Knowledge that Transforms

are strong, but just how strong and how effective we need to measure. Not only the factors that help determine consumer demand, but also consumers' perceptions of the extent to which marketed goods and services will satisfy their wants, are important questions relevant to purchase predictions. This again is a little explored frontier for systematic research. Exciting opportunities have presented themselves in such events as the development of a market for small cars. Until we have gathered more data of the kinds indicated and then studied the predictive value of relevant attitudinal material in combination with that of other variables, no verdict on their predictive value but a Scotch one would seem acceptable.

HIS article is concerned with the deterT minants of consumer saving in the short run. Preoccupation with consumer saving springs from a desire to predict year-to-year changes in the over-all level of economic activity and to show the effect of consumers' actions on these year-to-year changes. As its central theme, the article focuses on and demonstrates the existence of a strong relationship between tangible consumer investment and consumer saving. Potentially, the use of this relationship, which has been hitherto overlooked, may enable economists to achieve better predictions of consumer saving. The analysis is restricted to the consumer sector only. (Thus, this article does not deal with the essentially entrepreneurial investments of consumers'whose occupation classification is self-employed businessman or farm operator.) Supporting data are introduced from the I94952 Surveys of Consumer Finances. Deflationary saving, or in net claims, defined as disposable money less total expenditures for goods and services, has been taken as our dependent variable. This variant of saving does not count as saving purchases of tangible assets such as houses or cars; only changes in claims to assets are counted as saving. We chose this variable because we are in-' terested in short-run fluctuations in the level of economic activity. Deflationary saving measures the extent to which the actions of economic units increase (the case of negative or decrease (the case of positive the circuit flow of The usual variants of saving which define saving as change in net do not measure directly additions to or subtractions from the flow and thus are less suitable for our purposes.' If the magnitudes and movements of the net worth variants of saving were highly correlated with deflationary saving, they could be used as proxies for it, but this is not the case.2 The question may legitimately be raised as to why we do not consider the use of Milton Friedman's3 (or saving) concept. The answer is that the permanent consumption concept is designed to eliminate just those transitory elements of consumption (or in which we are chiefly interested, namely those which account for most of the year-to-year variation in saving. For a similar reason we have employed measured income rather than income.

JN an article in thisREVIEW in I947, James Tobin' investigated the relationship between interest rates and the quantity of money in this country for the period I9I9-47. His results appeared to conform extremely well to the Keynesian liquidity-preference hypothesis which asserts that the demand for idle balances is a decreasing function of the interest rate, and that the interest-elasticity of demand for idle balances approaches infinity as the interest rate approaches its institutional floor. Tobin's data are shown as the dots in the graph in Chart i.

T HE Census Bureau reported a decline from 7.2 per cent to 6.8 per cent in the seasonally adjusted ratio of unemployment to labor force between May and June I958 but noted that rate is subject to small statistical and other noneconomic fluctuations. The Bureau further discounted the May-June dip in its July Report, attributing it to technical difficulties arising from application of a percentage seasonal adjustment to the figure for June I958, but there was no hint of a solution to this problem on either occasion.' Less cautious interpretation in the popular press of this improvement in the unemployment rate2 invites the response that the ratio of adjusted employment to adjusted labor force was stable in this period at 92.8 per cent; this implied an unemployment rate unchanged at 7.2 per cent. Furthermore, continued stability in this residual rate would have been accompanied automatically by a climb in the official rate from 6.8 per cent to 8.o per cent by October I958. These figures offer a recent illustration of a chronic inconsistency in the three seasonally adjusted labor force series. Under certain conditions, this intuitively undesirable and systematic discrepancy may widen rapidly and obviously discredit the adjustment process. This note generalizes the criticism of the adjusted United States unemployment series, reports a bias in the standard procedure due to application of a misleading descriptive model, and suggests a residual estimate as an alternative. The general argument can be stated as a hypothesis: Because the standard adjustment is roughly proportional to the unemployment level it will tend to exaggerate the adjustment when unemployment is high and understate it when unemployment is low. Therefore when unemployment is high and the seasonal index increases, the adjusted series will tend to decline; a fall in the seasonal index will tend to produce a rise in the adjusted series. The opposite association or bias exists when unemployment is low. This bias is avoided by the residual alternative. The following section contrasts results of the standard and residual procedures; the second section presents the argument for the above hypothesis and for the superiority of the residual estimate.

T HE purpose of this paper is to reformulate the theory of pattern bargaining to enable it to account for variations among the results of wage bargaining within a given wage pattern. We shall start with an explanation of the inadequacy of the existing wage-pattern theory in view of the increased importance of variations within wage patterns. Next we shall introduce our proposed extension of the theory, the concept of minimum differentiation, which we shall illustrate with the aid of a specific example the differentiation of a wage pattern established by cost-of-living escalation. Finally, we shall test our hypotheses by comparing recent interindustry wage increases in the so-called auto-steel orbit.