Education and the Vulnerability to Food Inadequacy in Timor-Leste

Abstract

This paper adopts a simple empirical approach to estimate vulnerability to food inadequacy using cross-section data from the 2001 Timor-Leste Living Standard Measurement Survey. This measurement is based on the assumption that households are exposed to the same kind of shock. It is found that the distribution of vulnerability to food inadequacy to the education of household head is more significant than that to observed food poverty. The results support the argument that senior primary and tertiary education can help reduce the food risk that households face, in particular the risk that a household is undernourished. Thus, in Timor-Leste public spending on these forms of education can provide a form of support that favours the poor.

Notes

Helpful comments from an anonymous referee and an assistant editor are gratefully acknowledged, as is the encouragement of the Managing Editor. The authors are grateful to the International Fund for Agricultural Development (IFAD) for financial support through grant number ANU 927–ASARC, and Raghav Gaiha and Katsushi Imai for useful advice. The usual disclaimer applies.

¹ Luttmer (2002) also discussed measurement error in well-being and used two other indicators of such well-being: income and subjective answers.

² See Ligon & Schechter (2003) for another approach where vulnerability is considered as expected utility. This approach is suitable only for countries where panel data are available. See also Christiansen & Subbarao (2005), Christiansen & Boisvert (2002) and Scarmozzino (2006).

³ Otherwise, it may be necessary to use the pseudomaximum likelihood estimation, to take into account effects of clustering, which can be implemented by using “svy-intreg” in STATA (Gibson & Rozelle, 2003).

⁴ Or has a lower calorie intake than the minimum requirement as per the Timor-Leste Living Standards Measurement Survey.

⁵ The food poverty line is defined as the minimum expenditure needed to buy 2100 calories per person per day.

⁶ But, allowing for economies of scale can alter this basic correlation. However, there remains debate over how to measure these economies. Lanjouw & Ravallion (1995) indicate that the correlation between household size and poverty vanishes, in the case of Pakistan, when the size elasticity of the cost of living is 0.6. Anand & Morduch (1999) provide a succinct summary of analysis of the relationship between poverty and population.

⁷ It is difficult to gauge causality in cross-section data. To see the importance of the education variable we estimated the model (as in Table 6) without this variable. We also estimated the same model without the educational dummy variables. The signs and significance of the variables are largely unchanged, however the magnitudes are smaller and the fit of the equation is marginally poorer. These results are not reported here to conserve space.

⁸ We could have used the alternative threshold mentioned. This would have raised the proportion of the vulnerable but would not have the advantage that if a household is just at the poverty line and expects a mean zero shock it has vulnerability to poverty of 0.5 (Pritchett et al. 2000). We modelled vulnerability using both thresholds but do not include the results for lack of space. These results are available on request.

⁹ It should be noted that household responses are elicited from a particular individual within the household, rather than the household. As such, responses from wives may differ from those of husbands. Further, respondents may have different views on how food is shared within the household. Moreover, individual responses may themselves depend on who is conducting the interview. Thus, the results of the analysis undertaken may be sensitive to who responds and who conducts the interview. For more details, see Dercon (2004, 2005), Dercon et al. (2005) and Harrower & Hoddinott (2005). We are grateful to an anonymous referee for pointing out this argument and literature to us.

¹⁰ As in the case of Table 6, we proceeded to investigate the importance of the education variable by estimating a model (as in Table 9) without this variable. We also estimated the same model without the educational dummy variables. The signs and significance of the variables are largely unchanged, however the magnitudes are smaller and the fit of the equation is marginally poorer. These results are not reported here to conserve space.

¹¹ See World Bank (2003a) for methods of coping with food shortages in Timor-Leste.

¹² A similar result has been found by Jha et al. (2007) for the case of poverty in India.