Andreas Duus Pape

Associate Professor of Economics

Department of Economics, Binghamton University

Andreas Duus Pape studies agent-based modeling, decision theory, game theory, behavioral economics, and complex systems. He builds computational agents with psychologically realistic behaviors, such as similarity in a case-based model, and brings them to experimental and field data. He also applies these tools to environmental and public finance questions, such as taxation, as well as to collective resources and collective action, particularly from an Ostrom perspective.

Curriculum Vitae


Research Statement. Citations.

(Legend: ABM = Agent-Based Modeling. B= Behavioral Economics. DT=Decision Theory. GT = Game Theory. PUB=Public Finance/Policy.)

Topic 1: Agent-based Methods applied to Case-based Decision Theory

In this series of papers, I construct an agent-based model of Case-based Decision Theory and successfully explain empirical findings about human learning and game play. I also use this model to expand the literature about learning in macroeconomics.

  1. Pape AD, Kurtz KJ. Evaluating Case-Based Decision Theory: Predicting Empirical Patterns of Human Classification Learning. Games and Economic Behavior 82 (2013): 52-65.     ABM, DT, B.

    We introduce a computer program which calculates an agent's optimal behavior according to Case-based Decision Theory (Gilboa and Schmeidler, 1995) and use it to test CBDT against a benchmark set of problems from the psychological literature on human classification learning (Shepard et al., 1961.) We find (1) CBDT correct predicts the empirically observed relative difficulty and speed of learning in the human data, (2) 'Similarity' is decreasing in vector distance, which is consistent with evidence from Psychology, (3) the best-fitting parameters suggest humans aspire to an 80-80% success rate, and (4) Average similarity is rejected in favor of additive similarity.

  2. Guilfoos T, Pape AD. Predicting Human Cooperation in the Prisoner's Dilemma Using Case-based Decision Theory. Theory and Decision (2015): 1-32.     ABM, DT, B, GT.

    We use the computer program introduced above to show that Case-based Decision Theory can explain the aggregate dynamics of cooperation in the repeated Prisoner's Dilemma, as observed in the experiments performed by Camera and Casari (2009). We find CBDT provides a better fit than does the existing Probit model, which is the first time such a result has been found. We also find that humans aspire to a payoff above the mutual defection payoff but below the mutual cooperation outcome, which suggests they hope, but are not confident, that cooperation can be achieved.

  3. Pape AD, Xiao W. Case-Based Learning in the Cobweb Model.     ABM, DT, B.

    The macroeconomics learning literature uses bounded rationality and adaptive learning as an alternative to rational expectations to model the macroeconomy. We use the computer program introduced above to model bounded rationality, which is the first model which uses cased-based reasoning in macroeconomic models. When we apply this case-based learning approach to the Cobweb model, we find that market prices converge to the rational expectations prices if agents search for a better outcome with an intermediate level of persistence. If agents are insufficiently persistent, then multiple equilibria abound and the rational expectations equilibrium becomes a special case. On the other hand, if they are too persistent, they can search forever and never achieve convergence.

Topic 2: Public Goods and Common Resources Policy

In this series of papers, I use agent-based modeling and game theoretic techniques to provide insight into the management of public goods and common resources.

  1. Guilfoos T, Pape AD, Khanna N, Salvage K. Groundwater Management: The Effect of Water Flows on Welfare Gains. Ecological Economics 95 (2013): 31-40.     ABM, GT, PUB.

    We construct a spatially-explicit and agent-based groundwater model that has multiple cells and finite hydraulic conductivity to estimate the gains from groundwater management and the factors driving those gains. We calibrate an 246-cell model to the parameters and geography of Kern County, California, and find that the welfare gain from management significantly higher in the multi-cell model (27%) than in the bathtub model (13%). This shows the importance of this modeling approach to questions of social welfare.

  2. Pape, AD, Anderson, N, Guilfoos, T, and Schmidt J. An Agent-based Model of Tax Ceilings: Leviathan Extraction and Tax Payment Uncertainty. The Review of Behavioral Economics (3) 1. Special Issue.     ABM, GT, PUB.

    Special Issue Title: "Shedding Light on the Shadow of the Economy: Research Methods in Studies on Tax Behavior."

    In this paper, we present a new agent-based model that incorporates unique administrative data on property taxes to provide, for the first time, a detailed explanation of how tax payment uncertainty determines the extraction choice set of a Leviathan government. Our agent-based model is unique because it allows voters to forecast the future effects of alternative policies by simulating rational expectations. For a given referendum, we use the model to generate a joint distribution for all random variables as a function of the proposed policy. Voters then choose whether to support the referendum by calculating their expected utility with this distribution as a prior. Our model incorporates administrative data on tax-payment uncertainty from two American cities with different property tax assessment regimes and thus different profiles of tax- payment uncertainty across their populations. Our model suggests that voters in a city with a high-uncertainty assessment regimes would support a tax ceiling on their Leviathan government that is five times as restrictive as voters in a city with a low-uncertainty assessment regime.

  3. Pape AD, Seo M. Reports of Water Quality Violations induce Consumers to buy Bottled Water. Agricultural and Resource Economics Review. 44(1), 2015.     PUB

    The 1996 Safe Drinking Water Act Amendments require that water utilities send quality reports to customers. We test whether receiving WQRs of health violations increases purchases of bottled water. With new data, we find a larger response than previous studies, and, unlike previous studies, we disaggregate the intensive and extensive margins of demand changes. The order of magnitude of our results implies American consumers spend approximately $300 million dollars per year—about 4% of yearly bottled water expenditures— to avoid water quality violations.

  4. Anderson, N, Pape AD, Self-Imposed Constraints on Collective Action Under Uncertainty.     GT, PUB.

    Under what conditions does, or should, a collective of rational individuals support the imposition of a binding constraint on their own collective action? Our innovation is to allow citizen-taxpayers in a standard political economy model to be risk-averse and uncertain about the future average cost of the collective good, their own future income or wealth, and the future distribution of the tax burden. We show that if citizen-taxpayers face such uncertainty, an agency problem — broadly defined as a flaw in the political process that produces collective decisions incongruent with majority preferences — is neither necessary nor sufficient to justify a binding tax ceiling. This is a striking result, because it contradicts existing theory and can help explain why, empirically, some tax systems are more subject to ceilings than others.

Topic 3: Decision Theory, Behavioral Economics, Psychology

  1. Pape AD, Kurtz KJ, Sayama H. Complexity Measures and Classification Learning. Revise and Resubmit, Journal of Mathematical Psychology. 64 (2015): 66--75.     B.

    In classification learning experiments, test subjects are presented with objects which they must categorize. The correct categories, which are known to the experimenter, are functions of the characteristics (“dimensions”) of the objects, such as size, color, brightness, and saturation. The experiments measure the relative difficulty of learning different categorizations. One major factor which influences the difficulty of learning is whether the dimensions are easily distinguishable by the subjects. Separable dimension problems are ones in which humans can differentiate dimensions of the objects, e.g. size, color, and shape. Integral dimension problems are ones in which humans can not (easily) distinguish the dimensions, e.g. brightness, saturation, and hue. Psychologists reason that when dimensions are separable, humans develop logical rules about which dimensions matter, and when they are integral, humans do not (cannot) develop these rules.

    Feldman (2000) connected separable dimension learning to logical complexity. Logical complexity is a mathematical metric which ranks problems by how concisely the solution can be represented as a set of logical rules. Feldman showed that a logical complexity metric can accurately predict the relative difficulty of separable dimension learning problems for human subjects.

    We find the other half of Feldman’s story. In integral dimension problems, when humans cannot differentiate dimensions of objects and cannot form logical rule systems, humans fall back statistical information. Statistical complexity is a mathematical metric which evaluates the informational density of a problem statistically. Our conclusion is parallel to Feldman’s: While Feldman finds that logical complexity predicts learning difficulty in separable dimension problems, we find that statistical complexity predicts learning difficulty in integral dimension problems.

  2. Pape, AD. Action-Independent Subjective Expected Utility Without States of the World. Theoretical Economics Letters 3 (2013): 17-21.     DT.

    This paper develops an axiomatic theory of decision-making under uncertainty that has no state-space. The choice setting follows Karni: a set of effects (outcomes), a set of actions which induce these effects, and a set of real-valued bets over effects. In Karni's representation, a preference over action/bet pairs yields utility, which is action-dependent. In our representation, utility is action-independent. This is achieved by augmenting Karni's choice set with lotteries over actions. Identification is achieved similarly to Anscombe-Aumann, in which there are objective ``roulette'' lotteries over subjective ``horse race'' lotteries.

  3. Bose S, Ozdenoren E, Pape AD. Optimal Auctions with Ambiguity. Theoretical Economics. December 2006.     DT, GT.

    In this paper, we study the optimal auction problem allowing for ambiguity about the distribution of valuations, where agents are ambiguity averse. When the bidders face more ambiguity than the seller we show that (i) given any auction, the seller can always (weakly) increase revenue by switching to an auction providing full insurance to all types of bidders, (ii) if the seller is ambiguity neutral and any prior that is close enough to the seller's prior is included in the bidders' set of priors then the optimal auction is a full insurance auction, and (iii) in general neither the first nor the second price auction is optimal (even with suitably chosen reserve prices).

Topic 4: Public Policy

  1. Chaplin D, Pape AD, Turner M. Minimum Wages and School Enrollment of Teenagers: A Look at the 1990s. The Economics of Education Review 22 (2003): 11-21.    PUB

    In this paper we estimate the effects of higher minimum wages on school enrollment using the Common Core of Data, collected by the US Department of Education. These data cover the entire population of public school students in the United States. We find some evidence that higher minimum wages reduce teen school enrollment in states where students can drop out before the age of 18. This appears to be driven by the grade 9 to grade 10 transition. This suggests that minimum wages may have a substantial effect on teens’ schooling effort in these early grades but also that these unintended effects can be offset by policies that encourage continued school enrollment.


  1. Teaching Statement

  2. SOOT Survey results: