Doctoral Dissertation Research in Economics: Investigating the Impact of the 'Norm to Work' on Worker Power and Labor Market Outcomes

This project investigates the labor market implications of social norms regarding employment. Media narratives and existing research suggest that such norms are important drivers of worker behavior and also, potentially, macroeconomic outcomes like employment and wages. Recent coverage of the “great resignation” post-Covid has echoed this idea, claiming that a collective re-evaluation of the importance of work has shifted power toward workers. In particular, this project explores the idea that the “norm to work”—the social expectation that working-age adults should be employed—may erode worker power by making workers fearful of unemployment, and thus willing to accept lower-quality jobs or worse wages. To empirically test this hypothesis, the researchers develop a novel measure of work norms based on social media text data. From a policy perspective, this project sheds light on the delicate balance between promoting a healthy work ethic and stigmatizing the unemployed. To the extent that work norms do in fact diminish worker power, there are potential second-order effects of political and media narratives that stigmatize unemployment (e.g., President Reagan’s widely discussed “welfare queen” rhetoric). Similarly, policies that tacitly enshrine work norms, such as “workfare” programs or “right to work” laws, may have unintended consequences for the broader employed populace. Moreover, this project provides a methodological roadmap for constructing measures of other social norms using the vast trove of unstructured text data now available online. The researchers study the effect of the norm to work on worker power and labor market outcomes by first constructing a novel measure of the norm to work using machine learning methods applied to unstructured text data. They then empirically analyze the relationship between this measure and various outcomes, including labor force participation, employment, wages, and labor’s share of income, using economic data from Census surveys and the Bureau of Economic Analysis. The primary measure of the norm to work is based on a sample of about 20 million tweets found through keyword searches related to unemployment and attitudes toward work. Initially, a dataset of roughly 100,000 labeled tweets is obtained, which serves as a training and validation dataset to calibrate a machine learning model. This trained model is subsequently used to label the remaining tweets in the sample, which are aggregated into a geographically granular measure of the norm to work in the U.S. over the past decade. Using this metric, the researchers document the relationship between the norm to work and labor market outcomes at the city-year level, controlling for temporal and geographic characteristics. In order to better understand the direction of causality in this relationship, the researchers also conduct analyses incorporating more advanced panel data econometrics. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.