Dealer Activity and Macro Fundamentals: New Evidence from Hybrid Exchange Rate Models

We construct a new class of hybrid exchange rate models, combining macroeconomic fundamentals from a conventional Taylor rule with information from the foreign exchange interdealer market. We provide evidence that hybrid models have a superior model fit and produce more accurate in-sample predictions than their individual nested components. As part of our analysis, we employ a new market microstructure measure, based on submitted and cancelled limit orders, and document its significant impact on monthly exchange rate returns. We show that its effect is transitory and largely diminishes in a stylized out-of-sample forecasting exercise, while market order flow can improve short-term forecasts. Our comprehensive empirical assessment is based on one of the largest foreign exchange interdealer order flow datasets analyzed so far. It comprises 19 U.S. dollar and euro currency pairs and covers a sample period of more than 10 years.