Europeanization, EU Conditionality and Governance Quality: Empirical Evidence on Central and Eastern European Countries

A large body of work within the Europeanization and conditionality literature has examined EU’s effect on governance quality in Central and Eastern European countries (CEECs). This article aims to contribute to existing work by addressing two questions. First, was EU’s accession conditionality conducive to better governance in CEECs or was it designed to pick up already committed reformers? Secondly, during which period was the EU more influential on governance quality in CEECs: accession or membership? To address these questions, we propose a simple model of moral hazard and adverse selection; and draw on survey-based governance quality data to test its predictions. We report that the EU has offered candidate status to countries that already had better governance quality relative to their peers before 1997. Nonetheless, EU conditionality also had some positive effects on CEEC governance during the accession period, but this effect tended to disappear during the membership period. On the basis of these findings, we argue that EU’s impact on governance quality in CEECs has to be examined in conjunction with path dependence and therefore the perceived risk of ‘backsliding’ requires more detailed analysis.

The Europeanization literature examines the extent to which polities and/or public policy contents of EU member states or candidate countries converge towards a 'European norm'usually proxied by existing European Union (EU) rules, legislation and institutional frameworks. Using mainly a qualitative approach, this work has provided significant insights into the extent to which Europeanization has affected governance quality in Central and Eastern European Countries (CEECs).
On the other hand, there is also a well-developed body of work on the political economy of institutional reforms in the context of World Bank or International Monetary Fund (IMF) conditionality. This literature suggests that adherence to rules provides a credibility bonus that reduces the cost of reforms. Therefore, policymakers that have a commitment problem would be better off tying their hands with external rules. The implication for CEECs is that policymakers in these countries can reduce the cost of governance reform by tying their hands with EU conditionality.
Both strands have provided a rich account of the actors and causal mechanisms that must be taken into account when investigating the EU's impact on governance quality in CEECs since mid-1990s. Particularly, the Europeanization strand has also produced rich evidence on the distribution of EU's impact across countries, public policy contents and areas of public administration. However, the qualitative nature of the work have limited the scope for deriving aggregate, comparable and reproducible conclusions about EU's impact on governance quality across countries and/or over time.
This article aims to contribute to the Europeanization and conditionality debate in three ways. First, we demonstrate why the EU might grant accession country status or membership to eligible countries that already have a good track record with respect to reform and governance quality. To the extent that this is the case, it is necessary to disentangle the true effect of EU conditionality from the reform momentum that is due to path dependence. This is an issue that is not addressed either by the Europeanization literature or by the wider literature on external conditionality. Secondly, we argue that CEECs may be faced with a moral hazard problem that mirrors the EU's adverse selection problem: the reform effort in candidate countries may continue during the accession period when there is a probability of exclusion from membership, but it may diminish or disappear after the country becomes an EU member. Finally, reform efforts in CEECs can be expected to be influenced not only by the domestic cost of reforms, but also by strategic reactions of CEEC policymakers to changes in the rules of engagement with the EU. Put differently, it is necessary to disentangle the cost-related effects from perceptions-related effects caused by moral hazard.
To work out these arguments, we begin by providing an overview of the related literature and developing the rationale for explicit modeling of the EU-CEEC relations. Then, we present a simple model of utility maximization on the basis of 2 choice variables: (i) the level of reforms (R) by the candidate country; and (ii) the membership prospect (M) offered by the EU. The model suggests that the EU would tend to offer accession country status to countries that are already good reformers relative to the eligible group and that the accession country's reform effort would weaken after EU membership is secured. To test the predictions of the model, we draw on survey-based governance quality data compiled by two sources: the World Bank and the International Country Risk Guide. We report that the empirical findings are in line with the analytical predictions of the model. We conclude by summarizing the main findings and elaborating on their implications for the debate on the pre-and postmembership reform efforts in the CEECs.

Europeanization, Conditionality and Governance Quality in CEECs
The Europeanization literature analyzes the extent to which polities and/or public policy contents of EU member states or candidate countries converge towards the 'European norm' as a result of EU conditionality or rule transfers. (Bulmer and Radaelli 2004;Schimmelfennig and Sedelmeier 2004;Grabbe 2001 andand Goetz 2005). The work that focuses on CEECs reports that Europeanization was visible first and foremost in "linkage institutions" that link CEEC executives and EU authorities; and in "support institutions" that support the multiple platforms of interaction between the two sides (Ágh 2003;Fink Hafner and Lajh 2003;Nakrosis 2003;Zubek 2002;and Lippert et al. 2001). These findings were supported by Goetz (2005), who reports that change was visible at the level of "governmental and administrative actors, structures and procedures that were in charge of the accession negotiations, the transposition of the acquis, and the management of pre-accession funds." Europeanization was also reported in the reform of the civil service instigated by the EU through formulation of "baselines for civil service development" and provision of "technical and financial assistance [that] was made available through the SIGMA program and through PHARE projects" (Goetz 2005. See also Meyer-Sahling 2004;Scherpereel 2003).
A sub-set in this literature examines the mechanisms through which Europeanization has affected governance quality in CEECs. For example, Grabbe (2001) reports that the accession conditionality has influenced public policymaking processes in CEECs through a number of mechanisms, including: gate keeping (the threat of suspending accession negotiations); benchmarking and monitoring; provision of institutional templates; and financial assistance. On the other hand, Schimmelfennig and Sedelmeier (2004: 663) report that the effectiveness of rule transfer from the EU to the CEECs has been dependent on the credibility of EU conditionality and the domestic cost of rule adoption. Furthermore, they point out that the domestic cost of rule adoption has limited the effectiveness of EU conditionality in the context of democratic reforms more than it has in the context of acquis adoption.
Among this body of work, Goetz (2001; is the most pessimistic: he asserts that Europeanization is bound to remain limited/shallow because of weak 'uploading capacity' of CEECs, inadequate time for learning and socialization, and the restriction of engagement with EU institutions to mainly political, administrative and economic elites of the CEECs. However, in a more recent review of the literature, Goetz (2005) observes that Europeanization has affected polity, politics and public policies in CEECs; although these effects have remained shallower than it is the case in previous member states.
The Europeanization literature has made significant contributions to our understanding of reforms and governance quality in CEECs under EU conditionality. However, its qualitative nature has limited the extent to which its findings can be aggregated and/or used for comparative purposes. More importantly, the Europeanization literature rightly considers the effect of the domestic cost of reforms on reform outcomes, but does not problematize four issues that arise when the EU and CEEC policymakers interact strategically. First, what is the relationship between the pre-and post-1997 reform effort in CEECs? Secondly, what is the relationship between EU's offer of candidate country status in 1997 and the pre-1997 reform efforts of the CEECs? Third, how do policymakers in CEECs react to changes in the rules of the game that govern their relations with the EU? Finally, is it not necessary to distinguish between the domestic cost of reform and the moral hazard problem as potential determinants of reform effort in CEECs?
Some political economy work that adopts a strategic approach tends to address some of these issues explicitly. For example, Bronk (2002) examine EU's impact on CEEC governance quality from the perspective of credible commitments. He draws attention to a "credibility bonus" that policymakers in CEECs can enjoy by tying their hands under EU conditionality. In this perspective, EU's impact on reform efforts is not a result of simple rule transfer or compliance per se. Rather, it originates from EU's role as a commitment device that reduces the scope for discretionary (hence uncertain/reversible) policy choices by the political and administrative elite. This, in turn, increases the credibility of the commitment to reforms by providing 'credibility benefits' such as lower risk premiums on public and private borrowing from international markets and substantial inflows of foreign direct investment (FDI). Put differently, EU conditionality can instigate a virtuous circle, whereby tighter EU conditionality leads to higher levels of credibility and credibility benefits that, in turn, leads to higher levels of reform and governance quality.
This approach draws on a well-established research agenda in the political economy of macroeconomic policy design, dating back to Kydland and Prescott's (1977) seminal contribution on rules versus discretion in monetary policy. In this perspective, subscription to rules ties the policymaker's hands and reduces the scope for optimal policy adjustment in the face of shocks. However, rules still remain superior to discretion as the latter can be abused to maximize political gains. Given this scope for political opportunism and in the absence of a credible domestic commitment mechanism, external conditionality can emerge as a substitute commitment mechanism that would tie the policymaker's hands and produce efficient outcomes by preventing policy reversals. 2 Nevertheless, there is evidence suggesting that conditionality in IMF programs or in development aid has been less effective than what this line of research would imply. For example, the rate of non-compliance with aid conditionality from 1973-97 is reported by Moussa and Savastano (1999) as 54.5%. Interestingly, non-compliance has increased as the IMF has become more experienced in designing and negotiating conditionality. Hence, noncompliance increased from about 50% in 1970s to about 72% in the 1990s. 3 Other work examines the mismatch between IMF structural adjustment credits and the policy environment in recipient countries. It demonstrates that IMF credits tended to favor policy environments that are too weak, but tended to diminish in policy environments that are sufficiently good for funds to be effective (Collier and Gunning 1999). However, Dollar and Svensson (2000) report opposite findings and conclude that the primary role of the international financial institutions has been to pick up winners rather than increase recipient's commitment to reforms.
In the light of these findings, two questions arise with respect to EU's impact on governance quality in CEECs. First, does the EU pick up winnersthat is, does it grant accession status to already committed reformers? Or does it actually function as a commitment device that increases the pace and quality of governance reforms in CEECs? Secondly, when is it more likely for EU conditionality to function as a commitment device that would underpin the reform effort: during accession or during membership? To address these questions, we first propose a utility-maximization model that demonstrates the adverse selection and moral hazard problems faced by the EU and CEECs, respectively. Then, we test the predictions of the model, using survey-based governance data compiled by the World Bank and the International Country Risk Guide (ICRG).

Governance Reforms under Moral Hazard and Adverse Selection
The relationship between the EU and an accession country can be modeled as a utilitymaximization game with 2 choice variables for the actors: (1) the level of reforms (R) to be chosen by the accession country; and (2) the membership prospect (M) that the EU offers to the latter. Policymakers in CEECs and the EU will choose the level of reforms and membership prospect that would maximize their utilities in equations (1) and (2), respectively.
Finally, the level of reforms chosen by a CEEC and the membership prospect offered by the EU will determine the probability of membership (π). As specified in equation 3, higher levels of reforms and commitment by CEECs and EU respectively, lead to higher probability of membershipthat is, the partial derivatives of the membership probability function (π) with respect to (R) and (M) are positive. 4 Solution of the model is given in the Appendix. The variables are defined as follows: U a , U e = Utility functions of the accession country and the EU, respectively M = Membership prospect offered by the EU to the accession country R = Accession country's reform effort to comply with EU conditionality Z = Worth of membership for the accession country πZ = Expected value of membership, given the probability of membership (π) sZ = EU support received during the accession period as a fraction of Z (1-π)sZ = Expected value of EU support when membership does not materialize P = Benefits derived by the EU as a result of CEEC convergence towards EU norms/standards C a (R, R 0 ) = Cost of reforms for policymakers in CEECs C e (M, M 0 ) = Cost of membership for the EU π(M,R) = Probability of membership as a function of EU's membership commitment and candidate country's reform effort π M > 0; π R > 0 = Partial derivatives indicating that the probability of membership increases as the membership commitment and reform effort increase.
The cost functions [C a (R, R 0 ) and C e (M, M 0 )] are defined as quadratic functions that depend on the difference between the current and past levels of: (a) reform in the accession country (R-R 0 ); and (b) level of EU commitment to membership (M-M 0 ). Stated formally: In equation 4, the cost of reform for policymakers in accession countries is assumed to increase as the level of required reform (R) increases. This is due to political risk and/or adjustment costs associated with reforms, including resistance by veto points, increased Euroskepticizm , increased costs in the affected economic, business and public policy domains. However, the cost of reforms is assumed to be lower, the higher is the stock of past reforms (R 0 ). Higher past reform levels dampen the cost of current reforms either by reducing the level of reform required to satisfy EU conditionality or by enabling policymakers to exploit path-dependence dynamics.
Similarly, in equation (5), we assume that the cost of membership for the EU is higher, the higher is the level of current commitment (M). This may be due to political risks associated with negative perceptions of enlargement by EU nationals or budgetary costs of integrating a candidate country. On the other hand, the cost of membership is assumed to be lower, the higher is the level of past EU commitment to membership. This is because antienlargement actors, having observed EU's strong commitment in the past, would reduce their lobbying activities in the current period as they realize that their opposition is less likely to change the EU's stance. This relative withdrawal of the anti-enlargement lobbying activities, in turn, will increase the scope for the EU to signal a high level of commitment to enlargement.
To find the equilibrium levels of reform (R) and membership prospect (M) that would maximize the utilities of the accession country and the EU, we differentiate the utility functions (equations 1 and 2) and set them equal to zero. In this process, the cost functions in (1) and (2) will be replaced by their equivalents in (4) and (5). We provide the solution in the Appendix and reproduce only the equilibrium levels of reform (R) and (M) below.
From (6), we can see that the level of reform effort (R) by the candidate country is positively related to: (i) the benefit of membership (Z); (ii) reforms' contribution to the probability of membership (π R ); and (iii) the level of past reforms (R 0 ). However, the level of reform in the candidate country is negatively related to: (i) the level of EU support (s) for convergence reforms; and (ii) the marginal cost of reforms (h).
From (7), we can see that the level of EU commitment to membership (M) is positively related to: (i) the benefit of membership (Z); (ii) the level of EU support to the candidate country (s); (iii) the contribution of the membership prospect to the probability of membership (π M ); and (iv) the level of past commitment (M 0 ). However, the membership prospect is negatively related to the marginal cost of membership (d).
Given this setting, three questions arise. First, how would the EU choose the candidate country for membership from a group of eligible countries if it knows that the equilibrium level of reforms to be chosen by CEECs is related negatively to the cost of reform in the candidate country (h) and the level of EU support (s)? Secondly, what would be the reform effort of a candidate country when the latter knows that its reform effort increases the probability of membership with a factor (π R )? Third, what would be the reform effort of the member country when the latter knows that its reform effort has no effect on the probability of membership, which is now equal to 1? Answers to these questions can be best developed by pointing out the moral hazard and adverse selection problems faced by the parties.
From equation (6), the accession country is faced with two types of moral hazardone during the accession period and one during the membership period. During the accession period, the candidate country must deliver reforms that contribute to the probability of membership. Otherwise, the EU could withdraw the membership offer or delay its realization. This is the level of reform that must satisfy EU conditionalitythe so-called Copenhagen criteria for accession. However, this reform effort will be dampened by two factors: the level of EU support as a fraction of the membership benefits (s); and the incremental cost of reforms (h).
The combination of these factors represents the scope for moral hazard in CEECs. In other words, they represent the incentives for CEEC policymakers either to reduce the reform effort or to deliver less reform than what was agreed. This type of behavior will occur when the EU provides generous support towards the cost of convergence reforms (as s increases) or when the marginal cost coefficient (h) increases. Nonetheless, equation (6) also indicates that the candidate country's reform effort will remain positive as long as: (a) the value of EU aid/support remains a fraction of the membership benefits -i.e., as long as s < 1 in the numerator; and (b) the trade-off between R0 and this positive value will be boosted by the path-dependence dynamics of the past reform effortsi.e. by the contribution of R 0 to current reform efforts. Therefore, during the accession period, that the candidate country will maintain a positive level of reform effort that is related to its past track record, the EU's incentive package, and the perceived cost of reforms. However, the level of reform delivered during membership will fall below the level of accession reforms as membership probability is now 1 and the reform effort does not affect the chance or membership -i.e. π R = 0. Any reforms that may or may not be undertaken in CEECs will be a function of path dependence (R0) or other factors not included in the modelfor example rule/policy transfer through adoption of EU legislation, elite socialization, arbitrage through political and/or economic markets, etc. In other words, the linkage between EU conditionality and reform weakens even if it does not disappear altogether. Therefore, we can expect reform performance to slow down or stall after CEECs join the EU as full members.
On the other hand, the EU is faced with an adverse selection problem, which is due to its awareness of the moral hazard problem faced by the accession country. To minimize its loss when the moral hazard problems kicks in, the EU will be inclined to offer candidate country status and membership prospect mainly to countries with good track record for reforms. This is for two reasons. First, countries with good track record are less likely to renege on their reform commitments after membership or as EU assistance (s) becomes more generous. Secondly, even if they renege, the gap between EU norms/rules and candidate country's governance quality will be narrower compared to other potential candidates without a good track record. The narrower gap, in turn, enables the EU to minimize the risk of noncompliance by new members or that of gridlock in EU decision making due to divergent norms and values.
In the next section, we examine the extent to which these analytical predictions are consistent with the evidence on governance quality in CEECs and a control group of eligible countries in south-eastern Europe.

Empirical analysis: model specification, data quality and results
We specify two models to test the analytical conclusions derived above. The first is a logistic regression model that relates the probability of accession status to the level of past reform effort by a candidate country. The second is a random-effect panel data regression for estimating the effect of accession and membership periods on governance quality of 10 CEECs that had been included in the enlargement agenda of 1997. In addition to the governance data, we use data for control variables such as per-capita income, openness to trade, size of the stock of market (measured as a ratio of the market value of listed companies to GDP), and ratio of urban population to total population. Data for control variables is from World Bank Development Indicators (WDI) database; whereas governance data is from World Bank Governance Indicators (WGI) and International Country Risk Guide (ICRG) databases. Definitions of governance scores are in Table A1 and Table A2 in the appendix. Inclusion of per-capita GDP, openness to trade, size of the stock market, and size of the urban population as control variables with potential effect on governance quality is informed by the relevant literature (see, for example, Alonso and Garcimartin 2010;Al-Marhubi 2004;and Kandil 2003).
The governance measure for each country consists of the sum of scores for six governance dimensions. In the WGI data, the total score is the sum of scores for: voice and accountability; rule of law; regulatory quality; political stability; government effectiveness; and control of corruption. ICRG scores, on the other hand, measures governance quality along the following dimensions: democratic accountability; law and order; government stability; bureaucratic quality; ethnic tensions; and control of corruption.
Instead of governance quality scores as magnitudes, we use a relative measure that is equal to the difference between the country score and the group average. The group consists of 17 countries -11 CEECs that had been granted accession country status in 1997 (Bulgaria, Cyprus, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia, and Slovenia) and 6 south-eastern European countries (Albania, Bosnia and Herzegovina, Croatia, Macedonia, Serbia, and Turkey) that were not included in the enlargement agenda of 1997 but are considered eligible due to their European geography. Table A1 in the appendix summarizes the governance quality of CEECs relative to 17 countries eligible for accession status due to their locations in Europe. The evidence in Table A1 indicates the following: (i) the 10 CEECs that were granted accession status in 1997 had higher governance quality scores than the group in 1997; (ii) the 10 CEECs increased their leads over the group average from 1997 to the accession year (2004 for the majority and 2007 for Bulgaria and Romania); and (c) relatively higher governance quality in the 10 CEECs tended to stabilize or disappear after accession.
This summary evidence is line with the main predictions of the moral hazard and adverse selection model presented in section 3. However, the descriptive nature of this evidence does not allow causal inference. To do this, we first estimate a logistic regression model to establish the partial effect of past governance quality in CEECs on EU's decision to grant or withhold accession status in 1997. Then, we estimate a random-effect regression to establish the effect of EU conditionality on governance quality in CEECs during two periods: (i) the accession period from 1997-2004 (from 1997-2007 for Bulgaria and Romanian); and (ii) the membership period from 2004 (from 2007 for Bulgaria and Romania) to the last year for which data exists (2009). All estimations are based on panel data for the specified period.
We address two issues before reporting and discussing the estimations results. The first issue concerns the subjective nature of the governance quality scores and whether such scores can be used for empirical estimation. The second issue is the large confidence interval (i.e., the lack of precision) associated with governance quality scores compiled from different sourcesas the WGI scores are.
Concerns about the first issue are well-documented in the literature (see, for example, Kurtz and Schrank 2007;Knack 2006;Arndt and Oman 2006). These concerns relate to the "halo effect" discussed by Kurtz and Schrank (2007) and Kaufmann et al (2007) and the endogeneity problem discussed in the context of the relationship between governance quality and economic performance in general (Knack and Keffer 1997;Acemoglu et al 2001;Rodrik et al 2004, etc.). However, this endogeneity problem can be and has been addressed in the empirical literature. For example, Acemoglu et al (2001) have introduced instrumental variables (i.e., settler mortality rates in early colonial period) that are correlated with institutional quality but are not dependent on current economic performance. Knack and Keefer (1997), on the other hand, used a measure of ethnic cleavage and the number of law students as instrumental variables. Rodrik et al (2004) have used lagged values and conducted Granger causality tests to demonstrate that institutional quality measures determine economic performance rather than the other way round. Finally, Kaufmann et al (2007) demonstrate that economic performance (e.g., growth) is likely to impact on governance quality only in the long run.
The endogeneity problem does not affect the estimations in this article because we do not estimate to the effect of governance quality on economic performance (e.g. growth). On the contrary, governance quality in this paper is the dependent variable and we estimate the partial effects of its determinantsincluding past governance quality and other factors such as EU conditionality, per-capita income, openness to trade and financial development. However, the halo effect may contaminate the estimations in this article to the extent that contemporaneous values of per-capita income (or other economic variables) and governance quality are highly correlated. To minimize this risk, we have used two-year-lagged values of all economic variables.
The second issue concerns the statistical significance of the quality scores and whether they can be used for comparison across countries or over time. For example, Knack (2006) as well as Arndt and Oman (2006) warn that cross-country differences in scores or differences over time may not be statistically significant due to the margin of error associated with individual scores. This problem is especially acute when the reported governance scores consist of aggregated scores from different sources. We acknowledge that the WGI data we use in this article is particularly prone to this problem. We cannot eliminate this risk, but we control for its implications by reporting estimates using both WGI and ICRG data. The extent of convergence/divergence between the estimation results from the two datasets will enable us to assert or qualify the relevance of our findings.
Bearing in mind the elaborations above, we estimate the following logit model using unbalanced panel data for the period 1995-2003. The dependent variable is the odds ratio for accession status being granted to 10 CEECs and Cyprus in 1997 and the maintenance of this status until 2003; given that this status was withheld from other peers and given the economic and governance characteristics of each country in the sample. The model can be stated as follows: The variables are defined as follows: P(Accession it ) = 1 if the country was given candidate status in 1997; 0 otherwise. dmgq it-2 = difference between country i's total governance quality score and the group average score, lagged 2 years. dmpcgdp it-2 = difference between country i's per-capita GDP and the average per-capita GDP for the group, lagged 2 years. dmopt it-2 = difference between country i's openness to trade and the average openness to trade for the group, lagged 2 years. ε it = error term.
This model enables us to estimate the impact of past reform performance on the odds ratio for a country being granted a candidate status in 1997 and the maintenance of this status until 2004 for the following countries: Bulgaria, Cyprus, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia, and Slovenia. We estimate this model with a control group of other countries in the south-eastern periphery of the EU, consisting of Albania, Bosnia and Herzegovina, Croatia, Macedonia, Serbia, and Turkey. All independent variables measure the country-specific differences from the group average. Per-capita GDP is measured in constant US dollars with a base period of 2002; and openness to trade is measured as the ratio of the sum of exports and imports to gross domestic product in nominal values. Estimation results are given in Table 1 below. Table 1 Table 1 indicate that the model as a whole is well-specified and statistically significant. This is confirmed by the Wald test results indicating that the null hypothesis of jointly insignificant variables should be rejected at 5% significance for WGI data and at 1% for ICRG data. Looking at the main explanatory variable of interest -relative governance quality in the past (dmgq it-2 )we can see that the coefficient for the odds ratio is greater than 1 and statistically significant in both datasets. Hence, it is possible to infer that past reform performance increases the odd ratio for accession status being granted in 1997 and maintained throughout the period until 2003. This result is in line with the predictions of the analytical model in section 3 and has two important implications for the debate. First, the EU tended to sign accession partnership contracts with CEECs that were above-the-average reformers before the start of the accession period. Secondly, the past the relationship between past performance and the accession status continued to prevail during the accession period too. Hence, it is possible to conclude that the EU had 'picked up winners' (i.e., countries with governance scores) at the beginning of the accession period, but the maintenance of the accession country status was also related to subsequent improvements in terms of reforms and governance quality.

Results in
Table 1 also indicates that the EU tended to sign accession partnership contracts with CEECs that have relatively higher levels of openness to trade. This result is compatible with EU's emphasis on accession country's ability to cope with competitive pressures that would result from trade liberalization within the customs union and the single market. Finally, the coefficients for the relative per-capita income levels are marginally less than 1 and statistically significant only with the ICRG data. Therefore, it is possible to conclude that relative per-capita income levels were not a significant determinant of EU's choice of accession partners.
The analytical model in section 3 above also indicates that the change in governance quality may differ between the accession and membership periods because of the change in the strictness of the EU conditionalityeven though the level of past reform (i.e., governance quality in the past) may have a path-dependence effect. Therefore, we estimate a randomeffect model for the determinants of governance quality in two periods: the accession period from 1997-2004 (2007 for Bulgaria and Romania); and the membership period from -2009-2009. In this model, we regress the governance quality performance of the countries relative to the group average on the following determinants: (i) an EU conditionality dummy that captures the period of accession or membership; (ii) past levels of relative governance quality (to capture path dependence); and (iii) lagged values of economic/demographic variables that have been demonstrated to affect governance quality in the literature (e.g., per-capita income, openness to trade, market capitalization of listed companies, and ratio of urban to total population). The model is specified as follows: Here, dmgq, dmpcgdp and dmopt = as defined above.
EUdummy it = 1 if the country was given a candidate or membership status, depending on the time period specified above; 0 otherwise. dmmrktcap it-2 = relative size of the capital market, measured as market value of listed companies as proportion of GDP, lagged two years dmurbpop it-2 = relative size of urban population, measured as urban population as a proportion of population, lagged two years. u it = error term.
Estimation results for the accession and membership periods are given in Table 2 and  Table 3, respectively. In both tables, F-statistic for joint significance of the variables is less than the critical value of 5% and therefore it is admissible to examine the individual effects of the explanatory variables.
In Table 2, we have two main explanatory variables of interest: EU conditionality dummy for the accession period and relative governance quality in the past. The coefficient of EU dummy is positive in both datasets, but only the coefficient estimated with WGI data is significant. This lends support to the prediction of the analytical model about EU conditionality during accession, but this support needs to be qualified because of the weak precision associated with WGI data. With this caveat in mind, it is possible to conclude that there is some empirical support for the analytical model prediction that EU conditionality during the accession period is likely to have a positive effect on reform effort and consequent governance quality. Table 2

here)
However, EU conditionality during the accession period is not the only factor that affects governance quality in the CEECs: path dependence also has a positive and statistically significant effect in estimations with both datasets. The estimated coefficient indicates that an accession country's governance quality would be 1 or 3.5 units higher than the group average in the current period if its governance quality was 1 unit higher than the group average two years ago. In other words, relatively better performers in the past are likely to outperform the group average mainly because of path dependence. It is not possible to make consistent inference about the impact of economic and demographic determinants of governance quality as either the estimates are not statistically significant or their signs are inconsistent between the two datasets.
Lastly, in Table 3, we present the estimation results for the membership period -i.e., 2004-2009 for 10 CEECs and 2007-2009 for Bulgaria and Romania. The results during membership are consistent across both data sources and comparable with accession period results in Table 2. The estimated coefficients for the membership dummy are positive, but they are not statistically significant. Hence, we cannot infer that EU conditionality under membership has had an effect on governance quality and the reform effort that underpins it. This is in line with the analytical prediction of our model, which implies that reform effort by CEEC policymakers is likely to slow down as the latter is no longer a pre-requisite for increasing the probability of membership (which is now 100%). However, past reform performance (i.e., path dependence) is still a significant determinant of governance quality in CEECs during membershipwith the effect ranging between 1 -1.7. Table 3 here)

(Insert
The findings in Tables 1, 2 and 3 above can help fill in some of the gaps in the existing Europeanization and conditionality literature. First, the finding that accession status has had a positive effect on governance quality is in line with the conditionality literature that predicts a positive effect from EU conditionality to reform and governance quality improvement in CEECs. However, our estimations provide additional evidence that would qualify the assertions of the conditionality literature in two ways. First, EU conditionality does not function within a tabula rasa in CEECs; and the effect of conditionality must be separated from the effect of path dependence. The estimations results indicate that past performance is a strong and significant determinant of current governance quality in both the accession and membership periods. Secondly, the conditionality literature does not explain (or does not very much discuss) the following question: if EU conditionality is conducive to higher levels of reforms and governance quality, why does the EU not offer accession partnerships to all European countries? The analytical model in section 3 and the estimation results above provide some explanation: the EU exercises adverse selection in response to the risk of moral hazard by accession partners. Hence, the EU selects countries with already good track record in terms of reform and governance quality and its conditionality can then help policymakers in these countries resolve some of the commitment problems that may lead to reform reversals.
Secondly, the findings we report above also provide a wider perspective within which the findings of the Europeanization literature should be interpreted and evaluated. The Europeanization literature tends to be based on exogenous choice of reform effort by CEEC policymakers. In this approach, the reform effort of CEEC policymakers is a negative function of the domestic cost of reformsand this is compatible with our analytical model argument. However, the Europeanization literature also makes an assumption that contradicts the predictions of the analytical model above: reforms under EU conditionality are negative function of the strictness of EU conditionality.
We agree with the Europeanization literature that domestic cost of reform has a negative effect on reform effort, but were unable to verify this argument empirically due to lack of 'reform cost' data. However, the evidence we provide in Table 2 and 3 contradicts the Europeanization literature's thesis that strict EU conditionality is associated with weak reform performance. On the contrary, EU conditionality tended to be associated with higher reforms (i.e., better governance quality) when it was strict during the accession period, and the association has become insignificant during membership when conditionality is less asymmetric.
Thirdly, the positive effect of path dependence indicated by our findings can be placed in the context historical institutionalism and institutional economics literature. According to the positive feedback hypothesis in this literature, institutional structures tend to have a high level of persistence due to high costs of switching. High costs of switching, on the other hand, may be due to resistance of stakeholders who benefit from the existing institutional set up and/or high mobilization costs faced by constituents demanding change (Pierson 1996 and2000;Thelen 1999;Levi 1997). The positive feedback process may also be due to reactive sequences, which imply that a chain of temporally ordered events may occur as a reaction to antecedent events (Abbott 1983;Ebbinghaus 2005).

Conclusions and discussion
The impact of EU conditionality on the reform effort and consequent institutional quality in CEECs has attracted significant interest before and after EU enlargement towards these countries. Before enlargement, the focus was mainly on the extent to which EU conditionality or the Europeanization process had contributed to the reform effort and improved governance quality. After accession, the emphasis has been on whether EU membership would enable CEECs to build on pre-accession reforms or whether the disappearance of EU leverage would lead to 'backsliding'. The analysis above suggests that it is difficult to provide uniform and time-invariant answers to these questions. One of the reasons is that the outcome depends both on past performance and rules of engagement with the EU. The other reason is that it is highly difficult to devise a quick institutional fix for politics when it comes to long-term commitments (Posen 1993). Despite these caveats, the findings above can still support a number of conclusions that contribute to the ongoing debate.
One conclusion is that EU conditionality (i.e., the combination of incentives and constraints devised in accession partnerships and membership pacts) matters; but this is only one of the factors to be taken into account. In addition, the level of pre-existing commitment to reforms and the level of path dependence are also important factors. EU conditionality or the Europeanization process in accession countries can help policymakers maintain their commitments to reform in the face of frequent shocks and in the absence of a strong domestic commitment mechanism that combines both rules (e.g., a constitution) and values (e.g., commitment to democracy, accountability, meritocracy, etc.). Another conclusion is that the prospect of status change (e.g., the prospect of graduation from a peripheral transition country status to a member-state status with scope to influence EU decision making) is likely to be more effective in instigating commitment to reforms compared to side payments in the form of EU assistance.
The third conclusion is that the domestic cost of reform (i.e., resistance to reforms and the impact of such resistance on re-election probability) does enter the calculation of the policymakers and may reduce the level of reform performance. However, this cost includes a vector of subjective/strategic considerations that depend not only on the true cost of reforms but also on the rules of the game, the policymakers' perceptions and their ideological orientations. In addition, the domestic cost of reform is a negative function of the past reform effort. Given these qualifications, it may be misleading to explain the level of reform and consequent governance quality by referring to the cost of reforms as an exogenously-given quantity.
A fourth conclusion is that the impact of EU membership on reforms and consequent governance quality is uncertain. Therefore, it may be misleading to establish a causal link between the risk of backsliding (i.e., the probability of lower reforms and deterioration in governance quality) and the EU conditionality during the EU membership period. Recently, both international think-tanks and scholars working on Central and Eastern Europe have expressed concerns about this risk. For example, the Nations in Transition (NIT) project by Freedom House has alerted the relevant audience to the risk of backsliding in CEECs. In addition, a special issue of the Journal of Democracy (volume 14, no. 4, 2007) included 7 studies. Most of the contributors (for example , Rupnik 2007;Krastev 2007;Mungiu-Peppodi 2007;and Greskovits 2007) suggested that CEECs were experiencing some degree of backsliding and that the EU connection has played some role in this process. Whilst some point out the cost of convergence in economic policy as an explanatory variable, some point out the Euroskeptic reaction that brought populist forces and parties to power after accession.
The empirical findings presented in this article both support and qualify these qualitative observations. The evidence confirms that EU conditionality did not have a statistically significant effect on reform efforts and governance quality in CEECs during membership. As such, it lends indirect support to qualitative/descriptive evidence indicating stagnation or slow down in the pace of reform after EU membership. However, it also qualifies the backsliding argument in two ways. First, past performance matters and therefore any slowdown or deterioration in governance quality is bound to be country-specific. The persistence of stalling or the level of deterioration will depend on the level of governance quality achieved in the past. Ipso facto, the adverse impacts of stalling reforms or governance quality deterioration will also depend on how far each country had improved its governance quality during the accession period. Secondly, when analyzing the causes of backsliding, it is necessary to isolate the impact of EU membership from the wider impact of the external environment within which the risk of backsliding emerges. In the emerging literature on backsliding, this separation is not systematic and therefore there is a risk of mis-specification or ad hoc inclusion/exclusion of the relevant factors. The analysis in this article does not address this issue directly, but it points out the way in which it can be addressed analytically and empirically.    Captures perceptions of the extent to which a country's citizens are able to participate in selecting their government, as well as freedom of expression, freedom of association, and a free media Regulatory quality Captures perceptions of the ability of the government to formulate and implement sound policies and regulations that permit and promote private sector development.

Rule of law
Captures perceptions of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence. Political stability Measures perceptions of the likelihood that the government will be destabilized or overthrown by unconstitutional or violent means, including domestic violence and terrorism Government effectiveness Government effectiveness captures perceptions of the quality of public services, the quality of the civil service and the degree of its independence from political pressures, the quality of policy formulation and implementation, and the credibility of the government's commitment to such policies.

Control of corruption
Captures perceptions of the extent to which public power is exercised for private gain, including both petty and grand forms of corruption, as well as "capture" of the state by elites and private interests. All scores range from -2.5 to 2.5, with higher scores representing better governance quality. Source: http://info.worldbank.org/governance/wgi/mc_countries.asp Table A2: Governance data description and measures: ICRG Data Law and order Law and Order the strength and impartiality of the legal system, the popular observance of the law. Ranges from 0 -6.
Government stability This is an assessment both of the government's ability to carry out its declared program(s), and its ability to stay in office. Ranges from 0 -12.
Ethnic tensions This is an assessment of the degree of tension within a country attributable to racial, nationality, or language divisions. Ranges from 0 -6.
Democratic accountability This is a measure of how responsive the government is to its people. Ranges from 0 -6.
Corruption control This is an assessment of corruption within the political system. Ranges from 0 -6.

Bureaucracy quality
The institutional strength and quality of the bureaucracy. High scores are given to countries where bureaucracy has the strength and expertise to govern without drastic changes in policy or interruptions in government services. Ranges from 0 -4.
The higher the score, the better is the institutional quality and/or the lower is the risk. Source: http://www.prsgroup.com/ICRG_Methodology.aspx Solution of the utility maximization model.
The utility functions for the accession country (Ua) and the EU (Ue) are: The probability of membership as a function of CEEC reform and EU commitment: The variables are as defined in section 3 above.
The cost functions for the CEEC and EU depends on the difference between current and past commitments. The past commitment of each side is denoted by the variable with zero subscript. R 0 is the level of past reform (hence past governance quality) in CEECs and M 0 is the level of membership prospect offered by the EU at the beginning of the accession period. The difference (R -R 0 ) and (M -M 0 ) reflects the extra effort during each year of the accession period.
Replace the cost functions in (A1) and (A2) with their equivalents in (A4) and (A5): Differentiate (A1') with respect to reforms (R) and (A2') with respect to membership prospect (M), and set equal to zero to find the turning point.
From (6) and (7), the equilibrium conditions for the accession country and the EU are as follows: Now we can solve for the level of reform (R) and membership prospect (M) that maximize utility for both sides simultaneously. These levels are: From (A8), we can see that the level of reform is negatively related to EU support towards adjustment costs (s) and the coefficient of the marginal costs of reforms (h). These are the sources of moral hazard for accession countries. From (A9), we can see that the current membership prospect is a positive function of past EU commitment (M 0 ) and negative function of the coefficient of the marginal cost (d). These are the sources of adverse selection by the EU.