Comments

Omar Arias:

This paper addresses an important but insufficiently studied aspect of the persistent high poverty and inequality in Latin America and the Caribbean. As the authors point out, understanding mobility is very important to identify the determinants of the persistent high levels of inequality in the region. However, the issue has not received adequate attention in terms of data generation, research, and the policy agendas.

The United States and other industrialized countries have a long history of studies in this area, although the main focus has been on intergenerational mobility, particularly the extent to which socioeconomic success is transmitted from parents to children. This paper instead focuses on intragenerational income mobility, that is, the changes in a person's income level during his or her adult life course. The authors concentrate on studies that track the evolution of the incomes of the same individuals over time to see who gains and who loses during the economic growth process. This is closely related to the recent literature on pro-poor growth and vulnerability. The paper provides a succinct summary of the current state of knowledge on intragenerational income mobility in the region (largely produced by the authors themselves), documents its importance, and highlights areas for further work. The work should thus be of great interest to development academics and policymakers alike.

The paper reviews the main conceptual and practical challenges in measuring intragenerational income mobility with existing data in the region. This includes the possible ambiguities in using alternative mobility measures and the difficulties that measurement error in (noisy) incomes and panel data attrition pose to empirical analysis. The authors suggest possible ways to address these issues, including the advantages and disadvantages of alternative methodologies such as instrumental variables and pseudo-panels.

The paper clearly highlights the limitations of panel household survey data in Latin America and the Caribbean and the critical need to expand their use and coverage. Nevertheless, some additional methodological issues are important for deriving knowledge and policy lessons from existing studies using the few short (one- to two-year) panels available in the region and for advancing research in this important field. First, while mobility measures are useful, much care is warranted in drawing inferences based on such short panels. The evidence from studies such as those surveyed in the paper can establish whether a certain growth pattern (or a crisis) affected individuals differently, particularly those that start out in the low ranks of the income distribution. They are less reliable, however, for inferring whether a given country shows high or low intragenerational income mobility even compared to other countries. One of the key advantages of mobility studies for Latin America is that they allow the analyst to discern how much of the rise in or persistency of the region's high inequality is due to lifetime (or structural) inequality and how much to the volatility in measured annual earnings that inflates cross-sectional inequality. Such calculations cannot be done reliably with existing short panel data. In the case of the United States, studies have reached different conclusions regarding the level and evolution of intragenerational mobility depending on the length of the longitudinal data used to track incomes over time.¹

A second and related issue, which is mentioned in passing in the paper given space constraints, is the modeling and interpretation of intragenerational mobility earnings regressions. This is key to gauging the determinants of intragenerational mobility and hence the policy knowledge that can be derived from empirical research such as the Latin American and Caribbean studies reviewed in the paper. The following earnings Mincer equation illustrates the issues:²

where Z and χ denote the matrices of observable individual characteristics, u and e are the vectors of unobservable attributes, and their respective coefficients are the corresponding income returns. The corresponding change in earnings for the individual over a given period is

This formulation indicates that the change in an individual's log earnings results from a growth rate in earnings common to all persons; from changes in observed (Z) and unobserved (ε) individual time-varying characteristics (such as the accumulation of schooling, experience, or...