Modeling the Effect of Central Bank Communication on Foreign Exchange Volatility

Abstract

In Foreign Exchange Options trading desks, traders base their trades on predictions of volatility rather than directional spot movement, due to the dynamics of their delta-hedged positions. Because investing in an exchange rate involves borrowing one currency to buy another, movements in the Foreign Exchange market is based largely on movements in the interest rates of each nation. Since the financial crisis, interest rates have been carefully controlled by the central banks; thus, the Foreign Exchange market has been closely tied to the perceived sentiment of the central banks. Currently, by far the most common Foreign Exchange volatility model is the GARCH(1, 1) model. This study attempts first to characterize the speeches given by the Federal Reserve, the Bank of Canada, and the Reserve Bank of Australia since the crisis using the Latent Dirichlet Allocation (LDA) topic model. It then explores a potential improvement upon GARCH(1, 1) on days with Central Bank speeches, using Support Vector Regression, taking both the GARCH(1, 1) prediction and the LDA topic distributions as features, to model the volatilities of USD/JPY, USD/CAD, and AUD/USD. The results of this study suggest that, while it is important to take into account whether or not the central bank is giving a speech, the contents of the speech do not substantially impact FX volatility.