.svg)
Salary and Compensation Data for Economics Research
Pay is one of the hardest quantities to observe directly. Household surveys ask respondents to recall their earnings, and administrative earnings records are rarely released at a grain that identifies the employer, the occupation and the moment a wage was set. Compensation records drawn from job postings, disclosures and worker profiles attach a salary or a range to a named firm, a job title and a date, so economists can watch offered pay move within an employer and across geographies between one quarter and the next. The trade is that much of what these sources show is the offer rather than the wage eventually agreed, and that distinction has to be carried through the design.
How economists use salary and compensation data
Wage differentials across markets and occupations
Economists use posted and reported compensation to estimate how much of the pay gap between two labour markets survives once occupation, seniority and employer are held fixed, comparing otherwise similar workers who move between markets or who take comparable positions in different places. The design turns on having a pay figure attached to a named job title at a known date, which posted salary data supplies and recall based survey earnings do not. What it cannot recover is the negotiated outcome: a posted range is an offer, and the realised wage may sit anywhere inside it.
Labour demand and skill requirements after a policy change
A difference in differences on vacancies before and after a tax, licensing or employment law change measures whether firms respond on the hiring margin, and posted salaries and stated skill requirements separate a response in the number of openings from a response in what firms are willing to pay for them. Because postings are dated to the day, the design can distinguish an announcement effect from an implementation effect, which quarterly wage statistics collapse together. The panel only covers hiring that is advertised online, which skews toward larger employers and white collar roles and away from positions filled by referral.
Technology shocks and the price of skills
To test whether a new technology substitutes for or complements labour, economists build occupational exposure from a pre period composition of roles and then track vacancies and offered pay for exposed occupations against unexposed ones within the same firm. Compensation carries the half of that test quantities cannot: a firm that posts fewer roles and a firm that posts the same roles at lower pay look identical in vacancy counts alone. Exposure measures are constructed from job titles, so classification error in those titles propagates straight into the estimated effect.
Salary and compensation datasets available on Dewey
WageScape
Salary and compensation records that let economists compare pay across occupations, employers and locations at a stated point in time, rather than inferring it from self reported earnings in survey microdata.
View datasetLightcast
Labour market and workforce data built from job postings, pairing advertised pay with occupation codes and required skills, which is what makes it usable for studying how employers adjust hiring after a policy or demand shock.
View datasetRevelio Labs
Workforce data covering roles, headcount and departures alongside compensation estimates, so economists can follow pay and staffing within a firm over time, with the caveat that the pay figures are modelled rather than payroll records.
View datasetMore published work using salary and compensation data
Joseph Chami · Working Paper · 2025
Tatjana Dahlhaus, Reinhard Ellwanger, Gabriela Galassi, Pierre-Yves Yanni · AEA Papers and Proceedings · 2025
Martin Andersen, Kaden Grace · Working Paper · 2026
.svg)