AdobeStock_219834607.jpeg
VoxEU Column Labour Markets Gender

Salary history and employer demand

Salary history bans are increasingly popular in the US. Using a field experiment to investigate how salary history disclosures impact employer demand, this column finds that disclosing a high salary is often a signal of quality and yields higher salary offers. However, higher salaries can also make candidates too expensive to justify a callback. The authors find evidence that such policies can equalise some outcomes across genders, but sometimes as a result of reductions for men rather than raises for women.

Salary history bans have been adopted by 21 states in the US since 2016. The laws aim to reduce historical inequalities relating to gender, race/ethnicity, or other protected characteristics. Such bans are similar to other laws aiming to make decisions more fair by limiting the information employers can seek, such as ‘ban the box’ laws limiting criminal background inquiries (Doleac and Hansen 2020, Agan and Starr 2018) and laws banning employer credit checks (Bartik and Nelson 2016, Friedberg et al. 2017, Corbae and Glover 2018) or gender inquiries (Card et al. 2021, Kuhn and Shen 2021).

Although these bans regulate the seeking of information, they don’t always stop candidates from providing it. As such, the bans raise questions that economists have studied in research on voluntary disclosure. Classic theory (Viscusi 1978, Grossman and Hart 1980, Milgrom 1981, Grossman 1981, Autor 2008) explored ‘unravelling’, or the idea that market forces will compel firms or candidates to disclose information (such as a prior salary). Behind these results are the idea that silence would be a worse signal. As Jin et al. (2015) summarised, ‘no news’ might be perceived as ‘bad news’, because anyone possessing ‘good news’ would volunteer it immediately. However, more recent empirical/behavioural papers show that unravelling does not always play out as predicted (Luca and Smith 2015, Bederson et al. 2018, Jin 2005, Fung et al., 2007, Bederson et al. 2018).

In a recent paper (Agan et al. 2021), we study how salary history disclosures impact employer demand. We examine employers’ reactions to these disclosures, as well as how their reactions change depending on whether the candidate was asked (versus the candidate volunteering their salary history completely unprompted).

In most settings, the key behaviours are not random: employer’s don’t ask randomly, and workers don’t volunteer their salaries randomly. As such, we develop a unique field experiment to understand how salary disclosure causally impacts employer demand. Our approach extends the traditional resume audit method (used, for example, by Booth and Leigh 2010 and Patacchini et al. 2012) to incorporate recruiters. Employment in third-party recruiting in the US more than doubled between 2002 and 2019 (Cowgill and Perkowski 2021) while overall employment grew by 15%.

We hired 248 recruiters to evaluate 2,048 job applications for a software engineering position. Although the job opening and the applicants are secretly fictitious, the materials we gave the recruiters were based on realistic inputs in the software field (including realistic candidate histories and qualifications). Because we can control what the recruiters are shown, we randomly vary whether the job application form asks for salary history, and whether the candidate answers the question (or volunteers their salary history unprompted in another section of the application). Among those who disclose, we also randomly vary the amount of salary disclosed (in a way that mimics real-world salaries for our applicants, including the gender wage gap in software engineering). This setup allows us to ask for a variety of assessments about the candidates, including whether they should be called back about the job, what salary should be offered, how many competing offers the recruiter thinks they may have, and others. We call our field experiment a two-sided audit study because there is randomisation on both sides.

We find that recruiters respond to the salary amounts disclosed by candidates. Candidates who disclose higher previous salaries are presumed to be higher quality and receive higher salary offers from our recruiters, as shown in Figure 1. They are also viewed as having better competing offers, and thus the wage premium partly reflects beliefs about competing offers from rival employers.

Figure 1 

 

The recruiters we study are sophisticated in their inferences about historical salaries. There are many ways to earn a higher salary, and the candidate’s characteristics affect how recruiters interpret the salary disclosed. Some of our candidates come from firms with a high average wage. Some candidates come from lower wage firms, but are well paid within their firm’s distribution. Finally, some candidates (men) are simply the beneficiary of the gender wage gap.

Our experiment was designed to measure whether recruiters notice these distinctions and how they treat them differently. We found differences in how recruiters treat the scenarios above. An extra dollar earned from being well-paid inside a firm’s distribution was the most highly rewarded: an extra dollar earned from a good internal distribution was matched by $0.71. Next were the extra dollars earned from being at a high-wage firm ($0.67).

Finally, the recruiters were least responsive to dollars coming from the gender wage gap. Recruiters appeared to detect overpaid men and treat them differently than candidates who are well-paid for other reasons. Although the male gender premium was discounted, it was not entirely eliminated. For every extra dollar male candidates disclosed from the gender gap, recruiters matched dollars $0.48 – less than 1:1 and less than the amounts above, but still above zero.

Overall, having a high salary was a signal of quality and yielded larger salary offers. However, across many of our results, we find that salary history disclosures change beliefs about competing offers more than anything else we study (including beliefs about candidate quality). This likely drives one of our main results: it is possible for candidates to disclose ‘too high’ of a salary history and become too expensive to justify a callback. At some point, a worker’s competing offers can outpace the worker’s value. These workers – which include many highly paid men in our study – might be better off staying silent.

Finally, our study explores the meaning of silence. Some of the job applicants in our study did not disclose their salary at all. We find that recruiters appear to anticipate strategic disclosure (or silence) by candidates. On average, our recruiters interpreted nondisclosure as a negative signal of quality and outside options, and offered lower salary amounts to these candidates, as shown in Figure 2. The recruiters treat silent candidates as if their quality was below that of an average candidate (specifically around the 25th percentile). In this sense, they felt that ‘no news’ was ‘bad news’ on average.

Figure 2

 

We do find some exceptions to the finding about silence. First, very highly paid workers may be better off silent; as discussed above, a salary can be ‘too high’ and scare off some employers. Second, we find that recruiters penalise female silence less, and male silence more. One possible reason for this is that women are more uncomfortable disclosing their salaries at all levels (Agan et al. 2020, Cowgill et al. 2021). As a result, female silence appears to be interpreted as discomfort, rather than a strategic choice to withhold unflattering information.

Taken together, several themes appear in our findings. 

  • First, we see that salary history is interpreted as a measure of candidate quality. However, it is a much stronger signal of outside offers. 
  • Second, we see that employers are relatively sophisticated in their interpretation of salary history. They appear to anticipate strategic silence, they discount male wage premiums, and they anticipate that female silence may be non-strategic. Third, recruiters do not fully eliminate the gender wage gap, despite this sophistication. 
  • Finally, choices about salary history disclosure appear to involve tradeoffs on many dimensions. Candidates may need to weigh the benefits of receiving a callback at all versus lowering their probability of a callback but having a higher salary attached.

Our papers on salary history (Agan et al. 2021, Cowgill et al. 2021) identify strengths and weaknesses in the broader efforts to limit employers’ information. We find evidence that these policies can equalise some outcomes across groups. However, the equality sometimes comes from reductions for men, rather than raises for women. Choices to disclose may involve trade-offs between competing objectives. Information such as reporting a lower historical salary can help candidates’ chances of a callback, while harming their chances of getting a high salary (conditional on a callback). Finally, we also find multiple ways that both employers and job candidates can adapt to the limitations on information seeking. Candidates can adapt by disclosing even without being asked, and employers can adapt by inferring candidates’ undisclosed salaries based on other characteristics (particularly if they anticipate strategic silence).

Our paper suggests several next steps. First, our experiment examines the employer side of the unravelling story, but what about candidates? In a separate paper (Cowgill et al. 2021), we survey candidates’ disclosure choices around salary history. We find that candidates feel more pressure to disclose as others disclose, and over time we find that candidates are more likely to disclose (even without prompting). Our results suggest that silence about salary history is unravelling over time.

Second, our experiment raises questions about the accuracy of employer beliefs. Are silent workers really less productive? Do higher paid workers truly have better outside offers? These questions are beyond the scope of our paper, but may be important to understanding salary history bans.

Finally, our two-sided audit design has a number of applications outside of salary history bans. We mention two particularly useful features. First, our design can be used to randomise any employer-side hiring practice that might reduce discrimination. For example, a researcher could randomise the gender blindness of resume screens, or other innovations to reduce discrimination. Second, our design allows us to measure more outcome variables. While a traditional audit study examines only callbacks, our paper was able to elicit employers’ willingness to pay, beliefs about competing offers and many other variables.

References

Agan, A and S Starr (2018), “Ban the box, criminal records, and racial discrimination: A field experiment”, The Quarterly Journal of Economics 133 (1): 191–235.

Agan, A Y, B Cowgill, and L Gee (2020), “Do workers comply with salary history bans? Voluntary disclosure, adverse selection, and unraveling”, AEA Papers and Proceedings 110: 215-219.

Agan, A Y, B Cowgill, and L Gee (2021),  “Salary history and employer demand: Evidence from a two-sided audit”, SSRN.

Autor, D (2008), “The economics of labour market intermediation”, VoxEU.org, 30 October.

Bartik, A and S Nelson (2016), “Deleting a signal: Evidence from pre-employment credit checks”, MIT Department of Economics Graduate Student Research Paper 16-01, Chicago Booth Research Paper No. 19-23.

Bederson, B B, G  Jin, P Leslie, A J Quinn, and B Zou (2018), “Incomplete disclosure: Evidence of signaling and countersignaling”, American Economic Journal: Microeconomics 10(1): 41–66.

Booth, A and A Leigh (2010), “Do employers discriminate by gender in female-dominated occupations? Results from a field experiment”, IZA Discussion Paper No. 4690.

Card, D, F Colella, and R Lalive (2021), Gender preferences in job vacancies and workplace gender diversity, Technical Report, National Bureau of Economic Research.

Corbae, D and A Glover (2018), “Employer credit checks: Poverty traps versus matching efficiency”, NBER Working Paper 2500.

Cowgill, B, A Agan, and L Gee (2021), “Salary history ban and unravelling: Psychological costs and coarse beliefs”, Working Paper.

Cowgill, B and P Perkowski (2021), “Delegation in hiring: Evidence from a two-sided audit”, Columbia Business School Research Paper 898.

Doleac, J L and B Hansen (2020), “The unintended consequences of “ban the box”: Statistical discrimination and employment outcomes when criminal histories are hidden”, Journal of Labor Economics 38(2): 321–374.

Friedberg, L, R M Hynes, and N Pattison (2017), “Who benefits from bans on employer credit checks?”, Departmental Working Papers 1704.

Fung, A, M Graham, and D Weil (2007), Full disclosure: The perils and promise of transparency, Cambridge University Press.

Grossman, S J (1981), “The informational role of warranties and private disclosure about product quality,” The Journal of Law and Economics 24(3): 461–483.

Grossman, S J and O D Hart (1980), “Disclosure laws and takeover bids”, The Journal of Finance 35(2): 323–334.

Jin, G, M Luca, and D Martin (2015), “The behavioural economics of voluntary disclosure”, VoxEU.org, 22 July 22.

Jin, G Z (2005), “Competition and disclosure incentives: An empirical study of HMOs”, Rand Journal of Economics 36(1): 93–112.

Kuhn, P J and K Shen (2021), What happens when employers can no longer discriminate in job ads?, Technical Report, National Bureau of Economic Research.

Luca, M and J Smith (2015), “Strategic disclosure: The case of business school rankings”, Journal of Economic Behavior & Organization 112: 17–25.

Milgrom, P R (1981), “Good news and bad news: Representation theorems and applications”, The Bell Journal of Economics 12(2): 380–391.

Patacchini, E, G Ragusa, and Y Zenou (2012), “Unexplored dimensions of discrimination in Italy: Sexual orientation and physical appearance”, VoxEU.org, 14 November.

Viscusi, W K (1978), “A note on ‘lemons’ markets with quality certification”, The Bell Journal of Economics 9(1): 277–279.

735 Reads