Overnight exchange rate changes: an econometric approach using long memory processes.

Summary The aim of this thesis is to model the complex dynamics governing the variations of four main daily exchange rates - the pound sterling, the yen, the French franc and the German mark vis-à-vis the dollar - as well as their volatility. The approach adopted consists in considering the statistical properties of exchange rate variations before proposing a statistical model representative of the very recent techniques for formalizing long memory processes. The proposed statistical model makes it possible to evaluate simultaneously the degree of persistence of shocks on the level and the volatility of exchange rate variations. This model is then extended to take into account the role of monetary authorities. Using the statistical model developed, we analyze the effect of the interventions of the main central banks (the Federal Reserve, the Bundesbank and the Bank of Japan) on the variations and volatility of the deutsche mark and the yen, and thus judge the effectiveness of these interventions in the very short term. In terms of economic contribution, the application of long-memory processes has led to relatively enlightening results: future volatility shocks persist at all forecast horizons, even if they converge towards the mean in the long term, thus making it possible to better predict exchange rate volatility. This result is robust to the sampling period as well as to the choice of the distribution, unlike the long memory detected in the exchange rate variations. In terms of economic contribution, the study shows that official central bank interventions are successful in moving the market (especially when reported in the press) but in the wrong direction, in the sense that official purchases of dollars increase the volatility of the exchange rate and generally lead to a depreciation of the dollar.