To solve this problem, we need to exploit instrumental variables that generate plausibly exogenous variation in the incidence of TELs. A valid instrumental variable (instrument) must meet two requirements: first, theinstrument must be relevant; that is, the variation in the instrument is related to variation in the endogenous regressor; second, the instrument must be exogenous, that is, the variation in the instrument is not related to the error disturbance (Stock and Watson 2003). Applied to this study, the instruments must be directly related to the possibility of a state passing a TEL, but not directly related to other unobserved factors affecting the level of municipal revenues per capita. In other words, the instruments must affect municipal revenue only indirectly through the passage of TELs.