Household Income Dynamics And Poverty Traps In Rural El Salvador

A growing concern about effects of the economic strategies followed by developing countries along the lines of the Washington Consensus suggests that the recommended policy reforms are necessary but not sufficient conditions for poverty alleviation. Several authors have pointed out the possibility of growth with inequality and the fragile status of the poorest sectors of the population, which during the growth process may not be able to weather the effects of income shocks because of their isolation from markets and limited opportunities for diversification. Additional policy and institutional reforms, to address poverty and inequality, will require a better understanding of the dynamics of income generation. The rural areas of El Salvador offer an excellent laboratory to study income dynamics and poverty in the context of a growing economy. In the period between 1995 and 2001, rural households in this country faced major consecutive income shocks. In addition to drought, brought about by El Niño in 1997 and 2001, the rural economy endured the effects of Tropical Storm Mitch in early 1999, two major earthquakes in early 2001, and a steady fall, since 1997, in the international prices of coffee, a major export crop. Consequently, most rural households in El Salvador are expected to be out of their steady-state equilibrium. Longitudinal household income data would shed light about the dynamic path of returning to equilibrium. This paper takes advantage of a detailed data set of rural household characteristics and income generating activities, to test for nonlinearity in the process of movement toward equilibrium. Non-linearity would reveal the possible existence of low-income poverty traps. The data set contains information from a sample of representative households of the rural population of El Salvador interviewed four times about their activities during the years of 1995, 1997, 1999, and 2001. This information presents a unique opportunity for this type of analysis, as it provides consecutive observations of households over the period and as the data were collected with the purpose of computing household incomes. In El Salvador, as a consequence of systemic and idiosyncratic shocks, rural incomes exhibit great volatility. In spite of an overall growing trend in total income, inequality and poverty measures do not decline over the entire period. A non-decreasing trend in inequality and in the number of poor households between 1995 and 1997 is contrasted with a decreasing trend in both indicators between 1997 and 1999 and between 1999 and 2001. Around 20 percent of the extremely poor in 1995, those households unable to generate incomes equal to the cost of a basic basket of goods defined to maintain the nutritional status of an adult, still remained extremely poor 6 years later. This was the case even though some of them may have temporarily escaped extreme poverty in one or more of the intermediate years. Ten percent of the extremely poor in 1995 could not escape extreme poverty at all at any time. The persistent presence of some households in the extreme left-hand side of the income distribution seems to indicate that they may have fallen below a threshold level of income from which their asset composition and income generating capacity would make it difficult to escape. The presence of such a poverty trap can be econometrically tested by posing a non-linear income dynamic process that allows for multiple equilibria, one of which may be unstable. The appropriateness of such test is guaranteed by the use of a panel data set containing microeconomic data on household characteristics and a detailed enumeration of all income-generating activities, including those activities in which households engaged as a form of self-employment. The downward measurement error bias typical of income computations is well established in the literature. First-difference estimation takes care of this drawback while at the same time reducing the problem of correlation between lagged incomes and the error term. Attrition, another typical problem with longitudinal data, is shown not to be systematically related to the income process. Between 1995 and 2001, there was 28 percent attrition. However, attrition is not significant in both an income generation model, where income is regressed as a function of household characteristics, and in the non-linear dynamic income model that tests for the presence of multiple equilibria. The aparent existence of a low-income poverty trap in rural El Salvador may be explained by sizeable successive income shocks that may have depleted assets endowments for low-income households through the years and have eventually rendered them vulnerable and unable to recover. Government subsidies do not seem to be effective to solve this problem. The test of nonlinearity in the income process is as an indirect test of the effectiveness of the income-smoothing mechanisms available to rural households. In developing economies, income-smoothing mechanisms take the form of diversification of activities within the household, changes in labor market participation, migration, and access to formal credit as forms of preventing future fluctuations in income. The methodology used in this paper considers the success of these mechanisms by looking at the resulting income process instead of attempting to evaluate the effectiveness of each mechanism separately, thus capturing the whole spectrum of possibilities available to households in rural areas and avoiding the oversimplified analysis of individual mechanisms prevalent in the literature.