MARC 主機 00000nam 2200349 4500
001 AAI3402228
005 20110110132610.5
008 110110s2010 ||||||||||||||||| ||eng d
020 9781109742282
035 (UMI)AAI3402228
040 UMI|cUMI
100 1 Muller-Frank, Manuel
245 10 Essays on learning in networks
300 92 p
500 Source: Dissertation Abstracts International, Volume: 71-
05, Section: A, page: 1746
500 Advisers: Alessandro Pavan; Marciano Siniscalchi
502 Thesis (Ph.D.)--Northwestern University, 2010
520 This dissertation makes several contributions to the
theory on learning in networks. Chapter 1 provides a
formal characterization of the process of rational
learning in social networks. A finite set of agents select
an option out of a choice set under uncertainty in
infinitely many periods observing the history of choices
of their neighbors. Choices are made based on a common
behavioral rule. I find that if learning ends in finite
time and the choice correspondence is union consistent,
then every action selected by any agent once learning ends
is optimal for all his neighbors. I further provide
sufficient conditions such that every action chosen
infinitely often by an agent is optimal for all his
neighbors in the limit. If only common knowledge of
rationality rather than common knowledge of strategies is
assumed, the validity of the aforementioned results
depends on the network structure. If the network is
complete, the result of local indifference across
neighbors once learning ends still holds, while it can
fail in incomplete networks
520 Chapter 2 considers aggregation of private information in
networks where Bayesian agents announce their posterior
belief of an uncertain event to their neighbors in each of
countable communication rounds. I show by example that
complete networks can be inferior to incomplete networks
in terms of the quality of information aggregation under
certain circumstances. I then characterize sufficient
conditions on the informational structure for optimality
of complete networks
520 Chapter 3 establishes a fixed point convergence result in
Euclidean spaces. The result is used to extend existing
results in the literature on Non-Bayesian learning in
networks. In a Non-Bayesian learning framework, agents
announce their belief of an uncertain event to their
neighbors in each of countable communication rounds using
updating rules: the posterior they announce in a given
round is a function of the last period announcements of
their neighbors and themselves. I show that if the
updating rules are continuous and contracting, then the
beliefs of all agents in a connected social network
converge
590 School code: 0163
650 4 Economics, Theory
690 0511
710 2 Northwestern University.|bEconomics
773 0 |tDissertation Abstracts International|g71-05A
856 40 |uhttp://pqdd.sinica.edu.tw/twdaoapp/servlet/
advanced?query=3402228