Information and learning in foreign exchange markets

by D'Souza, Chris

Equilibriurn modek of the foreign exchange market cannot account for, or predict, most short-term movements in the nominal exchange rate. A related puzzle in international hance is the Large trading volumes in foreign exchange markets that are not accounted for by these models. Spdative trading will only eiost if agents in these markets have ciifferences in information or difEerences in expectations about movements in the exchange rate. Even then, pdectIy rational traders may stül be able to extract all privste information fkom equilibrium excharge rates. a result that would eliminate a.U the gains fiom trade. In this thesis, prices, returnç and trading volumes are fkst euamined in asset markets when agents are assumed to have expectations that are not Mly- rational. A dynamic mode1 oi speculative trading is presented in which agents have private, but imperfect, information about the fundamentals of the economy. In the bounded rationd environment, agents continue to leam about the tme probability distribution of the model. The analysis focuses on empincal regulazïties of the foreign e~change mmket. In contrast to the no-trade implication of the rational expectations equilibrium, positive trading persists in the learning environment of the model. Results suggest that rational e,upectationsand bounded rational mvironments have very simiiar timeseries properties. Both dispIay fat tails in returns, but neither envkonment provides evidence of non-stationary prices or volatility clustering in retunis. Lastly, while rational expectations prices follow a similar pattern as the underlying fundamentab of the economy, bounded rational prices do not. Next, a monetary model of the exchangerate is presented that, rinlikemost models in this class, provides an explicit role for trading in foreign exchange markets. Ut;ility rna;i8miPng agents choose an amount of wealth to be invested in foreign assets that is determined partly by private information about domestic monetary policy and partly by rïsk and rem factors. The dynamics of money growth rates across countries, in addition to clifferences in information about these growth rates, determine exchange rate dynamics in the rational expectations equilibrium of the model. Implications of the model include nominal exchange rates with unit roots, volatility clustering in exchange rate returns, and the existence of positive trading volumes. Ehther, a calibration exercise IIIustrates that a reasonable correspondence exists between the time series properties of the modd and the histoncal saxnple. The performance of new monetary mode1 of the exchange rate is then demonstrated. The Kaiman ater is empIoyed to estimate the forcingprocesses of the model in th& state space form. Prdirninary and final money stock data for Canada and the US. are decomposed into the various unobserved components of interest using dynamic factor analysis- These series, together with the monetary stock data, are then employed in the new monetary model of the exchange race to simulate a s- of nominal e~change rates. The sarnpIe path forecasteci by the mode1 is compd to the realized sampie path of the Canadian-US. nominal =change rate. Th & to all my fiends and fdy who provideci support during my Ph.D. studies. 1am greatly indebted to Gregor Smith for the central role he played in the guidance and supervision of my thesis. 1 wodd &O lüre to thank Charles Beach, Bevdy Lapham and Dan Bernhardt for helpfd sugggtions. Finally, financial mpport fiom the Queen's School of Graduate Studies and the Ontario Graduate SchoIarship pr* gram is gratefully acknowledged.