| ||||||||||||||||||||||||||||||||||||||||||||||||
This article appears in the following Oxford Review of Economic Policy issue: THE POLITICAL ECONOMY OF DEVELOPMENT [View the issue table of contents]
Corruption, institutions, and economic development
* Faculty of Economics, University of Cambridge, e-mail: toke.aidt{at}econ.cam.ac.uk
 |
Abstract |
|---|
Many scholarly articles on corruption give the impression that the world is populated by two types of people: the ‘sanders’ and the ‘greasers’. The ‘sanders’ believe that corruption is an obstacle to development, while the ‘greasers’ believe that corruption can (in some cases) foster development. This paper takes a critical look at these positions. It concludes that the evidence supporting the ‘greasing the wheels hypothesis’ is very weak and shows that there is no correlation between a new measure of managers’ actual experience with corruption and GDP growth. Instead, the paper uncovers a strong negative correlation between growth in genuine wealth per capita—a direct measure of sustainable development—and corruption. While corruption may have little average effect on the growth rate of GDP per capita, it is a likely source of unsustainable development.
Key Words: corruption • growth • sustainable development
I would like to thank Partha Dasgupta, Arye Hillman, Pramila Krishnan, Pierre-Guillaume Méon, Vania Sena, and a reviewer for helpful comments and suggestions. Comments from participants in the Centre for the Study of African Economies 2009 Conference on Economic Development in Africa are also much appreciated.
1 See, for example, Murphy et al. (1993) and Mauro (1995).
2 The discussion in this section is based on Aidt (2003). For alternative surveys of the literature, see Bardhan (1997), Jain (2001), and Svensson (2005).
3 See Hillman and Katz (1987) for a formal model of such chains.
4 We assume that the number of talented individuals in the population is larger than this and that 
(
H) – w (H) < 0.
5 See, for example, Nitzan (1994), Congleton et al. (2008), or Hillman (2009, ch. 2).
6 A counter-example in which corrupt activities, including the payment of bribes, were documented in great detail comes from Peru in the 1990s (McMillan and Zoido, 2004).
7 Corruption can also allow agents to circumvent efficient regulation. Bertrand et al. (2007), for example, study corruption in the driving licensing process in Delhi. They find that not only did the average applicant pay more than twice the official price, but many unqualified drivers ended up getting a licence, and they did so because they were willing and able to pay for the privilege.
8 See also Olken (2007) for an interesting field study of corruption in road projects in Indonesia.
9 At the national level several other objective measures of corruption are available. These include data on the number of officials convicted for corruption (see, for example, Alt and Lassen (2003) for a study of US states and Del Monte and Papagni (2001) for a study of Italian regions) and data on the amount of leakage from infrastructure projects in Italian regions (Golden and Picci, 2005).
10 In principle, it is also possible to deal with some of these issues by exploring movements in corruption over time within a country (Méndez and Sepúlveda, 2006). However, the scope for doing this is limited by the fact that country experts often build their perceptions about corruption up gradually over time. This introduces inertia in the corruption indices and it is doubtful how informative the time-series variation really is. It is also a problem that there are inconsistencies over time in the way they are constructed and that the time span for which the indices are available is relatively short (the ICRG index is available from the early 1980s, the TI and WB index are available from the mid-1990s, and the WBES index is available for only one year, 1999–2000). The later observation also implies that researchers trying to explain the growth experience from, say, 1960 to 2000 must assume that corruption at the time it is measured, say in the 1990s, affects this experience.
11 For a detailed discussion of sources, see Aidt et al. (2008).
12 Similar results can be obtained with the other perception-based indices (not reported).
13 The variable is the Voice and Accountability Index from the World Bank's ‘Governance Matters’ database (Kaufmann et al., 2005). This index measures the extent to which citizens of a country are able to participate in the selection of their government and able to hold it accountable for its choices. The index has been re-scaled to lay in the interval 0 (weak institutions) to 1 (strong institutions).
14 The instruments pass the tests for relevance and validity.
15 This variable is also from the World Bank's ‘Governance Matters’ database (Kaufmann et al., 2005). It is representative of the governance variables that Méon and Sekkat (2005) use to measure the quality of institutions.
16 In contrast to my specification, they do not include the governance variable as separate regressors.
17 In a related study, Méon and Weill (2008) study the impact of corruption on aggregate efficiency. They report that corruption reduces efficiency in societies with effective institutions, but improves aggregate efficiency in societies with ineffective institutions.
18 The Freedom House index of political rights codes countries according to several dimensions of political freedom, such as censorship, freedom of association, free elections, etc. The index ranges from 1 (most free) to 7 (least free). I have split the sample using 3.5 as the cut-off.
19 The scaling of the TI index matters greatly for this result. For example, if one inverts the TI index, e.g. by subtracting it from 11 (which is often done in applied work) so that it becomes increasing in corruption, the non-monotonic relationship reported in regression 5 is no longer statistically significant.
20 The regime-specific linear growth models are estimated using instrumental variables. The instruments are the Voice and Accountability Index and the index of ethno-linguistic fractionalization. Similar results obtain with ordinary least squares (OLS) and with alternative instruments. Moreover, there is no evidence of a non-monotonic relationship in any of the two regimes.
21 Restricting attention to the 42 countries for which the WBES index is available has little effect on the results presented in Table 1.
22 I am not the first to recognize this general point within the context of corruption. Gupta et al. (2002), for example, study the impact of corruption on various measures of inequality.
23 See World Development Indicators (various years, Table 3.15). The World Development Indicators use the term ‘genuine saving’. Here, I follow Arrow et al. (2004) and use the term ‘genuine investment’.
24 For details of how to estimate these deductions, see World Development Indicators (various years) or Arrow et al. (2004).
25 They use a ratio of 0.2 for industrialized countries and a ratio of 0.15 for developing and oil-rich countries.
26 I use the same instruments as in the growth regressions.
27 The TI index for India is 2.85, while that for the United Kingdom is 8.65.
28 This is based on the estimated coefficient on the TI index in regression 23 in Table 3.
29 As measured by the Freedom House index of political rights.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  C. Adam and S. Dercon The political economy of development: an assessment Oxf. Rev. Econ. Policy, June 1, 2009; 25(2): 173 - 189. [Abstract] [Full Text] [PDF]
|
|