Peer Groups and Employment Outcomes: Quasi-Experimental

In studies of employment outcomes, it is typical to relate outcomes to personal attributes, such as race and gender; less often do we consider the race and gender composition of the group in which the person works. Yet interactions between group members may have powerful effects on an individual's job performance. Although we might believe groups matter, it is difficult to credibly measure group effects because individuals usually self-select into groups or are assigned to them on the basis of unobservable characteristics. In fact, the dearth of evidence on this subject is likely explained by difficulties in isolating the impact of peer composition from the assignment mechanism. In this paper, we use confidential U.S. Army personnel records, and exploit the conditional random assignment of newly enlisted young adults by the Army to their first work units. Our dataset contains a panel of approximately 66,000 new enlistees during 2002, with 2 million person-month observations up to 2005. Our primary outcome variable is time to promotion and we observe the peer composition of each enlistee, providing us with a rare opportunity to estimate the effect of group composition on employment outcomes. Our estimates suggest that an increase in the percentage of female peers decreases the time to promotion of men relative to women, whereas an increase in the percentage of women among the leadership in a unit decreases the time to promotion of women relative to men. We find similar effects for race. An increase in own-race peers disproportionately increases time to promotion for blacks and Asians. Hispanics, on the other hand, experience reductions in promotion time when grouped with a higher fraction of Hispanics. All racial groups analyzed benefit from more own-race peers in leadership positions.