planation issued by the Babson Statistical Organization: "After considerable study of the different subjects, it seems clear that the subject most successful as an indicator. . . is the volume of bank clearings for the country, excluding New York. . But as it is always dangerous to use one subject alone and especially a subject reflecting surface movements, it is necessary to take bank clearings as an indicator only, and to check conclusions based upon it at the end of each year by all the important barometers of wealth which are reported annually, and again at the end of each cycle, as shown by the areas of the Composite Plot. Therefore, on our Composite Plot, the line X-Y is now drawn so as to make the areas equal,1 with special attention to the cycles." 2

In other words, without offering a detailed explanation, the X-Y line is now adjusted from time to time according to bank clearings, one of the twelve subjects used in obtaining the barometer figure, and, in the long run, the line is drawn so as to make the positive and negative areas equal. In last analysis therefore, the whole scheme turns upon the X-Y line, which is readjusted more or less in accordance with what the manipulator thinks that the chart ought to show.

(2) Brookmire's system. The other system of forecasting which I shall examine here is that of the Brookmire Economic Chart Co. In this system there are three composite indices and no single plot. No attempt is made to lay down rules that the indices must always react upon each other in the same way or that any hard and fast law is to be followed. It is recognized that many forces are at work which cannot be expressed statistically but which must be taken into consideration in judging the probable course of business conditions.

1 The italics are mine.

"How the Line of Normal Growth 'X-Y' of the Composite Plot is Located," Babson's Reports.

In obtaining the Business Index the following statistics are used: 1 total bank clearings in the United States, bank clearings exclusive of New York City, commodity prices, railroad gross earnings, new building (70 cities), pig iron production, pig iron price, price of Bessemer billets, unfilled orders of United States Steel Corporation. For the Stock Market Index, the average price of twenty railroad stocks and twelve industrials is computed; and for the Banking Index, use is made of loans, deposits, reserves, ratio of reserves to loans, and rate on commercial paper.2

The method of reducing these statistics to a common basis has been explained by Mr. Brookmire as follows: "In combining these banking indices it was necessary to create a common scale on which to place each index before averaging them all together. I decided to take a period beginning with 1900 and find the average figure for each index taken. This 'normal' or 'zero' point is the place where the points of each index used fall half above and half below the normal line. For example, the 'normal' or 'zero' point of the loans to deposits graph is 98.5 per cent for the period 1900 to 1912. This'normal' or 'zero' point is the starting point of the new combined index." 3 That is, the median is apparently used as the standard in working out the scale.

In criticizing this system of forecasting, attention is first to be called to the limited number of subjects included and to the omission of all crop statistics. But, here again, the main criticism lies against the technical methods used in making adjustments for seasonal fluctuations and for normal growth. For those statis

1 J. H. Brookmire, "Financial Forecasting," Moody's Magazine, January, 1914, p. 8.

Ibid., June, 1913, p. 444.

Ibid., June, 1913, p. 444.

tics which manifest a seasonal fluctuation, the seasonal variation is calculated and, before the index is prepared, the statistics are " compensated" in accordance with these calculations. Owing to the nature of the statistics a certain percentage of error must be involved in these calculations and compensations.

As regards "normal" growth, the rate of annual increase in those figures which are influenced directly by the progressive advance of the country is also calculated, and the figures are "stepped down" before using.1 Since so many diverse forces affect these statistics, a rate of "normal" annual increase can, at best, be only an approximation; whereas the system presumes to make a nice adjustment. Obviously both the "compensation" and the "stepping down " are somewhat arbitrary, depending more or less upon the judgment of the person preparing the index. A system in which the personal element is dominant, as in this case, is always open to doubt. It does not tell its whole story upon its face.

III. SUGGESTED METHOD OF OBTAINING INDICES

It is apparent, from the criticisms which have been made in the preceding section, that one of the fundamental problems in preparing indices of business conditions is to secure a common denominator which will allow for normal growth and seasonal fluctuations without leaving any of the adjustments or compensations to personal judgment or manipulation. Το achieve this end I suggest the following method.

For each subject let a monthly index number be obtained by dividing the actual figure for the month by the average for that month during the ten preceding

1 J. H. Brookmire, "Financial Forecasting," Moody's Magazine, June, 1913, p. 444.

years. This is illustrated by the following table, which gives the ten-year monthly averages, the actual figures, and the index numbers, for one item, namely bank clearings, exclusive of New York City; the period covered being the years 1913 and 1914. The figures for clearings are from Bradstreet's.

The ten-year average for the month of January, 190312, was $4,903,000,000; the actual amount for January, 1913, $6,739,000,000. Dividing the latter by the former, an index number of 137 is obtained. This

1 In millions.

means that in January, 1913, bank clearings were 37 per cent above the ten-year average for that month. The ten-year average for February, 1903-12, was $4,142,000,000 and the actual amount in February, 1913, $5,670,000,000, which also gives an index number of 137. Similarly for each month in 1913 the actual number is divided by the average for that month during the years 1903–12. For January, 1914, the actual amount, $6,687,000,000 is divided by $5,193,000,000, the ten-year average for January, 1904-13; and a similar base is used for the other months in 1914.

By means of this moving base the comparability between the index number for December, 1913, and that for January, 1914, is maintained. The basic months used in obtaining the index for January, 1914, bear the same relation to the basic months used in obtaining the index number for December, 1913, that the latter bear to the basic months for November, 1913. By using the ten-year monthly averages, seasonal fluctuations are automatically allowed for, and by always taking the ten preceding years as the base, provision is made for normal growth.

The ten-year monthly average represents a normal standard, whether the figures tend to increase or to fluctuate about a constant level. For the purpose in hand this moving base seems superior to a fixed base and certainly it is more reliable than any arbitrary scale. It may prove advisable to use a fifteen or a twenty-year period in determining the bases, in order to reduce the influence of exceptional years. The principle, however, will remain the same. The ten-year period facilitates the use of those statistics which have not been collected for a longer time and, from numerous experiments which I have made with a wide variety of statistics, the ten-year period appears to be satisfactory.