|
Field
Experimental Design (Public)
|
Before
I will invite 1000 job seekers who recently created or updated their resumes to take part in an incentivized survey. The survey will be framed as a pilot tool to help the platform tailor job openings to job seekers. We will inform the participants that the more accurate the answer they give the better the outcome of the job recommendation the platform can cater to them. We will first collect their resumes which include baseline characteristics such as demographics, past employment, expected salary for their next job, etc. After collecting baseline information, all participants will be entered into two modules where the sequence is random. The first module is the belief elicitation, and the second module is the discrete choice experiment.
Participants will be randomized into four treatment arms where I randomize the order of the modules and when information about young firms’ exit rates is shown to participants.
1. A randomly selected one-fourth will be in the control group 1 where we elicit the belief about young firms before they enter the DCE experiment. This arm will not receive a true distribution of young firms’ exit rates
2. A randomly selected one-fourth will be in the control group 2 where we elicit the belief about young firms after they finish the DCE experiment. This arm will not receive information about the true distribution of young firms’ exit rates.
3. A randomly selected one-fourth will be in treatment group 1 where we elicit the belief about young firms before they enter the DCE experiment. This arm will receive information about the true distribution of young firms’ exit rates after they finish the DCE experiment.
4. A randomly selected one-fourth will be in treatment group 2 where we elicit the belief about young firms before they enter the DCE experiment. This arm will receive information about the true distribution of young firms’ exit rates immediately after they finish the survey model and before they enter the DCE experiment.
The information about firms’ exit rates will be first presented in a simple message about the true exit rate of young firms in Thailand. Additionally, we will show dashboard information about firms’ exit rates in Thailand where participants can choose different parameters such as the firm’s location, industry, initial registered capital, etc. I will also send this information—or link to the dashboard—to their emails where they can always come back and see this information throughout the study period.
One week after they finish the survey, I will send out a list of real job ads that match the preferences of the DCE to job seekers. In this part of the experiment, treatment arm 1 and treatment arm 2 where no information about young firms’ exit rate will be our control group and treatment arm 3 and 4 where the information provision is given will be our treatment group. The recommendation list in one email will contain 10 job adverts: 4 young firms and 4 old firms that match the preferences extracted from the DCE and the other 2 random ads. The purpose of 2 random job ads is to test the validity of our DCE experiment.
During the email experiment phase, I will introduce another variation in the 2X2 design where I will include (public) accounting information about firms such as sale growth, interest coverage ratio, debt ratio, profitability, etc. Job ads that will be sent to participants will be randomized and stratified such that across treatment groups job ads are similar at the same time match the individual preferences of the job seekers.
|
After
Experimental Design
I will invite 800 job seekers who recently created or updated their resumes to take part in an incentivized survey.
The survey will be framed as a pilot tool to help the platform tailor job openings to job seekers.
We will inform the participants that the more accurate the answer they give the better the outcome of the job recommendation the platform can cater to them.
We will first collect their resumes which include baseline characteristics such as demographics, past employment, expected salary for their next job, etc.
After collecting baseline information, all participants will be entered into two modules where the sequence is random.
The first module is the belief elicitation, and the second module is the discrete choice experiment.
Participants will be randomized into two treatment arms where I randomize the order of the modules and when information about young firms’ exit rates is shown to participants.
1. A randomly selected one-half will be in the active control group where information about the firms' exit rate in Thailand from the Townsend Thai data is shown.
2. A randomly selected one-half will be in the treatment group where information about the firms' exit rate in Thailand from the department of business development (estimated by the authot) is shown
The information about firms’ exit rates will be first presented in a simple message about the true exit rate of young firms in Thailand with a breakdown of exit/suvival rate of each year after the birth of the firm. Additionally, we will show dashboard information about firms’ exit rates in Thailand where participants can choose different parameters such as the firm’s location, industry, initial registered capital, etc. I will also send this information—or link to the dashboard—to their emails where they can always come back and see this information throughout the study period.
One week after they finish the survey, I will send out a list of real job ads that match the preferences of the DCE to job seekers. This is the incentive mechanism that would align participants interests to the reseacher's goal. In this stage, I may randomly allocate both ads that match the preferences and ads that don't match the preference (from the DCE estimation) depending on the feasibility of the project (the constraint of the platform)
|