This paper proposes a way to estimate and test market microstructure models. The methodology entails taking advantage of the special structures these models impose and relating those characteristics to a state-space model. I develop, as a benchmark, a theoretical model of speculative trading between a market maker and an insider. In this model, the insider acts strategically with his long-lived private information which evolves randomly not only overnight but also during a trading day. I show that there exists a unique recursive linear sequential auction equilibrium as defined in Kyle (1985). To estimate the deep and shallow parameters of the model, the equilibrium restrictions are represented by a state-space model. Then, Kalman filtering is applied to construct a likelihood function of the observed price series as a function of the deep parameters. Using the depth-weighted quotes, maximum likelihood estimation is performed and the deep parameters are identified. I then test whether the inside information is revealed to the insider mainly overnight or gradually during a trading day. The results from five working days show evidence that the former information structure outperforms the latter. To illustrate how this methodology can be used, I then modify the benchmark model and study the Monday effect. The results are largely in contrast to the widely held belief that adverse selection problems are most conspicuous on Mondays. |