Political Capacity and Corruption Nexus: Re-Examining Evidence for Developing Countries

Abstract

This article examines the question of whether developing countries with strong political capacity have lower levels of corruption. Despite the ubiquity of literature on corruption, the relationship between a state’s political capacity and corruption has not been addressed by the existing academic literature. To measure the political capacity of a country, the authors have used relative political capacity (RPC), an indicator that gauges the effectiveness of governance by its ability to meet or exceed their expected extractive capabilities and its ability to implement a set of policy choices. On the one hand, politically capable and stable governments are in a position to pursue their political and economic goals, such as reducing corruption. On the other hand, a strong political capacity provides them with the opportunity for rent-seeking and corruption. This implies that a state’s strong political capacity can be either a ‘boon’ or ‘bane’ to implement a set of desired policy goals. Based on this assertion, the authors test the hypothesis of whether a strong relative political capacity increases or reduces the level of corruption. The analysis uses the ordinary least-squares and two-stage least squares methods for 98 developing countries to test the hypothesis. The findings suggest that the explanatory power of political capacity is at least as important as conventionally accepted causes of corruption, such as economic development, and democracy.

1. Introduction

Despite the burgeoning academic literature, the relationship between the government’s political capacity and corruption has been rarely examined. Many studies have considered the quality of institutions as a proxy for the state’s political capacity and have held poor governance and weak institutions responsible for causing corruption (Aidt 2009, 2016; Huntington 2006; Klitgaard 1988). However, debatable in their analysis is the use of an appropriate indicator measuring the quality of a nation’s institutions and governance capacity.

This study endeavors to re-examine the relationship between the quality of institutions and corruption. In doing so, however, the aim is also to find an objective measure of governance and political performance that is aggregate yet comprehensive, reflecting a government’s overall capacity and performance to pursue policy goals, both economic and political.

The existing studies have rightly recognized the importance of good governance and institutions in deterring corruption (Such as Aidt 2009; Lambsdorff 2003; Rose-Ackerman 2004). Likewise, the study of institutions, both theoretical and empirical, have merited the attention of contemporary political economists (Gerring and Thacker 2004; Lederman et al. 2005; Baig and Feng 2016); however, little work has been done to objectively and systematically evaluate the relationship between a nation’s overall political performance and corruption in an empirical, cross-national setting that focuses specially on developing nations.

For the majority of the developing nations, controlling corruption is an important policy goal. However, delivering policy goals is costly. In particular, reforming the civil service, improving the quality of design and implementation, and creating an accountability mechanism hinges on the financial stress as well as on political conditions. In this backdrop, it is imperative to examine whether having stable political and economic conditions that otherwise manifest the higher quality of public institutions leads to lower levels of corruption. Such an indicator of the quality of governance is required that captures both financial and political stability of a nation, which is discussed in the following section.

This paper will proceed as follows. In the next section, we briefly discuss the different measures of governance and political capacity. Section 3 discusses the available measures of corruption. Section 4 puts forward the theory and main arguments of this paper. Section 5 presents the data and empirical methodology, followed by the main results in Section 6. Section 7 concludes the paper.

2. Measures of Government’s Political Capacity

Existing literature suggests that the incidence of corruption depends crucially on the quality and capacity of political institutions (For instance Aidt 2009). However, defining governance and political capacity is itself a task complex to tackle. In a cross-national setting, constructing a one-size-fits-all governance capacity measure is not without problems. For instance, the quality and competence of civil services might not be of significant importance to many developing countries. Likewise, if we consider the quality of an economic or legal or political institution as a measure of governance capacity, as many existing studies have done, we may be, in fact, measuring only a fraction of what we intend to measure. Therefore, we are interested in finding an objective yet a broad measure that reflects various aspects of governance, including quality of bureaucracy, capacity to implement rules and regulations, quality and the credibility of the government’s commitment to policies, and such like factors. In doing so, we garner some wisdom from the existing literature that attempts to define governance and political capacity.

Among many indicators of governance and political capacity, Worldwide Governance Indicators (WGI) provided by the World Bank are widely used. WGI consist of six indices, namely regulatory quality, voice and accountability, control of corruption, rule of law, government effectiveness, and political stability. The World Bank defines governance as follows: it “is the manner in which power is exercised in the management of a country’s economic and social resources for development” (WDI, World Bank 2020). According to Arbetman-Rabinowitz and Johnson (2007) as cited in Baig and Feng (2016), justice, ethics and institutional quality, among other variables, are the main constituents of good governance. Among many studies that make use of WGI, Rock (2009) uses the ‘rule of law’, Aidt (2009) considers ‘voice and accountability index’ to be a representative measure of governance. Likewise, Al-Marhubi (2004) constructs a governance measure by taking simple average of any two indictors out of the six WGI. To measure the policy content and the quality of policy design, Baig and Feng (2016) consider ‘the governance effectiveness’ indicator. Organski and Kugler (1980), however, criticize WGI for being biased towards democracy. They argue that “an effective political system needs not be free, democratic, representative, or participatory.”

Some scholars use proxies for measuring political capacity, which include the ratio of paved roads to total road network (Arjona and Kalyvas 2009; De la Calle and Sánchez-Cuenca 2001), telephone lines per 100 people (Mann 1984), GDP per capita (Fearon and Laitin 2003), and income tax as a share of GDP (Lieberman 2002). However, these proxies are too parsimonious for more detailed research that is directly related to political capacity, possibly failing to capture some crucial aspects.

Unlike the aforementioned studies, Xin and Rudel (2004) argue that democracies, in themselves, are partially endowed with the capacities necessary for maintaining government legitimacy. They suggest a measure of revenue extraction through the broad-based taxation of citizens as a proxy measure for the ability of a state to mobilize its citizens and, in turn, to support the state. A large public sector reduces corruption, they argue. Their hypothesis is based on the premise that the public pays taxes when it believes that the government is not corrupt. Likewise, Rose-Ackerman (2005), endorsing the findings of Xin and Rudel (2004), suggests that low government legitimacy encourages citizens to shirk on taxes. On the other hand, among many dissenting voices are Lederman et al. (2005), who suggest a higher association of corruption with large government revenues. Given that we are interested in capturing the aggregate performance of political institutions, considering government revenue as a single economic variable may not be an appropriate yardstick to measure political capacities.

Related to the arguments of Xin and Rudel (2004), and Rose-Ackerman (2004) about the government’s legitimacy is the concept of Relative Political Capacity (RPC)1, developed by Organski and Kugler (1980). RPC is a measure that evaluates the ability of a government to extract resources from a population, given their level of economic development2 (Feng 2006). Likewise, RPC gauges governments’ efficiency by their ability to meet or exceed their expected extractive capabilities and their ability to implement a set of policy choices (Arbetman-Rabinowitz and Johnson 2007). Furthermore, politically capable and stable governments are in a position to pursue their political and economic goals. Thus, RPC reflects the durability and effectiveness of a nation’s institutional fabric. To capture the quality and capacity of state institutions, RPC seems to be an objective yet a comprehensive measure. One of the major components of RPC is relative political extraction (RPE). RPE approximates the ability of governments to appropriate portions of the national output to advance public goals Kugler (2018). The following model estimates RPE for developing countries:

$$\begin{gathered} TAX/GDP=\mathsf{\beta }1\left(Time\right)+\mathsf{\beta }2\left(Mining/GDP\right)+\mathsf{\beta }3\left(Agriculture/GDP\right)+\mathsf{\beta }4\left(Exports/GDP\right)+\mathsf{\beta }5\left(OECD\right) \\ +\mathsf{\beta }6\left(Inclusion Dummy\right)+\mu i \end{gathered}$$

(1)

RPE empirically tests the government’s productivity at the national, sub-national, and local levels. According to Kugler and Tammen (2012), RPE “represent(s) a political science counterpart to the long-valued, numbers-based GDP measure economists use to determine a nation’s growth or decline”.

In this backdrop, we suggest Relative Political Extraction (RPE) as an objective measure of the quality and capacity of a government to pursue a public policy goal. furthermore, in the contemporary polities, controlling corruption is one of the most pressing policy goals.

3. Measures of Corruption

Scholars have tried to give corruption an operational definition. Given its subjective nature, however, defining corruption is complicated. The three commonly used corruption indices are The International Country Risk Guide Corruption Index (ICRG) given by Political Risk Services Inc., the Transparency International Corruption Perceptions Index (CPI), and the World Bank’s Regulation of Corruption Index (WB). Misuse of government services for their own gain by government officials, including politicians, bureaucrats, and all civil servants, tends to be the only concept that makes sense for the empirical examinations using these indices. Any other conceivable description will be only a minor deviation from this one (Baig and Feng 2016).

4. Theory and Related Literature

Does a government’s political capacity affect corruption? This important question has not been addressed. This paper examines the question of whether weak governance performance is a key determinant of corruption in developing countries. While examining the effect of political institutions on corruption, existing literature (for instance, Groennendijk 1997) relies on a conventional principal–agent or a principal–agent–client model. The underlying principal in these models is that the voters (the principals) vote out the corrupt public officials (the agents), while retaining or voting in the honest ones. The principal–agent theory has two major limitations:

First, it refers to political institutions as democracies with the assumption that elections to elect public officials take place (Baig 2019). Without considering authoritarian regimes, however, any discussion on corruption in developing countries would be incomplete. This is because roughly half of the countries in the world, particularly in the developing world, can be classified as authoritarian in one way or another. Hadenius and Teorell (2007) have argued that the assumption of competitive elections in theoretical agency models have severely restricted the empirical sphere covered by these agency models.

Second, conventional economic and political analyses on corruption have an implicit assumption about the endogeneity of governance capacities in democracies and political institutions (Baig and Feng 2016). Scholars like Shleifer and Vishny (1993); Persson et al. (2003); Gerring and Thacker (2004); Lederman et al. (2005) have suggested that lower corruption is associated with democracies. However, this may not hold true for developing nations, given that India consistently ranks higher on the Corruption Perception Index (CPI) despite having a history of uninterrupted democracy. On the other hand, Singapore stands among the least corrupt countries despite ranking low on civil liberties indices. Having a democracy is not analog to having a higher political performance or better quality of institutions.

Likewise, many studies, such as Brunetti and Weder (2003), have investigated the hypothesis that a free press may serve as a potent check and balance against corruption. Investigating a wide range of nations, consistent evidence has been found of a strong association between increased press freedom and decreased levels of corruption. This finding is unaffected by changes in the specification or the sample, and the association does not seem to be affected by the precise measure of corruption or press freedom that is used. While examining the role of social media, Jha and Sarangi (2017) explored the connection between multi-way communication and corruption using social media. Using a cross-country study of over 150 nations, they established a negative association between Facebook penetration and corruption. They argue that the low-press-freedom nations had the highest association between Facebook and corruption.

In the backdrop of these limitations, we need a theory that redefines the principal-agent theory by taking authoritarian regimes into consideration and dropping the implicit assumption of high-quality institutions in democracies, which is ambitious. As discussed above, if not all democracies have lower levels of corruption, and if not all the autocracies have higher corruption, it would be worthwhile investigating if both democracies and autocracies with better institutional and policy framework have lower levels of corruption. In the onset, we discussed that RPC reflects the durability and effectiveness of institutions and the government’s capabilities to advance and implement policy goals. RPC also gauges a government’s efficiency by its ability to meet or exceed their expected extractive capabilities and its ability to implement a set of policy choices, and politically capable and stable governments are in a position to pursue their political and economic goals As controlling corruption is one of the most pressing policy goals, it is expected that higher RPC provides the government, either democratic or autocratic, with essential financial resources and stability, thus enabling them to enhance the efficacy of policy and institutions, which would raise the costs of corruption for public officials. This leads to the first hypothesis of the paper:

Hypothesis 1 (H1).

Higher political capacity has a positive effect on controlling corruption.

However, the effects of RPC on corruption can be counter-effective. A higher extraction of public resources and the political stability may have the temptation for public officials to follow a personal agenda. The public officials may prefer using public resources for private gains rather than enhancing the capacity of state institutions. As RPC enables governments to pursue their political and economic goals, public officials may choose an agenda based on personal rather than national interests. This leads to the second hypothesis of the paper:

Hypothesis 2 (H2).

The effect of RPC on corruption will be less significant if public officials pursue a personal agenda.

5. Empirical Strategy and Data

5.1. Empirical Strategy

Following You and Khagram (2005) and Jha and Sarangi (2018), we used OLS and 2SLS estimation techniques. According to Treisman (2000), using panel data for estimating variables having little or no enough variation (within country data) may not be the appropriate technique. You and Khagram (2005) suggest averaging the variables over the period under study to minimize the measurement error. We averaged all the independent variables over the years of 2002–2015. On the other hand, to minimize endogeneity, our main dependent variable, Corr-WB, forms the average of the years of 2016–2018. However, we also estimated several models where data for the dependent variable, either Corr-WB or Corr-TI, are for a single year, such as 2015, 2016, 2015, or 2018. Likewise, we estimated a single-year model with all the variables, both dependent and independent variables, for the year 2015. The decision to include or exclude control variables was based on motivations from the existing literature and their importance in any polity and economic system. However, we intentionally excluded some of the variables that have no clear policy implications.

For our base model, we estimated the following:

$$\begin{array}{l}\mathrm{Corruption}=\mathsf{\beta }1 \mathrm{Governance} \mathrm{Capacity} (\mathrm{Relative} \mathrm{Political} \mathrm{Extraction}, \mathrm{RPE})\\ +\mathsf{\beta }2 \mathrm{Democracy} (\mathrm{Freedom} \mathrm{House})\\ +\mathsf{\beta }3 \mathrm{Logged} \mathrm{GDP} \mathrm{per} \mathrm{capita}\\ +\mathsf{\beta }4 \mathrm{Trade} \mathrm{openness} \left(\mathrm{Index}\right)\\ +\mathsf{\beta }5 \mathrm{Rents} \mathrm{from} \mathrm{natural} \mathrm{resources} (\%\mathrm{GDP})\\ +\mathsf{\beta }6 \mathrm{Logged} \mathrm{GDP} \mathrm{per} \mathrm{capita}\\ +\mathsf{\beta }7 \mathrm{Logged} \mathrm{Population}\end{array}$$

For the rest of the model, we used variants of our base model. We added several additional control variables to this model, but not simultaneously, as presented:

$$\begin{array}{l}+\mathsf{\beta }8 \mathrm{Business} \mathrm{Freedom}\\ +\mathsf{\beta }8 \mathrm{Investment} \mathrm{Freedom}\\ +\mathsf{\beta }8 \mathrm{Monetary} \mathrm{Policy}\\ +\mathsf{\beta }8 \mathrm{Regime} \mathrm{Durability}\\ +\mathsf{\beta }8 \mathrm{Quality} \mathrm{of} \mathrm{Legal} \mathrm{Institutions}\\ +\mathsf{\beta }8 \mathrm{Quality} \mathrm{of} \mathrm{Economic} \mathrm{Institutions}\\ +\mathsf{\beta }9 \mathrm{Ethnic} \mathrm{fractionalization}\left(\mathrm{Index}\right)\\ +\mathsf{\beta }10 \mathrm{Religion} \mathrm{fractionalization}\left(\mathrm{Index}\right)\\ +\mathsf{\beta }11 \mathrm{Region} \mathrm{Dummies}\\ +\mathsf{\mu }\mathrm{i}\end{array}$$

The β’s are parameter estimates. The variables business freedom, investment freedom, monetary policy, regime durability, quality of legal institutions, and quality of economic institutions entered the model one at a time. All the models were tested for model diagnostics, such as linearity, multi-collinearity, normality of residuals, model specification, and heterokedasticity. All the regressions were estimated with White’s standard errors and covariance.

5.2. Data

Corruption: Our main dependent variable was the control of corruption index (Corr-WB) from the World Bank governance indicators. It ranges from −2.5 to +2.5, with a larger number indicating lower corruption. However, in the tradition of the existing literature, we rescaled the index by multiplying the original index by −1, (Corr-WB* − 1). On the new scale, a higher number indicates higher corruption. We also used the Transparency International’s index of corruption (Corr-TI) for a robustness check. It ranges from 0–10, where a higher number indicates lower corruption. We rescaled the original index (10-Corr-TI) so that a higher number indicates higher corruption. We used the average values of Corr-WB during the period 2002–2015 and 2016–2018 and single year values for the period of 2015, 2016, and 2015. For Corr-TI, we used average values of the years 2002–2015 and single year values of the year 2018.

Relative Political Capacity: Our variable of primary interest was the government’s political capacity. Data for Relative Political Capacity were taken from Kugler and Tammen (2012). Higher points on the scale show greater political capacity.

Religion and Ethnic Fractionalization: To measure religious fractionalization in a nation, we used a measure developed by Alesina et al. (2003). This indicator shows the probability of not belonging to the same religious group when two people are randomly selected from a given country. It ranges from 0 to 100 in percentage points, with a higher number indicating higher fractionalization. While the impact of cultural norms in corruption is well-studied, the link between individualism vs collectivism and corruption was studied by Jha and Panda (2017) Their study looked at individualism/collectivism and corruption in a wide range of nations. To prove causality, the research employed a measure of genetic distance between a country’s population and the U.S. population to instrumentalize the individualism/collectivism variable. Individualistic nations have lower levels of corruption (perception). This association is resistant to control factors and other corruption metrics.

Unlike Treisman (2000); Gerring and Thacker (2004); Rock (2009), who used an index of ethno-linguistic fractionalization to capture the effects of ethnicity, we used a measure of ethnic fractionalization developed by Alesina et al. (2003). The authors of this index argue that this index is more about racial fractionalization. A higher number on the ethnic fractionalization scale indicates higher fractionalization. Alesina et al. (2003) argue that ethnic fractionalization has “deleterious effects on institutional quality, economic policy, and economic development, whereas institutional quality is the transmitting channel through which ethnic fractionalization impacts corruption”, as cited in Baig and Feng (2016, p. 6).

GDP per Capita: Existing studies, for example, Gerring and Thacker (2004); Ades and Di Tella (1999); Goel and Nelson (2005), argue that higher economic development leads to lower levels of corruption. To control for the effects of economic development, we considered real GDP per capita.

Openness of Economy: In the tradition of the existing literature on corruption, we included a variable to capture the effects of openness of the economy. Openness is measured by total trade (sum of exports and imports) as a ratio of the GDP of a country.

Likewise, the existing literature also argues that the abundance of mineral resources leads to higher levels of corruption (Ades and Di Tella 1999; Treisman 2000; Rock 2009). To capture the effects of the mineral resources, we included a variable on mineral and natural sources, which was defined as the total natural resources rented as a fraction of a country’s GDP. Furthermore, in tradition with Alt and Lassen (2003) and Xin and Rudel (2004), we included a variable on population to control for the size of the country. Finally, two regions’ dummies, Asia and Africa, were included to isolate the region’s effects.

5.3. Additional Control Variables

In addition to the control variables discussed in the preceding sections, we included several economic and institutional variables in order to evaluate their influence on RPE and corruption. The purpose of considering these variables is to examine their effects as substitutes or compliments of RPE. A brief discussion on these variables is given in the following paragraphs.

Business freedom: The business freedom score encompasses the following 10 components, all weighted equally: (1) starting a business—procedures (number); (2) starting a business—time (days); (3) starting a business—cost (% of income per capita); (4) starting a business—minimum capital (% of income per capita); (5) obtaining a license—procedures (number); (6) obtaining a license—time (days); (7) obtaining a license—cost (% of income per capita); (8) closing a business—time (years); (9) closing a business—cost (% of estate); (10) closing a business—recovery rate (cents on the dollar). Data were from the Heritage Foundation. The business freedom index, averaged for the years 2002–2015, ranges from 0 to 100, where 100 represents the maximum degree of business freedom.

Investment freedom: This variable is an indicator of each country’s policies toward foreign and domestic investment. It captures the overall investment climate in a country. The country’s investment freedom ranges between 0 and 100, where 100 represents the maximum degree of investment freedom. Data were from the Heritage Foundation.

Monetary freedom: The score for the monetary freedom factor is based on two components: The weighted average inflation rate for the three most recent years and price controls. The weighted average inflation (WAI) rate for the three most recent years serves as the primary input into an equation that generates the base score for monetary freedom (MF). The extent of price controls is then assessed as a penalty of up to 20% subtracted from the base score. The country’s monetary freedom ranges between 0 and 100, where 100 represent the maximum degree of monetary freedom. Data were from the Heritage Foundation.

Regime durability: This variable is defined as the number of years since the most recent regime change or the end of transition period, defined by the lack of stable political institutions. Regime durability, averaged for the years 2002–2015, was calculated from Polity IV.

Legal Institutions: The legal institutions variable is a comprehensive measure of the overall quality of institutions in each country. Developed by Kunčič (2018), this variable is an average of the following individual indicators: (1) the property rights index from the Heritage Foundation; (2) the freedom of the press index as in indicator of the legal environment, calculated from the Freedom House; (3) Civil Liberties taken from the Freedom House; (4) Judicial independence, with data calculated from the Fraser Institute; (5) the Impartial courts index from the Fraser Institute; (6) the Protection of property rights index from the Fraser Institute; (7) the Law and order index calculated from the International Country Risk Group (ICRG); (8) Religion in Politics; and (9) Rule of Law. Both the indices were taken from the ICRG. The overall legal institutions index averaged for 2002–2015 ranges from 0 to 1, with a higher number indicating the higher quality of an institution.

Economic institutions: Like the legal institutions variable, the economic institutions variable is a comprehensive measure indicating the overall quality of economic institutions in each country. It is an average of the following individual indicators of the quality of economic institutions: (1) the Financial Freedom index calculated from The Heritage Foundation; (2) Freedom: the Business freedom index taken from the Heritage Foundation; (3) Regulatory Quality from the World Bank governance indicators; (4) Freedom of the Press is an indicator of Economic Environment from the Freedom House; (5) Freedom to own foreign currency bank accounts from the Fraser Institute; (6) Credit market regulations from the Fraser Institute; (7) the Labor market regulations index from Fraser Institute; (8) the Business Regulations index from the Fraser Institute; (9) Foreign ownership/investment restrictions from the Fraser Institute; and (10) Capital controls from the Fraser Institute. The overall economic institutions index averaged for 2002–2015 ranges from 0 to 1, where a higher number indicates a higher quality of institutions.

5.4. Instrumental Variables

Rose-Ackerman (2004) argues that less corrupt countries may have greater government legitimacy and thus higher revenue. To address this possible issue of simultaneous causation and the problem of measurement error and to make sure that the arrows point in the right direction, we also used 2SLS estimation to investigate the influence of RPE on corruption. However, we could not find a suitable instrument that is correlated with RPE but not with corruption. Instead, we used lags of RPE as instrumental variables.

Table 1. Correlation among RPE and Corruption.

Variables	WB	TI	TI	WB	RPE	RPE	RPE	RPE
	2000s	2000s	2018	2016–2018	1970s	1980s	2000s	1960s
CORR-WB 2002–2015	1
CORR-TI 2002–2015	0.8972	1
CORR-TI 2018	0.8809	0.8158	1
CORR-WB 2016–2018	0.9247	0.8171	0.9461	1
RPE 1970–1979	0.073	0.0968	0.0723	0.0912	1
RPE 1980–1989	−0.0332	−0.0181	−0.038	−0.0311	0.6925	1
RPE 2002–2015	−0.1797	−0.1541	−0.2751	−0.2494	0.3808	0.5763	1
RPE 1960–1969	−0.1192	−0.0743	−0.1001	−0.119	0.7383	0.4615	0.228	1

presents the correlations among averaged RPEs for various decades. The correlation between the RPE averaged for 2002–2015 and RPE averaged for 1960–1969 was r = 0.23, whereas the correlation between the RPE averaged for 2002–2015 and RPE averaged for 1970–1979 was r = 0.38. Furthermore, the correlation between the RPE averaged for 2002–2015 and RPE averaged for 1980–1989 was approximately r = 0.58. Interestingly, on the one hand, the correlation between the RPEs for the 2000s and 1980s was the highest amongst the RPEs for the average of various decades. On the other hand, the correlation between corruption indices and the averaged RPE for the 1980s was the lowest among averaged RPEs for various decades. Therefore, the RPE averaged for 1980–1989 served as our primary instrumental variable.

We did not consider the endogeneity of other variables, income per capita trade openness, and democracy and relied on existing literature, using instruments for these variables. However, to obtain an unbiased estimate for our variable of interest, we needed only to control for other variables that may have been correlated with the instrumental variables (You and Khagram 2005).

6. Empirical Evidence and Interpretations

Table 2. OLS regressions for the single-year versus averaged dependent variable.

Dependent Variable	Corr-WB				Corr-TI
	2002–2015	2015	2016	2016–2018	2002–2015	2018
	(1)	(2)	(3)	(4)	(5)	(6)
RPE	−0.19 *	−0.28 **	−0.28 **	−0.28 **	−0.47 **	−0.64 **
	(−1.99)	(−3.21)	(−2.94)	(−3.15)	(−2.88)	(−3.38)
Democracy	−0.074 **	−0.083 **	−0.044	−0.058 *	−0.093 *	−0.14 *
	(−3.01)	(−3.02)	(−1.59)	(−2.14)	(−2.03)	(−2.49)
GDPpc (Log)	−0.18 ***	−0.15 ***	−0.17 ***	−0.167 ***	−0.49 ***	−0.32 ***
	(−4.67)	(−4.12)	(−4.30)	(−4.34)	(−6.48)	(−4.02)
Rents	0.01 ***	0.01 ***	0.011 ***	0.011 ***	0.015 ***	0.017 **
	(3.70)	(4.31)	(4.19)	(4.40)	(3.60)	(3.15)
Trade Index	0.001	0.0002	−0.002	−0.001	0.006	−0.002
	(0.23)	(0.05)	(−0.56)	(−0.34)	(0.79)	(−0.22)
Population (Log)	0.004	0.02	0.002	0.01	−0.01	−0.004
	(0.19)	(0.93)	(0.10)	(0.49)	(−0.29)	(−0.07)
Constant	2.1 ***	1.8 ***	2.3 ***	2.1 ***	11.01 ***	10.03 ***
	(4.35)	(3.59)	(4.10)	(4.03)	(11.36)	(8.77)
Countries	93	93	93	93	93	93
Adj. R-sq.	0.46	0.49	0.4	0.45	0.5	0.42

Notes: t statistics are in parentheses. * p < 0.05; ** p < 0.01; *** p < 0.001 for two-tailed test. All the independent variables are for the average for the years 2002–2015. In all regressions, heterokedasticity –robust standard errors were used. Corr-WB = World Bank Control of Corruption Index. Corr-TI = Transparency International Corruption index. 2002–2015 = average over the years 2002–2015. 2016–2018 = averaged over the years 2016–2018. RPE = Relative Political Extraction.

presents the OLS regression results. Both single-year data and averaged data were used for the dependent variable. For the independent variables, averaged data for the years 2002–2015 were used. The Transparency International’s (Corr-TI) and the World Bank control of corruption index (Corr-WB) for the single years of 2016, 2018, and 2018 or the average for 2002–2015 or 2016–2018 were used in each regression as the measure of corruption, being the dependent variable in this analysis. Governance capacity, as measured by relative political capacity (RPE); per capita income (natural log of GDP per capita), the democracy score from the freedom house; and the trade openness index for their average values for 2002–2015 were used as explanatory variables, with the control used for the logged population.

6.1. Comparing the Dependent Variable for Various Years

In Table 2, we have compared measures of corruption for averaged data and single-year data. Independent variables were the average of 2002–2015 across all the specifications. In column (1) of Table 2, we have presented the results for the data averaged for 2002–2015 for both the dependent and independent variables. In column (2), the dependent variable is for 2016. Column (3) has the dependent variable for the year 2018. In column (4), Corr-WB is the average for the years 2016–2018. Next, in columns (5) and (6), we have replaced Corr-WB with Corr-TI. Corr-TI is the average for 2002–2015 for the model presented in column (6). Finally, in column (6), the dependent variable (Corr-TI) is for the data for 2018.

For the same dependent variable (Corr-WB or Corr-TI), the estimated coefficients for RPE, democracy, and per capita income always became larger in magnitude and more significant, whereas those for rents always remained same when data of the dependent variables for the following years were used. For Corr-WB (2002–2015) (as shown in column 1), RPE was significant at the 5% significance level and had its intuitive sign. For Corr-WB (2016) and (2018) and the average for 2016–2018, RPE was significant at the 1% significance level with its expected sign. The standardized coefficient for RPE increased from 0.14 to approximately 0.2 as we switched the dependent variable measures from 2002–2015 to 2016, 2018, and the average for 2016–2018. Note that the standardized coefficient for RPE was approximately the same in the case of Corr-WB for the years 2016 and 2018 and average of the years 2016–2018. For the dependent variable Corr-TI (2002–2015), and (2018), as shown in columns (5) and (6), respectively, RPE was significant at the 1% significance level with its intuitive sign. The standardized coefficient for RPE increased from 0.18 to approximately 0.23, as we switched the dependent variable (Corr-TI) from the average of the years 2002–2015 to the single year 2018.

Except for one specification, which was when the dependent variable (Corr-WB) was for the year 2018, the democracy variable was significant at, at least, the 5% significance level across all the specifications. For Corr-WB, the effect of democracy was highest for the model with the dependent variable for the average of the years 2016–2018, as shown in column (4) of Table 2. This was that a one-standard-deviation increase in the Freedom House democracy scale produced approximately a two-standard-deviations drop in the World Bank control of corruption index (Corr-WB). For Corr-TI, the effect of democracy was significant at 5%, and this effect became stronger when the corruption measure was for the year 2018. Column (6) shows that one standard increase in the aggregate effect of the democracy variable, averaged for 2002–2015, produced a 0.25-standard-deviations drop in the Transparency International’s corruption measure (Corr-TI), as shown in column (6).

The per capita income variable was significant at the 1% significance level across all of the six specifications. Unlike the democracy measure, the effect of per capita income on corruption was strongest when the dependent variable (both Corr-WB and Corr-TI) and independent variables were for the same year, i.e., 2002–2015, as shown in columns (1) and (5). For Corr-WB, a one-standard-deviation increase in per capita income averaged for 2002–2015 produced a 0.4-standard-deviation increase in the corruption measure averaged over the same years. For Corr-TI, a one-standard-deviation increase in per capita income averaged for 2002–2015 produced a 0.6-standard-deviation increase in the corruption measure averaged over the same years.

Regardless of the type of dependent variable used, the rents variable was significant at a 1% significance level across all of the six specifications. A one-standard-deviation increase in the rents variable produced approximately a 0.3-standard-deviation increase in both the corruption measures.

The trade index and the population have their intuitive positive sign but are not significant across all of the six models.

6.2. Models with Single Year Independent Variables

The purpose of this section is to show the significance and usefulness of using averaged data over single year data. In

Table 3. OLS regressions for a single year.

Dependent Variable	Corr-WB			Corr-TI
	2015	2016	2018	2015	2018
	(1)	(2)	(3)	(4)	(5)
RPE	−0.22	−0.21	−0.21	−0.26	−0.48
	(−1.82)	(−1.72)	(−1.57)	(−1.52)	(−1.81)
Democracy	−0.06 *	−0.06 *	−0.03	−0.09	−0.11 *
	(−2.18)	(−2.38)	(−1.11)	(−1.82)	(−2.16)
GDPpc (Log)	−0.11 **	−0.12 **	−0.15 **	−0.3 ***	−0.26 **
	(−2.69)	(−2.84)	(−3.22)	(−4.04)	(−3.06)
Rents	0.009 **	0.009 **	0.008 **	0.015 **	0.012 *
	(3.09)	(2.77)	(2.94)	(3.24)	(2.14)
Trade Index	−0.01	−0.01	−0.01	−0.01	−0.02
	(−1.75)	(−1.67)	(−1.61)	(−1.20)	(−1.61)
Population (Log)	0.01	0.02	0.005	0.01	−0.002
	(0.61)	(1.01)	(0.23)	(0.22)	(−0.04)
Constant	2.136 ***	2.014 ***	2.414 ***	10.16 ***	10.42 ***
	(4.06)	(3.74)	(4.43)	(10.69)	(9.03)
Countries	91	91	91	90	91
Adj. R-sq.	0.46	0.49	0.4	0.5	0.42

Notes: t statistics are in parentheses. * p < 0.05; ** p < 0.01; *** p < 0.001 for two-tailed test. All the independent variables are for the single year of 2015. In all regressions, heterokedasticity – robust standard errors were used. Corr-WB = World Bank Control of Corruption Index. Corr-TI = Transparency International Corruption index. RPE = Relative Political Extraction.

, we report results when all of the independent variables used were for the single year of 2015. Table 3 also shows that the estimated coefficients for RPE and other explanatory variables varied depending on whether Corr-WB or Corr-TI was used as the dependent variable. RPE was significant at only the 10% significance level when the single-year Corr-WB for 2015 and 2016 was used, whereas it was insignificant for Corr-WB for 2018. In contrast, RPE was significant at only the 10% significance level when the single-year Corr-TI for 2018 was used, whereas it was insignificant for Corr-TI for 2015. This suggests the presence of measurement errors in single-year independent variables as well as dependent variables. When the averaged data were used for independent variables, however, RPE was significant regardless of the corruption measure, with the control used for per capita income, democracy, trade index, natural rents, and population. The standardized coefficients for RPE were −0.16 for both Corr-WB (2015) and Corr-WB (2016). Similarly, the standardized coefficient for RPE was 0.15 for Corr-WB (2018). Likewise, for Corr-TI, the standardized coefficients for RPE were 0.1 and 0.17 for the year 2015 and the year 2018, respectively.

Just like RPE, the democracy variable was significant for the dependent variable (Corr-WB) for the year 2015 and year 2016. The democracy variable was, however, insignificant for Corr-WB for 2018. In contrast, democracy was significant when Corr-TI was used for the year 2018 and insignificant for the year 2015. The standardized coefficients for RPE were 0.16 for both Corr-WB 2015 and Corr-WB 2016 and 0.15 for Corr-WB 2018. For Corr-TI, the standardized coefficients for RPE were 0.1 and 0.17 for 2015 and 2018, respectively. If the difference in results among the various models is due to measurement error, averaging helps to reduce measurement errors, although it may not solve the problem of systemic bias in the corruption measure (You and Khagram 2005).

Next, per capita income and the rents variable were significant with the expected sign across all the models, regardless of the type of the dependent variable. On the other hand, both the population variable and the trade openness variable were insignificant across all of the five models.

6.3. The Influence of RPE on Corruption: Model with Additional Controls

We tested our hypotheses with a sample of 93 developing and emerging countries, for which both RPE for the years 2002–2015 and Corr-WB for the years 2016–2018 were available.

Table 4. OLS results for Corr-WB (2016–2018) for various models.

Dependent Variable	Corr-WB Averaged for the Years 2016–2018
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
RPE	−0.35 **	−0.45 ***	−0.36 ***	−0.29 **	−0.148	−0.34 ***	−0.29 **	−0.28 **	−0.22 *	−0.30 **	−0.29 **	−0.22 *
	(−2.88)	(−3.83)	(−3.68)	(−3.07)	(−1.35)	(−3.67)	(−3.00)	(−3.10)	(−2.59)	(−3.07)	(−2.87)	(−2.22)
GDPpc (Log)		−0.21 ***	−0.16 ***	−0.17 ***	−0.1 *	−0.14 **	−0.14 **	−0.16 ***	−0.1	−0.16 ***	−0.17 ***	−0.19 ***
		(−4.55)	(−4.06)	(−4.07)	(−2.16)	(−3.27)	(−3.21)	(−3.76)	(−1.99)	(−3.83)	(−3.79)	(−4.55)
Democracy			−0.1 ***	−0.06 *	0.023	−0.020	−0.051	−0.06 *	−0.05	−0.035	−0.06	−0.06 *
			(−3.95)	(−2.38)	(0.93)	(−0.69)	(−1.99)	(−2.03)	(−1.83)	(−1.39)	(−1.92)	(−2.32)
Legal Institutions					−2.43 ***
					(−4.99)
Econ Institutions						−1.41 **
						(−2.94)
Regime Durability							−0.01 **
							(−2.32)
Monetary Policy								−0.005
								(−1.74)
Business Freedom									−0.02 ***
									(−4.19)
Investment Freedom										−0.008 *
										(−2.51)
Regions											Yes	Yes
Other Controls	No	No	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	0.943 ***	2.527 ***	2.515 ***	1.970 ***	2.122 ***	2.284 ***	1.964 ***	2.371 ***	2.270 ***	2.496 ***	2.053 ***	2.707 ***
	(7.54)	(7.69)	(8.73)	(3.74)	(4.17)	(4.30)	(3.48)	(4.17)	(4.33)	(4.53)	(3.51)	(5.12)
Countries	93	93	93	93	93	93	93	93	93	93	93	93
Adj. R-sq	0.052	0.26	0.37	0.43	0.56	0.48	0.46	0.44	0.52	0.46	0.43	0.48

Notes: t statistics are in parentheses. * p < 0.05; ** p < 0.01; *** p < 0.001 for the two-tailed test. All the independent variables are averaged for the years 2002–2015. In all regressions, heterokedasticity –robust standard errors were used. Corr-WB = World Bank Control of Corruption Index. Corr-TI = Transparency International Corruption index. 2002–2015 = average over the years 2002–2015. 2016–2018 = average over the years 2016–2018. RPE = Relative Political Extraction.

presents the OLS regression results of various specifications for a sample of 93 countries. All the explanatory variables, including RPE, were averaged for the period of 2002–2015. The dependent variable, Corr-WB, was averaged for the period of 2016–2018. Table 4 presents results for various models with additional controls.

The simple correlation between the relative political capacity (RPE) and control of corruption (Corr-WB) was −0.35 and highly significant. However, after the inclusion of per capita income, the magnitude of the RPE coefficient increased, with the standardized coefficient of −0.32, whereas the per capita income was highly significant at the 1% significance level, with the standardized coefficient of approximately 0.5. (OLS2). This indicates that economic development is not a substitute for RPE to deter corruption. Likewise, the inclusion of the democracy variable in the regression, along with per capita income as shown in column (3) of Table 4, makes the magnitude of the coefficient of RPE decrease. The standardized coefficients for RPE, per capita income, and democracy were 0.26, 0.37, and 0.35, respectively. Next, in all the specifications from column (4) to column (12), we introduced a standard set of control variables, which included rents, the trade index, the logged population variable, the ethnic fractionalization variable, and the religion fractionalization variable. Except for the rents variable, all of the aforementioned control variables were insignificant across the eight specifications. However, they had their expected signs in most of the specifications. We have not reported these variables in Table 4; a yes for other controls at the bottom of each column refers to this standard set of control variables. However, we have reported other controls in Table 4. The standardized coefficients declined from −0.26 to −0.2 and 0.35 to 0.2 for RPE and democracy, respectively, when we introduced the standard set of control variables, as shown in column (4). On the other hand, the standardized coefficient for per capita increased from 0.37 to 0.38 with the inclusion of control variables.

Out of the 12 specifications, RPE was insignificant only in one case, which was when we introduced the legal institutions variable as an additional control variable along with the standard controls as shown in column (5). Technically, the variables RPE and Legal Institutions are not strongly related, as the correlation between the variables was r = −0.24. However, theoretically, the two variables are strongly related. The Legal Institutions variable is a comprehensive measure of governance effectiveness, as it is an average indicator of the quality of various institutions in a country. On the other hand, RPE captures a measure of the ability of a government to pursue and implement policy goals, as it reflects the extractive capabilities of nations as well as their economic and political stability Arbetman-Rabinowitz and Johnson (2007). Thus, the two variables are theoretically related. If the estimates in column (4) are correct, they suggest that the effect of RPE on corruption is captured by the Legal Institutions variable, which was highly significant at the 0.001 level of significance. The standardized coefficient for Legal Institutions was approximately −0.6. Another striking result after inclusion of the Legal Institutions variable was that the democracy variable became insignificant. The standardized coefficient of the per capita income variable (−0.19) dropped as we introduced the Legal Institutions variable, which was highly significant at the 0.001 level of significance. The standardized coefficient for Legal Institutions was approximately −0.6.

The situation was reversed, however, when we switched Legal Institutions with Economic Institutions, as shown in column (6). RPE was significant at the 0.001 significance level, with the standardized coefficient of 0.25. The Economic Institutions variable was significant at 5% significance, with a standardized coefficient of −0.35. As in the case of Legal Institutions, the democracy variable became insignificant with the inclusion of the Economic Institutions variable. However, per capita income remained significant, and its standardized coefficient increased from −0.19, as in the case of the model with Legal Institutions, to −0.23.

Next, for the model reported in column (7), we replaced the Economic Institutions variable with the Regime Durability variable, rather than keeping the standard set of controls along with the democracy and per capita income variables. Controlling for Regime Durability decreased the coefficient for RPE as compared to the model with Economic Institutions. However, it was highly significant at the 1% significance level. The inclusion of Regime Durability made the democracy variable significant at the 10% significance level, whereas per capita income remained significant at 1%. The standardized coefficients for RPE, per capita, and the democracy variables were −0.21, −0.32, and −0.18, respectively. Regime Durability was highly significant at the 1% significance level, with a standardized coefficient of −0.18.

RPE, per capita income, and the democracy variable remained significant at the 1%, 0.1%, and 5% significance level, respectively, when we switched the Regime Durability variable with the Monetary Freedom Variable, as reported in column (8). The standardized coefficients for RPE, per capita, and the democracy variables were −0.2, −0.37, and −0.2, respectively. Monetary Freedom was highly significant at the 10% significance level, with a standardized coefficient of −0.09.

When we introduced Business Freedom into the model in column (9), the coefficient of RPE declined; however, it was still significant at 5%. Coefficients of both per capita income and the democracy variable declined as well and were significant at the 10% significance level. Business Freedom itself was highly significant at 0.1 percent. In contrast, replacing Business Freedom with Investment Freedom magnified the coefficients for RPE and per capita income. However, the coefficient for the democracy variable decreased, and it was no longer significant. The Investment Freedom was significant at the 5% significance level.

In the last two columns, we introduced controls for regional influences. For the model in column (11), we controlled for Asia, and for the model in column (12), we included the region dummy for Africa. The impact of RPE on corruption was higher for countries in Asia than those in Africa. When the Asia dummy was switched with the Africa region dummy, the standardized coefficient of RPE dropped from −0.21 to −0.18. For Asia, the effect of democracy on corruption was significant at 10%, whereas for Africa, it was significant at the 5% significance level. Per capita income was highly significant at 1% for both Asia and Africa.

6.4. Results for 2SLS Estimations

Given that the OLS estimates for RPE can be biased due to measurement error in RPE, reverse causality from corruption to RPE, and omitted variables, we used RPE for the average of the years 1960–1969, 1970–1979, and 1980–1989, one at a time, as the instrumental variable for RPE for 2002–2015. The data on RPE for the decades other than 2002–2015 were not available for many developing countries, which shrunk our sample size to 76 countries.

Table 5. 2SLS results for Corr-WB and Corr-TI.

Dependent Variable	Corr-WB 2002–2015			Corr-TI 2002–2015			Corr-WB 2016–2018		Corr-TI 2018
Instruments	RPE60s	RPE70s	RPE80s	RPE60s	RPE70s	RPE80s	RPE80s	RPE80s	RPE80s
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
RPE	−0.79	−0.56 +	−0.39 *	−1.9	−1.04 +	−0.76 *	−0.3	−0.4 +	−0.8 +
	(−1.00)	(−1.82)	(−1.98)	(−1.29)	(−1.82)	(−2.05)	(−1.33)	(−1.66)	(−1.86)
Democracy	−0.02	−0.03	−0.04	0.003	−0.03	−0.03	−0.04	−0.01	−0.05
	(−0.60)	(−1.01)	(−1.22)	(0.03)	(−0.41)	(−0.57)	(−1.40)	(−0.43)	(−0.69)
GDPpc	−0.24 **	−0.22 ***	−0.21 ***	−0.62 ***	−0.55 ***	−0.53 ***	−0.17 **	−0.14 **	−0.3 **
	(−3.17)	(−4.59)	(−4.36)	(−4.17)	(−6.39)	(−6.10)	(−3.07)	(−2.71)	(−2.81)
Population (Log)	0.001	0.01	0.02	−0.04	−0.02	−0.01	0.02	0.004	−0.04
	(0.04)	(0.30)	(0.51)	(−0.55)	(−0.33)	(−0.19)	(0.65)	(0.16)	(−0.71)
Rents	0.01 *	0.01 **	0.01 ***	0.02+	0.02 **	0.02 ***	0.01 ***	0.01 ***	0.02 **
	(2.24)	(3.06)	(3.61)	(1.75)	(3.05)	(3.34)	(4.08)	(3.39)	(2.94)
Religion	0.37	0.29	0.22	0.44	0.10	0.004	0.06	0.12	0.1
	(1.21)	(1.67)	(1.51)	(0.67)	(0.26)	(0.01)	(0.35)	(0.73)	(0.29)
Ethnic	−0.37	−0.32	−0.29	−0.48	−0.28	−0.23	−0.02	−0.035	−0.23
	(−1.37)	(−1.63)	(−1.54)	(−0.84)	(−0.80)	(−0.67)	(−0.09)	(−0.18)	(−0.62)
Trade Index	−0.002	−0.002	−0.002	0.003	0.002	0.002	−0.004	0.002	0.01
	(−0.39)	(−0.46)	(−0.51)	(0.30)	(0.29)	(0.27)	(−1.05)	(0.46)	(0.86)
Economic Institutions								−1.229 *	−2.200 +
								(−2.32)	(−1.89)
Constant	3.100	2.707 **	2.414 ***	13.73 ***	12.10 ***	11.63 ***	2.056 **	2.267 **	10.67 ***
	(1.92)	(3.28)	(3.39)	(4.23)	(8.39)	(8.88)	(2.72)	(3.06)	(7.05)
Countries	76	76	76	76	76	76	76	76	76
Adj. R-sq.	0.29	0.39	0.43	0.16	0.42	0.45	0.43	0.46	0.41

Notes: t statistics are in parentheses. + p < 0.1; * p < 0.05; ** p < 0.01; *** p < 0.001.

shows results for the 2SLS estimations. For the specification in column (1), when we used Corr-WB for the average for years 2002–2015 as the dependent variable and RPE for the average of years 1960–1969 as the instrumental variable, RPE for 2002–2015 was insignificant, although it had the expected sign. One reason can be the weak correlation between the RPE for 1960–1969 and RPE for 2002–2015, which was approximately r = 0.23. Except for per capita income and the rents variable, which were significant at the 1% and 5% significance level, respectively, all the other control variables were insignificant. For the specification in column (2), when we used Corr-WB for 2002–2015 as the dependent variable and RPE for 1970–1979 as the instrumental variable, RPE for 2002–2015 was significant at the 10% significance level and had the expected sign. Again, except for per capita income, which was significant at the 0.1% significance level this time, all the other controls variables were insignificant. For the specification in column (3), when we used Corr-WB for 2002–2015 as the dependent variable and RPE for 1980–1989 as the instrumental variable, RPE for 2002–2015 was significant at the 5% significance level and had the expected sign. Again, except for per capita income and the rents variable, which were significant at the 0.1% and 1% significance levels, respectively, this time, all the other controls variables were insignificant. For the same dependent variable, when we used RPE for 1980–1989 (our primary instrumental variable), the RPE 2002–2015 was significant at 5%. As discussed previously, the correlation between RPE80s and RPE00s was highest among the RPEs for various decades. Similarly, the correlation between RPE80s and the corruption indices (both Corr-WB and Corr-TI) was negligible.

In the next three specifications, column (3) to (6), we used Corr-TI as the dependent variable for the same set of instrumental variables and control variables. Results were more or less identical. RPE 2002–2015 was not significant when we used RPE60s as the instrumental variable, was significant at 10% when RPE70s was used, and was significant at 5% significance level when we used RPE for 1980–1989 as the instrumental variable. Just like the findings in columns (1) to (3), all the variables, except GDPpc and Rents, were insignificant.

For the specifications in columns (7) and (8), we used the average of Corr-WB for 2016–2018 as the dependent variable. RPE 2002–2015, instrumented by RPE 1980–1989, was not significant, as shown in column (7). However, it turned significant at 10% when we entered the economic institutions variable into the model, as presented in column (8). The magnitude of the coefficient for RPE increased in the case where Corr-TI formed the dependent variable for the same set of controls and instrumental variables, as presented in column (9). In both the models, the economic institutions variable, GDPpc, and the rents variable remained significant with their intuitive signs.

In summary, 2SLS estimations showed a weak influence of RPE on corruption as compared to the OLS estimations. However, when instrumented by RPE for 1980–1989, the effect of RPE was still significant at the 5% significance level. Unlike the OLS estimations, which showed a strong effect of RPE averaged for 2002–2015 on corruption indices averaged for 2016–2018, 2SLS estimations showed a strong effect of RPE for the same years’ corruption indices. Another striking result was that the democracy variable remained insignificant in all of the nine specifications, although the sign was as expected.

7. Conclusions

This article argues that developing countries with a strong political capacity have lower levels of corruption. Despite the ubiquity of literature on corruption, the relationship between a state’s political capacity and corruption has not been addressed by the existing academic literature. To measure the political capacity of a country, the authors have used relative political capacity (RPC), an indicator that gauges the effectiveness of governance by its ability to meet or exceed their expected extractive capabilities and its ability to implement a set of policy choices. On the one hand, politically capable and stable governments are in a position to pursue their political and economic goals, such as reducing corruption. On the other hand, a strong political capacity provides them with the opportunity for rent-seeking and corruption. This implies that a state’s strong political capacity can be either a ‘boon’ or ‘bane’ to implement a set of desired policy goals. Based on this assertion, the authors tested the hypothesis of whether a strong relative political capacity increases or reduces the level of corruption.

Controlling corruption is an important policy goal for most of the governments in developing democracies, but delivering policy goals is costly. In particular, reforming the civil service, improving the quality of design and implementation, and creating an accountability mechanism hinges on the financial stress as well as on political conditions. As the relative political capacity reflects not only the states’ extractive capabilities, but their political and economic stability as well, the authors found that countries with higher revenue extraction capabilities were the ones to control corruption.

Based on their findings, the authors suggest that democracies, in themselves, are partially endowed with the capacities necessary for maintaining government legitimacy and that low government legitimacy fosters corruption, whereas strong government capacity inhibits corruption.

This comparative analysis of 98 developing countries using ordinary least-squares and two-stage least squares methods, with lags as instrumental variables, supports the authors’ hypotheses using different measures of corruption (the World Bank’s Control of Corruption Index and the Transparency International’s Corruption Perceptions Index). The explanatory power of political capacity is at least as important as conventionally accepted causes of corruption, such as economic development, and democracy.

Numerous studies have recognized the importance of good governance in mitigating corruption. However, they have failed to incorporate a comprehensive measure of governance to capture its effect on corruption. This paper contributes to our understanding of the governance and corruption in developing countries by considering a comprehensive measure of governance, the relative political extraction, which not only captures a country’s capacity to extract resources from its population, but also its capacity to implement important policy goals. The findings are in consistence with the theoretical arguments that the strong governance capacity of a country is a key factor in the fight against corruption. Furthermore, the magnitude of its effect is stronger than various other determinants of corruption, such as per capita income, democracy, rents from natural resources, and such like factors.

Further, the paper has re-examined the relationship between the quality of institutions and corruption. In doing so, however, the aim was also to find an objective measure of governance and political performance that is aggregate yet comprehensive, reflecting a government’s overall capacity and performance to pursue policy goals, both economic and political. The findings suggest that a higher relative political capacity in previous years has higher effects on mitigating corruption.