- Nageeb Ali (Penn State).
"Laws, Norms, and Authority"
- Itai Arieli (Technion).
"Private Bayesian Persuasion" with Yakov Babichenko
We consider a multi-agent Bayesian
persuasion problem where an in- formed sender tries to persuade a
group of agents to adopt a certain product. The sender is allowed to
commit to a signaling policy where she sends a private signal to every
agent. The payoff to the sender is a function of the subset of
adopters. We characterize an optimal signaling policy and the maximal
revenue to the sender for three different types of payoff functions:
supermodular, symmetric submodular, and a supermajority function.
Moreover, using tools from cooperative game theory we provide a
necessary and sufficient condition under which public signaling policy
- Kostas Bimpikis (Stanford).
"Cournot Competition in Networked Markets"
The paper considers a model of competition among
firms that produce a homogeneous good in a networked
environment. A bipartite graph determines which
subset of markets a firm can supply to. Firms
compete in Cournot and decide how to allocate their
production output to the markets they are directly
connected to. We provide a characterization of the
production quantities at the unique equilibrium of
the resulting game for any given network. Our
results identify a novel connection between the
equilibrium outcome and supply paths in the
underlying network structure. We then proceed to
study the impact of changes in the competition
structure, e.g., due to a firm expanding into a new
market or two firms merging, on firms' profits and
consumer welfare. The modeling framework we propose
can be used in assessing whether expanding in a new
market is profitable for a firm, identifying
opportunities for collaboration, e.g., a merger,
joint venture, or acquisition, between competing
firms, and guiding regulatory action in the context
of market design and antitrust analysis.
- Yang Cai (McGill).
Framework for Designing Simple Mechanisms" with
Nikhil R. Devanur, Matt Weinberg and Mingfei
A central theme in mechanism design is understanding
the tradeoff between simplicity and optimality of
the designed mechanism. An important and challenging
task here is to design simple multi-item mechanisms
that can approximately optimize the revenue, as the
revenue-optimal mechanisms are known to be extremely
complex and thus hard to implement. Recently, we
have witnessed several breakthroughs on this front
obtaining simple and approximately optimal
mechanisms when the buyers have unit-demand (Chawla
et. al. '10) or additive (Yao '15)
valuations. Although these two settings are highly
similar, the techniques employed in these results
are completely different and difficult to extend to
more general settings.
In this talk, I will present a principled approach to design simple and approximately optimal mechanisms based on duality theory. Our approach unifies and improves both of the aforementioned results, and extends these simple mechanisms to broader settings.
- Arun Chandrasekhar
Networks, Reputation and Commitment: Evidence from a
Savings Monitors Experiment" with Emily Breza
We study whether individuals save more when
information about their savings is shared with another
village member (a "monitor"). We focus on whether the
monitor's effectiveness depends on her network
position. Central monitors may be better able to
disseminate information, and more proximate monitors
may pass information to individuals who interact with
the saver frequently. In 30 villages, we randomly
assign monitors. Average monitors increase savings by
35%. A one-standard deviation more central monitor
increases savings by 14%; increasing proximity from
social distance three to two increases savings by
16%. The increased savings persist over a year after
the intervention's end, and monitored savers better
respond to shocks. In 30 other villages, savers choose
their monitors. Proximate and central monitors are
Information flows. 63% of monitors tell
others about the saver's progress. 15 months later,
others know more about the saver's progress and
believe she is responsible if the saver was assigned a
more central monitor.
- Laura Doval (Northwestern).
- Matthew Elliott (Caltech and
"Endogenous Financial Networks: Efficient Modularity
and Why Shareholders Prevent It" with Jonathon Hazell
We consider systemic risk in financial networks,
by examining the conflict of interest between debt- and
equity-holders. Through trading, banks can diversify their
idiosyncratic risks and avoid failures following small
shocks. However, the resulting interdependencies can cause
multiple failures after large shocks. A social planner
resolves this trade-off by creating a modular network
structure with fire breaks, thereby preventing failures
from small shocks while containing contagion. Socially
efficient networks favor debt-holders over equity holders,
meaning equity-holders can profitably trade away from
these networks. Moreover, profitable trades for equity
holders align their counter-parties' failures with their
own, creating systemic risk.
- Maryam Farboodi (Princeton).
"Meeting Technologies in Decentralized Asset
Markets" with Gregor Jarosch and Robert Shimer
We study decentralized trading networks where agents
differ in both their time-varying taste for an asset
and the constant frequency at which they meet
others. We demonstrate that fast agents endogenously
arise as intermediators whose net valuation of the
asset gets moderated through their exposure to
others. We show that allocating meetings in an
ex-ante asymmetric fashion across agents generates
higher welfare then a homogeneous distribution of
meeting frequencies, only if some agents
intermediate. We also characterize properties of the
market equilibrium in which ex-ante identical agents
choose their meeting rates, and show that an
equilibrium with symmetric meeting rates does not
exist. Finally we compare the properties of
equilibrium outcome with the planner allocation.
- David Hirshleifer (UC Irvine).
in Networks: It's a Matter of Degree" with Ben
Individuals' susceptibility to social
influence varies with their social network
positions. We show that the nature of this
relationship is critical to how population
traits evolve in the aggregate, and who has a
large weight in determining that evolution. When
per-neighbor susceptibility is increasing
(decreasing) in degree, low (high) degree nodes
have excess weight in moving the population
average. We show this in a social network model
that generalizes a standard model of trait
updating to allow susceptibility to depend upon
own degree, and show how this affects both the
short- and long-run population trait
evolution. We apply the model to the
transmission of sociability and promiscuity. The
model provides insight into why the friendship
paradox and majority illusion do not necessarily
result in disproportionate influence of the
popular or highly connected.
- Ben Golub (Harvard).
- Matthew Jackson (Stanford).
"The Friendship Paradox and Systematic Biases in
Perceptions and Social Norms"
The ``friendship paradox'' (Feld 1991) is the structural implication of networks that, on average, peoples' friends have strictly more friends than the average person in a network. In particular, the number of people who observe a given person is proportional to the number of connections that the person has. This can distort perceptions of norms and behavior if more popular people in a society behave differently from less popular people. As I show here there are two things that drive people with more friends to behave differently from people with fewer friends. The first is that in any setting with strategic complementarities, people with more friends are exposed to greater interaction and influence. The second is that people who benefit more from a given activity will tend to form more relationships as they benefit more from the complementarities. These two effects lead people with more friends to choose more extreme actions, which in turn feeds back via the the friendship paradox to increase overall perceptions of behavior and then via complementarities to amplify average behavior. These theoretical results are consistent with the multitude of studies finding that students (from middle school through university) consistently overestimate peer consumption of alcohol, cigarettes, and drugs. This in turn amplifies students' own behaviors, and can help explain problems with adolescent abuse of drugs and alcohol, as well as other behaviors. I also discuss how these results change in cases of strategic substitutes, where individuals overestimate free-riding by peers.
- Markus Mobius (MSR).
- Pooya Molavi (MIT). "Foundations
of non-Bayesian Social Learning" with Alireza
Tahbaz-Salehi and Ali Jadbabaie
In this paper, we study the problem of
boundedly rational learning in networks by taking an
axiomatic approach. As our main behavioral assumption,
we postulate that agents follow social learning rules
that satisfy imperfect recall, according to which they
treat the current beliefs of their neighbors as
sufficient statistics for all the information
available to them. We establish that as long as
imperfect recall represents the only point of
departure from Bayesian rationality, agents' social
learning rules take a log-linear form. Our axiomatic
approach enables us to provide a taxonomy of
behavioral assumptions that underpin various boundedly
rational models of learning, including the canonical
model of DeGroot. We then provide a complete
characterization of the asymptotic outcomes of social
learning in our environment. We introduce the notion
of group nonpolarization and show that it represents
the necessary and sufficient condition for information
aggregation. Our results on the asymptotic outcomes of
social learning extend to a fairly large class of
learning rules that admits as special cases both the
log-linear and DeGroot Models. Finally, we show how
the dispersion of information among different
individuals in the social network determines the rate
- Manuel Mueller-Frank (IESE).
"A General Model of Boundedly
Rational Observational Learning:
Theory and Evidence" with Claudia
This paper introduces a model of boundedly rational observational learning, which is rationally founded and applicable to general environments. Under Quasi-Bayesian updating each action is treated as if it were based only on the private information of its respective observed agent. We establish the long run learning properties of Quasi-Bayesian updating in a model of repeated interaction in social networks. Evidence from a laboratory experiment supports
Quasi-Bayesian updating and our theoretical predictions.
- Stephen Morris (Princeton).
"Expectations, Networks, and Conventions"
with Ben Golub
- Mallesh Pai (Rice University).
"Evaluating Forecasters" with Rahul Deb and Maher Said
- Evan Sadler (Harvard).
I study how the structure of a social network
affects the diffusion of a new product or
technology. A population of players interacts in a
random network. Initially, few are aware of the
opportunity to adopt a product, but a player
becomes aware if a neighbor adopts. Payoffs may
depend on neighbors' choices. To make this
analysis tractable, I extend recent mathematical
work on random graphs. A branching process
approximates the local structure of the network,
allowing us to characterize conditions under which
diffusion results in a large cascade, the size of
the cascade, and the rate of spread. This
characterization facilitates comparative statics
results for changes in the degree distribution,
information about neighbors' valuations,
clustering, and homophily. I apply the framework
to study Bertrand competition with word-of-mouth
communication and referral marketing.
- Vasilis Syrgkanis (Microsoft
"Bayesian Exploration" with Yishay Mansour,
Aleksandrs Slivkins and Zhiwei Steven Wu
We consider a ubiquitous scenario in the Internet economy when individual decision-makers (henceforth, agents) both produce and consume information as they make strategic choices in an uncertain environment. This creates a three-way tradeoff between exploration (trying out insufficiently explored alternatives to help others in the future), exploitation (making optimal decisions given the information discovered by other agents), and incentives of the agents (who are myopically interested in exploitation, while preferring the others to explore). We posit a principal who controls the flow of information from agents that came before to the ones that arrive later, and strives to coordinate the agents towards a socially optimal balance between exploration and exploitation, not using any monetary transfers. The goal is to design a recommendation policy for the principal which respects agents' incentives and minimizes a suitable notion of regret.
We extend prior work in this direction to allow the agents to interact
with one another in a shared environment: at each time
step, multiple agents arrive to play a Bayesian game,
receive recommendations, choose their actions, receive
their payoffs, and then leave the game forever. The
agents now face two sources of uncertainty: the
actions of the other agents and the parameters of the
uncertain game environment.
- Lones Smith (Wisconsin).
"Rushes in Large Timing Games" with Axel Anderson
and Andreas Park
We develop a continuum player timing game that
subsumes standard wars of attrition and pre-emption
games, and introduces a new rushes
phenomenon. Payoffs are continuous and single-peaked
functions of the stopping time and stopping
quantile. We show that if payoffs are hump-shaped in
the quantile, then a sudden "rush" of players stops
in any Nash or subgame perfect equilibrium. Fear
relaxes the first mover advantage in pre-emption
games, asking that the least quantile beat the
average; greed relaxes the last mover advantage in
wars of attrition, asking just that the last
quantile payoff exceed the average. With greed, play
is inefficiently late: an accelerating war of
attrition starting at optimal time, followed by a
rush. With fear, play is inefficiently early: a
slowing pre-emption game, ending at the optimal
time, preceded by a rush. The theory predicts the
length, duration, and intensity of stopping, and the
size and timing of rushes, and offers insights for
many common timing games.
- Philipp Strack (Berkeley).
"The Speed of Social Learning" with Matan Harel, Elchanan
Mossel and Omer Tamuz
We consider Bayesian agents who learn from
exogenously provided private signals, as well as the
actions of others. Our main finding is that
increased interaction between the agents can lower
the speed of learning: when two agents observe each
other, learning is significantly slower than it is
when one only observes the other. This slowdown is
driven by a process in which a consensus on the
wrong action causes the agents to discount new
contrary evidence. We also show that very large
groups of agents do not learn very quickly.
- Adam Szeidl (Central European
"Interfirm Relationships and Business
Performance" with Jing Cai
We organize regular business meetings for randomly selected managers
of young Chinese firms to study the effect of business networks on
firm performance. We randomize 2,800 managers into several groups that
hold monthly meetings for one year, and a ``no-meetings'' control
group. We survey all firms before, after, and one year after the
intervention. We find large, wide-ranging and persistent positive
effects of business meetings. (1) The meetings increase firm revenue
by 7.7 percentage points, and also increase profit, a management
score, employment, and the number of business partners. (2) These
effects persist one year after the intervention. We then exploit
additional randomizations to identify mechanisms. (3) Firms randomized
into groups with larger peer firms exhibit higher growth. (4) The
meetings help diffuse randomly distributed business-relevant
information, but diffusion is lower when group members are
competitors. (5) Managers create more business partnerships with, and
exhibit higher trust towards, those they meet every month than those
they meet at one-time cross group meetings. We discuss policy
implications for business associations in developing countries.
- Ali Tahbaz-Salehi (Columbia).
"Information Aggregation in Heterogeneous Markets" with Yaarit Even and Xavier Vives
- Omer Tamuz (Caltech).
"Anonymous Network Games" with Wade
We study a class of network games in which players are anonymous, in the sense that the game looks the same from each player's point of view. Fixing the local interaction, we compare the equilibria obtained on large finite networks and on infinite networks, and show that despite the fact that the game is locally indistinguishable to the agents, on infinite networks there is generally a richer equilibrium set.
- Leeat Yariv (Caltech).
Matching" with Mariagiovanna Baccara and SangMok
We study a dynamic matching environment where individuals arrive sequentially. There is a tradeoff between waiting for a thicker market, allowing for higher quality matches, and minimizing agents' waiting costs. The optimal mechanism cumulates a stock of incongruent pairs up to a threshold and matches all others in an assortative fashion instantaneously. In decentralized settings, a similar protocol ensues in equilibrium, but expected queues are inefficiently long. We quantify the welfare gain from centralization, which can be substantial, even for low waiting costs. We also evaluate welfare improvements generated by transfer schemes and by matching individuals in fixed time intervals