High-Frequency Principal Components and Evolution of Liquidity in a Limit Order Market

The paper applies a popular methodology of competing risks to the analysis of the timing and interaction between the Deutsche Mark/U.S. dollar transactions, quotes, and cancellations in the Reuters D2000-2 electronic brokerage system. Consistently with previous stock market studies, the bid-ask spread and market depth at the best bid and ask quotes are found to be major determinants of limit order market dynamics at ultra-high frequencies. Consistently with the microstructure approach to exchange rate determination, the signed transaction activity appears to be the main factor behind the limit order market dynamics at lower frequencies. Application of principal component analysis to the covariate indices of competing risks identifies five pervasive factors that capture 85% of the Reuters D2000-2 limit order book activity. The multifactor competing risks model substantially improves the quality of short-term probability forecasts for buyer- and seller initiated transactions, relative to popular moving average-type forecasting rules