Neil Duzett

The debate surrounding demographic compositions in organizations has intensified in recent years, spurred by claims that different compositions can group decision-making and performance. Yet despite significant attention and investment, evidence on the effects of group composition on group outcomes is limited. This paper provides new evidence in the U.S. city council setting nationwide. Leveraging quasi-random victories in close elections between different-race candidates, I use a regression discontinuity design to estimate the impact of racial compositions on city council communication and effectiveness. Applying natural language topic modeling and sentiment analysis techniques to 17,045 city council meeting transcripts, I show that changes in composition enhance information exchange, but also heighten communication frictions – increasing negativity in council meetings by 66%. In line with theories on effects in groups of different sizes, larger councils see greater shifts in discussion topics and only modest changes in negativity, while smaller councils see an increase in negativity five times greater than large councils. These competing communication effects and heterogeneity by council size are reflected in compositional effects on incumbent vote shares, a measure of council effectiveness. Changing composition improves future vote shares by 6.7 points in large councils, with little to no effect for smaller councils. Together, these results suggest that composition's impact on communication plays a key role in determining the effectiveness of groups.

We examine the transmission of lifespan across generations in the US using a unique dataset containing more than 26 million individuals born between 1880 and 1920. On average, 47 percent of men and 57 percent of women lived longer than their parents, though this varied across cohorts and socio-economic groups. The intergenerational persistence in lifespan is low across cohorts and socio-economic groups, and it is much smaller than persistence in socio-economic status. Moreover, persistence in lifespan and in socio-economic status are independent of each other. Lifetime well-being, which combines socioeconomic and lifespan measures, is less persistent than socio-economic measures suggest.

Using national data on local government elections in the U.S., I show that in close contests, candidates who do not match the majority gender of current officeholders are systematically disadvantaged. Near the victory cutoff, these candidates are 1.9 times as likely to lose as win. This gender penalty affects both women and men and is present in elections for city councils, school boards, and county legislatures. Evidence suggests the effect may be mediated by current officeholders steering campaign donations toward candidates who match their own gender. These findings highlight how those in power can shape the compositions of future leadership.