Background Paper to the

2000 Corruption Perceptions Index

Framework Document

Johann Graf Lambsdorff

Transparency International (TI) and

Göttingen University

September 2000

The Corruption Perceptions

Index is a composite index.

The data used this year were

compiled between 1998 and

2000. Comparisons to last

year’s index are not feasible.

This document explains

which data entered into the

index and how this data was

standardized and aggregated.

It is discussed how corruption

is defined by our sources and

how the perceptions gathered

relate to reality.

2

1. The methodologyTransparency International’s Corruption Per-ceptions Index (CPI) has assumed a centralplace in debates about corruption. It is used byeconomists, academics, business people andjournalists. As in previous years, we thereforeprovide this framework document which pro-vides an in-depth explanation of the method-ology. This document complements the pressmaterials and another background paper ex-plaining the precision of the results.

The goal of the CPI is to provide dataon extensive perceptions of corruption withincountries. This is a means of enhancing un-derstanding of levels of corruption from onecountry to another. It does not attempt to as-sess the degree of corruption practiced by na-tionals outside their own countries. This is aseparate phenomenon and a separate instru-ment, the Bribe Payers Propensity Index(BPI), was published last year for the firsttime.

In an area as complex and controver-sial as corruption, no single source or pollingmethod has yet been developed that combinesa perfect sampling frame, large enough coun-try coverage, and a fully convincing method-ology to produce comparative assessments.This is why the CPI has adopted the approachof a composite index. It consists of crediblesources using different sampling frames andvarious methodologies and is the most statisti-cally robust means of measuring perceptionsof corruption.

Objective versus subjective data

Unbiased, hard data is difficult to obtain andusually raises difficult questions with respectto validity. One such set of data has been as-sembled by the Crime Prevention and Crimi-nal Justice Division of the United Nations Of-fice at Vienna, [United Nations 1999]. This isa survey of national agencies in a large varietyof countries called the United Nations Surveyof Crime Trends and Operations of CriminalJustice Systems. The major goal of this inves-tigation has been to collect cross-nationally

comparative data on the incidence of reportedcrime and the operations of criminal justicesystems. The questionnaire consists of a seriesof questions which ask for data, primarily sta-tistical, on the main components of the crimi-nal justice system. The latest version of thissurvey relates to the years 1990 to 1994. Allnational data are derived from the official na-tional criminal statistics.1 However, the pre-cise legal definition of bribery and corruptioncan be different in each national context, thedifferences drawn between bribery, embez-zlement and fraud may be troublesome andthe statistical methodology of counting andaggregating used in each national agency candiffer considerably from that used elsewhere.Apart from this, countries such as Singaporeand Hong Kong have extremely high per cap-ita conviction rates for bribery. This lends it-self to the conclusion that the data are to alarge extent determined by the effectivenessand capacity of a country's judiciary in prose-cuting corruption. High levels in this case in-dicate the success of anti-corruption initiativesrather than high levels of actual corruption. Assuch problems commonly arise with objectivedata, international surveys on perceptionsserve as the most credible means of compilinga ranking of nations.

Sources in 2000

Prior to selecting sources guidelines havebeen set up which organize the underlyingdecision making process. These include theactual criteria that a source needs to meet inorder to qualify for inclusion as well as or-ganizational guidelines on how the final deci-sion is reached with the help of the Transpar-ency International Steering Committee. Thisprocess aimed at making the final decision astransparent and robust as possible. As a resultof this it was decided that the 2000 CPI in-cludes data from the following sources:

1 A full description of the methodology and thecomplete data can be obtained via internet at:http://www.ifs.univie.ac.at/~uncjin/wcs.html.

3

• Freedom House Nations in Transit (FH),• the Economist Intelligence Unit (EIU),• the Institute for Management Develop-

ment, Lausanne (IMD),• the International Crime Victim Survey

(ICVS),• Political Risk Services (PRS),• the Political and Eonomic Risk Consul-

tancy, Hong Kong (PERC),• the World Bank and European Bank for

Reconstruction and Development (WB),• the World Economic Forum (WEF).

Other sources have also been suggested forinclusion. An essential condition for inclusionis that a source must provide a ranking of na-tions. This condition is not met if a sourceconducts surveys in a variety of countries butwith varying methodologies. Comparisonfrom one country to another are not feasible inthis case and a ranking cannot be produced.Another condition is that sources must meas-ure the overall level of corruption. This isviolated if aspects of corruption are mixedwith issues other than corruption or if changesare measured instead of levels of corruption.

The 2000 CPI combines assessmentsfrom the past three years to reduce abruptvariations in scoring. Such changes might bedue to high-level political scandals that affectperceptions, but do not reflect actual changinglevels of corruption. Some sources, such asWB and ICVS, provided only one recent sur-vey. Others such as PERC, WEF and IMDconducted various surveys between 1998 and2000, which are all included. In addition to itsGlobal Competitiveness Report (GCR), theWEF also published the Africa Competitive-ness Reports (ACR) in 1998 and 2000, whichare also included.

While this averaging is valuable for theinclusion of surveys, it is inappropriate forapplication to the data compiled by countryexperts. Such assessments as compiled byPRS, FH and EIU are conducted by a smallnumber of country experts who regularlyanalyze a country's performance, counter-

checking their conclusions with peer discus-sions. Following this systematic evaluation,they then consider a potential upgrading ordowngrading. As a result, a country's scorechanges rather seldom and the data shows lit-tle year-to-year variation. Changing scores inthis case are the result of a considered judge-ment by the organization in question. To thengo back and average the assessments over aperiod of time would be inappropriate. On theother hand, in the case of elite or general pub-lic surveys an averaging over various yearsproduces a useful smoothing effect. Whilesome annual data may contain random errors,these do not necessarily carry over into thenext year, and their impact is decreased by theaveraging procedure. Overall, 16 sourcescould be included in the 2000 CPI, originatingfrom 8 independent institutions. The completelist of sources is presented in appendix.

The number of sources has decreasedslightly in 2000 as compared to 1999. Thiscame about as the older sources from 1997 byGallup International, the Wall Street Journaland the World Bank/University Basel havenot been updated. Since they are older than 3years they could no longer be included. WhileICVS updated its data, not all survey resultswere processed at the time the CPI was com-piled. Only a smaller sample of 11 countriescould thus be considered for the CPI. All inall, the number of countries in the CPI slightlydropped from 99 last year to 90.

Year-to-year comparisons

The CPI incorporates as many reliable and up-to-date sources as possible. One of the draw-backs to this approach is that year-to-yearchanges of a country's score do not only resultfrom a changing perception of a country's per-formance but also from changes in sample andmethodology. This is comparable to theproblem of designing a price index for a bas-ket of goods when the ingredients are chang-ing. The price index for one period cannot befully compared to that of the next since thebasket on which it is based has changed. Asimilar problem arises with the CPI. Some

4

sources are not updated and must be droppedas a result, while new, reliable sources areadded. With differing respondents and slightlydiffering methodologies a change in a coun-try's score cannot be attributed solely to actualchanges in a country's performance.

Accordingly, TI repeatedly stressesthat each year's index must be seen as the re-sult of the sum of all reputable sources avail-able at that time. Comparisons with the viewscollected in previous years can be misleading.In order to reduce the number of misleadinginterpretations of the CPI scores, the officialCPI table does not include the scores from theprevious year.

2. ValidityAll sources generally apply a definition ofcorruption such as the misuse of public powerfor private benefits, e.g., bribing of public of-ficials, kickbacks in public procurement, orembezzlement of public funds. Each of thesources also assesses the “extent” of corrup-tion among public officials and politicians inthe countries in question:

• The IMD asks respondents to assesswhether “bribing and corruption prevail ordo not prevail in the public sphere.”

• The WEF asks in its Global Competitive-ness Report “irregular, additional pay-ments connected with import and exportpermits, business licenses, exchange con-trols, tax assessments, police protection orloan application are common/ not com-mon.” In the Africa Competitiveness Re-port it was additionally asked how “prob-lematic the following areas are for doingbusiness: ... corruption” and “when firmsin your industry do business with the gov-ernment, how much of the contract valuemust they offer in additional or unofficialpayments to secure the contract?” For de-tails see [Lambsdorff and Cornelius:2000].

• The PERC asks “To what extent does cor-ruption exist in the country in which youare posted in a way that detracts from the

business environment for foreign compa-nies?”

• The EIU defines corruption as the misuseof public office for personal (or party po-litical) financial gain and aims at measur-ing the pervasiveness of corruption. Cor-ruption is one of over 60 indicators used tomeasure “country risk” and “forecasting.”

• The ICVS asks: "In some areas there is aproblem of corruption among governmentor public officials. During 1999, has anygovernment official, for instance a customsofficer, police officer or inspector in yourown country, asked you or expected you topay a bribe for his service?"

• The PRS determines a variable "Corrup-tion in Government" and assesses theoverall spread of corruption.

• FH determines the "level of corruption"without providing further defining state-ments.

• The WB asks various questions with re-spect to corruption. One group of questionsis directed towards determining the level ofgrand corruption and called “state cap-ture”. An aggregate measure for this vari-able is provided by the WB. To adequatelybalance this issue with the level of admin-istrative corruption responses to the fol-lowing question have been used: “It iscommon for firms in my line of business tohave to pay some irregular ‘additionalpayments’ to get things done.” For detailssee [Hellman, Jones and Kaufmann 2000]and [Hellman, Jones, Kaufmann andSchankerman 2000].

The terms "level", "problem", "prevalence","pervasiveness" and "commonness" arelargely identical. They all refer to some kindof “degree” of corruption, which is the alsoaim of the CPI. This common feature of thevarious sources is particularly important inview of the fact that corruption comes in dif-ferent forms. It has been suggested in numer-ous publications that distinctions should bemade between these forms of corruption, e.g.between nepotism and corruption in the form

5

of monetary transfers. Yet, none of the dataincluded in the CPI emphasize one form ofcorruption at the expense of other forms. Thesources can be said to aim at measuring thesame phenomenon.

It is important to note that largely noneof the sources differentiates between adminis-trative and political corruption, and that bothtypes of corruption are addressed equally bythe various questions posed. The IMD asksabout corruption in the public sphere. Thisinevitably includes both corruption in admini-stration and in politics, as they both constitutethe public sphere. The WEF addresses par-ticular areas where corruption can occur andin each of these, either politicians or adminis-trators can be the relevant actors. Similarly,corruption in government, as assessed by thePRS, also includes both types of corruption,since administration and politics are each partsof government structures. The same applies tothe broad definition used by FH. The EIU ex-plicitly notes that its assessments include cor-ruption among public servants and politiciansalike. This largely justifies a blending of po-litical and administrative corruption, sincethere is no strong evidence that countries dif-fer in the prevalence of the one type of cor-ruption over another. But the results by WBrepresent a challenge to the type of blendingexercised for the CPI. It is observed there thatin some countries grand political corruption ismore prevalent while in others small scaleadministrative corruption dominates. This lineof research is particularly promising for thefuture. Results are at this stage only availablefor transition economies and it will be crucialto observe whether they are also valid forother continents and robust to the usage of dif-ferent methodologies.The only source which does not clearly relateto large scale political corruption is the ICVS.But taking into account that this source (alsoin the past) well correlates with the othersources, there was no strong argument thatalso the extent of political corruption is notwell represented by this data. The data on cor-relations are provided in another background

paper which deals with the precision of theCPI.

Degrees of corruption

As we have emphasized, the CPI aims to as-sess the "degree of corruption". But this termcan suggest different interpretations, [Rose-Ackerman 1999: 4]. In order to confirm thevalidity of our approach, we must first clarifywhether this term is unambiguous. Imaginethe simple case that 10 percent of all publicservants take a bribe of $200 each, 5 times ayear in exchange for awarding a contract thatresults in a gain of $500 each for corrupt pri-vate contractors. Degree could either relate thefrequency of corrupt acts, the amount ofbribes paid or the overall gain that contractorsachieve via corruption.

While all of these definitions appear tobe valid, they need not fully correlate witheach other. For example, consider that a fewhigh-ranking public servants are taking largebribes on the one hand, as opposed to manypublic servants engaging in petty corruptionon the other. The total sum of bribes would beabout the same in both cases, but the fre-quency of corrupt incidents would doubtlesslybe higher in the latter case. Similarly, whencorrupt private competitors are in a strongbargaining position and do not allow much oftheir illegitimate gain to be shared with publicservants, the total amount of bribes may below while total gains from corruption arelarge. Clearly, absence of corruption would besimilarly defined in all three cases — i.e. fr e-quency, amount of bribes and value of rents— as being equal to or nearly zero.

Having taken this theoretical look atdegrees of corruption, we can now turn to theparticular definitions used by our sources.First, the questions asked by the ICVS, WEFand the WB (on adminstrative corruption) re-lates to the frequency of bribes paid. In con-trast to this, the questions by the PERC andthe WB (on state capture) hint at the damageto private business people caused by corrup-tion. The implication here might be that largebribes are particularly serious, while large

6

benefits for corrupt private business peoplemay not be. The questions posed by the IMD,PRS, FH and EIU provide no insight regard-ing an assessment of degree. The terms"level", "prevalence", "existence" and "perva-siveness" used there might refer to frequencyas well as the overall value of bribes involved.

In sum, the term “level of corruption”seems to include both aspects, frequency ofcorruption and the total value of bribes paid.Taking into account that the sources correlatewell with each other, we may conjecture thatat the moment there is little evidence that di f-ferences with respect to these two aspects arecrucial to the outcome of a survey. Either, re-spondents have a very homogeneous pre-specified idea of how to define the "degree ofcorruption" which influences their response,irrespective of the precise wording of thequestionnaire, or countries do not differ con-siderably with respect to the particular kindsof corruption that prevail there. More re-search, as the one by the WB, is required todeepen our understanding of the levels andtypes of corruption and the extent to whichcorruption differs between countries.

3. Perceptions and realityWhile the sources all aim at measuring thedegree of corruption, the sample design differsconsiderably. With the exception of thePERC, FH, EIU and PRS the sources mostlysample residents, who must rely on their per-sonal, local estimate (as opposed to an expa-triate’s external viewpoint) of the degree ofcorruption and the meaning of the term intheir own cultural context. Whether this dif-ference may lead to different outcomes stillrequires scientific study. For the purposes ofthe CPI, it added to the robustness of the re-sulting figures, since the data correlate wellwith other data. This correlation suggests thatthere being different samples makes no greatdifference to the results.

Of greater importance is the differencebetween polls on the opinions of experts andthe polls of the general public provided by theICVS. Whereas the general public may tend

to form views on the corruption (or lack of it)experienced in daily life, business people andrisk analysts are usually closer to high-levelincidents of corruption and may be in a betterposition to assess grand corruption. Further-more, elites may have a biased viewpoint to-wards corruption insofar as they might be lessnegative about forms of corruption which fa-vor their own group. Similarly, the generalpublic may be less negative about petty formsof corruption. To what extent the generalpublic deviates from an elite sample in its a s-sessment of corruption has not yet been thesubject of investigation and constitutes an im-portant area for future research. Various dataon perceptions by the general public used forthe CPI in previous years suggests no signifi-cant difference between these viewpoints.2

Interpreting perceptions

As the data collected relates to perceptionsrather than to real phenomena, it has to beconsidered whether such perceptions improveour understanding of what real levels of cor-ruption may be. This is necessary for the CPIto be a fruitful contribution to political debate,investment decisions and academic research.Since actual levels of corruption cannot bedetermined directly, perceptions may be allwe have to guide us. However, this approachis undermined to at least some extent, if theperceptions gathered are biased. Such a po-tential bias might originate from the particularcultural background of respondents. Depend-ing on whether the sample consist of locals orexpatriates, this suggests two potential biasesto be relevant.

2 Even when elite and general public viewpointsshow some differences, an aggregation of thesedata still makes sense, just like price levels forvarious goods can be aggregated to form somecombined price index. While the idea of creating aprice index would be to value a complete basket ofgoods, the idea of aggregating subjective datawould then be to obtain an assessment of the levelof corruption as seen by a broad and possibly het-erogeneous sample of respondents.

7

Imagine that being asked to assess thelevel of corruption, a local estimates a highlevel of corruption in the country of residence.Such an assessment would be a valid contri-bution to the CPI only if the respondent makesthe assessment as a result of comparisons withthe levels of corruption perceived in othercountries. But this is not necessarily the view-point taken by the respondent. A respondentmay also assign high levels by comparing cor-ruption to other (potentially less pressing)problems facing the country, or by evaluatingit according to a high ethical standard (e.g.which assumes any kind of gift-giving to apublic official to be corrupt and not culturallyacceptable). In the case of such an outlook, ahigh degree of observed corruption may re-flect a high standard of ethics rather than ahigh degree of real misbehavior. Perceptionswould be a misleading indicator for real levelsof corruption. This bias can occur particularlyif only locals are surveyed, each assessingonly the level of perceived corruption in theirown countries. If respondents are asked to as-sess foreign countries or to make comparisonsbetween a variety of countries, this biasshould not occur. Respondents will, in thiscase, compare a foreign country with theirhome country or with an even larger set ofcountries. They will be forced to apply thesame definition of corruption and make use ofthe same ethical standard for all countries,which produces valid comparative assess-ments. However, in this context a second typeof bias might arise, originating from the po-tential dominance of a particular culturalheritage in the sample questioned or becauseexpatriates lack a proper understanding of acountry's culture. If this happens, comparativeassessments might reflect disproportionatelythe perceptions of a particular culture. But theresults would be meaningless to locals whohave a different understanding and definitionof corruption. While samples which are domi-nated by a particular cultural heritage are sus-ceptible to this kind of bias, surveys whichquestion local residents clearly avoid this kindof bias.

The strength of the CPI rests with theidea that we include surveys which are notsusceptible to the first type of bias. Particu-larly these are EIU, PRS, FH and PERC.Since the data provided by these sources referto assessments by expatriates, they are subjectto a homogeneous definition of corruption anda consistent ethical standard. The CPI also in-corporates the data from the ICVS, IMD,WEF, and WB. Since these refer to assess-ments made by local residents, they are notlikely to represent the perception of a certaincultural heritage. The second type of bias canclearly be rejected for these sources.

Since the data from the EIU, PRS, FHand PERC correlate well with the other data,there seems to be no support for the sugges-tion that they might be influenced by the sec-ond type of bias. Similarly, since the data bythe ICVS, IMD, WEF, and WB correlate wellwith the other three sources, the notion thatthe first type of bias might be present isclearly not supported. The validity of thesources is mutually confirmed and prevalenceof the potential biases mentioned before canbe rejected. Our approach clearly suggests thatthe perceptions gathered are a helpful contri-bution to the understanding of real levels ofcorruption.

The role of the media

Another potential problem with the collectionof perceptions may arise from the possibilitythat respondents do not report their personalexperiences but rely on media coverage andreports obtained from others. Certainly thisinfluence cannot be excluded and necessarilycontributes to perceptions. Yet in its extremeform such an influence may suggest that re-spondents rely only on hearsay. The potentialproblem with this influence is that the asses s-ment of a country might then reflect the qual-ity of the press in uncovering scandals, andparticularly its freedom to do so. Countriesthat suppress a free press may escape a badreputation for corruption among their popula-tion. Such an influence would certainly un-dermine the validity of the CPI and its useful-

8

ness as an aid to understanding real levels ofcorruption.

Investigating whether such an influencemight affect the CPI, it is worthwhile to ob-serve that some sources may be more influ-enced by hearsay than others. It is particularlyinteresting that the question posed by theICVS clearly relates to personal experience asopposed to hearsay. Already in the past weobserved a large correlation of ICVS withother sources, which indicates that hearsaydoes not appear to be an important influencefor the overall CPI.

4. The indexWith the various sources having some differ-ences with respect to sample and date, a num-ber of ideas have been considered forweighting the sources before aggregatingthem. One possibility was to weight them ac-cording to the number of replies collected byeach source. However, this would mean thatsome surveys would obtain a large weight,particularly if seen against the expert asses s-ments conducted by PRS, FH and EIU. If thisline were pursued, it would mean that theviews of an individual selected at randomwould have the same quality as an expert as-sessment made after country-specific analysisand peer review. This approach was thereforenot convincing.

Another methodology for aggregatinggovernance data has recently been suggestedby Kaufmann, Kraay and Zoido-Lobaton[1999], based on a formal model. The authorsassume that each source is a noisy indicatorfor actual levels of corruption, which is the"unobservable component" they seek to de-termine. Based on this model an average scoreand a measure of precision is obtained for alarge variety of countries.3 Those sources

3 With respect to measures of precision, some notnecessarily realistic assumptions had to be intro-duced. The measures of precision therefore repre-sent a best-case scenario and are not unbiased. Inaddition, they neglect the standard deviation be-

which then better correlate with the resultingaggregate index receive more weight, whilethose which contribute less viably enter intothe index with less weight. The quality of thesources is therefore determined endogenouslyand is not an expert's opinion on a source'svalidity and reliability. While there may be apoint in taking this approach, weighting canbe biased if the sources are not independent ofeach other. It may occur that the sources thatare least independent — for example becausethey use other sources as their benchmark orsample people who have little first-hand expe-rience — are given higher weights than thosewho engage in discovering original insights.This weighting system would be in contrast toexperts' viewpoints regarding the quality ofsources. Given this disadvantage it was de-cided that this approach should not be adoptedfor the TI-CPI.

In the end, it remains preferable toadopt the simple approach of assigning equalweights to those sources which have beenfound to meet the criteria of reliability andprofessionalism. Other procedures may havetheir merits, but this simple averaging systemis easiest to explain to a broad public.

It was suggested in this context thatdata from various years provided by the samesource should not obtain the same weight asother data. One may adhere to the viewpointthat the data provided by an institution is in-dependent to that from another institution, butthe same independence may not prevail forsurveys originating from the same institution.But this argument may push too far an issuewhich is in fact difficult to assess, since alsoan institutions may lean on the data producedby others in reaching a conclusion. Since thematter of independence is therefore difficult toquantify, there was no clear argument inchanging the methodology used so far. As aresult of giving each survey an equal weight,some institutions obtain a larger weight than

tween sources, i.e. that precision should be lowerin countries where sources differ considerably intheir assessment.

9

others. While other approaches can certainlybe justified there is also some rational in this.It reflects previous recommendations of theTransparency International Steering Commit-tee that continuous annual surveys are supe-rior for our purposes than one-off surveys:they may have gathered more expertise inproviding their service and their inclusionhelps to avoid abrupt year-to-year changes inthe CPI. In addition to that, continuous annualsurveys may be seen to be superior to expertassessments because the methodology of pro-ducing data is more transparent and subject toa clear procedure.

Standardizing

Since each of the sources uses its own scalingsystem, aggregation requires a standardizationof the data before each country’s mean valuecan be determined. For all sources not alreadystandardized for the CPIs of previous years,the 1999 CPI was the starting point for thisprocess. It had a mean value of 4.61 and astandard deviation of 2.36. Each of thesources naturally had different means andstandard deviations. Yet standardization doesnot mean that each source is given the samemean and standard deviation, since eachsource covers a different subset of countries.Instead, the aim of the standardization processis to ensure that inclusion of a source consist-ing of a certain subset of countries should notchange the mean and standard deviation ofthis subset of countries in the CPI. The reasonis that the aim of each source is to assesscountries relative to each other, and not rela-tive to countries not included in the source.The aim here is that a country not be "pun-ished" for being compared with a subset ofrelatively uncorrupt countries, nor rewardedfor being compared with a subset perceived tobe corrupt. In order to achieve this, the meanand standard deviation of this subset of coun-tries must take the same value as the respe c-tive subset in the 1999 CPI.

An example can illustrate the standard i-zation. In 2000, IMD assessed Brazil with avalue of 2.57 on a scale between 0 and 10. At

first, a common subset of countries was de-termined, countries which belong to both theIMD 2000 and the 1999 CPI. The means andstandard deviations in each of these sourceswas determined. In the IMD 2000, thesecountries had a mean value of 4.79 and astandard deviation of 2.64, while in the 1999CPI, these countries had a mean of 6.07 and astandard deviation of 2.41. Standardizing thevalue for Brazil thus required subtracting 4.79from the 2.57,4 multiplying the result with2.41, dividing by 2.64 and adding 6.07. Theresult turns out to be 4.04, the standardizedvalue for Brazil. Applying this to all countriesin the subset, the standardized values thenhave a mean of 6.07 and a standard deviationof 2.41, the same values this subset of coun-tries had in the 1999 CPI. The same formula isthen applied to all countries covered in theIMD, including those that do not belong to thesubset described above. After this is done forall countries and all sources, the index is de-termined by computing the simple mean foreach country.5

The previous indices relied solely on atechnique of standardizing means and stan-dard deviations for the respective subsamplesof countries. It was observed in the past thatan alternative technique of matching percen-tiles would bring about largely identical re-sults. Matching percentiles is superior in com-bining indices which are differently distrib-uted. But, as it makes use of ranks and notscores of sources, it looses some of the infor-mation inherent in the sources. The generalapproach is therefore that it is preferable torely on the described standardization tech-nique, except where the distribution of asource clearly differs from that of the CPI.

4 In case a source assigns a higher score to morecorrupt countries, this value must be multiplied by-1.5 A final standardization must be undertaken, sincethe aggregate may again differ with respect tomean and standard deviation as compared to theprevious years index.

10

One such source with a clearly differentdistribution is ICVS. It was decided that forICVS the method of matching percentileswould be applied. For this technique again thecommon subsamples of the ICVS and the1999 CPI are determined. Then, the largestvalue in the 1999 CPI is taken as the stan-dardized value for the country ranking best byICVS. The second largest value is given to thecountry ranking second best, etc.

For IMD and PERC, this standardiza-tion procedure did not change the values sig-nificantly, since the data was already deliv-ered on a scale between 0 and 10. This con-trasts to the values provided by WEF who re-port the data on a scale between 1 and 7. TheWB provides two data on corruption, whichwere aggregated6 before being standardizedand included in the CPI. Likewise PRS andEIU provide assessments ranging between 0and 6 and between 0 and 4, respectively. Theoriginal data by Freedom House were notgiven in numerical format but a "broader al-phabetical grade" assigned to the respectivecategories. This implies that FH does not in-vite for a "cardinal" interpretation of their as-sessments which is required for the normalstandardization methodology. By making useof a methodology of matching percentilesonly the ordinal information by FH would beused. But also a normal standardization of thedata provides an indicator which correlates0.992 with the one obtained from matchingpercentiles. Given this it was decided to keepthe normal standardization technique.

Presentation

The 2000 CPI will include all countries forwhich at least three sources had been avai l-able. Some critics had argued in favor of ex-tending the index to include also countries forwhich less than three sources are available. Inthis case the CPI would include 151 countries.There are undeniable merits to this. A larger

6 Again, aggregation requires a standardizationprocedure by means and standard deviation.

list of countries would further facilitate theusage of the CPI in academic research. Therehas been an immense research activity basedon the CPI, but the limited number of coun-tries was sometimes felt to present a slightdisadvantage. Above that, it was observed thatlimiting the index to countries where suffi-cient information is available would excludeparticularly countries perceived to be corrupt,because information on such countries tendsto be scarce. It was argued that this may mis-lead the public.

But these arguments must be valuedagainst the respective disadvantages of a fur-ther expansion. In public debate, measures ofprecision are commonly not well taken intoaccount — irrespective of the immense effortTI has put into the presentation of the CPI inthe past. The method to avoid this mispercep-tion has been to restrict the index to thosecountries where sufficient information isavailable. These are countries where the mar-gin of error of the reported average score israther low. In contrast, countries with lessthan three sources are measured with largeimprecision. It therefore makes sense to stickto this established guideline and include onlythose countries for which at least three sourceswere available. Since those countries left outof the index are on average perceived to berather corrupt, there emerges an inadequatecomparison of a country to the rest of theworld — an interpretation which TI did notinvite for but which some media was engagedin. It may be worthwhile to note that all 151countries would on average score 4.3. Thisfigure may serve as a benchmark value. Par-ticularly it illustrates that countries not beingincluded into the CPI should not interpret thisas a particular type of qualification or dis-qualification.

Some media misinterpret the fact thatbeing worst in the index does not mean beingmost corrupt in the world. This year this posi-tion is occupied by … … .. But this misunder-standing would also not be alleviated by ex-panding the index, because the lowest positionis simply assigned to another country, while

11

the index would still not cover the wholeworld. Another suggestion was that the last 10countries, those with the worst scores, belisted alphabetically, without publishing theirscores, avoiding that a country is singled out.It would have been difficult to adequately pre-sent such a mixture of groups and ranks. Butmore pressing was the fear that a whole groupof countries may have been branded as theworst in the world, “tieing” for bottom place.This alternative was dropped as a conse-quence.

On the web-sites(http://www.transparency.org/ andhttp://www.uni-goettingen.de/~uwvw) weprovide further data for each country on thestandard error, as described above, and theresulting confidence intervals. Moreover,some observers may be interested in theamount of independent institutions that con-tributed to an average value and not only thetotal number of sources. This figure is alsoreported. Apart from that, the CPI continues torank countries and assign scores with onedigit, as we have done in the past. Alongsidethis data the standard deviation and the num-ber of surveys used for each country is re-ported.

In addition to this data, the main tablethis year for the first time provides the high-low range. This depicts the highest and thelowest values provided by our sources, so asto portray the whole range of assessments.However, no quick conclusions should be de-rived from this range to the underlying preci-sion with which countries are measured.Countries which were assessed by 3 or 11sources can have the same min-max range,but in the latter case the associated standarderror is much smaller. In order to arrive atsuch measures of precision, a more compre-hensive background paper is provided at ourwebsite.

ReferencesHEUNI (1998), "Crime and Criminal Justice

in Europe and North America.", Euro-

pean Institute for Crime Prevention andControl, affiliated with the United Na-tions (HEUNI), P.O. Box 161, FIN-00131 Helsinki. Ed. by K. Kangas-punta, M. Joutsen and N. Ollus. Thestudy can be obtained at:http://www.vn.fi/om/suomi/heuni/news/ccj.pdf.

Hellman, J., Jones, G. and D. Kaufmann(2000), ” 'Seize the State, Seize the

Day': State Capture, Corruption, andInfluence in Transition”, TheWorld Bank Policy Research WorkingPaper No. 2244, September:http://www.worldbank.org/wbi/governance.

Hellman, J., Jones, G., Kaufmann, D. and M.Schankerman (2000), “ MeasuringGovernance, Corruption, and StateCapture”, Policy Research WorkingPaper 2312, The World Bank April:http://www.worldbank.org/wbi/governance.

Kaufmann, D., A. Kraay and P. Zoido- Lo-baton (1999), "Aggregating GovernanceIndicators." World Bank Policy Re-search Working Paper No. 2195(Washington D.C.: The World Bank).

Lambsdorff, J. Graf and P. Cornelius (2000)Corruption, Foreign Investment andGrowth. In: Africa Competitiveness Re-port, ed. by K. Schwab, D.J. Sachs etal., joint publication of the World Eco-nomic Forum and the Institute for Inter-national Development, Harvard. OxfordUniversity Press 2000.

Rose-Ackerman, S. (1999), Corruption andGovernment. Causes, Consequencesand Reform (Cambridge: CambridgeUniversity Press).

United Nations (1999), “Global Report onCrime and Justice.” Report by theUnited Nations Office for Drug Controland Crime Prevention (New York: Ox-ford University Press).

12

Appendix: Sources used in the 2000 CPINumber 1 2 3

Source Political & Economic Risk Consultancy

Name Asian Intelligence IssueYear 1998 1999 2000Internet address http://www.asiarisk.com/Who was sur-veyed? Expatriate business executives

Subject asked Extent of corruption in a way that detracts from the business environment forforeign companies

Number of re-plies 280 40-50 per country 1027

Coverage 12 Asian countries 14 countries

Number 4 5 6

Source Institute for Management Development

Name World Competitiveness YearbookYear 1998 1999 2000Internet address www.imd.chWho was sur-veyed?

Executives in top- and middle-management; domestic and international compa-nies

Subject asked Bribing and corruption exists in the public sphere

Number of re-plies 2515 4314 4160

Coverage 46 countries 47 countries

Number 7 8Source Economist Intelligence Unit "International Working Group"

Name Country Risk Service and CountryForecast International Crime Victim Survey

Year 2000 1999/2000

Internet address www.eiu.com http://ruljis.leidenuniv.nl/group/jfcr/www/icvs/Index.htm

Who was sur-veyed? Expert staff assessment (expatriate) general public

Subject askedAssessment of the pervasiveness ofcorruption among politicians and

civil servants

During 1999, has any government of-ficial in your own country, asked you

to pay a bribe for his service?Number of re-plies Not applicable ca. 20,000

Coverage 115 countries 11 countries

13

Number 9 10Source World Bank and EBRD Freedom House

Name Business Environment and Enter-prise Survey Nations in Transit

Year 1999 1998

Internet address http://www.worldbank.org/wbi/governance http://www.freedomhouse.org/

Who was sur-veyed? Senior businesspeople Assessment by US academic experts

and FH-staff

Subject askedState capture and frequency of ir-

regular, additional payments to publicofficials

Levels of corruption

Number of re-plies 3000 Not applicable

Coverage 20 transition economies 28 transition economies

Number 11 12 13Source World Economic Forum

Name Global Competitiveness ReportYear 1998 1999 2000Internet address www.weforum.orgWho was sur-veyed? Senior business leaders; domestic and international companies

Subject askedIrregular, additional payments connected with import and export permits,business licenses, exchange controls, tax assessments, police protection or

loan application.Number of re-plies 3167 3934 4022

Coverage 53 countries 59 countries

Number 14 15 16Source World Economic Forum Political Risk Services

Name Africa Competitiveness Report International CountryRisk Guide

Year 1998 2000 2000Internet address www.weforum.org www.prsgroup.comWho was sur-veyed?

Senior business leaders; domestic and internationalcompanies

Expert staff assessment

Subject asked How problematic is corruption? irregular, addi-tional payments are required and large in amount.

Assessment of "corruptionin government"

Number of re-plies 582 1800 Not applicable

Coverage 20 countries 26 countries 140 countries