The sign of |A1| and |A2| depends on the sign and magnitude of α and β and is thus ambiguous.
Using Eqs. (25) and (26) the slopes of the best-response functions (24) are:
∂si _ β ( tikii + ki ) + α
∂ti = ∣A∣ ’
∂sj βtj kji - α
∂tf = Й ’
∂τ b00(gi)b00(gj)tjktji + b00(gi)b00(gj)(tiktii + ki)
∂)ti = Й
A.2 Proof of Proposition 3
Proposition 3 is proved by establishing two lemmata. Lemma 1 states conditions for td to be
greater, equal, or lower than tN . Lemma 2 presents a refined condition pertaining to cases (i),
(ii), and (v) listed in table 1.
Lemma 1: If siti ≥ 0 the capital tax rate under decentralized leadership exceeds the capital
tax rate under both-sided Nash behavior, i.e. td > tN . If siti < 0, the effect on the tax rate ti is
given by
-li τti
I b0(gi)⅛ I ≠- td
<
>
(27)
Proof: The optimal state behavior is characterized by Eq. (21). Thus, starting from the
Nash outcome tN , as characterized by Eq. (9), state i’s incentive to deviate from tN depends on
the terms Vgisiti and Vτiτti . By Eq. (5) Vτiτti = -liτti which following Eq. (18) is positive. Note,
Vgisiti = b0(gi)siti . If b0(gi)siti is non-negative (case (iii) and (iv)), state i can increase after-tax
labor income as well as local public good provision by setting td > tN . To prove the second
assertion, consider the opposite case. If siti < 0 and -liτti > (<) Ib0(gi)siti I, state i can improve
utility of the representative household by choosing td > (<) tN . For the special case siti < 0 and
-liτti = I b0(gi)siti I the opposite effects nullify each other and td = tN. □
27