leading indicators and the 'prime mover' view

By FRANK de LEEUW

Leading Indicators and the "Prime Mover" View

THE "prime mover" view of a coun-try's economy holds that fluctuationsin economic activity reflect, to an im-portant extent, movements in certainfundamental forces. These forces,or prime movers, include monetaryand fiscal policies, regulatory deci-sions, foreign economic developments,demographic shifts, new technologies,droughts or bumper crops, and a fewothers.

The prime mover view is a broadframework with many variants. Key-nesians, monetarists, and supply-siders, for example, all use the frame-work, although they disagree sharplyabout the relative importance of var-ious prime movers and about exactlyhow prime movers affect economic be-havior. Some variant of the primemover view implicitly underlies manystatements before the economic com-mittees of the Congress, reports toboards of directors on the economic outlook, projections of econometric mod-els, and theories of business cycles.

If the prime mover view is valid,then measures of prime movers oughtto make highly useful leading indica-tors of economic activity. Statisticalmeasures of monetary and fiscal poli-cies, foreign economic developments,and other prime movers ought to fore-shadow overall economic changes, un-less their impacts are extremely speedyor highly complex.

In fact, indexes of leading indicatorsuse very few prime movers. The 11components of the U.S. composite in-dex of leading indicators include only1—the deflated money supply. Indexesfor other countries similarly have littlerepresentation of prime movers. Mostindicators measure either (1) an earlystage of a production sequence, such

NOTE.—Thomas Holloway and Geoffrey Mooremade helpful comments on an earlier draft of thisarticle.

as new orders or permits; (2) a dimen-sion of activity that is rapidly respon-sive to changes in the economic envi-ronment, such as average weekly hoursor profits; or (3) a market value that ishighly expectation sensitive, such asstock prices or commodity prices.

It is possible to explain the good per-formance of these commonly used indi-cators by theories of the dynamic be-havior of a cost-minimizing firm.1 Theprime mover view, however, is not atall inconsistent with these theories; itsimply holds that prime movers aremajor underlying causes of changes,for example, in new orders or profits. Ifthe prime mover view is valid, indica-tors that measure prime movers couldprovide even earlier clues to changesin economic activity than the variablesthat usually appear on lists of leadingindicators.

Why, then, are prime movers largelyabsent from lists of leading indicators?Possibly the prime mover view, in spiteof its popularity, is wrong or incom-plete. Possibly statistical measures ofpiime movers have major shortcom-ings. Possibly there is some technicalflaw in the method of selecting leadingindicators.

This paper is in three major sec-tions. The first section summarizesthe composition of composite indexesof leading indicators for 21 countries,based on the work of the Organisa-tion for Economic Co-operation and De-velopment (OECD). The second sec-tion discusses eight possible reasonswhy prime movers seldom appear onlists of leading indicators. Some ofthe reasons, if true, are valid rea-sons why prime movers should be ab-sent from such lists. Other reasonsare defects in the way prime moversare measured. The discussion of these

1. Frank de Leeuw, "Toward a Theory of LeadingIndicators," in Leading Economic Indicators: New Ap-proaches and Forecasting Methods, ed. Kajal Lahiri andGeoffrey Moore (Cambridge, United Kingdom: Cam-bridge University Press, 1989). Forthcoming.

reasons suggests remedies for thesedefects that might improve the per-formance of prime movers as leadingindicators.

The third section reports on usingsome of these remedies to construct anexperimental prime-mover-based lead-ing index for the United States. The re-sults from using the index suggest thatthere is substantial scope for improv-ing the representation of prime moversin compilations of leading indicators.

The Composition of OECDIndexes of Leading

Indicators

The OECD has identified leading in-dicators iii many countries by usingprocedures similar to the ones devel-oped in the United States. The resultsare useful for exploring the represen-tation of prime movers among leadingindicators. Table 1 shows, for eachof 21 countries, the number of indexcomponents in each of 5 categories:Early stage indicators, rapidly respon-sive indicators, expectation-sensitiveindicators, prime movers, and otherindicators.

These categories, which representdifferent rationales for leads, are use-ful for understanding the underpin-nings of leading indicators. Formany indicators, however, placementin more than one category could be de-fended; the entries in the table repre-sent primary classifications only. Forthis and other reasons, the table pro-vides no more than rough indicationsof the importance of the five categories.

For the 21 countries as a whole,prime mover components include 18money supply series, 8 export series, 8terms of trade series, and 2 leading in-dexes for neighboring countries. Theseseries represent only 20 percent of the

Digitized for FRASER http://fraser.stlouisfed.org/ Federal Reserve Bank of St. Louis

August 1989

24 SURVEY OF CURRENT BUSINESS August 1989

total number of components.2 Primemover components are greatly out-numbered by components representingother rationales.

Reasons for the LowRepresentation of Prime

Movers and Some Remedies

This section lists eight possible rea-sons why prime movers might fail toappear in compilations of leading indi-cators. Some reasons are problems re-lating to the theoretical connections be-tween prime movers and economic ac-tivity; others are problems in the pro-cedure for selecting leading indicators;still others are problems related to theadequacies of available time series.

For some of the problems, there areno easy remedies. For example, if nodata are available for a prime mover,then it cannot be used as a leadingindicator, short of the lengthy processof developing new data that cover asubstantial time period. For some ofthe problems, there are easier reme-dies. For example, if the unexpected,but not the expected, movements in aprime mover influence economic activ-ity, then it may be possible to extractan approximate measure of the unex-pected movements from the measure ofthe prime mover.

1. Some prime movers may notinfluence economic activity

In part, this is a semantic problem.The definition of prime movers couldbe restricted to those forces that havea discernible impact on economic activ-ity. Then there would be no such thingas a prime mover that does not influ-ence economic activity.

Based on such a definition, however,we could never be sure whether a spe-cific variable was a prime mover or not.Some empirical tests would indicatethat it was a prime mover; others, that

Table 1.—Components of OECD Leading Indexes by Type of Rationale

[Number of indicators]

2. Columbia University's Center for InternationalBusiness Cycle Research, under Geoffrey Moore's lead-ership, publishes indexes of leading indicators eachmonth for 10 industrial countries—the first 8 countriespresented in table 1 plus Taiwan and Korea. Primemovers have even less representation in these indexesthan in the OECD indexes; they account for only 3 per-cent of the total number of components. This low repre-sentation is not the result of testing and rejecting primemover series, however; it is the result of initially select-ing components for these indexes that replicated thecomponents of the U.S. composite index, which includesonly one prime mover series.

Country

Australia .CanadaFrance . .Germany, Federal Republic ofItaly ....JapanUnited KingdomUnited States . . .

AustriaBelgiumDenmarkFinlandGreeceIrelandNetherlandsNorwayPortugal .Spain *Sweden . ,Switzerland ..,

Total:NumberPercent

Early stageindicators l

22

32

13

2221112

22311

3318

Rapidlyresponsiveindicators 2

411133

1

2121121

2413

Expectation-sensitive

indicators 3

32321222

12232123

212

3821

Prime movers 4

222122

1

212322331

221

3620

Otherindicators 5

12

21443

1232433322412

4927

1. Includes new orders (amounts and tendency surveys), unfilled orders, construction approvals and starts, new company formation, andvacancies.

2. Includes average hours, profits, inventory change and level (amounts and tendency surveys), and bottlenecks (tendency surveys).3. Includes stock prices, changes in raw materials prices, tendency surveys of expected production and expectations about the economic

situation.4. Includes money, deposits, exports, terms of trade, and composite indexes for foreign countries.5. Includes production series, retail sales, motor vehicle registrations, claims for unemployment insurance, layoffs and new hires, price indexes,

unit labor costs, credit ratios, interest rates, foreign exchange holdings, foreign trade balances, surveys of stocks, and employment.NOTE.—The OECD leading indexes are described in Organisation for Economic Co-operation and Development, OECD Leading Indicators and

Business Cycles in Member Countries, 1960-1985, Sources and Methods No. 39, January 1987. The author classified the components of the leadingindexes using information from this study.

it was not. A more useful definition ofa prime mover, therefore, is any funda-mental force that, according to a well-articulated theory, influences economicactivity. Based on this definition—theone adopted in this paper—most ex-perts will at least agree on a core list ofpotential prime movers, and empiricalmaeroeconomists can concentrate onassessing the impact of specific primemovers.

During the 1960's and 1970's, em-pirical maeroeconomists did attempt toassess two prime movers—monetarypolicy and fiscal policy—by conductingextensive empirical tests of the relationof GNP changes to measures of thesetwo policies. Measures of monetarypolicy turned out to have a significantinfluence on economic activity in mostof these tests, while measures of fiscalpolicy had a much more mixed record.Some participants in the debate con-cluded that fiscal policy is an unimpor-tant prime mover. Other participantsin the debate criticized the tests.3

During the 1970's, an importantcriticism was that economic activity

3. For a recent review of this debate, see Karl Brun-ner, "Fiscal Policy in Macro Theory: A Survey and Eval-uation," in The Monetary versus Fiscal Policy Debate,ed. R.W. Hafer (Totowa, NJ: Rowman and Allanheid,1986), 33-116; Alan S. Blinder, "Ruminations on KarlBrunner's Reflections," Debate, 117-126; and Robert J.Gordon, "Comment," Debate, 127-136.

influences policy variables, in addi-tion to being influenced by them (thiscriticism is discussed in point 2 be-low). More recently, some maeroe-conomists have emphasized the dis-tinction between expected and un-expected changes in policy.4 Theywould, presumably, criticize tests ofthe kind that were common 20 yearsago on the grounds that expectedchanges in policy may have very lit-tle influence on economic activity,whereas unexpected changes in policymay have a substantial influence.

The expected-unexpected distinctionsuggests that policy variables mightbe absent from leading indicators, notbecause they contain no informationabout future economic activity, but be-cause tests have not focused on thecomponent of them that leads—the un-expected component. There is no fullysatisfactory method of separating mon-etary or fiscal policy variables intoexpected and unexpected components,but there are crude methods that maydo a serviceable job. Applying thesecrude methods might improve the per-formance of some prime movers asleading indicators.

4. Robert E. Lucas, Jr., "Econometric Policy Evalua-tion: A Critique," in Carnegie-Rochester Series in PublicPolicy, vol. 1, ed. K. Brunner and A.H. Meltzer (NewYork: North Holland, 1976), 19-46.


August 1989

August 1989 SURVEY OF CURRENT BUSINESS 25

2. Economic activity may influence thebehavior of prime movers

If a prime mover is itself influencedby economic activity, then it may notmake a good leading indicator. For ex-ample, if policymakers varied tax ratesso as to reduce, but not to eliminate,fluctuations in economic activity, thentax policy would tend to be expansion-ary when the economy is contractingand contractionary when the economyis expanding. Goldfeld and Blindershow that the estimated impact of sucha policy may well be biased towardzero in regression tests.5 The policyappears to have little impact becausetwo-way interaction invalidates simpletests.

The possible dependence of primemovers on economic activity poses amajor challenge to the prime moverview. If monetary policy, for exam-ple, could be explained largely as asequence of predictable reactions torecent economic developments, thenthere would be no reason to think ofmonetary policy as different from con-sumption decisions, wage determina-tion, or some other component of eco-nomic behavior. A leading index basedon prime movers would have no specialappeal.

In fact, monetary policy is surelymore than simply a sequence of pre-dictable reactions. The answer to thechallenge that predictable policy reac-tions pose is to try to measure the en-dogenous component (the componentinfluenced by current economic activ-ity) of a prime mover and subtract itfrom the total. The remaining ex-ogenous component—often referred toas the "innovation"—should be suitablefor testing in conventional ways and forinclusion in a leading index.

A well-known example of this ap-proach is the separation of FederalGovernment receipts and expendituresinto a cyclical portion and a cyclicallyadjusted, or high-employment, portion.Separating endogenous responses frominnovations, however, need not be re-stricted to policy variables. Any vari-able that is not a definitional identitymay at times have important innova-tions. The search for prime moversshould involve both removing endoge-nous influences from policy variables

and trying to detect innovations inother variables.6

3. Prime movers may generate repeatingcycles

Theories of business cycles oftenleave open the possibility that a main-tained shock to the economy—for ex-ample, a change in the rate of growthof the money supply—will producecontinuing cycles in economic activ-ity, rather than a single cycle ormovement.7 If economies actually be-have in this way, then it may be dif-ficult or impossible to relate a changein economic activity to its underlyingprime mover cause, because the causecould have occurred many years ago.

Econometric models of the economysuggest that repeating cycles are not aserious problem. For the DRI model,Eckstein states that, in response to atypical disturbance, "the stock-flow ad-justment processes do create a secondcycle, but of much smaller magnitude,usually less than one-third as large asthe original cycle."8 Unpublished sim-ulations of the BEA econometric modeldisplay similarly heavy damping.

4. Binary comparisons may mask someindicators

Standard procedures for selectingleading indicators rely totally on bi-nary comparisons. Researchers clas-sify each time series as leading, coin-cident, lagging, or unclassified by com-paring its turning points with the turn-ing points in aggregate activity. If theweighted sum, or the ratio, of two se-ries makes an excellent indicator buteach series separately does not, re-searchers would detect this fact only ifthey happen to define the sum, or theratio, of the two series as a separateseries to be tested.

It is possible that potential lead-ing indicators—and especially primemovers—go undetected because ofthese procedures. It would hardly besurprising if monetary policy were thecause of some movements in economicactivity, changing terms of trade werethe cause of others, and tax policy werethe cause of still others. If that werethe case, it could easily happen that

5. Stephen M. Goldfeld and Alan S. Blinder,"Some Implications of Endogenous Stabilization Policy,"Brookings Papers on Economic Activity 3 (1972): 597-613.

6. For an influential development of these ideas, seeChristopher A. Sims, "Macroeconomics and Reality,"Econometrica 48 (January 1980): 1-49.

7. See R.G.D. Allen, Mathematical Economics, 2nded., chs. 7 and 8 (New York: Macmillan Co., 1960).

8. Otto Eckstein, The DRI Model of the U.S. Economy(New York: McGraw-Hill, 1983), 51-52.

these three prime movers individuallywould be "unclassified" using the stan-dard procedures.

The solution to this problem is touse multivariate, rather than bivari-ate, analysis. Multiple regression isthe best known, but not the only, mul-tivariate technique available. It doesnot easily lend itself to the emphasis onturning points that has characterizedleading indicator research, but that isnot necessarily a drawback.

5. Emphasis on turning points maymask some indicators

Under standard procedures for se-lecting leading indicators, the centralrequirement for a series to be classi-fied as leading is that it shows turn-ing points that lead turning points inaggregate economic activity. Withoutthis turning-point relationship, a se-ries will not be classified as leadingeven if it provides an excellent indi-cator of future slowdowns or accelera-tions. Furthermore, with this turning-point relationship, a series is likely tobe classified as leading even if it hasother, "false" turning points.

This heavy emphasis on turningpoints in aggregate economic activ-ity is controversial. Indicators of thestrength of expansions or contractionsas well as of turning points wouldsurely be useful. The emphasis onturning points probably stems from thelimitations of time-series techniquesthat were available 40 years ago, lim-itations that modern time-series anal-ysis and computers have largely over-come.

This emphasis on turning pointsmay not be more relevant to primemovers than to other indicators.Whether it is or not, experimentationwith other procedures—for example,multiple regression—seems worth pur-suing.

6, Some prime movers may be reflectedin other leading indicators

The classification of leading indica-tors into prime movers and other cat-egories is crude, as has been noted. Itis possible that the low representationof prime movers in lists of leading in-dicators is partly a matter of misclas-sification.

Concretely, some of the movementsin early-stage variables may reflectchanges in prime movers. New ordersfor equipment or for durable goods, forexample, include new orders for ex-ports and new orders for government


August 1989


purchase. Construction permits or con-tracts may also include construction forgovernment use—even when (as in theUnited States) they exclude construc-tion that government agencies them-selves undertake.

7. Data for some prime movers may notbe available by month

Potentially useful leading indicatorsmay be missed because data for themare not available or because they areavailable only for years or quarters.With few exceptions, candidates forleading indicator indexes have had tomeet the requirement of availabilityfor each month. Government-sectorindicators—to cite one example of thisproblem—may be missed because mostcountries publish government-sectorvariables in detail only by quarter.

This study does not restrict itself tomonthly series, because its purpose isto explore new ideas rather than to pro-duce a monthly indicator. Accordingly,the empirical section of the study usesquarterly, rather than monthly, data.For quarterly series that appear to beuseful leading indicators, it may bepossible to devise serviceable monthlyseries by using available monthly in-dicators together with interpolationtechniques.

8. Data for some prime movers maymeasure late stages in a productioncycle

A final possibility is that the avail-able data for some prime movers maymeasure changes in production or de-mand at late stages, long after the im-pact on economic activity has begun.If measures of early stages of produc-tion or demand were available for theseprime movers, they might be classifiedas leading indicators; however, mea-sures may be available only for a stageat which the lead has disappeared.

In the United States, for example,the national income and product ac-counts (NIPA's) measure Federal Gov-ernment purchases of goods at the timeof delivery. For goods with a longproduction period, orders and progresspayments have been influencing mar-kets for some time before the purchasesappear in the NIPA's. A desirable so-lution would be to develop measures ofthese prime movers at earlier stages ofthe order-production-payment process.

A summary of proposed remedies

To conclude this section, it is use-ful to summarize the remedies thatcould overcome some (but not all) of theproblems that underlie the failure ofprime movers to be represented in in-dexes of leading indicators. The num-ber in parentheses following each rem-edy refers to the reason, in the preced-ing listing, under which the remedy isdiscussed.

• Separate the changes in a primemover into expected and unex-pected components (1);

• Separate the changes in a primemover into exogenous (innova-tions) and endogenous compo-nents (2);

• Use multivariate comparisons inselecting indicators (4, 5);

• Use quarterly data when monthlydata are unavailable (7); and

• Develop measures of prime moversat early stages of the order-production-payment process (8).

A Prime-Mover-BasedLeading Index

This section will show how usingsome of the proposed remedies can im-prove the performance of prime moversas leading indicators. The section be-gins with a list of some widely usedprime movers. Because many of theseprime movers initially affect current-dollar income, the section then dis-cusses how to relate them to fluctu-ations in real activity. Next, it de-scribes the application of some of theremedies to the variables representingprime movers. The section concludesby presenting a tentative, quarterlyleading index based on prime movers.

Some prime movers

Monetary and fiscal policies are theprime movers that have received themost attention. There are three com-peting time series used to representmonetary policy: The monetary base,or currency plus reserves adjusted forreserve requirement changes; Ml, orcurrency plus checkable deposits; andM2, or Ml plus small-denominationtime and savings deposits and cer-tain other highly negotiable assets.The preliminary empirical work in thisstudy tested each of these three series.Because M2 consistently gave better

results than the others, the equationreported later in this section uses M2.

For fiscal policy, the familiar prob-lem of the dependence of Govern-ment receipts, expenditures, and debton current economic activity has anequally familiar solution—the deduc-tion of estimated cyclical effects fromthe actual figures. The study thereforeutilized BEA's cyclically adjusted Fed-eral expenditures, receipts, and debt.9

The equation reported later in this sec-tion uses only cyclically adjusted ex-penditures, because that is the onlyone of the three measures that yieldedstatistically significant results.

The study also tested a measureof defense activity at an early stagerelative to activity at a later stage.The measure—new orders for defenseproducts relative to shipments of de-fense products—is available beginningin 1968. Even after various methodsof smoothing, it did not prove signifi-cant in any regression tests, and it isnot discussed further.

To represent the impact of foreigneconomic developments on the U.S.economy, preliminary work utilizedboth transactions measures and pricemeasures. The transactions measureswere total U.S. exports, exports ofgoods, and exports of services. Theprice measures were the ratio of acrude petroleum price index to the U.S.producer price index and the ratio of afixed-weighted price index for importsrelative to one for GNP. Two of theseproved statistically related to economicactivity: Total exports and the relativeprice of imports.

For various reasons, the preliminarywork did not include any other primemovers. One candidate, regulatory de-cisions, is difficult to capture in oneor two time series. Two others, majorstrikes and farm supply shocks, rarelyaffect aggregate activity and probablydo so with very short lags. As fordemographic shifts and technologicalchange, it is at least plausible to sup-pose that they affect longrun growthmore than shortrun fluctuations—butthis proposition is a matter of contro-versy among macroeconomists. Morecareful and complete empirical workshould certainly test these, and per-haps other, prime movers.

9. Estimates based on a middle-expansion GNPtrend were used through the second quarter of 1974.Estimates based on a GNP trend corresponding to a6-percent unemployment rate—the only estimates pub-lished currently—were used starting in the third quar-ter of 1974, a quarter in which the middle-expansionunemployment rate trend was 6 percent.


August 1989


Nominal income and real activity

A number of prime movers—for ex-ample, monetary aggregates, Federalexpenditures, and exports—have theirinitial impact on nominal, or current-dollar, income. Because the focus ofthis study is on real economic activ-ity, it is necessary to transform theseprime movers into indicators of realactivity.

One approach is to deflate them us-ing a broadly based price index. That isthe way the official U.S. index of lead-ing indicators handles its one primemover component—the deflated moneysupply (M2). The obvious way to con-vert a nominal variable to a real oneis to deflate it; however, applying de-flation to a nominal leading indica-tor implies that price changes affectreal economic activity with the sametime pattern as changes in the nomi-nal prime mover, but with the oppositesign. There is, unfortunately, no rea-son to expect such symmetry.

A less restrictive, and thereforepreferable, approach to the relation be-tween nominal and real activity beginswith the assumption that the excess ofactual inflation above some normal orexpected level is related to the recentgap between actual output and trend,or potential, output.

The mathematical representation ofthis assumption is

(i) i>p = Dp6 + a(y-y*),,where DP is the actual inflation rate;DPe is the expected inflation rate, of-ten represented by lagged actual infla-tion rates; Y— Yt is actual minus trendoutput; and the subscript / indicates arecent time period or periods. The rateof change of nominal income, by defi-nition, is equal to the sum of rates ofchange of prices and of real income:

(2) £>y$ = DP + DY,

where DY$ and DY are rates of changeof nominal and real income, respec-tively.

Solving for DY gives

(3) DY = DY$-DPe-a(Y-Yt)i.

Equation (3) suggests a relationshipbetween the growth of real incomeand three sets of variables: (i) Thosethat influence nominal income change(DY$), (ii) the expected inflation rate(DPe\ and (iii) recent actual output mi-nus trend output •(Y—Yt). Incorporatingvariables (ii) and (iii) in a multivariate

analysis is a more satisfactory way ofdealing with nominal-real interactionthan simply deflating nominal primemovers.

The preliminary empirical work,therefore, utilized multiple regressionsof the general form

(4) DY = + b2DPM2i2+

where DY is the growth rate of realGNP, DPMln and DPM2i2 representlagged values of rates of change inprime movers transformed in appropri-ate ways, DP/s represents lagged val-ues of rates of change in a general priceindex, and (y-yt)/4 represents laggedvalues of actual minus trend output.

Applying some of the remedies

Two of the remedies proposed forimproving the performance of primemovers as leading indicators are us-ing multivariate, rather than bivari-ate, analysis and using quarterly datawhere monthly data are not available.The empirical work in this study im-plements these remedies by using mul-tiple regression analysis of quarterly(seasonally adjusted) data.

Another remedy is sepa-rating changes in a prime mover intoexpected and unexpected components.The present study used, as a simple ex-pedient, second differences as a proxyfor unexpected changes. For mone-tary aggregates, actual changes con-sistently gave better results than anymore complex transformations; how-ever, for the other prime movers, sec-ond differences gave significant re-sults, while actual changes or actualchanges less second differences did not.

Two other remedies are separatingthe changes in a prime mover into ex-ogenous and endogenous componentsand developing a measure of a primemover at an early stage of the order-production-payment process. To sepa-rate exogenous from endogenous com-ponents, the study used cyclically ad-justed, rather than actual, Federal re-ceipts and expenditures. To measureprime movers at early stages, the studytested (unsuccessfully) ratios of ordersto shipments for defense products.

This summary of remedies shouldmake clear that this study has not ex-hausted the possibilities. The resultsare nevertheless promising enough tosuggest that prime movers may bestrong candidates for an improved

leading index. Further work is clearlyfeasible and could have a substantialpayoff.

Developing a prime-mover-based leadingindex

This study reports only the finalequation used to develop a prime-mover-based leading index. A numberof alternative equations gave roughlysimilar, but not quite as close-fitting,results. For this preliminary explo-ration, it does not seem useful to reportthese equations.

The best fitting equation of thosetested is

(5) DY =0.0046 + 0.517DM 23

(2.2) (4.4)

0.229JQDCAE73

(2.8)Q.Q91DDEXP3 -(2.0)

- 0.7S1DPY6 -(-5.5)(-2,2)

0.045GAP6

(-1.6)

period of fit: 1 953:1-1 986 :IV

R2 = 0.301, £>-W = 1.9.

The numbers in parentheses are t-ratios. The subscript 3 refers to aver-ages of values for quarters t — 1, t - 2,and t— 3. The subscript 6 refers to aver-ages of values for quarters t— 1 throught — 6, Other averages did not fit as wellas these three- and six-quarter aver-ages. See the box for definitions andsources for the variables.

Equation (5) has several notable fea-tures. Its coefficients and significancelevels are higher for the money and in-flation variables, DM2 and DPY, thanfor the other variables. This resultis consistent with many of the equa-tions testing the relation of GNP tomonetary and fiscal policies in the1960's and 1970's. Unlike many ofthose equations, however, the coeffi-cient of M2 in equation (5) is far be-low 1.0— perhaps because of the veloc-ity changes in the 1980?s.

The value of R2 is low, perhaps be-cause prime movers do not affect manyof the forces that cause quarter-to-quarter changes in GNP growth rates


August 1989


An Index of Prime Movers Compared With the U.S. Composite Indexes of Leading andCoincident Indicators

CHART 2

1982 = 100100

Ratio Scale 1982 = 100100

90

80

70

60

90

80

70

60

Prime-Mover-Based Index

50 * 50

Coincident Index

40 ' I I I I I I 'I I I I I I I I I I I I I I I I I I I I I I I I ! I I ! I I I I ! I I I I I | | | | j | ! | , | | | ! 40

150

1951 53 55 57 59 61 63 65 67 69

150

Prime-Mover-Based Index

60 * ' ' I I I ' I ' I I ' I I I I I I I I I I ! I I ' I I I '' I I I I I I I I I I I I I I I I I I I ! I I I I I I I !

1970

U.S. Department of Commerce, Bureau of Economic Analysis

- 70

60


August 1989


Definitions and Sources for the Variables in Equation (5)

Variable

DY

DM2

DDCAE

DDEXP

DDPREL

DPY

GAP .

Definition

Change in the natural log of real GNP

Change in the natural log of M2

Second difference in the natural log ofcyclically adjusted Federal expenditures.

Second difference in the natural log of U.S.exports, national accounts basis.

Second difference in the natural log of anindex of relative import prices.

Change in the natural log of the fixed-weightedprice index for GNP.

Natural log of real GNP less natural log oftrend GNP.

Source

BEA, table 1.2

Federal Reserve Board, 1959-88;BEA, pre-1959

BEA, March 1986 and March1988 SURVEY OF CURRENTBUSINESS; pre-1955, author

BEA, table 4.1

BEA, table 7.1, ratio of importprices to GNP prices

BEA, table 7.1

BEA, table 1.2 and sources forDDCAE

NOTE. — References are to national income and product accounts tables published in the SURVEY OFCURRENT BUSINESS and its supplements, unless otherwise indicated.

(including measurement errors). Se-rial correlation of the residuals fromthe equation, however, is not a prob-lem; this suggests that the equationmay be a better guide to changes overseveral quarters than to changes fromone quarter to the next.

Predicted values from equation (5)form the basis for a prime-mover-basedleading index. There are three maindifferences between the predicted val-ues and the desired index. First,the measure of economic activity—realGNP—used in equation (5) differs inlevel and in amplitude of change fromthe indexes of leading and coincidentindicators with which it is desirableto compare the new prime-mover-basedindex. Second, the predicted valuesrefer to quarterly changes, whereasa leading index should refer to lev-els. Third, the predicted values arebased on distributed lags of the ex-planatory variables, whereas a lead-ing index uses current values of thosevariables.

Because of these differences,a prime-mover-based index requirestransformation of the predicted valuesfrom equation (5). The equation oftransformation is

(6)

0.003}

where PMIt is the index for quarter2, PMIt-i is the index for quarter i —1, Exp denotes exponentiation, andDYPREDt is the predicted value fromequation (5) based on current ratherthan lagged values for all variables.

The two coefficients in equation (6)are the result of trial and error. Theywere set so as to match, approxi-mately, the cyclical amplitude and av-erage growth rate of the official U.S.leading index. The first coefficient (4.0)mainly affects cyclical amplitude, andthe second coefficient (0.003) mainly af-fects average growth rate.

The performance of theprime-mover-based leading index

Chart 2 compares the prime-mover-based index and the official U.S. com-posite leading and coincident indexes.The following are some of the morenoteworthy features of the chart:

• The prime-mover-based index, likethe leading index, unquestionablyleads the coincident index;

• The average lead of the prime-mover-based index is at least aslong as that of the leading index;

• The prime-mover-based index doesnot have the pronounced falseleads that the leading index dis-plays in 1967 and 1984; and

• The prime-mover-based index hasmore irregular quarter-to-quarterswings than do the composite in-dexes.

In conclusion, it is important to re-iterate the tentative character of theprime-mover-based index. The indexfails to implement a number of sugges-tions that may prove useful—for exam-ple, the representation of other primemovers and alternative techniques fortrying to measure expected and unex-pected, or endogenous and exogenous,components of prime movers. Chart 2is based entirely on the latest revisedestimates of all variables, and it in-cludes only a few years outside of theperiod of fit of equation (5). It wouldnot be at all surprising for the forecast-ing performance of equation (5) to beworse than its performance in the bot-tom panel of chart 2.

In spite of these shortcomings, thenew index and the work leading up toit strongly suggest that much can bedone to improve the representation ofprime movers in indexes of leading in-dicators. In any attempt to bring eco-nomic theories and leading indicatorscloser together, further investigation ofprime movers should play a major role.


August 1989

leading indicators and the 'prime mover' view

Documents