Public administrators as politicians in office

Hessami, Zohal; Häcker, Timo; Thomas, Maximilian Article — Published Version Public administrators as politicians in office International Tax and Public Finance Suggested Citation: Hessami, Zohal; Häcker, Timo; Thomas, Maximilian (2025) : Public administrators as politicians in office, International Tax and Public Finance, ISSN 1573-6970, Springer US, New York, Vol. 32, Iss. 6, pp. 1960-1987, https://doi.org/10.1007/s10797-025-09882-z This Version is available at: https://hdl.handle.net/10419/333227 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by-nc-nd/4.0/ Vol:.(1234567890) International Tax and Public Finance (2025) 32:1960–1987 https://doi.org/10.1007/s10797-025-09882-z Public administrators aspoliticians inoffice ZohalHessami1· TimoHäcker2· MaximilianThomas3 Accepted: 17 January 2025 / Published online: 30 January 2025 © The Author(s) 2025, corrected publication 2025 Abstract We analyze whether mayors’ prior occupation in the local public administration matters for their performance. In theory, mayors’ professional background may shape their competence in bureaucratic tasks. We use the example of grant receipts for visible investment projects for which mayors must submit an extensive application to the state government. Our dataset includes 1,933 mayor elections (1993-2020) in the German state of Hesse to which we apply a sharp RD design for close mixedbackground races. Mayors’ background on average has no effect on grant receipts. Yet, public administrator mayors do attract more grants than outsider mayors when they are ideologically aligned with the council, raising the motivation to apply for grants in the first place. We conclude that the competence of public administrator mayors only matters when they are motivated to use it, i.e. this is an example where incentives are necessary for the effects of political selection to materialize. Keywords Public administration· Professional background· Administrative skills· Political selection· Mayors· Investment grants JEL Classification D73· D78· H71· H77· H83 1 Introduction What does it take to be a successful mayor? A growing literature provides insights on the effect of various politician characteristics such as age, gender or education on policy outcomes (Alesina etal., 2018; Baskaran & Hessami, 2025; Besley etal., 2011). Yet, we know little about specific competences, skills or experiences that may help mayors do a good job in office. Many mayor duties involve bureaucratic * Maximilian Thomas [email protected] 1 Ruhr University Bochum, CESifo Munich, IZA Bonn, Bochum, Germany 2 LSE, civity Management Consultants, Hamburg, Germany 3 Faculty ofSocial Science, Chair forSocial Policy andPublic Economics, Ruhr University Bochum, Universitätsstr. 150, 44801Bochum, Germany 1961 Public administrators aspoliticians inoffice tasks and require tons of paperwork and coordination within the local public administration and its many subdivisions. Experience in the local public administration may thus indeed be helpful.1 In some countries, a background in public administration is a requirement for specific public offices. In the US, there exist two main forms of government: council-manager and mayor-council governments. In both systems, a city manager will be appointed for whom it is customary to hold a Master of Public Administration (MPA) degree,2 In contrast, in most countries, e.g. in Germany, a degree or prior work experience in this field is not required for local public offices. Yet, descriptively, many mayor candidates (and elected mayors) happen to have a public administration background.3 The question arises whether expertise in local public administration is relevant for policy choices and mayors’ performance in office. We analyze this question in the context of 421 municipalities in the German state of Hesse where mayors are elected every six years. Hessian mayors are directly elected but have relatively few powers/competences (i.e. Hesse applies a mixture between the US council-manager and mayor-council system). In Hesse (and other parts of Germany), although there is no educational requirement many mayors have a background in public administration – indicated by a Diplom-Verwaltungswirt/in degree obtained from a University of Applied Sciences – and are often employed in the local public administration before being elected into office. In a regression discontinuity setting with close mayor elections between a candidate with a background in local public administration and a candidate with a nonadministrative background, we analyze the effect of mayors’ background on municipal outcomes. If these two types of candidates win the election, we refer to them as public administrator mayors or outsider mayors.4 We focus on an outcome variable where mayors in Hesse have discretion and bear the sole responsibility. Most local policies in Hesse are set by local councils. However, investment grants from the state level are acquired through a tedious application procedure for which mayors have to put together an extensive application package (Hessami, 2018). 1 A contrary argument could be made by saying e.g. that an outsider, i.e. a mayor that gets elected without this experience, may exert extra effort in these tasks to signal competence (Rogoff & Sibert, 1988; Rogoff, 1990). 2 In the council-manager system voters elect the council, which is the primary legislative body of the city. The council in turn appoints a city manager who carries out the directives of the council. In some towns, the council may additionally have an elected mayor. However, in the council-manager form of government, the mayor has few or no powers above and beyond those of the other councilors. In the mayor-council system, there is also an elected council and an elected mayor, but the mayor typically is equipped with more powers. A mayor-council government will also appoint a city manager. Regarding evidence on effects of these systems on policymaking, see Coate and Knight (2011), Deno and Mehay (1987), Enikolopov (2014), Vlaicu and Whalley (2016). 3 In our sample, this is the case for 50.7% of mayor candidates and 63.9% of elected mayors (see Table1). 4 36.1% of Hessian mayors are outsiders with backgrounds in business/finance (10.9%), law (7.1%), teaching/social work (5.2%), military/police (5.0%), industry (3.6%), self-employment (2.8%), as shown in Table1. 1962 Z.Hessami et al. Our setting is ideal to study the effect of mayors’ occupational background on their performance. First, Hessian mayors have relatively few powers and competences (compared to e.g. Southern German mayors). Thus, we can focus on a specific outcome (investment grants) rather than analyzing a large set of outcomes. Second, as Hessami (2018) finds, the switch from mayor appointments by local councils to direct mayor elections between 1993 and 1998 has boosted the diversity of mayors in terms of their educational/professional background.5 This gives rise to a sufficiently large number of mayor candidates in our sample that do not have a background in public administration. Third, compared to other German states, mayor elections in Hesse do not take place at a uniform date.6 Thus, if we do find any significant effects, these are unlikely to be driven by temporal patterns in investment grant allocation. Our paper is primarily related to RD studies on the effect of local politicians‘ professional background on fiscal outcomes. Beach and Jones (2016) find that "managerial" skills of city council candidates (proxied by business experience) in California do not have a causal effect on various local outcomes (expenditures, revenues, unemployment, administration salaries, vote shares). Similarly, Kuliomina (2021) finds no effect of Czech local councilors’ entrepreneurial background on municipal budget allocation, deficits or debt. Yet, Szakonyi (2021) shows that Russian mayors with a business background promote business-friendly policies by prioritizing economic infrastructure spending (e.g. building/maintaining roads). Likewise, Kirkland (2021) finds that US businessperson mayors allocate more resources to infrastructure projects. Overall, the results are mixed and appear to depend on the specific context.7 In contrast to our paper, the above literature focuses on the business background of local politicians. Since most tasks of mayors are, however, bureaucratic, it is natural to ask whether mayors’ public administration background affects municipal outcomes. One exception in the literature is Bordignon etal. (2020) which applies a difference-in-differences design to an Italian reform in the 1990s that granted some cities higher local fiscal autonomy. The authors show that more affluent regions 5 Before 1993, the mayor was appointed by the council via a two-thirds majority vote. The council was able to remove the appointed mayor from office. Since the reform, the mayor is elected by the people. 6 Sources of this variation are initial asynchronous dates for mayor appointments in 1946/1947 and 1970s municipal mergers. In line with the Hessian municipal code (§42, HGO), the appointment/election of a mayor must take place three to six months before the incumbent steps down. A third reason is that some mayors do not serve a full term (death, illness, retirement). If a term ends prematurely, the appointment/election must take place within four months. Hence, further variation is due to the autonomy that municipalities have in choosing the date. 7 Related literature studies the effect of mayors’ education level (not field) on policy outcomes. Avellaneda (2009) finds that more educated mayors collect more property tax revenues and increase social spending in Columbia. Alpino etal. (2022) show that college-educated mayors in Italy are more likely to implement progressive tax reforms. In Germany, Freier and Thomasius (2015) find no effect of mayors’ education on municipal finances. Earlier literature at the cross-country level shows that the level (Besley etal. (2011)) and field of education (Dreher etal., 2009) of political leaders affects economic development. Further, a large literature studies politician gender (see Baskaran and Hessami (2025) for a recent study and Hessami and daFonseca (2020) for a literature review) and more recently politician age (Alesina etal., 2018; Baskaran etal., 2024). 1963 Public administrators aspoliticians inoffice attract administrative experts, while poorer regions elect mayors with strong political skills. Thus, administrative skills are likely relevant for becoming a mayor in certain contexts. Do administrative skills/experience also matter for being a successful mayor? A second related strand of literature studies the role of politicians versus bureaucrats in policymaking. Bureaucrats are guided by career concerns, i.e. they strive to move up the career ladder within the local administration, often holding positions where they have high job security and enjoy indeterminate appointments (see e.g. Vlaicu and Whalley (2016)). The question of how to allocate different types of tasks across bureaucrats and politicians has been discussed in previous literature and may depend on available information and characteristics of these tasks (Alesina & Tabellini, 2007). In our context, a bureaucratic task with relatively visible (positive) welfare effects is assigned to an elected mayor. In some sense, a public administrator mayor was a bureaucrat before and then became a politician, whereas an outsider mayor is a pure politician in line with the definitions above. We, therefore, argue that the main difference between public administrator and outsider mayors is their competence in bureaucratic tasks, which may lead to different levels of success in acquiring investment grants. We make four substantive contributions to the literature. First, our unique dataset – combining official municipal election and fiscal data with hand-collected individual information on top-two mayor candidates in Hesse – and our RD design allow for a credible identification of the causal effect of mayors’ occupational background. Second, we employ a new outcome variable to analyze fiscal policy choices. Whereas most studies concentrate on spending or taxes, we focus on investment grant receipts. Mayors in Hesse are in charge of investment grant applications and these grants finance projects that are visible to voters. Third, our hand-collected dataset allows us to determine mayors’ background with e.g. the tertiary degree Diplom-Verwaltungswirt/in specifically dedicated to public administration. Our paper is the first to evaluate the relevance of mayors‘ background in public administration for municipal outcomes. Fourth, we also investigate how the effect on investment grant receipts varies with the political environment (e.g. unified/divided local government, political alignment with state government). Since some previous RD studies do not find significant effects of mayors’ characteristics, the specific political setting and resulting incentives likely matter. We find that mayors’ public administration background has no causal effect on investment grants receipts. A number of robustness tests confirm this null result. First, we show that the results are not sensitive to the scaling of the outcome. Second, the inclusion of controls does not affect the results, confirming the successful randomization in our RD design. Third, beyond the data-driven optimal bandwidths that we use in our main estimations, we also obtain a null result using various bandwidths between 5 and 15%. Finally, we conduct power calculations indicating that our study is sufficiently powered to detect moderate to large effect sizes. In two extensions, we analyze heterogeneity in the treatment effect according to the ideological composition of councils and alignment of the mayor‘s party with the state government. First, our results suggest that political alignment with the state government plays no role, confirming that the state government and its ministries 1964 Z.Hessami et al. have little discretion in rejecting/approving applications for investment grants based on political preferences. As described in Hessami (2018), applications are typically successful if there is money left in a specific year and if the application is complete and documents compliance with the funding guidelines. Second, we find that public administrator mayors receive more grants when they belong to the ideological camp that has the council majority (i.e. unified local governments). Public administrator mayors attract more than twice as many investment grants than outsider mayors when they have the support of the council majority. This corresponds to 38 Euros per capita more per year for a median-sized Hessian municipality with about 10,000 inhabitants and can therefore be considered as a substantial effect. While this effect is not based on explicit randomization of horizontal alignment and should thus be interpreted with caution, we can show that horizontal alignment is not systematically correlated with various municipality characteristics. We conclude that the competence of public administrator mayors appears to only make a difference for their performance when they are sufficiently (politically) motivated to use it. As investment transfers are highly visible to voters and at the same time it is difficult for voters to correctly attribute outcomes to political bodies, one interpretation is that public administrator mayors have a greater incentive to attract additional funds if they are supported by the council. Any electoral benefit in the next election will accrue to both the mayor and will spill over to his/her politically aligned council.8 Outsider mayors, on the other hand, have less competence/experience in dealing with bureaucratic application procedures and thus, even if the council is aligned, it does not matter for their effort/success in attracting grants. To test for mechanisms, we run additional estimations. Given diverse historical trajectories, mayor elections may take place at any point in time throughout council terms. Council terms, on the other hand, are synchronized across all Hessian municipalities. This allows us to test whether mayors are motivated to attract additional funds to ensure the reelection of ideologically aligned councils.9,10 Our results show 8 An additional explanation could be that public administrator mayors are more motivated to apply for investment grants when the council is politically aligned because it is easier to agree on projects of interest and because these grants are matching grants and for every approved grant the council needs to provide additional financing. Given our data, we are unfortunately not able to test this explanation. 9 We also test for political cycles in grant receipts across mayor terms. For mayors politically aligned with councils, we do not observe higher grant receipts before mayor elections (see TableA.10 in the online appendix). Thus, mayors in our sample are not so concerned about their own reelection, which is not surprising given that the reelection rate of recontesting incumbent mayors is 87.3% and 78.9% for public administrator and outsider mayors, respectively. In contrast, Hessami (2018) finds that elected mayors (compared to appointed mayors) in Hesse attract more investment grants when a mayor election is imminent. The main difference is that she applies a DiD approach including all mayor elections and focuses on the difference between the two types of selection for mayors, while we rely on a much smaller RD sample and compare public administrator with outsider mayors. 10 A referee pointed out that the higher electoral success of public administrator mayors may positively correlate with council alignment, creating endogeneity. For this to be the case, public administrator mayors would need to be more often affiliated with one of the major parties likely to hold council majorities. In our complete dataset, public administrator mayors (outsider mayors) are affiliated with one of the major parties in 62.4% (61.1%) of the cases (difference not statistically significant, p-value 0.57), while they run as independent candidates in 33.9% (31.5%) of the elections. The remaining share (<5%) are 1965 Public administrators aspoliticians inoffice that indeed in the subsample of unified governments public administrator mayors attract more grants when a council election is imminent. In a second set of estimations, we also show that these grant receipts are positively correlated with the reelection likelihood of existing council majorities. Our findings have important implications for our understanding of the role of politicians’ characteristics for their motivation and performance in office. While on average it appears that a background in local administration is irrelevant, there are circumstances that give rise to differences in the extent to which mayors make an effort and successfully attract investment grants. Even though we only look at one specific outcome, most of the duties of mayors in Hesse (or elsewhere) involve bureaucratic tasks. Hence, prior experience with local administration may indeed be helpful for doing a good job in office. Since descriptively the share of public administrator candidates and mayors is already high (69.3% of elected mayors have a public administration background), it does, however, not follow that a compulsory requirement for a public administration background is needed. Voters appear to be aware of this advantage and seem to take this into account in their voting decision. As a more general contribution to the literature, our results highlight that competence and experience in bureaucratic tasks (or broadly speaking political selection, i.e. politicians’ characteristics) only matter for politicians’ performance and municipal outcomes when politicians are sufficiently (politically) motivated to use them. Thus, our results can be seen as an example where political selection and (indirect) electoral incentives interact, while the existing literature typically studies these concepts in isolation. Finally, we illustrate that politicians’ incentives may go beyond their own reelection motive. In a setting where some degree of collaboration between two (or more) political bodies is helpful to be successful, politicians (who in any case enjoy a high incumbency bonus) may be motivated to create cycles in their performance that create positive spillovers for ideologically aligned political bodies. While previous literature argues, that these types of spillovers may exist within one office and within one party, we show that they also exist across offices and potentially across parties (but within an ideological camp in the left/right sense).11 11 For instance, Fowler and Hall (2014) and Lopesda Fonseca (2017) distinguish between personal and partisan incumbency advantages in the context of US state legislature and Portuguese mayor elections. Both find that partisan incumbency advantages (i.e. spillovers to candidates other than the incumbent) are not empirically relevant. Note, however, that in these settings incumbency advantages are much smaller than in our context. Footnote 10 (continued) minor parties not consistently aligning with either of the major parties. Thus, the proposed mechanism likely does not threaten our identification. 1966 Z.Hessami et al. 2 Background 2.1 Municipal finances Located in the center of Germany, Hesse represents the German average in terms of size and prosperity. In 2019, the population size was 6.3 million equalling 7.5% of the German population. Hesse consists of 421 regular municipalities, out of which 416 are organized into 21 counties, and five large cities which hold a special status.12 Local administrations substantially impact citizens’ daily lives. Municipalities provide a wide range of mandatory and voluntary public goods. Compulsory public goods include municipal childcare, primary schools, and civil protection. Voluntary provision includes swimming pools and sports venues, cultural and historical sites, theaters or public libraries. The provision of public goods at the municipal level is financed through a combination of municipal taxes, user fees, and transfers from higher levels of government. State-level transfers make up a large share of municipal revenues. Yet, municipalities are also granted considerable fiscal autonomy to generate own tax revenues by setting local business and local property tax rates or introducing new taxes (e.g. secondary home tax, tourist tax, dog tax). In 2019, Hessian municipalities in total had 52.2 billion Euros at their disposal. 2.2 State‑level investment grants The Hessian state government provides municipalities with several types of rulebased and discretionary grants. In this paper, we focus on investment transfers (Zuweisungen und Zuschüsse für Investitionen und Investitionsförderungsmassnahmen) intended for specific projects, such as the construction/modernization of local infrastructure (e.g., hospitals, childcare facilities, municipal elderly care facilities, public transportation, road construction, municipal drinking water systems, libraries, museums, and sport facilities). In 2019, the Hessian state government allocated 301.4 million Euros in total as investment grants. Investment grants are allocated annually via a formal application procedure (Antragsund Bewilligungsverfahren). In the second half of the previous year, the state government publishes calls for proposals via the state government gazette (Staatsanzeiger), detailing program objectives, eligible projects, and required documentation. To receive funding, mayors must submit complete applications. Grants are awarded if eligibility requirements are met and funds are available, with decisions made by the responsible ministry within weeks. Though earlier applications are more likely to be successful, unfunded but eligible applications can be carried forward to the next year. Municipalities may receive grants for several projects per year.13 12 These five cities (Darmstadt, Frankfurt, Kassel, Offenbach, Wiesbaden) fulfill both municipal and county responsibilities are therefore excluded from our analysis. 13 For more details on the grant application procedure, please refer to Hessami (2018). 1967 Public administrators aspoliticians inoffice Mayors are responsible for the preparation and submission of grant applications and can increase the likelihood that their municipality receives an investment grant (Hessami, 2018). First, mayors can stay informed about new funding opportunities by regularly checking the calls for grants. Second, preparing and submitting an application (or several ones) in the first place is necessary for obtaining any grant. Third, mayors can impact the likelihood of grant receipt by ensuring that the application is complete and meets all funding guidelines.14 While investment grants on average only account for 19.4% of total transfers, individual payments to municipalities can be substantial. For example, in 2019, the highest amount granted to a municipality was 435 Euros per capita. For a municipality with 10,000 inhabitants, this amounts to 4.35 million Euros. Despite being project-specific, these grants clearly enhance the financial possibilities of a receiving municipality. 2.3 Mayors andlocal politics There are two political institutions at the local level: the mayor and the council. Voters elect a municipal council at a uniform state-wide date every five years in March. Council size depends on municipal population size and varies which between 15 and 105 seats. The municipal code of Hesse (Hessische Gemeindeordnung) stipulates that the local council monitors the local administration and decides autonomously on all policy areas assigned by the federal and state constitutions to the municipal tier. Council seats are allocated to parties via a proportional rule. The party landscape in Hessian local council and mayor elections is dominated by the centre-right Christian Democratic Union (CDU) and the centre-left Social Democrats (SPD). Several smaller parties also compete in local council elections, including the Greens (environmental, socially liberal), the FDP (fiscally conservative, socially liberal), and the Left Party (socialists). Also, several far-right/-left parties, independent candidates and candidates backed by local voter initiatives contest in local elections. Multiple voter initiatives may run in a single municipality and tend to be successful, in rare cases securing seat majorities. Typically, multiple parties win seats, requiring coalitions to secure majorities. Even absent formalized coalitions, parties must cooperate to reach council decisions (Baskaran & Hessami, 2018). Councilors hold significant political power, while the mayor remains an important political figure. Since 1993, mayors in Hesse are directly elected for a six-year term.15 A state-wide election date does not exist. Instead, a mayor election is held 15 A reviewer has pointed out that the transition phase (1993-1997) from appointments to direct elections of mayors falls within the observation period. However, our close-election RD design ensures that we exclusively compare municipalities governed by directly elected mayors. It might in principle be that during the transition phase directly elected administrative mayors perceived heightened competition with neighboring municipalities led by appointed mayors, potentially increasing their reelection incentives and influencing their decision to apply for investment grants. To address this concern, we conduct an 14 Note that the allocation of these grants does not include a built-in equalization scheme across municipalities. 1974 Z.Hessami et al. density of the margin of victory at the threshold, suggesting that there is no manipulation of the treatment status. 4.2.2 Balance tests formunicipality characteristics A second assumption is that the selection of candidates close to the cutoff is uncorrelated with municipal characteristics that may matter for investment grant receipts. Thus, municipalities in which a public administrator candidate barely wins against an outsider candidate should not systematically differ from municipalities where an outsider candidate barely wins against a public administrator candidate. To formally test the success of our local randomization at the municipal level around the cutoff, we repeat our baseline estimations with eight different municipality characteristics as dependent variables (see TableA.3). None of the differences between municipalities with a public administrator and an outsider mayor are significant. 4.2.3 Unconfoundedness withother mayor characteristics A further concern are confounding effects due to other personal attributes of elected mayors. For instance, public administrator candidates might be concentrated in specific parties. Hence, the results that we obtain may be due to mayors’ party affiliation rather than mayors’ background. We test whether in close races victorious public administrator candidates differ from victorious outsider candidates in various attributes. Although "ability" cannot be measured directly, it is at least partly related with observable characteristics such as education. In TableA.4, we test for a discontinuity at the threshold forthe following characteristics of elected mayors: education, gender, age, and party affiliation. We do not find any significant differences between public administrator and outsider mayors who were elected in close races. Fig. 2 Descriptive statistics: Investment grants per capita. This figure illustrates descriptive statistics for investment grants per capita. Subfigure (a) shows the distribution of per-capita investment grants across municipalities. Subfigure (b) shows the development of average investment grants during the 1993–2020 period for municipalities governed by outsider mayors (dashed line) and municipalities governed by public administrator mayors (solid line) 1975 Public administrators aspoliticians inoffice 5 Results 5.1 Baseline We first graphically explore the relationship between mayors’ background and investment grant receipts. Figure4 plots investment grant receipts at different margins of victory. Positive values of the margin of victory imply that the public administrator candidate wins the election, whereas negative values imply that the outsider candidate wins the election. The dots are binned averages of the outcome variable. The solid lines plot flexible polynomials on both sides of the cutoff with 95%-confidence intervals shaded in grey. There is no discontinuity in investment grant receipts between municipalities with and without a public administrator mayor. Next, we explore the effect of public administrator mayors on investment grants with regressions based on Equation (2). Table2 collects RD estimates for both nonparametric (Model 1 and 2) and parametric (Model 3) specifications. Again, there is no significant difference.28 In the online appendix (see Table A.5), we also report results from OLS estimations based on the full sample. While the simple bivariate relationship between public administrator mayors and investment grant receipts is significantly positive, including controls (e.g. population size, population density, etc.) and municipality and year FE leads to insignificant results. This indicates that OLS estimation leads to an upward bias in the relationship. As stated in Sect.4, this might for instance be due to voters in poorer or more rural municipalities strategically electing public administrator candidates as mayors as they are likely better able to secure grants for investment projects. The fact that the significant relationship disappears once population controls are included indicates that these characteristics play a key role. 5.2 Robustness 5.2.1 Bandwidth choice We test the sensitivity of our results to bandwidth selection in Figure A.1 in the online appendix based on a variety of bandwidths, franging from 5% to 15% margins of victory. All estimated coefficients for the public administrator mayor dummy are close to zero and insignificant. 28 Even though the estimates are far from conventional significance level, the coefficients are slightly positive and of moderate size, ranging between 11% and 26% additional grants being attracted by public administrator mayors. In Sect.5.2.4, we conduct further tests to investigate whether our analyisis is underpowered and whether we would be able to detect small, moderate and large effects given our RD sample. 1976 Z.Hessami et al. 5.2.2 Scaling ofoutcome variable We test whether the null effect for investment grant receipts is sensitive to the scaling of the outcome variable using five alternative scalings (see Table A.7 in the online appendix). Models (1)–(5) present results for the extensive margin (dummy = 1 if a municipality received non-zero grants in a year), total investment grants, per capita investment grants, total per capita investment grants acquired during the entire term, and the difference in per capita investment grants between the last and first year in office. None of the coefficients are statistically significant. 5.2.3 Inclusion ofcontrols We also rerun our baseline RD model by including municipal controls (i.e., population size, population density, female population share and share of working-age population) in TableA.8. As expected, adding controls reduces standard errors. Model (2) is our preferred specification including fixed effects and controls. The coefficient is close to zero, confirming the baseline null result.29 Fig. 3 RD validity: Density tests for running variable. This figure presents a histogram of the running variable "Margin of victory" (Subfigure (a)) and the results of local density estimation according to McCrary (2008) and Cattaneo etal. (2018) (see Subfigures (b) and (c)). The solid line in Subfigures (b) and (c) represents the density of the running variable, while the shaded areasindicateconfidence intervals 29 We also repeat our baseline estimation with additional economic controls capturing the potential demand for investment grants, i.e. local employment share, local business tax revenues per capita, number of firms per capita, municipal debt per capita, and total grants received. The results remain 1977 Public administrators aspoliticians inoffice Fig. 4 Baseline RD plot. This figure shows the RD plot for the log of per capita investment grant receipts per municipality when the mayor in office has a public administrator vs. outsider background. The forcing variable is the margin of victory of the public administrator candidate for the mayor’s office in mixed-background races (where the top-two candidates have a background in public administration vs. outside of it). Observations to the right of the cutoff imply that a public administrator holds the mayor’s office. Each dot is the binned local average of investment grants raised by the municipality for the margin of victory. The solid lines are from a local linear smooth of the underlying observations. The grey-shaded areas are the 95% confidence intervals (Color figure online) Table 2 Baseline: Public administrator mayors and investment grants This table reports estimates from sharp RD estimations that relate the background of elected mayors to investment grant receipts. The sample covers all mayor elections in Hesse where the top-two candidates were of opposite backgrounds ("mixed-background elections"). The dependent variable is the log of investment grants per municipality, measured in constant (2015) per-capita values. The treatment variable is a dummy which equal to one if the candidate with an administrative background won the last election. We report local linear (Model 1 and 2) and quadratic (Model 3) regressions applying MSERD optimal bandwidths proposed by Calonico etal. (2019). Heteroscedasticityand cluster-robust standard errors are in parentheses. The unit of clustering is the municipality of the candidate. Stars indicate significance levels at 10% (*), 5% (**) and 1%(***) Dep. var.: Log(inv grants p.c.) (1) (2) (3) Public admin mayor 0.110 0.264 0.189 (0.368) (0.328) (0.518) Bandwidth type MSE optimal MSE optimal MSE optimal Bandwidth size 11.712 11.712 11.712 Polynomial Linear Linear Quadratic Time FE No Yes No N 605 605 605 Elections 105 105 105 Municipalities 103 103 103 Mean (SD) 3.17 (1.52) 3.17 (1.52) 3.17 (1.52) Footnote 29 (continued) unchanged (available upon request). Since these economic variables are not consistently available across all periods, their inclusion reduces the number of observations. For this reason, we do not include these additional controls in our baseline estimation. 1978 Z.Hessami et al. 5.2.4 Statistical power One concern could be that our analysis lacks statistical power. Therefore, we conduct ex-post power calculations for RDDs following Cattaneo etal. (2019b) in TableA.9 in the online appendix. In more conservative scenarios e.g., if the true effect of mayors’ background at the threshold were 20% of the standard deviation, we would be able to detect it with a 9.7% probability. However, we have sufficient statistical power to identify larger effect sizes. For instance, if the true effect were 80% of the standard deviation, we would detect it with a 72% probability. Thus, although we cannot rule out minor effects of mayors’ background on investment grants, our study is adequately powered to detect moderate to large effects. 6 Effect heterogeneity We explore extensions to our main regression to test for effect heterogeneity. We rerun our baseline RD model estimating fully interacted specifications, i.e. we interact "public administration mayor" and the "margin of victory" with a dummy that relates to the political environment.30 6.1 Vertical alignment: state governments State governments might allocate grants to politically aligned municipalities for electoral reasons. Baskaran and Hessami (2017) find that aligned municipalities receive more discretionary transfers in Hesse, specifically when the degree of local support for the state government is strong. Quinckhardt (2023) shows for eight German states that party alignment between a mayor and the state government increases transfers to municipalities by 20% per year. One could thus hypothesize that municipalities with a public administrator mayor who is from the same party as the state government receive more grants. We explore the role of partisan alignment with the state government and rerun our baseline RD model including and interacting a dummy equal to one if the mayor‘s party is in the state government with "public administrator mayor" and "margin of victory". The results are collected in Table3.31 The interaction effect is insignificant in all models and this null finding is robust across different bandwidth choices (see FigureA.2 in the online appendix). Hence, 30 Note that the political environment is of course not randomly assigned in the following analysis. Both state government alignment and unified local governments are more likely to emerge in a setting where there is less polarization and specific partisan preferences are more pronounced within a state or municipality. Nevertheless, it is of interest to analyze whether the baseline effect varies with the political environment. 31 Overall, for 177 municipality-year observations in our close mixed-background races, the mayor’s party is politically aligned with the state government (i.e. 33% of the sample). 103 and 74 observations refer to public administrator mayors and outsider mayors, respectively. 1979 Public administrators aspoliticians inoffice investment grant receipts do not depend on party ties between the mayor’s party and the state government,32 These results represent correlational evidence. In online appendix Table A.13, we show that vertically aligned municipalities are statistically different from nonaligned municipalities in several observable characteristics (population size, population density, debt, and local tax rates). Although we explicitly control for population size and density in Model(3), we acknowledge that (un)observed confounders could still introduce endogeneity.33 6.2 Horizontal alignment: local councils Mayors with a public administration background may be more motivated to attract investment grants if their party also holds the majority in the council (unified government). On the other hand, public administrator mayors may not be willing to leverage their administrative skills if faced with an opposing council majority (divided government). Our mixed-background RD sample ( ≤ 10% margin of victory) includes 51 municipality-year observations (24 and 27 observations for public administrator and outsider mayors) where the mayor’s party is aligned with the ideological camp of the council majority ( ≈ 9.5% of the sample).34 One concern is that municipalities in which the mayor and the council majority align ideologically might differ in their characteristics from municipalities without horizontal alignment and these differences could affect grant receipts. To address this issue, we compare municipalities with aligned local governments to all other municipalities in TableA.12: the only significant difference is in terms of population age structure, which we control for in our model. There are no differences in other characteristics, particularly fiscal indicators. Note that our results still need to be interpreted with caution and as mostly correlational since we cannot explicitly test for political strategies and other unobservables determining horizontal alignment. Next, we rerun our baseline RD model including and interacting a dummy equal to one if the mayor belongs to the ideological camp (left/right) which holds the 33 In additional estimations (available upon request), we also control for the fiscal variables mentioned here. The results are unchanged. Since these are simultaneously determined (endogenous) variables considered as “bad controls”, we prefer not to include them in the main estimations. 34 As mayor and council elections vary in their timing, council majorities may change during a mayor term. 32 One possible reason why, in contrast to Baskaran and Hessami (2017), we do not find a significant difference in the amount of transfers between municipalities according to state government alignment is that we use investment grants while Baskaran and Hessami (2017) focus on budget support transfers (Zuweisungen und Zuschüsse für laufende Zwecke). Mayors do not apply for budget support transfers and therefore the personal effort and competence of mayors plays no role for the allocation of these transfers. 1980 Z.Hessami et al. council majority with "public administration mayor" and "margin of victory".35 The results are collected in Table4. The positive and statistically significant coefficient for the interaction suggests that public administrator mayors attract more than twice as many investment grants than outsider mayors when they have the support of the council majority. This corresponds to 38 Euros per capita more per year for a median-sized Hessian municipality with about 10,000 inhabitants. The coefficient for the interaction effect is robust Table 3 Effect heterogeneity I: Aligned state governments This table reports estimates from sharp RD estimations that relate the background of the elected mayor to investment grant receipts. The sample covers all mayor elections in Hesse where the top-two candidates were of opposite backgrounds ("mixed-background elections"). The dependent variable is the log of investment grants per municipality, measured in constant (2015) per-capita values. The treatment variable is a dummy equal to one if the candidate with an administrative background won the last election. We interact the treatment dummy as well as the control function with a dummy for whether the mayor’s party belongs to the state government. Overall, in 177 municipality-year observations, the mayor’s party is politically aligned with the state government. Of these, 103 (74) observations refer to public admin (outside) mayors respectively. We report local linear (Model 1, 2, and 3) and quadratic (Model 4) regressions using MSERD optimal bandwidths proposed by Calonico etal. (2019). Municipal controls include population size, population density, and the working-age population share. Heteroscedasticityand cluster-robust standard errors are in parentheses. The unit of clustering is the municipality of the candidate. Stars indicate significance levels at 10% (*), 5% (**) and 1%(***) Dep. var.: Log(inv grants p.c.) (1) (2) (3) (4) Public admin mayor 0.021 − 0.112 − 0.060 0.335 (0.435) (0.298) (0.335) (0.607) State govt alignment 0.095 − 0.135 − 0.159 0.492 (0.601) (0.330) (0.368) (0.801) Public admin mayor x 0.239 0.293 0.284 − 0.388 State govt alignment (0.759) (0.497) (0.532) (1.031) Bandwidth type MSE optimal MSE optimal MSE optimal MSE optimal Bandwidth size 11.712 11.712 11.712 11.712 Polynomial Linear Linear Linear Quadratic Time FE No Yes Yes No Controls No No Yes No N 605 605 605 605 Elections 105 105 105 105 Municipalities 103 103 103 103 Mean (SD) 3.17 (1.52) 3.17 (1.52) 3.17 (1.52) 3.17 (1.52) 35 We categorize the SPD and the Greens as left-wing, and the CDU and the FDP as right-wing. Even though smaller parties and voter communities contribute to coalition building, the ideological positions at the local level are often more "pragmatic" and cannot be clearly classified into a particular camp. The dummy variable is equal to one if the camp of the mayor’s party also has the majority (50%+ 𝜂 ) of seats in the council. Note that by definition independent mayors can not be aligned with a local council majority. 1981 Public administrators aspoliticians inoffice to including time fixed effects (Model (2)), and municipal controls (Model (3)).36 In addition, FigureA.3 in the online appendix shows that the positive interaction term is robust to different bandwidths.37,38 7 Mechanism: Mayors’ indirect electoral incentives In this section, we run additional estimations to analyze why public administrator mayors that face an ideologically aligned local council attract more investment grants. 7.1 Investment grant cycles overcouncil terms Mayors in Hesse (and in other German states) are reelected at a very high rate and once in office they often stay there until retirement. This makes it unlikely that mayors are primarily motivated by their personal electoral incentives to apply for investment grants. Moreover, this would not explain why the competence of public administrator mayors (compared to outsider mayors) only makes a difference for grant receipts when the council is politically aligned. What seems more likely is that mayors have an interest in helping the ideologically aligned local council stay in power. This does not only reflect their personal ideological preferences as citizens of the municipality but also matters for the direction of future local policy choices. We test whether the amount of investment grants that public administrator mayors attract when faced with an aligned council varies over the council term, i.e. whether they are higher just before the next council election. Table5 collects the results for specifications where we include a dummy for one year and two years before the council election and interact this with the public administrator mayor dummy.39 Even though the sample is of course smaller and we therefore have to be careful in interpreting the results, we find that the baseline effect heterogeneity is driven 38 A reviewer has pointed to the negative Council alignment coefficient in Model (3) in Table4. This result is not directly related to our main hypothesis but is indeed interesting. One potential explanation could be that outsider mayors jointly governing with an aligned council are complacent and even reduce their effort in raising additional funds. However, note that the effect is only significant at the 10%-level in one of the four models (Model 3), and coefficient signs vary across specifications. Therefore, any interpretation must be treated with caution. 39 FigureA.4 in the online appendix illustrates this more clearly in an event-study plot going back up to four years before council elections: the effect indeed only kicks in one year before council elections. 36 For the quadratic polynomial (Model (4), the interaction is insignificant. However, due to the concerns expressed by Gelman and Imbens (2018), we focus on the linear specification as our preferred model. 37 For the sake of completeness, we also explore divided local governments (see TableA.6 in the online appendix). Our RD sample includes 53 municipality-year observations ( ≈ 9.8% of the sample) with a divided government. The results are, as expected diametrically opposed to those for unified governments: public administrator mayors who face an opposing council majority are less successful in securing investment grants. The interaction is robust to the inclusion of fixed effects (Model (2)) and controls (Model (3)), and the use of quadratic polynomials (Model (4)). FigureA.3 illustrates the robustness to different bandwidth choices. 1982 Z.Hessami et al. by higher grant receipts one year before council elections. The relevant interaction term coefficient is significant at the 10% level. Mayors may be uncertain about when grants will be received and therefore may submit applications “too early” which in some cases may lead to receiving the money earlier. For this reason, we also consider the effect two years before the council election: the coefficient is positive but insignificant.40 7.2 Investment grants andthereelection ofcouncil majorities In a second step, we test whether this hike in investment grant receipts matters for the electoral success of aligned councils. This requires that voters indeed are not Table 4 Effect heterogeneity II: Unified local governments This table reports estimates from sharp RD estimations that relate the background of the elected mayor to investment grant receipts. The sample covers all mayor elections in Hesse where the top two candidates were of opposite backgrounds ("mixed-background elections"). The dependent variable is the log of investment grants per municipality, measured in constant (2015) per-capita values. The treatment variable is a dummy equal to one if the candidate with an administrative background won the last election. We interact the treatment dummy as well as the control function with a dummy for whether the mayor belongs to the ideological camp (left/right) which holds the council majority. Overall, in 51 municipalityyear observations, the mayor’s party is politically aligned with the council majority. Of these, 24 (27) observations refer to public admin (outsider) mayors respectively. We report local linear (Model 1, 2 and 3) and quadratic (Model 4) regressions using MSERD optimal bandwidths proposed by Calonico etal. (2019). Municipal controls include population size, population density, and the working-age population share. Heteroscedasticityand cluster-robust standard errors are in parentheses. The unit of clustering is the municipality of the candidate. Stars indicate significance levels at 10% (*), 5% (**) and 1%(***) Dep. var.: Log(inv grants p.c.) (1) (2) (3) (4) Public admin mayor 0.020 0.159 − 0.254 0.199 (0.390) (0.341) (0.292) (0.560) Council alignment − 0.203 − 0.210 − 0.558* 0.782 (0.505) (0.413) (0.322) (0.545) Public admin mayor x 1.333** 1.667** 1.102** 1.155 Council alignment (0.656) (0.663) (0.463) (0.903) Bandwidth type MSE optimal MSE optimal MSE optimal MSE optimal Bandwidth size 11.712 11.712 11.712 11.712 Polynomial Linear Linear Linear Quadratic Time FE No Yes Yes No Controls No No Yes No N 605 605 605 605 Elections 105 105 105 105 Municipalities 103 103 103 103 Mean (SD) 3.17 (1.52) 3.17 (1.52) 3.17 (1.52) 3.17 (1.52) 40 On the other hand, if mayors do not consider potential lags, it could be that the money is accidentally received in the year of the council election or even one year later. If anything, this would mean that we underestimate the true effect as the investment grant receipts in the “control” period are inflated. 1983 Public administrators aspoliticians inoffice aware of who has written and submitted the grant application and therefore voters give the incumbent council majority credit for this success. In Table6, we investigate the relation between the investment funds raised and the probability of reelection for the ruling council majority. We use a linear probability model, with a dummy as dependent variable equal to one if the council majority defends its majority in the subsequent council election. The independent variable is the log of investment grants per capita; the estimated coefficients can therefore be interpreted as semi-elasticities. We find that a 1% increase in investment grants leads to a 4.4% increase in the reelection probability of the council majority.41 This suggests that the mayor indeed increases the reelection probability of the aligned local council majority in the Table 5 Mechanism: Budget cycles in investment grants across council terms This table reports estimates from sharp RD estimations that relate the background of elected mayors to investment grant receipts. The sample covers candidates all mayor elections in Hesse where the top-two candidates were of opposite backgrounds ("mixed-background elections"). The sample is limited to the subsample of unified governments. i.e. where the mayor and the majority of councilors belong to the same ideological camp. The dependent variable is the log of investment transfer per municipality, measured in constant (2015) per-capita values. The treatment variable is a dummy equal to one if the candidate with an administrative background won the last election. We interact the treatment dummy as well as the control function with a dummy for whether the local council is in the last two years (Model 3 and 4) or final year of the legislative period (Model 1 and 2). We report local linear regressions using MSERD optimal bandwidths proposed by Calonico etal. (2019). Municipal controls include the population size, population density, and the working-age population share. Heteroscedasticityand clusterrobust standard errors are in parentheses. The unit of clustering is the municipality of the candidate. Stars indicate significance levels at 10% (*), 5% (**) and 1%(***) Dep. var.: Log(inv grants p.c.) (1) (2) (3) (4) Period t-1 t-1 t-2 & t-1 t-2 & t-1 Public admin mayor 0.419 0.435 0.372 0.320 (0.389) (0.449) (0.585) (0.594) Period before council election − 0.734 − 0.591 − 0.157 − 0.088 (0.517) (0.487) (0.500) (0.514) Public admin mayor x 0.938 1.026* 0.462 0.681 Period before council election (0.689) (0.616) (0.784) (0.722) Bandwidth type MSE optimal MSE optimal MSE optimal MSE optimal Bandwidth size 12.229 12.229 12.229 12.229 Polynomial Linear Linear Linear Linear Controls No Yes No Yes N 102 102 102 102 Elections 28 28 28 28 Municipalities 27 27 27 27 Mean (SD) 2.92 (1.44) 2.92 (1.44) 2.92 (1.44) 2.92 (1.44) 41 In additional estimations, we repeat the analysis using a binary probit model. The estimated effects are similar. See TableA.11 and FigureA.5 in the online appendix.