« Outlines of a Social Security Deal? | Main | Praise of Steve Clemons »

January 31, 2005

Brunnermeier and Morgan on Clock Games

This week's Wednesday afternoon seminar looks very good:

Clock Games: Theory and Experiments

Markus Brunnermeier (Princeton) and John Morgan (Haas)

December 4, 2004

Abstract: Timing is crucial in situations ranging from currency attacks, to product introductions, to starting a revolution. These settings share the feature that payoffs depend critically on the timing of a few other key players--and that their moves are uncertain. To capture this, we introduce the notion of clock games and experimentally test them. Each player's clock starts on receiving a signal about a payoff-relevant state variable. Since the timing of signals is random, clocks are de-synchronized. A player must decide how long, if at all, to delay his move after receiving the signal. We show (i) equilibrium is always characterized by strategic delay--regardless of whether moves are observable; (ii) delay decreases as clocks become more synchronized and increases as information becomes more concentrated; (iii) when moves are observable, players "herd" immediately after any player makes a move. We then show, in a series of experiments, that key predictions of the model are consistent with observed behavior.

Posted by DeLong at January 31, 2005 01:52 PM


When and where? Secret handshake? It's not listed on the seminar list this week.

Posted by: ogmb at January 31, 2005 04:29 PM

Turns out Brad's clock is de-synchronized:

ECON 281: Wednesday*
February 9, 4-5:30 p.m., 608-7 Evans
John Morgan, UC Berkeley
"Clock Games: Theory and Experiments"
(joint with Department Seminar)


Posted by: ogmb at January 31, 2005 04:34 PM

For anybody not clued in, this is about when everybody is going to abandon the dollar. The post is saying that experiments in similar situations show that herding behavior is built into the psychology of the group.

Posted by: walter willis at January 31, 2005 04:56 PM

Brunnermeier & Morgan's paper on clock games:


Posted by: abc at January 31, 2005 07:15 PM

"The post is saying that experiments in similar situations show that herding behavior is built into the psychology of the group."

On reading the paper the problem seems to be that herding behavior is built into the game setup. In most of the potential applications they offer the phase transition is endogenous (ie only occurs if enough people herd), but in their setup the transition is a foregone conclusion and the only question remains when to jump ship -- with the optimal choice being the last one to get in the life boat. This differs markedly from their applications where optimal choice is to get in early provided the revolution succeeds, and not get in at all if it is bound to fail.

Still interesting paper. I hope I can stop by.

Posted by: ogmb at February 1, 2005 01:10 PM

"herding behaviour is built into the game setup" is a little harsh... it's true only in the way that a conclusion is already built into a set of premises.

"Herding behavior turns out to be the logical response given this sort of game setup" is probably a little bit fairer. If you don't know when everyone else is gonna act, then you turn out to be best off waiting and then making a jump quickly when you see someone else start to move. But everyone else is in the same boat, so you all tend to jump at once.


Posted by: meno at February 1, 2005 01:53 PM