Knowledge that Transforms

We examine the effects of an affirmative action policy at an elite Brazilian university that reserved 45% of admission slots for Black and low-income students. We find that marginally admitted students who enrolled through the affirmative action tracks experienced a 14% increase in early-career earnings. But the adoption of affirmative action also caused a large decrease in earnings for the university’s most highly ranked students. We present evidence that the negative spillover effects on highly ranked students’ earnings were driven by both a reduction in human capital accumulation and a decline in the value of networking.

The diffusion and use of new ideas is critical for producing innovations and realizing their potential. We explore how characteristics of innovators and potential adopters affect the adoption of important, new scientific ideas in networks. Using rich data on biomedical researchers and their networks, natural language processing, and a novel two-way fixed effects strategy, we find that new ideas introduced by female scientists are underutilized for two reasons. First, female innovators are less connected than men. Second, at short network distances, researchers (especially men) adopt women’s ideas less. Similar gaps hold for underrepresented racial and ethnic minorities.

We exploit the natural experiment provided by the Vietnam lottery draft to evaluate the intergenerational effects of fathers’ draft eligibility on children’s propensity to engage in risky behaviors during adolescence using the National Longitudinal Survey of Youth 1997. Draft eligibility increases measures of substance use, increases intensity of use, decreases age of initiation—particularly for marijuana—and increases measures of delinquency. We also show that draft eligibility led to changes in parenting that are conducive to these risky behaviors among children: it affects paternal parenting styles and attitudes toward the respondent, environmental aspects, and even maternal factors. Results are robust to alternative specifications and falsification diagnostics.

Many studies estimate the Frisch elasticity via two-stage least squares (2SLS) regression of hours changes on wage changes. However, 2SLS suffers from a power asymmetry problem that makes estimates far away from ordinary least squares appear spuriously imprecise. Hence, it is difficult for t-tests to detect a positive Frisch elasticity. We illustrate the problem using NLSY97 data. We estimate an elasticity of 0.60 for young men, but the t-test indicates that it is insignificant. The Anderson-Rubin (AR) test, which avoids the power asymmetry problem, indicates that it is highly significant. We argue that the AR test should be widely adopted in lieu of the t-test in instrumental variable applications.

There are 420 million young people in Africa today, and only one in three has a regular salaried job. We study how two common labor market interventions—vocational training and matching—affect the job search behavior of young workers. We do so by means of a field experiment tracking young job seekers for 6 years in Uganda’s main cities. Vocational training amplifies the job seekers’ initial optimism, leading them to search more intensively and toward high-quality firms. Adding matching has the opposite effect, plausibly because of low callback rates. These differences affect labor market outcomes in the long run.

To assess the child welfare impact of policies governing divorced parenting, such as child support orders, child custody assignments, and marital dissolution standards, one must consider their influence not only on the divorce rate but also on spouses’ fertility choices and child investments. We develop a model of fertility, parenting, and divorce, from which we derive estimates of parental preferences and a child cognitive ability production function, using data on parental time allocation, children’s cognitive attainment, and realized fertility and divorce. Family policies that reduce divorce are simulated to have significant negative impacts on both fertility and child development.

Using text data from curricula of the entire Dutch vocational education system, we extract verbs and nouns to measure basic cognitive and cross-functional skills, such as social, technical and resource management skills. This method allows us to uncover, at a fine-grained level, the skills middle-educated students learn at school. We then estimate the returns to these skills. Results show that resource management and social skills have the highest return. Importantly, the returns to skills are strongly conditional on sorting mechanisms. Our results fit in the literature that highlights the demand for complex interactive skills in environments that require more coordination.

Economists analyzing a well-conducted randomized controlled trial or natural experiment and finding a statistically significant effect conclude that the null of no effect is unlikely to be true. But how frequently is this conclusion warranted? The answer depends on the proportion of tested nulls that are true and the test's power. I model the distribution of t-statistics in leading economics journals. Using my preferred model, 65% of narrowly rejected null hypotheses and 41% of all rejected null hypotheses with |t|<10 are likely to be false rejections. For the null to have only a .05 probability of being true requires a t of 5.48.

We examine the relationship between income volatility and life expectancy in mid-sized U.S. commuting zones between 2006 and 2014. We use a commercial dataset, InfoUSA, to measure income volatility which we link to estimates of life expectancy by gender, county, race, and income. We find that higher income volatility in a county is associated with lower life expectancy, but only at the bottom of the income distribution and primarily for non-Hispanic Whites. Though we cannot extrapolate our findings to individual-level relationships, we do link them to existing literatures on place-based differences in mortality and the relationship between volatility and health.