Skip to main content

Two-stage least squares (2SLS) is a structural equation modeling technique that is used when the IV is correlated with the DV's errors.

Two-stage least squares (2SLS) is a regression technique used in structural equation modeling (SEM) which is primarily used when the dependent variable's error term is correlated with an independent variable.

To get around the potential bias created by this situation, 2SLS is estimated via two separate stages:

  1. First stage (the reduced form of the predictor): Regress the primary predictor, here made as an endogenous variable, on all instruments and all exogenous variables.
  2. Second stage: Substitute predicted values of exogenous variable in the original model and apply OLS.

A summary on this technique can be found here.