Do constitutions matter? Evidence from a natural experiment at the municipality level

In the following, I will mostly mention direct elections only (and ignore that this goes along with the fact that the mayor can be recalled only by referendum) to improve readability.

However, some restrictions apply (see Sect. 3).

Points relate to a multiplier on the tax base where 100 points represent a multiplier of one.

As rightly pointed out by an anonymous referee it might be that a small change in tax rate leads to significant changes in revenues if the tax base is large. However, the data shows that the associated changes in revenues are very small: given that in the dataset the tax on non-agricultural land represents just over 5 % of the municipalities’ total revenues, a decrease of 0.8 % in tax rates (i.e., by 2.06 points) will result in a ceteris paribus decrease in total revenues of less than 0.05 %.

The section of Blume’s book (2009) that is relevant to this research was also published as a paper (Blume et al. 2008).

The number of observations is between 29 and 112, depending on the regression.

This could be either the mayor elected by the council or the municipality’s manager, depending on the state.

Often mayors did not serve the full term which then led the community to have their following mayor terms start at different dates than previously planned. As more than 40 years had already passed since the end of the war when the introduction of the direct mayoral election was decided, this had occurred in many municipalities. Furthermore, the mergers and reforms of municipalities in the 1970s led to a further misalignment in term start and end dates across municipalities. Finally, in Rhineland-Palatinate the term length for full time mayors could vary from ten to 12 years. It remains to note that there are several reasons why a mayor might not serve a full term: death, choice of different career path (e.g., mayor Hans-Jürgen Boebnik of Saarbrücken resigned in 1999 after six years (instead of 10) to become head of the reserve bank of Saarland) or dismissal (e.g., mayor Hans-Joachim Jentsch of Wiesbaden was recalled in 1985 after majorities in the city council had changed).

Within the classes of presidential and parliamentary systems, there are large differences. For example, the French President has different (veto-)powers than the US President. The definition proposed by Persson and Tabellini (2002) thus focuses on what the authors regard as the key differences (i.e., confidence requirement and separation of powers) by accepting that different political systems can be ordered into these two classes. This also applies to the municipal constitutions of the three German states considered in this paper. Not only are the systems in these states different from those of countries, but they also differ between states (see Appendix A). Nevertheless, following the definition of Persson and Tabellini (2002), they can be roughly described as parliamentary before and as presidential after the reform.

Data on the first mayoral elections could not be obtained for the states of Lower-Saxony and Schleswig-Holstein, which also phase in direct mayoral elections.

An extended version of this dataset also included data for the 2256 member municipalities of joint municipalities, which was analyzed as a sensitivity check.

In Rhineland-Palatinate, the member municipalities of the joint municipalities are in charge of setting tax rates. For this reason, joint municipalities are excluded from the tax analysis. This leaves us with 49 municipalities in Rhineland-Palatinate for the tax analysis. Within this group, the first mayoral elections took place between 1994 and 2001.

For example, the tax amount for a house is based on the value of the house. This value is multiplied by a set number to identify the tax base. This base is then multiplied by the rates that municipalities set.

Points relate to a multiplier on the tax base where 100 points represent a multiplier of one.

I also estimated specification (6) excluding the economic strength variables for the trade tax to cover for the potential endogeneity of the economic strength variable. There was only a small change in the other parameters estimated. I did the same for specification (6) in Table 5 and also detected only a marginal change.

As pointed out by an anonymous referee, it could also theoretically be that a directly elected mayor needs more staff than one elected through the municipal council (e.g., because of greater responsibilities) or might be entitled to a higher salary. Given the institutional circumstances, these arguments can be ruled out, as the post filled by the directly elected mayor existed before the reform and his or her responsibilities were unchanged; the only rule that was altered was the way in which the mayor is appointed (and recalled where applicable).

Population density decreases total spending per capita, thus total spending in urban areas is lower. Communities with a larger tax base (and hence a high value of economic strength) spend significantly more. The share of the population under 15 has a strong positive effect on total spending. A one percentage point increase in this age group increases total spending by 23.90 euros. The share of the population over 65 does not have a significant influence.

A one percentage point increase in the share of the older population increases personnel expenditure by 2.66 euros per capita, while a one percentage point higher younger population reduces this spending by 3.14 euros.

I am grateful to an anonymous referee for pointing out that the adjustments discussed in this section might be driven by the effects of the political business cycle. To check for such business cycle effects, I ran an additional analysis using the second mayoral election after the constitutional switch. In this analysis, the dummies for the years around the election capture the effects of the political business cycle only. The analysis shows that the pre- and post-election adjustment effects for the second direct election are different from those for the first election. When estimated in one equation, either the adjustment effects for the second election are not significant when those for the first are or vice versa and/or of a different sign. This shows that the effects reported here are strongly influenced by the constitutional switch. However, interpretation should not focus on the precise magnitudes of the estimates but rather on the signs.

In municipalities in which no candidate seeks the position of the mayor in a public election, the mayor is elected by the council.

As for the smaller sample, I also estimated specification (6) excluding economic strength for the trade tax to control for the potential endogeneity of this variable. There was only a small change in the other parameters estimated. I did the same for specification (6) in Table 8 and detected a marginal change only.

In 1998, 12 municipalities had more than 50,000 inhabitants.