1 of 25

How do employers use compensation history: �Evidence from a field experiment

Moshe Barach (University of Minnesota)

John Horton (NYU Stern)

1

West Coast Economic Association

​

July, 2019

2 of 25

More and more municipalities are banning employers from using past wages in hiring. . .

2

Introduction Theory Setting Findings Discussion

  • Half of workers report current employer learned their wages from prior job (Hall & Krueger, 2012)

​

  • >80% of workers report current employer learned wages prior to extending an offer (Barach & Horton survey, 2016)

​

  • Public concern over path dependence in wages and use of this information in job screening

3 of 25

. . . Yet, business opposition to these laws is still high.

  • “In order to attract and retain them [educators] we need the ability to ask applicants about their salary history.” - San Jose Unified School District

​

  • “It effectively eliminates an employer’s ability to negotiate wage, as well as creates a new reason to sue.” - The Western States Trucking Association

3

Introduction Theory Setting Findings Discussion

4 of 25

Research question

How does the absence of wage history information change the hiring process?

​

4

Introduction Theory Setting Findings Discussion

5 of 25

Theory

  • Scholarship highlights the importance and effects of signal substitution in hiring, yet mostly ignores endogenous information acquisition.
    • Most empirical work on the role in information in hiring has come from audit studies and highlights the importance of signal substitution (Goldin & Rouse, 2000; Bertrand & Mullainathan, 2004; Autor & Scarborough, 2008; Dobbie et al. 2019)
    • Recent work highlights the potential downside of signal substitution (Henry & Jacobs, 2007; Starr, 2017; Doleac & Hansen, 2018)
    • Endogenous information acquisition has been overlooked in extant empirical work, yet has a long history in labor theory (Mortensen 1970; McCall 1970; Mortensen and Pissarides, 1999)

5

Introduction Theory Setting Findings Discussion

6 of 25

Research question revisited

  • How does the absence of wage history information change the hiring process?
  • the extent and intensity of information acquisition by employers?
  • the attributes of the workers evaluated and ultimately hired?
  • whether employers made a hire?
  • wage bargaining?
  • the match quality, if a hire was made?

6

Introduction Theory Setting Findings Discussion

7 of 25

Context: �Online labor market

7

Who?

77% of jobs are posted by a business

80% of jobs have 1-10 full time employees

50% have revenue over $100k/year

90% have revenue under $5M/year

​

What?

Remote Tasks: computer programming, graphic design, data entry, research, and writing

​

Where?

Top Employers: US, UK, France, Germany, Israel

Top Employees: US, Philippines, Russia, Bangladesh, UK

​

Introduction Theory Setting Findings Discussion

8 of 25

Posting a Job

Introduction Theory Setting Findings Discussion

9 of 25

Introduction Theory Setting Findings Discussion

10 of 25

Detailed Applicant History.

11 of 25

Experimental design.

PAST WAGES AVAILABLE

CONTROL EMPLOYER

  • Job Title: Lead data scientist
  • Contract Type: Hourly
  • Hourly Wage: $17.00
  • Total: $451.34
  • Time: December 2014 - present

PAST WAGES NOT AVAILABLE

TREATMENT EMPLOYER

  • Job Title: Lead data scientist
  • Contract Type: Hourly
  • Total: $451.34
  • Time: December 2014 – present

Introduction Theory Setting Findings Discussion

12 of 25

Experimental Details

12

The treatment level was the employer.

​

Treated employers could no longer observe recent hourly wage (measure of marginal productivity)

​

Ran on oDesk.com during September 2014

​

5,922 unique employers

2,948 assigned to control posted 4,661 job openings

2,974 assigned to treatment posted 4,815 job openings

​

Randomization was effective and the samples were well-balanced with respect to pre-randomization employer characteristics.

​

Introduction Theory Setting Findings Discussion

13 of 25

Balance

13

Introduction Theory Setting Findings Discussion

14 of 25

Hiring funnel and summary statistics

14

Introduction Theory Setting Findings Discussion

33.69 /Job

6.67 /Job

1.79 /Job

1.59 /Job

0.43 /Job

Applicants

Viewed

Contacted

​

Face-to-Face Interview *

​

Hired

20%

27%

89%

19%

1.8%

.6%

*Applicants who are questioned or asked for skype ID

15 of 25

Hiding past wages increases employer information acquisition …

15

  • Treated employers called-back 8.9% more applicants.
  • Asked 12.1% more applicants questions.
  • Invited 8.2% more applicants to face-to-face interviews.

​

Introduction Theory Setting Findings Discussion

  • Results from job-posting level poison model, controlling for category, prior jobs billed, employer prior spend, number of applications to job, number of recommended applications, average bid, indicator if job requires special skills.

16 of 25

…and [to some extent] intensity of employer information acquisition.

16

Introduction Theory Setting Findings Discussion

  • Message-threads from treated employer where 1.2 times as likely to ask a question,
  • But no more likely to set up an interview.
  • Results from messaged applicant level logit model, controlling for category, prior jobs billed, employer prior spend, number of applications to job, number of recommended applications, applicant’s bid, applicant’s tenure, applicant’s experience, and applicant’s prior feedback.

17 of 25

Characteristics of messaged applicants do not differ . . .

17

Introduction Theory Setting Findings Discussion

18 of 25

Bargain Hunting

Treatment induced bargain hunting.

  • In ‘beginner’ and ‘intermediate’ tiers call-back probability shifts towards lower wage applicants.
  • In the ‘expert’ tier we simply see less messaging of most expensive applicants.

​

​

18

Introduction Theory Setting Findings Discussion

Probability of call-back on applicant profile wage rate

19 of 25

Signal Substitution

Employers value prior jobs, prior earnings, and prior feedback.

​

Employers do not place high value on profile wage rate.

​

Across all analyzed ‘signals 'but tenure there does not seem to be any significant shift in weighting of signals.

19

Introduction Theory Setting Findings Discussion

Applicant characteristics on probability of call-back

20 of 25

. . . but hiding past waged increased hire-rate and shifted hiring toward cheaper applicants

20

Introduction Theory Setting Findings Discussion

21 of 25

Hiding past wages shifts bargaining power towards [lower wage] applicants . . .

21

Introduction Theory Setting Findings Discussion

  • Treated employers are no more likely to bargain over wages
  • But when they do, the ratio of applicant’s wage to the applicant’s initial bid is 9% higher.

22 of 25

. . . but has no measurable effect on employer outcomes

22

Introduction Theory Setting Findings Discussion

23 of 25

Winners & Losers among applicants

23

Introduction Theory Setting Findings Discussion

Notes: This figure shows the relationship between an applicants expected value from submitting a bid, and the treatment by the applicant's profile wage. The level of observation is the bid. Models all include an applicant level fixed effect. Heteroskedasticity-robust 95% confidence intervals are displayed.

24 of 25

Contributions

24

Employers use past wages as a screening shortcut for learning about productivity

  • Information contained in salary history is a substitute for more/ more intense screening of applicants.

Banning past wages will increase search/ screening costs for employers but. . .

  • Will lead to the hiring of more lower cost applicants
  • Will lead to savings on wages
  • Will shift bargaining power to applicants, and specifically lower wage applicants.

​

No evidence of worse job outcomes or undo burden on employers.

Evidence of these policies helping exactly who they are intended to help.

​

​

​

Introduction Theory Setting Findings Discussion

25 of 25

Thank you!

mbarach@umn.edu

25