Can one email change how workers feel about their pay?

Transparency turns a vague comparison into a rank

Workers often have an impression of where their pay sits. A searchable salary database replaces that impression with names and numbers. It may reveal an underpayment, correct an inaccurate belief or make relative position easier to infer.

In 2008, salary records for University of California employees became publicly searchable. Researchers used that moment to test what happens when access to comparison information becomes easier.

One random email made salary data harder to ignore

The sampling frame contained 41,975 employees at UCLA, UC San Diego and UC Santa Cruz. A randomly selected treatment group received an email explaining and linking to the new salary website. Controls did not receive that prompt before a follow-up survey.

The main analysis used 6,411 respondents whose survey answers could be joined to administrative salary data: 1,776 in treatment and 4,635 in control. Researchers compared each worker with the median in their pay unit and broad occupation, then measured salary satisfaction, job satisfaction and intention to look for another job.

The random treatment was an encouragement to browse. It did not change salary, force a website visit or verify exactly what every recipient learned. The results are effects of making the information more salient and accessible.

Workers below the median reacted

For workers paid below their comparison median, the information treatment reduced a combined satisfaction index by 0.063, approximately one-tenth of a standard deviation. It also increased the probability of saying they were very likely to make a genuine effort to find another job in the next year by 4.3 percentage points—about 20% relative to the comparison mean.

The negative response was larger farther below the reference point. In models that compared explanations, ordinal pay rank described the response better than the size of the absolute salary gap.

These outcomes were attitudes and intentions. The study did not establish that workers actually quit, found better roles or negotiated higher pay.

Workers above the median gained no equal lift

Employees above the median did not report a statistically detectable increase in satisfaction. Their estimates were small and consistent with no effect. The information response was therefore asymmetric: learning that one sat lower in the local distribution had a measurable cost, while learning that one sat higher did not produce a matching measured benefit.

The search outcome was a stated intention, not observed quitting or a verified job move.

That asymmetry matters for product design. Our editorial inference is that a comparison interface may create downside without an equal upside. More precision is not automatically more wellbeing.

One workplace cannot settle the transparency debate

This was one public university system at a particular moment of salary disclosure. Public-sector pay, comparison units and expectations may differ from private companies or independent work. Survey response was incomplete, and the experiment measured whether people were prompted—not whether each person understood the same reference group.

Treatment and placebo emails also reduced follow-up survey response by roughly four to five percentage points. Their similar effects suggest that receiving an extra email may itself have discouraged response, but the outcome analysis still conditions on who answered.

The study does not conclude that pay transparency is bad. Transparency may expose inequity or support bargaining and accountability, none of which was fully measured by the satisfaction survey. It shows one consequence that a complete assessment must include.

Rank needs an action path

Comparison data is easier to absorb when it answers more than where do I rank? A useful system can also show how pay is set, what the relevant role and experience bands are, how to question an error and what choices remain.

That shifts the emphasis from rank as identity to information as agency. The experiment shows how exact comparison can change attitudes, but it did not measure who workers could reach or how opportunities moved through their relationships. At a broader level, are places with more cross-class connection also places with greater economic opportunity?