In the age of Levels.fyi, Glassdoor and leaked compensation spreadsheets, a company can no longer assume its people are ignorant of what their colleagues earn. Return-to-office fights and rolling layoffs have made employees more attentive to fairness, not less, and every headcount cut is read partly as a statement about who was worth what. The instinct of many leaders is that sunlight is neutral, that showing the numbers will reassure the well paid as much as it unsettles the underpaid. The evidence says the ledger does not balance. A pay reveal is not a symmetric event, and treating it as one is the most common mistake in open-pay design. The costs and the benefits fall on different people in different sizes, and once you see the shape of that imbalance you stop asking whether to be transparent and start asking how.
The asymmetry, measured
The cleanest test comes from Card, Mas, Moretti and Saez (2012). Working with public salary data at the University of California, the researchers told a random subset of employees about a website listing colleagues' pay, which created genuine variation in who actually looked. Workers who discovered they were paid below the median for their unit reported lower job satisfaction and a stronger intention to look for another job. Workers who learned they were above the median showed no matching gain. The upside was flat, the downside was real. Comparison, in this setting, behaves like loss aversion: the number you did not want to see does more work than the number you were pleased to find. That single fact should reframe how a leader thinks about disclosure, because it means the average reaction to transparency hides a distribution that is skewed toward harm. A survey of average sentiment might report that pay information barely moved the workforce. Underneath that flat average sit two very different groups, one bruised and looking for the exit, the other indifferent, and it is the bruised group that determines whether the policy was worth running. Averages flatter transparency. Distributions indict it.
When people cannot see why
Card and his colleagues measured feelings. Breza, Kaur and Shamdasani (2018) measured output. In a month-long field experiment with manufacturing workers in India, they randomized whether teammates doing similar work were paid the same daily wage or different ones. When the productivity gap between coworkers was not observable, pay inequality cut output by roughly 0.45 of a standard deviation and attendance by about 18 percentage points, and it eroded cooperation even where cooperating served the workers' own interest. The decisive result is the exception. When workers could see that the better-paid colleague was in fact more productive, the same pay gap produced no adverse effect. Justifiability, not equality, was doing the work. Inequality that people can explain to themselves is tolerable. Inequality that looks arbitrary corrodes both effort and the willingness to help a neighbor.
The one place the upside shows up
Transparency is not uniformly corrosive, and the counterexample sharpens the point. Cullen and Perez-Truglia (2022) ran information experiments with 2,060 employees at a commercial bank in Southeast Asia. Employees badly misjudged their bosses' pay, underestimating a manager's salary by 14.1 percent on average. When they learned the boss earned more than they had assumed, they worked harder: a 10 percent rise in the perceived manager salary raised hours worked over the following 90 days by about 1.5 percent. Looking up motivates, because a boss's salary reads as a signal of what promotion is worth. Looking sideways demotivates, because a peer's salary reads as a verdict on the present. The direction of the gaze changes the sign of the effect, which is why lumping all pay information into one policy is a category error. The bank result also carries a quieter lesson: people are working from bad estimates in the first place, off by double digits about their own boss, so the choice is rarely between information and no information. It is between the guesses employees already carry and a corrected picture you can shape.
The practical read
Put together, these findings argue against the naive version of open pay and against blanket secrecy in equal measure. If you expose horizontal comparisons without explaining them, you will pay in the morale of everyone who lands below the median, and the people above it will not repay you. The move is to make differences legible before you make them visible. Publish the bands, the criteria and the evidence that a higher number reflects a larger contribution, so that a gap arrives already justified rather than as a raw insult. Show people the ladder they can climb more readily than the neighbor they cannot unsee. Sunlight is not neutral. It falls harder on the shaded side, and a serious pay policy is built for the shade.
Sources
- Card, Mas, Moretti and Saez (2012). Inequality at Work: The Effect of Peer Salaries on Job Satisfaction. Workers told they were paid below their unit's median reported lower satisfaction and higher intent to quit with no symmetric gain above the median, establishing the asymmetry at the heart of the piece. https://doi.org/10.1257/aer.102.6.2981
- Breza, Kaur and Shamdasani (2018). The Morale Effects of Pay Inequality. Unexplained pay gaps cut output and attendance and eroded cooperation, but justified gaps did not, showing that legibility, not equality, governs the damage. https://academic.oup.com/qje/article-abstract/133/2/611/4430649
- Cullen and Perez-Truglia (2022). How Much Does Your Boss Make? The Effects of Salary Comparisons. Learning a manager earns more than assumed raised subsequent hours worked, the counterexample that shows vertical comparison can motivate where horizontal comparison wounds. https://doi.org/10.1086/717891