In this paper, we develop a likelihood approach for quantification of qualitative survey data on expectations and perceptions and we propose a new test for expectation consistency (unbiasedness). Our quatification scheme differs from existing methods primarily by using prior information (perhaps derived from economic theory or well established empirical relations) on the underlying process driving the variable of interest. To investigate the properties of our novel quantification scheme and to analyze the size and power properties of the new expectation consistency test, we perform Monte Carlo simulation studies. Overall, the simulation results are very encouraging and show that efficiency gains from including prior information can be substantial relative to existing quatification schemes. Finally, we provide an empirical illustration. We argue that a nonlinear regime switching model, historically, provides a more adequate characterization of prices changes in the British manufacturing industry relative to a wide range of alternative linear representations. We then show how to use this information to quantify the qualitative survey data on price change expectations and perceptions from the CBI survey in the U.K. manufacturing industry. Our findings support * Comments and suggestions from John Carlson and an anonymous referee are gratefully acknowledged. The notation used follows the "new" standard proposed by Abadir and Magnus (2002). The software developed for this paper is available from the authors upon request.the existence of so-called ex post bias, i.e., that rationally formed forecasts might appear biased when compared to ex post realizations. As a result, tests of expectation consistency by comparing forecasts with realizations become invalid. The new test we propose, however, is not affected by the existence of ex post bias and by applying this statistic we cannot reject the expectation consistency hypothesis.