Monthly Estimates of Certain National Product Components, 1946-49

Abstract

C ONSIDERABLE progress has been made in U. S. national product and income statistics in recent years. One of the most useful improvements is that, since I939, quarterly estimates have been made available within two months following the close of the period. In addition to furnishing comprehensive indicators of general economic conditions more promptly than before, these quarterly estimates have made it possible to analyze many economic relations, involving lags shorter than a year, which could not be observed from annual data. There are, however, grounds for further improvements. This paper represents a very preliminary attempt toward partially filling one of the gaps, namely, monthly estimates of national product statistics.2 For the same reasons that quarterly estimates are desirable in addition to annual data, monthly estimates would be useful in providing a really up-to-date picture of economic developments and in supplying data for the study of many important economic relations which may involve lags even shorter than a quarter. In Section I, seasonally adjusted monthly estimates of the following components I of the gross national product are derived for the I946-49 inclusive: (I) personal consumption expenditures on durable and nondurable goods as a whole, (2) personal consumption expenditures on services, (3) gross private domestic investment in new construction, and (4) gross private domestic investment in producers' durable equipment.4 The estimated figures are presented in Table i and are plotted in Charts i and 2, together with the quarterly figures as reported by the U. S. Department of Commerce. In order to indicate a possible use of the monthly estimates, an attempt is made in Section II to find the lagged relation which, on purely a priori grounds, may be expected to exist between income and consumption. In spite of several statistical attempts to establish this lag on the basis of annual data, it seems that by no stretch of imagination could one believe that this particular lag is as long as a year. Since the pay period should be the most important single factor in the determination of this lag, and since by far the major portion of income payments in this country is on a weekly or bi-weekly basis, there are grounds for believing that this lag may be even shorter than a quarter. One would therefore be led to expect that the monthly figures would yield good lagged relations. On account of the difficulties inherent in drawing statistical inference from economic time series of a very short unit-period (in this case, a month), no definite conclusion is reached on this point. For what they are worth, the evidences found seem to indicate that the lag-by-one-month relation is about as good as the relation involving no lag at all.