Effective call admission control (CAC) methods are needed to provide bandwidth on demand and meet quality of service requirements. Current methods are often excessively strict, allowing available bandwidth to go unused, and depend on accurate estimation of traffic description parameters. We formulate and test a simple method for estimating cell loss rate from discretized queue size measurements and using the estimate in admission decisions. Simulation test results show that our estimation method will improve performance of current methods by correcting for incorrect traffic parameter estimations.