Research

The employee hiring decision is a classic example of information asymmetry, yet little is known about heterogeneity among firms in their ability to screen for worker quality. Using German linked employer-employee data, I measure the quality of new employees using fixed effects estimated from an Abowd, Kramarz and Margolis (1999) (AKM) model during their later-career years. Screening ability correlates with survey-based measures of screening intensity, and varies widely across firms: the difference between a 25th percentile and 75th percentile firm is 0.6 worker effect standard deviations. To explore how initial screening affects market-level learning about workers, I also propose and validate a new interpretation of worker fixed effects as reflecting market beliefs over productivity. This framework offers some key advantages over variables used in the earlier employer learning literature such as wages, which do not take into account firm-specific components of wages, or cognitive test measures, which are not generally available in administrative datasets and capture a limited component of ability. I then use this framework to show that being hired by a better-screening firm accelerates market-level learning about worker quality, which leads to faster assortative matching with high-wage firms.

Standard models of rational job search assume agents know the distribution of offered wages when deciding which jobs to accept. We test if incorrect beliefs about wages affect real-world job search behavior in a field experiment with 1100 senior-year undergraduate students in the graduating Class of 2023 at the University of California, Berkeley. Partnering with the Career Center, we present personalized information graphics on school-and-major-specific salary distributions to students in the treatment group. We first document novel evidence that even prior to labor market entry, errors exist in wage beliefs – some students overestimate the available distribution, while others underestimate the available distribution. Post-treatment, we find that students treated with correct information update their beliefs towards the truth, and this is reflected in changes in reservation wages. At the end of the school year, we find that in comparison to the control group, students who increased their reservation wage after treatment had higher total and base salaries conditional on employment, a result significant at the 5% level. However, these same students had a lower, but imprecisely estimated likelihood of being employed by June post-graduation. An opposite but symmetric effect occurred for students who decreased their reservation wage. Our results are consistent with job search models where workers with more optimistic expectations wait longer to accept a job, but accept higher wages. We use our experimental estimates together with the model to simulate the effect of pay transparency laws on labor market outcomes, including wage dispersion. Our paper suggests an economically important role for subjective beliefs about labor market conditions and shows the effectiveness of a light-touch information intervention on employment and earnings for first-time job seekers.

Worker fixed effects as estimated through the influential decomposition of wages into firm and worker effects by Abowd, Kramarz and Margolis (1999) (AKM) are typically interpreted as heterogeneous, time-invariant wage premia reflecting worker ability. But the extent to which these premia are indeed stable over time, and therefore the credibility of this interpretation, is unclear. I examine the persistence of worker effects using a representative sample of the German labor market from 1985-2017 using fixed effects separately estimated across 5 windows. Worker effects are highly persistent, with an autocorrelation between estimates 6 years apart of 0.78, especially for middle-to-high fixed effect workers. I then show that wage ‘scarring’ from adverse economic conditions at labor market entry largely results from firm effect scars. I estimate that less than one-fifth of the initial wage penalty is driven by permanent impairment of worker-specific factors and four-fifths by starting at firms paying lower wage premia. The recovery in wages is driven almost entirely by the recovery in firm fixed effects, although worker effects remain slightly impaired throughout.

Selective high schools in the Australian state of New South Wales (NSW) provide an opportunity for students to attend a public school with significantly higher-achieving peers – the average successful applicant scores more than two standard deviations higher on baseline numeracy tests than the state average. Competition for entrance into these schools is fierce, with public opinion attributing the superlative academic success of selective school students at least in part to the selective school environment. However, success could simply reflect selection, and not an effect of the school or peer group. In this paper, I employ fuzzy regression discontinuity designs on 18 NSW selective schools with varying gradations of selectivity to estimate causal effects of selective school attendance on performance in high-stakes university entrance assessments and participation rates in advanced coursework. This is the first such study of selective schools in NSW, which is home to the oldest and most extensive selective school system in Australia, using a newly matched dataset encompassing the school careers of three state-wide cohorts of selective school applicants. I find that receiving an offer to attend a selective school has only scattered and mostly insignificant impacts on overall student achievement and participation in advanced coursework. I do find suggestive evidence that selective schools benefit low socioeconomic status students, but that such students are typically underrepresented in selective schools, which has implications for Gifted and Talented education policy.