two-stage least squares

Two-stage least squares (2SLS) is a statistical method used to accurately estimate relationships between variables when there is concern about "endogeneity"—situations where an explanatory variable is influenced by other factors, making simple analysis misleading. In the first step, it identifies an external, unrelated variable (an instrument) that affects the problematic variable but not directly the outcome. In the second step, it uses this instrument to replace the problematic variable, helping to reveal a more accurate estimate of its true effect on the outcome. This process improves the reliability of causal inference in complex models.