Europeanization and the Soft Law Process of EU Corporate Governance: How Has the 2003 Action Plan Impacted on National Corporate Governance Codes?

Europeanization addresses the impacts of EU (European Union) membership on national politics and policies. Over time, new policy processes have been developed, such as the use of soft law and the open method of co‐ordination. What, though, are the consequences of these new processes for Europeanization? This article contributes to this under‐researched area by examining the extent to which EU soft law influences policy adoption at the national level. Our empirical application is corporate governance, an area of growing EU policy interest, with significant soft law elements. We analyse the extent to which the European Commission's 2003 plan to enhance corporate governance achieved its aim of 'co‐ordinating corporate governance efforts of Member States'. Our quantitative analysis of 95 national codes issued over 1992–2010 suggests that the Action Plan did influence national policy‐making, but that the degree of national policy alignment depends on when the corporate governance code was issued, where, and by whom.

Introduction EU (European Union) membership impacts on the Member States in numerous ways. Europeanization is a term which, in essence, describes the process by which EU decision-making manifests itself at the national level and the outcomes of EU decisionmaking (see also Radaelli, 2012, p. 1). Europeanization is, therefore, a concept that covers different facets of EU membership. Moreover, the precise meaning of the concept is contested (Olsen, 2002;Bache and Jordan, 2006c) and the chances of finding a single theory 'improbable' (Bache and Jordan, 2006c, p. 17, quoting the work of Bulmer and Radaelli). That said, Europeanization is particularly useful 'as an attention-directing device and a starting point for further exploration' (Olsen, 2002, p. 943).
Much of the proliferating research into Europeanization in recent years focuses on the transmission of, transposition of and compliance at the national level with EU hard law. Ongoing EU integration has, however, resulted in growing diversity in what constitutes EU decision-making. In particular, there has been growth in soft law processes manifested, for example, through the OMC (the open method of co-ordination). By definition, there is no legal obligation to implement EU 'policy' domestically, because the EU process has not yielded hard law to transpose (Saurugger, 2012). That said, soft law and the OMC are still intended to deliver policy change at the Member State level (Trubek and Trubek, 2005). Our attention in the present article is thus directed toward this relatively underdeveloped area of the Europeanization literature, giving us our over-arching research question: to what extent is domestic policy-making influenced by EU-level soft law?
Our empirical focus in this article is corporate governance, an area that has seen growing policy interest within the EU in recent years, for reasons elaborated on below. Our aim is to analyse the extent to which national corporate governance policies have been Europeanized. Specifically, we analyse the extent to which Member States' corporate governance codes have been adapted to address the 26 key priorities 1 laid down in the European Commission's 2003 Communication on company law and corporate governance (Commission of the European Communities, 2003). Corporate governance is particularly interesting in the context of Europeanization because, while the 2003 Communication contains both hard and soft law, most of the soft law elements relate to corporate governance and the 26 priorities. Henceforth we refer to the 2003 Communication as the Action Plan. This is the title of the section which sets out the soft law of corporate governance.
An important question raised in the Europeanization literature is measurement (Bache and Jordan, 2006c;Sifft et al., 2007;Exadaktylos and Radaelli, 2012). We explore how researchers can 'measure' Europeanization and subject it to quantitative analysis, by utilizing a unique dataset of national corporate governance codes issued over the period 1992 to (mid-) 2010. We subject 95 relevant national documents to a detailed content analysis, using as our textual codes the 26 corporate governance priorities from the Action Plan. We apply univariate and multivariate techniques to determine the extent to which the Action Plan has Europeanized domestic corporate governance policy-making, judged as the presence (or not) of the 26 priorities in national corporate governance codes.
We then disaggregate our data, first by country (between 'western' and central and eastern European EU members) and second by the type of body that issued the code (government body, stock exchange, industry body or composite issuer). These additional analyses are of particular relevance to understanding Europeanization, in terms of both processes and outcomes (Radaelli, 2012, p. 7). The first can offer insights into the impact on Europeanization of the EU membership application process, while the second can provide important information about the possible channels by which Europeanization occurs.
Further, by analysing convergence of corporate governance codes on priorities determined by EU-level processes, rather than analysing convergence per se between national codes, we are contributing to both the Europeanization and corporate governance literatures. Most of the work on convergence in corporate governance regimes looks either at convergence per se (are countries' regimes becoming more similar?) or convergence toward one particular model (often the 'Anglo-American' model, but see also below). In contrast, we seek evidence of convergence toward a set of policy-determined corporate governance priorities, negotiated within the multi-level governance structures of the EU. Thus, by centring our analysis on the EU Action Plan, we argue that any quantitative evidence of convergence toward the 26 priorities in national corporate governance codes 1 See Online Appendix Table 1 for details. will be, at the very least, suggestive of Europeanization. Disaggregating corporate governance policy adoption by country and by issuer then allows us to refine our initial, general, findings.
From this we derive specific research questions. Is there evidence that the Action Plan shaped Member States' corporate governance policy-making? Did the type of code issuer affect those policy responses? Were some Member States more responsive to EU-level discussion than others?
The article proceeds as follows. In Section I, we outline relevant literatures on Europeanization, soft law and the OMC and corporate governance policy, and we identify inter-related gaps in these literatures that we seek to address. Section II describes the data and Section III sets out our research design. Section IV presents the data analysis. We finish with a discussion of our findings and their implications for Europeanization and corporate governance, conclusions and suggestions for future directions that research can take.

I. Literature Review
The literatures on both Europeanization and corporate governance are extensiveunsurprisingly so, given the breadth of these concepts. A full exploration of both concepts is beyond the scope of any single journal article. In this section, we seek to identify briefly the main themes relevant to this article, as set out above. In so doing, our aim is to identify inter-related gaps in the two literatures that we seek to address.

Europeanization
Europeanization is a concept which has received much academic attention in the past two decades (major studies include Featherstone and Radaelli, 2003;Bache and Jordan, 2006a;Graziano and Vink, 2007;Stanek and Wach, 2015). Many definitions and meanings co-exist (Exadaktylos and Radaelli, 2012, p. 21;Bache and Jordan, 2006a, pp. 20-2). Meanwhile, as noted above, not only is there no single theory developed of Europeanization, but it is unlikely that one will be.
A helpful definition of Europeanization for our purposes is one that refers to impacts of EU membership on EU Member States, identifying who and what are affected. That is, it refers to outcomes and processes. Thus Europeanization is 'a process of change affecting domestic institutions, politics and public policy. Change occurs when political behaviour at the European Union (EU) level has a transformative effect on domestic political behaviour' (Radaelli, 2012, p. 1). Bache and Jordan (2006b, pp. 13-14) talk of Europeanization impacting on national polities, politics and policies. To keep our analysis focused and tractable, our focus is on national policies. We also note that Europeanization is broader than the EU and has a long history (Wallace, 2000). We stick with this term, however, even though it might be more accurate to talk of 'EU-ization' (Bache and Jordan, 2006c, p. 21).
One possible implication is that Europeanization is seen as a top-down process. That said, it is not something that just happens to countries. Studies have identified dynamics that are top-down, bottom-up and 'horizontal' between Member States (see, inter alia, Cowles et al., 2001;Featherstone and Radaelli, 2003;Bache and Jordan, 2006c;Mastenbroek and Kaeding, 2006). This reflects the complexities of the EU and its multiple policy-making processes, both formal and informal, and the multiple roles for Member States within them.
Another critical distinction is between hard law and soft law. Much of the Europeanization literature analysing policy transfer focuses on EU hard law transmission, transposition and compliance at the national level. Hard law exists where decisions taken at the EU level (whether through Regulations, Directives or Decisions) are legally binding on the Member States, who are obliged to develop a legislative response. In effect, the EU level develops a 'template' that Member States then 'download' (Dyson, 2007, p. 289). What, though, with soft law, where there is no such template and no hard law to transpose (Saurugger, 2012)? These are increasingly important questions which require further attention, as the use of soft law and other 'experimentalist governance' processes (Sabel and Zeitlin, 2008)

expands.
A key soft law process is the OMC. This 'bottom-up' process has facilitated discussion and co-ordination in policy areas that are of common interest, but which the Member States have not been willing to discuss under the ordinary legislative procedure. The intention, however, is still to deliver policy change at the Member State level (Trubek and Trubek, 2005). The OMC seeks to achieve this via a three-step process, which places the Member States at its heart: • jointly identifying and defining objectives to be achieved (adopted by the Council); • jointly established measuring instruments (statistics, indicators, guidelines); • benchmarking, i.e. comparison of the Member States' performance and exchange of best practices (monitored by the Commission). 2 Whether Europeanization refers to top-down, bottom-up or horizontal processes and dynamics, and to hard law or soft law, how do we know if, when and by how much Europeanization has happened? Much Europeanization research does not attempt quantitative analysis, yet measurement is an important part of the literature (Bache and Jordan, 2006c;Sifft et al., 2007;Exadaktylos and Radaelli, 2012). The challenge with all quantitative analyses, but especially where numerical data are obtained from text-based sources, is identifying features that can be quantified yet retain their original essence and meaning while maintaining suitability for quantitative analysis. This is an evolving area of Europeanization research to which we seek to contribute. As set out in the Introduction and explained further below, we have operationalized our notion of Europeanization with reference to specific terms relevant to the policy under focuscorporate governanceto which we now turn.

Corporate Governance
Corporate governance is 'the system by which companies are directed and controlled' (Cadbury, 1992, quoted  shareholders and directors' (Commission of the European Communities, 2003, p. 10). Corporate governance codes are thus instruments which seek to address those problems. Problems can arise from poor management or poor control of agents by principals, and can be made worse by difficult economic conditions. While corporate governance has long been a focus of policy and political attention, within the EU the UK's 1992 Cadbury Report is widely regarded as marking the starting point of rapidly accelerating interest in corporate governance (Aguilera and Cuervo-Cazurra, 2004;Cicon et al., 2012). The backdrop to the Cadbury Report was a difficult economic climate, scandals involving BCCI and Robert Maxwell 'and the controversy over directors' pay' (Cadbury, 1992, p. 9). This interest in corporate governance soon spread across continental Europe (Barker, 2010).
A decade on, the Action Plan emerged amid renewed concern about corporate governance, with more scandals breakingnotably with the collapse of Enron (in 2001), Worldcom (2002) and Parmalat (2003). The authors of the Action Plan were clear about the principal-agent problems: 'Shareholders own companies, not managementyet far too frequently their rights have been trampled on by shoddy, greedy and occasionally fraudulent corporate behaviour' (Commission of the European Communities, 2003, p. 7). Moreover, policy responses were also seen beyond the EU. In the United States, the Sarbanes-Oxley Act was passed in 2003. A few years before, the OECD (Organization for Economic Cooperation and Development) published its 'Principles of Corporate Governance' (OECD, 1999) following the Asian economic crisis, which they modernized in 2004 (OECD, 2004). The EU and OECD principles address similar broad themes. One key difference, central to the present article, is that the Action Plan makes explicit reference to the co-ordination of national policy efforts (see below).
The backdrop to the Action Plan also included around 40 EU-relevant national corporate governance codes adopted in the decade up to 2003 (Commission of the European Communities, 2003). Despite this, it was felt that more was needed. Indeed, there had been calls for 'the European Commission to … come forward with a recommendation for a European code of best practice' (Lannoo, 1999, p. 292). In preparing the Action Plan, the European Commission was responding to an invitation by the EU Competitiveness Council to respond to a report prepared by a high-level group of experts on company law and corporate governance. Moreover, this response was to be prepared 'in co-ordination with Member States' (Commission of the European Communities, 2003, p. 4), consistent with the OMC process. The result was four over-arching corporate governance principles: • enhancing corporate governance disclosure; • strengthening shareholders' rights; • modernizing the board of directors; • co-ordinating corporate governance efforts of Member States.
The first three of these were disaggregated into 26 individual priorities, which form the focus of our quantitative analysis (see below and Online Appendix Table 1). In this, the European Commission gave particular attention to non-legislative approaches (consistent with subsidiarity and proportionality). The Commission's approach was supported by the now-defunct ECGF (European Corporate Governance Forum). It started from the premise that there was no need for an EU Code. Rather, it would seek to achieve 'coordination and convergence of national codes' (Collier, 2007, p. 279).
In this article, we focus on the OMC element of benchmarking, analysing whether domestic codes changed to reflect common goals ('objectives' and 'indicators', in the language of the OMC) identified in the Action Plan. We do not consider information exchange, as this is primarily a horizontal process between countries, with the Commission acting as an information broker. Our focus is thus more on the vertical dimension of Europeanization.
As discussed later, we find that the priorities set down in the Action Plan did not simply reflect what was already in place in national corporate governance codes. Our analytical focus on convergence in national codes toward a set of EU priorities is thus an important contribution to the literature on corporate governance, as well as Europeanization. There is an extensive literature looking at convergence in national corporate governance codes (an excellent overview is provided by Yoshikawa and Rasheed, 2009; see also Aguilera and Cuervo-Cazurra, 2004;2009;Lütz, 2004;Lütz and Eberle, 2008, among many). This literature tends to focus either on convergence per se (are national corporate governance codes becoming more similar?) or convergence of national codes toward one 'model' of corporate governance. This is often analysed as the 'Anglo-American' model, although it can also be a comparison between different styles, such as Anglo-American and Continental European (Lütz, 2004, looks at UK, US and German systems, for example). Moreover, studies identify various drivers of convergence (for example, Yoshikawa and Rasheed identify financial market integration, goods market integration and diffusion of codes and harmonization of accounting rules, which they discuss separately but argue probably interact). Even where convergence is discussed with reference to EU countries' corporate governance codes, convergence per se is often the primary focus (Cicon et al., 2012). This, however, implies market-driven convergence, rather than EU policy-driven convergence (i.e. Europeanization).
The limited literature looking at EU Member States' corporate governance codes finds some evidence of convergence. This stands in contrast to the prediction of Cernat (2004, p. 161), who believed differences across EU countries would continue because EU decision-making was 'poorly equipped' to advance convergence. Despite this prediction, Cromme (2005, p. 363) saw a 'high degree of agreement' across the (then) 47 corporate governance codes in Europe, 'even though company law in the individual Member States continues to be characterized by significant differences'. Cicon et al. (2012) also found evidence of convergence between the codes of 23 (mainly EU) countries. Similar to Cromme, they also found a growing divergence of code content from national legal context. Cicon et al. (2012, p. 622) also analyse whether the content of codes differs by code issuer, something that had been 'unexamined in the literature' (see also Aguilera and Cuervo-Cazurra, 2009, p. 385;Zattoni and Cuomo, 2008). Cicon et al. (2012, p. 622) argue that this is important, given that because 'issuers differ in organisational priorities, internal politics, and the demands of their membership, they are likely to emphasise different themes in the design of their codes'. They find that code issuer does influence code content. By incorporating this dimension into our analysis below, we add not only to the corporate governance literature, but also to Europeanization. If code content differs by issuerinsofar as alignment with the priorities of the Action Plan variesthis may provide important insights into the processes and channels by which Europeanization occurs. Hermes et al. (2006) are the only other researchers of whom we are aware who look specifically at convergence toward the criteria laid down in the Action Plan. They find that domestic codes are 'not in full accordance' (p. 280) with it. Two limitations of their study, however, are that it was undertaken quite soon after the publication of the Action Plan, and they omit some countries that we are now able to include. Ivaschenko and Koeva Brooks (2008, p. 1) find evidence of convergence, but they focus narrowly on 'the dynamics of the voting premiums, a measure of the private benefits of control in a corporation'. Mavrommati and Papathanassiou (2006) are, to the best of our knowledge, the only authors who have discussed EU corporate governance policy with explicit reference to the OMC. They do not, however, analyse formally whether this has delivered policy convergence. In this article, we utilize a wider range of statistical tests, and a longer and wider dataset, than this limited existing literature on convergence in EU corporate governance policies.

Identifying Mutually Relevant Gaps in These Literatures
By focusing on convergence in EU Member States' corporate governance codes toward priorities set at the EU level, we are offering important insights into both the Europeanization and corporate governance literatures. First, with regard to Europeanization, we explicitly link discussion and policy responses at the EU and national levelswhich is at the heart of the conceptand we do so with a novel focus on soft law. Europeanization has been hypothesized as being more likely with hard law (Olsen, 2002, p. 933), so ours is an important addition to this literature. As such, we are looking for evidence that EU policy debate 'steers' Member States' actions (Stoker, 2000), within the multi-level governance setting of the EU (see, inter alia, Stephenson, 2013).
Second, we offer a related insight into the corporate governance literature. Our focus on convergence toward a pre-determined set of corporate governance priorities contrasts with much of the corporate governance literature, which is dominated by convergence driven by market forces and (unco-ordinated) diffusion (see, inter alia, Enriques and Volpin, 2007;Ivaschenko andKoeva Brooks, 2008, Deakin, 2009;Yoshikawa and Rasheed, 2009), rather than by policy.
Third, we offer further illumination on how soft law and the OMC work in different policy areas. With the Action Plan, the Member States and the Commission co-ordinated their efforts to develop the priorities (Commission of the European Communities, 2003, p. 4). This contrasts with, for example, the EU Lisbon Strategy, where national reform plans were drawn up and submitted to the Commission, or youth policy, where Member States determine objectives 'without the need for any European-level coordination of national action plans'. 3 These examples illustrate that policy design, implementation and monitoring under soft law can vary considerably by policy area: context is critical in any analysis of soft law processes.
It is also important to clarify what we do not analyse. We focus specifically on policy debate and policy-making at the EU and Member State levelswe do not look at firm-level implementation, monitoring and enforcement (see, inter alia, Wójcik 2006;Verga Matos and Faustino, 2012). Wymeersch (2013, p. 68)  jurisdictions'. With monitoring, corporate governance codes aim to improve existing corporate governance practices and are mainly applicable to listed firms. It is thus in the capital market that monitoring and enforcement of compliance with code provisions occurs (Lütz et al., 2011).

II. Data Description
Our primary dataset consists of 95 corporate governance codes and principles issued in EU Member States from the beginning of 1992 to the middle of 2010. These allow us to test whether European corporate governance policies included in the Action Plan are reflected in national corporate governance codes across time and across countries. The main source was the European Corporate Governance Institute (ECGI, 2010) which maintains a database of the full texts of international corporate governance codes and guidelines (a list of the 95 national documents is available from the authors on request). In addition to corporate governance documents classed officially as 'codes', we included those listing corporate governance principles and recommendations, because these include best practices of corporate governance, information relevant to an exploration of soft law.
We exclude hard law instruments, but also draft codes and codes with a very narrow focus (e.g. the role of pension funds as shareholders in publicly traded companies; non-executive directors) or a specific type of organization (e.g. NGOs (non-governmental organizations), pension funds, government departments). Such codes are designed to cover limited corporate governance priorities, so their inclusion could bias our results. Ireland is not included in our analysis because the documents published fall into this category (IAIM, 1999;CGAI, 2008;CGAI, 2010; see also Hermes et al., 2006).
Of the 95 codes included, 36 were published before 2003 and the Action Plan, of which 20 were the issuing countries' first codes. The majority of codes were therefore issued after the Action Plan was published. We also see that as first code issuance falls away, the number of revisions rises and remains high for several years (once started, countries continue to engage with corporate governance questions). See Table 1 and Online Appendix Figure 1.
The first code was the UK Cadbury Report (Cadbury, 1992). There is a jump in code issuance in the late 1990s. This is driven in 1998 mainly by early adopters revising existing codes, and in 1999 more by other countries establishing first codes. The year of the Action Plan, 2003, sees a new peak in the number of revisions issued, with only Finland and Lithuania issuing first codes. After 2007 all codes issued are revisions, the number peaking in 2008. This may indicate an initial response from Member States to one perceived cause of the economic crisis. Following Cicon et al. (2012), we also distinguish between the type of code issuer. Of the 95 codes, 30 had a stock exchange issuer, 24 a government issuer (national legislature or government commission), 24 an industry or trade association issuer and 17 a composite issuer, with representatives from at least two individual issue types. 4

III. Research Design
We use content analysis, combined with univariate and multivariate quantitative techniques (see below). Content analysis is particularly useful for our purposes, given that we have ready-made textual codes in the 26 corporate governance priorities in the Action Plan. This allows us to create our indicators in a simple but systematic way. A quantitative approach contributes to the empirical Europeanization literature, which is predominantly qualitative. By examining a large sample, with longitudinal variables over time and across Member States, we minimize the chance of case-selection bias (Moumoutzis and Zartaloudis, 2015;Töller, 2012). Consequently, this improves our ability to measure the scope of, and identify major trends in, Europeanization (Töller, 2010;. We use content analysis to construct a 'Coverage Index' variable that measures national policy responses to the Action Plan. Specifically, it measures the degree of alignment of national corporate governance codes with the 26 priorities of the Action Plan. A binary score has been applied -1 if the priority is present, 0 otherwise. We do not adopt a range of zero to one, to avoid subjective evaluations (Giner, 1995). The values obtained for each corporate governance principle are then aggregated and converted to a percentage. Therefore, the Coverage Index variable shows the percentage of priorities from the Action Plan contained in each national document. In this way, we seek to determine the extent to which the fourth principle in the Action Plan, 'co-ordinating corporate governance efforts of Member States', has been achieved.

IV. Quantitative Analysis of Member States' Codes
A two-stage quantitative methodology is used to analyse variations in the Coverage Index across time and countries. First, univariate analysis is carried out to explore the co-ordination of national corporate governance codes. We then compare the Coverage Index for the codes published before and after the Action Plan, as well as those issued by western European 5 and CEE (central and eastern European) 6 countries. Multivariate analysis is then applied to control for multiple factors that may affect the degree of national policy alignment to the Action Plan's priorities.

Univariate Analysis
The Action Plan does not seek agreement on a single EU corporate governance code, but, rather, to work toward 'a certain co-ordination of corporate governance codes in the EU' (Commission of the European Communities, 2003, pp. 11-12; see also Zattoni and Cuomo, 2008, p. 2). We analyse this utilizing the measures of convergence used by Starke et al. (2008). Table 1 presents Coverage Index statistics by country and by year. Most commonly, the Index in a country is stable or increasing over time. The former does not mean that a code does not introduce new information but, rather, that there is no change with respect to the Action Plan policy priorities. In some cases, the Index falls or fluctuates. One explanation, relevant to Finland, France and Greece, is that later codes were issued by industry or trade organizations, with narrower corporate governance interests than Notes: g: code issuer is a government institution (national legislature or government commission); t: code issuer is an industry or trade association; s: code issuer is a stock exchange; m: the code is issued by a group that contains representatives from at least two of the other groups. government. Another factor is that in Luxembourg, Spain and Sweden, notably, some aspects of corporate governance are now addressed via legislation. Sigma-convergence (Barro and Sala-i-Martin, 1995) represents a reduction in the dispersion across countries at a particular point in time. This is displayed in Table 1. There is a significant rise in the mean Index over 1996 to 2004, and again from 2007 to 2008, for reasons outlined earlier. The trend in the standard deviation is more or less flat, while the range falls dramatically in 2002 and, from 2005, exhibits smaller fluctuations. From 2002 the coefficient of variation exhibits volatility, but at a much lower level, mirroring the rise in the mean. There is thus modest convergence across national codes.
Beta-convergence (Barro and Sala-i-Martin, 1995) occurs when the growth rate is higher in countries with lower starting levels. The negative sign of the catch-up measure (À0.4583, in Table 1) shows that national codes with a lower initial Coverage Index have seen the Index grow more rapidly than national codes with a higher initial Index. That is, countries whose codes initially did not align much with the EU priorities have responded in subsequent revisions, resulting in the convergence between countries' codes toward the Action Plan's priorities that is seen over the period to 2010.
To explore this last finding further, we distinguish between countries whose codes were published before and after the Action Plan (see Online Appendix Table 2 for detailed statistics). Codes published before the Action Plan have a lower Index (less convergence on the Action Plan's priorities) than codes published after. In the former group there is also less convergence between countries than in the latter group, shown by the higher coefficient of variation. The standard deviations for the two groups of codes, however, are not statistically significantly different from each other. It is important to note in interpreting these results that the codes published before the Action Plan include the majority of countries' first codes. On the other hand, Bulgaria, Estonia, Latvia, Luxembourg and Slovenia published their first code after the Action Plan. For three of these countries, the Index score is higher than the mean for all countries' first codes, with Estonia's only marginally below.
This indicates that these late adopters introduced first codes that were much more in line with the Action Plan than the first codes of the early adopters of corporate governance principles. Subsequently, early adopters adapted their policies iteratively toward the Action Plan. This also indicates that the Action Plan did not simply formalize what already existed in Member States' codes, but introduced a wider range of measures, toward which countries had to work.
These findings may possibly be explained by differences between old and new Member States (Hermes et al., 2006). Details are shown in Table 1 (see also Online Appendix Figure 2). These show the date of publication of each country's first code, distinguishing between western and CEE countries. A strong trend is identifiable, other than for the UK, Bulgaria and possibly Estonia. This indicates that, on average, first codes published later have higher Indexes. This trend is particularly strong for the western countries.
When looked at in isolation, there is a much less clear pattern evident among the CEE countries. That said, six of the countries that joined the EU in 2004 and produced their first code before or in 2004 have a much higher Index in those first codes than was the case with the western countries. The corollary of this is that we see larger increases in the Index for the codes of western European countries after the Action Plan. For first CEE codes published even before the Action Plan, learning may be taking place vis-à-vis existing western countries' codes. For example, Romania's first code was published in 2000, before the Action Plan but after 13 codes, including 7 first codes, had been published in western countries.

Multivariate Analysis
The foregoing reveals several possible reasons that could explain the evolution of the Coverage Index between codes and over time. To analyse how these variables might impact on the evolution of the Index we undertake further, multivariate, analysis estimating Equation (1), taking into account the possible relationship between codes issued in the same country. To relax the assumption that the observations are independent, we cluster the data by countries.
CoverageIndex ¼ β 0 þ β 1 ÁWestern þ β 2 ÁAfter_AP þ β 3 ÁAfter_AP*Westernþ β 4 ÁExchangeIssuer þ β 5 IndustryIssuer þ β 6 ÁCompositeIssuer þ β 7 ÁOrderþ β 8 ÁNrevision þ β 9 ÁCode þ ε In Equation 1, the dependent variable is the Coverage Index. As the literature shows, relevant factors to include as independent variables are: whether the country is western or CEE ('Western' dichotomous variable); whether the code was published before or after the Action Plan ('After AP' dichotomous variable); and whether the Action Plan had the same influence on the codes from western or CEE countries ('After AP*Western' interaction variable). In addition, we include as control variables other factors that could affect national codes. We include variables relating to the type of issuer, expressed with respect to Government Issuer ('ExchangeIssuer', 'IndustryIssuer' and 'CompositeIssuer'); the number of codes issued in a specific country ('Nrevision' variable); and the order of code issued in the EU context ('Order' variable). Because the codes in the study include documents, codes and principles, we include the variable 'Code'. Table 2 presents our results. Consistent with our earlier findings, the Index is larger for codes issued by CEE than by western countries. There are not, however, significant (1) differences between CEE Indexes published before or after the Action Plan. These results together suggest that when CEE countries were still candidates, their codes reflected what was already being discussed in the OECD and individual western EU countries' codes. We find that, after the publication of the Action Plan, codes in the western countries become more aligned with the Action Plan's priorities than those issued in CEE countries.
In addition, Table 2 shows that issuer type is a significant variable, affecting the level of the Coverage Index. Codes issued by a government commission align more closely to the priorities of the Action Plan than those issued by an industry association or national stock exchange. The robustness of the results was investigated using models b and c in Table 2. Model b considers EU legal systems, as these may reflect different corporate governance models and practices. According to Cicon et al. (2012), CEE countries correspond to the Former Socialist legal system (omitted dummy variable), with western countries divided into Common Law, German Civil, French Civil and Scandinavian Civil. Our results reveal that the codes issued under Former Socialist legal systems cover more of the Action Plan's priorities than the codes issued under other legal systems, with the codes issued under German and Scandinavian Civil legal systems the least aligned with the EU initially. Model c repeats this analysis, excluding those codes that were issued when CEE countries were still candidates. Both models show that after the publication of the Action Plan there is a greater alignment with the Action Plan's priorities for those codes issued in the western countries, supporting our previous findings with similar levels of significance and the same signs on variables.

Discussion and Conclusions
Europeanization describes the impacts of EU membership on both the processes by which EU policy-making manifests itself at the national level and the policy outcomes of those processes. Most of the Europeanization literature focuses on hard law processes and policies, yet the EU is expanding its use of soft law processes. We have sought to explore empirically and quantitatively whether soft law processes also influence national policy-making. We have utilized a unique dataset on national corporate governance policy measures introduced at the national level, derived from a content analysis of 95 national documents issued between 1992 and (mid-) 2010. The textual codes were the 26 corporate governance policy priorities laid down in 2003 via the 'Action Plan' (Commission of the European Communities, 2003).
Earlier, we identified three specific research questions, to which we are now able to offer initial answers. We find evidence that the 26 corporate governance priorities laid down in the Action Plan did influence national corporate governance policy-making, with Member States' corporate governance codes converging on the priorities of the Action Plan over time. We also find that different corporate governance code-issuing bodies aligned priorities with the 26 priorities to different degrees, and that there was cross-country variation in the extent to which Member States aligned their corporate governance codes to the 26 priorities.
Countries whose early corporate governance codes had less similarity with the (subsequent) Action Plan converged toward it after 2003. Most of the countries whose early codes showed little similarity with what was to come were 'western' Member States.
The fact that we detect convergence with the priorities of the Action Plan after 2003 suggests that the priorities included in the Action Plan were not merely a reflection of pre-existing western EU countries' priorities. We thus argue that this is suggestive of Europeanization rather than convergence per se, with countries' policies taking a statistically significant step toward the priorities laid down in the Action Plan.
Another finding reinforces this conclusion. For EU countries in total, we find no statistically significant difference between national corporate governance codes before and after 2003. When analysed separately, however, western countries exhibit significant catch-up, but not the CEE 'new' Member States. CEE corporate governance policies, even those pre-dating the Action Plan and the accession of most countries in 2004, had closer alignment with the Action Plan.
This difference between older and newer Member States supports the expectation of such differences expressed by Dyson (2007, p. 292). One possible interpretation is that as these countries went through economic and political transition and prepared for EU membership, they looked west for inspiration on soft infrastructure policy measures, including corporate governance. Policy design was, if not strictly tabula rasa, relatively free of institutional path dependency pressures. Several western countries already had corporate governance policies in place (if not, at that stage, well aligned with what the Action Plan would become). This offers interesting insights into one channel of Europeanization. Even before accession, applicant countries with different economic and political legacies looked to 'Europeanize' their policies.
We also find evidence that government-issued codes are more closely aligned with the Action Plan priorities than codes from other issuers. This may reflect that governments have wider policy interests than stock exchanges and industry bodies. Additionally, governments are directly involved in negotiating soft law policies, which may influence their willingness to engage with soft law outcomes (Olsen, 2002, p. 933). This may also explain another feature of our results. After 2003, western countries' codes continue to converge toward the priorities in the Action Plan, eventually becoming more convergent than the codes of CEE countries. The latter, however, as non-members at the time, would not have been involved directly in creating the Action Plan.
These observations lead us to an extremely important general conclusion, which also points to possible directions for future research: with soft law, context matters. In contrast to hard law, where the 'issuer' is de facto the national legislator, our research has shown that, for soft law corporate governance policies, the issuer has a significant impact on the measures laid down. An important task for future research is, therefore, to test the generalizability of our findings across different policy areas. This leads to an important observation regarding Europeanization. EU Directives will vary in their detailed implementation across Member States. With corporate governance, however, we have found that differentiated implementation can also occur within Member States, dependent on code issuer. This is in addition to the differences we have identified by country or country-group, distinguished by 'old'/'new' Member States and by domestic legal system. Future research can also develop several issues not addressed here. We focus on policy-making. One direction would be to investigate later stages of the policy process, notably firm-level implementation, monitoring and enforcement. A further enhancement would be to analyse in detail what is happening at the level of individual corporate governance priorities. Such detailed investigation could explore whether or not the convergence we identify involves clustering around certain types of corporate governance policy priority in different countries or country-groups. Finally, and related to monitoring, enforcement and individual policy priorities, is the question of Member States following up soft law with hard law policy responses. We hope this initial analysis will encourage more researchers to explore these avenues, as we try to deepen our understanding of Europeanization through soft law, different channels of Europeanization, and how Member States are developing their corporate governance regimes.