This paper provides a theoretical and empirical analysis of the effect of cost-based reimbursement (CBR) on hospital costs and charges. It takes issue with previous analyses which have treated CBR as paying economic costs plus a mark-up, and have concluded that the mark-up is too small to significantly distort hospital decision-making. The basic thesis here is that if reimbursement is based on costs, accounting costs become a price to cost-paying patients, and will be optimized to maximize revenue. A hospital serving both cost and charge-paying (private) patients can set two price schedules. Accounting profits (ratio of charges to costs) are not a measure of economic profit but of relative prices to these two groups of patients. In the absence of constraints from regulation or patient co-payment, the optimum level of accounting costs would be infinite. In practice, the Medicare reimbursement formula links allowable costs to charges received from charge-paying patients. This formula creates incentives for the hospital to raise charges above the single-price, profit-maximizing monopoly level. This inflationary effect of the Medicare formula does not presuppose that Medicare pays less than full cost. The empirical analysis of hospital laboratory costs and charges generally supports the predictions; for other departments, the conclusions are consistent but more tentative because of data limitations. Overall, evidence suggests minimal cross-subsidy between cost and charge-paying patients. Comparisons of cost and charge levels in for-profit, voluntary non-profit and government hospitals are presented, but it is emphasized that inferences about relative efficiency and profitability cannot be drawn from accounting data, given the incentives created by CBR.