Poverty in Mexico, 1990-2020

James Pick, University of Redlands
Maria-Antonieta Rebeil, Universidad Iberoamericana

The object of the present paper is first to analyze quantitatively the changes in the magnitude and profile of poverty in Mexico from 1990 to 2000, and second to perform statistical projections through the year 2020, with the goal of assessing future transformations of poverty in the nation. This paper has the following research questions. 1. What are the correlates of poverty at the state level in Mexico in 1990? 2. What are the correlates of poverty at the state level in Mexico in 2000? 3. Based on regressions determined in 1.and 2., what is the forecast level of poverty in Mexico in 2020? 4. What are the implications of the results from the study for present and future government policy decisions? Methodology. The paper utilizes data of the decennial Mexican Censuses of Population and Housing of 1990 and 2000 (INEGI, 1992, 2002). These censuses are considered sufficiently accurate for data analysis The unit of analysis for the study is the state. It is selected because analysis at the state level allows coverage of the entire nation, which has greater import than studying a single state or region. The dependent variable is the Poverty Index. This is constructed from a weighted average the following seven components: illiteracy, primary occupation, water scarcity, plumbing scarcity, electricity scarcity, earth floor, and crowding. The present index is similar to the “Indice de Marginación” (Progresa, 1999). The main change in the present Population Index is in the choice of weights. The principal methodological technique is stepwise regression analysis. The list of the regression variables and their definitions are presented the full paper. The independent variables, listed below, were selected as important ones based on prior studies of poverty, some of which have been discussed in the literature section, and on their practical important in influencing poverty. Based on prior literature studies, the expected directionality of effects of the independent variables on poverty may be summarized as follows: Indep. Variables expected to Indep. Variables expected to Direction of effect unknown have + links to poverty. have – links to poverty small sized localities (+) secondary education (-) migration (?) fertility (+) religion (?) infant mortality (+) low level occupation (+) indigenous population (+) common law marriage (+) no health services (+) The methodology to forecast the national poverty level to year 2020 is regression-based forecasting (Makridakis, Wheelwright, and Hyndman, 1997; Hanke, Reitsch, and Wichern, 2001). In particular, we first determine the most significant correlates of state poverty for 1990 and 2000. We determine if the predictors are similar or different for the two years. If they are similar, it implies that the most significant poverty predictors have not changed during the 1990s, and we can adopt the year 2000 regression model or one close to it to forecast the two decades up to the year 2020. If the predictions are different, the next step in this forecasting methodology is to estimate the future values of the independent predictor variables (Makridakis, Wheelwright, and Hyndman, 1997; Hanke, Reitsch, and Wichern, 2001). This is done by exponential extrapolation of the 1990-2000 national rates of change to the period 2000-2020 and by exponential extrapolation of trends in more lengthy national historical series for attributes, if they are available. Exponential extrapolation applies compounded annual rates of change to a year 2000 base value. The estimated 2020 levels of the predictors are substituted in the forecasting regression equation to obtain estimated poverty values for 2020. In the paper we apply the regression model to independent variables which consist of the national average for the 32 states. Results and Discussion. The results for the state average values shows a moderate reduction in the poverty index from 2000 to 2020. The three alternatives yield similar values that result in average reduction of 9.59 percent. The computations based on attribute values aggregated for the nation lead to somewhat greater reduction in national poverty level by 2020. This comparison run results in an average reduction in poverty of 14.4 percent. Both these reductions are less than the poverty reduction that occurred during the decade of the 1990s. There are several reasons the drop is less dramatic. The main reason that state averaging of different reductions can lead to sharp differences. This is what may have happened in the 1990s i.e. populous states had a sharper lowering in indigenous and/or low level occupations. The erratic nature of state lowering is reflected in the large coefficients of variation for indigenous in 1990 (143.9) and 2000 (125). However the coefficient of variation has dropped in the 1990s, and it is likely to continue to fall, perhaps to well under 100, so that further drops in percent indigenous will vary less among states. A second reason for the lesser drop is that between 1990 and 2000, the relative weighting of indigenous to low level occupations in the regression coefficients changed by 2.6 times. This means that the relatively slower rate of reduction of indigenous is much more reflected in the forecasts. From a practical standpoint, this is important i.e. that indigenous proportion in Mexico appears stabilized and it accounts for an increasing proportion of poverty. This may slow government efforts to reduce poverty, unless it can effectively transform the level of indigenous standards of living to approach more the mainstream. This paper demonstrates that at the state level in Mexico, two predictor variables are of overriding importance – indigenous population and low level occupations. Of the eligible variables, these more than any others point directly to the lowest ends of the socioeconomic scale. Hence, it is not surprising that these are tied to poverty. The two factors are able to predict around three quarters of the variation at the state level. The mix of the two factors has changed from 1990 to 2000, with indigenous being emphasized relatively more. At the same time, over the past 60 years, the proportion of indigenous population in Mexico has fluctuated down and up, but has not changed much from the period’s beginning to end. One reason is that although the indigenous population has become more assimilated, it has also had significantly higher fertility rates. Politically, the plot of the indigenous people is highlighted by Mexican social protest movements, particularly the Zapatistas. From the standpoint of this mathematical analysis, reduction in proportion of indigenous population would reduce poverty. Certainly, however, the appropriate measures are not to reduce the proportion of indigenous people, but to increase their standard of living to approach national norms. In conclusion, the research questions can be answered as follows: What are the correlates of poverty at the state level in Mexico in 1990? The significant correlates, of the ten independent variables studied, are low level occupation and indigenous population. What are the correlates of poverty at the state level in Mexico in 2000? The significant correlates, of the ten independent variables studied, are indigenous population and low level occupation. Based on regressions determined in 1.and 2., what is the forecast level of poverty in Mexico in 2020? The forecast level of poverty in Mexico in 2020 is between 5.9 and 6.3 Poverty Index, versus actual levels of 6.9 in 2000 and 11.0 in 1990. The lesser rate of drop in poverty is because indigenous population has risen as the dominant predictor variable, but historically in the twentieth century and in the 1990s, it fell relatively slightly. It is not expected to fall greatly in the next two decades, due to the higher fertility rates of indigenous people and low expectation of assimilation. What are the implications of the results from the study for present and future government policy decisions? The implications are that the Mexican federal government, and also the states, need to focus their efforts on geographical areas of high indigenous and high extent of low level occupation. Those efforts would benefit by extending from the traditional efforts to improve health and illiteracy to include more emphasis on infrastructure improvements and job/work enhancement for poor people. Poverty remains a huge national problem in Mexico. There are 44 million poor according to SEDESOL (Vargas and Muñoz, 2001), nearly half of the population. In the U.S., by contrast the poor constituted 13.1 percent of individuals in 2000. Furthermore, the poverty line is much lower in Mexico than in the U.S. Mexico’s great thrust in the past decade towards globalization and an open economy must be tempered by the need to pay attention to its huge poor population. The scale of work to be done to reduce poverty is immense, certainly larger than the scope of the U.S. “War on Poverty” of the 1960s. The present research provides insights to understand poverty and its correlates at the state level. The present methods can be extended to analyze and forecast poverty at the municipio level throughout the nation. More specific information for smaller regions can further help target particular policy initiatives and actions tailored to specific zones. Note: Reference available from authors and in full paper version.