1. Introduction

The recent COVID-19 pandemic has come at overwhelming health and economic costs to Latin America, Mexico in particular.¹ In August 2020, Mexico was among the top ten countries in terms of infections and the top five in terms of deaths per one-hundred thousand inhabitants.² To contain the spread of the virus, governments implemented lockdown policies of various degrees.³ Although in Mexico, the federal government did not implement full-lockdown measures, “stay-athome” recommendations were put in place. Inevitably, these measures caused a sharp reduction of activity, a fall in employment and income, and a rise in poverty and inequality.⁴ In this paper, we analyze the impact of the economic dislocation on poverty, inequality, and income mobility in Mexico, and compare it with the impacts in the three other largest Latin American countries: Argentina, Brazil, and Colombia.⁵ In addition to lockdowns to control infection rates, governments in these three countries have introduced new or expanded social assistance measures to varying degrees. In Mexico, there has been no such expansion.

Based on the economic sector in which household members work, we use microsimulation to estimate potential income losses at the household level, using the National Survey of Household Income and Expenditure (Encuesta Nacional de Ingresos y Gastos de los Hogares, ENIGH, 2018).⁶ The simulations first identify individuals whose income is “at risk” because they work in sectors in which COVID-19 has reduced or eliminated activity. We base our determination of at-risk income on the economic sectors in which one works. We assume that income derived from work in sectors that are “essential” is not at risk, while any income earned in “nonessential” sectors is at risk. For Mexico, we use the International Labour Organization (ILO) definition of essential sectors. We aggregate this at-risk income to the household level and then simulate actual losses using a range of two key parameters: the share of households with at-risk income that actually lose income and, of those who lose income, the share of at-risk income lost. We allow both parameters to range from zero to one-hundred percent, yielding one-hundred possible outcomes. To narrow our focus to reasonable possibilities, we choose a combination of the two key parameters that yields a decline in per capita income that comes closest to the International Monetary Fund (IMF) World Economic Outlook forecast from October 2020: 9.0 percent for Mexico.⁷ Even here, there are multiple possibilities of which we present two extremes, one in which a smaller proportion of households lose a large share of their income and another in which a larger number of households lose less income. To complete the analysis, we construct a simulated income distribution that incorporates the losses we estimate and compare it with the ex-ante distribution. In addition to comparing standard distributional statistics for each income distribution, we find it especially useful to examine income losses conditional on one’s position in the ex-ante distribution.⁸

This paper makes several contributions. First, most existing exercises that predict the impact of COVID-19 on poverty assume that income losses are proportional across the income distribution.⁹ Two exceptions are the Economic Commission for Latin America and the Caribbean (^{ECLAC, 2020}) and ^{Hufmann and Najera (2020)}, who combine data from several surveys -including post-COVID telephonic surveys and use small-sample and matching methods to predict postCOVID incomes and extreme poverty. Based on existing information, however, the distribution of income is changing -and changing fastduring the pandemic.¹⁰ Our use of microsimulation allows us to relax the equal loss assumption and so incorporate distributional changes in the analysis. In particular, we use techniques analogous to nonanonymous growth incidence curves to describe income losses across the ex-ante income distribution. Second, we carry out the analysis not only at the national level but also by ethnicity and gender of the household head, and by urban and rural areas. Third, we test the sensitivity of our results to two assumptions: the definition of income “at risk” and the size of the aggregate contraction.

Our findings show that the potential impact of the economic dislocation on inequality and poverty can be quite large. The Gini coefficient could rise between 1.3 and 3.7 points. The increase in the number of poor (measured with the national poverty line) could be between 7.5 and 8.7 million. In contrast, a distribution-neutral simulation would result in an increase of 6.8 million individuals. In addition, contrary to many people’s priors, the non-anonymous growth incidence curves show that income losses could be greater in the middle deciles of the ex-ante distribution. That is, losses are more pronounced for the moderate poor and those vulnerable to falling below the poverty line rather than among the poorest. This is so because social assistance programs and consumption of own production represent a larger share of total gross income for the poorest; essentially, these two items put a “floor” for their incomes. The economic dislocation is likely to cause a smaller increase in poverty for rural areas because it is here where consumption of own production -as a share of gross incomeis the largest. Because the indigenous population is concentrated in rural areas, the rise in poverty among the indigenous population is potentially lower than for the non-indigenous one. The expected rise in poverty is similar for male and female-headed households.

Our exercise has some important caveats. The microsimulations do not take into account behavioral responses or general equilibrium effects, so they yield first-order effects only. The depth and duration of the crisis are still uncertain, and the Mexican economy could end up contracting by more (or less) than the IMF’s October 2020 projections. Our results depend on the specific assumptions we make about income sources that are “at risk” and the extent to which losses are concentrated or dispersed across households. Finally, our exercise does not include social assistance measures undertaken by the Mexican states (or other subnational entities).¹¹

2. Data and methodology

For our simulations, we use the ENIGH 2018, the most recent income expenditure survey available for Mexico. We use gross income per capita as the welfare indicator. Gross income is defined as labor income plus rents, private transfers, pensions, consumption of own production, the rental value of owner-occupied housing and government cash transfers before any direct taxes. We update gross incomes for Mexico to 2019 by the rate of growth of the Gross Domestic Product (GDP) per capita for 2019 multiplied by a so-called pass through of 0.85.¹² Also, for Mexico, we update gross incomes to take into account the significant reforms introduced to the cash transfers system in 2019.¹³

We obtain our estimates by simulating potential income losses at the household level using the updated ENIGH. The simulations first identify individuals whose income is “at risk” because they work in sectors in which the COVID-19 pandemic have reduced or eliminated activity. We base our determination of at-risk income on the economic sectors in which one works. We assume that income derived from work in sectors that are “essential” is not at risk, while any income earned in “nonessential” sectors is at risk. For Mexico we use the ILO definition of essential sectors.¹⁴ At the household level, the at-risk incomes also include rental incomes and incomes of informal street vendors (regardless of the sector in which they work). We aggregate this at-risk income at the household level.

Regarding incomes that are not-at-risk, besides including all the employment not counted as at risk according to the above definitions, we make two additional assumptions. First, incomes from cash transfers programs, consumption of own production, imputed rent, social security pensions, public employment, and private transfers (e.g., remittances) are not affected. Second, we do not consider the income of white-collar workers who are CEOs, managers and researchers with internet access at home to be at-risk even if they work in nonessential sectors.¹⁵

Figure 1 shows the composition of per capita household income by centile of the pre-crisis income (per capita) distribution across seven categories: cash transfers, consumption of own production, imputed rent, social security pensions, government salaries, other incomes notat-risk, and incomes at-risk. This is shown at the national level and disaggregated by urban and rural areas. The population in urban areas, defined as those living in localities with 2,500 inhabitants or more, represents 75.5 percent of the country’s total population.

Notes: Rural is defined as a locality with less than 2,500 inhabitants; 24.5 percent of the population is rural.

Source: Authors’ calculations based on ENIGH (2018).

Figure 1 Composition of per capita household gross income 

There are several results to note. First, the share of income that is not at risk is not equal across the income distribution as many studies assume, nor is it uniformly decreasing in income as it would be if the poorest were most at risk. Rather, it is U-shaped with the greatest risk in the middle of the income distribution rather than either extreme. The very poorest households have an income floor (albeit low) that protects an important share of their income. This income floor comes from three sources: cash transfers, consumption of own production, and imputed rent. Consumption of own production is especially notable in rural areas.

After at-risk income is identified, we simulate potential losses using a range of two key parameters: the share of households with atrisk income that actually lose income and, of those who lose income, the share of at-risk income lost. Households who actually lose income (from the set of households with at-risk income) are randomly selected. We allow both parameters to range from zero to one-hundred percent (in 10 percent intervals), yielding a ten-by-ten matrix of possible income losses (table 1).

Cells in table 1 show the range of possible per capita gross income losses (as a proportion of ex-ante gross income) of all households as we vary both the probability that households lose at-risk income (down the rows) and the share of that at-risk income they lose (across the columns). For example, in the 10 percent-20 percent cell of this matrix, we show the fall in income in percent corresponding to the case in which 10 percent of the households (with at-risk income) lose 20 percent of their income each (and so on). The possible losses are very wide indeed, ranging from near zero to near 30 percent of precrisis income. For macroeconomic consistency, we narrow our focus to outcomes that have income losses similar to the IMF’s October 2020 World Economic Outlook projections for the decline in GDP per capita, highlighted in gray in table 1.¹⁶ From that “iso-loss” curve, we choose two scenarios: either a smaller proportion of households lose much income, or a larger proportion of households lose smaller amounts of income. In the selected “concentrated losses” scenario, 30 percent of households lose 100 percent of their income while in the “dispersed losses” scenario, 100 percent lose 30 percent.

3. Results

Impact on income mobility

The inequality and poverty comparisons above are anonymous. By (re-)ranking households from poorest to richest in each distribution, they do not consider the income trajectories of individual households. But those income trajectories are of considerable interest when income losses (or gains) differ, perhaps greatly, among households as they do here. To describe those trajectories, we use the nonanonymous growth incidence curves (GIC).²⁰

We estimate that households across the entire income distribution could be worse off on average (regardless of the scenario) after the COVID-19 impact, which is not surprising (figure 2). We find that the worst effects may not be on the poorest, but those (roughly) in the middle deciles of the ex-ante income distribution: the moderate poor and those vulnerable to falling below the poverty line if subject to an adverse shock. This result does not downplay the negative effect of the economic crisis on poor households. Even if poorer households lose less in relative terms, the impact on living standards could be devastating, especially for the extreme poor.²¹

Source: Authors’ calculations based on ENIGH (2018).

Figure 2 Non-anonymous growth incidence curves Panel (a) “Concentrated losses” 

The trajectories, however, could be quite different for rural in comparison to urban areas. The black lines in figure 2 show the GIC after the effect of the COVID-19 crisis for urban (solid line) and rural (dashed line) areas. First, except for the top, the fall in incomes is lower for rural areas throughout the distribution and for either scenario. The reason why households in the top (10 or 20 percent, depending on the scenario) fare better in urban areas is associated with the larger share of incomes from public employment and pensions, two sources assumed to remain unaffected. The larger participation of incomes from agricultural employment, consumption of own production, and cash transfers in rural areas explains why for the rest of the distribution rural areas are hit less. Second, within each area, the poorest urban households are actually among those who get hurt the most. In contrast, for rural areas the impact is U-shaped. The latter reflects that poorer households in rural areas have a larger share of their income coming from social assistance and consumption of own production (figure 1).

4. Conclusions

Our microsimulations show that the potential impact of the economic dislocation on inequality and poverty can be quite large. In addition, contrary to many people’s priors, the non-anonymous growth incidence curves show that income losses could be more pronounced for the moderate poor and those vulnerable to falling below the poverty line rather than among the poorest. This is so because social assistance programs and consumption of own production represent a larger share of total gross income for the poorest, especially in rural areas. In essence, these two items put a “floor” for the incomes of the poorest of the poor (the poor in rural areas). The economic dislocation is likely to cause a smaller increase in poverty for rural areas because it is here where consumption of own production -as a share of gross incomeis the largest. Because the indigenous population is concentrated in rural areas, the rise in poverty among the indigenous population is potentially lower than for the non-indigenous one. The expected rise in poverty is similar for male and female-headed households.

^{Lustig et al. (2020)} estimate the increase in poverty due to the COVID-19-induced economic dislocation in Argentina, Brazil, Colombia, and Mexico. Their estimates suggest that Brazil and Mexico could face the largest increase in the number of poor among these countries. However, they also find that the expanded social assistance governments have introduced in response to the crisis could have a large offsetting effect in Brazil and Argentina. In Colombia, the mitigation is quite modest. In Mexico, it is nil: the Federal government has provided no (at least, not as of the writing of this article) additional social assistance in the wake of the crisis.