How Do Firms Search For Workers Over Time?

How to cite

Hoffman, Mitchell; Cullen, Zoe; Koenig, Felix. 2023. "How Do Firms Search For Workers Over Time?." AEA RCT Registry. AEARCTR-0011117. https://doi.org/10.1257/rct.11117-2.1

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Mitchell Hoffman* , Zoe Cullen , Felix Koenig

* Primary investigator

Registry ID
AEARCTR-0011117
DOI
https://doi.org/10.1257/rct.11117-2.1
First registered
Last updated
Status
In development
Trial dates
2023-03-29 – 2024-03-27
Countries
United States of America ()
Registry record
http://www.socialscienceregistry.org/trials/11117

Abstract

Many firms report trouble filling vacancies, with many jobs filled. How do firms adjust wages and non-wage policies over time to fill their jobs? When firms are provided with an opportunity to automatically adjust features of the job such as pay and skill requirements, are they interested in doing so and under what conditions? To address this issue, we will study two randomized controlled trials (RCTs) conducted by a nationwide staffing platform. In the first RCT, firms are presented with different dynamic adjustment policies that the firm is considering. We will examine whether firms choose dynamic wage and quality adjustment features and under what conditions. The second RCT has already been completed by the Platform. In this RCT, some unfilled job slots are randomly assigned bonus wages, allowing estimation of the labor supply elasticity to the firm.

Experimental design

FIRST RCT:
Employers on the Platform are invited to participate by email. They are informed that they will be asked to make decisions in different potential scenarios involving hard-to-fill jobs:

All worker adjustment: Employers are first asked about demand for dynamic wage adjustment when the wage increase applies to all workers, including those who have already accepted the job.

Marginal adjustment: 12 hours before job start time, increase pay by 10%, but where the adjustment only applies for workers accepting the job in the final 12 hours.

Marginal adjustment to median: Same as marginal adjustment, but where pay is increased to the median pay rate for completed jobs in an employer's area in this quarter.

Marginal, added pay framed as hiring bonus: Same as marginal adjustment, but where the added pay is framed as a hiring bonus instead of a wage increase.

Marginal, additional pay covered by Platform: Same as marginal adjustment, but where the added pay is paid for by the Platform.

Businesses are asked whether they wish to relax who can perform their jobs 12 hours before the job start time, specifically, whether they wish to allow lower pool workers to perform the job and whether they wish to drop any required badges.

In addition, businesses are asked about demand for having their jobs guaranteed to be filled for a randomized price.

Businesses are asked for their beliefs regarding fill rates, such as the impact on fill rates of raising pay by 10%. In addition, businesses are also asked whether they would be interested in a program that would improve their fill rate by the same amount as raising pay by 10%, but where they paid the Platform directly instead of raising pay. The offered increase in fill rate is equivalent to the business's belief about the impact of a 10% pay increase.

Finally, businesses are randomly provided with information from the first RCT about the responsiveness of job filling to bonus wage increases applied to some jobs by the Platform. Using accurate information framed differently, some businesses will receive information suggesting that responsiveness is fairly low and other businesses will receive information suggesting that responsiveness is fairly high.

All the different randomizations (price of guaranteed fill, information provided) are cross-randomized with one another.

SECOND RCT:
Some unfilled job slots are randomly assigned bonus wages by the Platform. This secondary RCT has already been conducted.

Design details

Randomization method
In the first RCT, randomization will be done in the office by a computer. In the second RCT, randomization was performed using the commercial A/B test software used by the Platform.
Randomization unit
In the first RCT, randomization is at the level of the HR manager. In the second RCT, randomization is at the job request level.
Clusters
FIRST RCT: The experiment will be emailed to all roughly 8,000 requesters / HR managers active on the Platform in the last few years. Based on typical response rates on the Platform, We expect 400-1,200 managers will respond. SECOND RCT: The experiment had 5,525 job postings.
Observations
Design is not clustered.
Arms
FIRST RCT: All randomizations will be done evenly 50/50. For example, half of managers will receive higher cost of guaranteed fill and half of managers will receive the lower cost of guaranteed fill. SECOND RCT: There are 3,059 job postings in Treatment and 2,466 job postings in Control.
Primary outcomes
In the first RCT, our primary outcome is employer demand for dynamic wage adjustment. Specifically, we are interested in employers' choice of whether to have their pay increased for jobs that are unfilled before the start time. In the second RCT, our primary outcome is the share of job slots that are filled.
Secondary outcomes
There are two secondary outcomes for the first RCT: 1. Employer demand for adjustment of required skills, that is, adjustment of who is allowed to perform the jobs. This is done by asking about demand for including lower pool workers (i.e., workers who have lower ratings and higher no show rates on the Platform) and by asking about badge requirements. 2. Employer beliefs about the probability that their job gets filled, both before and after exposure to treatment. There are no secondary outcomes for the second RCT.

Cited by

No citation data available for this registration yet.

Keywords