Why stated values drift from behaviour, and how to measure the gap
Declared values are easy to publish and hard to test. A forecast-style method, borrowed from research on forecasts and surveys, shows where the gap sits.
Most organisations can state their values. Far fewer can show, from their own records, that those values shaped a decision in the past year. The gap between the two is common, and it is rarely measured. None of the five sources on our fact sheet addresses leadership values or value-behaviour gaps directly. What follows therefore draws on how researchers measure forecasts, opinions and observed conduct elsewhere, plus our own judgement about how a board can apply those methods. Each source is named where we rely on it.
How it works
A stated value is a commitment written in words: integrity, customer first, learn from failure. An observed behaviour is something that leaves a trace in a record: a budget line, a promotion, a rejected proposal, a founder who stopped funding a project. The gap is the distance between the two. It is only useful once both sides have been measured on the same terms. Most boards measure the first and assume the second.
Research bodies show the difference clearly, even when they are not writing about companies. Brookings says on its website that it upholds the highest standards of research quality and professional ethics. That is a declared standard. The statement itself gives no behavioural data, so a reader cannot tell from it whether the standard is met. It can only tell you what is claimed.
Pew Research Center shows the same split in opinion surveys. In a 37-country survey, Pew reports pessimism about AI's impact on jobs and economic inequality. That is a stated attitude. It tells you what people say they expect, not what they have done about it. By contrast, Pew's content analysis of health and wellness influencers found that they promote a product or service in 38% of posts. That is observed behaviour, counted from the posts themselves.
- Gap between 2024 US healthcare spending and 2010 projections, per NBER
- 15%
- US medical spending as a share of GDP in 2024, up from 17.2% in 2010 (NBER)
- 18.0%
- Health and wellness influencer posts promoting a product or service (Pew)
- 38%
Measurement brings its own uncertainty. Pew's Courtney Kennedy describes the margin of error as the range within which the true answer likely falls. The same logic applies to any internal gap. A single percentage for how often a value was honoured needs a range around it, and that range may be wider than the finding itself. Pew also reports that preelection polls captured the direction of this year's election but generally understated the size of the Democrats' advantage. The direction was right and the magnitude was wrong. A gap measure can fail in the same way.
The most usable method on the fact sheet comes from NBER work by Cutler and Klarnet on US medical spending. They compare actual expenditure with the 2010 actuarial forecast from CMS, then adjust for differences that were anticipated when the forecast was made. What remains is the part of the gap that was not expected. That two-step structure is the one to borrow. First record what was promised. Then separate what was foreseeable from what was not.
The method transfers to values with one change. In a health forecast, the forecast is an actuarial projection made years earlier. For a value, the forecast is the commitment itself, read as a claim about what would be done. Write the claim down in terms a record could contradict. We put customers first cannot be contradicted by anything, which is why it survives every annual review. We do not approve a price change on a contract without the account lead's sign-off can be checked against the approval log.
Who it affects
The people most exposed to a value-behaviour gap are those who rely on a published commitment to make a decision. A board member reading a values statement before approving a senior hire is one. A founder deciding whether to keep funding a venture is another. The NBER field experiment by Bailey, Fehder, Floyd, Hochberg and Lee, published as Working Paper 34755, is a randomised, multi-year, multi-site study of innovation-driven entrepreneurs. It tests whether training changes quit decisions. The fact sheet does not give its results, so we draw no conclusions from it.
The study matters here for a narrow reason. Quitting is a behaviour that can be recorded, dated and compared with what a founder said they valued. The NBER summary says entrepreneurship training aims to improve founder success and may also help founders recognise when to stop investing. A founder who says they value disciplined capital allocation, and keeps funding a stalled project, shows a gap in exactly the place a record can capture it. The decision to stop is a useful test because it is rarely made for free.
Boards carry the heaviest version of the problem. A chair who approves a values statement each year is making a stated commitment on behalf of people who will never read the wording. The chair's test is simple to state and slow to apply: name one decision from the past year that would have gone differently if the stated value had governed it. If no such decision can be named, the value is probably a description of intent, not a rule for conduct.
World Bank programmes, which the fact sheet links to developing countries in Africa and Latin America, describe their data work as essential to guiding decisions, monitoring progress and evaluating impact. For a leader, the useful part is the stated purpose. Indicators exist to track outcomes against goals. A goal that is never measured against outcomes has not been tested, whatever its wording.
What leaders get wrong
- Counting statements. They measure how often a value appears in speeches and handbooks, not how often it decided something.
- Using a survey as a record. Staff saying the culture is open is a stated attitude. It is not evidence about how a particular decision was made.
- Reporting a point. A single figure with no range invites the board to trust a difference the data cannot support.
- Choosing the measure after the result. If the metric is picked once the answer is known, the gap will look as small as the metric allows.
- Ignoring what was foreseeable. The NBER method separates anticipated differences from unanticipated ones. Skipping that step blames behaviour for conditions that were known in advance.
Most of these errors come from the same place. The measure was chosen because it was available, not because it tests the value. Availability is a fair reason to start, but a poor reason to conclude. A board that receives a clean dashboard should ask which decision the dashboard would have failed to show. If the answer is none, the dashboard is measuring activity, and the values remain untested.
What to do about it
Start with three declared values, not twelve. For each one, write down the decision record that would show it in practice: a budget reallocation, a rejected proposal, a promotion, an exit, a deal that went against the easy option. Then write the forecast, meaning the behaviour you would expect if the value were actually held. A value is a forecast about decisions, and it can be checked like one. Keep the wording short enough that a stranger could tell whether it was met.
Set the measurement period before you look at the data. A year is a common choice, but the right length depends on how often the decisions in question occur, and that is a judgement to make from your own records. Count what you can find. Say plainly which decisions had no usable record. A gap you cannot measure is itself worth reporting to the board, because it shows where the values have no evidence behind them.
Measure the gap with someone who did not write the values statement. That person does not need to be sceptical of leadership. They need access to records that leaders do not usually read: approval logs, the reasons given for rejected proposals, exit interviews, and the dates on which decisions were revisited. The result should be reported as a range, with the forecast, the realised figure and the adjustments for conditions known in advance. Keep those adjustments visible, so the board can disagree with them openly.
If your declared values include sustainability commitments, the IFRS Foundation's FSA Credential may be worth knowing about. It has two exams and assesses the materiality of sustainability information for corporate performance and investment analysis. It is a useful reference point when you judge which sustainability disclosures matter, though it is not a method for measuring values. A sensible first measure, in our judgement, is one the board can repeat next year. A gap measured once is a finding. Measured twice on the same terms, it becomes something a board can act on.