Higher and higher? Performance pay and wage inequality in Germany

Abstract

Performance pay is of growing importance to the wage structure as it applies to a rising share of employees. At the same time, wage dispersion is growing continuously. This leads to the question of how is the growing use of performance pay schemes related to the increase in wage inequality? German Socio-Economic Panel (SOEP) data for the years 1984–2009 confirm the large increase in the application of performance pay schemes. This in turn led to an upward shift of the wage distribution by about one log point. However, it did not contribute to the growth in wage inequality. Even though wage inequality grew within the group of employees who receive performance pay, it grew even more so within the group who do not. Still, the wage difference between both wage schemes remained flat over the distribution. The empirical analysis employs sequential decompositions in a quantile regression framework.

Keywords:

• performance pay
• wage structure
• quantile regression
• sequential decomposition

JEL Classification:

• J31
• J33
• C21

Acknowledgements

I am grateful to Juan J. Dolado, Bernd Fitzenberger and Marie Paul for fruitful discussions. For comments and suggestions, I thank seminar participants at the University Carlos III de Madrid and Freiburg University as well as at the XV. Encuentro de Economía Aplicada en A Coruña and the 10th International German Socio-Economic Panel User Conference. Further, I thank the scientific committee of the SOEP User Conference for choosing this article for the Joachim R. Frick Memorial Prize 2012 for the best paper presented at the conference. This research was started when the author was visiting Universidad Carlos III de Madrid. Financial support from the University of Freiburg and the German Academic Exchange Service (DAAD) is gratefully acknowledged. All errors remain my own responsibility. An additional online appendix to this article is available at https://www.empiwifo.uni-freiburg.de/discussion-papers-1.

Notes

1. The longest panel study on performance pay for Germany is provided by Pannenberg and Spiess (2009) for the period from 1991 to 2000. However, they do not model the contribution of the growing use of performance pay to growing wage dispersion.

2. Performance pay plays a special role in Germany, given the background of the strong German system of industrial relations (Jirjahn, 2002, p. 158). Compared to collectively negotiated wages, performance pay is more flexible. Therefore, it was seen as a way to increase the competitiveness of German firms and thus to reduce unemployment (Jirjahn, 2002, p. 163).

3. The most recent available wave at the time of writing is from 2010, which refers to pay components in the year 2009.

4. See https://www.empiwifo.uni-freiburg.de/discussion-papers-1.

5. Qualitatively similar results are obtained from different subsamples of the data (available upon request).

6. Another question in the Socio-Economic Panel (SOEP) data explicitly asks for performance evaluations by the supervisor in the years 2004 and 2008. According to this, the share of employees whose performance was evaluated in the year 2004 ranges between 25% (Cornelißen et al., 2011) and 31% (Grund and Sliwka, 2010), depending on the exact specification of the data set. On the firm level, Berger et al. (2011) report that 37% of all firms use performance-related pay.

7. Wages are defined as real log hourly wages.

8. In the following, the notation will abbreviate these time periods, mentioning only the starting year.

9. Further explanation to the sequential decomposition can be found in Fortin et al. (2010). This method has been applied, e.g. by Antonczyk et al. (2009, 2010).

10. The applications of the decomposition method from DiNardo et al. (1996) in Lemieux et al. (2009) and Heywood and Parent (2009) decompose the wage difference between performance pay and non-performance pay jobs. As an intermediate step, they simulate a wage distribution in the absence of performance pay jobs. This step is avoided here because the changes in the wage distribution over time are modelled instead. As an intermediate step, the share of PP jobs is held constant at 1986 levels – not at zero as in DiNardo et al. (1996).

11. How the coefficients in the PP job and in the non-PP job scheme evolved will be considered in the additional decomposition results later in this section.

12. Alternative orders of decomposition will be considered in the sensitivity checks in the web appendix.

13. An alternative matching procedure will be explained as a sensitivity check in the web appendix.

14. Inference is based on 100 bootstrap replications, applying a block cluster bootstrap where individuals are resampled and using all observations over time for the resampled individuals.

15. Here, following Chernozhukov et al. (2008), the predicted wages from the quantile regressions form the basis. Because this smoothes out the error term, the total difference is not as erratic as in Fig. 1. This does not change the results (available upon request).

16. Collective bargaining coverage cannot be identified from the data and therefore cannot be investigated further in this study.

17. This result is very robust to variations in the estimation procedure, see the web appendix.

18. Pannenberg and Spiess (2009) also find that over the period from 1991 to 2000, wage inequality as measured by the coefficient of variation increased within both pay regimes. However, they do not differentiate between the coefficients and characteristics effect.

19. The decomposition for the years 1986–1989 follows analogously.