Authors: Michel van der Wel, Albert J. Menkveld,
and Asani Sarkar

We develop a new likelihood-based approach to signing trades in the absence of quotes. This approach is equally efficient as the existing Markov-chain Monte Carlo methods, but more than ten times faster. It can address the occurrence of multiple trades at the same time and allows for analysis of settings in which trade times are observed with noise. We apply this method to a high-frequency data set of thirty-year U.S. Treasury futures to investigate the role of the market maker. Most theory characterizes the market maker as an uninformed, passive supplier of liquidity. Our findings suggest, however, that some market makers actively demand liquidity for a substantial part of the day and that they are informed speculators.