Do lobbyists represent the priorities and preferred policies of some citizens better than others? The notion of bias in interest representation has long been central to political science, but the literature has struggled to specify the normative idea of biased representation, both theoretically and empirically. To address this problem, we propose a novel approach to operationalizing and measuring the impact of lobbying on political equality.
Historically, most work in this tradition has focused on the descriptive representation of different kinds of organizations. This means counting how many organizations of different types exist (or have access to policy makers) and assessing the extent to which this distribution is compatible with some – often implicit – notion of fairness. Distinctions have been made between organizations representing diffuse and specific interests, producers and consumers, organizations with or without ordinary citizens as members, and business groups vs. ideational groups. These studies tend to find bias in the form of specific business interests being able to mobilize more lobbying resources than organizations representing diffuse public interests and people of lower socio-economic status (Schattschneider Reference Schattschneider1960; Schlozman, Verba and Brady Reference Schlozman, Verba and Brady2012; Walker Reference Walker1991). However, as Lowery and Gray (Reference Lowery and Gray2004, 9) pointed out, ‘without a standard – some assessment of what an unbiased interest system might look like – against which to compare these numbers, they can tell us little about bias in interest representation’. Similarly, two decades ago, Baumgartner and Leech (Reference Baumgartner and Leech1998, 83), in their review of the literature on bias and diversity in the interest group system, concluded that ‘[i]n spite of a great range of interesting and provocative research results, few strong conclusions emerge from this literature, partly because scholars cannot agree on a shared point of reference’ (see also Lowery et al. Reference Lowery2015). This inherent deficiency in the literature has yet to be resolved.
How to study equality and bias in interest representation?
We propose to focus on substantive rather than descriptive representation (Pitkin Reference Pitkin1967). By substantive representation, we refer to the degree to which lobbying resources are allocated in line with citizens’ priorities and preferred policies. Bias in terms of political equality occurs when the lobbying community represents some citizens better than others with respect to the messages forwarded to policy makers. Thus studying substantive representation means focussing on what lobbyists do – rather than who they are – and comparing that to citizens’ attitudes.
We suggest that the most straightforward way to assess both attitudes and lobbying efforts is to use survey data to ask citizens about their policy views and interest groups about their lobbying activities. This approach to operationalizing representational bias is not perfect. Surveys’ ability to capture peoples' real concerns, or lobbyists’ real activities, may be debated. However, our approach overcomes several problems that have beset previous research (Lowery et al. Reference Lowery2015).
First, our method avoids making assumptions about whose attitudes different groups represent. This is especially problematic for organizations without individual membership, such as business groups, companies and institutions (Schlozman, Verba and Brady Reference Schlozman, Verba and Brady2012, 274). Representational claims of membership groups may also be difficult to validate. According to Lowery et al. (Reference Lowery2015, 1,226), it is ‘more often assumed than demonstrated’ that organizations actually represent the interests they claim to represent. Indeed, when studying the activities of advocacy groups, Strolovitch (Reference Strolovitch2007) found that despite these groups' commitment to social and economic justice, their activities were disproportionally directed towards issues that benefited advantaged subgroups of their constituencies. Our approach mitigates these problems by directly comparing the activities of the organizations with citizens' preferences.
Second, a problem with the traditional descriptive approach is that some segments of society may not be able to mobilize at all. The views of those citizens will be missing in studies that count different types of (existing) groups (Baumgartner and Leech Reference Baumgartner and Leech1998, chap. 5). The approach we suggest takes a representative sample of the public as a baseline for assessing representation.
Third, using citizens' attitudes as the standard is a more valid approach to assessing (in)equality in interest representation than attempts to predefine which ‘significant’ interests should be counted (C.f. Berry in Lowery et al. Reference Lowery2015, 1,216). Even if it were possible in a certain context to identify the relevant interests, it would still be necessary to define a standard for how many organizations (per interest) are needed for effective representation (Hojnacki in Lowery et al. Reference Lowery2015). As Schlozman, Verba and Brady (Reference Schlozman, Verba and Brady2012, 275) explain, ‘In achieving equality of political voice, how much of the total organizational space should be allocated to organizations based on race? Sexual orientation? Attitudes toward capital punishment or the rights of homeowners?’ When measuring substantive representation using lobbying activities and public opinion data, the evaluation is straightforward: Do the lobbying activities represent the priorities of some citizens better than others, according to these citizens’ own assessments of their priorities?
Fourth, focusing on lobbying activities, instead of organizational characteristics, provides a reasonable operationalization of the impact of resource inequalities on bias in interest representation (Cf. Baumgartner et al. Reference Baumgartner2009, chap. 10). Rather than estimating the access to different types of resources of different organizations (such as staff, financial resources and members), and making assumptions about the use of these resources, we propose to measure the amount of lobbying being pursued – and for what ends.
In sum, our approach assumes that what matters in the end is the message received by policy makers – rather than who delivered it – and how that message corresponds to citizens' attitudes. Although this approach stops short of studying actual policy impact, we believe it provides a conception of representation that is more relevant to the notion of effective participation in democratic theory (Dahl Reference Dahl1989), and a more direct way of operationalizing and empirically measuring political inequality generated in the interest groups system.
We build on a more recent strand of research studying the link between public opinion, interest groups and public policy (Flöthe and Rasmussen Reference Flöthe and Rasmussen2018; Gilens and Page Reference Gilens and Page2014; Kimball et al. Reference Kimball2012; Lax and Phillips Reference Lax and Phillips2012; Rasmussen and Reher Reference Rasmussen and Reher2019; Rasmussen, Carroll and Lowery Reference Rasmussen, Carroll and Lowery2014). Only a few of these studies have addressed the question of inequality between citizens in interest group representation. Gilens and Page (Reference Gilens and Page2014) find no association between the preferences of high-income or average-income citizens and interest groups – neither mass-based nor business-oriented groups. Similarly, Kimball et al. (Reference Kimball2012) find limited differences in how well American interest groups represent the agenda priorities of low income and high income citizens. They conclude that the lobbying agenda fails to represent the concerns of citizens from all income groups.
We contribute to this literature not only by introducing a new approach to measuring bias, but also by broadening the sources of political cleavages that are studied and by going outside the United States. In addition to looking at class differences based on income, we also include education, gender and urban/rural residence in our analyses. These are overlapping characteristics of citizens that we know tend to generate differences in political opinion. What we know less about is the extent to which these characteristics correlate with higher or lower levels of substantive representation by the interest group community.
Empirical analyses
Our empirical case is Sweden, which we consider a most likely case for the fair representation of citizens' interests. Sweden is a comparatively egalitarian country both in terms of socio-economic status and gender equality. It has a long tradition of mass-based organizations enjoying substantial political influence (Pontusson Reference Pontusson2005). Thus we expect to find relatively little bias compared to many other political systems.
Our goal is descriptive rather than causal. We do not seek to explain why some citizens are better represented than others, or to disentangle the inherently endogenous link between public opinion and the policy positions of organized interests. The empirical question we ask is how the Swedish interest group community substantively represents the views of different groups of citizens. In Gerring's terms, this amounts to an associational descriptive analysis. Such analyses ‘has causal implications, to be sure. However, the descriptive patterns are important, in and of themselves’ (Gerring Reference Gerring2012, 726). Furthermore, in this particular case we believe the why question is relatively well understood. The drivers of bias in interest group systems have been found in the logic of collective action (Olsen Reference Olson1965), the socio-economic inequalities of political participation (Verba, Lehman Schlozman and Brady Reference Verba, Schlozman and Brady1995) and unequal access to economic resources for different types of interests (Schattschneider Reference Schattschneider1960). Previous research has struggled most to specify the what question – what does bias look like?
We propose to study two aspects of substantive representation – agenda representation and policy representation. By agenda representation we mean the correspondence between public views about what are the most important societal problems and the total amount of lobbying activities that are directed towards those problems. Similarly, we define policy representation as the congruence between citizens' views on which policy proposals should be implemented and the amount of lobbying activities that are being directed towards (and against) these proposals. We refer to the Appendix for the analyses of agenda representation and focus here on policy representation.
The Surveys
We combine existing opinion polls administered during the period 2011–14 with information from an original survey of the Swedish interest group population conducted in 2015. We use public opinion data from the Swedish National Election Studies Program and the Samhälle Opinion Media (SOM) Institute (Berg and Oscarsson Reference Berg and Oscarsson2014; Göteborgs universitet, SOM-institutet 2016). We gather data on a total of forty-seven questions relating to policy issues that were debated in the Swedish public sphere during the study period, and that fell within policy areas that are important to the public (see the Appendix for details).
The population of interest groups was identified through three sources: (1) all incoming letters and e-mails from organized interests to government offices in 2011, (2) all responses to proposals referred for consideration by the government in 2011 and (3) all organizations that organized events during Almedalsveckan in 2011 (a major annual political ‘fair’ in which hundreds of interest groups participate).
We identified a total of 1,723 interest organizations in this way. After cleaning the data (for example, removing groups that were no longer active, did not provide any contact information, etc.) 1,534 groups remained to which the survey was sent. Of these, 650 answered the survey, giving a response rate of 42 per cent. Comparing the overall population of groups with those that responded, there was little bias in terms of group type: 48 per cent of the respondent groups were coded as idea groups, 15 per cent as business groups, 13 per cent as representing professionals and 11 per cent as hobby/leisure groups.
In the survey, interest groups were asked which policy areas they were active in. Groups that indicated activity in a certain area were also asked to give their views on policy issues that related to that area. For example, a group that indicated that it lobbied in the area of education and research would be asked to answer related questions, such as whether the Swedish educational system should rely more on charter schools, or whether private profit should be allowed in the educational sector.
Assessing Policy Representation
To assess the extent to which interest groups represent the public's view, we took into account how much time the groups invested in lobbying, which activities they performed, in which areas they were active, and in which direction they were lobbying on the policy issues in these areas. We asked the groups which lobbying tactics they were using, and how often they used them. To measure activities of ‘inside’ character, we used the question:
During the last 12 months, how often has your organization actively sought access to the following institutions and agencies in order to influence public policies? Cabinet ministers (including their assistants/cabinets /political appointees); elected members from the parties of parliament; national civil servants working in the Prime Minister's Office; national civil servants working in departmental ministries.
To measure ‘outside’ activities we used the question:
During the last 12 months, how often has your organization engaged in the following activities to try to affect or influence public policies? Organize press conferences or distribute press releases; publish research reports; active involvement in media debates such as giving interviews, editorials, opinion letters; place advertisements in newspapers and magazines; contact journalists to increase media attention; stage protests involving members and supporters (strikes, consumer boycotts, public demonstrations); publish statements and position papers on your own website; organize a conference of experts and other stakeholders.
The response options to these questions included ‘we did not do this’, ‘we did this at least once’, ‘we did this at least once every three months’, ‘we did this at least once a month’ and ‘we did this at least once a week’. Our measure of the amount of lobbying efforts performed represents the sum of activities per week. Thus a response indicating that a particular group lobbied the ministries once a month, published research reports once a year, and published statements and position papers once a week would add 12/52 + 1/52 + 52/52 = 1.25 activities. We also took into account the number of policy areas in which a group was active. If a group indicated that it was active in three different policy areas, we would divide its total amount of lobbying activities by three when measuring how many activities went into each policy area from this particular group. Thus in the previous example, if this group had indicated that it was active in the areas of human rights and environmental issues, 0.625 (1.25/2) activities per week would be added to each of these two areas.
In the analyses we use the aggregated figures for all interest groups and in relation to all types of lobbying activities. Thus our measure captures the proportion of the total amount of lobbying activities performed by the Swedish interest group community that goes into lobbying different policy areas and favors different policy proposals. An alternative approach would be to weigh different lobbying activities according to a predefined index of potential policy impact, for example giving different weights to inside and outside strategies. We have refrained from doing that here to avoid arbitrary choices that may steer the results.
The questions posed in the surveys generate answers on 4-, 5- and 11-point scales. For the public, we derive measures of the balance between the share of respondents indicating values above (those in favor of the policy) or below (those against the policy) the midpoints on the scales. For example, one question that has been debated extensively in Sweden is whether the national government should take responsibility for schools from local governments. On a five-point scale, ranging from a very good proposal to a very bad proposal, 65 per cent of the respondents found this to be a very (1) or fairly (2) good proposal, while 14 per cent saw it as a very (5) or fairly (4) bad proposal. Thus, the net balance was 51 percentage points in favor of national state-run schools.
We subsequently compare the balance in the public's issue attitudes with the amount of lobbying activities pursued by interest groups that were active in the area. Looking again at the question of state-governed schools, the distribution of lobbying activities pursued by groups that were against and in favor of shifting from local to national government control over schools was very similar to that of public opinion (11 per cent and 63 per cent, respectively – a net balance of 52 percentage points). The difference between the lobbyists and the public for this particular item is therefore 1 (52–51).
The policy issue with the largest divide between the lobbying efforts of the interest groups and the view of the public as a whole concerned a policy proposal designed to reduce the opportunities for private companies in the health care sector to profit from their activities. A majority (52 per cent) of the lobbying hours put in by interest groups engaged in this policy area came from those that found the proposal to be very or fairly bad, while 35 per cent found it to be a very or fairly good proposal (thus a net balance of 17 percentage points against the proposal). Among the general public, however, a large majority (72 per cent) found the proposal to be very or fairly good, while only 11 percent regarded it as a very or fairly bad proposal (a net of 61 percentage points in favor of the proposal). Thus the difference here is 78 percentage points (61–(−17)).
Findings
Overall, we find that the priorities and preferences of the general public and the activities of the interest groups correlate. For policy representation, comparing the net balances, we find a significant relationship between public opinion and lobbying activities for the forty-seven issues (r = 0.54, p = 0.00). Whether this is a strong or weak correlation may be discussed, but it is clear that the Swedish lobbying community overall promotes the views of the general public – taken as a whole – on salient policy issues through their lobbying efforts.
However, both agenda and policy representation via the interest group system are strikingly unequal when we compare different groups of citizens. Figure 1 shows the correlations between the lobbying activities and attitudes of eight different overlapping sections of the public: men and women, rural and urban residents, lowly and highly educated citizens, and citizens with low and high incomes. The figure reports the differences in the proportion supporting a policy vs. the proportion opposing the policy for the public (X-axis) and for the interest groups (Y-axis).Footnote 1
Figure 1. Substantive policy representation with different sections of the public
Socio-economic status strongly affects the extent to which the interest group system represents the policy attitudes of Swedish citizens. The correlation between attitudes and lobbying activities is almost twice as strong for the highly educated (0.64) as for the lowly educated (0.33). The citizens with the highest income are by far the best represented (0.72). Our measure of policy representation also indicates a difference in the advantage of the urban population (0.56) compared to the rural population (0.46). Men (0.55) are slightly better represented than women (0.51). In the Appendix, we show that these patterns referring to policy representation are very similar to our findings with regards to agenda representation. Swedish lobbyists allocate more resources towards policy problems that are perceived to be more important by citizens with a high income and high level of education compared to those with a lower socio-economic status.
Discussion
We have argued that substantive representation is a promising way to conceptualize and measure the normative idea of a non-biased interest group system, compared to descriptive representation, which has been more common in the literature. By focusing on the message – in terms of the priorities and policies actually being pursued – rather than the character and resources of the messenger, we capture the content of the ‘pressure’ being applied by the interest group system. Furthermore, using public opinion surveys allows us to directly measure citizens’ preferences, which can be used as a benchmark with which to evaluate bias in terms of effective participation – the idea that citizens ought to have an adequate and equal opportunity to express their preferences as to the final outcome (Dahl Reference Dahl1989, 106–114).
Our empirical analyses found that the Swedish interest group system is clearly connected to the public as a whole. However, some citizens are better represented than others. The findings indicate differences between the urban and rural populations (to the disadvantage of the latter), while gender differences are small. Most strikingly, we find strong disparities based on socio-economic status. The fact that such differences appear in Sweden, where we would expect them to be comparatively weak, indicates the general persistence of ‘upper-class bias’ in interest group systems (Schattschneider Reference Schattschneider1960). These results highlight the importance of strong impartial public institutions, political parties, and free and fair elections as counterweights to the influence of interest groups.
The results also highlight the importance of discussing the normative criteria relating to bias. While we have assumed that the interest group system should be proportionally representative of the views of the public (C.f. Mansbridge Reference Mansbridge1999), others may not agree. One could, for example, argue that civil society should instead have a separation-of-powers function, and play the role of a counterweight to elected institutions (Habermas Reference Habermas1962/1989; see also Lax and Phillips Reference Lax and Phillips2009), in particular with respect to guarding minority rights and future generations.
Supplementary material
Online appendices are available at https://doi.org/10.1017/S0007123421000041.
Data availability statement
Data replication sets are available in Harvard Dataverse at: https://doi.org/10.7910/DVN/SF8XNP
