1
What makes an interest group successful?
An examination into the determinants of financial interest group success
David van Oostveen Student Number: 10289437 Supervisor: Gijs Schumacher Second reader: Joost Berkhout MA Political Science: Political Economy
University of Amsterdam June 2017
2 Acknowledgements
I would firstly like to thank my supervisor Gijs Schumacher for his help. He offered stern but friendly advice and this was especially invaluable for the more technical aspects of my thesis. Other teachers at the UVA were always receptive towards meeting me to discuss ideas or answer questions about their own work. Special thanks to my friends and family who supported me in this time of relative isolation. Additional gratitude for my parents who were patient listeners and always ready to bounce ideas back and forth. A special additional note of love for my girlfriend, who was ever supportive and offered calming words when I was frustrated.
I must admit that this was the most stressful, illuminating and rewarding project during my academic career, leaving me torn between happiness at new breakthroughs and low-key depression at
seemingly unsurmountable errors. After a due break, I hope to encounter more challenges such as these.
3 Abstract
This thesis examines the determinants of financial interest group lobbying success at the Basel Committee on Banking Supervision (BCBS). While accused of dispersing private sector preferences in international financial regulations, there has been little research into the objective measurement of private sector influence at the BCBS. In this thesis it is argued that financial interest groups are influential due to their economic size, their information supply, and the diversity of interest groups that attend the BCBS consultations. These grounded assumptions are tested as three separate hypotheses, each hypothesis arguing that there will be statistically significant impact of the
determinants on the lobbying success of interest groups. These hypotheses are tested using a newly constructed database containing the interest groups participating in the public consultations of the BCBS from 2003-2016. The findings of this thesis were inconclusive, since no statistically significant relationships were found. The lack of statistically significant results might however be one of the most relevant findings of this thesis. The BCBS consultations might be unsuited for automated text analysis due to the presence of a multidimensional structure of conflict. Further explanations for the results include differences in preferences between various financial interest groups, and/or that most BCBS documents are already privately contested before they are publically disputed.
4
Table of contents
Contents
What makes a lobbyist successful? ... 1
An examination into the determinants of financial interest group success ... 1
Table of contents ... 4
1. Introduction ... 5
2. Methodology ... 9
3. Theoretical framework ... 13
4. Data collection and analysis ... 17
5. Preference attainment ... 21 a. Validation check ... 22 6. Determinants ... 28 7. Statistical analysis ... 30 8. Conclusion ... 38 9. Bibliography ... 39 10. Appendix ... 43
5
1. Introduction
Since the 1970’s, finance as an industry has become increasingly global and central to the world economy. Financial trade across borders was initially largely unregulated and unsupervised. This led to a series of bankruptcies and subsequent coordination problems between several national
jurisdictions where these banks were operating. To regulate and address these transnational coordination problems, several transnational institutes were founded (Goldbach 2015).1 The Basel committee (BCBS) was founded in 1974 to ensure more transnational harmonization of banking standards (Goldbach 2015; Woods 2005).
It initially served as an informal meeting forum for banking supervisors from the ten founding countries but gradually transformed into an institutionalized regulatory authority with a written charter. Nowadays it fulfills three distinct functions; First, sharing information about best practices, second, harmonization of the setting and content of national laws on finance, and third, enhancing cooperation by stressing the enforcement of national rules and sharing intelligence (Slaughter 2004, p 54-69).2
Seats in the Committee are occupied by important supervisors from the member countries and decisions are made on the basis of consensus. Formally, decisions have to be approved by the Governors and Heads of banking Supervision (GHOS), which is comprised of central bank governors and heads of banking supervision.3 However, decisions made by the Committee are always approved by the GHOS (Buchmüller 2008, p 19-20). Since country membership of both is the same, the
Committee is the crucial place of decision making. The Committee steers various ad-hoc working groups, which are comprised of national regulators who are temporarily attached to a working group. Its secretariat is composed of only twenty people, of which only five are permanent staff members (Goldbach 2015, p 34-35)
One of its more surprising and enduring aspects is its lack of a legally binding charter, meaning its regulations and standards are not binding on members. Despite this, the Basel Committee’s standards and regulations are adopted worldwide (McKeen-Edwards & Porter 2013; Helleiner & Pagliari 2011; Young 2011; Singer 2007; Goodhart 2011). This adoption stems from its reputation as promoting and harmonizing good practices worldwide in addition to being seen as measures of good financial governance by the World Bank and the IMF (Young 2011; Barth et al. 2006).
The Basel Committee is thus quite different from other transnational governance institutions. It has no binding regulations, it is comprised of unelected officials, and it lacks a large staff. However at the same time the BCBS fulfills a vital role in standard and regulation setting for one of the biggest and most important industries worldwide (Goodhart 2011)
From its inception in 1974 until today, the Basel Committee has increasingly allowed financial firms to provide information and input on the regulations and standards the Basel Committee has been designing. A hallmark for cooperation with financial firms was the decision to rely on internal
1 An immediate impetus can be found in the failure of Bankhaus Herstatt and the US Franklin Bank. Both
failures led to problems for other banks with whom both had finished trades but funds from both were not forthcoming due to bankruptcies. Involved nations argued over who should have jurisdiction and responsibility while other banks veered towards bankruptcies.
2 A sometimes fourth mentioned function is increasing the trust and cooperation between supervisors due to
the regular exchange of information and practices (Kapstein 1996)
3 The initial membership was confined to the ten G-10 countries that founded the BCBS. There has been an
incremental process in adding more developing countries to the BCBS and GHOS. See
6 bank risk management calculations for setting capital standards. The BASEL I standards incorporated this in 1996, with Value-At-Risk (VAR) models from banks partially determining the banks’ own capital requirements (Underhill et al. 2010). The increasing reliance on statistical risk models led the BCBS to form close working relations with technically advanced banks in developing new regulations (Goodhart 2011). From 1996 onwards the BCBS also started to hold public consultations to receive input on consultation papers outlining the proposed new regulations and standards for financial firms. These public consultations quickly became an important avenue where firms and interest groups could communicate their preferences to the BCBS.
This increased access of banks and their interest groups to one of the premier issuers of regulations and standards designed exactly for these banks has led to criticisms from researchers. These researchers stated that banks were increasingly able to ensure rules were made for their private gain, at the cost of the public interest. Researchers such as Underhill, Helleiner, Porter, and others advocated the argument that the BCBS had essentially become complicit in dispersing financial firms’ preferences on global regulations (Underhill & Zhang 2008; Helleiner & Porter 2009; Blom 2011).
Researchers identified several reasons for the easy access and influence of the financial industry on the BCBS and other rule issuing institutions. Pagliari, Young and Chalmers, and others pointed to the limited representation of other interest groups next to financial interest groups at important rule issuing institutions (Pagliari & Young n.d.; Chalmers 2015; Underhill et al. 2010; Scholte 2013). This limited representation of other organizations is partially the result of the
generally opaque nature of financial regulations for the general public, which makes it hard for other organizations to find public support in lobbying at rule issuing institutions (Woll 2013; Pagliari & Young 2013; Klüver 2011; Pagliari & Young 2012a). Culpepper, Bell and Hindmoor noted that the economic size of firms and their importance to national economies and employment were a key reason for their influence (Bell & Hindmoor 2015; Culpepper & Reinke 2014). The existence of vast financial reserves also enables financial firms and interest groups to sustain long and informative lobbying campaigns at these institutions (Hacker & Pierson 2010).
The easy access of financial interest groups and firms to rule issuing institutions such as the BCBS and the disjointed and weak involvement of other civil society actors has led some researchers to argue that the financial industry is essentially dominant in financial regulatory politics (Underhill & Zhang 2008; McKeen-Edwards & Porter 2013; Anheier 2014). Case studies by other researchers have illustrated that although the financial industry often enjoys dominant representation at rule issuing institutions, it is not always able to translate its stated preferences into policy. Clapp and Helleiner showed how an alliance of agricultural companies and NGO’s was able to successfully lobby the US Congress in the wake of the financial crisis to regulate the (agricultural) derivatives market, against the wishes of the financial industry (Clapp & Helleiner 2012). Kastner detailed how an alliance of publically elected officials and consumer protection organizations were able to politicize the formerly uncontested issue of financial consumer protection but that the financial industry still managed to weaken significant parts of the regulations in the passage and implementation phase (Kastner 2014). Helleiner and Thistlethwaite explicitly highlight how the financial crisis enabled a coalition of NGO’s and policymakers to re-regulate the carbon market, which is highly dependent on fincial firms to function. After extensive interviews with BCBS officials and representatives from banks, Kevin Young illustrated how the financial industry had unprecendented access to the BCBS in the BASEL II
agreement of standards and regulations in 2003. This unprecendented access however, led to only limited influence or lobbying success in the three cases that Young studied (Young 2013). Young even found that regulations became stricter after the financial industry had expressed its wishes for less regulation on one of the cases he studied (Young 2012, p 680).
7 empirical studies, which studied interest group mobilization and interest group influence at financial governance institutes. Young and Pagliari examined the diveristy of interest groups at stakeholder consultations in national jurisdictions as well as the BCBS and the International Organization of Securities Commissions (IOSCO). They analyzed 13,466 comment letters from 292 consultations in 58 different governance bodies and found that business groups dominate financial sector consultations, with 91.56% of the comments letter originating from business groups in pre-crisis consultations (pre-2008) and 84.92% in post crisis (post (pre-2008) consultations (Pagliari & Young 2012, p 92). They do note that that there is significant diversity within business groups but do also awknowledge that financial interest groups are skilled at building a coalition with other business groups. They subsequently gain the support of other business groups by pointing out the distributional consequences of regulations on the financial sector for end-users of financial products (Pagliari & Young 2012, p 53-54).
Chalmers examined the mobilization of interest groups by looking at 203 consultations with 2395 unique actors at the European Securities and Market Authority (ESMA) from 2002-2013. He explained the differerences in mobilizations by looking at institutional opportunity and
demonstrations effect. The institutional opportunity was measured by examining whether actors got enough time to prepare, if multiple consultations were held in quick succession and the avenues through which they could comment. Demonstration effects were exogenous shocks like the financial crisis, in which the crisis highlights the cost of regulatory failure to a broader public and incentivizes policy makers to re-regulate industries in order to satisfy angry voters. Crises should thus raise the diversity of interest groups that mobilize to participate in consultations. His results showed that greater institutional opportunity next to exogenous shocks led to a greater diversity of mobilized interest groups (Chalmers 2015, p 497). However, exogenous shocks only led to a more diverse mobilization if they were considered in tandem with increased issue salience, measured as mentions in relevant press of consultations or highlighting the costs of an absence of regulation.
The most relevant large n study to this thesis was Heike Klüver’s book ‘Lobbying in the European Union’ published in 2013. In this book Klüver examines the determinants of lobbying success in the European Union with relatively novel automated text analysis methods. By analyzing the policy positions of the initial consultation paper, the comments from interest groups, and the final policy paper, Klüver was able to identify which interest groups were succesful in the
consultations. She found a clear correlation between the success of interest groups and the measure of economic power, information supply, and citizen support that these interest groups posses. (Klüver 2013).
Examining what makes financial interest groups successful in attaining their preferences at the Basel Committee is the specific goal of this thesis. The BCBS has held public consultations since 2003 and this makes it possible to partially emulate Klüver’s approach (Klüver 2013). This thesis will pay specific attention to three possible determinants of financial interest groups success: economic size, diversity of interests, and information supply. Economic size and information supply are tied to organizations themselves, while the diversity of interest groups is tied to the specific issue or
consultation. Most of these determinants are considered in the literature as influential to determine interest group success, but there have been no attempts to systematically assess their importance in determining financial interest groups success at the Basel Committee.
Contrary to Heike Klüver’s work, the officials of the Basel Committee are not remotely accountable to voters or to other control mechanisms (Underhill 2014). This lack of accountability coupled with a usual absence of public attention for financial policymaking makes it rather
unpredictable which determinants are most influential in determining interest group success at the BCBS (Underhill & Zhang 2008; Underhill 2015; Goldbach 2015). The expectation is that the diversity of interest groups and information supply will have largely the same impact on the BCBS as on other policymaking institutes. The diversity of interest groups is likely quite low due to the complicated
8 nature of financial policymaking and its absence in the public debate but its effect should be largely the same as at EU policymaking institutes. Diverse interest groups deliver their own perspectives on prospective policies and their impact at the BCBS, which could possibly be in opposition to financial interest groups (Chalmers 2015; Klüver 2013; Pagliari & Young 2015). The BCBS is similar to other institutes in the sense that it also has to develop policy on a wide range of interdependent complex issues. To ensure they have enough information they are dependent on information from relevant interest groups, which means that information supply plays a similar role at the BCBS as at other institutions (Chalmers 2012; De Bruycker 2016). This is in opposition to the expectation for economic size, which is expected to have a different dynamic at the BCBS and possibly, and a different impact as well. Economic size in Klüver’s research was influential due to its ties to employment and/or economic growth, which is linked to electoral support (Klüver 2013). In the case of the BCBS, electoral support is less important for policymakers, as they are not dependent on re-election. The BCBS does try to develop regulations and standards that do not overly impact economic growth, and employment. Its staff regularly tries to estimate the impact of its new regulations on the global economy (Angelini & Clerc 2011). This is also reflected in its attempts to facilitate a stable financial environment that also facilitates economic growth.4 When financial interest groups complain about prospective regulations, they emphasize the possible broader economic consequences (Elliott 2010; Institute of International Finance 2010).
Research question:
What are the determinants of financial sector lobbying success at the Basel Committee from 2003-2016?
Hypotheses:
H1: A larger source of information supply will translate into higher success rates for financial interest groups
H2: Larger financial interest groups have more success than smaller ones
H3: A lower diversity of actors will result in higher success rates for the financial interest groups This thesis will start off by detailing its methodology, explaining the workings of automated text analysis as well as how data on the determinants is collected. The shortcomings of these methods and several solutions to mitigate these shortcomings will be contemplated. In the theoretical framework the importance of these determinants and the role these have played in earlier literature will be contemplated. The data collection and analysis chapter will first detail how the data was collected and structured. The second part of this chapter discusses the results from the analysis. In the conclusion the results are summarized and the potential usefulness of this thesis for future research is contemplated.
4 The Basel Committee describes its own work as: ‘The Basel Committee on Banking Supervision (BCBS) is the
primary global standard setter for the prudential regulation of banks and provides a forum for cooperation on banking supervisory matters. Its mandate is to strengthen the regulation, supervision and practices of banks worldwide with the purpose of enhancing financial stability.’
9
2. Methodology
This thesis relies upon multiple methods, with the main method being the measuring of the attainment of preferences by interest groups. Measuring the attainment of preferences is one of the three methods used most often in interest group research (Dür 2008b). The other two are process-tracing and the attributed influence method. Process-process-tracing is more suited to small n level studies, as it requires a researcher to examine in depth the channels through which influence is exerted. Process-tracing studies also use a variety of approaches to collect information to triangulate multiple points of influence.5 However, the informational requirements and multiple methods confine
researchers to using small samples, thus reducing the generalizability of the results (Dür 2008b; Beyers et al. 2008a). A second problem is that lobbying is by nature a partially secretive business, which makes it hard for researchers to uncover all channels of influence (Baumgartner et al. 2009).
The other method, assessing attributed influence works through surveys that ask (expert)-respondents to give a self-assessment of their influence or assess the influence of peer interest groups. One mayor problem of this approach is that it relies on the judgments of interested parties themselves, which might overstate or to understate their or other parties’ influence, based on strategic considerations. Interest groups might overstate their influence to signal their continued efficiency and relevance to (prospective) members or understate their influence in order to hide their true influence from a broader public. Dür and Bievre encountered NGO’s that complained about the influence of industry interest groups and lamented their own lack of influence while industry interest groups argued the exact opposite in the EU (Dür & De Bièvre 2007). The results from self-assessment studies also measure the perception of influence instead of actual influence (Klüver 2013). Since this thesis focuses on an objective measurement of lobbying success and its correlation to several determinants, this would be a significant reason not to use the attributed influence method. A third more practical disadvantage would be that the response rate for some studies is quite low, and the chances as a master-student to get a higher response rate might be correspondingly lower.6
This thesis will focus on measuring the degree of preference attainment. In their comments on consultative documents of the BCBS, financial interest groups state their preferences regarding new regulations and standards. The final policy document is often different from the consultative document because it has taken some suggestions, warnings, and problems from the comments into account, e.a. the BCBS has taken some preferences into account.
The software program R is an excellent help as it offers many different options due to the wide availability of additional packages. It allows the execution of automated text analysis through the ‘readtext’ and ‘quanteda’ packages, in addition to providing statistical analysis to measure the correlation between the ‘success’ of interest groups and the identified determinants. R is utilized to first remove all unnecessary text from the document that does not convey relevant information. This represents an more straightforward method than Klüver’s approach, whom had to convert all PDF files to plain text files and use multiple programs to alter the characteristics of the texts (Klüver 2013, p 82). By using the wordfish scaling tool, it is possible to display the position of the comments as well as the consultative and final policy paper on a policy scale.7 This scale itself does not have any meaning but show if the positions of the three kinds of documents have changed and in what direction.8 If the position of the final policy document is closer to the comments of some interest groups (closer than the consultative paper), these interest groups have managed to influence the BCBS and have (partly) attained their preferences.
5 Semi-structured interviews and surveys are often used in addition to document analysis (Dür 2008b) 6 The response rate for the study by Dúr and Bievre was 48% (Dür & De Bièvre 2007, p 87), Heike Klüver
attained a response rate of 38% (Klüver 2013, p 68)
7 The wordfish tools is integrated within the quanteda Package.
8 Researchers can attach meaning to the left and right side of the policy scale, but this requires an in depth
10 This is immediately one of the advantages of this method; it provides an objective
measurement of influence. There are further advantages. First, it is well suited to large n-scale studies as automated text analysis allows analysis of large volumes of documents. Second, this approach captures multiple channels of influence, as it compares the final policy document against the preferences in the comments (Dür 2008a; Klüver 2009; Slapin & Proksch 2008).9 This does imply that this thesis focuses on examining the degree of interest group success and which determinants correlate highly with this success, at the expense of black-boxing the exact process through which this influence is exerted. There is recent research on the shift of financial interest groups from vetoing new regulations to agenda-setting and self-regulation to avoid being regulated (Young 2013; Young & Young 2014). Nevertheless, financial interest groups send in regular comments to
consultations from the Basel Committee to influence the outcomes. While the influence that causes some policy proposals to disappear from the agenda is left unexamined, in this thesis the influence that interest groups exert when they push for self-regulation or delayed implementation is
examined. More importantly this thesis is explicitly researches the measure and reason why financial interest groups are successful, not about their (shifts in) lobbying strategies.
Klüver’s original approach was criticized by Bernhagen et all for not utilizing the full potential of her raw data due to her dichotomous expression of interest group success. Klüver made a
distinction between successful and unsuccessful interest groups (Bernhagen et al. 2014, p 206). To remedy this thesis takes up Klüver’s suggestion that future researchers could also express the success of interest groups by calculating the relative change ‘in the distance between the Commission and an
interest group from t0 to t1 as a percentage of the original distance at t0’ (Klüver 2009, p 547 italics
added). This ensures that it is possible to differentiate between interest groups that had a large measure of success and interest groups with lower measures of success.
Bunea and Ibenskas offered more fundamental criticisms on Klüver’s work and automated text analysis utilized for interest group analysis in general in 2015 (Bunea & Ibenskas 2015, p 434). Bunea and Ibsenskas argued that using words as a unit of analysis ignores the contextual nature of words. Furthermore, they partially disagree with using quantitative text analysis for analyzing interest group comments because in that context, few words can convey substantive information about groups policy positions (Bunea & Ibenskas 2015, p 343). Furthermore they argue that Klüver wrongfully assumes that most consultations are one-dimensional and since scaling with wordfish results in wordfish placing every document on a single policy scale, this results in incorrect policy scores (Bunea & Ibenskas 2015, p 446).
However, their criticism suffered from shortcomings that Klüver immediately addressed in the same issue of European Union Politics (2015). First, Klüver argues that despite the technical nature of these consultations, the vocabulary that interest groups use is still affected by their policy position on the consultation. Wordfish is able to take this into account by attaching word weights and word fixed effects to the word stems it analyzes. High word weights signal that a word occurs very often in some texts but not in others. Word fixed effects control for the fact that some articles and conjunctions are used very often in all texts. Thus words with high fixed effects do not indicate any policy stance (as all actors employ them regularly), while words with high word weights do indicate differences on policy stances (Klüver 2015, p 460). Klüver directly disputes the importance that Bunea and Ibenskas attribute to the multidimensionality of policy consultations. She argues that the multidimensionality is not an obstacle as long as the multidimensionality does not evoke “a
multidimensional structure of conflict” (Klüver 2015, p 461, italics added). She argues that most of
her own studied consultations are of a unidimensional conflict nature by pointing out two other studies. The large study by Baumgartner et all also indicated that the majority of policy debates (98 policy debates studied in total) have an almost unidimensional structure of conflict (Baumgartner et al. 2009). Klüver points out that even Bunea and Ibenskas in their own critical article find that the most important policy dimension identified by Klüver accounts for 80% of the variation in interest
9 Even if the comments themselves do not persuade the BCBS, we can still see their preferences reflected in the
11 group positions (Klüver 2015, p 462).
Furthermore, Klüver stresses that she is highly aware of the potential validity problems that come with using automated (unsupervised) text analysis. Like Grimmer and Stewart she points out that if the text is not influenced by ideological domination or certain policy stances, wordfish will fail to separate actors on the basis of their ideological stances (Grimmer & Stewart 2013, p 27; Klüver 2015, p 460). To validate the output of automated text analysis, she proposed three different validity checks: hand coding, using different quantitative text analysis programs and utilizing external data.10 In her previous work, Klüver used the three different methods in conjunction to assess the validity of the wordfish results, whereas Bunea and Ibsenskas only use hand coding in addition to using a different quantitative text analysis program.11
The first part of this thesis focuses on the degree of preference attainment that financial interest groups obtain. The comments from the interest groups on the sampled consultations are collected, as well as the consultative document and the final policy document from the Basel
Committee. All non-substantial text is removed and the remaining text is equalized by utilizing R with the Readtext and Quanteda packages.12 After having transformed the raw text, R is used to produce word frequency matrixes of the documents. This means that R will produce an excel sheet with the (stemmed) terms of the documents and count how many times they appear. After doing this for all the sampled comments and Basel Documents in a consultations, the wordfish scaling tool will be used to assess each documents’ policy position.
Wordfish is a statistical scaling method that uses word frequencies to place documents within a single dimension of a policy space (Slapin and Proksch, 2008). It does not use any anchoring (reference) documents. The method assumes that word frequencies follow a Poisson distribution, which has a single parameter (λ) representing both the mean and the variance. The functional form of the model is the following:
Figure 1: Wordfish functional model
13
”Where Yij is the frequency of word j in the document of the interest group. Α is a set of document fixed effects that controls for the length of documents. ψ is a set of word fixed effects that control for the fact that some words (even after removing non-substantial text) occur much more often than other words. β is an estimate of a word-specific weight capturing the importance of word j in
discriminating between policy positions and ω is the estimate of actor i’s policy position. The entire right-hand side of the equation is estimated by an expectation maximization algorithm. In order to identify the model, α 1 and the mean of all policy positions of actors are set to 0 and the standard deviation of all policy positions is set to 1. Confidence intervals are obtained using a parametric
10 The external data was comprised of a short survey amongst the participants in one consultation where she
asked respondents about the positions of other interest groups in the same consultation. The estimates derived from the answers were checked against her automated text analysis and hand coding and the scores attributed by both methods were in agreement 78.75% of all scores (Klüver 2013, p 78).
11 Klüver also question the hand coding procedure of Bunea and Ibenskas, as it is not based on a previously
well- established coding technique whereas hers was based on the Comparative Manifesto Project (Klüver 2009; Klüver 2013; Klüver 2015)
12 Punctuation, numbers, symbols are all removed, the remaining text is converted to lowercase, spelling is
converted to English (instead of American-English spelling) and all words are reduced to their stems.
12
bootstrap.” (Klüver 2009, p 538, italics added).
This ensures that it now becomes possible to distinguish which actors have been successful in changing the final policy documents position towards their position (relative to the consultative documents position). To address the dichotomous coding criticism by Bernhagen et all, the relative change in the distance between the Commission and an interest group will be calculated by taking the distance from t0 to t1 as a percentage of the original distance at t0 (Bernhagen et al. 2014). This score thus becomes the first dependent variable (dv1). To see if there is a significant difference in results, Klüver’s initial, dichotomous expression of preference attainment will also be employed. In this expression a 1 means that an interest group has shifted the committee’s position towards its view, and a 0 means that it has not succeeded in doing this. This will be the second dependent variable (dv2).
All interest groups in all consultations with their associated preference attainment scores will be added in one excel sheet. The second part of this analysis will be added to this excel sheet to ensure that the diversity, word count, and economic size/employees are all in one dataset.
As pointed out by Klüver, Ibenskas and Bunea amongst others, there needs to be a validity check of the wordfish results to ensure that they actually represent the documents positions and are not an artifact of the structure of the texts themselves and dismiss the possibility that texts are not affected by policy stances (Klüver 2009; Bunea & Ibenskas 2015; Klüver 2015; Grimmer & Stewart 2013). Manual hand coding will be performed for all actors in consultation 31. Consultation 31 was chosen as it contains a variety of different interest groups. A variety of policy scores can thus be expected with different interest groups. Two additional reasons were the average number of
participants, which makes hand coding feasible and the topic of the consultation, which was a typical BCBS topic.14 The coding scheme will be informed by the method of the Comparative Manifesto Project as Klüver suggested (Klüver 2015, p 260; Werner 2014). The unit of analysis in this hand coding exercise will be natural sentences, as these provide equally valid content estimates as quasi sentences, but perform better on reliability scores (Däubler et al. 2012).
The second part of this thesis relies on statistical analysis between the interest groups and the organizational level and issue level determinants identified in this thesis. The easiest determinant to calculate is the diversity of interest groups. A modified version of the International Standard Industrial Classification Scheme (ISIC rev.4) will be used, a United Nations scheme for classifying actors in diverse economic sectors (Pagliari & Young 2012a; Chalmers 2015).15 After studying the wide variety of interest groups, 4 broad categories were chosen, within which there are 18 different organization types. After classifying all interest groups per consultation, the diversity of interest groups of all 39 consultations and per individual consultation will be calculated. Actor diversity can be calculated by using a Herfindahl-Hirschmann Index (HHI).16 The HHI index is commonly used to calculate the dominance of firms per industry, but has also been used in the context of interest group representation (Rasmussen & Carroll 2014; Baroni et al. 2014; Chalmers 2015). The actor diversity will be compiled in an excel sheet. A shortcoming of this approach is that a high HHI score does not explain which types of industry groups are dominant. A consultation could thus have a very high HHI score but central banks could actually have been dominant in a consultation instead of financial interest groups. This shortcoming is partially solved in this thesis by the manual counting of all
14 The consultation contemplated the exact way that capital requirements for members of centralized clearing
parties should be calculated.
15https://unstats.un.org/unsd/cr/registry/isic-4.asp. Pagliari and Young have a slightly different categorization
than Chalmers. This thesis will have a slightly different categorization as well, due to the different categories of group that respond to consultations at different rule issuing institutions.
16 HHI is measured as the sum of the squared proportions of actors belonging to each of the 18 actor categories
considered in this analysis. The index ranges from 0 to 1, with values closer to 0 indicating greater actor diversity and values closer to 1 indicating less actor diversity. See (Chalmers 2015; Rasmussen & Carroll 2014; Lowery & Gray 2015)
13 different types of interest groups per consultation, which allows a list of which groups actually dominate every consultation.
Economic size will be measured by the operating revenue per year in dollars and the number of employees.17 Both of these express information about the size of a firm. These will be displayed per firm in an excel sheet. Each excel sheet covers one year. Unfortunately, this data is only available for banks and financial holdings. This means that the number of observations for this determinant will be limited.
Information supply is measured by the number of words on interest groups consultation submissions, after removing text passages devoid of substantial information. The word frequency matrixes produced by R will feature the number of words per comment letter and is thus easily added to the preference attainment excel sheet.
Once all data is collected, a multiple linear regression will be performed to examine the correlation between the dependent variable (dv1) and the independent variables (operating income, employees, word count and diversity). The presence of the dichotomous second dependent variable allows a logistic regression to be performed between the second dependent variable (dv2) and the independent variables, since the second dependent variable is a categorical variable.
3. Theoretical framework
Measuring the influence of financial interest groups starts by explaining how this thesis defines financial interest groups and influence. Interest groups in this thesis are defined by three key factors: organization, political interests, and informality (Beyers et al. 2008b).18 An interest group is thus an organized actor that tries to influence policy or regulations and is comprised of private interests (Beyers et al. 2008b).19 Financial interest groups are comprised of banks/holdings and membership organizations such as the International Institute of Finance (IFF). Financial interest groups are thus defined according to their economic sector but a distinction in category between different kinds of banks and financial holding companies will be added, as their preferences are often different.20 The turnover and the number of employees of the different kinds of banks will also be vastly different.
Influence in this thesis is understood as ‘the ability to shape political decisions in line with the interests groups preferences’ (Dür 2008, p 561). This means that the focus is on the first face of power, the ability to affect a collectively binding decision. This means that this thesis will not examine the second face of power (agenda-setting) or the third face of power (preventing other actors from acting on their genuine interest) (Lukes 1986). Recent research has shown that financial interest groups have increasingly focused on setting the agenda at the Basel Committee instead of outright protest against new regulations (Young 2013). Despite these developments, it is still essential to investigate to which degree financial interest groups are able to influence the eventual formulation and adoption of policy. Lobbying success can thus be seen as policy change caused by the lobbying efforts of interest groups.
Attainment of preferences alone however, does not necessarily reflect influence.
Organization level characteristics (economic size and information supply), issue level characteristics
17 The number of employees is often defined as the number of Full Time Employees (FTE’s).
18 As opposed to their organizational characteristics, since this thesis question aims to explain the impact of the
behavior of interest groups, as opposed to their constitutions or patterns of mobilization.
19 Private interests in the sense that it is not trying to occupy a public office.
14 (diversity of actors and issue salience), institutions characteristics (accessibility and openness) in addition to simple luck all play a role in the outcome of policy decisions (Klüver 2013; Chalmers 2017). If certain kinds of actors are systematically successful in shaping policy, this should result in a statistically significant effect between interest group success and these characteristics.
The diversity of actors affects the outcomes of policy consultations due to the variety of interests and perspectives that are presented. The presence of interest groups that counter the view of business are partially responsible for the failure of business to accomplish its aims (Berkhout et al. 2015; Chalmers 2015). Consultations that are dominated by financial interest groups, should thus see these financial interest groups attaining their preferences statistically more often. However, despite a recent increase in diversity at regulatory institutions, this thesis expects the Basel Committee’s consultations to have a low diversity overall (Pagliari & Young 2012b). Opposition to financial interest groups will mostly be voiced by actors from other industries, as opposed to critical NGO’s or trade unions. However, since the Basel Committee issues rules on a variety of different subjects and areas of finance, some consultations could feature a much larger diversity in interest groups due to either the subject or the wider applicability of the rules under discussion.
Economic size could benefit financial interest groups because of two factors: structural power and providing the means to sustain lobbying campaigns. The structural power argument has a long history. In its classic form it is framed as the reliance of governments on business interest to sustain investment and growth (Lindblom 1977; Hall 1986). Business and financial interest thus occupy a crucial and powerful position in capitalist societies. The original theory of structural power has received fair amounts of criticism over the years, because it left little room for alternative explanations of conflicts between business and states (Hacker & Pierson 2002). Later scholars have continually refined and explored how structural power is constituted, how it is influence is brought to bear, and when structural power declines (Culpepper 2011; Bell & Hindmoor 2015). Structural power functions differently at the Basel Committee, as it is comprised of regulators and supervisors from member states. While these regulators and supervisors do not dependent on good economic results for their re-election, they are sensitive to growth and employment concerns. This means that powerful and large financial firms can point to negative effects for global growth and employment if certain measures are disadvantageous to them (Pagliari 2012; Institute of International Finance 2010).
Economic size is a further advantage because financial resources and manpower grant financial interest groups the ability to sustain lobbying campaigns. Reading the consultations, preparing elaborate responses and delivering them to the Basel Committee all require significant resources and expertise. The biggest banks and holdings in addition to their membership
organizations will probably have the submissions with the highest word counts.
A third determinant of lobbying success is the information supply of interest groups. Policy producing transnational institutions such as the Basel Committee, as well as the European Securities and Markets Authority (ESMA), are generally understaffed while being burdened with developing policy on multiple interdependent complex issues (Goldbach 2015). This makes these transnational institutions partly reliant on interest groups to supply them with information on existing problems, flaws in the design of regulations, and advice on how to solve problems (Chalmers 2012; De Bruycker 2016). The size of the information supply is partly dependent on the volume of economic size that interest groups possess, as it is costly to acquire the necessary expertise and allocate manpower to the process of responding to the Basel Committee. Measuring them separately will reveal the separate impact of both while measuring them together will reveal whether or not they are interlinked.
The determinants influence the potential success of the interest groups in different ways. Two of the determinants are characteristics at the organizational level; which are economic size and
15 the level of information supply. Here the characteristics of the interest groups influence their
potential lobbying success. The other determinant, diversity, is characteristic of the consultations, which affect the lobbying success of all the participating interest groups in the respective
consultation. However, as stated before, these determinants might prove to influence each other or all be influenced by other factors. Other factors would include the 2008-2009 financial crisis, after which financial interest groups will be temporarily unable to affect regulations to a significant degree and a broader diversity of interest groups could participate in the consultations (Chalmers 2015) . Although the conceptualization of interest group influence in this thesis closely resembles Klüver’s conceptualization, consultation properties are added. The diversity of a given consultation is a result of interest group behavior, e.a. to participate or not to participate. The diversity of a consultation is ultimately a characteristic of the consultation itself, which might have an impact on the ability of varying interest groups to influence the eventual regulations.
Figure 2: Klüver’s conceptualization of interest group influence
16 Figure 3: Conceptualization of interest group influence in this thesis
Outcome Reason Explanation Observable Implication Convergence of policy output with actor’s policy preference
Luck Influence
Factors not related to Actor Actor Properties Consultation
properties Systemic effect of actor properties Systemic effect of consultation properties No systemic effect of actor or consultation properties
17
4. Data collection and analysis
The consultations are the main source of data for the text analysis. Since these consultations are publically accessible, it was possible to collect the initial consultation document on which interest groups comment, the comments of the interest groups, and the final policy document. Due to the large number of comments, the browser plugin ‘downthemall’ was employed to automatically download these. Afterwards there was a check if the number of downloaded comments matches the number of comments counted by hand for each consultation. A dataset of all consultations was built in excel, displaying the year of the consultation, the number of involved interest groups and the titles.21 Additionally, a database was compiled in excel detailing the unique actors, which category they belonged to, the consultations on which they responded and the number of times they responded in total. This last excel sheet is essential to connect a unique ID number to the unique actors. This ensures that all documents can be tied to a single actor with the organization level determinants.
Some of these consultations are unusable for text analysis for three reasons: because there is no final policy document yet, they are followed up by an additional consultation or there is a very low number of interest groups that responded which makes the results of text analysis unreliable
(Proksch & Slapin 2009).22 The list of usable consultations is still quite large. 39 consultation which attracted 1776 comments from 787 unique interest groups. Similar problems as Klüver were encountered when examining the 1776 comments. Some comments were written in a foreign
language, which meant that wordfish was unable to accurately scale them. There were also a number of PDF’s that consisted of scanned letters where the quality was so bad that optical recognition software proved unable to accurately transform these letters into readable text.23 In total there were 25 unusable comments. Most of these unusable comments (15) were all part of the first usable consultation concerning Basel II in 2003.24 This means a total of 1751 usable comments was left.
All documents were in PDF format. R was unable to analyze 26 documents. The problem was either the formatting of the documents (too many tables, pictures and fonts) or the fact that some PDF’s were scanned from letters. To solve this problem, these documents were transformed to word or plaintext formats by using ‘pdfconverter assistant’, which uses Optical Character Recognition (OCR) to transform scanned PDF’s into plain text. On regular PDF’s, the program ‘AntFileConverter’ was used, which strips all formatting from PDF’s and only leaves plain text. The results of each file were inspected to see if the transformation affected the text negatively.25
After each PDF proved readable, Wordfish was used to calculate the position for the consultative document, the comments, and the final document. These were sorted from the first consultation to the last consultation.26
The diversity of actors is calculated on the basis of the diversity of interest groups involved in
21 This dataset stretches from 2001-2017, listing 66 consultations on which 2965 comments were delivered. 22 Most consultations after 2016 haven’t been transformed into policy yet.
23 With the quality of these texts being so unreadable, manually rewriting them proved impossible. 24 This might point to a further professionalization of interest groups, as there are almost no comments in
foreign languages after this consultation. It might also represent an institutionalization of the public
consultation process, indicating that interest groups understand that submissions in foreign languages are not taken into account.
25 In some cases the process scrambled the text, misplacing letters and incorrectly spelling words. These files
were manually corrected.
18 these consultations. This calculation results in a number of excel tables with the HHI scores for each sampled consultation.
Table 1: Actor diversity and HHI index for consultation 18. table 18. Distribution of actor types (Financial
conglomerates) 2011
Category type frequency % HHI
1. Financial Service Activities 1. Monetary intermediation 3 12 0,12
2. Central Banks 0 0 0
except insurance and pension
funding 3. Activities of holding companies 0 0 0
4. Trusts, funds and similar financial
entities 0 0 0
5. Financial interest groups 7 28 0,28
6. clearing 0 0 0
2. Insurance, re-insurance and
pension 6. Insurance 0 0 0
Funding activities, except
compulsory 7. Re-insurance 0 0 0
social security 8. Pension Funding 0 0 0
9. Insurance interest groups 9 36 0,36
3. Activities Auxiliary to
financial 10. Acitivities Auxiliary to Insurance 0 0 0
service and insurance
activities and pension funcing
11. Fund management activities 1 4 0,04 12. Credit rating Agencies 0 0 0
4. Other 13. Government, regulatory and 1 4 0,04
enforcement organizations
14. Legal and Accountancy 0 0 0
15. Press 0 0 0 16. Individuals 3 12 0,12 17. Other 1 4 0,04 18. Critical Ngo's 0 0 0 Total 25 100 1 0,2416 27
Data on the economic size of banks and holdings is provided by the Bureau van Dijk
database.28 Economic size is defined as operating income in this database (in dollars), in addition to the number of employees. This means that economic size is computed by taking the gross income and deducing the operating expenses, and depreciation of assets. However, this database has a few
27 The HHI score in this table was thus obtained by performing the following calculation:
0,12^2+0,28^2+0,36^2+0,04^2+0,04^2+0,12^2+0,04^2 = 0,2416.
19 omissions concerning operating income as well as the number of employees. The missing data was added manually by sorting through various annual reports from the associated banks and holdings. The number of employees was often directly stated, operating income required more work to calculate.29 The operating income for multiple years required converting multiple currencies. The historical currency data from the XE website proved very useful.30 By taking the date from the annual reports it was possible to look up the value of the currency at that time and convert it to dollars. The data was sorted on economic size per year in different excel sheets. Regretfully, the van Dijk
database only spans from 2007-2016 and thus there is no data on economic size and employees for the consultation in 2003.31
Figure 2: Calculation of Operating Income.
32
The information supply is measured per interest group per consultation. After using R to remove the non-substantial text and make word frequency tables for each consultation, the word count per actor was noted in each consultation in excel.
The data from the Wordfish analysis combined with the determinants forms one excel sheet. All 39 usable consultations were coupled with the participating interest groups. The number of categories was reduced from the initial 18 to 11 to improve the readability of the results. Since this thesis focuses on financial interest groups, the results for the other groups are less important.33 The word count is available for all interest groups and the diversity scores for all consultations. The economic size and employees are only available for banks and holdings from 2008-2016 and are thus missing in the second consultation. The position of each document is noted left of each group. The first dependent variable in this dataset is the relative change in distance between the Basel Commission and the interest group submission from t0 to t1, divided by the original distance at t0. The second dependent variable is based on a dichotomous coding of interest group success, with a 1 representing the situation where the Committee’s position got closer to the interest groups position, and a 0 representing the opposite. The second dependent variable is a categorical variable, which will also allow a logistic regression to be performed. See table 2 for an example of the wordfish position scores.
Table 2: Results of wordfish scaling for Consultation 4 position group
1 -0,49208 Consultative
29 Either as the number of employees or as FTE’s (Full Time Employees) 30http://www.xe.com/currencytables/
31 It proved impossible to add these to the database, due to the reduced availability of companies’ annual
reports from 2003. Most banks and holdings only have publically available annual reports of the last 10 years on their website.
32http://www.myaccountingcourse.com/financial-ratios/operating-income
33 Some categories also contained a minimal number of interest groups, such as the microfinance, law, and
20 2 -0,47504 Final 3 0,269492 algorithmics 4 -0,25039 boa 5 4,049659 cbairt 6 0,244628 cbamr 7 -0,24005 fbf 8 -0,32341 gs 9 -0,19515 hbosirt 10 -0,18882 hkabirt 11 -0,17293 hkabmr 12 -0,31409 isda1 13 -0,27184 jba 14 -0,29504 jpmorgan 15 -0,07174 markit 16 -0,29001 nvb 17 -0,34216 scbirc 18 -0,32573 trmr 19 -0,31532 zkirc
21
5. Preference attainment
In this chapter the position of documents resulting from the Wordfish treatment and the dependent variable are discussed. As explained earlier, the dependent variable is computed by taking the relative change in distance between the Basel Commission and the interest group submission from t0 to t1, divided by the original distance at t0.
The position score has a minimum of -11 and a max of 6.11, with a mean of -0.00132. A look at all position scores indicates that the minimum and the maximum are outliers and most scores are between 3 and -3. A potentially problematic result is that only 23 consultations show a difference in position of the consultative and final document of at least 0.07. Consultative and final documents of the other consultations have only a minimal difference in their position. Comparing the position scores of the comments against the Basel committee’s position in these 13 consultations will thus not provide much useful information. Non-significant position changes also affect the dependent variables scores for the comments that are contained within those consultations. Thus, a separate analysis of the 39 consultations and the 23 consultations will be run to see if the results show a sizable difference in the importance of the determinants.
Table 3: Descriptive statistics of all consultations
=========================================================================== ======
Statistic N Mean St. Dev. Min Max --- Number 1,792 47.0552 56.9548 1 273 position 1,792 -0.0013 0.9894 -10.3360 6.117 1 dv1 1,712 -0.1776 1.6054 -12.2882 10.488 2 dv2 1,713 0.9720 0.1651 0 1 Operatingincome 339 24,538,597 22,179,29 4,500 102,908,439 Employees 339 1,515.2990 26,009.2800 1.0000 478,980 Wordcount 1,714 2,070.8070 2,925.7330 13 40,108 Diversity 1,790 0.2819 0.0717 0.1211 0.5800 Consultation 1,790 21.2380 14.9056 2 51 ---
The scores on the dependent variable mostly resemble the distribution of the position scores, due to the fact that they are directly calculated from the position scores.
22 A look at the 23 consultations where there is a significant difference between the consultative and final document shows clear differences in the position and dependent variable scores. The outliers are less extreme but the mean of both scores is roughly similar.
Table 4: Descriptive statistics of 23 consultations
=========================================================================== ====
Statistic N Mean St. Dev. Min Max --- Number 874 24.6762 19.6680 1 100 position 874 -0.0006 0.9878 -5.5712 5.4579 dv1 828 -0.1644 1.5902 -12.1862 9.4882 dv2 828 0.9589 0.1986 0 1 Operatingincome 191 26,449,287 23,081,397 4,500 102,908,439 Employees 191 2,607.2910 34,650.6800 1.1470 478,980 Wordcount 828 1,757.9440 2,234.5450 36 21,282 Diversity 874 0.3076 0.0789 0.1994 0.5800 Consultation 874 27.1453 12.2990 6 50 ---
a. Validation check
To check whether the position scores, and by extension, the dependent variables are valid, a hand coding exercise was performed on Consultation 31. Consultation 31 was chosen due to its average number of participants (28), which makes hand coding feasible. An additional reason was that many categories of actors are represented within this consultation. This showed that actors belonging to different categories show different position scores. Furthermore, the subject of the Consultation was typical of the usual topics being addressed by the BCBS, as it discussed the new capital requirements banks that are trading through Centralized Counterparties (CCP’s). A final reason is the significant change in positions between the consultative and final policy document.
The wordfish scaling tool produced the following position scores for the submissions from the interest group in consultation 31:
Table 5: wordfish position scores for consultation 31. 1 0,56072 Consultati ve 2 0,800875 Final 3 0,009389 ause 4 -0,03119 barcl 5 0,261137 cba 6 0,191505 ccoil 7 0,258102 ccp12 8 0,430993 cg 9 0,530895 db 10 -0,11469 duba
23 11 0,466202 each 12 0,525964 ec 13 0,268053 fbf 14 0,700381 fia 15 -0,22486 gbic 16 -1,064 hkaob 17 0,168853 ja 18 0,071272 jba 19 -1,47863 jpmc 20 0,295497 jscc 21 -0,17108 jse 22 0,267782 lchc 23 0,326077 no 24 0,391671 sb 25 -4,48348 scb 26 0,374023 se 27 0,431472 ubs 28 0,237075 verbandd erautom
The hand coding started off with a thorough reading of all documents within the
consultation. After reading these it became possible to identify the main issue/dimension of conflict. The main conflict was whether the capital requirements for banks that are trading through central counterparties (CCP’s) should be higher or lower. This helped to build a classification scheme with 30 categories that split into a positive and a negative category.All sentences that could not be
attributed to a category were grouped in the ‘others’ category. The unit of analysis were natural sentences although sometimes sentences were split up in two or even three parts if they contained multiple, discrepant statements. The pro percentages were subtracted from the negative
percentages. A more negative score represents lower capital requirements, a more positive score represent higher capital requirements.
Table 6: Hand coding classification scheme for consultation 31 Higher capital requirements
overall category pro anti
1 Disincentives negative positive
2 Collateral negative positive
3 Default fund negative positive
4 CEM negative positive
5 Trade Exposure negative positive
6 NEMM negative positive
7 Contribution negative positive
8 Margin negative positive
24
10 Waterfall negative positive
11 Stress test negative positive
12 Complexity negative positive
13 Kcca negative positive
14 Capitalisation negative positive
15 Leverage negative positive
16 QIS positive negative
17 Conservative positive negative
18 Cover positive negative
19 Risk sensitive positive negative
20 NIMM positive negative
21 Clearing members positive negative
22 Tranche positive negative
23 Incentives positive negative
24 Ratio positive negative
25 Clearing positive negative
26 IOSCO positive negative
27 OTC derivatives positive negative
28 CCP positive negative
29 Risk
Weighting positive negative
30 Risk Management Positive Negative
31 Other
A thorough reading of the documents lend credibility to the significant difference in position scores for the consultative and final document. The final document clearly rejects earlier proposals from the consultative document and most of the preferences of the actors responding. A contradictory
observation is that the final policy document moves away from the position of the vast majority of the interest groups, the BCBS is thus seemingly unaffected by lobbying efforts. An exception can be found in the following three excerpts below, as the BCBS does take one important preference into account.
Figure 3: Consultative document mentioning Cover 1 and Cover 2 requirements
34
Figure 4: Interest group protesting the use of Cover* requirements for calculating capital requirements
25 35
The cover* requirements would entail significant increase in the contributions of clearing members to CCP’s due to its peculiar way to calculate default funds. All interest groups mention that it would create a disincentive to use central clearing and would create unduly high capital requirements.
Figure 5: Final Document absence of the mentioning of the Cover 1 and Cover 2
36
The final policy document shows that the interest groups have seemingly succeeded in their effort to persuade the BCBS to drop the cover* requirements. Generally, the BCBS in its final policy document has insisted on stricter capital requirements, despite lobbying by these interest groups. The BCBS has persisted in the use of NIMM to calculate the capital requirement, without removing the CCP’s own resources from the calculation. Most elements of the ‘Tranche’ and ‘Ratio’ approach are also used in the final policy document, despite heavy criticism on both.
The hand coding scores below have been calculated by subtracting the number of negative sentences from the positive sentences, divided by the total number of sentences (excluding the others
category). This results in a different scale than the Wordfish estimates, but it is still possible to
35 An excerpt from the submission of the Deutsche Bank, p 3. There are varieties on this requests in almost
every interest group submission.
T The rest of the document was searched for any mentions of the cover* requirements and for the associated
formula given in the consultative document. Neither appears in the final policy document and it is not replaced under another name. Excerpt from the final policy document, p 9.
26 compare both scores by looking at the order in which each group is placed. The order is highly similar, despite the smaller distances between each group, which is caused by the alternative way of calculating the scores. This order confirms the validity of the Wordfish estimates but there are two interesting divergences. First, the submission from ’verbandderautom‘, the automotive industry, is lobbying on a totally different issue than all other interest groups, making both the wordfish and hand coding position score of this group unreliable. Two, the submission from the ‘scb’, the Standard Chartered Bank, was very small, which also resulted in a very different score than wordfish estimate. However, these two divergences are less important for the rest of this thesis, since the financial interest groups usually write long submissions and do contest the BCBS mainly on one similar conflict dimension.
Table 7: Hand coding (left) and wordfish (right) position scores consultation 31 0,6134 Final 0,800875 Final 0,4 verbandderautom 0,700381 fia 0,447826 fia 0,56072 Consul 0,310345 Consul 0,530895 db 0,2981 db 0,525964 ec 0,29414 ec 0,466202 each 0,263623 ubs 0,431472 ubs 0,241543 each 0,430993 cg 0,236754 cg 0,391671 sb 0,23575 sb 0,374023 se 0,20154 lchc 0,326077 no 0,1935 no 0,295497 jscc 0,18934 se 0,268053 fbf 0,18563 fbf 0,267782 lchc 0,176471 cba 0,261137 cba 0,17543 jscc 0,258102 ccp12 0,16243 ccp12 0,237075 verbandderautom 0,14235 ccoil 0,191505 ccoil 0,112353 ja 0,168853 ja 0,052145 barcl 0,071272 jba
27 0,037037 ause 0,009389 ause -0,08108 jba -0,03119 barcl -0,11765 jse -0,11469 duba -0,15385 duba -0,17108 jse -0,15824 gbic -0,22486 gbic -0,16129 scb -1,064 hkaob -0,24566 hkaob -1,47863 jpmc -0,40541 jpmc -4,48348 scb
28
6. Determinants
This chapter serves to discuss the results of the determinants for all consultations. The first
determinant is the diversity index (HHI score), which is the only issue level determinant amongst the three determinants.
As explained earlier, by calculating the HHI index it becomes possible to shortly summarize the relative dominance of one group of actors in a consultation. Traditionally, the HHI scores are interpreted in the following way:
An H below 0.01 (or 100) indicates a highly dispersed industry. An H below 0.15 (or 1,500) indicates a concentrated industry.
An H between 0.15 to 0.25 (or 1,500 to 2,500) indicates moderate concentration. An H above 0.25 (above 2,500) indicates high concentration37
The HHI scores for all consultations range from a low of 0.121094 to a high of 0.58, with a mean of 0.2819. The missing entries (NA’s = Not Available) are due to the spaces necessary to differentiate between consultations.
Table 8: Descriptive statistics of all consultations
=========================================================================== ======
Statistic N Mean St. Dev. Min Max --- Number 1,792 47.0552 56.9548 1 273 position 1,792 -0.0013 0.9894 -10.3360 6.1171 dv1 1,712 -0.1776 1.6054 -12.2882 10.4882 dv2 1,713 0.9720 0.1651 0 1 Operatingincome 339 24,538,597.0000 22,179,295.0000 4,500. 102,908,439 Employees 339 1,515.2990 26,009.2800 1.0000 478,9800 Wordcount 1,714 2,070.8070 2,925.7330 13 40,108 Diversity 1,790 0.2819 0.0717 0.1211 0.5800 Consultation 1,790 21.2380 14.9056 2 51 ---
This would seemingly resemble the results from earlier studies, showing that financial interest groups and banks have a relative dominance in their access and representation at regulatory bodies. However, this result on its own obfuscates that in some consultations, other firms and interest groups enjoy a relative dominance vis-a-vis financial interest groups. In 32 consultations, banks and financial interest groups clearly dominate the consultation. In 7 consultations other groups, such as clearing corporations and insurance groups, have a dominant representation. The dataset with all unique actors and the consultations that they have responded on presented an opportunity to see which other sectors responded most often to the consultations. Of the 100 actors that responded most often, over 25 are actors are clearing corporations, credit rating agencies, accountants, and individuals. This is quite significant, as these 100 actors are responsible for approximately half of the total number of responses (899 out of 1777).
The financial membership groups were composed of international and national organizations. The international groups often delivered comments together, while national
