This paper considers the performance of government forecasts--by the Congressional Budget Office (CBO) and the Office of Management and Budget (OMB)--from the standpoint of bias and efficiency. The results are mixed. On the one hand, their performance has not differed significantly from that of a private forecaster, Data Resources, Inc. (DRI). Further, even when the sample period is broken down into "pessimistic" and "optimistic" periods, forecast errors have such large standard errors that it is difficult to conclude that the forecasts exhibit any underlying bias. On the other hand, the government forecasts fail statistical tests of efficiency. In particular, forecast revisions exhibit significant serial correlation and strong seasonality. These results suggest that government revenue forecasts could, in principle, convey more information than they do at present. Also, the large standard errors of the forecasts stand at odds with the use of point estimates for policy purposes. Because the budget process ignores forecast uncertainty, the "best" forecasts for budget purposes need not be the most accurate point estimates; it might well be appropriate, for example, for forecasts to reflect a pessimistic bias. Thus, the requirements of forecast efficiency and those of a poorly conceived budget process are inconsistent.