Why Do People Choose Freelance Work Arrangements? Evidence From a Randomized Trial

2024-03-13

2024-06-03

Willingness to pay for traditional employment instead of independent contract work; propensity to choose traditional employment

See full analysis plan.

Beliefs about net pay relative to gross pay; reported expenses

We estimate individual-level preferences for traditional employment versus freelance work using the Bayesian Adaptive Choice Experiment (BACE) framework developed in Drake, Payró, Thakral and Tô (2024). We have respondents choose between 18 sequential pairs of jobs. Key to our design is that in each comparison, one job is always a freelance independent contractor job and the other is a traditional employment job Each job is additionally characterized by a pay amount (discussed further below), a degree of control over the performance of one’s job, control over schedule, remote work possibility, the probability the job ends before 1 year—these are varied in each set of options in order to estimate one’s willingness to pay for each job feature separately from a willingness to pay for a W-2 job per se (holding other attributes constant). To fix comparisons, we state that neither job offers health insurance or retirement benefits and that the weekly number of hours are the same in each job, and anchor pay levels to their actual current pay level. We randomize participants into three groups that vary the order of questions asked and the extent of information provided about the distinction between freelance work and traditional employment.

Our RCT randomly assigns respondents into one of three experimental conditions that vary what information is provided about the nature of independent contract arrangements whether the pay amount is presented as a gross amount before taxes and expenses or a net amount after taxes and expenses. In particular, respondents are randomized into three conditions at the beginning of the study:

In a control group, jobs options are described as either 1099 independent contractor jobs or as W-2 employment jobs and list the gross pay before taxes and expenses for each job in the 18 choice pairs. We ask detailed questions about expenses in one’s current job (asking workers who are employees to imagine what it would cost to cover expenses themselves if they had to) after data collection in the discrete choice experiment has concluded.

In a first treatment condition, we ask the same detailed questions about expenses but do so before beginning the pairwise comparisons, and additionally provide information about typical costs of various expense items. We then explicitly highlight key differences in how the law treats employees and contractors (i.e. social insurance, tax responsibility, and legal protections) and provide expected taxes and expenses for sample independent contract and employment jobs with the same gross pay. After conducting a knowledge check and assessing beliefs about net pay, we then direct respondents to make the 18 job comparisons, again presenting gross pay for each option.

In a second treatment condition. We provide the same information about how the law treats employees and contractors as in condition B. We assess beliefs about net pay, but do not provide information about taxes and expenses, nor do we ask questions about expenses until after the discrete choice experiment has concluded (as in the Control condition A). In the discrete choice experiment, however, we present participants with net earnings after taxes and expenses for each option rather than gross pay amounts.

This study is implemented in an extension of the Entrepreneurship in the Population (EPOP) survey conducted by NORC, which is fielded in the AmeriSpeak survey panel. The Qualtrics extension survey was developed in collaboration with the Princeton Survey Research Center and is being administered by NORC.

Randomization done within Qualtrics.

Individual.

No

3000 individuals.

3000 individuals.

1000 in control, 1000 in treatment arm 1, 1000 in treatment arm 2