Many social scientists believe that effects of policies or interventions vary for one individual to another. Existing approaches to the estimation of treatment heterogeneity require researchers to observe and specify moderating variables. However, moderators are often unknown, unobserved, or mismeasured. This paper proposes a nonparametric Bayesian approach that uncovers heterogeneous treatment effects even when moderators are unobserved. The method employs a Dirichlet process mixture model to estimate the distribution of treatment effects, and it is applicable to any settings in which regression models are used for causal inference. Empirical applications demonstrate how the method offers new insights. It discovers an unobserved cleavage in Americans' attitudes toward immigrant, an omitted moderator for the effect of indiscriminate counterinsurgency violence, and the form of heterogeneity in the effect of voter audits on voter buying. An application to a study on resource curse also shows that the method finds the subset of observations for which the monotonicity assumption of instrumental variable analysis holds.
Do mixed member electoral systems provide the ``best of both worlds''? We examine whether candidates in the proportional representation (PR) tier of these systems take policy stances closer to their party's position while candidates in the single member district (SMD) tier adopt policy positions to appeal to their districts preferences. We exploit a comprehensive panel survey of all candidates for the Japanese Upper and Lower House elections between 2003 and 2010 and estimate the policy positions of each candidate over time using Bayesian item response theory model. Our analysis suggests that candidates' policy positions vary substantially within parties. In addition, the two major parties appear to converge in the economic policy dimension during this period. In general, we find limited evidence consistent with the ``best of both worlds'' hypothesis. In the economic policy dimension, there is tentative evidence that Lower House SMD candidates are responding to their districts' preferences while PR candidates are closer to the position of the median party member. In the main foreign/security policy dimension, however, SMD candidates are not particularly responsive to their districts' preferences.
Instrumental variable estimation is a long-established means of reducing endogeneity bias in regression coefficients. Researchers commonly confront two problems when conducting an IV analysis: the instrument may be only weakly predictive of the endogenous variable, and the estimates are valid only for observations that comply with the instrument. We introduce Complier Instrumental Variable (CIV) estimation, a method for estimating who complies with the instrument. CIV uses these compliance probabilities to strengthen the instrument through up-weighting estimated compliers. As compliance is latent, we model each observation's density as a mixture between that of a complier and a non-complier. We derive a Gibbs sampler and Expectation Conditional Maximization algorithm for estimating the CIV model. A set of simulations shows that CIV performs favorably relative to several existing alternative methods, particularly in the presence of small sample sizes and weak instruments. We then illustrate CIV on data from a prominent study estimating the effect of property rights on growth. We show how CIV can strengthen the instrument and generate more reliable results. We also show how characterizing the compliers can help cast insight into the underlying political dynamic.
Information about insurgent groups is a central resource in civil wars: counterinsurgents seek it, insurgents safeguard it, and civilians often trade it. Yet despite its essential role in civil war dynamics, the act of informing is still poorly understood, due mostly to the classified nature of informant "tips.'' As an alternative research strategy, we use an original 2,700 respondent survey experiment in 100 villages to examine attitudes toward the Guardians of Peace program, a widespread campaign by the International Security Assistance Force in Afghanistan to recruit local informants. We find that coethnic bias --- the systematic tendency to favor cooperation with coethnics --- shapes attitudes about informing and beliefs about retaliation, especially among Tajik respondents. This bias persists even after adjusting for additional explanations and potential confounding variables, suggesting that identity considerations such as coethnicity may influence attitudes toward high-risk behavior in wartime settings.
Reciprocity plays a key role maintaining cooperation in society. However, little is known about the neural process that underpins human reciprocity during social interactions. Our neuroimaging study manipulated partner identity (computer, human) and strategy (random, tit-for-tat) in repeated prisoner's dilemma games and investigated the neural correlate of reciprocal interaction with humans. Reciprocal cooperation with humans but exploitation of computers by defection was associated with activation in the left amygdala. Amygdala activation was also positively and negatively correlated with a preference change for human partners following tit-for-tat and random strategies, respectively. The correlated activation represented the intensity of positive feeling toward reciprocal and negative feeling toward non-reciprocal partners, and so reflected reciprocity in social interaction. Reciprocity in social interaction, however, might plausibly be misinterpreted and so we also examined the neural coding of insight into the reciprocity of partners. Those with and without insight revealed differential brain activation across the reward-related circuitry (i.e., the right middle dorsolateral prefrontal cortex and dorsal caudate) and theory of mind (ToM) regions [i.e., ventromedial prefrontal cortex (VMPFC) and precuneus]. Among differential activations, activation in the precuneus, which accompanied deactivation of the VMPFC, was specific to those without insight into human partners who were engaged in a tit-for-tat strategy. This asymmetric (de)activation might involve specific contributions of ToM regions to the human search for reciprocity. Consequently, the intensity of emotion attached to human reciprocity was represented in the amygdala, whereas insight into the reciprocity of others was reflected in activation across the reward-related and ToM regions. This suggests the critical role of mentalizing, which was not equated with reward expectation during social interactions.