Corruption and Democracy

Abstract

Most models and empirical tests of the relationship between corruption and democracy assume corruption falls as democracy matures. Yet recent theoretical developments and case evidence suggest an inverted U relationship between the two, although so far no one has tested for this relationship. By drawing on a panel data set covering a large number of countries between 1982–1997, substantial empirical support is found for an inverted U relationship between the durability (age) of democracy and corruption. Fortunately, the turning point in corruption occurs rather early in the life of new democracies – between 10–12 years.

Acknowledgements

I would like to thank Sumana Shrestha, Danny Tang and two anonymous references for their assistance and helpful comments on an earlier draft of this paper.

Notes

1. Schneider and Schmitter (2004: 67–68, 84) measure the consolidation of democracy (CoD) in terms of the degree to which elections are free, fair, regular and contested, as well as by political agreements on the rules governing association formation, territorial division of competencies, and ownership of the media. Figure 1 plots the relationship between CoD and a World Bank measure of corruption (Kaufmann et al., 2007). The World Bank measure has been rescaled so that an increase in the variable implies more corruption. The predicted value for corruption in Figure 1 is based on the OLS regression equation: CORRWB = − 9.73 + 6.76 Log (CoD) – 1.07 Log (CoD)². Both regression coefficients are statistically significant at the 0.01 level (t = 3.16 and 3.63) as is the equation F statistic (F = 16.48). Adjusted R² is 0.56.

2. Schneider and Schmitter (2004: 63) label this Dahl's procedural minimum.

3. Linz and Stepan (1996: 10) and Diamond (1999: 111–112) identify the rule of law as one of the critical elements in the consolidation of democracy.

4. Both Linz and Stepan (1996: 10–11) and Diamond (1999: 93–96) stress the importance of effective government. As Linz and Stepan state, ‘Modern democracy … needs … a functioning state and a state bureaucracy considered useable by the new democratic government’ (Linz and Stepan, 1996: 11).

5. Schneider and Schmitter (2004: 68) emphasise the behaviours of political actors; Linz and Stepan (1996: 6) emphasise behaviours and attitudes, while Diamond (1999: 69) emphasises the behaviours, norms and beliefs of political actors. All three emphasise the importance of the time it takes for actors to learn democracy and become habituated to it.

6. What follows draws on Gurgur and Shah (2005: 6–7).

7. This process is the outcome of factional politics within parties (McCargo and Ukrist, 2005: 73–74).

8. Thailand's corruption score rose by 12 per cent between 1996–2000, but fell by 35 per cent thereafter (Kaufmann et al., 2007). An OLS regression of Thailand's corruption score on the duration (DUR = age in years) of democracy and its square yielded the following CORR = 2.80 + 0.03 DUR − 0.009 DUR². The t values for DUR and DUR² are 3.73 and −5.26 respectively. Both are statistically significant at the 0.05 and 0.01 level respectively. Adjusted R² = 0.75.

9. Indonesia's corruption score rose 140 per cent between 1996–2002 before declining by 72 per cent through 2005 (Kaufmann et al., 2007). The OLS regression of corruption on the duration of democracy (DUR) and its square yielded the following: CORR = 3.25 + 0.26 DUR − 0.05 DUR². Both regression coefficients have the expected signs and are statistically significant at the 0.05 level and adjusted R² is 13 per cent.

10. Freedom House measures political rights by asking: ‘Is the government free from pervasive corruption?’ ‘Has the government implemented effective anticorruption laws … to prevent, detect, and punish corruption…? Is the government free from excessive bureaucratic regulations, registration requirements, or other controls that increase opportunities for corruption? Are there independent and effective auditing … bodies that function without … political pressure or influence? Are allegations of corruption by government officials thoroughly investigated and prosecuted without prejudice…? Are allegations of corruption given wide and extensive airing in the media? Do whistleblowers, anticorruption activists, investigators, and journalists enjoy legal protections that make them feel secure about reporting cases of bribery and corruption? What was the latest Transparency International Corruption Perceptions Index score for this country?’ (Freedom House, 2007b).

11. With respect to corruption and civil liberties, Freedom House (2007b) asks two questions: ‘Are bribes or other inducements needed to obtain the necessary legal documents to operate private businesses?’ and ‘Is entrance to institutions of higher education or the ability to obtain employment limited by widespread nepotism and the payment of bribes?’

12. There is no overlap in definition or measurement of autocracy and democracy (Marshall and Jaggers, 2002: 15).

13. Linz and Stepan (1996: 10) argue that in consolidated democracies, all significant actors respect and uphold the law.

14. As Linz and Stepan say, ‘To protect the rights of citizens and … deliver the … services that citizens demand, a democratic government needs to be able to exercise … its claim to the monopoly of the legitimate use of force… … it would have to tax compulsorily. For this it needs a functioning state and a state bureaucracy usable by the new democratic government’ (Linz and Stepan, 1996: 11).

15. Diamond (1999: chapter 5) emphasises how the successful practice and experience of democracy contributes to changes in political culture that are supportive of democracy.

16. Linz and Stepan (1996: 3–7) emphasise how successful experiences in reaching agreement on the forms of democracy contribute to changes in behaviour and attitudes that reinforce support for democracy. They also emphasise how successful experiences with democracy habituate all political actors to resolve political conflict through democracy.

17. Schneider and Schmitter (2004: 85) include time in their measurement of consolidated democracy and argue that the ‘…extent of consolidation is positively related to the amount of time that democratic institutions have been in place.’

18. Treisman (2000) found that the number of years of uninterrupted democracy is negatively correlated with corruption. An OLS regression of the log of the Schneider and Schmitter (2004) consolidated democracy variable (LCoD) on the log of the average number of years of uninterrupted democracy (LADUR) yields the followings: LCoD = 3.23 + 0.45 LADUR. Both the regression coefficient on LADUR (t = 3.69) and the equation F statistic (11.47) are significant at the 0.01 level and adjusted R² = 0.30 suggesting that consolidation of democracy is dependent on time.

19. Knack and Keefer (1995) in IRIS (n.d.) define bureaucratic quality to indicate the degree to which public sector bureaucracies have ‘an established mechanism for recruitment and training,’ have ‘autonomy from political pressure’ and have ‘strength and expertise to govern without drastic changes in policy or interruptions in government services' when governments change.

20. Knack and Keefer (1995) in IRIS (n.d.) define the rule of law to reflect ‘…the degree to which the citizens of a country are willing to accept the established institutions to make and implement laws and adjudicate disputes.’ Higher scores indicate ‘… a strong court system’, while lower scores indicate ‘… a tradition of depending on physical force or illegal means to settle claims.’

21. Figure 2 is based on the following fixed effects panel regression for democracies: BQ = − 1.34 − 0.61 LDUR + 0.35 LDUR² − 0.032 LDUR³ + 0.58 LYN. The t values for LDUR, LDUR² and LDUR³ are −2.40, 2.45 and −1.71. These values are significant at the 0.05, 0.05, and 0.10 levels. Adjusted R² is 0.73 and the equation F is statistically significant at the 0.01 level (F = 133.00).

22. An OLS regression equation of the relative government wage (RGW) from van Rijckeghan and Weder 1997: 40) on real GDP per capita (YN) for a sample of 22 developing countries, between 1982–1994, yields RGW = 0.57 + 0.0001 YN. The regression coefficient (t = 3.71) on YN and the equation F statistic (13.79) are statistically significant at the 0.01 and adjusted R² equals 0.38.

23. Knack and Keefer (1995) in IRIS (n.d.) define ethnic tensions to measure ‘…the degree of tension within a country attributable to racial, nationality, or language divisions. Lower ratings are given to countries where racial and nationality tensions are high… Higher ratings are given to countries where tensions are minimal…’. ETHTEN was rescaled so that a rise in tensions is reflected in a higher score.

24. Huntington (1991: 16–21) identifies two waves of reversal of democracy that were centred in the tropics. Van de Walle (2001: chap. 3) details the emergence and consolidation of authoritarian neo-patrimonial regimes in sub-Saharan Africa. Monga 1997 and Ottaway 1997 have identified the role of ethnic strife in undermining democracy there.

25. The first stage regression on LDUR for Equation 2 in Table 3 is given by LDUR = − 1.96 + 0.01 PROT −0.96 LAT + 0.58 LYN. The t values are 7.42, −4.03, and 19.48 respectively and they are significant at the 0.01 level. Adjusted R² is 0.45. The equation F is 42.46 and it is significant at the 0.01 level. The first stage regression on LDUR² is given by LDUR² = − 14.14 + 0.07 PROT −2.44 LAT + 2.80 LYN. The t value for the regression coefficient on PROT is 8.52, that on LAT is −1.94, and that on LYN is 17.85. All three coefficients are statistically significant at the 0.01, 0.05 and 0.01 level respectively and adjusted R² = 0.45. The equation F is 43.56 and it is significant at the 0.01 level.

26. Country specific fixed effects require estimation of between 74 and 84 country specific coefficients.

27. These panel regression equations are available from the author.