Assistant Professor
University of California, Berkeley
Department of Economics
kariv[at]
[Home] [Contact]
[CV] [Research] [Teaching]
RESAERCH
PUBLISHED AND FORTHCOMING PAPERS
-
Bayesian Learning in Social
Networks, with Douglas Gale, NYU. Games and
Economic Behavior, November 2003
Abstract. In this paper, we extend the standard model of social learning in
two ways. First, we introduce a social network and assume that agents can only observe
the actions of agents to whom they are connected by this network. Secondly, we
allow agents to choose a different action at each date. If the network
satisfies a connectedness assumption, the initial diversity resulting from
diverse private information is eventually replaced by uniformity of actions,
though not necessarily of beliefs, in finite time with probability one. We look
at particular networks to illustrate the impact of network architecture on
speed of convergence and the optimality of absorbing states. Convergence is
remarkably rapid, so that asymptotic results are a good approximation even in
the medium run.
-
Observational Learning Under
Imperfect Information, with Boğaçhan
Çelen,
Abstract. This paper explores Bayes-rational sequential decision making in a
game with pure information externalities, where each decision maker observes
only her predecessor's binary action. Under perfect information, the martingale
property of the stochastic learning process is used to establish convergence of
beliefs and actions. Under imperfect information, in contrast, beliefs and
actions cycle forever. However, despite the instability, over time the private
information is ignored and decision makers become increasingly likely to
imitate their predecessors. Consequently, we observe longer and longer periods
of uniform behavior, punctuated by increasingly rare switches. These results
suggest that the kind of episodic instability that is characteristic of social
behavior in the real world makes more sense in the imperfect-information model,
and that the imperfect information premise provides a better theoretical description
of fads and fashions.
-
Distinguishing Informational Cascades from Herd
Behavior in the Laboratory, with Boğaçhan
Çelen,
Abstract. This paper reports an experimental test of how individuals learn
from the behavior of others. By using techniques only available in the
laboratory, we elicit subjects' beliefs. This allows us to distinguish informational
cascades (convergence of beliefs) from herd behavior (convergence of actions).
By adding a setup with continuous signal and discrete action, we enrich the
ball-and-urn observational learning experiments paradigm of Anderson and Holt
(1997). We test a model that explains subjects' behavior as a form of
generalized Bayesian behavior that incorporates limits on the rationality of
others. We find strong evidence that, in Bayesian terms, subjects put too much
weight on their own information and too little weight on the public
information. Put differently, subjects are overconfident in the precision of
their private information. To put the observed behavior into perspective,
we use a simple modification of the Bayesian model, which provides a framework
that enables us to understand individual behavior in the laboratory.
-
An Experimental Test of Observational Learning
under Imperfect Information, with Boğaçhan
Çelen,
Abstract. To explore the difference between social learning
under perfect and imperfect information, this paper takes an experimental look
at a situation in which individuals learn by observing the behavior of their
immediate predecessors. Our experimental design is based on the theory of
Çelen and Kariv (Observational Learning Under Imperfect Information) and
uses the procedures of Çelen and Kariv (Distinguishing Informational
Cascades from herd Behavior in the Laboratory) with the exception that the
history of actions observed by subjects is different. We find is that imitation
is much less frequent when subjects have imperfect information, even less
frequent than the theory predicts. Further, while we find strong evidence that
under perfect information a form of generalized Bayesian behavior adequately
explains behavior in the laboratory, under imperfect information behavior is
not even consistent with this generalization of Bayesian behavior. To reconcile
this with the conclusions under perfect information, we undertake a
modification of the model that abandons the assumption of common knowledge of
rationality.
-
Behavioral Aspects of Learning in Social
Networks: An Experimental Study, with Syngjoo Choi,
NYU, and Douglas Gale, NYU. Advances
in Applied Microeconomics,
Volume 13, Behavioral and Experimental Economics, 2005, edited by
Abstract. Networks are natural tools for
understanding social and economic phenomena. For example, all markets are
characterized by agents connected by complex, multilateral information
networks, and the network structure influences economic outcomes. In an earlier
study, we undertook an experimental investigation of learning in various
three-person networks, each of which gives rise to its own learning patterns.
In the laboratory, learning in networks is challenging and the difficulty of
solving the decision problem is sometimes massive even in the case of three
persons. We found that the theory can account surprisingly well for the
behavior observed in the laboratory. The aim of the present paper is to
investigate important and interesting questions about individual and group
behavior, including comparisons across networks and information treatments. We
find that in order to explain subjects' behavior, it is necessary to take into
account the details of the network architecture as well as the information
structure. We also identify some black spots where the theory does least well
in interpreting the data.
-
Financial Networks, with Douglas Gale,
NYU. American Economic
Review, Papers & Proceedings, May 2007,
97(2), pp. 99-103.
Abstract. Apart from centralized exchanges such as
the NYSE, most financial transactions take place in networks where one or more
intermediaries link the initial seller and final buyer. This paper
presents a model of financial networks, in which financial exchange is
intermediated by traders who form a chain of links between the initial owner of
the assets and ultimate owner of the assets. Networks are incomplete in the
sense that each trader can only exchange assets with a limited number of other
traders. The greater the incompleteness of the network, the more intermediation
is required to transfer the assets between initial and final owners.
Intermediation takes time and time is costly, so incompleteness constitutes a
potentially important market imperfection. The cost and uncertainty of trade in
networks may give rise to other problems and, in extreme cases, lead to a
market breakdown. The results are applicable not just to financial networks but
to any model of exchange which shares the same basic network structure.
-
Revealing Preferences Graphically: An Old
Method Gets a New Tool Kit, with Syngjoo Choi, UCL, Ray Fisman, Columbia
B-School, and Douglas Gale,
NYU. American Economic Review, Papers & Proceedings, May
2007, 97(2), pp.153-158.
Abstract. This paper describes the necessary tools,
both methodological and analytical, for providing a comprehensive
individual-level analysis of decision-making under risk. Two distinctive
features of the paper are the new experimental technique, and the application
of the tools of the theory of consumer demand to individual decision-making in
the laboratory. To characterize an individual's
decision-making under risk, it is necessary to generate many observations per
subject over a wide range of choice sets. An innovative graphical interface was
developed for this purpose, where subjects see on a computer screen a
geometrical representation of a portfolio choice problem. Subjects choose
portfolios through a simple point-and-click. This intuitive and user-friendly
interface allows for the quick and efficient elicitation of many decisions per
subject under a wide range of choice scenarios. The experimental platform and
analytical techniques that have been developed can also be applied to many
types of individual choice problems.
-
Individual Preferences for Giving, with Ray
Fisman, Columbia B-School, and Daniel
Markovits, Yale Law School. American Economic Review, May 2007, 97(2), pp. 153-158.
(Previously distributed in three different papers titled Individual
Preferences for Giving, Pareto Damaging Behaviors
and Distinguishing Social Preferences from Preferences
for Altruism.) [Appendix I][Appendix II][Appendix III][Appendix IV][Appendix V][Appendix VI]
Abstract. We utilize graphical
representations of Dictator Games which generate rich individual-level data.
Our baseline experiment employs budget sets over feasible payoff-pairs. We test
these data for consistency with utility maximization, and we recover the
underlying preferences for giving (tradeoffs between own payoffs and the
payoffs of others). Two further experiments augment the analysis. An extensive
elaboration employs three-person budget sets to distinguish preferences for
giving from social preferences (tradeoffs between the payoffs of others). And
an intensive elaboration employs step-shaped sets to distinguish between
behaviors that are compatible with well-behaved preferences and those that are
compatible only with not well-behaved cases.
-
Consistency
and Heterogeneity of Individual Behavior under Uncertainty,
with Syngjoo
Choi, UCL, Douglas Gale,
NYU, and Ray Fisman,
Columbia B-School). American Economic Review, December 2007, 97(5), pp. 1921-1938. (Some of the
results reported here are also distributed in Substantive and
Procedural Rationality in Decisions under Uncertainty.) [Appendix I] [Appendix II] [Appendix III]
[Appendix IV] [Appendix
V] [Appendix VI] [Appendix
VII] [Appendix VIII]
Abstract. By using graphical representations
of simple portfolio choice problems, we generate a very rich data set to study
behavior under uncertainty at the level of the individual subject. We test the
data for consistency with the maximization hypothesis, and we estimate
preferences using a two-parameter utility function based on Faruk
Gul (1991). This specification provides a good
interpretation of the data at the individual level and can account for the
highly heterogeneous behaviors observed in the laboratory. The parameter
estimates jointly describe attitudes toward risk and allow us to characterize
the distribution of risk preferences in the population.
-
Sequential Equilibrium in Monotone Games:
Theory-Based Analysis of Experimental Data, with Syngjoo Choi,
UCL, and Douglas Gale, NYU.
Version: February 28, 2008. Forthcoming, Journal
of Economic Theory. [Appendix]
Abstract. A monotone game is an extensive-form
game with complete information, simultaneous moves and an irreversibility
structure on strategies. It captures a variety of situations in which players
make partial commitments and allows us to characterize conditions under which
equilibria result in socially desirable outcomes. However, since the game has
many equilibrium outcomes, the theory lacks predictive power. To produce
stronger predictions, one can restrict attention to the set of sequential
equilibria, or Markov equilibria, or symmetric equilibria, or pure-strategy
equilibria. This paper explores the relationship between equilibrium behavior
in a class of monotone games, namely voluntary contribution games, and the
behavior of human subjects in an experimental setting. Several key features of
the symmetric Markov perfect equilibrium (SMPE) are consistent with the data.
To judge how well the SMPE fits the data, we estimate a model of Quantal
Response Equilibrium (QRE) (McKelvey and Palfrey 1995, 1998) and find that the
decision rules of the QRE model are qualitatively very similar to the empirical
choice probabilities.
WORKING PAPERS
-
Substantive and Procedural Rationality in
Decisions under Uncertainty, with Syngjoo Choi, UCL, Douglas Gale, NYU, and Ray Fisman, Columbia
B-School. Version:
Abstract. We report a laboratory experiment
that enables us to study systematically the substantive and procedural
rationality of decision making under uncertainty. By using novel graphical
representations of budget sets over bundles of state-contingent commodities, we
generate a very rich data set well-suited to studying behavior at the level of
the individual subject. We test the data for consistency with the maximization
hypothesis, and we recover underlying preferences using both nonparametric and
parametric methods. We find considerable heterogeneity in individual behaviors
across subjects. In spite of this heterogeneity, we identify prototypical
heuristics that inform subjects' decision rules. To account for these
heuristics, we propose a type-mixture model based on Expected Utility Theory
employing only combinations of three heuristics which correspond to the
behavior of individuals who are infinitely risk averse, risk neutral, and
expected utility maximizers with intermediate risk aversion. This links the
procedural rationality that is evident in the data to substantive rationality,
and supports the use of Expected Utility Theory for both normative and
descriptive purposes.
-
An Experimental Test of Advice and Social
Learning, with Boğaçhan
Çelen, Columbia B-School, and Andrew Schotter, NYU.
Version: November 5, 2007.
Abstract. Social learning is the process of individuals learning by observing
the actions of others. In the real world, however, although people learn by
observing the actions of others, they also learn from advice. This paper
introduces advice giving into a standard social-learning problem. The
experiment is designed so that both pieces of information - actions and advice
- are equally informative (in fact, identical) in equilibrium. Despite the
informational equivalence of advice and actions, in the laboratory, subjects
are more willing to follow the advice given to them by their predecessors than
to copy their actions. In addition, when advice is given subject behavior is
more consistent with the prediction of the theory. Consequently, advice is both
more informative and welfare improving.
-
Trading in Networks: A Normal Form Game
Experiment, with Douglas Gale,
NYU. Version:
Abstract. This paper reports an experimental study of trading networks, in
which exchange is intermediated by traders who form a chain of links between
the initial owner of the assets and ultimate owner of the assets. Traders
choose bid and ask prices and trades are executed by the computer once subjects
have submitted their strategies. Networks are incomplete in the sense that each
trader can only exchange assets with a limited number of other traders. The
greater the incompleteness of the network, the more intermediation is required
to transfer the assets between initial and final owners. The uncertainty of
trade in networks constitutes a potentially important market imperfection. As a
result, the inferences subjects must draw in order to make optimal decisions
are quite subtle. Nevertheless, we find that the competitive prices can account
for the pricing behavior observed in the laboratory in variety of networks and
trading protocols. Furthermore, significant differences can be identified in
the pricing behavior of subjects in different networks, and different trading
protocols lead to different dynamics. .
-
Estimating Ambiguity Aversion in a Portfolio
Choice Experiment, with David Ahn,
Berkeley, Syngjoo Choi,
UCL, and Douglas Gale, NYU.
Version: December 17, 2007. [Appendix I] [Appendix II] [Appendix III]
[Appendix IV] [Appendix V]
Abstract. We report a laboratory experiment
that enables us to estimate four prominent models of ambiguity aversion --
Subjective Expected Utility (SEU), Maxmin Expected Utility (MEU), Recursive
Expected Utility (REU), and α-Maxmin Expected Utility (α-MEU) -- at
the level of the individual subject. We employ graphical representations of
three-dimensional budget sets over bundles of Arrow securities, one of which
promises a unit payoff with a known probability and two with unknown
(ambiguous) probabilities. The sample exhibits considerable heterogeneity in
preferences, as captured through parameter estimates. Nonetheless, there exists
a strong tendency to equate the demands for the securities that pay off in the
ambiguous states. This feature is more easily accommodated by the α-MEU
model than by the REU model.
-
Social
Learning in Networks: A Quantal Response Equilibrium Analysis of Experimental
Data, with Syngjoo Choi, UCL, and Douglas Gale, NYU. Version: May
14, 2008. [Appendix I] [Appendix
II] [Appendix III]
Abstract. Social learning describes any
situation in which economic agents learn by observing the behavior of others.
Social learning in networks applies this idea to situations in which agents
observe the other agents to whom they are connected in a social network.
Whether agents can rationally process the information available in a network is
ultimately an empirical question. This paper reports an experimental
investigation of learning in networks and uses the theoretical framework of
Gale and Kariv (2003) to interpret the data generated by the experiments. We
find that the theory can account for the behavior observed in the laboratory in
variety of networks and informational settings. To explicitly allow for the
possibility of errors in our theoretical model, we adapt the model of Quantal
Response Equilibrium (QRE) of McKelvey and Palfrey (1995, 1998) and find that
its restrictions are confirmed.
-
Network
Architecture, Salience and Coordination, with Syngjoo Choi, UCL, Douglas Gale, NYU, and Thomas Palfrey, Caltech.
Version: February 25, 2008. [Appendix I] [Appendix II]
Abstract. This paper reports the results of
an experimental investigation of monotone games with imperfect information.
Players are located at the nodes of a network and observe the actions of other
players only if they are connected by the network. These games have many
sequential equilibria; nonetheless, the behavior of subjects in the laboratory
is predictable. The network architecture makes some strategies salient and this
in turn makes the subjects' behavior predictable and facilitates coordination
on efficient outcomes. In some cases, modal behavior corresponds to equilibrium
strategies.
-
Overconfidence and Informational Cascades.
Version: March 21, 2005 (under revision).
Abstract. This paper combines behavioral economics and social learning. We
test how robust the theory is to the well-known behavioral phenomenon of
individual overconfidence (mistaken private information perception,
specifically overvaluing). In the context of social learning, overconfident
agents overweigh their private information relative to the public information
revealed by the decisions of others. Therefore, when following a herd, they
broadest more of the information available to them. However, overconfidence
trades the additional information revealed by overconfident decisions against
more information that is being suppressed by perfectly rational decisions.
Thus, the presence of overconfident individuals intensifies the free-rider
problem of rational individuals. With the help of numerical simulations, we
show that, from a social perspective, the presence of overconfident agents
cannot improve decisions accuracy or break the poor information flow intrinsic
to erroneous uniform behavior.