average rate in the long, and almost unprecedentedly rapid, peace-time movement from 1896 to 1915.
In any one year the movement seldom reaches 12% or an average of 1% per month. In the whole pre-war period, 1890-1915, of 25 years for which we have figures of the United States Bureau of Labor Statistics this happened only twice, the figures then being 13% and 14%.
We have monthly figures beginning only with 1900. From these we find that, beginning with January, 1900, and taking every other month up to the end of 1915, the successive jumps of the index number by bimonthly intervals were not over 1% in two cases out of three, were not over 2% in nine cases out of ten, and were not over 3% in 31 cases out of 32.
Our problem, as already stated, is how best to deal with such a tendency by selecting, as ideally as is open to us, the other four factors.
First consider the ideal brassage. This is scarcely capable of exact formulation. Evidently 3% would permit a full adjustment in almost all cases. But, as the calculations in "H" below will show, even a 1% brassage will be adequate for all practical purposes and other calculations which I have made show that there is remarkably little difference in the results between 1%, 2%, 3%, and 4% brassages.
To fix a figure, let us call the ideal brassage 1½%.
The ideal adjustment is, evidently, that which will tend exactly to correct the deviation on which it is based, thus bringing the index number back to par (except as further deviated by further tendency, and this of course is apt to be in either direction).
This ideal adjustment depends on what influence that adjustment has on the index number. If the influence is less than in the standard case the adjustment might advantageously be greater and vice versa. For instance, if the adjustment is 2% per 1% of deviation, this will just correct the deviation when the influence of that adjustment is ½% per 1% of adjustment. For
