public sector, since (2.20) and (2.21) imply that this will increase an outsider’s
expected economic gain from voting for that candidate.13 On the other hand, a
relatively high public sector wage rate makes the creation of public sector jobs
more expensive by requiring a higher tax rate. Hence politicians must trade off
the political gain from high public sector wages and public sector job creation
against the political cost of having to raise taxes. The next section analyses the
resulting political equilibrium.
3. Political equilibrium, tax competition and rents
3.1. Political equilibrium
In political equilibrium the fiscal policy variables W , G,andτ are set so as to
maximise the probability of election victory (2.20), subject to the government
budget constraint (2.22) and the recruitment constraint (2.23). The first-order
conditions for the solution to this problem are derived in Appendix 2. When the
constraint W ≥ w is not strictly binding, these conditions can be shown to imply
that
u0g
= (α+δ ) (1
u0p,
ε≡
μ n — 1 ∖ τk0
n k,
(3.1)
g0 (α)
ug
up
u0g (1 + δ)
= +++δ ∖ -
α+ αδ
1+
α(n
1)
(1 — α) (n
+)l (W)
FL,
(3.2)
where ug ≡ u (W + rk} and up ≡ u (w + rk} are the total utilities of private con-
sumption for public and private sector workers, respectively; ug ≡ u0 (W + rk}
and up ≡ u0 W + rk} are the corresponding marginal utilities; and ε is the nu-
merical elasticity of the tax base with respect to the tax rate.14
To understand the effects of tax competition on public sector efficiency, it is
useful to start by considering the benchmark case of autarky where no interna-
tional capital mobility is allowed. The world economy will then function like a
13When W > w, we have U0 > Uv, so from (2.21) we get puo = U g- Up > 0. It then follows
, g p, . g ∂α 1-αi .
from (2.20) and (2.21) that 'A = Ug — Uo > 0.
14 Note that ε is a general-equilibrium elasticity, allowing for the impact of a change in the
domestic tax rate on the world interest rate. Specifically, the tax base elasticity is defined as
d(k(r+τ)) τ
ε ≡---
dτ k
k0 ∙ (dτ + ∂ Tdτ) τ
dτ k
/ n — 1 ∖ τk0
n k.
where we have used the symmetry assumption plus equation (2.13) to derive the last equality.
17