Corruption perception indices: a comparative analysis.

Ahmad, Naved

1. INTRODUCTION

The empirical literature on corruption has used data on corruption from three different sources: (i) investigative reports, (ii) newspapers, and (iii) surveys or questionnaire-based data. Some studies on corruption are based on case studies and newspaper reports. Studies by Wedeman (1997); Wade (1982) and Alam (1996) fall in this category. While these studies have presented an in-depth analysis of corruption, they do not examine a large sample of countries. Moreover, the investigative reports require detective work and sometimes connections with people in high echelons in order to expose corruption. Unlike investigative reports, access to survey data on corruption enables researchers to study corruption for a large sample of countries, but at the same time, raises questions about their subjectivity) However, the subjectivity of these indices is often justified on the ground that corruption is illegal in nature, and hard to measure directly.

Empirical studies on the causes of corruption after the mid-1990s have used several corruption indices from Business International, International Country Risk Guide, Peter Neumann and his collaborators at Impulse, Transparency International, and World Competitiveness Report. More than one corruption index has been used in most of these empirical studies. For example, Ades and Di Tella (1997) used corruption indices from WCR and Peter Neumann and his collaborators at Impulse (1994). Treisman (2000) recently utilised four corruption indices: three from Transparency International (1996 to 1998) and one from Business International. While empirical literature on the causes of corruption using these indices continues to surge, it is imperative carefully to examine what exactly these indices portray.

I have selected four sources: (i) World Competitiveness Report WCR (1990, 1992, 1994, 1996), (ii) Transparency International TI (1995 to 1998), (iii) International Country Risk Guide ICRG (1982 to 1995), and (iv) the World Bank WB (1996) data to examine this issue. The following section presents various sources of corruption indices. The analysis will be performed in two phases. In Section 2, I present a comparative analysis based on rank correlation across various sources and over time. To strengthen my analysis, Section 3 presents a regression analysis to show whether these indices produce similar results for a common set of countries. There are only 20 countries for which the data are available from these sources. I run regressions on a common set of independent variables encompassing variables from winners' and losers' side. The main purpose of these regressions is carefully to examine the relationship between the corruption indices and all the independent variables. This empirical exercise will help determine the association between the corruption indices and the independent variables. The final section concludes.

2. SOURCES OF CORRUPTION INDICES

The data on corruption and other risk factors are now available for a large number of countries. These indices are sold to banks, multinational companies and international investors. The first set of data comes from Business International (BI), now incorporated into The Economist Intelligence Unit. The survey asks about "the degree to which business transactions involve corruption or questionable payment". The assessment reports are completed by the staff members of Business International working in the survey countries. The replies are examined at Business International's regional and corporate headquarters. The survey covers 68 countries for 1980-1983 period (one observation per country for the entire period). The index ranges from 0 (corrupt) to 10 (clean).

The second set of corruption data comes from World Competitiveness Report (WCR), which is published by the Institute for Management Development. The report compares and ranks countries on 224 criteria under 8 categories. These eight categories are domestic economy, internationalisation, government, finance, infrastructure, management, science and technology, and people. In this survey, one of the questions asks top and middle management about the prevalence improper practices, such as bribery or corruption in the public sphere. The exact question about corruption varies slightly from year to year. In 1989, the question involves "the extent to which the country prevents corruption". In 1990, the question dealt with "the extent to which government regulations prevent improper practices in the public sphere" and in 1991, and 1992, the question addressed "the extent to which improper practices such as bribing and corruption do not prevail in the public sphere".

The index ranges from 0 (corrupt) to 10 (clean). For the four years, 1993-96, the survey included 37, 45, 48, and 46 countries, respectively. One advantage of the WCR data over the BI data is that the WCR data contain responses of people who have intimate knowledge of the business practices in each of the surveyed countries.

In a similar fashion, in 1997, the Political and Economic Risk Consultancy (PERC) in Hong Kong surveyed 280 expatriate business executives in 12 Asian countries, asking "to what extent does corruption exist in the country in which you are posted in a way that detracts from the business environment for foreign companies?" The index ranges from 1, which represents a situation in which corruption does not exist, to 10 which represents a completely corrupt environment.

Unlike the BI and WCR, the indices constructed by DRI/McGraw-Hill Global Risk Service and International Country Risk Guide (ICRG) were based on assessments made by their staff members after an in-depth country analysis and discussion, although this method is somewhat less transparent to outsiders. The DRI/McGraw Hill Global Risk Service covers 105 countries for 1995. Unlike the BI index, the ICRG index is prepared annually. It covers 1982 to 1995 and, depending on the year is available for 88 to 129 countries.

The most interesting contribution to the corruption index is the Gallup International Survey for 1997. On average, 800 individuals from the general population are interviewed in 44 countries, either in person or on the telephone. Question 5 of the "global" portion of the questionnaire is as follows: "From the following groups of people, can you tell me for each of them, if there are a lot of cases of corruption, many cases of corruption, few cases, or no cases of corruption at all. The groups listed afterwards were "politicians," "trade unionists," "public officials," "policemen," "businessmen," "judges," "ordinary citizens," "clergy/priests," and "journalists." For each country the replies to the 5 categories ("a lot," "many," "few," "none" and "no answer") were aggregated, yielding data by categories.

Similar to the Gallup International data, Guttingen University conducted a survey via the Internet to generate a corruption perception index (Internet Corruption Perception Index). Between January and June 1997, Internet users with interest in the topic of corruption were asked to complete an interactive questionnaire. There were 246 responses to the questionnaire. Internet users were asked the following question: "You enter a public office which is authorised to grant licenses and permits (e.g. the license to conduct business). After you have waited for a long time you are expected to pay a bribe and are told that otherwise you will not receive the license. According to your perception, in which countries may this (i.e. the asking for bribes by public officials) happen? On the other hand, where do you consider it to be unlikely?" Three choices, "often," "sometimes," and "rarely," were listed at the end of the question.

The indices discussed are based on the perception of heterogeneous respondents with different nationalities. Peter Neumann and his collaborators at Impulse, a German business firm, conducted a survey in 1994 in which only German businessmen (exporters) were interviewed who were normally involved in the trade with each of the countries. They were asked to indicate the number of deals that involve corrupt payments and the estimate of kickback per deal as a percentage of the deal value. One hundred and three countries were included in the survey. These indices are less subjective because of the nature of the questions asked. Each respondent had to give an estimate of the kickback per deal. Moreover, the respondents were taken from a homogeneous group of people (German exporters) with practical business experience in each country.

Transparency International and Guttingen University combined several corruption indices to produce a composite index of corruption. This index is available for 1995 to 1998. In 1997, an index (TI-Corruption Index) was compiled for 52 countries from 7 sources. The seven sources were two surveys from the Institute for Management Development in Lausanne, Switzerland (World Competitiveness Report), one from the Political and Economic Risk Consultancy Ltd. in Hong Kong (Asian Intelligence Issue No. 482), one by Gallup International, two assessments by DRI/McGraw-Hill (Global Risk Service) and the Political Risk Services, East Syracuse, New York (International Country Risk Guide), and a survey conducted at Guttingen University via the Internet (Internet Corruption Perception Index). From Gallup International survey only 4 groups "politicians," "public officials," "policemen" and "judges," were used in this index because the other groups mentioned in the Gallup International survey do not fit the definition of corruption as the misuse of public power for private benefits. The index is a "poll of polls". Recently Transparency International has released corruption rankings for 90 countries for the year 2000.

There are four statistics given for each country. The first is its position in the TI-index. The second figure is the overall integrity ranking (out of 10). Ten stands for a highly clean country, while zero is for a country where business transactions are entirely dominated by kickbacks, extortion, and bribery. No country scored a ten or a zero. The third figure indicates the number of surveys in which the particular country has been included (from 4 to 7). The fourth figure indicates the variance of the different sources. The minimum number of surveys in which a particular country is included was reduced from 4 to 3 in 1998. The 1997 index has an average variance that is almost one third lower than in 1996, making it far more reliable. A higher variance indicates a higher degree Of deviating opinions, with some respondents placing the country much higher and others much lower on the overall scale.

All the sources of corruption indices discussed above focus on perceptions of foreign firms only. These indices represent either the perception of staff members (external viewpoint) or the perception of people working in organisations located in those countries (internal viewpoint). Unlike these sources, the World Bank conducted a worldwide survey of the private sector. The survey questionnaire measured the uncertainty of government policies. Section three of the questionnaire asks the degree to which corruption is problematic for doing business. The survey dealt with both domestic and foreign business firms. The indices are available for 67 countries for 1996 and range from 1 to 6.

Several important characteristics of the World Bank data on corruption make these indices very useful. First, unlike all other corruption indices, the World Bank data include the perception of people who work in domestic firms i.e., firms with no foreign participation. Second, the data are also available for specific departments such as the police, customs, and the judiciary. Third, this data set provides measures of corruption from several perspectives. For example one question asks respondents to rate how corruption is problematic in doing business. The other question asks respondents to rate on a scale of 1 to 6 whether these businesses accept bribes. Still another question asks about the pervasiveness of bribery.

In the following section, I will present correlation coefficients among the corruption indices discussed in this section.

3. CORRUPTION INDICES AND RANK CORRELATION

A few recent survey studies on corruption have documented various sources of corruption indices. The explanation of these corruption indices has already been discussed in Section 2. Nevertheless, I will briefly mention the types of corruption reflected by these indices. Virtually all sources define corruption as an abuse of public office for private benefit. These indices reflect the behaviour of public officials and politicians. While these indices theoretically define corruption in a same fashion, they do not guarantee that the rankings they generate are consistent. This section investigates whether these rankings produce consistent results and whether they are consistent over time.

The rank correlation coefficients are given in Table 1. The correlation coefficients between ICRG95 and WCR96 (0.80) and between WCR96 and WB96 (0.82) are higher than the correlation between ICRG95 and WB96 (0.56). The high correlation between WCR96 and ICRG95 may be due to the fact that both indices focused on firms or businesses engaged in foreign activities. On the other hand, the correlation between WCR96 and WB96 is higher because these indices represent the internal viewpoint about corruption. In contrast, the correlation coefficient between WB96 and ICRG95 is only 0.56. One plausible reason for this low correlation is that the ICRG95 corruption indices include foreign firms with external viewpoints about corruption, whereas WB96 corruption indices represent internal viewpoints with concentration in local business firms.

Although the above indices evidence a close association among themselves, the question is whether they rank countries consistently over time. To ascertain this, rank correlation coefficients of the various corruption indices from the same sources, but for the different time periods have been calculated. The rank correlation for ICRG, WCR, and TI corruption indices are presented in Tables 2, 3, and 4 respectively.

The indices are highly correlated between any two consecutive surveys. It is evident from these tables that the value of correlation coefficient decreases as the time span between the two indices expands. Notice that the minimum value of correlation coefficient in Tables 2, 3 and 4 is 0.68 [ICRG (1982) and ICRG (1995)], which is high. This demonstrates that the corruption rankings are persistent over time.

These rank correlation coefficients produce very close estimates of how these corruption rankings are correlated over time. Furthermore, I categorised countries into three groups: namely clean (0-2), partly corrupt (2-7), and corrupt (7-10) (2). Tables 5, 6, and 7 present the changes in corruption rankings for WCR (1994 to 1996), TI (1997-1998), and ICRG (1982 to 1995) respectively. Table 5 reveals that only 8 countries out of 43 have succeeded in moving from one category to another. The results are mixed when the changes within groups were considered. Fifteen countries increased their rankings, whereas rankings of fourteen countries have gone down. Six countries did not change rankings. Most of the countries whose rankings had increased or decreased were European countries. It is evident from Table 5 that very few countries have succeeded in moving from one category to another. As far as increase (or decrease) in corruption rankings within categories is concerned, the results are mixed.

I observed similar results in changes in corruption rankings using Transparency International indices between 1997 and 1998. Four countries out of 52 were successful in changing their categories. Sixteen countries increased their rankings, whereas rankings of nineteen countries have gone down. Thirteen countries did not change rankings. The data in Table 6 indicate that many countries do change their rankings within categories. This may be because corruption rankings might be affected with certain sudden changes in government policies. These changes do not cause the public to overlook government corruption altogether, but they do have some impact on corruption rankings.

The above analysis revealed that few countries have managed to move from one category to another, suggesting the entrenched nature of corruption. To determine whether corruption exerts a persistent nature over a longer period of time I have, using corruption indices from ICRG, categorised countries into three groups: (1) clean (0-2), (2) partly corrupt (2-7), and (3) corrupt (7-10).

Of the 87 countries for which data are available for 1982 to 1995, 53 countries remain corrupt (partly corrupt or clean), suggesting persistency in corruption rankings. For a corrupt country, it is difficult to change people's perception about corruption because it takes time for the government to make people believe that they are sincere in their efforts to mitigate corruption. In other words, perceptions about corruption exhibit a self-generating property. As Tirole (1996) and Tanzi (1994) argue, corruption is likely to be found in countries where corruption existed for some time as compare to countries where corruption is relatively new.

Table 7 confirms that perceptions about corruption do not change quickly. It shows that only three European countries improved their ranking from one category to another between 1982 to 1995. Similarly, out of 21 Latin American countries, only 3 have achieved improvements from one category to another.

4. CORRUPTION PERCEPTION INDICES: A REGRESSION ANALYSIS

Although the above analysis suggests that these corruption indices are highly correlated and stable over time, investigators continued using at least two indices to support their hypotheses. To widen my analysis, I have run regressions on the same set of independent variables using several corruption indices for a common set of countries.

A corrupt transaction always creates winners and, almost, always losers. The current empirical literature on causes of corruption has, however, failed to analyse corruption from winners' and losers' perspectives. In this analysis, I have included factors from winners' and losers' point of view.

From winners' side, I have included two factors that generate rent, i.e., the size of the government, and an index of regulations. The share of government consumption in GDP measures government size. It can be intuitively argued that large governments create large bureaucracies, which, in turn, provides more opportunities for graft. I expect a positive relationship between government size and corruption. Husted (1990) has used this measure as a determinant of corruption. The index of government regulation measures the extent of regulations imposed by government regarding business operations, price controls, foreign trade (exports and imports), labour regulations, foreign currency regulations, tax regulations, and safety and environmental regulations. This index is a sum of seven indices. These seven indices are (i) regulations for starting business and new operations, (ii) price controls, (iii) regulations on foreign trade, (iv) labour regulations, (v) foreign currency regulations, (vi) tax regulations and/or high taxes, and (vii) safety or environmental regulations. I expect a positive relationship between the composite index of regulations and corruption.

From the losers' side, I have included an index of bureaucratic competition, newspaper circulation (level of information), urbanisation, average years of schooling, index of political liberty. (3) All these variables enhance the ability of losers to take countervailing actions against corrupt officials. The index of bureaucratic competition measures the extent to which people can obtain fair treatment by avoiding corrupt officials. Level of information is another factor that helps people take countervailing actions. The print media keep people well-informed about public officials who are misusing their powers. I have used newspaper circulation per 1000 persons as a measure of the level of information. Urbanisation deters corruption. People living in urban areas can raise their voice against corruption through the platform of various organisations such as association of producers, traders, importers, and consumers. These associations will increase the effectiveness of direct countervailing actions. (4) Urbanisation is defined as the share of urban population in the total population. I expect a negative relationship between urbanisation and corruption. Educational level is likely to enhance the awareness of the people about their rights and increase their ability to fight against their losses from corruption. Average years of schooling at age 15 and above in the total population measures the level of education. Finally, democracy can also increase the effectiveness of countervailing actions. An index of political rights is used as a measure of democracy. In democratic societies, the weak bargaining power of public officials because of decentralised political power may enable losers to resist corrupt officials with ease.

The dependent variable, corruption index, is taken from various sources. I have included three indices from Transparency International (1996 to 1998), one from the International Country Risk Guide (1995), one from the World Competitiveness Reports (1996), and one from the World Bank (1996). These indices cover a wide range of perceptions of corruption, including internal and external viewpoints and domestic and foreign firms' perspectives. Ordinary least square technique is used to estimate the coefficients of independent variables. Table 8 presents the regression results.

An examination of the regression results leads to several convincing arguments. First, in all regressions, these independent variables explain more than 50 percent of variations in corruption, Second, with one exception, the relationship between the independent variables and the corruption index is consistent across all sources. The coefficient for newspaper circulation is positive only in the case of ICRG (1995). Third, the coefficient of the measure of government regulations has expected positive sign in all cases. Finally, bureaucratic competition, the average years of schooling, and an index of political liberty have the expected negative signs in all cases, whereas the coefficients of urbanisation and government size have wrong signs in all cases.

5. CONCLUDING REMARKS

An attempt has been made to examine various corruption indices that are used in current empirical literature, especially with regard to the causes of corruption. The primary purpose of this analysis is to demonstrate that these corruption indices not only produce similar results, but that the results are consistent over time. I approach this task by first presenting rank correlation coefficients among these indices and then categorising countries into three groups to analyse their rankings over time. Finally, I regress these indices on the same set of independent variables for a common set of countries. The results reveal that these indices are correlated among each other and are stable over time. In addition to the rank correlation, the regression results confirm that these indices yield similar results. Thus, using any one of these sources would be sufficient for the determination of the causes of corruption. However, the regression results must be considered with caution because these results may be influenced by the small sample size. Nonetheless, this analysis has shed some light on the consistency of the corruption indices across various sources and over time.

Appendices

APPENDIX A

WORLD COMPETITIVENESS REPORT (WCR) 1990

Clean: Singapore, Denmark, New Zealand, Netherlands, United Kingdom.

Partly Corrupt: Hong Kong, Switzerland, Finland, Sweden, Germany, Canada, Norway, Ireland, Australia, United States, France, Belgium, Luxembourg, Japan, Austria, Malaysia, Taiwan, Portugal, Turkey, Mexico, Spain, Thailand.

Corrupt: India, Greece, Hungary, Indonesia, Italy, Brazil.

WORLD COMPETITIVENESS REPORTS (WCR) 1992

Clean: Singapore, New Zealand, Finland, Denmark, Sweden, Netherlands, Switzerland, Canada, United Kingdom, Australia.

Partly Corrupt: Germany, Ireland, Norway, Austria, United States, Hong Kong, Japan, France, Belgium, Luxembourg, Hungary, Taiwan, Portugal, Malaysia, Turkey, Spain, Greece, Mexico.

Corrupt: Thailand, Italy, India, Brazil, Indonesia.

WORLD COMPETITIVENESS REPORT (WCR) 1994

Clean: New Zealand, Denmark, Singapore, Sweden, Finland, Canada, Ireland, Australia, Norway, Switzerland, United Kingdom.

Partly Corrupt: Netherlands, Germany, United States, Austria, France, Hong Kong, Portugal, Malaysia, Belgium, Luxembourg, Japan, Spain, Mexico, Taiwan, Greece, Turkey.

Corrupt: Hungary, Thailand, Indonesia, India, Italy, Brazil.

WORLD COMPETITIVENESS REPORT (WCR) 1996

Clean: Denmark, Singapore, Finland, New Zealand, Norway, Australia, Ireland, Canada, Sweden, United Kingdom, Netherlands.

Partly Corrupt: Switzerland, Luxembourg, United States, Germany, Austria, Hong Kong, Japan, France, Belgium, Portugal, Taiwan, Malaysia, Hungary, Spain, Turkey, Italy.

Corrupt: Greece, Brazil, Thailand, Indonesia, India, Mexico.

TRANSPARENCY INTERNATIONAL 1995

Clean: New Zealand, Denmark, Singapore, Finland, Sweden, Canada, Australia, Switzerland, Netherlands, Norway, Ireland, United Kingdom, Germany.

Partly Corrupt: Chile, United States, Austria, Hong Kong, France, Belgium, Japan, South Africa, Portugal, Malaysia, Argentina, Taiwan, Spain, Hungary, Turkey, Greece, Colombia, Mexico.

Corrupt: Italy, Thailand, India, Philippines, Brazil, Venezuela, Pakistan, China, Indonesia.

TRANSPARENCY INTERNATIONAL 1996

Clean: New Zealand, Denmark, Sweden, Finland, Canada, Norway, Singapore, Switzerland, Netherlands, Australia, Ireland, United Kingdom, Germany.

Partly Corrupt: United States, Austria, Japan, Hong Kong, Belgium, Chile, France, Portugal, South Africa, Malaysia, Greece, Taiwan, Hungary, Spain.

Corrupt: Brazil, Colombia, Philippines, Indonesia, India, Venezuela, China, Pakistan.

TRANSPARENCNY INTERNATIONAL 1997

Clean: Denmark, Finland, Sweden, New Zealand, Canada, Netherlands, Norway, Australia, Singapore, Switzerland, Ireland, Germany, United Kingdom.

Partly Corrupt: United States, Austria, Hong Kong, Portugal, France, Japan, Chile, Spain, Greece, Belgium, Hungary, Italy, Taiwan, Malaysia, South Africa, Brazil, Turkey, Thailand, Philippines.

Corrupt: China, Argentina, Venezuela, India, Indonesia, Mexico, Pakistan, Colombia.

(1) See Fisman and Gatti (1999); Husled (1999); Mauro (1995); Rijckeghem and Weder (1997); Tanzi (1998) and Tanzi and Davoodi (1997).

(2) The list of countries and their categorisation are presented in Appendix A.

(3) See Alam (1995).

(4) The discussant, Mr Daniyal Aziz, argued during the general discussion that urbanisation fosters conditions that are conducive to corruption. The same argument was put forward by Meier and Holbrook (1992) who argue that in an urban environment, family and religion lose their social control, which, in turn, reduces the ability of losers to take countervailing actions against corruption. The investigators measured corruption as the number of public officials convicted of corruption charges. The problem with this measure is that it may be a proxy for the effectiveness of the reporting system and cannot take care of those corrupt activities that are not reported, as discussed in Knack and Keefer (1995). It is likely that urbanisation has provided more opportunity to uncover corruption charges as compared to rural areas, and therefore it is unclear whether urbanisation induces public officials to become involved in corruption or if it helps to uncover already corrupt officials.

Naved Ahmad is Assistant Professor of Economics, at the Institute of Business Administration, University of Karachi, Karachi.

Author's Acknowledgements: I am thankful to Prof. M. Shahid Alam, Prof. John Adams, and Prof. Alan Dyer for their helpful suggestions. I bear the sole responsibility for any remaining errors.

Table 1

Rank Correlation:

(WCR96, ICRG95, TI96, WB96)

 WCR96 ICRG95 T196 WB96

WCR96 1.000
ICRG95 0.8035 (44) 1.000
T196 0.9644 (43) .8739 (53) 1.000
WB96 0.8224 (22) .5612 (47) .8385 (29) 1.000

Note: Figures in parenthesis are the number of observations.
WCR = World Competitiveness Report, ICRG = International Country
Risk Guide, TI = Transparency International, and WB = World Bank.

Table 2

Rank Correlation:

(World Competitiveness Repor7 1990, 1992,1994, 1996)

 WCR90 WCR92 WCR94 WCR96

WCR90 1.000
WCR92 .9378 (34) 1.000
WCR94 .9093 (34) .9514 (36) 1.000
WCR96 .8892 (33) .9482 (35) .9386 (43) 1.000

Figures in parenthesis are number of observations.

Table 3

Rank Correlation:

(Transparency International 1995, 1996, 1997, 1998)

 T195 T196 T197 T198
T195 1.000
T196 .9784 (40) 1.000
T197 .9274 (41) .9594 (47) 1.000
T198 .9478 (41) .9559 (53) .9813 (52) 1.000

Figures in parenthesis are number of obsen at ions.

Table 4

Rank Correlation:

International Country Risk Guide (ICRG) 1982-1995

 Icrg82 Icrg83 Icrg84 Icrg85

Icrg82 1.000
Icrg83 .9819 (88) 1.000
Icrg84 .9179 (88) .9624 (90) 1.000
Icrg85 .8748 (88) .9131 (90) .9579 (111) 1.000
Icrg86 .8713 (88) .9124 (90) .9545 (111) .9959 (129)
Icrg87 .8582 (88) .8988 (90) .9420 (111) .9871 (129)
Icrg88 .8336 (88) 8707 (90) .9208 (111) .9724 (129)
Icrg89 .7837 (88) 8194 (90) .8730 (111) .9370 (129)
Icrg90 .7824 (88) .8143 (90) .8488 (111) .9132 (129)
Icrg91 .7877 (88) .8135 (90) .8430 (111) .8866 (128)
Icrg92 .7602 (88) .7805 (90) .8060 (111) .8020 (128)
Icrg93 .7149 (88) .7415 (90) .7734 (111) .7448 (128)
Icrg94 .7162 (88) .7435 (90) .7752 (111) .7446 (128)
Icrg95 .6899 (88) .7165 (90) .7418 (111) .7113 (128)

 Icrg89 Icrg90 Icrg91 Icrg92

Icrg89 1.000
Icrg90 .9789 (129) 1.000
Icrg91 .9442 (128) .9672 (129) 1.000
Icrg92 .8294 (128) .8499 (129) .9082 (129) 1.000
Icrg93 .7619 (128) .7759 (129) .8358 (129) .9358 (129)
Icrg94 .7570 (128) .7719 (129) .8339 (129) .9317 (129)
Icrg95 .7142 (128) .7304 (129) .7940 (129) .8883 (129)

 Icrg86 Icrg87 Icrg88

Icrg82
Icrg83
Icrg84
Icrg85
Icrg86 1.000
Icrg87 .99 (129) 1.000
Icrg88 .9715 (129) .9829 (129) 1.000
Icrg89 .9357 (129) .9464 (129) .9660 (129)
Icrg90 .9116 (129) .9211 (129) .9391 (129)
Icrg91 .8850 (128) .8922 (128) .9089 (129)
Icrg92 .7994 (128) .8048 (128) .8229 (128)
Icrg93 .7437 (128) .7473 (128) .7643 (128)
Icrg94 .7432 (128) .7465 (128) .7643 (128)
Icrg95 .7063 (128) .7138 (128) .7315 (128)

 Icrg93 Icrg94 Icrg95

Icrg89
Icrg90
Icrg91
Icrg92
Icrg93 1.000
Icrg94 .9968 (130) 1.000
Icrg95 .9621 (130) .9763 (130) 1.000

Figures in parenthesis are the number of observations.

Table 5
Changes in Corruption Rankings (WCR: 1994, 1996)
Changes 1994 to 1996
(One Category to Another)

 Improve Decrease Total

Africa/Middle East
Asia/Pacific Region 1 1
Europe 4 2 6
Latin America/Caribbean
North America 1 1
Total 5 3 8

(Within Categories)

 Improve Decrease Unchanged Total

Africa/Middle East 1 1
Asia/Pacific Region 6 5 1 12
Europe 8 6 2 16
Latin America/Caribbean 1 2 1 4
North America 1 1 2
Total 15 14 6 35

Table 6

Changes in Corruption Rankings (TI: 1997-1998)
Changes 1997 to 1998

(One Category to Another)

 Improve Decrease Total
Africa/Middle East
Asia/Pacific Region 1 1
Europe 1 1
Latin America/Caribbean 1 1
North America 1 1
Total 3 1 4

Table 6

Changes in Corruption Rankings (TI: 1997-1998)
Changes 1997 to 1998
(One Category to Another)

 Improve Decrease Total

Africa/Middle East
Asia/Pacific Region 1 1
Europe 1 1
Latin America/Caribbean 1 1
North America 1 1
Total 3 1 4

(Within Categories)

 Improve Decrease Unchanged Total

Africa/Middle East 1 1 2
Asia/Pacific Region 8 6 14
Europe 3 11 8 22
Latin America/Caribbean 4 2 2 8
North America 2 2
Total 16 19 13 48

Table 7

Changes in Corruption Rankings (ICRG: 1982-1995)
Changes 1982 to 1995
(One Category to Another)

 Africa/ Asia/
 Middle Pacific Europe
Unchanged
 Clean 1 7 14
 Less Corrupt 7 2 1
 Corrupt 2
 Less corrupt to clean to 1 1
 less corrupt
 Less corrupt to corrupt to less
 corrupt
 Sub-total 11 9 16
Improvement
 Less corrupt to clean 2 1 3
 Corrupt to less corrupt 4 5
 Corrupt to clean 1
 Corrupt to clean to less corrupt I
 Corrupt to less corrupt to clean 4
 Less corrupt to corrupt to less
 corrupt to clean 1
 Corrupt to less corrupt to clean
 to less corrupt 1
 Sub-total 14 6 3
Decrease
 Clean to less corrupt 1 1
 Less corrupt to corrupt 2
 Sub-total 3 1
 Total 28 16 19

 NA LA/CAR Total
Unchanged
 Clean 2 1 25
 Less Corrupt 1 11 22
 Corrupt 2
 Less corrupt to clean to 1 3
 less corrupt
 Less corrupt to corrupt to less
 corrupt 1 1
 Sub-total 3 14 53
Improvement
 Less corrupt to clean 1 7
 Corrupt to less corrupt 3 12
 Corrupt to clean 1
 Corrupt to clean to less corrupt 1
 Corrupt to less corrupt to clean 4
 Less corrupt to corrupt to less
 corrupt to clean 1
 Corrupt to less corrupt to clean
 to less corrupt 1
 Sub-total 4 27
Decrease
 Clean to less corrupt 3 5
 Less corrupt to corrupt 2
 Sub-total 3 7
 Total 3 21 87

Table 8

Regression Results: WB96, T198, T197, T196, WCR96, and 1CRC95

Independent Variables WB96 TI98 T197

Regulations 1.66 *** 1.05 *** 0.92 ***
Bureaucratic Competition 0.38 ** -0.60 ** -0.56 **
Average Years of Schooling -0.044 -0.098 0.083
Government Consumption -0.063 0.15 ** 0.13 **
Urbanisation 0.010 0.009 0.011
Political Liberty -0.34 -0.57 ** -0.74 ***
Newspaper Circulation -0.003 * 0.001 0.001
Adjusted [R.sup.2] 0.78 0.74 0.76
Number of Observations 20 20 20

Independent Variables T196 WCR96 ICRG95

Regulations 0.68 * 1.09 ** 0.57 **
Bureaucratic Competition -0.55 * 0.65* -0.53 **
Average Years of Schooling -0.18 -0.11 -0.16
Government Consumption -0.17 ** -0.21 ** -0.08 **
Urbanisation 0.1127 0.012 0.018
Political Liberty -0.32 0.39 0.47 **
Newspaper Circulation -0.003 -0.002 0.0009
Adjusted [R.sup.2] 0.65 0.60 0.56
Number of Observations 20 20 20

* 10 percent level of significance.

** 5 percent level of significance.

*** 1 percent level of significance.

Results are adjusted for heteroscadasticity.

The 20 countries are Austria, Canada, Colombia, France, Germany,
Hungary, Ireland, Italy, Malaysia, Mexico, Poland, Portugal, South
Africa, Spain, Switzerland, Turkey, U.K, U.S.A, USSR, and Venezuela.