Lobbying during government formations: do policy advocates attain their preferences in coalition agreements?

Abstract

Elections produce shifts in power and policy that give lobbyists incentives to influence the policy plans of new governments, but very little is known about such lobbying. This study directly observes lobbying during government coalition negotiations and its consequences for coalition agreements by studying the letters that policy advocates send to the (in)formateur steering coalition formation negotiations. While political parties are crucial for the preference attainment of lobbyists, the analysis shows that advocates that are traditional allies of a negotiating party tend to benefit more from making a request in line with the preferences of that political party than other advocates. This seems to be especially the case when advocates represent a constituency that is important to a party’s electoral strategy, suggesting that the policy implications of ties between parties and organized interests are determined by more than the presence of historical ties between parties and groups alone.

Keywords:

• Lobbying
• interest groups
• coalition agreement
• political parties
• elections

Democratic elections can produce changes in the party composition of government and future policy (Mansbridge 2003). On average 60% of the pledges that political parties make in their manifestos are implemented by new governments (Thomson et al. 2017). In many Western democracies the political parties that form coalition governments outline their policy plans in coalition agreements (Müller and Strøm 2008), which strongly constrain future legislative action by the government (Moury 2011; Schermann and Ennser-Jedenastik 2014; Zubek and Klüver 2015). This influence on future policy making makes coalition agreements attractive documents for policy advocates to influence. However, there are no existing studies of the impact of lobbying on coalition agreements. The goal of this article is therefore to study the conditions under which policy advocates can attain their preferences when trying to influence coalition agreements.

The existing literature on coalition agreements suggests that the direct influence of lobbying on coalition agreements is limited. A main purpose of such agreements is to reduce uncertainty about future actions by the other parties in the coalition (Klüver and Bäck 2019; Schermann and Ennser-Jedenastik 2014; Müller and Strom 1999; Timmermans 2003). Given this primary focus and the (time) pressure on negotiating politicians, it seems unlikely that organized interests can exert much influence over the negotiations.

This article argues, however, that there are two ways by which advocates can attain their preferences. Firstly, because policy advocates whose policy requests are in line with the policy position of a party entering the government coalition are more likely to attain their preferences. Secondly, politicians negotiating on the agreement also pursue another goal in addition to reducing uncertainty about the future actions of their coalition partners: maximizing their party’s ability to implement its preferred policies in the new coalition (Eichorst 2014). When reaching compromises on specific policy pledges made during the campaign, politicians face uncertainty about which of their policy plans would be (un)popular with their voters and supporters. The second expectation in this article is therefore that political parties will rely on policy advocates with whom they share historical ties to provide them with such information. (e.g. Allern et al. 2007; De Bruycker 2016; Öberg et al. 2011). The signals of these traditional allies of political parties can therefore help them increase their likelihood of preference attainment in coalition agreements by emphasizing the popularity of specific campaign promises.

The hypotheses are tested using a dataset that covers lobbying after the 2017 Dutch general election. During the negotiations about a new government coalition, 775 advocates sent letters containing specific policy requests to the (in)formateur chairing the negotiations. They offer a direct observation of the policy requests of most active interest groups and other lobbying organizations. Hand-coding of the letters using methods developed by studies on pledge-fulfillment by political parties (Thomson et al. 2017) identifies 1,200 unique policy requests that are analyzed in the article. Coding whether these policy requests were fulfilled in the coalition agreement helps identify whether advocates attained their preferences (Dür 2008). In addition, the policy requests of advocates are compared to the election manifestos of the negotiating parties to determine whether the party had a policy position on the request.

The findings reinforce the view that political parties are crucial for the preference attainment of policy advocates in coalition agreements. The multilevel regression models support the expectation that advocates whose policy requests are in line with policy positions held by negotiating parties are more likely to attain their preferences. Results are more mixed when assessing the impact of historical ties between groups and parties. Some policy advocates with ties to a political party are more likely to benefit from a policy request in line with a policy position from an allied party than other advocates, specifically business advocates who benefit more from overlap with the main liberal right-wing party than other advocates. However, the analyses indicate that these ties do not always increase the preference attainment of advocates, and a discussion of the findings suggests that the usefulness of the ties may depend on the electoral strategies of political parties, rather than historical ties alone. The article therefore presents a mixed picture of lobbying influence on coalition agreement negotiations: while there are some indications of such influence, advocacy influence remains constrained, making it hard to establish whether lobbying after elections is more or less effective than at other stages of the policy cycle.

These findings therefore align with and contribute to studies of the ties between groups and parties (Allern and Bale 2012; Otjes and Rasmussen 2017; Thomas 2001), which suggest that historical alliances between groups and parties have weakened in recent decades (Christiansen 2012) and have been replaced by more ad-hoc cooperation (Rasmussen and Lindeboom 2013). In addition, the study highlights an overlooked channel that interest groups use to try to influence policy making in Western European political systems: lobbying during coalition agreement negotiations. Finally, the results are important for studies of coalition agreements (e.g. Bäck et al. 2017; Schermann and Ennser-Jedenastik 2014; Müller and Strøm 2008), as they highlight the limited but sometimes important role played by non-party actors in coalition negotiations in multiparty democracies.

Coalition agreements and their appeal to lobbyists

In Western-European democracies, political parties that form government coalitions write coalition agreements in up to 80% of all formations (Eichorst 2014), typically including election pledges which the new government plans to implement (Schermann and Ennser-Jedenastik 2014; Timmermans 2003). Crafting these agreements is attractive to negotiating parties for at least three reasons. Firstly, they help manage the diverging policy preferences of coalition partners by outlining policy plans for the future government. Secondly, they reduce uncertainty about and opportunism in the actions of future government partners and their freedom to shift policy into their preferred direction. Finally, they help parties explain the trade-offs and choices made to parties’ audiences (their members or voters). Since not all coalitions have to solve these issues to an equal extent (their preferences may diverge more or less, for example), coalition agreements vary in the extent to which they are formalized, and range from very short documents containing few policy-specific details to long formalized agreements that outline comprehensive policy plans. Some contain not just policy plans, but also outline the ‘rules of the game’ within the coalition (Müller and Strøm 2008).

In spite of this variation, the majority of European coalition agreements contain rather comprehensive policy plans: they are crafted after elections and most of the content is indeed policy related (Müller and Strøm 2008: 174–9). While it may depend on their degree of formality and policy content, coalition agreements affect the legislative activity of the new government. In a study of Belgium, Italy and the Netherlands, Moury (2011) shows that 30% of all important cabinet decisions relate directly to the coalition agreement, with up to 50% of such decisions constrained by the agreements in some way. Polish cabinets also implement on average 60% of the policy plans that they outline in coalition agreements (Zubek and Klüver 2015), and election pledges by Austrian and Dutch political parties that are included in the coalition agreement are more likely to be turned into policy than those that are not (Schermann and Ennser-Jedenastik 2014; Thomson 2001). Bäck et al. (2017) also show the limiting effect of coalition agreements on government spending across Western Europe. There is thus strong evidence that both elections and coalition agreements play an important role in determining future policy change.

Their influence on (future) policy making makes coalition agreements interesting targets for lobbyists. However, existing research does not directly study advocacy influence on coalition agreements. This is likely to be due to the fact that these are, by their nature, inter-party negotiations, making political parties both the most powerful actors and the logical object of study. However, this does not preclude other actors like policy advocates from also exerting some influence over the agreements. In addition, if we accept the normative standard that politicians ought to implement the policies on which they were elected (Mansbridge 2003), it makes sense first to evaluate the extent to which they do so (e.g. Thomson et al. 2017). However, if policy advocates affect whether and which election pledges end up in policy and lead parties to deviate from their pre-election promises, this may have important democratic implications.

What is more, there is empirical evidence that interest groups in Norway do indeed attempt to influence coalition negotiations (Allern and Saglie 2008). Two interviews conducted for this study also suggest that both the run-up to general elections and the coalition formation period are very important to Dutch interest groups¹. Moreover, while not directly related to coalition agreement negotiations, Brown (2012) shows that groups spend more on lobbying in the transition period between the election and inauguration of an American president than in the period before or after it. Although he does not systematically study preference attainment, he offers numerous examples of policy plans influenced by policy advocates – further emphasizing the importance to policy advocates of the period directly after election.

Lobbying coalition agreement negotiations

In spite of these incentives for interest groups to try to affect the negotiations of coalition agreements, there are reasons to expect that actually exerting influence over the negotiations is difficult. For one, political parties in most European countries are funded through state subsidies, and private contributions make up on average only 10% of the revenue streams of European parties (van Biezen and Kopecký 2017). At least compared to the US, this should reduce direct incentives for politicians to implement requests by advocates that may have supported a party’s campaign.²

More importantly, negotiating parties have limited incentives to accommodate policy requests from policy advocates. Uncertainty about the cooperation with coalition partners is a major challenge in government coalitions (Laver and Shepsle 1990) and writing coalition agreements is a means for the negotiating parties to reduce this uncertainty (Moury 2011; Müller and Strom 1999; Timmermans 2003). Especially when policy preferences diverge, politicians use a public coalition agreement to reduce uncertainty and the possibility that coalition partners pull policy too much in their preferred direction. Negotiators therefore face the complicated task of finding ways to agree on policy and reach compromises, as well as the need to distribute cabinet portfolios (Laver and Budge 1992).

Hence, political parties in negotiations are unlikely to be very receptive to lobbying, which means one may generally expect that advocates are unlikely to attain their preferences in the coalition agreement. However, advocates may still see their preferences included in coalition agreements. Firstly, parties are especially unlikely to be receptive to policy requests about issues that were not part of their own campaign promises and not already part of the negotiations: after all, granting such a lobby request would mean bringing more issues to the negotiating table that the coalition partner may potentially disagree with and which might further complicate the substantive compromises that have to be reached. One may therefore expect that advocates are more likely to see their policy requests fulfilled when at least one negotiating party holds a policy position in line with their request.

However, existing studies show that political parties put more emphasis on policy issue areas (like the environment or migration) where they disagree with their coalition partners, to prevent them from moving policy in an undesirable direction (e.g. Klüver and Bäck 2019). This might lead to the expectation that advocates’ preferences are especially likely to be reflected in the coalition agreement if one party shares a policy position with an advocate, and another opposes it (as the issue area is more likely to feature in the agreement). However, most lobbying occurs on much more specific policy proposals than the policy areas studied in the literature. Even when negotiating parties disagree on the policy area, like environmental policy, and therefore discuss it in the agreement, they will still want to signal to voters that they intend to act on the issue and are therefore likely to include the specific policy plans they do agree on, for example: closing down a specific coal powered power plant. In short, while policy areas are more likely to feature in a coalition agreement when the negotiating parties disagree, this is not necessarily the case for the more specific policy preferences that policy advocates lobby for.³ Therefore, the following expectation can be formulated:

H1: Policy requests by policy advocates are more likely to be fulfilled if the proposed policy position was part of a coalition party’s manifesto.

In addition, coalition agreements serve another purpose for politicians, as they can provide advocates with leverage to successfully lobby the negotiations: the function of ‘advertising’ is how entering the government coalition allows the party to implement its election promises, to both their voters and party members (Eichorst 2014; Müller and Strøm 2008; Timmermans 2003). Since parties that enter government often lose seats at the next election (Müller and Louwerse 2018), it is important for them to show which election pledges they are able to implement by entering the coalition (Eichorst 2014). Politicians have to decide on which of their electoral pledges they are willing to reach compromises and will want to implement those pledges that are popular with the general public and/or their voters. Coalition negotiations are generally closed off and take place under considerable time pressure (Timmermans 2003). As a consequence, the regular ways of gauging voter preferences like media coverage, debates in parliament, and consultations are not as readily available to politicians. Policy advocates can therefore try to use their lobbying to signal which election pledges are especially (un)popular with the voters of the negotiating parties.

Advocates that have strong historical ties with political parties should be especially able to use this mechanism. Studies of the relations between interest groups and parties emphasize the importance of such ties (Allern et al. 2007; Otjes and Rasmussen 2017; Thomas 2001). They highlight the historical and institutional relations between trade unions and social democratic parties (Allern et al. 2007; Allern et al. 2019), as well as those between other types of party and groups, such as employers’ organization DA and the conservative parties (e.g. Christiansen 2012) and the environmental movement and Green parties (e.g. Blings 2020).

In each of these cases, the interest group and the party it shares ties with were and often continue to be able to offer each other resources that make a long-term exchange relationship mutually beneficial from a rational-institutionalist perspective (Allern et al. 2007; Christiansen 2012; Öberg et al. 2011). Parties can offer interest groups they share historical ties with a way to influence political decision making, even if this influence remains an untested assumption in the literature (e.g. Allern and Bale 2012; Otjes and Rasmussen 2017; Thomas 2001). Traditionally such groups had the ability to deliver voters for political parties through their members. However, some groups’ ability to do so may have declined over time (Allern et al. 2007; Christiansen 2012; Öberg et al. 2011) and contacts between groups and parties have generally become more ad-hoc (Rasmussen and Lindeboom 2013). Still, such historical ties do persist to this day (e.g. Allern et al. 2019) and while not on all points, parties’ historical allies are likely to still share a similar ideological outlook. In addition, such groups may still have members that politicians will be aiming to represent and from whom they will want to secure support for the new government’s coalition agreement. The longer-term cooperation between such traditional allies will also mean that they are more likely to have access to the negotiating parties: either through existing institutional integration between the groups and the party, or ‘simply’ because their cooperation means representatives of the groups and parties will move in similar networks. The second hypothesis is therefore:

Hypothesis 2: Policy requests are more likely to be fulfilled if they align with the positions of parties that are historically allied with the advocate than if they align with the positions of parties that are not historically aligned with the advocate.

Case selection and research design

The hypotheses are tested using a dataset covering the 2017 Dutch general elections, which was collected as part of the GovLis project.⁴ The election led to a long government formation process. After the election it quickly became clear that four parties would be required to achieve a majority coalition government (the norm in Dutch politics). From the start, the liberal right-wing party Volkspartij voor Vrijheid en Democratie (VVD: 33 out of 150 seats in parliament) that would provide the new prime minister, the more centrist liberal Democraten 66 party (D66: 19 seats) and the center-right Christian democrats of the Christen Democratisch Appèl (CDA: 19 seats) were very likely to be part of any majority coalition. After negotiations between these three parties and the Green-Left party failed, the smaller center-left Christian Union party (CU: 5 seats) joined the negotiations and finally the government coalition following a 225 day negotiation. The number of negotiators during the formation was relatively small: only two representatives from each for the parties and the (in)formateur (the person chairing the negotiations) were present on most days. Negotiations about some policy areas were prepared by specialized members of parliament, and the main negotiators were in touch with other members from their parties and invited policy advocacy organizations, advisory bodies, and ministerial departments to the negotiations on some days.

The parties outlined their detailed government agenda and policy plans in a post-electoral coalition agreement of over 35,000 words. The negotiation period and final agreement were lengthy compared to both previous Dutch formation processes and internationally (for a discussion, see: Timmermans 2003). The agreement is likely relatively formalized compared to other agreements and contains a comprehensive set of detailed policy plans (see Müller and Strøm 2008). While there is variation across coalition agreements, and the selection of a single case comes with trade-offs, the majority of coalition agreements tend to contain specific policy plans, be mainly focused on policy, and are concluded after the election – making the 2017 Dutch coalition agreement a case that occurs relatively often across Western Europe (Müller and Strøm 2008). Still, especially where coalition agreements contain less specific or formalized policy content, coalition agreements may be less attractive objects of lobbying.

However, the described theoretical mechanisms are likely to hold in other countries, at least for post-electoral coalition agreements. For one, both Dutch and non-Dutch negotiators alike will use coalition agreements to reduce uncertainty about the behaviour of their coalition partners. The constraining effects of coalition agreements on policy have also been demonstrated in other European countries (Bäck et al. 2017; Moury 2011; Schermann and Ennser-Jedenastik 2014; Thomson 2001; Zubek and Klüver 2015). Similarly, the tendency to aim to use the coalition negotiations to implement pre-election promises also applies beyond the Dutch context. In fact, Dutch politicians manage to implement their election pledges in policy at a rate that is about average in Western democracies (Thomson et al. 2017: 535).

It is therefore likely that lobbying to affect coalition agreements takes place after more elections than the one observed here. In the Netherlands, previous elections with generally shorter formation periods attracted similar or higher numbers of lobbying letters than the 2017 election.⁵ Moreover, in the only study to describe this kind of lobbying in Europe, some Norwegian interest groups also indicated that they tried to influence the coalition agreement (Allern and Saglie 2008: 94). Given that Norwegian coalition negotiations last an average of just 6.5 days (Golder 2010), it is likely that lobbying to influence coalition agreements occurs in other settings with longer formations, too. Of course, where negotiations are quick, there may be less opportunity for interest groups to exert influence. Media coverage also shows that this type of lobbying is indeed common in at least some other Western-European countries.⁶ Finally, the Dutch coalition negotiations involve a rather small number of negotiators. On the one hand this may make it harder for the average interest group to influence the negotiations when compared to countries like Germany, where many more negotiators are involved in the negotiations. On the other hand, it may mean that in the Netherlands, influence is relatively concentrated around those advocates that do have access.

Measuring the policy preferences of advocates

In order to observe the policy preferences of advocates lobbying during the formation period, the analysis relies on the letters they sent to the (in)formateur⁷ outlining policy requests for the following government period and the coalition agreement. 2017 is the first time these letters have been made publicly available.⁸ Following an observational definition of lobbyists/advocates, the study includes all actors who observably tried to exert influence by sending letters (Baroni et al. 2014). A total of 775 policy advocates sent letters containing specific policy requests. In itself, this is an indication that advocates do indeed actively lobby after elections. It also provides us with a snapshot of the policy requests of Dutch advocates during the formation period.

The mechanism of being able to send letters to an informateur is specific to the Dutch context and may make it comparatively easy to (try to) contact the negotiating parties and has increased the amount of observed lobbying. However, the expectation is not that these letters directly gave policy advocates influence over the final coalition agreement. Given their large quantity, it seems unlikely that the (in)formateur would read all the letters. Instead, they provide an excellent opportunity directly to observe the policy preferences of a large number of Dutch policy advocates and their issue priorities during the formation period. The interviews undertaken for this study also confirm the impression that even if these letters do not directly influence the negotiations they do contain the preferences of the interest groups sending them – forming the basis for lobbying in other ways. These may include inside strategies, like contacting political parties and negotiators, or civil servants involved in drafting the agreement. Lobbyists also use outside strategies (e.g. Kollman 1998) like commentary in the media or protests. For example, primary school teachers’ organizations sought the media and successfully protested for higher wages during the 2017 coalition negotiations. Unlike the instrument of sending letters to an (in)formateur, these strategies are also available in other countries, so while more advocates’ lobbying (preferences) may be observed in the case studied than in other countries, they still provide the best available observation of the policy requests of a wide range of advocates who tried to influence the negotiations.

All letters sent to the (in)formateur were hand-coded to detect the policy requests made by advocates. Drawing on methods developed by studies on the pledge-fulfillment of political parties (Thomson et al. 2017), a request is coded if it meets the following criteria. Firstly it has to be an explicit request for the future government to take action or for the coalition agreement to include something, containing a marker word such as ‘request’, ‘ask’ or ‘demand’. Secondly, it has to be possible for the request to be fulfilled theoretically in the coalition agreement and the requirements for fulfillment have to be specified in the request itself⁹ (Thomson et al. 2017). Examples of requests included would be a demand to increase subsidies for daycare services, or to close all coal-powered power plants. On the other hand, more general requests to ‘make policy greener’, or enforce ‘stricter immigration laws’ would be excluded, because it is impossible to determine from the request alone whether any part of the coalition agreement fulfills these requests. Requests to prioritize an issue in the coalition agreement are also included, but the results presented here are not conditional on their inclusion (see Online appendix 6).

Measuring preference attainment

While the goal of this study is ultimately to study the influence of policy advocates, the approach taken here stops short of claiming to observe influence directly (Dür 2008; Mahoney 2007). Instead, the study relies on an approach called ‘preference attainment’ (Dür 2008) or ‘lobbying success’ (Mahoney 2007), which compares the rates at which different groups of policy advocates got what they wanted. An advocate is thus considered as having attained their preferences (but not necessarily influential) if their policy request is fulfilled in the coalition agreement. The final measure is dichotomous and outlines whether an advocate did not (0) attain their preferences, or did, somewhat to fully (1). Online appendix 1 contains more information about the measurement of the dependent variable and inter-coder reliability.

The measure used here overcomes several drawbacks typically associated with studies of preference attainment (Klüver 2013: chapter 3). Firstly, the policy requests of advocates are observed directly, rather than retrieved through interviews or surveys. Although letters may also have been sent to signal engagement to the members of some advocacy groups, it is likely that the policy requests outlined in the letters contain the policy priorities of the advocate. This assumption was also confirmed in two interviews with major interest groups. Secondly, the fact that the letters were sent to the (in)formateur with the goal of influencing the coalition agreement means that there is a good fit between the policy requests and the measure of preference attainment. Thirdly, since the requests are formulated by the advocates themselves, the advocates’ own formulations are used to assess preference attainment, instead of often used pre-defined sets of issues that are on the legislative agenda, or formulated by researchers.

Party policy positions

In order to identify the policy positions of political parties on requests, the first step was to code whether a request was present in the election manifesto of a negotiating party (2 coders and 88 coded units with Krippendorff alpha: 0.95.) and then code whether the party position was in line with the request (Krippendorff alpha: 0.73). To test H1, that advocates are more likely to attain their preferences when their request is in line with the policy position of a negotiating party, a binary measure is used that captures whether at least one political party supported the issue prior to the election (in its manifesto).

Parties with historical ties to interest groups

In order to assess hypothesis 2 about historical ties, the article relies on ties between business advocates (employers’ organizations and firms) and two major right-wing political parties (see Christiansen (2012) for a discussion of similar ties in Denmark). The first of these parties is the liberal VVD party. The most economically right-wing of the major parties that regularly participate in Dutch government, the VVD maintains strong ties with business actors. While cooperation between the VVD and private sector advocates was never as institutionalized as between trade unions and social-democratic parties in some Nordic countries (e.g. Allern et al. 2007), the party has maintained strong ideological and interpersonal links with large businesses and employer organizations since its foundation in the 1940s. One of its founders was very active in a major employer organization and ties between employer organizations, business advocates and the party remained relatively strong throughout the decades (Lucardie 1986). What is more, such ties persist to the present day. As an example, VVD MPs are much more likely to have previous work experience in the private sector than those of other parties,¹⁰ its party leader and prime minister during the 2017 election Mark Rutte’s previous career at multinational Unilever providing an example. These strong historical and interpersonal links between the VVD and private sector actors, as well as its identity as a party for entrepreneurs (for example in its 2017 election manifesto, see also Lucardie 1986), make it likely that the party is more receptive to policy requests by business advocates than other advocates.

Secondly, the Christian democratic center-right CDA party identified itself less clearly as pro-business – at least in its 2017 election manifesto where it, for example, advocated reducing the role of market forces in health care. At the same time, it arguably shares stronger historical and organizational ties with the main employer organization VNO-NCW, than even the VVD. VNO-NCW is a merger between two employer organizations, one of which has Christian roots. Like the Christian parties that the CDA is a merger of, NCW was part of the Christian pillar of post-war Dutch society, and historical ties between the organization and the party were close (for an example illustrating this cooperation, see: Hordijk 1988). While it is likely these ties may have weakened somewhat over the decades (like in Scandinavia: see Öberg et al. 2011), there is also evidence that these ties continued at the time of the 2017 negotiations. For example, the chairman of VNO-NCW in 2017 was a member of the CDA and its previous chairman served as a senator for the party. In 2019, the party appointed a chairman who also serves as the secretary of VNO-NCW. Hence, the CDA shares strong links to business advocates that should make it easier for these advocates to contact the party’s politicians and make them relatively receptive to their requests.

Hypothesis 2 is tested by interacting a binary variable (right-wing support) indicating whether a request was supported by either the VVD or the CDA, with a binary variable that indicates whether the request is made by a business group/firm, or a non-business actor. The analyses also look at the interaction effect of both parties separately to explore differences between them and the nature of their ties to business actors. This measure is a relatively crude way to operationalize the ties between business advocates and these two political parties, as it does not focus on specific advocacy organizations. If anything, however, this should make it harder to observe an effect of these ties on preference attainment. There were no Social Democratic or Green parties in the cabinet, which means that the influence of their ties with other group types cannot be tested. There were also not enough requests by religious interest groups to reliably model possible shared constituencies between them and the Christian Union or CDA party.

Control variables

The analyses contain three control variables. Firstly, advocates may access and influence the coalition negotiations directly. The interest group literature tends to assume ‘that groups with political access are on average more likely to be influential than groups without such access’ (Binderkrantz et al. 2017: 307). The analyses control for the possibility that rather than party-group ties, it is simply advocates with access to the negotiations who attain their preferences. Access may have enabled policy advocates to influence the negotiations in two ways. Firstly, some advocates are invited to the negotiation table. To control for this, the daily calendars of the (in)formateur are analyzed to code who secured such a meeting. Secondly, civil servants from government departments also visit the negotiations and are involved in drafting some of the text of the coalition agreement. That is why – based on the coding by Berkhout and colleagues¹¹ – the twenty advocacy organizations with the most access (defined as having a meeting with a government minister, or invitations to round-table hearings in parliament) are also coded as having access. The two measures are combined in a single binary variable.

Advocates often formed coalitions when sending letters, most likely in order to signal the broad support for their requests. Previous studies show mixed or conditional effects of coalitions on preference attainment (Junk 2019; Mahoney and Baumgartner 2004). To control for the possibility that these coalitions affect the preference attainment of advocates, the analyses include a control for coalition size, which is a count of the number of advocates that sent a specific letter. The third control variable captures whether an advocate’s request was in favour of changing the status quo. Studies of the United States document a ‘status quo bias’ (Baumgartner et al. 2009), and policy advocates who defend the status quo are more likely to attain their preferences, at least at the national level in Western Europe (Rasmussen et al. 2018). Even if 94% of the requests in the data are to change the status quo, this bias may still persist in the setting studied here. Online appendix 2 provides more descriptive information about the dataset and requests to change the status quo.

Modeling strategy

The unit of analysis is a policy request (nested) in a letter by an advocate to predict whether an advocate did not (0) or somewhat to fully (1) obtain it preferences, requiring logistic regression. This means that if a letter is sent by three advocates, each request also features three times in the data. To capture variation at the level of letters and because the coalition-size variable is measured at this level, random intercepts are fitted for letters (Steenbergen and Jones 2002).

Analysis and results

In order to test hypothesis 1 (that requests by advocates are more likely to be fulfilled, i.e. attain their preferences, if the policy request was part of a coalition party’s manifesto), a first step is to consider the descriptive statistics alone. Around a third of all requests were supported by at least one political party (34%). There were only 30 out of 1,201 unique non-procedural policy requests to change the status quo (or around 2.5%), that were fulfilled when no political party actively supported the advocate’s position in their election manifesto. This offers initial support for H1: in order for advocates to get their requests for policy change included in the coalition agreement, a party having a policy position in line with their policy request is close to a necessary condition.

Turning to the multilevel logistic regression models predicting preference attainment presented in Table 1, the importance of political parties for advocates’ preference attainment is underlined further. Model 1 shows that the relationship between a request being present as a policy position in a political party’s manifesto and preference attainment is both strong and significant. The predicted probability of preference attainment when no party holds a policy position in line with the request in its manifesto is around 23% and increases to 64% when at least one political party has a position in line with the request. Online appendix 4 demonstrates that taking into account whether another party also opposed the request does not substantively alter this finding. While measures of preference attainment are often used as an indicator for influence, this may not be the case here: advocates may either simply be ‘lucky’ in the sense that a party had a policy position in line with their request, or they may have successfully influenced the party’s manifesto before the election.

Table 1. Multilevel logistic regression models predicting whether a policy advocate attained their policy preferences.

	(1)	(2)	(3)	(4)
Party support	2.57***
	(0.15)
Right-wing support		1.43***
		(0.20)
Right-wing support* Business		1.30***
		(0.32)
VVD support			0.63*
			(0.28)
VVD support* Business			2.81***
			(0.45)
CDA support				2.25***
				(0.24)
CDA support* Business				0.31
				(0.36)
Business	0.19	−0.25	−0.22	0.15
	(0.21)	(0.23)	(0.21)	(0.22)
Controls
Access	0.51	0.38	0.42	0.37
	(0.38)	(0.36)	(0.37)	(0.36)
Coalition size	0.06*	0.06+	0.05	0.05+
	(0.03)	(0.03)	(0.03)	(0.03)
Pro policy change	−3.84***	−3.62***	−3.53***	−3.41***
	(0.35)	(0.35)	(0.35)	(0.34)
Constant	1.35***	1.90***	1.97***	1.66***
	(0.36)	(0.36)	(0.37)	(0.36)
Letter random intercepts	Yes	Yes	Yes	Yes
Number of requests	2,281	2,281	2,281	2,281
Number of letters	346	346	346	346
AIC	1,954	2,123	2,176	2,122
BIC	1,994	2,169	2,222	2,168

+ p < 0.10,

*p < 0.05,

**p < 0.01,

***p < 0.001.

Secondly, models 2 through to 4 then test hypothesis 2, that advocates that make requests that are in line with the policy positions of traditionally allied parties are more likely to attain their preferences. The empirical implication of this expectation was that a request in line with the policy preferences of the center-right parties (the liberal VVD and the Christian democratic CDA) in the coalition should be more strongly related to preference attainment for business advocates than other advocates. The interaction effect in model 2 supports this expectation: the effect of a request in line with a policy position of these right-wing parties is stronger for business advocates than for other types of actors.

However, when the analysis is split by party in models 3 and 4 it becomes clear that this correlation is driven by the VVD: the interaction effect between business advocates’ requests in line with the VVD’s policy position in model 3 is positive and significant, whereas the interaction effect is much smaller and insignificant for the CDA in model 4. Figure 1 shows the increase in the predicted probability of preference attainment for different advocates when a party has a position in line with their request. The figure underlines that while the increase is stronger for business advocates that make a request in line with the policy positions of right-wing parties (grey triangles) than non-business advocates, this effect is driven by the VVD (blue squares). The VVD seems to distinguish between requests from business and non-business advocates. Importantly, the increase in predicted probability that comes from making a request in line with a VVD position is only 12 percentage points for non-business advocates and just over 50 percentage points for business advocates. On the other hand, the CDA (orange dots) does not significantly distinguish between the two sets of advocates: both business and non-business advocates see an increase in the predicted probability of attaining their preferences of around 40 percentage points when their request is in line with a CDA policy position.

Figure 1. The marginal effect of gaining the support from a party on preference attainment.

Note: The marginal effect is expressed in increases in predicted probabilities for different sets of actors, based on Table 1. Calculated for both right-wing parties (grey triangles, model 2), the VVD (blue squares, model 3) and the CDA (yellow dots, model 4) [Colour on line]. All other variables kept at their mean. 95% confidence intervals.

These results partially support hypothesis 2: whereas there is a clear effect for the VVD in support of the hypothesis, the absence of the expected interaction effect for the CDA goes against the expectation.

Additional analyses

Unpacking these results, it may be the case that rather than historic ties between a party and a set of advocates, it may be the importance of the constituency an advocate represents to the political party’s electoral strategy that matters: VVD’s 2017 election campaign was relatively pro-business and suggested, for example, tax cuts for business actors, making business advocates (representatives of) an important electoral constituency to the party. On the other hand, the CDA’s campaign was more critical of free-market forces and proposed, for example, to reduce the role of the market in the health care system, which should reduce the importance of the constituency represented by business advocates. These impressions are underlined by data from the Comparative Manifestos Project (Volkens et al. 2019): where 3.6% of the VVD’s manifesto consisted of positive mentions of the free market economy, the comparable figure was 1.3% for the CDA. In addition, the VVD’s position on the CMP’s market economy index was for example 5.4 (more pro-market), compared to 2.6 for the CDA. Similarly, the CDA made greater reference to economic planning and very positively referenced welfare state policies. These differences in campaign strategy, or at least issue emphasis and position, may help to explain why the VVD attached additional value to requests by business advocates where the CDA did not.

To further assess this, Online appendix 5 also repeats the analyses from Table 1, looking at trade unions rather than business advocates. Although declining, trade unions still represent large numbers of members (the largest trade union Federation FNV having around 1 million members in 2017). If parties were aiming to ‘please’ the largest share of the public possible, rather than specific (electorally relevant) constituencies one would expect to find a significant interaction effect. The fact that no such effects appear emphasizes that the interaction between the VVD and business is about more than just pleasing the largest number of voters possible. Hence, the importance of the constituency that a group represents may in part depend on the electoral strategy of a political party, rather than just the historical and personal links between an advocate and a party or the number of members represented by the advocate. That possible explanation would also align with findings that historical ties between groups and parties have weakened (Allern et al. 2007; Christiansen 2012) and cooperation has become more ad-hoc or pragmatic (Rasmussen and Lindeboom 2013).

Finally, turning to the control variables, only the effect of defending the status quo is strong and significant in all models (Baumgartner et al. 2009). Access has a positive but insignificant relationship with the likelihood of preference attainment. The size of a coalition has a positive effect on preference attainment, but it is not significant across different model specifications – underlining previous findings of more conditional effects of coalitions on preference attainment (Junk 2019; Mahoney and Baumgartner 2004).

Robustness tests

The online appendices outline a number of checks to ensure the robustness of the findings and control for alternative explanations. Firstly, it may be the case that the issue ownership of specific political parties (Petrocik et al. 2003) makes certain requests more attractive to a party than others. Since business advocates are likely to make requests about economic issues, the importance of economic issues to the VVD may mean that the interaction effect in model 3 is an artefact of this. Online appendix 3 therefore includes an interaction between making a request in line with the policy position of the CDA and VVD and whether the request was in the field of economics and taxation to exclude the possibility that the party’s ownership of economic issues explains the results regarding hypothesis 2. The models show that VVD support is more valuable for advocates’ requests about economy and taxation than requests in other policy areas.

However, this effect is independent of the interaction between requests by business advocates and VVD positions. This indicates that the findings in Table 1 are not just the result of issue ownership by the VVD party. However, even business requests about non-economic issues may be about specific policies that were more salient to the VVD than requests by other advocates: something that the analysis cannot fully preclude. Future research could investigate whether a party’s ideology matters in this regard. Online appendix 6 shows that results remain unchanged when analyzing only requests for policy change. Online appendix 7 reruns the analyses using robust standard errors instead of multilevel modeling to account for the clustering of observations in letters and individual advocates. This leaves the substantive results unchanged.

Conclusion

Are lobbyists able to attain their preferences in coalition agreements? In line with expectations derived from the literature on coalition agreements, this article shows that making requests in line with the policy positions of political parties is crucial for the preference attainment of policy advocates: requests are generally not fulfilled if they are not first present in the election manifestos of political parties (H1). At least to the extent that these party manifestos make explicit the mandate of political parties (Thomson et al. 2017), most policy requests are only implemented after such a mandate is obtained through elections. To the extent that political parties are expected to implement the policy platforms on which they were elected (Mansbridge 2003), this is good news in democratic terms. After all, policy advocates most likely do not hinder the implementation of election promises that parties were elected on in coalition agreements, or introduce new policies into the coalition agreement that were not previously featured in a negotiating party’s manifesto.

Secondly, the article shows that policy advocates that make policy requests that are in line with the policy positions of negotiating parties they share historical ties with are more likely to attain their preferences (H2). The analyses demonstrate that this mechanism does not necessarily apply to all sets of parties and groups that share such ties, however. Whereas there was an effect for business advocates and the VVD, business advocates did not benefit more from making requests that were in line with the policy positions of the CDA than other types of advocates – even though the CDA (traditionally) has strong organizational ties with business groups. This unexpected result might be a consequence of the different campaign strategies these parties followed, which may have meant that business advocates represented constituencies that were more salient to the VVD than the CDA. Either way, the findings indicate that the policy implications of ties and contacts between interest group and parties are nuanced, and that the impact of these ties on policy making requires further theorizing. Future studies could therefore focus on other political parties and interest groups to investigate further when relations between parties and interest groups matter for policy making. Although more difficult, they could also seek to separate ideological congruence and historical ties with a party to further study this question.

The findings demonstrate the general importance of including elections and the policy changes they help produce into the study of lobbying success outside the US context (see also: Binderkrantz 2015; Farrell and Schmitt-Beck 2008). While this study shows that there are reasons to expect some (limited) lobbying influence on coalition negotiations, future studies may also focus on whether policy advocates are more or less influential during elections and coalition formations than at other stages of the policy cycle (see also: Binderkrantz 2015).

The fact that the Dutch 2017 formation was particularly protracted may have made it relatively likely that policy advocates were able to affect the negotiations. At the same time, similar numbers of letters were sent during shorter previous negotiations in the Netherlands and interviews with two interest groups conducted for this study also suggest that lobbying before and after the election is an important strategy to these organizations.

The coalition agreement studied here is likely to be relatively formalistic and high in policy content (Müller and Strøm 2008) and this may have made it more attractive for policy advocates than other coalition agreements. Still, many coalition agreements are policy rich and coalition agreements have been shown to shape future policy making across a range of countries, making them attractive to advocates beyond the Netherlands (e.g. Moury 2011).

The finding that advocates are especially likely to attain their preferences when they make requests that are in line with the policy preferences of political parties can of course mean they were simply lucky to have the same policy preferences as an incoming government party. On the other hand, they may have already influenced the election manifesto of the party as it was written. This would provide lobbyists with another way to introduce their requests into both coalition agreements and final policy. There is also evidence from interviews with the writers of manifestos and groups that policy advocates do indeed use this strategy in Austria, Norway and Ireland (Allern and Saglie 2008; Däubler 2012; Dolezal et al. 2012), which requires further research. Of course, a negotiating party may have prioritized a given election promise over another as a result of election lobbying.

Finally, the result that advocates may be able to affect the calculations made by politicians negotiating the coalition agreement is important for studies on how coalition agreements are produced. It highlights the role of non-party actors in a literature that is understandably predominantly focused on the role of the negotiating parties, but which does not consider the role of lobbying or other actors like the administration (e.g. Klüver and Bäck 2019; Müller and Strom 1999; Timmermans 2003).

Acknowledgements

The manuscript benefitted from feedback from Anne Rasmussen, Linda Flöthe, Dimiter Toshkov, and Wiebke Junk, as well as comments on presentations at Institute of Public Administration of Leiden University, the Politicologenetmaal 2018 in Leiden and the ECPR General Conference in 2018, Hamburg, and the anonymous reviewers of West European Politics. The author is grateful to Janna Goijaerts for excellent research assistance.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This article is part of the GovLis project and was funded through NWO VIDI Grant 452-12-008 and Danish Council for Independent Research Sapere Aude Grant 0602-02642B.

Notes on contributors

Jeroen Romeijn

Jeroen Romeijn is a PhD candidate at Leiden University. His research focus is on the representative role of political parties and other possible aggregators of public preferences like interest groups. [[email protected]: [email protected]]

Notes

1 In the autumn of 2018, semi-structured interviews were held with representatives and employees of political parties involved in the drafting of the 2017 manifestos of two major political parties, as well as with employees from two major interest organizations.

2 Although even in the US the evidence of the effect of campaign donations on lobbying success is mixed (McKay 2018).

3 To test whether this is indeed the case, Online appendix 4 shows that there is indeed a positive correlation between the number of parties that have the same policy position as the advocate’s request and the likelihood of preference attainment.

4 For more: www.govlis.eu

5 See the official evaluation of the 2017 formation: https://zoek.officielebekendmakingen.nl/kst-34700-64.html

6 See for example: Finland: https://www.hbl.fi/artikel/gron-lobbyist-lamnar-regeringsforhandlingarna/ and Germany: https://www.zeit.de/wirtschaft/2009-10/lobbyisten-koalitionsverhandlungen.

7 The informateur chairs the negotiations during most of the negotiations. It is only in the final stages that the formateur (typically the leader of the party providing the prime minister) chairs the meetings.

8 Through https://www.kabinetsformatie2017.nl

9 Akin to what Thomson et al. (2017) call a ‘narrow’ definition of a pledge.

10 See: https://www.vn.nl/de-haagse-banencarrousel/

11 Compiled by scholars at the University of Amsterdam and based on a report published in the magazine Vrij Nederland: https://www.vn.nl/lobbyclubs-schaduwmacht/