Do Governments Put Their Money Where Their Mouth Is? Policy Adoption And Administrative Resource Provision in 15 OECD Countries

Abstract

Government programs often fail because administrative actors receive insufficient financial and personnel resources for their implementation. Despite the importance of resource provision for policy implementation, we know very little about when and why implementers are equipped with the resources they need. We examine the conditions under which new policies go hand in hand with resource increases for the administration. We match data on policy adoptions and budgetary changes in the area of social policy for 15 European countries over 30 years (1990 to 2020). The analysis reveals that governments tend to provide more financial resources when 1) an issue is prominently discussed among parties, and when 2) institutional fragmentation is low. Moreover, governments provide fewer additional resources for policy implementation 3) when their chances of getting re-elected are low due to intense political competition. These findings contribute to our broader understanding of democracy and public administration.

Insufficient resources are one of the key reasons why policies fail to achieve their intended effects (Lipsky 1980; Pressman and Wildavsky 1984). If administrators do not possess sufficient resources, they find it challenging to execute and enforce policies effectively. While such capacity deficits have been recognized as the “Achilles’ heel of governance” (Howlett and Ramesh 2016), we know surprisingly little about when public authorities get the resources they need and when they do not (Dasgupta and Kapur 2020). This blind spot results from the perception of administrative capacities as a “macro-level” and largely “static” phenomenon that applies to the public sector as a whole and that hardly varies over time (Addison 2009; Williams 2021). However, this simplified view cannot explain why, even in the same country or sector, some policies come with more staff and money for their implementation while others do not (Knill, Steinbach, and Zink 2024).

This article seeks to address this lacuna. We explore the conditions under which new policies go hand in hand with the provision of additional administrative resources. To do so, we match data on policy adoptions and budgetary changes in the area of social policy for 15 European countries from 1990 to 2020. Our analysis reveals important new insights into the “politics of administrative resource provision.” First, governments tend to provide more resources for implementation when the newly adopted policies address salient issues that are the object of intense inter-party debate. When issues are salient, it is important for governments to fulfil their policy promises. Second, governments tend to allocate more resources for policy implementation in contexts where institutional fragmentation is limited. Low fragmentation makes it easier for voters to trace implementation deficits back to the government and provides fewer opportunities for opposition actors to obstruct the provision of administrative resources. Third, governments tend to provide fewer additional resources when political competition is intense. In these settings, governments’ chances of getting re-elected are smaller and they hence care less about implementation deficits that usually only reveal themselves at a later point in time. In contexts of intense electoral competition, governments face strong pressures to adopt new policies to show responsiveness to public demands, while (unpopular) budgetary implications become an afterthought. The replication of our analysis for environmental policy suggests that these findings primarily apply to areas where citizens are directly affected by policies and thus face immediate disadvantages if these policies are not properly implemented. Future research on the “politics of administrative resource provision” thus needs to develop and test sector-specific explanations.

These insights contribute to political science in three ways. First, we provide a first theorization of the link between policy change and administrative resource provision. While previous research has acknowledged the importance of administrative resources for policy implementation, it has not yet addressed the question of when and why governments back their policies with additional funding. Second, we introduce a new conceptual approach for capturing the extent to which new policies come with extra resources. Until now, and to the best of our knowledge, there are no conceptual attempts to assess whether the decisions for policy production and resource expansion are decoupled or aligned. Third, we offer an empirical analysis of the link between social policy reforms and budgetary changes for 15 European countries over 30 years (1990 to 2020). Overall, our analysis stresses the need to conduct more research on a consequential but hitherto neglected aspect of the opinion-policy link in democracies. What we term “empty responsiveness” is a situation where politicians adopt public policies in response to political demands, but these policies fail to make the promised impact due to lacking resources for policy implementation.

The structure of this article is as follows: The first section begins with a short overview of the literature on policy implementation and administrative resources. The second section then theorizes variations in resource provision for policy implementation across different contextual conditions. The third section introduces our empirical case and discusses the measurement of our dependent and independent variables. The fourth section presents the analysis and discusses the generalizability of our findings by extending our analysis to the case of environmental policy. The fifth section concludes.

Perspectives on Policy Implementation and Administrative Resources

Administrative capacities, i.e., the amount of available resources for a given set of tasks, importantly influence the implementation effectiveness of public policies (Hill and Hupe 2021; Meier and O’Toole 2003). Only administrations that possess sufficient staff, expertise, and financial resources can transfer policies from “paper” into practice (Limberg et al. 2021). Administrations “handicapped by inadequate financial resources will be an unlikely vehicle for effective policy implementation” (Ringquist 1993, 1026). This is true regardless of the type of policy that needs to be implemented (for a discussion, see Wegrich 2021). If implementing authorities must enforce regulatory measures prescribing a certain behavior, compliance with these rules needs to be supervised, controlled, and, if necessary, coerced. In environmental policy, for instance, implementing authorities need inspectors to carry out on-site visits to ensure that the required emission-abatement technologies are installed, and that emission limits are not exceeded (Sevä and Jagers 2013; Steinebach 2022). Likewise, if implementing authorities need to deliver social programs and other public services, they require sufficient staff and equipment to handle and process applications and to decide on citizens’ eligibility. The importance of administrative capacities for social policy delivery became particularly evident during the COVID-19 crisis, when a higher demand for welfare support and assistance coincided with a shortage of administrative personnel; a situation that resulted in long waiting periods for “immediate” help programs with severe implications for those in need (Nyashanu, Pfende, and Ekoenyong 2020). The literature on street-level bureaucracy suggests that administrators develop “coping practices” that allow them to handle their lack of resources and the resulting stress levels (Lipsky 1980). Implementers may start to prioritize certain tasks, for example by focusing only on cases and aspects that are easy to handle (Winter and Nielsen 2008; Kaplaner and Steinebach 2023). Overall, however, and irrespective of how administrators try to deal with an “overload,” insufficient administrative resources eventually have clear and negative implications for the success of policy implementation (Tummers et al. 2015; Fernández‐i‐Marín et al. 2024b).

These considerations suggest that the adoption of new policies and the associated rise of implementation workloads should ideally result in the expansion of administrative resources so that the newly adopted policies can work effectively. This is all the more so because recent research suggests that administrations in OECD democracies usually work at their respective capacity limits, i.e., they do not possess unused reserve capacities that can be employed to cover increasing implementation workloads (Fernández‐i‐Marín, et al. 2024a). Yet despite the importance of administrative capacities for policy implementation, we know very little about variations in administrative resource provision, i.e., why implementing authorities sometimes get what they need to make newly adopted policies work and why they sometimes do not.

This deficit is due to the dominant “macro-level” perspective on administrative capacities and resources in the existing literature. In most implementation studies, administrative capacities are not systematically measured or compared. Available administrative resources are simply deemed to be (in)sufficient depending on the implementation outcome (see, e.g., Domorenok, Graziano, and Polverari 2021; Terracciano and Graziano 2016). If systematically measured and compared, capacities are typically considered as a phenomenon that applies to a country’s public sector as a whole and that hardly varies over time. This perspective is particularly pronounced in studies on policy implementation in the EU context (Hille and Knill 2006; Steunenberg and Toshkov 2009). These studies expect that implementing authorities are sufficiently equipped to transpose or apply EU policies domestically depending on country-wide and broad-based indicators such as GDP per capita or the governance indicators by the World Bank (ibid.). This simplified and static view, however, cannot explain why some policies fail due to insufficient resources in the same country (or sector) while others do not.

A notable exception in this context is the work by Dasgupta and Kapur (2020), who show that administrative resource provision in India systematically varies across administrative entities—with less money and personnel being provided to entities where the political responsibility for implementation is overall less clear. The following sections thus seek to advance our knowledge of when implementing authorities get what they need to make policies work and when governments fail to adequately invest in their administration.

Theory

In a perfect world, newly adopted policies should go hand in hand with the provision of additional administrative resources needed for their effective implementation. In reality, however, this is not always the case. In the aggregate, this reality implies a growing mismatch between the policies in place and the administrative capacities available for policy implementation (Adam et al. 2019). There are several reasons why governments may not automatically equip the policies they adopt with the resources needed for their implementation. One reason, for instance, may be that the central government may simply not know what is required to make policies work locally due to “information asymmetries” and lacking channels of vertical feedback and exchange (Matland 1995; Fernández-i-Marín et al. 2024d; Knill, Steienbacher, and Steinebach 2021). More importantly, however, there are different “incentive structures” that guide the production of policies on the one hand and the provision of administrative resources on the other.

Policies are governments’ primary problem-solving tool because they allow them to deal “with issues and problems as they arise” (Orren and Skowronek 2017, 3). Policies can be relatively easily adapted to changing problem constellations. Their flexibility and forward-looking nature are also why political actors have historically developed strong incentives to use them to please and show responsiveness to their supporters. At the same time, governments have little to gain from investing in the expansion of the administrative apparatus. Modern governments face fundamental ideological and fiscal barriers to constantly expanding the public sector. In times of fiscal austerity (e.g., Blyth 2013), governments are under strong political pressure to do “more with less.” Moreover, global financial markets restrict governments’ ability to extract resources from citizens and businesses (Streeck and Schäfer 2013). And while citizens want governments to protect them against an increasing range of threats, they are often unwilling to pay additional taxes for this purpose (Ansell 2019).

Given these considerations, our “baseline” assumption is that there is usually not a one-on-one relationship between policy adoptions and corresponding resource provisions. The arguments developed in the remainder of this section thus theorize the variation in resource provision, i.e., why and when new policies are “coupled” with resource expansions and when they are not. In general, we expect the link between policies and resource provision to depend on whether governments can simply adopt new policies to show responsiveness to public demands or whether they also have to make sure that the adopted policies deliver on their promises and produce “tangible” effects. In contexts where effective policy implementation is an afterthought for politicians, new policies should be frequently decoupled from resource expansions. In contexts where deficient policy implementation is costly for politicians, new policies should more often go hand-in-hand with resource expansions. Based on these considerations, we consider the influence of three variables: 1) issue salience, 2) the intensity of electoral competition, and 3) the level of institutional fragmentation.

Issue Salience

“Issue salience” refers to the degree to which a particular issue is perceived as important or relevant by individuals, groups, or the public as a whole. It reflects the prominence or visibility of an issue in the public discourse or opinion. There is a general consensus in the literature that issue salience matters for public policy (Shapiro 2011; Wlezien and Soroka 2016; Rasmussen, Reher, and Toshkov 2018). Existing research has shown that governments are more likely to adopt policies in response to issues that receive high levels of public or political attention (Burstein 2003; Klüver and Spoon 2016). Yet issue salience does not only play a role in the production of policies but also in their later implementation. Spendzharova and Versluis (2013), for instance, find that governments implement policies more quickly if a given policy issue is salient. With higher issue salience, citizens, media, or political opponents are much more likely to pay attention to whether a given policy exists “on paper only” or if it also delivers the expected results in practice (Lieberherr and Thomann 2020). Therefore, when politicians consider an issue to be publicly salient, they are more likely to care about whether their policies actually deliver on their promises. A crucial aspect that determines whether policies achieve their intended effects is the provision of the resources needed for their proper implementation. We can hence expect that the greater the issue salience of a given issue, the more resources administrations get for implementation.

H1: The greater the issue salience, the higher the resource provision for policy implementation.

Intensity of Electoral Competition

Electoral competitiveness is a key explanatory construct in research on political incentives and behavior. Politicians elected in competitive settings are said to be more responsive to their median constituents (see, e.g., Ansolabehere, Snyder, and Stewart 2001) while voters, for their part, turn out to vote in greater numbers (Selb 2009). However, electoral competition is also the primary reason for why politicians are considered to be short-term thinkers who care much more about their actions’ immediate effects rather than about their long-term consequences (e.g., Jacobs 2011). We expect that electoral competition affects the link between policies and resource provision for two reasons. First, when electoral competition is high, i.e., when the incumbent government faces a high risk of losing its majority in the next elections, governments have a greater incentive to produce policies that demonstrate their responsiveness to the people’s demands (Dubois 2016; Schulze 2021). At the same time, they can be expected to be eager to keep associated costs and budgetary implications low or present them as unproblematic, as new policies may require unpopular tax increases or higher debt levels. Second, in a situation of high electoral competition, governments are less concerned about their policies’ (long-term) consequences given that their chances of re-election are lower. As Gratton et al. (2021, 2965) highlight, the “political horizon and bureaucratic efficiency [i.e., the time it takes until a policy reform becomes “visible”] jointly determine a politician’s incentive to propose low-quality reforms.” Hence, when the political horizon is short (due to a high likelihood of losing office in contexts of intense electoral competition) and it takes some time until a policy is implemented and produces “tangible” effects, incentives for providing additional resources can be expected to be rather weak. For these reasons, we can expect that electoral competition increases the chance of “empty rule growth” (Adam, Knill, and Fernández-i-Marín 2017, 262), i.e., a situation where policies are produced but not (or insufficiently) backed with the resources needed for their implementation. ¹

H2: The greater the electoral competition, the lower the resource provision for policy implementation.

Institutional Fragmentation

We expect that the level of administrative resource provision also depends on the institutional setup in which governments operate. There are two primary reasons for this. First, the degree of institutional fragmentation influences the likelihood of the government being held accountable for ill-designed policies and shortcomings in implementation (Hinterleitner, Honegger, and Sager 2022; Bach and Wegrich 2019). Blame attribution becomes more difficult as the level of institutional fragmentation becomes greater, i.e., the more different institutions are involved in the policy-making process (Jensen and Mortensen 2014). Governments face a lower risk of being blamed for implementation deficits when there are many veto points, implying that responsibility for a policy (and its effectiveness) is shared by and dispersed across many actors. Second, a more fragmented institutional setup typically implies that more political actors need to be involved to reach an agreement. This can provide opponents of certain policies with additional avenues for obstructing undesired policies, either by vetoing required budgetary adjustments or by simply delaying the provision of administrative resources (Tsebelis and Chang 2004; Saeki 2009). The recurring negotiations over the U.S. debt ceiling, which necessitate the approval of multiple legislative actors, are a prime example of how the need for broad consent can be exploited to curtail financial backing for policies deemed undesirable. Such maneuvers not only impact the policy directly but can also strain administrative capacities, thereby affecting the effective implementation of the initiatives in question.

H3: The greater the level of institutional fragmentation, the lower the resource provision for policy implementation.

Research Design

We examine the link between policy adoption and resource provision for a sample of 15 European countries in the area of social policy. The countries analyzed are Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, the Netherlands, Norway, Portugal, Spain, Sweden, and Switzerland. The countries under scrutiny exhibit substantial variation across theoretically relevant dimensions such as levels of institutional fragmentation (veto players), the structure of electoral competition, and their financial capacities. For a detailed overview, please refer to table 3 in the online appendix. While our sample primarily consists of advanced democracies in Western Europe, we are confident that our findings can also be extrapolated to a broader population, including advanced democracies in North and South America, Australia, and Eastern Europe. ²

We focus on social policy because this is an important case in implementation research. For example, the two seminal contributions by Lipsky (1980) and by Pressman and Wildavsky (1984) both focused on welfare state policy and social service delivery. Moreover, a simple “Web of Science” search reveals that over the past decade, about 25 political science articles per year have dealt with the issue of policy implementation in the context of social policy. ³ In addition, social policies (e.g., regarding unemployment, pensions, or child benefits) are typically executed by central (federal) agencies and not, as in other areas, by subordinate entities located at lower levels of government. This facilitates the comparative assessment of whether new policy changes imply budget expansions for the administration (see the next two sections).

To assess whether new social policies come with additional resources, we need to combine two measures: One that captures whether the government adopts new policies; and one that assesses the level of resource provision. Moreover, we need to specify how we measure the explanatory variables that we identified in the theoretical section.

Measuring Policy Change

There are multiple approaches for assessing public policies and their changes (Hall 1993; Schaffrin, Sewerin, and Seubert 2014). The literature typically distinguishes between policy changes that affect the number of policy targets (who/what is addressed?), the policy instruments applied (how is a target addressed?), or the settings of the respective instruments (how broad/generous is a specific welfare scheme?) (Knill, Schulze, and Tosun et al. 2012). For the purposes of this article, we focus on policy changes that affect the number of policy targets addressed and the policy instruments applied. We do so because we can reasonably assume that these types of policy adoptions actually come with an additional case—and workload for the administration that (ideally) requires some level of “compensation,” i.e., more money or staff.

Based on this measurement, we also take account of the fact that there may be policy changes that come with a reduction in the overall number of policy targets or instruments. Although such instances of policy dismantling have been found to be extremely rare (Bauer and Knill 2014), our approach is sensitive to such developments because we consider yearly net changes (new adoptions minus potential dismantling) in a country’s social policy portfolio.

The data collection was carried out in the context of the ACCUPOL project. ⁴ We identified changes in policy targets and policy instruments by scrutinizing all relevant national legislation adopted throughout the observation period in the countries under study. Additional checks on data reliability were carried out using legal commentaries and secondary literature. A detailed coding manual helped to systematically extract relevant information (policy targets and instruments) from the legal documents. Overall, we identified 25 policy targets that spread across the three social policy subfields of unemployment, retirement, and children. The policy targets identified include, amongst others, regular unemployment, temporary unemployment due to bad weather conditions, retirement for individuals or married couples, child benefits (family assistance), and payments for childbearing. Furthermore, targets include birth, children, and juveniles. Moreover, we identified seven different policy instruments. These are universal benefits/allowances, means-tested benefits, contributions/fees, tax exemptions/subsidies, bonuses/grants, retention periods, and one residual category (“others”). In sum, combining policy targets and instruments leads to a total of 175 policies (25 targets*7 instruments) that can be adopted (refer to section A3 and A4 in the online appendix for an overview).

Measuring Changes in Administrative Resource Provision

Public administrations need different things to make policies work in practice (for a broader discussion on administrative capacities, see Lodge and Wegrich 2014; Moynihan 2022). Financial resources are especially crucial in this regard as they allow administrations to employ well-trained staff and procure adequate equipment. This claim is strongly supported by informal background interviews with implementers in five of the examined countries carried out in the context of the ACCUPOL project. Interviewees consistently emphasized the pivotal role of financial resources in order to ensure effective implementation. For example, implementers indicated that “achieving objectives always depends on the resources an office gets” and that “the solution of addressing everything with more money and staff definitely helps.”

We thus focus on changes in budget allocations as an indication of resource provision for policy implementation. For the budget data, we use the EUROSTAT database. This database allows us to distinguish between different sectors of government (local, state, and central) and budgetary positions (social protection, public safety, environmental protection, housing, health, etc.). Our final measure is thus social budget growth (numerator) per additional policy (denominator), with higher values implying a greater provision of financial resources.

Changes in social budgets are indicated as percentage changes. Changes in the number of policies are captured in the following manner: By combining our policy targets (25) and instruments (7), we can form a two-dimensional policy space. This space is considered zero if no policies (target-instrument combinations) are addressed, and 100% covered if the entire policy space is occupied. Changes in this portfolio space can be quantified as percentage point variations. For instance, if the portfolio coverage shifts from 5% to 8%, we would record a 3-percentage point change. ⁵

Working with fractions inherently presents the challenge of having zero in the denominator, which is why we exponentiate it. By exponentiating the denominator, we also introduce a “proportionality factor” into our analysis. Specifically, changes in policies are given slightly more weight compared to changes in social spending. There is a valid reason to assume a “non-linear” effect of additional policies. Minor policy adjustments can often be managed with existing resources. However, more substantial policy changes—measured by adding multiple policies to portfolios—are more likely to call for increased government compensation.

In the online appendix, we provide several additional analyses showing that this transformation does not affect our final results (see figure 17 in the online appendix). In 76% of our observed cases, there are no year-to-year policy portfolio changes, leaving room for policy change in approximately 24% of the cases. Among these instances of policy change, we find a predominant trend toward expansion, with 79% reflecting increases in the policy portfolio. The remaining 21% correspond to policy cutbacks. The scale of policy change exhibits considerable variation, with policy changes expanding the policy portfolio size up to 4.08 percent points. In contrast, instances of policy dismantling are marked with a more modest scale, with a decrease of 1.02 percent points maximum. This aligns well with the scholarly consensus that pursuing policy dismantling and retrenchment is politically challenging (Pierson 1994).

Figure 1 presents our dependent variable for different expenditure and social policy change constellations over time. The line in the four boxes consistently represents the average annual expenditure changes in our sample. Each of the four boxes, by contrast, depicts different policy change scenarios, ranging from 0 (no change) to 3 (strong policy expansion). This visualization illustrates that the measure of our dependent variable varies based on the scale of policy changes and the level of expenditure adjustments. Specifically, smaller compensation values follow smaller expenditure changes or more drastic policy changes.

Figure 1 The dependent variable for different expenditure and policy changes

Our analysis effectively captures variations in expenditure, portfolio changes, and their respective ratios within the spectrum of observed values. However, it must be noted that our approach has its constraints and might not accurately reflect outcomes when there are drastic changes in the numerator (expenditure) or the denominator (policy changes). Akin to a clinical thermometer, which is precise within the defined bounds of human body temperature but cannot gauge more pronounced temperature swings in the external environment, our approach works and is robust for the range of values we observe, but might require adaptations when applied to other empirical settings.

Our data on social policy change provides us with the date of adoption, pinpointing the exact moment when politicians finalize their decisions on a particular issue. This detail is pivotal for our research, as our objective is to understand the conditions prompting these political determinations. Yet it is important to recognize that, in many cases, after a policy’s adoption, there is often a temporal “gap” before policies become operative and do begin to influence the budgetary allocation. Therefore, our budget changes are “lagged” by one year when calculating the ratio between budget growth and policy expansion.

A challenge for our analysis is that the budgetary implications of new policies not only include the administrative resources required for their implementation (i.e., an agency’s “bureau budget”) but also the “direct” policy costs, such as the amount of child benefits paid out to families (i.e., an agency’s “program budget,” see Dunleavy 1985). ⁶ We thus separately control for the fact that some social policy adoptions have bigger resource implications than others as they imply bigger changes in welfare state generosity (e.g., more/less money paid out to the unemployed or pensioners). Welfare state research usually works with replacement rates to capture the level of social policy generosity (Kvist, Straubinger, and Freundt 2013). Replacement rates describe the proportion of income from work that is replaced by welfare benefits. By considering changes in replacement rates and in the number of beneficiaries, we can identify the budgetary changes caused by changes in direct policy costs (i.e., changes in the program budget) and “distil” the financial resources added for implementation (i.e., changes in the agency’s bureau budget). According to this logic, the amount of direct policy costs affecting the public budget is determined by 1) the exact level of welfare benefits and 2) the number of people allowed to claim them. This approach allows us to separately analyze whether newly adopted social policies come with changes in the program budget or in the bureau budget. To systematically account for both aspects in the statistical analysis, we control for changes in pension and unemployment replacement rates and multiply them by the number of retired people (above the age of 65) and unemployed people, respectively. The data required to compute both variables can be taken from Scruggs (2022) and the World Bank (2022) and is presented in figure 1 of the online appendix.

Overall, our measurement approach reveals that the provision of additional administrative resources varies across the different policy adoptions under scrutiny. While policy adoptions happened in 25% of our observations (country-years), about half the policy adoptions involved above-average budget changes (45%) while the other half did not (55%). While our approximation of changes in countries’ agency budgets of course does not capture all the resources required to effectively implement newly adopted policies, we demonstrate in the online appendix that our measurement is positively correlated with more broad-based administrative capacity indices (Fernández-i-Marín, et al. 2024a; refer to table 6 in the online appendix). Moreover, our background interviews further provide support for the developed measurement approach. The interviews not only indicate that implementers experience great variation in administrative resource provision for newly adopted policies. ⁷ The variation in our budget data also aligns with the variation observed in the interviews. Specifically, interviewees from countries where recent policy changes were uncompensated (Ireland, Italy, Portugal) strongly complained about lacking resources and associated implementation problems, while interviewees from countries where recent policy adoptions were compensated (Denmark, Germany) did not complain in a similar fashion (see section J in the online appendix).

Capturing Issue Salience

There are different approaches to measuring issue salience (Dennison 2019). Some approaches measure issue salience as “party system salience” and look at how vibrantly issues are discussed among parties (see, e.g., Abou-Chadi, Green-Pederson, and Mortensen 2020; Williams and Hunger 2022). Green-Pedersen and Mortensen (2010) describe this as the “party-system agenda”, i.e., “a hierarchy of issues, to which the relevant actors must pay attention” (260), and this irrespective of whether they own the issue or not. The second approach focuses on “public salience.” It considers issues to be salient if a high or growing number of citizens start to worry about an issue (Rasmussen, Reher, and Toshkor 2018; Wlezien 2005). Which of these salience measures is more suitable in view of our analytical interest? As previous research has shown, there is considerable “friction” in the nexus between public opinion and policy output (Baumgartner et al. 2009). ⁸ One could thus expect that governments are more likely to respond to (that is, provide resources for) issues that are prominently on the political agenda rather than to issues that only citizens consider to be important. According to this argument, we consider party system salience the more likely measure to explain variation in administrative resource provision.

To assess party system salience, we measure the average mentions of social issues (variable “504, Welfare state”) relative to the overall length (number of quasi-sentences) in party manifestos. ⁹ This information is provided by the Manifesto Project Database (MPD) (Volkens et al. 2021). It is important to stress that because our approach is to measure common perceptions of the importance of issues across the entire party system, the parties are not weighted by size. This way, we also make sure that our measure is different from the government’s position. This approach aligns well with previous works on party system salience (Abou-Chadi, Green-Pedersen, and Mortensen 2020). We additionally measure “public salience” using Eurobarometer data. Eurobarometer surveys are conducted regularly across Europe, asking citizens to indicate their top priority issues. Here, we construct issue salience as the share of people who identify social issues as the most important issues.

Capturing the Intensity of Electoral Competition

We assess the intensity of electoral competition based on two interrelated factors. From the government’s perspective, it is important to know whether 1) voters will change their vote from one party to the other and 2) whether these vote shifts will ultimately make a difference for the electoral outcome, i.e., the legislative seat share. While the first factor can be influenced by the government’s actions, the latter is determined by a country’s electoral rules, institutions, and the geographic distribution of the electorate (Chen and Rodden 2013). Kayser and Lindstädt (2015) combine these aspects in a single index of electoral competition to estimate the “perceived loss probability” of the parties in government. This index is readily available online.

Given that Kayser and Lindstädt (2015) estimate a government’s loss probability from the perspective of the first day after an election, their measurement benefits from exogeneity from the exact policy measures taken. In other words, we can largely exclude the possibility of reverse causality, i.e., that electoral competition is high because the government has produced policies but did not provide sufficient resources for their implementation.

Capturing Institutional Fragmentation

To account for institutional fragmentation, which facilitates blame avoidance and provides additional opportunities for resource blockage, we concentrate on the number of veto points within a country. Veto points indicate how difficult it is to pass legislation and to change the status quo. Higher numbers of veto points indicate a greater tendency toward shared responsibility. In this regard, Henisz (2000) provides a measure of institutional fragmentation that considers the number of veto points in a given polity derived from the constitutional setup and the various actors controlling these veto points.

Additional Variables

In addition to our key explanatory variables and the “direct” policy costs (as opposed to the costs for administration), we also test for other influences on our outcome variable. First, we control for the general ideology of the government. Given that the provision of resources to the administration ultimately implies a bigger state apparatus and thus a greater need for creating tax revenues, one might expect differences between left-leaning and right-leaning parties (Cusack 1999). ¹⁰ To take account of the influence of organized interests, we measure corporatism levels across countries. Previous research has shown that organized interests not only try to influence policy-making; they also try to influence policy implementation (Christensen 1993). To measure corporatism levels, we rely on the dataset provided by Jahn (2016). To account for the impact of a country’s economic context and fiscal capacity, our models incorporate GDP per capita as a measure of economic wealth and the public deficit-to-GDP ratio to gauge budgetary flexibility. Moreover, we control for the size of—as opposed to the change in—a country’s social policy portfolio. It is conceivable that the administration still has some “slack” if there are fewer policies that require implementation but that the demand for more resources increases with a greater policy stock. And last, it is possible that the provision of administrative resources might not necessarily change in response to new policies, but instead due to reform activities within the administration itself. To tackle this potential issue, we incorporate controls for administrative reforms that occurred independently of policy changes, drawing on data provided by Trein and Magetti (2021). This dataset identifies the number of annual reform endeavours, like the establishment or merger of public authorities in the realm of social policy.

Empirical Analysis

In the following, we empirically assess the theoretical expectations developed earlier. First, we showcase the outcomes from our primary model, followed by elaborations on added model specifications. Subsequently, we delve into the bureaucracy’s role in administrative resource provision and assess the broader applicability of our findings by branching out to the realm of environmental policy.

General Results

We explain the year-to-year changes of the ratio between policy expansion and budget growth with the help of a linear model that controls for unequal variances (heteroscedasticity, clustered errors) by country. To account for time dynamics, we include an autoregressive component of order one (AR1). We model all these components using probabilistic programming and estimate the parameters using Bayesian inference. We have not explicitly incorporated fixed-effects into the model. Our focus is on the ratio between changes in social policies and corresponding budget allocations. Given this approach, baseline differences between countries are inherently accounted for and balanced out. The use of Bayesian inference eases the use of missing data. ¹¹ Moreover, Bayesian inference allows us to report the results in terms of probabilities and does not require the assumption that the data comes from a sample of potential other realizations of countries and years. We use standard weakly informative priors for the main parameters of interest. All variables are standardized to half a standard deviation so that the effect sizes can be directly compared (Gelman 2008). The exact model description reads as follows:

Figure 2 presents the results of our analysis. A positive coefficient indicates a higher level of resource provision, i.e., stronger budget growth per additional policy. The analysis shows that all three factors theorized—issue salience measured as party system salience, electoral competition, and institutional fragmentation—make a significant difference for the level of resource provision. First, we see that higher issue salience comes with stronger provision of additional financial resources. This provides support for our first hypothesis ( H1 ). Second, our analysis reveals that stronger political competition implies overall less resource provision. It seems that the more fierce the competition, the more governments tend to “overproduce” policies while minding less about the long-term implications of their policy decisions. This confirms our second hypothesis ( H2 ). Finally, the results also provide support for our third hypothesis ( H3 ) that greater levels of institutional fragmentation (veto points) involve a weaker “coupling” of policy and budgetary changes. This finding is very much in line with the observation made by Dasgupta and Kapur (2020, 1332) in the context of social service delivery in India, namely that the overburdening of public authorities “occurs … because of a lack of adequate electoral incentives originating from unclear political responsibility for implementation.”

Figure 2 Explaining resource provision for social policy implementation (1990 to 2020)

Note: The dependent variable is the ratio between social budget growth (numerator) and policy expansion (denominator), with higher values implying a greater provision of financial resources per additional policy. Highest posterior densities (HPD) of the main parameters of interest (β) (95% credible interval). All parameters are standardized to two standard deviations and can therefore be roughly interpreted as the effect of an increase in one interquartile range.

The results related to our first hypothesis do not persist when we assess issue salience via public salience instead of party system salience (refer to figures 5 through 8 in the online appendix). In line with our initial reasoning, this finding suggests that governments are more inclined to allocate additional administrative resources to issues that are subject to intense inter-party contention rather than to issues that primarily resonate with citizens’ concerns. In addition, it is important to note that in the model presented in figure 2, issue salience and institutional friction are significant only at the 90% confidence interval. In the sections that follow, we conduct several additional analyses to test for the robustness of our findings.

With regard to control variables, our results indicate that greater portfolio size comes with more administrative resource provision, while higher levels of debt and GDP per capita (logged) lead to less administrative resource provision overall. Having the same signs for debt and GDP per capita seems contradictory at first sight. However, it may simply indicate that after some level of economic prosperity has been reached, the expectation toward social protection grows faster than what the state can effectively afford (Karceski and Kiser 2020). In addition, our analysis suggests that administrative reforms go hand-in-hand with proportionality between budgetary and policy changes.

A challenge for our analysis is that our findings might be affected by a reversed relationship. For instance, it might be the case that—contrary to what we argued—it is not fierce political competition that leads to a reduced provision of administrative resources but rather the opposite effect, i.e., that governments fear election defeat because of lacking administrative capacities to implement the policies they promised. Although the operationalization of the electoral competition variable already tries to address this problem by looking at the government’s (future) loss probability at the beginning of the executive term, the issue of reverse causality cannot be entirely discarded. According to Leszczensky and Wolbring (2022, 837), a maximum likelihood structural equation model (ML-SEM) “offers the [best] protection against bias arising from reverse causality under a wide range of conditions.” We thus replicate our previous analysis using a cross-lagged panel model with country-fixed effects. Figure 3 shows that our main findings are robust to this modification.

Figure 3 Explaining resource provision for social policy implementation (1990 to 2020)

Note: Cross-lagged panel model with country-fixed effects. The dependent variable is the ratio between social budget growth (numerator) and policy expansion (denominator), with higher values implying a greater provision of financial resources per additional policy. Highest posterior densities (HPD) of the main parameters of interest (β) (95% credible interval). All parameters are standardized to two standard deviations and can therefore be roughly interpreted as the effect of an increase in one interquartile range.

So far, we assumed a plain one-year time difference (lag) between the formal adoption of a policy and the provision of additional resources. In reality, however, the time between new policies and their budgetary implications may vary across countries due to different underlying bureaucratic procedures that regulate budgetary decision-making across countries (Reddick 2003). To account for these unobserved cross-country differences, we calculated country-specific lags. ¹² As shown in figure 4, our main findings remain robust to this modification of the main model, with one exception: while institutional fragmentation points in the expected direction, it is no longer significant. A plausible explanation for this could be that added veto points do not merely give policy opponents the chance to sidestep resource allocation but primarily to postpone it. In other words, as soon as we account for longer time lags between policy adoption and resource provision, the “negative” effects of veto points on administrative resource provision become substantially smaller.

Figure 4 Explaining resource provision for social policy implementation (1990 to 2020)

Note: Country-specific time-lags. The dependent variable is the ratio between social budget growth (numerator) and policy expansion (denominator), with higher values implying a greater provision of financial resources per additional policy. Highest posterior densities (HPD) of the main parameters of interest (β) (95% credible interval). All parameters are standardized to two standard deviations and can therefore be roughly interpreted as the effect of an increase in one interquartile range.

Discussion 1: What about the Bureaucracy?

Up to this point, our analysis primarily focused on “top-down” political factors that influence the likelihood of adequate resource provision to implementing actors. However, governments may not only face political pressures for resource provision; also the bureaucracy may act as a “resource demander.” In general, bureaucrats can be assumed to have a strong interest in receiving the resources necessary to carry out the tasks they are charged with (Cohen 2021). They are also experts in their respective fields and usually know which resources are required to effectively implement a policy (Molenveld et al. 2020). However, the ability of implementers to articulate their resource needs and exert pressure on the government hinges on their access to policy formulators (Knill, Steinebach, and Zink 2024). We gauge this “coupling” between the policy-making ministry (department) and the implementing authorities using the managerial autonomy index provided by Jordana, Fernández-i-Marín, and Bianculli (2018, 527), which captures “the administrative separation between agencies and existing ministerial structures.” In this scenario, we expect that implementing agencies with closer ties to policy-making ministries are more likely to obtain additional resources compared to those that are only remotely connected.

Another consideration is whether the bureaucracy effectively receives the resources allocated by the policy-initiating government. After all, it is possible that subsequent governments attempt to counteract their predecessor’s policies by withholding the resources necessary for their execution. As Moynihan (2022) argues, the “deconstruction of the administrative state”—as opposed to the dismantling of public policies—has become a popular method to counteract the policy ambitions of political opponents. This is relevant for our analysis since we are typically expecting a (short) time lag between policy-making and the allocation of resources. Given this delay, government changes can occur between the stages of policy formulation and resource provision. One way to consider the potential instrumentalization of the bureaucracy in our analysis is to examine whether there are major policy divergences between consecutive governments. ¹³ To achieve this, we measure the ideological disparity between two consecutive governments on the left-right spectrum by utilizing the data from the Manifesto Project.

As shown in figure 5, our results suggest that a more pronounced administrative separation between implementing agencies and ministerial structures, leading to limited avenues for bureaucrats to directly voice their concerns, is associated with reduced administrative resource provisions. ¹⁴ Another interpretation of our findings is that the “politico-administrative separation” signifies institutional fragmentation during the implementation phase. We note that reduced compensation is linked to instances where implementation is less directly controlled by the government, allowing for blame deflection in the event of maladministration. This interpretation aligns well with the prior observation that institutional fragmentation of the politico-administrative apparatus, both at the policy-making and the implementation stage, is a crucial driver of non-compensation. In contrast, our analysis reveals no discernible effect stemming from the ideological differences between two consecutive governments.

Figure 5 Explaining resource provision for policy implementation (1990 to 2020)

Notes: Extended model. The dependent variable is the ratio between social budget growth (numerator) and policy expansion (denominator), with higher values implying a greater provision of financial resources per additional policy. Highest posterior densities (HPD) of the main parameters of interest (β) (95% credible interval). All parameters are standardized to two standard deviations and can therefore be roughly interpreted as the effect of an increase in one interquartile range.

Discussion 2: Towards A General Theory of the Politics of Administrative Resource Provision?

Until now, we tested different explanations that we expected would affect the level of administrative resource provision when governments adopt additional policies. We focused on the area of social policy given that it constitutes an important case with dozens of scholarly publications per year focusing on social policy delivery, implementation, and administration. We found that issue salience (understood as party system salience), electoral competition, and institutional fragmentation influence the extent of resource provision for newly adopted policies. The crucial question, at this point, is whether these findings can travel to other policy areas or whether the extent of resource provision in other policy areas is determined by different factors. In other words: does our explanation constitute a “general” theory of administrative resource provision or is it specific to the policy area under study?

An underlying expectation of Hypotheses 1–3 is that citizens can assess both the degree of resource provision and, crucially, the effectiveness of policy implementation. While for example citizens can be reasonably expected to assess whether they receive all the promised family benefits on time, citizens’ ability to identify implementation deficits may be much lower in policy areas where they are not the primary target group and where a higher level of technical understanding is necessary. In these contexts, the extent of administrative resource provision may be less contingent on factors such as issue salience, electoral competition intensity, or institutional fragmentation but, instead, may be primarily influenced by the presence of actors other than citizens who can oversee and assess the quality of policy implementation.

To test for this argument and thus the generalizability of our previous results, we replicate our main analysis for environmental policy. ¹⁵ Most environmental policies are directed at businesses, and whether these policies deliver on their promises (e.g., less air emissions or water pollution) is difficult for ordinary citizens to assess, especially in the short term. At the same time, environmental non-governmental organizations (ENGOs) play a major role as “watchdogs” in environmental policy implementation (Eilstrup-Sangiovanni and Sharman 2021). As highlighted by Li et al. (2021, 2), “ENGOs are the primary driving force for vigilance and advocacy of the environmental policy.”

As in the previous analysis, we identified the environmental policy targets and instruments under study (see section I in the online appendix) by scrutinizing national legal repositories and databases. The data on budgetary changes comes (again) from the EUROSTAT database. Unfortunately, and in contrast to social policy, there is no data available that provides information on the direct policy cost of the individual environmental policy measures taken (as opposed to administrative costs). We thus opt for a more “crude” approach and use a simple dummy variable to distinguish between policy instruments that, by design, should involve stronger (e.g., subsidies and taxes) and weaker (e.g., regulatory and information-based measures) budgetary implications. In addition to the previous analysis, we include a variable that captures the number of national ENGOs per capita. The data comes from the online library of the United Nations (2022) and has been collected and coded by Li et al. (2021).

Figure 6 presents the results of our analysis for the area of environmental policy. In contrast to the analysis for the area of social policy (see again figure 1), issue salience, electoral competition intensity, and institutional fragmentation do not reach the significance level. At the same time, the analysis reveals that the level of resource provision substantially increases with the strength of national ENGOs. This finding confirms the supposition that the factors explaining administrative resource provision vary by policy area and depend on who is in the position to evaluate the proper implementation of public policies.

Figure 6 Explaining resource provision for policy implementation (1990 to 2020).

Note: The dependent variable is the ratio between environmental budget growth (numerator) and policy expansion (denominator), with higher values implying a greater provision of financial resources per additional policy. Highest posterior densities (HPD) of the main parameters of interest (β) (95% credible interval). All parameters are standardized to two standard deviations and can therefore be roughly interpreted as the effect of an increase in one interquartile range.

Another crucial aspect to consider when developing a general theory on the politics of administrative resource provision are the potential interactions between the independent variables in our analysis. For example, issue salience and electoral competition might reinforce each other. In figure 18 in the online appendix, we provide additional analyses that explore the interplay between our primary independent variables. We discover no significant interactions among them. At the same time, all our main effects remain significant. This further bolsters the robustness of the main findings.

Conclusion

This article started from the generally acknowledged assumption that government programs often fail on the ground because the public administration receives insufficient financial and personnel resources for policy implementation. We thus examined the conditions under which newly adopted policies added to the existing policy stock coincide with resource increases for the administration in charge of policy implementation by matching data on policy adoptions and budgetary changes for 15 European countries in the area of social policy from 1990 to 2020. The analysis revealed that governments tend to provide more financial resources 1) when the newly adopted policies address salient issues that are the object of intense inter-party debate and 2) when the level of institutional fragmentation is low, limiting possibilities for obstructing resource provision and deflecting responsibility for the lack thereof. Moreover, governments provide fewer additional resources for policy implementation 3) when their chances of getting re-elected are low due to intense political competition and they therefore do not have to dedicate much thought on whether their policies will eventually have their intended effect. While the impact of issue salience (measured as party system salience) and electoral competition remains consistent across all models, the influence of institutional constraints exhibits some variability depending on the exact model choice. Last but not least, the analysis suggests that 4) these findings primarily apply to policy areas where citizens can relatively easily assess and directly experience whether policies are properly implemented. In policy sectors where this is not the case, resource provision for newly adopted policies more strongly depends on organized interests that keep an eye on governments.

Our analysis constitutes a first step towards illuminating the “politics of administrative resource provision,” a still poorly understood phenomenon that crucially influences governments’ success in achieving the objectives spelled out in legislative statutes (Zacka 2022). While we presented factors that account for administrative resource provision in the area of social policy, more research is needed to assess these factors’ generalizability. Our study focuses on European countries, where social policy has traditionally been a public domain of the state. In some other countries, in contrast, social protection tends to be a more “private” affair, where governments often shape the overarching regulatory framework but delegate essential implementation activities to the private sector. These differences suggest that we might see varying dynamics between policy-making and its subsequent execution. Future research should additionally consider “bottom-up” factors that determine how successful public administrations are in demanding additional resources from their political superiors so that they can effectively carry out their policy implementation mandate (Knill, Steinbacher, and Steinebach 2021). Zooming in on the negotiations about whether and to which extent implementing actors are “compensated” for additional policies may be an interesting case for studying bureaucratic politics (Bach 2021).

Moreover, our analysis contributes a new perspective to the literature on democratic responsiveness, which studies the “chain of responsiveness” that connects public opinion to public policy (Powell 2004). Empirical studies usually treat policy adoption as the “end link” in this chain, while leaving aside the question of whether the policies adopted in response to changes in public opinion are also effectively implemented (Wlezien and Soroka 2016). By highlighting the reasons why policies may not be equipped with the resources required for their implementation, our analysis points to a hitherto neglected form of “empty” democratic responsiveness where policies are adopted but are unlikely to have the promised effect because of insufficient resource provision. In this regard, it is particularly striking that while electoral competition has generally been found to increase democratic responsiveness (in terms of policy adoptions), our results indicate that intense electoral competition actually comes with a lower probability that the adopted policies will be effectively implemented. Future research on democratic responsiveness should thus zoom in on the quality of policy implementation (and its determinants) to assess whether governments really are responsive to their citizens.