Recent research has suggested that the long-observed negative association between seniority and pay among college faculty largely reflects below-average research productivity of senior faculty-a possibility that most earlier studies did not examine.Overlooked in both waves of studies, however, is match quality. Because the higher quality of the faculty/institutional match implied by higher seniority should, all else equal, result in higher salaries, failure to account for match quality inflates the estimated returns to seniority. Indeed, that positive bias, the authors find, is roughly equal in magnitude to the negative bias caused by failure to account for research quantity and quality. When they account for both match quality and faculty research productivity in an analysis of data on economics faculty at five research universities over a 21-year period, the authors estimate that, holding experience and other factors constant, the penalty for twenty years of seniority is 16% of salary. aculty pay structures have long been of interest to economists.1 Consistentwith findings for nonacademic markets, faculty pay has been found to rise with experience and to be positively related to measures of productivity.
Unlike in other markets, how-'Early studies of academic pay include Cohn (1973), Siegfried and White (1973), Ferber (1974), andJohnson and Stafford (1974).ever, there is evidence that returns to seniority are commonly negative for faculty. One proposed explanation for this anomalous finding emphasizes the monopsony power of most colleges and universities; another singles out the effects of "raiding," in which high-quality faculty are bid away, resulting in a negative correlation between seniority and unmeasured faculty productivity.Recently, Moore, Newman, and Turnbull (1998) argued that estimated returns to seniority in previous studies are biased downward because of the failure of these studies to control for quality of faculty research. Consistent with this claim, they showed that the coefficient of seniority became less negative once controls for research quality were included in the empirical model. Indeed, when the coefficient of seniority lost statistical significance, they concluded that the puzzle had been resolved: there is no need to explain negative