A New Look at the
NBER/BEA 'Leading Indicators'

  1. Conceptual Flaws in the NBER/BEA "Leading Indicators"
  2. Criteria for More Appropriate Indicators
  3. Chart Presentations for More Incisive "Visual Analysis"

A. Conceptual Flaws of the NBER/BEA Indicators

Economic "science" and policy are now in such serious disarray that they are often the butt of derisive jokes by businessmen, journalists, students and even economists themselves: "If X number of economists were laid end to end they'd never reach a conclusion;" "A panel of 6 economists will have 10 points of view;" "If you are unemployed it's a recession; if I am unemployed its a depression;" etc. Much of the public doesn't understand even the basic concepts of present economic "science" and even feel it's irrelevant to the real world they live in. A main reason is conceptual confusions and complexity reminiscent of the Ptolemaic (earth-centered) theory of astronomy before Copernicus recognized that the earth revolved around the sun rather than vice versa.

The "Composite Index of Leading Indicators" -- developed by the National Bureau of Economic Research (NBER) and now published by the Department of Commerce's Bureau of Economic Analysis (BEA) in the "chart book" section of the Survey of Current Business (SCB) -- is a good example of these conceptual confusions. It is generally supposed to "forecast" turns in the "business cycle" three to six months in advance. But as a forecaster it is unreliable and even misleading. Since 1958, its NBER-reported "lead" has varied from 8 to 18 months, not 3 to 6. With such poor performance, the wonder is that it has taken so many years for user disillusionment -- eloquent testimony to the continued powerful prestige of the NBER, the still-accepted authority for dating "business cycle" turns.

The impending transfer of calculation and publication of these indicators to the Conference Board presents a window of opportunity for their long-needed fundamental restructuring. Here is a very brief summary of their most basic conceptual flaws.

1)  The NBER/BEA concept of an inherent "business cycle" is an obsolete myth.

There are undoubtedly cyclical tendencies in a "free enterprise" economy -- cumulative unsustainable structural changes which inherently tend to end "expansions" (including a tendency to inflation, which induces the Federal Reserve to "step on the brakes") before the economy reaches a conceptually and economically appropriate definition "full" employment, and then a downward "spiral" interaction between inventories and other business investment, consumer credit and incomes during subsequent "contractions" until the Fed again decides to switch from brakes to accelerator.

However, instead of a genuinely autonomous "business cycle," there is good empirical evidence that at least since WW II every U.S. recession and every recovery has been caused mainly by specific, identifiable actions (or intentional inactions) of government policy- makers, and that the inappropriate nature and timing of these policy decisions have been a key cause of the observed irregularity of the supposed "business cycle."

2)  The NBER/BEA "business cycle" conceptual framework uses a functionally inappropriate "static" absolute standard of reference for its empirical analysis of lead/lag relationships.

This simplistic ("black and white") framework has only two phases:

In this conceptual framework, 10 straight years of stable "expansion" at a 0.1% rate of real growth without any NBER-designated contraction could result in an unemployment rate like 1982-83 -- the worst depression since WW II.

The conceptually appropriate growth-trend standard of reference. When the basic trends of both labor force and productivity are continually growing, as in the U.S., the economy's inherent growth-trend is the most conceptually-sound definitional standard of reference for describing the general condition of the economy. That perspective permits far more functionally significant, conceptually precise and empirically measurable descriptive terms:

3)  When viewed in a conceptually appropriate growth-trend perspective, the NBER/BEA "Composite Index of 11 Leading Indicators" is not a "leading" indicator at all, but rather a very good COINCIDENT indicator.

Almost all its peaks are very close to the corresponding ORE/PUR quarterly peaks -- 1959:2, '68:3, '73:1, '78:4, '80:4, '89:1. In most cases they are even closer than the NBER/BEA "Composite index of 4 coincident indicators" (#920, p.C-7).

Moreover, the leading indicator composite shows three additional peaks -- '66:1, '84:1 and '94:4 -- which also coincide with ORE/PUR peaks but are not NBER-designated as such because they were followed only by an ORE/PUR (and TRGR) recession, not by an NBER- designated contraction.

The BEA indicator charts do not include the ORE/PUR or GDP gap, presumably because these concepts are foreign to the NBER's "business cycle" conceptual framework. However, the dates of ORE/PUR peaks can be visually approximated on the BEA GDP chart (#55 p.C-14) by drawing a peak-to-peak line from '68:3 to '73:1, then to '89:1 and projecting this to '95.)

4)  The apparently erratic lead-times of the NBER/BEA "leading" indicators are due mainly to the inappropriateness of its "business cycle" conceptual framework and absolute standard of reference.

Most of its component series are, like the ORE/PUR itself, functionally related to the economy's overall growth rate. Conceptually, a series' growth rate must decline to zero before its absolute value even begins the decline which turns an ORE/PUR (and TRGR) recession into an NBER contraction. The quicker this transition occurs, the shorter will be the fictitious "lead-time" of the NBER/BEA indicator. The "business cycle" conceptual framework simply assumes a more consistent pattern of economic fluctuation than usually occurs in the real world. When economic recessions and recoveries are actually determined primarily by economic policy errors, there is no good reason to expect consistency.

5)  Conceptually, none of the index's 11 component series should be considered genuine leading indicators because they do not effectively reflect any of the key functional causes of economic fluctuations

(Some of these are listed in Section II.) Functionally, most of these indicators represent effects, rather than causes. (Of course, in the theory of autonomous "business cycles," each "effect" is also a contributing "cause" of the next stage of the "cycle.")

Three of the present 11 components are not even reliable coincident indicators.

When presented as ratios to an appropriate growth-trend, some of the other BEA "cyclical indicators" may turn out to be even more reliable leaders than those now so designated.

6)  The BEA presentation of the component series uses a variety of scales and mixes absolute-value growth-trend series (#5,8,19,20,29,106) with non-trend series (#1,32,83, 92,99) in the same chart-panels.

This makes it difficult to relate them to each other or to recognize their functional significance.

B. Conceptual Criteria for More Appropriate Indicators

1)  The basic policy assumption: "laissez-faire" acceptance of a fictitious "business cycle" vs a systematic effort to manage the economy for stable full-employment economic growth.

The NBER "business cycle" conceptual assumption of an autonomic cycle implicitly ignores the key role of government policy interventions. But in the real world, active government interventions - - or intentional policy inactions -- are a primary cause of recessions and sub-capacity operation.

The most destabilizing policy inactions result from the political refusal to view the economy as an integrated "business enterprise," and the related fact that maintaining structural balance and stable growth actually requires comprehensive, systematically- coordinated macroeconomic management of money, credit and structural balance -- rather than relying so heavily on the Fed's present crude ad hoc "accelerator" and "brake" interventions. This political refusal is perhaps most clearly evident in the way the 4% unemployment policy target legally mandated by the Humphrey/Hawkins (H/H) "Full-Employment and Balanced Growth Act of 1978" has been effectively ignored or explicitly rejected, in practice, by almost all government agencies ever since it was enacted.

\Development of reliable leading indicators requires explicit recognition of the importance of government policy interventions and parallel recognition of the times and places where intervention or intentional non-intervention is a key determining factor in the rate of economic growth. Government actions are often themselves the best "leading indicators."

2)  The growth-trend perspective and definition of "potential"

In the 1920s when the NBER first developed its "business cycle" analysis there were few reliable economic statistics, so there was probably justification for its initial use of an absolute standard of reference. But with today's much more adequate statistics, there would seem to be no scientific reason for continuing to use this static standard of reference.

It is of course possible to calculate a series' growth trend by some sort of moving average of its own actual values. But to be most analytically useful most series' growth-trends should be based directly or indirectly on a conceptually and statistically sound definition of the economy's overall productive potential and ORE/PUR.

The Congressional Budget Office (CBO) includes in its semi-annual Economic and Budget Outlook two charts with a growth-trend perspective:

But even these charts are conceptually inappropriate, on three counts: the concept of "potential GDP" is too low, the "gap" concept is inverted, and the GDP chart has an arithmetic scale.

Definition of "potential" GDP. The dictionary defines "potential" as "possible but not yet [or infrequently] realized." That is the way manufacturing capacity is defined in calculating the familiar capacity utilization rate. However, the CBO now defines "potential" in terms of the "Non-Accelerating Inflation Rate of Unemployment" (NAIRU). That is two percentage points higher than the 4% unemployment policy target legally mandated by the 1978 H/H Act and generally accepted before the severe mid-'80 to mid-'84 depression. Thus, in the CBO's "GDP Gap" chart its "potential" was exceeded in 5 of the 21 years since the 1973 economic watershed, and in 16 of the 24 years before that. That clearly poses a key question: is the CBO/NAIRU definition of "potential" GDP really appropriate as either a scientifically useful analytical standard of reference or as a political policy goal?

Producing our huge WW II GDP brought unemployment down to near 1%. From mid-1951 to mid-1953 (during the Korean War) we had only 3% unemployment without inflation. For 14 straight years before the 1973 -- the economic watershed when world economic policy became focused more on inflation than on optimizing output and employment -- demilitarized Japan and Germany both kept their unemployment close to 1%, and the rest of the industrial world close to 2%. Thus, it seems clear that the CBO's NAIRU definition seriously underestimates the American economy's real "potential" for output and employment.

NAIRU is also not an appropriate standard of reference for evaluating the economy's current operating rate. The present high rate of unemployment is a significant cause of the present federal deficit. The present two percentage points over the H/H 4% rate costs the budget about $150 a year -- enough to nearly eliminate the deficit and/or finance such badly needed public investments as infrastructure and education-system rehabilitation. High unemployment is also generally recognized as a significant cause of the present high (and internationally embarrassing) rates of crime, poverty, welfare dependency, race prejudice and race conflict, urban decay, excessive income inequality, illiteracy and inadequate skills training, and a host of other socially damaging "negative indicators." Can our country really afford to continue its political acceptance of the NAIRU definition of "potential" output and employment? On this question the requirements of economic science, business profitability, fiscal responsibility, national pride, social peace and religious morality all tend to coincide.

3)  The "political" implications of the growth-trend perspective and definition of potential.

Scientifically, it should be no more difficult to develop an analytically sound estimate of potential GDP than it is to estimate manufacturing capacity. In a mismanaged roller-coaster economy which seldom approaches very close to really full utilization of its productive resources they both have conceptual definition and empirical estimating problems. But the congressional and public debates regarding the Humphrey-Hawkins Act, and even current debates on monetary policy, show that choosing a precise definition of the whole economy's potential output has a "politically sensitive" implication: focusing analytical attention on "full" employment and "potential" output implicitly suggests the desirability of trying to achieve that goal by more effective economic management, something that conservatives tend to oppose. In today's inflation-phobic environment the CBO deftly avoids that problem by its NAIRU definition of "potential" and by turning the normal GDP gap chart upside down, so that areas above the "potential" line don't correspond with the unemployment rate but indicate inflationary "excess" output and employment which implicitly need to be reduced by even tighter monetary policy.

But trying to avoid that political problem by using a conceptually inappropriate definition of potential, or a static absolute standard of reference, has led the NBER, BEA and CBO to absurdly unscientific results.

4)  Genuine leading indicators reflect basic causal relationships

Because Wall Street "Fed watchers" and many other forecasters recognize that the Fed largely controls the rate of both nominal and real economic growth, their "forecasting" is mainly concerned with forecasting Fed policy. [3] Thus, the genuine leading indicators are those which effectively reflect the functionally-causal factors which the Fed does (or should) take into account in making monetary policy, such as:

However, since the effective lead-time of money growth over GDP growth is so short (usually less than one quarter), and since monetary policy may be constrained by other considerations than the GDP growth rate, this key monetary indicator needs to be supplemented by other series which indicate the more basic structural changes which the Fed takes into account, particularly those which indicate how fast the economy's output is approaching its currently-effective capacity -- i.e. whether demand is growing faster than supply and thus tending to generate temporary inflationary pressures. (The more gradual the "soft-landing" approach to "full-employment" output, and the more effectively government policy maintains a stable National Credit Balance (NCB -- the balance between the total supply and demand for credit), and facilitates the growth of adequately skilled labor supply, the more likely that growing capacity can keep comfortably ahead of growing demand.)

Other key indicators include:

C. Chart Presentations for more incisive "visual analysis"

Inventory-Adjusted GDP (IAGDP) -- estimated current trend value. Sophisticated inflation analysis uses "core" value price indexes to exclude volatile food and energy prices. Similarly, for calculating the ORE/PUR, MDR and structural ratios, and for dating "cyclical" turning points, the raw value of current GDP is too volatile because of inventory fluctuations. These are caused mainly by disruption of the normal flow of durable goods from manufacturers to wholesalers, retailers and consumers, which is caused in turn mainly by sharp and unexpected changes in monetary policy, consumer credit, and related competitive sales efforts (discounts, rebates), etc. Final sales is a better measure than raw GDP, but has the same problem in lesser degree. A sophisticated inventory-adjusted trend value of GDP would be much more analytically useful, and cause less volatile Wall Street and policy reactions.

The basic presentation format: ratios to growth-trend value. In this growth-trend perspective, each component indicator which is essentially a growth series (present #5,8,19,20,29,106) should be presented and analyzed mainly on a growth-trend basis -- as a % of IAGDP (or of potential GDP) or, where more appropriate, as a % of its own long-run growth-trend value. "Peak" and "trough" then refer to the peak of the growth-trend value rather than the absolute value. This presentation also facilitates the linking and analytical comparison of convenient short-run (1-15 year) charts and longer-run historical charts.

Compact presentation facilitates visual analysis. When growth series are shown as ratios to growth-trend values many related series can be presented close together, and in conjunction with growth-rate series. This not only saves space but also facilitates visual analysis of their turning point leads and lags and structural secular trends (if any). For example, if the GDP Gap and unemployment rate are placed very close together, the roughly one quarter lag of the unemployment rate will be obvious, and if the present "leading" indicator series are presented more appropriately, all 11 could probably be shown to better analytical advantage on a single page.

>Absolute value charts. Where it is also useful to view growth series in terms of their absolute values, analysis is facilitated if they are presented on a common ratio scale (unlike BEA charts). When a "growth-rate protractor" is included with such charts, a clear-plastic parallel-line overlay makes it possible both to analyze growth rates of a given series between points of the user's own choosing and to compare growth rates between different series, with considerable precision.

"Normal-line" shadings. The peaks of many functionally useful indicators tend to correspond closely with ORE/PUR peaks. But if their values (and thus the timing of their peaks) can be confidently determined only several months after the decline has begun they may not be useful as leading indicators unless their charts also give a clear indication of when their values first exceed the level that the Fed (and other economic policy-makers) consider structurally-appropriate. Thus, the chart for each indicator should be shaded above and below its structural-balance "normal" line to facilitate easy "visual analysis," with, wherever appropriate, special shading above an empirically-determined policy-related "warning" line.>

  1. Since it is not yet clear which of these terms is the most appropriate and easily remembered way to express this concept, the combined expression ORE/PUR is used in this paper.
    (back to ref 1)

  2. Parenthetical numbered references are to the Cyclical Indicators pages in the Survey of Current Business.
    (back to ref 2)

  3. cf. CBO, Economic and Budget Outlook, 1/95:
    p. 15-16: "Risks to the CBO Forecast," and
    p. 13-14: "Alternative Outlooks."
    (back to ref 3)

Written: October, 1995
To IEA home page