This paper reevaluates the quantitative performance of the standard labor-market matching model developed by Mortensen and Pissarides with special attention to the behavior of vacancies, one of the key variables in the model. I first estimate trivariate vector autoregressions with gross worker flows and vacancies and identify an aggregate shock by imposing only minimal sign restrictions on the responses of worker flows and employment growth and no restrictions on the response of vacancies. The data strongly suggest a hump-shaped and persistent response of vacancies. The calibrated model, on the other hand, predicts that vacancies respond to aggregate shocks with no delay and are not persistent even though an aggregate productivity shock is assumed to be highly persistent. These problems in vacancy behavior also cause gross flow series to exhibit counterfactual cyclical properties.