Labour market information & human capital investments: The effect of message framing on training behaviour

The information intervention is with flex workers at a large Dutch employment agency investigating the effect of (framed) labour market information on training intentions and actual training behaviour.

Specifically, we provide (framed) information on the potential outcomes of training, such as the opportunity to deepen or update knowledge, the chance to work in a different occupation, the prospects for a future wage increase, the potential for increased job satisfaction, and the opportunity for a new job.
The information we provide participants is discussed during three focus groups consisting of individuals with similar characteristics as our target group. Moreover, we will run a pilot to pre-test. There will be a control group that receives no labour market information, but do receive an email to inform individuals about the training portal, and a link to the portal, and three treatment groups:

• General labour market information (in a neutral way) on the current changes in the labour market
• General labour market information + information highlighting the benefits of engaging in training (gain-framed message)
• General labour market information + information highlighting the costs of not engaging in training (loss-framed message)

2023-11-05

2024-11-30

We measure individuals training behaviour. More specifically, we will investigate a) individuals’ training intentions and b) actual training behaviour (1, 3 or 6 and 12 months after the information intervention) and c) potential mediating variables. Moreover, we plan to conduct a heterogeneity analyses (d)


a) Training intentions
• Engagement: Clicking behaviour in the information treatment email (on the link to the training portal)
• Training website behavior: how many training-related pages has been visited, time spent on training-related pages, search queries on Randstad website for information about the Randstad training options, the number of applications / registrations for a training
• Self-reported training intentions via an additional survey


b) Training behaviour
• Including registration person (teamleader vs student), registration date, category of course, specification of course, self-chosen (self-planned) vs mandatory (blended, planned), institution providing the courses, costs of the training, subsidy for the training (yes/no)


c) Mediator variables assessed via an additional survey
• Perceived benefits of training
• Perceived need to update skills
• Perceived fear of automation


d) Heterogeneity analysis
• Interaction terms between the treatment dummies and the following personal characteristics will be included: age, gender, employment characteristics.

We seek to understand individuals' job search patterns and outcomes. More specifically, we are interested in whether individuals engage in job search, the frequency with which they view vacancies, the number of times they apply, the number of successful placements and the preferred roles they seek. We aim to pay additional attention to broad job search behavior, job search outside of the current occupation. Earlier studies (Altmann et al., 2018, Belot et al., 2019) have found that labor market information results in a broader job search behavior, however, does not result in a larger probability to receive job offers outside of their own occupation. This could be due to an insufficient set of skills and knowledge needed to be successful in broad job search. While this experiment focuses on the effect of information provision on training behavior, we will also analyse whether the information leads to broad job search and whether the success of this can be explained by training behavior.


Ref:

Altmann, S., Falk, A., Jäger, S., & Zimmermann, F. (2018). Learning about job search: A field experiment with job seekers in Germany. Journal of Public Economics, 164, 33-49.

Belot, M., Kircher, P., & Muller, P. (2019). Providing advice to jobseekers at low cost: An experimental study on online advice. The review of economic studies, 86(4), 1411-1447

The proposed study will take place within a prominent Dutch employment agency and will target a cohort of 80,000 registered flexible workers. The experiment consists of two parts: 1) an information intervention delivered by mail with a link to the training portal, and 2) an additional survey supplementing the labour market information with an additional short survey. Both segments of the experiment have the same experimental set-up, with one control group and three treatment groups.
More specifically, the three treatment groups will look the following way:

• General labour market information (in a neutral way) on the current changes in the labour market
• General labour market information + more specific labour market information highlighting the benefits of engaging in training (gain-framed message)
• General labour market information + more specific labour market information highlighting the costs of not engaging in training (loss-framed message)

Not available

Email distribution will follow a rule-based randomisation, using criteria including odd/even days and months of birth, to ensure the robustness of the experimental design. This rule will guide the allocation of participants to different treatment groups, with individuals being randomly assigned to specific groups within the prescribed rule parameters. Each of the four groups, including one control group and three treatment groups, will comprise 25% of the total population. The groups will remain the same for both studies, the information treatment (email 1) and the survey (email 2). As each individual opens the emails independently, and considering that the employees work in different companies, we expect to avoid spillover effects. Nevertheless, the opening and clicking behaviour of the emails will be closely monitored in order to detect any abnormal patterns that might indicate the occurrence of spillover effects.

Individual level of randomisation

No

Cluster: individual level
The emails will be sent to ca. 80,000 individuals:

Pilot study: 8,000 individuals
Main study: 72,000 individuals

Estimation based on 5% opening of the information treatment emails and/or response to the survey (we only receive data of individuals that opened the email and/or responded to the survey) Pilot study: 400 individuals Main study: 3600 individuals

Full sample that email will be send to ca. 80,000 individuals.

1/4th: control group
1/4th: treatment group 1
1/4th: treatment group 2
1/4th: treatment group 3