This paper proposes a new method to estimate the extent of job competition between workers with different schooling levels. We estimate the structural parameters of a matching function generalised to incorporate crowding out effects. We use flow data out of unemployment containing information on the level of educational attainment of the worker, but not on the level of schooling required by the employer for the job. The method therefore avoids the bias induced by mismeasurement in the educational requirements. Applied to Belgian data, we find evidence of significant crowding out among dismissed workers, particularly at the highest schooling levels.