of Labor keeps monthly records of unemployment, but up to the present time these have been published only after so long an interval as to give them little more than historical interest.

For lumber some scattered statistics of production and shipments are published and also some price statistics. Unfortunately the quotations for lumber prices in trade papers are not altogether reliable. Judging from

the statistics given in Part IV of the Bureau of Corporations' Report on Lumber, accurate price statistics for certain grades of lumber, especially for the common grades of fir and pine, would be as valuable indices as are the price statistics of other commodities.

Newspaper and book-paper prices are regularly published, but they too seem to be unreliable. Furthermore, paper is sold largely upon contracts extending over a year or more, so that the prices are somewhat inflexible. The American Pulp and Paper Association has been collecting reports of production and these were for a time published. From such material as is available, it appears that the paper trade is sensitive to fluctuations in general business conditions. The volume of advertising which the newspapers and magazines carry varies with business prospects and the size of the publications is thereby affected. When business is brisk there is also a greater demand for paper for posters, circulars, advertising booklets and for other purposes. For advertising itself some statistics are available, but not enough to be of much service as yet.

The National Association of Wool Manufacturers began in December, 1913, to collect quarterly reports of the number of cards, combs, spindles, and looms in operation and idle in the woolen and worsted mills. If these reports are continued, they should prove

1 In the Paper Trade Journal.

• Printers' Ink gives monthly tables.

valuable indices, even if they are not upon a monthly basis.

It is apparent, I think, from what has been stated in the foregoing paragraphs, that there is now abundant material for experimentation on this subject of business indices. In order to use these statistics properly some common basis of comparison is needed, which will not only provide a common denominator but which will also take into account the seasonal fluctuations. It is of vital importance to know whether an increase or a decrease represents a normal seasonal fluctuation or whether it represents a fundamental change in conditions. We now turn to a critical examination of the attempts which have been made to provide such a common denominator and to construct business barometers.

II. SOME METHODS OF BUSINESS FORECASTING

The systems of business forecasting which are now in use are open to criticism in two directions: (1) their selection of statistics and (2) their statistical methods. Such criticism does not imply a lack of appreciation of the useful service done by these "barometers." Their pioneer work has been especially valuable in creating amongst business men a more wide-spread interest, and a broader recognition of the fact that crises and depressions are not caused by politics or accidents.

(1) Babson's Composite Plot. One of the best known business barometers is that prepared by Mr. Roger W. Babson, who also publishes a very serviceable compilation of monthly statistics on his Desk Sheet. Statistics for twelve subjects are used in the preparation of this barometer,- (1) immigration, (2) new building, (3) liabilities of business failures, (4) bank clearings, exclu

sive of New York City, (5) Bradstreet's index number for commodity prices, (6) surplus reserves of the New York Clearing House banks, (7) foreign money rates, (8) domestic money rates, (9) conditions of crops, (10) idle cars, (11) political factors, (12) stock market conditions. The first four are grouped together as representing mercantile conditions, the second four as representing monetary conditions, and the third four as representing investment conditions.

From what has been said in the preceding pages it is evident that these statistics vary greatly in significance. Immigration, for example, is a much less reliable index than bank clearings or domestic money rates, and idle car statistics are altogether unsatisfactory. Furthermore, the methods of obtaining statistics for three of the subjects are open to serious criticism. In order to get an index for foreign money rates the official rates of the Bank of England, Bank of France, and Reichsbank are averaged. Such an average does not seem to me statistically sound, since the policies of these banks are by no means the same. The Bank of France, for instance, sometimes puts a premium upon gold deliveries instead of changing the discount rate. For crops only corn and wheat statistics are used. The cotton crop,

which provides about one-fourth of our exports and affects so large a section of the country, is not included. The estimated crops of corn and of wheat, in bushels, are added together, despite the fact that in this way corn is given a weight four times that of wheat, which sells at considerably higher prices per bushel and is more of a cash crop. Corn should be given a weight not over twice that of wheat. As previously stated, the figures for total production seem to me less satisfactory for this purpose than the average yield per acre. "Political factors," finally, cannot be measured statistically,

and to include such a subject indicates a startling disregard for scientific method. An index on such factors could, at best, be only guess work.

Both Babson's selection of subjects and his treatment of the figures are open to criticism. If only twelve subjects were to be used in preparing the business barometer, these twelve should have been the most sensitive and the most trustworthy. Babson's selection seems to me to fall far short of that requirement. It is especially notable that no strictly industrial statistics are used. The selection of subjects, however, is open to less criticism than the methods of manipulating the statistics.

In order to secure a common basis of comparison for these diverse denominations and to eliminate the effects of seasonal fluctuations, a set of intermediary "scale" figures was worked out.1 Taking immigration for illustration, a table of scale figures was prepared for each month. For January the highest and lowest figures for the month of January during the years of 1898-1908 were found, - 18,300 in 1901 and 56,200 in 1905. The range between these two figures was taken as equal to 100 points. The difference between the two actual figures (37,900) was divided by 10. By adding this quotient, 3,790, to 18,300, the point ten "degrees above the lowest was found, and by repeating the process the entire scale was built up in arithmetical progression until it reached the highest actual figure, 56,200. The same scheme was used in working out a scale for each month. For February the lowest and highest figures for immigration in the month of February, 1898-1908, were found and a 100 point scale similarly ascertained, and so on for the other months. Thus there is a separate scale for each subject for each month.

1 "Preparing the Composite Plot," Babson's Reports, 1912.

1

To quote Mr. Babson's own explanation: -"We then arrange the scale figures in column, placing zero over the column whose average approximates most closely to the average conditions of the years 1903 and 1904, that is the depression following the 1903 panic. This date is taken arbitrarily as the starting point of the Barometer. We then place our index figures in series to the left and right of zero. If the volume of business increases so as to go beyond the scale, higher scale figures are added, using the same arithmetical progression as at first, so that the actual condition of the years 1898-1908 serves as a constant by which to compare succeeding years. Scales similar to this one on immigration have been prepared for all subjects."

As an example of the way in which the immigration scales for January, February, and March are worked out the following table is given.

On each scale the range would not necessarily be from -40 to +60, but in every case it would have a range of 100 points, with the lowest actual figure for that month, 1898-1908, at the bottom, the highest actual figure at the top, and "zero" fixed by the figures for 1903-04.

1 "Preparing the Composite Plot," Babson's Reports, 1912.