Corruption Perceptions Index: Methods, Rankings & Impact

Introduction: Quantifying Corruption in a Global Context

Country comparisons serve many purposes: economic, political, social, educational, and more. Many countries — and likely all developed countries — conduct country comparisons focused on international trade and overall national economic status. The collection of international economic data has been increasingly influenced by sophisticated strategy and technique, largely because national fiscal markets are globally linked and multinational corporations engage at high rates with foreign supply chain vendors (Podobnik et al., 2008). Measures such as the gross domestic product (GDP), the consumer price index (CPI), general government gross and net debt, and purchasing power parity (PPP) are all used to understand the economic status, monetary exchange, and other fiscal dynamics of different countries ("WEO," 2014).

Discourse among economists is strongly skewed toward ratios, curves, slopes, and derivatives — each carefully constructed to reveal patterns and trends that would not otherwise be accessible or interpretable ("WEO," 2014). It is precisely because the data are quantitative that they are so comparable; moreover, statistical procedures are used to weight differences that might inadvertently skew outcomes and to calculate the relationships between various measures, such as the CPI and GDP (Shao et al., 2007). For example, to accomplish surveillance of other countries, the World Economic Outlook (WEO) produces a country database containing data on each country's currency, the type of national account used, historical and current data on national accounts, whether chain-weighted methodology is used, and historical and current data on the consumer price index. The Consumer Price Index (CPI) is a term used by economists to reflect the changing monthly data on "prices that urban consumers pay for a representative basket of goods and services" ("Bureau of Labor Statistics," 2014).

Another organization that uses country-based data is the Organization for Economic Co-operation and Development (OECD). The OECD is known for examining patterns related to international trade and business. Its mission is to "promote policies that will improve the economic and social well-being of people around the world" ("OECD," 2014). The OECD measures national productivity and the flow of global investment and trade. It also sets international standards on topics as disparate as agriculture, taxation, pension systems, and chemical safety.

From its policy experience and the country data it collects, the OECD designs and recommends policies that can improve quality of life. At the core of OECD work is "a shared commitment to market economies backed by democratic institutions and focused on the well-being of all citizens" ("OECD," 2014). Tandem objectives of OECD work include making "life harder for the terrorists, tax dodgers, crooked businessmen and others whose actions undermine a fair and open society" ("OECD," 2014). Toward these goals, the OECD holds a Convention on Combating Bribery of Foreign Public Officials in International Business Transactions. Each participating country agrees to treat foreign bribery as a crime for which individuals and enterprises are held responsible. The Convention is therefore instrumental in curbing the export of corruption globally, since roughly two-thirds of global exports and nearly 90% of foreign direct investment outflows are directly tied to the 41 signatory countries. To increase its influence and effectiveness, the OECD Working Group on Bribery conducts follow-up reviews of nine to ten countries annually. While the OECD emphasizes fact-based policymaking and implementation, an organization known as Transparency International serves as an aggressive watchdog.

Transparency International is one of many non-governmental organizations (NGOs) that monitor and publicize the performance of political groups and corporations. The type of performance that Transparency International focuses on is the level of corruption in countries and territories around the world. The instrument it uses to measure and communicate perceived levels of corruption is the Corruption Perceptions Index (CPI).

Appendix: Calculation Steps for the CPI Index Scores

The calculation of the index entails the following steps:

Step 1. To enter the index, individual responses from business people's opinion surveys are averaged by country. When more than one question is used, the simple average score across questions is first calculated for each respondent, and then the average score by country is calculated.

Step 2. Because each source uses its own scaling system, the data must be standardized before entering the index. Rescaling is carried out in two steps.

Step 2.1 — Matching Percentiles. The first step consists of standardizing the scores using "matching percentiles." This technique uses the ranks of countries reported by each individual source (but not the scores). The method allows all reported scores to be denominated in common and comparable units within the same bounds, enabling proper aggregation within the CPI bounds of 0–10. However, while this method is useful for combining variables with different distributions, some information is lost in the process.

Standardization is only required for data that have not been used in previous editions of the CPI. Data used in the previous year's index are already standardized and enter the calculation of the current edition with those standardized values. The matching percentile technique proceeds as follows, labeling the individual survey or assessment as Source Y:

2.1.1. Select a master list. This master list is the pool of values from 0 to 10 to which rankings of Source Y will be matched. The master list is based on the previous year's scores — for the 2010 CPI, the master list is the TI CPI 2009.

2.1.2. Identify common countries. Only countries included in both the master list and assessed by Source Y are used in standardization. Countries appearing only in the master list or only in Source Y are not used.

2.1.3. Rank countries by Source Y score. Countries identified in the previous step are ranked from lowest to highest perceived level of corruption according to their Source Y score.

2.1.4. Retain only rank positions from Source Y. For each country, only its position in the ranking is kept from Source Y.

2.1.5. Sort the master list. Scores for the common set of countries in the master list are sorted from the lowest to the highest perceived level of corruption.

2.1.6. Delink scores from countries in the master list. Scores in the master list are linked to positions in the ranking rather than to specific countries.

2.1.7. Match master list scores to countries by ranking. The country ranking first in Source Y (lowest perceived corruption) receives the highest master list score (lowest perceived corruption), and so on down the ranking. For countries not included in the previous CPI edition, the score is set through linear interpolation between the scores of the two neighboring countries, taking into account the distances to those neighboring scores in the original source.

2.1.8. Handle ties. The matching-percentile technique is not designed to handle ties in rankings or master list scores. The following rules apply:

If two countries are ranked in the same position by Source Y but have different master list scores, both receive the simple average of their two scores.

If two countries are ranked differently by Source Y but have the same master list score, both receive a new score calculated through interpolation that takes into account the scores they receive from Source Y and the scores in both the master list and Source Y from their upper and lower neighbors.

Step 2.2 — Beta-Transformation. The second step of the rescaling process applies a beta-transformation to the matched scores obtained in Step 2.1. The beta-transformation increases the standard deviation of these values to counter the statistical effect by which the matching-percentiles technique produces a smaller standard deviation each year. The transformation uses the Cumulative Distribution Function of a variable that follows a beta distribution. The alpha and beta parameters are set so that the mean and standard deviation of the index match those of the master list. For the CPI 2010, the parameters were set at α = 1.121 and β = 1.1454.

The final CPI score for a country is the average of these transformed values across all sources in which it appears. Only countries assessed by three or more sources are included in the index.

Confidence Intervals. Confidence intervals indicate the reliability of the CPI scores. They provide the range within which the true value of the estimated CPI score is plausibly thought to fall. The width of a confidence interval provides information on the level of uncertainty of the true value: the wider the interval, the less precise the estimated score. In general, one can state with 90% confidence that the true value of a corruption perception score lies within the constructed 90% confidence interval, corrected in an appropriate manner.

Intervals for the CPI 2010 are built as follows: each source provides an assessment of corruption perceptions in a given country in a given year, and each source is assumed, on average, to correctly capture the underlying phenomenon — that is, to provide an unbiased estimate. However, each source is a somewhat noisy measure of the phenomenon; otherwise all sources would agree on the assessment of every country. The variation (or disagreement) across sources can be exploited to estimate how precise those sources are, and therefore how precise the index estimate is.

Each source could have provided a slightly different value due to random noise. The approach adopted to calculate the confidence intervals is bootstrapping. This method exploits the variation across sources in the evaluation of a given country to create alternative scenarios in which slightly different values are provided by each source for the same country. For each country, 10,000 samples were drawn with replacement from the observed values of the individual sources — that is, alternative configurations of values were constructed starting from the set of values actually observed in the data.

References

Cobham, A. (2013, July 22). Corrupting perceptions: Why Transparency International's flagship corruption index falls short. Foreign Policy.

Newman, M. E. J. (2006, May 26). Power laws, Pareto distributions, and Zipf's law (p. 11). Retrieved from http://arxiv.org/PS_cache/cond-mat/pdf/0412/0412004v3.pdf

Podobnik, B., Shao, J., Njavro, D., Ivanov, P. C., & Stanley, H. E. (2008). Influence of corruption on economic growth rate and foreign investment. The European Physical Journal B, 63(4), 547.

Saisana, M., & Saltelli, A. (2012). Corruption Perceptions Index 2012 statistical assessment. European Commission Joint Research Centre.

Shao, J., Ivanov, P. C., Podobnik, B., & Stanley, H. E. (2007). Quantitative relations between corruption and economic factors. The European Physical Journal B, 56(2), 157.

Taleb, N. N. (2010). The black swan: The impact of the highly improbable (2nd ed.). New York, NY: Random House.

Wilhelm, P. G. (2002). International validation of the Corruption Perceptions Index: Implications for business ethics and entrepreneurship education. Journal of Business Ethics, 35(3), 177–189.

Bureau of Labor Statistics. (2014, December). Consumer Price Index. Retrieved from http://www.bls.gov/cpi/

Transparency International. (2014). Corruption Perceptions Index 2013. Retrieved from http://www.transparency.org/research/cpi/overview

Transparency International. (2010). Corruption Perceptions Index 2010: Long methodological brief (Report). Transparency International.

Organization for Economic Co-operation and Development. (2014). About the OECD. Retrieved from http://www.oecd.org/about/

World Economic Outlook. (2014, October 7). International Monetary Fund. Retrieved from https://www.imf.org/external/pubs/ft/weo/faq.htm