relevant elasticity threshold is below 0.5. However, the traditional tax competition outcome is
not welfare-inferior for all elasticity values below 0.5. Intuitively, for ∣e∣ → 0 the inefficiency
inherent to the tax competition equilibrium becomes small. In contrast, with decentralized
leadership the own tax-revenue effect tiktii + ki is large (as ktii < 0 is small), while the cross
tax-revenue effect tj ktji becomes negligible (as ktji = -ktii is small). A tax rate hike thus induces
a stark rise in state i’s public funds, relative to the neighbor state. Due to the federal gov-
ernment’s concern for horizontal equity the imbalance yields a severe loss of transfers in state
i. Consequently, state i,s incentives to tax capital are weak when ∣e∣ → 0, resulting in starker
efficiency losses relative to tax competition with no federal intervention.
7 Conclusion
This paper analyzes how the pre-commitment ability of government entities affects the efficiency
of resource allocation. The analysis demonstrates that if states are able to pre-commit, the
capital tax rate can be higher or lower relative to the capital tax rate chosen under the Nash-
conjecture. The prospect to receive more federal funds may increase capital taxes above the
level which prevails in tax competition. In contrast, capital taxes are unambiguously chosen at
a higher, but still inefficient level if the federal level can pre-commit. On the expenditure side,
public good consumption choices exhibit a reinforced tendency to “race to the bottom” (relative
to simultaneous decision-making).
Contrasted with the traditional tax competition equilibrium, the paper shows that a federal
intervention robustly (i.e. irrespective of the governments’ pre-commitment capacity) improves
welfare upon tax competition when incentives to compete for mobile capital are sufficiently
strong (high capital tax base elasticity).
The analysis does not resort to corrective grants to address the tax competition externality.
The design of corrective policy when state governments cannot pre-commit is well explored in
24