The state of the art in network analysis of policy making

There are two main bodies of literature on network analysis: the ‘policy network’ approach (Marsh and Rhodes Reference Marsh and Rhodes1992), which considers networks as heuristic devices, hence tends to be rather descriptive (Dowding Reference Dowding1995), and the quantitative sociological approach, which offers the opportunity to systematize and formalize the analysis of networks. SNA is a well-established approach emphasizing the relational aspects of social interaction. It is often presumed, for instance, that employing a network perspective allows for improving on classic hierarchical accounts in social science. This is supposed to legitimise a transcendence of hierarchies in describing policy-making. This is where the frequent reference to networks as ‘diffuse’ and ‘non-hierarchical’ comes from. Yet, formal SNA provides good evidence that network structure is not uniform in social systems and localised concentrations of interaction (clusters), often creating a network hierarchy, are the norm in social exchange (see Wasserman and Faust 1994; Scott Reference Scott2000; Carrington et al. Reference Carrington, Scott and Wasserman2005). Maintaining ties to others entails a cost. Actors therefore tend to be selective when making such resource investments. Indeed it can be argued that this is the core concern of their strategic interaction with other actors. Furthermore not all ties have the same strength or value while actors are unlikely to have the same status or resources within a network. Political action is contingent on an actor's formal authority and their political capital. And while influence can be perceived as diffuse, decision-making is reserved for those with the authority to make decisions and is therefore concentrated (Christopoulos Reference Christopoulos2008). A formal network analysis advances this classic conceptualisation by factoring in an actor's relational environment, in other words, an assessment of their opportunity structure as affected by their relational space.Footnote ¹

In a realistic representation of a network, where actors have a finite number of contacts or ties, their individual perspective is inevitably limited. They have a constrained network horizon that realistically is limited to confidently knowing the relations of their own contacts (Friedkin, Reference Friedkin1983). The information they have on the nature of transactions among other actors is constrained by this horizon. This inevitably creates actor bias. This is partly the reason why political capital is an important actor reputational resource. Most actors do not have the opportunity to transact directly with one another and cannot therefore depend on personal trust for their interaction. While in small, close-knit, long established networks of policy-making, most actors are acquainted with one another or are aware of others by reputation, in wider issue networks network horizon limitations is an instrumental limitation.

Finally, even in the most elementary network of three actors, a triad, not all ties can be assumed positive or reciprocated with the same intensity. Indeed, networks are often the locus of conflict for competing policy ideas. Policy communities are therefore more likely to be clustered and factionalized than ‘dense’ or ‘diffuse’ assumptions imply. Often the actors of interest are those who play brokerage roles or those who are most central within a particular cluster. This is where formal network analysis can offer valuable insights about influence by looking at brokers, information flows and sub-clusters within a wider network (Knoke et al. Reference Knoke, Pappi, Broadbent and Tsujinaka1996 and Pappi and Henning Reference Pappi and Henning1999).

Trust is frequently mentioned as a condition of network interaction. But trust should more legitimately be seen to underwrite reciprocity or underpin network structure (Coleman Reference Coleman1990). When trust is a key parameter of the theoretical framework of an investigation a key consideration is to determine if actor ties or information flows are strong or weak and whether they play a bridging or bonding role (Putnam Reference Putnam2000). The flow of information or influence has to be associated with the status of relevant actors in the network and fundamentally, with constraints and opportunities afforded by network structure, as can be detected through an actors' centrality and brokerage roles.

In the exploratory analysis that follows we employ a network approach, to look at the relationship between influential intermediating institutions and brokerage. Are exceptional intermediators those with the most central position in the network? We also explore the relevance of an actors' extroversion to their political influence.

The configuration of the banking regulation network in the EU

Amongst all the financial services, the oldest and most developed network in the EU is the banking regulation network (Allen and Gale, Reference Allen and Gale2000; Story and Walter Reference Story and Walter1997). This network is part of, or at least intersects with, the international banking regulation network (see Coleman Reference Coleman1996), centred on the Basel Committee on Banking Supervision (BCBS), which was created at the Bank for International Settlements in 1974 to deal with the Herrstatt crisis (Wood Reference Wood2005). Countries are represented in the Committee by their central bank and also by the authority with formal responsibility for the prudential supervision of banking business, whenever this is not the central bank. The Committee's members come from Belgium, Canada, France, Germany, Italy, Japan, Luxembourg, the Netherlands, Spain, Sweden, Switzerland, the United Kingdom and the United States. As explained below, this Committee is important for the cases studied because the CRD was initially negotiated within its framework and agreed to by the national authorities represented in it. A distinction is often made between bargaining and arguing a policy (see Beyers Reference Beyers2008). Some of the exchanges in the interest intermediation network we have studied here are aimed at convincing other actors in the network. In that respect actor positions would not have been adequately explained if we had applied a bargaining model.

The banking regulation network in the EU in its earlier configuration was formed after the re-launch of financial market integration following EMU and the FSAP in 1999 (for an earlier configuration of the network, see Josselin 1997). The FSAP was a five-year plan that contained 42 legislative measures to promote financial market integration in the EU (Mügge Reference Mügge2006). The negotiations of these legislative measures broadened the banking network already present in the EU. As far as the composition is concerned, there are two substantive differences between the ‘old’ and the ‘new’ banking networks: the later version comprises two additional supranational authorities, namely, the EP and the European Central Bank (ECB) and, since 2004, the network has also included the so-called Lamfalussy Committees (see see Coen and Thatcher Reference Coen and Thatcher2005, De Visscher et al. Reference De Visscher, Maiscocq and Varone2008, Quaglia Reference Quaglia2008b).

In the 1980s the EP did not have the power of co-deciding EU legislation, hence it was marginal during the decision making process. From the late 1990s onwards as its power under the co-decision procedure widened, the EP has become a major target for financial sector lobbying, while it also appears deliberately to seek interaction with industry (for example soliciting information and technical resources, see Greenwood Reference Greenwood2003). The ECB, created in 1999, is also influential whenever issues concerning financial stability, for example capital requirements, are discussed in the banking sector (Quaglia Reference Quaglia2008c). After 2004, when the so-called Lamfalussy process was set up (Quaglia Reference Quaglia2007), the network extended to the Lamfalussy committees in the banking sector. These committees, however, were established after the directive discussed in the case study was already in the process of being agreed.Footnote ²

The network encompasses a large number of national and EU banking associations, as well as private financial enterprises. Following a broader trend in the EU, the network has substantially been opened up to industry (See Arrequi et al. Reference Arrequi, Stokman and Thompson2004 and Eising, Reference Eising2004). Indeed, the Commission and the EP implicitly supported the formation of the network by encouraging consultation with policy stakeholders as part of the agenda of good governance and better regulation recommended by the Report of the Committee of Wise Men (2001). There is also some evidence that the Commission and the EP prefer to deal with EU associations, which are regarded as more representative, than with national associations or private firms, which are regarded as more ‘parochial’ (Bouwen Reference Bouwen2004) having narrower national or sectoral interests. Sometimes, even further aggregation is sought at the EU level. For example, in the banking sector, the Commission strongly encouraged the six main EU banking associations to form a common platform, the European Banking Industry Committee EBIC, so as to present a coherent position when interacting with the Commission. The catalyst of this initiative was the negotiation and implementation of the CRD (interview, Brussels, March 2007).

The network also includes national associations, many of which have increased their external lobbying activities vis-à-vis EU bodies and other international bodies, such as the BCBS (see Speyer Reference Speyer and Schuppert2006, Underhill and Zhang Reference Underhill and Zhang2008), besides lobbying at the national level. For example, all national associations are generally members of EU umbrella associations and some national associations have offices in Brussels. All the main German banking associations have offices in Brussels, sometimes working very closely with, or even sharing the office space with EU banking associations.Footnote ³ This reflects some permanent cleavages in the German banking system, where there are three main categories of banks and banking associations representing them: the savings banks, the cooperatives and the private banks (Deeg Reference Deeg1999). These banking groups compete fiercely against each other at the national level as well as in EU regulatory fora (Grossman Reference Grossman2006).

By contrast, the British Bankers Association (BBA) and London Investment Banking Association (LIBA) do not have offices in Brussels, as they feel sufficiently well represented by their national government negotiating in Brussels (interview, London, December 2005). However, in 2005, the City of London decided to open its own office in Brussels to elevate its presence in the regulatory milieu with a view to a more direct influence at the policy process. The Italian banking association (ABI) has one representative in Brussels. The American investment banks based in London and a couple of big private German banks have their own offices or representatives in Brussels.

The banking regulation network does not include consumer protection organizations, even though there is increasing awareness that the under-representation of consumers promotes a negative perception of the network by outsiders or by marginal actors. However, even when consumer organizations are consulted and temporarily included in the network, their input and influence is limited by their lack of economic resources, technical expertise, personnel etc (IIMG 2007a).

Negotiating the Capital Requirement Directive

Capital requirements were regulated by existing EU legislation that was issued throughout the 1990s and largely incorporated by the Basel 1 Accord into EU directives (see Underhill Reference Underhill and Underhill1997). When negotiations began on the Basel II Accord, the member-states agreed that the new capital requirements framework would be incorporated into EU legislation by the amendment of the existing directives, the Codified Banking Directive 2000/12/EC and the Capital Adequacy Directive 93/6/EC, through the recasting procedure. The CRD is articulated on three pillars: requirements for an internal capital assessment by financial institutions (exposure to ‘credit’, ‘market’ and ‘operational’ risks), a supervisory review process conducted by supervisors to evaluate the risk profile of each institution, and market discipline.

Let us now look first at the policy-making process through which the CRD came into being and second at the most controversial issues in the negotiations (see also Quaglia Reference Quaglia2008a). In November 1999 the Commission's ‘review of regulatory capital requirements for EU credit institutions and investment firms’ was issued for consultation to financial services practitioners, market analysts, consumer groups, member-states and other interested parties. The consultation was undertaken principally at the national level via competent authorities. However, the Commission also received comments directly from EU-level associations. The consultation exercise complemented the consultation undertaken by the BCBS, which was launched in June 1999. The second Commission's consultation document on review of regulatory capital for credit institutions and investment firms was issued in February 2001. The consultation document was designed to be read in conjunction with a similar consultation on the new Basel Capital Accord launched by the BCBS in January, but concentrated on issues of particular EU concern.

In November 2002, the Commission published a working document based closely on the draft legislative proposal developed for the implementation of a new capital adequacy regime. The working document was intended to provide the basis for a structured dialogue with representative organizations at both the EU and the national levels. At the EU level, this dialogue took place directly between the Commission and European representative organizations from both the financial services sector and other sectors including small- and medium-sized enterprises and consumers. At the national level, the dialogue was co-ordinated by the national supervisory authorities.

In July 2004, the Commission proposed new capital requirements for banks and investment firms. In September 2005, the EP voted to approve the proposed CRD for credit institutions and investment firms. In October 2005, an agreement was reached in the Council, without the need for a second reading by either the EP or the Council. By June 2006, the CRD was formally adopted by the Council and the EP (2006/49/EC).

Some of the issues that were raised during the negotiation of the Basel II accord were also contentious in the negotiation of the CRD. Other issues, such as those concerning the use of credit rating agencies to assess credit risk and the measurement of operational risk were mainly negotiated and settled in Basel (see Wood Reference Wood2005, Underhill and Zhang Reference Underhill and Zhang2008), hence they were not raised in the negotiations of the CRD. One of the main points for discussion was the implications of the new capital rules relating to the terms of, and access to, bank credit for small and medium-sized enterprises (cf Wood Reference Wood2005, CEC 2004: 23). This issue was prominent for Germany (see BDB 2003; ESBG 2003) as well as Italy (ABI 2003) and to some extent France, countries where SMEs are perceived to be the backbone of the national economy. These member states were keen to set in place a low risk weight for lending to SMEs. The issue of the risk weight for residential real estate was also an issue in that German actors considered the proposed level as too high (ZKA 2003: 17), whereas other participants considered it to be too low (CEC 2004: 33).

A second point for discussion was the trading book review, a measure agreed by the Basel Committee and the International Organisation of Securities Commissions (IOSCO) in July 2005 (BCBS 2005b) and later incorporated into EU legislation. Essentially, the trading book introduced advanced rules for trading activities in which several investment banks and investment firms in the City of London are involved (HMT 2003). More generally, there was the issue of the modifications necessary for the application of the CRD to investment firms (APCIMS 2003: 5).

On certain issues, the CRD made a number of adaptations of the Basel II rules, some of which were rather contentious. The CRD enhanced the role of the ‘consolidating supervisor’ for the supervision of EU cross-border groups, namely, the national supervisor in the member state where the group's parent firm is authorised. Banks with cross-border activities (British-based banks and large private German banks were at the forefront on this) were largely in favour of the establishment of a ‘lead supervisor’ (EBF 2003a: 11, BBA 2003: 9), with greater decision making power as compared to what was proposed for the consolidating supervisor (interviews, London, December 2005, Frankfurt, January 2007). Cross-border banking groups pointed out the disadvantage of duplicating compliance costs, due to the fact that under the consolidating supervisor their activities would still be subject to a somewhat different national system of supervision (CEC 2004: 7, BBA 2003a: 8).

For the British banks and their regulators, a specific concern related to the initial Commission proposal requiring the calculation of capital requirements at a ‘solo’ and ‘sub-consolidated’ level, as well as at the ‘aggregate’ holding company level. Fundamentally, the ‘solo’ model insulates the principal regulated entity from other members of its group, whereas the ‘consolidated’ model allows regulation and supervision to be applied to the top tier (i.e. parent or holding companies) of the group covering all members that provide financial services (BBA 2004, BBA 2003: 49, see also CEC 2004: 10). British banks, in conjunction with the British authorities, wanted the application of the ‘solo’ model to remain a possibility in the UK by virtue of the EU legislation (interview, London, December 2005).

Another relatively minor yet extremely controversial issue was the treatment of ‘intra-group exposure’, as the German savings and co-operative banks wanted zero-risk weighting for this within their sub-sector, and directed intense lobbying at the EP in order to achieve this objective (interviews, March 2007). This provision lowers the capital requirements of savings and co-operative banks, giving them a competitive advantage, and thus it was challenged by private banks in the UK (BBA 2003a: 12), especially in Germany (interviews, Frankfurt, January 2006, Berlin, April 2008, Brussels, March 2007).

Analysis: Centrality and brokerage in the CRD network

What can we infer from the graph relational structure in graph 2? We employ the most widely used algorithms for our analysis as available in Borgatti et al. (Reference Borgatti, Everett and Freeman2002) to further explore relational structure. The French, Italian and British components appear relatively isolated. While the European Banking Federation (EBF) and the German financial regulator, the Bundesanstalt für Finanzdienstleistungsaufsicht (BaFin) and Deutsche Bundesbank (hence forth, Bundesbank), appear to be the major brokers for peripheral groups. The EBF is a broad umbrella association that represents also the French and Italian banks – hence its apparent role as a broker of peripheral groups – though it does not include the German saving banks, which are the pillars of the European Savings Banks Group (ESBG). The brokerage role of the German regulator is somewhat unexpected, but can be explained by the fact that Italian and French associations had several policy preferences in common with the German associations such as seeking a low risk weight for lending to SMEs (see ABI 2003).

The European associations and the German institutions are firmly entrenched in the middle of this network and are strongly interconnected to one another. The implication is that this group is more capable to co-ordinate positions. If network structure matters, such a core group should be effective in orchestrating effective lobbying action by exploiting its brokerage opportunities of peripheral actors. This is confirmed by the fact that German banks and their regulators were able to have many of their regulatory preferences incorporated into the CRD, such as a favourable risk weight for lending to SMEs and residential mortgages, and the zero intra group exposure rules as elaborated below. The positions of the EBF were somewhat more diluted because it is an umbrella association that has to bring together the views of very different parts of the industry, not only in terms of geography but also market segments (interviews, Brussels, March, June 2007). Hence, its ability to influence the outcome of negotiations was impaired by the lack of a coherent and consistent position.

In terms of their centrality, we found that all Europe-wide associations, three German and one British actor demonstrate above average centrality. These actors are the BaFin-Bundesbank, the Zentraler Kredit Ausschuss (ZKA) and the BBA. The ZKA is the German peak association that brings together all the various German banking associations. On the one hand, this contributes to explaining its level of centrality. In other words, German organizations are to some extent overrepresented in the network because the banking sector in Germany is more heterogeneous than in other countries (Deeg Reference Deeg1999). On most issues, because of its umbrella role, the KKA tends to present diluted positions (interview, Frankfurt, January 2006, September 2007, Berlin, April 2008), and this affects its ability to influence the outcome of the negotiations. So, the ZKA successfully lobbied for a favourable treatment of the lending to SMEs, an issue supported by all German banks. By contrast, the German banks were divided on the issue of intragroup exposure, as evidenced by the fact that the ZKA did not mention this issue in its position paper (ZKA 2004).

The BBA is well resourced, traditionally being central within lobbying networks of financial services. These resources were successfully deployed to retain the provisions of the solo supervision in the CRD and in drafting the provisions of the trading book. Reportedly, the British Treasury mobilised on this in the Council and direct links were sought with MEPs (interviews, London, December 2005, April 2007).Footnote ⁸

The BaFin, which is the German supervisory authority for the entire financial sector, and the powerful German central bank, the Bundesbank, intensively liaised with several German banking associations, as well as with other national counterparts and EU institutions (interview, Frankfurt, January 2006). This accounts for its above average centrality.

But centrality alone does not provide a fair representation of the relational advantages and constraints for actors.Footnote ⁹ The first and most obvious observation when examining the network graphs is that there is a strong core-periphery structure. European and German associations are in the core of the relational space while all other actors appear relatively peripheral. This is evident in Graph 2 where the core component is in the top right-hand corner of the graph as well as in Table 1 where the Blocked Adjacency Matrix statistics are depicted. One can observe that the core group has a density of interactions of 0.750, that there is very little interaction between the core and peripheral group and that the peripheral group has a much lower density of 0.167, implying that peripheral actors do not interact much with one another (see Doreian et al. Reference Doreian, Batagelj, Ferligoj, Carrington, Scott and Wasserman2005).

Table 1a: Core-Periphery Statistics, Blocked Adjacency Matrix

Table 1b: Density Matrix

Note: Blockmodelling algorithm as implemented in UCINET VI (Borgatti et al. Reference Borgatti, Everett and Freeman2002). The algorithm groups actors on the basis of their ties to others. The top left quadrant in this instance comprises the nine actors that belong to a core group with high level of within-group ties. Within block densities indicate a strong core-periphery structure.

To have an indication of which actors have an exceptional position in the network and by implication likely exceptional behaviour we run a P1 log-linear probabilistic estimation model (Holland and Reinhardt, Reference Holland and Leinhardt1981, Iacobucci, Reference Iacobucci, Wasserman and Faust1994). This can help us identify how exceptional each actor is regarding their outgoing and incoming ties with other actors. Outgoing ties are assumed to be related to expansiveness – seeking to influence others – and incoming ties to popularity – being influenced by others. We should note that the co-efficients produced are logarithmic probabilities, a one unit increase implies a 2.7 increase in probability. Six actors stand-out. These include the EAPB, EBF and ZKA, who are very popular but do not make attempts at influencing others. This can be explained by the fact that they are ‘umbrella associations’ and their members attempt to link up with them. The first two associations are located at the EU level, hence the national associations that are their members (amongst which, the BDB and the BBA for the EBF) are eager to link up with them. The German members of the ZKA (amongst which, the BDB) are also keen to link up with this German peak association. Other three actors that stand out are the BaFin-Bundesbank, the BDB and the BBA who have the exact opposite relational profile as compared to the three actors previously mentioned, since they are strongly expansive, attempting to influence others, but are not popular. Indeed, the BBA and DBD have a consolidated lobbying experience in EU affairs and it is unlikely they would be targeted by other actors trying to influence them, whereas the opposite is indeed the case. These groups of institutions seem to play distinctively different and potentially complementary roles in the way they relate to the CRD network. So, the picture that emerges from the P1 model is of two types of exceptional actors those that are attempting to influence and those that are a key nexus of influence. In the analysis of their brokerage roles that follows we try to show the relational advantage of each position.

Table 2: P1 statistics. Expansiveness and Popularity Parameters

Note: Routine as implemented in UCINET VI (Borgatti et al. Reference Borgatti, Everett and Freeman2002). G-Square 251.18 (587 D.F.), theta −4.99, rho 8.32.

A basic issue in brokerage is who are these actors that benefit by connecting others? Brokers can be seen as the girders that bind different components in a network. A direct brokerage statistic identifies those actors who provide the sole route between others or who join other actors that have otherwise weak (non-reciprocated) ties to one another.Footnote ¹⁰ Three actors from those earlier identified as having an exceptional structural position also emerge as exceptional in their brokerage role (Table 3). From those we termed popular the EAPB is in a position of brokerage in 47.2 per cent of their relations. The ZKA has 50 per cent of their brokerage identified as direct. While from those we termed significantly expansive only the Bafin-Bundesbank demonstrate substantial direct brokerage with 33 per cent of their network ties in that role. The EAPB indeed acts as broker on behalf of the national associations of public banks, which, with some exceptions (the German public banks, which play a crucial role in the EAPB), are less able to interact at the EU level and articulate their preferences vis-à-vis other EU actors. The implication is that these two institutional actors are significant for the network as they connect others who have limited alternative access to the network, but also that they are likely to be more powerful than other actors, since they could control information or control access to network pendants. They can be seen as the actors most capable of exploiting the structural holes (Burt, Reference Burt2005) of the relational structure.

Table 3: Direct (Honest) Broker Indices, normalized scores

Notes:

Pairs: number of pairs the actor has in the network. Pure brokerage: No alternative tie between any pair of alters joined by broker. Weak brokerage: One directed tie allowed between pairs of alters joined by broker. Non-brokerage: Alters who have tie to broker have 2-way tie with each other as well.

Coefficients rounded to two decimal points.

Computation and output notes from UCINET VI (Borgatti et al. Reference Borgatti, Everett and Freeman2002).

Evidence of the relevance of brokerage can be gleaned by scrutinising the case study material. The EAPB, the EBF, the ZKA, the BDB and the BaFin-Bundesbank were successful in securing a favourable risk weight for credit to SMEs, a position supported also by the French and Italian associations and their national public authorities. The EAPB and the German public banks were successful in securing the ‘intragroup exposure’ provisions.Footnote ¹¹ The ZKA, the BDB and the BaFin-Bundesbank managed to secure a favourable risk weight for residential mortgages. The BBA was successful in having the ‘solo’ supervision model accepted in the CRD. The EBF and the BBA influenced the negotiations on the trading book.

Arguably, these actors were able to have their policy preferences incorporated into the final legislation not acting on their own but, instead, through network brokerage. In all these cases, those actors that had an overt brokerage structural position in the network were able to influence the eventual outcome. This is not to say that the actors concerned always managed to achieve what they wanted but that they influenced the agenda at a critical stage in the debate of the directive. Their network position allowed them to punch above their weight. For example, the BBA, the BDB and the EBF were only partly successful in enhancing the role of the consolidating supervisor, as this measure was opposed by several national public authorities, amongst which is the Bafin-Bundesbank, as well as the German public banks represented in the ZKA and the EAPB.