Information Frictions in Job Descriptions

I partner with a large Indian job board to examine the preferences of first-time jobseekers over amenities that are typically missing from job posts. First-time jobseekers are shown hypothetical job posts with randomized amenities. These amenities include those mandatory for recruiters to report on the partner platform (i.e. salary, work from office vs. work from home, etc.) and typically “missing” amenities, described below.

2023-06-22

2023-09-30

The goal is to measure the effect of including typically "missing" amenities on hypothetical job posts on jobseekers’ responses to the question:

- Would you apply to this job? [1 = Yes, 0 = No]

All jobseekers will answer this question for each hypothetical job they choose to assess.

A random subset of jobseekers will answer the following questions for each hypothetical job they choose to assess:

- If you applied, do you think HR would give you an offer? [1 = Yes, 0 = No]
- If HR is ready to give you a job, how interested are you in this job? [1 = Not interested, 10 = Very interested]

For these jobseekers, we will also construct a summary index of their responses to all 3 questions.

First-time jobseekers who are active on the partner platform are sent a message with a survey link. They are asked to assess hypothetical jobs with randomized amenities, and in return, they will receive job recommendations that reflect their preferences as measured through their assessments of the hypothetical jobs. Upon opening the survey, jobseekers are equally randomized to view hypothetical job posts with one of the following subsets of missing amenities:

- Working individually or in a team, job training
- 5, 5.5, or 6 day work week, company-provided transportation or travel allowance
- Good company culture, no salary delays, company verification

On each hypothetical job post, the typically “missing” amenities are randomized to be displayed or not. For some amenities, conditional on being displayed, the value is randomized. Amenities that are mandatory to report on the partner platform are always displayed and their value is randomized.

Jobseekers are equally randomized to be asked 1 or 3 questions for each hypothetical job, as described in “Primary Outcomes” and "Secondary Outcomes". They are also equally randomized to view the typical “missing” amenities before or after the “Job Description” section of the hypothetical job post.

Jobseekers randomized to view hypothetical jobs with good company culture, no salary delays, and company verification are also equally randomized to receive:

- No additional instructions
- An explanation that these characteristics have been checked with employees
- An explanation that these characteristics have been checked with employees plus a blue tick verification symbol

Not available

Randomization is conducted by the survey software.

Randomization occurs at both the individual level and for each hypothetical job post:

Individual level:

- Subset of typically missing amenities to view on posts:
-- Working individually or in a team, job training: 1/3
-- 5, 5.5, or 6 day work week, company-provided transportation or travel allowance: 1/3
-- Good company culture, no salary delays, company verification: 1/3, crossed with an equal randomization across
--- No additional instructions
--- An explanation that these characteristics have been checked with employees
--- An explanation that these characteristics have been checked with employees plus a blue tick verification symbol

- For each hypothetical job, answering:
-- 1 question: 1/2
-- 3 questions: 1/2

- Order of the hypothetical job post content:
-- View the typical “missing” amenities before “Job Description” section: 1/2
-- View the typical “missing” amenities after “Job Description” section: 1/2

For each hypothetical job post:
- Amenities that are mandatory for recruiters to report on job post: value is randomized
- Whether or not typically missing amenities are displayed
- Conditional on being displayed, the value that some of the typically missing amenities take.

Analyses conducted at the individual-hypothetical job level will be clustered at the individual level.

Yes

Because the number of clusters depends on jobseeker response rates to the survey, it is not possible to predict the exact number of clusters. However, the target is 3,000 first-time jobseekers who assess at least 1 hypothetical job.

Update 2023_8_31: During the data collection of the initial 3,000 first-time jobseekers, a filtering issue was discovered in the sample generation procedures that unintentionally excluded a large portion of first-time jobseekers who fit the study criteria, but had recently joined the partner platform. The filtering issue was partially solved in the final survey distribution of the main sample (survey wave 8), which nearly doubled the number of jobseekers who were distributed the survey. The jobseekers who were no longer erroneously excluded assessed at least 1 hypothetical job at nearly double the average rate of the previous 7 survey distributions. With the greater sample size of jobseekers who were sent the survey plus higher response rates, the target number of clusters was surpassed.

In the main sample, we also found heterogeneity in the preferences of first-time jobseekers by the amount of time since joining the partner platform. To further explore this heterogeneity and because a large portion of recent joiners were erroneously excluded, we will collect a follow-up sample of first-time jobseekers who had joined the platform within 60 days of the survey distribution. The target number of clusters for this follow-up sample is 2,000 first-time jobseekers who assess at least 1 hypothetical job. All other study conditions remain the same. The survey distribution for this follow-up sample is scheduled to start today (August 31, 2023). At the time of this update, no surveys for this follow-up sample have been distributed.

Because the sample size depends on how many hypothetical jobs each jobseeker chooses to assess, it is not possible to predict the exact number of observations. Each jobseeker will rate up to 20 hypothetical jobs.

It is not possible to predict the exact number of individuals within each survey condition in advance, but jobseekers who open the survey are randomized to conditions at the individual level as described in section “Randomization Unit.”