4

of the wage data, no index running over the whole period with the average of 1890-1899 as a base being computed. Fortunately enough information was given to enable Dr. Rubinow to construct such an index. Using these figures in conjunction with the new index numbers of the retail prices, he computes a series representing real wages over the period 1890-1912. Table 1, taken from his article, brings his figures together. Chart I, accompanying it, presents graphically his findings.

Year

TABLE 1.-COMPUTATION OF INDEX OF REAL WAGES, 1890-1912.1

These figures even more than the ones previously computed by the Bureau of Labor offer the wage-earner small cause for rejoicing. And yet the data from which these figures are computed leave, we are forced to admit, much to be desired. No one, perhaps, more than Dr. Rubinow himself realizes the need of caution

For detailed information as to method of procedure see AMERICAN ECONOMIC REVIEW, vol. IV, pp. 793-817. 5 P. 811.

1890 91 92 93 94 95 96 97 98 99 1900 61 02 03 04 05 06 07 08 09 10 11 12

in drawing conclusions from data so limited. He warns us as follows:

Before these sweeping conclusions may be accepted, their general trustworthiness should be carefully scrutinized. It will be argued with justice that they are based upon an indifferent assortment of statistical data. Even up to 1907 the "index of real wages" derived from the above table is much more gloomy than the official index published in Bulletin 77 for the last time.

It is necessary, therefore, to examine the table critically comparing it column by column with the similar table published in Bulletin 77 (p. 7). Such an examination will show that the index of "hours of labor," although independently computed here, is practically identical with the official index (in 1907, 95.1 against 95.0). The same is true of the index of wages per hour (129.3 as against 128.8 in 1907); and as a result the derivative indexes of "full time weekly wages" are not very far apart (in 1907, 123.0 against 122.4).

The real difference is found in the "retail prices of food" (in 1907, 125.9 against 120.6 in the old index). This difference, however, it will be remembered, has not been introduced by the writer. It is the new "official" index of Bulletin 105 as compared with the old "official" index of Bulletin 77.

And he also says:

P. 812.

See Table I, p. 320. & P. 801.

It is possible that the omitted articles are those which have not risen in price quite as violently as some of those retained. The prices of coffee, tea, molasses, rice, prunes, salt beef or pork, etc., have surely not been so sensitive to price increase as were sirloin steak, pork chops, or strictly fresh eggs. It may be questioned how far any changes in the price of such luxuries as sirloin steak, pork chops, and eggs "strictly fresh," would affect the status of the wage-worker's family. From this point of view the earlier figures would appear to be somewhat more reliable, yet we have the official statement of the bureau (Bulletin 105, p. 6) that "these fifteen articles represent approximately two thirds of the expenditures for food by the average workingman's family."

The present writer, therefore, no more desires to make, than he feels prepared to sustain, a sweeping condemnation of Dr. Rubinow's study. He has only praise for the work. The importance and timeliness of Dr. Rubinow's subject, and the shadow of suspicion cast upon his results by their difference from the old series, prompts the desire, however, to "scrutinize carefully the general trustworthiness of the figures." The preparation of his wage index, though accomplished by combining what would seem rather heterogeneous material, was by him step by step carefully scrutinized and cautiously done, and, on the whole, would seem above serious criticism. The credibility of the series is also greatly enhanced by its close correspondence with the older index.10 It appears reasonably probable, therefore, that the wage figures are fairly representative.

With the relative prices of food, however, the situation is somewhat different. Here not only have we a very limited list of articles included-a fact in itself justifying some skepticism as to the representativeness of the material-but our suspicions are also aroused by the knowledge that the new index covering the 15 articles by no means agrees with the older list covering 30. Is there no way by which we may test the trustworthiness of this new series? A comparative study of the two index series may throw some light on the subject. Proceeding, let us first discover, if we may, why the two series differ. Since the two are based on the same price lists, the difference must be due either to the difference in methods of computation11 used, or to the fact that the old series includes articles which the new one omits. In principle the two methods are essentially different, and in general the later method is certainly to be preferred. The question confronting us, however is, Has the difference in methods given rise in this case to substantially different results?

To recompute the old series using the new method is, of course, out of the question. A comparison of the relative price of several individual articles as computed by the old and by the new method is given in Table 2. Each of these relative prices was made up from several quotations. A glance at the table shows that the difference in no case is significant. When the variation in methods gives rise to no significant differences in the relative prices of the individual articles, although in the case of each article the price was compounded of numerous samples, it is not likely that any serious discrepancy in the series as a whole would be due to method. To this evidence add the fact that for a period of eight years the old and new series almost coincide, although the differing methods apply to the whole period, and it is obvious that we must look elsewhere for the cause of the difference in the curves. On the other hand, what evidence, other than general impressions regarding prices, have we that the divergence is due to "a one-sided selection of articles"? To answer this question I have prepared series of indexes 12 covering the omitted articles (see Table 3).

TABLE 2.-RELATIVE RETAIL PRICES OF SEVERAL ARTICLES AS COMPUTED BY OLD AND NEW METHODS.

O. S. signifies Old Series; N. S. signifies New Series.