Direkt zum Inhalt

3. Juni 2015

SOEP Brown Bag Seminar

Trends of unemployment scarring over time in Germany

Termin

3. Juni 2015
13:00 - 14:00

Ort

Eleanor-Dulles-Raum
DIW Berlin im Quartier 110
Room 5.2.010
Mohrenstraße 58
10117 Berlin

Sprecher*innen

Martina Dieckhoff (WZB Berlin)

To date, we find abundant evidence that unemployment leads to reduced wages upon re-employment (e.g. Arulampalam, 2001; Burda and Mertens, 2001; Gangl, 2006) and also negatively affects other job quality outcomes (Brand, 2006). We still lack a systematic account of scarring trends over time, however. One central goal of our contribution is to fill this gap and investigate how scar effects in Germany have evolved over time and to unveil the institutional and macro-economic mechanisms behind observed trends. Moreover, existing work has rarely considered effect heterogeneity in unemployment scarring and changing compositions of the unemployed over time. By not considering these, substantial parts of the observed temporal trends with respect to average scar effects could be due to changes in the composition of the unemployed. By contrast, in our study we deliberately investigate the role of composition in explaining change over time. We believe that our focus on time trends on the one hand and on compositional change on the other, will provide us with the new opportunity to test existing theoretical explanations for unemployment scarring and thus has the potential to offer important insights into the mechanisms that drive cumulative disadvantage following unemployment.

Using data from the German Socio-economic Panel 1985-2013 we estimate the effects of a recent unemployment experience on subsequent employment outcomes. In particular, we focus on the scar effects of unemployment on reemployment probabilities and (conditional on being reemployed) on subsequent wages and occupational status. The longitudinal nature of the data affords us with the possibility to examine the employment outcomes for workers who experience a spell of unemployment and compare them to those of (otherwise similar) workers who did not experience unemployment. Based on this empirical approach, one can extract the average re-employment probabilities as well as the wage and occupational status penalties that are associated with unemployment and investigate the changing nature of unemployment scarring over time. Furthermore, we are able to test for change in compositional effects over time. By so doing, some more light is shed on the important issue of heterogeneous effects of unemployment, which clearly goes beyond looking at average treatment effects. We achieve these goals of the empirical analyses by combining a new matching method called ‘Coarsened Exact Matching’ (CEM) proposed by Iacus, King & Porro (2008) with the difference-in-differences approach (Heckman et al., 2007). This analytical strategy allows us to control for selection based on observable and unobservable characteristics. We are estimating the ‘treatment effects’ of unemployment covering an observational period of more than 25 years.

Our preliminary results confirm the findings of prior work that unemployment leaves substantial scars. Scars are not only observed shortly after the unemployment spell, but also in the mid-term (4 years after the unemployment incidence) suggesting that the labour market disadvantage caused by unemployment is rather persistent. With respect to the evolution of these scar effects we show that there is no secular trend over the time period under study. However, there exists considerable temporal variation in the size of scarring effects. The specific pattern of this temporal variation can be partly brought in line with the business cycle, but parts of the observed temporal effects are also to be explained by fundamental reforms of the German unemployment benefit system. Finally, we investigate the role of compositional effects. Drawing on the idea of decomposition techniques and using a reweighting approach, we can examine whether we would observe the same scarring trends if our population had stayed the same. Our findings suggest that in some periods the estimation of average treatment effects is affected by compositional change. In particular, while in the mid-90s the specific composition of the unemployed at that time leads to an estimation of larger average treatment effects regarding wages, from the mid-2000s the opposite is the case. A detailed inspection of these specific time periods suggest that this is mainly due to an over-proportional share of high-skilled workers and men, in the mid-90s, as well as their underrepresentation from the mid-2000s. This result underlines the necessity to consider effect heterogeneity as well as changing compositions when investigating the scarring effects of unemployment.
(Paper joint with Johannes Giesecke)

Kontakt

keyboard_arrow_up