Knowledge that Transforms

N previous issues of this Review,1 one of the present authors presented material to demonstrate the occurrence of a seventeen-toeighteen year cycle in the development of the United States for the period I825-I933. This cycle was designated the transport-building cycle and was found to exist in such comprehensive and strategic series as transport development, immigration, urban population growth, bituminous and anthracite coal production, pig iron production, wholesale prices, and building, and in the general growth of Chicago. At approximately the same time as that material was presented, the outstanding work of Professor Edwin Frickey, Economic Fluctuations in the United States,2 appeared in which the conclusion was reached that within the geographic and temporal setting, of the study, there is analytical evidence of the presence of one, and only one, definite pattern of fluctuation. 3 And a few years later the monumental volume, Measuring Business Cycles, by Professor Wesley Mitchell and Professor Arthur Burns 4 was published in which the authors maintained that there was no compelling reason at the present time, nor even any real justification, for organizing cyclical measures of our time series on the assumption that business cycles undergo cyclical swings within periods of long building ' which periods correspond closely to the periods of transportbuilding cycles. In the light of such statements by these authorities, the existence and significance of transport-building cycles might be doubted, a priori. But upon further scrutiny, the statistical material embodied in Frickey's study is found to lend considerable support to, rather than discredit, the hypothesis of transportbuilding cycles in United States experience. And further it is found that Burns and Mitchell arrive at the cited conclusion by employing a test which is neither good nor relevant. First consider Frickey's data. For his series to show evidence of a transport-building cycle, averaging roughly seventeen-to-eighteen years in duration, it might be expected, following Schumpeter's and Hansen's classifications of cycles, that each transport-building cycle should include two Juglar 6 or major 7 cycles. Further, since it has been maintained that the correspondence is of such a form that the depression phase of every other Juglar or major cycle coincides with the depression phase of a transportbuilding cycle,8 it should be expected that in series in which Juglar or major cycles are superimposed upon transport-building cycles, every other Juglar or major cycle depression would be unusually severe and long. More specifically, since Frickey's study covers the period i866I9I4, and since troughs of transport-building cycles have been observed to have occurred roughly during the years, I860-64, I875-79, I894-Io90, and I9I4-i8, depressions of unusual length and severity should be expected in various time series some time during the years I875-7Q and I8Q4-I90oo.9

HE career of the Steel Industry Fact-Finding Board's Report and Recommendations, which were issued on September IO, I949, has been a curious one. The Board recommended that the steelworkers withdraw their demand for a general wage increase, and that a system of private retirement benefits, supplementary to those afforded under the Security Act and financed entirely by employers, as well as against death, disability, and hospital and surgical expenses, be established throughout the industry by collective bargaining between employers and the Union. The recommendations were immediately accepted by the Union and the Report was widely hailed in the public press as a statesmanlike document. The steel industry, after an ill-conceived and hopeless show of resistance on behalf of employee as well as employer contributions, surrendered on the basis of the Bethlehem pension settlement, the terms of which were somewhat more generous than those recommended by the Board. Meanwhile Ford Motor Company had concluded its pension settlement with the auto workers, and it had become clear that retirement benefits would be the main issue in collective bargaining during I949 and I950. The reasoning which led the Board to its recommendations was simple: Retirement benefits and social insurance are necessary; government has not made adequate provision; therefore negotiated plans be incorporated in collective bargaining agreements. Social insurance and pensions, the Board said, should be considered a part of normal business costs to take care of temporary and permanent depreciation in the human 'machine', in much the same way as provision is made for depreciation and insurance of plant and machinery. This obligation be among the first charges on revenues . . . So long as Government fails to provide an adequate amount, industry take up the slack. This was the extent of the Board's general appraisal of a private, companyby-company social security system. Whether, in view of customary claims of human dignity, the analogy between man and machine provides a satisfactory ethical basis for social protection is perhaps not too important. But whether private industry is in a position to offer such protection that is, whether industrial retirement and insurance plans can measure up to the need is highly important and is a question to which the Board did not address itself. The inherent weaknesses of private pensions will be reviewed at a later point in this article, which is limited to the retirement problem. It will be shown that even on the most optimistic assumptions the majority of the gainfully employed will not be covered by industrial retirement plans of the type which the Board has sponsored; that among those theoretically covered, the majority will receive no benefits, or partial ones at best; and that among those receiving benefits, many will pay the price of remaining on jobs which have become unsuitable. Clearly these are not the characteristics of a sound retirement system. And yet it may well be that the Report of the Daugherty Board, and the subsequent union-management agreements such as those at Ford and Bethlehem, are playing a crucial role as catalysts in social politics. In the months following the issuance of the Report, the basic inadequacies of private plans were often noted. Moreover, inasmuch as pensions payable under recent agreements are net of Old Age and Survivors Insurance benefits under the Security Act, employers were given a strong motive to press for larger and more inclusive OASI benefits. Promises to modernize the Act had been reiterated in Republican and Democratic campaign platforms for many years, and proposals to that effect had long been shunted about in Congress. Now redoubtable adversaries of the Welfare State joined with union leaders in seeking immediate

THE rate of economic development of the Soviet Union is a subject of considerable interest to western economists. To date, the most important single measure of this development has been the official Russian index of the physical volume of production of Russian industry published in various Soviet sources for the I920's and I930's. In the past few years, there have been several articles by western economists devoted to a discussion of the defects of this index.2 The main criticisms are these. First, the index represents gross value of industrial output rather than value-added. Second, the index is subject to an inflationary bias on two main counts: (i) prices of the baseweight year I926-27 were unduly high for many newly produced industrial goods which continued to be valued at these inflated prices even after more efficient production methods ha:d brought their prices more in line with the overall price structure of the economy; had a later year been used as a base-weight year the resulting index would have revealed a smaller rise than the existing index does; (2) the so-called constant prices of I926-27 include an increasing number of prices for later years so that, under conditions of rising prices, an artificial inflation of the index results. Soviet economic authorities themselves have long voiced dissatisfaction with the official index expressed in prices of I926-27, and in I948 it was decided to abandon the I926-27 based index in favor of an index expressed in current wholesale prices and corrected for price changes by means of a wholesale price index.3 It is not known whether the new method is in actual use, whether the old official index will be recalculated by the new method or, if this is done, whether the results will be made public. Despite these reasons for dissatisfaction with the old official index, western economists have made no attempt to date to construct independently a more satisfactory index of Soviet industrial output.4 Although Societ statistical sources contain an adequate number of series for output in physical units (i.e., tons, square meters, etc.), a means for combining these in a single index of industrial output has been lacking. This is due to the deliberate suppression by Soviet authorities of the necessary price data. No systematic presentation of price data has been made for any year, and such scattered prices as are available are largely for years prior to I930. No price index has been published for the years after I93I. Value of output is never expressed in the prices of the current year but always in the so-called constant prices of I926-

T HIS article reports the results of a statistical analysis of the merger movement in the United States since I94o.' Merger activity has long been recognized as having a major impact on the competitive structure of the American economy.2 It is also generally recognized that the high degree of industrial concentration which characterizes our economy today stems largely from the two waves of merger activity in the late nineteenth century which were typified by the formation of the Standard Oil Trust in I879 and, later, of the billion dollar United States Steel Corporation at the turn of the century. The major merger movement of the I920'S still further increased prevailing levels of concentration. With these historical precedents and in the face of the already high degree of industrial concentration illustrated by Chart I,3 the resurgence of large-scale merger activity in the United States since I940 raises issues of major national importance. In manufacturing and

STATISTICS books do not fall from heaven at random. They come in runs. For a shockingly long period no adequate textbooks on modern statistics were available, so that the interested student had to refer directly to the technical journals and to pursue each topic in many loosely connected installments. All that is changed now; a number of first-rate books have appeared and more are promised every day. Postwar periods seem to offer a favorable time for the consolidation of knowledge. In the I920's appeared such excellent works as Whittaker and Robinson, The Calculus of Observations, the Rietz symposium, Handbook of Mathematical Statistics, Arne Fisher, Mathematical Theory of Probability, and a number of other works. Good as these were in their day and valuable as parts of them still remain, it is nonetheless true that they do not present a fair picture of modern statistical theory. R. A. Fisher's Statistical Methods for Research Workers-whose first edition goes back to this period -remains an indispensable book; but there has always been something anomalous about the situation in which students and workers had to depend upon a cook-book, however useful the recipes contained therein; and no student can do justice to the truly fundamental contribution of R. A. Fisher without going back to his numerous research memoirs. Textbooks are important to the wives and children of their authors, but they have an importance that goes beyond that. Nowhere is this more important than in the field of statistics, which represents a house divided against itself. The bone of contention is of course nathematics. This subject rears its ugly head even in a field like economics. These days a young student can aspire to become a first-rate economic theorist without knowing any mathematics, but this is doing the job the hard way and he must be just that much more brilliant to make his mark. Still he can do it. In the field of statistical theory, this is simply not possible. Statistical theory is a branch of applied mathematics and no one can get anywhere in the subject without recognizing this basic fact. Moreover, the amount of mathematics needed in statistics is immeasurably greater than that required for even mathematical economics: the statistician must know the properties of many basic mathematical functions (exponential, trigonometric, Fourier transforms, and worse), not just be able to talk about f (x). The economist can go most of the way on the mathematics in R. G. D. Allen, Mathematical Analysis for Economists, plus a little matrix algebra and difference equations; but not so the statisticians. All this is realized by the members of the Institute of Mathematical Statistics, who in recent years have set up committees to remedy the present deplorable state of the teaching of college statistics. At many institutions this is assigned to young men, who teach statistics not because they like to do so or are prepared to teach the subject, but simply because they are young and at the bottom of the ladder and unable to avoid the task. There is no shortage of elementary textbooks on descriptive statistics to make life tolerable for the student and barely tolerable for the instructor; indeed the latter may choose between psychological, economic, business, educational, and biological statistics texts. It would not be true to say that these texts contain no mathematics; invariably they have a table of logarithms and squares, and a number of error formulas, of which many are usually wrong. As a genus, these works fail to distinguish between sample statistics (e.g., sample standard deviation) and population parameter (e.g., standard deviation of the universe sampled); they are shaky in the concept of degrees of freedom, and where they introduce Student's distribution, chi square, and other so-called small sampling theory, it is often too obvious where this discussion has