Revisiting the causal effect of education on political participation and interest

ABSTRACT

Many studies suggest a relationship between education and political participation, but only some address causality. We add to this by re-examining the German case. For identification, we exploit an exogenous increase in compulsory schooling, and use data from the National Educational Panel Study (NEPS). The data enable analyses that do not rely solely on the conversion of school-leaving qualifications into schooling duration but use the individuals’ actual length of schooling as part of their educational biographies. Our results indicate that the well-known association between education and political participation partially reflects causal effects.

KEYWORDS:

• School reform
• political participation
• IV estimation
• returns to education
• Germany

Acknowledgements

We are grateful to two anonymous referees and the editor for valuable comments that helped to improve the paper. We also thank Jacqueline Kroh and Susanne Elsas for substantial support in the preparation of the biographic data of NEPS, as well as participants of the 2018 Annual Conference of the Scottish Economic Society in Perth, Scotland, the XXVII Meeting of the Economics of Education Association in Barcelona, Spain, the 4th Meeting on Empirical Microeconomics and Applied Microeconometrics in Bamberg, Germany, and the DGS Spring Meeting in Berlin, Germany, for helpful comments and recommendations. All remaining errors are ours.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The datasets analysed during the current study are available in the National Educational Panel Study (NEPS) repository, https://www.neps-data.de/.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Notes

1 In addition, Van de Werfhorst (2014) argues that education aims at preparing individuals for challenges on the labor market, ensuring equal opportunities with respect to the access to education, and sorting individuals into educational tracks according to their interests and talents to ensure optimal production of knowledge and skills.

2 Two more technical concerns are: (1) attenuation bias that results from potential measurement errors in years of education, which may distort the estimates to zero, and (2) social desirability in the interview. If more highly educated individuals are more likely to give – from their perspective – the most appropriate answer, irrespective of actual behavior or underlying attitudes, the relationship between education and political participation may be overestimated (Bernstein, Chadha, and Montjoy 2001).

3 We, however, depart from Helbig and Nikolai (2015) in one case. For Bavaria, our data strongly suggests that the reform was implemented in the school year 1968/1969 (see supplemental material – Figure A1), yet Pischke and von Wachter (2008), and Cygan-Rehm (2022) use the same timing for the reform in Bavaria.

4 In Bavaria, the school year already started in autumn.

5 This paper uses data from the National Educational Panel Study (NEPS): Starting Cohort Adults, doi:10.5157/NEPS:SC6:9.0.1. From 2008 to 2013, NEPS data was collected as part of the Framework Program for the Promotion of Empirical Educational Research funded by the German Federal Ministry of Education and Research (BMBF). As of 2014, NEPS is carried out by the Leibniz Institute for Educational Trajectories (LIfBi) at the University of Bamberg in cooperation with a nationwide network. For further information about the National Educational Panel Study see also Blossfeld and Roßbach (2019) and NEPS Network (2018).

6 Political efficacy can be defined as ‘both the belief that the potential voter can influence what the government does (external efficacy), and the belief that the potential voter has the competence to understand and participate in politics (internal efficacy)’ (Jackson, 1995, 280).

7 The NEPS surveys of this wave were conducted before and after the 2013 federal election, which took place on 22 September 2013. For interviews conducted before the election, we linked turnout in the previous 2009 federal election, and for interviews conducted after September 22, we linked turnout in this election. We account for this by including an additional control indicating whether the survey was conducted before or after the 2013 election. Since we are interested in voting behavior in general and control for other individual characteristics as well as for temporal and regional trends, the two different reference points should not pose a significant problem.

8 We chose this time frame to maintain a sufficient number of observations. We conducted further analyses with narrower time bands, specifically 7 and 4 years before and after the reform, respectively. However, the sample sizes get too small for reliable analyses.

9 This may be the case, for instance, for older individuals who went to school but did not graduate. The latter can also be the case for persons who attend a special school.

10 We lose another 102 observations doing this. We also repeated all estimations with cut-off points at 1,5 years or one year. This does not change our main results, but the coefficients are less precisely estimated because of the smaller sample sizes.

11 In all our regressions, the F-values exceed the conventional weak instrument threshold of 10 for both groups.

12 The first stage coefficients in all our regressions are larger for the whole sample than for the target group. This appears counterintuitive because students in the upper secondary school track where not affected by the reform. To some extent, this may be due to the time frame – 10 years before and after the reform – we have to use to obtain an adequate sample size. It is difficult to say whether social developments, and especially the expansion of education that began in the 1960s, drive this finding, especially since we control for birth cohorts in the regressions. In additional analyses, we find that the first stage coefficients converge at smaller time periods around the reform date. However, the sample sizes are then too small to conduct reliable analyses.

13 It is unclear to us whether the different variations of the schooling duration as shown in Figure 3 cause the differences in the coefficients. However, we will not go into more detail here, as this would go beyond the scope of this paper.

14 In a prior version of our paper, we did not remove the observations that we suppose are likely prone to measurement error and did not find any causal effects for this sample.

15 The results for the larger sample show the same pattern.

16 As mentioned earlier, this can lead to measurement error in the first stage estimations. We perform reduced form estimations and find that the R² value is much lower (R²= 0.0087) compared to the R² values when using the information on the actual place of graduation (R²= 0.0617 for self-reported schooling, and R²= 0.0422 for the generated indicator).

17 This overall picture further does not change if we use individuals’ self-reported schooling duration or if we additionally include the deviations between generated and self-reported schooling. Results for these additional analyses are available upon request.

18 We do not provide these results here, but they are available upon request.