Edward Renshaw
Professor of Economics
State University of New York at Albany
One of the most popular extensions of Keynesian economics is the celebrated IS-LM framework, which was invented by Sir John Hicks in (1937). The popularity of this model has decreased since the late 1970s, however, since it doesn't do a very good job of coming to grips with the problem of inflation. Some economists are inclined to either ignore the IS-LM framework altogether, or relegate it to the status of an unnecessary side- show that belongs in an appendix. It can be a useful starting point for discussing the topic of monetary targeting, however, starting with Poole's (1970) popularization of the IS-LM model to compare targeting of the money supply with the alternative of targeting the interest rate.
I find the IS-LM framework to also be of some value in discussing the problem of parameter "identification" in a world where the Fed has a propensity to "lean against the wind" (see Essay 5) and in providing students with a better appreciation of why short term interest rates fluctuate in a procyclical manner. It can also be used as a starting point for deriving an interest rate reaction function rather than an aggregate demand curve.
In a book documenting his own model of the US economy, which endeavors to do a better job of separating the behavior of households and business enterprises, Ray Fair (1984, p. 93) describes the standard IS-LM model as consisting of the following ten equations: "(1) a consumption function in income and assets (the level of assets is exogenous), (2) an investment function in the rate of interest and income, (3) an income identity, where income is consumption plus investment plus government spending, (4) a real money demand function in the rate of interest and income, (5) a money supply function in the rate of interest (or the money supply taken to be exogenous), (6) an equilibrium condition equating money supply to money demand, (7) a production function in labor and the capital stock (the capital stock is exogenous), (8) a demand for labor equation equating the marginal product of labor to the real wage rate, (9) a labor supply function in either the money wage (the "Keynesian" version) or the real wage (the "classical" version), (10) an equilibrium condition equating the supply of labor to the demand for labor. These ten equations determine the following ten unknowns: consumption, investment, income, demand for money, supply of money, demand for labor, supply of labor, the price level, the wage rate, and the interest rate."
The celebrated downward sloping IS curve depicting equilibrium states where investment is always equal to saving (or injections equal leakages) can be obtained by first substituting the consumption function (1) into the income identity (3) to obtain a Keynesian type of multiplier equation similar to those discussed in Essay 1. The next step is to substitute an investment function (3) into this multiplier equation.
One of the problems that is encountered at this point is that economists haven't had a great deal of luck in demonstrating that gross private domestic investment and some of its more important components, such as changes in business equipment and inventories, are negatively related to the interest rate (Essay 5).
An upward sloping LM curve depicting equilibrium in the money market can be obtained by substituting an expression for the supply of money into the Baumol or Latane demand for money equations discussed in Essay 6. The problem that one encounters at this juncture is that the more publicized demand for money equations are not very stable. Large shifts in the IS curve, however, do make it easier to identify an upward sloping LM curve than a downward sloping IS curve. Accelerator type shifts in the IS curve are most pronounced in the vicinity of economic recessions.
In December 1992 NBER's dating committee finally decided that the most recent recession ended in March 1991 and had a duration of only eight months. If you believe this assessment there have only been two shorter recessions in the history of business cycle analysis, the seven month recession following World War I and the six month recession from January to July of 1980.
One of the problems with NBER's dating system is that it doesn't provide any insight as to the severity of the recessions or the time that was required for the economy to recover to a new historic high. This problem can be ameliorated, to some extent, by focusing one's attention on the behavior of gross domestic product (GDP) expressed in constant dollars and other statistical series that have acquired the reputation of being "coincident" indicators of economic activity. See Table 7.1.
Strong demand for US exports caused industrial production to peak out in September 1990, two months after the official NBER peak in business activity, and to only decline by a modest 4.2 percent at the NBER Trough in March 1991. The 25 month depression in industrial production from September 1990 to October 1992, however, was of median duration for the post 1948 period.
The 34 month depression for payroll employment, on the other hand, established a new record for depressed activity that was six months longer than the 28 month depression associated with the 1981-82 recession. The prolonged period of no growth in payroll employment probably cost President Bush his job and gave rise to the immortal political slogan, "it's the economy, stupid."
The 36 month depression in real personal income less transfer payments was exceeded only by the 37 month depression from November 1973 to December 1976. The seven quarter depression for real GDP from 1990-92 was only one quarter less than for the 1973-75 recession, when the US unemployment rate soared upward by more than four percentage points.
The more important point to note in connection with Table 7.1 is that post World War II recessions in economic activity are not getting shorter, if we view them from the perspective of how long it takes a coincident indicator to recover to a new historic high. The average duration of the last four recessions is higher for each of our indicators than the average duration for the first five recessions since 1947. The most impressive difference in average duration is for real personal income less transfer payments. This indicator was depressed on the average for less than 13 months during the five economic recessions from October 1948 through January 1971. During the last four recessions real personal income less transfer payments remained below its peak value for 27 months, on the average.
Since the Fed stopped pegging interest rates in 1951 it has always let the discount rate on new issues of 91 day Treasury bills decline by at least 20 percent during employment recessions (Table 7.2) and has almost always allowed the yield to decline during any year when the unemployment rate increased by more than .1 percentage points. The only exception is 1980 when a brief recession failed to reduce the CPI inflation rate below the double digit level.
The Fed's policy with regard to the real, inflation-adjusted return on Treasury bills, however, has varied considerably during the employment recessions which have occurred in the post World War II period. The average real rate of return during a recession has varied by more than nine percentage points and ranged from a negative return of 4.94 percent for the 1974-75 recession to a positive return of 4.21 percent for the 1981-82 recession. See the last column of Table 7.2.
The interesting question is whether we can learn anything of importance from this type of experimentation. In Table 7.3 employment recessions are rank ordered in terms of the average real rate of return on T-bills. The data are not inconsistent with the hypothesis that low rates of return on liquid assets may help to shorten employment recessions by encouraging people to "beat inflation" by investing in new houses, cars and other goods and services with enduring value.
Employment recessions have also been shorter the greater the lead time for the peaking out of industrial production relative to employment and the steeper the decline in payroll employment during any one of the first three months of the recession (Renshaw 1992, Tables 1.24 and 1.25). Since much of the reduction in real GDP during an economic recession can be explained by a reduction in inventory investment, it is not unreasonable to suppose that the sooner the peak in industrial production and the steeper the decline in factory employment, other things equal, the sooner an inventory recession will be over. When the inflation adjusted return on T-bills is included in a multiple regression with these two variables to help explain the duration of employment recessions, however, its coefficient turns out to be statistically insignificant.
In the post 1947 period the duration of employment recessions has been about equal to 15 months minus the monthly lead time for industrial production at its peak relative to the peak in payroll employment. See Table 7.4. The implication would seem to be that a year or more may be required to turn an industrial recession around and that lower interest rates may not be very effective at halting a cyclical decline in payroll employment that begins before or at about the same time as the decline in industrial production.
There is also not much evidence to support the conclusion that a big decline in short term interest rates, or a low average real rate of return on T-bills during a recession, are very effective at stimulating a rapid first year recovery from an employment recession. See the first and last columns of Table 7.3.
There may even be a silver lining to grudging reductions in short term interest rates in response to a recession since the following expansions in payroll employment have tended to be longer when the inflation adjusted return on T-bills during the recessions has been positive instead of negative. The only exception is the 58 month expansion in payroll employment from April 1975 to March 1980 when the 12 month CPI inflation rate was allowed to soar upward from about six percent during the first year of the expansion to almost 15 percent one year before the employment peak. See Tables 7.2 and 7.3.
Economic recessions are the only sure cure for inflation. In the post 1947 period employment recessions have reduced the CPI inflation rate from a little less than one percentage point, for the mild recession of 1960-61, to almost seven percentage points for the prolonged recession of 1981-82. The reduction in the inflation rate one year before the employment peak to one year after the employment trough has been about equal, on the average, to the percentage point increase in the civilian unemployment rate from its cyclical low to its cyclical high. The higher the 12 month inflation rate at the employment peak, the greater the drop in the inflation rate, if the increase in the unemployment rate was about the same. See Table 7.5. Including the duration of the payroll recessions as an independent variable in the regression used to identify the parameters in this table doesn't improve the adjusted R square. This would suggest that prolonged employment recessions are not very helpful at reducing the inflation rate.
In a (1986) paper, Keim and Stambaugh found that the spread between the yields on low grade corporate bonds and one-month Treasury bills was of some value in predicting financial returns in both the stock and bond markets. Since 1941 the following year financial returns for the S&P index have usually been positive after a spread of 2.85 percentage points or more for the December yield on Moody's Baa Corporate bond index minus the last offering yield for 91 day Treasury bills. See Table 7.6.
While yield to maturity curves have done a fairly good job of identifying some of the stock market crashes that have been associated with economic recessions, they aren't very reliable in warning investors about the possibility of a stock market correction in the midst of prosperity. A person who purchased a portfolio similar to the S&P index at the end of 1976, when the yield spread in Table 7.6 was equal to 4.82 percentage points, would have suffered a financial loss of 7.6 percent during the "correction" of 1977.
The slope of the yield to maturity curve also has the distinction of not providing investors with any warning about the stock market corrections of 1984, 1987 and 1994, when the Fed engineered other increases in short term interest rates in the current or preceding year to prevent inflation from getting out of hand.
Fair, Ray (1984). Specification, Estimation, and Analysis of Macroeconometric Models(Cambridge, MA: Harvard University Press).
Hicks, John R (1937). "Mr Keynes and the Classics: A Suggested Interpretation," Econometrica, vol. 5, no. 2, 147-159.
Keim, Donald and Robert Stambaugh (1986). "Predicting Returns in the Stock and Bond Markets," Journal of Financial Economics, (17), 357- 90.
Hotelling, Harold (1925). "A General Mathematical Theory of Depreciation," Journal of the American Statistical Association, September.
Poole, William (1970). "Optimal Choice of Monetary Policy Instruments in a Simple Stochastic Macro Model," Quarterly Journal of Economics, 84(May), 197-216.
Renshaw, Edward (1992). The Practical Forecasters' Almanac(Burr Ridge, Illinois: Richard Irwin).
Dates of Recessionary Peaks in Duration to a New High Recovery ---------------------------------- ------------------------------- IP EMP PI-TR RGDP IP EMP PI-TR RGDP July 48 Sep. 48 Oct. 48 48-4 21 22 18 5 July 53 June 53 June 53 53-2 20 23 16 6 Mar. 57 Mar. 57 Aug. 57 57-3 23 25 11 5 Jan. 60 Apr. 60 May 60 60-1 21 20 9 5 Oct. 69 Mar. 70 Nov. 69 69-3 26 18 14 6 Nov. 73 Oct. 74 Nov. 73 73-4 37 16 37 8 May 79 Mar. 80 Jan. 80 80-1 26 10 10 4 July 81 July 81 Aug. 81 81-3 26 28 25 7 Sep. 90 June 90 Apr. 90 90-2 25 34 36 7 Average Duration to a New High Recovery First Five Recessions 22.2 21.6 13.6 5.4 Last Four Recessions 28.5 22.0 27.0 6.5
Source of basic data: The Survey of Current Business, October 1994 and January/February 1996.
Duration Unempl. Ave. Real
Date of Employ. in Rate T-Bill Rate Inflation Rate T-Bill Rate
Peak Trough Months Peak Peak Trough Peak Trough Peak-Trough
(1) (2) (3) (4) (5)n (6)n (7)n
Sep. 48 Oct. 49 13 3.8 1.09 1.04 6.5 3.5 .68
Jun. 53 Aug. 54 14 2.5 2.23 .89 1.0 -.5 .51
Mar. 57 May 58 14 3.7 3.14 1.05 3.6 .3 -.76
Apr. 60 Feb. 61 10 5.2 3.24 2.41 2.0 .9 1.01
Mar. 70 Nov. 70 8 4.4 6.71 5.29 6.0 3.5 .48
Oct. 74 Apr. 75 6 6.0 7.24 5.69 12.0 6.1 -4.94
Mar. 80 July 80 4 6.3 15.53 8.13 14.7 10.7 -4.57
July 81 Nov. 82 16 7.2 14.70 8.01 10.7 3.2 4.21
June 90 Feb. 92 20 5.3 7.74 3.84 4.7 3.2 1.22
(5)n. The twelve month inflation rate for the all item consumer price index at the employment peak.
(6)n. The twelve month inflation rate for the all item consumer price index one year after the employment trough.
(7)n. Average value for the decline months obtained by subtracting the 12 month CPI inflation rate from the T-bill rate each month.
Source of basic data: The old cyclical indicator section of the Survey of Current Business. The CPI inflation rates for the preceding 12 months are calculated from an index with a base of 1967 = 100 for the 1946- 87 period and a base of 1982-84 = 100 for the years since 1987.
Date of Average Real --Duration in Months-- % Change Employment
Employment T-Bill Rate Employment Following 12 Months After the
Peak Peak-Trough Contraction Expansion Employment Trough
(1)n (2) (3) (4)
July 81 4.21 16 91 3.5
June 90 1.22 20 ? 1.4
Apr. 60 1.01 10 109 3.0
Sep. 48 .68 13 44 8.7
June 53 .51 14 31 4.6
Mar. 70 .48 8 47 2.1
Mar. 57 - .76 14 23 5.2
Mar. 80 - 4.57 4 12 1.9
Oct. 74 - 4.94 6 58 3.7
(1)n. Average value for the decline months obtained by subtracting the 12 month CPI inflation rate from the T-bill rate each month.
Source of basic data: Table 7.2.
Date of Peak for
---------------------- Monthly Lead Time Duration of Recession
Industrial Payroll Industrial Actual Predicted
Production Employment Production -----in months-----
(1) (2) (3)n
Nov. 73 Oct. 74 11 6 4
May 79 Mar. 80 10 4 5
Oct. 69 Mar. 70 5 8 10
Jan. 60 Apr. 60 3 10 12
July 48 Sep. 48 2 13 13
Mar. 57 Mar. 57 0 14 15
July 81 July 81 0 16 15
July 53 Jun. 53 - 1 14 16
Sep. 90 Jun. 90 - 3 20 18
(3)n. The predicted duration is equal to 15 months minus the monthly lead time for the industrial production peak in column (1).
Source of data: The old cyclical indicator section of the Survey of Current Business.
Decline in Inflation Rate
Economic % Point CPI -------------------------
Recession Increase Inflation Actual Predicted
Civilian Rate at
Unemployment Recessionary
Rate Peak
(1) (2) (3)n (4)n
1948-49 4.5 6.5 3.0 4.4
1953-54 3.6 1.0 1.5 1.6
1957-58 3.8 3.6 3.3 2.7
1960-61 2.3 2.0 1.1 .7
1969-70 2.7 6.0 2.5 2.4
1973-75 4.4 12.0 5.9 6.1
1980 2.2 14.7 4.0 4.8
1981-82 3.6 10.7 7.5 4.9
1990-91 2.7 4.7 1.5 2.0
Average Values 3.3 6.8 3.4 3.4
(3)n. The actual 12 month CPI inflation rate at the employment peak in column (2) minus the 12 month CPI inflation rate one year after the employment trough in Table 7.2.
(4)n. The predicted decline in the CPI inflation rate is equal to the increase in the civilian unemployment rate from its cyclical low to its recessionary peak in column (1) plus one-third of the peak CPI inflation rate in column (2) minus 2.3 percentage points.
Year Last Offering December Yield Yield Spread Following Year
Yield for 91 Moody's Baa Col.(2) Minus Financial Return
Day Treasury Bond Index Column (1) S&P Index
Bills
(1) (2) (3)n (4)
1941 .31 4.38 4.07 19.4
1942 .36 4.28 3.92 25.7
1943 .37 3.82 3.45 19.3
1944 .37 3.49 3.12 35.7
1960 2.12 5.10 2.98 26.6
1970 4.80 9.12 4.32 14.1
1971 3.73 8.38 4.65 18.7
1974 7.11 10.63 3.52 36.9
1975 5.21 10.56 5.35 23.6
1976 4.30 9.12 4.82 - 7.6*
1977 6.14 8.99 2.85 6.4
1981 11.69 16.55 4.86 20.4
1982 7.98 14.14 6.16 22.3
1983 8.94 13.75 4.81 6.0*
1984 7.84 13.40 5.56 31.1
1985 7.04 11.58 4.54 18.5
1986 5.68 9.97 4.29 5.7*
1987 5.73 11.29 5.56 16.3
1990 6.78 10.43 3.65 30.0
1991 3.91 9.26 5.35 7.4
1992 3.22 8.81 5.59 9.9
1993 3.06 7.69 4.63 1.3*
1994 5.56 9.10 3.54 ?
(3)n. Cases where the yield spread is equal to 2.85 percentage points or more.
*Poor financial returns which were associated with rising interest rates and stock market corrections in the midst of a business expansion.
Source of basic data: Standard and Poor's Securities Price Index Record, Moody's Municipal & Government Manual, and the Economic Report of the President.
Go on to Essay 8:
Return to the Introduction