Abstract:  Two-year community colleges enroll nearly half of all first-time undergraduates, but to ambiguous effect: low persistence rates and the potential for diverting students from 4-year institutions cast ambiguity over 2-year colleges’ contributions to upward mobility. Developing a new multivariate instrumental variables approach applied to linked administrative education and earnings records, I disentangle the net impacts of 2-year college access into two competing causal margins: significant value-added for 2-year entrants who otherwise would not have attended college, but negative impacts on students diverted from 4-year entry. 2-year access particularly boosts the upward mobility of disadvantaged students, who experience less 4-year diversion.  

• "On the Use of Outcome Tests for Detecting Bias in Decision Making"  (comment and reply) 

with Ivan Canay and Magne Mogstad  (R&R, Review of Economic Studies)

  

Abstract:  The decisions of judges, lenders, journal editors, and other gatekeepers often lead to disparities in outcomes across affected groups. An important question is whether, and to what extent, these group-level disparities are driven by relevant differences in underlying individual characteristics, or by biased decision makers. Becker (1957) proposed an outcome test for bias leading to a large body of related empirical work, with recent innovations in settings where decision makers are exogenously assigned to cases and vary progressively in their decision tendencies. We carefully examine what can be learned about bias in decision making in such settings. Our results call into question recent conclusions about racial bias among bail judges, and, more broadly, yield four lessons for researchers considering the use of outcome tests of bias. First, the so-called generalized Roy model, which is a workhorse of applied economics, does not deliver a logically valid outcome test without further restrictions, since it does not require an unbiased decision maker to equalize marginal outcomes across groups. Second, the more restrictive "extended" Roy model, which isolates potential outcomes as the sole admissible source of analyst-unobserved variation driving decisions, delivers both a logically valid and econometrically viable outcome test. Third, this extended Roy model places strong restrictions on behavior and the data generating process, so detailed institutional knowledge is essential for justifying such restrictions. Finally, because the extended Roy model imposes restrictions beyond those required to identify marginal outcomes across groups, it has testable implications that may help assess its suitability across empirical settings.

with Marianne Bertrand and Magne Mogstad  (Accepted, Journal of Labor Economics)

Abstract:  We study the impacts of a major reform to vocational secondary education that aimed to move beyond the tradeoff between providing occupational skills and closing off academic opportunities. Norway’s Reform 94 integrated more general education into the vocational track, offered vocational students a pathway to college, and increased access to apprenticeships. We identify reform impacts through a difference-in-discontinuity research design applied to linked population registries. The reform substantially increased initial vocational enrollment, but with divergent consequences by gender. Overall, the reform succeeded at improving social mobility, particularly for disadvantaged men, but it somewhat exacerbated the gender gap in adult earnings.

with Brent Hickman  (new draft in progress)

Abstract:  Students who attend different colleges in the U.S. end up with vastly different economic outcomes. We estimate relative value-added of individual colleges to disentangle causal college impacts from selection bias in producing these large outcome disparities. Linking administrative high school records, college applications, admissions decisions, enrollment spells, degree completions, and quarterly earnings spanning the Texas population, we identify relative college value-added by comparing the outcomes of students who apply to and are admitted by the same set of institutions, as this approach strikingly balances observable student potential across college treatments and renders our extensive set of covariates irrelevant as controls. Methodologically, we develop a framework for interpreting value-added estimates under various restrictions on treatment effect heterogeneity, generalizing the constant-effects approach typically employed in the value-added literature. Empirically, we estimate a relatively tight, though non-degenerate, distribution of value-added across the wide diversity of Texas public universities. Selectivity is a poor predictor of value-added in our sample: attending a more selective public university over a less selective one yields only a fleeting earnings premium that fades to a precise zero after a few years in the labor market, and it reduces the likelihood of completing a STEM degree. Non-peer college inputs like instructional spending more strongly predict value-added, especially conditional on selectivity. Educational impacts predict labor market impacts: colleges with larger earnings value-added also tend to be colleges that boost persistence, BA completion, and STEM degrees along the way. Finally, we probe the potential for (mis)match effects by allowing each college's value-added to vary flexibly by student characteristics, including race, gender, family income, and pre-college measures of cognitive and non-cognitive skills. At first glance, Black students appear to face small negative returns to attending more selective colleges, but this pattern of modest "mismatch" is entirely driven by the availability of two large historically Black universities with low selectivity and above-average value-added. Across the non-HBCUs, Black students face similar returns to selectivity, and indistinguishable value-added schedules more generally, compared to their peers from other backgrounds.