Amalgamation and size (increase) – two faces of the same coin or not?

As Gendźwiłł et al. (Reference Gendźwiłł, Kurni, ewicz and Swianiewicz2020: 2) state, ‘cross-national comparative studies of municipal amalgamations (…) are rare’ since almost all contributions are single-case studies. From a non-exhaustive but accurate review of over 100 articles published on the size, merger, and local democracy relations since the early 2000s, little less than 90% focus on a single country.Footnote ²

Moreover, the existing comparative analyses are usually devoted to single out the drivers of territorial reforms (Baldersheim and Rose, Reference Baldersheim and Rose2010; Askim et al., Reference Askim, Klausen, Vabo and Bjurstrøm2017) or to exploring the different strategies (Steiner et al., Reference Steiner, Kaiser, Eythórsson, Kuhlmann and Bouckaert2016) and paths in their implementation (Sancton, Reference Sancton2000; Swianiewicz, Reference Swianiewicz2010; Reingewertz, Reference Reingewertz2012) – rather than their consequences. Evidence on the impact of the amalgamation reforms on different aspects of local government remains, at best, unclear and unsystematic (Swianiewicz et al., Reference Swianiewicz, Gendźwiłł and Zardi2017; Tavares, Reference Tavares2018). However, recently, some studies have started addressing the consequences of mergers on some aspects of the local government. They primarily focus on the economic performances – considering, for instance, the aggregate effects of consolidation and size on efficiency gains and cost/expenditure reduction (Reingewertz, Reference Reingewertz2012; Bel and Warner, Reference Bel and Warner2016; Blesse and Baskaran, Reference Blesse and Baskaran2016; Blom-Hansen et al., Reference Blom-Hansen, Houlberg, Serritzlew and Treisman2016; Harjunen et al., Reference Harjunen, Saarimaa and Tukiainen2021), on the costs of public services provision (Solé-Ollé and Bosch, Reference Solé-Ollé and Bosch2005), on the economies of scale (Dollery et al., Reference Dollery, Byrnes and Crase2008; Soukopová et al., Reference Soukopová, Nemec, Matějová and Struk2014; Drew et al., Reference Drew, Kortt and Dollery2016), or local taxation (Allers and Geertsema, Reference Allers and Geertsema2016).

Voter turnout variation has also raised scientific interest as a potential consequence of amalgamation. A large majority of the studies on this aspect observe a detrimental effect of the enlarged size on electoral participation (Fritz and Feld, Reference Fritz and Feld2015; Koch and Rochat, Reference Koch and Rochat2017). Other studies show that information costs for electoral participation are higher (Lapointe et al., Reference Lapointe, Saarimaa and Tukiainen2018), thus implying lower turnout rates – which is more evident in the first election after a merger (Koch and Rochat, Reference Koch and Rochat2017; Heinisch et al., Reference Heinisch, Lehner, Mühlböck and Schimpf2018). Studies revealing a detrimental effect of size increase on turnout are mostly Continental-European cases – Germany (Fritz and Feld, Reference Fritz and Feld2015; Roesel, Reference Roesel2017), Austria (Heinisch and Mühlböck, Reference Heinisch and Mühlböck2016; Heinisch et al., Reference Heinisch, Lehner, Mühlböck and Schimpf2018), Switzerland (Ladner, Reference Ladner2002; Koch and Rochat, Reference Koch and Rochat2017), Netherlands (van Houwelingen, Reference van Houwelingen2017; Allers et al., Reference Allers, de Natris, Rienks and de Greef2021) – but the same negative effect was also observed elsewhere, e.g., in Japan (Horiuchi et al., Reference Horiuchi, Saito and Yamada2015), and Israel (Zeedan, Reference Zeedan2017).

Nonetheless, the overall picture is not entirely consistent: a null effect was observed in South Africa (Cameron and Milne, Reference Cameron and Milne2013), Japan (Mabuchi, Reference Mabuchi2001), and Finland (Lapointe et al., Reference Lapointe, Saarimaa and Tukiainen2018); while some positive, albeit limited, effects were found in the Netherlands (Kraaykamp et al., Reference Kraaykamp, van Dam and Toonen2001), Canada (Hicks, Reference Hicks2006), and Denmark (Bhatti and Hansen, Reference Bhatti and Hansen2019). Southern Europe shows few but, again, mixed pieces of evidence. While in Italy, Bolgherini and Mollisi (Reference Bolgherini and Mollisi2021) find a positive effect on turnout after amalgamation, in Portugal, Rodrigues and Tavares (Reference Rodrigues and Tavares2020) observe a negative impact – although referring to sub-municipal units.

Significant cross-country inconsistencies are therefore evident. They may reflect specific conditions in individual countries, such as legal/institutional or political factors. Thus, we posit that these inconsistencies derive from the single-country focus and the lack of a comparative perspective, which this article instead provides. Moreover, literature so far has devoted its attention to the (positive or negative) effect of amalgamation on turnout. We argue that, while interesting, this strategy is somewhat limited, as the actual effect of the process of amalgamation results from the specific characteristics of that particular process. Indeed, different countries show different features in the amalgamation policy process, such as average municipal size (both pre- and post-merger), timings, and type of mergers. Thus instead, we believe it is more fruitful to comparatively investigate the different amalgamation processes in different countries to make sense of the apparent idiosyncrasies by including various features of the reform process and assessing their predictive power over voter turnout.

Theoretical framework

This article aims at making sense of the apparent idiosyncrasies in the effects of municipal amalgamation on municipal electoral participation emerged in the literature, by investigating this variance in a comparative cross-country setting and focussing on the explicative role of the features of the amalgamation processes. Are there characteristics of these processes within the various countries (which could not be controlled for in case studies) that account for (part of) the turnout variation? And which ones? Furthermore, turnout after amalgamation varies significantly within country as well. Our comparative dataset is helpful to investigate this too, again focussing on the characteristics of the specific process of amalgamation occurring in each municipality – in which direction, to what extent, and under which circumstances do these elements impact turnout?

With this in mind, in this section, the theoretical framework employed to conduct our empirical investigation is outlined. It combines the above-reviewed literature arguments about size, municipal amalgamations, and their specific (national and local) features with classical predictors of electoral participation in municipal elections. First, we consider some characteristics of the amalgamation process at the national level, which are constant within each national electoral wave but vary across countries and by different national waves within our dataset. Analysing these, serves to evaluate the explicative ability of the type of reform enacted by each country over turnout and to assess under which circumstances amalgamations have different effects on municipal electoral participation. Amalgamation reforms can occur in different ways. The literature commonly stresses two main types of merging processes: compulsory vs. voluntary and massive vs. gradual. Massive indicates a comprehensive reform involving all municipalities in a country, while gradual refers to incremental, step-by-step reform. A compulsory amalgamation reform implies a top-down imposed decision, while voluntary reform entitles the single (groups of) municipalities to decide (with guidelines varying in strictness) if, when, and with whom to merge.Footnote ⁵

Previous studies have found these elements to be particularly important (Steiner et al., Reference Steiner, Kaiser, Eythórsson, Kuhlmann and Bouckaert2016). Top-down reforms are more likely to meet resistance at the local level (Baldersheim and Rose, Reference Baldersheim and Rose2010) since they can be considered as power-gathering by the central government. On the contrary, resistance to amalgamations may decrease if the reforms are introduced bottom-up (Eythórsson, Reference Eythórsson, Baldacchino, Greenwood and Felt2009). Voluntary amalgamation processes are considered more successful. When the initiative is bottom-up, the process towards the amalgamation is more rapid (ibidem). Moreover, Steiner et al. (Reference Steiner, Kaiser, Eythórsson, Kuhlmann and Bouckaert2016) argue that top-down compulsory amalgamations are likely to trigger conflicts. This could imply a higher level of dissatisfaction. We still need to understand how this translates into effects on turnout. It could be both ways, as Denters et al. (Reference Denters, Goldsmith, Ladner, Mouritzen and Rose2014: 231) stress the unpredictability of the direction of (dis)satisfaction on the likelihood to vote since voting may be an expression of both satisfaction and dissatisfaction. However, we believe that in the case of low-stake, second-order municipal elections, dissatisfaction leads more probably to de-mobilisation than to mobilisation. Municipal elections are usually considered second-order elections (Reif and Schmitt, Reference Reif and Schmitt1980). Hence, regarding the electoral consequences of post-merger conflicts and dissatisfaction, we can reasonably expect apathy and a consequent de-mobilisation, because of the amalgamation, to prevail over mobilisation. Therefore, in general terms, to recall Hirschman’s (Reference Hirschman1970) classic concepts, we assume an exit strategy (de-mobilisation) as prevalent over a voice strategy (which might, in turn, lead to an increase in electoral participation). Thus, we expect a depressive effect of more conflictual (forced) amalgamations on turnout. Hence, we hypothesise that:

Analogous mechanisms can be expected for the overall scope of the amalgamation process, namely the portion of national municipalities involved in it. We argue that the scope of a particular amalgamation reform wave may be (inversely) related to the reaction triggered by it. We start by considering that the reaction to the policy decision to amalgamate is prevalently negative,Footnote ⁶ consistently with most previous studies that found a negative variation in voter turnout after amalgamation. Against this backdrop, voters might be expected to react less negatively when the whole country is affected (massive process), while they could oppose more fiercely the amalgamation when they are ‘alone’ (gradual process) in suffering the loss of their previous municipal home.Footnote ⁷ The reasoning behind this claim can be traced back to Wilson’s (Reference Wilson1973) classic argument that conflict tends to be lower when the costs of a policy are distributed rather than concentrated, as individuals feel less (or more) closely related to the policy. This classic claim has been widely applied to policy feedback studies, which emphasise the importance of perception of proximity as a key element in understanding citizens’ reactions to a policy, also with specific reference to the micro level (Larsen, Reference Larsen2019).

As for the previous hypothesis, how this translates into turnout variation is somewhat ambiguous. It depends on whether dissatisfied voters will turn out to the polls more (mobilisation hypothesis) or less (de-mobilisation hypothesis) than the others. Again, we argue that de-mobilisation is the most likely translation of such dissatisfaction in low-stake, second-order municipal elections. Thus, we expect that:

We now shift to the local features of the amalgamation. A crucial element is clearly the size of the constituent units, as many of the fundamental mechanisms discussed in the literature review are connected (directly or indirectly) with it. Hence, considering that the role of a different size (either absolute or relative) of the merged local units on the reform outcomes remains severely underexplored, we aim to analyse exactly this ‘heterogeneity of effects’ (Gendźwill et al., Reference Gendźwiłł, Kurni, ewicz and Swianiewicz2020: 13). We articulate the size-effect on turnout by employing the distinction between the input and output sides of municipal size introduced in the previous section.

First, we look at the output side by starting from the classical size-and-democracy argument posed by Dahl and Tufte (Reference Dahl and Tufte1973). Smaller size should favour higher turnout levels by enhancing political efficacy, while larger size should lead to lower turnout by putting a larger distance between citizens and institutions. As anticipated, this relation has been repeatedly tested and very often confirmed. Quite straightforwardly, our hypothesis thus sounds as:

Besides the output side of size, we consider two aspects of the input side: the number of constituent units and the cross-unit population polarisation, which refers to the size relations among pre-amalgamation municipal units. Lapointe et al. (Reference Lapointe, Saarimaa and Tukiainen2018) and Koch and Rochat (Reference Koch and Rochat2017) both found that turnout was primarily affected in the small pre-merger municipalities. Considering this, we need to understand how these patterns translate into turnout variation in the post-amalgamation municipal units. While focussing on the pre-merger municipalities, the mechanisms highlighted by these studies can nicely serve our aim of building a theoretical framework that accounts for turnout variations in post-amalgamation units. On the one hand, we can expect higher population polarisation among pre-amalgamation constituent units (namely, mergers occurring among a bigger unit and much smaller ones) to lead to a stronger effect on turnout propensity operating on a limited population (that of the smallest units). In the opposite case of low population polarisation (when merging units are more balanced in population), mergers may have a weaker effect on turnout propensity, which however operates on a larger population.

Still, we need to consider the overall expected effects on turnout taking place in the final units. Some studies have specifically looked at this aspect. Jakobsen and Kjaer (Reference Jakobsen and Kjaer2016) found that (in Denmark) the struggle for representation in amalgamated municipalities is affected by the relative sizes of the former jurisdictions, with higher mobilisation when municipalities of similar size amalgamate. Moreover, the degree of mobilisation of the amalgamated units is more similar in medium cases, in contrast to amalgamated jurisdictions characterised by concentration.Footnote ⁸

This evidence leads us to hypothesise a dual dynamic linking cross-unit population polarisation to turnout variation, a dynamic that presents maximum and minimum expected points. On the one hand, we can expect that turnout after amalgamation will be highest when polarisation is lowest – i.e., when equal-size municipalities are merged (lowest depressive effect). On the other hand, the highest depressive effect on the overall post-amalgamation municipality is not expected to be found at the opposite extreme (i.e., when polarisation is highest). In such a case, the strong depressive effect on turnout operates on a too tiny fraction of the overall post-amalgamation unit (namely, voters of the small units that represent a smaller portion of the final overall population) to be appreciated, and hence, instead, a depressive effect of intermediate strength is expected. To appreciate a remarkable decrease in the post-amalgamation municipality (which can be expected to produce the highest depressive effect on turnout), the population on which the depressive effect operates needs to be large enough. The expected trend thus assumes a sort of v-shape.Footnote ⁹ In formal terms, we then expect that:

Besides the cross-unit population polarisation aspect, amalgamations result from the merger of a varying number of municipalities. This amalgamation feature is also an important element to include in our framework, as recently suggested by Allers et al. (Reference Allers, de Natris, Rienks and de Greef2021). The reasoning behind this is that, in line with the instrumental voting theory, (some) voters turn out to the polls hoping to affect policy outcomes with their vote. From this perspective, amalgamation reduces the probability that one’s vote matters. This reduction increases as more jurisdictions are included. Moreover, the (perception of) internal political efficacy (Niemi et al., Reference Niemi, Craig and Mattei1991) is expected to decrease with the increase of size (here, the number of involved units) since the marginal weight of a single vote on a larger aggregate tends to vanish (Lassen and Serritzlew, Reference Lassen and Serritzlew2011).Footnote ¹⁰ Therefore we hypothesize that:

As anticipated, our theoretical reflection needs to consider classic predictors of electoral participation in municipal elections, regardless of the presence of amalgamations. Gaardsted (Reference Gaardsted2002) argued that a cross-national comparison of the relation between size and local electoral participation is especially interesting when the countries compared differ for structural variables, such as size of the local authority, electoral system, and scope of government – which is exactly the case of our set of 10 countries. For instance, Gaardsted (Reference Gaardsted2002: 858) pointed out that the relatively high autonomy and scope of functions of local governments in the Nordic countries may be related to relatively high local election turnout. Furthermore, as Cancela and Geys (Reference Cancela and Geys2016) stated, on top of population size and composition, concurrent elections and the electoral system play an important role in explaining turnout in subnational elections. More generally, municipal elections are perfect examples of second-order elections (Reif and Schmitt, Reference Reif and Schmitt1980). As such, their turnout is expected to be influenced by the powers of the municipal government, the placement within the national electoral cycle, and the number of municipalities voting together (Bolgherini et al., Reference Bolgherini, Grimaldi and Paparo2021).

Are these features also relevant in cases of municipal merge? Our empirical investigation will reveal the factual answer to this research question. Nonetheless, the core of our scientific contribution lies in the assessment of the effects of the features of the amalgamation process on turnout. Hence, we do not formulate specific hypotheses on these classic predictors of electoral participation in municipal elections as their (dis)confirmation would not add very much to the existing literature. Instead, our framework includes these elements (as our regression models will) as controls, to make sure that our evidence on the hypotheses concerning the effects of the (local and national) amalgamation process is as robust as possible and to assess the impact of the amalgamation process’ characteristics net of the effects of all traditional turnout predictors. Although they do not vary before or after the amalgamation, they can be expected to interfere with the effects that amalgamation features have on voter turnout. It is therefore appropriate, when assessing the theoretically expected effects of the different amalgamation features, to control for them.

Finally, we believe that national political culture needs to be considered, as it may influence the electoral reactions to the amalgamation processes across different countries. Namely, polities more prone to passively accept decisions by the authority may show lower variations in municipal turnout after the introduction of a territorial policy change like the municipal amalgamation.

Data, design, and methods

To empirically assess the outlined hypotheses about the variegated effects of the municipal amalgamation processes on turnout in municipal elections, we have assembled an original dataset consisting of ten European countries –Albania, Austria, Denmark, Finland, Germany, Iceland, Italy, Luxembourg, the Netherlands, and Norway.Footnote ¹¹ The pool of countries includes Northern, Southern, Eastern, and continental European countries, and it is large enough to perform a multi-level analysis – required given the nature of our data.Footnote ¹² Moreover, the selection of these national cases was driven by the necessity to maximise variance on the amalgamation processes’ crucial features highlighted by our theoretical framework– voluntary v. compulsory; massive v. gradual; more or less municipal merging units; more similar- or more different-sized units. Finally, we were forced by data availability – crucially on both the municipal elections and the amalgamations enacted.Footnote ¹³ We have included all municipalities amalgamated in the period 1991–2019 in the selected countries.Footnote ¹⁴

Our dataset features 1,204 post-amalgamation municipalities, constituting the units of analysis of this study, which have merged 4,348 pre-amalgamation municipalities (see Table 1). Thus, on average, each post-amalgamation unit is formed by roughly four constituent municipalities. However, we note relevant cross-country variance. Germany and Albania merged, on average, more than four municipalities into one, Denmark is in line with the pooled mean value, and all remaining countries merged no more than three units.

Table 1.

Municipalities involved in amalgamation processes included in our dataset and turnout variation

Source: Authors’ computation from official electoral data.

Note: Appendix B, Table B2 reports pooled descriptive statistics about both the dependent and independent variables of the regression models (presented below) specifically for the 1,174 municipalities included in them.

For each country, we have consistently measured the electoral data, which will constitute our dependent variable, along with the predictors – both at the lower (municipal) and at the higher (country) level – emerging from our theoretical framework. A discussion about the specific operationalisation of our DV is necessary. Our hypotheses revolve around voter turnout in municipal elections after amalgamation. Consistently with previous amalgamation literature (trying to assess the treatment effect of being amalgamated), it is here operationalised as the turnout variation between pre- and post-amalgamation municipal elections. Although we argue that no single amalgamation treatment effect exists – rather, there are different treatments with different expected effects on turnout – we need to look at the variation to disentangle these different treatment effects. Consequently, we constructed our main DV by subtracting the turnout percentage in pre-amalgamation municipal elections from the turnout percentage in post-amalgamation elections.Footnote ¹⁵ Post-amalgamation turnout percentage was calculated by dividing the number of voters that went to the polls by the number of registered voters in the post-amalgamation unit (and then multiplying by 100).Footnote ¹⁶ We computed the turnout percentage in the pre-amalgamation elections by dividing the sum of the participating voters in the merging municipalities (multiplied by 100) by the sum of the registered voters in those municipalities.Footnote ¹⁷

Overall, the process of municipal amalgamation results in a decline in electoral participation in municipal elections (−2.9 percentage points on average). However, Table 1 reveals remarkable variations across countries. In Denmark and the Netherlands, turnout decreases are way above the mean value (15 and 9 points drop on average); in Italy, Austria, and Albania, on average, turnout decreases similarly to the overall mean; in the remaining countries, the average variation is virtually null or even positive (i.e., turnout increased).

This variance in the turnout after amalgamation (which is not limited to a cross-country comparison – as also indicated by the high values of the standard deviation for each national mean, last column) is particularly interesting to our analytic purposes. Indeed, this article attempts to model this variance according to the predictors highlighted by our theoretical framework with a rigorous cross-national multi-level empirical investigation.

We focus on the amalgamated municipalities only (and not on the comparison between merged and non-merged units),Footnote ¹⁸ as we are interested in the different effects on turnout of different treatments (i.e., different amalgamation processes) – and not in isolating a unified effect of treatment. We have argued that no such treatment exists, precisely as the treatments are different in nature. Conversely, varying expected consequences of the differences in the type of treatments can be meaningfully empirically verified. To this end, we choose a most-different-system design, in which amalgamation is a constant across all the municipal cases included in the dataset, which vary remarkably in terms of the predictors expected (according to our theoretical framework) to influence the impact of amalgamation on electoral participation. Our design strategy is methodologically particularly well-suited to empirically assess how different constellations of municipal amalgamation features impact turnout in the subsequent local-level election, which is the analytic goal of our article. To such end, we need to maximise variance on these process features.

Moving now to the specific measurement choice for our predictors, we point out that they consist of municipal-level variables (measured on post-amalgamation units) and country-level variables (on which values for post-amalgamation units are clustered by country – though often differentiated within country by year of election).Footnote ¹⁹ As selected predictors are at two different levels, and our DV is continuous, we will perform multi-level mixed regression with random intercepts at the country level, which allows the estimation of the effects for the country-level indicators, while best coping with the multi-level clustering of the observations.Footnote ²⁰

We discuss here the measurement strategy for the predictors highlighted by our theoretical framework, whose effects on municipal turnout are in the hypotheses, while the online Appendix A reports similar information concerning the control variables included in our models. Starting with the municipal-level variables pertaining to the characteristics of the amalgamation process in each municipality, we included the number of merging municipal units and the population polarisation. The former simply counts the number of merged municipalities, ranging from 2 to 30. The latter implements the threefold classification proposed by Jakobsen and Kjaer (Reference Jakobsen and Kjaer2016), which distinguishes between amalgamation among equals (score = 3), characterised by one large municipality (=1), or mixed (=2) according to the relative size of the largest, second-largest and smallest pre-amalgamation municipalities. Namely, ‘one large’ represents cases of high cross-unit polarisation – when the largest of the former municipalities is more than twice as large as the second largest; the amalgamation is ‘mixed’ in cases of medium polarisation – when the largest of the former municipalities is more than twice the size of the smallest, but less than twice as large as the second largest; and ‘equal’ is when polarisation is low – the largest of the former municipalities is less than twice as large as the smallest municipality (Jakobsen and Kjaer, Reference Jakobsen and Kjaer2016: 216).Footnote ²¹ At the country level, we include a dummy indicator scoring 1 for municipal elections following forced amalgamations (when a higher-ranked authority has, if willing, the power to impose the amalgamation) and 0 for voluntary amalgamation processes (when either the municipal authority or the citizens have the final say on amalgamation)Footnote ²² ; and a continuous predictor indicating the scope of the specific amalgamation wave within the country. This is calculated as the fraction of the registered voters involved in the amalgamation wave on the overall national voters.Footnote ²³

Findings

We now present and discuss the empirical evidence concerning the hypothesis testing for the effects of the features of the municipal amalgamation over turnout (variation) after amalgamation. Through these analyses, we can answer the focal research question of this study – what accounts for differences in turnout at municipal elections after municipal amalgamations? Are there significant effects for the characteristics of the amalgamation process? Which ones?

The relevant models – also including a control for the flowing of time within the period of observation (days since January 1, 1991) – are reported in Table 2. In a nutshell, they reveal the decisive role of the (national and municipal) characteristics of the amalgamation process. Compared to a baseline model (Model 1) only including the multi-level specification and the control variables, we find that the addition of the national features of the amalgamation (Model 2) reveals how, when the amalgamation is imposed by the national government, net of everything else, this translates into a highly statistically significant negative effect (4.5 percentage points large). The significant negative effect for the dichotomous indicator separating forced amalgamation processes from voluntary ones is a strong empirical confirmation for H1. Voters appear demotivated to participate in the local electoral process when they endure an imposed amalgamation as opposed to when they are included in the decision-making process.

Table 2.

Predictors of turnout variation compared with previous municipal elections in post-amalgamation municipalities

P-values in parentheses. *P < 0.05, **P < 0.01, ***P < 0.001. ICC coefficient for an analogously specified multi-level null model is 0.32.

However, this is counterbalanced by the large, highly significant, positive effect of the continuous indicator measuring the overall (nation-wide) scope of the amalgamation wave. The coefficient indicates that one municipality included in a country-wide reform of municipal boundaries (fully massive reform, 1) has a turnout variation around 12 points higher than a municipality identical on all other aspects but merged alone (gradual reform, almost 0).Footnote ²⁴ This empirically corroborates H2. Voters appear to be particularly dissatisfied with their amalgamation when this is isolated, which brings de-mobilisation and depresses turnout. Yet, they are much less prone to feel alienated when they are not alone in losing their own old municipal entity. In other words, gradual amalgamations are particularly detrimental for turnout compared to large-scale reforms.

Furthermore, the addition (in Model 3) of the local characteristics of the amalgamation process, namely the output and input sides of municipal size, confirms the robustness of the effects of national-level amalgamation features. Moreover, Model 3 indicates the importance of the input-side features – number of units and cross-unit population polarisation. In fact, we find an empirical confirmation for H4, hypothesizing a v-shaped effect of polarisation on turnout. In particular, our analysis unveils that the baseline mixed type has the most depressive effect on turnout, precisely as H4 predicts.Footnote ²⁵ To recall, the mixed type is intermediate in terms of polarisation – namely, the largest of the former municipalities is more than twice the size of the smallest, but less than twice as large as the second largest. On the other end, the two extreme types (‘one large’ and ‘equal’) both show higher levels of voter turnout. The significant positive coefficient for the ‘equal’ amalgamation type suggests that, when similar-sized municipalities are merged, a power struggle between the constituent units (for instance, to control the resources of the new municipal entity) might arise, and that this struggle might have a mobilising effect on voters. This is in line with our arguments that voters in smaller municipalities that are merged with much larger ones do feel less motivated than before to participate in local electionsFootnote ²⁶ ; but that, in order to appreciate the depressive effect on turnout variation, the smaller units need to be not too small – otherwise, the lower participation rate of their former inhabitants risks passing unnoticed/unperceived.Footnote ²⁷ It shall be noted that the effect of moving from ‘mixed’ to ‘one large’ is only marginally positive and the P-value is not significant. However, we deem this not to contradict the overall confirmation of the v-shaped hypothesis as the coefficient is in the expected direction, its P-value close to the .1 slight significance threshold, and considered the lower number of observations used to compute that P-value (not including the ‘equal’ type cases). Moreover, H4 expects the ‘one large’ type to fall in between ‘mixed’ and ‘equal’ (see Appendix B, Figure B1), which is exactly what is found in Model 3. Furthermore, Model 4 (without control variables, see below) indicates a full confirmation for H4 in terms of both statistical significance and ranking of the effects.

Moreover, Model 3 shows that the number of units constituting the post-amalgamation municipality has a positive and significant effect, indicating that the more units are merged, the less turnout tends to decrease. This is contrary to H5, which thus must be rejected. This piece of evidence might appear counterintuitive at a first glance. However, we argue that it can be interpreted as the manifestation of a mechanism possibly traced back to instrumental voting theory. Indeed, as Allers et al. (Reference Allers, de Natris, Rienks and de Greef2021) point out, the amalgamation is also an increase in voters’ power to affect policy through their vote, as the post-amalgamation municipality has more resources than pre-amalgamation ones. This is even more true the higher the number of the amalgamating units. Hence, the positive and significant effect emerging for the number of merged units suggests that voters weigh the increase in value of their vote in terms of higher resources at stake more than the loss of power of their vote due to its dispersion in a larger municipal unit.

Surprisingly, the indicator for the final dimension of the post-amalgamation unit (output-side size), while showing an effect in the expected negative direction, does not yield a significant effect. Thus, we have to reject H3: once national and local characteristics of the amalgamation process are included, overall size is not a significant predictor of turnout in municipal elections. This is a very relevant piece of evidence, which suggests a careful reconsideration of the classic argument about the relation between (local) size and democracy, seminally proposed by Dahl and Tufte.

Finally, Model 4 serves as a crucial robustness test for our hypothesis testing. By replicating the full model without the control variables, we make sure that no result is due to the encompassing control strategy adopted. Model 4 shows the robustness of our findings.Footnote ²⁸ All significant effects are confirmed, while post-amalgamation population is clearly not relevant, thus making the rejection of H3 even stronger. Moreover, Model 4 corroborates the empirical confirmation for H4, as the coefficient for the effect of experiencing a ‘one large’ amalgamation (as opposed to the ‘mixed’ type) is fully statistically significant – but still lower than the ‘equal’ v. ‘mixed’ effect, exactly as expected. In short, in the case of a ‘mixed’ amalgamation, turnout variation will be roughly 2 points lower than in an ‘equal’-type amalgamation, and roughly half of that lower than a ‘one-large’ amalgamation. Thus, exactly as H4 argues, the least detrimental effect on turnout emerges for minimum cross-unit population polarisation values, and the most detrimental effect is found at medium polarisation values.

Overall, our empirical evidence provides a clear confirmation of the explicative role of the features of the specific amalgamation process over municipal electoral turnout. As we have hypothesised, different types of amalgamations do translate into different effects over electoral participation, which vary according to predictable dynamics responding to clear mechanisms. This is something that only a large cross-country comparative investigation, such as the one at hand, could unveil.

Our empirical analyses also reveal some interesting points concerning the classic predictors of electoral participation in municipal elections, included here as control variables. Net of all other variables in the models, the characteristics of the municipal electoral system do not play a role. Both the coefficients for the proportional electoral system and the direct mayoral elections are far from suggesting any statistically relevant effect. The composed marginal effect of the (linear and quadratic indicators for the) placement of the municipal elections within the national electoral cycle is also not significant.Footnote ²⁹ Moreover, national political culture is not significant.Footnote ³⁰

On the other hand, the powers of the municipal level yield a statistically significant effect. In line with the second-order elections theory, the coefficient is positive, indicating that, also after amalgamation, lower local autonomy depresses turnout. Furthermore, a highly significant positive effect emerges for horizontal simultaneity, which confirms that, also with specific reference to post-amalgamation municipalities, holding their municipal elections together with more municipalities increases electoral participation.