Stigma is usually described in terms of its effect on people after they disclose a mental health problem. Its larger practical effect happens earlier, at the point where somebody decides what to say.
Disclosure is a calculation with real consequences
Deciding to describe a symptom accurately involves weighing consequences for employment, immigration status, custody arrangements, insurance and how one will be treated afterwards.
Those consequences differ enormously between people, which means the same symptom carries a different disclosure cost depending on who has it.
A person who understates a symptom in that situation is making a reasonable assessment of their circumstances rather than failing to engage.
Under-reporting is uneven across groups
Because the costs differ, the degree of under-reporting differs too, and it tends to be greatest where the consequences are most severe.
This means recorded rates of mental health problems partly reflect willingness to report rather than actual prevalence.
Comparing groups on such figures can therefore invert the real picture, showing the lowest rates precisely where disclosure is most costly.
Services are planned from the data that results
Commissioning decisions, staffing and funding allocations are built on recorded contacts and recorded diagnoses, which are downstream of all these decisions.
A population that presents less often generates less recorded need, and less recorded need generates less provision.
The result is a self-reinforcing loop in which the groups least able to disclose end up with the services least designed for them.
The form of the question changes the answer
Response rates differ markedly between anonymous questionnaires, self-completed forms in a clinic and questions asked aloud by a clinician.
Wording matters as well, since questions about specific experiences generally elicit more disclosure than questions using diagnostic language.
This is why prevalence estimates for the same condition can vary widely between studies that were, on the surface, asking the same thing.
Reducing stigma changes the numbers before it changes the problem
Successful anti-stigma work tends to increase recorded rates in the short term, because more people are describing what was already there.
That rise is frequently reported as a worsening crisis, when part of it is improved measurement of a situation that had been undercounted.
Distinguishing the two requires looking at measures less sensitive to disclosure, which is one reason researchers track service contacts and outcomes alongside self-reported symptoms.