The Macro-Political Context and Interest Groups ’ Access to Policymakers

The article explores how macro-level political factors in conjunction with micro-and meso-level factors affect interest-group access to policymakers. The analysis is conducted based on two original data sets: a population ecology database of Czech, Hungarian, Polish and Slovenian national-level energy policy, healthcare and higher education organizations, and an online survey of these populations. Combining the two data sets allows us to investigate both polity-, population-and organizational-level factors. As the sampled countries have recently experienced democratic backsliding, we also test the effect of closing deliberative structures. The analysis reveals that the political process influences access: legislative fractionalization affects access positively, while the closure of deliberative structures has a negative effect. Nevertheless, the political contextual factors are mediated through variables at both the population (e.g. the size of latent constituency) and organizational (e.g. expertise provision) levels, as well as the meso-level of interorganizational cooperation.

negatively affects access. That is, in policy domains with higher interest-group density it is more difficult for individual organizations to access policymakers, even if individual-level factors remain strong predictors. In other words, in crowded interest communities more groups are vying to gain the ear of policymakers, which potentially poses constraints on their means of access. This is an important finding because it shows that access is not entirely driven by organizational-level characteristics. Indeed, access varies by policy domain and possibly by interest-group population. Yet we still know relatively little about how the macro-political context in conjunction with other population-and organizational-level factors affects access. This lack of focus on the political process in explaining interest-group access is all the more the surprising in light of the well-established scholarship on the importance of the political opportunity structure in social mobilization (Kriesi 2004).
This article adopts the political process model of the political opportunity structure framework (Kitschelt 1986;Kriesi 2004;Meyer and Minkoff 2004) to explore how the macro-political context affects interest-group access to policymakers. However, our theoretical framework also explores how the constraints and characteristics of the national political system are mediated through population (e.g. the size of the potential constituencies of organizations or the number of interest groups in a policy area) and organizational-level factors. Furthermore, we also account for intergroup cooperation at the meso level.
We take advantage of two unique data sets: a comprehensive population ecology data set covering all national-level energy, healthcare and higher education policy organizations in Czechia, Hungary, Poland and Slovenia between 1989 and 2020, and a 2019-2020 comparative interest organization survey of these organizational populations, focusing heavily on political access, perceptions of influence and interest intermediation structures.
Crucially, our sample allows us to examine how the closure of political opportunity structures due to rising authoritarianism affects interest-group access. The four Central and Eastern European (CEE) countries have recently been subject to significant democratic backsliding, albeit to widely varying degrees (Sata and Karolewski 2020;Vachudova 2020). The decline of democratic quality severely constrains the deliberative component of democracy (Coppedge et al. 2022a), potentially leading to a closure of the political opportunity structure (Gerő et al. 2020).
This article continues as follows: the next section sets up the theoretical framework and hypothesizes the effects of macro-political and population-level factors on access to policymakers. The third section describes the sample and research design, and presents the data and variables. In the fourth section we present the analysis, before summarizing the findings and conclusions in the last section.
The potential effects of macro-political (polity and population-level) factors on access While it is theoretically possible that interest groups may wield structural power over public spending by their mere existence in a certain sector (see , Dür 2008: 561), access is generally seen as a key indicator of political influence in the policy process. Exclusion for an interest group means a diminished ability to defend its interests (Binderkrantz et al. 2015). Interactions with the executive, bureaucrats, members of parliament or regulatory authorities enhance the position of organized interests (Eising 2007). Before exerting influence, groups must have access that facilitates 'the exchange of policy-relevant information with public officials through formal or informal networks' (Beyers 2004: 213). Actual influencethat is, the degree to which interest groups can shape policy outcomes to match their preferencesis notoriously difficult to measure, particularly with survey data. Nevertheless, without some form of access to policymakers, interest groups essentially have little chance of shaping policy outcomes.
In the lobbying literature, access is generally explained by organization-level factors such as personnel and financial resources (e.g. Bernhagen 2013; Dür and De Bièvre 2007). Other studies focus on the type of interests and/or constituency represented, demonstrating that business lobby organizations and in general groups representing concentrated, special interests win out against cause groups representing civic, diffuse interests (e.g. Olson 1965). Organizational professionalizationthat is, highly trained staff, leadership and strategic actionis also a key organizational-level characteristic explaining access and influence (McGrath 2005). Information (expertise) provided by organizations to policymakers has also proven important in explaining access (Bouwen 2004;Chalmers 2011).

Theorizing the effect of polity-level factors on interest groups' access
The openness of a political system to societal interests is suspiciously missing from the above-mentioned research agenda. In the study of social mobilization and protest, macrostructural, as well as issue-specific, environmental-level variables have been treated as decisive factors since the publication of Peter K. Eisinger's (1973) seminal article, 'The Conditions of Protest Behavior in American Cities'. Although devised as a framework for the study of contentious politics, political opportunity structure (POS) models have been applied not only in research on political participation (Vráblíková 2014), but also in interest-group research. As a framework for social mobilization, POS has been operationalized, for example, to explain interest-group formation rates, as well as populations by David Meyer (1993), David Meyer and Debra Minkoff (2004), and Anthony Nownes (2004). However, POS has yet to be tested on interest-group access.
In its original form, POS explained the occurrence of mass protest precisely with the lacking (only partially open) structures and processes providing access to policymaking arenas (Eisinger 1973). There is, nevertheless, a framework within POS that emphasizes political institutional structures in explaining the strategies of different stakeholders and the resulting opportunity structure: the political process approach (Kriesi 2004). It was first conceived by Herbert Kitschelt (1986), who concentrated on the input (policymaking) and output (policy implementation) sides of political systems with several factors on both dimensions determining openness to societal groups. As formalized and operationalized by Hanspeter Kriesi (2004), the political process approach provides an opportunity to derive testable hypotheses on how the macro-political context affects interest-group access. It emphasizes that the separation of power between the executive, legislature and judiciary is decisive. The structure of the party system also matters as fragmented party systems are more open. However, it also matters whether the parties are disciplined unitary actors or internally divided ones, regardless of their number (Kriesi 2004).
In view of our country selection (see below), we focus on two polity-level factors potentially affecting interest-group access: party systems and democratic backsliding. First, we assume, in line with Kitschelt (1986), that party system properties matter. In a seminal study, David Lowery and Virginia Gray (1995) found that party systems influenced macro-level characteristics of interest-group systems, namely their densities. We argue that vibrant partisan competition may also facilitate access to policymakers. Our rationale largely follows that of Elin Allern et al. (2021): when political parties compete with multiple other parties for electoral support, they have greater incentives to reach out to interest organizations than parties facing little competition. Dynamic partisan competition may force parties and political actors in general to become more flexible when interacting with organized interests, as the costs and benefits of close ties with singular groups become uncertain in a competitive environment (see Allern et al. 2021). Similar incentives apply to interest groups: amid partisan fragmentation and competition: it is likely more expedient to have multiple 'irons in the fire' than to focus on one single party. Although there are highly fragmented party systems with highly closed and familiar government formation patterns (e.g. Switzerland or Sweden), in post-communist democracies fragmentation is highly correlated with open and innovative government formation (Bértoa and Enyedi 2016).
That is, in CEE countries with more fragmented party systems, interest groups have an incentive to diversify their lobby strategies and access parties across the political spectrum due to fluctuating partisan coalitions. Equally importantly, in more fragile party systems, governments and governing parties equally have strong incentives to include organized interests in policymaking (with an eye on the constituents of organized interests) due to more intense partisan competition. The legislative arena is of pivotal importance for access as members of parliament have a direct electoral mandate as opposed to the executive (in parliamentary systems, that is) and the bureaucracy (Kitschelt 1986). Therefore, we operationalize the openness of the party system as the fragmentation of mandates across competing parties in parliament.
Hypothesis 1: Higher fragmentation of parliamentary mandatesthat is, more partisan competitionpositively affects interest-group access to political decisionmakers.
While Kitschelt (1986) draws attention to the processes of forming policy compromises and consensus, Ruud Koopmans and Hanspeter Kriesi (1995) emphasize the importance of 'prevailing strategies' employed by established political actors to deal with challengers (i.e. societal actors with their own political agenda). They distinguish between exclusive (repressive, confrontational, polarizing) and integrative (facilitative, cooperative, assimilative) strategies. Political stakeholders in consensus democracies are more likely to employ integrative strategies, whereas majoritarian democracies are characterized by exclusive strategies. While Koopmans and Kriesi (1995) consider this an example of a measurable operationalization of the cultural model within the political process approach, we find it to be a useful starting point to integrate democratic backsliding into our framework.
Democratic backsliding represents a systemic challenge for interest groups as government-civil society relations may become increasingly exclusive. Labelled by Robert Sata and Ireneusz Karolewski (2020) as 'caesarean politics', illiberal governments may engage in clientelism characterized by elaborate systems of rewards and punishments, the predominance of informal networks and state capture, potentially resulting in atrophic government-civil society relationships. Illiberal incumbents also nurture their own civil society base. Béla Greskovits (2020) and Grzegorz Ekiert (2019) explored the grassroot support of illiberal incumbents, the emergence of 'illiberal civil society organizations' and networks aligned with authoritarian and nationalist objectives.
In the meantime, the social and political space for independent organizations is increasingly shrinking. Illiberal incumbents may restrict funding of organizations and harass interest groups obstructing their policy agenda (Buyse 2018; Carothers 2016), and weaken or even abolish formal interest intermediation mechanisms (Olejnik 2020). Authoritarian governments generally employ a threepronged strategy to constrain critical NGOs: restricting funding of their projects and operations, harassing NGOs that receive foreign funding and targeting international groups that support civil society organizations (Carothers 2016). However, there is evidence that the general closure of the POS associated with backsliding also forces interest groups to disengage from the political system (Gerő et al. 2020). Hence, altogether, democratic backsliding and the associated infringements on social mobilization may result in the 'squeezing of the civic space' (Buyse 2018) in which organized interests operate, making it more difficult for (non-ideologically aligned) groups to access policymakers.
Hypothesis 2: Democratic backsliding affects access to policymakers negatively.
Theorizing the effect of population-level factors on interest groups' access However, within the macro-political context, it is not only the characteristics of and changes at the polity level that potentially affect interest groups' ability to access policymakers. Organizations are active in a multitude of areas with very different aims (often opposing onesthink about business organizations and trade unions in a specific industry). There are interest groups that represent large, wellorganized, and socially and politically influential constituencies (e.g. a national medical chamber), but also groups with small or large but dispersed constituencies. Moreover, groups operate in sectors with vastly different financial resources. Groups might be affected differently by de-democratization based on the policy area they operate in. Indeed, a recent article argues that authoritarian governments in backsliding democracies apply sector-specific strategies against civil society organizations (Gerő et al. 2023).
And this is precisely the starting point of the human ecology approach in interest-group scholarship. The population ecology framework postulates that groups are members of populations that are aggregates of organizations, which are 'relatively homogenous in terms of environmental vulnerability' (Hannan and Freeman 1977: 934). In other words, organizational selection happens at the level of populations and not at the level of individual organizations. Population ecology models are mostly used to explain the vital rates of interest organizations, that is, their density, diversity, foundation and mortality rates (Gray and Lowery 1996;Hannan and Carroll 1992). As Joost Berkhout et al. (2015) summarized, the underlying theoretical assumption of the population ecology approach is that the number of organizations is constrained by 'the availability of organizational resources, relatively independent of mobilization rates and dependent on the pre-existing density of organizations ' (2015: 465). Density is a key variable because the supplies of potential organizational resources (organizers, members, donors, constituents) become exhausted (Hannan and Freeman 1989: 132-133). That is, as the number of organizations grows, so does the competition for resources (Hannan and Carroll 1992).
Density, however, not only affects mobilization rates (i.e. formation and mortality) but also presumably organizational lobbying behaviour. Hanegraaff et al. (2020) found indeed that the more interest groups are active in a policy domain, the lower their access is to policymakers. Hence, higher densities had crowding-out effects on organizations. We apply the energy-stability-area (ESA) model of organizational density (Gray and Lowery 1996). Accordingly, organizational diversity and density are determined by the carrying capacity of the organizations' environment and not by organization-level variables. The number of potential constituents corresponds to the area term of the model, the 'habitat' or size of population of the interest organization constituency. The energy term corresponds to the current and potential government actions of direct concern of interest (e.g. government spending).
We posit that larger (potential) constituencies should affect interest groups' access to policymakers positively. For example, even if not every employee in a large industry/sector is unionized, unions active in such industries nevertheless should have a greater chance of accessing policymakers. Similarly, we assume that more intensive government actions in an area (e.g. spending, but also the number of laws and regulations) would prompt organizations to mobilize (and lobby) and lead to higher access. Population ecology models assume that larger (latent) constituencies and higher government actions in an area would lead to higher social mobilizationthat is, higher group formations and larger densities (as the resources are more abundant)but larger densities eventually lead to more competition among interest groups (and consequently lower potential access in a population).
Hypothesis 3: The greater the interest-group population density, the more difficult it is for individual groups to access policymakers.
Hypothesis 4: The size of potential constituentsor 'latent constituency'affects interest-group access to policymakers positively.
Hypothesis 5: The intensity of current and potential government actions in the policy area of direct concern of an organization affects interest-group access to policymakers positively.
It is important to note that Hanegraaff et al. (2020) only found density to negatively affect access in older interest-group systems. Their sample consisted of survey responses from two mature (Belgium and the Netherlands) and two younger, postcommunist (Slovenia and Lithuania) interest-group systems, whereby density only had a 'crowding-out' effect in the former two. Our sample consists exclusively of post-communist interest-group systems. However, we operationalize density differently (see below).

The importance of intergroup cooperation (the 'meso level')
We also hypothesize the effect of domestic cooperation between interest groups on access. Marcel Hanegraaff and Andrea Pritoni (2019) recently found that coalition formation was the 'weapon of the weak'. That is, interest groups with waning influence and at a risk of organizational failure are more likely to cooperate with others. Indeed, it was also observed in lobbying scholarship that intergroup cooperation is one of the most important strategies organizations employ to reduce environmental uncertainty (Pfeffer and Salancik 2003). We find cooperation to be an important 'meso level' variable, which may be very relevant in the context of democratic backsliding.
Hypothesis 6: Intergroup cooperation affects interest-group access to policymakers positively.

Organizational-level factors (control variables)
We control for several 'standard' organizational-level factors that may affect access to policymakers: financial and personnel resources, expertise, professionalization and longevity. As noted above, organized interests have different material and nonmaterial resources that affect their and their counterparts' strategies (Dür and De Bièvre 2007). Financial and personnel resources are particularly important for access across all political arenas (Binderkrantz et al. 2015: 96). Legal or technical expertise are also crucial resources explaining the influence of business lobbyists (Chalmers 2011).
Professionalization and longevity may also affect interest groups' ability to access policymakers. According to Heike Klüver and Sabine Saurugger (2013), professionalized interest groups are characterized by creating and retaining positions requiring 'a high degree of qualification in terms of training and relevant working experience ' (2013: 193). As regards longevity, with growing age, organizations might have created long-term bonds with political decision-makers (Fraussen et al. 2015;Klüver and Saurugger 2013). We also test for interest-group type following Mancur Olson's (1965) classical distinction between concentrated and diffuse interests, as concentrated interests generally have a structural advantage over diffuse groups in terms of monitoring the actions of policymakers.
To test whether these 'standard' variables derived from the Western European and US context would explain the variance in access in the post-communist context is also an important research goal of this article. There is evidence that 'insider', network-based strategies are prevalent in CEE. For example, Sándor Gallai et al. (2015) found that organizations formed during the Communist-era in particular tend to prefer governmental and civil relations over formal advocacy and mobilization strategies. Brigitte Horváthová and Michael Dobbins (2019) found that in energy policy the Hungarian government networks closely with preferred business groups with frequent access, while opponents of governmental policy are largely excluded. Maciej Olejnik (2020) calls this phenomenon by illiberal governments in CEE 'patronage corporatism': an elaborate system of rewards and punishments, the predominance of informal networks and state capture by the government and allied special interests.

Sample
Our sample consists of nationally active Czech, Hungarian, Polish and Slovenian healthcare, higher education and energy policy organizations. To define interest groups, we consistently applied three attributes inherent to interest groups: organization, political interest and informality (Beyers et al. 2008). That is, interest organizations strive to influence policy outcomes of their concern on behalf of constituencies or a general political idea not by seeking public office but through interactions with politicians and bureaucrats.
The first data set contains the formation and mortality rates, and density of different populations in these policy domains between 1989 and 2019. We then conducted an online survey based on the population ecology data set among those organizations active during survey fieldwork (between March 2019 and May 2020). The three selected policy areas are diverse and not interrelated, thus increasing the generalizability of the findings. All represent a large portion of public budgets and include both public, non-state and business interest groups, albeit to a varying degree. Finally, the interest-group landscape includes both concentrated interests (e.g. energy producers, healthcare providers/physicians, professoriate) and more diffuse interests (environmental groups, healthcare consumers, students) (see Olson 1965). We identified a total of 1,345 active national-level interest organizations active as late as 2019 (see Supplementary Material Table A1). As a rule, we collected data from public registries of civil society organizations. We crosschecked the data with internet searches, lists from parliaments and different ministries that invited organizations to various committees, meetings and interest intermediation bodies. We used the same set of keywords in all four languages. As in healthcare and higher education there are highly specialized professional, patient and student groups, we made a standardized list of medical professions and higher education disciplines to improve comparability. Regrettably, in many cases, we could not determine the exact date of dissolution, only the last date of activity or that the organization was active/inactive during data collection (2019).
We coded the organizations to smaller domain-specific populations to control for the environmental pressures these organizations aggregately face. Energy groups are divided into fossil, nuclear, renewable (both sectional and business groups in these latter populations) and environmental protection organization populations (we added only environmental groups dealing with clean energy and/or air pollution issues). Higher education organizations are grouped based on whether they represent general higher education faculty/employee interests (labour), student interests, institutional interests (e.g. rectors' conferences) or the interests of specific scientific communities (e.g. political science). Healthcare groups were sorted into five populations: business (e.g. pharmaceutical), institutional (e.g. hospitals), nonmedical staff, medical doctors and medical professional, and patient interest-group populations. Thus, we distinguished between 13 populations across three policy domains, and a total of 52 national-level interest-group populations in the four countries (see Supplementary Material, Table A1).
Subsequently, we conducted an online survey targeting the active organizations between March 2019 and May 2020 in the four national languages. The multiplechoice questions with a comment option addressed their interactions with parties, parliaments, regulatory authorities and policy coordination with the state. We invited a total of 1,264 organizations and received 427 responses (see Supplementary Material, Table A2), which corresponds to a 33.7% response rate. However, the response rates varied by country with 51.8% in Slovenia, followed by 35.3% in Hungary, 33.6% in Czechia and 24.6% in Poland. Organizations perceived as critically important (e.g. major students' organizations, large labour unions, medical chambers, main energy business organizations etc.) were contacted more intensely, and nearly all responded. In general, there is a bias in the sample towards organizations that are primarily advocacy, lobby and interest representation groups over organizations that are alsooccasionallyinvolved in consultations on policy and regulatory issues, but whose main focus is not directly policy related (for a detailed breakdown of the response rates per policy area and populations see Dobbins et al. 2023).
The four countries selected are usually treated in comparative politics as similar cases, and they certainly are alike in many important respects. For example, they all joined the EU in 2004 (Vachudova 2005), boast a high value-added manufacturing export sector, enjoy high foreign direct investment (FDI) inflows and are characterized by comparatively high welfare spending (Bohle and Greskovits 2012). However, they show a considerable variance on several factors of theoretical importance. Importantly, their party systems exhibit different levels of institutionalization, with the Hungarian and Slovenian the most and the Polish the least institutionalized, whereas the Czech system occupies a middle position (Bértoa 2020). They also experienced different levels of democratic backsliding. According to the 2021 Nations in Transit Report by Freedom House, Hungary and Poland underwent the steepest decline in democratic quality ever recorded (Csaky 2021). In fact, Hungary was downgraded first from a stable to semi-consolidated democracy, and eventually in 2020 to a 'transitional hybrid regime'. Poland is still rated by Freedom House as a semi-consolidated democracy. Czechia and Slovenia are both rated as consolidated democracies. Nevertheless, Czechia experienced some regression in democratic quality under the Babiš government between December 2017 and December 2021 (Buštíková ). Between March 2020 and June 2022, the quality of democracy also deteriorated in Slovenia under the third premiership of Janez Janša (Novak 2023). This decline in Slovenia is, however, not reflected in our data as our fieldwork period in summer/autumn 2019 and spring 2020 mostly preceded Janša's premiership.
The nations' lobbying and election finance regulations also differ. Czechia has privately funded elections and weak lobbying regulations (McGrath 2008). Unlike Czechia, Poland has publicly funded elections. Moreover, Poland operates a register for professional lobbyists since 2005. However, lobbying in practice is mainly carried out by associations not required to register and reporting requirements are easily circumvented (Kwiatkowski et al. 2016;Rozbicka et al. 2021). In Hungary elections are publicly funded; lobbying activities since 2010 are, however, only loosely regulated (Laboutková et al. 2020). In Slovenia, lobbying has been more heavily coordinated since 2010 by law-regulating corruption. However, fewer than 100 lobbyists had registered by 2019 (Rozbicka et al. 2021).

Variables and research design
Our measurement for political 'access', our dependent variable, is a composite index of four survey questions measuring interest-group access to the governing parties, regulatory authorities and the level of participation in parliamentary committees/hearings, and includes a control item, an overall evaluation of the level of policy coordination/political exchange with the state. We believe these variables best reflect the main channels of interest-group access and proximity to the state. The items are measured on a 1 to 5 scale, 1 being 'no access' or 'extreme difficulties in access' and 5 being 'full access' or 'very easy access'. We added together the scores of the four questions, which resulted in a continuous scale from 1 to 20. We conducted a principal component analysis and a scale reliability analysis based on the four survey questions to validate our access index. Both analyses confirmed that the four indicators measure the same underlying construct: the principal component analysis resulted in only one factor with an eigenvalue above 1, and the Cronbach's alpha reliability coefficient for the four items was 0.72.
We base our democratic backsliding measurement on the deliberative component index of the V-Dem indices on democratic quality (Coppedge et al. 2022a). The component index is composed of indicators measuring the 'extent to which political elites give public justifications for their positions on matters of public policy, justify their positions in terms of the public good, acknowledge and respect counter-arguments; and how wide the range of consultation is at elite levels' (Coppedge et al. 2022b: 54). That is, the deliberative component index expresses changes in the contextual, polity-level institutional context crucial for interest representation. The index is a continuous scale ranging from 0 (low) to 1 (high).
Instead of observing a single year, it is more instructive to look at the change in the deliberative component of democratic quality since 2004, the year of EU accession for the four countries in our sample. The regional trend is in decline: between 2004 and 2021 the deliberative component index decreased on average with 0.12 points across the 11 post-communist EU member states (Coppedge et al. 2022a). In all four countries in our sample the deliberative component is To test the fragmentation of mandates across parliamentary parties, we use the index of legislative fractionalization proposed by Douglas Rae (1968) and calculated and provided by Klaus Armingeon et al. (2021). The index measures the distribution of seats by number of parties. The index takes values between 1 (maximal fractionalization) and 0 (minimal fractionalization). The party systems in our sample differ quite substantially. The Hungarian legislature is the least fractionalized (0.52), followed by the Polish (0.64), Czech (0.79) and the Slovenian (0.85).
We operationalize population density using our population ecology data set (2018 densities). Our variable measures the relative density across the 13 populations and three policy domains in four countries. We divided the number of interest groups in a respective population by the number of all interest groups in the policy domain (for each country separately) and multiplied it by 100, thereby facilitating interpretation of the results. This way, our density variable shows the relative density of a population as a proportion (%) of the number of all interest groups in its respective policy domain.
For the area term of the ESA model (latent constituency) for the respective populations we generally relied on statistics on the logarithm of sectoral employment per 100,000 inhabitants. However, employment was not in every case a fitting or a possible indicator of constituency size. To test H5 on the effect of the intensity of current and potential government actions (the energy term or resources), we relied on public expenditure as a proportion of GDP for the sector (see Table A3 in the Supplementary Material for details).
The organizational-level variables are all constructed from our online survey. As our backsliding variable is at the country level (see above) and we only have four countries in our sample, its variance is necessarily low. Since Hungary (and partly also Poland) has a much lower V-Dem deliberative component score than the rest of the sample, there is particularly a strong possibility that our results are driven by the Hungarian scores. One obvious solution is to estimate a model with country dummies. However, we have a survey-level variable that measures the (perceived) openness of the government towards policy consultations. We asked our respondents about the frequency of government consultations in their respective area of activity (that is, not the frequency of consultations with their organization!). We use this item to create an organizational-level proxy for backsliding. The frequency of government consultations in a policy area is measured on a five-point scale: never, once a year, twice a year, monthly or weekly. For our multivariate analysis, we dichotomized the variable: it takes the value 1 if the government holds no or only yearly consultations on an organization's field of activity (148 organizations, 47%), and the value 0 in every other case (that is, twice a year, monthly or weekly; 166 organizations, 53%).
For professionalization, we calculated the organization-specific total score of five variables based on this question: To what extent does your organization focus on the following activities as opposed to 10-15 years ago (or since its founding, if founded more recently)?
The wording hence assumes that CEE organizations were largely un-professionalized approximately 15 years ago (Borragán 2006). Similarly to Klüver and Saurugger (2013), we included personnel resources (i.e. paid staff) in our composite professionalization variable, as we believe having employees significantly reflects organizational development. This aspect is supported by other studies that emphasize hiring professionalized staff as a key fundamental of professionalization (e.g. Heylen et al. 2020).
Due to the extreme spread of staff numbers spanning from 0 to nearly 6,000, we opted for a categorical variable (1-5) based on quintiles. Our aggregate variable for organizational professionalization is thus constructed as the mean of the abovementioned variables and a categorical variable for staff scaled 1-5. A principal component analysis and a scale reliability analysis confirmed that the five indicators measure the same construct (only one factor had an eigenvalue higher than 1, the Cronbach's alpha of the six items is 0.74). Our composite professionalization variable takes the values from 1 to 30.
The importance of the expertise that interest groups provide was based on three survey questions regarding economic, scientific and legal expertise. All were measured on a three-point scale: 1unimportant, 2somewhat important, 3very important. For the analysis, we created a dummy variable that takes the value of 1 if an organization gave the answer somewhat or very important to any of the three questions.
We coded concentrated and diffused interests as a dummy variable with the value 1 given to concentrated interests (for the coding scheme see Table A4 in the Supplementary Material). Organizational longevity is measured on a logarithmic scale. The variable measuring domestic cooperation is based on four questions regarding the frequency of organizational cooperation with others in fundraising, representation on advisory boards, issuing joint statements and joint political/policy strategies. All questions were measured on a three-point scale (0 = never, 1 = occasionally, 2 = frequently). Our composite cooperation variable is a sum of these four variables ranging from 0 to 8. For descriptive statistics of the variables employed see Table A5 in the Supplementary Material.

Descriptive analysis
Out of a total of 428 respondents across the four countries, 369 answered at least one of the four access items that compose our composite access index. The proportions of respondents by country remain similar in the subsample. That is, Czech and Slovenian respondents remain slightly overrepresented (the share of Czech groups drops from 28.2% to 24.5%, while the share of Slovenian respondents remains 27.1%) and Hungarian and Polish respondents slightly underrepresented (though their share slightly increases from 22.6% and 22% to 23.8% and 24.1%, respectively). The already stated bias towards more directly policy-oriented organizations, however, becomes somewhat more pronounced.
We graphed the dispersion of our composite political access index per country to see in which countries interest groups have lower access to policymakers (Figure 2). The median is 9 on the 20-point scale among Czech, Polish and Slovenian respondents and 7 among Hungarian ones. Furthermore, the interquartile range in Hungary is between 5 and 9, as opposed to 6 and 12 in Czechia and Poland, and 6 and 11 in Slovenia. Furthermore, Hungary has the clearest outliers: it has all three organizations with the highest composite access index scores (18 out of the maximum of 20).
It is also instructive to explore in which arena Hungarian interest groups experience less access. While Hungarian interest groups often reported that access was very difficult, the pattern of access is particularly different in one venue: the parliament. Some 64.9% of Hungarian survey respondents reported that they do not participate in parliamentary hearings and committees at all (see Figure 3; for the breakdown in the other decision-making arenas see Figures A1-A3 in the Supplementary Material). The closing of the opportunity structure in Hungary affected the legislature the most. The Hungarian parliament has been dominated by disciplined constitutional majorities of Viktor Orbán's party since 2010. For example, 2011the first full year of Orbán's perpetual constitutional majorities saw 356 laws and a new constitution adopted, whereas in the previous legislative cycle between 2006 and 2010 the total output amounted to 591 acts (Várnagy 2012). The legislative juggernaut has yet to rest. The new constitution, the Basic Law was amended 10 times since its adoption in 2011. Recently, Orbán entirely abolished parliamentary control. Since November 2020, according to the government, there has been a 'state of danger' in Hungary, which allows the government to rule by decree. Although it was originally introduced because of the COVID-19 pandemic, the government has not ended this exceptional regime. Indeed, the most recent amendment of the constitution allowed the government to declare another state of danger, this time because of the war against Ukraine (conveniently precisely when the one introduced because of the pandemic could not have been prolonged any longer) (Pásztor 2022a, 2022b). Our survey results thus lend support to Kitschelt (1986), who emphasized that the capability of the legislature to develop policies independently from the executive is a decisive factor determining the openness of a political system to societal demands and organized interests.
However, the importance of access to policymakers through the legislature is also very low in Slovenia, which had not experienced backsliding until our survey fieldwork period (see Figure 1). At the same time, the Slovenian legislature is the most fractionalized of the four. However, because the Hungarian legislature is the least fractionalized, and Hungary experienced the highest backsliding in its deliberative democracy score (see Figure 1), and the Hungarian respondents reported on average the lowest access, our country-level variable models might be driven by Hungary alone. Moreover, as legislative fractionalization and backsliding are almost perfectly negatively correlated (−0.95), they can only be tested in separate models (see below). Again, this very high correlation between the two variables is strongly explained by the Hungarian scores. Yet since our sample comprises only four countries, we now turn to our organizational-level proxy for democratic backsliding: the perceived frequency of consultations in an organization's respective field of activity. Interestingly, almost 28% of Slovenian organizations report no consultations with the government in their field of activity. However, 50% of the Hungarian respondents to this question report only yearly consultations, while 37.1% of Czech and 35.4% of Polish respondents to this question report that there are monthly consultations in their field of activity with the government. If we sort the respondents to this question per policy field, we find that energy policy is the least excluded policy field, only 8.1% reporting no government consultations at all, and 29.1% reporting monthly consultations. Healthcare and higher-education groups give a mostly similar assessment of the frequency of consultation. Nevertheless, healthcare organizations assess somewhat higher frequency, with 50.9% reporting two consultations a year or monthly ones, whereas the corresponding proportion among higher education groups is 40%. Not surprisingly, there is a pronounced difference in access between groups based on their general assessment of the frequency of consultations. The mean composite access index of the organizations that reported no or only yearly government consultations is 8.7 compared with 12.7 of those that reported more frequent ones (biannual, monthly or weekly).

Statistical analysis
We estimated seven ordinary least squares (OLS) regression models on the macro-, meso-and micro-level determinants of access with robust standard errors (Table 1). 1 Since we are interested in how the impact of organization-level characteristics is ultimately mediated by macro-and meso-level factors, we begin in Model 1 with organizational-level (control) variables, before 'moving up' to the meso and macro levels. It shows that the coefficients of the professionalization index and the expertise dummy have a significant effect on access. Organizations providing any expert knowledgescientific/technical, legal or economicenjoy approximately 2.3-points higher access on our 20-point composite access index, which is a substantial advantage. A 1-point increase on the 30-point scale professionalization index is associated with a 0.145-point increase on the access index. That is, one standard deviation on the professionalization index corresponds to about a 1.2-point change on the access index. Both coefficients remain robust across all models (professionalization loses significance only in Model 7, which has the fewest observations).
In Model 2 we add the meso level, that is, domestic intergroup cooperation, which is significant and robust across all models. A 1-point change on the 9-point domestic cooperation index corresponds to approximately a 0.65-point change on the 20-point access index, which is a substantial effect. This already sheds some doubt on the widespread assumption that post-communist policymaking is characterized by informal networks based on loyalty, as the data show that organizational development (in terms of professionalizing, expertise provision and networking with like-minded organizations) can strongly boost access.
In Model 3 we introduced the population ecology variables. The coefficient of relative population density is insignificant but is in the expected direction. This Notes: Standard errors in parentheses. Linear regression models, robust standard errors. * p < 0.05, ** p < 0.01, *** p < 0.001.
lends support to Hanegraaff et al. (2020), who arrived at the same result on a different sample and measuring density on the basis of a survey item: though density exhibits a negative relationship with interest groups' access to policymakers in postcommunist democracies, these younger interest-group populations do not seem to be saturated yet. Thus, we reject H3. Nevertheless, both the coefficients of the logarithm of the latent constituency size (i.e. the area term) and government expenditure as a percentage of GDP (i.e. the energy term) are significant and robust (H4 and H5). However, only the coefficient of the area term (the logarithm of the size of the latent constituency) is in the expected direction, that is, positive. Though the effect is moderate (see Figure 4), the size of the potential constituency of interest groups facilitates access to policymakers (H4). Hence, CEE governments indeed seem more attentive to groups representing broader parts of the population, and not just particularistic interests. Government expenditure as a share of GDP in a policy area interestingly affects access mildly negatively (see Figure 4). We can only speculate why this is so: it might be that government expenditure leads to a hierarchical relationship between the interest groups and the government in a sector, with the government having the upper hand. As a robustness test, we estimated a model only with the three population-level variables and obtained essentially the same results as in the models with the organizational, intergroup cooperation and polity-level variables all added (see Table A7 in the Supplementary Material). How does the openness of the political process influence access to policymaking venues? We noted that the variable change in the V-Dem deliberative democracy index is highly (negatively) correlated with legislative fractionalization (almost perfectly: −0.95). Therefore, we tested them separately (Models 4 and 5), controlling for all individual-level factors, domestic cooperation, governmental resources, constituency size and relative population density. The political process models partially support our theoretical framework: only the coefficient for legislative fractionalization (H1) is significant, while the one for backsliding (H2) does not have a significant effect on access (both are in the expected direction, though). A 1-percentage point increase in legislative fractionalization translates into 0.07 points higher access score (Model 5) (see Figure 4). Thus, more partisan dynamics boost interest-group access.
We also ran a regression model with country dummies controlling for all organizational and population-level (ESA) factors (Model 6). Slovenia was set as the baseline category, as it did not experience significant regression in its deliberative democracy component index until 2019. The regression confirms what we see in the boxplots: interest groups in Hungary have significantly lower access than the rest of the sample; the composite access index for Hungarian groups is on average ceteris paribus approximately 1 point lower than in Slovenia, the reference category. In Models 5 and 6 the coefficient of the dummy for concentrated interest reaches statistical significance, which tentatively supports Olson's (1965) assumption about the advantage of groups representing smaller but well-organized and influential interests over diffuse, more encompassing organizations. However, the finding is not robust, and the effect is rather modest.
In Model 7, we introduced our individual-(organizational-)level proxy variable for backsliding. Those organizations reporting no or only yearly consultations in their respective field of activity with the current government (that is, not between their own organization and the government!) score 2.65 points lower on the composite access index on average than those reporting at least two consultations a year. That is, the (perceived) closing of the political opportunity structure negatively affects access to policymakers across the sample.

Discussion
Based on two new data sets, we explored how systemic properties of political systems affect the capacity of interest groups to access policymakers on a sample of 13 populations across three policy domains and four CEE EU member states.
We based our analysis on the political process approach of POS theory (Kitschelt 1986;Kriesi 2004). We derived two crucial variables affecting interest-group access: the fragmentation of party systems (H1) and democratic backsliding (H2), conceptualized as a general closure of the POS. Following Hanegraaff et al. (2020) we also hypothesized population-level factors: density, governmental expenditure, latent constituency size and several organizational-level variables (professionalization, expertise, finances etc.). Finally, we introduced a 'meso level' into the analysis regarding networking/cooperation between interest groups.
We believe that our analysis generated some substantial new insights that largely confirmed our most important theoretical expectations, above all the contextual and population-level hypotheses. First, legislative fractionalization (H1) increases access, lending support to Allern et al. (2021): when the partisan landscape becomes more crowded, not only political parties but also the whole political system become more open to interest-group input.
However, our organizational level model on backslidingoperationalized as the perceived openness of the government to consult interest groups in a policy field generallyshows that the closing opportunity structure has an adverse effect on interest-group access to policymakers. Hence, we see a certain degree of detachment of illiberal governments from lively engagement with civil society actors and thus a closure of the political process. Controlling for all organizational-, population-and country-level factors, the coefficient of our survey item on political closure signalled large differences in the level of access between those that reported at least biannual governmental consultations in their field activity and those that reported only yearly or no consultations at all. The latter group scores ceteris paribus almost 2.65 points lower on the 20-point composite access index than the former. As our data set comprises the three most significant backsliders in the deliberative component of democracy among the CEE EU member states (Hungary and Poland in absolute, and Czechia in relative terms as of 2019/2020), this finding provides further evidence as to the corrosive nature of the closure of the POS on interest-group access independently from national contexts.
Our analysis also aimed to contribute substantially to the debate on how population-level factors, namely density, latent constituency size and resources, affect access. Hanegraaff et al. (2020) found only a significant and negative effect for population density in mature interest-group systems (Belgium, the Netherlands) and not for the post-communist ones (Slovenia, Lithuania). Our analysis lends support to their finding. Although the coefficient of density is in the expected (that is, negative) direction, the effect is statistically insignificant.
However, latent constituency size (the 'area term') has a modest but positive, significant and robust effect on interest groups' access to policymakers. This finding is particularly important in the context of de-democratization, as it signals that even illiberal governments must consider giving access to interest organizations representing larger segments of society (significant professional and business groups, union confederations etc.). The significantly negative effect of resources (the 'energy term'), that is, governmental expenditure on interest groups' access as a proportion of GDP in a sector, lends support to what Sabina Avdagic (2005) called an 'inverse-dependence' relationship between unions and the state in CEE (with the state having the greater power). Although our study's scope is not state-labour relations, the results nevertheless imply that this might be a more general phenomenon between interest groups and governments in the region.
So how can organizations deal with potentially adverse macro-level conditions at the micro and meso levels? Our findings show that, regarding the effect of backsliding, both the coefficients for professionalization and domestic cooperation stay significant at the individual and meso levels, respectively. Hence, even in the context of closing political opportunity structures, potentially hostile governments and increasing population density, professionalization and cooperation with other groups are strong resources against exclusion. On the one hand, this 'optimistic' finding indeed reflects the 'coming of age' of CEE organizations, many of which can avoid becoming permanently incapacitated by means of a stronger focus on internal professionalization and cooperation with other groups. In other words, civic mobilization, when conceptualized as interorganizational cooperation and a focus on internal organizational development, can be a powerful means to counteract democratic backsliding and sustain links between civil society and policymakers despite a closing political space. On the other hand, access through intergroup cooperation in authoritarian contexts may be motivated by coalitions of government-friendly players who help grant access to outsider organizations through their loyalty and personal contacts. Hence, who helps whom gain access requires further exploration.
Despite some data constraints, we believe that our analysis meaningfully enhances interest-group research, as it shows that the access to policymakers is not entirely driven by individual factors, but is equally significantly determined by the macro-political context mediated through population-level factors and intergroup cooperation. Studying new democracies with fragile and volatile political institutions provided us with an opportunity other scholars usually studying interest-group systems based on older democracies simply do not have. However, as many 'Western' democracies also experience increasing electoral volatility, the decline of neo-corporatist structures and higher political polarization, we believe that neither political institutional structures nor political cultural characteristics can be treated as constants any longer. The political context must be brought back into the focus of the research programme on interest groups and lobbying.
Specifically, future research should focus on which organizations still enjoy privileged access to backsliding governments. For example, to what extent are backsliding governments nurturing specific organizations to advance their ideological agenda? Moreover, we still know little about the relationships between political parties and organized interests in the context of backsliding. It would be particularly interesting to explore which types of interest groups are occupying ideological spaces abandoned by (democratically backsliding) political parties, and how both parties and interest groups strategically position themselves in terms of political representation of various segments of society, and how this in turn affects political access.