The Impact of Corruption On Economic Growth: A Nonlinear Evidence

Several cross-country studies have found that corruption slows growth, but these ndings are not universally robust. Therefore, the questions to be addressed are to what extent corruption can be tolerated and at what threshold it has a detrimental effect on an economy. This article investigates the impact of corruption on economic growth by testing the hypothesis that the relationship between these two variables is nonlinear. In this article, a panel data analysis has been used to examine 65 countries over the 1987 to 2011 period. Our ndings are that corruption can have a positive effect on growth. The results indicate that beyond an optimal threshold, both high and low corruption levels can decrease economic growth. Under this optimal threshold, a moderate level of corruption, dened by the point of reversal of the curve of the marginal corruption effect on growth, could have advantages for economic growth.


Introduction
Empirical literature in the eld has consistently reported a negative correlation between economic growth and corruption. These studies have shown that developed countries are known by low corruption levels and a relatively high growth rate (Cooper & al, 2006), and by contrast most developing countries are known by high poverty and corruption levels (Chetwynd & al, 2003;Umbreen and Saadat, 2015).
The novelty of the empirical contribution is that we estimate a non-linear growth model that allows for threshold effects. To this end, we will use the method proposed by Beck and Katz (1995) who suggested estimating linear models of time-series-cross-section (TSCS) data by ordinary least squares (OLS). For this, they proposed the panel-corrected standard errors model (PCSE).
The paper is structured as follows: section 1 presents a review of both the theoretical and empirical literature; section 2 presents the econometric model and the main results followed by a discussion of the ndings in the nal section.

Literature Review
The theoretical and empirical literature on corruption has generated a rich debate over the last 40 years.
This literature can be summarized in two opposing theories. The rst assumes that corruption "lubricates the economic cycle" or "greases the economic wheel" and produces the most e cient economies ( Mauro (1995) detects a weak statistical signi cance between corruption and economic growth. However, this signi cance disappears once investment rate is introduced in the model. Mo (2001) nds that corruption negatively affects economic growth. However, the additional introduction of variables like investment to GDP ratio, political stability and human capital weakens or eliminates the signi cance of this negative impact. Aidt et al. (2008) show that the impact of corruption on economic growth depends on institutional quality. Trabelsi & Trabelsi (2020) show that beyond an optimal threshold, both high and low corruption levels can decrease economic growth. Under this optimal threshold, a moderate level of corruption, de ned by the point of reversal of the curve of the marginal corruption effect on growth, could have advantages for economic growth.
All these studies indicate that corruption may have either positive or negative effects on economic growth, making the issue ambiguous and con rming the non-linearity of the relationship between corruption and growth. However, one must ask to what extent can corruption be tolerated and from what threshold would it become destructive to the economy. The questioning is motivated by the fact that studies don't test whether there is a growth-enhancing or growth-reducing level of corruption and not one study thoroughly identi ed the corruption level that will allow an optimal growth.

Description of data
Corruption is not the only factor that affects economic growth (Barro, 1991and Brunetti, 1997, Lambsdorff, 1999. Other control variables are also relevant (Fernando & al, 2016). According to theory and on the basis of arguments cited in the literature, we propose economic growth depends mainly on investment, in ation and trade openness.
The study is based on a panel data set over the period 1987-2011 for 65 countries taken from the World Development Indicators (Growth rate, Foreign direct investment, In ation & Trade). The ICRG index has been obtained from the Quality of Government Institute, the Transparency International and International Country Risk Guide published by Political Risk Services group. It measures the risk involved in corruption rather than the perceived level of corruption.
The descriptive analysis for the full set of 65 countries appear in Table 1. It shows that average economic growth is 3.68% with an average corruption index of 3.32. . This is a quadratic function based on the hypothesis that the impact of corruption on growth is not always negative and that a moderate corruption level could have advantages for economic growth.
In order to verify this, a cross-sectional framework is used in which growth rate and the ICRG index are observed only once for each country. The scatter plot ( gure below), using the tted Kernel curve illustrates and con rms the hypothesis that the relationship between corruption and economic growth ( tted values) is nonlinear. The curve is clearly increasing in the middle range of corruption and decreasing where corruption is least and most.
Past studies have used a panel of 5-year averages and the system GMM estimator because this choice reduces, in general, short run uctuations and resolves the endogeneity due to time invariant effects; but this method will not address endogeneity due to the possible interactions between higher growth rates and greater resources to combat corruption, or other time varying effects. Levin & Satarov (2000) and Paldam (2002) have presented evidence for the existence of both types of endogeneities.
Recently, the empirical studies characterized by having repeated observations over time on some countries are resolved by others models. In this study, we will follow the Beck and Katz (1995) methodology who suggested estimating linear models of time-series-cross-section (TSCS) data by ordinary least squares (OLS) and they proposed the panel-corrected standard errors (PCSE) estimator.
The results for GDP growth using the PCSE estimator are reported in Table 2.
It can be seen that corruption negatively affects (-0.9967357) economic growth unlike the square coe cient of corruption which positively affects (0.1793806) economic growth. The signi cance of Icrg² coe cient con rms the non-linearity of this model and shows the presence of a threshold above which there will be a change of sign.

Results And Discussion
The concave function ( Figure) may be interpreted in the following way. Corruption, that facilitates tax evasion, has two types of effects in economics. It offers households an opportunity of tax savings that can be consumed or invested, as tax evasion leads to a transfer of public resources to private agents (Tanzi &Davoodi, 2000 andCerqueti &Coppier, 2011). This could improve growth up to a certain threshold. The optimal threshold represents the reversal point of the curve; otherwise, the country may suffer underdevelopment like several countries immersed in corruption.
This corruption, if signi cant, will reduce state resources because of productive public spending which will lead to a loss in economic growth, which sooner or later will lead to an uprising calling for establishing democratic principles and good governance.
This result may surprise those who advocate the negative effects of corruption but it can be explained by the fact that administrative delays resulting from absence of "bribes" paid in a corrupt economy may dampen economic growth and reduce economic development.

Conclusion
The aim of this paper is to examine the impact of corruption on economic growth. The empirical literature that reported a linear relationship between corruption and economic development failed to differentiate between growth-enhancing and growth-reducing levels of corruption.
In our study, we have presented evidence that suggests the existence of hump-shaped relationship between corruption and growth, which shows the existence of a non-linear relationship between these two variables. This non-linear results show that growth increases at middle-corruption and decreases as nations'achieve higher level of governance (low corruption). In other words, the results indicate that higher or lower levels of corruption negatively affect growth. Minimum corruption can be bene cial to economic growth. This con rms some theories that assume that corruption "lubricates the economic cycle" and produces the most e cient economies. However, this lubricating effect has a threshold beyond which it becomes a threat to economic growth. Conversely, lack of corruption may be a mechanism that slows down growth.