The constraint on bank supply of demand deposits, 524. — Aggregative properties, 529. — Conclusions, 529.

This paper is concerned with the validity of the conventional t tests on regression coefficients when there is serious multicollinearity between the explanatory variables. It is well known that increasing multicollinearity causes the true standard errors of regression coefficients to rise. The crucial question, however, is whether the conventional formulas will in practice reflect this rise. The purpose of this note is to show that the conventional t tests will in practice reflect this rise. But this note also points out the danger involved in mechanically dropping variables from multiple regression equations by t tests because t values of the regression coefficients may not be significantly different from zero when the true (population) values of these coefficients are in fact not zero, if the explanatory variables are highly intercorrelated.

This paper analyzes the workings and potentials of French planning as a prototype model of modern capitalist planning. Its principal concern is methodological. How can we analyze, categorize, compare, and criticize planning processes? The French planning process is not a streamlined design of smoothly fitting parts. Its formal structure tells little about its functional structure. Its explicit targets do not define its operational role. The plan is a collection of activities which have never been integrated into a single, coherent process. That is perhaps why there has been so much confusion about the way it operates; it operates in several ways at once. The French plan has two principal components. Each is a complex system possessing a powerful logic of its own. Each is based on a different planning model and each model implies a radically different conception of the political function of planning. Each pulls the plan in a different direction. The first component is a complex institution of daily, pragmatic state intervention in the activities of the major industries. The second is a formally coherent set of output targets-the general resource allocation plan.

In the February 1967 issue of the Review we presented estimates of a relationship between changes in help-wanted advertising normalized for growth in the labor force, and changes in the unemployment rate, and new hires, and a dummy variable reflecting the phase of the business cycle [11]. Two notes by Burch and Fabricant [2] and Gujarati [3] have expanded upon our findings. Each paper presents an alternative model and finds that in its own model the relationship between the unemployment rate and the amount of help-wanted advertising is not stable over the period 1951 through 1966 or 1968. The Burch-Fabricant paper tests for the shift in the coefficient of the reciprocal of the unemployment rate before and after 1957. The Gujarati paper tests for the difference in the slope of the unemployment coefficient in ten different business cycle phases from 1951 to 1968. Both papers seem to explain a greater fraction of the variance of the dependent variable than does our model. However, our dependent variable is the change in the normalized help-wanted advertising index; in the other two papers it is the level. By the usual standards, our ]R2 of 0.82 is at least as good as the somewhat higher correlations produced in the level equations. When the two tests suggested respectively by Burch and Fabricant, and Gujarati were applied to our model, the coefficient of the unemployment rate change variable proved to be stable. We first added a variable which is zero up to 1957 and equal to the change in the unemployment rate after 1957; in effect, the regression coefficient of AU is allowed to change its value in 1957. If this variable is estimated along with the other variables in our model, a simple t test will indicate whether the coefficient did in fact change after 1957. Our original equation for the period 1951-1966, second quarter, is reproduced as (1) below. The estimated equation with the dummy variable added is shown as (2):