Gig Rules: The Political Economy of Labor Market Regulations

How to cite

Luo, Hong; Oberholzer-Gee, Felix. 2024. "Gig Rules: The Political Economy of Labor Market Regulations." AEA RCT Registry. AEARCTR-0013651. https://doi.org/10.1257/rct.13651-1.1

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

Hong Luo* , Felix Oberholzer-Gee

* Primary investigator

Registry ID
AEARCTR-0013651
DOI
https://doi.org/10.1257/rct.13651-1.1
First registered
Last updated
Status
Completed
Trial dates
2024-05-31 – 2024-06-28
Countries
United States of America ()
JEL codes
D72, J3, J8
Registry record
http://www.socialscienceregistry.org/trials/13651

Abstract

In response to the emergence of online platforms that match workers with specific tasks such as ride-sharing, coding, or handy work, lawmakers, regulators, and sometimes even voters created new rules to govern these novel forms of work. The regulations often seek to strike a balance between the benefits of gig work, primarily flexibility and autonomy, and the advantages of traditional forms of employment. In this paper, we study the welfare effects of a wide range of hypothetical job designs, using ride-sharing as our example. We ask three questions: First, if current and prospective drivers were to choose freely between different job designs, which features would be particularly attractive to them? Second, what are the design elements of jobs that find broad political appeal among voters? And third, what are the welfare consequences of imposing politically popular job designs on current and prospective drivers? To answer these questions, we use conjoint survey experiments to solicit preferences for job designs from both likely voters and prospective drivers. In our analysis, we first estimate voters’ and drivers' preferences for each job attribute. Equipped with these estimates, we then predict the expected voter support for all possible job designs. Zooming in on designs that secure a majority of votes, we estimate drivers’ decisions to drive as well as their well-being under these working conditions. Our project aims to expand the horizon on what ridesharing could look like and the welfare consequences of having voters and their representatives decide labor market regulations.

Intervention

Our data are generated by two sets of conjoint survey experiments, in which we ask voters to vote between a random hypothetical job versus the status quo and drivers about their willingness to drive personally for a random hypothetical job. Please see the attached pre-analysis plan for the details.

Experimental design

For the voting survey, we draw a random sample that represents the general voting-age population. For the driving survey, we draw three different samples: individuals who worked for companies like Uber or Lyft in the past; current drivers for these types of companies, and a third group, which we call likely drivers. Likely drivers consist of people who might consider driving in the future, perhaps not under the current working conditions but under a broad range of possible ride-sharing job designs. In our analysis, we combine the current and past drivers due to their limited numbers.

We conceptualize jobs as bundles of attributes. A job design consists of ten attributes including compensation, work-related benefits (e.g., healthcare), and other elements such as work flexibility and the predictability of work schedules.

The design of the two conjoint survey experiments is similar. We ask each participant to complete eight tasks. In Task 1, the respondents evaluate a job profile that describes the working conditions that are typical of many of today’s ride-sharing jobs. In Tasks 2-7, the participants compare two job profiles: Job A represents the status quo, the profile that the participants saw in their first task. Job B is a hypothetical, randomly populated job profile that differs from the status quo along multiple dimensions. To facilitate the comparison, we highlight in boldface the attributes of Job B that differ from Job A. The first seven tasks are identical across voter and driver surveys. The only difference is that we ask respondents to evaluate the job designs as voters (“would you vote for or against…”) or drivers (“would you personally be willing to drive…”). The eighth task differs across the two surveys.

Please see the attached pre-analysis plan for the details of our study design.

Design details

Randomization method
Randomization done in office by a computer
Randomization unit
Each respondent is asked to compare the status-quo job and six randomly-populated hypothetical jobs.
Clusters
Not clustered
Observations
1. For the voting survey, the goal is to collect about 3000 completed surveys. We plan to use the forced-choice data in our main specification for this survey. Each survey respondent answers six forced-choice questions. We will, thus, have 18,000 observations on pair-wise choices. 2. For the likely-driver survey, the goal is to collect about 1000 completed surveys. We plan to use the independent-rating data in our main specification for this survey. Each survey respondent answers 7 rating questions. We will, thus, have 7,000 observations on pair-wise choices. 3. For the current/past driver surveys, the goal is to collect about 550 completed surveys in total. Each survey respondent answers 7 rating questions. We will, thus, have 3,850 observations on pair-wise choices.
Arms
Our study uses a conjoint survey methodology. We are interested in the marginal value of changing a job attribute from its default value to an alternative value. For each survey, the number of observations for the default value and each alternative value of a job attribute will depend on the number of levels of this job attribute. For example, for the minimum-hour restriction, the default value is "Drivers can freely choose how many hours they work each week." There are three alternative values: "Drivers are required to work at least 10 [or 20, 40] hours per week." For the voting survey, because we have 18,000 effective observations for the voting survey, each of the four groups (one default and three alternatives) includes 4,500 observations. Please see additional details in the pre-analysis plan.
Primary outcomes
Our primary interest is the marginal value of each job attribute, separately for likely voters, likely drivers, as well as current and past drivers.
Secondary outcomes
We are also interested in conducting counterfactual analyses based on our estimates of the marginal values of job attributes. Specifically, we are interested in predicting the expected voter support for all possible job designs. Zooming in on designs that we expect to be supported by a majority of voters, we estimate drivers’ decisions to drive as well as their well-being under these working conditions.

Cited by

No citation data available for this registration yet.

Keywords