TI 1997 Corruption Perceptions Index
Questions & Answers
The Trends:
Is (e.g.) Nigeria the most corrupt country in the world?
No! Nigeria is perceived by contributors to be the most corrupt in this ranking of 52 countries. Keep in mind that some countries not included here are likely to score worse than Nigeria. Also the perception of corruption must not necessarily reflect the real level of corruption. Also keep in mind that a low performance in the index does not make corruption a justifiable. Comparing the scores for 1996 and 1997 it may also appear that the perceived level of corruption is strongly decreasing. It should be noted however, that the 1997 is much more precise with a variance of only 0.16 points. This means that there is even stronger agreement that Nigeria scores lowest among the 52 countries ranked.
How should your results be interpreted?
There are four figures given for each country. The first is its overall integrity ranking (out of 10). A ten stamds for a highly clean country while zero equals a country where business transactions are entirely dominated by kickbacks, extortion, bribery, etc. No country scores a perfect ten or an outright zero. The second column gives the 1996 score. The third figure indicates the number of surveys in which the particular country has been included (i.e. from 4 to 7). The fourth column indicates the variance of the different sources in 1997. The 1997 CPI has an average variance that is almost a third lower than in 1996, making it far more precise. A high number indicates a high degree of devia-ting opinions with some placing the country much higher and others much lower on the overall scale.
Has worldwide corruption increased or decreased?
Evaluations about worldwide trends are likely to bring about biased opinions. Cultural settings are likely to differ considerably over time and between diffe-rent surveys and differing perceptions may be due to a change in awareness rather than real corruption. Since such trends are difficult to assess they have been neutra-lized in our data. All surveys are normalized to the same mean and variance, hence only their comparative cross-country information is used. This implies that with our data no "world-wide-trend" can be determined. In fact, a worse score of one country may be due to a situation, where all other countries have improved and vice versa.
What does a low score mean for a country?
The CPI is a measure of lost development opportunities. According to a recent study by Professor Shang-Jin Wei of the Harvard School of Government based on the findings of the CPI 1996, a clear link has now been established between the level of corruption and foreign direct investment. The study shows that the difference in corruption levels from that of Singapore to that of Mexico is equivalent to raising the marginal tax rate by over twenty per cent. A one percentage point increase in the marginal tax rate reduces inward foreign direct investment by about five per cent.
The Methodology:
How do you measure the degree of corruption?
The TI-Corruption Perceptions Index is a joint initiative of Transparency International and Göttingen University and is updated once a year. The major operational work is conducted by Dr. Johann Graf Lambsdorff at Göttingen University, Germany. The index is an assessment of the corruption level in 52 countries as perceived by businesspeople, risk analysts and the general public. To the extent that any country has a problem with its ranking, this lies not with the index but rather with the perception that contributors polled apparently have of that country. This may not always be a fair reflection of the state of affairs. We are grateful to all corporations and individuals who supported us with their evaluations.
Which sources did you use?
The index is a "poll of polls". It has been prepared using seven sources, including two surveys from the World Competitiveness Yearbook, Institute for Management Development, Lausanne; one from the Political & Economic Risk Consultancy Ltd, Hong Kong; one by Gallup International; two assessments by DRI/McGraw-Hill Global Risk Service and the Political Risk Services, East Syracuse, NY; plus finally a survey conducted at Göttingen University via internet (http://www.uni-goettingen.de/~uwvw) which gives contributors the possibility for anonymous contributions and also directly approaches employees of multinational firms and institutions.
What does the variance indicate?
The high variance of South Africa of 3.08 implies that 66% of the scores range between 3,3 and 6,8 (this is computed by taking the square root of the variance and by adding and subtracting this value from the average score to obtain the upper and lower bound), likewise 95% of the scores range between 2,55 and 8,55. Apparently, the average score is only to a very limited degree an assessment of the observed degree of corruption. The low variance of Thailand on the other hand indicates, that 95% of the scores range between 2,54 and 3,98. There seems to be a coherent impression of the degree of corruption in Thailand. Deviating scores can on the one hand be due to diverging perceptions regarding what has to be interpreted as corrupt and to different experiences made with respective countries. It can also be due to objective difficulties in assessing the "right" score. In countries where some institutions still resist corruption while others are openly engaging in illegitimite practices an assessment is all the more difficult. A high variance may in this respect also represent a heterogenous state of affairs. The high variance of Belgium and South Korea are noteworthy.After a series of scandals they may represent insecurities in the process of reassessing these countries. On the other hand, the variance of Hong Kong and South Africa may be regarded as insecurities with respect to future developments.
How were your sources assembled?
The 1996 index was the starting point for assembling the new index. Each of the 7 sources has been normalized to the same mean and standard deviation the respective subgroup of countries had in the 1996 index. Hence, the inclusion of a survey which only scores a subgroup of countries affects only the scores between those countries and not the performance of the subgroup in relation to other countries. Each of the seven sources have been assigned the same weight. Accor-ding to the respective quality of the sources, this appeared plausible. However, there is no "objective" weight which can be applied to the sources and a different weighting may be justifiable. With equal weights, the simple average has been calculated from the normalized data. Since taking the average changes the mean and variance of the data, the final results have again been normalized to the same mean and variance the countries were having in the 1996 index. The data by DRI/McGrawHill required a so-called "monotonous linearisation" before inclusion. A second order polynomial has been fitted with the help of regression analysis. The Gallup data required a more sophisticated treatment since the raw data set consists of replies to the categories "A lot", "Many", "Few", and "None", which represents categorical data. For each of the groups of people "politicians", "public officials", "policemen" and "judges" the factor scores have been determined by correspon-dence analyis. Other groups mentioned in the Gallup International survey (e.g. ordinary citizens) have not been included since they do not fit into our definition of corruption as the misuse of public power for private benefits and because their results in the Gallup International survey have been of lower quality. Similarly our own results (Internet Corruption Perceptions Index) had to be analyzed with the help of correspondence analysis.
How reliable is the index in measuring real levels of corruption?
Perceptions may vary randomly with those voicing them. An indicator as to whether perceptions refer to some real world phenomenon can be obtained however, if perceptions are consistently reproducing similar assessments. There-fore, an indicator of the overall performance of the ranking in measuring real levels of corruption can be drawn from the high correlation of the various sources. The following table reports the correlation-coefficients of the sources.
Corr-coeff |
WCY 1997 |
WCY 1996 |
PRS 1997 |
DRI 1997 |
ICPI 1997 |
PERC 1997 |
Gallup 1997 |
---|---|---|---|---|---|---|---|
WCY 1997 |
1,00 |
0,97 |
0,72 |
0,80 |
0,95 |
0,95 |
0,89 |
WCY 1996 |
1,00 |
0,73 |
0,81 |
0,97 |
0,93 |
0,82 |
|
PRS 1997 |
1,00 |
0,73 |
0,78 |
0,68 |
0,78 |
||
DRI 1997 |
1,00 |
0,85 |
0,64 |
0,70 |
|||
ICPI 1997 |
1,00 |
0,90 |
0,88 |
||||
PERC 1997 |
1,00 |
0,88 |
|||||
Gallup 1997 |
1,00 |
This year, the reliability of the data further improved, which can be observed by comparing the correlations between the sources, which is often even higher than last year. The variance for the individual countries (as reported in the main table, column 4) is lower this year. The geometric mean of these figures amounted to 0,9 last year and dropped to 0,67 this year. The idea of combining these different sources into a single index further serves to strengthen the index with respect to validity and reliability. This strategy has been lately endorsed by Lancaster (Emory University) and Montinola (New York University), "Toward a Methodoloy for the Comparative Study of Politican Corruption". In: Corruption and Reform: An International Journal, forthcoming 1997. The reliability of each figure is finally strengthened by including only countries which have been subjected to at least four polls at the minimum. Last not least, most sources focus on responses from businesspeople and risk analysts who are usually close enough to actual incidences of corruption and who can adequately recognize corruption when they see or experience it. The Gallup International survey differs from the other sources in so far as the general public is surveyed, i.e. people who sometimes may not have first hand experience with corruption and who may have limited capa-bili-ties in comparing levels of corruption between countries. The methodo-lo-gi-cal scepticism this implies is not reflected in the results. The high correlation of this source with the other sources is particularly useful in endorsing our results. Keep in mind, that the ranking does not try to assess corruption with a precision of double digits. Some sources are very rough in their evalua-tion and the inclusion of two digits exceeds the precision of the original data. We still included two digits in our presentation in order to allow for a transparent interpretation of the
Which countries have been included?
Because of the nature of the index it has only been possible to include countries who have themselves been subject of a certain number of polls. To the extent that the list does not include many countries, it is because the polls surveyed did not include them. The minimum number of sources for the CPI is four. Malperformance of a single survey may hence be balanced by the inclusion of at least three more surveys. This way we were able to include 52 countries into the 1997 index, slightly less than in 1996. We hope to broaden the scope of the index in future years.
How precise are your figures?
The numbers of surveys and the variance of the results give a hint on how good the results are. The higher the number of scores and the lower the variance of the results, the more trustworthy is the score for a particular country. However the figures we produce cannot be regarded as objective. Rather they represent the subjective evaluation of businessmen, risk analysts and the general public. With a low amount of surveys and a high variance the figures become weaker and special attention has to be put to the interpretation.
Why does your index not produce objective figures?
An objective approach is almost impossible. Corruption is by definition hidden. There is of course objective data created by the justice system and the media. However, this data rather measures the effectiveness of the media in discovering and reporting scandals and how independent and well-trained the judiciary is in prosecuting. An efficient and incorruptible jurisdiction may result in a high rate of convictions. Instead of acknowledging this, "objective" data would "punish" such a country with a bad score.
How up-to-date is your assessment?
Six surveys are from 1997, with one going back to 1996. This makes the CPI very up-to-date. Nonetheless, the perception by contributors may refer back to experiences made long before the assessment is made. The high amount of data which are necessary for reliable scoring, make it necessary to use some older data. The high correlation between the surveys indicates, that corruption scores usually do not change quickly. The misdeeds of an earlier government may shade that country's perfor-mance in the ranking in the first few years even if fundamental reforms are on their way. This should be kept in mind when our results are interpreted.
The General Approach:
How do you define corruption?
Corruption is the misuse of public power for private benefits, e.g. the bribing of public officials, taking kickbacks in public procurement or embezzling public funds. The external surveys we included were mostly very close to this definition. The index tries to assess the degree, to which public officials and politicians in particular countries are involved into corrupt practices.
Is corruption part of the culture in some countries?
Culture and ethics naturally vary between countries. However in all environments corruption is an illegitimate behaviour. Neither is it sensed as legitimate by those delegating power to politicians and public officials. Nor can politicians or public officials claim to be legitimately empowered to corrupt acts. Therefore corruption is necessarily accompanied by secrecy. Corruption cannot prosper in highly transpa-rent environments. Obfuscation is an instrument applied by those who engage in corrupt behaviour to provide the necessary discretionary power. This approach may help in providing a comparative viewpoint towards corruption.
How can you compare countries?
Corruption is defined by some researchers as a particular public reaction to political/administrative behaviour rather than an illegitimate act as such. Looking for appropriate definitions, this approach assigns a much more active role to the public perception and reactions towards corruption. A high degree of observed corruption may in such an approach reflect a high standard of ethics and a rigid application of rules rather than a high degree of real misbehaviour. A cross-country comparison of levels of corruption would hence not be applicable since the underlying standards of ethics may not correspond between countries. However, the sources we included with the exception of the Gallup International survey put a high effort on comparative judge-ments. This way, each respondent is using his or her viewpoint to assess different countries, bringing about a valid approach for a comparative viewpoint. The high correlation of the assessment by the general public in the Gallup International data is noteworthy as such. Allthough the point of view provided there may refer to purely local standards of ethics, the results are so close to the other sources that there are good reasons to assume the perception of what is regarded as corruption to be more global than many thought it might be.
How are the exporting countries contributing to corruption?
The ranking only scores the observed behaviour of public officials and politicians, that is, mainly the passive part of taking bribes. However, exporting industries also contribute to the development of corruption abroad. This is not taken into account in this ranking. Therefore our index is not a fair assessment of the respon-sibility associated with corruption in international trade. There are a number of industrialised countries which may be successful at keeping their house clean and highly engaged in bribing foreign officials bear a much higher burden.
How can the 1997 Index be compared to the 1996 Index?
Scores for most countries did not change significantly, in the past year. A worse or better perfor-mance is indicated by the average score, not by the "rank". In case a country has not been included in the 1996 ranking as is the case for Luxemburg, Costa Rica, Uruguay, Romania and Vietnam, there is no number reported. Some countries (i.e. Jordan, Ecuador, Egypt, Uganda, Cameroon, Bangladesh and Kenya) included in 1996 are missing this year due to a lack of data. A lower score indicates a worse performance, whereas a higher number indicates a perception of improvements. However, the following needs to be pointed out: Differences in the two scores which are close to zero (e.g. between 0 and 0,3) do not indicate a significant change. A different score does not necessarily indicate a different performance. Allthough the sources are highly correlated and reliable, each source employs a different methodology and thus produces ist own framework of evaluation. We cannot therefore rule out that the perception of a country's integrity may also vary with the availability and choice of sources. The sources range back to 1996 and perceptions may rely on experiences made before that time. A comparison between the 1997 and the 1996 index does therefore not necessarily reflect political developments made between 1996 and 1997. A changing performance may be due to actual regime shifts and a trend towards increasing or decreasing corruption over time. Keeping in mind the remin-ders mentioned above, the comparison between the 1997 and the 1996 index highlights the following developments: Europe is experiencing major changes with mediterranean countries (Italy, Spain and Portugal) catching up while Belgium is downgraded considerably. Denmark is now clearly dominating the list. Noteworthy for Asian countries is the downgrading of South Korea, while Pakistan is catching up again from a low score in 1996 after the ouster of the Bhutto administration. In Latin America, Bolivia, Argentina, Mexico and Colombia experienced a further decrease in their score, while improvements have been perceived in Brazil. Costa Rica which is included for the first time now tops the list of Latin American countries ahead of Chile, which also scored lower this year. Many African countries could not be included due to a lack of data. Nigeria is still bottom, and although its very low score in 1996 appears to have improved, in fact what probably happened was that a consensus has emerged that its "true" was a little higher in the spread of the variance but that at the same time it merits its ranking.
Overview: Sources and the adjustments
Back to Corruption Perceptions Index 1997
© Transparency International & Dr. Johann Graf Lambsdorff, Göttingen University, 1997
The Corruption Perceptions Index (CPI) is published jointly by
TI and Dr. Johann Graf Lambsdorff,
Volkswirtschaftliches Seminar, Universität
Göttingen, Platz der Göttinger Sieben 3, 37073 Göttingen, Germany;
Tel: 49-551-397298,
Fax: 49-551-392054,
e-mail: [email protected];
WWW: www.uni-goettingen.de/~uwvw
For any press enquiries please contact