4.1 Worldwide Governance indicators
To measure institutional development, we employ the Worldwide Governance Indicators (WGIs) of the World Bank produced by Kaufmann et al. (2013).9, 10 The WGIs consist of six composite indicators capturing governance perception. The six dimensions of governance are Voice and Accountability (VaA), Control of Corruption (CoC), Government Effectiveness (GE), Political Stability and Absence of Violence (PSNV), Rule of Law (RoL) and Regulatory Quality (RQ). The WGIs are composed of several hundred variables obtained from surveys of firms and households and subjective assessments collected by commercial business providers, non-governmental organizations, multilateral organizations and other public sector
7 Our results were obtained using R 2.15.2 with the packages plm 1.3-1, lmtest 0.9-32 and car 2.0-19 (R Core Team 2012; Croissant, Millo 2008; Zeileis, Hothorn 2002; Fox, Weisberg 2011).
8 See also Deutsche Bundesbank (2012, pp. 23–25) for an empirical application to a similar sample as well as Alesina et al. (2010), both of which apply a fixed-effects estimator.
9 For a detailed description of the WGI methodology, see Kaufmann et al. (2010).
bodies. The variables are clustered along the six dimensions of governance by an unobserved components model. By following this method, it is possible to construct margins of error, which indicate the underlying uncertainty, and assign weights according to the informative signal of the source (Kaufmann et al. 2010, pp. 5–11). The WGIs are normally distributed, with zero mean and ranging approximately from –2.5 to 2.5. Table 1 describes the six dimensions of the WGIs.
Table 1 Description of the six dimensions of the WGIs Voice and Accountability
(VaA)
Capturing 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 free media.
Political Stability and Absence of Violence/
Terrorism (PSNV)
Capturing perceptions of the likelihood that the government will be destabilized or overthrown by unconstitutional or violent means, including politically motivated violence and terrorism.
Government Effectiveness (GE)
Capturing 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.
Regulatory Quality (RQ) Capturing 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 (RoL) Capturing 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.
Control of Corruption (CoC)
Capturing 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.
Source: Kaufmann et al. (2010, p. 4)
The indicators cover the years 1996–2012 and are available on a two-year basis until 2002 and on a yearly basis subsequently. Hence, we have to handle the missing data problem for the WGIs in 1997, 1999 and 2001. Since the lack of these WGIs depends neither on their value nor on the values of other variables in the dataset, they are missing completely at random (for a discussion of missing data assumptions and their consequences, see Cameron, Trivedi 2007, pp. 923–941).
There are two simple ways to handle missing data in this case: listwise deletion11 and mean imputation. The former has several disadvantages. Since 3 out of 17 observations (time dimension) are missing, deletion leads to a substantial reduction in the total number of observations. Even more information is lost if we calculate the year-to-year changes in the
10 The field of economic governance analyses the “performance of different institutions under different conditions, the evolution of these institutions, and the transitions from one set of institutions to another” (Dixit 2008).
Important measures of institutional quality are the Worldwide Governance Indicators (WGIs).
11 Listwise deletion means reducing the sample to complete observations (Cameron, Trivedi 2007, p. 925).
WGIs, which indicate institutional development. Then the sample is essentially cut to 2003–
2012. One possibility to mitigate this problem is to calculate two-year changes in the variables and skip the information for the years 2003, 2005, 2007, 2009 and 2011. This is carried out in Section 6 as a robustness check. The second way to handle missing data is mean imputation. Thereby, missing WGIs are replaced by the average value of the previous-and following-period WGIs for each country. The disadvantage is that this could have an impact on the distribution of the WGIs and therefore could affect the covariances with other variables. Nevertheless, we favour the latter way to handle the missing WGI problem to retain as much information as possible.
There are several potential problems with the WGIs to discuss as we use the cross-section and time dimension in our analysis.12 As the WGIs are constructed to have a zero mean in each period, comparisons of WGIs over time could be a problem. Kaufmann et al. (2007, pp.
3–4) argue that this could indeed be problematic for absolute changes in the WGIs.
However, relative comparisons of individual countries or country groups are not affected and valid, even if the world averages have changed over time. Indeed, the world averages of the underlying sources show little evidence of significant trends, as Kaufmann et al. show in previous works. Hence, this allows the interpretation of relative changes as absolute changes in individual or groups of countries (Kaufmann et al. 2007, pp. 3–4). A further point of criticism is that the WGIs might be too imprecise to yield sensible comparisons over time or countries. This criticism could be applied to every governance indicator because of measurement errors. However, the WGIs aggregate the existing indicators and hence their information about governance. Above that, margins of errors are computed, which allows the testing for significance in differences (Kaufmann et al. 2007, pp. 10–11).
4.2 Status dummy variables
To indicate the status or official relationship of the countries of our sample with the EU and the euro area, we construct a set of dummy variables. The sample covers 56 countries, among which are 33 European countries, which have been at least a potential candidate at some point in time (1996–2012) according to the classification in Table 2. The remaining 23 countries are other OECD countries and other European and Central Asian developing countries as defined by the World Bank. A full list is presented in the appendix.
12 This and other critiques are discussed in Kaufmann et al. (2007).
Table 2 Classification of the status dummy variables
Status Abbreviation for Classification Source MBEA Member State in
the euro area
EU Member State at Stage Three of the Economic and Monetary Union, i.e. Member State in the euro area
(European Central
EU Member State with derogation, i.e. a Member State preparing to adopt the euro but has not yet done so (other than Sweden)
(European Central Bank 2012, p. 64;
European Union 2012)
ACEU Acceding country for the EU
Country that has signed the treaty of accession (European
Commission 2012) CCEU Candidate
country for the EU
Applicant country for EU membership that has been granted candidate country status by the European Council
Countries of Central and Eastern Europe, which have signed Europe agreements;
countries of the Western Balkans involved in the stabilization and association process, which are not yet candidate countries; 6 Western Balkans countries were identified as potential candidates during the Thessaloniki European Council summit in 2003; the European Council confirmed a clear European perspective for Kosovo in 2008; in 2009 Iceland applied to join the EU
(European
Notes: Many countries were granted candidate status at European Council meetings in December. As one should expect no effect for the respective year, our dummy variables generally display all the changes in the status occurring during the months November and December in the following year. Beyond that, we assign all the Member States that adopted the euro in 1999 the status CCEA until 1998. EU Member States that have been granted exemption from participating in the third stage of the Economic and Monetary Union (i.e. the United Kingdom and Denmark) and Sweden, which is de facto not willing to introduce the euro (see European Central Bank 2012, p. 64; European Union 2012), are not considered here.
The notation and classification of a country’s status are based on the official specifications of the European Union and the European Central Bank. However, they are not identical in the case of Sweden and the potential candidates. Originally, the EU named countries involved in the stabilization and association process in the Western Balkans potential candidates, which are not yet official candidate countries (European Commission 2012). We extend this term to all the countries involved in a pre-accession strategy of the EU according to the definition in Table 2. Sweden, which is officially a EU Member States with derogation, is not considered here, as Sweden is de facto not willing to introduce the euro. For example, it did not participate in the European Exchange Rate Mechanism (ERM and ERM II) in the relevant period.