Experimental Design Details
The purpose of this experiment is to collect data about people's work schedule preferences with the goal of learning about their WTP for alternative work arrangements and to examine differences in WTP by gender and age. There are two parts to the experiment. The first is a hiring experiment and the second is a survey.
Hiring experiment: As part of a labor market survey that we are conducting, we randomly offer applicants for interviewer positions choices between traditional 9-5 office positions and “worker-friendly” alternatives including the possibility of making their own schedule, choosing the number of hours of work, the option of working from home, and combinations of these features. We also offer a choice between a traditional job and an “employer-friendly” schedule that allows the employer to set the work schedule with only one week of anticipation. We randomly vary the difference in the wage between the traditional option and the alternative and recorded which position was chosen by applicants. This approach allows us to estimate the entire distribution of willingness to pay across job applicants for these alternatives relative to traditional arrangements.
The sequence of the experiment is as follows:
We post job ads in a job board in randomly selected cities for call center telephone interviewer positions. The ads have a link that redirects job searchers to a job application page set up specifically for this research. The application asks applicants a number of demographic and work history questions. During the application process we ask applicants to choose between two positions. One of the positions is (almost always) a traditional 9-5 job. The second is a randomly selected work alternative. We do not tell applicants that these positions are available, but we also do not explicitly state that the question is hypothetical. Specifically, we tell them:
"Tell us which of the following two positions you prefer. The type of work is the same in both jobs. Please click on each job title in order to review the work descriptions.
It is important that you read the position descriptions carefully so you can indicate your preference below.
List position 1 here
List position 2 here
If you were selected for both positions, which one would you prefer? Write your preferred position number in the box below. (Regardless of your choice, you will be considered for all open positions at [org]. Your choice will not affect whether you receive a job offer. It will only be reviewed after hiring decisions have been made.) If you are not interested in either position, simply click on "No thanks, this isn't for me."
The difference in the wages between position 1 and position 2 will vary at random by the amounts (in $) -5,-4,-3,-2.75,-2.5,-2.25,-2,-1.75,-1.5,-1.25,-1,-.75,-.5,-.25,0,.25,.5,.75,1,1.25,1.5,1.75,2,2.25,2.5,2.75,3,4,5. We set the wages such that the maximum wage over position 1 and 2 is the same for all applicants.
Position 1 and Position 2 each have a randomly assigned position number. When the applicants choose one of the positions they have to write down the position number in a box (we only accept valid position numbers). To overcome order and position number effects we randomly vary the order of the position 1 and 2 text, but we do not vary the order of the position numbers.
Applicants either choose one of the two positions or opt out. We record choices made. After applicants have applied we hire a subset of them for our telephone interview positions. In practice we give them the highest of the wages they observed and the maximum flexibility to do their work.
The attached document lists the exact language of the comparisons.
We anticipate that some applicants will be inattentive and not read the position descriptions carefully. The presence of inattention will bias our WTP estimates. To correct for this we seek to estimate the inattention rate. We do this in three ways. First, we present comparisons that are highly dominated such as an employer-friendly flexible job that pays $5 less than traditional job. Second, we compare the rate of people choosing the lower paid job when the same descriptions are posted but with different wages. Third, we ask people to tell us whether they chose the 9-5 job or an alternative. All three approaches are designed to estimate the fraction of applicants making "mistakes". We can then use these estimates to correct the WTP measures. Our plan is to rely mostly on the first of these measures (comparison of dominated options).
Survey: In addition to the hiring experiment, we plan to conduct a survey that asks a nationally representative group of individuals questions about their work schedules. In the survey we will also pose some hypothetical scenarios for the respondents to answer. For each scenario we ask respondents to decide if they want a more or less flexible alternative and we vary the wage difference (proportionally) between them. See attached document for additional detail on the comparisons being made. As seen in the attached document we also make a comparison of two alternatives where one is clearly dominated in order to measure the inattention rate.