610 likes | 645 Views
Regime change. Ec 517 October 2, 2017. Dynamics of complementarity and learning. Boom and busts in business cycles Switch of political regimes Eastern Europe Arab world Two structural properties
E N D
Regime change Ec 517 October 2, 2017
Dynamics of complementarity and learning • Boom and busts in business cycles • Switch of political regimes • Eastern Europe • Arab world • Two structural properties • Strategic complementarity implies that the marginal agents are in the tail of the distribution • The observation of the tail of the distribution provides little information on the entire distribution
Interactions with externalities and complexity • Boom and busts in business cycles • Financial crises • Switch of political regimes or general opinion • Eastern Europe • Herding behavior
Structural models of interactions • “Canonical” model • Minimalist, abstract • Interactions are caused by externalities • Two types of externalities • Payoff externality (coordination) • Information externality Imperfect observation of the state Learning from others’ actions (social learning, informational cascades, herding)
Coordination • Positive externality: strategic complementarity • The action of others increases my incentive to act • Bank runs • Speculative attacks against a fixed exchange regime • Change of political regime, etc…
Stylized model • Large number of individuals, each chooses, 0-1 action with an individual cost • Gross payoff of action: mass of actions • Net payoff • Distribution of individuals with costs • Each individual knows his own c and has an estimate of the others’ costs • If agents with cost small than c* invests, and F is the cdf, the payoff is F(c*)-c • Under perfect information, Nash equilibrium is F(c*)=c*
Stylized model • Large number of individuals, each chooses, 0-1 action with an individual cost • Gross payoff of action: mass of actions • Net payoff • Distribution of individual costs
Static equilibrium • Strategy • Equilibrium • With sufficient strat. compl., Multiple equilibria
Problem of the “choice” of equilibrium • Global games (Carlsson and Van Damme, Morris & Shin) • (very) Small uncertainty on the structure of the economy • No common knowledge
Evolution (dynamics), learning • Global games (Carlsson and Van Damme, Morris & Shin) • (very) Small uncertainty on the structure of the economy • No common knowledge
Imperfect information • Individuals have a strategy • They observe only • Impact of the shift of the c.d.f • Aggregate activity shows the tail of the distribution • Property inherent to strategic complementarity
Model of regime choice • Kuran: “Public lies private truths”. • Continuum of people, total mass normalized to one. • Y is the mass of people choosing action 0 (speak against) • Mass that supports the regime is 1-Y • Individual is defined by his preference for the regime • Payoff of individual with parameter c : • The difference between the payoffs is
Representation Two (stable) equilibria L1 and H1 The structure (cdf) evolves randomly and slowly
Evolution of the structure of costs • The distribution of costs is “placed” by a parameter that evolves by a random walk (with regression to the mean) Unique equilibrium with low activity Multiple equilibria (under perfect information) Unique equilibrium with high activity T=60
Evolution of beliefs and actions • Individuals form their beliefs from • Their private information (cost c) • The history of aggregate activities (mass of total actions in each past period) • Individuals have a monotone strategy (invest in period t iff the own cost is smaller than ct* (can be proven to be iteratively dominant under the structure of information) • Unique equilibrium
Under imperfect information • On the left, the true state of nature and the public belief (from history) • On the right, the payoff of agent with cost c if every agent (uncertain mass) with cost lower than c chooses to invest
Delays • Assumption: individuals can delay • Delay has a cost • But with delay, more information • Tradeoff • Impact of the cost of delay (equivalent to longer period) • Does a lower cost (shorter period) facilitate coordination ?
Impact of the cost of delay • No under the following argument: • Because of strategic complementarity, the marginal agent is in the tail of the distribution. • Little information • If more agents act, more information that reduces the incentive to act: more incentive to delay • Waiting for information reduces information • Under some plausible assumption, delay prevents an equilibrium with high activity (work with Lucia Esposito).
Social learning • Definition for economists: learning for the observation of what others • Do • Say • Learning about what? • State of nature that affects the payoff of the observed and me (possibly in different ways.
Rational (Bayesian) learning • Requirements • Common knowledge on • The prior distribution of the states of nature • The mechanical properties of the private signals • The payoff functions of the others (how they take action as a function of the information: the common prior and their private signal) • What is not known is the private information (signal) of others. • One’s action is a signal (noisy) on one’s private information
General properties • Because learning is rational (Bayesian), the beliefs (probability distributions about the state) converge (in probability). • Consequence of the Martingale Convergence Theorem • But beliefs may not converge to the truth • More “pathologies” in social learning when individuals are selfishly rational.