Whilst much research has been conducted to evaluate the merits and demerits of Transparency International’s (TI) CPI (Thompson and Shah, 2005; Luo, 2006; Andersson and Heywood, 2009; Laufer and Warren, 2009; Saha et al., 2012; Donchev and Ujhelyi, 2014; Ulman, 2014; Vadlamannati and Cooray, 2017; Gilman, 2018; Hart, 2019), there has been considerably less attention paid to the role that ratification of key global anti-corruption policy conventions has on country scores within the CPI. It is within this context that this paper seeks to evaluate whether early ratification of the UNCAC has had an impact on country scores within the CPI.
The CPI enjoys widespread use as a tool to measure corruption within countries. For instance, there have been a number of attempts to evaluate the effect that a country’s CPI score or rank in the table has on a wide range of other variables across a number of disciplines, ranging from corporate social responsibility (Luo, 2006), environmental studies (Koyuncu and Yilmaz, 2009), medicine (Gadit, 2011), economics (Dreher and Herzfeld, 2005; Ulman, 2014; Sparks et al., 2014; Manu and Patel, 2018), engineering (Sohail and Cavil, 2008), military science (Georgiev, 2013), religion (Charles et al., 2013), and gender studies (Alatas et al., 2009), to name but a few examples.
It is within this context that in an attempt to evaluate whether there is, in fact, a relationship between early ratification of the UNCAC and a country’s CPI Scores, a panel of 178 countries using data gathered from the period 2004 – 2018 has been brought together. There are a total of 2,638 country-year observations contained within the CPI scores data, and a total of 25,806 data points across all variables contained within the dataset.
Given the focus on the score, as opposed to the rank a country receives, this paper uses an OLS simple linear model to regress the score received each year with the addition of a lagged CPI explanatory variable as a means of capturing dynamic effects in political processes which may not be captured by a simple CPI score and as a method for ridding the model of autocorrelation (Dougherty, 2016, p. 402). The focus on the score was to ensure that changes in CPI performance were captured, wherein the use of a rank may change due to the inclusion of other countries, or significant changes in rank by countries that were lower, which may then affect the rank of a country where their score does not change. It was decided that evaluating country scores captured a more reflective picture. The dependent variable was then regressed against a dichotomously coded indicator variable which looked at whether a country was an early ratifier of the UNCAC or not. This was based on whether a country ratified the UNCAC before December, 2008 or afterwards. In this context, the analysis focuses on viewing early ratification through the lens of policy diffusion and more specifically, the notion of ‘leaders’ and ‘laggards’ with regard to the diffusion of policies. The staggered ratification process further allows for an analysis which incorporates the notion of ratification ‘leaders’, i.e. those that ratified pre-2008, and ‘laggards’, i.e. those who ratified after 2008 (Shipan and Volden, 2008, p. 843). This allowed for an analysis that could calculate the impact early ratification had, given two important developments following this period; firstly, the introduction of an implementation and enforcement review for the UNCAC in 2009, which changed the expectation that a country could expect following ratification, and secondly, the methodology changes made to the CPI in 2012. Similarly, as can be seen in Figure 1 earlier, a considerable majority of countries had ratified the convention by December 2008; indeed, 68% of countries had ratified the convention by this date (Cole, 2015). This diverse diffusion in the way of early ratification meant that we could effectively control for regional variations, which would otherwise make findings more difficult to interpret. Similarly building on previous policy diffusion literature, it can be seen that early ratification prior to 2008 serves as an effective gauge into whether countries which could be considered ‘leaders’ – by ratifying the convention early – would benefit from an improvement in their country scores at later stages.
The choice of Ordinary Least Squares simple linear model allowed for an analysis which could explore the relationship between the dependent variable, in this case, a country’s CPI score, and an independent variable that looked at whether a country ratified the UNCAC early, i.e. prior to 2008. This method provides an estimation of the relationship by minimising the sum of the squares in the difference between the observed and predicted values of the dependent variable, which is configured as a straight line. The inclusion of control variables allows us to hold all other factors constant and address the problem of confounding (Pearl, 2009, p. 116). The use of a lagged dependent variable allows for the model to more accurately map the temporal dynamics of the dependent variable and to take account of the impact of previous dependent variable values as an insight into current and future dependent variable values. It does so alongside understanding that, in line with Dougherty (2016, pp. 454- 457) and Keele and Kelly (2006, p. 5), the use of lagged dependent variables can also serve to correct for potential autocorrelation. Alongside this, the use of a model with a lagged dependent variable is, as stated by Dougherty (2016, p. 402), “often attractive because it permits the representation of the process to have plausible dynamic properties without necessarily giving rise to the problem of multicollinearity”.
Whilst there are a myriad of methodological challenges within the CPI, it continues to find widespread use in discussions surrounding corruption and currently serves as the most effective composite index of perceptions of corruption, and indeed as a benchmark of expected levels of corruption within a country. Given its structure, it perhaps serves as the best gauge for the salience that ratification of the UNCAC has across the constituent indices that constitute the CPI.
Of particular relevance to this paper, as outlined by Marcos et al. (2017), is that the CPI is not dominated by any individual sources. Whilst this is important with regards to the statistical coherence of the CPI, it should also more thoroughly support the case for exploring how such a wide-ranging global instrument such as the UNCAC has impacted upon country scores within the CPI. Indeed, it should support the case for analysing whether this can be seen across the sources which contribute towards the development of the CPI, as outlined in Appendix 3, and the impact that policy changes and ratification of international conventions may have upon country scores within the CPI; specifically focused on those countries who ratify such conventions early and who could be considered leaders in seeking to tackle corruption.
A number of scholars, most notably Kolstad and Arne (2011), posit that whilst the use of perceptions of corruption data has its drawbacks, it remains valid in such research. Indeed, Kolstad and Arne (2011, p. 19) go as far as to state that “[w]hile the indices of corruption employed capture perceived rather than actual corruption levels, this reflects limitations in data availability, not in analytical approach. The empirical approach used would be perfectly applicable to analysis using other corruption indices, should these attain wider country coverage.” It is within this context that we seek to use the CPI as the best available measurement methodology to collate data on perceived levels of corruption.
Early Ratification of the United Nations Convention Against Corruption
Building on the work of Cole (2015) who sought to evaluate significant, increased attention paid to corruption through examining the volume of anti-corruption conventions and treaties, this paper takes the most wide-reaching convention, namely the UNCAC, and analyses it through the lens of early ratification of the convention as the dichotomously coded independent variable, thus building on the approach used by Heather et al. (2010) who similarly used dummy coded early adoption variables in an analysis that looks at adoption of an international treaty. Due in part to this widespread ratification, alongside its unique development process, the UNCAC serves as a benchmark through which countries can be analysed on their commitment to tackling corruption, by being ‘leaders’ in ratification of a key international instrument targeted towards reducing corruption on a level which is universal as all countries were eligible to ratify the convention. As outlined previously, the wide-ranging provisions contained within the UNCAC provide an insight into the level of commitment required by states prior to ratification.
We can see that the convention while relevant to the research question contained within this paper, mainly due to its widespread ratification and ambitious targets, does in fact face serious challenges with regards to both the review mechanism, which evaluates the co-option of the policy prescriptions into domestic legislation, which would thus provide a more nuanced picture of how countries are facing up to the challenge of corruption, alongside whether it goes far enough in tackling corruption (Hechler et al., 2011; Saha et al., 2012). Such a nuanced picture, and indeed the varying motivations for early or late ratification, may serve to inform a more balanced interpretation of the results of this analysis. The independent variable is structured as an indicator variable, with 1 being those countries which were early ratifiers of the UNCAC, and 0 being those who were not.
This analysis of the relationship between early ratification of the UNCAC and its impact upon CPI scores uses a number of social, political, and economic control variables, as highlighted in previous research (Achim, 2016; Alatas et al., 2009; Charles et al., 2013; Cole, 2015; Dreher and Herzfeld, 2005), into perceptions of corruption in order to minimise their potential impact upon this analysis.
|Democracy||Building on the work of Kolstad and Arne (2011), we can see that democracy is correlated with a reduction in the level of corruption and serves as an effective control of corruption considering the stage of democracy. To this end, this paper included a dummy variable using data from Freedom House. The dummy variable gave a score of 1 for Free countries, and a 0 for Partly Free and Not Free countries. As a large body of research points towards ‘the stage of democratic development’ being important in the control of corruption, it was understood that only those countries which were considered Free using the Freedom House classification system could be considered democracies.||Freedom House, 2019|
|Whilst there continues to be a widespread academic debate on the merits of the HDI in measuring levels of development vs. GDP data (Brinkman and Brinkman, 2011, p. 451; Dervis and Klugman, p. 75; Islam, 1995), there is albeit more consensus around the impact of poverty, low life expectancy (and more specifically poor health outcomes), and poor education in creating the conditions for corruption (Dreher and Herzfeld, 2005, p. 11; Mauro, 1998, p. 267; Gupta et al., 2002). It is for this reason that this paper seeks to control for the three primary areas under which the HDI covers, as outlined in the United Nations Development Program (2018) namely;|
1) Life expectancy index: this looks at the health and longevity of a country’s population;
2) The education index: this component evaluates the average school life expectancy of children of school age and mean years of schooling of the adult population; and
3) The index of gross national income: this component evaluates the gross national income per capita in US dollars.
|GDP Per Capita (Purchasing Power Parity||Another variable used to control the level of development is the use of GDP per capita purchasing power parity (PPP). GDP PPP has particular merit due to its ability to compare generalised differences in living standards. GDP PPP further takes into account the relative cost of living and inflation rates of each country to provide a more nuanced picture of the level of development. A number of scholars show that the level of development is correlated to the level of corruption within a country (Shabbir and Anwar, 2007; Shleifer and Vishny, 1993). This variable is used as an alternative to HDI.||World Bank, 2019|
|OECD Member||As outlined within Ross (2003, p. 241), the idea of using OECD membership is to control for Western biases, a critique of which CPI often faces. Through using a dummy coded variable for OECD membership, we can largely control for those countries typically considered ‘Western’ (Rehman and Naveed, 2007, p. 34).||OECD, 2019|
|Protestantism||A number of studies have demonstrated that corruption is less severe in predominantly Protestant societies. To this end this paper uses the percentage of population that identifies as protestant based on data from the Correlates of War, World Religion Project dataset (Sandholtz and Koetzle, 2000, p. 44; Treisman, 2000, p. 405; You and Khagram, 2005, p. 146).||World Religion Project, 2019|
|Inequality (Gini)||You and Khagram (2005, p. 146), Gupta et al. (2002), and Gyimah-Brempong (2002) show that income inequality correlates with increased corruption. To this end, this paper uses a measure of gross income inequality, expressed as Gini coefficients.||World Bank, 2019|
|Region||Using data from Wahman et al. (2013) and Teorell et al. (2019), this variable is a politico-geographic classification of world regions into ten distinct categories. The values are derived through both an analysis of geographical proximity (with some exceptions) and demarcation done by area specialists who have contributed to regional understandings of democratisation||(Wahman et al., 2013)|
Teorell et al. (2019),
|Country||Due in part to the unique and long-term country-specific effects which are often not captured in the variables above, such as the impact of culture, history and institutional structure in a country’s CPI score, it was decided to include a dummy coded variable to control for country-specific effects.||CPI, 2019|
Table 2: Control Variables used.