Suppose the true regression model relating X1 to Y should also contain a second independent variable X2. Show mathematically the amount of omitted variables bias if I estimate a regression containing only X1 as an independent variable. When is omitted variable bias not a problem?