THE National Corruption Perception Survey (NCPS) by Transparency International Pakistan 2025 has once again confirmed what many Pakistanis suspect: people believe corruption is widespread, deep-rooted and structurally embedded across the state. Police, tenders and procurement, and the judiciary top the list of perceived offenders. However, a staggering 66 per cent of respondents say they did not personally face a situation requiring them to pay a bribe in the last year. A majority believe their purchasing power has declined, that provincial governments are more corrupt than local ones, and that anti-corruption institutions need oversight themselves. In the health sector, respondents describe corruption as having a “very high impact” on people’s lives. Across nearly every question, citizens express mistrust of public services, of state capacity, of political neutrality, of institutional fairness.
Year after year, Transparency International’s findings shape headlines and political rhetoric. Politicians pick up select statistics to batter their rivals. Television screens host endless discussions on whether corruption has increased or decreased. Academics cite the results as useful indicators of public sentiment. Commentators and columnists use NCPS numbers to summarise, and sometimes oversimplify, public mood. The report has become a reference point in national discourse, often treated with the same authority as an official statistical document.
But it is time we interrogated this unquestioned acceptance. The NCPS is an ambitious exercise, but ambition alone does not make it methodologically sound. One can argue that the underlying methodology is flawed to the point that the results cannot reasonably claim to represent the state of corruption perceptions in Pakistan. The entire exercise, as currently designed, defies the most basic standards of social science. And this matters, because flawed data does not merely misinform, it reshapes public discourse, reinforces political narratives, and feeds into a cycle of self-perpetuating pessimism.
The sampling strategy is the first and most serious concern. The survey claims to employ “multistage stratified cluster sampling”, but in practice it appears neither stratified nor random. The design allocates an identical sample size to each province, 1,000 respondents, regardless of population share. Thus, Punjab, home to over half the country’s population, receives the same weight as Balochistan, with barely six per cent of the population. This is not stratification; it is arbitrary equal allocation.
Districts are selected by zoning rather than genuine probability proportionate to size. Within districts, clusters are not randomly selected households but public spaces; markets and parks, chosen at convenience. Respondents are approached on the spot, quota-balanced by gender and disability, but not randomly sampled from the population. This is not probability sampling; it is convenience sampling with quotas. A survey built on non-random selection cannot generate national or provincial estimates, cannot calculate margins of error, and cannot claim representativeness. Yet the NCPS reports ±3pc margins of error, statistically impossible under such a design.
It is time we interrogated the unquestioned acceptance of the NCPS.
Who appears in these public-space interviews? Those who happen to be out in markets during enumerator hours: overwhelmingly urban, mobile, younger and more male than the actual population. Who is missing? Women confined to domestic spaces, the elderly, rural households far from public venues, those without mobility, and millions who simply do not frequent the selected locations. The resulting sample excludes large segments of society, yet the results are presented as reflections of national opinion.
But even if the sampling were scientifically rigorous, and it is not, we would still face a deeper problem: corruption perception is not corruption. Measuring perception is not the same as measuring incidence. Perception is influenced by media narratives, political scandal, elite gossip, personal frustration, and public cynicism. Globally the most prominent perception-based measure, Transparency International’s Corruption Perceptions Index has been criticised for precisely this reason. Experts note that perception indices can reflect “noise” rather than reality, and can reinforce stereotypes about corruption rather than reveal actual patterns. Professor Paul Heywood describes the CPI as “good, bad, and ugly”: good for starting conversations, bad for meaningful comparison across countries, and ugly when used as a proxy for real corruption.
Particularly problematic is the temptation to compare perception scores over time. Public perception does not move in tandem with reality. A country can improve procurement processes, strengthen oversight, and reduce opportunities for rent-seeking, yet see no improvement or even a worsening in perception indices. Conversely, perceptions can improve simply because scandals fall out of public sight.
This problem applies to NCPS. When we claim “corruption has increased since last year”, what are we really measuring? Are we capturing an increase in corruption, or an increase in the visibility of corruption? Are we measuring declining trust in government, or an actual surge in the misuse of authority? Are respondents reacting to their own experiences, or to a month of televised political warfare? Perception does not distinguish between structural corruption, petty bribery, predatory bureaucratic behaviour, and political vendettas. It collapses complexity into a single sentiment: things feel corrupt.
Corruption perception surveys should be treated as what they are: barometers of public frustration, not forensic diagnoses of governance. They can tell us how hopeless people feel, how disillusioned, how angry but not how corrupt the country actually is. For measuring corruption itself, we need better indicators, administrative data, procurement transparency metrics, audit reports, and sector-specific indicators. The tragedy is not that NCPS exists; the tragedy is that we treat it as canon. Pakistan needs credible, scientifically designed surveys, if any. It needs stratified random sampling, probability-based respondent selection, proportionate provincial weights, and transparent methodology. Without these, NCPS risks misleading more than informing.
Public frustration with corruption is real and deserves attention. But the tools we use to measure it must be credible. A flawed mirror, no matter how frequently consulted, cannot reveal the truth. Only when we begin grounding our analysis in scientific methods, not convenience sampling and perception-based headlines, can we hope to build a serious, evidence-based conversation about corruption in Pakistan.
The writer is an associate professor of economics at the Lahore School of Economics.
Published in Dawn, December 17th, 2025
