Working papers

The selection of beneficiaries in public policy is a central issue in economics, as it involves
the efficient allocation of scarce resources. This paper analyzes key targeting strategies using PROBEMS, a cash transfer program for high school students in Mexico. We examine
three dimensions of beneficiary selection. First, we compare two allocation mechanisms: a
centralized system based on a proxy-means test and a decentralized school-based ranking
process. We find that school rankings prioritize academic merit over poverty, with a moderate risk of resource capture. Second, we examine how misreporting affects the centralized
system. It not only distorts poverty indicators but also changes the weight assigned to specific assets in the proxy-means algorithm. Third, we apply machine learning methods to
assess how targeting would differ if the program prioritized outcomes rather than need, as
is typical with proxy-means tests. Relative to the centralized proxy-means test, we estimate
that recipient assignments would differ by 40% under school-based rankings, 49% using
corrected data, and 39% with outcome-based machine learning targeting. These results
underscore the tradeoffs between accuracy, fairness, and manipulation risk when targeting
social programs.

We study distributional and educational impacts of shifting from means-tested to semi-universal
antipoverty programs, leveraging a natural experiment in Mexico. In 2019, the Mexican program
PROGRESA was replaced by the semi-universal Benito Juarez Scholarships (BBJ), resulting in an
eligibility base expansion at the cost of transfer amounts. We give evidence that this resource
reallocation was regressive, but that the marginal households that benefited from the new targeting
strategy might display educational benefits. Using a difference-in-differences and a regression
discontinuity design, we employ a variation in the BBJ targeting design to compare a semi-universal
to a proxy-means strategy. Employing administrative and survey information, we focus on the basic
education level since BBJ employs both targeting strategies at this level. Our DiD findings suggest
that the semi-universal targeting reduces dropout rates by 0.76 percentage points at the elementary
level, while having non-significant reductions in middle school. Our RDD estimates do not display
significant changes.

This paper studies how a conditional cash transfer program with an embedded healthcare
component might induce relevant tradeoffs. By improving a previously employed identification
that exploits a sharp discontinuity in progresa’s initial locality-level rollout, we estimate a
sizable 12% increase in outpatient visits at public clinics, mainly driven by children and adults
aged 20-49. This translates into improvements in reproductive healthcare and screenings for
chronic diseases. While the attendance of non-targeted elderly members is not affected, the
program does seem to impose costs by increasing congestion and reducing self-reported quality.
This suggests that the benefits of this policy lever may carry negative unintended effects to the
non-beneficiary population.

This paper studies the consequences of a human error during the school choice mechanism of
Mexico City public high school system. Using the reported rank-ordered lists of the applicants,
we estimate their indirect utilities and generate cardinal measures of the error’s impact on the
applicants’ utilities. In particular, we found that the median welfare change is negatively correlated
with the educations of the applicants’ mothers (proxy for socioeconomic level).

The long-term effects of Mexico’s conditional cash transfer program, PROSPERA, on poor households are of great interest to both policymakers and academics. This paper analyzes the long-term effects on the welfare of the original participating households and their descendants, approximately 20 years after the program’s inception. To complement other studies that analyze effects on education and health, this analysis focuses on a utilitarian definition of welfare and employs two empirical strategies. The first uses the 1997–2000 experiment as the cleanest source of variation, although it is limited. The analysis shows that, by 2017–2018, the descendants of the original beneficiary families are more likely to form their own households, to migrate to different localities, and to have more durable goods and higher consumption expenditure than their control counterparts. The second strategy confirms and expands these findings using a difference-in-differences methodology that exploits the program’s expansion across localities and the age of individuals as sources of variation. This second approach covers a much broader and more representative sample and captures information on self-reported vulnerability regarding food consumption. The results confirm the overall positive outlook in terms of durable goods and reduced food vulnerability. Perhaps most interesting and relevant for evaluating the program’s success is that it improved intergenerational mobility. Using the 1997–2000 experiment, the analysis shows that young adults who benefited improved in terms of education, asset ownership, and income relative to their parents. It appears that the descendants of the original beneficiaries are improving their relative position in terms of assets and income.

This paper empirically examines the educational selectivity of United States immigrants and of those that return to their source country. Data from the 1970 to 2000 U.S. Census and the 2010 American Community Survey are employed. Ten countries are selected for the study based on their historical and contemporaneous importance on U.S. migration. The results generally indicate positive selection on educational attainment of recently-arrived immigrants, being China, India, and Philippines the most prominent examples. Mexico does not show evidence of positive or negative selection, but their immigrants’ selectivity has worsened through time. Historically, the educational selectivity of returning migrants accentuated the positive selection of those migrants that stay in the United States in most countries’ cases. However, patterns of selection among migrants that stay have recently changed. A more detailed analysis with data from the last decade finds evidence of positive selection of immigrants staying in the U.S. for the Mexican and Philippines’ case, as well as negative selection for the Chinese. Trends of returning migration are also analyzed by gender, age, naturalization status, and migration spell duration. Mixed evidence of selection trends is found.