Why do individuals do gig work?

Our study examines why individuals choose to participate in the gig economy. We hypothesize that an important driver of labor supply to gig work is the degree of working-time flexibility it offers relative to traditional employment. We conceptualize working-time flexibility along two dimensions: (i) control over the number of hours worked and (ii) control over the timing of work.

To study these preferences, we design a bespoke survey that elicits workers’ valuations of working-time flexibility, earnings, and other non-wage job amenities. Embedded in the survey is a conjoint experiment in the spirit of Maestas et al. (2023), in which respondents choose between hypothetical job offers that vary randomly in wages and non-wage amenities. Details on the amenities and randomization procedure are provided under experimental details.

The survey is administered to three samples: (1) a representative sample of gig workers employed by online food-delivery platforms in Germany and their subcontractors; (2) a sample of gig workers employed by ride-share platforms and their subcontractors; and (3) a representative sample of lower-skilled non-gig workers. All samples are drawn from the administrative social security data of the Institute for Employment Research (IAB) in Nuremberg, Germany.

The German institutional setting is particularly well suited for this study because many food-delivery gig workers are formally or semi-formally employed by a small number of large delivery platforms and their subcontractors, making them identifiable in administrative data. Ride-share drivers are also frequently employed, although their identification in administrative records is more challenging.

The conjoint experiment generates choice data that allow us to estimate workers’ willingness to pay (WTP) for non-wage job amenities, particularly working-time flexibility. We compare WTP estimates across the different samples and test whether gig workers place a higher value on flexibility, which may help explain why workers sort into gig-economy jobs despite lower average pay relative to other lower-skilled occupations requiring similar formal qualifications.

Maestas, N., Mullen, K. J., Powell, D., Von Wachter, T., & Wenger, J. B. (2023). The value of working conditions in the United States and implications for the structure of wages. American Economic Review, 113(7), 2007-2047.

External Link(s)

Registration Citation

Citation

Friedrich, Martin et al. 2026. "Why do individuals do gig work? ." AEA RCT Registry. May 18. https://doi.org/10.1257/rct.18587-1.0

Sponsors & Partners

Experimental Details

Interventions

Intervention(s)

Intervention Start Date

2026-05-13

Intervention End Date

2026-08-31

Primary Outcomes

Primary Outcomes (end points)

Job choice

Primary Outcomes (explanation)

Respondents to a survey conjoint experiment choose between 10 pairs of jobs, yielding choice data.

Secondary Outcomes

Secondary Outcomes (end points)

Secondary Outcomes (explanation)

Experimental Design

Each respondent is presented with 10 pairs of hypothetical job offers (Job A and Job B) and is asked to indicate which job they would accept if they were searching for employment and had received multiple competing offers.

Job offers vary in both wages and non-wage job attributes. In each choice pair, gross pay is randomized around the respondent’s actual wage by multiplying the respondent’s actual wage by a random rescaling parameter drawn from an approximately normal distribution with a mean of 1 and a SD of 0.01. To avoid implausible values, randomized wages are censored at 75% and 125% of the respondent’s last reported wage.

The primary non-wage attribute of interest is schedule flexibility. We distinguish between five work-scheduling regimes:
(1) The worker sets the schedule and can change it at short notice.
(2) The worker sets the schedule but can only change it with advance notice.
(3) The company sets the schedule and it is fixed.
(4) The company sets the schedule and can change it with advance notice.
(5) The company sets the schedule and can change it at short notice.

In addition, job offers may vary along the following dimensions:
- Contracted weekly working hours (10, 20, or 40 hours)
- Availability of voluntary additional hours (possible vs. not possible)
- Whether evening or weekend work is common
- One-way commuting time (15, 30, 45, or 60 minutes)
- Type of employment contract (permanent vs. temporary)

In each job pair, wages are always randomized, and two of the six non-wage attributes are randomly selected for randomization. The remaining non-wage attributes are fixed at the values reported for the respondent’s current job, which are elicited immediately prior to the experiment.

Respondents are instructed to assume that the two job offers are identical in all respects except for the attributes highlighted in yellow. This assumption also applies to job characteristics not explicitly mentioned in the experiment.

Experimental Design Details

Not available

Randomization Method

Randomization done in office by a survey programming software

Randomization Unit

We randomize job attributes across 10 job pairs shown to each individual

Was the treatment clustered?

No

Experiment Characteristics

Sample size: planned number of clusters

The treatment is not clustered.

Sample size: planned number of observations

We target 2,250 interviews with gig workers and 1,500 interviews with lower-skilled non-gig workers. Gig workers are intentionally oversampled because we expect higher rates of employment attrition among gig workers relative to workers in traditional employment. In some specifications, we may restrict the analysis to respondents who are still engaged in gig work at the time of the survey. The oversampling strategy is intended to yield an approximately balanced sample of gig workers and lower-skilled non-gig workers in the experimental analysis even if we impose this restriction. Assuming respondents complete an average of 8 out of 10 job-choice pairs, we expect to observe approximately 18,000 choices from gig workers and 12,000 choices from lower-skilled non-gig workers.

Sample size (or number of clusters) by treatment arms

We target 2,250 interviews with gig workers and 1,500 interviews with lower-skilled non-gig workers. Gig workers are intentionally oversampled because we expect higher rates of employment attrition among gig workers relative to workers in traditional employment. In some specifications, we may restrict the analysis to respondents who are still engaged in gig work at the time of the survey. The oversampling strategy is intended to yield an approximately balanced sample of gig workers and lower-skilled non-gig workers in the experimental analysis even if we impose this restriction. Assuming respondents complete an average of 8 out of 10 job-choice pairs, we expect to observe approximately 18,000 choices from gig workers and 12,000 choices from lower-skilled non-gig workers.

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Supporting Documents and Materials

IRB