Edward Renshaw
Professor of Economics
State University of New York at Albany
In recent years there has been a revival of interest in leading economic indicators on the part of academics (Lahiri and Moore 1991) and some improvements in indicator construction and interpretation that might aid forecasters and policy analysts to make better use of the cyclical indicators that were pioneered by the National Bureau of Economic Research (NBER) in the 1920s and 30s.
In 1961 the Commerce Department's Bureau of the Census began to publish Business Cycle Developments which featured some of these indicators. In 1968 this publication was expanded to include previously unpublished composite indexes and its title was changed to Business Conditions Digest. In 1972 responsibility for the cyclical indicators was turned over to the Bureau of Economic Analysis. In April 1990 the indicators were integrated into the Survey of Current Business. In December 1995 the responsibility for maintaining and publicizing the composite indexes of leading, coincident and lagging indicators was turned over to the Conference Board.
In this essay we will provide a brief history of important changes in the construction of the index of leading economic indicators and then address such issues as how to reduce the number of false or highly premature indications of an impending recession. Persons who are worried about a possible recession should keep a close eye on the prime rate charged by banks and one of the index of leading indicators' most interesting components, new private housing units authorized by local building permits. One should also look for noteworthy declines in two or more of the most interest sensitive components of real GDP: expenditure for consumer durables, producers' durable equipment and residential construction.
The Commerce Department's first index of leading economic indicators was seriously distorted by a failure to deflate some of its components which were measured in current dollars. At the December 1969 peak in business activity the preliminary lead time for the index was only three months and at the November 1973 peak in economic activity the index registered a new all time high. The composite index was substantially overhauled in 1975 to correct this problem and at the January 1980 peak in business activity the new index had been depressed for a total of 15 months.
The success of the revised index after 1980, however, was impaired by other objectives. Zarnowitz and Boschan noted in (1975) that there are six criteria which are often used in selecting and assessing cyclical indicators: (1) Economic significance--how well understood and how important is the role in business cycles of the variable represented by the data? (2) statistical adequacy--how well does the given series measure the economic variable or process in question? (3) timing at revivals and recessions--how consistently has the series led (or coincided, or lagged) at the successive business cycle turns? (4) conformity to historical business cycles--how regularly have the movements in the specific indicator reflected the expansions and contractions in the economy at large? (5) smoothness--how promptly can a cyclical turn in the series be distinguished from directional change associated with shorter (mainly irregular) movements? and (6) currency or timeliness--how promptly available are the statistics and how frequently are they reported?
Many of the components in the composite index of leading indicators are not as trendy as real GNP or GDP. In order to satisfy objective (4) and make the index of leading economic indicators do a better job of conforming to the upward trend in what most economists would consider our best overall measure of economic activity, the Commerce Department added .142 percent to the index of leading economic indicators each month to achieve an overall targeted trend of .261 percent per month. This trend adjustment sometimes delayed the down turn in the index of leading indicators enough to obscure the risk of a recession. At the July 1981 business peak the new index of leading economic indicators had been depressed for only three months and at the July 1990 peak in economic activity it registered another historic high at a time when six of its eleven components had been depressed for 25 months or more.
In November 1993 the trend adjustment was eliminated. See the discussion by Green and Beckman in the October (1993) Survey of Current Business. This adjustment and a reweighting of the components to make the unadjusted lead times for the composite index conform more closely with the lead times for the median component of the revised index can be considered a major improvement. The index of leading economic indicators which was taken over by the Conference Board, however, was still plagued by a problem of false signals--which led Nobel Laureate Paul Samuelson to once quip that the index of leading economic indicators has correctly predicted 9 of the last five recessions.
In Table 14.1 we show some adjusted lead times at business peaks for the Conference Board's revised index of eleven leading economic indicators. The lead times are associated with cumulative declines amounting to one percent or more from the index's own peak value to those peaks identified by the National Bureau of Economic Research. A cumulative decline of this magnitude does help to reduce the number of false signals. Two false or rather premature signals, that are associated with tight monetary policy and stock market "corrections" in the midst of prosperity, can still be identified in the 1959-96 period, however.
One way to cope with the possibility of false signals is to ignore declines in the index of leading economic indicators until after the Fed has ceased to worry about a recession and has begun to raise short term interest rates in an effort to prevent an acceleration in the inflation rate. In the post 1947 period the US has never experienced a peak in business activity until three months after the average prime rate charged by banks has increased by one third or more from its cyclical low.
Once this amount of financial stringency has been imposed on the economy one should pay closer attention to one of the leading index's most interest sensitive components: the number of new private housing units authorized by local building permits.
Of the twelve components in the Commerce Department's original index of leading economic indicators, which was first released in November 1968, only three (the average workweek in manufacturing, residential building permits, and stock prices) have survived without substantial modification.
The average weekly hours of nonsupervisory workers in manufacturing provides a good example of an enduring component of the index of leading indicators that has always peaked out in advance of an economic recession but has sometimes not declined enough to provide a reliable warning of its occurrence (Renshaw 1992, Table 1.20).
The monthly change in the number of housing units authorized by local permit-issuing places has a better track record. The collected data are related to the issuance of permits and not the start of construction, which frequently occurs several months later and sometimes not at all.
In 1989 this analyst noted that residential building permits is the only component of the index of leading economic indicators that has consistently declined by at least twenty percent before each of the economic recessions in the post 1947 period. For the five cases where the number of housing units authorized by local building permits exceeded 1800 thousand units there was at least a 35 percent decline in the number of authorized units before the next recessionary peak identified by NBER. This set of relationships turned out to provide the only model or rule of thumb in Lahiri and Moore's collected compendium on Leading Economic Indicators that did a good job of identifying the July 1990 peak in business activity shortly before it occurred. See the last column of Table 14.2 for an updated version of this rule of thumb.
In May 1995, after the Conference Board's revised index of leading economic indicators declined by a little over one percent, the number of authorized building permits was only down about 17 percent and was heading upward in response to lower mortgage rates.
While none of the individual components of real GDP are very reliable recessionary indicators, history would suggest that a noteworthy loss of investment confidence in more than one interest sensitive sector of the US economy will often spread to other sectors and help to tip the economy into a recession.
This point is illustrated in Table 14.3, which shows the lead times from NBER peaks during business expansions for coincident annualized quarterly declines in real consumer durables amounting to more than five percent, declines in real producers' durable equipment over three percent, and declines in inflation adjusted residential construction amounting to more than nine percent--after there has been a sustained increase in the prime rate charged by banks of a third or more.
Other investment categories with a propensity to "collapse" during economic recessions (such as changes in business inventories and non residential structures) are ignored on the grounds that they tend to be lagging indicators at recessionary peaks. We also ignore small and large singular declines in consumer durables, producers' durable equipment and residential construction. If the overall economy is growing at a respectable pace, a large decline in only one investment category is more likely to be a temporary phenomenon and not spread to the rest of the economy.
All of the recessionary peaks experienced from 1947 to 1995 were preceded by at least one double loss of investment confidence. The longest lead time, following a one-third increase in the prime rate from its cyclical low, is ten months. This lead time is associated with the third quarter of 1952 when the production of many goods was disrupted by a 53 day strike by the United Steel Workers from June 3 to July 24.
The other lead times in column 5 of Table 14.3 occurred in the comparatively narrow range of from zero to seven months. Half of the lead times are in the very narrow range of from zero to only two months. For the three cases where there was a triple loss of investment confidence (1957-2, 1979-4, and 1990-2) the recessionary lead times are all in this narrow range.
One of the problems with such short lead times is that they are not likely to provide policy makers with enough time to implement measures that might prevent a recession from occurring or at least prolong the life of a business expansion. It makes sense, in any event, to try to explain the duration of the lead times in column (5) of Table 14.3 on the basis of what has been happening to other variables with leading indicator properties. The three month percentage changes in residential building permits and contracts and orders for plant and equipment in the last two columns of this table may be of some value in that regard.
Some of the most persuasive evidence in support of a longer run pattern to stock prices has been provided by the National Bureau of Economic Research. In a massive compendium which was published in 1961, Geoffrey Moore found that common stocks were excelled as a leading indicator of business cycles by only the net change in the number of operating businesses. Stock prices were classified as being a leading indicator 31 times, roughly coincident 14 times and a lagging indicator only five times.
The most publicized weakness of stock prices as an indicator of economic activity is a propensity for them to have predicted more economic recessions than have actually occurred. This point is illustrated in Table 14.4 where cumulative reversals amounting to 5.5 percent or more in the monthly average daily closing prices for the S&P composite 500 stock price index are used to define peaks and troughs in the index.
This reversal percentage is the smallest percentage that one can use and still eliminate all of the abortive stock market rallies which have occurred in the midst of recent economic recessions. Once the economy is in a recession and the stock market has recovered 5.5 percent on a monthly average basis the S&P index has consistently continued up for an additional gain of at least 25 percentage points before the index reversed itself on the downside by 5.5 percent. (See the gains in column 5 of Table 14.4 which are marked with a double asterisk.)
It will be noted that there have been 24 downside reversals since May 1946 and only nine recessions. Fifteen of these downside reversals occurred in the midst of a business expansion and were followed by an upside reversal of at least 5.5 percent before the economy became mired down in a recession. There are seven cases where a downside reversal of 5.5 percent was followed by a recession and two cases where a reversal of this magnitude did not occur until after the business peaks of January 1980 and July 1991.
From January to April of 1994 the monthly values for the S&P composite stock price index declined 5.446 percent and came very close to adding another contraction of 5.5 percent to the 24 downside reversals in Table 14.4. The disqualification of this contraction enabled the S&P index to exceed the old record established in January 1966 of 43 months of sustained rise in stock prices (without a dip of 5.5 percent or more) when the monthly values for the S&P index recovered to a new historic high of 481.92 in February 1995.
Suppose one was able to identify business peaks at the end of the month in which they occurred. Would it then be too early to purchase common stock? The data which are presented in Table 14.5 show that the S&P index has then proceeded to lose from 3.8 to 35.1 percent of its value on a daily closing value basis before bottoming out during the nine recessions since 1947. Since the mild recession of 1960-61 the S&P index has consistently lost more of its value after the business peaks than before the business peaks.
While much attention has been paid to the unreliability of stock prices as a predictor of economic recessions it is less well appreciated that this market is a very superior predictor of economic recoveries. If we ignore the money supply, M-2 expressed in constant dollars, which is so trendy as to have not exhibited a distinct trough during some recessions, stock prices can claim the distinction of being the only component of the index of leading economic indicators that has consistently led economic recoveries in the post World War II period.
The unadjusted recovery lead times for the revised index of eleven leading economic indicators published in the October 1994 issue of the Survey of Current Business range from a low of one month for the 1973-75 and 1969-70 recessions to a high of 10 months for the 1981-82 recessions. The unadjusted lead times for stock prices, on the other hand, fall in the comparatively narrow range of from three to eight months. The lead times have been longer for those recessions associated with major wars. See Table 14.6.
While the good year approach to economic forecasting is not as well developed as the "poor growth year" approach, which was emphasized in The Practical Forecasters' Almanac and illustrated in Table 11.3, it can play a complementary role in helping one to ignore false indications of a possible recession.
In Table 21.5 interest rates and some inflation indicators are used to identify years which have (so far) never been followed in at least the post 1948 period by a year containing a recessionary peak in economic activity.
Since the beginning of World War II there has never been a peak in economic activity after any year when the Fed allowed the yield on new issues of 3-month Treasury bills in Table 17.4 to decline by six percent or more.
During the same period there has never been a business peak until at least 23 months after any year with an annual decline in both the December- December inflation rate for the CPI and the last offering yield in December for new issues of 91 day Treasury bills. If this relationship continues to hold there won't be another recessionary peak in business activity until at least November of 1997.
Fed watchers should also keep a close eye on the prime rate charged by banks. If there is no offsetting increase during the remainder of 1996 the first quarter decline in this indicator would suggest that there won't be another recession until at least 1998.
Since 1923 the US economy has never experienced a recessionary peak in business activity after any year when the growth rate for M2 or Friedman and Schwartz's M4 accelerated by more than one percentage point on a December-December basis. To the extent that individuals are saving for a near term purpose, such as the purchase of a new automobile, this sort of saving behavior should help to bolster the economy in the following year.
M2, which includes general purpose money market mutual fund balances and small denomination time deposits as well as all of the components of M1, has been a part of the Conference Board's composite index of leading economic indicators since 1979. From 1988-94 the M2 growth rate sagged from 5.5 to only .9 percent as many investors shifted more of their savings to stock and bond funds. During 1995, however, the M2 growth rate recovered to 4.5 percent.
Another statistical indicator with an impressive ability to distinguish between peak and non peak years for the US economy is the fuel segment of the crude materials for further processing component of the producer price index. Year-year declines in this index have (so far) always been followed by a non peak year. See those "x" mark the spot years in Table 21.5 that are captioned with the symbol "FP".
Rapid increases in stock prices have also helped to prolong a business expansion. Since the great bull market of the 1920s there has never been a peak in business activity after any year when the financial return for the S&P index in Table 19.1 was over 34 percent. Nor has there been a peak year after any year when the financial return for this index increased by 16 percentage points or more. There are many economists who believe that consumption is a function not only of current income but may also depend on wealth which was acquired in previous years. (See Essay 1.)
In Table 14.5 it was shown that since the mild recession of 1960 the stock market has consistently lost more of its value after a recessionary peak in business activity than before the peak. Let us hope that these indications of a good year for the US economy are correct and that the slow down in the economic growth rate that many economists are expecting for the second half of 1996 won't be the prelude to a near term recession.
Green, George and Barry Beckman (1993). "Business Cycle Indicators: Upcoming Revision of the Composite Indexes," Survey of Current Business, October, 44-51.
Lahiri, Kajal and Geoffrey Moore (1991). Leading Economic Indicators: New Approaches and Forecasting Records(New York: Cambridge University Press).
Moore, Geoffrey (1961). Business Cycle Indicators(Princeton: Princeton University Press), Vol. 1, Table 3.2, p. 56.
Renshaw, Edward (1992). The Practical Forecasters' Almanac(Burr Ridge, Illinois: Irwin Professional Publishing).
Zarnowitz, Victor and Charlotte Boschan (1975). "Cyclical Indicators: An Evaluation and New Leading Indexes," Business Conditions Digest, May, v-xiv.
Dates of Index of Leading Indicators
---------------------------------- ------------------------- Adjusted
Index Index Down NBER Index Index NBER Index Lead Time
Peak 1.0% Peak Trough Peak Peak Trough in Months
(1) (2) (3) (4) (5) (6) (7) (8)n
May 59 Oct. 59 Apr. 60 Mar. 60 75.1 73.8 73.6 - 6
Mar. 66 June 66 ----- Feb. 67 85.1 ---- 83.2 ---
Apr. 69 July 69 Dec. 69 Apr. 70 87.4 85.7 83.9 - 5
Feb. 73 June 73 Nov. 73 Feb. 75 92.4 90.6 83.9 - 5
Oct. 78 Dec. 78 Jan. 80 Apr. 80 93.0 90.1 87.0 -13
Apr. 81 June 81 July 81 Mar. 82 90.9 89.8 88.7 - 1
June 88 Mar. 89 July 90 Jan. 91 100.8 99.6 97.8 -16
Dec. 94 May 95 ----- May 95 101.6 ---- 100.4 ---
(8)n. Months from the NBER business peak in Column (3) to the month when the revised index of leading economic indicators was down one percent in Column (2).
Source of basic data: http://www.stat-usa.gov/BEN/ebb2/bea/bci-01a.prn The BCI datafiles are provided to the STAT-USA EBB by the Conference Board as a public service.
Thousands of Housing
Peak Months for Units
----------------- -------------------- Lead Time in Months
Housing Economic Permit Economic Percent -------------------
Permits Activity Peak Peak Decline Unadjusted Adjusted
(1) (2) (3) (4) (5)n
Oct. 47 Nov. 48 1476 1076 27.1 -13 - 1
July 50 July 53 2282 1248 45.3 -36 -28
Feb. 55 Aug. 57 1946 1187 39.0 -30 -11
Nov. 58 Apr. 60 1729 1232 28.7 -17 - 5
Feb. 69 Dec. 69 1689 1306 22.7 -10 0
Dec. 72 Nov. 73 2688 1557 42.1 -11 0
Jun. 78 Jan. 80 2065 1333 35.4 -19 - 1
Sep. 80 July 81 1545 974 37.0 -10 - 1
Feb. 84 July 90 2043 1108 45.8 -77 - 3
Dec. 93 1502p
(5)n. Months from the next economic peak identified by the NBER after the building permit index has been down 20 percent for two months in a row if the permit peak was less than 1800 units and down 35 percent for two months in a row if the permit peak was over 1800 units.
p equals a preliminary estimate of the peak which may turnout to be too low.
Source of basic data: An index (1987 = 100) in the October 1995 issue of the Survey of Current Business is multiplied by 15.348 to produce unit values for the 1947-53 period. The seasonally adjusted estimates for more recent months is based on BCI datafiles provided to the STAT-USA EBB by the Conference Board as a public service. This data can be accessed at: http://www.stat-usa.gov/BEN/ebb2/bea/bci-05.prn
Prime Annualized Lead Time ---Three Month % Change--
Quarter Rate Up Percentage Changes in Months Residential Contracts,
Ratio -------------------- from NBER Building Orders Plant
CD PDE RC Peak Permits and Equip.
(1)n (2) (3) (4) (5) (6) (7)
1948-3R 1.33 --- -8.7 -12.3 - 2 -17.8 -17.8
1952-3R 1.50 -24.1 -47.3 --- -10 13.8 28.4
1953-2R 1.62 - 6.4 - 4.4 --- - 1 - 7.3 -20.0
1957-2R 1.33 - 8.0 - 4.4 - 9.3 - 2 .8 -11.9
1959-4R 1.43 -19.1 --- -11.3 - 4 - 1.2 - 4.3
1969-4R 1.89 --- - 3.1 -26.3 0 - 8.5 -13.7
1973-2R 1.58 - 6.4 --- -17.8 - 5 - .6 3.4
1979-2R 1.86 - 8.7 - 4.9 --- - 7 - 4.6 -11.7
-4R 2.45 -10.6 - 7.0 -14.9 - 1 -23.4 .0
1981-2R 1.80 -15.3 --- -14.5 - 1 -17.5 - 4.3
1989-4R 1.40 -11.4 -10.0 --- - 7 8.2 21.8
1990-2R 1.33 -11.8 -10.8 -18.0 - 1 -11.0 -10.9
(1)n. The monthly average prime rate at the end of the quarter expressed as a ratio to the monthly cyclical low value for the prime rate in the preceding business expansion.
The annualized growth rates in columns (1), (2) and (3) are based on GDP components expressed in 1987 dollars from 1947-58 and on percentage changes in chained values linked to 1992 dollars since 1958.
R identifies quarters ending after there has been a sustained increase in the average prime rate charged by banks amounting to a third or more from its cyclical low. All of these cases have (so far) been followed by a recessionary peak in business activity within zero to ten months.
Source of basic data: The annualized growth rates for the real GDP components are from the National Income and Product Account Table 8.1 obtained from National Income and Product Accounts of the United States, Volume 1, 1929-58 and the Survey of Current Business, January/February, 1996. The monthly data for the prime rate, residential building permits and contracts and orders for plant and equipment are from the cyclical indicator section of the Survey of Current Business, October 1995.
Date of Cycle S&P Stock Index Months Duration % Change in Index
-------------------- --------------- --------------- -----------------
Peak Trough Peak Trough Rise Decline Rise Decline
(1) (2) (3) (4) (5) (6)
May 1946 Nov. 1946 18.70 14.69 30 6 65.0 21.4
Feb. 1947 May 1947 15.80 14.34 3 3 7.6 9.2
July 1947 Feb. 1948 15.77 14.10 2 7 10.0 10.6
June 1948 June 1949 16.82 13.97 4 12 19.3 16.9*
June 1950 July 1950 18.74 17.38 12 1 34.1** 7.3
Jan. 1953 Sep. 1953 26.18 23.27 30 8 50.6 11.1*
July 1956 Feb. 1957 48.78 43.47 34 7 109.6** 10.9
July 1957 Dec. 1957 48.51 40.33 5 5 11.6 16.9*
July 1959 Oct. 1960 59.74 53.73 19 15 48.1** 10.1*
Dec. 1961 June 1962 71.74 55.63 14 6 33.5** 22.5
Jan. 1966 Oct. 1966 93.32 77.13 43 9 67.8 17.3
Sep. 1967 Mar. 1968 95.81 89.09 11 6 24.2 7.0
Dec. 1968 June 1970 106.48 75.59 9 18 19.5 29.0*
Apr. 1971 Nov. 1971 103.04 92.78 10 7 36.3** 10.0
Jan. 1973 Aug. 1973 118.42 103.80 14 7 27.6 12.3
Oct. 1973 Dec. 1974 109.84 67.07 2 14 5.8 38.9*
July 1975 Sep. 1975 92.49 84.67 7 2 37.9** 8.5
Sep. 1976 Mar. 1978 105.45 88.82 12 18 24.5 15.8
Aug. 1978 Nov. 1978 103.92 94.71 5 3 17.0 8.9
Feb. 1980 Apr. 1980 115.34 102.97 15 2 21.8 10.7*
Nov. 1980 July 1982 135.65 109.38 7 20 31.7** 19.4*
Oct. 1983 July 1984 167.65 151.08 15 9 53.5** 9.9
Aug. 1987 Dec. 1987 329.36 240.96 37 4 118.0 26.8
Jun. 1990 Oct. 1990 360.39 307.12 30 4 49.6 14.8*
*Stock market declines that were associated with economic recessions.
**Rises following economic recessions.
Source of Basic Data: Economic Report of the President.
Date of Closing Values S&P Index % Decline
---------------------------- ------------------------ -----------------
Market Business Market Market Business Market M. Peak B. Peak
Peak Peak Low Peak Peak Low B. Peak M. Low
6/15/48 11/30/48 6/13/49 17.06 14.75 13.55 13.5 8.1
1/05/53 7/31/53 9/14/53 26.66 24.75 22.71 7.2 8.2
8/02/56 8/30/57 10/22/57 49.74 45.22 38.98 9.1 13.8
8/03/59 4/29/60 10/25/60 60.71 54.37 52.30 10.4 3.8
11/29/68 12/31/69 5/26/70 108.37 92.06 69.29 15.1 24.7
1/11/73 11/30/73 10/03/74 120.24 95.96 62.28 20.2 35.1
2/13/80* 1/31/80 3/27/80 118.44* 114.16 98.22 3.6 14.0
11/28/80 7/31/81 8/12/82 140.52 130.92 102.42 6.8 21.8
07/16/90 7/31/90 10/11/90 368.95 356.15 295.46 3.5 17.0
*Only case where the stock market peak occurred after the business peak.
Source of Basic Data: Standard and Poor's Security Price Index Record.
Date of NBER Business Unadjusted Lead Times Adjusted Lead Times at
--------------------- --------------------- ----------------------
Peak Trough Months to Months to Business Business
Peak Trough Peak Trough
(1) (2) (3)n (4)n
Nov. 48 Oct. 49 30 4 27 3
July 53 May 54 6 8* 3 5
Aug. 57 Apr. 58 13 4 0 -1
Apr. 60 Feb. 61 9 4 2 2
Dec. 69 Nov. 70 12 5* 6 2
Nov. 73 Mar. 75 10 3 8 2
Jan. 80 July 80 - 1 3 -2 1
July 81 Nov. 82 8 4 -2 2
July 90 Mar. 91 1 5* -1 3
(3)n. Months to the business peak after the monthly average value for the S&P composite stock price index has declined five percent or more from its preceding cyclical high and maintained that degree of downness until after the next cyclical trough in the S&P index.
(4)n. Months to business trough after the monthly average value for the S&P index has increased the 5.5 percent needed to offset abortive rallies in the midst of an economic recession.
*Lead times that may have been influenced by major military operations.
Source of Basic Data: BCI series 19, the Survey of Current Business, October 1995.
Go on to Essay 15:
Return to the Introduction