CPI vs AMI: How the Two Indices Compare

TL;DR

The CPI (Corruption Perceptions Index, Transparency International, since 1995) measures public sector corruption via expert surveys. The AMI (Academic Misconduct Index, 2026) measures academic misconduct via mixed data sources. The AMI uses CPI methodological principles (country-level scoring, periodic updates, open licensing) while differing in scope and inputs.

What the CPI does

The Corruption Perceptions Index is the longest-running global integrity index. Key features:

• Coverage: 180 countries (current version)
• Scoring: 0–100 scale, 0 = highly corrupt, 100 = very clean
• Methodology: composite of 13 expert and business surveys
• Publisher: Transparency International
• Update frequency: annual since 1995
• Licence: open, with attribution

The CPI is the most widely cited integrity index globally. It is referenced in academic literature, policy documents, business risk assessments, and journalism in every country it covers.

What the AMI does

The Academic Misconduct Index is a younger, narrower index focused on academic integrity:

• Coverage: 39 countries (v1.5, May 2026; expanding)
• Scoring: two-axis system (Prevalence P, Response Quality R), each 0–100
• Methodology: composite of live data (Google Trends, Retraction Watch, FOI), surveys (ICAI/McCabe), and literature
• Publisher: independent (Booth, F.)
• Update frequency: irregular currently; periodic
• Licence: CC BY 4.0

Methodological similarities

The AMI takes several methodological principles from the CPI:

Composite index design

Both indices combine multiple data sources into a country-level score. Neither relies on a single measurement.

Country-level scoring

Both score countries rather than institutions or individuals. The unit of analysis is the country.

Rescaling within set

Both rescale scores within the country set rather than producing absolute measurements. A change in country coverage changes the scale.

Transparency

Both publish methodology documents that allow reproduction. Both license data for reuse.

Periodic update

Both are versioned and updated. Older versions remain valid for historical comparison.

Methodological differences

Data sources

CPI: 13 expert and business perception surveys, weighted and combined. All inputs are subjective expert assessments.

AMI: live data sources (Google Trends, Retraction Watch, FOI), self-report surveys (ICAI/McCabe), and literature-derived estimates. Mixed objective and subjective inputs.

Scoring structure

CPI: single score per country, 0–100.

AMI: two-axis system (P, R) with quadrant placement. The two-axis design allows distinguishing high-prevalence-low-response from low-prevalence-high-response situations.

Scope

CPI: public sector corruption (general).

AMI: academic misconduct (specific).

Maturity

CPI: 30 years of methodology development, established expert respondent base, well-understood inter-annual variation.

AMI: first version 2026. Methodology will mature; current scores carry larger uncertainty than mature CPI scores.

What each does well

CPI strengths

• Time series: 30 years of comparable data
• Coverage: 180 countries
• Expert respondent base: mature, diversified
• Recognition: cited in policy and media globally
• Funding stability: established institutional sponsor (Transparency International)

AMI strengths

• Live data: Retraction Watch and Google Trends provide current signals rather than survey lag
• Two-axis design: captures the prevalence-response interaction directly
• Dimension breakdown: D1–D6 plus R-Score components allow policy-specific analysis
• Open methodology: source code and dataset published

What each does less well

CPI weaknesses

• Perception-incidence gap: measures perceptions, not confirmed corruption
• Expert sample bias: respondents are typically business and policy elites
• Annual update lag: signal slow to capture rapid change
• Single score: cannot distinguish high-corruption-low-response from other patterns

AMI weaknesses

• Methodology immaturity: first version 2026, known limitations including Norway anomaly
• Coverage: 39 countries (vs CPI's 180)
• No expert panel yet: planned for v2.0
• Survey data age: McCabe surveys are 2002–2015; pre-AI era
• Detection-incidence confound: stronger detection countries report more cases

How the AMI plans to evolve

The AMI methodology document indicates v2.0 will:

• Add expert perception surveys modelled on the CPI
• Expand country coverage
• Add country-level subject vulnerability scores
• Improve language-disambiguated Google Trends signal interpretation
• Address the detection-incidence confound more directly

The progression mirrors the CPI's history: the CPI's first version in 1995 covered 41 countries with relatively simple methodology; the current 180-country index with mature expert panels developed over decades.

Using both together

CPI and AMI scores correlate substantially. Countries with high CPI scores (clean) generally have low AMI Prevalence and strong Response. Countries with low CPI scores often have high AMI Prevalence.

The correlation is not perfect — academic misconduct can occur in countries with otherwise strong governance, and vice versa. Researchers and analysts using both indices together can identify cases where the two diverge, which are often the most informative for understanding country-specific dynamics.

Sources

• Transparency International CPI methodology
• AMI v1.5 methodology document
• CPI historical archive (1995–present)

Full methodology | Download dataset

Related

Read the full methodology