Exchange Rates, Information, and Crises
- Author(s): Fernholz, Ricardo Turrin
- Advisor(s): Gourinchas, Pierre-Olivier
- et al.
In this dissertation, I theoretically investigate how the actions of central banks affect the information and beliefs of rational agents. I focus primarily on my models' equilibrium predictions during crisis episodes, including situations in which agents' actions are strategic complements. The first two parts explore the implications of central bank transparency during foreign exchange interventions and develop dynamic models in which investors are heterogeneously informed about both interventions and fundamentals. In the first part, the benchmark two-period model presents the main result that transparency can often exacerbate any misalignment between the exchange rate and fundamentals. This is a consequence of two distinct effects of transparency. First, transparency reveals some information about fundamentals to investors (the truth-telling effect). Second, transparency increases the precision of the exchange rate as a signal of those fundamentals that remain unknown (the signal-precision effect). If a central bank announcement reveals little information about fundamentals, then this second effect dominates and transparency magnifies exchange rate misalignment. In effect, partial information revelation is worse than no information revelation. An important implication of this result is that a policy of ambiguity can increase the effectiveness of intervention to support a declining currency during times of crisis. In the second part, the benchmark model is extended to an infinite horizon and also expanded into a Bayesian signalling game. In both cases, I demonstrate that the principal results do not change.
In the third part, I examine a global coordination game in which the information of the agents is manipulated before it reaches them. I assume that the regime is imperfectly informed about the underlying state of fundamentals and must trade off the cost of biasing the signals of agents upwards with the benefit of being more likely to defeat the agents' attack. The main conclusion is that the effect of information manipulation depends on the extent to which the regime is better informed about the outcome of the game. In the limit as both the regime and the agents' information becomes arbitrarily precise, the effectiveness of information manipulation depends on whether the regime learns about fundamentals faster than the agents. If agents learn faster than the regime, then information manipulation is ineffective and incurs costs for the regime without diminishing the size of coordinated attacks. In these cases, the regime prefers to take no action but cannot credibly commit to do so.