Omitted Variable Bias in OLS
Problem statement
The true data-generating process is
$$Y = \beta_0 + \beta_1 X_1 + \beta_2 X_2 + \varepsilon$$
where $\varepsilon$ is mean-zero noise independent of $X_1$ and $X_2$. You run OLS of $Y$ on $X_1$ alone -- dropping $X_2$ entirely.
Derive the expected value of the resulting OLS estimate $\tilde{\beta}_1$. Under what conditions is the bias zero? What determines the sign of the bias?
Show hint
What happens if you substitute the true model into the single-regressor OLS formula?
Relevant roles: Data Scientist, Applied Scientist and Research Scientist.