Why Do We Tenure? Analysis Of A Long Standing Risk-Based Explanation
University of Washington – Department of Finance and Business Economics
University of California, San Diego (UCSD) – Rady School of Management
Edward Dickersin Van Wesep
University of Colorado at Boulder – Department of Finance
May 18, 2016
Using a sample of all academics that pass through top 50 economics and finance departments between 1996 and 2014, we study whether the granting of tenure leads faculty to pursue riskier ideas. We use the extreme tails of ex-post citations as our measure of risk and find that both the number of publications and the portion that are “home runs” peak at tenure and fall steadily for a decade thereafter. Similar patterns holds for elite (top 10) institutions, for faculty with longer tenure cycles, and for promotion to Full Professorship. We find the opposite pattern among poorly-cited publications: their numbers steadily rise after tenure. The decline in both the quantity and quality of publications points to tenure incentivizing less effort in publishing rather than more risk-taking.
Why Do We Tenure? Analysis Of A Long Standing Risk-Based Explanation – Introduction
The granting of tenure has been enshrined since at least 1940 as a basic tenet of academic contracting,1 and is now pervasive among American universities: every one of U.S. News and World Report’s top 500 colleges in the United States has some kind of tenure-granting system. But is tenure part of an optimal contract? The costs of tenure should be clear: employment guarantees remove one of the strongest incentives for effort that workers face. But there are potentially many benefits: academic freedom allows faculty to research and teach what they consider important, without fear of firing. If this allows for more innovative research, as theory suggests that it might, then tenure may be optimal, even in the face of the loss of effort incentives.
The argument is simple: if society benefits most from “home run” ideas – e.g., Ackerlof (1970)’s market for lemons or Einstein (1905)’s theory of special relativity – then academic contracts should provide incentives for risk-taking, whether it takes the form of (i) arguing for ideas that run contrary to popular or academic opinion, (ii) writing about novel topics, which may die in obscurity or may be seminal, or (iii) simply working on a difficult problem for which effort and talent are no guarantee of success. Tenure protects researchers from dismissal that might accompany the ridicule or, more likely, the lack of citations and publications that accompany failed risks. As shown in Manso (2012), motivating innovation requires rewarding success, but also protecting researchers in the case of failure. In the absence of tenure, academics have an incentive to “play it safe”, which may be personally optimal, but will rarely lead to ground-breaking ideas.
To determine the effect of tenure on academic effort and risk-taking, we measure both the quantity and quality of academic publications in the years surrounding tenure. We hand collect a sample of all academics who pass through economics or finance departments at top 50 US schools between 1996 and 2014. From this sample of over 2000 faculty we consider two variables in the years before and after each academic receives tenure: the number of publications, and the number of “home run” publications, where a home run is a paper that is among the 10% most cited of all papers published that year.2 The number of publications is a measure of the quantity of output, and the number of home run publications is a measure of the quantity of highly influential output, which is presumably more likely to result from risky ventures.
We find that both variables have values that peak at tenure and decline thereafter. The average number of annual publications falls approximately 25% within five years of tenure, a statistically and economically significant drop of approximately a quarter of a publication per year. The average number of home runs falls by more than 30% per year over the same span. Further, the probability that any given publication is a home run falls post-tenure as well. This likelihood also peaks in the tenure year, and falls by two to five percentage points within five years of tenure. Given that only 20% of papers in the sample are home runs, this is a highly significant drop in quality.
Together, these numbers suggest that, if risk-taking does increase post-tenure, these risks are not successful in producing home run papers. Instead, the results are most consistent with academics increasing effort over time up to the tenure year, and reducing their efforts thereafter. Further, the reduction in effort is not driven by reducing the number of lower quality publications, and focusing upon higher quality work. Instead, it appears that academics can exert effort toward both quantity and quality, and reduce both post-tenure.
To support this (dismal) conclusion, we evaluate three alternative explanations for the publication and citation data. A first possibility is that productivity naturally increases post-PhD for some number of years before declining, and that the year tenure is granted just happens to be the natural peak in productivity. Perhaps the length of a typical tenure clock was chosen specifically to match peak academic productivity. This possibility seems unlikely, as people finish their PhDs at a variety of ages, and the peak age for productivity likely depends upon the discipline, but the tenure clock is set at the university level.4 Theoretical problems aside, we want to more precisely evaluate this possibility, and separate the sample into academics who received tenure five, six, seven, etc., years after finishing their PhDs. The number of observations in each group are much smaller than for the full sample, so there is considerably more noise in our estimates. Nonetheless, within each sub-group, productivity, measured both with the production of papers and the production of home run papers, peaks in the years surrounding the granting of tenure, and declines thereafter. Academics who receive tenure after five years see their productivity fall starting in year six, while those who receive tenure in 10 years see their productivity fall starting in year 11. These results are hard to reconcile with an explanation that does not relate specifically to tenure. Given that those receiving tenure only five years after getting a PhD are probably superior researchers, one would not expect their productivity to be falling at a time when the productivity of less talented researchers is rising.
See full PDF below.