Pakistan’s corruption ranking improves, yet data warrants closer look

• CPI ranking can contribute to broader country-risk picture.
• IMF says corruption, governance weaknesses affect investment.
• Transparency says scores more meaningful than CPI ranks.

ISLAMABAD: Pakistan’s score on Transparency International’s Corruption Perceptions Index (CPI) has improved by one point, but an examination of the data and methodology behind the ranking raises questions about how the score is produced, how much of it can be independently verified and how closely international perceptions correspond with the live experience of ordinary Pakistanis.

The questions matter beyond the ranking itself. Corruption and governance indicators form part of the wider information available to foreign investors, businesses, lenders and international institutions when assessing country risk. The CPI is not an investment or credit rating, nor an IMF lending condition, but it can contribute to the broader country-risk picture.

The IMF’s 2025 Governance and Corruption Diagnostic for Pakistan says corruption and governance weaknesses significantly affect economic growth and investment. Citing research, it said every one-point improvement in the CPI is associated with a 4% increase in a country’s investment rate, while stressing that this is an association, not proof of direct causation.

Transparency International’s 2025 CPI, released in February 2026, gave Pakistan 28 out of 100, ranking it 136th among 182 countries, compared with 27 and 135th among 180 countries previously. Thus, the score rose by one point while the numerical rank fell by one place. TI says scores are more meaningful than ranks because changes in the number and performance of countries can affect rankings.

What is behind Pakistan’s 28 score?

Pakistan’s 2025 CPI score was not based on a survey of Pakistani citizens. TI uses 13 external data sources globally, with each country assessed using at least three; Pakistan had eight. The CPI measures perceived public-sector corruption among experts and businesspeople.

Five of Pakistan’s eight source scores remained unchanged, two declined and only V-Dem increased substantially. Bertelsmann stayed at 21, EIU at 18, Global Insights at 32, PRS ICRG at 33 and World Bank CPIA at 39. WEF fell from 33 to 32 and WJP from 26 to 25, while V-Dem rose five points from 14 to 19.

The published 2025 source scores total 219, giving an average of 27.375, compared with TI’s published CPI score of 28. This is not evidence of a mathematical error because TI uses unrounded underlying scores. The exact calculation therefore cannot be reproduced from the published whole-number figures. The key transparency questions are the exact unrounded average and how close it was to the threshold for a score of 28, particularly because V-Dem accounts for the visible improvement.

Why did V-Dem move five points?

The five-point rise in V-Dem’s Pakistan score is the only substantial positive movement among the eight CPI sources. V-Dem uses expert assessments and statistical modelling, and its datasets can be revised as new versions are released.

The key issue is what changed in Pakistan’s assessment: whether the number or composition of experts changed, underlying indicators were revised, historical observations were updated, or the model or methodology changed. The exact unrounded V-Dem score supplied to TI is also not publicly disclosed in the CPI data.

TI referred these questions to V-Dem, which produces the underlying data. With seven of Pakistan’s eight published source scores unchanged or lower, the five-point V-Dem increase accounts for the only substantial improvement in the source-level data behind Pakistan’s one-point CPI increase.

Eight sources, eight different methods

The eight inputs are also not eight identical surveys. The World Economic Forum uses an executive opinion survey. The World Justice Project combines household and expert information. Bertelsmann relies on expert assessments. V-Dem uses expert coding and statistical modelling. The EIU, Global Insights and PRS use country-risk assessments, while the World Bank CPIA is an institutional assessment.

They therefore involve different respondents, experts, analysts, definitions, time periods and levels of public disclosure. The TI standardises the different source scores and combines them into one country score.

Equal treatment after standardisation is a methodological choice. It does not by itself establish that the CPI is flawed. But it raises a basic question: should assessments produced from fundamentally different types of evidence carry the same arithmetic weight in determining a country’s final corruption score?

A second question concerns independence. The eight organisations are institutionally separate, but that does not necessarily mean their information is statistically independent. Different assessors may rely on the same court cases, government documents, media reports, international developments or expert networks.

That does not prove duplication or bias. It simply means that “eight sources” should not automatically be understood as eight completely independent measurements of corruption.

The “2025” score does not represent one year of common evidence

The underlying assessments do not necessarily cover the same period. Some use 2024 information, others particular months of 2025, while some reflect longer assessments. The CPI labelled 2025 therefore does not represent eight simultaneous measurements of calendar-year 2025 corruption.

A late-2025 reform may not appear immediately in every source, while earlier events may continue influencing assessments. A one-point movement should therefore not automatically be attributed to a particular government or reform.

How much of the underlying evidence can be checked?

TI publishes its methodology and explains how sources are selected, standardised and combined, but acknowledges that some underlying information is private or proprietary. This creates a distinction between methodological and data transparency: researchers can understand the CPI’s design but cannot reproduce every Pakistan source score from original observations.

Meaningful scrutiny would require precise country-level data, assessment dates, respondent or expert numbers and types, response rates, sampling and weighting methods, missing-data treatment, unrounded scores, and changes in methodology or expert pools.

The domestic picture is different

The difference between perception and experience is clearer in Pakistan’s own surveys. Transparency International Pakistan (TIP) conducts separate National Corruption Perception Survey (NCPS) and has clarified that it does not collect data or calculate Pakistan’s global CPI score.

The 2025 NCPS covered 20 districts across all four provinces and received 3,989 completed responses. Police were identified as the most corrupt sector by 24%, followed by tendering and procurement at 16% and the judiciary at 14%. 66% said they had not paid a bribe for a public service in the previous 12 months, while 77% were dissatisfied with government efforts to combat corruption.

TIP says it changed from non-probability convenience sampling to a multistage stratified cluster design in 2025. About 1,000 respondents were allocated to each province, with respondents selected at public venues including markets, parks, government buildings and community gathering points.

The published methodology does not detail individual selection within venues, treatment of refusals, replacements or selection probabilities. Equal provincial allocation also requires weighting to represent Pakistan’s population, since the provinces differ in size. TIP does not provide enough information to reproduce the national weighting or selection probabilities.

PHC had earlier found serious problems in TIP survey

A judicial precedent also shows why methodology and data transparency matter in domestic corruption surveys. In April 2024, the Peshawar High Court ordered Transparency International Pakistan to recall and republish its 2023 National Corruption Perception Survey after examining findings concerning the Khyber Pakhtunkhwa judiciary. The court found deficiencies in “ethical and methodological rigour” and held the findings concerning the KP judiciary not credible.

The court questioned TIP’s handling of responses to whether respondents had paid a bribe to access public services. TIP reported 760 respondents who said they had paid a bribe, but the court noted that responses from another 788 respondents had not been adequately explained. It described this as a major data-analysis flaw undermining the finding’s reliability and validity. The court also criticised TIP for failing to disclose that perception does not necessarily mean reality.

The court ordered TIP to recall the report and issue a revised version containing the relevant data and public-opinion ratios.

The judgment does not establish that TIP’s 2025 NCPS has the same weaknesses; its methodology is substantially different. It does, however, provide a documented example of why sampling, responses and data treatment need to be fully disclosed when survey findings are used to report corruption levels.

Sample size alone does not guarantee accuracy

TIP increased its 2025 sample to almost 4,000 respondents. A large sample can reduce sampling error but does not automatically remove selection bias. The methodology cites an approximately three percentage-point margin of error and a cluster design, but design effect, weighting and respondent-selection probabilities are needed to independently assess national precision.

Several questions also have fewer responses than the headline 3,989. TIP does not prominently show separate missing-response categories for every question or fully explain how missing responses were treated. These issues do not establish that TIP manipulated its findings. They establish that the published information is insufficient for an independent survey expert to fully reproduce the national estimates.

What do Pakistanis actually experience?

The difference between perception and live experience is also visible in a separate survey by Ipsos and the Federation of Pakistan Chambers of Commerce and Industry (FPCCI). The Index of Transparency and Accountability in Pakistan (iTAP), based on fieldwork in December 2025 and January 2026, interviewed more than 6,000 respondents across 82 urban and rural districts and over 195 tehsils, with a separate sample of about 300 respondents from government institutions.

The survey found that 68% believed bribery was common in public institutions, while 27 % said they had personally faced a bribe demand. Similarly, 56% believed nepotism was widespread, compared with 24% who reported personally experiencing favouritism undermining merit. While 59% believed public servants became rich through illegal means, only 5% said they had personally observed such illicit enrichment.

These figures neither establish that corruption is low nor disprove the CPI. They show that perception and personal experience measure different aspects of corruption. Citizens, business executives and international analysts may therefore assess the same problem through different types of evidence.

Why the distinction matters

The distinction is often lost when the CPI is reported simply as saying Pakistan is the “136th most corrupt country”. A score of 28 does not mean 28% of Pakistanis are corrupt, public money is stolen, government contracts are corrupt or citizens have paid bribes. It measures perceived public-sector corruption and is an indicator, not an audit.

Dismissing it because it is perception-based would also be wrong. The IMF’s 2025 diagnostic identified governance and corruption vulnerabilities in public financial management, procurement, regulation, accountability and rule of law, saying these constraints private-sector development and economic performance. The issue is whether the evidence behind such indicators is sufficiently transparent for meaningful scrutiny.

A number with an economic consequence

Foreign investors and financial institutions do not make decisions on the CPI alone; they assess wider economic, political, financial and institutional risks. Corruption and governance form part of that country-risk picture. The IMF links governance weaknesses and corruption to investment and growth, and conducted its Pakistan diagnostic because these weaknesses have macroeconomic consequences.

The CPI therefore has economic significance beyond a media ranking. The issue is not whether Pakistan should reject the ranking, but whether a number that can contribute to international perceptions of Pakistan provides enough source-level information for researchers, policymakers and the public to understand how it was produced.

The bottom line

The available evidence does not establish that TI manipulated Pakistan’s CPI score or TIP its national survey. Three measurement systems produce different pictures: the CPI measures perceived public-sector corruption through eight external assessments; TIP’s NCPS measures domestic perceptions using a substantially changed 2025 methodology whose full selection and weighting process is not publicly reproducible; and Ipsos-FPCCI’s iTAP measures perceptions alongside reported personal experiences.

Pakistan’s CPI remained within a narrow range: 28 in 2021, 27 in 2022, 29 in 2023, 27 in 2024 and 28 in 2025. The latest increase coincided with a five-point V-Dem rise, WEF and WJP declines and no change in the other five sources.

The questions are what evidence produced 28, how much can be independently checked, and how closely it reflects citizens’ experience.

What Transparency International says

Transparency International was asked 16 detailed questions about Pakistan’s 2025 CPI. On the rise from 27 in 2024 to 28 in 2025, TI referred to its technical methodology but did not provide the unrounded scores or exact arithmetic. It said V-Dem was the only source showing a positive change; WJP and WEF declined slightly. TI added that Pakistan’s measurement error became smaller because V-Dem had previously provided its lowest source score.

Asked why V-Dem rose from 14 to 19, TI referred the question to V-Dem, which produces the underlying data. TI said none of Pakistan’s historical CPI changes has been statistically significant. It described the five-year trend as “stagnant”, with Pakistan consistently in the bottom quartile. TI also confirmed that it did not conduct a sensitivity test excluding V-Dem, saying the CPI uses all source data and does not simulate results by removing individual country observations.

On the different source methodologies, TI said all measure aspects of public-sector corruption and are highly correlated, referring to an independent statistical audit by the European Commission’s Joint Research Centre. TI said it cannot formally test statistical independence because of confidentiality constraints and instead relies on methodological diversity as a proxy.

Regarding different reference periods, TI referred to its Source Description. On EIU, Global Insights and PRS ICRG, it said they provide the data points needed for the CPI but did not provide details of their proprietary assessments or separate verification process.

TI said an individual country score cannot be reproduced in isolation because rescaling depends on global means and standard deviations. It said, however, that the full CPI has been independently reproduced and statistically audited by the European Commission.

TI said some sources, including V-Dem, WJP and WEF, may have different sample sizes, but their methodologies and questionnaires have remained stable.

On what the CPI measures, TI said no independently verified measure of actual corruption exists and described the CPI as a perception-based proxy for public-sector corruption. It acknowledged limitations and said the CPI does not capture illicit financial flows, private-sector corruption, money laundering, or citizens’ direct experiences and perceptions.

TI also said the CPI does not measure individual government institutions, so it could not identify a Pakistani institution responsible for the 2025 improvement and referred that question to TIP.

TI, therefore, stands by its methodology and says the CPI has been independently reproduced and statistically audited. Its responses also confirm that Pakistan’s one-point improvement was not statistically significant, no V-Dem exclusion test was conducted, statistical independence could not be formally tested and the CPI remains a perception-based proxy rather than a direct measure of actual corruption.

Originally published in The News