We develop a deep learning model to detect emotions embedded in press conferences after the Federal Open Market Committee meetings and examine influence of detected on financial markets. find that, controlling for Reserve’s actions sentiment policy texts, positive tone voices Reserve chairs leads significant increases share prices. Other variables also respond vocal cues from chairs. Hence, how...

A principal owns a firm, hires an agent of uncertain productivity, and designs dynamic policy for evaluating his performance. The observes ongoing evaluations decides when to quit. When not quitting, the is paid wage that linear in expected productivity; claims residual After players secure fixed outside options. I show equilibrium Pareto efficient. For broad class performance technologies, det...

How do social interactions shape collective action, and how are they mediated by networked information technologies? We answer these questions studying the Temperance Crusade, a wave of anti-liquor protest activity spreading across 29 states between 1873 1874. Relying on exogenous variation in network links generated railroad accidents, we provide causal evidence driving diffusion movement, rai...

This paper combines a data-rich environment with machine learning algorithm to provide new estimates of time-varying systematic expectational errors (“belief distortions”) embedded in survey responses. We find sizable distortions even for professional forecasters, all respondent-types overweighting the implicit judgmental component their forecasts relative what can be learned from publicly avai...

This paper investigates the trade-offs between local elites and state agents as tax collectors in low-capacity states. We study a randomized policy experiment assigning neighborhoods of large Congolese city to property collection by chiefs or agents. Chief raised compliance 3.2 percentage points, increasing revenue 44 percent. Chiefs collected more bribes but did not undermine morale trust gove...

Justified communication equilibrium (JCE) is an refinement for signaling games with cheap-talk communication. A strategy profile must be a JCE to stable outcome of nonequilibrium learning when receivers are initially trusting and senders play many more times than receivers. In the model, counterfactual “speeches” that have been informally used motivate past refinements messages actually sent. S...

We show that rational but inattentive agents can become polarized, even in expectation. This is driven by agents’ choice of not only how much information to acquire, also what type information. present optimal acquisition, and subsequent belief formation, depends crucially on the agent-specific status quo valuation. Beliefs systematically update away from realized truth with same initial belief...

We investigate information aggregation and competition in a delegation framework. An uninformed principal is unable to perform task herself must choose between one of two biased imperfectly informed experts. In the focal equilibrium, experts exaggerate their biases, anticipating an ideological winner’s curse. show that having second expert can benefit principal, even when equally or more than f...

We show that the treatment effect estimated by standard methods such as regression discontinuity analysis or difference-in-differences may contain a transient “learning effect” is entangled with long-term of treatment. This learning occurs when variable interest agents’ efforts, and control correspond to success failure: failure gives agents information about how much their effort matters, cons...

We develop an equivalence between the equilibrium effects of incomplete information and those two behavioral distortions: myopia, or extra discounting future; anchoring current behavior to past behavior, as in models with habit persistence adjustment costs. show how these distortions depend on higher-order beliefs GE mechanisms, they can be disciplined by evidence expectations. finally illustra...