Proposition 1: Under centralized leadership the tax mix becomes more efficient relative to
the Nash outcome, i.e. tc > tN and τc < τN.
At this point it might be informative to relate the result to the analysis in Boadway and
Keen (1996) and Boadway et al. (1998). In these contributions the federal government can
replicate the unitary nation optimum by pre-committing to its policy. Besides the assumption
of pre-commitment, the result is driven by a perfect tax base overlap.20 Setting a federal tax
rate, which, when summed up with the local tax, yields the overall level of taxation chosen in a
unitary state, the federal government can mitigate any inefficiency in local tax policy.
With vertical tax base independency, the federal government is only indirectly able to con-
trol lower level governments’ policy choices (via the state reaction function). The unitary state
outcome of a zero labor tax would eliminate labor supply distortions, but in the presence of
decentralized capital taxation would also expose the public expenditure level to a downward
pressure exerted by fiscal competition among states. Weighting both effects the federal govern-
ment provides transfers (and therefore uses labor taxes) up to the point where the FMCPF,
adjusted for the adverse impact on state capital taxes, is equated to the SMCPF.
The implication for the level of public good provision is:
Proposition 2: Under centralized leadership local public good provision decreases relative to
the Nash level, i.e. gc < gN.
Proof: The proof involves a comparison of the SMCPF evaluated at (tN, τN) and at (tc, τc),
respectively. Based on the first-order condition (9) we then infer how public spending levels will
20The modelling choice implies that federal and local taxes are perfect substitutes in the private agents’ decision
problem. This is substantially different to our analysis where both taxes do not interact - see Eq. (12).
15