Chapter 8 Vulnerability to Poverty: A Microeconometric Approach and Application to the Republic of Haiti

Abstract

This paper investigates vulnerability to poverty in Haiti. Research in vulnerability in developing countries has been scarce due to the high data requirements of vulnerability studies (e.g. panel or long series of cross-sections). The methodology adopted here allows the assessment of vulnerability to poverty by exploiting the short panel structure of nested data at different levels. The decomposition method reveals that vulnerability in Haiti is largely a rural phenomenon and that schooling correlates negatively with vulnerability. Most importantly, among the different shocks affecting household's income, it is found that meso-level shocks are in general far more important than covariate shocks. This finding points to some interesting policy implications in decentralizing policies to alleviate vulnerability to poverty.

Addressing High Food Prices: A Household Vulnerability Analysis in Rural Burundi

“…Such a three-step methodology was further implemented, including information on covariate shocks (Christiaensen and Subbarao, 2005), by utilizing historical information on price yields and production (Christiaensen and Sarris, 2007), allowing for instrumental variables for crop income (Sarris and Karfakis, 2006), and estimating with maximum likelihood allowing for multilevel-data (Günther and Harttgen, 2009;Jadotte, 2010).…”

“…In formal terms, recalling and readjusting equations [5.1] to expand the 3-step methodology with multilevel analysis (Goldstein, 1999;Günther and Harttgen, 2009;Jadotte, 2010) Controlling for household heteroskedasticity guarantees the estimates to be efficient, thus allowing a proper estimation of a household's probability of being 65 Kamanou and Morduch (2002) propose a non-parametric approach focusing on Monte Carlo simulations of future consumption distribution, based on bootstrapping observable shocks. This model is yet based on the assumption that the shock distribution would be the same across all households, regardless the different circumstances they face, or, to put it differently, not allowing for heteroskedasticity (Chaudhuri, 2003).…”

“…Moreover, provided the marginal effects of the regressors on the ex-ante mean and variance of future consumption can differ in sign granted such a flexible heteroskedastic specification (Just and Pope, 1979), it could be possible to infer in risk-coping ability (Jadotte, 2010), firstly including consumption smoothing (Christiaensen and Subbarao 2004).…”

“…where K represents the covariates X ijk , Z jk , and W k at every single level, and [5.11] is modeled as a non linear-function of K (Jadotte, 2010), to avoid negative variances.…”

“…where v jk and u k represent shocks at the province and sous-colline level affecting household consumption (Jadotte, 2010).…”