Edward Renshaw
Professor of Economics
State University of New York at Albany
Since the publication of Stocks as Long-Term Investments, by Edgar L. Smith in (1925), every reputable investigator has been forced to conclude that the financial returns associated with representative stock price indexes have outperformed bonds over long periods of time. See Jeremy Siegel's (1994) book for a more recent review of the numbers and some of the literature supporting this conclusion.
While most of the theories that have been developed by professors of economics and finance to explain stock prices aren't very useful at discriminating between good and bad years to own equities, they may still be of some value in helping one to detect changes in the standards that have been employed by investors over time in valuing the earnings and dividends associated with representative stock market averages.
Most of the more advanced textbooks on corporate finance contain at least a few paragraphs devoted to the interpretation of an actuarial formula that has been of some value in projecting the financial returns for the S&P index from the end of one year containing a peak in business activity to the next.
If the dividend price ratio remains the same and if there is no change in the dividend growth rate, the actual financial return (dividends plus price appreciation expressed as a percent of price at the beginning of the period) will be equal to the current dividend yield plus the dividend growth rate. Gordon (1962) was among the first to promote this relationship.
In Table 5.1S we use this type of formula to project the financial returns for the S&P composite stock price index between years of peak prosperity. The actual returns in the following period, which are shown in column (6) assume that all dividends are reinvested at the end of each year and that there are no transaction costs.
It will be noted that the actual return of - 3.0 percent per year from the end of 1929 to the end of 1937 was quite a bit less than the projected average return of 16.5 percent at the end of 1929. From 1937 through the 1960s, when the stock market was still recovering from the trauma of the great depression of the 1930s, the actual returns associated with the S&P index were often closely related but consistently greater than the projected returns. Since 1969, however, the actual returns through the following year containing another business peak have sometimes been less than the projected returns.
At the end of 1990 this projection scheme was predicting a compound average return of 10.6 percent for the current business expansion which turned out to not be very different from the actual return of 11.5 percent for the four year period from 1991-94.
It should be noted, however, that the dividend yield in column (3) of Table 5.1S has been trending downward since 1937. When that happens the financial returns will be temporarily enhanced if the dividend growth rate remains the same. In the long run, however, a lower dividend yield can be expected to reduce the financial returns associated with common stock.
One of the more interesting points to note in connection with the dividend yields in column (3) to Table 5.1S is that there has never been a yield under 3.4 percent at the end of any year containing a peak in business activity. There have been several occasions when investors experimented with dividend yields of less than three percent in the midst of recessions and business expansions but they have (so far) all been wiped out by bear markets either before or shortly after the business expansion was over. The unanswered question is whether this will continue to be the case in the future?
The US tax code makes it advantageous for some corporations to use part of their excess earnings to purchase other companies and buy back their own shares rather than increase their dividends. Stock buy-backs, however, help to support the notion that there is a law of diminishing profitability associated with internal investments since it wouldn't be rational for a firm to repurchase its own stock, under most circumstances, if it could reinvest its retained earnings at a higher expected return internally.
While the drift toward lower dividends may be rational, it is not necessarily a good omen for the stock market in either the short or the long run. History suggests, in any event, that investors should be cautious whenever the monthly dividend yield for the S&P composite stock price index has either established a new record low or tied the preceding low. (See those dividend lows identified with an asterisk in Table 5.2S.)
In September 1929, before the memorable crash which occurred in October of that year, the dividend yield for the S&P index fell to 2.92 percent and the index then proceeded to loose more than 85 percent of its value.
The S&P stock price index more than doubled in value between June 1, 1932 and July 1933, pushing the dividend yield back down from an historic high of 10.30 percent to only 2.95 percent. This short lived, but rather spectacular bull market was followed by a less severe correction of not quite 25 percent.
During the bull market of 1961, when investors became fascinated with growth stocks that don't pay much dividends, the monthly average yield on the S&P index fell to a new low of 2.84 percent and the stock price index then proceeded to loose more than 25 percent of its value in the early stages of what was to turnout to be the longest expansion of economic activity in the history of business cycle analysis.
In 1971 President Nixon did what he had promised never to do and imposed a system of wage and price controls on a peace time economy. Corporations, for the most part, cooperated by investing increased earnings in new plant and equipment rather then using them to pay more dividends. By January 1973 the dividend yield on the S&P index had established a new record low of only 2.69 percent. This record was followed by the most severe stock market correction in the post World War II period.
On October 19, 1987, after leveraged buy-outs and speculative enthusiasm had pushed stock prices up enough in August to tie the previous dividend low of 2.69 percent, the S&P index experienced a record loss of more than 20 percent in a single trading day.
Since the economic recession of 1990-91 business enterprises have been using most of the recovery in corporate earnings to purchase computers and other types of equipment which might reasonably be expected to improve labor productivity and make US enterprises more competitive in the world market place. The expectation that dividend increases would follow and the lowest short term interest rates on savings accounts since the early 1960s encouraged many investors to continue a process of shifting more of their financial assets from maturing CDs to common stock without much regard for what happened in 1973 and 1987 when the yield on the S&P dipped to only 2.69 percent.
Low dividend yields will often persist for a prolonged period of time when the US economy is recovering from an economic recession and inflation is not regarded as being a serious threat. When low yields are accompanied by a policy of tight money and rapidly rising interest rates, however, one has all the ingredients that are necessary for a stock market correction.
The stock market correction of 1994 turned out to be rather mild and never did push the monthly yield associated with the S&P index above three percent. The unanswered question is whether investors will be content with a yield of less than three percent when it becomes clear that the US economy has slipped into another recession. Since the mild recession of 1960-61 the S&P index has declined more after the recessionary peaks in business activity designated by the National Bureau of Economic Research than before the peaks.
The standards that have been used by the financial community to value the earnings and dividends associated with representative stock market indexes have been subject to a great deal of variation over time.
From 1936-57 the December yield on Moody's high grade corporate bond index was consistently less than the December yield for the S&P composite stock price index. Since 1957, however, the December yields on high grade bonds have been greater than the yields for the S&P index.
From 1961 to 1981 one could have identified all of the loss years for the S&P composite stock price index with a simple risk premium defined as the earnings-price yield on the S&P(measured by the average earnings over the last three years divided by the closing price in the last year) minus the December yield on Moody's high grade corporate bond index with only a few false signals (Renshaw 1992, Table 3.24).
During this period the financial returns associated with the S&P index were quite a bit lower, on the average, in the years following a risk premium under .35 percentage points than the financial returns that could have been obtained from long term bonds.
In 1979, however, Nobel Laureate Franco Modigliani and Richard Cohn criticized financial analysts for using a nominal rate of interest to capitalize corporate earnings and suggested that investors had been systematically undervaluing the stock market by 50 percent. "In the presence of inflation," they argued, "one properly compares the cash return on stocks, not with the nominal return on bonds, but with the real return on bonds." The gradual acceptance of this new standard for valuing corporate earnings helped to set the stage for the bull market of the 1980s, which has now gone down in history as the second-best decade to have owned stock since the S&P index was extended back to the 1870s.
A more modern year-end risk premium is presented in Table 5.3S for the S&P index which assumes that its associated earnings should be discounted by a real interest rate. We start by calculating an earnings-price ratio for the index based on the highest earnings for the current or any previous year expressed as a percent of the closing value for the index. A real risk premium is then obtained by subtracting the December yield for Moody's Aaa corporate bond index from the high earnings-price ratio and adding to this residual the December-December inflation rate for the all item consumer price index.
Basing the risk premium on the highest observed earnings for any calendar year probably over states the "true" risk premium for some years when earnings are depressed but enables one to largely avoid negative risk premiums for the S&P index since it was first computed on a daily basis in 1928. Negative risk premiums are not very plausible if investors believe that stocks are more risky than bonds in the short run.
At the end of 1986, however, we do observe a risk premium of -.49 percentage points for the highest earnings price ratio approach. The negative premium should probably be attributed to crashing oil prices which lowered the inflation rate for the all item CPI to only 1.13 percent at a time when the inflation rate for the all item CPI without energy was still increasing at a 3.8 percent rate. The negative risk premium for the more publicized CPI, however, might have provided some investors with a timely warning with regard to the "bursting stock market bubble" of 1987.
From 1929-82 there were only three cases of a real year-end risk premium less than two percentage points (1961, 1971 and 1972) in Table 5.3S and no cases of a risk premium less than one percentage point. Since 1982 there have been nine cases of a real risk premium under two percentage points and six cases of a risk premium under one percentage point. From 1991-94 the year-end risk premiums in column (6) of Table 5.3S were consistently less than one percent and did not change very much from one year to the next, compared to the fluctuations in risk premiums that occurred before the last recession. The implication would seem to be that some investors may now be using something akin to this type of risk measurement in managing their own portfolios.
Lower dividend yields and smaller inflation adjusted risk premiums, other things equal, should help to narrow the longer run return differential for stocks and bonds.
For the four years when the December risk premium in column (6) was under .55 percentage points (1983, 1986, 1991 and 1992) the following year financial returns for the S&P index were 6.3, 5.2, 7.4 and 9.4 percent respectively. In the 1995 edition of Johnson's Chartsit is reported that corporate bonds would have provided investors with financial returns (of 16.4, -.3, 9.1 and 12.4 percent) which were higher during three of the four following years.
The bull market of 1995 has been fueled in part by a reduction in long term interest rates and an impressive increase in corporate earnings in 1994 and during the first half of 1995. Lower interest rates are beneficial to investors in both stocks and long term bonds in the short run. It is by no means certain, however, that long term interest rates will continue to move downward.
The data in column (2) of Table 5.3S indicate that big increases in corporate earnings are eventually followed by periods of stagnant or declining earnings. From 1948-1979 stock prices were propelled upward by a more than doubling of the inflation adjusted earnings associated with the S&P index. From 1979-94, however, there was only a slight increase in the inflation adjusted earnings for this index. See column (3) of Table 5.4S.
In a world where there is no growth in the inflation adjusted earnings for representative stock price indexes, the average dividend yield on common stock will have to exceed the real inflation adjusted coupon rates on new issues of long term bonds, for one to be confident that stocks will continue to outperform bonds in the long run. See the appendix to this essay for the rationale behind this conclusion.
From 1981-94 there was only one end of the year dividend yield associated with the S&P index (for 1990) that was in excess of the December yield on Moody's high grade corporate bond index in column (4) of Table 5.3S minus the CPI inflation rate in column (5). The implication would seem to be that the managers of many pension funds, that don't have to pay taxes on capital gains, are now betting--whether they know it or not--that the US economy has returned to a bright new era where corporate earnings will continue to increase faster than the CPI inflation rate.
Accounting for differences in financial returns is an undeveloped science in part because of the problems that are encountered in measuring and compounding the financial returns for assets with different life expectancies but also because there is no generally accepted method of measuring risk premiums. The data presented in Tables 5.3S and 5.4S would strongly suggest, however, that most, if not all of the superior performance of stock over long term bonds from 1979-94 can probably be attributed to a dramatic reduction in the real risk premiums associated with common stock.
While the financial returns for stocks have varied more from year-to- year than the financial returns that have been calculated for bonds, it is by no means clear that a diversified portfolio of stocks will continue to be considered more risky than a portfolio of long term bonds. There is mounting evidence to suggest that the stock market is more stable now than it used to be (Essay 2S) and that the financial returns associated with long term bonds are more variable than was formerly the case. Many long term bond issues were clobbered more severely than representative stock market indexes during the Fed's preemptive strike against the possibility of accelerating inflation in 1994.
If the stock market is perceived as not being more risky than the bond market and continues to provide many investors with tax deferment advantages, it wouldn't be very surprising if the prices paid for equities eventually get bid up to the point of providing a smaller before tax return, on the average, than bonds.
The financial return for a stock, Rs, will be equal to its price at the end of the holding period, P, plus the dividends received during the holding period, D, when this sum is expressed as a percent of the price at the beginning of the holding period, Pb:
Rs = (P + D)/Pb = P/Pb + D/Pb (1)
Let us assume that a corporation wants to keep the dividend yield, D/Pb, constant. To accomplish that objective over a long period of time, the dividend growth rate, D/Db, will have to be equal to the stock price growth rate, P/Pb. Replacing P/Pb in equation (1) with D/Db allows one to derive a close approximation of the dividend growth model that was used to project the financial returns for the S&P index from one business peak to the next in Table 5.1S:
Rs = D/Db + D/Pb (2)
To maintain a constant dividend yield, corporate earnings must increase at least as fast in the long run as dividends. If earnings were to increase less rapidly than dividends, the dividend payout ratio will eventually exceed one and create an untenable deficit problem for the corporation. By a process of further substitution one can show that the longer run financial return will be equal to or less than the earnings growth rate, E/Eb, plus the dividend growth rate, D/Pb.
In a world where earnings are expected to increase at the same rate as consumer price inflation, CPI/CPIb, we can show by a process of further substitution that the real return on stock in the long run will be equal to a firm's dividend yield. And if bonds are perceived as being about as risky as stock, the real return on stock should be about the same as the real return on bonds in the long run.
With this type of reasoning in mind it is easier to appreciate why a dividend yield below the real return on long term bonds must be offset by a growth rate for corporate earnings that is in excess of the CPI inflation rate if stocks are to provide investors with a longer run return that is equal to the return on bonds. In the short run, however, there could still be lots of variation in the returns associated with these two asset classes.
Gordon, Myron (1962). The Investment, Financing and Valuation of the Corporation(Business One Irwin), Chapter 4.
Modigliani, F. and R. Cohn (1979). "Inflation, Rational Valuation and the Market," Financial Analysts Journal, March 1979, pp. 24-44.
Renshaw, Edward (1992). The Practical Forecasters' Almanac(Burr Ridge, Illinois: Irwin Professional Publishing).
-----"Is the Stock Market More Stable than It Used to Be?" The Financial Analysts Journal, tentatively scheduled for publication in the September/October 1995 issue.
Siegel, Jeremy (1994). Stocks for the Long Run(Burr Ridge, Illinois: Irwin Professional Publishing).
Smith, Edgar (1925). Stocks as Long Term Investments(New York: Macmillan).
S & P's Composite
Years of Stock Price Index Yield in Dividend Projected Actual
Peak ----------------- Percent Growth Average Average
Prosperity Year Associated Rate in Return for Return in
End Dividends Percent the Following Following
Price per Year Period Period
(1) (2) (3)n (4)n (5)n (6)n
1926 $.69
1929 21.45 .97 4.5 12.0 16.5 -3.0*
1937 10.55 .80 7.6 -2.4 5.2 9.6
1944 13.28 .64 4.8 -3.1 1.7 8.6
1948 15.20 .93 6.1 9.1 15.2 17.1
1953 24.81 1.45 5.8 9.3 15.1 17.2
1957 39.99 1.79 4.5 5.4 9.9 16.9
1960 58.11 1.95 3.4 2.9 6.3 8.5
1969 92.06 3.16 3.4 5.5 8.9 4.6*
1973 97.55 3.38 3.5 1.7 5.2 6.5
1979 107.94 5.70 5.3 9.1 14.4 11.9*
1981 122.55 6.63 5.4 7.8 13.2 16.0
1990 330.22 12.10 3.7 6.9 10.6 11.5p
1994p 459.27 13.18 2.9 2.2
(3)n. Column (2) expressed as a percent of column (1).
(4)n. Compound average annual growth rates for the associated dividend figures in column (2), 1926-29, 1929-37, etc.
(5)n. Column (3) plus column (4).
(6)n. Compound average annual financial return (price appreciation plus dividends associated with the S&P index) from the end of the year in question to the end of the next year of peak prosperity.
*Actual return in the following period was less than the projected return. For these three cases there was both an increase in the dividend yield in the following period and a slump in the dividend growth rate.
p represents a preliminary estimate based on the compound average growth rate for the four year period 1991-94. One would hope that the next year containing a recessionary peak in business activity won't occur for many years in the future.
Source of basic data: Standard and Poor's Security Price Index Record.
Date of Dividend Yield Closing Value S&P Index
------------------------------ -------------- ------------------------
Dividend S&P Dividend At At After Interim After
Yield Index Yield Low High Dividend Low Dividend
Low Low High Low High
Sep. 1929 6/ 1/32 June 1932 2.92 10.30 30.16 4.40 4.43
July 1933 4/28/42 Apr. 1942 2.95 8.47 9.95 7.47 7.66
Nov. 1961 6/26/62 June 1962 2.84* 3.78# 71.32 52.32 54.75
Oct. 1965 10/ 7/66 Oct. 1966 2.91 3.76# 92.42 73.20 80.20
Nov. 1968 5/26/70 July 1970 2.92 4.20 108.37 69.29 78.05
Jan. 1973 10/ 3/74 Aug. 1982 2.69* 6.32 116.03 62.28 119.51
Aug. 1987 12/04/87 Oct. 1990 2.69* 4.01 329.80 223.92 304.00
Jan. 1994 4/ 4/94 Dec. 1994 2.69* 2.91# 481.61 438.92 459.27
July 1995p 2.50*p
*Cases where the dividend yield either established a new record low or tied the preceding record low.
#Cases where the stock market correction occurred in the midst of prosperity and where the high yield for the S&P index is not associated with an economic recession.
p equals a preliminary record low estimate that may turnout to be too high.
Source of basic data: Standard & Poor's Security Price Index Record.
Year Closing Earnings High Moody's Percentage
Value for the Earnings December CPI Real Change
S&P S&P Price Aaa Corp. Inflation Risk S&P
Index Index Ratio Bond Yield Rate Premium Index
(1) (2) (3)n (4) (5) (6)n (7)
1957 39.99 3.37 9.05 3.81 3.02 8.26 -14.3
1958 55.21 2.89 6.56 4.08 1.76 4.24 38.1
1959 59.89 3.39 6.04 4.58 1.50 2.96 8.5
1960 58.11 3.27 6.23 4.35 1.48 3.36 - 3.0
1961 71.55 3.19 5.06# 4.42 .67 1.31 23.1
1962 63.10 3.67H 5.82 4.24 1.22 2.80 -11.8
1963 75.02 4.02H 5.36 4.35 1.65 2.66 18.9
1964 84.75 4.55H 5.37 4.44 1.19 2.12 13.0
1965 92.43 5.19H 5.62 4.68 1.92 2.86 9.1
1966 80.33 5.55H 6.91 5.39 3.35 4.87 -13.1
1967 96.47 5.33 5.75 6.19 3.04 2.60 20.1
1968 103.86 5.76H 5.55 6.45 4.72 3.82 7.7
1969 92.06 5.78H 6.28 7.72 6.11 4.67 -11.4
1970 92.15 5.13 6.27 7.64 5.49 4.12 .1
1971 102.09 5.70 5.66 7.25 3.36 1.77 10.8
1972 118.05 6.42H 5.44 7.08 3.41 1.77 15.6
1973 97.55 8.16H 8.36 7.68 8.80 9.48 -17.4
1974 68.56 8.89H 12.97 8.89 12.20 16.28 -29.7
1975 90.19 7.96 9.86 8.79 7.01 8.08 31.5
1976 107.46 9.91H 9.22 7.98 4.81 6.05 19.1
1977 95.10 10.89H 11.45 8.19 6.77 10.03 -11.5
1978 96.11 12.33H 12.83 9.16 9.03 12.70 1.1
1979 107.94 14.86H 13.77 10.74 13.31 16.34 12.3
1980 135.76 14.82 10.95 13.21 12.40 10.14 25.8
1981 122.55 15.36H 12.53 14.23 8.94 7.24 - 9.7
1982 140.64 12.64 10.92 11.83 3.87 2.96 14.8
1983 164.93 14.03 9.31 12.57 3.80 .54* 17.3
1984 167.24 16.64H 9.95 12.13 3.95 1.77 1.4**
1985 211.28 14.61 7.88 10.16 3.77 1.49 26.3
1986 242.17 14.48 6.87 8.49 1.13 -.49* 14.6
1987 247.08 17.50H 7.08 10.11 4.41 1.38 2.0**
1988 277.72 23.76H 8.56 9.57 4.42 3.41 12.4
1989 353.40 22.87 6.72 8.86 4.65 2.51 27.3
1990 330.22 21.34 7.20 9.05 6.11 4.26 - 6.6
1991 417.09 15.97 5.70 8.31 3.06 .45* 26.3
1992 435.71 19.09 5.45 7.98 2.90 .37* 4.5**
1993 466.45 21.88 5.09# 6.93 2.75 .91* 7.1**
1994 459.27 30.63H 6.67 8.46 2.67 .88* - 1.5
(3)n. The highest earnings for the current year or any previous year expressed as a percent of the closing value for the S&P index in column (1). The # marks for 1961 and 1993 identify the two lowest year-end earnings price ratios in the post 1929 period.
(6)n. Column (3) minus column (4) plus column (5).
H indicates the record earnings that are used to compute the high earnings price ratios in column (3).
*Years when the risk premium in column (6) was less than 1.00 percentage points.
**Price appreciation for the S&P index following real risk premiums in Column (6) that are less than .60 percentage points. The price appreciation in these years has (so far) always been positive but less than the longer run average.
Source of basic data: Standard and Poor's Security Price Index Record and the Economic Report of the President.
Financial Returns
Year Earnings CPI Real ----------------------------
(dollars) (1982-84=100) Earnings Current Year Following Year
(1) (2) (3)n (4) (5)
1928 1.38 ---- ---- 41.9 - 7.9*
1929 1.61 17.3 9.31 - 7.9P -23.9
1948 2.29 24.1 9.50 5.4P* 17.8
1949 2.32 23.8 9.75 17.8 30.5
1950 2.84 24.1 11.78 30.5 23.4
1955 3.62 26.8 13.51 31.0 6.4*
1962 3.67 30.2 12.15 - 8.8* 22.5
1963 4.02 30.6 13.14 22.5 16.3
1964 4.55 31.0 14.68 16.3 12.3
1965 5.19 31.5 16.48 12.3 -10.0
1966 5.55 32.4 17.13 -10.0 23.7
1968 5.76 34.8 16.55 10.8 - 8.3*
1969 5.78 36.7 15.75 - 8.3P 3.5
1972 6.42 41.8 15.36 18.7 -14.5*
1973 8.16 44.4 18.38 -14.5P -26.0
1974 8.89 49.3 18.03 -26.0 36.9
1976 9.91 56.9 17.42 23.6 - 7.2*
1977 10.89 60.6 17.97 - 7.2 6.4
1978 12.33 65.2 18.91 6.4 18.4
1979 14.86 72.6 20.47 18.4 31.5
1981 15.36 90.9 16.90 - 4.8P* 20.4
1984 16.64 103.9 16.02 6.0* 31.1
1987 17.50 113.6 15.40 5.7* 16.3
1988 23.76 118.3 20.08 16.3 31.2
1994 30.63 148.2 20.67 1.3 ?
(3)n. Column (1) divided by column (2).
P signifies a financial return associated with a year containing a peak in business activity.
*Identifies the poorest financial return in either the current or following year when the earnings for the S&P have achieved a new record after being depressed for a year or more.
Source of Basic Data: BCI series 320 and Standard and Poor's Security Price Index Record.
Go on to Essay 6S:
Return to the Introduction