This paper analyzes the demand for liquidity by US business firms using individual data. The theoretical background is the money demand model as proposed by Baumol and Tobin, and further extended by Miller and Orr (1966). Accordingly, a firm's money demand depends on firm size, on labor costs and on the cost of money. In this paper, the theoretical model has been modified to allow for firms' and industry heterogeneity. We use COMPUSTAT annual data for the period 1982-2000: the length of the time interval considered allows to treat measurement error which may affect the estimates. From the balanced and unbalanced panel estimations with fixed effects we infer that there are economies of scale in the demand for money by US firms, because the estimated sales elasticities are smaller than unity. In particular, the estimated sales elasticities are lower than those found in the previous literature, suggesting that the economies of scale in the demand for money by the US business firms are even bigger than formerly thought.