What Drives Reskilling Decisions? Evidence from a Discrete Choice Experiment with Unemployed Jobseekers

This is a survey experiment consisting of five parts:
1. Occupational preferences
2. Training decisions under different scenarios with varying training characteristics
3. Beliefs about preferred occupations
4. Training decisions under different scenarios with varying job-related characteristics
5. Additional questions
The experimental design is described in detail in the pre-analysis plan.

2024-11-18

2025-06-30

Our primary outcomes are meant to capture interest in undertaking occupational training programs for in-demand occupations (interest in reskilling). We want to explore how interest in reskilling varies depending on i) the features of the training program and/or ii) the characteristics of the job the training prepares participants for, and/or iii) jobseekers demographics, work history and beliefs.

We will define interest in reskilling by using respondents' answers to the different discrete-choice scenarios they see.

First, we will compute two main variables of interest in reskilling: (i) the share of scenarios in which a respondent picked a training course in a shortage occupation as first choice, and (ii) whether they ever picked a training course in a shortage occupation as first choice in any of the given scenarios. As the choice scenarios are split into two sets - one of 5 choice scenarios (which vary the occupation targeted by the training) and one of 3 choice scenarios (which keep the targeted occupation) - we will define these reskilling variables by pooling together all the choices, as well as separately for the two sets of scenarios.

Second, we will also consider an alternative outcome encompassing “interest in trainings” more broadly. This measure will be defined as: (i) the share of scenarios in which a respondent picked any training course, and (ii) the respondent having ever picked any training course as first choice in any of the given scenarios. However, Given the hypothetical nature of our experiment, we expect fewer people to pick the no training option compared to reality, and thus training interest to be biased upward. We will therefore pay particular attention to estimates pertaining to interest in reskilling, and consider interest in training as rather exploratory.

Third, in our discrete choice model pertaining to training characteristics (for the set of 5 scenarios), we will quantify the increase in marginal utility and the willingness to pay for reskilling by considering as the "reskilling" option the training course which targets a respondent's (less preferred) shortage occupation.

Finally, in our discrete choice model pertaining to job-related characteristics (for the set of 3 scenarios), we will fix the occupation to the second-preferred, and estimate the willingness-to-pay for a set of job-related characteristics. The estimates will be benchmarked against our WTP measures for reskilling found in the previous paragraph.

A. Random variation in information on labor shortage
In order to study possible frictions to the demand for training, we introduced experimental variation within the survey right before the discrete choice experiment. Respondents are randomized into the following two groups:
1. Shortage Information Group: before asking for their training choices, respondents are informed that the occupation for training B (one of the training options) is in shortage.
2. Shortage Control group: no additional information is provided before training choices.

B. Random variation in training (and job-related) characteristics
The discrete choice experiment contains random variation within and between subjects. Within subjects, we randomized the characteristics of training programs. Between subjects, we randomized the block of choices seen by a person.

C. Random variation in information on employer involvement in trainings
At the end of the survey, jobseekers receive information on trainings that seem to match their preferences. For a randomly selected half of respondents, we mention that some of these trainings are organized in collaboration with employers.
1. Employer Involvement Information Group: respondents are told that some of these trainings are organized with employers.
2. Employer Involvement Control group: respondents are not told that some of these trainings are organized with employers.

Not available

Randomization is done within the Qualtrics survey.

Individual respondents.

No

Target= 3,000 individual respondents.

Target= 3,000 individual respondents.

Half of respondents will be in the shortage information treatment (target: 1,500 treated, 1,500 controls).
An orthogonal half of respondents will be allocated to the employer involvement information treatment (target: 1,500 treated, 1,500 controls).
In the discrete choice with varying training (resp. job-related) characteristics, respondents are allocated to one of three (resp. four) blocs of 5 scenarios with given sets of characteristics.

See pre-analysis plan.

PAP_SurveyExperiment-Forem.pdf