Is corruption grease, grit or a gamble? Corruption increases variance of productivity across countries

Abstract

The effect of corruption on economic growth has attracted interest in empirical development economics. The conventional view of corruption as impediment for growth has been challenged by the ‘grease-on-the-wheels’ hypothesis. We take a new perspective on the issue and suggest corruption as macro risk, referred to as a ‘gamble’ hypothesis. Using cross-country data and two alternative indicators of corruption, we find corruption to be a significant driver of heteroscedasticity in total productivity. This supports the new gamble hypothesis. We also note some misleading interpretations in the previously published frontier applications. To avoid these shortcomings, we apply a flexible semi-nonparametric estimator.

Keywords:

• productivity
• corruption
• heteroscedasticity
• economic growth
• semi-parametric regression

JEL Classification:

• O47
• O17
• C14

Notes

¹ The terms environmental variables and z-variables are also commonly used in the literature.

² The distinction between productivity and efficiency is worth clarifying. Productivity is defined as the ratio of output to input (or the index of inputs). Output efficiency is defined as the ratio of the observed output and the maximum output. Productivity change can be decomposed into the components of efficiency change and technical progress, and possibly some other components (see, for example, Färe et al., 1994; Kuosmanen and Sipiläinen, 2009).

³ For a more extensive treatment of SFA, we refer to the book by Kumbhakar and Lovell (2000).

⁴ StoNED is an abbreviation of Stochastic semi-Nonparametric Envelopment of Data (Johnson and Kuosmanen, 2011, footnote 3).

⁵ This issue is related to the notion of scaling property (Wang and Schmidt, 2002; Alvarez et al., 2006) in SFA models. The KGM model does not satisfy the scaling property, as Wang and Schmidt (2002) point out.

⁶ Nonparametric DEA applies a similar axiomatic nonparametric frontier, but in contrast to StoNED, DEA neglects the noise term altogether. We consider this a major limitation in the present context. Further, regressing DEA efficiency estimates on contextual variables is known to be problematic even if one assumes away noise (see Simar and Wilson, 2007, for a sharp critique of two-stage DEA). The problems concern particularly the statistical inferences. For these reasons, we do not consider DEA appropriate for estimating the effect of corruption on the level of productivity, let alone its variance.

⁷ In the 1990 data, the CPI index is available for a subset of 48 countries. Using this subset of 48 countries, the results are rather similar than with the larger sample. These results with the sample of 48 countries can be found from the online appendix.

⁸ The causality from exports to productivity is subject to debate. For example, Yamada (1998) does not find strong evidence for the causality, but mentions that the degree of causality may differ across countries (see also Konya, 2004).

⁹ Empirical analysis was conducted using the following software: Stata 11 (OLS and SFA) and GAMS using MINOS solver (StoNED).

¹⁰ Murphy et al. (1991, 1993) suggest rent-seeking and corruption as possible explanations for the misallocation of talent.

¹¹ For example, Gerring and Thacker (2004) argue that centralized political systems reduce the prevalence of corruption largely because fragmented systems create more opportunities for corruption and the path of accountability is often more clearer in a centralized system (see also Tavits, 2007). In contrast, centralized systems have been criticized on the basis of reducing the amount of checks and balances, and competition among bureaucrats, thus inducing more chances on corruption. Thus, any micro-level actions against corruption are ineffective in the decentralized case and the system should be centrally restructured. In contrast, in the centralized case, the benefits of centralization may be questioned. Note that by centralized system, Gerring and Thacker (2004) refer to a system that has a strong national government and which follows parliamentarism. In such systems, the electoral threat should impose some checks on political corruption. On the opposite, one can expect that dictatorial regimes are more likely to exhibit risk-seeking corrupt behaviour than democratically elected governments as dictatorial regimes are not subject to electoral threat (Quinn and Woolley, 2001).