-
International Journal of Forecasting 17 (2001)
333348www.elsevier.com/ locate / ijforecast
A framework for measuring international business cyclesa ,
b*Anirvan Banerji , Lorene Hiris
aDirector of Research, Economic Cycle Research Institute, 420
Lexington Avenue, Suite 1645, New York, NY 10170, USAbProfessor of
Finance, C. W. Post /Long Island University and Senior Research
Scholar, Economic Cycle Research Institute,
720 Northern Blvd., Brookville, NY 11548, USA
Abstract
The classical measurement of business cycles, growth cycles, and
growth rate cycles lies at the foundation for theunderstanding of
macroeconomic dynamics in open market economies. This essay
presents a framework for analyzing andforecasting cyclical behavior
in economic activity, employment, and inflation. The framework is
extended to foreign tradeand important domestic sectors of an
economy such as manufacturing, services, and construction. This
multidimensionalframework, which allows for a more in-depth
analysis, serves as a model to be developed on a comparable basis
acrosscountries. Business cycle and growth rate cycle reference
chronologies, which have been determined for the majoreconomies,
are presented in this context. 2001 International Institute of
Forecasters. Published by Elsevier Science B.V.
Keywords: Business cycles; Growth rate cycles; International
reference cycles; International reference chronologies;
International cycledates; Turning points; Cyclical analysis;
Forecasting; Leading indexes; Long leading indexes; Short leading
indexes; Economic sectors;Inflation cycles; Employment cycles;
Foreign trade; Exports; Imports
1. Introduction for a much more fine-grained, nuanced analysisof
economic cycles. In fact, in the decades since
While the index of leading economic in- the LEIs creation, the
general approach hasdicators (LEI) is still popularly perceived as
the been refined under the guidance of its creator,chief
forecasting tool available for predicting Geoffrey H. Moore, and
applied in a consistentUS recessions and recoveries, it is not
widely manner to a variety of economies.known that the classical
National Bureau of One important finding is that leading
in-Economic Research (NBER) approach may be dicators of business
cycles, when used in theapplied within a multidimensional framework
form of growth rates, also lead cyclical turns in
the growth rates of coincident indicators. Infact, it is useful
to perform complementarycyclical analyses in terms of growth rate
cycles.
*Corresponding author. Tel.: 1 1-212-557-7788; fax: In order to
identify these business cycles and1 1-212-557-9874.growth rate
cycles, we first present the definingE-mail address:
[email protected] (A.
Banerji). characteristics of economic cycles, including
0169-2070/01/$ see front matter 2001 International Institute of
Forecasters. Published by Elsevier Science B.V.PI I : S0169-2070(
01 )00089-9
-
334 A. Banerji, L. Hiris / International Journal of Forecasting
17 (2001) 333 348
cyclical co-movements and stable sequences of and recoveries).
By the end of the 1960s,leads and lags in cyclical indicators.
however, many industrial economies had not
Next, a multidimensional framework that experienced a recession
for many years. Thispermits more in-depth monitoring through a led
some observers to question whether thecloser look at three key
aspects of the economy business cycle was still in existence
(Bronfen- aggregate economic activity (the traditional brenner,
1969). Subsequently, there was a movedomain of the LEI), inflation,
and employment among students of business cycles to study the is
described. Aggregate economic activity growth cycle, which based
cyclical analysis onmay be divided into foreign trade and domestic
the deviations in economic activity from trendeconomic activity.
The latter, in turn, can be (Mintz, 1969). A few years later when
thesubdivided into the major sectors of the OECD developed leading
indicators for itseconomy services, manufacturing and con- member
countries, it decided to monitor thesestruction. Each of these
areas exhibits distinct growth cycles. Growth cycle analysis
alsoeconomic cycles, marked by the co-movement formed the basis for
the international economicof many economic activities within each
area. indicators (IEI) project (Klein & Moore, 1985)Also,
within each of these areas, specialized started at the NBER in the
early 1970s.coincident and leading indexes can be designed While
growth cycles are not hard to identifyto anticipate the distinct
cyclical movements. in a historical time series, they are difficult
toWhile this framework has been fully fleshed out measure
accurately on a real-time basis (Bos-for the US economy, with
composite indexes for chan & Banerji, 1990). This is because
the trendeach aspect and sector currently available, the over the
last couple of years must be estimated,framework is still in the
process of being fully and these trend estimates tend to be
veryextended to the other major economies. unstable near the end
(Cullity & Banerji, 1996).
The international extension of this framework This difficulty
makes growth cycle analysis lesswith respect to economic activity
is then pre- than ideal as a tool for monitoring and
forecast-sented, and comparable business cycle and ing economic
cycles in real time, even though itgrowth rate cycle reference
dates for the US and still useful for the purpose of historical
analysis.other major economies are defined. Comparable By the late
1980s, the use of growth ratesets of coincident and long leading
indexes for cycles for the measurement of series, whicheach of
these major market economies are also manifested few actual
cyclical declines but didpresented. show cyclical slowdowns, was
introduced
In sum, this paper shows how the classical (Layton & Moore,
1989). Like the step cycleindicator approach to forecasting can be
refined introduced by Mintz (1969), the growth rateand applied in a
consistent fashion to many cycle was based on the growth rate of
economiceconomies within a multidimensional frame- activity.
However, unlike the step cycle, it didwork, allowing for more
in-depth analysis as not presume that the growth rate changed
inwell as greater breadth of application. steps. The growth rate
cycle was based on the
six month smoothed growth rate concept,which avoids the sort of
extrapolation of thepast trend needed in growth cycle analysis.
This2. Business cycles and growth rate cyclessmoothed growth rate
is based on the ratio of
Leading indicators were originally designed the latest monthly
figure to the average of theto anticipate traditional cyclical
downturns and preceding twelve months. Cyclical turns in
thisupturns in economic activity (i.e., recessions growth rate
define the growth rate cycle
-
A. Banerji, L. Hiris / International Journal of Forecasting 17
(2001) 333 348 335
(Banerji, 1999). Turning points in the growth activity. During a
cyclical upswing, the im-rate cycle could be identified by the same
provement in economic activity spreads fromobjective procedures
(Bry & Boschan, 1971) one firm to another, from one industry
toused to identify business cycle turning points. another, from one
region to another. Moreover,
The growth rate cycle not only avoids the these spreading
movements snowball over time,problems of trend estimation presented
by the as an expansion or contraction unfolds.growth cycle in the
case of real-time moni- Another important characteristic of a
cyclicaltoring, but also shares key cyclical characteris- upswing
or downswing is its persistence. Atics exhibited by the business
cycle. Most move in cyclical indicators that is
pronouncedimportantly, the cyclical turns in the broad and
pervasive, but lasts only 2 to 3 months,measures of aggregate
economic activity in the does not qualify as a cyclical movement.
Tech-form of output, income, employment, and sales nically, a
cyclical upswing or downswing has tocluster together, whether
viewed in the frame- persist at least 5 months (Bry &
Boschan,work of business cycles or growth rate cycles. 1971), but
most last much longer.Moreover, when growth rates are used, the
Finally, cyclical changes tend to be pro-analysis still generates
stable sequences of nounced in magnitude, compared with non-leading
indicators that anticipate coincident cyclical fluctuations, or
noise. In particular, theindicators. magnitude of the upswings or
downswings must
What has emerged in recent years is the be comparable to those
exhibited in previousrecognition that business cycles, growth
cycles cyclical episodes.and growth rate cycles can all be
monitored in a Of the three Ps, pervasiveness, or the
co-complementary fashion. However, of the three, movement of many
indicators, is necessarilybusiness cycles and growth rate cycles
are more one that can be defined only with respect tosuitable for
real-time monitoring and forecast- many series considered together.
The other twoing, while growth cycles are more suitable for Ps can
be considered, however, for individualhistorical analysis (Klein,
1998). time series, for which cyclical turns specific to
the series can be identified.2.1. Pronounced, pervasive and
persistentswings in levels and growth rates 2.2. The clustering of
cyclical turning points:
evidence of co-movementsThe identification of an economic
fluctuation
as a cyclical movement is based essentially on The robustness of
the classical approach tothe three Ps, i.e., whether the movements
are cyclical analysis is based on the clustering ofpronounced,
pervasive and persistent. The con- cyclical turns in the coincident
indicators, ascept of the three Ps is not a new idea as it is well
as the ability of leading indicators toinherent in the notion of
the three Ds (duration, anticipate cyclical turns in the economy.
Thedepth and diffusion), which are established objective
identification of such cyclical turningcriteria for measuring the
severity of a recession points is critical to cyclical analysis.
Since(Fabricant, 1972). However, while the three Ds leading
indicators are meant primarily to fore-apply only to cyclical
downturns, the three Ps cast business cycle turning points, the
identifica-apply to both upturns and downturns. tion of turning
points in time series is a sine qua
A fundamental feature of a free market non for appropriate
evaluation of forecastingeconomy is the cyclical diffusion or
pervasive- performance. In order to identify cyclical upsw-ness of
movement of indicators of economic ings and downswings, and the
turning points
-
336 A. Banerji, L. Hiris / International Journal of Forecasting
17 (2001) 333 348
that demarcate them, an algorithm for the cyclical turns that
the differences in the threeidentification of turning points, based
on a aspects and various sectors of an economy cansystematic
codification of the judgmental pro- be fully appreciated. One key
reason for dis-cedures, is used. Such a procedure was devised
tinguishing among cycles in aggregate econ-three decades ago (Bry
& Boschan, 1971) omic activity, inflation and employment is
thatshortly after the creation of the LEI, and was while cyclical
turns cluster quite clearly withinused for decades at the NBER.
each of these aspects, the resulting cycles are
The objective, though not mathematically different from each
other. Similarly there issimple, definition of turning points given
by Bry strong evidence of clustering of turns withinand Boschans
algorithmic formulation of the services, manufacturing, and
construction, andclassical NBER procedure makes it possible to the
resultant cycles are quite distinct.evaluate the performance of
indicators in termsof timing of cyclical turns. The
computerizedBryBoschan algorithm was used extensively in 3. The
multidimensional frameworkthe years following its creation (e.g.,
Klein &Moore, 1985). Other users of this algorithm The
multidimensional framework for thehave included King & Plosser
(1989), who analysis of economic cycles applies to both theprovide
a description of the procedure. As levels of economic activity and
their growthWatson (1994) has pointed out, the BryBos- rates.
However, the latter are particularly rel-chan procedure provides a
good way to define evant for the service sector and many
interna-turning points since it is based on objective tional
economies, which may exhibit relativelycriteria for determining
cyclical peaks and rapid growth without absolute declines.
Whiletroughs. the focus in the 1960s and 1970s remained on
The output from the BryBoschan program is cycles in economic
activity or its growth rate,used as the basis for the determination
of by the early 1980s it had become clear thatreference
chronologies as well. In the NBER cyclical activity was not
one-dimensional, andtradition, the turning points are determined
for that there were cycles in other important aspectsall the major
measures of aggregate economic of an economy that were worth
monitoring,activity, with specific reference to output, in-
specifically, inflation and employment.come, employment and sales.
The turning points Of course, the LEI was used to predict cyclesare
then clustered, i.e., the reference cycle in aggregate economic
activity or economicturning point is chosen on the basis of the
best growth. A key step forward in the evolution ofconsensus among
the turning point dates for the multidimensional framework was the
sepa-these individual indicators. The business cycle ration of
leading indicators with long leads fromand growth rate cycle dates
identified by such those with short leads (Cullity & Moore,
1990).an objective procedure are characterized by the The former
may be combined into a longthree Ps pronounced, pervasive, and
persistent leading index, and the latter into a short
leadingmovements in the cyclical indicators between index, while
coincident indicators are combinedturning points. into a coincident
index. Such a set of long
This clustering of turning points reflects the leading, short
leading and coincident indexescyclical co-movement of many economic
ac- results in a sequential system for monitoringtivities that is
the hallmark of an economic cyclical developments.cycle. It is in
the context of such clustering of Because inflation exhibits cycles
distinct from
-
A. Banerji, L. Hiris / International Journal of Forecasting 17
(2001) 333 348 337
economic activity, a composite leading index tightness in
employment, financial, or domesticmay be constructed specifically
to lead inflation or foreign goods markets, and thus measurecycles.
Since employment cycles differ from underlying inflationary
pressures. Such in-business cycles as well, just as inflation
cycles flation-specific leading indicators can be com-do, a
composite leading index may also be bined into a Future Inflation
Gauge (FIG).constructed specifically to predict employment Similar
composite leading inflation indexescycles. Independent
corroboration of the validity have also been developed by other
researchersof such a multidimensional cyclical structure in
(Niemira & Klein, 1994). Other researchersthe economy comes
from Stock and Watson have also focused on the inflation cycle,
as(1998), who conducted a large-scale factor distinct from the
business cycle (Roth, 1986;analysis of a large number of US time
series. Ivanova et al., 2000).They found that according to the
factor load- The distinction between cycles in economicings, the
first factor related to measures of growth and inflation once again
came to the foreeconomic activity and in particular measures of in
the late 1990s, when the US economyoutput similar to industrial
production. The experienced several years of non-inflationarysecond
(and orthogonal) factor had strong factor growth. During this
period, the FIG accuratelyloadings on money, earnings, and prices,
which predicted subdued inflation, even as the leadingare of course
linked directly to inflation. The indicators of economic growth
correctly forecastthird factor had heavy factor loadings on vari- a
robust economy. In this case, overall inflation-ables related to
employment. This suggests that ary pressures, as measured by the
FIG, werethe proposed multidimensional cyclical frame- kept in
check by imported disinflationarywork covers the three most
important distinct pressures, even as joblessness fell and
domesticdimensions of the US economy. inflation pressures
climbed.
The framework was further extended with the In general, while it
is true that cyclicalrecognition that the economic activity aspect
of downturns in inflation follow growth slow-the economy can be
divided into foreign trade downs about 70 percent of the time,
growthand domestic activity. The latter can be sub- slowdowns
actually follow inflation downturnsdivided into three major sectors
services, the other 30 percent of the time. In other
words,manufacturing, and construction. economic growth or the
unemployment rate
alone can be imprecise predictors of the timingof cyclical turns
in the inflation cycle. Hence the3.1. Inflation and employmentneed
for leading indicators of inflation distinct
In the aftermath of the stagflation of the late from leading
indicators of economic activity.1970s, it was quite obvious that
there could be Another important aspect of the economy isimportant
divergences between cyclical be- employment. In the early 1980s,
Moore and hishavior in economic growth, in inflation, as well
colleagues developed a specialized leadingas in employment. While
these variables are index to anticipate changes in
employmentloosely related, it became increasingly clear that
conditions (Moore, 1981). The Coincident Em-they could exhibit very
different cyclical timing. ployment Index is designed to move in
step
These developments gave a fresh impetus to with the cyclical
movement in employmentthe development of an index of leading in-
conditions, while the Leading Employmentdicators of inflation
(Moore & Kaish, 1983). Index is designed to anticipate these
employ-These indicators typically reflect the degree of ment
cycles. In early 1990, the Leading Em-
-
338 A. Banerji, L. Hiris / International Journal of Forecasting
17 (2001) 333 348
ployment Index forecast a sharp and recession- act in concert,
and quicker employment growthary rise in the jobless rate, and was
the key to is accompanied by faster price rises in com-
1Geoffrey Moores prediction of a recession modity prices, rapid
expansion of credit and(Sebastian, 1990) that newer econometric
meth- increasing delays in supply lines. The pervasive
2 influence of all these factors usually leads to anods failed
to predict. The recession did start ininflation upswing.July 1990,
but the governments LEI did not
3 However, at certain times the link betweenpredict the
recession either (Wessel, 1990)employment growth and some of these
otheruntil much later in the year.indicators may be broken. This
may happen, forAlthough strong employment growth can be ainstance,
because of central bank action orsource of inflationary pressure,
it is by no meansbecause of adverse cyclical developments inthe
only factor that determines future inflation.other countries. In
these cases, the inflationaryThere are many other factors, such as
growth ininfluence of high employment would be morethe prices of
industrial materials and imports,than counteracted by other
disinflationary orcredit growth and tightness in supply lines.even
deflationary influences, resulting in lowMuch of the time, the
influences of these factorsinflation in spite of tight labor
markets.
1The importance of the employment indicators to Moores The link
between cyclical swings in employ-prediction is evident from the
description in Sebastian ment growth and inflation during the
postwar(1990), from The Wall Street Journal dated March 9: era has
broken down several times. During theGeoffrey Moore, who at 75
years of age has had a hand
stagflation of the late 1970s, employment de-in declaring many
modern recessions, gives his opinionclined in 1976 and again in
1978, while inflationeven without being asked. Mr. Moore, director
of Colum-
bia Universitys Center for International Business Cycle
continued to climb. Also, there were severalResearch, recently
noted the centers employment index episodes when employment grew
strongly with-has begun signaling recession . . . Newly
pessimistic, Mr. out causing inflation. Out of 13 upswings inMoore
puts the odds of a recession at two-to-one in the
employment during the postwar period, 10 werefirst half. If we
escape recession in the first half, Idfollowed by an inflation
upswing within a yearchange the odds to fiftyfifty in the second
half. Mr.or so. In two cases, the inflation upswing startedMoore
admits that the centers long-leading index its
main predictor is still very strong. I had some difficulty even
before the upswing in employment growthreconciling that. Maybe it
means a recession will be very began. However, in three cases, in
1980, inbrief if it comes. 1991, and most recently in 1996, a
sustained2In the same article, Sebastian (1990) wrote: The
just-
upswing in employment was accompanied by areleased January
reading on the NBERs new recessionsustained inflation downturn. The
evidence isindex puts the chance of recession at a measly 3% in
the
next 6 months . . . Our new index would be consistent clear,
therefore, that employment cycles andwith the fourth quarter of
1989 being the worst of quarters inflation cycles are quite
distinct from eachin this cycle, says James Stock, . . . one of the
new other. For that reason, in the case of inflation asindexs
creators.
well as the case of employment, the notion was3Wessel (1990),
writing in The Wall Street Journal datedthat, by combining
indicators pertaining to aDecember 3, noted, The governments
economic forecast-single critical macro-economic dimension ofing
gauge is finally flashing the recession sign, but the
warning comes after what most economists say was the economic
performance, results would be morebeginning of the downturn . . . A
rule of thumb developed precise and, hopefully, produce longer and
moreby University of Michigan forecaster Saul Hymans is that
reliable leads for forecasting (Klein, 1993).three consecutive
declines in the index signal a recession. It is interesting to note
that US FederalThe index dropped in August and September, but it
wasnt
Reserve policy during the Greenspan years hasuntil Friday that
the Commerce Department revised thefigures for July to show a
decline for that month, too. been remarkably well correlated with
cycles in
-
A. Banerji, L. Hiris / International Journal of Forecasting 17
(2001) 333 348 339
the FIG (Banerji & Klein, 2000). According to parts,
representing the major sectors of theCoons (2000), the FIG, which
is a measure of economy.inflationary pressures, has reliably
predictedchanges in the direction of the federal funds rate 3.2.1.
Foreign tradetarget at least since Greenspan was named Fed The
traditional leading indicator approachChairman in 1987. Coons
further observes that
evolved mainly in the context of closedno indicator or
forecasting method is without
economies. Yet, with the increasing globaliza-its flaws, but
this combination (the FIG) seemstion of the economy, it has become
more
to distill from the relevant information available important to
be able to anticipate the cycles inat any time what is knowable
about the future foreign trade that can significantly impact
thedirection of interest rates. domestic economy.Thus, the leading
indexes of employment and The economic growth aspect of the
economyinflation have chalked up impressive achieve-
can be split up into two major divisions, onements in the past
decade. Five months before
relating to the domestic economy, and the otherthe 199091 US
recession began, the employ-
to foreign trade. In order to monitor trade flows,ment index
predicted a sharp increase in un-
the Leading Exports Index (LExI), based onemployment consistent
with a recession, which
exchange rates and the long leading indexes forbetter-known
forecasting tools were unable to US trading partners, was developed
so that the4foresee. The inflation indicators, on the other LExIs
growth rate could anticipate cycles inhand, accurately anticipated
every directional US exports growth. Similarly, the growth rate
ofchange in a successful Federal Reserve policy
the Leading Imports Index anticipates cycles induring a period
when many observers invoked a US imports growth, and the Leading
Tradenew paradigm to explain the apparent delink- Balance Index
anticipates cycles in the US tradeing of US growth and inflation.
balance (Hiris & Guha, 2001).3.2. Economic activity and
growth
3.2.2. Major sectors of the domestic economyComposite indexes of
leading, coincident, Traditional leading indexes are designed
to
and lagging indicators have long been con- anticipate cycles in
overall domestic economicstructed for the purpose of monitoring
cycles in activity. However, the major sectors of theeconomic
activity and growth. However, econ- domestic economy do not always
move in theomic activity, one of the three aspects of the same
direction, and it can be of value toeconomy, can be broken down
into several subdivide the domestic economy into its three
sectors manufacturing, services and construc-tion corresponding
to the broad division of4In February 2001, the Leading Employment
Index growth gross domestic product into goods, services and
rate plunged to a 19-year low, pointing to a
recessionarystructures.increase in unemployment in 2001. Along with
recession-
The manufacturing sector is known to beary declines in most of
the other U.S. leading indexes, thisresulted in ECRIs prediction of
a recession in 2001. As of highly cyclical, exhibiting its own
expansionsJune 2001, most of the key coincident indicators used by
and contractions in economic activity. Cycles inthe NBER to define
U.S. recessions, such as industrial the manufacturing sector are
tracked by theproduction and employment, had declined in a way they
Coincident Manufacturing Index, which is madehad done only during
earlier recessions. However, con-
up of coincident indicators of manufacturingcensus forecasts
suggested that the economy would averta recession. activity. The
Leading Manufacturing Index,
-
340 A. Banerji, L. Hiris / International Journal of Forecasting
17 (2001) 333 348
made up of leading indicators specific to manu- inflation and
employment, exhibit distinctfacturing activity, anticipates these
cycles. characteristics. Specifically, they each exhibit
In recent decades, the service sector has pronounced, pervasive,
and persistent move-become the increasingly dominant part of the
ments in terms of levels, growth rates, or both.US economy,
accounting for more than half ofoutput and four-fifths of
employment. Although 3.2.3. Evidence of cyclical behavior:
stablethe service sector does not typically exhibit the sequences
of leadsfamiliar cycles of expansion and contraction, it Leading
indicators, whose cyclical turns con-does show cyclical speedups
and slowdowns in sistently anticipate cycles in the
coincidentgrowth. The application of the concept of indicators,
provide further evidence supportinggrowth rate cycles to analyze
the service sector the existence of stable patterns of cyclicalof
the economy is, therefore, appropriate movements. This is true
whether the indicators(Layton & Moore, 1989). The growth rate
of are viewed in terms of levels or growth rates,the Coincident
Services Index, made up of and whether it is overall economic
growth orcoincident indicators of services activity, aspects or
sectors of the economy that aremonitors the current state of the
service sectors examined. At the Economic Cycle Researchgrowth rate
cycle, while the growth rate of the Institute (ECRI), founded by
Geoffrey H.Leading Services Index anticipates those cycles. Moore,
composite leading and coincident index-
The third slice of the domestic economy is es were created for
all of these aspects andthe highly cyclical construction sector.
The sectors of the US economy.Coincident Construction Index and
Leading As Table 1 shows, the ECRI leading indexesConstruction
Index track these cycles in an for each aspect of the US economy as
well asanalogous fashion. for each major sector consistently lead
the
Each of these sectors of the domestic corresponding coincident
indicators at cyclicaleconomy, as well as foreign trade, along with
turning points (by convention, leads have nega-the two other key
aspects of the economy, i.e., tive signs and lags have positive
signs). On
Table 1aUnited States, 10 leading indexes timing at cyclical
peaks and troughs
Index Number of Number of Leading indexescyclical cyclical
average leads (months) attroughs peaks
Troughs Peaks Overall
Long leading 9 9 2 6 2 11 2 8Short leading 9 9 2 2 2 10 2
6Leading employment 9 8 2 3 2 10 2 6Leading manufacturing 9 9 2 3 2
7 2 5
bLeading services Leading construction 9 8 2 4 2 5 2 5Leading
trade balance 4 5 2 14 2 13 2 13
bLeading imports bLeading exports
Future inflation gauge 12 11 2 12 2 11 2 12a Source: Economic
Cycle Research Institute, New York City.b These indexes do not show
clearly cycles in terms of levels, but do exhibit growth rate
cycles (see Table 2).
-
A. Banerji, L. Hiris / International Journal of Forecasting 17
(2001) 333 348 341
average, the Long Leading and Short Leading a cyclical turn is
usually obtained from a longindexes lead business cycle turning
points by leading index (Cullity & Moore, 1990). In thisnearly
a year at peaks and half a year or less at regard, they are
superior to traditional leadingtroughs. A similar asymmetry of
leads is seen in indexes. In practice, because of the normal
datathe performance of the Leading Employment publication lags and
the inevitable delay inIndex. The Leading Manufacturing and Con-
recognizing a turning point after it occurs, thestruction Indexes
have average leads of about effective leads of most traditional
leading index-half a year. The Future Inflation Gauge and es are
often whittled down to virtually zero.Leading Trade Balance Index
have leads that Because of their longer lead times, long leadingare
roughly symmetric at peaks and troughs, and indexes are less
susceptible to this problem, andclose to a year. are therefore more
useful to policy makers and
As Table 2 shows, the leading character of others who need to
take early action before athese indexes is retained when they are
ex- cyclical turn. Such long leading indexes are nowamined in terms
of growth rates. However, the available for over a dozen countries
other thanleads are now generally shorter and more the United
States.symmetric. In fact, all the above-mentionedindexes lead by
about half a year, on average, atboth peaks and troughs. The growth
rates of the 4. The multidimensional framework in anLeading
Services Index and the leading indexes international contextof
exports and imports, which do not exhibitcycles in level terms
(Table 1), do show clear Wesley Mitchell long ago recognized
thecycles in growth rate terms, and lead by about need for the
study of business cycles in thehalf a year at both peaks and
troughs. international context. More than seventy years
It should be noted that the earliest warning of ago, he wrote,
For theoretical uses, there is
Table 2aUnited States, 10 leading indexes timing at growth rate
cycle peaks and troughs
Index Number of Number of Average leads (months) atgrowth rate
cycles growth rate cycles
Troughs Peaks Overalltroughs peaks
Long leading 13 14 2 7 2 7 2 7Short leading 14 14 2 6 2 3 2
4Leading employment 14 15 2 6 2 6 2 6Leading manufacturing 13 14 2
5 2 6 2 5
bLeading services 12 12 2 6 2 6 2 6Leading construction 14 13 2
5 2 8 2 6
bLeading trade balance Leading imports 12 13 2 7 2 8 2 7Leading
exports 5 5 2 6 2 6 2 6
cFuture inflation gauge a Source: Economic Cycle Research
Institute, New York City.b Since the trade balance can be both
positive and negative, growth rates of this measure are not
appropriate.c The level of the future inflation gauge is already
designed to anticipate cycles in the rate of inflation (see Table
1), i.e.,
the growth rate of CPI.
-
342 A. Banerji, L. Hiris / International Journal of Forecasting
17 (2001) 333 348
needed a systematic record of cyclical alterna- focus on data
from a single country and ations of prosperity and depression,
covering all limited time period. Unfortunately, businesscountries
in which the phenomena have ap- cycles follow many different
patterns, and cy-peared, and designed to make clear the recurrent
cles may arise out of the combined contributionfeatures of the
fluctuations (Mitchell, 1927). of a number of economic processes.
The contri-Three quarters of a century later, Basu and butions of
different economic processes changeTaylor confirmed, a robust and
useful theory of from turning point to turning point. Also, it
isbusiness cycles should be able to account for impossible to
predict which of these cyclicalthe pattern seen in the long-run
data for many processes will trigger the next turn. Thus,countries
(Basu and Taylor, 1999). single-country data, especially over a
period of
Some of the most important substantive ad- one or two decades
(perhaps three or fourvances in measuring business cycles have been
cycles), is simply not enough to permit thein the extension of the
classical indicator analy- coverage of the diversity of these
processes.sis approach to a number of economies. More What is
preferable to such single-countryspecifically, a major achievement
has been the data, but not always available to most
research-finding that many of the same indicators lead ers, is the
data on cycles in a large variety ofeconomic cycles in a diverse
collection of free-market economies. While the structuraleconomies
with different structures. This en- differences can obviously be an
issue, theables the design of robust systems of indicators,
commonalities can be very revealing in guidingwhich can work in a
variety of economies and the choice of a robust set of leading
indicators.in the same economy through major structuralchanges.
4.1. Comparable cyclical indicators across
Leading and coincident indexes, similar to bordersthose
developed for the United States, wereinitially constructed for
other major market The classical approach represented by
Mooreeconomies by Klein and Moore (1985). A and his colleagues has
never been to start with afurther important enhancement of these
efforts large list of potential leading indicators, andongoing at
ECRI is the development of a system choose from them just on the
basis of the bestof cyclical indicators which are comparable
statistical fit to a limited amount of single-across borders. Moore
and his associates fol- country data. The guiding principle has
alwayslowed a rather strict procedure in that they been conceptual
in that the list of indicators thatassembled data from the other
countries on the are to constitute a leading index must contain
asame types of economic processes that have representative sample
of all the key processesproved to be leading indicators in the US.
In a that are known to contribute to the economicsense, each
country for which these systems of cycles being targeted. These
have been based onindicators were constructed served as an out-of-
the detailed cyclical analysis of many freesample test for the
selected set of leading market economies. The choice of leading
in-indicators. Such a procedure does not necessari- dicators is
then made from all the available timely identify the best leading
indicators for each series that reflect the key processes
mentioned,country over a given time span. What it does primarily on
the basis of their performance atyield, however, is a robust set of
leading cyclical turning points. Despite the relativeindicators.
brevity of the time series, the approach of using
There is the tendency for virtually all roughly equivalent
indicators across countries,economists who work on leading
indicators to instead of choosing the indicators according to
-
A. Banerji, L. Hiris / International Journal of Forecasting 17
(2001) 333 348 343
Table 3aTiming at business cycle peaks and troughs, long leading
indexes, 13 countries
Country Number of Number of Long leading indexesbusiness cycle
business cycles average leads (months) attroughs peaks
Troughs Peaks Overall
US 9 9 2 6 2 11 2 8Canada 2 2 2 14 2 12 2 13Germany 4 4 2 10 2
10 2 10France 4 4 2 2 2 9 2 6UK 3 3 2 13 2 20 2 17Italy 3 2 2 11 2
12 2 11Switzerland 4 4 2 15 2 13 2 14Sweden 4 3 2 7 2 10 2 9Japan 2
3 2 12 2 10 2 11Korea 2 2 2 7 2 1 2 4Australia 6 5 2 7 2 15 2
11Taiwan 1 1 2 12 2 10 2 11N.Z. 6 6 2 5 2 4 2 4
a Source: Economic Cycle Research Institute, New York City.
the degree of statistical fit, has led to success in has worked
successfully in the vast majority ofdeveloping economic indexes for
countries as these countries as it has in the US.diverse as China,
Jordan and India. While new The full multidimensional framework
de-countries are being added to the list on an veloped for the US
economy can be successful-ongoing basis, it is very important to
note that ly applied, in principle, for other marketso far, the
selected group of leading indicators economies as well. This effort
is already under-
Table 4aTiming at growth rate cycle peaks and troughs, long
leading indexes, 13 countries
Country Number of Number of Long leading indexesgrowth rate
cycle growth rate cycles average leads (months) attroughs peaks
Troughs Peaks Overall
US 13 14 2 7 2 7 2 7Canada 9 10 2 6 2 5 2 5Germany 8 7 2 12 2 10
2 11France 9 8 2 4 2 7 2 6UK 8 8 2 5 2 11 2 8Italy 4 3 2 13 2 13 2
13Switzerland 6 7 2 12 2 8 2 10Sweden 6 6 2 4 2 9 2 6Japan 11 11 2
7 2 8 2 7Korea 6 6 2 9 2 8 2 8Australia 13 13 2 7 2 6 2 6Taiwan 9 9
2 6 2 5 2 5N.Z. 7 6 2 6 2 5 2 6
a Source: Economic Cycle Research Institute, New York City.
-
344A
.Banerji,L
.Hiris
/International
Journalof
Forecasting17
(2001)333
348
a During the period for which data are available (1984present),
the Chinese economy experienced no business cycle recessions.
Source: for the United States, National Bureauof Economic Research
(NBER). For other countries, Economic Cycle Research Institute, New
York City. Note: Shaded cells represent periods for which data are
not available.
Table 5aBusiness cycle peak and trough dates, 18 countries,
194898
-
A.Banerji
,L.H
iris/
InternationalJournal
ofForecasting
17(2001)
333348
345a Source: Economic Cycle Research Institute, New York City.
Note: Shaded cells represent periods for which data are not
available.
Table 6aGrowth rate cycle peak and trough dates, 18 countries,
19492000
-
346 A. Banerji, L. Hiris / International Journal of Forecasting
17 (2001) 333 348
way. In fact, like the US economy, at least a What is remarkable
about the consistent leadsdozen other market economies exhibit pro-
shown in Tables 3 and 4 is that virtually thenounced, pervasive and
persistent movements in same indicators are used for the long
leadingboth the levels and growth rates of aggregate indexes for
every country, showing how robusteconomic activity, which are
characteristic of these cyclical patterns remain in spite of
thebusiness and growth rate cycles. structural differences among
the different
As Table 3 shows, the long leading indexes economies. The
robustness of these leads alsofor each of these countries show
clear leads at suggests that these same long leading indexespeaks
and troughs of the business cycle. As can be used as reliable
leading indicators ofTable 4 shows, the growth rates of these long
economic cycles. It follows, therefore, that theseleading indexes
also consistently lead the peaks long leading indexes are likely to
remain reli-and troughs of their respective growth rate able even
after the economy being monitoredcycles, by about half a year to a
year on undergoes significant structural changes.average.
4.2. Reference chronologies 5. ConclusionsIt is important to
understand the vital impor- It has been shown that economic cycles
can
tance of a proper set of reference dates for take the form of
business cycles or growth rateinternational business cycles and
growth rate cycles. It is important to note that in both
casescycles. Those reference chronologies really they are
characterized by the pronounced,define the cycles that we seek to
compare and pervasive and persistent movements of cyclicalcontrast
across the countries, so it is critical that coincident indicators.
Such movements are thethey should be selected on the basis of a
hallmark of cyclical patterns seen in a wideuniform set of
procedures based squarely on the range of market economies, as well
as in criticalNBER approach long used in the US. The aspects of the
US economy, and sectors of theend-result of these efforts is
displayed in Tables US economy. In all of these cases, it is also55
and 6. These reference chronologies can now possible to design
composite leading indexes toserve as benchmarks for use by other
research- anticipate these cyclical turning points ers who seek to
perform cross-country com- another characteristic of cyclical
processes.parisons of cyclical patterns. The same set of indicators
used to create the
The business cycle peaks and troughs for US Long Leading Index
can be used to putcountries other than the United States were
together long-leading indexes for a variety ofchosen by applying
the same NBER-type clus- market economies. Not only do these
economiestering approach, and these reference cycle dates exhibit
business cycles and growth rate cycles,are shown in Table 5. The
reference cycle dates but these cycles are consistently anticipated
byfor the growth rate cycles for the same countries cycles in the
long leading indexes for all thosewere also chosen using analogous
procedures, countries.and are shown in Table 6. Both sets of
reference Thus, what the research has shown is thecycle dates were
determined by the Economic durability of the phenomena known as
businessCycle Research Institute under the guidance of cycles and
growth rate cycles, and the robust-Geoffrey Moore. ness of the long
leading indexes that anticipate5 these cycles in countries that
have differingThe latest updates to these reference chronologies
areavailable at www.businesscycle.com. economic structures.
-
A. Banerji, L. Hiris / International Journal of Forecasting 17
(2001) 333 348 347
The success of the multidimensional frame- associate editor of
this journal, for meaningfulwork for the US strongly suggests the
same and insightful comments.approach should be fully extended to
othercountries. In addition to the coincident and long
Referencesleading indexes of economic activity for all
theeconomies, future inflation gauges have already
Banerji, A. (1999). The three Ps: simple tools for moni-been
constructed for the United Kingdom, Ger-toring economic cycles.
Business Economics 34(3), 72
many, France, and Japan. Also, a short leading 76.index, as well
as leading indexes of exports, Banerji, A., & Klein, P. A.
(2000). Father of leadingimports and the trade balance have been
con- indicators: the legacy of Geoffrey H. Moore. Interna-
tional Journal of Forecasting 16(2), 277282.structed for the UK,
while a short leading indexBasu, S., & & Taylor, A. M.
(1999). Business cycles inand a leading manufacturing index have
been
international historical perspective. Journal of Econ-developed
for Japan.omic Perspectives 13(2), 4568.Over the last few decades,
the application of Boschan, C., & Banerji, A. (1990). A
reassessment of
classical approaches to indicator analysis has composite
indexes. In: Klein, P. A. (Ed.), Analyzingbeen both extensive and
intensive, and has modern business cycles, M. E. Sharpe, New
York.
Bronfenbrenner, M. (Ed.), (1969). Is the business cycleyielded
robust systems of indicators that haveobsolete, Wiley-Interscience,
New York.produced accurate turning point forecasts. An
Bry, G., & Boschan, C. (1971). Cyclical Analysis of
Timeimportant step has been the development of a Series: Selected
Procedures and Computer Programs,multidimensional framework, which
provides National Bureau of Economic Research, New York.much more
nuanced and fine-grained insights Coons, J. (2000). Fed is poised
to lower rates. The
Huntington Economic Monitor 15(21), 13.into the different
aspects of an economy thanCullity, J., & Banerji, A. (1996).
Procedures for construct-earlier work on economic indicators. The
work
ing composite indexes: a re-assessment. Presented at theof the
past few decades, therefore, has mademeeting on OECD leading
indicators, 1718 October,
significant strides in fulfilling Mitchells dream Paris.of a
systematic record of cyclical alternations Cullity, J., &
Moore, G. H. (1990). Long-leading andof prosperity and depression,
covering all coun- short-leading indexes. In: Moore, G. H. (Ed.),
Leading
indicators for the 1990s, Dow Jones-Irwin, Homewood.tries in
which the phenomena have appeared,Fabricant, S. (1972). The
recession of 196970. In:and designed to make clear the recurrent
fea-
Zarnowitz, V. (Ed.), The business cycle today, NBER,tures of the
fluctuations (Mitchell, 1927) to New York.serve the purpose of
serious theory-building. Hiris, L., & Guha, D. (2001). A
cyclical framework for
forecasting trade flows. In: Fatemi, K. (Ed.), Interna-tional
public policy and regionalism at the turn of thecentury, Pergamon,
Amsterdam.Acknowledgements
Ivanova, D., Lahiri, K., & & Seitz, F. (2000).
Germaninterest rate spreads as predictors of German inflationThe
authors gratefully acknowledge the im-and business cycles.
International Journal of Forecast-
measurable debt to their mentor, Geoffrey H. ing 16(1),
3958.Moore, whose pioneering work on a great King, R. G., &
Plosser, C. I. (1989). Real business cycles
and the test of the Adelmans. Unpublished manuscript,variety of
leading indexes inspired the develop-University of Rochester.ment
of the framework described herein. We
Klein, P. A. (1993). Geoffrey H. Moore and dynamicfurther
acknowledge the contribution of ourstatistical methods.
International Journal of Forecasting
colleagues at ECRI who helped guide the 9(1), 3137.frameworks
evolution. In addition, we thank Klein, P. A., & Moore, G. H.
(1985). Monitoring growthJohn B. Jones, a referee, and Kajal
Lahiri, an cycles in market-oriented countries: developing and
-
348 A. Banerji, L. Hiris / International Journal of Forecasting
17 (2001) 333 348
using international economic indicators, Ballinger for Watson,
M. W. (1994). Business cycle durations andNBER, Cambridge, MA.
postwar stabilization of the US economy. American
Klein, P. A. (19980. Swedish business cycles assessing Economic
Review 84, 2446.the recent record. Working paper 21. Federation of
Wessel, D. (1990). Leading index of the economy declinesSwedish
Industries, Stockholm. again October drop by indicators was fourth
in a row;
Layton, A. P., & Moore, G. H. (1989). Leading indicators
recession now signaled. The Wall Street Journal, De-for the service
sector. Journal of Business and Econ- cember 3, A2.omic Statistics
7(3), 379386.
Mintz, I. (1969). Dating postwar business cycles: methods
Biographies: Anirvan BANERJI is Director of Researchand their
application to western Germany, 195067. of the Economic Cycle
Research Institute (ECRI), whereOccasional Paper No. 107. NBER, New
York. he served earlier as Co-Director of Research with
Geoffrey
Mitchell, W. C. (1927). Business Cycles: The Problem and H.
Moore. Prior to that, for over a decade, he worked withIts Setting,
NBER, New York. Moore at Columbia Universitys Center for
International
Moore, G. H. (1981). A new leading index of employment Business
Cycle Research (CIBCR). His recent publica-and unemployment.
Monthly Labor Review, July, 44 tions, primarily in the area of
economic cycle analysis,47. have appeared in the International
Journal of Forecasting,
Moore, G. H. and Kaish, S. (1983). A new inflation the Journal
of Economic and Social Measurement andbarometer. The Morgan
Guaranty Survey, July. Business Economics.
Niemira, M., & Klein, P. A. (1994). Forecasting financialand
economic cycles, John Wiley, New York. Lorene HIRIS is Professor of
Finance at the C.W. Post
Roth, H. L. (1986). Leading indicators of inflation. Econ-
Campus of Long Island University. She is also Senioromic Review,
Federal Reserve Bank of Kansas City Research Scholar at ECRI where
her research and publi-71(9), 320. cations focus on the cyclical
behavior of credit, inflation,
Sebastian, P. (1990). No recession yet, says panel of and trade.
Previously, she worked with Geoffrey H. Mooreexperts: group
tracking business cycle sees slow growth as a Research Associate at
CIBCR.ahead. The Wall Street Journal, March 3, B3B.
Stock, J. H., Watson, W. W. (1998). Diffusion
Indexes.Unpublished manuscript, NBER.