the term on the right-hand side. Suppose, by negation, that the inequality is strict, that
is the individual strictly prefers to engage in signaling. By continuity considerations,
an individual with ability slightly lower than w will then also choose to signal, which
would violate the definition of our equilibrium. Thus, we obtain a contradiction.
Clearly, the equality in (5) implies that the signaling constraint in (4) has to bind,
which in turn implies that zs (w) = Z > zNS (w). Mild parametric restrictions on the
utility function in (1) guarantee the existence of a threshold level of contributions, Z ,
which satisfies the equality in (5). For example, when the functions u, h and v are
logarithmic, it is straightforward to verify that the equality in (5) is satisfied.8 As
∂Vs(τw)/∂Z < 0 whereas by construction, ∂Vns(τw)/∂Z = 0, there exists a unique
value of Z for which the equality in (5) is satisfied (for any given set of tax
parameters). To establish the existence of an equilibrium, we need to verify that for
any ability w < w, individuals choose not to engage in signaling; whereas, for any
ability level w > w, individuals do engage in signaling. There are two cases to
consider. Consider first the case where zNS (w) ≥ z. In such a case, Vs (w) > Vns (w),
by construction. By virtue of the strict concavity of the functions u, h and v, charitable
contribution, z, is a normal good; hence, as Z > zNS(w), this may only hold true for
w > w, which is consistent with the separating equilibrium presumption. Consider
next the case where z' (w) < Z. In order to prove the result, it suffices to show that
the following condition holds:
(6) ∂Vs(w)/∂w-∂Vns(w)/∂w ≥ 0.
8 To see this, note, that whenZ = zNS, it follows that Vns(w) < Vs (w) ; whereas, when
Z = [w (1 -1) + T]/(1 + .s`), namely the individual spends her entire potential income on charitable
contributions, consumption and leisure drop to zero, hence Vs ( w) = -∞, which obviously implies that
Vns (w) > Vs (w). Existence follows by the intermediate value theorem.
10