Principals or puppets? Analysing variation in local political parties’ social policy positions

The decentralisation of policy responsibilities from the national to the local level in the domain of social policies is meant to facilitate a better fit to local conditions, and, consequently, inspire local variation in social policy positions. This article examines two questions: (1) to what extent do Dutch local party branches’ social policy positions deviate from their national mother party and local peer parties and (2) do local conditions explain this deviation? To answer these questions, we developed a dataset including 168 local party manifestos from 27 strategically selected municipalities and 8 national party manifestos. Our analyses show limited deviation in local parties’ positions compared to their national mother party and other local branches of their national mother party. This suggests that the social policies addressed in the party manifestos of local parties seem to reflect a process of institutional isomorphism. Furthermore, the limited deviation that does exist in local parties’ social policy positions is not convincingly larger in municipalities (1) that are smaller, (2) that have higher social benefit dependency, or (3) that have high vote shares for local independent challengers. This is contrary to what can be expected based on the contingency theory.


Introduction
National governments often justify decentralisation by pointing to a tighter fit between policy measures and the problems those policies aim to address (Costa-Font and Greer 2013, p. 6). In line with this approach, it can be argued that local parties are better equipped to tailor policies to local needs as they are better informed about local circumstances (Boogers and Voerman 2010;Clark 2004). Consequently, considering the variation in needs and circumstances across municipalities, deviation in social policy positions from national parties is expected. This resembles the so-called contingency theory (Donaldson 2001), claiming that a misfit with local conditions inspires the feedback needed to fine-tune policies to local conditions. With the widespread trend towards decentralisation of social policy responsibilities and the devolution of authorities from national to local governments (Borghi and van Berkel 2007;de Vries 2000;Falleti 2005), local branches, instead of national political parties, become responsible for developing social policies.
However, adjustments to local circumstances could generate tensions between their role as political players in the local arena on the one hand, and as representatives of their national mother party's ideology on the other (Geser 1999). Moreover, insights from institutional isomorphism theory (Dimaggio and Powell 1983) suggest that local deviations in social policy positions are not self-evident. Various institutional pressures exist in the type of complex organisational environments in which local political party branches must operate (Boogers and Voerman 2010). More specifically, pressures to align with the central party's preferences may impede the adjustments to local circumstances that local party branches try to make. As such, institutional pressures could hamper achieving the desired fine-tuning of decentralised social policies to local conditions. The 2015 decentralisation of the Participation Act in the Netherlands (for a more elaborate description see : Dijkhoff 2014), provides a good opportunity to examine whether decentralisation of social policy leads to the desired local political fine-tuning or not. The Participation Act obliges municipalities to provide social welfare services, such as matching, training, and subsidised employment", as well as "income support for less than fully disabled persons who have no employment history (as opposed to former employees)" (Vermeulen 2017, p. 128).
Our main research questions are twofold: (1) to what extent do Dutch local party branches' social policy positions deviate from their national mother parties and local peer parties and (2) do local conditions explain this deviation?
As decentralisation is still in its early stage, we focus on local party manifestos drawn up for local elections to answer these questions, as these are the documents in which local variation will first manifest itself. The production of local party manifestos has been viewed as the indicator of party politicisation of local governments (Ashworth 2000;Jones 1975). "These manifestos outline policies that parties will enact once elected to legislative or executive office" (Lowe et al. 2011, p. 124). Using party manifestos allows us to examine whether local party branches operate as local 'principals' or national 'puppets' in determining their positions concerning the new social policy responsibilities. We also think that, despite the one-policy and one-country focus in this research, our findings could provide relevant insights for other types of decentralised policies and for the politics of decentralising social policies from national to local governments in other countries.
We developed a unique data set to answer our research questions. It includes relevant information on local party manifestos for more than two dozen strategically selected Dutch municipalities that vary on important aspects, e.g. population size, social benefit dependency, political constellation, and region (e.g. Breeman et al. 2015). We followed a widely used technique (cf. Cole 2005; Gross and Jankowski 2018) to systematically code and analyse the social policy positions reported in those manifestos. The dataset contains 168 manifestos of local party branches, in addition to national party manifestos from the eight largest political parties.
The constitutional design of the Netherlands is a decentralised unitary system in which each of 380 municipalities has its own democratic legitimacy (Andeweg and Irwin 2005;Breeman et al. 2015). All municipal councils are directly elected every 4 years on a fixed day. Whilst the electoral support for independent local parties has increased, local party branches still make up the lion's share of local electoral support in the Netherlands (Boogers and Voerman 2010).
This article is structured as follows: Sect. 2 explores the literature on institutional isomorphism and contingency theory in order to find theoretical scenarios for how social policy positions are likely to be determined by local parties. Section 3 outlines the data and methods used. Section 4 presents the results, and, in Sect. 5, we discuss the implications of our findings.

Local party's social policy positions: contingency versus institutional pressures
Two specific theories inform scenarios for local party behaviour pertaining to the decentralisation of social policies. Institutional isomorphism theory emphasises considerations which would 'force' local party branches to follow their national mother parties, while contingency theory points to conditions that would 'force' local parties to adapt social policy positions to local conditions.

Institutional isomorphism
The desire to adapt to local conditions might not be as self-evident for local party branches as it seems (Boogers and Voerman 2010). Since political parties represent specific ideologies and normative viewpoints, local party representatives will generally hold like-minded ideas. Accordingly, a situation can arise in which social policy positioning of national parties and their local branches shows a strong resemblance, regardless of differences in local conditions. Institutional isomorphism theory suggests that three specific institutional pressures are key drivers of public organisation behaviour and design in their quest for legitimacy: coercive, normative, and mimetic (George and Desmidt 2014). Coercive pressures reflect the imposition of organisational patterns by the authority on which organisations depend. Normative pressures are associated with adopting patterns or norms which are considered appropriate. Finally, mimetic pressures result in adopting patterns of, supposedly, successful organisations (Dimaggio and Powell 1983). Due to one or more of these pressures, a situation of so-called 'institutional isomorphism' emerges: i.e. a situation where all organisations look rather similar (Dimaggio and Powell 1983;George et al. 2018).
At least two sources of isomorphism may lead to congruent positions among local party branches. First, the ideological congruence of local party branches with their national mother party can inspire normative isomorphism. Shared values will, in this scenario, translate into similar social policy positions. Second, national party offices can inspire coercive isomorphism by facilitating the drafting of local party manifestos by organising meetings and supplying information, and sometimes even templates (cf. George and Desmidt 2014;Ashworth et al. 2007). Which specific form of isomorphism is possibly in play cannot be inferred from the data available, but generally isomorphism theory tells us to expect that local branches from the same national mother party demonstrate similar social policy positions in their manifestos.

Contingency theory
Contingency theory claims that contextual characteristics are key drivers of organisational behaviour (Donaldson 2001). Contingency determinants refer to pressures that originate from outside the control of the organisation, such as economic developments (Donaldson 1987). Previous research using contingency theory reveals that organisations adapt their structures to create a 'fit' between the organisational characteristics and context conditions (Donaldson 2001;George and Desmidt 2014). As argued by Donaldson (2001, p. 12), "structural change occurs in response to contingency change and is triggered by the feedback effect from low performance caused by misfit". This feedback effect resonates with the basic assumption of policy decentralisation: local politicians and policy makers have a better view of the specific conditions that a given policy aims to address. This implies that the positions of local branches of national parties may differ from the ideological position of their mother party to create a better fit with local conditions: "if local parties wish to participate in local government, they must adopt policies relevant to the community they aspire to represent" (Clark 2004, pp. 40-41). Various local conditions could be relevant in determining local policies (Geser 1999), but for the aim of this research-determining national-local position deviation rather than ideological positioning of parties through time (e.g. Adams et al. 2004Adams et al. , 2009)-we focus on three likely candidates: population size, dependency on social benefits, and the salience of local independent parties.
Population size is possibly relevant in the delegation of sovereignty from voters to popular representatives (Bovens 2007). We suggest that the accountability mechanism associated with this sovereignty is more salient in municipalities with a smaller population size than in municipalities with a larger population size, as politicians in the former operate closer to citizens. Consequently, they are more likely to be held accountable than politicians in municipalities with a larger population size. Therefore, it can be expected that local party branches in smaller municipalities are more likely to modify their manifesto in line with local circumstances (cf. Stoker 1991). As a result, we expect that deviation in social policy positions from the national mother party and local peer parties occurs more often in municipalities with a smaller population size.
Based on the party competition theory (Budge and Farlie 1983), the importance of a topic could also be influential for social policy position development. Therefore, the second context condition that is possibly relevant is the share of citizens depending on social benefits. High shares lead to more pressure on local parties to address this social problem and, thus, to tailor their social policy positions to local conditions-Pogorelis and colleagues have already demonstrated something similar for local elections in Scotland and England (2005). In short, we expect that deviation in social policy positions from the national mother party and local peer parties occurs more often in municipalities with higher levels of social benefit dependency.
We expect that the third context condition relevant for local party behaviour is the salience of independent local parties. Independent parties can play a distinctive role in local politics as they have no ties with national parties. They can clearly focus on local issues and are 'said to be more responsive than local party branches to local issues' (Boogers and Voerman 2010, p. 78). This representation of local issues would pose no threat for local party branches if local independent parties were marginal players, as was roughly the case until the 1990s. However, their large upswing since then has created anxiety among local party branches. In line with the threat mechanism theory, we expect that the salience of local independent challengers could increase the (perceived) risk that local party branches might lose seats in the next election (Boyne et al. 2001). Consequently, local party branches facing strong competition from local independent parties feel more inclined to modify their manifesto to local circumstances. Hence, we expect that deviation in social policy positions from the national mother party and local peer parties occurs more often in municipalities where independent local parties receive high vote shares.

Data and methods
A dataset with Dutch local party manifestos does yet not exist. Therefore, we travelled great lengths to develop one that contains all necessary information needed to empirically scrutinise the expectations outlined above. Our dataset allowed us to examine if variation in local social policy positions existed, and if so, whether this variation was patterned in line with the contingency theory. In order to do so, we measured social policy positions in two ways: topic presence and position-taking.

Case selection and coding
First, we followed Van de Bovenkamp and Vollaard (2017) in strategically selecting municipalities that contrasted in characteristics. 1 Following the selection process of Andeweg and Irwin (2005), Breeman et al (2015), and Van de Waart and Brouwer (2015), we selected nine small, nine medium, and nine large municipalities that showed substantial variation in social benefit dependency, political constellation at the time of the manifesto development, and the region of the Netherlands in which they are located. 2 This brought the total number of municipalities in our dataset to 27.
The party manifestos of the eight largest national parties and their local branches in the selected municipalities were used to determine Participation Act positions. Considering that not all parties participated in local elections in each selected municipality and that some manifestos were unavailable, in total, 8 national and 168 local party manifestos were collected and analysed. 3 Online Appendix 1 shows which party manifestos are included in our analyses.
The inductively developed coding scheme can be found in Table 1. Only after intensively reading position statements of national political parties on the Participation Act, and multiple discussions in face-to-face meetings between the authors of this article, the coding scheme became a predefined format for analysing the local branches.
We first found seven regularly occurring main topics and identified whether these were mentioned in a particular manifesto or not: (1) Decentralisation of the  (7) Welfare experiments. We then discerned 30 subtopics within the seven main topics that covered a range of relevant issues concerning the Participation Act. For example: do parties address decentralisation in their manifesto? To whom do they assign responsibility for the implementation? Is some performance required in return for receiving benefits? And what should be done about welfare fraud? With this codebook, we determined the 'social policy positions' of local political party branches.

Operationalisation
The 'social policy positions' of local political party branches consist of two elements: 'topic presence' and 'position-taking' on the Participation Act. Following Breeman et al. (2015), topic presence measures the share of manifestos that pay attention to a main topic or subtopic for each of the eight political parties included. As such, topic presence was used to measure whether political parties deviate in topic presence per subtopic from: (1) their national mother party, or (2) their peer local branches from the same national mother party in other municipalities. It is thus based on discerning manifestos with code 'no' on a given subtopic from manifestos with code 'yes' (see Table 1). Code 'no' represents party manifestos that did not present a position on a given subtopic, whereas code 'yes' represents party manifestos that did present a position. Two types of deviation in topic presence were discerned: (1) local versus national deviation and (2) local versus local deviation. Local versus national topic presence deviation measures the share of subtopics local branches (i.e. local manifestos) deviate on from the mother party. Local party branches deviated from the national mother party if one emphasised a subtopic, while the other one did not. Local versus local topic presence deviation indicates the share of subtopics local branches deviate on from what the majority of local branches of their own party reports (50% of the local branches plus 1). 5 This means that the local majority's topic presence is a dichotomy: the majority addresses a topic, or it does not. Local party branches deviated from the majority of local branches of their own party if the former emphasised a subtopic, while the latter did not, or vice versa.
The second element of social policy positions is position-taking. It measures whether local political parties have a different position on a subtopic than: (1) their national mother party, or (2) their local party peers (local branches of the same national mother party). As depicted in Table 1, 26 of the 30 subtopics have three code options: there is no position (code A), there is position B, or there is position C. The remaining four subtopics have four code options. A party deviates from its mother party or local majority if its code deviates from the code of the mother party or local majority. For example, a local party claims to be in favour of basic income experiments (row 29: code B), whereas the national mother party or the majority of local branches of their own party claim(s) to be opposed to this idea (code C).
Local versus national position-taking deviation is measured in the exact same way as the measurement of local versus national topic presence outlined above. Local versus local position-taking deviation is measured as the share of subtopics local branches deviate on from the majority of local branches of their own party.
To scrutinise our expectations, we linked deviation in topic presence and position-taking to three municipality characteristics: population size, social benefit dependency, and vote share for local independent parties. Population size ranged from 26,055 (Wassenaar) to 200,914 (Almere); this being the number of inhabitants in 2017 (Statistics Netherlands 2017a).
Social benefit dependency is measured by means of the score of a municipality on a composite indicator-ranging from 3.5% (Lansingerland) to 8.9% (Kerkrade)comprising three items. The first item measures the unemployment rate per municipality and ranged from 4.2% (Barneveld) to 8.3% (Almere) in the labour force in 2017 (Statistics Netherlands 2017b). The second item measures the share of inhabitants relying on welfare (bijstand) between 2016 and 2017 and ranged from 1.3% (Almere) to 6.7% (Nijmegen) (Association of Dutch Municipalities 2018). The third item measures the share of inhabitants relying on a disability pensions (arbeidsongeschiktheiduitkering) and ranged from 3.8% (Lansingerland) to 13.8% (Kerkrade) in 2016 (Ministry of Health, Welfare and Sport, 2016). The three items form a reliable scale (Cronbach's α = 0.781), with an explained variance of 74.4% and factor loadings ranging from 0.77 (reliance on disability pension) to 0.88 (unemployment rate) to 0.93 (reliance on welfare), respectively.

Results
We present our findings on social policy positions in three steps. First, we discuss whether we observe similarity or deviation in topic presence. This is followed by exploring the associations between topic presence deviation on the one hand and the local conditions of population size, social benefit dependency, and share of votes for local independent parties, on the other. Finally, we explore the association between position-taking deviation and those local conditions. Figure 1 demonstrates topic presence for all parties on the main Participation Act topics (determined by addressing at least one of its subtopics). The order of the topics is determined by the topic presence in national party manifestos. Figure 1 signals two clear patterns. First, although almost all manifestos pay attention to some aspects of the Participation Act (88%), we see that, if the data is disaggregated into main topics, both at the national (black dots) and local (bars) level, the attention paid to any of the topics is rather low. Tellingly, on average, each main topic is emphasised by less than 50% of the national parties. This percentage is even lower for the local branches (43%).

Topic presence
Second, there seems to be an association between the presence of main Participation Act topics in national and local manifestos: topics that are addressed most at the national level, 'Participation Jobs' (100%) and 'Reintegration Trajectories' (63%), are also addressed most in local manifestos (respectively, 70% and 70%). Attention paid to the 'Decentralisation of the Participation Act' (national: 0%; local: 24%) and 'Compulsory Job Applications' (national: 38%; local: 23%) are relatively low at both levels. Finally, in sharp contrast to the national level (63%), local manifestos hardly stress 'Welfare Fraud' (23%). Figure 2 shows the similarity in topic presence between national party manifestos and local party manifestos per party. Of course, this means that, at the national level, topic presence is a dichotomy: a given party addresses a certain topic or does not. We ordered the topics on the horizontal axis as follows: the topics addressed in a national-level manifesto are depicted to the left, while those not addressed are depicted to the right. The bars, again, depict the share of local manifestos that addressed a certain topic. The order of the topics is determined by the topic presence in local party manifestos.
Three clear patterns arise from Fig. 2. The first two concern the local versus national topic presence, whereas the final pattern concerns the local versus local  Percentage of local party branches addressing the main topic in their manifesto. Dotted line: left from line are topics addressed by the mother party, right from the line are topics not addressed by the mother party topic presence. First, the association between national and local topic presence proves stronger when the data is analysed per party (Fig. 2) than in general (Fig. 1). For almost all parties considered, topics addressed in national party manifestos are more often also addressed in local manifestos than topics unaddressed in national party manifestos, and vice versa. However, the strength of this association between national and local topic presence seems to vary across parties. The strongest relationship can be found in the Reformed Political Party (SGP): topics (not) addressed in the national manifesto are (not) addressed by nearly all local SGP branches (Fig. 2c). The Christian-Democratic Appeal (CDA), on the other hand, shows the weakest relationship: its branches less often address topics addressed by the national party than topics unaddressed by the national party (Fig. 2e).
Second, notwithstanding these associations, a limited number of local branches also deviate from their national party manifestos in terms of topic presence. For example, 'Compulsory Job Applications' is a relevant topic for local People's Party for Freedom and Democracy (VVD) branches, whereas the mother party does not mention it. The Labour Party (PvdA), on the other hand, shows the exact opposite situation: 'Compulsory Job Applications' is addressed by the national party, but is barely mentioned by her local branches. Now we turn to the analysis of local versus local topic presence. By focusing solely on the bars, we observe that local branches not only show strong resemblance in topic presence to their national party, but also to their party peers. Local branches completely agree on topic presence if the share of local branches addressing a main topic is either 0% (no local branch addresses the topic) or 100% (all local branches address the topic). As an illustration, the local Socialist Party (SP) branches show strong similarities in topic presence. Topics are either addressed by more than 60% or less than 20% of the local branches. For example, 'Participation Jobs' is addressed by 85% of the local SP branches, whereas 'Welfare Experiments' only by one of the local SP branches (Fig. 2h).
Thus, the general patterns reveal strong resemblance in topic presence, both for local versus national topic presence and for local versus local topic presence within a party. Hence, institutional isomorphism seems to be in play in the formulation of social policy positions by local party branches. However, to some extent, we also observe deviation in topic presence. The next section explores whether this can be linked to local conditions.

Topic presence deviation by local conditions
The observation that there is some topic presence deviation leads to the follow-up question: under which circumstances do local party branches deviate? Here, we will scrutinise our central expectations that in smaller municipalities, in municipalities with higher levels of social benefit dependency, and in municipalities with higher shares of votes for local independent challenger parties, both local versus national topic presence deviation and local versus local topic presence deviation will be larger. The limited statistical power accompanying a dataset of 27 cases makes focussing on levels of significance not informative, and we will consequently merely focus on the size and direction of the regression coefficient for the theoretical interpretation of our findings. Figure 3a demonstrates the relationship between population size and both types of topic presence deviation. Each dot depicts topic presence deviation per municipality. The black dots represent the average share of subtopics local branches deviate on from their national party in terms of topic presence per municipality (e.g. a score of 30% means that in a municipality, local branches, on average, deviate on 9 out of the 30 subtopics from their mother party). The white dots represent the share of subtopics that local branches deviate on from the majority of their party peers per municipality (e.g. a score of 7% means that in this municipality, local branches on average deviate on 2 out of the 30 subtopics from the majority of their party peers). Figure 3a suggests two things. First, there is a relationship between local versus national topic presence deviation and population size. This relationship is in the expected direction: local party branches in smaller municipalities are more likely to diverge in terms of topic presence from their mother party than local party branches in larger municipalities. This coincides with the contingency scenario. Second, the relationship between local versus local position-taking deviation is in the opposite direction than expected. In other words, party branches in municipalities with a larger population size are more likely to diverge from their peer parties' topic presence than party branches in municipalities with a smaller population size. Thus, the local versus local topic presence deviation findings are contrary to what was expected and, as such, are not in line with our interpretation of the contingency theory scenario. Figure 3b portrays the results for social benefit dependency and shows that the relationship between this dependency and local versus national and local versus local topic presence deviation is not in the expected direction. Party branches in municipalities with higher social benefit dependency are less, instead of more, likely to diverge from their mother party and peer parties in terms of topic presence. In short, the relationship between social benefit dependency and local variation in topic presence deviation is contrary to the contingency scenario.
Finally, Fig. 3c depicts the relationship between the share of votes for local independent challengers and both types of topic presence deviation. Although the relationship between local versus national topic presence deviation and that share is in the expected direction, we consider it too weak to convincingly corroborate the expectation that local party branches deviate more in municipalities with a larger share votes for local independent challengers. Similarly, the relationship between local versus local topic presence deviation and the share of votes for local independent challengers is far too weak to be interpreted in accordance with the contingency scenario.
All in all, most findings clearly demonstrate that deviation in topic presence does not vary across municipalities in the way that contingency theory predicts. At least, not in the case of the contingencies of social benefit dependency and the salience of local party challengers. Only the population size associates with topic presence deviation according to the contingency theory scenario. And even then, this is only in the case of local versus national topic presence deviation, and marginally so at that.

Position-taking deviation by local conditions
In addition to choosing whether to focus on a specific topic, local political parties can also differ from their mother party, as well as from the majority of local branches of their own party, in the position they take on a specific topic. This is measured by means of position-taking. As outlined in the operationalisation section above, we apply the same analyses on local versus national position-taking deviation and local versus local position-taking deviation as we did for topic presence deviation. Consequently, first we analyse whether we observe deviation in positiontaking and, subsequently, explore if the observed deviation is related to municipal conditions. We merely depict the relationships between the two types of deviation and the local conditions, as these provide all the necessary information (Fig. 4a-c).
Position-taking deviation proves to be low: local versus national position-taking deviation ranges from 4% to 10%, and local versus local position-taking deviation ranges from 2% to 9% (Y-axes). This low level of deviation in position-taking should be considered when analysing its relationship with possible predictors. Figure 4a shows the relationship between both types of social position-taking deviation and population size that are not in line with the contingency theory scenario: i.e. we observe more deviation in municipalities with a larger population size. Figure 4b contrarily reports findings that are in line with that scenario. Especially local versus local position-taking deviation seems to be strongly related to social benefit dependency. Hence, local party branches are more likely to diverge from their mother or peer parties in municipalities in which this dependency is higher. The relationships reported in Fig. 4c are also in accordance with the contingency theory scenario: local party branches are more likely to diverge in position-taking from their mother party, and especially their local peer parties, in municipalities where the share of votes for local independent parties is higher. Overall, three out of the six patterns depicted in Figs. 3 and four out of the six depicted in Fig. 4 can be interpreted according to the contingency theory scenario. Yet, only one of the seven relationships in accordance with that scenario is substantial in strength: local to local position-taking deviation by social benefit dependency. In sum, there is very limited evidence that local parties, generally, aim to tailor their social policies to local conditions, at least not to the three municipal conditions that we focussed on. 6

Conclusions and discussion
In this article, we addressed two questions. To what extent do Dutch local party branches' social policy positions deviate from their national mother party and local peer parties? And do local conditions explain this deviation? To answer those questions, we strategically selected 27 Dutch municipalities, and subsequently created a dataset of 168 local party manifestos and 8 national party manifestos, all from 2018. Even though we searched for maximum variation in relevant municipal conditions, we observed that topic presence and, especially, position-taking in local party branches manifestos are (still) strongly associated with the topic presence and position-taking of the national mother party and peer local branches.
As far as local party branches do deviate, it can hardly be interpreted according to the contingency theory scenario. We do not see that, generally, deviation is larger in municipalities that are smaller, that have higher levels of dependency on social benefits, or that have higher vote shares for independent local challengers. The only relationship convincingly in line with that scenario is that, in the limited cases in which local party branches do deviate from the local party branches' majority, it occurs in municipalities with higher dependency on social benefits. Additionally, in the case of the relationship between population size and deviation, it might be that the findings are suppressed by the mechanism that allows local party branches in larger municipalities to have the power and the resources to act more independently from their mother parties than in smaller municipalities. But overall, the social policies addressed in the party manifestos of local party branches seem to reflect institutional isomorphism more than adaptation to local contingencies.
There are at least two ways in which institutional isomorphism could have played a role in this: (1) normative isomorphism, i.e. the similarities reflect correspondence in normative viewpoints and party ideology, and (2) coercive isomorphism, i.e. they reflect imposition by the mother party. The first way is more in line with previous findings that the ideological bases of parties are good predictors for local party positions, even in times of decentralisation (Pogorelis et al. 2005). The second way may be informed by the existence of the so-called 'basic party programs' (basis verkiesprogramma's). These are drawn up by the national mother party as an example for local party branches to use in drafting their manifestos. This support might, however, result in copying and pasting this basic programme and only adjusting the bare minimum to make it fit the local circumstances. However, it is beyond the scope of this research to determine how institutional isomorphism could have played a role in drafting party manifestos of local party branches.
Besides the role of institutional isomorphism, it might be that local conditions are hardly related to local party positions on social policies in accordance with the contingency theory scenario because local elections are determined by national as opposed to local issues. "Local election results overwhelmingly reflect national swings of opinion for or against the incumbent government" (Stoker 1991, p. 52). Institutional differences could also play a role. These might, for instance, account for why Pogorelis et al. (2005) found that poor economic conditions prompted position deviation in Scotland and Wales, while higher levels of dependency on social benefits did not in our study. Future research focussing on similar questions but other institutional settings than the Netherlands could shed more light on this. Moreover, party characteristics could be more relevant than municipal characteristics for local party positions on social policies; for instance, whether there is inclusive or exclusive internal party democracy as described by Lehrer (2012).
Two other reasons could also account for our finding that deviation in topic presence and position-taking on social policies by local party branches is minimal. First, despite our focus on the first stage in local policy development following the decentralisation (party manifestos, instead of coalition agreements or policy making), it could be that our analyses came too soon to detect contingency pressures. This would imply that local variation in social policy positions has yet to materialise. Second, it might be that it already materialised, but remained hidden from view because of our focus on manifestos. On the one hand, the party manifesto "should be the first source of policies for a newly elected government" (Ashworth 2000, p. 13, cf. Hofferbert andBudge 1992). On the other hand, we cannot rule out the possibility that our central finding reflects discrepancy with actual positions pursued during policy discussions in local government. Future research could shed light on the empirical merits of these two interpretations of our unexpected results.
Furthermore, some limitations of this study need to be considered. First, we solely focused on the Netherlands. Consequently, our conclusions might be idiosyncratic. We do not know how far our findings travel beyond the Dutch case; perhaps policy decentralisation in other countries actually have led to substantial policy position variation across municipalities, as contingency theory predicts. Second, the smallest type of municipalities is not included in our dataset, while including these municipalities might alter the results. Also, we recognise that local circumstances other than population size, social benefit dependency, and the salience of local independent challengers might be relevant for determining policy position variation across municipalities. Moreover, because a relevant dataset does not exist, we had to compose one ourselves by strategically selecting a sub-set of Dutch municipalities. It is, of course, possible that another sub-set would lead to different results. Finally, future research may complement the study of manifestos with in-depth interviews with local politicians in order to research how local manifestos are drawn up. This could reveal whether, and how, institutional isomorphism plays a role.
Finally, to the best of our knowledge, we are the first researchers to provide an extensive empirical analysis on the relationship between national mother parties and local party branches in the process of social policy decentralisation. Our analysis of the Dutch case indicates that, four years after decentralisation, local party branches still largely operate as puppets instead of principals in formulating social policies. We caveat this with an important footnote that inevitable limitations of data and time warrant care in drawing conclusions that are too far-reaching in this matter.