Working Papers


Race and Declining Labor Force Participation

Abstract: I investigate the causes of declining labor force participation in the U.S. I find that the changing age structure of the population fully explains the decline in white labor force participation. African-Americans, who were more likely to participate in the labor force in 1970, have grown less likely, even when controlling for age. Using a logistic regression with reasonable controls, I find that African-Americans now have a 5 percentage point lower labor force participation rate than whites. I propose various explanations for this, including health, immigration, labor demand and discrimination, but I find incarceration to be the most compelling.

A graph:




Retrospective Voting: A Vector Autoregressive Approach

Abstract: I run a vector autoregression on daily stock market data and presidential approval data to estimate the responsiveness of voters to economic data. As opposed to quarterly data, the daily time series allows has more power and allows for better estimation of the duration of the effect. I find that stock market movements used to Granger-cause rises in presidential approval, but the relationship is no longer holds. The decline over time is statistically significant, and I date it to 1980. I attribute this to an increase in ideological voting and a decline in retrospective voting. I also use a Bartik instrument to try to estimate this effect, but find that it is underpowered.

A graph:




Gender and Economists on Twitter

Abstract: I collect descriptive statistics of economists' behavior on Twitter. I find that economists interact more with others of their own gender than with the opposite gender. This holds for a variety of modes of interaction on Twitter. I also describe the differences in interactions by gender.

A graph:




Gerrymandering: A Sufficient Statistic

Conventional measures of gerrymandering are not useful descriptions of its impact across all vote shares. I show how the efficiency gap breaks down as a measure of gerrymandering when each party's vote share is not 50%. I show the usefulness of plotting a modeled relationship between vote share and seats won, and examining the partisan symmetry of the distribution. Then, I offer two statistics. The first is the vote share required for a party to win 50% of seats, which exists elsewhere. The second fully describes the legislative impact of the distribution of seats by vote share, using the marginal impact on the ideology of elected officials. This is the only reasonable way to deal with measuring gerrymandering in the context of large partisan victories.

A graph: