which the classification is identical, or nearly so. For many more, despite differences in classification, it still seems possible to draw a comparison. A second difficulty is inherent in the nature of the census figures. The Census Bureau, in issuing its blanks, does not instruct manufacturers to base "value of product" upon actual net returns from sales. Instead, value is to be computed by multiplying the number of units produced by the average price during the period. Undoubtedly, the basis of computation actually used by individual manufacturers varies considerably. Special difficulties arise when "sales" take the form of inter-branch transfers, as when a factory transfers merchandise to a regional sales branch. In such cases, "value" will be based upon the price recorded in the company's books, and may well range all the way from actual cost of manufacture to the nominal wholesale price level prevailing at the time. Nevertheless, in the majority of returns, it is probable that the reported "value" is a close approximation of the net income from sales. Moreover, census data have the advantage of covering an entire industry rather than a small selected sample. "Net value" as computed from census reports must be recognized as constituting a good approximation rather than an absolutely accurate figure. Whereas differences between the Bureau of Labor Statistics price and the census net value at any one time may not be significant, an examination of their relative behavior over a period of years should afford a valuable clue to the reliability of the Bureau. of Labor Statistics data. Comparisons for sample items have been presented in chart I. Indexes for additional commodities are compared in table IV. Except where noted, the items mentioned in the following discussion are shown in table IV. Indexes have been based upon the 1929 average as 100. This base was selected in order to show more clearly the relative extent to which the Bureau of Labor Statistics price and the computed net value declined during the depression. In some cases, the Bureau of Labor Statistics price and the computed value show widely different trends. In others, significant but narrower differences exist. In the rest, the movement of the two indexes is virtually parallel. These three groups will be considered separately. Cases of Wide Disparity Very wide differences between the trend of the Bureau of Labor Statistics index and computed net value occur in the case of two standard chemical products-66° Baumé sulphuric acid (chart I) and bone black. The Bureau of Labor Statistics indexes for both of these products are extremely rigid. In the case of sulphuric acid no change in price was reported from 1928 through 1936. The price of bone black has remained perfectly stable from 1923 through 1936. In each of these two cases the computed net value shows a totally different picture. The net value of 66° Baumé sulphuric acid dropped 20 per cent from 1929 to 1933 and that for bone black showed an approximately equal decline. The explanation of this disparity seems clear. The Bureau of Labor Statistics prices are obtained, not from direct reports, but from trade paper quotations. These quotations may apply to very small purchases, but large quantities are sold on the basis of individual negotiations. In most cases sales are on the basis of annual contracts and the nominal quotation is significant only as the point from which bargaining starts. A somewhat similar picture is presented in the case of petroleum asphalt. The Bureau of Labor Statistics index for this product is also rigid. It is also based upon a trade paper quotation. Although the quotation itself dropped about 16 percent during the depression the actual plant net value dropped approximately twice as far-32 percent. Moreover, the decline in net value had persisted for 4 years (from 1927 to 1931), before the nominal quotation reflected what was happening. There is a wide discrepancy between the Bureau of Labor Statistics quotation for aluminum ingots and the price trend as computed from census data (chart I). To some extent, this may be explained by the fact that the Bureau of Labor Statistics data are for virgin aluminum, 99 percent plus pure, while census figures are for both primary and secondary metal, whose purity may fall as low as 98 percent. This difference, however, is not of major significance. The Bureau of Labor Statistics index is based upon quotations in the American Metal Market. The price of secondary aluminum is also reported by this publication. It is noteworthy that the census figures, which include both primary and secondary metal, are even lower than those for secondary aluminum alone. Moreover, the Aluminum Co. of America proclaims the policy of maintaining no fixed prices, but of dealing with each customer independently. It may be inferred, then, that the Bureau of Labor Statistics quotation for the virgin metal is purely nominal. A very different market condition exists in the case of men's dress shirts (chart I). Here again the Bureau of Labor Statistics index is extremely rigid. The Bureau of Labor Statistics quotation shows only a nominal drop during the depression, while the average net value as computed from census data declined more than 36 percent from 1929 to 1933. In this case the product classification is not identical. The Bureau of Labor Statistics data are for a shirt made from a carefully specified material, while the census figures relate to all shirts. The significant factor in this case is the peculiar structure of the market. Men's shirts, like many other lines of apparel, are traditionally sold in certain fixed price classes. Thus, in the retail market, there are $2.95 shirts, $2.50 shirts, $1.95 shirts, $1.69 shirts, and so on down. Corresponding to these retail prices are wholesale prices per dozen which show similar variations. Shirts are not sold at intermediate levels. The wholesale prices will be either-say, $18 or $15 per dozen, but not $16.50 per dozen. The product is manufactured to a price. During the depression a shirt of quality and workmanship which originally sold for $1.95 could be purchased by the consumer for $1.69 or less. Quotations, however, apparently remained rigid. Changes in price took the form of changes in workmanship and style, rather than in the traditionally established wholesale price range. On the other hand, the census index undoubtedly exaggerates the extent of the price drop. During the depression there was a marked shift of consumers from the better to cheaper garments. It may be assumed safely that the average shirt purchased during 1932 was not comparable in quality with the average sold during 1929. The true course of the market probably lay somewhere between the Bureau of Labor Statistics and the census index. Wide discrepancies also appear in the case of certain less rigid indexes. Hydrated lime and portland cement are examples of this character. Both of these indexes, while not flexible, display a moderate response to market changes. However, additional factors, such as changes in terms or special concessions, apparently exist which the Bureau of Labor Statistics data fail to reflect. The drop in plant net value during the depression as computed from Bureau of Mines data, was considerably wider than that shown by the Bureau of Labor Statistics for quoted prices. From 1929 to 1932 the Bureau of Labor Statistics index for cement declined approximately 15 percent, while that based on Bureau of Mines figures fell 30 percent. In the case of hydrated lime, the Bureau of Labor Statistics decline was 14 percent as compared to 24 percent for the Bureau of Mines data. Even flexible prices do not appear to be exempt from this type of variation. The Bureau of Labor Statistics index shows the price of yellow pine lumber as very flexible, yet it registered a decline of only 20 percent from 1929 to 1932, whereas census data indicated a drop of 32 percent. In this case, of course, the product classification is not strictly comparable. Nevertheless, the general parallelism in the course of prices between 1919 and 1929 makes it seem probable that the comparison is reasonably valid. Moreover, there is ample evidence that lumber producers often grant special prices not in accord with listed quotations in order to dispose of their product. In each of these cases, therefore, the Bureau of Labor Statistics index indicated an apparent rigidity greater than the true price structure warranted, and failed to reflect adequately the extent of the price decline during the depression. Cases of Moderate Disparity Narrower, but appreciable differences between Bureau of Labor Statistics price quotations and computed net values may be noted in a number of the cases illustrated. Thus, for polished plate glass, the Bureau of Labor Statistics publishes two series, while only one is available in the series based on the Census of Manufactures. The table shows a close general correspondence between Bureau of Labor Statistics series no. 2 and census figures. During the depression, however, the census shows a more rapid decline in prices from 1929 to 1931, than does the Bureau of Labor Statistics. By 1935 the indexes were again practically identical. Presumably, again, the Bureau of Labor Statistics quotation failed to reflect changes in terms and special concessions. Census figures for book paper and wood screws also show a somewhat wider decline than do Bureau of Labor Statistics. In neither of these cases is there strict product comparability and, therefore, only limited reliance may be placed upon the comparison. However, from 1921 to 1931, the Bureau of Labor Statistics and Census prices for book paper run closely similar courses. An interesting comparison is available in the case of sand-lime brick (chart I). The Bureau of Labor Statistics index for this commodity is quite rigid. Nevertheless, the comparison shows that it fully reflected the drop in net value which took place between 1929 and 1932. The decline in the Bureau of Labor Statistics index, however, took place a year later than did that in the Bureau of Mines figures. It seems likely that the price decline first took the form of indirect or secret concessions. When these become established, the quoted prices followed the full extent of the true decline. Cases of Close Correspondence In a very considerable number of cases, Bureau of Labor Statistics price series and net values computed from Census and Mines data show very close correspondence. Cases of this sort occur both for rigid and for flexible prices. For example, steel rails, concrete reinforcing bars, structural steel (chart I), and pig iron all show a close parallelism between the Bureau of Labor Statistics and Census prices. Other rigid, "adminis tered" prices such as salt cake (chart I) exhibit similar characteristics. The same is true of complex manufactured products such as farm machinery. The close correlation shown by the two series for hay loaders, grain binders, and mowers is very striking. The same may be said of such diverse commodities as mixed fertilizer, fire brick, washing machines, and window glass. Highly flexible prices such as those for denims. (chart I), canned peaches, and dried peaches, show an almost perfect correspondence between Bureau of Labor Statistics and Census figures. In certain cases, although quantitative data are not available, there is evidence that the Bureau of Labor Statistics price quotation is entirely accurate. This is true, for example, of iron ore, the quoted price for which is known to be rigidly adhered to. TABLE IV. Comparison of price indexes, Bureau of Labor Statistics and Census of Manufactures or Bureau of Mines, 1919–36 Year Bureau of Labor Census Labor Census Labor Census Labor Census Sta tistics Conclusion On the basis of the evidence presented, it may be concluded that Bureau of Labor Statistics data fail to present an accurate picture of price movements in the case of certain commodities. This seems particularly true of heavy chemicals, such as sulphuric acid, and of other commodities such as petroleum asphalt nd aluminum for which the source of information is rade paper quotations. Even where price series are ased upon reports by manufacturers, the quotations may be largely nominal during periods of severe economic stress, as in the case of lime and cement. For products such as men's shirts, which are particuarly hard to define precisely, the Bureau of Labor Statistics series may fall far short of accuracy. On the other hand, it is significant that Bureau of Labor Statistics data do not appear to exaggerate the rigidity of the price structures of important commodities and commodity groups such as, for instance, steel rails and agricultural machinery. Although series based upon direct reporting are probably, on the whole, more accurate than some of those derived from trade journals, there seems no reason for generally rejecting the latter with the possible exception of the chemical group.1 Trade paper quotations for such items as iron ore and steel rails have been shown to be reliable. For commodities sold on open markets, prices listed in trade publications may be presumed to be perfectly accurate. These observations make the use of caution in dealing with individual price series imperative. However, they do not preclude the use of Bureau of Labor Statistics wholesale price data as statistical bases for broad economic investigations. In analyses of price rigidity and amplitude of price movement, it becomes necessary to place emphasis upon broad and consistent relationships and to avoid relying upon small differences in absolute figures. Yet, after all due allowance is made for the factors demanding caution, very marked and significant differences still remain between the behavior of rigid and flexible prices. For the statement and interpretation of such different types of price behavior, Bureau of Labor Statistics series can be regarded as furnishing an acceptable basis. 14 The Bureau of Labor Statistics is now in the process of revising its chemical series. 79418-39- -13 |