How do job seekers react to information about wage and job benefits?

In the field experiment, some of the job seekers who will visit the job vacancy websites managed by our implementing partner will be able to access simple information about wages and job benefits that are associated with the companies posting job ads on the websites.

The belief-elicitation survey experiment replicates a similar intervention to estimate how job seekers' belief are effected by the provision of information.

2023-03-06

2023-05-31

The main objective of our experimental design is to measure how much the interest of job seekers in job ads changes as different wage and job benefit information is displayed. The primary outcomes that we use to measure job seekers’ interest in a job ad are
whether the job seeker clicks and opens the webpage associated with the job ad in order to view extra details about the vacancy and about the firm that posted it;
whether the job seeker applies for the job through the platform.

In a separate project, we plan to use our experimental design to measure how transparency about wage and job benefit information affect the search behavior of job seekers on the platform and the characteristics of the jobs they apply for.

The primary outcomes in the belief-elicitation survey experiment are the respondent's beliefs about the wage and benefits associated with each of the job ads that they see.

Secondary outcomes that we plan to use to measure job seekers’ interest in a job ad are:
whether the job seeker clicks and opens the webpage associated with the firm posting the job ad;
whether the job seeker clicks and opens the third-party website displaying reviews (from employees) about of the firm posting the job ad;
whether the job seeker adds the job ad to a watchlist;
whether the job seeker clicks and opens the webpage showing the original job ad on the firm’s website.

The secondary outcomes in the belief-elicitation survey experiment are, for each job ad, the respondent's beliefs about the benefits not displayed, their chance of getting the job had they applied, the workplace atmosphere in the firm, and the wage after five years.

Job seekers who will visit the job vacancy websites managed by our implementing partner will be randomized in different treatment groups, and some of them will be able to access simple information about wages and job benefits that are associated with the companies posting job ads on the websites.
Jobseekers in the control group will not be shown any extra information other than the one typically displayed on the websites.
Other job seekers will instead be able to see information about the wages that companies posting job ads typically pay for given positions. This information is recovered from a third-party website that collects reviews from the employees of those companies. Specifically, in order to generate random variation in the wages shown across treatment groups, some job seekers are shown average wages while others are shown median wages.
On top of wage information, other job seekers will also be able to see information about some of the job benefits that these companies offer. This information is also recovered from the third-party website. In each of these treatment groups, job seekers are given information about the availability of three benefits (for example, in one group they are given information about flexible working hours, home office, and childcare). Job seekers are shown an icon indicating the presence of a given job benefit in a given company only as long as a given percentage of the reviewers on the third indicated the presence of that benefit. Specifically, in order to generate random variation in the benefits shown across treatment groups, some job seekers are shown that a benefit is present as long as at least 20% of the reviewers indicated its presence, while for others we use a threshold of 50%. Across different treatment groups, we provide information about a total of twelve different benefits (although each job seeker is shown only three): flexible working hours, home office, childcare, parking, public transportation, company car, canteen, food allowance, health services, doctor, coaching, and employee events.

In the belief-elicitation survey experiment, job seekers who will visit the job vacancy websites managed by our implementing partner will see a banner that invites them to participate in a survey. Those who click on the banner, are redirected to a Qualtrics survey. Job seekers who consent to participate are asked to indicate the type of job and the region in Switzerland that they are interested in, and in the following page they are shown 8 real job ads (sampled from ads from the field experiment) customized to the preferences that they just expressed, using the same layout as in the platform. For each of the job ads that they see, they are asked to express their beliefs about the wage paid for that position and the benefits available at that firm, and an additional question about their chance of getting the job had they applied, the workplace atmosphere in the firm, and the wage after five years (we randomize which of these three questions we ask).
As in the field experiment, respondents in the control group will not be shown any extra information other than the one typically displayed on the websites. Other respondents will instead be able to see information about the wages and/or benefits available (the same used for these ads in the field experiment). Moreover, in one of the treatment group, we explicitly inform respondents about the source of the information about wages and benefits.

In the field experiment, the randomization of the job seekers into different treatment groups takes the form of an A/B test conducted with Google Optimize over different variants of the job vacancy platform’s websites.

In the belief-elicitation survey experiment, the randomization is done on Qualtrics.

The intended randomization unit for the field experiment is a job seeker. More specifically, the randomization is done using the users’ cookie identifier detected by Google Analytics.

The randomization unit for the belief-elicitation survey experiment is an individual respondent.

No

300,000 job seekers for the field experiment

2,000 respondents for the belief-elicitation survey experiment

300,000 job seekers 2,000 respondents for the belief-elicitation survey experiment

We will run our experiment over the course of three months. Each month, we expect about 100,000 job seekers to participate in the experiment. Each month, we will randomize job seekers into eight equally-sized groups. Treatment groups will change from month to month according to the schedule below.

A) March
1) Control group
2) Average wage group
3) Median wage group
4) Average wage, flexible working hours, home office, childcare (20% threshold) group
5) Average wage, parking, public transportation, company car (20% threshold) group
6) Average wage, canteen, food allowance, coaching (20% threshold) group
7) Average wage, childcare, health services, doctor (20% threshold) group
8) Average wage, flexible working hours, coaching, employee events (20% threshold) group

B) April
1) Control group
2) Average wage group
3) Median wage group
4) Average wage, flexible working hours, home office, childcare (20% threshold) group
5) Average wage, flexible working hours, home office, childcare (50% threshold) group
6) Median wage, flexible working hours, home office, childcare (20% threshold) group
7) Median wage, flexible working hours, home office, childcare (50% threshold) group
8) Childcare, home office, flexible working hours, average wage (20% threshold) group

C) May
1) Control group
2) Average wage, public transportation, canteen, health services (20% threshold) group
3) Average wage, company car, doctor, employee events (20% threshold) group
4) Average wage, childcare, food allowance, employee events (20% threshold) group
5) Average wage, flexible working hours, home office, parking (20% threshold) group
6) Average wage, home office, parking, company car (20% threshold) group
7) Average wage, public transportation, food allowance, health services, (20% threshold) group
8) Average wage, doctor, canteen, coaching (20% threshold) group

We plan to run our the belief-elicitation survey experiment for at least a couple of months. Each month, we expect about 1,000 job seekers to participate in the survey. We will allocate the sample into five treatment groups as follows:
1) 3/7 Control group
2) 1/7 Average wage group
3) 1/7 Flexible working hours, home office, childcare (20% threshold) group
4) 1/7 Average wage, flexible working hours, home office, childcare (20% threshold) group
5) 1/7 Explicit source + Average wage, flexible working hours, home office, childcare (20% threshold) group