University of Southern California Dissertations and Theses

A study of the special problems of short-run economic prediction

(USC Thesis Other)

A study of the special problems of short-run economic prediction

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content A STUDY OP THE SPECIAL PROBLEMS OF SHORT-RUN
ECONOMIC PREDICTION
A Thesis
Presented to
the Faculty of the Department of Economics
University of Southern California
In Partial Fulfillment
of the Requirements for the Degree
Master of Arts
by
Jack Earl Smith
June 19^9
UMI Number: EP44690
All rights reserved
INFORMATION TO ALL USERS
The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
Dissertation Publishing
UMI EP44690
Published by ProQuest LLC (2014). Copyright in the Dissertation held by the Author.
Microform Edition © ProQuest LLC.
All rights reserved. This work is protected against
unauthorized copying under Title 17, United States Code
ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, Ml 48106-1346
Lc ‘n Sts-i
This thesis, written by
JACK.EARL SMITH..... 3 S °
under the guidance of h.X3... Faculty Committee, i r ^ ,
and approved by a ll its members, has been ) I ^
presented to and accepted by the Council on [/V ^
Graduate Study and Research in partial fu lfill
ment of the requirements fo r the degree of
........ MASJEB...OP.ARTS.........
E._S. Bogardus
Dean
D a t e -..... Jana-l-W...
Faculty Committee
f C^Tjhairman
TABLE OF CONTENTS
CHAPTER PAGE
I. THE NATURE OF THE PROBLEM: AN ORIENTATION ... I
The problem ................. 2
Statement of the problem ......... 2
Importance of the study .......... I j .
Definitions of terms used ......... 6
The short-run .................. 6
The activity indicator .......... 7
Business activity ..... ........... • * 8
Organization of the study ......... 8
II. MONETARY ACTIVITY INDICATORS IN THE SHORT-RUN . 12
Interest rates ...................... 13
The long-term rate......... l6
The short-term rate ............ 18
Conclusions ••••.. ................ 21
Savings and investment . .............. 22
Stocks............................... 23
Conclusions........ 35
Bonds ............................. 36
Conclusions ............ I 4 . 2
Banking measurements.................... I 4 . 3
Bank demand deposits .......... I 4 . 3
Bank debits ........................ I j J j.
Bank clearings........................ I 4 . 6
CHAPTER PAGE
Commercial loan volume.................. k - 7
Conclusions........................... ^9
National income and the gross national product 50
The composition of the gross national
product 53
Use in forecasting ............... 57
Conclusions ......... .......... 62
III. PHYSICAL ACTIVITY INDICATORS IN THE SHORT-RUN . 6 1 * .
Production............................ 65
Durable goods ........................ 66
Non-durable goods .... ............... 70
Agricultural production . ............. 7 1 1 -
Conclusions ........... 76
Trade volume .................... 73
Domestic ........................... 73
Foreign .......... 30
Retail prices ......................... 82
Wholesale prices.......... 8 1 * .
Conclusions....................... 86
IV. THE INDEX NUMBER IN FORECASTING............. 89
The danger of trend extrapolation......... 90
The composite index...................... 93
Inherent limitations of the index number ... 97
V. ERROR INTRODUCED BY RANDOM OCCURRENCE IN THE
SHORT-RUN ................... 101
Political............................... 103
CHAPTER PAGE
Social ......................... ..... 106
Acts oT God ........................... 108
Conclusions.......................... 109
VI. CONCLUSIONS OP THE STUDY . .................. Ill
Short-run forecasting problems: a pattern? . Ill
Present study and progress............. ll l ^ .
BIBLIOGRAPHY............................. 119
LIST OF FIGURES
FIGURE PAGE
1. The Movement of Stock Prices and the General
Price Level in the United States Since 1875 • 28
2. A Comparison of 90 Stocks and Business
Activity...................... 33
3. Stock Prices and Bond Yields, 1875-1939 . . • • 38
I } . . Bank Clearings, Bond Yields, and Pig Iron
Production in the United States Since 1857 . 4-8
5. Durable and Hon-Durable Good Production in the
United States Since 1900 .......... 72
CHAPTER I
THE NATURE OP THE PROBLEM: AN ORIENTATION
Business-cycle theory and business forecasting have
been customarily regarded as sufficiently distinct to
warrant separate study. There is more than convention and
precedent behind such divided treatment. A natural division
exists between the subject matter of the two fields and the
observance of Its boundries facilitates the understanding
of, as well as the investigation into, economic change*
Formal business-cycle theory analyses and attempts
to explain the cause and effect relationships between basle
economic forces as well as their resultant cyclical manifes
tations. To this end it is often forced into the abstract
and hypothetical and referral to practice from theory can
become difficult. Such departure from the real to the im
agined can be a useful analytical device and, though artifi
cial, is justifiable. Most theoretical work in the business-
cycle field has been aimed at establishing relationships
and has only secondarily charged itself with the responsi
bility of having direct applicability to the problems of the
businessman. This is not so much an indication of disinter
est on the part of the theorist as it is a characteristic
Of his task. A theory designed to describe interaction
2
between basic economic forces will likely be too general
to be of use in explaining the meaning of the unique situa
tion existing at a particular period*
The realm and importance of business forecasting
may, hence, be seen to lie in the consideration of that
which business-cycle theory necessarily neglects. The pur
pose of forecasting is to assemble current business inform
ation in such form as will make possible a prediction of
future economic occurrence. Special treatment of existing
relationships rather than rigid adherence to formal economic
theory is invariably indicated. It is this emphasis on
particular as opposed to general process which, distinguishes
forecasting from business-cycle theory. Such emphasis is
the essence of this study.
I. THE PROBLEM
Statement of the problem. The businessman is regular
ly required to make decisions which will affect the future
output and pricing policies of his firm. Current commitments
may predetermine the future activities or alternatives open
to the business. It is vital, therefore, that the executive
authority of the firm estimate prospective conditions wisely.
The extreme value of an accurate forecast of future business
activity in this connection is evident.
3
The field of enterprise in which a firm is engaged
will determine what kinds of information are necessary in
order that decisions foe intelligently made. It will also
dictate how far in advance the forecast need penetrate. So
may foe seen the two requisites necessary to the preparation
of a tailor-made forecast for a special segment of the
economy.
There are certain kinds of information that will be
of interest to the businessman no matter what his field,
however. One of the most important of these is information
regarding prospective general business activity. No econ
omic enterprise can for long be independent of the effects
of the general state of health of the economy, whatever
character it may from time to time assume. The level of
general business activity is, of course, closely related
to general economic health and vigor. It is for this
reason that an accurate forecast of the former will hold
as widespread an interest as would a preview of the latter.
Many businessmen are interested, moreover, in pros
pective business activity for the period which has been
commonly termed the short-run. Though the short-run will
later be defined for the purpose of this thesis, it may here
be said that it rarely exceeds one year.
4
Though there is widespread need of and interest in
forecasts of business activity for the short-run, there has
been relatively little direct attention paid it by litera
ture in the field* Most emphasis has been placed upon the
techniques of forecasting for longer periods* The reason
for neglect of the short-run has been the smaller theoret
ical interest which it holds. The basic forces at work
within the economy rarely work themselves out in the length
of time encompassed by the short-run, and, hence, theoret
ical analysis can be most unrewarding*
With full recognition of the many manifest difficul
ties, the following thesis investigates the nature of some
of the problems which make short-run forecasting of business
activity so hazardous. Some of the problems of short-run
forecasting are, of course, common to forecasting in general.
The study will give scant attention to, or even omit, such
general problems in favor of giving more thorough treatment
to the special problems of short-run forecasting with which
it is primarily concerned.
The problems to be studied are of wide variety and
differing complexities. An appropriately planned approach
has been thought necessary and its organization is to be
found at a later part of the Introduction.
Importance of the study* The foregoing section has
suggested, and it will here be contended, that the special
problems constituting the focus of interest of this thesis
are of unquestionable significance. It hardly need be
mentioned that the techniques and arts of forecasting busi
ness conditions are as yet crude. As a result the accuracy
of forecasting work has been unimpressive. In light of the
immense value of an accurate prediction of future business
activity, any effort expended in an attempt to clarify pre
diction problems seems sustainable.
Many books on business-cycle theory and forecasting
as well as periodicals in kindred fields have included
information on, or have made allusion to, the problems of
short-run forecasting. Were this not the case a study as
broad as the present one would have been prohibitively
difficult. In most cases, however, treatment has been
limited to a consideration of a single problem or group of
problems .
The attempt of this study has been (1) to assemble
under one cover an account of the special problems of short-,
run forecasting of business activity; and (2) to conduct
the investigation in such fashion as permits the recognition
of the problems in patterns, where such appear to exist.
The value of the study, thus, is expected to be in its sum
mation and arrangement qualities. Its significance rests on
the assumption that an altered exposition of existent infor
mation may have new value by virtue of form and, more ambi
tiously, permit arrival at new and useful conclusions*
II. DEFINITIONS OF TERMS USED
The short-run. The short-run, as the term will here
inafter be used, does not refer to either the market period,
short-run or intermediate period of conventional economic
theory* These periods are usually defined in terms of the
alternative actions open to the producer*- while the short-
run used of this paper will refer to chronological time* As
such the term is generally substitutable with both the
“short-period” and “short-term” of forecasting literature.
It is evident that "short-run” for one businessman
may well be a rather long period in the eyes of another.
The need of this study is a definition of short-run which
will be broad in the sense of being general and, in addition,
be definite in time sense* Short-run forecasts have not
been commonly identified with a particular time period, and
it has, thus, seemed permissible to formulate an arbitrary
1 See Robert Fettengill, Price Economics (New York;
Ronald Press, 19^8), p. 191.
definition for the purpose at hand.
The majority of businessmen are especially interested
in prospective business conditions for the ensuing twelve
month period. This is so because the majority of business
men are producers, wholesalers, or distributors of non
durable goods. Operational commitments in this area are
generally for periods no more than a few months in advance
and hence there is intense concern over possible turns in
business activity over like periods.
Unless otherwise stated, therefore, the outside limit
of a short-run forecast as considered in this study will be
taken as one year. The delimitation inherent in this def
inition has greatly facilitated the investigation and, it
is felt, is open to no serious criticism*
The activity indicator. Activity indicators shall
be interpreted as being measures, monetary or physical, of
current economic quantities of recognized forecasting value.
As such, activity indicators are identical to what are common
ly termed “business barometers’ 1. The advantage of the former
lies in its lesser connotation of forecasting value. The
shortcomings of the word barometer for forecasting have beai
carefully noted by Bratt.2
^ Elmer G. Bratt, Business Cycles and Forecasting (Chi
cago: Hi chard D. Irwin, Inc., 19i|B), p. 351]-. „ Bratt says, “In
the proper sense, the series measured is as truly a business
barometer as the series which leads. It may well be that the
best practice would be to avoid entirely the word barometer in
talking of business measures."
8
Business activity. The meaning of business activity
shall be taken to be the measure of the vigor of the eeon-
mic community. There are many indexes of general business
activity currently available, most of which are composed of
numerous production, trade or financial series. Ho parti
cular index is referred to by the present definition al
though it should be made clear that business activity has
been considered to include trade as well as production. An
index of activity such as that regularly appearing in the
Federal Beserve Bulletin would be ruled out as considering
production data only. Such indexes as the Hew York Times
Weekly Index of Business Activity and Barron* s Index of
Production and Trade are good indicators of general business
activity as defined for the purpose of this paper.3
III. ORGANIZATION OF THE STUDY
The special problems confronting the short-run fore
casting of business activity might be investigated in a num
ber of ways, all satisfactory. The danger in all approaches
would seem to lie in emphasizing some problems over those
3
For a statistical breakdown of the more common in
dexes of business activity see: Frederick Groxton and Dudley
Gowden, Applied General Statistics. (New York: Prentice-^Hall,
Inc., 19ho), p. 631-3h.
equally important as well as omitting some classes of pro
blems altogether* The organization of the study to follow
faces the same pitfalls as would other plan3 of approach
but it does, it is contended, minimize the possibilities of
outright blunder.
If the forecasting qualities of the common business-
aetivity Indicators were accurately known and their portent
skillfully manipulated it would be possible, today, to make
dependable economic predictions. Institutional causes would,
of course, prevent one hundred percent accuracy but the re
sults would be incomparably superior to those now being
obtained.
The very shortcomings of these business-activity in
dicators suggest a method of analyzing the forecasting pro
blems now extant. The method is, namely, that of appraising
the defects and limitationsof the activity indicators now
in use as well as Investigating their special inadequacies
for the short-run.
To this end the present study treats a number of the
common activity indicators in the two classifications common
ly made,monetary and non-monetary.^- Considering the problems
^ Wesley Clair Mitchell, Business Cycles and Their
Causes (Berkeley and Los Angeles: University of California
Press, 19i|i), p, 188,.
10
presented by some indicators while neglecting those pre
sented by others represents an arbitrary choice but an
attempt was made to include the most commonly used. The
general process of analyzing the various indicators has
been:
1. To review published information, whether in book
or periodical form, with regard to establishing the use of
the particular indicator in short-run forecasting.
2. To stress the inadequacies and deficiencies,
where apparent, of particular indicators.
3. To determine, in conclusion, the short-run fore
casting problems presented by the indicator under analysis.
Two separate chapters of the thesis have been devoted
to investigation of two broad problems of particular signi
ficance to short-run forecasting. These are the problems
of index number use in forecasting and, also, the problems
presented by random occurrence.' Heither of these can, of
course, be studied by the indicator analysis outlined above
but their importance demands their inclusion and treatment.
The final chapter is devoted to summarizing and re
lating the special problems of short-run business activity
forecasting as revealed by the study at large. These pro
blems do, it is maintained, fall into certain patterns and
such are outlined* ©ads chapter, also, will give exposition
to present controversy and progress in areas closely related
to that of the present study.
CHAPTER I I
MONETARY ACTIVITY INDICATORS IN THE SHORT-RUN
The modern economy has become so complex that it is
a practical impossibility to gain an understanding of its
operation by reference to physical production and distribu
tion alone. Money and money substitutes are much more than
a "veil" hiding exchanges of goods and services and are of
vital importance as factors of basic change in their own
right. Accordingly, many business-cycle theory authorities
have chosen to give prime emphasis to the pecuniary phase of
economic activity. The justification of such an approach
has been well stated by Mitchell:
The reason for . . . staying upon the ’ money surface
of things’ in analyzing business cycles, rather than
attempting to penetrate beneath to the motives that
actuate economic conduct, is . . . that ... modern
economic activity is immediately animated and guided,
not by the quest of satisfactions, but by the quest
of profits. Therefore business cycles are distinctly
phenomena of a pecuniary as opposed to an industrial
character. ... Businessmen refuse to complicate
their problems by going back of the dollar to that for
which, the dollar stands, and he who would understand
what they are doing must treat their action as it is.l
There are writers who do not weight monetary motives
1 Ibid.. pp. 187-88
13
quite so heavily as does Mitchell,2 at least for purposes of
forecasting. Pew would allege monetary factors to be of but
minor importance in a profit economy, however.
The remainder of the present chapter is devoted to
an investigation of the forecasting problems presented by
monetary indicators of business activity in the short-run,
I. INTEREST RATES
So very much has been written concerning interest
rates and so wide has been the variance of opinion that it
seems that one could conjure and develop, with substanti
ating documentation at each step, very odd theories indeed*
The present section has no position to champion but, instead,
has endeavored to select with impartiality and balance the
more commonly advanced opinions as to the value and use of
interest rates as indicators.
The usefulness of interest rates as forecasting im
plements rests, as is the case with other indicators, upon
the degree of accuracy with which they presage future econ-
mie change. Interest rates, as a group, cannot be readily
^ Norman J, Silberling, The Dynamics of Business (New
York: McGraw-Hill Book Co., Inc., I9I 4. 3), p. 571, Silberling
expresses a widely, held sentiment when he says, MA forecast
of the extent of the correction about to occur in.business
activity would be inadequate if it did not extend well beyond
banking conditions or the state of the security markets and
such fundamental factors as the building cycle, international
conditions, the state of agriculture, and raw-material inven
tories on hand in various processes.1 1
assigned a single index of value for this purpose* They
consist of a variety of different rates for various kinds
of loans as well as being distinctive as to the time for
which such loans are extended* As. aright be expected, some
rates are of much more use for short-run prediction than
are others*
Writing in 1931» Haney stated:
One general rule is that, sooner or later* (italics
in the original] low money rates are followed by busi
ness expansion and high money rates by business re
cession* While the time lag is uncertain, this rule
is helpful in business forecasting*3
Here, it seems, is a restatement, with allowance for
time lags, of the theoretical economic concept of interest*
If interest rates were but the prices which equate the de
mand and supply for loanable funds of different categories,
they might well, as Haney suggests, be of considerable help
in business forecasting. Interest rates, unfortunately for
purposes of prediction, do not readily respond to changes in
the demand and supply for funds. There are a number of
reasons why this is so, as we shall see*
Recent years have witnessed a substantial change in
the interest rate as an economic force. The United States
3
L. H. Haney, Business Forecasting (Mew York: G-inn
and Company, 1931), p.. 219*
15
government has accumulated an enormous national debt and,
consequently, has assumed increased stake and influence in
domestic financial matters. Government policy has been to
manipulate interest rates to such artificially low levels
as would assure low debt service charges, The result has
been a low and relatively stable long-term rate as well as
greater stability in the case of most other rates* In
creased stability has meant lessened sensitivity to business
activity fluctuations and, hence, reduced value as an in
dicator.
This new character of the interest rate has caused
many writers to relegate the interest rate to a position of
but secondary importance as an indicator of change. Pew
are more positive than Knight:
To the ... question as to what interest rates
have to do in an effective causal sense with the
course of events of the cycle the answer undoubtedly
is ’very little1I To be sure, interest is an element
in cost of production, and interest rates, as well
as wages show an important lag, . , , Its effects
are important ’in the long-run’ but not.for the per
iods for which businessmen can or do make plans , , , 4 -
It would not be correct to state that the interest
rate lost its value as an indicator overnight. Variations
^ Prank H, Knight, "The Business Cycle Interest and
Money," Review of Economic-Statistics.Vol. 23, p, 60,
16
In the prevailing rate were "smoothed out" as far back as
the early Twenties by action of the Federal Reserve Board*
The advent of the Great depression found very little re
sponsiveness remaining between the interest rate and Invest
ment activity. Speaking of this period, H. C. Wallich
points out:
In the United States, interest rates declined to
very low levels without succeeding in arresting the
collapse of the early thirties or in restoring nor
mal activity later on. It was, as Professor Hansen
has pointed out, something in the nature of a labor
atory te3t with very discouraging conclusions for
monetary policy,5
Most business-cycle theorists as well as experts in
the forecasting field distinguish between long-term and
short-term rates in evaluating the worth of interest as an
indicator. The short-term rate is ordinarily identified as
the bank rate of interest on commercial paper of from two
to six month maturity. Long-term rates, for the most part,
are those prevailing for bonds both government and private.
The long-term rate. The long-term interest rate
offers, as has been pointed out,6 little as a device for the
^ H. C. Wallich, "Changing Significance at the Inter
est Rate," American Economic Review. Vol. 36, December, I9I 4 . 6,
p. 762. .
6 Cf. p. 15.
17
short-run forecaster. It does seem to correspond In a
poorly understood manner with the long-run price level, a
relation noticed by Tinbergen and Fisher,7 The relation
ship between the long-term rate and the state of business
activity are probably remote. According to Tinbergen the
long-term rate of recent years gives no positive evidence
O
of cyclical variation at all.
The reason for the unresponsiveness of the long-term
rate might well be laid at the door of government regulation
and manipulation. The effort of the government to minimize
the service charge on the federal debt has resulted in un-
realistically low rates upon the funded debt of private
corporations. The policy of many producers, especially
those engaged in heavy manufacturing, seems to have been
one of disregarding the long-term rate in connection with
7 J. Tinbergen, "Some Problems in Explanation of
Interest Rates,” Quarterly Journal of Economy, Vol. 6l,
May, 1 9 k - 71 p. 43k* The long-term rate is here observed to
exhibit a ten year lag with respect to the long-run price
level. Tinbergen's feeling might well be that of the
forecaster as he queries, "... I ask, is this not rather
long?”
Q
Ibid., p. l j . 3 2 . Tinbergen refers to the three to
four year "American” cycle*
18
capital outlay expenditures. Manufacturers seem, in many
cases# to base their calculations on a "normal” rate of re
turn for the particular industry in which they are engaged,
This rate of return may be quite independent of the long
term market rate of interest,9 The rationality of such be
havior may be questioned but it is not the immediate concern
of the forecaster to do so. It is of importance to note,
however, that historical evidence does not substantiate the
alleged causal relationship between interest rate and in
vestment for a vital segment of the economy.
The short-term rate. The short-term rate has been,
and remains, considerably less stable and more sensitive
than the long-term rate. Short-term rates have given "tip-
offs" of turning-points in cyclical activity, but the gen
eral performance of such rates as indicatorsof impending
change has been spasmodic. Those who emphasize money have
the 1929 downturn as an excellent example of the short-term
or commercial paper rate operating as an indicator of change,
9 F. P. Lutz, "Interest Rate and Investment in a
Dynamic Economy," American Economic Review. Vol. 35, Decem
ber, 19^5# p. 082. As Lutz describes# , f it is unfortunately
inpossible to judge how widespread is the custom of calcula
ting with a standard rate of "interest" which is indepen
dent of the level of the market rate. .It seems reasonable
to suppose that the practice is particularly widespread
among companies which, because their interest payments are
of negligible magnitude, are not sensitive to changes in
the interest rate."
Commercial paper rates became substantially stiffer a full
year in advance of the Crash.10
The short-term rate, however, is not usually so por-
tentuous. Though more flexible than the long-term rate,
it is nevertheless sluggish in its response to money market
conditions. Furthermore, the causal relationship between
the short-term rate and investment activity can be most
misleading. Banks may raise the short-term rate by limit
ing short-term credit, for example. Though this would
evidence less confidence on the part of banks it would not
be an Indication of the sentiments of entrepreneurs. In
addition, and of great importance, the current supply of
money in relation to the demand for cash bears heavily
upon the short-term rate.11
The aforementioned shortcomings of short-term rates
as indicators of current business movements have been widely
observed and much discussed. The conclusions of various
authors are interestingly different. After, a rigorous
econometric analysis of some length and complexity, Tinber-
Norman J. Silberling, Dynamics of Business (New
York: McGraw-Hill Book Co., Inc., 19h3, p. 296. ' ” .
11 ^
J. A. Estey, Business Cycles. Their Nature. Cause.
and Control (New York: Prentice-Hall, I94I, p. 290-91.
20
gen summarized the short-term interest rate case as follows:
Shorter cyclical movements show themselves es
pecially in short-term rates where they are very
pronounced. They lay somewhat behind the general
cycle, as is expressed in the G_ curve of the
Harvard Barometer . . .12
Prom this it is fair to deduce that the short-term
rate might be a valuable indicator for short-run forecasts.
The general cycle to which Tinbergen refers is the three
to four year “American” cycle. The shorter movements about
the "American” cycle are of critical importance to the
short-run forecast and it is this movement Tinbergen here
relates to the short-term rate.
The conclusions of P. P. Lutz, on the other hand,
allow for but negligible connection between business activ
ity and the short-term rate. In his words:
... in a dynamic world a change in the (short
term) interest rate will not affect the calculations
of a trader (or a manufacturer) in such a way as to
induce him to reduce or increase his inventory hold
ings *13
12 Tinbergen, op. eit., p. i^-32. Se also, M. Kalecki,
"The Short-Term and Long-Term Rate of Interest,” Review of
Economic Statistics. Vol. 33, May, 19ljJL, p. 99•- In this
article M. Kalecki relates the short-term rate to-money vel
ocity. The esqpression i3 of the following form:
MV (a) = T
Where:
. M = the quantity of money and money
substitutes
V = the velocity of turnover of M
d * the short-term rate
T = transactions in dollars
1^
J Lutz, op. cit., p. 830.
21
Though reference Is here made to inventory holdings,
the tenor of the Lutz thesis is such as to permit extension
of his comment to general business conditions as well*
The position of Mitchell\is intermediary between
the extreme positions of Kaleeki and Lutz. Mitchell would
relate the value of the short-term rate as a forecasting
device to the phase of the cycle. The short-term rate is
contended to lag behind business activity until an upswing
is well under way 1 ^ 4 - but precedes it on the downswing. 15
There would, thus, be danger In using the short-term rate
as an Indicator when the cyclic phase approached a peak or
trough.
Conclusions. The relative treatment given here to
the interest rate is somewhat more extended than its limited
value as an indicator in the short-run warrants. It has
been accorded such emphasis due to the wide attention being
given the interest rate in its relation to the marginal
efficiency of capital. 3-6
The interest rate as an indicator of change in the
3*^ Mitchell, op. cit*, pp. 18-19,
^ Ibid.. p. 135*
3-6 John Maynard Keynes, The General Theory of Employ
ment Interest and Money (Kew York: Hareourt Brace and Co.,
1935), p. 183-04* Also see Bratt, op. cit.. pp. 223-2£.
short-run has been seen to have shortcomings which pose
the following problems to the forecasters
1. Interest rates are subject to variance due to
factors other than investment activity, that is, -the in
fluence of the Federal Government and private financial
institutions.
2. Interest rates, especially the long-term rate,
are sluggish and unresponsive to business activity in the
short-run — the causal relation is, in fact, uncertain.
3. The lag or lead of interest rates with respect
to general business activity is indefinite and tends to
vary with the phase of the cycle.
II. SAVINGS AND INVESTMENT
Changes in investment and saving do, essentially,
cause and constitute the cycle. The major swings, both
upward and downward, can be explained by the lags and un
balance of the savings-investment arrangement.17 It is
quite understandable, therefore, that the business forecaster
has seized upon indicators of savings and investment varia
tion as one of his most significant analytical tools. Mo-
23
digliani has stated, in fact, that the ratio between in
vestment and savings is a key to the whole problem of
economic forecasting*^
Although aggregates indicative of mass movements of
savings and investment are of use in forecasting, major
interest has centered about variations in the segments
thereof* It seems desirable at this point to give consid
eration to the use and problems characteristic of some of
these.
Stocks* The stock market has been one of the most
dependable of all change indicators over a period extend
ing as far back as the Civil War. Variations in the prices
of stocks have most always been reflected by subsequent
changes in the level of business activity. There have been
instances, such as the 1926-1929 period, where speeulatory
activity has caused stock indicators to be misleading. Qr-
3-® Franco Modigliani, "Fluctuations in the Saving
Ratio, A Problem In Economic Forecasting," Social Research,
Vol. l i | * , December, 1947# p. ip-4* Modigliani states his
view thus, "Currently, atteng>ts are being made to estimate
investment in advance; in the case of domestic investment,
estimates are based on entrepreneurs reports on their in
vestment plans* But even if such estimates should prove to
be only moderately accurate, there Is much to be gained from
knowing what level of investment will be required to produce
any given level of income and employment. . .
It is clear, therefore, that the relation between
saving and income holds the key position in economic fore
casting*"
2 k
dinarily, however, stock price movements have had real and
relatable significance.
Unlike certain other of the indicators, stock prices
can be causally as well as functionally related to business
change. Prom the day of the earliest professional fore
casting services, such as the Brookmire Barometer, the
stock market has been viewed as an arena in which present
business activity was ’ ’ discounted” in anticipation of future
trade movements. It can. thus be maintained that market
movements contribute to, as well as act as an indicator of,
the level of business activity of future periods. A con
trary viewpoint, namely that the stock market is a baro
meter and not a cause of future business activity, is
fundamentally untenable. As James Ross has stated:
The rationale of the stock market as a barometer
and not a cause of the volume of business rests on
the theory that the large holders, officers and dir
ectors, who, by * inside1 ' examination of costs,
prices, inquiries and orders for future delivery,
estimate future earnings and by speculative market
action cause the stock prices to indicate in their
movements what earnings will be sometime later. The
fatal weakness of this explanation lies in the time
element. The average lead of the Dow-Jones index of
industrial stocks of 10 business upturns is 8.3
months while the average lead of the index of business
downturns is 6.5 months. How can insiders know so
far in advance that business will be better or worse
or what the earnings will be?3-9
3*9 James A. Ross, Jr., Speculation. Stock Prices and
Industrial Fluctuations (New York: Roland Press Go*, 193&77
p. 352.
25
Causal relationship, If established, would be of
great Interest to the forecaster* The fact that it has
not been, however, detracts but little from the usefulness
of the well verified functional relationship*
Practical use of changes In stock prices as a fore
casting Indicator Involves making a selection from the
various indexes or averages of stock fluctuations* Al
though a given stock or group of stocks may seem superior
to averages for short periods, a well balanced index is
more reliable* There regularly appear several score of
stock market averages including combined averages of stocks
of all kinds as well as separate averages for industrials,
rails, and utilities. Some of these are simple arithmetic
averages of the prevailing prices and priee changes of the
selected stocks while others are mathematically weighted
by a number of different methods.
It is quite surprising, in light of the different
formulations used, that the various indexes have been so
close in their variations* Rarely can they be found to
have moved in opposite directions for more than a day or so
at a time. Hence, though the sensitivities differ, the
forecaster of general business conditions will have but
little to choose between them*20
20 Haney, o£. cit*. p. 133.
The great value of stock prices as an indicator is
not simply that they lead business activity fluctuations
but that they do so both during upswings and downswings.
Thus stock fluctuations have been especially useful in the
predicting of turning-points, the highest achievement of
the forecasting art. Stock fluctuations have, in addition,
a record of leading changes in business activity by a great
enough time to enable their inclusion as part of the in
fo mat ion necessary to the preparation of the balanced fore
cast. The average time of this lead has been noticed to
differ rather considerably and appears to be greater in
the upswing than in the downswing.^ Disagreement between
authors as to the length of this lead can probably be ex
plained by difference in method of calculation.
P. R. Macaulay of the National Bureau of Economic
Research has compiled one of the most remarkable and most
complete accounts of stock prices and business activity yet
to appear.22 ^fr© study is complete for the years 1857 to
1935» inclusive. The average of a selected group of rail-
21 Of., p. 2l p .
22 P. R. Macaulay, Bond Yields. Interest Rates and
Stock Prices in U.S. Since~~ lH56. New York: National Bur
eau of Economic Research, 193°)• See Chart 21 for other
series in addition to the adaptation for stocks alone repre
sented by Figure 1.
27
road stocks has been adjusted for trend and minor fluctua
tions and has been plotted, on a logarithmic chart, against
The Snyder Index of General Prices, Since the variation of
railroad stocks is most generally in aceord with movements
of stock priees as a whole, they may be accepted as a fair
indication of stock fluctuations at large. Also, the
Snyder Index of General Prices has been observed to move
with such indexes of general business activity as Barron*s.
We may, thus, make a comparison which is both valuable and
in keeping with the definitions of the present investiga
tion.
It should first be noticed that Macaulay*s plot of
the two trends — see Figure I — gives ample evidence of
the rather consistent lead of the turning points of stock
prices over those of business activity. Though the lead
has varied from extremes of three to thirteen months, it
has ordinarily been between four and nine months.
Though no horizontal rulings appear in the original
and the vertical axis is not sealed in units, the logarith
mic plot enables a comparison of the rate of change in
stock prices as contrasted to that of general business
activity. The angle of inclination and declination of both
the stock and the Snyder Index are decidedly similar. The
importance of this comparison Is, of course, In demonstra-
FIGURE 1
THE MOVEMENT OF STOCK PRICES AND THE GENERAL
PRICE LEVEL IN THE TJ . S. SINCE 1875
(Vertical Scale Logarithmic)
E
K
1 9 2 5 1335 1 3 3 0 m 1920 1895 1 9 0 0 1905 1885 1890 1 8 7 5 1880
E z Plot of American R. R. Stocks IInverted)
K z Plot of Snyder's Index of General Prices in the U* S. (Inverted)
From Chart 21 p. 218, F* R. Macauley The Movement of Interest Rates,
Bond Yields and Stock Prices in the United States Since 1856•
ting that the lead of the stock is consistent on both the
upswings and downswings*
Further,, and of more current interest, it can be
seen that fluctuations of the stock in recent years has
been an exaggeration of the corresponding fluctuations of
the Snyder Index*
Though Figure 1 is in excellent form for its pur
pose, that of showing the lag of business activity with
respect to variation of a representative stock, it is
important to notice its limitations. The rhythmic beauty
of graphics of this type can be most deceiving — especi
ally when the data, as in this case, is concerned with
economic occurrence. The curves, to repeat, are but trends
of the actual historical data, while the rising secular
trend has been neglected entirely. The inverted logarith
mic plot distorts the absolute magnitude of the amplitude.
Only the time axis is in units requiring no interpretation.
Having, then, established some of broader relation
ships between stock and business activity movement, it is
appropriate to deal with certain problems of forecasting
associated thereto. As previously stated,23 the index of
stock price selected is usually unimportant. An exception
30
to sueh a practice exists for periods of critical uncer-
stocks caused several prominent indexes to be insensitive
to generally falling stock prices in I929* The error was
due to unrepresentative weighting and is a potential danger
presented by any index of fixed composition*
Even a well weighted index of stock prices will be
a poor indicator of prospective business activity if there
is rampant speculation in the market* Speculation, of
course, leads to over or under valuation. Evidence indi
cates that even speculatory activity will not long eause
stock prices to deviate from the longer term trend. Even
so, the short-run forecasting value of stock prices during
such periods has been negligible.^
Another factor affecting the significance of stock
indexes is the condition of the short-term money market.
A change in confidence or-in policy on the part of the call
loan bankers can mean either contracted or expanded part
icipation by those operators dependent upon margin. Chances
of error from this source are great if marginal trading is
a large portion of the total. Present restrictions on
tainty. Haney notes that a heavy weighting of "blue chip"
period „ w
31
margin reduce the chances of error from this source appre-
the effect of the changes of the market upon its own sub
sequent changes. This may also be stated as the reaction
of the market to its own fluctuations. It is one of the
hardest of all effects for the forecaster to deal with as
it is psychological In nature and, hence, not subject to
ready measurement. Conditions surrounding the 1929 Crash
give elaborate testimony to the irrational, yet powerful
reaction to a market fluctuation of major proportions.
Willford King gives good expression to the exasperating
circumstances then extant:
What mystified observers generally at that time
was the fact that this decline should occur at a
time when the industrial fabric of the nation was
still In prime condition. Factories were in excel
lent repair and equipped with the most modern ma
chines. . . . Almost everywhere, the workers were
willing and able to perform their tasks efficiently,
and yet, with every condition apparently favorable
to the maintenance of prosperity, the business
nation gradually slowed down to a
economic activity, not to say forecasting. The lesson in
^ W. I. King, Causes of Economic Fluotuation-
Possibilities of Ant i c ip at ion and Control. (Hew York: The
Ronald Pres s , 1 9 3 6) , p. 3 i|_*
ciably.
Finally, and perhaps most important of all, there is
Full explanation is probably outside the realm of
32
regard to use of stock prices as an Indicator would seem
to be that the rate of change of prices can be more im
portant than their absolute level. The experienced and
alert forecaster may sense the situation and compensate
for it. Failing this, he errs as no mechanieal technique
exists.
A recent, and cautioning, experience with stock
price indicators has been the “unruly" behavior of the
market since the end of World War II. Commencing with
August of 19^5, stock indicators can be noticed to have
deviated very considerably from the traditional behavior
briefed in this section.
Reviewing the record of this period, an editor of
Business Hews expressed the opinion that the market "is
still OK as a forecaster."26 Such opinion was not without
qualification as his words testify:
In the meantime, the record cited here see Figure 2
does not necessarily mean that the stock market has
lost its touch as a business forecaster. It may
simply mean that it is in the process of becoming a
high-speed tipster on the immediate business outlook.
You want to be very much heads-up in using the
stock market as a business guide. Itfs off on a new
and tricky tack.27
"Stock Market as a Business Forecaster," Business
Hews. April 19, 19^1-7, p. Il6. _
^ hoc. cit.
33
FIGURE 2
A COMPARISON OF 90 STOCKS AND BUSINESS ACTIVITY
(LAST WEEK OF AUGUST 1945 = 100)
SO
From Business News; Standard and Poor’s Corporation.
2405, U niv ersity Bookstore, Los A ngeles
A study of Figure 2 reveals that the market was off
on a "new and tricky tack" indeed. It is not possible to
locate a comparable situation in the recorded history of
stock and business performance.28 A later article by the
editor contained the reassurance that no decline in busi
ness activity anticipated by the stock market by more than
seven months has ever amounted to much.29
Experts have been unable to agree as to what the
new relation between stock price indexes and business
activity signifies. A poll was conducted in the fall of
194-6 in which the question was asked, "What, in your opin
ion, does the decline in the prices of corporation stocks
signify?".30 The responses of fourteen well known author
ities, including economists George Edwards and Walter E.
Spahr, were received. Agreement seems to have been (1)
that the prevailing market condition in its relation to
business activity was highly atypical, and (2) that positive
judgment as to significance, at the time, would be diffi
cult and dangerous.
Whatever the meaning of the changed relationship,
Cf., p. 28, Figure I.
^ "Stock Market as a Business Forecaster: reply,"
Business Hews. May 3, 194-7* p. 116.
U.S. Hews. Vol. 21, October 4* 194-6, PP* 32-34-*
Also, U.S. Hews. Vol. 21, October 1, 194. 6, pp. 30-32.
35
the forecaster Is forced to recognize it* Whether or not
the stock price index has lost its reliability as an indi
cator of change is a question to be answered by future
development 9 •
Conclusions * The stock market has been seen to have
the finest record, historically, of any indicator of busi
ness activity fluctuations. It does, however, have certain
limitations and deficiencies as an indicator of change in
the short-run. Stated in the form of problems confronting
the forecaster:
1. Stock prices and stock price indexes are subject
to fluctuation by factors other than the "discounting" of
the future frequently alleged to give causal relation to
the market as an indicator. These include speculatory
activity, short-term credit expansion or contraction, and
Irrational mass psychological reaction.
2. A well adjusted trend of the cyclical fluctuation
of stock prices shows a rather consistent lead over corres
ponding variations of business activity. Variation about
the cyclical trend may well lead the short-run forecaster
astray.
3. The lead of stock prices over business activity
movements varies with the phase of the cycle.
i j . . A new and as yet poorly understood relation of
36
the market to business has developed since World War IX,
The traditional dependability of stock prices as an indi
cator has been under question although, of course, this
does not constitute a denial of continuing causal relevancy.
Bonds, Observers of business trends have long been
aware of the decided correlation between changes in bond
prices and yields and the level of business activity. As
in the case of stocks, bond prices and yields have been
badly misleading during particular periods. The record
reveals that most such variances have occurred during war%
however. By and large, bond trends have corresponded to
business trends. They have, therefore, received consider
able study with regard to their utilization as forecasting
tools.
A distinguishing characteristic of past movements
of bond prices has been their great stability relative to
stock price fluctuations. One authority suggests that the
reason for this is due to the nature of the security itself.31
3^- W. H. Husband and J. C. Dockeray, Modern Corpora
tion Finance. (Chicago: Richard D.Irwin, Inc., I9I 46), p. f?87.
As Husband and Dockeray compare movements of bonds as con
trasted with stocks, "As has been stated on other occasions,
the price of a share of stock Is considerably influenced by
earnings. These, in turn, are greatly affected by cyclical
movements and it is not surprising to find that stock prices
fluctuate to an astonishing degree. Bond Prices on the con
trary do not respond so quickly to earnings but tend to
reflect the basic factors of risk and prevailing rates of
interest." Italics not in the original
37
Having fewer speculative features, bonds will not rise as
fast in a boom market. Being a claim on assets, they have
not, on the other hand, descended so low during depressions
as have stocks. Bond prices are not, thus, as sensitive in
amplitude as are stock prices. This relation is apparent
in Figure 3 in which bond yields have been substituted for
bond prices. Such a substitution is commonly made for pur
poses of comparison as bond yields closely follow bond
prices and are of greater prediction significance*
Quite as important as the changes in the absolute
level of bond prices has been the meaning behind such
changes. Bond prices may, for example, change several
points at certain phases of the cycle and attract little
attention thereby. Again, a change of but a fraction may
be of critical interest and importance and can be the in
stigator of important change in other segments of the mar
ket.
As pointed out by Haney and others, the trend of
the prices of bonds is one of the major indications of
change in stock prices.32 There are several reasons for
32- Haney, op, cit.. p. 30&, Haney emphasizes the
general dependability of this characteristic. Also see,
Leonard Ayres, Turning Points in Business Cycles.(Macmillan,
1939)» P* Burns notes here that in twenty-five observed
cycles of bonds and stocks the former precede the latter to
peaks by an average of 5.3 months and to troughs by an
average of 2.1 months.
3 8
FIGURE 3
STOCK PRICES & BOND YIELDS
1875 - 1935
Z 3 Q
230
96
/30
80
30
4‘ S 20 BONDS SOI
l « % % % » % i \
Based on Babsonchar t, Babson*s Reports Inc*
2405, U niv ersity Bookstore, Los A ngeles
39
this, the two main ones being (1) prevailing money rates
have a definite effect upon the prices of bonds and exert
an influence over the stock market, and (2) it is purchase
of securities for investment that affects the bond market
and, though not commonly emphasised, purchase of stock
reflects investment as well as speeulatory motives. Bond
price trends therefore, are of use to the forecaster as a
check on stock price movements* They have been used separ
ately, each receiving independent interpretation, and as
cheeks upon each other.
Though bond prices customarily lead stock prices,
historical evidence shows the amount of the lead to be both
variable and erratic. The bond market, for example, has
almost always led a major downswing of stock prices. The
amount of lead has varied from approximately two to eighteen
months. The forecaster of business activity for the short*
run could ill afford to place heavy weight on bond prices
as an exclusive indicator. Using the extreme lead, that of
eighteen months, a past prediction of a downturn in business
activity could very easily have been in error by two yeara
The lead of bond prices over stocks has ordinarily been
between four and eleven months, however, and more often of
real prediction value than not,
.Another interesting feature concerning the relation
ho
between bond and stock prices has been their almost simul
taneous recovery from slumps. This phenomena is noted by
Haney and is evident in the charts of Macaulay.33 From the
forecasters point of view, the significance of this rela
tion is but little unless his intention has been to forsee
an upswing in stock* The business activity forecast would
be less affected as it ordinarily experiences a lag behind
stock movements*
Numerous other relations have been noticed which
relate, directly or indirectly, bond prices to conditions
of general business. It seems best go forgo discussion of
them at this point in favor of discussing the other and
equally important bond change indicator, the yield.
The significance of the yield on high grade bonds
to business vigor has been historically great. The yield,
and reference is here made to average yields, is perhaps
the best basis upon which to estimate the rate at which
fixed incomes are being capitalized. A high rate on
securities of this class has customarily been a bad sign
for business health, and understandably so. It is indica-
33 Macaulay, op, cit.. p. 218. As Haney points out,
the ma^or interest of this Htrough“ characteristic is in
relation to forecasting changes in,general stock prices.
tive, among other things, of high interest on real-estate
and industrial development loans and has often proceeded
a contraction in these fields. A low rate, on the other
hand, is customarily regarded as an encouraging sign al
though this has not been invariably so. Such a rate,
whether high or low, should be classified as a long-term
rate and is subject to severe limitation for forecasting in
the short-run.3 ^ 4 -
The yield on bonds is of more use in another connec
tion, however. They are valuable as a base on which to
evaluate prevailing money rates. It has often happened
that a discrepancy between the average yield on high grade
bonds and the current rates on commercial-paper has been a
presager of violent adjustment. As Haney has put it, "When
time-money and commerei al-paper rates are much above the
yield on bonds, they are high, regardless of what their ab
solute percentages may be, and vice versa."3^ Such malad
justment then, may well be indicative of impending reaction
in one direction or the other. A chart of inverted bond
yields is a graphic device used by many forecasters and,
properly interpreted, can give valuable clues concerning
future business trends*
Conclusions, Bond prices and yields have a better
than average dependability record as indicators of change
in the securities market. Their value for the forecasting
of general business conditions can also be established but
the case is stronger for the long-run than the short-run.
Deficiencies of bond yields and prices as indicators
for the short-run forecast include:
1. Bonds, especially bond yields, are rather in
sensitive to the fluctuations of business activity which
are shorter than the "three-to-four year American cycle”.
This constitutes a serious shortcoming for short-run fore
cast purposes,
2. Bond yields are identifiable as long-term rates
and as such are often difficult to relate to the cycle.
3. Understanding of the meaning behind a bond price
or yield change is often more important than the absolute
value of same. Such changes, unfortunately for prediction
purposes, are due to government manipulation and other non-
market forces. This constitutes a serious handicap to all
but "inside" or "on-the-spot" forecasters.
k3
III. BANKING MEASUREMENTS
By far the greater part of the business trans
actions of the nation are handled through banks. Statis
tics pertaining to the operating characteristics of finan
cial institutions have, accordingly, been valuable indica
tors of the level of business activity. Gertain aspects of
banking activity, interest rates and loans in the security
markets, have already been treated and will purposely be
avoided from the considerations of the ensuing section.
Bank demand deposits. The level of deposits in
commercial banks have an indirect but important use in busi
ness forecasting. It has been common practice to relate
the amount of demand deposits with the reserves behind them
and, evaluating the ratio, determine the general condition
of the banking system. The logic of sueh comparison in time
of crisis is obvious but its usefulness is yet broader.
A low ratio has usually been indicative of a sound
banking system with a satisfactory margin of safety. To
the forecaster the meaning has been favorable as it will, in
most eases, be accompanied by generous loan policies and low
interest rates. If expectations are optimistic, the basic
ingredients making for prosperous business conditions are
assembled and the outlook, ceteris paribus, is bright. A
high ratio, on the other hand, may be followed by a calling
of loans, a raising of the interest rate on loans, and a
k k
contraction or consolidation of loan volume,
Bratt mentions that some forecasters have depended
heavily on hanking figures as a basis of prediction and
have, in this connection, often used demand deposit series.3&
Demand deposits have more often than not led general busi
ness activity and their level, as we have here seen, can be
symptomatic of important basic conditions affecting the
economies’ health. One would have often been in error in
the past, however, had he depended on the level of deposits
alone. The source of funds composing demand deposits can
be fully as important as their amount. It is for this
reason that the forecasting value of demand deposits has
been questioned on a theoretical basis.37 They are, never
theless, an indication of the spendable funds of the nation
and are of current if not prediction value.
Bank debits. Another of the banking figures which
has received the attention of the forecasters is bank
debits. Debits are entered on individual accounts for each
check paid and constitute statistically useful and readily
available data. The supposed value of bank debits for fore-
36 Bratt, o£. cit., p. I 4 . 32.
37 Loc. cit.
casting has been as an indication of payment for goods and
services. Unlike clearing figures, each bank debit repre
sents a check against a personal account and is, hence, free
from much of the duplication present in the former*
Bank debits in raw form have been of little value
to forecasters for a number of reasons. For one thing,
money values have no direct relation to the quantities of
the market place. It is difficult to compensate for this
handicap as the composition of debit balances is so com
plex. In addition, many debits are purely "financial
items' 1 in spite of the fact that only individual accounts
are involved. This occurs due to remittances between
branches of the same firm and in connection with security
and real estate dealings*3®
Another limitation of the bank debit as a forecast
ing device has been the remote relation between debit
balances and known volume of sales. On several occasions
the annual volume of debits has been ten-fold that of the
"gross national product".
Albert G. Hart, Money Debt & Economic Activity
(Hew York: Prentice-Hall, Inc., lgljB"), p. 161. In regard
to debits Hart states, "But ’debits’ do not directly measure
payments for goods and services. They include many purely
’financial* items— including remittances between branches
of the same firm, duplicate payments In connection with
security and real estate dealings, check cashed, and so
forth* The dollar amounts of debits is so much above the
dollar volume of known sales of goods and services that we
can only rub our eyes and assume that most of the ’trans
actions’ recorded by debits are fictitious^"
Without correction, then, bank debit balances as
released by the Federal Reserve Board and the Survey of
Current Business are but poor approximations to the rate
of domestic payments. It is necessary to seasonally ad
just original data and make allowance for speculatory
tendencies affecting the dollar volume and value.
Thus corrected, debit balances have exhibited a
tendency to vary with general business activity. Further
more, their weekly publication gives them a definite "time1 1
edge over much other indicator data. Compared with bank
demand deposits, debits have long been used in calculating
"turnover of bank deposits". The velocity figures thus
obtained have, interestingly, failed to substantiate the
quantity theory of money.39
Bank clearings. Of somewhat less forecasting value,
but still of interest, have been the weekly figures on
bank clearings as reported by Bradstreet1 s. Dun*a Review,
and other periodicals. Being money figures, clearing
balances have the same shortcomings as debits. In addition
O Q
Wilson Wright, Forecasting for Profit, (New York:
John Wiley & Sons, Inc., 19^7), p. 50. This does not, of
course, raise question of the quantity theory of money as
the latter is a truism. Rather, it shows the faulty nature
of present velocity formulations and/or statistics.
1 4 - 7
they have several Inadequacies of their own. One of the
worst of these is the duplication effects of multiple
clearings of a single check which tend to exaggerate bank
activity. A contrary affect will be introduced by checks
which may not be cleared at all such as payroll checks.
In view of these and other errors, levels of bank
clearances must be cautiously employed by the forecaster.
The raw data, as in the case of bank debits, is customarily
adjusted so as to eliminate unwanted characteristics. A
correction has ordinarily been made for seasonal variation
and, most important for this particular indicator, the data
is adjusted for speculatory movements. The latter may be
well compensated for by eliminating the clearing figures
of banks in the large financial centers. So adjusted, and
with allowance for the changing value of the dollar, clear
ing volume has a fairly respectable record of variation
with other indexes of economic movement. See Figure i j . ,
following page. It exhibits a lag with respect to business
activity and this characteristic, plus that of its complex
ity, makes it generally unsuitable for mechanical short-run
forecasting.
Commercial Loan Volume. Still another of the banking
figures of forecasting value is the volume of commercial
loans. These are listed in the “other loans" category of
48
FIGURE 4
BANK CLEARINGSV BOND YIELDS* AND PIG
IRON PRODUCTION IN THE U. S.
1857 - 1935
(Vertical Scale Logarithmic)
AMEIJCAN S.R. 1
(Inverted)
ONDS
BANK
TSID1 01
SANK CLEARINGS
::n m r York city
CLSAR-
; bgs outside n . y
(DefLated)
pig : : ron iroduc
IN THE UNITED S
(Trend A < 3 juste
1860 1870 1880 1890 1900 1910 1920 1930 1940
Based on Macaulay, Chart 22, p. 223.
Federal Reserve Bulletin and hence, are readily and period
ically available. The particular merit of commercial loan
volume has been supposed to rest in the fact that it repre
sents business, not speculatory, loans.
In actuality, loans of this type have had but in
direct relation to changing business conditions. Like
clearing volume, commercial loan volume has a tendency to
lag general levels of business activity. Their main wortl*
in consequence, would seem to be as an index of demand of
funds for non-speculative purposes. Some of the mechanical
forecasting techniques find such information of use in the
evaluation of conclusions arrived at by use of more sensi
tive indicators. Other than this, application has been
limited.
Conclusions. Banking activity measurements are in
wide use and have the valuable common asset of being rapidly
compiled and widely distributed. Forecasters usually find
it necessary to adjust and correct the raw data as published,
but are fortunate to obtain statistics so well categorized
and concerning so many items.
Even with such marked attributes as these, banking
information has definite handicaps as a forecasting indica
tor. Those which are particularly to be noted by the short-
run forecaster of business activity include:
50
1. Banking measurements, as a group, exhibit a
definite lag behind movements of business activity. Their
main use has been as a general indicator of money and credit
conditions ,
2, The dollar value of the different bank measure
ments may tell little in itself. The forecaster must
determine the reason behind changes in balances and amounts.
3* Some of the banking measurements, such as bank
clearings and commercial loan volume, seem to have but in
direct relation to changes in business conditions. In this
respect they must be regarded as insensitive even though'
they fluctuate over considerable ranges in response to
other conditions.
IV. NATIONAL INCOME AND THE GROSS NATIONAL PRODUCT
The monetary indicators thus far discussed in this
chapter are alike in that each represents a level of activity
in some segment of the economy. They are of value to the
forecaster as he can, by inferring relationships, estimate
future change. The indicators of the present section are
somewhat different in that they are aggregates of the money
payments- made within the economy. As such they are some
what more complex and, in the opinion of many current
Si
writers,
4. 0
offer a much improved basis upon which to base
forecasts.
Before entering into a discussion of the nature and
use of the Gross National Product, (GNP), one should have
well in mind what is, and is not, included therein. It
should first be stated that GNP is a derived quantity.
National Income data as supplied by the Department of
Commerce is ordinarily increased by an amount equal to the
current consumption of durable capital such as producers*
equipment. The resulting figure is the dollar value of
GNP as such is calculated by the National Bureau of Econ
omic Research.
There are, however, two different formulations of
GNp in use, the other being the ’ ’ gross national produet at
market prices1 1 of the Department of Commerce. Kuznets es
tablishes the differences in the following words:
Thus, while at the final product level, national
income or net national product is the sum of the
flow of goods to consumers and net capital forma
tion, and our gross national product is the sum of
the flow of goods to consumers and gross capital
formation, the Department of Commerce's gross nation
al product at market prices is the sumi of (a) the
flow of goods to consumers, minus government ser
vices to consumers, (b) gross capital formation under
private auspices, and (c) all government expenditures
Cf., p. S7>
for commodities and services.^* (Dr. Kuznets is a mem
ber of the Research Staff of the National Bureau of
Economic Research, Inc.)
The reason for gross instead of net concepts is of
particular pertinence to the present study* Gross figures,
it has been contended, are of superior value in determin
ing changes in the volume of business activity in the short
run. Kuznets states the advantage as follows:
The reason for this deliberate duplication is
that, in practice, the distinction between the need
for durable capital for replacement and the demand
for durable capital for additions is quite tenuous,
in the short run* Within a relatively short period,
the capacity of an item of durable equipment is
elastic; and in few, if any, items does physical
deterioration compel replacement, leaving no discre
tion to the entrepreneur. If, therefore, we wish
to understand short term variations in the flow of
durable capital, we should measure it gross rather
than net, since short term decisions, whether of
private or public entrepreneurs, are more likely to
be in terms of replacement and additional demand
Commerce and National Bureau of Economic Research defini-
ings (New York; National Bureau of .Economic Research, Inc.,
I9 P '), p* 118.
combined than between capital for replacement and
capital for new additions*42 ^Italics in the orig:
all
For the purposes of this paper it has not been
thought necessary to make a choice between Department of
Ibid., pp. 117-18.
tions, Hie treatment to ensue Is descriptive of the overall
complexion of GNP and is not detailed enough to necessitate
accepting one definition or the other.
The composition of the gross national product. The
GNP is divided, statistically, into four major portions,
(1) personal consumption expenditures, (2) gross private
investment (3) net foreign investment and ( i * . ) government
purchases of goods and services. The money value of each
category is available in quarterly publications of the
Department of Commerce,
The first of these subdivisions, money spent on
personal consumption, is further divided Into (1) durable
goods, (2) non-durable goods and (3) services. The com
parison of the amounts in each category evidences the
pattern of consumer spending at any given time. In addition,
one may compare the total of consumer spending with the
money which has been spent in any of the other three main
parts of GNP or the smaller subdivisions thereof.
The main use to the forecaster of a break down of
personal consumption expenditures would seem to be that of
estimating future demand for "wage-goods". Year after year
the expenditure of this class have bulked much larger than
any of the other components of GNP, ranging from over one-
half to approximately four-fifths of GNP itself. The great
A
bulk of purchasing power of the nation will evince change
through the figures of this item* The value to those es
timating prospective business activity need hardly be
emphasized.
The second main part of GNP, gross private domestic
investment, is also subdivided into three parts. They are
(l) new construction, (2) producers* durable equipment, and
(3) change in business inventories. Here are presented the
figures indicating the current volume of capital formation.
It is this item, it should be noted, which makes for the
main difference between GNP and the net national income or
product* Being gross figures, the values of this category
do not allow for the depreciation and replacement amounts
taken into consideration in the compilation of the latter.
Net national income, in short, allows for net capital form
ation only,^-3
Past records reveal that the year to year variation
of gross private domestic investment is very great. In
relation to total GNP, such investment has been as low as
a percent and one-half in 1932 and as high as twelve percent
in 1 9 i | . 6 . The data tends to bear out the dislocation role
commonly assigned the irregular rate of heavy capital
I*3 of., p. S3.
55
formation.
Hot only have gross private investment figures been
at wide extremes over the years, they have frequently in
creased or decreased by one-hundred percent in successive
years. Here, then, is a component of GNP that changes so
rapidly and to such degree that it must be given careful
consideration in even the shortest of forecasts.
The third major division of GNP, net foreign invest
ment, has been given separate classification because it is
the only category dealing with other than domestic payments.
Its percentage part of GNP is less significant than most
of the sub-divisions of the other major parts. It has or
dinarily composed less than one percent of the total prod
uct and, with "change in business inventories", is the only
category which is frequently of negative value. The range
of variation lias been, historically, very great. The signi
ficance of the variation is relatively small, however, due
to the slight share of foreign investment in total G^P. it
would seem that most forecasts could neglect foreign invest
ments entirely and yet introduce but little error. This
would be especially true for the short-run situation.
The final major part of GNP is government purchases
of goods and services. It is subdivided into two classifi
cations, federal and state and local. The "federal"
category is still further subdivided into (l) war expendi
tures, (2) non-war expenditures, and (3) government sales,
both domestic and foreign* Total “federal" expenditures
are calculated by adding war and non-war costs and diminish
ing the sum by the amount of government sales* This, then,
is the only net amount in GNP, excluding the negligible 1
foreign investment item.
The second subdivision of government purchases is
entitled "state and local" and represents all of the expend
itures of the governments subsidiary to the United States
Government. It is by all odds the most stable item in the
entire GNP having increased by only thirty percent while
the GNP total doubled during the years 1929 to I9I 4 . 6 inclu
sive.
The percentage part of state and local expenditures
in total GNP is relatively small being approximately one-
twentieth* The tendency has been for the proportion to
become progressively smaller, moreover, as the dollar value
of GNP has increased at a much faster rate than have state
and local expenditures.
Though one of the lesser components of GNP, payments
to these lower levels of government are large enough to
warrant the forecaster*s attention. It is difficult to see
how they could po3e much of a problem, however, a s , , by and
large, they are a known quantity and tend to change but
f ?7
slowly.
Use in forecasting. The strong point of GNP figures
for forecasting purposes is their inclusiveness. They are
representative, as is no other single group of data, of
the total activity of the eeonomy. By analyzing the various
components of GNP the forecaster sees the existing pattern
of expenditures and can hope to isolate enough meaningful
relationships to enable accurate prediction.
Of course the purpose to which GNP figures are put
can be as varied as are the needs of the forecast. One may
forecast for a single industry or for the economy as a
whole. Again it may be desired to estimate the growth and
future size of the economy for some distant year. This
would mean estimating the size of the GNP for the selected
year and, in fact, is a kind of analysis frequently con
ducted in connection with "full employment" surveys.^4-
The type of forecast of concern here is, of course,
that of business activity. GNP has been employed in this
respect for some time and with varying results. Bratt
states that false predictions of the kind made thus far on
^ Bratt, o£. cit., p. I j i j . . As Bratt points out, such
surveys are usually focused upon a period several years in
advance and is, hence, of much longer range than the fore
casts of interest to this thesis.
$8
a basis of GNP analysis can now be avoided due to recent ‘
and more comprehensive methods.^ He does, furthermore,
give an outline of what is alleged to be not only a superior
forecasting program but one which offers hope of accurate
results. As Bratt introduces it:
A program is outlined in this section which
promises to produce moderately effective business-
cyele forecasts. If this i3 true, it will indeed
mark a distinct shift from the mediocrity of past
results . ............................. .
The belief that business-cycle forecasts can
now be reasonably adequate, in the face of such an
unsatisfactory showing in the past, is founded on
the following considerations: (1) employment of a
pattern-of-relationship framework which makes possi
ble the tracing of the influence on total activity
of any particular factor and the cross-checking
of various assumed relationships; (2) substantially
improved data which purport to measure some of the
most important cyclical processes; and (3) public
interest in developing any other measurements neces
sary to make adequate cyclical forecasts,^
The Bratt forecasting procedure is an analysis of
five distinct steps: (1) Listing of limiting and rein
forcing forces in parallel columns, (2) Establishing pre
liminary models of the formation and distribution of GNP
for the forecasted period, (3) Checking for conformity with
other patterns of relationship, ( I j . ) Tracing the effects of
the limiting and reinforcing forces implied in the forecast,
ibid.. p. 438.
^ Ibid., pp. 437-39.
S9
and ( f>) Confronting forecasts with actual developments,
conducting surveys, and developing new data.V7
While this process is not the only method by which
one might prepare a forecast from GNP, and though it may
not prove to be the best, nevertheless it does seem that
any successful analysis would have to take the steps men
tioned by Bratt by one method or another. They might be
considered as the minimum number as any fewer would neces
sitate taking less than full advantage of the theoretical
possibilities of GNP. Let us then look at the shortcomings
and pitfalls of GNP forecasts as an order.
An unavoidable hazard, it would seem, would be that
of "... choosing the method most suitable for indieating
the presence or lack of causal relationship."^ j f the
various GNP components can be found to bear causal relation
with one another or, indeed, with GNP itself, an important
step will have been taken. In practice it would seem that
establishing such relationship would be most difficult and,
if accomplished, might be but transitory in nature and hence
of little use in the preparation of subsequent forecasts.
^7 ibid.. pp. J 4 . 39—1 | - 3.
Stanley Lebergott, "Forecasting the National Prod-
uet," American Economic Review. Vol. 3f>> March, I9l 4 . f i > > p. 73.
6o
Her© the forecaster will have no check upon the validity
of his methods other than the future he wishes to predict.
Having assigned relation, the forecaster must cal
culate both magnitude and direction of change to be ex
pected in the GNP components over the period of the fore
cast, Summing up the various amounts of payments thus
derived, the future GNP is determined. It may be so
different from the amount anticipated that the forecaster
will deem it wise to re-evaluate his whole scheme. A con
stant check and recheek of assumption against assumption
typifies the entire analysis.
There are, in addition, certain shortcomings of
forecasting by use of GNP which technique alone cannot
eliminate. These are the characteristics of GNP data
itself. The figures are always several months old when
published by the Department of Commerce and are in current
dollar terms rather than scaled to a price base. This
necessitates deflating or inflating the money amounts with
reference to a current price index if the physical signifi
cance of the components is desired. Short-run forecasts
may not require this correction as the money values, for
short periods, may exert more influence on the business
scene than the basic economic forces which they more or less
accurately reflect.
The representativeness of the components of GNP as
6 l
indexes of current business activity is not beyond question*
As Kuznets cautions:
The usefulness of each of the many gross national
products that could be defined and measured ...
lies in the validity of the assumption that the
sectors selected as strategic and best understood
in terms of their gross activity are indeed deter
minants of changes in total output and better studied
in terms of gross than of net national product.49
The forecaster, it would seem, should not dwell upon
GNP values to the exclusion of other currently available
data of perhaps equal importance. Should not, for example,
the payment to agriculture bulk larger in some estimates
than net foreign investment even though the former is ex
cluded from the GNP tally?
Statistical criticism might be levied at length
against Gnp composition. Such would be to little avail,
however, as many imperfections continue due to a lack of
financial means rather than the fact they have escaped
recognition. Prospects for more extensive information
gathering will no doubt be dependent upon how valuable such
information is commonly held to be. As Kuznets sees the
situation:
t
Likewise, statistical problems arising from lack
of data will continue to hamper the estimator until
^•9 Kuznets, op. cit.. p. 120.
62
society becomes more cognizant of the need and takes
the Initiative in seeing that they are gathered
currently. Because of the huge cost of collecting
nationwide data, they have been gathered in the past
(usually by the government) only when society be
came convinced that the problems for whose solution
they seemed essential were crucial. Consequently
the accumulation of data has lagged behind the emer
gence of problems calling for quantitative analysis.50
Although this comment was intended to refer partic
ularly to income statistics, it might well be extended to
apply to all the monetary sources with which this chapter
has been concerned.
Conclusions. The Gross National Product (GNP) is
the sum of the gross expenditures of four major areas of
the economy. As such it is probably the best available
compound of statistical data representative of nationwide
business activity.
*
The components of GNP do not, however, include all
of the expenditures made and are inexact to an incalculable
degree. Furthermore, the techniques of forecasting based
on GNP data are relatively undeveloped. As a result the
accuracy of economic prediction by GNP analysis has been
mediocre.
S° Ibid., p. 138.
63
Though, gross figures are alleged to be of special
value in determining change in the short-run, 51 {j up figures
and prediction techniques derived therefrom present special
problems as well. Prominent among those to be noted are:
1. GNP figures are several months old when released
and are quoted in current dollar terms,
2. The components of GNP may not be the most stra
tegic elements on which data is available. Some of the
components are sluggish in reacting to changes in business
activity, a serious handicap for short-run prediction usage.
3. GNP figures are neither inclusive nor exact,
involving as they do both estimates and duplications ,52
I } . , °Pattern-of relationship1 1 forecasting techniques
are relatively new and are, as yet, in process of develop
ment, The practical value of GNP data for purposes of
short-run prediction is presently conjectural.
51 cf., pp. 52-53.
52 Joseph Mayer, "Deficiencies in the Dross-National-
Product Concept as a National Measure,0 Journal of Political
Economy. Vol. £3, December, 19^5* P« 357” This is a major
deficiency in Mayer’s opinion. He states, ° . . . there are
two fundamental reasons why the gross-national-product con
cept would seem to be a defective national measure. The first
reason — that it contains certain avoidable duplications.
The second reason ... it contains two major segments that
are non-comparable ...
CHAPTER I I I
PHYSICAL ACTIVITY INDICATORS IN THE SHORT-RUN
Though many writers use an almost exclusively mone
tary approach to the understanding of forecasting problems,1
the level and direction of physical activity cannot be
ignored. Purely pecuniary indications do not tell the whole
story of economic change and one cannot hope to achieve full
understanding through study of money receipts and payments
alone.
The proper place of data relating to physical change
in the modern forecast is still in dispute as, indeed, are
forecasting techniques in general. Postwar forecasting
blunders brought forth a rash of articles which charged
neglect of relative prices and physical magnitudes as major
causes of the errors. The words of Frank R. Garfield are
typical of the sentiment of such charges:
Thus a really comprehensive approach to the fore
casting problem in any period seems to call for much
more than a statement of possible expenditures and
receipts. At the same time in a more comprehensive
approach the analysis would include much more material
on inventories, production, and consumption in physical
terms . ............ ....... ......
There is a special need for more physical volume
data on production, productivity, inventories, and con-
sumption*2 ^Italics not in the original}
Such an assertion would have been considered self-
evident fifteen years ago* The prevailing popularity of
forecasting on a basis of GNP figures and Keynesian economic
concepts now makes it a topic for spirited debate* The
present investigation is not obliged to make a choice in
order to maintain logical consistency and will not do so*
Instead, it is sufficient that data pertaining to
physical volume and value are available and are being used
as at least partial basis for the preparation of forecasts.
As such, the data constitutes an indicator within the defini
tion of this paper. The problems associated with its use
in forecasting business activity are, hence, clearly within
the domain of this study.
I. PRODUCTION
The production of goods is-* of course, the first and
perhaps the most essential step in the economic process.
If money were “neutral" and our credit institutions did not
exist we might explain much of economic change by direct
^ Prank R. Garfield, "Transition Forecasts in Review,"
American Economic Review. Vol. 37, May 1 9 f y . 7 , p. 79.
66
analysis of the production process. Many of our economic
ills are but manifestations of basic unbalance in the form
of over-production, underproduction, or misdirected produc
tion.
The forecaster will not ordinarily have the time nor
will he find it profitable to appraise current activity in
all the avenues of production. It has become common, there
fore, to subdivide productive activity into three major
groups: durable goods, non-durable goods, and agricultural
products.
Durable goods. Durable goods are of two fundamental
classes, (1) Producers* durable goods, usually in the form
of capital equipment, and (2) Consumers * durable goods,
usually found in the form of housing, automobiles, or long-
lasting household appliances.
Briskness of enterprise in the producer good field
will always be of interest to the forecaster and near turn
ing points warrants his critical attention. It is a well
known fact that production of capital equipment tends to
taper off near a cyclical peak while activity in other
sectors may still be on the increase. A changing rate of
production of producer durables may, hence, be of utmost
67
importance .3
It is also to be noticed that changes in activity in
producer durables may have significance beyond number of
units produced and the dollar value thereof. Heavy com
mitments can usually be interpreted as an indication of
optimism and is a favorable sign. Less likely, but yet a
possibility to be considered, excessive activity in producer
durables may mean severe limitation of consumers* goods and
a resultant boost in the prices thereof. The situation of
the moment will determine the significance of change more
than can a historical analogy to absolute production values
of past times.
In practice, the forecaster may well focus attention
upon production figures of the “heavy" industries. Some
which have reflected activity in producer durables to a fair
degree of accuracy include blast furnace activity, sheet
steel production, pig iron production and unfilled orders
for railroad locomotives. The variation of production levels
in these fields during the course of the three to four year
0American" cycle has been very wide. Replacement of pro
ducers* capital can be delayed indefinitely during slack
3 Estey, op. cit.. pp. 171-72. Estey points out that
the producers goods industries may be in "trouble” even
though orders are on the increase if the rate of new orders
is declining.
68
times and production in the basic industries has decended
to less than twenty percent of optimum capacity.
The usefulness of producer durable figures for short-
run prediction purposes will,,hence, change with the phase
of the cycle. By themselves they are too fragmentary to be
depended on. Regard must always be given to meanings which
the data itself may shroud.
Consumer durables are another category of the pro
ductive array from which considerable forecasting informa
tion has been gleaned. A high level of purchases of com
modities of durable nature is usually a prosperity condition
as they, like producers' durables, may be used in excess of
normal life expectancy when business activity is at low ebb.
The two most reliable indexes of consumer durable
production tempo have been automobile production levels and
construction of residential housing. The latter has been
particularly responsive to changes in business activity,
experiencing as it does a clearly'distinguishable short-
period variation as well as a longer-period movement which
is sensitive to general cyclical amplitude trends. Clarence
Long of Princeton University published the results of a
rather exhaustive survey of the statistical background of
this relationship in 19i f 0. A partial summary of the con
clusions in his words:
69
Association between building and business turn
ing points proved to be close: average disparity
was six months, with few turns more than a year
apart: and one-to-one correspondence was high, with
only two cycles in each of business aiid building
not matched by corresponding cycles in the other.
No real tendency for building to lead business on
the upturn and downturn could be found, in spite of
the fact that the number of turns in which building
led was double the number in which building lagged.
When discount is made for the fact that permits
anticipate actual building by about four months the
two to four(th) month lead of building permits
becomes insufficient to establish any effective
lead of actual building.h- ("Misspelling in the ori
ginal ? J
though a definite sensitivity is, thus, demonstrable,
absence of a lead characteristic limits forecasting value
of building activity data. This would especially hold for
purposes of the short-run forecast.
A variety of other measures are useful in fathoming
the extent of production in the consumer durable line.
Household furnishings constitute a considerable part of
consumer durable production in dollar value as do the multi
tude of electrical appliances for home use. By and large,
however, the purposes of the general business forecast
would seem well served by a survey of tendencies in the
major elements alone.
Clarence D. Long, Building Cycles and the Theory
of Investment. (Princeton: Princeton University Press, 19k0),
pp. 102-03.
70
Non-durable goods. The two principle sources of data
relating to the production of non-durable goods are the
Federal Reserve Bulletin and the Federal Reserve Bank of
New York Index. The former offers a single composite index
of non-durable good manufactures while the latter presents
tabulated data on both producers! non-durables and con
sumers 1 non-durables.
As might be expected, the cyclic variation of non
durable manufactures is much less than that for durables.
Non-durable goods as a group are more likely to be replaced
at a regular rate than are durables. The more or less
regular replacement rate keeps the production rate from
dropping as far during periods of depression while the in
elastic nature of the demand for many non-durables damps
the upswing of booms.
Nevertheless, non-durable good production varies
enough and is sensitive enough to general business condi
tions to be of use as a forecasting indicator. Producer
non-durables are more responsive to business fluctuations
than are consumer non-durables and the rate of manufactures
varies over a wider range.£ Even so it is the latter which
5 Arthur F. Burns and Wesley G. Mitchell, Measuring
Business Cycles (New York: National Bureau of Economic Re
search, 19P77p . Producers non-durables and consumers
non-durables are here graphed for comparison with several
other statistical series.
71
has been the more dependable for forecasting usage as its
dollar volume is much greater.
Major turning-points in non-durable good production
have been frequently coincident with the turning-points of
both durable good production and industrial common stock
prices. This is equivalent to stating that non-durable
production fluctuations usually experience a distinct lead
over corresponding changes in business activity. The re
lationship was especially marked during the Thirties as
the charts of Arthur Burns and Wesley Mitchell testify.^
If we look still further back, however, it appears
that non-durables have ordinarily experienced both upper and
lower turning-points before durables. It would seem that
either the Thirties are not representative of the usual re
lationship between durable and non-durable production or
that the long-standing relationship between the two is chang
ing — perhaps both. The changing lead-lag relation as well
as the relatively great variation of durable production is
clearly evident in Figure 5.
Another characteristic of non-durable production
which has been of interest to forecasters is the close con
nection It has been observed to bear with respect to the
6 Burns and Mitchell, loc. cit.
72
FIGURE 5
DURABLE AND NON-DURABLE GOOD PRODUCTION
IN THE UNITED STATES SINCE 1900
(1913 s 100)
PERCENT
DURABLE
GOODS
-140
'-120
100
non-durab:
GOODS
80
60
1905 1940 1910 1930 1935 1915 1920 1925
Based on data computed by Leonard P. Ayres and the Federal
Reserve Board.
>
73
physical volume of trade* The later might be expected to
follow the former as, after all, what has not been produced
cannot be traded. The opposite has, in fact, been the rule*
Silberling has attempted to trace the reason behind the
seemingly reversed causal relationship:
How are we to explain this tendency for producers
of goods entering into retail merchandising to vary
their operations, usually ahead of the changing phases
of total manufacturing employment and still farther
ahead of the monetary income flow into the hands of
wage earners and, indeed, of the general population?
Is it possible that consumers or large groups of con
sumers begin to reduce the volume of their retail
purchases long before it becomes apparent in the
dollar value of merchants1 sales? Is it possible that
a converse tendency develops in expanding demand after
a trade recession? May there be changing phases of
consumer propensity to spend or ability to acquire
staple merchandise that originates highly sensitized
cycles in certain branches of manufacturing before
other branches are affected? If this is true, it
would be a reasonable inference that, in view of the
highly volatile movements of such industries as tex
tiles, there would be abrupt changes in employment
and therefore in payroll income, occasioned by these
variations and that these might contribute to further
changes in demand impinging upon still other products
and ultimately upon many branches of manufacturing
and hence upon the general course of payroll disburse
ments and national income.7
Whatever the explanation might be, the sequence has
been repeatedly verified. It is probably sufficient for
prediction purposes to have acquaintance with the functional
? Silberling, o£. cit.. pp. i|i>7-68.
Ik
relation alone.
Agricultural production. Agricultural production
statistics, including anticipations of.crop size, have long
been recognized as important indicators of business activity
trends. It is difficult to adapt agricultural information
into the appropriate form for forecasting purposes however.
Farm products vary as to value between different crops as
well as to customary ranges in price and output.
Analysis of the economic meaning of given levels of
crop output requires particular skill, experience and caution.
As is well known, the aggregate dollar value of the nation's
agricultural produce may be greater in a year of general
crop failure than in a year of bumper harvests. As the
crop value is very decidedly a supply and demand phenomenon,
it is necessary for the forecaster to estimate demand be
fore arriving at the proper dollar evaluation.
Fortunately much of this information is available in
computed form. The Bureau of Crop Estimates of the Bureau
of Agricultural Economics regularly publishes bulletins
describing the amount and condition of various crops. It
further ventures a forecast as to what the final value of
the total crop will be. The predictions of this bureau
have enjoyed a relatively high degree of accuracy and have
been alleged to be quite helpful to farmers and businessman
' 7 5 ?
alike,8
Aside from the ever-present uncertainty as to crop
size and value, there exists a marked seasonality of agri
cultural data. Statistically speaking, the data must be
weighted differently for predictions probing into different
seasons. It hardly need be mentioned that government parti
cipation via parity price programs may lead to reversal of
predictions based on a presumption of a free market,
The lead and lag relationship of agricultural prices
with respect to general business activity indexes has not
been consistent. Though the turning-points of agricultural
prices have both led and lagged corresponding turning-points
in business activity, a slight lag has been more frequent.
Prices, it should be noticed, fluctuate much more widely
® Roger W. Babson, Business Barometers. (Babson Park:
Babson*s Reports, Inc., 1939)* P* 207, See also, Vladimir
P, Timoshenko, The Role of Agricultural Fluctuations in the •
Business Cycle. (Ann Arbor: University of Michigan, School
of Business Administration, Bureau of Business Research,
1930)* As Babson states, "But in the subject of crops, not
only does the government publish a report on their amount
and condition in various stages from planting to the be
ginning of harvest, but it makes a prediction for the
benefit of business interests of what the total crop is
likely to be. It has been well proved that this forecast
made by the government is better than any forecast which
at the present time can be made by any association of mer
chants or bankers independently."
76
than do prices for non-agricultural products.
A great number of valuable comparisons can be made
between agricultural data and data of other segments of the
economy. Such comparison, as a matter of faet, is probably
the most effective use to be made of farm information as an
indicator. Two important points of any such analysis would
seem to be: (1) To consider the available supply of crops
on the market, not current production alone, and (2) To be
constantly informed as to the relative prices of agricul
tural products rather than absolute value alone.
Finally, the regional bias of agricultural fortunes
should be stressed. Though crop conditions and prices may
.be of little immediate interest to large financial and
manufacturing centers, they are of critical import to busi
ness prospects in rural areas. The general forecast must
allow for this diversity of interest. It would be easy,
for example, to formulate an overly pessimistic forecast
in light of crop failures which constitute little threat.
The condition might be too localized to have broad signifi
cance or else be nullified by ample commodity reserves or
government action. Here, again, the necessity of having
knowledge of the market is apparent.
i
Conclusions. Physical production levels as well as
the dollar values representing same have been observed to be
77
sensitive to general business activity. The balanced fore
cast will seldom ignore production factors.
Production data does, however, have certain marked
deficiencies as an indication device of economic prediction*
The principle one would seem to be that physical volume
alone cannot measure production value. Other shortcomings
of special interest to short-run forecasting:
1. The lead or lag of production data usually varies
with phases of the cycle. The amount is either not con
sistent or, as is the case with agricultural prices, it is
not possible to identify lead or lag characteristics at all.
2. Variations in production levels must frequently
be fathomed by changes in composite indexes which may not
be weighted so as to show representative changes. The
alternative, that of following changes in a single series,
is subject to a similar error.
3. The meaning behind changing production levels or
prices may not be certain. Analysis of same may require
time which the short-run forecast cannot afford.
i j . . Being indicators of supply alone, the ’ ’ barometers"
of this section must be used in conjunction with those in- -
dicating trends in other segments of the economy.
78
II. TRADE VOLUME
The logic behind use of trade information as an in
dication of economic change is evident. The functioning
of the distribution process determines, theoretically, what
the producer can sell and at what prices. Interruptions
in the flow of goods to ultimate consumers presage fluctua
tions in business activity. This will be so whether the
stoppage is physical or is induced by inflexible prices.
Indicators of trade activity are numerous. This
section will treat them generally under four headings:
(1) Domestic, (2) Foreign, (3) Retail prices, and ( i j . ) Whole
sale prices.
Domestic. In addition to such monetary indicators
of trade activity as were considered in the last chapter,
there are several well known indicators which stress the
physical aspects of the trading process* Although some of
these are expressed in the form of prices and are money
amounts, they do at least represent a direct valuation of
goods and services. As such, they are much more relatable
to physical trade volume than, say, banking figures.
Probably no indicator of physical trade volume has
been more widely used by forecasters than has that of car-
loadings. In several respects car-loadings are almost per-
79
feet data for prediction purposes. They are composed of
"natural” units, the car-load, not statistically abstracted
units. They are available on a weekly basis and are classi
fied according to railroad.line and type of cargo. Here
then would seem to be a very comprehensive indicator as
almost all products involve rail shipment at some stage of
their production or distribution*
There are, however, several notable shortcomings of
car-loading data as prediction information. Haney mentions,
in substance, the following: (1) Mo allowance is made for
the different tonnage capacities of the various kinds of
rolling stock, (2) Mo allowance is made for the length of
the haul, (3) Car-loading figures are heavily weighted with
mass shipments of low relative value such as coal, and ( i f . )
Transport carried by truck is assuming an increasingly signi
ficant part of the total.9
For purposes of short-run forecasting there is an
additional deficiency more important than any mentioned
above. Gar-loadings tend to follow business activity changes
but sluggishly, there being a general trend relationship
rather than a sharp sensitivity. This plus the fact that
9 Haney, dj). cit.. pp. 85-86.
S '
80
turning-points in car-loadings lag both industrial produc
tion and general business activity would seem to relegate
the data to secondary importance as an indicator for the
short-run.
Sales volume, both retail and wholesale, is another
of the important trade activity indicators. There is no
better method to measure balance of the basic forces of
exchange than to compare physical sales volume with that of
production. Movements tending toward over or under pro
duction will be evidenced by such process as will abnormal
or subnormal purchasing power. Unfortunately for short-run
forecast purposes, sales volume shows little or no lead
over business activity.
Somewhat similar to sales volume figures are data
relating to orders or unfilled orders for commodities. Both
of these series lead production trends and business activity.
Their shortcoming as forecasting information is that they
are usually heavily weighted with durable goods and hence
are somewhat removed from the pulse of general business
activity. The indicator value would seem to be mainly as
comparative and evaluating data for more responsive series.
Foreign. Measurements of the volume and value of
foreign trade have long been useful prediction devices of
the forecaster. The Department of Commerce publishes monthly
81
figures concerning the value of commodity imports and ex
ports. Also of importance is data indicating the physical
volume of such trade as such measurement is not subject to
the distortion of changing international monetary conditions.
An obvious weakness of foreign trade data for domes
tic prediction use is the difficulty of distinguishing
whether current changes in trade levels are due to reactions
in the home or foreign economy. Both volume and value are
subject to variation by reason of such changes as tariff
or quota regulations rather than by market forces. A care
ful interpretation of foreign trade data must always be
made.
Import data must be handled with special care as it
is subject to two characteristic errors: (1) Many goods are
brought into the country of which no record is made, and
(2) Importers tend to undervalue the goods shipped them in
order to pay as small duty charge as possible. Export
figures, on the other hand, should be scrutinized so as to
ascertain the relative balance between agricultural pro
ducts and manufactures. This is necessary in order that the
forecaster may correctly foresee specific surpluses or
scarcities on the domestic market.
Generally speaking, however, it would not seem that
foreign trade movements can be given much weight in the
82
short-run forecast. Both the quantity of goods involved
and the money Value of the transactions are small relative
to total economic movements. Also, foreign trade fluctua
tions have not been closely related to business activity
change. Exports, for example, may be seen to be forty-
fourth in a list of forty-six series arranged in order of
sensitivity by Burns and Mitchell.3 -®
These drawbacks notwithstanding, Roger Babson has
»
stated:
Figures, then, on imports. exports and the bal
ance of trade, when tabulated each month serve as
a wonderful barometer for discerning present ^ condi
tions and for forecasting future conditions.Il~~
[Italics in the original]
Babson*s comment is of interest because of its unique
ness. His opinion is not commonly shared by other writers
nor is it supported by statistics.
Retail prices. One can ordinarily obtain a fair
indication of the relative abundance of the different goods
offered consumers by comparing respective retail prices. The
Burns and Mitchell, o£. clt.. p. 101. These series
and their respective sensitivities are tabulated with re
spect to average lead or lag at trough and at peak*. An
average of both trough and peak averages is also given and
is the sensitivity figure referred to above.
Babson, ©£. cit.. p. 169.
83
pattern of consumer expenditures will ordinarily experience
little change over the short-run and the proportional dis
tribution of the consumer dollar between goods of different
classes will be known. Changes in relative prices will
therefore mean either changing relative abundance or chang
ing consumer taste, The former will be the more likely in
the short-run. Of course scarcities may be induced by a
withholding of goods from the market as well as by genuine
shortages and vice versa. The forecaster needs to know the
"why” behind observed retail price changes.
It does not seem that the trend of the general re
tail price level will be of much help in the preparation
of forecasts of any period, be it long or short. Retail
price levels are quite insensitive to changes in business
activity and have frequently dipped but slightly when a
major recession was underway. The reason for such sluggish
response is probably the varying leads and lags of the
prices of different retail goods with respect to business
activity fluctuations. The prices of some commodities com
posing a retail price index may be rising while prices of
other commodities in the index are falling. Thus while the
economy may be experiencing marked adjustment the retail
price index may show little change due to cancellation
effects.
The aforementioned faults of retail prices indexes
have caused many forecasters to resort to use of but a few
of the available retail price series. The cancellation
effects are thus minimized and sensitivity is increased.
One of the. most satisfactory series has been that of
apparel and textile prices. They have the advantage of
being more elastic than the food items which usually bulk
heavy in a retail price index, and, in addition, constitute
a significant part of total retail sales value. Elasticity,
paradoxically, happens to be the major drawback as well.
The extreme variation has resembled that in the durable
goods industries.
By and large it would seem that retail prices are
one of the least useful of the price indexes. Principle
forecasting value is probably limited to the representa
tion of general stability which will be revealed by an
analysis of retail price structure* Though this might
constitute an indication of minor importance during f , firm, ,
market periods, it would be of acute interest in the un
stable market situation which usually precedes a turning
point.
Wholesale prices. Wholesale prices are subject to
some of the same shortcomings as forecasting data as are
retail prices. Index compilation difficulties, for example,
are almost identical. There are, however, several dlstinc-
85
tions to be made both as to the nature of the data and to
the forecasting usage thereof.
Of particular interest for purposes of the short-run
forecast is the definite lead of wholesale price indexes
over both retail priees and general business activity. Al
though the amount of lead has been variable, it has usually
been in excess of a month. The practical advantage of this
characteristic is somewhat diminished by the fact that
wholesale price figures are not available for several weeks
after being recorded.
Wholesale price changes have been of considerable
value in prediction of future trends for specific indus
tries, A change in wholesale prices is usually followed by
an appropriate change in the rate of production, the amount
of lag being dependent on the production flexibility of the
given industry. The forecaster of general business condi
tions may not find much guidance in this phenomenon unless
wholesale prices as a group are moving predominately up or
down. This has seldom occurred historically with the excep
tion of relatively short and sharp recession movements.
Of but incidental interest to the present paper but
nevertheless worthy of notice is the long-period variation
of wholesale prices. No other commonly used series has so
closely followed the "Long Wave” first recognized by
86
Soiethoff• Kondratieff, who later analyzed the wave, has
4
estimated that the length fluctuates between forty-seven
and sixty years.12 Neither this nor shorter wholesale price
cycles appear to be of such periodicity or regularity of
amplitude as would encourage their use in short-run fore
casting.
The main use of wholesale prices in general fore
casting work has been in the detection of overproduction
or underproduction in major industrial or agricultural
lines. If the maladjustment is extreme or extends into
many fields, a reaction may be imminent and, hence, of
prime importance to the forecaster of the short-run. More
often wholesale prices will be compared with other indica
tors in order to judge general economic stability rather
than to anticipate a turning point.
Conclusions. The various trade volume indicators
Nikolai D. Kondratieff, ’ ’ The Long Waves in Econ
omic Life”, The Review of Eeonomic-Statistics. Vol. 17, (A
translation of "Die langen Wellen der Konjunktur,M which
appeared in Archiv fur Sozialwis3enschaft und Sozialpolitik.
Vol. 56, no. 3, 1926), p* 111* Speaking of the wholesale
price cycle, Kondratieff states, H ... three great cycles
are present in the movement of the wholesale price level
during the period since the end of the 1780fs, the last of
which is only half completed. The waves are not of exactly
the same length, their duration varying between i j . 7 and 60
years.
87
Inform the forecaster of the size and/or change of the
flow of the nation’s produce into the hands of ultimate con
sumers. Bottlenecks in this distributive process will be
manifested by changes in the trade volume figures which
describe it.
Practical difficulties in compiling and analyzing
trade volume data plus the theorectical limit of it being
but one phase of economic activity somewhat mitigate pre
diction value. Some of the deficiencies which can be
pointed out as special handicaps to the preparation of the
short-run forecast are:
1* Trade volume series are notably insensitive to
short-run changes in business activity.
2. The various indexes of trade volume, both price
and physical, are quite difficult to properly weight* An
improperly weighted index will not, of course, indicate
representative changes. The short-run forecast will suffer
if too much attention is paid short-period movements in
such indexes even though the longer trend movements of the
index prove accurate,
3. Lead or lag relationship, where evident, seems to
be of but small time duration* It is not of a constancy
which assures dependability such as, say, the pre-World
War II stock market.
i f . . Causal relation between indicator change and gen
eral business change is frequently complex and is subject
to variation as a result of new consumer tastes, modes of
transportation, government regulation, et cetera.
CHAPTER IV
THE INDEX NUMBER IN FORECASTING
Index numbers have been a powerful analytical tool in
business forecasting* Original economic data may well be
in such form as to make detection of general trends most
difficult* If the data be presented in the form of an in
dex, however, it becomes possible to compare data of differ
ent time periods and to throw in relief such trend character
istics as may exist.
An index may be constructed from either physical or
monetary data* It may be composed of data relating to a
single product or price or have several hundred components*
The index structure, further, may give equal weight to each
component or may weight each in accordance with supposed
relative importance* All of such forms, as well as combina
tions thereof, have been used in business forecasting work.
The index number must be carefully used as an indica
tor of change* It can mislead or be misrepresentative to
the unwary forecaster* The potential dangers of the device
are of particular concern in the short-run. This chapter
endeavors to describe index number limitations and pitfalls
In relation to the objectives of the present thesis.
The danger of trend extrapolation. The practical
use of index numbers in forecasting work makes desirable,
where possible, projection of past trends into the future#
The implicit assumption in such cases is that future devel
opments will occur at a rate and magnitude identical to
that of the past. The assumption will almost never be
valid and the forecaster, eognizant of the latent possibil
ity of error, attempts to compensate the projection so as
to allow for new faetors. It will often happen, of course,
that the forecaster has no way of knowing what new factors
will be at work. In such cases it has been common to pro
ject, or extrapolate, past trends into the future with no
modification. Frederick Mills has cautioned quite con
cisely:
The fact should be clearly recognized that pro
jection, or extrapolation, represents a guess, justi
fied only on the assumption that a proper line of
trend has been fitted and that the same conditions
that affected the series in the past will prevail in
the future# A change in conditions, the introduction
of new elements, renders the projection invalid.1
The method of extrapolation differs according to the
accuracy desired. For some purposes it may be permissible
to simply sketch a projection of a graphically presented
1 Frederick G. Mills, Statistical Methods (Hew York:
Henry Holt and Company, 1938)* P. 278*
91
trend into a future time period. Again the forecaster may
think it necessary to "fit” a curve to past trend data,
derive the formula by mathematical means, and mathematically
project the ’ ’ fitted1 ' curve into future months. In this
connection it is interesting to note that some authorities
have visualized the curves of economic fluctuations as a
composition of different sine curves. If the wave lengths
of the different sine curves can be found and the amplitude
determined, forecasting becomes no more than a mechanical
manipulation. As King puts It:
All that Is necessary is to produce each sine
curve independently, and then add together the read
ings from the various sine curves at the date for
which the forecast is sought. The sum of the read
ings will approximate the figure desired.2
It may be seen that the task might become most
formidable for even a single series. Unless simplified
from the complex conditions of reality, an analysis and
prediction of general business activity by this method
would probably be impossible In any allowable period of
time.
King, op. cit.. p. 2 5 i | - . Also see, 0. W. Blackett,
A Method of Isolating Sinousoidal Components of Economic
Time Series. . fUni veir si tv of Michigan School of Business
Administration, 1933)* It should be recognized that the
extrapolated curve of aggregated sine waves need not be of
regular amplitude or period. A rather irregular curve will,
In fact, result due to the differing periods, phase rela
tions and amplitudes of the component sine waves.
92
The unmodified extrapolation not only assumes that
no new factors will be introduced but also that the old
factors will continue in like relation to one another*
Woytinski, well aware of the assumption, suggests a pro
cedure for evaluating the worth of a projected curve:
... formulas will defy any extrapolation unless
it is assumed that the interrelation of observed
features will remain the same in the future as in
the period surveyed. In practice, after having ex
trapolated such formula, one must use his judgement
in deciding whether the result is good.3
Not only must the forecaster be concerned as to
constancy of functional Interrelationship, but also must
he consider whether or not such relations are causal. A
close correlation between various functional relationships
is not ordinarily sufficient to establish causal connec
tion. Otherwise the forecaster lays himself open to redlcu-
lous error as Woytinskir' S analogy describes:
Nearly any time series may be expressed as a
function of other time series; excellent fits may
be obtained between such series as the number of
visitors in our national parks and the number of
automobile accidents in Australia or divorces in
Paris. Unfortunately, these functions are purely
descriptive of observed data and permit no extrapo
lation.h-
3 W. S. Woytinski, "What Was Wrong in Forecasts of
Postwar Depression?" Journal of Political Economy. April,
19i}-7» P* llf 8. Woytinski here refers to the formula of a
"regression" equation which, being fitted to historical
data, allows for no relationships of the future not present
in the past.
^ Loc. cit.
93
At times the forecaster may have reason to believe
that past trends will experience new form in the future due
to anticipated new factors or changes in the interrelation
ships of previous factors. He will accordingly adjust the
extrapolated projection. Even so the projection is sub
ject to error by reason of new factors not anticipated and
by changed interrelationship of old factors which cannot
be forseen.
Finally, and of special importance to the short-run
forecast, a projected curve will almost always be a "smooth”
curve and will not show fluctuations about the general
trend. These shorter variations may be of utmost importance
to businessmen who make commitments for but a few weeks or
months in advance. The danger, if not temptation, will
always be present to "smooth1 * out the very variations that
should be predicted.
The composite index. As a single index can seldom
be depended upon to accurately indicate the future state
of business activity, the forecaster has come to use highly
complex Indexes composed of many constituent indexes. By a
proper selection of significant series and by correctly
weighting each in proportion to relative importance, the
econoraist-statistician formulates a composite index theo
retically superior to any single index for prediction
purposes.
Here, truly, is a job for the technician. A rare
combination of knowledge of economic affairs and a mastery
of advanced statistical and mathematical methods are called
for. Further, even the ''ideal" index of today will be in
theoretically, in constant state of flux and the "ideal"
index would be in constant need of revision.
than a hazy standard of perfection. Statistically, it
would meet the time-reversal test as does Professor Fisher's
Ideal Formula and, in addition, would serve equally well as
an index of price or quantity changes — the faetor-reversal
close to meeting these qualifications, indexes of general
business activity least of all.
commodity in the index
q signifies the quantity of that
commodity exchanged.
See Chapter XI, Irving Fisher, The Making of Index Numbers
(Boston & New.York: Houghton Mifflin Co., 1927). Also, William
L. Crum and Alson C. Patton, Economic Statistics (New York:
McGraw-Hill Book Co., 1928), pp. 296-97.
need of adjustment tomorrow. Relative weighting is,
In practice the "ideal" index is probably no more
test.5 Certainly few of presently used indexes can come
5 Professor Fisher's Ideal Formula "353" meets both
of these tests:
Where:
i refers to the given year
o refers to the base year
p signifies the price of a particular
The forecaster of general business conditions may
elioose to give heavy weight to the several currently pub
lished indexes of business activity or may prefer to con
struct an index of his own. In case the latter choice is
made, no less than five major steps will be involved: (1)
A selection of a representative group of statistical series
must be made, (2) Sources of data must be carefully inves
tigated, (3) Selection of a base year or base period must
be made, ( I } . ) The method of combining the data must be
decided, and (f?) A system of weighting must be evolved.
Error— major error--may be introduced at each and every
step because of unavailable data or other reason beyond the
forecasters control#
No matter how judiciously an index has been con
structed it must be experimentally tried before being
depended upon. The mathematical tests referred to above
do not give positive confirmation of the worth of an index.
There is no substitute for a trial and error test of a new
index in the capacity for which it has been designed. The
structure and composition will be the artistic creation of
the economic statistician as there Is no formal methodology
for such formulation. As Davis and Nelson describe the
task:
As a matter of fact, the practical application
of the theory of index numbers rests in a very
fundamental way upon the actual data. For the
solution of the problem of how to obtain desired
data and how many items to include, no mathemati
cal formula, of course, exists.&
An index designed to enable short-run forecasting
of business conditions would, ideally, have certain charac
teristics which an index designed for longer periods would
not, or need not, have. Criteria of design should include:
(1) The statistical series selected should be responsive
to minor fluctuation in business activity, (2) Component
data should be available on a weekly or monthly basis.
Semi-annual or annual data are likely to give undesirable
time bias to the index, (3) Other things being equal, data
requiring little or no processing is preferable to that
which must be deflated, deseasonalized or statistically
separated from a broader classification. Adherence to
these criteria would give, respectively, the following
characteristics, (1) Sensitivity to minor business activity
fluctuations, (2) Up-to-the-minute responsiveness to the
changing complexion of the business scene, and (3) Both
economy and dispatch in the preparation process.
^ Harold T. Davis and W. F. C. Nelson, Elements of
Statistics. (Bloomington: The Prineipia Press, 1935)#
Inherent limitations of the index number* Most
texts on mathematical and business statistics make a point
of stressing the limitations of the index number as an
analytical device. Index numbers, it is held, are subject
to a number of weaknesses both mathematically (theoreti
cally) and in practical application. An index number
designed to represent the price or quantity changes through
time of but one commodity will evince— and this is the
simplest case— both imperfections. Errors compound rapidly
as many different index numbers are massed to form a
multiple-commodity or composite index of the kind used in
forecasting general business conditions. Though mention
has here been made of some of problems of composite index
number construction, there are other inherent failings of
the index which structural refinement alone cannot eliminate.
First, and most obvious, index numbers suffer the
bias common to all mathematical averages; a single fre
quency of extreme value exerts more than proportional in
fluence upon the result. Although a number of statistical
techniques have long been in use which tend to minimize
errors of this type, they cannot be completely avoided and
will be considerable if the value3of the data happen to be
dispersed over a wide range. Professor Fisher’s The Making
of Index Mumbers gives careful treatment to this and similar
98
biases of the index number as an average.? This remarkable
book, as a matter of fact, remains today, as it has for
many years, the definitive work on business index numbers.
A second group of index number defects are also
primarily mathematical in character. Included would be
such shortcomings as have been alluded to throughout the
discussion of this thesis; the problems of proper weighting,
proper selection of a base period, selection of responsive
and causally related series and achieving appropriate
balance between monetary and physical series. Of course
problems of this order have structural aspects as well as
involving inherent index weaknesses.
Finally, and in a sense the most complex of all, one
may ask, ”Ihat does a change in the value of an index num
ber signify?1 1 . A substantially correct answer would be that
changes in index number value signify relative changes in
the prices or quantities with which the index deals. Such
an answer does not stress, as it should, that such changes
are not entirely value changes inasmuch as value Is not
entirely determined by price and quantity. As Edward Lewis
? Fisher, o£. cit.. see especially, Chapter V,
’ ’ Erratic, .Biased, and Freakish Index Numbers”.
99
states:
In a word, we may not analyze the change in
value into a price component and a quantity com
ponent which together determine completely in a
numerical sense the change in money aggregate.
Rather, we may compute indexes of the effect of
two important but not exclusive factors in that
change.°
This distinction may be proved very important for
an index composed of but a few commodities, the error chance
being inversely related to the number of commodities in
the index.
Though the various faults of index numbers cast a
measure of doubt upon any forecasting technique which uses
them, the inadequacies are especially marked in the short-
run. Indexes are composed of data which is gathered over
a considerable period of timej a week, a month, often more.
Such periods are rather large in comparison to the brief
span of the short-run forecast, say six months. Almost
unavoidably, therefore, the index number will be but a
crude gage of prospective change in the short-run. As
Irving Fisher remarked in regard to a similarly disappoint-
O
Edward E. Lewis, "Some Basic Problems in Index-
Number Theory" Economic Essays in Honor of Wesley Glare
Mitchell, (New York: Columbia University Press, 1935>),
pp. 273-714-.
ing index number characteristic:
No clock can keep time to the second if it jumps
only once in a minute, or once in an hour. Such
a clock must invariably be in error most of the
time, although, from the clock itself, we cannot
say how much.9
9 Pisher, op. cit., p. 115.
CHAPTER V
ERROR INTRODUCED BY RANDOM OCCURRENCE IN THE SHORT-RUN
The most elaborate forecasting scheme may be produc
tive of but mediocre results in the presence of what have
been termed “random disturbances". These are the accidental
and irregular variables which frequently figure more prom
inently in economic change than do fundamental economic
forces. Although no theory seeks to explain all such
variations there has been considerable research into the
nature of particular types.
Authorities differ as to how rewarding, to fore
casting, an investigation of random causes can prove. Jacob
Marschak seems to have been skeptical when he stated, "In
the presence of random variations, the problem is analogous
to that of weather prediction."1 As random variations are
always present it would seem that the foregoing comment
could be extended to apply to business forecasting at all
times and for all periods. The analogy may in fact show
economic prediction to be an easier task than it is as
^ Jacob Marschak, "Economic Structure, Path, Policy,
and Prediction," American .Economic Review. Vol. 37. Mav.
19^7, P. 81. "
University of Southern C&tffeatHe iJB S ttS !
102
meteorological forecasts have enjoyed a better accuracy
3core*
Other authorities, principally mathematicians, have
refused to accept random variations in statistical series
as completely unpredictable. Most notable in the field
of economics has been the work of Professor Slutzky of
Russia, Slutzky has identified two distinct classifica
tions of “chance series” (1) Those in which the probability
of the appearance in a given series, of a certain value of
the variable, depends on previous or subsequent values of
the variable, and (2) Those in which it does not,. 2 The
first have been called ”coherent" random variables and the
second “incoherent" random variables,3 Coherent random
variables in various economic series are claimed to experi
ence a definite periodicity and it is therefore reasonable
to believe that fluctuations of such variables are predic
table.
Prediction possibilities of such variables are no
more than proofs of theoretical mathematics at present,
^ E* Slutzky, "The Summation of Random Causes as the
Source Cyclic Processes," Econometrlca. Vol. £, 1937* p. 107.
3 Loc. cit.
103
however, There are two reasons why this must he so, (1)
The data needed to bring realism into the equations would
be the essentially unitless measurements of psychometrics
and sociometrics, and (2) Some of the required data would
have to be of a causal relevancy not yet established --
indeed this is the reason why the word ’ ’ random" is being
used.
Work of this type may well be called the "pure
science" of economic prediction and offers, perhaps, hope
that the forecasting techniques of the future may be so
refined as to permit inclusion of variables of a class
that have thus far been necessarily neglected. The pur
pose of the present chapter has been to survey, in brief,
the characteristics of some of the random fluctuations
which have in the past plagued the forecaster.
Political, All of the political factors bearing
upon economic change are not, of course, to be termed
random. Such well established laws as those governing
interstate trade and monopolistic combinations can usually
be regarded as constant or fixed for the periods of most
forecasts. Political institutions are in constant pro
cess of alteration but major changes have usually been
forseeable or have taken so long that it has been possible
to ignore them in the short-run.
lOlf
A logical division or two parts can be made of
governmental influence in economic affairs. They are (1)
Action which contributes to unbalance of the economy, and
(2) Action which tends to ameliorate the effect of unbal
ancing forces. Heither of these actions need be random in
nature nor need they be mutually exclusive, but what of
those which are?
Government, or political, influence of the first
type may take the form of heavy purchases or sales of com
modities and ean result in distorted market conditions.
If, as has here been assumed, such programs are embarked
upon with little or no advance notice they constitute an
element of surprise for which the forecast cannot have
allowed. The forecast, accordingly, becomes vitiated to
the extent of the importance of the omitted factor. Heavy
government activity of this sort has been common during
war periods but is more damaging to the peace time forecast
where it will normally be less expected. Intervention may
also occur as an effect of foreign-aid program sales of
governmental surpluses, new public works programs, et cetera.
Of course the market may be similarly disturbed due to
sudden changes in tariff barriers, shifting international
exchange rates and other political activity of international
scope.
The second type of political aetion may result in
105
unfavorable immediate effects although it may have been
intended to mitigate the consequences of recognized un
balancing forces. The' larger part of such action will be
either in the form of regulations governing market activity
or will concern monetary manipulation and/or credit con
trols. Policy enactments of such sort may be completely
unforseeable, being adopted to meet such exigencies as in
flationary or deflationary movements, runaway credit expan
sion or contraction or pure political maneuvering. In both
this and the preceding category it should be noticed that
a secession of government or political influence is as dis
turbing a random cause as is increased influence.
It Is true that forecasters can somewhat reduce
error induced by "erratic" political happenings by the
simple expedient of culturing "inside" contacts. In this
connection it is interesting to notice that practically all
of the major forecasting services are either located in
Washington, D. C., or have on-the-scene correspondents. The
increasingly important role of government in economic affairs
has caused information concerning pending legislation or
decrees to become a requisite of forecast preparation. As
one writer states:
The forecaster must live with the fact that the
government is virtually committed to much more
positive action in times of marked economic change
and to much more concern over the behavior of our
large scale private economic institutions than
the past might indicate.4 -
Though it has been suggested that there may be wave
like movements of general political atmosphere^ the best
policy for the short-run forecast would seem to be one of
treating each political situation as unique. The short-
run forecast, in particular, can be completely misleading
if political factors have been ignored or poorly estimated.
Social. Various social conditions exert a constant,
and to a large extent unpredictable, influence upon business.
There are some essentially social conditions such as popula
tion growth which change so slowly that no consideration
need be given them in even the longest of regular fore
casts. Others such as clothing styles may change in a
period of weeks.
It does seem that many random occurrences of social
origin would fall under the "coherent” classification of
Professor Slutzky.& Even if a recognition of periodicity be
^ Eliot J. Swan, "Pitfalls in Forecasting Business
Conditions," (Federal BeserveBank of San Francisco), p. 71*
A. M. Schlesinger, "Tides of American Polities"
Yale Review. Winter Issue, I9I 4.O. Schlesinger identifies
periods of predominate liberalism or reaction in the United
States from 1765 to 19i|0 without regard to political party
in power. The duration of eaeh phase is noted to vary con
siderably, ranging from ten to thirty years.
6 Of*, P* 102.
107
denied ’ ’ coherent” variables it seems likely that they are,
in another sense, determinate. Meed a new clothing fad,
to cite an example given above, "surprise” the forecaster?
Might he not have been advised of the pending change in
clothing demand by way of advance advertising campaigns, '
fashion shows of new creations, et cetera? Perhaps there
are, as Slutzky and others have inferred, a large class of
so-called ’ ’ random” variables which differ from "known”
variables only in the more obscure nature of "causal" data
At any rate the bulk of socially induced random
fluctuations appear attributable to "non-economic” behavior*
Though prices be steadily reduced while purchasing power
remains constant there will not always be an increase in
consumption, for example. It may even happen that consump
tion will actually decrease. From the forecasters stand
point there will have transpired an unfavorable turn of
business activity as a result of two fundamentally favorable
factors, i.e. attractive prices and undiminished purchasing
power. If the forecast proved, understandably, to be In
error in such instance, the forecaster might, in apology,
speak of changing patterns of consumption or high time-
elasticity of demand. At present the excuse might well be
an acceptable one to experts and laymen alike. Statisti
cally speaking, a random occurrence would have claimed
another forecast*
108
Reversals of the type hypothecated above have been
all too common. It would seem that the solution of pro
blems involving mass behavior factors require abilities
beyond those ordinarily expected of the economic statisti
cian*?
Acts of God. Random fluctuations due to “Acts of
God“ are almost by definition of an unpredictable character.
They are ordinarily thought to include the natural catas
trophes such as earthquakes and floods as well as the multi
tudinous forms of pestilence which irregularly and errati
cally snuff out the lives of humans and economically valuable
animals and plants.
The forecast, especially the short-run forecast, may
be completely upset by occurrence of this sort. Though
there is no way of eradicating errors thus induced, there
is good reason to believe that they are becoming ever less
important to the general forecast. Factors which tend to
substantiate this opinion are (1) A vast majority of such
occurrences are of but local concern and do not have a major
? Babson, op. clt.. pp. 239-i|.7. Probably no well
known authority has been more concerned with social condi
tions as causes and effects of business activity than Roger
Babson.
109
nationwide economic impact; ( 2) Hie effects of catastrophes
are being softened by new types of insurance designed to
safeguard against threats of most every conceivable nature;
and (3) "Acts of God” are actually becoming a more exclusive
category. Outbreaks of smallpox, for example, no longer
constitute a real danger to the supply of labor.
Whatever the trend, however, scientific forecasting
is compelled to accept the errors of this source as an un
known but irreducible minimum.
Conclusions. Hie whole class of random variables
are generally conceded by forecasters to be indeterminate
but ever present economic factors. Though mathematical and
statistical techniques have been advanced for the handling
of irregular factors of statistical series, none are as
yet in practical use for short-run forecast preparation.
Some types of variations presently temed random may
not, however, be so elusive as the word indicates. This is
a hopeful sign to short-run forecasting as erratic fluctua
tions constitute a larger source of error for short-period
prediction than for any other. This is because there will
not ordinarily be a balance between favorable and unfavor
able random fluctuations in the short-run and a forecast for
the period may be expected to prove optimistic or pessimistic
due to such irregular Occurrence alone, nevertheless the
high-aceurancy short-run forecast will have to await adequate
110
treatment of these difficult variables* As one writer states:
It is necessary to forecast the unusual and the
use, of and by themselves, of presumably stable func
tional relationships between so-called independent
and dependent, or autonomous and induced, factors
may lead to grave errors, especially during times of
economic instability when forecasts aremost likely
to be of critical importance to their users. In
the short-run* no significant economic magnitude can
safely be assumed to be entirely derived from other
economic quantities.8 [italics not in the original]
Q
Swan, 0£* cit., p. 71.
CHAPTER V I
CONCLUSIONS OP THE SHOT
Short-run forecasting problems: a pattern? The
present study of the special problems of short-run economic
prediction has been primarily focused upon the deficiencies
of the analytical tools used by forecasters; the change in
dicators or "barometers”, Though forecasting errors are
not limited to the shortcomings of change indicators alone,
such shortcomings impose a limit of accuracy which technique
alone cannot transcend. In this sense, the deficiencies
of indicators of change pose problems to all forecasters
regardless of the method they use.
Investigation of the various indicators reveals a
number of weaknesses whieh are rather commonly shared.
Quite similar weaknesses appear in indicator after indica
tor, differing sometimes only in degree. The study has not
evaluated all of the indicators which have been and are
being used, but the selection has been diversified enough
to encourage the claim that forecasting indicators as a
whole present a recognizable pattern of problems in the
short-run. It is not likely that only one pattern exists
but the conclusions of the present study suggest the follow
ing organization:
112
I. Time Factors
1, Indicator fluctuations lead business
activity but:
a. The amount of lead is uncertain.
b. The lead is so small as to be of
little practical value to forecast
ing,
e. Lead appears to vary with the phase
of the cycle,
2, Indicator fluctuations lag business
activity and are, therefore, of little
prediction value,
3, Indicator fluctuations lead or lag, in
termit ant ly, Performance is not con
sistent enough to be dependable.
II. Sensitivity of Response
1. Fluctuations of indicator values are great
er than (less than) and tend to exagger
ate (minimize), changes in business activ
ity.
2. Indicator values change in rough corres
pondence to variations in business activity
The response, however, is sluggish and
ill defined.
3. Indicator values are sensitive to condi
tions other than those of general busi
ness and may poorly reflect same.
Ill* Interrelationship Between Indicators and Business
Activity.
1. Functional relationship evident but:
a* Causal basis is fragmentary or
obscure.
b. Is subject to wide or indeterminate
variation.
113
2. Causal relationship likely but;
a. Is complex or indirect.
b. Is not evidenced functionally.
e. Appears to vary due to time or
circums tance .
IV. Information Difficulties
1. Available data insufficient or in
appropriate.
2. Statistics not published frequently
enough or not readily available.
3. Published data Involves duplication
or omission; perhaps both.
i j . . Important economic factors exist for
which no data exists— the random causes.
The above will be identified as an arrangement
pattern of indicator deficiencies according to the charac
teristics of the fault. Somewhat more difficult would be
a classification pattern based upon the frequency or common
ness of the various indicator faults. Most difficult of
all, and perhaps most significant, would be an arrangement
of indicator shortcomings In a pattern of relat ive impor
tance.
The underlying assumption which gives worth to such
classifications, or to studies such as the present one, is
that identification and clarification of the problems of
1 1 1 * .
forecasting may make it possible to avert some of the errors
which result from ignorance or misunderstanding of their
nature, Though the forecaster unquestionably would put to
good use more information than is currently available to
him, it is also true that he might make better use of that
which he has. 1 The future progress of forecasting may depend
as heavily upon a proper organization and understanding of
facts now obtainable as upon information yet to be compiled
and techniques yet to be developed.
Present study and progress. Mo comprehensive survey
has been made of the accuracy of general business forecasts
in recent years. A number were conducted in the Thirties,
however, the results of which were hardly such as to estab
lish forecasting as a science.2 There is no reason to be
lieve that business activity forecasting at the present
1 Garfield, oj>. cit., p. 79* Garfield states, 1 1 . . .
forecasters have not yet utilized data available: there is
opportunity, for example, for more detailed study by parts
of totals sometimes arrived at in casual fashion."
^ S, L. Andrew and H. I I I . Flinn, "Appraisal of Econ
omic Forecasts" Journal of the American Statistical Associa
tion. Vol. 25, March, 1930 Supplement. Also, G. V. Cox,
An Appraisal of American Business Forecasts (Chicago: The
University of Chicago Press, 1930)* and, E. W. Pettee,
"Short-Term Price Forecasting, 1920-1929" Journal of Busi
ness of the University of Chicago. Vol. 10, July, 1936.
time ia much improved over that of the periods studied. Of
course the art of forecasting has progressed if it has been
able to maintain but constant accuracy as prediction diffi
culties have certainly increased.
There are at present two principal positions regard
ing a logical basis for forecast formulation (1) Prediction
based on a study of relative prices and other market statis
tics; and (2) Prediction based upon an analysis of aggre
gate payments and receipts; the "gross national product" or
GNP forecast. Relative price methods have been in use since
the very beginning of scientific forecasting while predic
tion by money balance analysis has little more than a decade
of precedence.
Most forecasters who follow the relative price ap
proach also subscribe, to a greater or lesser degree, to
the existence of rhythmic or cyclical variations of econ
omic phenomena. A brief and positive statement of this
position is made in the following words of Professor Slutzky
Those investigators of economic life are right
who believe in their acumen and instinct and sub
scribe to at least an approximate correctness in
the concept of the periodicity of business cycles.3
3 Slutzky, op. cit.. p. 19. Slutzky in this article
takes particular issue with the "no-cycle" viewpoint of
Wesley Mitchell and his school of business cycle theorists.
ll6
"Money balanee" forecasters, by and large, place
little credence in such sentiment as well as the technique
of prediction by historical analogy o f which it is sugges
tive.
The conflict, and such it is, is at present un
resolved. There are, unfortunately, emotional aspects to
the disagreement due to the fact that GUP forecasts make
use of certain of the Keynesian aggregate income concepts.
Post World War II forecasts in the United States were badly
in error and some of the worst were made by GNP analysis.
In rebuttal to certain of the relative price advocates who
made no effort to conceal their disapprobation, L. R. Klein
stated:
Some economists claim that the wrong predic
tions show that the entire theoretical economic
model and the methodology of forecasting are wrong
... such claims are unfounded and errors in fore
casting may have nothing to do with the validity
of many of the underlying theories ... we shall
also attempt to show that these more accurate pre
dictions do not prove that their methods are super
ior to those that failed.^
Though forecasts based on relative price analysis
were, as Klein admits, more accurate than GUP forecasts
^ L. R. Klein, "A Post-Mortem on Transition Predic
tions of National Product" The Journal of Political Economy,
^ol. August, 19^6, p. . 289* .
117
they still rated only "Poor to Pair" on the Pettee rating
scale.5
Whatever method proves superior in future years the
need of accurate forecasts has already been established.
Forecasting blunders such as the post World War II example
are of more than academic interest. As Edward Dewey and-
Edwin Dakln "remonstrate:
When a people finds that predictions of many
financial advisers, statesmen, historians, and
other proclaimed experts are seldom better than
the predictions of the astrologers, our social
sciences have demonstrably not been earning their
way. It is time for action.°
The following words of Frank Garfield exhort a dili
gence of forecaster with which all can agree:
The task of forecasting developments in the
period ahead provides . . . opportunity for the
exercise of all the ingenuity which forecasters
may have developed over recent decades in the
use of analogies, formulae, and, above all analysis
— analysis at once broad enough to cover all
major elements in the situation and detailed
enough on critical points to be illuminating.7
£ Pettee, ©£_. cit., p. 296-300. The Pettee rating
scale recognizes seven categories of accuracy ranging from
"Very good" to "Very poor". The criterion is the time by
which a turning-point is forseen, the fourth category "Fair"
being given for anticipation of a turning-point by six
months .
6 Dewey and Dakin, Cycles, The Science of Predi ction.
frarry Holt and Company, 194-7) > P • x.
7 Garfield, o j > . cit.. p. 80.
118
It may well be, as Garfield’s appeal infers, that
the forecaster as a man and researcher is the weakest link
of all.
BIBLIOGRAPHY-
BIBLIOGRAPHY
A, BOOKS
Ayres, Leonard, Turning Points in Business Cycles. New York:
Macmillan, 1939* 211+ pp.
Babson, Roger W., Business Barometers. (Babson Park:
Babson*s Report, Incorporated, 1937)* 373 PP*
Blackett, 0, W., A Method of Isolating Sinousoidal Com
ponents of Economic Time Series. University of
Michigan School of Business Administration, 1933.
58 PP.
Bratt, E. G., Business Cycles and Forecasting. Chicago:
Richard. D. Irwin, Inc., I9I +8 edition.
Burns, Arthur F., and Wesley C. Mitchell, Measuring Business
Cycles. New York National Bureau of Economic. Re
search, 1938. 580 pp.
Clark, John Maurice, Strategic Factors in Business Cycles.
H. Wolff Company, 1934* 238 pp.
Cox, G. V., An Appraisal of American Business Forecasts.
Chicago: The University of Chicago Press, 1930.
Croxton, Frederick and Dudley Cowden, Applied General Statis
tics. New York: Prentice Hall, 19i+b. 954- PP*
Crum, William L. and Alson Currie Patton, Economic Statistics
New York: McGraw-Hill Book Company, Inc., I928. 1+93 PP
Davis, Harold T. and W. F. C.Nelson, Elements of Statistics.
Bloomington: The Frincipia Press, 1935*~J+2i}. pp.
Dewey, Edward, and Edwin Dakin, Cycles, the Science of Pre
diction. Harry Holt and Company, 19I + 7. 255 PP*
Estey, J. A., Business Cycles. Their Nature. Cause, and Con
trol. New York: Prentice Hall, I9I+I. Bbk- PP*
Fisher, Irving, The Making of Index Numbers. Boston and New
York: Houghton Mifflin Company, 1927. 538 pp.
Haney, L. H., Business Forecasting. New.York: Ginn and
Company, 1931. 378 pp.
Hart, Albert G. Money Debit & Economic Activity. New York:
Prentice Hall, Inc., 19^8. £$8 pp.
Husband, William H* and James G. Dockeray, Modern Corpora-
tion Finance. Chicago: Richard D. Irwin, Inc.,
19W. “pp.
Kalecki, M., Study in Economic Dynamics. Farrar and Rine
hart, I9I 44. 92 pp.
Keynes, John Maynard, The General Theory of Employment
Interest and Money. New York: Harcourt Brace and
Company, 1935. ¥>3 PP.
King, W. I., Causes of Economic Fluctuations— Possibilities
of Anticipation and Control. NewYoik: The RonaLd
Press, 1938. 353 PP•
Kuznets, Simon, National Income: A Summary of Findings.
New York: National Bureau of Economic Research, Inc.,
19^6 • l¥l- PP *
Long, Clarence D., Building Cycles and the Theory of Invest
ment. Princeton: Princeton University Press, 194-0.
239 PP.
Macaulay, F. R., Bond Yields, Interest Rates and Stock
Prices in U.S. Since 185>6. New York: National Bureau
of Economic Research, 1938* 5$9 PP*
Mills, F. G., Price Q.uantity Interactions in Business Cycles.
National Bureau of Economic Research, 19^'* T i j . 0 pp.
______ , Statistical Methods. New York: Henry Holt and Com-
pafly, 1938. "7I p pp.
Mitchell, Wesley Glair, Business Cycles and Their Causes.
Berkeley and Los Angeles: University of California
Press, 194-1. 226 pp.
Pettengill, Robert, Prioe Economics. New York: Ronald Press,
I9I 4 B. 4-$3 pp.
Roos, C. F., Techniques of Forecasting. Econometric Institute
Funk and Wagnalls Company, 194. 8.
Ross, James A., Jr. , Speculation. Stock Prices and Indus
trial Fluctuations. New Yoik : The Roland Press Go.
1938.
121
Silberling, Norman J., Dynamics of Business. New York:
McGraw-Hill Book.Go., Inc., 19^-3* 759 PP*
Sufrin, Sidney G., Labor Policy and the Business Cycle.
Washington: American Council on Public Affairs,
I9I 4 . 3 . 52 pp .
Timoshenko, Vladimir P., The Role of Agricultural Fluctua
tions in the Business Cycle. Ann Arbor: University
of Michigan School of Business Administration,
Bureau of Business Research, 1930* 89 PP*
Wright, Wilson, Forecasting for Profit. New York: John
Wiley & Sons., Inc., I9I +7• 173 PP*
B. PERIODICAL ARTICLES
Barron's "Sjjjock Prices Lag Behind Rise in Industrial Out
look, 25:11, November 26, 19l}.5*
Buchanan, N. S., "Anticipations and Industrial Investment
Decisions,'1 American Economic Review. 32;li|.l-l55*
supplement. 19^2.
Garfield, Frank R., "Transition Forecasts in Review,"
American Economic Review. 37:71-80, May, 19^7*
Gilbert, R. V., and V. Perlo, "The Investment-Factor Method
of Forecasting Business Activity," Econometrica,
10:311-316, July-October, 19k^*
Hart, A. G., "Anticipations, Business Planning, and the
Cycle," Quarterly Journal of Economics, j i > l :273-297,
November-August 1936-37.
Kerchner, D. H., "Possibilities and Limitations of Investorss
Services,".Journal of Business of the University of
Chicago, Oetober, 1937> 299-321, and January, 193$T
21- 1 4 .3 .
Kalecki, M., "The Short-Term and Long-Term Rate of Interest,"
Review of Economic Statistics. 33:97-99, May, I9I 4 JL.
Klein, L. R., "A Post-Mortem on Transition Predictions of
National Product," The Journal of Political Economy.
55:289-308, August, 1914. 6.
122
Knight, Prank H*, "The Business Cycle Interest and Money,"
Review of Economic Stati sties, 23:53-67, May, 194-1.
Kondratieff, Nikolai D., "The Long Waves in Economic Life,"
The Review of Economic Statistics, 17:105-15* (A
translation of "Die Lange Wellen der Konjunktur,"
which appeared in Archiv fur Sozi&lwiasenschaft und
Sozialpolitik, Vol. 56,No. 3, 1925) pp. 573-609*
Lebergott, Stanley, "Forecasting the National Product,"
American Economic Review. 35*59-80, March, 19^5*
Lutz, P. P., "Interest Rate and Investment in a Dynamic
Economy," American Economic Review, 35*811-830,
December, 19^5•
Marschak, Jacob, "Economic Structure, Path, Policy and
Prediction," Atnerican Economic Review. 37*81-84,
May, 194?. .
Mayer, Joseph, "Deficiencies in the Gnoss-National-Produet
Concept as a National Measure," Journal of Political
Economy, 37:71-80, May, 191+7*
Modigliani, Franco, "Fluctuations in the Saving Ratio, A
problem in Economic Forecasting," Social Research,
ll+*1+13-1+20, December, 191+7*
Neisser, H., "Investment Fluctuations as Cause of the Busi
ness Cycle," Social Research. 1 + : 1 + 1 + j 0-1+6o, November,
1937.
Richter, F. E., "Short-term Forecasting of General Business,"
Journal of the American Statistical Association.
Vol. 23, March supplement, 1928.
Roos, Charles F., "Economic Measures--The Basic Indexes and
Forecasting.Techniques," New Yoik, The Econometric
Institute, Inc., October, I9I + 7.
. "Some Problems of Business Forecasting," Proceedings
of the National Academy of Sciences. March, 1929 .
Schlesinger, A. M., "Tides of American Politics," Yale Review,
Winter Issue, I9I + 0.
123
Slutzky, E., "The Summation of Random Causes as the Source
Cyclic Processes," Econometrica, 5 > :105>**l46, April,
1937.
"Stock Market as a Business Forecaster," Business News,
April 19, 1947, p. Il6.
"Stock Market as a Business Forecaster: reply," Business
News, May 3, . 1947, p. 31.
Swan, Eliot J., "Pitfalls in Forecasting Business Conditions,
(Federal Reserve Bank of San Francisco) Proceedings
of the Pacific Coast EconomicsAssociation, 1948.
Tinbergen, J., "On a Method of Statistical Business-Cycle
Research: a reply," Economic Journal, ?>0 :l4l-15>4*
March, 1940.
. "Some Problems in Explanation of Interest Rates-"
Quarterly Journal of Economy, 61:397-438, May, 1947.
U.S. News, 21:32-ip., October 4, 194-6*
U,S. News, 21:30 -32, October 1, 194-6.
Wallich, H. C., "Changing Significance at the Interest Rate,"
American Economic Review, 36: 761-787, December, 194-6.
Woytinski, W, S., "What Was Wrong in Forecasts of Postwar
Depression?" Journal of Political Economy, f>£:l42-15>l,
April, 194-7.
C. ESSAYS
Lewis, Edward E. "Some Basie Problems on Index-Number
Theory,” Economic Essays in Honor of Wesley Clair
Mitchell. New York: Columbia University Press, 1935.
pp 271-93.
D. UNPUBLISHED MANUSCRIPT
Franck, H. J., "Time Lags in Adjustment to Economic Events.1 1
Unpublished Master1 sThesis, TheUniversi ty of Southern
California, Los Angeles, l§4o. 80 pp.
30
to such a practice exists for periods of critical uncer
tainty. Haney notes that a heavy weighting of "blue chip"
stocks caused several preminent indexes to be Insensitive
to generally falling stock prices in 1929* The error was
due to unrepresentative weighting and is a potential danger
presented by any index of fixed composition*
Even a well weighted index of stock prices will be
a poor indicator of prospective business activity If there
is rampant speculation in the market. Speculation, of
course, leads to over or under valuation. Evidence indi
cates that even speculatory activity will not long cause
stock prices to deviate from the longer term trend. Even
so, the short-run forecasting value ©f stock prices during
such periods has been negligible
Another factor affecting the significance of stock
indexes is the condition of the short-term money market.
A change in confidence or.in policy on the part of the call
loan bankers can mean either contracted or expanded part
icipation by those operators dependent upon margin. Chances
of error from this source are great If marginal trading is
a large portion of the total. Present restrictions on
2k
Haney, op. cit.. p. 378. See also the I918-I92I
period and the early Thirties of Figure I.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Inflation and a national wage policy in peacetime

PDF

A historical development of organized labor's opposition to immigration

PDF

Theories of the passing of capitalism

PDF

Population changes and economic welfare

PDF

Some contributions of Thomas R. Malthus to modern economic thought

PDF

An economic analysis of taxable capacity and its determining factors with special reference to the United States

PDF

The economics of air cargo transportation

PDF

Some basic modifications of American property taxation

PDF

An analysis of the institutional economics of John Rogers Commons

PDF

The collective farm workers in Russia

PDF

Papermaking, printing, and economic growth to 1500 A.D.

PDF

A survey of mediation techniques

PDF

A comparative study of price controls in the United States from 1942 to 1952

PDF

Comparison of travel durations by sex: Is there really a "meaningful" difference?

PDF

An analysis of recent expert estimates of maximum world population at a minimum adequate standard of living

PDF

The history of China's note issue

PDF

A comparative study of the current tax systems in the United States and Great Britain

PDF

Pawnbrokerage in California

PDF

An investigation of the unionization of teachers in Southern California

PDF

Economic and administrative aspects of capital budgeting in municipalities

Asset Metadata

Creator Smith, Jack Earl (author)

Core Title A study of the special problems of short-run economic prediction

Degree Master of Arts

Degree Program Economics

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Economics, General,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Pollard, Spencer D. (committee chair), Anderson, William H. (committee member), Garis, Roy L. (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c20-436732

Unique identifier UC11265164

Identifier EP44690.pdf (filename),usctheses-c20-436732 (legacy record id)

Legacy Identifier EP44690.pdf

Dmrecord 436732

Document Type Thesis

Rights Smith, Jack Kerl

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA