Growth, Crisis and Resilience: Consumption, Resource Use and Investment Responses to Economic

Growth, crisis and resilience: RN6.doc (rev. 1/01)

Growth, crisis and resilience: household responses to economic change in rural Southeast Asia

Evidence from Northern Thailand

Research Note No. 6: Social capital, vulnerability and resilience: a closer look

1

Growth, crisis and resilience: RN6.doc (rev. 1/01)

In this study we examine the combined effects of the Thai economic crisis and the 1998 El Niño drought on households in Mae Chaem, an upland and highland district of northern Thailand. Primary data are drawn from household surveys conducted in 1998 and 1999. We quantify the impact of the twin crises on household incomes, and examine household responses in terms of land use, broad consumption patterns, labor market participation and migration, and investments in such assets as human capital (schooling) and land quality.

Nearly all households in our sample derive income from agriculture or from agricultural labor. In addition, many have non-farm income sources such as trade and remittances from out-migrant family members. During the economic boom of the 1980s to the mid-1990s, the rural economy of Mae Chaem was transformed. The share of land devoted to subsistence crops (mainly rice) diminished greatly as markets and infrastructure development encouraged agricultural commercialization. Crops such as feed and seed corn and vegetables, produced primarily for external markets, occupied an increasingly important economic role. Factor markets were also radically altered as non-farm and migrant labor opportunities multiplied and formal-sector credit institutions emerged. Agricultural cooperatives and even formal banking became prominent credit sources, and remittance incomes from seasonal and permanent migrants emerged as a major source of rural income.

Naturally, not all households engaged with the formal market economy to the same degree, and the modes of engagement also differed. Some farms specialized completely in high-value commercial crops, while others allocated land so as to ensure adequate household rice supply in addition to cultivation of crops for market. Still other households specialized in other ways, including a significant fraction that began to form a kind of rural proletariat, earning income mainly through off-farm and non-farm labor.

The 1998 drought reduced all crop yields. The economic crisis raised input prices, causing most farmers to reduce the use of inputs such as fertilizer, although product prices were less directly affected. Lower farm profits made it difficult to service debt and pressured farmers to cut cash costs, for example by substituting family or exchange labor for hired labor. For some households, remittance income and/or non-farm employment opportunities diminished. Many households failed to produce enough rice to satisfy home consumption requirements and were forced to borrow.

Overall, the twin crises can be expected to have had differential effects on households according to their initial net wealth and the types and diversity of their income sources. Facing a common set of triggers— the drought, and market changes associated with the economic crisis— households display differing degrees of vulnerability. This heterogeneity, revealed largely in the pattern of household responses to the shocks, is of great importance. It provides clues as to how traditional and non-traditional social security mechanisms functioned during a major economic downturn. The study of household-level responses can generate insights for the design of policies and institutions for the protection of rural welfare during similar shocks in the future.

In our research we define measures of household wealth and its growth, relating these to households' initial resource bases and to their investments in land, education, businesses, migration and other non-farm activities. We then examine the ways in which each household was affected by the shocks, using both objective measures such as crop yields and remittance flows, and subjective measures as reported by households themselves. We use the information gathered to formulate and test hypotheses, and ultimately to draw conclusions, about the vulnerability of households to unexpected shocks, and the likely severity of their effects.

In terms of outcomes, our main focus is on changes in household investments in land, education and migration. In doing so, however, we find that another less tangible asset, 'social capital', apparently plays an important role in explaining variation in outcomes across otherwise similar households. Some forms of social capital, and in particular the web of informal relationships ('social networks') that link households of different kinds, may provide a safety net for some crisis-hit households, enabling them not merely to smooth consumption through the crisis, but even to maintain investments according to plan. This finding, if robust, has significant policy implications for Thailand, since it creates opportunities for the design of formal social safety nets that complement informal mechanisms, rather than replacing them at potentially far greater cost. It also spawns a new research agenda, in particular begging questions about the precise definition of social capital, and how it can be created, in the Thai context.

It is clear that along with the economic changes occurring in Mae Chaem during the decade of high economic growth there were significant social changes occurring as well. For many households, growth was accompanied by increasing reliance on formal economic institutions and market-based relationships than had previously been the case. Though this reliance on formal (and usually extra-village) links indicates another form of integration into the larger Thai society, it may also have resulted in increased exposure to risk. Participation in formal intra-village groups such as village loan funds, and market-based extra-village links to traders and banks was important (and helpful) for the income increases and wealth accumulation for certain villages and households. Participation in these groups may have been a way to gain information and to reap the economic benefits of growth through increased autonomy and diminished reciprocal obligations to networks of family and friends. But by sacrificing more traditional forms of mutual help and insurance in order to “get ahead”, households may have been left exposed to an unexpected economic downturn.

In a preliminary exploration of factors contributing to vulnerability, we constructed a measure of major household responses to the crisis and examined its relationship with a set of household characteristics. Table 1 shows a list of all reported responses. Among these, we judged changes in health expenditures, schooling, migration and forest product collection (all relative to a normal year) to be 'major' responses, that is, those representing disinvestment and thus likely to have long-term welfare effects. The vulnerability measure was a simple binary variable taking a value of 1 if a household reported any 'major' response, and 0 otherwise. Of 165 usable observations in our analysis, 65 reported a major response.

Table 1 Crisis response by sample households

aPercent of households reporting

We then constructed a set of variables representing household characteristics that might be expected to be associated with variation in the degree of vulnerability to a shock. These included the value of non-land assets; farm area; percentage of income derived directly from agriculture; ethnicity, sex, age, and educational attainment of the household head; the dependency ratio (number of working adults as a fraction of total household membership), and binary variables for households reporting health problems or family crises during the year. To these conditioning variables we added two proxies for social capital: the number of groups to which a household belongs, and the number of people a household reported it could count upon for help in a crisis. The former, 'groups', is intended to capture membership of formal organizations; the latter, 'help', the strength of informal network ties. We then used statistical analysis to test the hypothesis that any of these variables was associated with vulnerability. The statistical technique applied is known as ordered probit analysis. This technique enables us to test hypotheses about the statistical relationship between each variable representing household characteristics and our household-level measures of the severity of the shock and of their responses to it.

Our findings, when analyzed separately by village type, revealed a pattern that is both intriguing and informative. Overall, group membership is higher among wealthier households, and the same households tend to have more widely distributed network ties, including a higher proportion of ties to the formal or market economy than to family, co-villagers and clan members. This contrast emerges only in part because of differences that might exist between (poor) Karen and (less poor, or wealthy) Northern Thai and Hmong. The same pattern is as clearly seen within the sub-sample of poor (mainly Karen) villages as it is for the sample as a whole. When we used quantitative information on group membership and network ties in econometric analyses seeking associations with measures of vulnerability, we found the following.

First, the social capital variables were consistently more significant explanators of vulnerability than most other variables, including measures of household wealth, ethnicity, land endowments and income sources.

Second, the pattern of statistical significance varied across groups (for a summary of this aspect of the results see Table 2). Among the Karen, group membership counted did little to reduce vulnerability in the face of a shock, whereas network ties played a measurable mitigating role. For wealthier households, both group membership and network ties were significantly associated with diminished vulnerability. The pattern that emerges is one of wealthier households having both a greater asset base (including social capital) upon which to draw in times of trouble, but also a more diversified one, with the implication that at the community level, a more diversified safety net is one less prone to collapse in the face of a major shock. Among poor households, by contrast, group membership counted for little when all households were equally badly affected. Moreover, from our qualitative sub-sample we find that the "achieved" network ties of poor households are largely to employers rather than to a deeper or broader network. Both the drought and the economic collapse caused farm labor demand to decline in Mae Chaem, so for poor households these forms of achieved social capital also served poorly as safety nets during a crisis.

One potential explanation for the puzzling result is that group membership represents in some form the degree of integration of each household into formal Thai economic and social institutions. A similar hypothesis arose from a study by Robert Townsend (1995), who found an apparent decline of local insurance systems associated with entry into the larger market economy in the northern Thai context. In our sample there is a weak negative correlation between the 'group' and 'help' variables, consistent with a substitution of formal for informal relationships. During growth phases the former may be more appealing as mechanisms to obtain credit and other services without the need to take on reciprocal obligations. During a crisis, however, the lack of reciprocity of formal relationships may undercut the ability to smooth consumption and protect investments, e.g. by renegotiating the terms of debt repayment or sharing incomes and responsibilities.

Another interesting result from our statistical analysis could shed more light on the social capital puzzle. Though the majority of our respondents are from the northern Thai ethnic majority, there is a significant group of Karen, a hill-tribe group traditionally at the margins of the Thai polity. From analysis of sub-samples of the data we found that relatively asset-poor Thai households and relatively asset-rich Karen households were more likely to report major responses to the crisis. This finding supports the integration hypothesis in that it is these two groups of households who are likely to be participating in many groups but be marginalized within those groups. We can assume that most Thai would have relatively equal access to formal institutions at the village level. However, poorer households may still be marginalized by group structure or function. These relatively new groups are likely still developing working procedures and if the national struggle for democracy in Thailand is any indication then we can expect that even formal institutions do not guarantee equitable access. They are likely based on past hierarchical relationships and dominated by village elite interests. For Karen households, access may be even more restrictive and so only the richer Karen are able to participate. But as with the poor Thai, they may be marginalized within more formal institutions by virtue of their ethnicity or economic status. Again, these indications need to be explored more fully before any conclusions are definitively drawn.

1

Growth, crisis and resilience: RN6.doc (rev. 1/01)

Table 2: Significance tests on social capital variables from ordered Probit analysis

Note: "resilience" and "response" are dependent variables as described in text; "groups" is a count variable of group membership by household; "help" is a count variable of informal networks. "+" and "–" indicate statistically significant (at 5%) associations having the sign shown; "x" indicates no statistically significant association.

Authors: Ian Coxhead and Jean Geran, University of Wisconsin-Madison. This research was funded by a grant from the Ford Foundation and benefited from resources provided by the Thailand Development Research Institute Foundation. For more information visit or send email to .

1