108
M. Tirelli
extended system of equilibrium first-order conditions, and let F(ξ; ω) = 0
denote such a system (see Section 4 for the precise definitions). The num-
ber of equations in the system is n, equal to the sum of the number n , of
endogenous variables and the number S of policy instruments. The no-tax
equilibrium set is
E = {(ξ,ω) : F(ξ; ω) = 0, t, = 0},
and, when restricted to Ω*, it is a manifold. Next, define the function V :
E → RH such that V (ξ; ω) = (u1 (x1 ), ... , uH(xH)) is the vector of
utility functions. Further, let
G(ξ ; ω) =
F(ξ ; ω)
V (ξ; ω)
To prove Theorem 1 it suffices to show that there exists a marginal tax change
dtI , in an open neighborhood of tI = 0, such that the new equilibrium
(ξ + dξ) is Pareto superior, that is, Dξ V = {du1, ... , duH) ^ 0. This can
be done by proving that G behaves locally as a submersion, i.e.,
gg ( ξ ' t'∙,Λ-( DF ( ξ ' ∙t' ; ω )ʌ
DG(.ξ,t ∙ ω = (dv ( ξ.,t ; ω ) )^j,ω) ∈ E
has full row rank. In fact, observe that, when the subspace spanned by the
rows of DG is equivalent to Rn+H, it includes vectors whose first n entries
are zero and the remaining H positive:
Dξ , Fdξ , + DtI, Fdt1 , = 0,
Dξ, Vdξ, + DtI, VdtI, > 0
for some tax reform, dt', and corresponding changes in the endogenous
equilibrium variables dξ, .7
An equivalent, but instructive, way to describe our proof is as follows.
Assume that G is a submersion at a no-tax equilibrium, for an economy ω.
Then, rDG (ξ,, t', ω) = 0 if and only if r = 0, i.e., the following system
has no solution,
r^GG(ξ,, t1 ; ^
у Ii r Ii-1
= 0.
(1)
This has a natural interpretation in terms of standard welfare analysis. As-
sume that a planner chooses (ξ,, t') such that first-order (necessary) con-
ditions for the optimality of the equilibrium allocation hold: the welfare
function we have in mind is W = ɪ^h ah^h {χh} + μF (ξ,, t' ; ω), with
7 Constrained suboptimality is equivalent to the existence of a feasible direction of tax
reforms in the sense of Guesnerie (1977).