CASE M. SPRENKLE (1969), in a refreshing departure from elasticitygrubbing that constitutes most empirical work on money demand, examined predicted vs. reported levels of cash holdings in firms, using equation derived by Baumol (1952) and Tobin (1956) as his predictor. Sprenkle concluded that compensating balance requirements, and not asset management decisions taken to cover transaction needs, are most important data in explaining amount of cash held by business firms. In this note, I argue that his results do not support his claim of showing the uselessness of transaction demand models. Rather, they stem from (a) inadequacy of Baumol-Tobin (B-T) model as a representation of cash flows, and (b) irrelevance of cash balance data contained in financial reports to empirical research on money holdings. I also show that Sprenkle's heavy reliance on importance of compensating balances is misplaced. II. TRANSACTIONS DEMAND MODELS AS PREDICTORS He sets out to show little . . . [the B-T model] . . . really explains, how subject to error results of theory are, and how fruitless more sophisticated versions of theory are apt to be. He shows that model does a poor job of predicting level of cash holdings in large business firms, and concludes with obiter dictum that compensating balances determine levels of business cash holdings. The Baumol-Tobin model depicts cash flow as a sequence of k receipts per year, each one being followed by a steady stream of payments which just exhaust it. That representation is a priori unreasonable. A large part of any firm's cash transactions are with other firms, so few firms, if any, can show cash flow pattern assumed in B-T model. Also, direct evidence from a few firms shows a random mixture of odd-sized daily net receipts and payments. In those observed cases, running mean of daily cash activity rapidly converges to a value close to zero; daily flows have no significant underlying periodicity, and certainly no trend or drift.' The B-T model, by contrast, assumes that cash account level is extremely periodic, if it is left unadjusted. * NSF research support is gratefully acknowledged. This note is a spin-off from a longer paper that has been presented before several helpful audiences. Thanks are owed to R. Clower, M. Darby, J. Kindahl, A. Leijonhufvud, L. Meyer, R. Schmalensee, C. Sprenkle, and Jack M. Guttentag, for useful comments. The usual disclaimer on sources of error applies. ** Professor of Economics, University of California, San Diego. 1. One day is ideal interval at which to observe cash flows, since returns on short-term securities can be realized on a daily basis, and banks monitor demand deposit accounts of their customers once per day.