MODEL

Our two-date economy features two countries, a foreign government (FG), and a date-t domestic government (DG_t). FG can be interpreted either as an individual government or a group of governments such as the European Union, or an international organization such as the International Monetary Fund. There is a project that the governments can undertake at each of dates one and two; r _t = 1 indicates that the project is undertaken at date t, and r _t = 0 indicates that it is not. The project could represent the domestic country’s accession to an international organization such as the EU, the launch of a common currency, a climate agreement, or a region’s participation in a federation or national union.

At both dates, the project generates a value v_F for FG. The value of the project to DG_t depends on the identity of the political party that holds power. We consider a two-party setting that features a relatively friendly party with date-one valuation $\overline{v}$, and a relatively hostile party with date-one valuation $\underline{v}$. These project valuations can be interpreted as flow payoffs enjoyed at each date from the moment that the agreement is signed. If the project is not undertaken at date t, each agent receives a date-t payoff that we normalize to zero. All project valuations are common knowledge. Assumption 1 sets out the structure that the FG derives a higher value from the project than the relatively friendly government, which, in turn, derives a higher value from the project than the relatively hostile party.

All agents weight date-one payoffs by 1 − δ ∈ (0, 1) and date-two payoffs by δ. For example, 1 − δ could represent the time between the initial signing and the next election: when δ is large, negotiations take place relatively close to the election, after which there will be an opportunity to renegotiate the initial agreement.

At the outset of negotiations, participation by DG₁ in the project with FG implies a transfer $s_1\in \mathbb {R}$ from FG to DG₁. In the EU accession example, s ₁ ≥ 0 could represent a standard package of benefits, such as tariff reductions or a share of regional development funds that is awarded to a new member state upon joining. By contrast, s ₁ < 0 could reflect formula-based budgetary contributions made by the domestic government in exchange for its participation in the project. Alternatively, it could reflect monetary or fiscal convergence criteria that DG₁ must satisfy to accede, such as the Stability and Growth Pact. The precise value of s ₁—and whether it is positive or negative—does not play a role in our results. We focus on the most interesting setting, in which neither the relatively friendly nor relatively hostile party derives a positive date-one value from entering into an agreement on these terms, but the FG derives a strictly positive date-one value from the project taking place at the initial terms:

We allow for negotiations between the countries in which FG encourages DG_t to participate by offering more favorable terms. These negotiations unfold as follows. At date one, FG is the proposer, and DG₁ is the receiver.Footnote ¹¹ FG makes an initial offer b ₁ ≥ s ₁, which is a concession that it will give to DG₁ if and only if it participates in the agreement at that date.Footnote ¹² In the EU accession example, b ₁ could represent additional concessions and carveouts on labor market or financial sector regulations, budget contributions, or a more generous share of regional development funds. After receiving the offer b ₁, DG₁ chooses r ₁(b ₁) ∈ {0, 1}, where r ₁(b ₁) = 1 indicates that the project is implemented at date one and r ₁(b ₁) = 0 indicates that it is not.

Between dates one and two, the date-one domestic government DG₁ may be replaced by a new domestic government DG₂, according to a process that we describe below. After DG₂ is realized, all domestic agents are hit by a common additive preference shock λ to the payoffs they derive from the project. We assume that this publicly observed preference shock is drawn from a uniform distribution with support [ − σ, σ]. This shock can capture an unanticipated worsening of the economy—unemployment may increase, labor unions may organize industrial unrest, or there may be civil unrest. Alternatively, new information may come to light. For example, in 2004, an audit by the incoming Greek government found that, under a previous PASOK administration, the government’s statistics agency had misreported the country’s debt and deficit figures to qualify for entry into the European single currency.

We first assume that date-one negotiations do not affect domestic election outcomes. Thus, DG₂ is relatively hostile with exogenous probability $\Pr (\underline{v})\in [0,1]$, and relatively friendly with probability $\Pr (\overline{v})=1-\Pr (\underline{v})$. This captures a benchmark in which the election outcome is insensitive to the negotiation outcome. We later endogenize DG₂’s project valuation via an election, where electoral outcomes may depend on (1) whether the project was implemented at date one, and the terms of the initial bargain; (2) how voters make voting decisions (prospectively or retrospectively); and (3) the set of feasible replacements. We assume that there is sufficient variation in the domestic preference shock λ:

Assumption 3 says that there is enough uncertainty about the common domestic preference shock λ that (a) it could exceed the expected surplus from the project between FG and the relatively project-friendly DG₂ with valuation $\overline{v}$; and (b) it could be even lower than the expected value for the relatively hostile DG₂ with valuation $\underline{v}$ from participating in the project at the initial standing offer, s ₁.

After λ is realized, the initial terms for the project can be renegotiated, or if agreement was not reached at date one, the governments can try again. The inherited date-one terms serve as the reversion point s ₂ for date-two bargaining. Thus, if the project was implemented at date one with transfer b ₁, the status-quo transfer is s ₂ = b ₁; this transfer will be made at date two if the project is again implemented and new terms are not agreed upon. For example, Thatcher’s renegotiation of Britain’s EU budget rebate persisted from 1984 until 2005. If, instead, the project was not implemented at date one, then the status quo transfer (i.e., starting point for date-two negotiations in which the governments try again) is s ₂ = s ₁.

With probability θ ∈ [0, 1], DG₂ proposes the new terms, and with probability 1 − θ the FG makes the proposal. The parameter θ could reflect intrinsic bargaining power or institutional features of the agreement that determine who can initiate renegotiations. We allow for arbitrary θ ∈ [0, 1] to emphasize that results do not depend sensitively on the distribution of future bargaining power.Footnote ¹³ The agent realized as proposer at date two can propose a new transfer, ${b}_2\in \mathbb {R}$. If the date-two receiver accepts, this becomes the new date-two transfer. Otherwise, the inherited terms from past negotiations remain in force, so that b ₂ = s ₂. Next, DG₂ decides whether to quit the agreement and receive its outside option of zero or to execute the agreement given the date-two terms. FG then makes the agreed-upon transfer if and only if DG₂ executes the agreement by implementing the project.

The expected lifetime payoff of a domestic agent with date-one project valuation v is

$$\begin{eqnarray*} && (1-\delta )r_1(v+b_1)\nonumber\\ && \quad +\,\delta \sum _{v^{\prime }\in \lbrace \underline{v},\overline{v}\rbrace } \Pr (v^{\prime })\int _{-\sigma }^{\sigma }r_2(v+b_2+\lambda )f(\lambda )\,d\lambda, \end{eqnarray*}$$

where f(λ) is the density of the domestic preference shock, λ. Here r ₁ ∈ {0, 1} is the date-one domestic government’s initial decision to implement the project (r ₁ = 1) or not (r ₁ = 0), r ₂ ∈ {0, 1} denotes the project outcome at date two, and b ₂ denotes the date-two transfer from FG when the project is implemented at date two; that is, when r ₂ = 1. Note that domestic agents care about date-two policy outcomes regardless of who holds office at that date. In addition to deriving project-related payoffs like any other domestic agent, we assume that each domestic political party derives an office-holding benefit of w > 0 at any date that it holds office.

The analogous expected payoff of FG with project valuation v_F is

$$\begin{eqnarray*} &&(1-\delta )r_1(v_F-b_1)\nonumber\\ && \quad +\,\delta \sum _{v^{\prime }\in \lbrace \underline{v},\overline{v}\rbrace } \Pr (v^{\prime })\int _{-\sigma }^{\sigma }r_2(v_F-b_2)f(\lambda )\,d\lambda . \end{eqnarray*}$$

One may observe that FG’s project valuation does not evolve over time. This assumption eases presentation and analysis, allowing us to focus on the effects of uncertainty about DG₂’s valuation v ²_D. One can also interpret the FG as the IMF or the World Bank, whose leadership is not expected to change over the course of negotiations.

POLICY OUTCOMES AT DATE TWO

We start by analyzing the long-term consequences of date-one outcomes. If the project was implemented at date one, that is, if r ₁ = 1, then the status quo transfer s ₂ is the transfer b ₁ that DG₁ accepted. If the project was not implemented, that is, if r ₁ = 0, then the status quo transfer that serves as the starting point for date-two negotiations is s ₂ = s ₁.

Because there are no bargaining frictions, the project will be implemented at the terminal date t = 2 if and only if the associated surplus is positive; that is, if and only if

(1) $$\begin{equation} v_D^2+\lambda +v_F\ge 0 \iff \lambda \ge -(v_D^2+v_F). \end{equation}$$

Even though the date-two implementation decision does not depend on date-one actions, the division of the surplus depends on (a) the status quo transfer and (b) the shock realization λ.

Suppose, first, that DG₂ has a high enough project valuation v ²_D + λ that it would receive a positive payoff from implementing the project when it receives the status-quo transfer s ₂:

(2) $$\begin{equation} v_D^2+\lambda + s_2 \ge 0\iff \lambda \ge -(v_D^2+s_2). \end{equation}$$

With probability θ, DG₂ is recognized to propose a modification to the inherited terms, s ₂. Because DG₂ prefers higher transfers, it never proposes a transfer b ₂ < s ₂. Further, a proposal that raises the transfer to b ₂ > s ₂ will fail: if Equation (2) holds, FG recognizes that DG₂ will implement the project even if the initial agreement is not amended. As a result, FG would reject the amendment, because a threat by DG₂ to renege on the inherited agreement is not credible. With residual probability 1 − θ, FG gets to propose a modification. Although FG would like to negotiate a reduced transfer, DG₂ will refuse such amendments—it prefers to maintain the existing terms, which offer more favorable concessions in return for implementing the project.

Suppose, instead, that DG₂ anticipates a negative value from implementing the project at the status-quo transfer; that is, Equation (2) fails. This means that it would prefer not to implement the project at date two unless the initial terms were amended to a higher transfer. Suppose, first, that the surplus from agreement is positive; that is, Equation (1) holds.

With probability θ, DG₂ gets to propose a modification to the inherited terms. If FG rejects the proposal, the project will end when Equation (2) does not hold, giving FG a payoff of zero. Thus, DG₂ can renegotiate the date-two transfer from s ₂ to the larger transfer b ₂ = v_F. That FG is held to its participation constraint is not essential—what matters is that there is a discontinuity in the terms that DG₂ can obtain when its threat to break the existing agreement is credible; that is, at the threshold on λ defined in Equation (2). With probability 1 − θ, FG is, instead, recognized. Since Equation (2) fails, FG must offer DG₂ a larger transfer to secure its participation. It then raises the transfer from s ₂ to b ₂ = −(v ²_D + λ), leaving DG₂ with value v ²_D + λ indifferent between implementing the project and quitting, allowing FG to claim the rest of the surplus for itself.

Finally, if the date-two surplus from agreement is negative; that is, if Equation (1) does not hold, then no amendment will be agreed upon, as the joint surplus from implementing the project is negative. The project will not be implemented and all agents receive date-one payoffs of zero.

The expected date-two payoff of a domestic agent with date-one project valuation v who anticipates that the DG₂ will have project valuation v ²_D and face status quo transfer s ₂ is thus

(3) $$\begin{eqnarray} &&V_D(v, v_D^2,s_2) = \int _{-(v_D^2+s_2)}^{\sigma }(v+s_2+\lambda ) f(\lambda )\,d\lambda \nonumber\\ &&\quad +\,\int _{-(v_D^2+v_F)}^{-(v_D^2+s_2)}(v-v_D^2+\theta (v_D^2+\lambda +v_F))f(\lambda )\,d\lambda . \nonumber\\ \end{eqnarray}$$

The expected date-two project payoff of FG given s ₂ when it faces DG₂ with valuation v ²_D is

(4) $$\begin{eqnarray} && V_F(v_D^2,s_2) = \int _{-(v_D^2+s_2)}^{\sigma }(v_F-s_2)f(\lambda )\,d\lambda \nonumber\\ && \quad +\,\int _{-(v_D^2+v_F)}^{-(v_D^2+s_2)}(1-\theta )(v_D^2+\lambda +v_F)f(\lambda )\,d\lambda . \end{eqnarray}$$

A transfer of power from a friendly date-one domestic government DG₁ to a more hostile date-two domestic government DG₂ (i.e., from $\overline{v}$ to $\underline{v}$) carries two implications. First, it increases the prospect that DG₂ can renegotiate the initial terms to a more favorable arrangement. Second, it lowers the total surplus of the date-two negotiating parties. As a result, there will be situations in which a hostile DG₂ will fail to reach an agreement with FG in contexts where a more project-friendly DG₂ would have successfully concluded the negotiation.

Discussion: The bargaining protocol is starker than necessary for our main results. What is crucial is that the terms that the domestic government obtains at date two improve as its valuation of the project falls, relative to the status quo offer. This improvement in terms holds regardless of the distribution of date-two bargaining power, θ ∈ [0, 1]. When the domestic government holds date-two proposal power, a more hostile representative can renegotiate the status quo transfer from s ₂ up to b ₂ = v_F. When, instead, the FG holds proposal power, its offer holds the date-two domestic government to its participation constraint, but its transfer b ₂ = −(v ²_D + λ) still increases as the domestic government becomes more hostile; that is, as v ²_D decreases. A more hostile representative not only captures the upside of larger concessions—it also mitigates against the downside of subsequent appropriation.

POLICY OUTCOMES AT DATE ONE

Exogenous Power Transitions. In our benchmark setting, the date-two domestic government’s (DG₂’s) valuation does not hinge on the date-one policy outcome. At date one, the FG makes a proposal to the domestic government DG₁, which is either relatively friendly ($v_D^1=\overline{v}$) or relatively hostile ($v_D^1=\underline{v}$). DG₁ accepts the offer, that is, r ₁(b ₁) = 1, if and only if

(5) $$\begin{eqnarray} &&(1-\delta )(v_D^1+b_1) \nonumber\\ &&\quad +\,\delta \sum _{v_D^2\in \lbrace \underline{v},\overline{v}\rbrace }\Pr (v_D^2) \bigg [{\bf 1}[v_D^2=v_D^1]w+V_D(v_D^1,v_D^2,b_1)\bigg ]\nonumber \\ &&\ge \; \;(1-\delta )0 \nonumber\\ &&\quad +\,\delta \sum _{v_D^2\in \lbrace \underline{v},\overline{v}\rbrace }\Pr (v_D^2) \bigg [{\bf 1}[v_D^2=v_D^1]w+V_D(v_D^1,v_D^2,s_1)\bigg ],\nonumber \\ \end{eqnarray}$$

where we recall that w > 0 is the office rent that is enjoyed if and only if the incumbent is reelected; that is, v ²_D = v_D ¹. Thus, the FG’s date-one proposal solves

$$\begin{eqnarray*} &&\hspace*{-4.34pt} \max _{b_1\ge s_1}\,\,\,(1-\delta )r_1(b_1)(v_F-b_1)\\ &&\hspace*{-4.34pt} \quad +\,\delta\sum _{v_D^2\in \lbrace \underline{v},\overline{v}\rbrace } \Pr (v_D^2)V_F(v_D^2,r_1(b_1)b_1+(1-r_1(b_1))s_1), \end{eqnarray*}$$

subject to the participation constraint that r ₁(b ₁) = 1 if Equation (5) holds, and r ₁(b ₁) = 0, otherwise.

Strikingly, uncertainty about who will hold future domestic power has no effect on both (1) whether an agreement is signed and (2) how the surplus from an agreement is divided between the governments. In particular, the static and dynamic conditions for a date-one agreement coincide. Thus, Proposition 1 states that a date-one agreement is signed if and only if such an agreement is efficient from the perspective of the date-one negotiating parties.

To understand the result, let Δ(v ¹_D, v_D ²) be the ex ante expected date-two surplus from the perspective of the date-one bargaining parties, when the date-one domestic government DG₁ has project valuation v ¹_D and they anticipate that DG₂ has valuation v ²_D:

(6) $$\begin{eqnarray} \Delta (v_D^1,v_D^2)&=&{\bf 1}[v_D^2=v_D^1]w \nonumber\\ &&\,+\int _{-(v_D^2+v_F)}^{\sigma }(v_D^1+\lambda +v_F)f(\lambda )d\lambda . \end{eqnarray}$$

Crucially, this surplus does not depend on the date-two standing offer, s ₂. In particular, the total date-two surplus arising from an agreement is the same as the surplus in the event of disagreement. Thus, the total surplus from a date-one agreement versus no agreement is unrelated to its terms:

(7) $$\begin{eqnarray} &&(1-\delta )(v_D^1+v_F)+\delta \Pr (\underline{v})\Delta (v_D^1,\underline{v})\nonumber \\ &&\qquad+\,\delta \Pr (\overline{v})\Delta (v_D^1,\overline{v})-(1-\delta )(0+0)\nonumber \\ &&\qquad-\,\delta \Pr (\underline{v})\Delta (v_D^1,\underline{v})-\delta \Pr (\overline{v})\Delta (v_D^1,\overline{v})\nonumber \\ &&\hspace*{-10pt}=(1-\delta )(v_D^1+v_F). \end{eqnarray}$$

Because there is a constant surplus at each date, the surplus across dates is also constant, and its division represents a pure conflict of interest between the date-one negotiating parties. Starting from an offer that gives DG₁ its reservation payoff, suppose that FG can benefit from making larger initial offers that buttress its future negotiating position vis-à-vis an anticipated date-two domestic government. This could arise if both date-one governments expect a significantly more hostile DG₂ and the election is sufficiently imminent that FG’s immediate losses from a larger transfer today are outweighed by its expected future gains. Whenever a more generous offer raises FG’s total expected payoff, however, the constant total expected surplus implies that this gain must come at the expense of DG₁, which therefore prefers to reject the offer.

Thus, when agreement is reached, FG extracts all surplus from agreement. Equation (7) reveals that the total surplus is positive if and only if the total static surplus is positive: uncertainty about the future has no effect on whether an agreement is signed. Note, however, that the transfer from FG to DG₁ does not solve the static participation constraint that v ¹_D + b ₁ ⩾ 0, but rather the dynamic participation constraint given by Equation (5).Footnote ¹⁴

For simplicity, we assume that the FG makes the offer at date one. If, instead, the domestic government, DG₁, makes the initial offer, the conditions for agreement in Proposition 1 still apply, but now the domestic government extracts all surplus.

Exogenous power transitions create a constant total surplus between the FG and the date-one domestic government. So long as the static surplus from an agreement is positive, the FG can and will wish to induce the domestic government’s participation. But, there is no scope for both governments to benefit from more generous offers—so if and only if the date-one surplus is positive, (1) an agreement is signed and (2) the discounted total expected surplus is fully extracted by the FG.

To facilitate a clear and tractable benchmark, we assume that the sole dynamic linkage across periods is that date-one agreements determine the date-two standing offer, s ₂. In practice, date-one agreements affect the total surplus through other channels. For example, the possibility of participating in the project at date two might depend on whether an agreement was formed at date one. In the Supplemental Appendix, we consider this and related contexts, showing that while the conditions for an agreement to be signed may depart from our benchmark condition, all surplus is nonetheless extracted by the FG whenever an agreement is reached.Footnote ¹⁵

Endogenous Power Transitions. We now consider an electoral contest between dates one and two in which domestic voters, who differ in their project valuations $v\in \mathbb {R}$, observe the date-one negotiation outcome and then cast ballots in favor of their most-preferred date-two government.

Imminent elections have a polarizing effect on date-one negotiations between FG and DG₁. In the benchmark setting, initial negotiations are driven by the conflict of interest between the date-one negotiating partners over the surplus division. When elections respond to initial agreements, two other conflicts are critical: the policy and rent-seeking conflict between the domestic incumbent and its possible successors, and both domestic and foreign governments’ conflict with the domestic electorate. As a result, initial agreements no longer solely serve to divide the surplus: depending on whether DG₁ is relatively friendly or hostile, initial agreements may themselves change both the division and the size of the surplus from agreement.

Given status quo agreement s ₂, a voter with valuation v prefers a date-two domestic government that, from her perspective, induces the best date-two negotiation outcome, that is, that solves

$$\begin{equation*} \max _{v_D^2}V_D(v,v_D^2,s_2)\nonumber , \end{equation*}$$

where s ₂ = b ₁ if the project was implemented at date one with transfer b ₁, and s ₂ = s ₁ if it was not implemented at date one. The uniform distribution over the preference shock λ yields

Lemma 1. Given an inherited status quo agreement s ₂, a domestic voter with project valuation v prefers to elect the hostile government if and only if

(8) $$\begin{equation} v\le \frac{\underline{v}+\overline{v}}{2}+(v_F-s_2)\equiv \hat{v}(s_2). \end{equation}$$

A domestic voter’s induced preference for the friendly or hostile party depends on (1) her desire that a party reach an agreement with the FG in the same circumstances where she would value the project and (2) her desire to extract more generous transfers from the FG in exchange for implementing the project. The first aspiration depends on the voter’s valuation, but the second applies to all voters regardless of ideology, since all voters share a common value in extracting greater surplus from the FG.

A domestic voter is more intrinsically aligned with the friendly party whenever $v>\frac{\underline{v}+\overline{v}}{2}$: the friendly party is relatively more likely to reach agreements with the FG in circumstances where the voter would prefer an agreement to no agreement. Yet, Lemma 1 reveals that this domestic voter may nonetheless strictly prefer to vote for the hostile party!

To understand why, suppose that θ = 1, so that DG₂ always has proposal power in any renegotiation.Footnote ¹⁶ A voter anticipates a positive prospect that either a friendly or a hostile DG₂ will renegotiate the standing offer s ₂—indeed, with uniform preference shocks, each party is equally likely to renegotiate the standing offer in equilibrium. However, voters also anticipate that each party renegotiates with FG in different circumstances, with the hostile party renegotiating when voters attach a higher valuation to the project than when the friendly party is renegotiating.

First, if $\lambda \in [-(\overline{v}+s_2),-(\underline{v}+s_2)]$, a friendly DG₂ maintains the standing offer s ₂, but a hostile DG₂ renegotiates the offer from s ₂ to v _F. Nonetheless, both domestic governments ultimately pursue the project. In this context, a voter with valuation v + λ derives a positive payoff difference from the hostile DG₂ versus the friendly DG₂ of

(9) $$\begin{equation} v+v_F+\lambda -(v+s_2+\lambda )=v_F-s_2>0. \end{equation}$$

The voter’s date-two project valuation (v + λ) does not affect her relative value because both governments would pursue the project; regardless of her valuation, she perceives that the hostile party is superior, because it pursues the project and extracts a higher surplus in the same context that the friendly party would pursue the project at the standing offer.

Second, if $\lambda \in [-(\overline{v}+v_F),-(\underline{v}+v_F)]$, a friendly DG₂ renegotiates the offer from s ₂ to v_F, but a hostile DG₂ instead walks away from the project. A voter with project valuation v + λ derives a payoff difference from the friendly DG₂ versus the hostile DG₂ of

(10) $$\begin{equation} v_F+(v+\lambda )-0. \end{equation}$$

Like the hostile DG₂ in the first case, a friendly DG₂ in the second case appropriates FG’s value from the project, v_F. However, unlike the first case, this transfer is partially offset by the fact that a friendly government pursues the project in circumstances where a hostile government—and, possibly, the voter—wants to abandon it. When v + λ < 0, the friendly government’s continuation of the project generates a loss that is only partially offset by the transfer, v_F. So, for example, voter $v=\frac{\underline{v}+\overline{v}}{2}$’s expectation of the payoff difference Equation (10) at the time of the domestic election is zero.Footnote ¹⁷ Since the payoff gain Equation (9) is strictly positive, a voter whose project valuation is equidistant from the two parties strictly prefers to support the hostile party.

Note that a voter’s induced preferences over date-two governments differ from those of an agent who shares her project valuation, v, but chooses a date-two domestic government to maximize total expected date-two surplus between that voter and FG. Such an agent would prefer the hostile government if and only if $v\le \frac{\overline{v}+\underline{v}}{2}$.Footnote ¹⁸ The reason for this divergence is that a voter does not value total surplus, but rather her share of the surplus. This highlights a possible source of inefficiency in domestic election outcomes that are sensitive to a country’s external negotiations.

Voters’ induced preferences over date-two representatives are manipulable by both date-one governments. FG can manipulate a voter’s tradeoffs via its initial offer, b ₁ ≥ s ₁: more generous offers—if accepted—steer voters toward the more project-friendly party. But DG₁ can also manipulate voters’ tradeoffs via its choice to accept or reject the offer, r ₁(b ₁) ∈ {0, 1}: rejecting an offer bequeaths a worse status quo, steering voters toward the more hostile party. How these concerns affect the prospect of initial agreement, and the division of the surplus, will depend on the policy conflict between parties, between the parties and their electorate, and between all domestic agents and the FG. We now show how these conflicts resolve.

Henceforth, we assume that the distribution of voters’ project valuations has a unique median, v ^med. The single-peaked structure of induced preferences then implies that the voter with this median valuation is decisive in an election: for any standing offer s ₂, the hostile party wins if and only if $v^{\text{med}}\le \frac{\underline{v}+\overline{v}}{2}+(v_F-s_2)\equiv \hat{v}(s_2)$. We assume that, at date one, both the FG and domestic parties are uncertain of the future median voter’s project valuation:

Uniform uncertainty is not essential for our results, but it facilitates tractable comparative statics (e.g., on v ^e and α). There are many natural interpretations. For example, the larger is α, the more uncertain are date-one negotiating parties about the preferences of the domestic electorate. Conditions (1) and (2) imply that there is enough uncertainty about voter preferences that each party wins with positive probability given any standing offer, s ₂ ∈ [s ₁, v _F]. The probability that the majority-preferred date-two government is the relatively hostile domestic party is therefore

(11) $$\begin{eqnarray} \Pr ( v^{\text{med}}&\le& \hat{v}(s_2))=\frac{\hat{v}(s_2)-(v^e-\alpha )}{2\alpha } \nonumber\\ &=&\frac{\frac{\underline{v}+\overline{v}}{2}+v_F-s_2-(v^e-\alpha )}{2\alpha }. \end{eqnarray}$$

The electoral consequences of a more favorable date-two status quo s ₂ differ starkly for the friendly and hostile domestic parties. As s ₂ increases—for example, when the date-one domestic government extracts a larger transfer in exchange for pursuing the project—the hostile party’s electoral prospects fall, and the friendly party’s electoral prospects rise. This is a key difference with Wolford (Reference Wolford2012), who assumes that a domestic incumbent’s re-election prospects rise with the surplus it extracts from an FG in a pre-election negotiation. Our framework highlights that when initial negotiation outcomes become the reversion point in future elections, voters’ induced preferences may generate the opposite relationship between an incumbent’s negotiated share of the surplus from agreement and its reelection prospects.

We earlier showed that when power transitions are exogenous, total expected surplus is unaffected by the initial agreement. This is no longer true when date-one outcomes can alter electoral outcomes. To see why, recognize that from the perspective of the date-one bargaining parties, the expected date-two surplus derived from a status quo s ₂ is

(12) $$\begin{eqnarray} && \Pr ( v^{\text{med}}\le \hat{v}(s_2))\Delta (v_D^1,\underline{v}) \nonumber\\ && \quad +\,\Pr ( v^{\text{med}}>\hat{v}(s_2))\Delta (v_D^1,\overline{v}), \end{eqnarray}$$

where Δ(v ¹_D, v_D ²) (defined in Equation (6)) is the ex ante expected date-two surplus from the perspective of the date-one bargaining parties when DG₁ has project valuation v ¹_D and DG₂ has valuation v ²_D. Thus the relative total surplus from an agreement (versus no agreement) is

(13) $$\begin{eqnarray} &&(1-\delta )(v_F+v_D^1)+\delta (\Pr ( v^{\text{med}}\le \hat{v}(b_1))\nonumber\\ &&\quad -\Pr ( v^{\text{med}}\le \hat{v}(s_1)))(\Delta (v_D^1,\underline{v})-\Delta (v_D^1,\overline{v})). \nonumber\\ \end{eqnarray}$$

Lemma 2 highlights how the relative total surplus from an agreement changes with the terms of the agreement, depending on whether DG₁ is relatively friendly or relatively hostile.

To understand why, notice that the change in relative surplus between the date-one negotiators from increasing the transfer from b ₁ to a higher offer b′₁ is

$$\begin{eqnarray*} &&\delta (\Pr ( v^{\text{med}}\le \hat{v}(b_1^{\prime }))-\Pr ( v^{\text{med}}\le \hat{v}(b_1)))\\ &&\quad\times\,(\Delta (v_D^1,\underline{v})-\Delta (v_D^1,\overline{v})) . \end{eqnarray*}$$

If DG₁ is friendly, that is, if $v_D^1=\overline{v}$, then the second bracketed term is strictly negative; if instead DG₂ is hostile, that is, if $v_D^1=\underline{v}$, then the second term is strictly positive. However, higher transfers also encourage domestic voters to support the friendly party in the polls, so that

$$\begin{eqnarray*} b_1^{\prime }>b_1 &\Rightarrow& \hat{v}(b_1^{\prime })< \hat{v}(b_1) \\ & \Rightarrow & \Pr ( v^{\text{med}}\le \hat{v}(b_1^{\prime })) < \Pr ( v^{\text{med}}\le \hat{v}(b_1)). \end{eqnarray*}$$

The relatively hostile party values retaining office—both for office rents and because of policy differences with the relatively friendly party—so more generous transfers indirectly reduce its value from a date-one agreement, while the opposite holds for the friendly party. The change in surplus reflects both office rents and policy differences, as each party values its role in negotiating agreements. As office rents w rise, the consequences for increasing or decreasing the surplus are compounded, but they do not vanish as w becomes very small, since policy conflicts remain.

Lemma 2 highlights the polarizing effect of domestic elections on conflicts and confluences of interest between the date-one negotiating parties. In the benchmark setting with exogenous elections, different offers change the division of the surplus, but not its size. In contrast, when elections are sensitive to negotiation outcomes, offers affect both the surplus size and its division.

We now characterize date-one negotiation outcomes. DG₁ accepts an offer; that is, r ₁(b ₁) = 1, if and only if

(14) $$\begin{eqnarray} &&(1-\delta )(v_D^1+b_1)\nonumber \\ &&\quad +\,\delta \Pr (v^{\text{med}}\le \hat{v}(b_1)) ({\bf 1}[v_D^1=\underline{v}]w+V_D(v_D^1,\underline{v},b_1))\nonumber \\ &&\quad+\,\delta \Pr (v^{\text{med}}>\hat{v}(b_1))({\bf 1}[v_D^1=\overline{v}]w+V_D(v_D^1,\overline{v},b_1))\nonumber \\ &&\quad\ge\,\delta \Pr (v^{\text{med}}\le \hat{v}(s_1))({\bf 1}[v_D^1=\underline{v}]w+V_D(v_D^1,\underline{v},s_1))\nonumber \\ &&\quad+\,\delta \Pr (v^{\text{med}}>\hat{v}(s_1))({\bf 1}[v_D^1=\overline{v}]w+V_D(v_D^1,\overline{v},s_1)). \nonumber \\ \end{eqnarray}$$

Thus, FG’s date-one proposal solves

(15) $$\begin{eqnarray} &&\hspace*{-5.78pt} \max _{b_1\ge s_1}\,\,\,(1-\delta )r_1(b_1)(v_F-b_1)\nonumber \\ &&\hspace*{-5.78pt}\quad +\delta \Pr (v^{\text{med}}\le \hat{v}(s_2(r_1(b_1),b_1)))V_F(\underline{v},s_2(r_1(b_1),b_1))\nonumber \\ &&\hspace*{-5.78pt}\quad +\delta \Pr (v^{\text{med}}>\hat{v}(s_2(r_1(b_1),b_1))) V_F(\overline{v},s_2(r_1(b_1),b_1)),\nonumber \\ \end{eqnarray}$$

subject to the participation constraint that r ₁(b ₁) = 1 if Equation (14) holds, and r ₁(b ₁) = 0, otherwise, and the date-two status quo offer is s ₂(r ₁(b ₁), b ₁) = r ₁(b ₁)b ₁ + (1 − r ₁(b ₁))s ₁. We first characterize date-one negotiation outcomes between the FG and the hostile party.

With a responsive electorate, more than a positive static surplus is necessary for the governments to reach an initial agreement. With a hostile domestic government, more generous offers reduce the governments’ anticipated future surplus. The reason is that more generous offers lower the prospect that the hostile party retains power, denying it both the chance to capture office rents w and the ability to steer future negotiations. In a dynamic setting where elections do not respond to negotiations, an agreement would be signed whenever date-one surplus is positive. Now, however, sufficiently imminent elections preclude a date-one agreement, for any positive date-one surplus, if office-holding motives are strong. Finally, because one government’s gain must constitute a loss to the other, the FG appropriates all surplus whenever an agreement is reached, as in the benchmark setting. Proposition 2 highlights that more proximate elections can make impossible an agreement between FG and the hostile DG₁ that otherwise could have been secured; that is, even when the static surplus from agreement is positive.

Matters are very different when DG₁ is the friendly party with project valuation $\overline{v}$:

In a dynamic setting where elections do not respond to negotiations, a positive static surplus is necessary and sufficient for the governments reach an agreement. But with endogenous turnover, a positive static surplus is no longer necessary: more generous offers bolster the re-election prospects of the friendly DG₁, raising its chances of gaining office rents w as well as the ability to steer subsequent negotiations in its favor. When elections respond to negotiating outcomes, the surplus from agreement itself changes with more generous offers: the date-one negotiating parties’ joint desire to exclude the hostile party by shaping voters’ induced preferences creates a confluence of interest that may facilitate agreement despite a negative static surplus. As elections draw nearer, the static conflict between FG and friendly DG₁ pales in significance to the joint interest of both governments in using date-one outcomes to steer voters’ induced preferences in favor of re-electing the incumbent. If office-holding motives are strong enough, imminent elections facilitate a date-one agreement for any date-one surplus—positive or negative. Proposition 3 highlights that more proximate elections make possible an agreement between FG and the friendly DG₁ that otherwise could not have been secured.

Proposition 1 establishes that in a setting with exogenous turnover, a date-one agreement is signed whenever it is efficient to undertake the project, that is, when the (dynamic) surplus from an agreement is positive, from the perspective of the date-one negotiating parties. In the Supplemental Appendix, we highlight that in a setting with endogenous turnover (that is, Propositions 2 and 3), date-one negotiations may fail even when there exist transfers from FG to DG₁ for which the joint surplus from an agreement is strictly positive.Footnote ¹⁹

With exogenous turnover, or when DG₁ is relatively hostile, FG appropriates all surplus from agreement. By contrast, Proposition 3 shows that when DG₁ is relatively friendly, an imminent election may induce FG to offer DG₁ strictly positive surplus. The reason is that when negotiations are conducted close to the election, FG’s interest in promoting the friendly DG₁’s reelection may lead it to make more generous offers that sway voters toward the incumbent.

This raises a basic question: Conditional on securing a date-one agreement, which date-one party—the hostile party or the friendly one—extracts greater transfers from the FG? On the one hand, a friendly DG₁ enjoys a strictly positive surplus from the agreement, while a hostile DG₁ is held to its participation constraint. On the other hand, the friendly DG₁’s participation can be more easily secured than the hostile DG₁’s participation. Our next result provides an unambiguous resolution to this question:

The result reflects that the friendly party derives a higher surplus from agreements than the hostile party simply because its participation can be bought more cheaply by the FG. The hostile DG₁’s participation constraint is more stringent than the analogous constraint for a friendly DG₁, so whenever FG derives no surplus from an agreement, the result is immediate. Suppose, instead, that the friendly DG₁’s participation constraint is slack when FG advances its most preferred offer. This offer, b*₁, solves the first-order condition associated with Equation (15). Recall that FG and the hostile DG₁ face a pure conflict of interest: any gain for one must come at the expense of the other. If b*₁ is most preferred by FG, its value is strictly increasing in an offer b ₁ ∈ [s ₁, b*₁]—and so, the hostile party’s value relative to rejection is strictly decreasing on this interval. It follows that to induce the hostile DG₁’s participation, FG must overextend itself relative to its most preferred offer; that is, its offer must exceed b*₁.

Thus, even at date one, voters face a tradeoff with a more hostile domestic government. If DG₁ is too hostile, negotiations will break down. If, instead, it is very friendly to the project, it may agree to relatively ungenerous terms. In the event that the friendly DG₁ reaches agreement with FG, its conflict with voters rises with its value from holding office w, because it becomes willing to accept ever-worse offers to improve its electoral prospects relative to securing no transfers. With the hostile DG₁, the consequences of a greater office concerns are less clear-cut. On the one hand, conditional on securing agreement, greater office motivation makes the hostile party demand more transfers to compensate for its diminished electoral prospects resulting from an agreement. This gives the hostile party commitment power to reject offers that the friendly party would accept. On the other hand, a near-exclusive concern for retaining office may preclude agreement between a hostile DG₁ and FG.

In the Supplemental Appendix, we compare equilibrium transfers when an agreement is reached in our benchmark setting of exogenous turnover (i.e., Proposition 1) with the corresponding transfers when an agreement is reached in the context of endogenous turnover (i.e., Propositions 2 and 3). We show that whenever FG’s transfer to friendly DG₁ is larger with endogenous turnover than in our benchmark setting, FG’s transfer to the hostile DG₁ is also larger whenever an agreement is reached. Moreover, if the relative value of holding date-two domestic power (w) is sufficiently large, then—relative to the benchmark with exogenous turnover—FG’s date-one transfer that ensures DG₁’s participation in the setting with endogenous turnover decreases, while the transfer that is needed to satisfy hostile DG₁’s participation increases.Footnote ²⁰

Consequences of Changes in Domestic Politics. We now ask how changes in the preferences of the two domestic parties affect FG’s preferred initial interior offer; that is, the interior offer that solves FG’s objective Equation (15). FG’s responses turn on the answers to two questions: How does the change affect FG’s relative value from steering the subsequent election toward the friendly party? And how does the change affect FG’s ability to influence the electoral outcome? Recall that, under Assumption 4, we assume that v ^med is uniformly distributed on [v ^e − α, v ^e + α].

Suppose that one of the domestic political parties grows more inclined toward the project; that is, either $\underline{v}$ or $\overline{v}$ rises. If that party later wins office, FG calculates that the party’s threat to walk away from the agreement is now less credible, since it values the agreement by more. This encourages FG to respond with lower transfers.

However, the party preference shift also alters the electoral competitiveness of the two parties. Recall that a voter who is indifferent between the parties has project valuation

$$\begin{equation*} \hat{v}(s_2)=\frac{\underline{v}+\overline{v}}{2}+(v_F-s_2)\nonumber . \end{equation*}$$

Absent any change in the negotiation settlement, a higher $\overline{v}$lowers the electoral competitiveness of the friendly party by shifting $\hat{v}$ to the right, raising the prospect that the median domestic voter will favor the hostile party. Conversely, a higher $\underline{v}$raises the electoral competitiveness of the hostile party. Absent any change in FG’s offer, these shifts place the friendly party at an increased disadvantage. This encourages FG to respond with larger transfers.

Finally, FG’s value from promoting the friendly party’s reelection depends on the wedge $\overline{v}-\underline{v}$ between the two parties’ bargaining attitudes. When the friendly party grows even more favorably disposed to the project, the wedge grows, raising FG’s stake from steering the domestic election in the friendly party’s favor, encouraging FG to raise its transfer. By contrast, when the hostile party moderates, the wedge shrinks. This reduces FG’s stakes, encouraging FG to lower its transfer.

These calculations relate to the value placed by FG on using higher offers to buttress its future negotiating position. But whether higher offers can have a meaningful impact on the election depends on the sensitivity of voters’ choices to offers. With uniformly distributed uncertainty, the density of v ^med evaluated at the threshold $\hat{v}(s_2)$ is $\frac{1}{2\alpha }$: electoral outcomes are more sensitive to offers when α is lower.

When α is large, election outcomes are insensitive to offers, so FG’s return from using higher transfers to steer the election toward the friendly party is low. In this case, FG’s response to an improvement in the attitude of either party reflects that, conditional on holding office, that party is less likely to successfully renegotiate the terms. This encourages FG to reduce its offer. When α is small enough, election outcomes become sensitive to offers, and the FG can more effectively steer the domestic electorate in favor of the friendly party by way of a more generous offer. We now show how, depending on FG’s value from promoting the election of the friendly party, this may lead to either more or less generous date-one agreements.

Figure 1 illustrates these findings. If α is small (panel (a)), election outcomes are very sensitive to negotiation outcomes, so FG responds to increases in $\overline{v}$ with increased offers, to promote the reelection of the friendly DG₁. But, if α is large (panel (c)), election outcomes are relatively insensitive to higher offers, so FG responds with lower offers, since improvements in the friendly party’s bargaining attitude make it a more pliant negotiating partner when it is retained.

FIGURE 1.

Illustration of how FG’s most preferred proposal varies with the friendly party’s project valuation. Parameters: δ = 1, v _F = 3, θ = .5, σ = 3, $\underline{v}=-3$, v ^e = −1, s ₁ = 0, and $\overline{v}\in [-\frac{3}{4},0]$. Panel (a) corresponds to high electoral return from more generous offers, (b) to intermediate electoral return, and (c) to low electoral return. The thresholds described in Proposition 5 are $\overline{\alpha }_*\sim 4.62$, $\overline{\alpha }^*\sim 5.54$, and $\overline{v}^*(\alpha )=.5 (-3 \alpha +\sqrt{3} \sqrt{\alpha (7 \alpha -8)}-6)$.

Finally, when α is intermediate (panel (b)), the election is only moderately sensitive to international negotiations. When $\overline{v}$ and $\underline{v}$ are very close, the two parties are almost indistinguishable from FG’s perspective. As a result, increases in $\overline{v}$ only modestly increase FG’s value of promoting the reelection of the friendly party. In conjunction with the reduced electoral returns from raising its offer (since $\alpha >\overline{\alpha }_{*}$), FG prefers to respond to a higher $\overline{v}$ with smaller transfers.

As the friendly party grows even more favorably disposed to the project, that is, $\overline{v}$ rises, FG’s tradeoffs change. The increasing wedge $\overline{v}-\underline{v}$ in valuations between the domestic political parties raises FG’s stake in promoting the electoral success of the friendly party. In conjunction with the non-trivial electoral returns from raising its offer (since $\alpha <\overline{\alpha }^{*}$), FG responds to a higher $\overline{v}$ with larger transfers.

Distinct considerations drive FG’s response when the hostile party’s valuation $\underline{v}$ rises:

Proposition 5. (Hostile Party’s Valuation Increases.) Suppose the hostile party is initially electorally competitive, in the sense that

(16) $$\begin{equation} v^e-\underline{v}<v_F-s_1, \end{equation}$$

Then, if the hostile party’s project valuation $\underline{v}$ increases, FG’s most preferred offer decreases. Otherwise, there exist at most two thresholds $\underline{\alpha }_{*}$ and $\underline{\alpha }^*$ such that if $\alpha < \underline{\alpha }_{*}$, FG’s preferred offer increases and if $\alpha >\underline{\alpha }^*$, FG’s preferred offer decreases. For $\alpha \in [\underline{\alpha }_{*},\underline{\alpha }^*]$, there exists $\underline{v}^*(\alpha )$ such that FG’s preferred offer is increasing in $\underline{v}$ if and only if $\underline{v}\le \underline{v}^*(\alpha )$.

Increasing the hostile party’s project valuation $\underline{v}$ has three effects. First, conditional on winning office, the hostile party is a more pliant negotiator. Second, FG’s stakes in the election fall, since the expected difference in the bargaining stances of the two parties falls. Third, the hostile party wins more votes, since its platform moves closer to the friendly party’s platform; that is, $\frac{\underline{v}+\overline{v}}{2}$ moves to the right. The first two effects encourage FG to reduce its offer, while the third encourages it to raise its offer to offset the hostile party’s increased electoral advantage.

The difference $v^e-\underline{v}$ represents the intrinsic expected misalignment between the hostile party and the electorate. When this misalignment is large, voters worry about the risk that a hostile DG₂ will fail to reach agreement, causing the project to be abandoned. However, when condition Equation (16) holds, this risk is outweighed by the additional surplus that could be extracted from renegotiating a standing offer s ₁ up to v _F, and which a hostile government is more likely to secure. We say that the hostile party is “competitive” when condition Equation (16) holds.

Condition Equation (16) holds in Figure 1. When the hostile party is competitive, its behavior conditional on winning office dominates FG’s calculation. As $\underline{v}$ rises, FG understands that, if elected, the hostile party will be less credible in its threats to unilaterally quit at the inherited terms. Thus, it responds with lower transfers.

When the hostile party is initially relatively uncompetitive, changes in fundamentals that improve its electoral prospects weigh more heavily on FG. If α is small, the election outcome is sensitive to date-one transfers, so FG responds to a higher $\underline{v}$ with more generous offers to offset the hostile party’s increased electoral advantage. If, instead, α is large, FG lowers its transfer, because it understands that efforts to influence the election via its offer would be ineffectual.

Finally, when α is intermediate, the election outcome is only moderately sensitive to offers. Again, when $\underline{v}$ rises, the hostile party becomes more electorally competitive. But if $\underline{v}$ is very close to $\overline{v}$, FG regards the two parties as almost indistinguishable. This lowers FG’s stake from using higher transfers to compensate the friendly party’s reduced competitiveness. In conjunction with the reduced returns from raising its offer (since $\alpha >\underline{\alpha }_*$), FG prefers to respond to a higher $\underline{v}$ with smaller transfers. If, instead, the hostile party’s valuation $\underline{v}$ is initially far less than $\overline{v}$, FG anticipates a large wedge between the bargaining postures of the two parties, raising its stake in partially offsetting the friendly party’s increased disadvantage due to an increase in $\underline{v}$. In conjunction with the nontrivial electoral returns from raising its offer (since $\alpha <\underline{\alpha }^*$), FG prefers to respond to a higher $\underline{v}$ with larger transfers.

Proposition 5 contrasts strikingly with the main predictions of the two-level games literature, in which agreements signed between DG₁ and FG are subject to ratification by a distinct domestic agent. Suppose, for example, that the domestic country has a ratification requirement, that the relatively hostile party holds ratification authority, but initial negotiations with the FG are conducted by the relatively friendly party. This could arise with a divided government in which the friendly party holds executive authority, but the hostile party controls the legislative upper chamber. In that setting, a more hostile ratifier allows the friendly DG₁ to extract more concessions from FG, since FG recognizes that the hurdle for successful ratification increases as the ratifier becomes more intrinsically opposed to the project.

In the context of elections with subsequent renegotiation, however, Proposition 5 highlights that the opposite outcome may arise. The reason is that FG anticipates a lower prospect that the hostile party replaces the friendly party for any standing offer. This (1) reduces FG’s perceived urgency of using transfers to shift voters’ induced preferences in favor of the friendly government via larger initial concessions and (2) reduces FG’s need to buttress its future negotiating position against the date-two domestic government, which is more likely to be relatively friendly for any given standing offer. Thus, a more hawkish domestic opposition party may reduce the concessions extracted by friendly DG₁, reversing the empirical prediction of the Schelling conjecture in settings where electoral considerations are first order (Milner Reference Milner1997).

Conversely, we could evaluate the consequences of greater uncertainty (α) on the part of the date-one negotiating parties about the preferences of the domestic electorate. First, higher α lowers FG’s anticipated returns from offering more transfers to secure friendly DG₁’s reelection. This encourages FG to lower its offer. Second, more uncertainty raises the possibility both of an even more project-averse median voter (i.e., lowers z ^e − α) and an even more project-friendly median voter (i.e., raises z ^e + α). There exist primitives for which the former consideration weighs more heavily in FG’s calculation: in this case, FG may prefer to offer even more generous transfers, not to facilitate the reelection of the friendly DG₁, but rather to buttress its future negotiating position against a more likely hostile successor.Footnote ²¹

EXTENSIONS AND ROBUSTNESS

In the Appendix, we pursue several extensions, a subset of which we briefly outline.

More Domestic Alternatives. Our benchmark analysis considers a two-party system. In the Appendix, we allow the median voter to select a date-two domestic government that has any valuation $v_D^2\in [\underline{v},\overline{v}]$. This may reflect a setting with purely office-motivated parties that can commit to any policy.Footnote ²² Given a standing offer, s ₂, a voter with valuation v most prefers a DG₂ with valuation v − (v _F − s ₂). This is also true in our benchmark setting; but there voters must choose from one of two possible alternatives. In the extension, by contrast, the median voter can always achieve her most preferred date-two alternative. This, however, can feedback to adversely affect date-one agreements and hence expected voter payoffs.

Our result for exogenous election outcomes (Proposition 1) that static and dynamic conditions for a date-one agreement coincide extends directly to this setting. In the two-party setting with endogenous elections, however, Lemma 2 shows that, depending on the valuation of DG₁, the total surplus from agreement is either strictly increasing, or strictly decreasing, in the standing offer s ₂. With more than two alternatives, the relationship between standing offers and surplus is more subtle. When the median voter’s valuation is distributed uniformly on [v ^e − α, v ^e + α] and $v^e-\alpha -(v_F-s_1)>\underline{v}$ and $v^e+\alpha <\overline{v}$, the project valuation of the median’s preferred date-two representative is contained in $(\underline{v},\overline{v})$. We have:

To understand the result, recall that DG₂’s valuation in the event of an agreement is $\hat{v}_D^2(b_1)=v^{\text{med}}-(v_F-b_1)$, so that the total expected date-two surplus between the date-one governments is

(17) $$\begin{eqnarray} &&\int _{v^e-\alpha }^{v^e+\alpha }\,\,\int _{-(v_F+\hat{v}_D^2(b_1))}^{\sigma }\frac{1}{2\alpha }\left[{\bf 1}[\hat{v}_D^2=v_D^1]w\right.\nonumber\\ &&\quad\left.+\,(v_D^1+v_F+\lambda )f(\lambda )\,d\lambda \, \right]d v^{\text{med}}. \end{eqnarray}$$

This surplus is maximized—both via policy rents and policy payoffs—when $\mathbb {E}[\hat{v}_D^2(b_1)]=v_D^1$, that is, when b ₁ = v ¹_D + v_F − v^e. Thus, any DG₁ with valuation in $(\underline{v},\overline{v})$ has both a partial conflict of interest with FG, and a partial confluence of interest. To the extent that more generous offers push the expected valuation of DG₂ up and toward the initial valuation of DG₁, the governments are aligned. But as more generous offers push the expected valuation of DG₂ above the initial valuation of DG₁, the conflict between governments intensifies.

In our base two-party setting, the absolute degree of alignment between DG₁ and FG, that is, the values of v ¹_D and v _F, does not affect the sign of the impact of more generous offers on the surplus: what matters is the relative alignment, that is, which party is most closely aligned. In contrast, with many possible succeeding parties, the expression for b* in Lemma 3 reveals that absolute degrees of conflict between the governments and voters determines the extent to which DG₁ and FG are sufficiently aligned in their dynamic interests to achieve date-one agreements.

Limited Policy Commitments.Footnote ²³ Our core analysis presumes that parties cannot commit to platforms prior to entering office: voters anticipate that parties will choose the bargaining stance that maximizes their expected payoffs once they enter office. In the Appendix, we consider a related setting in which, between dates one and two but before learning the shock to voter preferences, the friendly and hostile parties may each commit to a bargaining posture—that is, a party may commit to negotiating at date two as if it had some intrinsic value v. Consistent with Calvert (Reference Calvert1985), platform differentiation arises in equilibrium: the hostile party commits to a more hostile bargaining stance than the friendly party. There is, however, a degree of moderation by the parties, because they trade off the prospect of winning—which calls on them to adopt a posture that is closest to the expected median’s most preferred posture—with their intrinsic policy motivation. Thus, the hostile party’s promised platform lies to the right of its most preferred bargaining posture (prior to learning λ) and the friendly party’s platform lies to the left of its most preferred bargaining posture.

Retrospective Voting. Our base analysis presumes that voters are forward-looking, that is, voting for the party that will secure the best anticipated negotiation outcomes. In the Supplemental Appendix, we consider retrospective voters who reward or punish the incumbent according to a linearly increasing function of their date-one payoffs. Specifically, we assume

$$\begin{eqnarray*} &&\Pr (\text{reelect incumbent}|\text{ date-one outcome})\nonumber\\ &&\quad=\max \lbrace 0,\min \lbrace a+\beta r_1(v^{\text{med}}+b_1),1\rbrace \rbrace \nonumber . \end{eqnarray*}$$

Here a reflects electoral aspects that do not depend on international negotiations, and β captures the salience of the negotiations in the election—when β is large, the date-one domestic government’s electorate fortunes are more sensitive to negotiation outcomes.

In our two-party benchmark setting, we show that if (1) international negotiations are sufficiently salient and (2) domestic parties are sufficiently polarized, in the sense that

$$\begin{equation*} \beta (\overline{v}-\underline{v}) >\frac{1+\theta }{2}\nonumber , \end{equation*}$$

then an analogue of Proposition 2 holds: if DG₁ is hostile, that is, with value $\underline{v}$, then either no agreement is signed or FG holds hostile DG₁ to its participation constraint. When $\overline{v}-\underline{v}$ is large, FG’s value from steering voters toward the friendly party is large, and when β is large, the election outcome is especially sensitive to the date-one outcome. These are the circumstances in which the conflict of interest between FG and hostile DG₁ is greatest.

With prospective voters, a hostile DG₁ refuses more generous offers because they harm its electoral prospects. With retrospective voters, the FG refuses to make more generous offers to the hostile party because they advance its electoral prospects. Thus, the conflict of interest between the date-one negotiating parties is fundamental and does not hinge on the sophistication or foresight of the electorate. In contrast with prospective voters, however, with retrospective voters, a friendly DG₁ may secure a larger date-one transfer than a hostile DG₁.