The Regrettable Loss of Mathematical Molding in Econometrics - Mechanism and Causality in Biology and Economics - page 61

elsewhere in the system. The problem of listing the causal factors and the problem
of invariance are closely related in which the requirements of economic-theoretical
as well as statistical and mathematical significance all have equal weights.

2 Tinbergen's Business-Cycle Schemes

In the 1920s, empirical business-cycle research mainly consisted of constructing so-
called barometers to forecast business cycles. That is to say, their research focused
on investigating whether certain economic time series were correlated. If there is a
lag between correlated time series, then it is possible to forecast the course of one
time series with the aid of the other.
The Harvard Committee on Economic Research (run by Charles J. Bullock,
Warren M. Persons, and William L. Crum) owed its international fame to such a
“barometer” based on three indices of the business cycle, the so-called A - B - C
curves. These three indices represented “speculation” ( A ), “business” ( B ), and
“money” ( C ) and were lag correlated. B lagged about 6 months behind A , and C
lagged about 4 months behind B . Therefore, A could forecast B and both A and B
could forecast C .
Tinbergen opposed the nontheoretical character of the Harvard barometer. His
very first scientific publication, “Over de Mathematies-Statistiese Methoden voor
Konjunktuuronderzoek” ( 1927 ; On Mathematical-Statistical Methods of Business-
cycle Research), was a review of this kind of business-cycle research. It criticized
the Harvard approach for not being based on any kind of causal theory. Moreover,
Bullock et al. ( 1927 , p. 79) had admitted that their method was not based on any
theory whatsoever; on the contrary, the curves were “derived solely from observa-
tion of the facts.”

Causal relations have, indeed, received increasing attention from us; but no theory of
causation or of time relation between cause and effect ever entered into the construction
of the index. (Bullock et al. 1927 , p. 79)
In addition, they observed “how foreign to actual experience are fixed mechani-
cal, or exact mathematical, relationships in the economic world” (p. 79).
Tinbergen ( 1927 ) claimed that the aim of correlation analysis should ultimately
be the recovery of causal connections, such as Karl G. Karsten's “theory of
quadrature” had suggested. Karsten's ( 1926 ) had shown the existence of two
“cumulative relations” between the three Harvard barometer indices, which he
interpreted as causal relationships. In the first place, he found, using correlation
analysis, that the cumulative values of the Harvard B -index parallel those of the
Harvard A -index, with a lag of 3 months:

X

t

B i ¼ A tþ 3 (4.1)

i¼ 1