Household Vulnerability and the CCT Within the Risk-Sharing Framework

Abstract

This chapter focuses on the empirical analyses of risk-sharing and household vulnerability to idiosyncratic shocks in rural Mexico. It aims to test the full risk-sharing hypothesis proposed by Townsend (1994), and to examine whether the conditional cash transfer (CCT) program reduces vulnerability within the risk-sharing framework using the same ENCEL rural household panel dataset for 2003 and 2007 applied in the previous chapters. Despite the rich information, the data have not been fully utilized given the lack of pure control groups. Drawing on Townsend’s (1994) risk-sharing model, the empirical results reject the hypothesis of full risk sharing at village level but confirm that risk-sharing functions serve better in securing basic needs such as food. Additionally, in this chapter, I define vulnerability as the lack of ability to smooth consumption because of liquidity constraints caused by restricted access to insurance and/or credit markets. The regression results of risk-sharing models considering the phase-in effects of the three treatment and control groups show that the risk-sharing function, reinforced by longer exposures to the CCT program, serves to mitigate vulnerability of poor households.

Appendices

Appendix 4.1

4.1.1 Variables

• Household real per capita food consumption: First, I construct each household’s weekly food consumption by summing up the reported amount of weekly food consumption and the estimated weekly self-consumption. Then, I divide the household’s weekly food consumption by the number of household members to ascertain the per capita weekly food consumption. In estimating self-consumption, I first calculate the median state price of each item using each household’s reported weekly purchase and the expenditure on the item. Then, I multiply the value of reported self-consumption by the estimated unit median price of the state. Per capita food consumption is deflated by the annual average food CPI. (Banco de México Estadísticas: http://www.banxico.org.mx/estadisticas/index.html); June 2001 = 100.

• Household real per capita total consumption: I construct a household’s real per capita total consumption in the same way as food consumption, using the reported weekly total consumption of food and non-food items. Per capita total consumption is deflated by the annual average general CPI.

• Household real per capita income in 2003 and 2007: This includes all household members’ wages, pensions, bonuses, monetary institutional transfers (including CCT), agricultural sales, and non-agricultural sales. It excludes personal transfers (including remittances), non-labor or irregular incomes, such as the sales of assets (e.g., houses, cars, and home electronics), inheritance, lottery, gifts, and donations. Personal transfers are excluded because they are more likely to reflect ex post adjustments to shocks, as Skoufias (2007) argues. The reported units for each income source vary from daily, weekly, and monthly to annual. Thus, I estimate the weekly amount of each income source and sum these up to estimate weekly household income. Then, I divide the weekly total income by the number of household members and deflate it by the annual average general CPI. Households that have any type of unreported income sources are dropped from the sample.

• Household real per capita income in 2001 and 2002: This consists of the sum of the household head and spouse’s retrospective weekly wage income divided by the number of household members, which is deflated by the average annual general CPI.

• Education dummies: Primary, secondary, or highschool refers to those who have enrolled in a primary, secondary, or high school, regardless of whether they graduated. Technical education refers to those who have enrolled in any technical or vocational school, including teacher’s college. University education includes those who have enrolled for a university and higher education (including those who graduated from university and have entered into or graduated from the post-graduate level).

• Household demographic variables: The total number of household members refers to the members who live in the same house. It excludes those who live separately for more than one year, whose stay is temporary, and who have expired. The dependency ratio is the proportion of household members under 14 years and over 65 years of age (non-labor force) to the number of household members aged 15–64 years (labor force).

• Local tortilla price: Drawing on Attanasio et al. (2009, 2013), I first calculate the median village (locality) price of tortilla using each household’s unit price (per kilogram), which is derived by applying the same method used to estimate self-consumption. I exclude median village prices above 20 pesos per kilogram for 2003 and 25 pesos for 2007. Then, I calculate the mean and standard deviation for the remaining median village tortilla prices, which are 4.7 and 2.1 for 2003 and 7.9 and 1.8 for 2007. I use each village’s median price as a local price of tortilla if the median price is between the mean value ± the standard deviation (2.6–6.9 for 2003 and 6.1–9.7 for 2007). The median price of a village with less than three households reporting the purchases of tortilla is also automatically dropped. Local prices that do not meet the criteria are replaced by the corresponding upper level (municipal) median prices, which are calculated using the same method as that for the village median price. I use the state median price as the local tortilla price in case the municipal median prices do not fulfill the criteria mentioned above. Of the sample, 30% was replaced by municipal median prices and 17% by state median prices in 2003, and 24 and 16% of the sample was replaced by municipal and state median prices, respectively, in 2007. The means (weighted by the number of households) of the local tortilla prices are 4.9 pesos per kilogram for 2003 and 8.1 pesos per kilogram for 2007. These estimated prices are quite reasonable since the means of the state median prices (weighted by the number of households) are 5.2 and 8.2 pesos. The local price changes for 2003 and 2007 are deflated by the food CPI.

Appendix 4.2

See Appendix Table 4.7.

Table 4.7 Regression results of the first stage of Eq. (4.12)

Appendix 4.3

See Appendix Tables 4.8 and 4.9.

Table 4.8 Regression results of Eq. (4.12)

Table 4.9 Regression results of Eq. (4.12)