Long-run effects of unemployment training programmes: 5% higher earnings
|Date:||15 December 2022|
Recently, professor Gerard van den Berg and professor Johan Vikström (IFAU Institute and University of Uppsala, Sweden), published an article in Econometrica. The journal belongs to the top 5 of economics journals worldwide and publishes on a range of economic and econometric subjects.
Most important findings
Van den Berg and Vikström developed a new method to estimate long-run effects of treatments, for example participation in an active labor market program for the unemployed. They applied this method to a large Swedish vocational training program and found that the long-run effects are much larger than previously thought. Even after 10 years, average earnings are up to 5% higher than otherwise. This means that the program was actually very beneficial for the unemployed. Older econometric evaluation methods fail to detect this.
They realized that long-run effects of treatments and policies to help poor folks and unemployed individuals may not have been correctly assessed. Depending on the application, this may have severe implications for policy decisions based on older methods. Van den Berg explains: “Our new method is particularly useful for treatments or program participation in “adverse states” such as unemployment or sickness spells or poverty, so there is a very general applicability. As our research was published in the best academic journal in economics in the world, we expect a wide exposure and impact.”
Underestimated effects of training
Van den Berg continues: “Existing studies have under-estimated the positive effects of training for the unemployed. We already knew that the short-run effects on reintegration into work are very small. Because older methods to estimate long-run effects on earnings did not find positive effects either, training for the unemployed got a really bad reputation. People talked about it as an expensive way to keep the unemployed off the streets.
Our results show that, every year, tens of thousands of individuals in Sweden have been helped by the training, not just for a few years but basically throughout their entire subsequent career.”
The difference between the new method van den Berg and Vikström used and previous methods is actually quite subtle. “Suppose you compare long-run outcomes of participants who start a treatment at a certain point in time to the outcomes of individuals who don’t enter the treatment at that moment but who are otherwise comparable. Then you should not ignore that some in the latter group may start the treatment at a later point in time, before the moment they lose their eligibility. The key challenge here is that people who are comparable at some point in time may not be comparable at a later point in time. Our method develops a way of weighting the relevance of the not-treated as “control-group” individuals.”
Their current research was inspired by another research project that van den Berg did before. In that project he studied effects of losing a child (e.g. in a car accident) on long-run economic and health outcomes of the parents. “This is an emotionally challenging topic and the model setting was a bit different. But one thing we learned was that with modern register data it is feasible to study long-run effects at a very detailed level. For this research, we had a dataset of over 700,000 individuals who we could follow up to 15 years. At the time we mostly looked at average effects on the affected families but we did find differences depending on the gender of the child and of the remaining siblings. In future research we aim to dig deeper into that to understand such differences.”
For more information please contact Gerard van den Berg (gerard.van.den.berg rug.nl).