Pay Inequality in Academic Departments Doesn't Just Redistribute Money — It Reshapes What Kind of Science Gets Done

105K faculty across 93 universities over 22 years: higher unexplained pay gaps correlate with more papers and citations but less disruptive, less novel research.

Imagine a karaoke contest where the judges score on pitch accuracy. Contestants who belt out crowd-pleasing standards — songs everyone can evaluate — will always outscore the person improvising something weird and new. The judges aren't biased; the scoring system simply cannot see originality in real time. That's the core mechanism this paper identifies inside academic departments: when salary gaps widen beyond what rank, seniority, and contract type explain, the resulting tournament selects for legible, incremental science and penalizes the disruptive kind. The committed claim: pay dispersion in U.S. academic departments is not merely a distributional outcome but an evaluative structure that channels intellectual effort toward conventional, well-recognized lines of inquiry and away from paradigm-shifting work. This is not the first paper to study pay dispersion's effect on output quantity; it is, however, the first to decompose the direction of scientific work along novelty and disruption dimensions using large-scale administrative payroll data linked to bibliometric records. The dataset is genuinely impressive in scope. Cao assembles administrative salary records for 105,050 tenure-line faculty in 4,839 departments at 93 U.S. public universities spanning 2004–2026, linked to publication histories via OpenAlex. For each department, a local pay equation is estimated using rank, contract basis, and seniority — the observable "ladder." The key independent variable is the Gini coefficient of the residual salary variation that this ladder cannot explain. This unexplained dispersion is the tournament signal: the rewards that flow to some faculty but not others for reasons the formal hierarchy doesn't capture. The results cut in a specific and uncomfortable direction. Departments with higher unexplained pay dispersion see more papers and more citations — the tournament works on quantity. But those same departments produce work that scores lower on disruption (the CD index family), lower on novelty, and higher on conventionality. The compensation premiums above the ladder flow disproportionately to scholars with broad collaborations and topical breadth — precisely the profile that maximizes legible, evaluable output rather than deep, frame-breaking inquiry. Tournament theory isn't wrong about effort; it's blind to the dimension of effort it selects on. Methodologically, Cao is admirably honest about limitations. The design is observational: department-level pay dispersion is not randomly assigned, and the paper cannot fully separate treatment effects (the tournament changes behavior) from selection effects (departments that reward conventional work attract conventional scholars). Fixed effects at the department and individual level help, but the causal arrow remains contestable. The disruption and novelty measures (CD index variants, combinatorial novelty scores) are themselves debated in scientometrics — they're the best available, not ground truth. The intellectual contribution is the reframing. Tournament theory and social comparison theory have argued for decades about whether inequality helps or hurts performance, treating "performance" as a single axis. Cao's move is to point out that scientific work has multiple axes — some legible to contemporaries, some not — and that tournaments systematically reward the legible axis. This is a structural argument about how evaluative regimes shape knowledge production, and it lands squarely in the organizational sociology tradition of work by Merton, Bourdieu, and more recently Uzzi and colleagues on novelty penalties in science. The practical upshot is sobering. If pay inequality in academic departments functions as an evaluative filter — rewarding breadth, collaboration networks, and citation-friendly incrementalism while penalizing deep, disruptive, slow-payoff work — then the current drift toward wider salary dispersion in U.S. universities is not neutral with respect to the kind of knowledge the system produces. Twenty years of this trajectory means more papers, more citations, and a slower rate of paradigm shifts. The assembly line runs faster; the blueprints stay the same.