Stephanie W. Wang

Curriculum Vitae [PDF]

Research interests

  • Experimental Economics; Behavioral Economics/Game Theory; Information Economics

  • Stephanie W. Wang


  • Belief Updating in Sequential Games of Two-Sided Incomplete Information: An Experimental Study of a Crisis Bargaining Model (with Dustin H. Tingley) [PDF] [Supplementary Materials]
    Quarterly Journal of Political Science, 5(3), 243-255 (2010)

  • Incentive Effects: The Case of Belief Elicitation from Individuals in Groups [PDF]
    Economic Letters, 111, 30-33 (2011)

  • Does Recession Reduce Global Health Aid? Evidence from 15 High-Income Countries, 1970-2007 (With David Stuckler, Sanjay Basu, and Martin McKee)
    Bulletin of the World Health Organization, Forthcoming

  • On Eliciting Beliefs in Strategic Games (with Thomas R. Palfrey) [PDF]
    Journal of Economic Behavior and Organization, 71, 98-109 (2009)

  • What Kind of Memory Supports Visual Marking? (With Yuhong Jiang) [PDF]
    Journal of Experimental Psychology: Human Perception & Performance, 30(1), 79-91 (2004)

  • Working papers

    Speculative Overpricing in Asset Markets with Information Flows (with Thomas R. Palfrey) [PDF]
    Revise and Resubmit, Econometrica


    The paper derives and experimentally tests a theoretical model of speculation in multi-period asset markets with two underlying states of the world and information flows. The specula- tion arises from the traders’ heterogeneous posteriors as they make di?erent inferences from sequences of public information, and this leads to overpricing in the sense that price exceeds the most optimistic belief about the real value of the asset. We find consistent speculative overpricing with both incomplete and complete markets, where the information flow is a grad- ually revealed sequence of imperfect public signals about the state of the world. We also find strong evidence of asymmetric price reaction to good news and bad news, another important feature of equilibrium price dynamics under our model. The markets we ran with a relaxed short-sales constraint exhibit significantly less overpricing.

    Dynamically Optimized Sequential Experimentation (DOSE) for Estimating Economic Preference Parameters (First author, with Colin F. Camerer and Michelle Filiba) [PDF]
    Revise and Resubmit, American Economic Review


    Dynamically optimized sequential experiments (DOSEs) for estimation of risk preferences start with a distribution of beliefs about risk preference parameters, and a set of questions, then dynamically choose the question that maximizes information gain considering previous answers. Applying the method to the 10-question set of Holt and Laury (2002) and the 140-question set of Sokol-Hessner et al. (2009) to measure risk-aversion and loss-aversion shows that DOSE sequences create a 50-70% increase in speed of inference about parameter from fewer questions. Simple DOSE designs could be useful in complex environments with highly-distractible groups like internet groups, children, CEOs and monkeys.

    Imperfect Choice or Imperfect Attention? Understanding Strategic Thinking in Private Information Games (with Isabelle Brocas, Juan D. Carrillo, and Colin F. Camerer) [PDF]
    Revise and Resubmit, Review of Economic Studies


    In experiments, people do not always appear to infer the information of other players from their choices. To understand this thinking process further, we use "Mousetracking" to record which game payoffs subjects look at, and for how long, in games of private information with three information states, which vary in strategic complexity. Subjects often deviate from Nash equilibrium choices, converge only modestly toward equilibrium across 40 trials, and often fail to look at payoffs which they need to in order to compute an equilibrium response. When cluster analysis is used to group subjects according to lookup patterns and choices, the clusters appear to correspond approximately to level-3, level-2 and level-1 thinking in level-k cognitive hierarchy models. Deviations from Nash play are associated with failure to look at the necessary payoffs. The connection between looking and choices is strong enough that the time durations of looking at key payoffs can predict choices, to some extent, at the individual level and at the trial-by-trial level.

    Patience Auctions: The Impact of Time vs. Money Bidding on Elicited Discount Rates (with Christopher Y. Olivola) [PDF]
    Under Revision


    We introduce two auction-based methods for eliciting discount rates. In these "patience auctions", participants bid the smallest future sum they would prefer -or- the longest time they would wait for a reward, rather than receive a smaller, immediate payoff. The most patient bidder receives the delayed reward; all others receive the immediate payoff. These auctions, in addition to offering certain potential methodological advantages, allow us to compare discounting when participants' attention is focused on the temporal vs. monetary dimension of delayed rewards. We find that the best-fitting model and rate of discounting differ between these two theoretically equivalent bidding methods.

    Prediction in Networks


    In firms, neighborhoods, and many other environments, individuals form probabilistic beliefs about the state of the world based on private information and the beliefs revealed by others who are connected with them in their communication network. We experimentally test the benchmark predictions for belief dynamics under common knowledge of Bayesian updating in several ten-agent networks (star, chain, circle, core-periphery, and complete) with multiple rounds of communication. We document significant differences in belief accuracy and convergence across network structures. Furthermore, because each individual in our experimental sessions were put into each of the network structures and into different positions within the same network, we can identify the relative influence of potential behavioral biases such as overweighting of own signal, double-counting the same source (persuasion bias), and optimism/pessimism on individual level belief updating.

    Visual Eavesdropping: Shifting Attention and Changing Expressed Social Preferences (First author, with Colin F. Camerer and Antonio Rangel)


    Subjects choose between two options in three types of simple distribution games: allocation to self and two other subjects, allocation to self and one other subject, and allocation to two other subjects. All payoffs in the distribution games are initially hidden and we generate a rich dataset of payoff lookups and allocation choices using mousetracking. We find that a subject's lookup patterns, specifically the ratio of time spent looking at others' payoff to own payoff, is significantly and negatively correlated with that subject's percentage of selfish choices. Using a novel screensharing technique where one of the subjects receiving the allocation can observe the decision maker's clicks (but not choices), we also study the impact of social image concerns on expressed social preferences. We find that this "visual eavesdropping" significantly alters both lookup patterns and choices. Subjects spend more time looking at the potential payoffs to the observer compared to that of the non-observer. They also shift away from the selfish choice. We also find evidence for selfish and maximin but not utilitarian preferences under the Charness-Rabin specification.

    Not All Nudges are Created Equal: Moral Sentiments and Contribution to Public Goods (with Margaret A. McConnell)


    We provide experimental evidence consistent with morally motivated charitable giving. We find that giving subjects with a suggested contribution amount increases the willingness to give and that framing the suggestion with moral language further increases contributions. We develop a simple theoretical framework in which providing a suggested contribution activates individuals' concern for their own self-image by providing a benchmark for what a generous individual would do. Within our model, moral framing language converts the individual's self-image benefit from their contribution into a binary good and bad classification given the benchmark. However, the lack of an explicit benchmark gives individuals the moral wiggle room to adopt a self-serving one. The behavior we see in our experiments closely follow our model's predictions.

    Yes, I Knew It All Along: Visual Focus and Hindsight Bias (with Daw-An Wu, Shin Shimojo, and Colin Camerer)


    Agents are shown blurred images that slowly clarify to a given point. They are then incentivized to report the likelihood that the image contains a human. Principals are primed with the image in total clarity before going through the agent's task. They are incentivized to guess the percentage of agents who guessed a human. We demonstrate that the principals exhibit the visual hindsight bias, that is, they overestimate the agents' ability to correctly parse the image. We also find a selective de-biasing effect for human images when principals are shown movies of the agents' eye gazes overlaid on the images.