Is it true that only 4% of crimes are attributable to mental illness

1. What the “4%” claim actually refers to

When commentators say “only 4%” or similar, they are usually invoking population-attributable-risk estimates from epidemiological studies that calculate the share of violent incidents that would be removed if a specified condition (here, major mental illness) did not exist; one review reports roughly 3% attributable to major mental disorders in some samples and another paper cited an estimated 4.3% for major mental illness in a general-population calculation ^{[1] [3] [2]}.

2. Why estimates vary: definitions, samples and methods

Variations arise because studies use different definitions (any mental disorder vs severe mental illness such as schizophrenia or bipolar), different outcomes (all crime, violent crime, homicide, or justice-system contact), and different populations (general population vs detainees or forensic samples); for example, an England and Wales analysis estimated 5.3% of violent incidents were committed by people with severe mental illness in 2015–16, while Canadian and US analyses have returned figures near 3% for non–substance-use disorders in some datasets ^{[4] [2] [1]}.

3. The outsized role of substance use and comorbidity

Across reviews and primary studies, substance use disorders—alone or combined with mental illness—explain a far larger share of violence: estimates often show substance abuse accounting for roughly 30–34% of the population-attributable risk, and comorbid substance use plus mental illness adding additional risk ^{[1] [3] [2]}. Many studies find that when substance use and socioeconomic confounders are controlled for, the independent effect of mental illness on violence shrinks substantially ^{[1] [5]}.

4. Crimes directly caused by psychiatric symptoms are rarer still

Detailed interview-based research of offenders with diagnosed disorders shows that a minority of offenses are directly and primarily driven by psychiatric symptoms—one study coded 429 crimes among 143 offenders and concluded most offenses were not directly preceded by symptoms, and an APA-reported study found only 7.5% directly related (rising to 18% if “mostly related” cases were included) ^{[6] [7]}. That distinction matters: being an offender who has a mental disorder is not the same as committing an act primarily caused by acute psychiatric symptoms.

5. Conflicting claims and advocacy-driven narratives

Some advocacy groups and policy reports emphasize higher figures for particular outcomes—e.g., a Treatment Advocacy Center report claims people with schizophrenia and bipolar disorder account for about 10% of homicides and a larger share of mass killings—which contrasts with population-attributable estimates and can reflect selection bias in homicide samples or differing aims to push for civil-commitment or treatment policies ^[8]. These divergent numbers reflect differing agendas, sample frames, and what is being measured (homicide vs all violent crime).

6. Bottom line and limits of the evidence

The evidence supports saying that a small percentage—commonly cited in the 3–5% range—of violent crimes are attributable to non–substance-related mental illness in broad population studies, while substance use and social determinants account for a much larger share; however, results vary by country, crime type, and method, and some data and claims (especially about homicides or mass killings) diverge from these averages and are influenced by sample selection and advocacy aims, so no single percentage can capture the full complexity ^{[1] [3] [2] [4]}. Reporting and policy should therefore distinguish between (a) prevalence of mental illness among offenders, (b) crimes directly caused by psychiatric symptoms, and (c) the population-attributable proportion—each yields different numbers and policy implications ^{[6] [7]}.