Question

Question 1

Suppose you are running a regression of crime rates (dependent variable) on police expenses (explanatory variable). If there is simultaneous causality, i.e. higher crime rates draw in more police resource, which leads to higher police expenses, then the estimated beta coefficient on police expenses will:

Question 1 options:

	be consistent
	have a positive bias.
	be unbiased
	have a negative bias

Question 2 (1 point)

Using a binary variable as the dependent variable means that:

Question 2 options:

	The adjusted R-squared is no longer a meaningful statistic.
	Errors are always heteroskedastic
	The explanatory variables cannot be binary variables.
	Both (a) and (b)

Answer 1

1) Have a positive bias. If Y causes X but at the same time X also causes Y, this problem arises. The explanatory variable gets correlated to the error term, thus creating a bias. This causes the OLS estimates to be biased and inconsistent. The amount of bias is positive as the explanatory variable is positively related to the error term.

2) Errors are always heteroskedastic. By its inherent construction, heteroskedasticity is always present in such models.