Research

Understanding how workers value various job attributes is of fundamental policy importance. The sensitivity of labor supply decisions to wages has long been a natural focus in economics. However, understanding preferences towards non-wage amenities is also crucial since non-wage amenities substantially affect the total value of a job. This paper provides new evidence about worker preferences towards wages and the type of work performed on jobs (i.e., tasks), addressing three key empirical challenges that may be present in revealed preference approaches: limited measurement of non-wage amenities, a lack of information about rejected offers, and a potential selection issue arising through an endogenous search decision. We address these three difficulties by combining unique data access with the specification and estimation of a novel choice model. Detailed information about job tasks/wages, rejected offers, search behavior, and job outcomes comes from the Berea Panel Study, which closely followed a group of college graduates during the early post-college portion of their careers. Our model addresses selection by explicitly accounting for the joint nature of the search and job choice decisions.

Results show that wages strongly influence job transitions, but that workers also exhibit clear preferences for tasks: high-skilled tasks are significantly more valued than low-skilled tasks, with some evidence of preference ordering over people, information, and objects tasks. Importantly, we show that analyses lacking the three key data features that we exploit – detailed job-level task information, rejected offers, and search effort – produce very different estimates of both wage and task preferences. We emphasize the empirical relevance of non-random selection into the sub-sample of workers who receive new job offers, operating through search decisions that depend on the unobserved attributes of current jobs. Finally, we use our results to examine differences between wage inequality among workers and a broader measure of inequality that considers the non-pecuniary benefits of job tasks.

Unique data from the Berea Panel Study provides new evidence about fertility outcomes before age 30 and beliefs about these outcomes elicited soon after college graduation. Comparing outcomes and beliefs yields a measure of belief accuracy. Individuals who are unmarried and not in relationships at age 24 are extremely optimistic about the probability of having children, while married individuals have very accurate beliefs. Novel attractiveness measures are central for understanding fertility beliefs and outcomes for females but not for males. Marriage is a mechanism that is relevant for understanding differences in beliefs, outcomes, and misperceptions across relationship and attractiveness groups.

Using data from the 1970s, Hammermesh and Biddle (1994) find that unattractive workers suffer substantial wage penalties, with this unattractiveness penalty seemingly larger for males than females. Using unique data from the Berea Panel Study, we find that large unattractiveness penalties in wages remain for a sample of more recent college graduates, but find no evidence of gender differences in this penalty. Extending the analysis, we find large unattractiveness penalties are also present for a variety of non-wage outcomes, including a measure of overall well-being. Further investigation reveals that the unattractiveness penalty in overall well-being can be largely explained by unattractiveness penalties in outcomes such as earnings, health, marriage, and children. We also provide evidence about college students’ perceptions of these unattractiveness penalties. Our results show that unattractive individuals face an extensive set of disadvantages.

Many economic decisions depend on individuals' beliefs about uncertain future life events, such as marriage or childbearing. A common approach is to assume Rational Expectations (RE) and infer these beliefs from outcomes, given a specific information set. However, recent evidence suggests that this assumption does not hold in many contexts. This paper introduces a decomposition method to assess why RE-implied beliefs deviate from stated beliefs. I show that the differences between these two sets of beliefs can be separated into three components, two of which are influenced by the choice of information set. Importantly, I demonstrate that adding more variables to the information set does not necessarily reduce this difference. I apply this framework to fertility expectations among young adults, comparing stated beliefs about childbearing with predictions based on observable factors such as marital status and demographics. The results highlight how the choice of information set can affect the accuracy of RE-implied beliefs and offer guidance for applied work when belief data are unavailable.