measured; new commodities be introduced more promptly; and that research divisions be established in the price collection agencies whose major function should be the development of methods for coping with quality changes. As for the Consumer Price Index, the Committee further recommended that this index be extended to include single persons as well as families and should cover rural nonfarm as well as urban workers; a more comprehensive index should be made for the entire population, not only for the wage and salary earners. Because of some of these problems, it was suggested that thé official published price indexes be reported in terms of full percentage points rather than in tenths of percentage points as is now done —again a simple conclusion that is in agreement with the more general comments made above ( p. 63 ).
The principle is constantly violated by the Bureau of Labor Statistics and, as a consequence, by the leading newspapers. It it grotesque to see The New York Times, for example, often reporting on its front page that "consumer prices" have "risen" or "fallen" by 1/10 of 1 percent without any qualifying word about the significance of this change in a mere index of doubtful validity. The temptation to take index numbers, whether of prices, production, or anything else, at face value is, of course, enormous. How could it be that a number that has been ground out after so many steps, operations, computations, etc., all based on a great deal of theory, should not be correct and free from error, or in other words not be significant to the last digit to which it is given? But the idea that as complex a phenomenon as the change in a "price level," itself a heroic theoretical abstraction, could at present be measured to such a degree of accuracy is nevertheless simply absurd. So are the "inferences" that the country undergoes inflation or the reverse, as the case may be.13
18 It is astonishing that in the United States changes in stock market prices are not only reported, but also interpreted, in absolute terms. For example, financial papers will report that the prices of a number of stocks rose, say, $1.00, neglecting the percentage significance of this change or throwing the burden of the correct interpretation upon the reader. Such restrictive reporting is as misleading as is the "conclusion" ( cf. page 64 ) that by virtue of such price changes the aggregate "value" of all stocks has changed by the amount of the price change per stock times the number of shares outstanding.
A closer look at just one aspect of the many problems in this area will indicate both the nature of the difficulties and the possible reasons for the lack of a concrete measure of error. Some of the data used in computing the indexes are ordinarily derived almost entirely from a highly complex network of samples. While it is natural to ask/therefore, how far the observed values can be expected to deviate from a "complete population" index, in order to have the sampling precision of the index, in numerical terms, such measures of precision are not available for any of the currently prepared price indexes. The Bureau of Labor Statistics has taken the position that this type of error is probably small in relation to systematic error (e.g., the error due to the constantly changing universe of commodities available to the consumer) because parts of the index structure depend on judgment and are "therefore" not capable of qualification of errors. Some of the investigations of the Committee suggested that for the overall index the procedural error may dominate the sampling error, but in their opinion, this does not mean that sampling error can be ignored. Empirical studies of procedural error are almost as lacking as those of sampling error. In answer to criticism on this point, the Bureau of Labor Statistics has pointed out, as already mentioned above, that they are not trying to measure the level of prices, but rather their change.14 But it would seem that this would make an estimate of sampling error even more essential since, in this case, it is likely that this kind of error will be more important than the procedural error.
It should be pointed out that while the Price Statistics Committee offers an extensive program for the Bureau of Labor Statistics, the adoption of these recommendations is not necessarily imminent. First, an increase in die accuracy of data often may demand additional expenditures for the collection of the data, and the Bureau of Labor Statistics must get its funds from Congress. Second, there is some reluctance on the part of the
14 This position has been taken by Ewan Clague, Commissioner of Labor Statistics, in a "Comment" on the Kruskal-Telser article, op. cit., pp. 280-284.
Was this article helpful?