Separation of Powers, Line Item Veto and the Tax Level: Evidence from the American States Draft 1

Line item veto, a feature present in most American States, gives the governor the power to veto single appropriation items from the budget. Its effects on the tax level, however, are still controversial in the empirical and theoretical literature (cf. Holtz-Eakins (1988) and Besley and Case (2003)). Line item veto is mostly a time invariant feature and to asses its effects previous studies have interacted it with political control variables such as a divided government. The endogenity problems that arise from using a political variable to explain a policy variable, however, have not been dealt with in these studies. We use three empirical approaches to tackle the problem and show that line item veto does have a significant negative effect on the tax rate in the States: diffs-in-diffs estimation with instrumental variables (election results in lower offices at the state level), regression discontinuity design, and a dynamic panel. Our prior on its effects comes from adapting the separation of powers model by Persson, Roland and Tabellini (2000) to the American States setup: we add line item veto and an executive. Our model delivers a clear prediction on the tax level, on the amount of public good, and on the importance of group specific transfers.