The Baseline of Foreign Policy Signaling

We treat extant research on trust-building through foreign policy signals as our baseline. In Supplementary Appendix A.1, we introduce a two-period model that first privately draws each state’s intentions (using probabilities $ {p}_A,{p}_B $ ) and then iterates the simultaneous-move trust-building model visualized in Table 1 over two periods. In later sections, we analyze the effects of domestic (or other independent) signals by manipulating features of the first-period choice based on our substantive arguments about why domestic policies are different.

We advance the simultaneous-move model for two reasons. First, it harnesses the power of incrementalism (Ashworth, Berry, and Bueno de Mesquita Reference Ashworth, Berry and de Mesquita2021, 58–9) by adding one parameter (independence, introduced next) to the framework common across much trust-building research (Acharya and Ramsay Reference Acharya and Ramsay2013; Jervis Reference Jervis1978; Kydd Reference Kydd2000; Reference Kydd2005; Schultz Reference Schultz2005; Yoder and Haynes Reference Yoder and Haynes2025). Second, it captures the strategic problem we hope to study. In the cases that interest us, such as enduring rivalries, arms control, and major power–client relations, both sides have the ability to exploit the other by adjusting their policy before the other realizes and can take countermeasures (e.g., Braumoeller Reference Braumoeller2008; Glaser Reference Glaser2010; Waltz Reference Waltz1979). Thus, at every moment, each worries the other is cheating them, creating pressure to cheat first. If we instead sequenced choices, it would artificially resolve the trust problem for whoever moves second.Footnote ²

In Supplementary Appendix A.2, we analyze the baseline model for perfect Bayesian equilibria (PBE). Consistent with Kydd (Reference Kydd2005) and others, we find many PBE—including mixed and pure strategy equilibria—in overlapping parameter ranges. We report them all in Supplementary Appendices. In the article, we focus on equilibria that survive an efficiency refinement (PBE-ER). This rules out any equilibria if, under the same parameter ranges, all types of all states would prefer to play a different equilibrium (i.e., it is Pareto dominated given initial beliefs). This is substantively appealing for the cases that motivate us because security-seekers often want to end competition because it is inefficient. It also holds technical appeal. Empirical implications of all trust models are hampered by multiple equilibria in overlapping parameter ranges. One reason we are confident in our main conclusion is that regardless of whether we apply a refinement or not, we cannot rationalize trust-building in the international signaling model when initial trust is low. Yet, as we show below, we can rationalize trust-building in these ranges when we model domestic signaling opportunities. Further, in the domestic signaling model, only trust-building equilibria survive refinement when initial trust is low, and they reliably do so, suggesting that these results are most robust in interesting parameter ranges.

The results of our baseline analysis are summarized in Figure 1. Panel (b) describes the equilibria that can survive the PBE-ER. Two points are notable. First, all equilibria are symmetric, pure strategy equilibria. Second, given we start with the same baseline structure as others (e.g., Kydd Reference Kydd2005), the three PBE-ER we find are consistent with past scholarship on trust-building through international costly signaling. We are most interested in the trust-building equilibrium. In it, security-seekers always cooperate in the first period, but greedy states do not. Then security-seekers cooperate in the second period only if their counterpart cooperated in the first. We call it a trust-building equilibrium because each player learns the other’s motives from first-period actions. When they both cooperate, they come to trust each other, facilitating second-period cooperation.Footnote ³

Figure 1.

Equilibria that Survive Refinement in the Standard Trust-Building Game

Note: Panel (a) plots three equilibria that can survive an efficiency refinement given $ H=1.5,L=-0.1,k=0.9 $ . These equilibria uniquely survive refinement in their plotted parameter ranges. Panel (b) describes the strategy profiles. Since these are symmetric strategies, we present them for one state. We report inefficient pure and mixed strategy PBE in Supplementary Appendices A.2 and A.2.1, respectively.

Panel (a) plots the conditions under which the equilibria listed in panel (b) survive as a function of each player’s initial trust levels. The conditions reflect findings in past work (Kydd Reference Kydd2005).Footnote ⁴ Here, we emphasize the boundaries of the trust-building equilibrium (which we shade dark red in this and all future plots).Footnote ⁵ We find there is a minimal level of trust necessary to sustain trust-building. Crucially, this lower bound is not a function of the efficiency refinement.Footnote ⁶ Rather, we cannot sustain first-period cooperation in any PBE with low levels of initial trust because when at least one state initially believes the other is very likely greedy, that state defects in period one. The other side, regardless of its true motives and level of trust, also defects because it understands exploitation is certain. Because it is impossible to build trust in the first period, this also assures we cannot sustain second-period cooperation. Therefore, when initial trust ( $ {p}_i $ ) is low, the unique PBE in the baseline model is mutual defection in both periods. This finding explains why trust-building is difficult, especially for enduring rivals or other highly distrustful states. Over time, rivals may genuinely come to desire peace as a result of exhaustion, shifting values, or leadership change. But they may be too fearful to take the first step toward trust-building because they are mistrustful of their rival and fear that cooperative gestures will be exploited. This motivates our question: when states start out deeply mistrustful of each other, how can they initiate trust-building?

The Domestic Advantage: Payoff Independence

We argue that many domestic policy choices have an unexplored advantage that enhances their capacity to forge trust: their level of payoff independence. In brief, independence refers to the extent to which a state’s payoffs (i.e., the combination of benefits and costs accrued) from a choice depend upon what a foreign counterpart does.Footnote ¹³ The concept of independence becomes relevant whenever a state faces a domestic or international policy problem and can choose among different policy options to address it. A state’s payoff from each option always depends on its own preferences. Yet sometimes the payoff also depends, at least partially, on what option a foreign counterpart chooses.

We conceptualize the degree to which choices have payoffs that are independent from or dependent on a foreign counterpart’s actions as a continuum. At one extreme, a state’s choice is maximally independent if the value it gets from selecting either policy option is the same no matter what its foreign counterpart does. For example, suppose State A was faced with nationwide protests (the policy problem) and considering whether to repress or negotiate with protesters. If A’s value from either repressing or negotiating does not depend on anything B does, then A’s decision about how to respond to the protests is fully payoff independent. If instead, A’s payoff from this decision depends somewhat on B’s response (e.g., whether B lodges diplomatic criticism or imposes sanctions), then the choice would only be partially independent. Put another way, if a state can calculate its own payoffs from each policy option without considering what any other state will do, then the choice is fully independent. But if the payoffs depend slightly, partially, or highly on what another does, then the decision is slightly, partially, or highly dependent.

Existing research into the security dilemma assumes choices are highly dependent (Kydd Reference Kydd2005). For example, in the arms control variant, the immediate payoff A gets from the decision to reduce arms (cooperate) depends on whether B also reduces arms (cooperates) or not (defects). This assumption is reasonable because the value states accrue from many international policy options previously studied are heavily determined by what rivals do.

In contrast, domestic choices often have greater independence. Examples of domestic choices that are plausibly near-fully independent include reforming social welfare, reducing domestic regulations, or changing land use policy. The associated costs and benefits have little to do with the choices foreign countries make. But choices about immigration policy and press freedom could be moderately independent. Easing immigration standards could result in different levels of immigration depending on other states’ policy choices. The costs and benefits of permitting press freedom could depend somewhat on whether foreign counterparts engage in influence campaigns (Levin Reference Levin2021). This last example also highlights that States A and B need not be choosing among identical options in trust-building (or even among options that are both domestic or both international).

Overall, we argue that domestic choices tend to be more independent than foreign policy choices on average. We also believe few domestic choices are as dependent as the choice to cut or build arms that the trust-building literature typically focuses on. However, not all domestic (foreign) policy choices are fully independent (dependent), and some foreign policy choices are even more independent than some domestic choices. Table 3 summarizes how various domestic and international choices can vary in independence. To be clear, these are on-average codings, and we encourage researchers to apply contextual knowledge in determining how independent actions are in historical cases.

Table 3.

Examples of Choices with Varying Degrees of On-Average Independence

Note: High, medium, and low codings roughly approximate where choices fall on an independence continuum.

We will use our formal model to show that more independent choices, which tend to be domestic, have an important advantage when it comes to trust-building. As noted earlier, the primary barrier to trust-building with international signals is the fear of being cheated. However, this fear assumes that one state’s value for taking a trust-building action hinges on what the other state does (i.e., full dependence). If the payoff from taking an action is independent of what the other does, this fear diminishes, and security-seekers face little risk when they make choices that reflect their genuine cooperative preferences.Footnote ¹⁴ For example, a state shifting policy in response to a domestic crisis would not fear that a counterpart would be able to greatly exploit its choice.Footnote ¹⁵ At the same time, greedy states are more easily identified because they can no longer claim, “I am not cooperating because I am afraid that you will cheat me.” Therefore, domestic choices that are both independent and correlated with security-seeking or greedy international preferences can play a crucial role in international trust-building.Footnote ¹⁶

Crucially, independence does not mean that policy choices are costless. It means that the benefits and costs are not contingent on a counterpart’s behavior. In the absence of dependence on the counterpart’s behavior, the payoffs of choices will depend even more heavily on a state’s own preferences. This means that it would entail an even larger opportunity cost for a state to play against type, promoting behavior that reveals more information about type.

The Strategic Implications of Independence for Trust-Building

We introduce variation in payoff independence into the baseline two-period trust game described earlier. We conceptualize the level of independence of each state’s choices using the continuous parameters $ {\beta}_A $ , $ {\beta}_B\in \left[0,1\right] $ . We apply these parameters only to the first period of the model. We omit the independence parameters from the second period because they are effectively equal to 0—that is, the choices are fully dependent, ensuring the second period is identical to the classic trust problem.Footnote ¹⁷

We visualize the updated game with independence in Table 4. When $ {\beta}_A $ and $ {\beta}_B $ both equal 0, the model converges to the classic model of trust-building through international actions presented earlier (Table 1 and Figure 1). In this classic model, $ A $ ’s value for cooperating or defecting depends on a combination of $ A $ ’s type and $ B $ ’s choice. In contrast, when $ {\beta}_A $ equals 1, it means that $ B $ ’s choice in the first period has no impact on $ A $ ’s value from cooperation or defection in the first period.Footnote ¹⁸ Rather, $ A $ ’s first-period value from cooperation depends entirely on $ A $ ’s preferences.Footnote ¹⁹ When $ {\beta}_A $ and $ {\beta}_B $ are between these values, it means both states’ trust-building payoffs are partially dependent.

Table 4.

First-Period Payoffs in the Game with Independence

Note: These updated payoffs only apply to the first period. Second-period payoffs are as in Table 1. Both are equal when $ {\beta}_A={\beta}_B=0 $ .

We then solve for all the PBE of our model in Supplementary Appendix A.3 under the assumption that the level of independence can vary.Footnote ²⁰ Treating the baseline model as the counterfactual, we preview the effects of introducing and increasing independence in Table 5. The top row explains how introducing independent signaling actions alters the conditions for equilibria found in the baseline model. The second and third rows explain why new PBE-ER emerge when we introduce independence in the choices of one or both sides. As the table indicates, introducing independence has many nuanced implications. We can no longer support the commonly studied tragic equilibria when independence is high; the trust-building equilibrium originally identified now survives under lower levels of initial trust, and several new equilibria, including mixed strategy and asymmetric trust-building equilibria, emerge.

Table 5.

Equilibrium Changes from Introducing Independence

We explore all these nuances in the technical appendices. Here, we focus on establishing our primary claim: introducing independence facilitates trust-building under conditions of deep initial mistrust. To establish this claim as clearly as possible, we initially focus on the symmetric pure strategy trust-building equilibrium (shaded dark red in Figure 1). This was the only trust-building PBE-ER established in the baseline model, and it continues to survive as we introduce variation in independence, allowing for a clear comparison.Footnote ²¹ Proposition A.6, reported in the Supplementary Appendix, formally characterizes the conditions under which we can sustain this equilibrium given that payoff independence can vary. Here, we describe the empirical implications of increasing independence by contrasting when trust-building occurs under high independence, relative to the baseline condition. Results 1a and 1b described below follow from a comparative static analysis of the trust-building equilibrium, which is generally presented in Supplementary Appendix A.3.2.

Result 1a: Independence and trust-building given initial mistrust. Starting at the baseline ( $ {\beta}_i=0 $ ), as independence increases, the minimum level of initial trust ( $ {p}_i $ ) necessary to sustain a trust-building equilibrium decreases.

In the baseline model, we could not sustain trust-building equilibria when initial trust was low because the security-seeker faced two strategic problems that prevented her from taking the initial trust-building action. Increasing independence increasingly resolves both of these problems. The direct problem was that the security-seeking A worried that B would cheat her. As the independence of A’s choice increases ( $ {\beta}_A $ ), A’s value is increasingly determined by her type, and less by how B reacts. Since security-seekers directly benefit from cooperation, security-seeking A’s value for cooperation is increasing with independence, regardless of concerns that B is greedy. The indirect problem was that the security-seeking A understood that B was deeply suspicious of her. Thus, even though A was a security-seeker, A also understood that B’s suspicion of A would drive even security-seeking B to defection. Increasing the independence of B’s choice ( $ {\beta}_B $ ), indirectly incentivizes A to cooperate because A now believes that security-seeking B will cooperate in spite of B’s initial mistrust. As in the baseline model, both states weigh the tradeoffs of being exploited in the first period against the potential value of cooperation in both periods. But as independence increases, mistrustful security-seekers are increasingly willing to risk cooperation because they are less sensitive to the cost of being exploited (the direct effect), and believe they are less likely to be exploited by their rival (the indirect effect).

We visualize the effect of increasing two-sided independence in Figure 2. The figure assumes intermediate levels of independence for both states ( $ {\beta}_A=0.6,{\beta}_B=0.7 $ ), but otherwise holds the $ k,H,L $ values at the same levels as in Figure 1, allowing for direct comparison. Substantively, these intermediate levels of independence could reflect a scenario where both sides face domestic reform opportunities. Yet because not all independent choices are domestic, it could also reflect that one side faced a domestic choice and the other faced a reasonably independent foreign policy decision.

Figure 2.

Pareto Optimal Equilibria Given Two-Sided Independence

Note: As in Figure 1, we set $ H=1.5,L=-0.1,k=0.9 $ in panel (a). We additionally assume intermediate values $ {\beta}_A=0.6 $ , $ {\beta}_B=0.7 $ , which implies $ {\beta}_A\sim {\beta}_B>\frac{k}{1+k-L} $ . As in Figure 1, we shade trust-building dark red and suckers gray. Panel (b) describes the strategy profiles and color codings for the new PBE-ER that emerge. PBE-ER are unique in plotted parameter ranges

At the highest levels of initial trust, we still cannot sustain trust-building equilibria. The reason is that independence does not influence the greedy type’s incentives to conceal his true preference, which is to cheat his rival in the second period. However, looking at the lowest levels of initial trust reveals a surprising result:

Result 1b: Independence threshold. When both players’ choices are sufficiently independent (i.e., $ {\beta}_A,{\beta}_B>\frac{k}{1+k-L} $ ), a trust-building equilibrium always exists for states that start out with the highest possible level of confidence that the other is greedy (i.e., $ {p}_i\to 0 $ ).

The independence threshold represents the point where the maximum payoff a security-seeker can assure herself (i.e., her minmax) comes from selecting cooperation rather than defection.Footnote ²² Substantively, consider a case where a minor power A faces the decision to increase press freedom or not. State B (say the U.S.) could exploit this decision by meddling with a newly free press. But, in practice, B’s policy choice has only a small effect on A relative to the domestic implications of allowing press freedom. Thus, A’s value from promoting freedom does not depend much on how B reacts. Instead, A’s decision hinges on whether A actually wants press freedom or not. In this case, we might say that A’s independence threshold is met because even if A was so mistrustful of B that A was certain B would exploit A’s new press freedom, it would not influence A’s decision. If A was the type that truly wanted these reforms, A would implement them and simply accept that B would meddle.

The threshold’s existence demonstrates that Result 1a is not overshadowed by extreme initial mistrust. Rather, as independence increases, the strategic incentives to cooperate in the first period continuously increase for low-trust types until they dominate the concerns that cause security-seekers to defect. The threshold is decreasing in the cost of being exploited ( $ k $ ) and the benefit a security-seeker accrues from cheating her rival ( $ L $ ). While this threshold can vary, Figure 2 demonstrates that, under plausible costs and benefits, we meet it for even intermediate levels of independence.

Figure 2 assumes both sides’ trust-building actions meet the independence threshold. In practice, one state may have an independent trust-building opportunity, but the other may not. This could arise if A is more sensitive to exploitation than B. For example, democracies can be exposed to election rigging by promoting press freedom, but autocracies cannot. It could also arise because different states have different available trust-building options. If one state starts with free markets, it cannot use market reform to engender trust. This might leave that state with only more dependent options, like arms reductions, to signal its security-seeking preferences. Can trust-building survive when independence is lopsided? Figure 3 considers this case. The plot assumes that B’s trust-building action is moderately independent and B is moderately trusting of A ( $ {\beta}_B,{p}_A $ are moderately high). Meanwhile, it allows A’s level of independence ( $ {\beta}_A $ ) and initial trust in B ( $ {p}_B $ ) to vary. On the furthest left side of the plot, we observe the counterfactual to the baseline, where, like in the baseline, $ {\beta}_A=0 $ . But unlike in the baseline, $ {\beta}_B $ is higher. As the plot moves to the right side, $ {\beta}_A $ increases, and the parameter ranges cross over those analyzed in Figure 2.

Figure 3.

When Is Trust-Building Possible Given Variation in $ \beta $ ?

Note: Continues to assume $ k=0.9,H=1.5,L=-0.1 $ . Assumes B’s decision is moderately independent $ {\beta}_B=0.7 $ , and B has moderate initial trust $ {p}_A=0.5 $ . Plots PBE-ER given variation in the independence of A’s choice and initial trust. As in Figures 1 and 2, we shade trust-building dark red, mixed strategy light red, and suckers gray. The new thatched space is the asymmetric semi-tragic equilibrium (panel b). The new diagonal striated space is the asymmetric trust-building equilibria (panel b). Where the lines and shading intersect, there are multiple PBE-ER. Inefficient (including mixed strategy) equilibria are reported in the Supplementary Appendix.

We note two points. First, as with prior plots, the symmetric trust-building equilibrium arises in the dark-red shaded area. Examining how this equilibrium expands as a function of $ {\beta}_A $ visualizes Result 1a in action. When A’s decision is fully dependent ( $ {\beta}_A=0 $ ), there is a lower bound on the symmetric trust-building equilibrium. As $ {\beta}_A $ increases, this lower bound diminishes. A reaches the independence threshold when $ {\beta}_A=0.45 $ . From this level of independence onward, we can sustain symmetric trust-building at the lowest levels of initial trust.

Second, the bottom left corner of the plot reveals a surprising equilibrium:

Result 1c: Asymmetric trust-building. When trust and independence are lopsided (e.g., $ {\beta}_A,{p}_B $ small but $ {\beta}_B,{p}_A $ large), we can sustain asymmetric trust-building equilibria. When the scope of lopsidedness is extreme ( $ {\beta}_A,{p}_B\to 0 $ ), this equilibrium is unique.

This is a trust-building equilibrium because security-seeking A conditions second-period cooperation on B’s first-period choice, and A infers B is trustworthy if B cooperates in the first period. However, it carries several interesting properties. First, B knows that A always defects in the first period. Still, if B is a security-seeker, B accepts cooperation with certain exploitation. Second, first-period exploitation is necessary to forge trust and cooperation in the second period. If B does not accept exploitation, A infers that B is greedy, and second-period mutual defection is assured.

Sustaining this equilibrium requires lopsidedness in both independence and trust. State A must have a sufficiently dependent choice and be sufficiently distrustful that it will never cooperate in the first period. In contrast, B’s choice must be sufficiently independent that security-seeking B will tolerate certain exploitation. B must also have at least moderate levels of initial trust. The reason is that B does not learn anything from A’s first-period decision because A defects regardless of A’s type. Thus, if B starts out too suspicious, then B is unwilling to cooperate in the second period. But B’s initial trust cannot be too high either. If it is, then the suckers dynamic takes hold and greedy types play cooperation to cheat A in the second period.

So far, we have found unique PBE-ER. But Figure 3 shows some overlap between efficient asymmetric and symmetric equilibria. The reason is that while both sides do equally well in a symmetric equilibrium, all asymmetric equilibria assure one side initially exploits the other. The exploiter prefers asymmetric equilibria and the exploited prefers symmetric equilibria.

Overlapping equilibria can complicate empirical predictions. Here, they strengthen our overall claim. To summarize what we have found, when trust-building decisions are dependent (as assumed in previous work), we cannot sustain trust-building PBE if at least one side has low levels of initial trust. However, when trust-building decisions are at least moderately independent, not only can we sustain trust-building equilibria at low levels of initial trust, but only trust-building equilibria survive refinement. When both sides face independent choices, there is a clear prediction given low initial trust: symmetric trust-building. When the trust-building options available to both sides are lopsided, the result is more complicated. We often sustain both asymmetric and symmetric trust-building equilibria. But we do not find non-trust-building equilibria survive refinement.Footnote ²³ Thus, either way, our model predicts some form of trust-building even in lopsided cases with low levels of initial trust.

Salience of the Trust-Building Exercise

Existing trust-building work considers variation in the salience of different international policies. In classic models, the salience of the first-period international choice must be neither too low nor too high relative to the second-period choice. Kydd argues that for any given level of initial trust, there is a first-period choice with exactly the right salience to signal reassurance without exposing the sender to too much risk (Kydd Reference Kydd2005). Therefore, Kydd concludes that security-seekers, in principle, can always find a right-sized signal to build trust when starting from any initial trust level.

Skeptics argue it is difficult to finely calibrate international gestures to satisfy this “Goldilocks” problem (Lieber Reference Lieber2011; Montgomery Reference Montgomery2006; Rosato Reference Rosato2014). Goldilocks problems can be exacerbated by problems with interpretation. It is often hard to notice if a rival dismantles a few missiles given regular changes in force posture, and even harder to interpret the salience of such a gesture. These interpretation problems likely impede trust-building between enduring rivals because the calibration window is small when initial trust is low (see Yoder and Haynes Reference Yoder and Haynes2021).

We now show that trust-building through independent (usually domestic) choices does not succumb to calibration problems. We introduce a salience parameter, $ \theta >0 $ , into the model. Larger values of this continuous parameter indicate that the second-period choice is increasingly important relative to the first-period choice. For example, when $ \theta =2,1/2,1\hskip-0.10em , $ it means that the second-period choice is twice as, half as, and equally important as the first-period choice, respectively. We test the robustness of the trust-building equilibrium to the inclusion of salience in Supplementary Appendix A.4. Here, we describe the primary empirical implication of that result:

This finding has two substantive implications that contrast international and domestic choices. First, trust-building via domestic (or other independent) choices entails an easier calibration task than typically studied international trust-building. Given sufficient independence, no first-period decision is too salient for security-seekers to make the choice that signals their type. Therefore, rather than the baby steps toward establishing trust that we would expect to see with classic international signaling, states can potentially build trust quickly through high-stakes domestic choices. Second, domestic and international actions can work together. States can build initial trust via domestic choices when calibration problem makes international signaling initially intractable. Since domestic choices are imperfectly correlated with international preferences, they may not completely resolve the trust problem. But they may increase trust enough to make international signals, such as arms control agreements, viable.

Mapping the Model to the Case

In the Cold War case, we code the goal of security-seekers as rapprochement. Thus, second-period mutual cooperation would involve arms control and other security and economic agreements (ending the Cold War). Defection would involve continuing or intensifying the arms race or security competition, to include reneging on arms control or other agreements. We argue that the case conditions match the parameter ranges for symmetric trust-building (Supplementary Appendix B maps the model to the case in tabular form). We code both sides as genuinely security seeking by the late Cold War, consistent with classic research on the effects of both “New Thinking” and structural constraints (Brooks and Wohlforth Reference Brooks and Wohlforth2000; English Reference English2000; Risse-Kappen Reference Risse-Kappen1994).

Even though both states desired rapprochement, we code trust between the U.S. ( $ {p}_{USSR} $ ) and the Soviet Union ( $ {p}_{US} $ ) as very low in the early 1980s.Footnote ²⁴ In his first press conference as commander-in-chief, Reagan stated that the Soviets “reserve unto themselves the right to commit any crime, to lie, to cheat,” to achieve their goals (Gwertzman Reference Gwertzman1981). Soviet General Secretary Leonid Brezhnev, in 1980, labeled the U.S. an “unreliable partner” (FRUS 2018, 1977–1980, Volume XII, No. 166). Even the reform-minded Gorbachev was dismayed by U.S. defense policy, claiming in 1985 that the Americans “promise the world stability but in reality strive to wreck the military balance” (Eaton Reference Eaton1985).

Given mutual distrust between states that genuinely desire second-period mutual cooperation, our theory predicts that a symmetric trust-building equilibrium can emerge if both rivals have sufficiently independent and salient trust-building choices to make. By contrast, if only dependent trust-building choices were available, then they would not make them. Although Washington and Moscow had discussed arms control from the early 1980s onward, the benefits had low independence (low $ {\beta}_i $ ), creating an impediment to progress under low trust. In contrast, the Soviets’ domestic policy choices had moderate-to-high independence (moderate-to-high $ {\beta}_{USSR}\hskip-0.15em $ ). The payoffs from more lenient treatment of dissidents, greater free speech, and reduced central planning of the economy (the Soviets’ first-period choice in our model) were not greatly dependent on Western policies. For example, the main benefits that Soviet leaders expected from glasnost, a policy of greater openness and freedom of speech, were reduced corruption and increased efficiency. These benefits did not depend much on any Western responses. We also classify Soviet domestic choices as high salience. Policies such as glasnost and perestroika, which introduced some market forces into the Soviet economy, created a fundamental shift in Soviet political culture and governance, with direct implications for the Kremlin’s governing ideology and political control.

Finally, domestic and international preferences were sufficiently correlated to enable learning. The Cold War had a strong ideological component, pitting a liberal, capitalist West against an illiberal Communist Bloc. Soviet domestic reforms reflected a shift in intrinsic values (English Reference English2000). Though the Reagan Administration would have preferred the USSR to become a true democracy, even liberalizing reforms within an autocratic context indicated greater compatibility in international preferences. Overall, given the relatively high independence and salience of Soviet domestic choices, and the correlation between domestic and international preferences, our theory predicts that these choices had the characteristics necessary to launch trust-building.

Initial Trust-Building through Soviet Domestic Choices

East–West trust in the early 1980s could hardly have been lower. President Reagan believed that the Soviet Union was bent on global domination. To Reagan, the Soviets had exploited the U.S. during the 1970s, opportunistically jumping out ahead in the arms race and engaging in military adventurism while misguided Western leaders sought détente. As Reagan prepared to take office, he confided in a friend that “I don’t really trust the Soviets and I don’t really believe that they will really join us in a legitimate limitation of arms agreement” (Wilson Reference Wilson2014, 17). Similarly, in 1983, Secretary of State George Shultz stated that “Soviet actions have come into conflict with many of our objectives” and lamented Moscow’s penchant for “stretching a series of treaties and agreements to the brink of violation and beyond” (FRUS 2021, 1981–1988, Volume IV, Soviet Union, January 1983–March 1985, No. 61).

In 1985, Mikhail Gorbachev became General Secretary of the Communist Party of the Soviet Union. Gorbachev and his advisors desired to end the costly U.S.–Soviet competition (Brands Reference Brands2014). However, the combination of low trust and low independence inherent in defense policy made it difficult to initiate trust-building through costly international actions such as arms control. Reagan administration officials were wary of arms control with the Soviets, believing that “the Soviet Union had violated every arms agreement it had ever signed” (Wilson Reference Wilson2014, 66). As a result, arms control proposals in the early 1980s were often more about public relations than genuine attempts to build trust (Colbourn Reference Colbourn2022). Though the Soviets, especially by the time Gorbachev reached office, greatly desired reductions in defense spending, unilateral arms reductions would have raised security concerns and potentially damaged Soviet prestige. Therefore, arms control could not proceed.

Fortunately, domestic choices offered another means for initiating reassurance. Consistent with our theory, a shift in Soviet policy on dissidents and minorities became an early signal that the Kremlin might be a trustworthy partner for the West. A notable case involved Andrei Sakharov, the Soviet nuclear physicist and Nobel Peace Prize-winning dissident. According to Shultz, the U.S. point person on Soviet diplomacy, Sakharov’s release from internal exile in 1986, “affecting a man of towering intellect and moral authority, made an impact on some of Gorbachev’s most severe skeptics” (Shultz Reference Shultz2010, 1095). More systematic policy changes reinforced Western beliefs. Jack Matlock, the senior Soviet expert on the Reagan National Security Council (NSC) and later the U.S. ambassador in Moscow, highlighted their impact on Shultz’s views:

From 1986 to 1988, the number of exit visas issued to Soviet Jews (“refuseniks”) increased from 1,000 to 80,000, a shift which the CIA called “remarkable” (Brands Reference Brands2014, 137). Illustrating the moderate-to-high independence of the issue, the Soviets exhibited little concern that these moves could be exploited by the West. Shevardnadze even invited Shultz to provide Moscow with a list of potential émigrés for the Soviets to consider (Shultz Reference Shultz2010, 986).

Gorbachev eventually pursued broader reforms, glasnost and perestroika, to increase political and economic freedom. Glasnost focused on transparency and openness. For example, in 1986 Soviet leaders green-lighted the release of the film Repentance, which critically represented Stalin, a “bombshell” that would “change our social system” (Taubman Reference Taubman2017, 248). By 1987, glasnost was “spreading like wildfire on the steppe,” and had led to something closer to freedom of speech and freedom of the press (Taubman Reference Taubman2017, 314). Under perestroika, new laws legalized private enterprise (1986), allowed state enterprises to determine output on the basis of demand (1987), and permitted co-operatives (1988). As with emigration policy, the Soviets did not seem particularly concerned that perestroika and glasnost left the USSR vulnerable to exploitation by the West. On the contrary, Shultz and Gorbachev enjoyed open discussions about economic policy (Wilson Reference Wilson2014, 132–3).

Perestroika and glasnost further increased Western trust in the Soviet Union. As early as April 1986, Soviet reforms had sparked “prominent” deliberations in the U.S. government about the possibility that the Soviets were truly changing (Savranskaya, Blanton, and Zubok Reference Savranskaya, Blanton and Zubok2010, 116).Footnote ²⁵ By 1987, Soviet reform had convinced senior U.S. policymakers that international cooperation was possible. During a meeting between Gorbachev and Thatcher in late 1987, journalists observed that Thatcher “placed almost as much emphasis on the Soviet leader’s internal reforms as on the superpower talks, considering perestroika and glasnost as evidence of a determination which also promised progress in East-West negotiations” on arms control and other security issues (Naughtie Reference Naughtie1987). Reagan concurred, subsequently saying that Gorbachev’s book, Perestroika, which outlined the Soviet leader’s vision, made him hopeful about Washington–Moscow relations (Matlock Reference Matlock2004, 294).

International-Level Trust-Building Begins

Soviet domestic reforms not only built trust, but facilitated international cooperation requiring more dependent choices, including the INF Treaty. The conclusion of the INF Treaty reflected, to be sure, important concessions by Gorbachev (e.g., including the SS-23 in the deal, de-linking missile defense). However, its signing was aided by the trust that Gorbachev had generated with glasnost and perestroika. During his April 1987 visit to Moscow, Shultz and Gorbachev discussed Soviet economic reform in depth, “establishing greater trust” and helping to make “the prospect of the elimination of INF a reality” (Wilson Reference Wilson2014, 133). To whet Reagan’s appetite for a summit in Washington to sign the INF Treaty, Shultz reported from Moscow that “the Soviet Union is changing” (Leffler Reference Leffler2007, 399).

When Gorbachev eventually came to Washington to sign the INF Treaty in December 1987, he received a hero’s welcome. Gorbachev interpreted the success of the Washington summit as evidence that his reformist domestic program changed perceptions of the Soviet Union abroad. Briefing the Politburo afterwards, Gorbachev observed:

Soviet domestic reform also built trust with the U.S. Senate, which would go on to ratify the INF Treaty. As the influential chairman of the Senate Armed Services Committee, Sam Nunn, observed, “the advent of Gorbachev, glasnost, and perestroika has undeniably improved the overall climate for the conduct of superpower relations” (Nunn Reference Nunn1988, 3).

In 1988, the Kremlin’s domestic reforms provided even stronger evidence that the Soviet Union had fundamentally changed. On the eve of Reagan’s May visit to Moscow, the Kremlin released a set of “theses” for the upcoming 19th Party Conference, which indicated that Gorbachev wanted to further liberalize the Soviet system. Matlock, then U.S. ambassador to Russia, was “electrified” when he read it (Savranskaya, Blanton, and Zubok Reference Savranskaya, Blanton and Zubok2010, 110). The following day, the ambassador told Reagan that “the Soviet Union will never be the same” (Matlock Reference Matlock2004, 296). Just days later, in Red Square, Reagan declared that he no longer saw the Soviet Union as an “evil empire.”

The 1988 19th Party Conference proved another milestone in the Kremlin’s reform program. Though falling far short of democracy, the conference resulted in political liberalization unprecedented in the Soviet context. Gorbachev secured popular elections at lower levels of government and greater judicial independence and rule of law. The conference “dealt only briefly with international and security affairs” (Garthoff Reference Garthoff2000, 361). The lack of attention to foreign policy is notable both because it speaks to the relative independence of domestic reform and because a party conference focused on domestic policy had such a big impact on Western perceptions of the Soviet Union. The conference contributed further to Washington’s appetite for cooperation with Moscow; high-level diplomacy with the Soviets thereafter “expanded rapidly” (Matlock Reference Matlock2004, 306). In September, Gorbachev cemented his reforms by removing several conservatives from high-ranking posts. The senior CIA Soviet analyst during the mid- to late-1980s identified these events as a critical juncture for Western perceptions of the Soviet Union:

All told, domestic reforms undertaken by the Kremlin from 1986 to 1988 played a crucial role in creating an atmosphere in which international cooperation was easier to sustain. In addition to the INF Treaty, the U.S. and the Soviet Union concluded a number of smaller but meaningful agreements on issues as varied as monitoring nuclear tests, peaceful nuclear energy, fishing rights, space exploration, cultural and educational exchanges, and maritime navigation (Garthoff Reference Garthoff2000, 353). Although U.S.–Soviet rapprochement slowed during the transition between the Reagan and Bush presidencies, Bush declared in fall 1989 that “[t]he world will be a better place if perestroika succeeds” and laid out more than a dozen proposals to increase U.S.–Soviet cooperation, including lifting trade restrictions and supporting Soviet efforts to join the General Agreement on Tariffs and Trade (GATT) as an observer (Engel Reference Engel2017, 297). Soviet policymakers saw this as “the end of economic warfare” between the two states (Engel Reference Engel2017, 298). U.S.–Soviet cooperation continued to bear fruit even as the Soviet Union disintegrated and revolution swept through Eastern Europe. The U.S. and the Soviet Union concluded the Treaty on Conventional Forces in Europe and the Strategic Arms Reduction Treaty (START), and the Kremlin further demonstrated its benign intentions through domestic choices by showing restraint toward separatists in the Baltics.

Overall, given the independence and high salience of Soviet domestic choices, our theory predicts that trust-building should have been possible despite low initial trust and that Soviet domestic choices should have had a substantial impact on Western perceptions. This is what we observe. As shown above, domestic reform was perhaps the earliest indicator of changing Soviet intentions. Moreover, domestic policies, including Soviet treatment of dissidents, glasnost, and perestroika, were seemingly as or more important than foreign policy in convincing key Western officials that the two sides had compatible goals. By increasing trust, Soviet domestic reforms thus contributed to an unprecedented level of East–West cooperation.

View from the Soviet Union

On the other side, how did U.S. and Western policies contribute to Soviet trust? Here, we briefly highlight one important choice: the Western response to Soviet economic and political reforms.Footnote ²⁶ Gorbachev came to power in 1985 with security-seeking intentions, but nonetheless worried that leaders like Reagan were too stuck in the Cold War mindset to be partners in peace.Footnote ²⁷ When Gorbachev initiated his domestic reforms to revitalize the Soviet Union, the West might have wholly dismissed these efforts. Instead, Western governments largely encouraged these reforms, which reassured the Soviets of their trustworthy intentions.

We classify this Western choice as moderately independent ( $ {\beta}_{US} $ ). The choice to respond positively to Gorbachev would bring more benefits if Gorbachev followed through on reforms. However, even if he had not, the U.S. and its partners would still have gained important benefits. They would have seemed magnanimous on the world stage, demonstrating that they were not implacably hostile to the Eastern Bloc. In addition, anti-Communist leaders like Reagan and Thatcher would, in particular, benefit from showing they were not recklessly hawkish (Goldfien Reference Goldfien2025; Schultz Reference Schultz2005). Indeed, public approval for Reagan’s Soviet policy surged when he pivoted to a more moderate stance (Nincic Reference Nincic1988). Therefore, this choice—while more dependent than Soviet domestic reforms—was still independent enough that Western leaders like Reagan and Thatcher were willing to make it even in the context of substantial distrust. We also classify the choice as moderately salient ( $ {\theta}_{US} $ ) because it concerned the West’s orientation toward its primary geopolitical rival and affected the political standing of Western leaders. Since this was an international choice, the Soviet leadership could also be confident that it correlated strongly with future willingness to cooperate internationally.

Ultimately, the choice to encourage, rather than dismiss, Soviet domestic reforms succeeded in reassuring the Kremlin that the U.S., the U.K., and other Western countries could be partners in international cooperation. For example, as Gorbachev considered further liberalization of the Soviet economy in early 1987, George Shultz—who had earned a PhD in Economics from the Massachusetts Institute of Technology and run the Bechtel engineering company—offered counsel. According to one historian, “Distinctions between Milton Friedman and John Kenneth Galbraith were less important than an American secretary of state’s message of two states confronting common challenges, aspiring for common results, and establishing greater trust. Gorbachev and those around him began to sense that the most conservative and anticommunist presidential administration of the Cold War was actually out to help them,” not out to get them (Wilson Reference Wilson2014, 133).

The West’s other arch-capitalist leader, Margaret Thatcher, also supported Gorbachev’s reforms (Brown Reference Brown2020). Following a visit to Moscow in early 1987, Thatcher wrote to Reagan, “I am firmly convinced it is in our interest to encourage him [Gorbachev], especially in his endeavours to create a much more open society.”Footnote ²⁸ This attitude also contributed to Gorbachev’s confidence in Western intentions. Writing about Thatcher’s contributions to the end of the Cold War, Gorbachev reflected that she “was genuinely interested in what was happening in our country. She closely, and with astonishing command of detail, followed perestroika and glasnost, and sincerely wished for our process of change to succeed.” This attitude led Gorbachev to assess that Thatcher—echoing the British prime minister’s famous judgment of him—was a “person one can deal with” (Gorbachev Reference Gorbachev2013).

Our analysis of the U.S.–Soviet case shows that it is possible for domestic and international choices to work in tandem to build trust, even as the role of domestic choices is particularly crucial. It also offers a rejoinder to accounts that critique the U.S. for being too slow to build trust and cooperate with Gorbachev (e.g., Braumoeller Reference Braumoeller2013). Our concept of independence explains why trust-building likely required starting with reforms in the Soviet Union rather than ambitious arms control agreements. Yet we show that the U.S. was not simply passive before the INF Treaty was signed in late 1987. Rather, U.S. and Western encouragement of Gorbachev’s domestic reforms was a meaningful and conscious policy choice that effectively reassured the Soviet leader and promoted further East–West cooperation.

Mapping the Model to the Korea Case

When a group of military officers ousted the democratically elected government of South Korea in early 1961, it threw U.S.–South Korea relations into uncertainty. To establish the potential for trust-building to resolve this uncertainty, we begin by mapping the situation to our model (see Supplementary Appendix B for a more formal mapping of how the case corresponds with the equilibrium plotted in the bottom-left of Figure 3). Since the U.S. was staunchly anti-Communist, we define security-seeking intentions as a shared anti-Communist orientation. Therefore, the goal for cooperation among security-seekers is blunting the spread of Communism.

We code U.S. trust in the new South Korean regime as low (low $ {p}_{ROK} $ ). Even before the coup, Washington had worried about societal malaise and Communist influence in South Korea. The coup plotters were not immediately seen as preferable to the deposed government of Prime Minister Chang Myon (Brazinsky Reference Brazinsky2009). Indeed, their intentions were questioned. The coup leader, Park Chung-hee, had prior Communist ties, leading one U.S. intelligence estimate to comment that “we cannot rule out the possibility that [Park] is a long-term Communist agent, or that he might redefect” (FRUS 1996, 1961–1963, Volume XXII, Northeast Asia, No. 224). However, trust was lopsided. Park’s military government believed that it could trust in U.S. support provided that it was seen as anti-Communist and committed to real reform and modernization (high $ {p}_{US} $ ), even as it understood that the U.S. did not yet trust it (Kennedy Reference Kennedy1988, 214).

The independence of the choices available to the two sides was also lopsided. The most meaningful way the U.S. could signal its willingness to cooperate was to give diplomatic recognition and aid to the new government. This choice was highly dependent (low $ {\beta}_{US} $ ) because it would be embarrassing for the U.S. and a waste of resources if Seoul ultimately joined the Communist camp. In contrast, the main choices that the new Korean government had available to signal its anti-Communist credentials were domestic. Because Park’s regime would reap the domestic rewards of these policies regardless of what the U.S. did, they can be considered to have high independence (high $ {\beta}_{ROK} $ ). For both sides, the choices available were quite salient and likely to be strongly correlated with a future desire for anti-Communist cooperation.

Given the lopsided nature of both trust and the independence of choices, our theory predicts that an asymmetric trust-building equilibrium would be most likely to emerge in this case. Specifically, we would expect Seoul to cooperate by pursuing anti-Communist domestic policies, while Washington defected by withholding recognition and aid in the first round. This initial asymmetric cooperation would build enough trust to facilitate mutual cooperation in the subsequent round.

Asymmetric Trust-Building and International Cooperation

Consistent with expectations, an asymmetric trust-building period ensued after the coup, in which the U.S. withheld official recognition and support from the new regime, while the Park government engaged in domestic reforms that proved appealing to Washington. The U.S. understood that recognition and support for the new military government would confer legitimacy, the benefits of which would be highly dependent on the character of the new regime. Thus, American officials in Korea quickly disavowed any support for the coup (FRUS 1996, 1961–63, Volume XXII, Northeast Asia, No. 213), and Washington adopted a “cautious attitude of wait-and-see” (FRUS 1996, 1961–1963, Volume XXII, Northeast Asia, No. 216).

For the Park regime, anti-Communism at home and reforms aimed at developing the economy and rooting out corruption were generally independent. The benefits of these reforms largely derived from the Park government’s own values. Though the Park government sought American support, it also had strong views of its own about how South Korea should be organized (Brazinsky Reference Brazinsky2009). Park and his government undertook anti-Communist policies and swiftly, if undemocratically, implemented modernizing reforms across South Korea’s economy and society.

Park’s early moves greatly encouraged Washington. In a telegram from the U.S. embassy in Seoul in October 1961, the reassurance felt by American officials is palpable. Ambassador Samuel Berger wrote that the regime “has taken hold with energy, earnestness, determination and imagination, albeit with certain authoritarian and military characteristics” (FRUS 1996, 1961–1963, Volume XXII, Northeast Asia, No. 244). Further, “vigilance against communist subversion and quality and volume of anti-communist propaganda have greatly improved.” Berger concluded that the new regime “offers much hope” (FRUS 1996, 1961–1963, Volume XXII, Northeast Asia, No. 244). While it is unclear whether the U.S. ultimately preferred the Park government to its predecessor, it is clear that Park’s actions increased U.S. trust relative to the beginning of his tenure.

Efforts to improve economic performance—even by nondemocratic means—and a demonstrated anti-Communist orientation not only built trust with the U.S. but facilitated mutual cooperation going forward. The positive impression that the military government had made early in its tenure led the Kennedy administration to host Park at the White House in late 1961, which “played a significant role in stabilizing and legitimizing the South Korean regime” (Brazinsky Reference Brazinsky2009, 120–1). Notably, this is the sort of high-dependence action that can be hard to contemplate under low trust. Looking ahead, the U.S. could—despite the Park regime’s “shortcomings,” that is, authoritarian style—expect that American aid would “be more effectively used than by any previous government” and that the administration could “go to Congress this spring in good conscience… that our continuing massive support is well justified” (FRUS 1996, 1961–1963, Volume XXII, Northeast Asia, No. 244). For its part, the Park regime cooperated with the U.S. on the international stage by, for example, hosting U.S. troops on its soil and sending South Korean soldiers to support the American war effort in Vietnam. Thus, initial asymmetric trust-building led to eventual mutual cooperation.