1 Introduction

The concept of corruption involves cultural, legal, philosophical, and economic aspects. Accordingly, the definition of corruption can vary depending on the research carried out, depending on the perception and subjectivity of the topic. For Tomaszewski (2018), industrialised and democratic countries have a different approach and perspective than other economies and cultures. However, in this study, we assume that corruption is the act of public agents in infringement of the laws and norms established to serve private interests, to the detriment of society.

In the economic context, a key question is whether high levels of corruption can be partly responsible for the slowdown in economic growth. Furthermore, does the size of government interfere with the level or growth of economic activity? Is there a transmission channel for corruption to affect the economy?

Despite important findings, there is no consensus in the literature regarding the determinants of corruption and its impact on economic activity. In this sense, corruption and its consequences have received considerable attention from international organisations and governments around the world, such as the development of governance indicators and specific laws and regulations to define and guide the conduct of public and private agents (Bação et al. 2019).

Different international organisations, such as the United Nations (UN), have increased their efforts to combat corruption through different programmes. For the UN, corruption not only distorts the decision of individuals and firms, but also constrains investment, inhibits competition, and hinders economic growth. For the World Bank (WB), the identification and control of corruption have assumed priority status in the political and institutional environment and for the need for reform over the last two decades.

Despite these arguments, there is no clear convergence of results regarding the effects of corruption on economic activity. Different studies present two hypotheses to understand the relationship between corruption and economic growth. The first hypothesis describes corruption as an obstacle to economic growth. This hypothesis is described as “sanding the wheels” (Aidt 2009; Hoinaru et. al. 2020; Nur-tegin and Jakee 2020).

The other hypothesis shows a positive connection between corruption and economic growth: the “greasing the wheels” hypothesis. This positive relationship occurs in situations where there is excessive bureaucracy and inefficiencies which hinder the development of new businesses. Corruption is therefore seen as the “second-best solution”, due to the distortions caused by the malfunctioning of public institutions. In other words, corruption can enhance economic growth when economic agents pay bribes to circumvent bureaucracy. Fisman and Gatti (2006) analyse the formation of transactions, involving acts of illegality between bureaucrats and entrepreneurs.

Using a panel of 48 developed and developing countries for the period of 2012–2019,Footnote 1 our study focuses on the response of economic activity to the effect of corruption, as measured by the corruption perception index (CPI). In addition, we investigate whether the size of the government interferes with this relationship.

Our study thus contributes to the literature in several different ways: (i) by offering new findings and filling a gap by reviewing the effects of corruption on economic activity, as many studies use the CPI before the methodological change (2012) and/or combining the data before and after this period. However, the scores are not comparable before 2012 (Gründler and Potrafke 2019), and hence, (ii) we analyse the effect of corruption (CPI) on the level and growth of economic activity, controlling for different configurations.

First, based on dynamic models, we study the effect of corruption on the level and growth of GDP per capita (pc) for the full sample. Subsequently, we classify the countries into medium–big and small governments. Accordingly, countries with an average annual index (2012–2019) below the 33% lowest results are classified as small, otherwise, medium, and large governments. In this way, we examine whether the size of the government matters. Second, we deepen the research and investigate whether the level of economic development makes countries susceptible to the effects of corruption. For this purpose, we initially use the full sample and then control for the level of economic development and government size. Third, for a full sample and, controlling for the size of governments and also the level of development, we examine whether the effectiveness of public services influences the effect of corruption on the level and growth of GDP pc. Finally, we analyse whether private investment is a potential transmission channel for corruption.

Our results indicate an adverse effect of corruption on the level and growth of GDP per capita. Furthermore, developing economies, regardless of government size, benefit less from reducing corruption, while government size is not sufficient to explain the influence of corruption on economic activity, although the level of effectiveness of public services is crucial. Lastly, we find that private investment declines with increasing corruption and can be a potential channel of transmission of corruption.

The remainder of the paper is organised as follows. Section 2 presents the literature review. Section 3 presents the data and estimation strategy. Section 4 presents the empirical analysis, and Sect. 5 concludes.

2 Literature

Since the late 1990s, governments have included combating corruption as one of their government goals, highlighting a phenomenon, which for a long time has not received much attention. In a similar vein, Abreu (2011) points out that economies have experienced an increase in events where corruption has been evident, after decades of flexibility and modernisation of expanded markets on a global scale.

The literature provides at least two main reasons for associating government size with economic inefficiency and corruption. First, large governments can hinder economic growth. Some researchers point out that the size of the government is negatively related to GDP growth, as they tend to consume resources from the economy without producing significant effects, while other studies indicate that there is an optimum level for the size of the government (Di Mateo, 2013; Dzumashev, 2014, Afonso and Schuknecht 2019, and Afonso et al. 2020a, b).

Second, governments with greater participation in the economy can be inefficient, and by interfering in different areas, they compete with and exclude private activities, which are more efficient in the opinion of some specialists. Such a scenario thus creates an environment, which is conducive to corruption and inefficiencies in economic activity. Although many experts defend free markets as being the mechanism for allocating resources, Di Mateo (2013) argues that a considerable proportion of economic inputs is defined outside the market environment and is therefore influenced by government institutions. For the author, the decision is not a trivial one, but there is a need to investigate why, when, and where resource allocation should take place. Once these questions are answered, we can then proceed to define the participation and role of government in economies.

Due to the multidisciplinary nature of the question, scholars highlight the complexity and challenges which arise from the different perspectives of corruption. Hayashi (2012) highlights three dimensions of this phenomenon: legalist, mercantilist, and the concept of a public good. For Bobbio et al. (1998), an aspect of corruption is associated with a transaction or deal between at least two agents. There must be the corruptor and the corrupted, usually with the offer of a promise that favours the interests of the corruptor. The United Nations (UN) clearly defines the activities which it considers to be directly linked to corruption, such as bribery, fraud, embezzlement, nepotism, extortion, and the use of inside information by a public agent for private benefit (Hayashi 2012).

These illegal activities affect the economy, the welfare, and the health of the population. Achim et al. (2020) investigated the association between corruption and the health of the population of 185 countries (high-income and low-income). They found evidence that corruption affects both physical health and mental health. Furthermore, the results indicate high levels of corruption affect the physical health (life expectancy and mortality rate) of individuals in low-income economies more intensely than in high-income countries. Conversely, corruption has more intense effects on mental health (happiness) in high-income countries. In addition to highlighting the relevance of the theme, these findings can help to improve national and international policies, considering the culture, economic level, and the other features of each country.

Regarding the causal chain of events in economic activity, Mo (2001) and Pulock (2010) emphasise that the embezzlement of public funds acts as barriers to the entry of new firms into the markets, which consequently affects innovation and reduces productivity and economic growth. Ahmad et al. (2012) stress that over the last 30 years, different studies point to corruption as being a factor, which is capable of changing the goals of public institutions to benefit agents and private institutions. They emphasise that corruption can also inhibit investments, which in turn makes public administration more expensive for society.

Mauro (1995) also suggests that corruption is negatively associated with economic growth and highlights that the direction of causality starts from corruption, rather than the other way around. This finding contradicts some arguments that there would be a “boomerang effect”, or that corruption can influence and be influenced by other variables.

Heckelman and Powell (2010) indicate that the effects of corruption depend on the degree of development of national institutions. Using a different approach, based on regression analysis, and controlling the variable economic freedom, they argue that corruption fosters economic growth when economic freedom is restricted. On the other hand, as economic freedom is more present and robust, this positive effect decreases significantly.

The empirical literature provides important results notably for developed countries, such as the USA and the European Union, although some are not conclusive. To expand the sample and bring new evidence, researchers have examined whether the effect of corruption varies in different regions and whether a country can affect its neighbours through the “spillover effect” of corruption. D’Agostino et al. (2016b) analysed African nations and found that a negative correlation between corruption and economic and economic activity. In a similar vein, Hoinaru et al. (2020) investigate how corruption and the shadow economy affect economic and sustainable development. Using a large database (185 countries) between 2005 and 2015, the authors provide empirical evidence of the harmful impact of corruption on the shadow economy and economic development in countries around the world, confirming the “sand the wheels” perspective. Similarly, Achim and Borlea (2020), Kirchler (2007), and Schneider and Buehn (2018) provide important findings on corruption, money laundering, and the shadow economy.

These findings are in line with the most widely observed argument in newspapers and morally accepted and disseminated by governments and international institutions. Accordingly, these results support the hypothesis “sanding the wheels”, or that corruption slows down innovations, distorting the economic system, and it is, therefore, harmful to economic growth.

However, some studies point to opposite conclusions—where corruption is the driving force of economic growth (Leff, 1964). Kaufman and Wei (2000) considered that corruption can have a lubricating effect on economic gears in certain circumstances, which supports the “greasing the wheels” hypothesis. Their findings support the argument that corruption is beneficial to the economy, due to the excessive set of regulations, rules, and bureaucracies that render the system inefficient. In this sense, corruption acts to overcome these obstacles and stimulate economic growth. Huang (2016) studies the causality between corruption and economic growth in Asia–Pacific countries. For most of the countries studied, there is no clear evidence of causality between corruption and economic growth. The results suggest that the use of anti-corruption policies to promote economic development may be ineffective. In addition, the author indicates that the “grease the wheels” perspective is supported for South Korea.

As stressed in corruption studies, the shadow economy can be seen as a way to encourage economic activity ("grease the wheels") and circumvent bureaucracy and government controls. De Soto (1989) describes the challenges of the Peruvian economy and a government with an insignificant presence, where the shadow economy organises its own market system.

On the other hand, there are studies (Hoinaru et al. 2020) that point to negative and destructive effects ("sand the wheels") and the consequences of these crimes on economic growth (Schneider 2021). However, measuring the size of the shadow economy, as well as corruption, is not a trivial task, but recent studies (Medina and Schneider, 2018) discuss methods and approaches and they manage to measure the shadow economy, evaluating a large sample (158 countries) over 1991 to 2015.

There is still the possibility of a hybrid behaviour between corruption and economic growth. Some economists argue that the relationship between variables has an inverted U shape, such as in countries in the early stages of economic and social development—where there is no fertile environment for corruption. During the intermediate stages of development, the opportunities for corruption increase, whereas corruption declines again in the case of a highly developed society (Ahmad et al. 2012). In addition, Méndez and Sepúlveda (2006) study the effects of corruption on growth using measures of political freedom. Controlling some economic variables in a group of countries with political freedom, the authors point to a non-monotonic relationship between corruption and growth.

As highlighted, there is no conclusive evidence about the effects of corruption, as aspects such as the degree of governance, economic development, culture or whether the size of the government exceeds a maximum scale can interfere with the efficiency of spending (Dzhumashev 2014a, b; Dzhumashev, 2016). In this sense, Blackburn and Forgues-Puccio (2009) point out that the phenomenon of corruption does not have a single shape, or that it spreads in social relations in a single way and, therefore, the effects on growth performance are not homogeneous. Thus, the authors examine the problem from a different perspective, that is, they assess why the phenomenon of corruption has different formats across countries. They emphasise that one possibility is the level of organisation of the corruption network within each country. Therefore, this evidence serves as a warning for anti-corruption policies and that it is necessary to examine the nature of corruption before indicating a single, general solution.

In addition, a frequent association is that governments with high participation in the economy (“Big Governments”) are bureaucratic and inefficient, and are therefore, a breeding ground for corruption. Some authors (Egger and Winner 2005; Dzumashev 2014) argue that corruption improves economic efficiency when the size of the government is above the ideal level. The choice of government size is a delicate issue and requires a sensitive balance. For Alesina and Angeletos (2005), the choice creates an impasse for policymakers, whereby small governments do not correct inequalities in the economic system, while a large government increases corruption. They also point out that public spending for low-income agents in developing countries is often misdirected, increasing corruption.

In contrast, Kotera et al. (2012) identify a positive association between the size of the government and corruption, for democratic countries. Empirical papers have indicated that large governments can increase participation in the economy and still reduce corruption, as they have a system of checks and balances (Billger and Goel 2009; La Porta et al. 1999).

Bobbio et al. (1998) advocate that the greater the scope of institutionalisation, the higher the chances for the emergence of corruption. This is the cornerstone of the argument of the authors, who point out that greater government participation relative to the private sector is harmful. However, Bobbio et al. (1998) also indicate that this is not a sufficient condition, as the pace of expansion as well as the social and cultural characteristics and the maturity of the institutions are crucial factors. Accordingly, corruption is less prevalent in countries with institutional stability than in unstable societies, as these tend to have less robust and established institutions. In this sense, Méon and Sekkat (2005) take a step further and analyse the quality of governance associated with corruption. For these authors, the influence of corruption on economic growth is negative (statistically significantly) in countries with low-quality political institutions. In the same line, Aidt et al. (2008) investigate the connections between corruption and economic growth and how institutional quality in different governance regimes affects this relationship. Through a theoretical model, the authors highlight the role of institutions and show that the specific characteristics of the regime in each country are crucial. Despite the difficulty of measuring institutional quality, the authors use the Voice and Accountability indexFootnote 2 (World Bank) and the findings indicate that corruption has little or no impact on economic growth, when institutions are fragile, highlighting the "grease the wheels" perspective. On the other hand, in regimes with high-quality political institutions, the effects of corruption on the growth of economic activity are negative. Many studies investigate the effects of corruption and its determinants, but fail to define which channels are used by corruption to affect the economy and the welfare of society. Zakharov (2019) examined the relationship between corruption and fixed capital investment in Russian regions. The author argues that corruption slows economic growth through different channels and identifies domestic investment in physical capital as the main channel. In addition, corruption fosters uncertainty (Mauro, 1995), which makes agents more cautious and leads them to postpone or reduce investment. This effect spreads across different sectors, resulting in a reduction in the economic activity (Baker et al. 2016; Bernanke 1993).

Along the same line, Ahmed and Alamdar (2018) measured the impact of corruption and the budget deficit on private investment in the Pakistani economy. They point out that corruption has an adverse and significant effect on private investment and stress the importance of the transmission channel in developing economies.

Table A2 depicts other studies that confirm the findings described above and provides new information on the effects (positive and negative) of corruption on economic activity. We highlight the isolated effect of corruption and the interaction with public spending, especially military spending in developed countries and in less frequent samples, such as Peru, African countries, and post-communist countries.

The lack of convergence and imprecision of the results is not only due to the complexity and multifaceted nature of corruption, but also to the data and methods used. The first issue is that many studies use the World Bank’s Corruption Perception Index (CPI) and ignore the technical recommendation of the World Bank. The problem is that the 2011 CPI scores are not comparable with the 2012 CPI scores, as the methodology used before 2012 means that the CPI scores are not comparable over time.Footnote 3 The second issue concerns panel models with fixed effects, where Gründler and Potrafke (2019) identify that “in particular, including fixed period effects in panel data models does not solve the incomparability problem because the CPI in individual years before the year 2012 included data for different components and time periods to measure perceived corruption across continents”.

In addition to the problems observed with the CPI, we must also be careful when using the Control of Corruption Index (World Bank) because it this index has been criticised on account of various methodological issues and the incomparability of a sub-index over time and across countries, as many country classifications come from different sources of information (Gründler and Potrafke 2019).

3 Estimation strategy and data

3.1 The magnitude of corruption

The literature provides different approaches to measure corruption, with one of the measures used as a proxy for the control of corruption being the one developed by the World Bank. In addition, the Corruption Perception Index (CPI) is widely used, which was developed by Transparency International and has the advantages of counting on a wide coverage and applies a consistent methodology for cross-country studies. An alternative strategy is to observe corruption directly, as highlighted by Olken and Pande (2012), although these authors clarify that their coverage is restricted and is more specific.

We employ the CPI to measure the effect of corruption on economic activity in this paper. CPI classifies more than 100 countries by perceived levels of corruption in the public sector, ranging from 0 to 100 (0 is perceived as being most corrupt).

3.2 Government size

There are different alternatives for classifying countries by size of government. To establish the presence or participation of the government in the economy, Afonso et al. (2005) divided the countries into three sizes (small, medium, and big), according to the public expenditure-to-GDP ratio (G/GDP). Small governments have a G/GDP ratio of less than 40%, medium ones less than 40%, and big governments more than 50%. The final consumption of the general government can be split into two different categories. To assess the effects of governments’ footprint on the economy, we use two classifications available on the World Bank database, namely (1) General Government Final Consumption (% GDP) and 2) Expenses (% GDP).

We choose these two metrics or indices because they present different aspects of government consumption. The General Government Final Consumption index represents the individual and collective services provided by the government and includes the remuneration of public servants, the final consumption of government goods, and services expenditures on national defence, but it excludes the part of government capital formation. When evaluating this indicator, which does not include expenses such as interest and pensions, we avoid classifying a government as “Big” just because it incorporates interest payments or invests heavily in social projects. On the other hand, and to ensure a broader perspective, the Expenses index includes social benefits, interest and subsidies, grants, and rent and dividends.

In this paper, we divided the sample of the two indicators into two groups: small and medium–big governments. Accordingly, governments with an average annual index (2012–2019) below the 33% lowest results are classified as small, while values for medium–big governments are above 33%. Regardless of the measure used for government size, the countries analysed have the same classification (small or medium–big) between 2012 and 2019.

To investigate the influences of corruption on economic activity, we select 48 developed and emerging countries. Due to the lack of recent data for some countries, we choose a sample that provided a good diversity of levels of economic activity and CPI scores, using corruption indexes from the World Bank database. Due to the change in methodology of the CPI (2012), the series just range from 2012 to 2019 and used the results of the comparable CPI over time.

The data are provided by the World Bank and Organisation for Economic Cooperation and Development. The series are based on CPI, GDP per capita (international dollar 2017), gross fixed capital formation per capita (GFC pc) (constant 2010 US$),Footnote 4 and labour force participation rate (% of the total population aged 15–64), which we call “Labour per capita” (L pc). We also use the Government Effectiveness Index (GEFF) to assess the quality of public service, as well as General Government Final Consumption expenditure (% of GDP) and Expenses (% of GDP) to rank the government size.

Tables 1 and 2 and Figure A1 present the averages (GDP per capita and CPI) for the 48 countries, as well as the descriptive statistics and the relation between the variables. We observe that there is a positive relationship between the reduction in the corruption (increase in the CPI) and an increase in the level of GDP per capita (see Figure A1).

Table 1 CPI and GDP per capita (2012–2019)
Full size table
Table 2 Descriptive statistics: individual samples (2012–2019)
Full size table
Fig. 1
figure 1

Source: Authors’ calculations and the World Bank

GDP per capita and corruption perception index (CPI).

Full size image

In addition, Annex Figure A2 illustrates the amplitude of the Corruption Perception Index per country throughout the period under analysis.

To investigate the transmission channel of corruption, we use the data provided by the International Monetary Fund (IMF) for the Investment Stock and Capital Stock Data (1960–2015). We use the corruption perception index (CPI) with private investment (gross fixed capital formation) and GDP, both in billions of international dollars in 2011. Table 3 indicates the average value of private investment and GDP for the 48 countries (2012–2015).Footnote 5

Table 3 Private investment and GDP (2012–2015)
Full size table

4 Empirical analysis

4.1 Model specification

The GMM approach enables us to incorporate a certain superiority of the dynamic estimators, in comparison with the static estimators, and it also controls the endogeneity of the lagged dependent variable in a dynamic model, especially when we identify a correlation between the explanatory variables and the error term. In addition, GMM controls omitted variable bias and unobserved panel heterogeneity.

The empirical literature on dynamic models (GMM) tends to use first differences transformation (FD), which is attributed in part to the results of Arellano and Bond (1991). Later on, Arellano and Bover (1995) presented a transformation (forward orthogonal deviations or FOD) as an alternative to the first difference transformation.

On the other hand, Phillips (2019) argues that, initially, there would be no reason to worry about the transformation technique and that the results indicate that two different transformations can lead to the same generalised method of moments (GMM) estimator. However, the same author also points out that in situations where the estimators based on these two transformations differ, the simulations suggest that the estimators obtained by FOD have better properties than those obtained by FD. Hayakawa (2009) suggests a similar result, indicating that, in many simulations, the FOD-GMM estimator performs better than the DF-GMM model. As an additional advantage for studies of panel models with gaps, Roodman (2009) highlights that the use of orthogonal deviations maximises the sample size.

We use different configurations for the GMM model (FOD) panel, as we examine not only the isolated effect of corruption, but we also control for the size of government, the degree of development of countries, and the effectiveness of public management. Therefore, our standard specification of the dynamic model for GDP pc (yit) can be defined as:

$$Y_{{it}} = \alpha Y_{{i,~t - 1}} + X_{{it}} \beta + e_{t} + ~u_{{it}}$$
(1)

where α is a scalar and β is the vector of coefficients (kx1). In this basic structure, Yit is the dependent variable (per capita GDP) and X´it represents the vector of explanatory variables (1xk).Footnote 6 As in different models of panel economic growth (Mankiw et al. 1992; Islam 1995), we include, in addition to the effect of corruption (CPI), two macroeconomic variables, that is, gross capital formation (GFC pc) and labour (L pc) to assess the effect on the level of economic activity and its growth. The subscript i indicates the countries across the time periods (t). The terms uit and et represent a composite error, where the random component of the variation in our independent variable is derived from the idiosyncratic error (uit) and the time-invariant error, et. It is this term that we investigate when we analyse fixed and random effects, and also whether it is correlated with Xit, or not. Finally, we introduce the lagged dependent variable Yi,,t-1 as a determinant for the dynamic panel concept and take advantage of the time series dimension.

Accordingly, the AR(1) α coefficient represents the persistence or memory of the process that affects Yit. In addition, to deal with the issue of endogeneity, we use dynamic models (GMM), and we identified an adequate strategy for instrumental variables; that is, we use models transformed by the forward orthogonal because it has better coefficient properties (in relation to the first differencing) and our models have gaps.

In the basic configuration, we assume that the maximum sample period (t) is equal to 8 years with 48 countries (i). To estimate our models based on the GMM approach, we select the option orthogonal deviations as a transformation method to eliminate the effect of the specification. In addition, the GMM specification is in line with the Arellano-Bond 2-step. Finally, we use the CPI and labour force and gross fixed capital formation lagged variables as instruments (Anderson and Hsiao, 1982).

4.2 Results and discussion

In this section, we examine the effects of corruption on economic activity, as well as the assumptions mentioned in the literature. To gain intuition, we start with an analysis of the average results between 2012 and 2019, and subsequently, we examine the results in the light of dynamic panel models.

The corruption proxy is represented by the Corruption Perception Index (CPI) and it indicates the level at which corruption is perceived by entrepreneurs and analysts, as described in the previous sections. Using GDP pc (income proxy), as measured by the average between the years 2012 and 2019, the results indicate that the higher the level of corruption, the lower the level of economic activity (see Fig. 1). For instance, the Brazilian economy has relatively lower levels of GDP pc and has higher levels of corruption. We also observe that countries such as Portugal (which is situated below the trend line) would expect the level of income (per capita) to be higher.

We initially use three dynamic panel approaches: fixed effects, OLS, and GMM. Although the focus of our study is the GMM approach, we also include the two other approaches to compare results and confirm patterns in the relationship between variables. Unlike the GMM approach, one of the limitations of the other approaches is that they may not necessarily address issues such as the endogeneity of explanatory variables. Despite the inaccuracies in the estimators of these additional approaches, the results point to similarities with the outputs from the GMM model, especially in the case of the fixed effects approach Table 4 presents the results for GDP in level (GDP pc) and its variation–D(GDP pc).

Table 4 Dynamic models (OLS, fixed effects, and GMM)
Full size table

We find that an increase in the corruption (the CPI ranges from 0 to 100, with 100 being the least corrupt) hinders economic activity (both level and growth). The OLS approach does not point to significant results; however, the GMM and the fixed effects approaches support the hypothesis that an increase in the CPI score (reduction in the corruption) stimulates the level of economic activity. The same can be seen for increases in the labour force and gross fixed capital formation. With regard to the growth of GDP pc, only the GMM model indicates significant results. These findings are in line with the “sanding the wheels” hypothesis (Nur-tegin and Jakee 2020).

Moving forward, we use only the GMM approach and two sets of modelsFootnote 7 to examine the effects of corruption on the level and growth of GDP pc. The first model uses the lagged dependent variable, CPI, and labour force pc as explanatory variables. In turn, the second model includes gross fixed capital formation pc in the list of explanatory variables.

After analysing the full sample, we divide the countries according to the government’s spending share of GDP and investigated whether the size of government matters. Table 5 indicates that the participation of the government in the economy interferes with the result (statistically significant). Thus, small governments benefit relatively more from reducing corruption, while countries with larger governments have less benefit. These findings are statistically significant for the level of economic activity and for growth of GDP (Model 2), where the dummy variable for the smaller government size interacts positively with less corruption to foster economic growth.

Table 5 GMM—small and medium–big governments
Full size table

It appears that the hypothesis that countries with large governments are excessively bureaucratic and inefficient can be accepted. In this sense, corruption would be an alternative for agents to overcome these obstacles and stimulate the economy. Nevertheless, this result could be premature, and there is, therefore, a need to evaluate this evidence carefully. Maybe the nature of the problem is not necessarily linked to the size of the government, but rather to the maturity or development of countries and institutions.

To further check whether the size of the government matters, and to carry out more in-depth research, we ask whether, apart from the size of the government, the degree of development of the economies is a relevant factor. Therefore, we split the sample into developed and developing economies, according to the World Economic Situation and Prospects (2012, 2014, 2015, and 2016).Footnote 8

From Table 6, from the subgroup of small governments, it can be seen that the two sets of models used suggest again that the increase in the corruption is harmful to the economic activity (level). In addition, the findings emphasise that in countries with small governments and developed economies, reducing corruption has an additional positive effect on economic growth. The models do not point to robust and definitive evidence with regard to the effect of economic development on GDP pc (level).

Table 6 GMM—small governments, developed, and developing economies
Full size table

Table 7 shows that for large governments, the control and reduction in the corruption also foster economic activity. However, we find no significant evidence that economic development interferes with the effect on per capita GDP.

Table 7 GMM—medium–big governments, developed, and developing economies
Full size table

Our results indicate that corruption is an adverse factor for economic activity, as well as for the growth of GDP pc. In addition, we find that the size of government matters, especially for developing economies.

The question brings up the dilemma presented by Alesina and Angeletos (2005). For small governments can be less corrupt, which thus creates conditions for an increase in GDP, however, they do not always address the different demands of society, as they fail to adequately correct market failures, inequalities, and social imbalances. On the other hand, large governments respond to agents' expectations, but they can incur more bureaucracy and corruption. The authors point out that many policymakers accept the cost of corruption—as it is often the only way to reduce inequalities and generate better conditions for an economically vulnerable population.

Despite the findings, we find no clear evidence that larger governments benefit from corruption because they are less efficient and more bureaucratic. These results are in a similar vein as some papers which suggest the possibility of increasing the size of the government and thus reduce corruption (Kotera et al. 2012; Billger and Goel 2009). In this case, there would be no direct association of the size of the government with ineffective and bureaucratic public management and the consequent increase in the corruption.

If the size of the government does not necessarily lead to the low effectiveness of public management, and therefore to corruption, then what are the correlations between the effectiveness in public management, perception of corruption (CPI), and government size (Gov. size)?

To answer this question, we use another governance indicator, the World Bank Government Effectiveness index (Kraay et al. 2010).Footnote 9

Table 8 shows that the correlation between government size, CPI, and GEFF is around 34% and 38%, respectively (an average of 48 countries between 2002 and 2019). The governance and CPI indexes have strong correlations, while government size does not have such a significant correlation. To better understand how corruption relates to the size of the government and the level of effectiveness of public policies, we investigate whether any pattern exists which provides new evidence. Annex Figures A3, A4, A5 highlight how the effectiveness of public management is related to government size and the CPI, and they support the findings of econometric models. Small and large governments can score high or low for corruption, and thus, the size of public administration does not seem to be a sufficient condition.

Table 8 Correlation matrix: government size, government effectiveness (GEFF), and corruption (CPI)
Full size table

A similar pattern can be seen when examining the public management effectiveness index (GEFF). In addition to not finding clear evidence that smaller governments are more effective, we note that developing countries perform worse for the CPI, which suggests that the level of economic development could be a determining factor.

In addition, for the 48 countries, out of the 10 lowest scores for effectiveness in public management (2002–2019), 8 of these are developing countries. On the other hand, out of the top 10, a total of 9 are developed economies.

Lastly, to assess whether those countries with higher government effectiveness also benefit more from a decrease in the corruption, we classify the models at two levels: low and medium–high. Therefore, scores below 33% of the lowest results of the sample mean (2012–2019) are classified as low GEFF, with the rest being medium–high GEFF.

Confirming the indications presented, the results suggest that countries with low levels of governance (low GEFF) do not achieve the same benefit in terms of economic growth when compared to those that stand out in terms of the quality and effectiveness of public services.

One hypothesis that can be considered is that corruption gains ground in countries that are less effective in public management, although not necessarily in those which have high expenditures relative to GDP. Another hypothesis is that high-income countries benefit from more instruments to increase efficiency and control corruption.

GEFF captures the quality of public services and policy implementation, based on perception or subjective measures which are taken from surveys of firms, households, and specialised analyses produced by different organisations (Kraay et al. 2010). Table 9 highlights the effects of corruption for two groups of countries: low GEFF and medium–high GEFF. CPI captures the impact of corruption in countries with a high perception of the effectiveness of the public sector (base group), while the dummy variable computed by CPI x Low GEFF indicates the differential effect of corruption in countries with low GEFF scores.

Table 9 GMM—government effectiveness: low and medium–high GEFF
Full size table

We observe that the control and reduction in the corruption increase per capita income (level and growth) in countries with a high GEFF score. On the other hand, countries with low performance in public management do not achieve the same effect in terms of economic growth. These findings are in line with the hypothesis that consumers and firms base their decisions on the perception of government performance (Kraay et al. 2010), and accordingly, agents that believe that the system is inefficient and corrupt can postpone or interrupt new investments, which consequently hinders economic growth.

In addition, as suggested by Afonso et al. (2020a, b), we further investigated government spending efficiency with two other objective metrics: total Public Sector Performance (PSP) and Public Sector Performance Opportunity (PSP-OP).

First, we use the Public Sector Performance Opportunity (PSP-OP) indicator, which is derived from performance in areas such as education, administrative, health, and public infrastructure. (PISA scores, life expectancy, and cardiovascular diseases are representative of the indicators used.) Second, we evaluate the PSP indicator, which is computed as the average between PSP-OP and an indicator that evaluates three government functions (Musgravian), namely allocation, distribution, and stabilisation (Afonso et al. 2005).

For this purpose, we create a cross-country panel dataset, covering a sample of 36 countriesFootnote 10 for the period between 2012 and 2017. Tables A3 and A4 (in Appendix) present the results for PSP and PSP-OP, respectively. The results suggest that higher public sector performance has a positive effect on GDP pc (level and growth) for both models.

Looking at country groups, those countries with high performance in public management (above the sample average) tend to perform better in terms of economic development (GDP pc) after they reduce corruption (see Tables A5 and A6 in Appendix). By contrast, countries with low performance in the public sector have worse economic results or do not benefit from the decrease in the corruption.

4.3 Private investment and corruption: the transmission channel

We analyse the effects of corruption in the previous section, whilst controlling for different variables, without examining how the effect is transmitted to economic activity. One of the channels presented in the empirical studies is the negative impact on private investment.

Despite the relevance of the topic, the literature does not provide sufficient evidence regarding those channels or instruments that definitively reveal the relation between corruption and investment. This negative association has been the subject of debates and studies over the years (Mauro 1995); despite being an intuitive topic, the results to date are inconclusive, with some authors indicating that the effect of corruption on investment is not statistically significant (Shaw et al. 2011).

In order to contribute to the empirical literature, we investigate the relationship between corruption and private investments, by comparing different types of econometric models, namely static (OLS and fixed effects) and dynamic (OLS, fixed effects, and GMM) models. As we highlight in the previous sections, the main focus in the literature is on the GMM model; however, the other models confirm and provide new information for our analysis.

As in the previous models, we analyse 48 countries and used the CPI as a proxy for corruption. However, due to the unavailability of more recent data for private investment, our sample only covers the period from 2012 to 2015.

Table 10 highlights the results of the static models (OLS and fixed effects) and indicates problems of serial correlation and the insignificant effect of corruption on private investment (OLS). On the other hand, the fixed effects model is more promising, in that it suggests an adverse influence of corruption on investment.Footnote 11

Table 10 OLS, fixed effects, and GMM. Corruption and private investment
Full size table

Despite providing information and confirming some patterns of behaviour, the OLS and fixed effects (static) models suffer from some issues regarding the quality of the estimators. On the other hand, the fixed effect and GMMFootnote 12 models present interesting results, which indicates the existence of a potential transmission channel.

5 Concluding remarks

The phenomenon of corruption is long-lived and is present in different areas of scientific knowledge, being linked to the philosophical perspective and moral and political degeneration, as well as the effects on the economy and welfare. This study contributes to this debate by investigating the effects of corruption on the economy, as well as its role as a potential transmission channel.

The findings of our study indicate that corruption has a negative effect on the economy—specifically on the level and growth of GDP pc. Our results regarding the impacts of corruption are broadly consistent with the “sanding the wheels” hypothesis.

We also find that the size of the government matters. Large governments register less benefit from reducing corruption than small governments. This isolated finding would support the hypothesis that large governments are bureaucratic, inefficient and that therefore corruption is an option to circumvent such obstacles.

However, this result can obscure the real reason for the effects of corruption and further research indicates that the level of economic development associated with the size of the government can provide another conclusion.

The findings of our research highlight that small governments in developed economies benefit relatively more from reducing corruption than in the case of developing economies. Maybe this result stems from the degree of maturity of institutions in developed countries and/or the fact that they generally have in place more resources to inhibit and control corruption. In addition, large governments can have positive effects after reducing corruption. These results are consistent with some studies, such as that of Kotera et al. (2012).

Consequently, the hypothesis that large governments generate higher bureaucracy, higher inefficiency, and the consequent breeding ground for corruption is not confirmed. The models indicate that countries with low quality of public services have low responses to the growth of GDP, after reducing corruption and that low efficiency in public management can be the main factor which is responsible for generating conditions for corruption to circumvent barriers and that this corruption consequently stimulates economic activity.

Finally, we find that private investment is negatively affected by an increase in the corruption. This is an important finding, which confirms the understanding that corruption slows down innovations and distorts the economic system, and consequently, that it is detrimental to economic performance. This result points to a potential transmission channel, which negatively affects the growth of GDP.

Our study contributes to the growing corpus of research which shows that corruption has a complex character and that it needs to be analysed in the light of not only the size of government but also of the level of economic development and the effectiveness of public policies. In addition, based on dynamic models (GMM—panel data), our study fills a gap in the literature by examining the effects of the CPI on the level and growth of GDP per capita, using different control variables.

Despite its varied contributions, our study has some limitations. An important finding suggests that developing economies benefit from or are not affected by increased corruption. Regardless of the debate on the moral issue, this result needs to be treated with extreme caution, bearing in mind that this paper and many others only partially examine the problem when it comes to only investigating economic growth and that other important factors were not analysed, such as health, happiness, human development, income inequality, and poverty. In this sense, whilst accepting that nowadays a certain level of corruption can lead to a perpetual vicious cycle of inefficiencies and corruption (Alesina and Angeletos 2005) which could even have a positive effect on economic growth, it must be stressed that corruption can bring a high cost to society in the future.

In this sense, future research should consider alternative methods to measure economic performance, in addition to investigating other indexes which are capable of classifying government size, such as the number of public servants per capita. With regard to the transmission channel of corruption, it would be interesting to use a larger sample and to disaggregate private investment in different sectors. Another suggestion is to analyse the effects of corruption based on the PVAR approach (Bação et al. 2019), which could contribute substantially to understanding the implications of corruption on economic variables.

Our paper points out to the negative relationship between corruption and economic growth in a sample of economies with international representativeness. However, some aspects such as the relationship between trust in public institutions and other elements of social capital and the size/efficiency of the government can be investigated with alternative approaches. One possibility is to develop an objective indicator using data envelopment analysis (DEA) and widely publicised economic indicators, which allows to expand the sample (time and space). Therefore, a potential future research involves examining the effects of different indicators of social capital (objective and perception measures) for different types of efficiency and sizes of government. Thus, future work could explore further the effects on economic activity, considering the heterogeneity of countries, for example, considering income inequality. One option would be to use the pooled mean group estimator (Pesaran and Smith 1995; Pesaran et al. 1999).