Endogeneity
A situation in a statistical model where an explanatory variable is correlated with the error term.
What is Endogeneity?
Endogeneity occurs when a predictor variable in a regression model is correlated with the model's error term, meaning that the variable is not independent of the unobserved factors affecting the dependent variable. This can arise from omitted variable bias, measurement error, or simultaneous causality (where the dependent and independent variables influence each other).
Why Endogeneity Matters
If endogeneity is present, standard ordinary least squares (OLS) regression estimates will be biased and inconsistent, leading researchers to incorrect conclusions about the magnitude or direction of causal relationships.
Example
A researcher assessing the impact of police presence on crime rates faces endogeneity because cities with higher crime rates might hire more police; thus, police presence and the unobserved drivers of crime (in the error term) are correlated.
Common Mistakes
- Assuming that a significant correlation between two variables implies a causal effect without testing for endogeneity.
- Believing that simply adding more control variables completely resolves all sources of endogeneity, such as simultaneous causality.