Background: Risk factors for violent behaviour may differ depending on whether this begins before (VBO) or after (VAO) the onset of schizophrenia spectrum disorder. However, previous studies have been limited by selective samples of forensic patients and crude outcome measures. Methods: The sample consisted of 1013 patients with schizophrenia spectrum disorders recruited from various treatment settings across the Netherlands. Putative risk factors and outcomes were measured with standardised instruments. We used logistic regression models to compare patients with VBO (n = 48), patients with VAO (n = 130) and nonviolent (NV) patients (n = 708) on each risk factor, adjusting for sex and age. Results: Patients with VBO more often lived in a socially disorganised neighbourhood than NV patients (adjusted odds ratio [aOR] 3.3, 95 % confidence interval [CI] 1.3–8.0) and patients with VAO (aOR 3.3, 95 % CI 1.1–9.6). Clinical risk factors were more prevalent in patients with VAO than in NV patients, with substance misuse (aOR 1.5, 95 % CI 1.0–2.3), impairments in executive functions (aOR 1.6, 95 % CI 1.0–2.4), poor impulse control (aOR 2.4, 9 % CI 1.5–3.6), delusions (aOR 1.5, 95 % CI 1.1–2.3) and lack of illness insight (aOR 1.5, 95 % CI 1.0–2.2) reaching statistical significance. Patients with VBO were also more likely to have poor impulse control than NV patients (aOR 2.6, 95 % CI 1.3–5.1). Conclusion: Strategies to predict and prevent violence in schizophrenia spectrum disorders should distinguish between VBO and VAO.