Search results

Incomplete information exacerbates the problems inherent in collective action. Participants cannot efficiently coordinate their actions if they do not know each other's preferences. I investigate when ordinary communication, or cheap talk, may resolve mutual uncertainty in collective action problems. I find that the efficacy of communication depends critically on the relationship between contributions and the value of the joint project. The incentive barriers to honesty are highest when every contribution increases the project's value. Participants then have a strict incentive to say whatever would induce others to contribute the most, so cheap talk lacks credibility. By contrast, when contributions may be marginally worthless, such as when the project has no value unless contributions hit a certain threshold, communication may help participants avoid wasted effort. Using these findings, I identify which collective action problems in politics might benefit from communication and which require more expensive solutions to overcome uncertainty.

AbstractFormal models commonly characterize interstate bargaining as dichotomous, ending in either war or peace. But there are many forms of coercion—including supporting rebel groups, sanctions, and cyberattacks. How does the availability of intermediate policy options affect the incidence of war and peace? We present an analysis of crisis bargaining models with intermediate policy options that challenges conventional results about the relationship between private information and negotiation outcomes. In our "flexible‐response" modeling framework, unlike in traditional crisis bargaining models, we find that greater private war payoffs may be associated with a lower probability of war or worse settlement values. When intermediate options are available, the relationship between the private efficacy of war and the private efficacy of these other options largely determines equilibrium outcomes. By utilizing the tools of mechanism design, we derive game‐form–free results on how private information shapes international conflict, regardless of the precise negotiating protocol.

AbstractThe development of parliamentary constraints on the executive was critical in Western European political history. Previous scholarship identifies external wars as a key factor, but with varying effects. Sometimes, willing monarchs granted parliamentary rights in return for revenues to fight wars. Yet at other times, war threats empowered rulers over other elites or caused states to fragment. We analyze a formal model to understand how external wars can either stimulate or undermine prospects for a contractual relationship between a ruler and elite actors. We recover the standard intuition that war threats make the ruler more willing to grant parliamentary rights in return for revenue. Our key insight is that war threats also affect the bargaining position of elites. A previously unrecognized tension yields our new findings: stronger outsider threats increase pressure either on elites to fund the ruler or on the ruler to accept constraints—but not both simultaneously. Elites with immobile wealth depend on the ruler for security. War threats undercut their credibility to refuse funding for an unconstrained ruler. By contrast, war threats make elites with mobile wealth and a viable exit option unwilling to fund a hopeless war effort. Only under circumscribed conditions do war threats align three conditions needed for parliament to arise in equilibrium: ruler willingness, elite credibility, and elite willingness. We apply our theory to posit strategic foundations for waves and reversals of historical European parliaments.

AbstractMany enduring questions in international relations theory focus on power relations, so it is important that scholars have a good measure of relative power. The standard measure of relative military power, the capability ratio, is barely better than random guessing at predicting militarized dispute outcomes. We use machine learning to build a superior proxy, the Dispute Outcome Expectations (DOE) score, from the same underlying data. Our measure is an order of magnitude better than the capability ratio at predicting dispute outcomes. We replicate Reed et al. (2008) and find, contrary to the original conclusions, that the probability of conflict is always highest when the state with the least benefits has a preponderance of power. In replications of 18 other dyadic analyses that use power as a control, we find that replacing the standard measure with DOE scores usually improves both in‐sample and out‐of‐sample goodness of fit.

In a recent article in this journal, Chapman presents a formal model of the informational role played by international institutions. Unfortunately, the equilibria given in the article are incorrect. In this article, we identify the errors in the analysis of Chapman and solve for correct equilibria to the model. Our results show little support for the empirical implications derived in the original article. Contrary to these original findings, we find that there may be no relationship between an institution's policy position and its effect on domestic public opinion or the likelihood that leaders will consult the institution. [Reprinted by permission of Sage Publications Inc., copyright holder.]

In a recent article in this journal, Chapman presents a formal model of the informational role played by international institutions. Unfortunately, the equilibria given in the article are incorrect. In this article, we identify the errors in the analysis of Chapman and solve for correct equilibria to the model. Our results show little support for the empirical implications derived in the original article. Contrary to these original findings, we find that there may be no relationship between an institution's policy position and its effect on domestic public opinion or the likelihood that leaders will consult the institution.

The effect of conditioning on an additional covariate on confounding bias depends, in part, on covariates that are unobserved. We characterize the conditions under which the interaction between a covariate that is available for conditioning and one that is not can affect bias. When the confounding effects of two covariates, one of which is observed, are countervailing (in opposite directions), conditioning on the observed covariate can increase bias. We demonstrate this possibility analytically, and then show that these conditions are not rare in actual data. We also consider whether balance tests or sensitivity analysis can be used to justify the inclusion of an additional covariate. Our results indicate that neither provide protection against overadjustment.

Violence against civilians in civil war is widely thought of as a strategic choice by combatant groups. We argue that a common strategic logic of competition underlies diverse theories of civilian victimization. We develop a theory of strategic complements in victimization, hypothesizing that an armed group's propensity to victimize civilians will increase with its expectation that its competitors will act likewise. We test this argument by structurally estimating a formal model of strategic interdependence between armed groups using data from the Colombian civil war. Our findings indicate that strategic expectations are responsible for a substantial amount of violence against civilians: the two major combatant groups would have systematically victimized civilians in at least 9% fewer municipalities if they had expected no violence by their rival. Examining causal mechanisms, we also find that victimization in the Colombian case was more likely aimed at controlling civilians than at influencing peace negotiations.

Abstract
Large language models (LLMs) offer new research possibilities for social scientists, but their potential as "synthetic data" is still largely unknown. In this paper, we investigate how accurately the popular LLM ChatGPT can recover public opinion, prompting the LLM to adopt different "personas" and then provide feeling thermometer scores for 11 sociopolitical groups. The average scores generated by ChatGPT correspond closely to the averages in our baseline survey, the 2016–2020 American National Election Study (ANES). Nevertheless, sampling by ChatGPT is not reliable for statistical inference: there is less variation in responses than in the real surveys, and regression coefficients often differ significantly from equivalent estimates obtained using ANES data. We also document how the distribution of synthetic responses varies with minor changes in prompt wording, and we show how the same prompt yields significantly different results over a 3-month period. Altogether, our findings raise serious concerns about the quality, reliability, and reproducibility of synthetic survey data generated by LLMs.