LAWS OF NATURE IN ECONOMIC AFFAIRS 233

It will be seen that the believers m the three-and-a-half-years cycle and
the believers in the nine-years cycle were each sceptical about the reality
of the others' article of faith, and that both schools alike were still more
sceptical about the reality of the 'Kondratieff cycle' with a reputed wave-
length of about fifty years. At the same time it will be realized that, mid-
way through the twentieth century, the data were still scanty indeed;
for, even if the earliest occurrences of the phenomena themselves were
anterior to the last decade of the eighteenth century, there was at any
rate no adequate evidence for them earlier than that date; and it is mani-
fest that, within the span of 160 years running from A.D. 1790 to A.D.
1950, there had not been time for the completion of more than forty-five
three-and-a-half-years cycles, more than seventeen nine-and-one-flfth-
years cycles, or more than three fifty-years cycles. Even if the count of
instances of the shortest and therefore most numerously represented of
these three kinds of cycles were to be multiplied by reckoning as so many
separate data the simultaneous epiphanies of one and the same occur-
rence in different geographical provinces of the Western World, the total
number would still stand in three figures.

'In the sense in which the term is used here—recurrences of prosperity,
recession, depression, and revival in the business activities of countries
taken as units—the total number of past business cycles may well be less
than a thousand. For business cycles are phenomena peculiar to a certain
form of economic organisation which has been dominant even in Western
Europe for less than two centuries, and for briefer periods in other regions.
And the average cycle has lasted five years, if we may trust our data. Of
the whole number of cases to date, the 166 cycles we have measured form
a significant fraction. . . . We should be glad to have a larger sample; but
the present one constitutes an appreciable fraction of its "universe".'1

The exiguousness of the quantity of data obtainable during the first
half of the twentieth century for the investigation of business cycles even
of the shortest wave-lengths had deterred investigators from attempting to
apply to fluctuations in economic human activities a method of 'periodo-
gram analysis' that had been found to work in the natural sciences. In
the, so far, brief history of an Industrial Western Society the series
were still too short; it was still uncertain whether these series were
strictly periodic; and it was also still uncertain whether, if there were
genuine periodicities, these were maintaining themselves over long
enough periods for 'periodogram analysis' to be feasible.2 Yet, short of
going to these mathematical lengths, the investigators of business cycles
resorted to mathematical devices that would have horrified any conven-
tionally heterodox contemporary Western historian.

*The procedure adopted in ascertaining secular trends is usually em-
pirical in high degree. Starting with a time series plotted to convenient
scale on a chart, the statistician seeks to find for that one series, within the
period covered by his data, the line which best represents "the long-time
tendencies" shown by the plotted curve.... The technical process usually

1 Mitchell, W. C.: Business Cycles, the Problem and its Setting (New York 1930,
National Bureau of Economic Research, Inc.), PP- 395~6 and 397-

2 See Mitchell, ibid., pp. 259-60.

B 2915. XX 12