Gene expression involves a series of single molecule events and therefore may be subject to large random fluctuations. As these fluctuations are thought to be important in many biological phenomena such as clonal population heterogeneity and incomplete penetrance of transgenes, probabilistic descriptions of the underlying dynamics are often necessary to accurately model biological systems. A number of approaches based on the chemical master equation have been proposed over the past few years for formulating and simulating stochastic models of biological systems. However, little effort has been devoted to the equally important problem of parameter estimation. In this work, we present an algorithm for estimating parameters in stochastic models. While our approach can be generalized to any number of systems, our focus is on genetic networks and the unique challenges associated with estimating parameters for these systems. We have applied our algorithm to two experimental systems: feedback regulation of the tetracycline resistance gene and the phage Lambda genetic switch. We demonstrate that a combination of single cell (e.g. quantitative microscopy and flow cytometry) and population measurements are necessary for precise parameter estimates. We also discuss the application of bootstrap techniques for evaluating confidence intervals on the parameter estimates.