You've built an anti-corruption training program. You've rolled out policies, run workshops, and sent reminders. Now someone asks: "Is it working?"
If your answer relies on asking employees whether they think corruption has decreased, or if you're tracking your company's score on a global corruption index, you're measuring the wrong thing. This confusion persists because we've inherited a flawed assumption: that corruption is something you can define once and measure everywhere.
It's not. Here's what you're getting wrong, and what to measure instead.
Myth 1: You Need a Universal Definition of Corruption
The Reality: Corruption is a value judgment, not a neutral fact.
When the drafters of the U.N. Convention Against Corruption sat down to define the term, they concluded that "corruption is a fluid concept, signifying different things to different people." Instead of forcing a definition, UNCAC lists specific acts countries must criminalize: bribery, embezzlement, and others.
Even Transparency International's definition ("the abuse of entrusted power for private gain") is circular. You're defining corruption with corruption.
The word originated as a descriptive term (from the Greek phthora, meaning decay), and it still works that way in some contexts. A corrupted computer file is a verifiable fact. But when you call a business practice corrupt, you're making a judgment about whether it deviates from what you believe is right. Reasonable people disagree about where that line sits, which is why 117 different definitions exist and none has won universal acceptance.
For your training program, this means: stop trying to measure whether your workforce thinks "corruption" has improved. They're each defining it differently.
Myth 2: Cross-National Corruption Indices Tell You If Your Program Works
The Reality: These tools aren't built to measure your progress.
Your company operates in multiple countries, and you're tracking its Corruption Perceptions Index score year over year to see if your ethics program is making a difference. The problem: that's not what the CPI measures.
The CPI exists for advocacy, to generate attention and pressure governments. It aggregates perceptions from business executives, risk analysts, and country experts. It's designed to compare countries against each other, not to track whether a specific intervention worked.
UN agencies, Sweden's Quality of Government Institute, and researchers have repeatedly warned against using these indices to measure change over time. The methodology shifts, the respondent pool changes, and external factors can swing scores regardless of what your company does.
If you're using the CPI or similar indices to justify budget requests or report program outcomes, you're building on sand.
Myth 3: Employee Perception Surveys Are Enough
The Reality: Perceptions don't track conduct.
You survey employees annually: "Do you believe corruption is a problem in this organization?" Scores improve, and you declare success. But what changed?
Perception surveys measure comfort, not behavior. Employees might feel better because you talk about ethics more often, not because fewer people are paying bribes or falsifying records. In authoritarian markets, employees might underreport problems out of fear. In cultures with strict enforcement, they might overreport because standards are higher.
Perceptions matter for understanding workplace climate, but they can't tell you whether corrupt conduct has decreased. You need direct measures of the conduct itself.
Myth 4: One Number Can Capture Corruption Risk
The Reality: Aggregating disparate measures into a single score obscures what's actually happening.
Some measurement systems combine responses to different questions into one composite number. The higher the number, the theory goes, the less corruption.
This approach treats corruption as a measurable, objective phenomenon. It's not. Corruption is what philosopher W.B. Gallie called an "essentially contested concept," like democracy or social justice. Reasonable people will never converge on a single definition, so any claim that a composite score "measures corruption" is a hypothesis that can't be empirically validated.
For your compliance program, a single score hides the details you need. Are kickbacks increasing while embezzlement decreases? Is fraud concentrated in one region or business unit? A composite number can't answer these questions.
Myth 5: You Can't Measure What Matters Without a Definition
The Reality: You can measure specific conduct without defining corruption universally.
Here's the shift: stop measuring corruption. Measure corrupt conduct.
UNCAC's drafters understood this. They couldn't agree on what corruption meant, but they could agree that bribery and embezzlement were wrong. Those are facts you can verify: a payment was made or it wasn't, funds were diverted or they weren't.
Your training program should do the same. Identify the specific acts you want to prevent: bribes to foreign officials, conflicts of interest not disclosed, falsified expense reports, procurement fraud. Then measure those directly.
What to Do Instead
1. Define the conduct you're trying to prevent.
List the specific acts your program addresses. Don't use "corruption" as a catch-all. Use "bribery of government officials," "kickbacks to vendors," "undisclosed conflicts of interest."
2. Track conduct-based metrics within your organization.
Measure what you can verify: reports of bribery, substantiated conflicts of interest, disciplinary actions for fraud. Track these over time and by business unit. When the U.S. Internal Revenue Service wanted to know if fraud detection in a cash assistance program was working, they didn't survey perceptions. They measured how much fraud their system caught and tested whether it was missing more.
3. Use perception data carefully.
Employee surveys can tell you whether people feel safe reporting misconduct or whether they understand your policies. But don't conflate comfort with compliance. If perception scores improve but your conduct metrics don't move, investigate why.
4. Measure program reach and quality, not abstract outcomes.
Track completion rates for your anti-bribery training. Measure whether employees can identify a bribe scenario in a quiz. Monitor whether managers are conducting the required due diligence on third-party intermediaries. These process measures tell you whether your program is functioning, even if corrupt conduct is rare enough that year-over-year changes are statistically noisy.
5. Accept that some things can't be measured precisely.
If kickbacks are happening, your detection system might not catch them all. That's not a reason to give up on measurement. It's a reason to be honest about what your data shows and what it doesn't. Build multiple indicators that triangulate toward the truth.
The goal isn't a perfect corruption score. It's a program that reduces specific harms and a measurement approach that tells you whether you're succeeding. You can't manage what you can't measure, but you also can't measure what you can't define. So stop trying to measure corruption. Measure the conduct that matters, and build your program around that.
