Multiple streams framework, decision rules and municipal mergers

Inspired by the “garbage can” model of organizational decision-making (Cohen, March, and Olsen Reference Cohen, March and Olsen1972), the multiple streams framework (MSF) assumes that problems, solutions and politics constitute “streams” flowing through political systems. MSF assumes that policymaking is fundamentally ambiguous, that there are many possible solutions to the same problem, and that problems are social constructs, not objective facts (Herweg, Zahariadis, and Zohlnhöfer Reference Herweg, Zahariadis, Zohlnhöfer, Weible and Sabatier2017, 22). The uptake and framing of problems thus reflect political agency. Policy problems are simply “deviations from people’s understanding of some desired state of affairs” (Béland and Howlett Reference Béland and Howlett2016, 222), and any number of conditions can be construed as a problem meriting political attention. However, time constraints and cognitive limitations mean that policymakers can only devote their attention to so many problems at a time. A basic feature of policymaking is therefore competition for policymakers’ attention and capacity to make decisions.

Kingdon (Reference Kingdon1984) identifies three mechanisms that can lead a problem to gain political attention and inspire a potential change in policy. Sudden, unpredictable, and harmful focusing events, such as a terrorist attack or a pandemic, can push other issues aside and rise to the top of the political agenda. Interest group campaigns are normally conducted in order to draw attention to the group’s pet problem. We investigate the third mechanism identified by Kingdon, namely changes in relevant problem indicators.

Media coverage, routine monitoring activities, and commissioned studies are sources of indicators that enable political decision-makers to assess a problem’s magnitude (Kingdon Reference Kingdon1984, 96). Problem indicators can be taken up and used by “problem brokers” (Knaggård Reference Knaggård2015, 451) in the political system. The selection and interpretation of indicators are key aspects of successful problem definition for actors who are competing for the attention of policymakers in order to effect policy change.

A commonly recognized problem in MSF research is that the transformation of a condition into a problem is hard to predict, as is the subsequent occurrence of a decision-making opportunity. A problem’s objective severity does not automatically correlate with the degree of political attention it receives (Baumgartner, Jones, and Wilkerson Reference Baumgartner, Jones and Wilkerson2011, 953; Dearing and Rogers Reference Dearing and Rogers1996, 91). As noted by Beland (Reference Beland2016, 234), “the window exists in the perceptions of the participants (…) some actors may think a window of legislative opportunity is open while others doubt (…) that the time for reform has arrived.” Indeed, MSF does not even rest on the assumption that policy change is decided on the basis that there is a problem to be resolved. It is just as likely that changes in problem indicators may be used by decision-makers to justify a decision that they wanted to make anyway.

Relatedly, MSF is underdeveloped as regards testable predictions, thus making it difficult to prove it wrong (Kuhlmann Reference Kuhlmann, Zohlnhöfer and Rüb2016; Cairney and Jones Reference Cairney and Jones2016; van der Heijden et al. Reference van der Heijden, Kuhlmann, Lindquist and Wellstead2021). It has been suggested that in order to improve predictability, more MSF research should be devoted to the institutional contexts of policy processes (Sabatier Reference Sabatier and Sabatier1999; Zahariadis Reference Zahariadis2016, 8). As noted by Koebele (Reference Koebele2021, 613), “institutions impact expectations about the roles that different political actors play in processes of policy change and the types of policy proposals that are more likely to be viable.”

In what follows, we respond to calls for increasing attention to be paid to the institutional context by analyzing the associations between problem indicators and ATS change in countries which have different ways of distributing the authority to make decisions about municipal mergers. As noted in the introduction, in some countries mergers require local consent; in others mergers can be imposed by national governments (Swianiewicz et al. Reference Swianiewicz, Gendźwiłł, Houlberg and Klausen2022). This implies that for ATS change to occur, problem indicators would have to grab the attention of decision-makers at the appropriate level of government. Assuming that policymakers at different levels pay unequal attention to different problem indicators, the decision rules in each country may systematically affect ATS change.

In comparing ATS change in countries with centralized and dispersed decision-making rules (“decision rules”) in relation to territorial choice, we depart from a fourfold classification of the institutional contexts for merger decisions provided by Swianiewicz et al. (Reference Swianiewicz, Gendźwiłł, Houlberg and Klausen2022, 55) into two classes: one with centralized and one with dispersed decision rules. The key difference is that it is only countries in the centralized class that allow decision-makers at central level to change the ATS in one fell swoop. In countries in the dispersed class, ATS change is always an aggregate of merger decisions taken at a regional or local level of government.

How would a centralized/dispersed institutional setting affect the potential of various problem indicators to gain the attention of relevant decision-makers? As already mentioned, our expectation is that the uptake of problem indicators and subsequent merger decisions at the level of central government differs from those at lower levels of government. In a study of waterway management, Koebele (Reference Koebele2021) argues that decentralized decision-making institutions necessitate more prolonged policy windows during which attention to a problem must be sustained. Perceptions of problems may be different in relation to the “little windows” at local level in comparison to the “big windows” at central level (Exworthy and Powell Reference Exworthy and Powell2004, 265), possibly because participation in problem definition and solution identification tends to be less differentiated, more inclusive and less dominated by specialists at local level than central level (Robinson and Eller Reference Robinson and Eller2010, 211–212).

As local and regional governments are directly impacted by problem indicators, they might be quicker than central authorities to pick up on them. It is also possible that local and regional decision-makers are primarily attentive to problem indicators of particular relevance to their own jurisdiction, whereas decision-makers at central level are more attentive to system-wide problem indicators. Moreover, the fact that central governments command large bureaucracies with specialized professional staff means that it is more likely that technically complex problem indicators communicated through channels other than the mass media will catch the attention of centralized decision-makers than those at local- or regional level.

At the same time, it is likely that the institutional “stickiness” of jurisdictional borders will tend to be strongest at the local level. As it is individual units that are institutionalized and not the entire administrative system, local decision-makers should be less attentive to problem indicators related to the local government system as a whole than central-level decision-makers. Local decision-makers have a strong aversion to mergers due to the institutionalization of individual local government units (Dente Reference Dente, Dente and Kjellberg1988), and policymakers at national level presumably care less about preserving the existing municipal borders for their own sake.

All in all, there is good reason to study if and how the transformation from problem indicators into ATS change depends on whether the decision-making prerogative over municipal amalgamations is centralized or dispersed. However, the inconclusiveness of the theoretical argument and the lack of directly relevant empirical research means that there is little basis, a priori, for formulating a directional hypothesis that assumes a stronger problem-change association (across different types of problems) in countries with either centralized or dispersed decision-making rules. Our theoretical strategy is therefore to pay close attention to differences between countries with centralized and dispersed decision-making rules when testing three hypotheses relating to problem indicators often used by reformers to justify local government amalgamations: fiscal stress, demographic change and increased scope of local policy.

In terms of fiscal stress, the necessity of increasing economic efficiency is a predominant framing device used by both central and local governments to justify mergers with reference to a temporary fiscal crisis or poor economic development over time. In addition, a widely held theoretical expectation in economics and political science is that reducing the number of local authorities is beneficial in achieving economies of scale (Tavares and Rodrigues Reference Tavares and Rodrigues2015; Silva and Bucek Reference Silva and Bucek2014; Oates Reference Oates1972), even though empirical evidence for a direct relationship between size and costs is inconclusive (Gendźwiłł, Kurniewicz, and Swianiewicz Reference Gendźwiłł, Kurniewicz and Swianiewicz2021; Blom-Hansen et al. Reference Blom-Hansen, Houlberg, Serritzlew and Treisman2016, 2). ATS change (i.e. reduction of the number of municipalities) can therefore be assumed to be associated with fiscal stress, specifically recession or low GDP growth. Our hypothesis is that:

Demographic change in the form of urbanization has been associated with ATS change for several reasons. Migration from rural/non-urban to urban areas puts the jurisdictional structure under pressure. With cities expanding due to urban sprawl and neighboring settlements becoming contiguous conurbations, a growing need arises for integrated planning and policy development over a correspondingly larger area, particularly in terms of land use and transportation (Sharpe Reference Sharpe1979; Meligrana Reference Meligrana2004). Furthermore, city governments might seek to incorporate neighboring jurisdictions in order to counteract freeriding on amenities funded by the city (Olson Reference Olson1969, 482). Territorial restructuring of fragmented metropolitan areas into single jurisdictions has therefore been in evidence in cities across Europe: Hannover and Barcelona for example (Zimmerman, Galland and Harrison Reference Zimmermann, Galland and Harrison2020, 31–32). At the same time, population decline erodes the tax bases of non-urban local authorities and diminishes their economic robustness and functional capacity (Bennett Reference Bennett1993; Hanes Reference Hanes2015). We thus assume increased urbanization to be a problem indicator that increases the likelihood of ATS change. Our hypothesis is that:

Starting in the 1970s, scholars have interpreted ATS change as being associated with the functional expansion of the welfare state (Sharpe Reference Sharpe1979). The merging of local authorities was viewed as part and parcel of the devolution of responsibility for welfare production to local government. Without expansion, local authorities would not have the requisite fiscal, professional, and managerial capacity to deliver universal, rights-based welfare services (Norton Reference Norton1994; Kjellberg Reference Kjellberg1985, 224, 226). For example, the comprehensive amalgamation reforms in Denmark, Norway, and Sweden in the 1960s and 1970s were justified by reference to local authorities having become too small relative to their portfolio of tasks (Hesse and Sharpe Reference Hesse, Sharpe and Hesse1991). The same was the case for post-2000 reforms and reform initiatives in Ireland (Quinn Reference Quinn2015), Norway (Klausen, Askim, and Christensen Reference Klausen, Askim and Christensen2021), and Denmark (Mouritzen Reference Mouritzen, Baldersheim and Rose2010). Comparative studies show that, in some cases, local policy scope has been increased prior to, and in other cases parallel to or shortly following, changes to the ATS (Swianiewicz et al. Reference Swianiewicz, Gendźwiłł, Houlberg and Klausen2022, 134). Irrespective of the order of the two elements, i.e. whether functional expansion has been implemented or is simply anticipated, increased policy scope at the local level is a problem indicator that central decision-makers may pick up and use to justify comprehensive amalgamation reform, and that local decision-makers may engage in to justify enlarging their own municipalities. Our hypothesis is that:

To sum up, the analytical model addresses (a) the associations between three problem indicators and ATS change, and (b) the interactions these associations have with the institutional setting. The theoretical contribution is thus in part to systematically test the three problem indicator hypotheses, supplementing existing partial tests in the literature on local government reform. However, the more novel and important contribution is our exploration of whether the problem/change associations vary depending on the institutional distribution of the final say in decisions concerning municipal mergers.

Research design

Hitherto, the vast bulk of MSF studies have been case studies (Rawat and Morris Reference Rawat and Morris2016; Jones et al. Reference Jones, Peterson, Pierce, Herweg, Bernal, Raney and Zahariadis2016), with just a few exceptions (Liu, Lindquist, and Vedlitz Reference Liu, Lindquist and Vedlitz2011; Robinson and Eller Reference Robinson and Eller2010; Travis and Zahariadis Reference Travis and Zahariadis2002). Qualitative designs have been appropriate in view of the need for thick empirical evidence to capture the complex MSF framework in its entirety. However, for the purposes of our study, a quantitative large-N design seems more appropriate. Considering the broad range of local contingencies that might affect ATS change in each particular case, it can be assumed that systematic correlations between our problem indicators and the occurrence of ATS change are quite weak, and therefore more likely to be observed by large-N than small-N designs (Gerring Reference Gerring2017, 54). An added benefit of our analytical strategy is that the data available pertaining to theoretically relevant problem indicators enables the use of “objective” (registry-based) indicators, an approach suggested by Jones and Baumgartner (Reference Jones and Baumgartner2005, 208) as well as by Zohlnhöfer, Herweg and Zahariadas (Reference Zohlnhöfer, Herweg, Zahariadis, Weible and Workman2022, 41–43). On a more general level, our study responds to a call from leading authors in the field for more rigorous, quantitative, large-N testing of key assumptions behind MSF (Herweg et al. Reference Herweg, Zahariadis, Zohlnhöfer, Weible and Sabatier2017) in order to improve external validity.

The application of a large-N analytical strategy to test the MSF simultaneously entails certain delimitations. As suggested by Herweg, Zahariadis and Zohlnhöfer (Reference Herweg, Zahariadis, Zohlnhöfer, Weible and Sabatier2017), it would not be possible for a large-N empirical design to realistically analyze the entire MSF framework in all its complexity, but rather it should focus on partial testing of framework assumptions. We consequently limit our analysis to the relationship between problem indicators and policy change, assuming a link between indicators, political focus and decision-making. An unavoidable consequence is that other constituent parts of the MSF are omitted.

Data and methods

Our empirical analyses consider 39 European countries between 1990 and 2020. Compared to, for example, Askim et al. (Reference Askim, Klausen, Vabo and Bjurstrøm2017), we move beyond Western European countries and extend the time frame for the analysis. Our data is also more detailed than that of previous comparative studies of ATS change, as countries are represented by a yearly time series. As the unit of analysis is the country-year, we use a time-series cross-sectional research design. The panel data set used in the subsequent analyses is well balanced, with the exception of some data covering the early 1990s.

Descriptive statistics are presented in Table 1 and in the Appendix (Table A1). Additionally, the online Supplementary Material contains one graph per country visualizing the dependent variable and the three main independent variables.

Table 1. Incidences of administrative-territorial system (ATS) change in 39 European countries, 1990–2020

* Classification of decision rules adapted from Swianiewicz et al. (Reference Swianiewicz, Gendźwiłł, Houlberg and Klausen2022, 42). Differences in periods covered stem from the differences in data availability.

The dependent variable is instances of ATS as measured by a substantial territorial consolidation of the municipal tier. We rely on the Local Autonomy Index (LAI) data set (Ladner, Keuffer, and Bastianen Reference Ladner, Keuffer and Bastianen2021) and trace the number of municipalities in each country year by year. The total number of municipalities in the 39 countries decreased by about 13% between the early 1990s and 2020. In line with previous studies (Askim et al. Reference Askim, Klausen, Vabo and Bjurstrøm2017; Swianiewicz et al. Reference Swianiewicz, Gendźwiłł, Houlberg and Klausen2022), we consider a decrease in the number of territorial units exceeding 5% to be substantial (in the additional analyses, we also check alternative thresholds). We cover not only easily identifiable one-off changes, such as the 2007 Danish reform, but also incremental changes, such as the continual consolidations in the Netherlands and Iceland (see Figure 1).

Figure 1. Incremental (Iceland, left panel) versus one-off (Denmark, right panel) ATS change. Note: Dashed lines mark observations identified in accordance with the “>5% rule,” i.e. years in which the accumulated decrease in the number of municipalities exceeds 5%.

We labeled a country-year as an event of substantial ATS change if the accumulated decrease in the number of municipalities in that particular year exceeded 5%, either from the first year in the time series or from the last substantial ATS change. This resulted in the identification of 61 instances of substantial ATS change, coded as a dummy variable (1 = change, 0 = no change) (see Table 1).

With a binary dependent variable and panel data structure, we use a logistic regression model with random effectsFootnote ¹ for countries and robust standard errors clustered by countries. The three problem indicators are represented by the lagged variables as we expect that decision-makers, whether at central or local level, need some time to formulate their perception of problems before deciding to initiate or accelerate ATS change. Our baseline formulation uses a two-year “window” in which we monitor the indicators of fiscal stress (recession), urbanization dynamics and increased policy scope. In order to manage outliers in these variables and simplify interpretation of the results, we dummy-coded “fiscal stress” and “policy scope.”

Fiscal stress was measured by a dummy variable with the value 1 if a country was in recession (i.e. a negative GDP growth rate) in at least one of the preceding two years t − 1 or t − 2. We based our coding on the World Bank database (annual percentage growth rate of GDP at market prices based on constant local currency).

Urbanization was measured as a positive change in the proportion of urban population between t − 1 and t − 2. Higher values of this indicator are mainly the result of depopulation of rural areas due to internal migration. The data was derived from the World Bank (2023) database.

Increased policy scope was measured using LAI data based on an expert survey conducted by Ladner et al. (Reference Ladner, Keuffer and Bastianen2021). Policy scope represents the range of functions for which local authorities assume responsibility for service delivery. LAI’s policy scope index covers education, social assistance, healthcare, land use, public transport, housing, police and care functions. We recoded the LAI data into a dummy variable that adopts the value 1 for any increase in the value of the policy scope index between t − 1 and t − 2. A consequence of this operationalization is that the test of H3 is only partial. While we do record instances of policy scope increasing prior to ATS change, we do not capture instances of policy scope increasing in tandem with or subsequent to ATS change. The test should therefore be considered conservative.

We supplemented our models with control variables. We firstly controlled for each country’s territorial fragmentation in t − 1 (measured by the logged average population size of a local government unit; lower values represent a high level of territorial fragmentation). We also added a dummy variable representing the year in which national parliamentary elections were held. As more substantial changes in the ATS are highly contested, the assumption is that governments in centralized settings tend to avoid introducing risky reforms that could impact adversely on their re-election prospects. In dispersed settings, where decisions are made at the sub-national level, a similar motivation might affect the decisions of local party branches and local or regional leaders affiliated with national parties.

As our hypothesis implies that the institutional setting moderates the relationship between problem indicators and the occurrence of ATS change, we interacted the explanatory variables with the binary variable representing the distinction between centralized and dispersed settings. The use of dummy variables facilitates the interpretation of models by means of interaction terms.

Table 1 shows the 17 countries with centralized decision rules for amalgamating local authorities and the 22 countries with dispersed decision rules. Occurrences of ATS change (>5% reduction in the number of local authorities) are indicated in parentheses.

The classification of countries in terms of decision rules was based on a recent comparative study (Swianiewicz et al. Reference Swianiewicz, Gendźwiłł, Houlberg and Klausen2022, 55). In countries with centralized decision rules (corresponding to the “nationalised” class in the previous study), the legal-constitutional set-up and administrative traditions permit mergers that are centrally imposed. Although mergers in these countries have historically been implemented as a mixture of (more or less) voluntary and imposed mergers, the key classification factor was that the central government held a legal mandate to impose mergers and had historically used that mandate.

Countries with dispersed decision rules comprise the remaining three classes in Swianiewicz et al.’s (Reference Swianiewicz, Gendźwiłł, Houlberg and Klausen2022) classification. Firstly, in countries classified as “formally local,” the legal-constitutional setting explicitly prohibits the imposition of mergers by higher-level governments, which means that merger decisions are always made at the local level. Secondly, in the “customarily local” group of countries, long-standing norms and traditions have effectively ruled out imposed mergers - sometimes for centuries - although a higher-level government did hold a legal mandate for making such decisions. Thirdly, the “regionalised” group comprised three of the federations included in the sample (Germany, Austria and Switzerland), as well as the UK and the Netherlands. Mergers in these countries have historically been decided through mixtures of local voluntarism and imposition by provincial-, state (Länder)- or country-level governments, without the involvement of federal or national governments.

We recognize that the centralized/dispersed classification subsumes a broad variety of scenarios in relation to ATS change. For instance, whereas some countries with centralized rule systems have implemented broadly based reforms through a central decision and with little involvement or influence from local authorities, reforms in several other centralized countries have comprised an initial phase where local authorities were allowed to choose their partners, or even defer the merger decision altogether. Whereas in some countries with centralized decision rules, mergers are always initiated by the government, other countries have experienced mergers as a result of local initiatives in periods between government-initiated reforms. For a more thorough discussion of alternative scenarios, see Swianiewicz et al. (Reference Swianiewicz, Gendźwiłł, Houlberg and Klausen2022, 56–59).

In the next section, we first present our baseline models, which provide the empirical tests for our hypotheses, and subsequently proceed to the supplementary analyses, which serve as robustness tests and verify the sensitivity of our results to the principal arbitrary decisions we made at the operationalization stage.