oxford University Press. LaScala, E. A, Johnson, F. W., Gruenewald, P. J. (2001). Neighborhood characteristics of alcohol-related peqestrian injury collisions: A geostatistical analysis. 2(2), 123-134. Leatherman, T. L., Thomas, B. (2001). Political ecology and constructions of environment in biological a,nthropology. In C. L. Crumley (Ed.), (pp. 113-132). Lanham: Altamira Press. Leyland, H., Groenewegen, P. P. (2003). Multilevel modeling and public health policy. 31,267-274. Lightstone, A S., Dhillon, P. K., Peek-Asa, Kraus, J. F. (2001). A geographic analysis of motor vehicle collisions with child pedestrians in Long Beach, California: Comparing intersection and midblock incident locations. 7, 155-160. Lin, N. (2001). Cambridge: Cambridge University Press. Link, B. G., Phelan, J. (2004). Social conditions as fundamental causes of disease. issue), 80-94. 249

PAGE 262

Little, M. A. (2002). Human biology, health, and ecology of nomadic Turkana pastoralists. In W. R. Leonard M. H. Crawford (Eds.), (pp. 151-182). Cambridge: Cambridge University Press. Little, P. D., Smith, K., Cellarius, B. A., Coppock, D. L., Barrett, C. B. (2001). Avoiding disaster: Diversification and risk management among East AfticanHerders. 32,401-433. Lohman, T. G., Roche, A. F., Martorell, R. M. (1988). Champaign, Illinois: Human Kinetics Books. Macintyre, S., Ellaway, A. (2003). Neighborhoods and health: overview. In Kawachi L. F. Berkman (Eds.), (pp. 2042). New York: Oxford University Press. Marfin, A. A., Peterson, L. R., Eidson, M., Miller, J., Hadler, J Farello, C., et al. (2001). Widespread West Nile virus activity, Eastern United States, 2000. 7(4), 730-735. Margai, F. L. (2001). Health risks and environmental inequality: A geographical analysis of accidental releases of hazardous materials. 53(3),422-434. Marriott, P., Erdene-Ochir, B. (2004). Ulaanbaatar: UNDP. McLafferty, S. (2003). GIS and health care. 25-42. Mearns, R. (2004). Sustaining livelihoods on Mongolia's pastoral commons: Insights from a participatory poverty assessment. 35(1), 107-139. Michael, Y. L., Berkman, L. F., Colditz, G. A., & Kawachi, (2001). Living arrangements, social integration, and change in functional health status. 153(2), 123-131. Moran, E. F. (2000). Chapter 8. Human adaptability to grasslands. In E. F. Moran (Ed.), (pp. 219-257). Boulder: Westview Press. Moran, R. A., Butler, D. S. (2001). Whose health profile. 11(1),59-74. 250

PAGE 263

Morinaga, Y., Tian, S.-F., Shinoda, M. (2003). Winter snow anomaly and atmospheric circulation in Mongolia. 23, 1627-1636. Morrow, B. H. (1999). Identifying and mapping community vulnerability. 23(1), 1-18. Mulder, M. B., Sellen, D. W. (1994). Chapter 11. Pastoralist desicionmaking: A behavioral ecological perspective. In E. Fratkin, A. Galvin A. Roth (Eds.), (pp. 205-229). Boulder: Lynne Rienners Publishers. Musgrove, P. (1987). The economic crisis and its impact on health and health care in Latin America and the Carribean. 17(3),411-441. Neupert, R. F. (1999). Population, nomadic pastoralism and the environment in the Mongolian Plateau Norovlin, B., Byambatogtokh, B., Bates, J., Serdula, M. K., Kaufmann, R., Woodruff, B. A., et al. (2003). Nutrition Research Center, World Health Organization, UNICEF Center for Disease Control and Prevention. NSOM. (2004). Ulaanbaatar: National Statistical Office of Mongolia. NSOM, UNFPA. (1999). Ulaanbaatar: National Statistical Office of Mongolia UNFPA. O'Brien, K., Eriksen, S., Schjolden, A., Nygaard, (2003). (working paper No. 38). Oslo: Center for International Climate and Environmental Research O'Brien, K., Eriksen, S., Schjolden, A., Nygaard, (2004). (working paper). Oslo: Center for International Climate and Environmental Research. O'Brien, K., Leichenko, R. M. (2000). Double exposure: Assessing the impacts of climate change within the context of economic globalization. 251

PAGE 264

O'Brien, K., Sygna, L., Haugen, 1. E. (2004). Vulnerable or resilient? A multi scale assessment of climate impacts and vulnerability in NOlway. 64, 193-225. Olmos, S. (2001). (foundation paper). Oslo: Climate Change Knowledge Network. Orth-Gomer, K., Rosengren, A., Wilhelmsen, L. (1993). Lack of social support and incidence of coronary heart disease in middle-aged Swedish men. 55(1),37-43. Paddle, J. 1. (2002). Evaluation of the Haemoglobin Colour Scale and comparison with the HemoCue haemoglobin assay. 813-987. Pargal, T., Huq, M., Gilligan, D. (1999). (working paper). Washington D.C.: The World Bank. Parker, E. B., Campbell, 1. L. (1998). Measuring access to primary medical care: Some examples of the use ofGISs. 4, 183-193. Pelling, M. (2003). Sterling, VA: Earthscan. Penninx, B. W. J. H., Tilburg, T. V., Deeg, D. J. H., Kriegsman, D. M. W., Boeke, A. J. P., Van Eik, 1. T. M. (1997). Direct and buffer effects of social support and personal coping resources in individuals with arthritis. 44(3), 393-402. Perry, B., Gesler, W. (2000). Physical access to primary health care in Andean Bolivia. Pine, J. & Diaz, J. H. (2000). Environmental health screening with GIS: Creating a community environmental health profile. 9-15. Portes, A. (1998). Social capital: Its origins and applications in modem sociology. 24, 1-24. Productivity Commission of Australia. (2004). (commission research paper). Melbourne: Productivity Commission. 252

PAGE 265

Putnam, R P. (1995). Bowling alone: America's declining social capital. 6, 65-78. Putnam, R. P., Leonardi, R, Nanetti, R. (1993). Princeton: Princeton University Press. Ravsal, (2003). (Study report). Ulaanbaatar: JEMR Rosengren, A., Wilhelmsen, L., Orth-Gomer, (2004). Coronary disease in relation to social support and social class in Swedish men. 25, 56-63. Rosero-Bixby, L. (2004). Spatial access to health care in Costa Rica and its equity: A GIS-based study. 58, 1271-1284. Rothauge, A. (1998). Rome, Italy: FAO. Sari, M., de Pee, S., Martini, E., Herman, S., Sugiatmi, Bloem, M. W., et al. (2001). Estimating the prevalence of anaemia: A comparison of three methods. 79(6), 506-511. Schareika, N. (2003). Rome, Italy: FAO. Seeman, T. E. (1996). Social ties and health: The benefits of social integration. 6, 442-451. Siurua, E., Swift, J. (2002). Drought and zud but no famine (yet) in the Mongolian herding economy. 33(4),88-97. Skees, J. R., Enkh-Amgalan, A. (2002). (working paper). Washington D.C.: The World Bank, East Asia and Pacific Region. Smit, B., Pilosofa, O. (2001). (report). Cambridge: Intergovernmental Panel on Climate Change. Speer, S. A., Semenza, 1. Kurosaki, T., Anton-Culver, (2002). Risk factors for acute myeloid leukemia and multiple myeloma: A combination of GIS and case-control studies. Stack, C. (1979). NY: Basic Book. 253

PAGE 266

Subramanian, S. V., Jones, K., Duncan, C. (2003). Multilevel methods for public health research. In I. Kawachi & L. F. Berkman (Eds.), (pp. 65-111). New York: Oxford University Press. Swift, J. (1999). Rome, Italy: F AO, Sustainable Development Department. Tanser, F. C., Ie Sueur, D. (2002). The application of geographical information systems to important public health problems in Africa. 1, 1-9. Templer, G., Swift, J., Payne, P. (1993). The changing significance of risk in the Mongolian pastoral economy. 33, 105-122. Theorell, T., Blomkvist, V., Jonsson, H., Schulman, S., Berntorp, E., Stigendal, L. (1995). Social support and the development of immune function in human immunodeficiency virus infection. 32-36. Tobin, G. A., Montz, B. E. (1997). NY: The Guilford Press. Treiber, F. A., Baranowski, T., Braden, D. S., Strong, W. B., Levy, M., Knox, W. M. (1991). Social support for exercise: Relationship to physical activity in young adults. Tserendash, S., Erdenebaatar, (1993) Performance and management of natural pasture in Mongolia. 33,9-15. UNDP. (2001). Ulaanbaatar: UNDP. UNDP. (2002). UNDP. UNDP. (2003). Ulaanbaatar: Government of Mongolia UNDP. Vincent, K. (2004). (working paper). Norwich: Tyndall Centre for Climate Change Research. Vogt, T. M., Mullooly, J. P., Ernst, D., Pope, C. R, Hollis, J. F. (1992). Social networks as predictors of ischemic heart disease, cancer, stroke and hypertension: Incidence, survival and mortality. 45(6),659-666. 254

PAGE 267

Waldinger, R. (1995). The "Other Side" of embededdness: A case study of interplay between economy and ethnicity. 555-580. Waller, L. A, C. A (2004). Hoboken, New Jersey: John Wiley Sons, Inc. Waller, R., Sobania, N. W. (1994). Chapter 3 Pastoralism in historical perspective. In E. Fratkin, K. A. Galvin E. Roth (Eds.), (pp. 45-68). Boulder: Lynne Rienner Publishers. Watts, M. (2000). Chapter 16. Political ecology. In E. Sheppard T. J. Barnes (Eds.), (pp. 256-274). Massachussets: Blackwell Publishers. Weichselgartner, J. (2001). Disaster mitigation: The concept of vulnerability revisited. Weinberger, M., Tierney, W. M., Booher, P., Hiner, S. L. (1990). Social support, stress and functional status in patients with osteoarthritis. White, G. F., Haas, J. E. (1975). Cambridge, MA: MIT Press. White, K. S., Ahmad, Q. K., Anisimov, 0., Amell, N., Brown, S., Campos, M., et al. (2001). (report). Cambridge: Intergovernmental Panel on Climate Change. 255