Higher inequality in Latin America: a collateral effect of the pandemic

ABSTRACT

This study explores the evolution of inequality in Latin America during the COVID-19 pandemic by using primary data from household and employment surveys collected in 2020. First, we discuss the trends in inequality in the region from 1992 to 2020. Next, we estimate regression models to examine how the changes in demographics and education levels might be correlated with changes in income distribution. Finally, we use a panel regression model with fixed effects for 16 countries in the region to identify how the socioeconomic context might help explain the changes in income inequality. The empirical findings suggest that inequality increased by a statistically significant 2% between 2019 and 2020. We obtained significantly heterogeneous results when disaggregating by gender, urban/rural location, and sector of economic activity. Remittances had a modest effect, while government transfers helped to prevent more significant disparities in half the countries studied. Our estimations show that the decline in employment levels – due to the economic contraction caused by COVID-19— is associated with increases in income inequality that might gradually diminish with the recovery.

KEYWORDS:

• Inequality
• COVID-19
• Latin America
• gaps
• remittances
• transfers

JEL CLASSIFICATION:

• D63
• I14
• I32
• I38
• O15

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/02692171.2023.2200993.

Notes

1. The topic is relevant within each country but also at a global level because countries with less capacity to respond to the pandemic can suffer greater economic impacts. Deaton (2021) and Ferreira et al. (2021) analyse this aspect.

2. According to data from Latinobarómetro, between 2017 and 2018 support for democracy declined in 14 countries in Latin America, and 71% of Latin Americans stated that they are dissatisfied with democracy in their country. Only 32% approved of the performance of their current government. According to the same source, in 2018 only 12% of the population of the region considered their economic situation to be ‘good’ or ‘very good’. Furthermore, the share of the population stating that income distribution is ‘unjust’ or ‘very unjust’ increased from 69% to 80% between 2013 and 2018.

3. For Panama, the data from 2000 to 2019 were obtained from a previous standardisation of data collected by the SEDLAC (Socio-Economic Database for Latin America and Caribbean) by the LAC Equity Lab of the World Bank, while data for previous years were based on household surveys. In the case of Peru, data from 2000 to 2003 are from the SEDLAC by the LAC Equity Lab.

4. Although the objective of the standardisation is to allow for comparability between countries and years, there are cases of under-reporting, under-representation of regions or socioeconomic groups, or other particularities of the survey.

5. The averages reported in this section are unweighted.

6. The study by Londoño and Székely (2000) presents a series that starts in 1980 for a smaller set of countries that shows that the increase in the 1990s is a continuation of the increase in inequality observed between 1980 and 1990.

7. The difference between the average Gini for 1999–2000 and the average Gini for 2019 is statistically significant (we calculated standard errors for the Gini for each year and compare the point estimates considering a 95% confidence interval).

8. The increase in these two countries is presumably due to increases in inequality in the intermediate deciles, given that the changes in deciles 1 and 10 on the extremes suggest that the dispersion of income declined.

9. These results are in line with a recent study by Furceri et al. (2022), who analyse the effect that pandemics had on income distribution over the past two decades. Their results suggest that past pandemics – although they were on a smaller scale than COVID-19—are associated with increases in inequality due to an increase in the concentration of income in the wealthiest deciles. In addition, the authors carry out simulations to approximate the magnitude of the current health crisis. They find that the impact on the distribution of income varies substantially between countries and depends largely on the initial distribution of income, the set of public mitigation policies, and the particular characteristics of each country.

10. The categorical education variables are created for the population of 18 years and older, and are classified as no school, primary, upper secondary, higher secondary, and tertiary education depending on each education system. The no school value is the baseline category. For the average age, the categorical variables include age ranges of 18–29 (baseline category), 30–44, and 45–59 years old.

11. We focus on per capita income of the household as the traditional indicator to measure income inequality, which is why we also include a weight for household size. Rodríguez-Castelán et al. (2016) follow a similar strategy to analyse the dynamic of inequality in Latin America during the first decade of this century.

12. For Chile, the dependent variable is the logarithm of weekly income because the 2020 survey did not ask about weekly hours but did ask about labour income.

13. The data for the base year for comparison with 2020 are for 2019 for all countries except for Chile (2017) and Mexico (2018).

14. The results in this second group of countries can seem counterintuitive because it would be expected that the premium with respect to the primary sector would diminish given that the secondary and tertiary sectors were the most affected. However, because the specification includes only persons who are wage earners, it is not possible to identify what the labour market dynamic was for each sector. For example, persons leaving the labour market could have been those whose salaries were greater in the primary sector or whose salaries were smaller in the secondary and tertiary sectors.

15. Acevedo et al. (2021) study how the pandemic affected the labour markets of Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, Mexico, Paraguay, and Peru. They find that between the first and third quarter of 2020, the working-age population that left the labour market in the region increased by 6% points and unemployment increased by 1% point, while informality declined by 5% points. That is, contrary to past crises where informality acted as a shock absorber, during the COVID-19 pandemic, informality as a share of the labour force participation declined, as a large number of informal (rather than formal) workers exited the labour market and became inactive. The dynamics of the formal sector might be explained by several factors, including (i) labour regulations that impose firing costs; (ii) uncertainty about the duration and depth of the health crisis that complicated estimating the costs and benefits of maintaining jobs; (iii) employment support measures introduced by different governments; and (iv) the preference of some employers to reduce the hours of activity per job instead of reducing the production plan, and thus avoiding firing costs. The dynamics of the informal sector might be explained by several factors, including the following: (i) Informality is specifically characterised by the lack of health insurance, so attention to contagion risks may be weaker, leading to longer recovery times (and inactivity) or less effective care due to the saturation of public services; (ii) Even in cases where government authorities have implemented economic support mechanisms to cushion the drop in employment, the informal population, which is outside tax and other public registries, is more difficult to identify and locate, and therefore unlikely to benefit from active policies.

16. This is a simple estimate that does not take into account collateral effects or behavioural changes that might occur if households experienced changes in their sources of income as simulated.

17. For El Salvador, Guatemala, Honduras, Jamaica, and Mexico, 95% of remittances come from migrants living in the United States (Ratha et al. 2021).

18. The independent variables are from the World Bank’s World Development Indicators, Barro and Lee (2013) and, for education, Lee and Lee (2018). The data are fitted using a linear extrapolation.

19. The dependency ratio could generate a temporary increase in inequality because higher-income households experience reductions in fertility and dependency ratios before poorer households.

20. If we believe that the omitted variables are uncorrelated with the explanatory variables that are in the model, then a random effects model is probably best because it will produce unbiased estimates and the smallest standard errors. However, if we believe that the omitted variables are time-invariant, with time-invariant effects, and correlated with the variables in the model, the fixed effects model would be most appropriate. See Wooldridge (2010) for a more thorough discussion.

Additional information

Funding

The work was supported by the Inter-American Development Bank