Overlapping Distribution Model
Discuss group differences without stereotyping individuals
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 96%
The model separates a population-level tendency from a claim about every member of a group. Richard Reeves uses height as the intuitive example: saying men are taller than women describes an average, not a world in which every man is taller than every woman. The same logic can be applied carefully to traits such as risk-taking, competitiveness, or crying. First identify the distribution-level claim, then make the overlap explicit, ask whether the average difference is relevant, and finally refuse to use it as a shortcut for judging an individual. This permits discussion of measured differences without converting them into hierarchy, destiny, or discrimination.
Origin
Reeves attributes the underlying idea to a formulation he associates, uncertainly, with Hans Rosling. Extracted from The Diary of a CEO.
Core principles
- 01Group averages do not determine individual traits
- 02Distributions can differ while overlapping substantially
- 03A difference is not evidence that one group is better
- 04Judge each person on individual evidence
How to run it
- 1
Define the comparison
Specify the populations and the trait being discussed. Keep the claim narrow enough to test rather than turning it into a statement about identity.
Pro tip Use an observable measure whenever possible.
Watch out Do not smuggle several different traits into one broad label.
- 2
State the average
Describe the tendency as an average or probability, not as a universal rule. Make the level of the claim explicit.
Pro tip Use phrases such as ‘on average’ or ‘somewhat more likely.’
Watch out Avoid absolute language such as ‘all’ or ‘always.’
- 3
Surface the overlap
Acknowledge that many people in both groups fall in the same range and that exceptions are expected. This keeps the distribution visible rather than reducing it to two boxes.
Pro tip Use a familiar analogy, such as height, before discussing a contentious trait.
Watch out Overlap does not mean the averages are identical.
- 4
Test relevance
Ask whether the measured difference actually matters for the decision or policy under discussion. A real difference may still be irrelevant.
Watch out Do not infer practical importance from statistical difference alone.
- 5
Judge the individual
When acting toward a person, replace the group prior with evidence about that person. Never use the group average to deny an individual opportunity.
Pro tip Design selection processes that measure the trait directly.
Watch out Population evidence is not a verdict on a person.
In the wild
A founder believes different average risk preferences may add value to a leadership team. Instead of assigning risk tolerance by gender, the founder measures each candidate’s decisions under uncertainty, acknowledges broad overlap, and builds the team from demonstrated complementary judgment.
→ The team gains varied risk perspectives without stereotyping candidates.
Reeves explains that listeners understand ‘men are taller than women’ as an average claim. They do not infer that every man is taller than every woman, which shows how two distributions can differ and still overlap.
→ A familiar example makes the distinction between averages and individuals clear.
Common mistakes
Turning an average into a rule
A group tendency does not establish what any particular person is like. Treating it as deterministic creates the stereotyping the model is meant to prevent.
Treating difference as hierarchy
A difference does not show that one trait or group is superior. Reeves explicitly separates equality of value from sameness.
Assuming every difference matters
Even a well-supported average difference may have no relevance to the decision at hand. Test practical relevance before acting on it.
Is it for you?
Best for
It is best for discussing demographic patterns while preserving fair individual judgment.
Not ideal for
It is not ideal when reliable population-level evidence is absent or the decision concerns a known individual.
From the transcript
“they're not completely separate or completely the same they just they have overlapping distributions”
“never use that as a way to discriminate against an individual”
From the episode
The Gender Expert: Men Are Emotionally Dependent On Women & We're Treating Them Like Malfunctioning Women! Richard Reeves