Nudging Disability Insurance Recipients into Partial Employment

The previous literature has brought consistent evidence that many Disability Insurance (DI) bene ciaries retain a considerable capacity to work (Autor & Duggan, 2006; Bound, 1989; French & Song, 2014; Maestas, Mullen, & Strand, 2013). In light of this, governments in several countries have developed return-to-work programs that allow DI recipients to combine benefi ts and work income. Kostol & Mogstad (2014) have shown that this type of policy reduces the cost for social security while enhancing the welfare of DI recipients.

Such program has been implemented in Belgium in 1996 offering DI recipients to return to work on a part-time basis and keep a share of their bene fits. Participation remains, however, relatively low with only 11% of long-term DI recipients (NIHDI, 2019). Among the reasons for low participation, our discussions with stakeholders have highlighted the lack of accessible information about the program rules and incentives. The most motivated individuals would find the information needed after extensive research on their mutuality fund's website, only to discover that it is of relatively high complexity and often described in juridical terms. This echoes the findings of Bhargava and Manoli (2015) who showed that informational complexity is an important contributor to the puzzle of low participation in government-sponsored programs.

With that in mind, we designed a low-cost informational intervention in collaboration with the Belgian National Institute for Health and Disability Insurance (NIHDI). A nudge was added to an existing communication addressed to all benefi ciaries upon their entry to long-term disability (i.e. after 12 months on DI).

This type of informational intervention has been shown to be a cost-effective way to increase program participation in other contexts (Bhargava & Manoli, 2015; Liebman & Luttmer, 2015). For sickness and disability insurance, we are only aware of two other studies that implement nudges in the form of information meetings organized by the social security administration (Johansson & Lindahl, 2013; Markussen, Roed, & Schreiner, 2018). In comparison, our nudge within an existing communication and with the help of an online calculator should be highly cost-effective since it does not involve case-workers or require in-person meetings.

2021-01-13

2022-12-31

The data used to construct the outcomes will be drawn from administrative registers of the NIHDI. The main outcomes of interest include:
- Participation in return-to-work program (binary);
- Elapsed time between entry to long-term disability and start of return-to-work program (# months), conditional on participating in the latter;
- Volume of work (% of full-time workload), conditional on participating in the return-to-work program;
- Type of employment (employee or self-employed), conditional on participating in the return-to-work program;
- Quarterly amount of DI bene fits (euros);
- Time on long-term disability (# months since entry);
- Exit from DI (binary).

The evaluation will rely on a randomized experiment that shall start in early 2021. From that date on, the NIHDI will send one of the three types of letters to the benefi ciaries upon their entry into long-term disability (i.e. after 12 months on DI). The entire mailing process is completely automatized. A dedicated machine prints, folds and inserts the letters in pre-addressed envelopes. Thus, workers at NIHDI are fully blind with regards to the random allocation.

Randomization will be conducted based on the last two digits of the Belgian national identity number of the benefi ciary who is entering long-term disability. One-third of the new entrants (national id numbers ending in 01-33) will receive the old version of the letter that was used so far by the administration (control group, C). One-third of the new entrants (national id numbers ending in 34-65) will receive the new letter with the fi rst version of the nudge (treatment T1). One-third of the new entrants (national id numbers ending in 66-97) will receive the new letter with the second version of the nudge (treatment T2). All treatment and control groups are therefore of equal size in terms of number of letters sent out.

Randomization will be performed at the individual level.

Yes

The experiment will comprise three clusters, one for each Belgian community: Dutch-speaking, French-speaking and German-speaking.

We aim to treat the universe of new bene ficiaries who enter long-term disability during the year 2021. Our partners at NIHDI have indicated that the number of new entrants should approximate 60,000, that is 20,000 observations per treatment arm and 20,000 in the control group.

Our partners at NIHDI have indicated that the number of new entrants should approximate 60,000, that is 20,000 observations per treatment arm and 20,000 in the control group.

Our simulations show that a sample size such as the one discussed above would be suficient to detect a statistically signi cant effect of the nudge on the probability to participate in the return-to-work program of 1 percentage point.