Job Search - an Information Experiment

n/a

We conduct a field experiment on a large Chinese online job board. The intervention takes place through an online survey (our main survey) that are pushed to job seekers through in-app messages. Participation is voluntary but compensated, which is common for online surveys carried out on the job board. The survey elicits their intended occupation, their beliefs (and their confidence in such beliefs) about the number of applications per vacancy in this occupation in the past month (referred to as “past prior beliefs” hereafter) as well as in the next month (referred to as “future prior beliefs” hereafter).

Then, the survey implements a 2x2 design that varies whether information about last month’s labor market competition is provided (info-treatment) and whether daily goals about the number of jobs they plan to apply for in the upcoming month are elicited (goal-treatment). Finally, future beliefs (referred to as “future posterior beliefs” hereafter) are once again elicited after information provision and/or goal setting. They are also elicited again in the no-information treatments to control for repeated elicitation effects.

At the time of writing this pre-registry file, it also appears that the job board has upgraded its administrative data recording system so that we may have access to further behavioral data besides what job ads they apply for that encompasses what job ads they open (but do not apply for), including the date/time they do so. We may also have access to the date, start and end time when job seekers use the app and, on the other side of the labor market, whether employers contact job seekers in response to their application via the internal message system. However, the availability and quality of this data are still uncertain.*

A month after the main survey has been completed by all participants, we conduct an online follow-up survey, which is pushed to all previous survey respondents through app message and/or cellphone text message, to collect their job search outcomes. However, this additional dataset is likely to be relatively small due to attrition and its purpose is mainly to provide robustness checks and additional supportive evidence for the main analysis.

*We had collaborated with the job-board in a previous study (He, Neumark and Weng, 2021). Since then, the job board has introduced novel new features and increased the tracking of their users. How viable some of these new measures will be is unclear at this point of time. We will comment on the potential analysis in our analysis plan below.

2021-09-13

2021-12-31

- Future posterior beliefs: beliefs about the level of competition in their occupation after the intervention.

- Search effort: number of applications made following the intervention, number of job ads read but not applied to (if data from job board is available), time spent searching for jobs on the job board (if data available).

n/a

- Goals: intended daily number of applications.

- Reservation wage: proxied by the wage offers of jobs that the job seeker applies for.

- Occupation switching: proportion of applications made in the intended occupation, etc.

- Job search outcomes: employers’ responses to job seekers’ application (if data from job board is available); number of interviews, number of offers, whether a job was accepted, accepted wage, etc. (elicited in the follow-up survey and subject to strong attrition).

n/a

We conduct a field experiment on a large Chinese online job board and analyze the effect of information provision and goal setting on job seekers’ search behavior.

Experimental setting – the online job board and job seekers:
We conduct the experiment on an online job board, which is one of the largest nationwide online job boards in China. The job board posts tens of millions of job openings per year with more than 100million registered job seekers, of whom millions are active each day. The job board specializes in white-collar, high-education jobs, and job seekers are highly educated. Employers post job ads that are produced using a standard template and that list important information about the job and employee requirements.

In order to search and apply for jobs on the job board, job seekers need to first register and provide individual information to construct a standardized résumé. The required information includes, among other things, gender, birth date, location, education and work/internship experience, current employment status, and type of job sought with location, intended industry, occupation and salary range specified.
As researchers, we have full access to the job seekers’ (anonymized) résumé (extracted from the administrative database at the time the survey is completed) as well as for which jobs they apply for both one month before and one month after the survey (extracted from the administrative database after one month the survey is completed).

After registering and logging in, job seekers can click to open job ads that are listed on employers’ web pages or presented to them after searching the system using a search bar. When they open an ad, they see a full-page description and can then click the “Send Résumé” button to send their pre- generated résumé to the employer to apply for that job. The employer receives the résumés immediately. At the same time, a chat box is created so that the employer and the job seeker can interact freely with each other if they would like to. However, it is common that most further interactions take place outside of the job board. Consequently, it is unlikely that we learn (via the job board) whether a job seeker receives an interview or job offer and whether such an offer is accepted.


Sampling:

Our surveys are run in cooperation with the job board and features their official company name, which is shown to job seekers. We define a job seeker as active if she applied for at least one job on the job board yesterday. On the first day of the survey, the job board draws the population of all active job seekers of that day and pushes the main survey through an in-app message to all of them. On subsequent days, the job board redraws the population of all active job seekers and pushes the survey to those eligible. The survey will be terminated when the sample size reaches our target.


Treatments:

For an explanation of the treatments, please refer to the intervention section above.

Randomization is done by the online survey system. We stratify based on job seekers intended occupation (elicited as the first question of the survey).

The unit of randomization is at the individual job-seeker level.

Yes

We cluster by intended occupations. The planned number of clusters is 54.

20 000 job seekers

20 000/4

As the sample is large enough, power for economically significant effects is not a concern. But if needed, we will do so later after the data is collected and standard deviations are known.