The Mexican economy experienced three major economic events during the last decade. First, the implementation of the North American Free Trade Agreement (NAFTA) starting on January 1, 1994, was a major milestone in the opening of the Mexican economy. Second, Mexico suffered its worst economic crisis since the Great Depression in late 1994 and throughout 1995. Third, the U.S. recession beginning in late 2000 resulted in a major decline in Mexican manufacturing employment, particularly in the maquiladora industry, where more than 200,000 jobs were lost between 2000 and 2002 (INEGI 2004). In addition, several major political events occurred during the last decade. On the same day NAFTA came into effect, there was open rebellion in the state of Chiapas. Also in 1994, Mexico's presidential candidate Luis Donaldo Colosio was assassinated in Tijuana in March. Later that year, in September, Mexico's attorney general was assassinated. Mexico selected Vicente Fox as its president in 2000. Fox was the first Mexican president since 1928 who did not belong to the Partido Revolucionario Institucional (PRI). Fox promised substantial economic and political reforms. In brief, Mexico experienced considerable economic and political instability during the last decade.
Changes in income distribution often occur slowly over long periods of time, but periods of significant instability may result in distributional effects during the short run. Mexico's political and economic instability during the last decade could be expected to affect its distribution of income in some obvious and not so obvious ways. In this paper, we address three specific questions regarding income distribution in Mexico. First, in what ways, if any, did the distribution of total money income in Mexico change during this turbulent decade? Second, did Mexico's income distribution in urban and rural areas change in different ways during the decade? Third, was the instability of the last decade associated with meaningful differences in the distribution of wage and nonwage income in Mexico?
The paper is organized into three parts. After this brief introduction, the data set and its limitations are described. Next, we discuss briefly our choice of the Gini coefficient as a measure of distribution and a method proposed by Shujie Yao (1999) for decomposing the Gini coefficient into various subgroups. Finally, we present empirical results and an interpretation of those results.
The Data and Their Limitations
Data for this study were obtained from sample surveys conducted by INEGI (Instituto Nacional de Estadistica, Geografia e Informatica 1992-2002). INEGI is Mexico's equivalent of the U.S. Census Bureau. Since 1992, INEGI has conducted sample surveys that are in many ways similar to the Current Population Surveys (CPS) conducted by the U.S. Census Bureau. The surveys are conducted every other year in even-numbered years. The title of the surveys is revealing: Encuesta Nacional de Ingresos y Gastos de los Hogares (ENIGH), or the National Survey of Income and Expenditures of Households. The intent of the surveys is to produce national estimates of income and expenditures for households, families, and individuals, but the surveys also provide estimates of many other variables including educational attainment and employment characteristics of the population.
In this study, we used the data on money income of households even though the survey collects data on nonmoney income such as in-kind income and imputations of the value of owner-occupied housing. Other studies (Attanasio and Szekely 1998 and Bouillion et al. 2003) have used the in-kind income, but our review of the data provided a certain level of discomfort due to the odd ranges obtained and due to the subjective nature of the data. In addition, for years where we could compare our results with those of other studies we found our computations to be somewhat similar to those using the imputing values. For instance, Orazio Attanasio and Miguel Szekely (1998) estimated the Gini for Mexico in 1996 to be 0.53, and the estimate presented below is 0.52. The study used data from the 1992, 1994, 1996, 1998, 2000, and 2002 surveys. Ideally, the surveys would have been conducted in all years. In this study, for example, the most severe year of the 1994-96 recession (1995) had to be omitted from the analysis. Although similar surveys were conducted as early as the 1950s, they were not conducted on a scheduled basis until 1992. Since 1992, the surveys have been administered using a consistent sample selection procedure and similar survey instruments and with careful post-sampling evaluation studies. The surveys were conducted at approximately the same time each year.
Economists should always be skeptical of income...