دوستانتان را حفظ کنید: اثر شبکه های اجتماعی محلی بر نتایج سرمایه انسانی کودک

Social networks can affect demand for human capital investments by relaxing household time or budget constraints or by defining and reinforcing human capital preferences. However, empirically identifying the effect of social networks on human capital investment is usually problematic because households self-select their networks in ways that may be correlated with their abilities to make these investments. In Northern Ugandan Internally Displaced Persons Camps, networks were not entirely self-selected. Rebel activity, which forced households into camps in 2002, disrupted pre-existing social networks in ways that were exogenous to household human capital preferences. This paper uses the exogenous variation in network disruption to identify a positive impact of networks on children's long-term nutritional outcomes.

Local social networks – the people that a household spends time with on a daily basis – can critically influence time-sensitive household behaviors. In particular, these local networks may affect the level of households' daily investments in children's nutrition, which is determined largely by the interaction of time-consuming care-giving practices and food and health resource availability. In most cases, it is impossible to demonstrate whether or not these networks ultimately improve child nutritional outcomes because households self-select into networks in ways that may be correlated with their abilities to invest in their children's nutrition. Using data collected in Northern Uganda, this study takes advantage of an exogenous disruption to households' local social networks, caused by a rapid escalation of a long-standing civil conflict, to show that the presence of a household's local network improves nutritional outcomes of the household's youngest members. Recent empirical work has demonstrated the importance of a household's social network to a number of economic outcomes and decisions, such as access to employment (e.g. Ioannides and Loury, 2004, Magruder, 2010, Munshi, 2003 and Topa, 2001), participation in social programs (Aizer and Currie, 2004 and Bertrand et al., 2000), and retirement investment decisions (Duflo and Saez, 2003). In developing countries, the role of social networks may be more important as a means of overcoming market failures or the absence of other institutions. Trust relationships function in the absence of contract-enforcement mechanisms (Fafchamps, 2004 and Greif, 1993), and networks provide insurance and credit systems where formal markets do not (Besley, 1995, Grimard, 1997, Rosenzweig, 1988 and Townsend, 1995). Finally, networks are a significant source of information about technology, health care, and returns to investment (Conley and Udry, 2010, Leonard, 2007 and Yamauchi, 2007), and affect use of health care technologies (Oster and Thornton, 2009). This paper focuses on a specific type of network, the local social network, that influences households' investments in child nutrition through daily contact with household members. The cumulative effect of these daily investments improves nutritional outcomes. A households' social network in general can influence the households' ability to make nutritional or health investments by expanding the household resource pool or reducing investment risks. Local social networks increase investments in child nutrition in ways requiring more frequent contact and stronger ties, such as through coordinating daily activities and framing and reinforcing certain health-seeking behaviors. For example, members of sufficiently large localized networks benefit from economies of scale in human capital production by sharing child care and home production activities. Local social networks also can increase nutritional investments by restricting the household's resource allocation choices to those that are potentially better for young children.1 As the network grows, the ability of the network to monitor and reinforce these norms and the possibilities for coordination also increase. Studies of social networks' relationship to any outcome or household behavior suffer from serious identification problems stemming from the fact that households are not assigned to networks, but choose whether to join, choose whether to live in places where there are networks, and choose whether to form networks. Some unobservable aspects of a household's ability to invest in child health, such as income, access to formal credit, or preferences for child health or health knowledge, may be correlated with the household's ability to form and maintain networks. Moreover, measuring the size of the network that might influence nutritional investment is difficult. Some past studies have used neighborhood or ethnic group to proxy for the network (for example, Aizer and Currie, 2004 and Bertrand et al., 2000), however, not all members of a neighborhood or ethnic group may influence a household's demand for health, introducing an errors-in-variables problem. In this study, I overcome both the problem of self-selection into networks and of network measurement using data from Ugandan internally displaced persons (IDP) camps. In these camps, the size of a household's network has an exogenous component. The civil war that forced households into IDP camps significantly disrupted all aspects of families' lives, including whether they ended up living near the people with whom they had previously shared their lives.2 Some households ended up with smaller local social networks for reasons out of their control. Because the civil war spread unpredictably over the region, households were forced to move to IDP camps at different times and under different circumstances. Villagers living in areas of heavy fighting moved to the camps first and were generally more likely to find space to build with friends and family. These households typically suffered less disruption to their networks. However, within months of the conflict's intensification, the military evacuated the remaining villages, giving households 48 h to move into camps. The chaos associated with such a large influx into the camps made coordination with friends and family virtually impossible. As a result, some part of the local network that a household was left with in the camp was exogenous to the households' preferences for social network size and its ability to make nutritional investments. I also address the network measurement problem by collecting details about the size of households' farmers' groups, which are self-selected agricultural labor teams, before and after the displacement. As in many parts of Africa, Northern Ugandan farmers' groups play a critical role in defining and reinforcing household child-rearing behaviors and in supporting members in daily child care activities. Households share daily farm and household tasks, including meal preparation and child care, so members depend on each other to ensure income and for child health production. Group cohesion is based on reciprocity of labor and therefore demands long-term, trusting relationships. Given the financial and social significance of farmers' groups, households can easily recall in an interview the members of their farmers' group prior to displacement and where members lived during displacement. Therefore disruption to the farmers' group provides an easily-measured indicator of disruption to the networks that matter most to child outcomes. Moreover, while households did form new networks within the camps, quantitative and qualitative data collected for this study show that these networks were not significant influences on household child-rearing behaviors in camps compared with networks established before displacement. While initial variations in disruptions to households' social networks are exogenous to household preferences, some households may have been able to coordinate with pre-displacement farmers' group members during transition to live nearby in the camp. If the ability to coordinate is correlated with human capital preferences, then estimates of network size impacts based on an uninstrumented measure of network disruption may be biased. Moreover, some households may have relied more strongly on households that were not part of their farmers' group to support human capital investment. These network members would not be included in the network measure based on farmers' group disruption, presenting a potential errors-in-variables problem. Therefore, I use village-level fragmentation measures, which are unlikely to be correlated with households' human capital and network preferences, to instrument for the farmers' group disruption. I infer variations in contributions to child health and nutrition from variations in children's height-for-age z-scores (HAZ). HAZ is a cumulative indicator of nutritional status and, for preschool-age children, is a function of lifetime dietary intake and morbidity, and genetics. Local network size can positively impact HAZ by increasing household contributions to nutrition and health. I find that an increase in local network size of one household (or roughly 25% of the average local network) leads to between a 0.06 and 0.22 standard deviation improvement in HAZ.

This study shows that small changes in households' network size yield substantial improvements in HAZ for preschool-age children, demonstrating the role of networks in increasing health and nutrition investments. In particular, I find that maintaining an additional pre-displacement network member as part of the in-camp network leads to a .20 HAZ improvement. As the average in-camp local network size was under 4 households, a marginal increase in network size is equivalent to increasing the average network by approximately 25%. In most settings, identifying network effects is impossible because households self-define and self-select networks in ways that may not be observable. For this study, data collected in IDP camps allowed me to identify the members of households' self-defined and self-selected networks and then to isolate an element of exogenous variation in network size across households. I measure the in-camp social network as the number of households that were part of a respondent's farmers' group before displacement, a group identified as influencing human capital investment in the village, who remained part of the respondent's local network in the camp. However, since households may be able to affect their in-camp network size by restoring network links in the camp, even the in-camp network size may have an endogenous component. Thus I use the ways in which a civil conflict fragmented villages in camps to explain exogenous variations in households' resulting in-camp network sizes. Since village fragmentation explains part of in-camp network size and was exogenous to household preferences for network size and households' ability to invest in child health, measures of village fragmentation can be used as instruments for in-camp network size, allowing me to identify a causal network effect. This study considers a particular type of network, the local social network, that has two important attributes: daily contact and long-standing ties. Farmers' group members who did not live in the same block of the camp as the respondent did not have the same effect on child health outcomes as those who lived nearby and had daily contact. Additionally, in-camp neighbors who did not have long ties with the respondent did not appear to influence households' behaviors in the same way that farmers' group members did. The daily contact with close ties affects child nutritional outcomes by impacting daily nutritional investments in these children. While the ability to detect network effects on child outcomes relied on the camp context, the mechanisms through which networks can impact child outcomes can apply widely. First, the possibility of networks increasing access to information or financial resources or reducing household constraints has been demonstrated in the context of informal insurance and access to employment, which involve periodic network influences. The findings in this paper suggest that networks may also improve access to resources for daily investments, as children's linear growth requires daily nutritional inputs. The explanation that networks restrict choices has been used to explain the persistence of racial inequality (e.g. Loury, 2002), in which individuals may make choices that appear economically detrimental in order to maintain group identity. The evidence presented above shows that network restrictions on choice may lead households to make better choices where their children are concerned. Investments in early childhood nutrition lead to increased educational attainment, improved cognitive functioning, better adult health outcomes and higher productivity and wages. Additionally, the effects of inadequate early childhood nutrition, which include poor adult health outcomes, reduced educational attainment, impaired cognitive functioning, and lower productivity and wages, are largely irreversible (e.g. Martorell et al., 1994). In developing countries, where food resources are scarce, social networks can play a critical role in ensuring that resources are directed to young children during critical windows so that households can realize these future benefits. In Northern Ugandan IDP camps, the stakes are particularly high as the opportunities for human capital development are limited for all age groups. The effects of households' networks on early childhood nutritional outcomes could be critical to lasting economic recovery as households rebuild their lives.