The reality is, at least according to one analysis of the stock market, most returns come from dividends, and over long periods stocks do not perform better than inflation (see “How to get active in long-term investing,” December 2011). These concepts, which run counter to popular assumptions, are rooted in the worst of times: The 1929 and 1987 crashes, the 2000 Internet bubble, the 2000-02 bear market, the 2008 banking disaster and the current euro crisis.
This understanding is important when formulating strategies to predict stock market returns based on fundamental and economic data. We’ll test these ideas on S&P 500 index data from Feb. 28, 1930 through July 29, 2011. We’ll begin with a quick look under the hood of the index itself.
The S&P 500 (SPX/INX) is an index of U.S. equities. It includes 500 U.S. large-cap companies and is accepted as a reliable gauge of the health of the U.S. market. All the stocks in the S&P 500 are listed on the New York Stock Exchange or Nasdaq.
The stocks in the S&P 500 index are not necessarily the largest companies, but those with the largest market capitalizations (at least $5 billion). Companies are selected based on liquidity, sector and other factors.
The significance of the index doesn’t just sit with its popularity among retail traders. Exchange-traded funds (ETFs) and index-linked funds base much of their holdings on the S&P 500. Companies moving into and out of the index can cause fluctuations based on balancing-related sales and purchases alone. Also, because the index is considered a benchmark by which actively traded funds are judged, its rise and fall can spark shifts in large trader sentiment.
To calculate the S&P 500, we first need to calculate the market capitalization of each company. This is simply the total outstanding shares multiplied by the stock price. The sum of these values is referred to as the total market capitalization of the index. The weighted market capitalization for each company comes from dividing each market capitalization by the sum total capitalization and multiplying the result by 100.
According to Standard & Poors, on Dec. 20, 2011, the S&P 500 had a market capitalization of $11.238 trillion. The four largest stocks in the index were Exxon, Apple, IBM and Chevron (see “Big money,” below). For example, Exxon had 4.790 billion shares outstanding. With the stock trading at $82, Exxon’s market capitalization was $329.780 billion. The stock’s weighted capitalization with respect to the index itself was 3.49%.
One closely watched calculation is the S&P 500 earnings data. The weighted capitalization figure is used to arrive at this number. First, we calculate the earnings per share (EPS) of each company by dividing each company’s net earnings by its number of outstanding shares. Next, we multiply the EPS by a weighted factor for a given stock to get a weighted EPS. We then calculate the S&P 500 earnings by summing the weighted EPS values for each company.
Earnings are relative over time because of inflation, so $2 per share in 1980 is not the same as $2 per share in 2011. Hence, we must adjust earnings for inflation. This constant dollar earning figure compares the real value of the price/income from the nominal value point of view. For instance, this method compares the purchasing power of a typical first job in 1957 with the purchasing power of a typical first job in, say, 1986. Likewise, it determines whether a gallon of gasoline in 1972 was costlier than today by comparing the hours of work required to earn the money needed to buy the gas.
Converting nominal to constant dollar values makes the comparison of the salary data possible. Using the Bureau of Labor Statistics inflation calculator, we can see that $5,000 in 1957 is equal to $19,502 in 1986 dollars, and that $18,000 in 1986 is equivalent to $4,615 in 1957 dollars.
Robert Shiller is a well-known analyst who has popularized what has become known as the Shiller P/E Ratio. For this calculation, Shiller uses quarterly earnings from Standard & Poors’ website and the monthly average of daily closes for the S&P 500 index. From the quarterly data, he interpolates monthly earnings. The daily closes and earnings figures are adjusted for inflation via the consumer price index (CPI) to get real earnings.
The Shiller P/E Ratio has been used and interpreted for the S&P 500 for years. Its current value is about 20.75, which is 30% higher than long-term average value of approximately 16. This suggests stocks in the S&P 500 are overvalued.
Real earnings model
Inflation-adjusted real earnings can be used to build a predictive model of S&P 500 price action. While we can’t consider this a standalone trading model because of its long-term nature and the estimation of earnings data, we can demonstrate its value as a measure of market sentiment. We will start our analysis using the S&P 500 cash index without any adjustment for dividends.
The core premise of the model is that earnings momentum leads price. We can build a simple intermarket divergence model using this relationship. Here are the rules:
- If real earnings are in an uptrend and the S&P 500 is in a downtrend, we buy
- If real earnings are in a downtrend and the S&P 500 is in an uptrend, we sell
And here’s the code:
Sub SP500_EarningSimple (MkLen,IntLen)
Dim Mk_Osc As BarArray
Dim Int_Osc As BarArray
Mk_Osc=Close-Average(Close,MkLen,0)
Int_Osc=Close Of independent1-Average(Close Of independent1,IntLen,0)
If Int_Osc>0 And Mk_Osc<0 Then Buy(“”,1,0,Market,Day)
If Int_Osc<0 and mk_osc>0 Then Sell(“SE”,1,0,Market,Day)
End Sub
We use 8.0 for the MkLen input and 22 for IntLen. We test this from Feb. 28, 1930 to July 29, 2011 with no deduction for slippage and commission. Results are in “Earnings-based model” (below).
On the long side, the model does well, winning 71% of its trades. The short side is profitable, but it’s only successful 21% of the time. When we take a closer look, we see the reason for the low winning percentage on the short side. Many of the short trades are up 5% to 10% before the trade reverses. Divergence between
S&P 500 prices and real earnings does work on the short side as an entry, but waiting for the reversal as a stop and reverse system is a poor trade exit.
Other factors may be at play, as well. Earnings are mean reverting, particularly on a stock index such as the S&P 500; companies won’t go out of business, they just will be removed from the index. We used this additional concept to develop a more complex model (code and results, below).
Sub SP500_RESystem(MkLen,IntLen,PEBuyLev,PESellRatioLev,ShortTarget)
Dim Mk_Osc As BarArray
Dim Int_Osc As BarArray
Dim RateModel As BarArray
Mk_Osc=Close-Average(Close,MkLen,0)
Int_Osc=Close Of independent1-Average(Close Of independent1,IntLen,0)
If Int_Osc>0 And Mk_Osc<0 Then Buy(“”,1,0,Market,Day)
If Int_Osc<0 and mk_osc>0 Then Sell(“SE”,1,0,Market,Day)
If EntryPrice<>0 Then
If Close/EntryPrice<1-ShortTarget ThenExitShort(“ShortTarget”,”SE”,1,0,Market,Day)
disablesignalafterexit(“SE”,”ShortTarget”,Not(MarketPosition=1))
End If
If Close Of independent1/Close[1] Of independent1Sell(“SEPE”,1,0,Market,Day)
If Close Of independent1/Close[1] Of independent1>1 And Close of independent1Buy(“LEPE”,1,0,market,day)
End Sub
Here’s the system in simple English. We buy if real S&P 500 earnings are below 12 but are flat or rising. We also exit the short entries based on divergence at a target 5% sell off. Once we hit this target, we do not make this divergence sell again until the system goes long. We used a 22-month moving average for the S&P 500 and a 24-month moving average for earnings. Finally, if earnings drop more than 5% in a given month, we also go short.
We test on the same time frame as the first system, with no deductions for slippage or commissions. Results are shown in “Super system?” (below).
We have more than 70% winning trades. More impressive, almost 70% of the trades are on the short side. We make more than 3,100 points since 1930. Buy and hold was about 1,200 points during that period. This is without even considering dividends.
These rules have not had a losing year since 1988. Our analysis period is 927 months while our system spent 677 months in trades and 250 months out of the market. This is more than 20 years in cash. Not only does so much time in cash limit risk, but even if we assume an average of 2% interest during these periods, it increases the growth rate by 40%.
Price to book
While the market’s view of any company may be captured in the P/E ratio, the return on equity is what drives shareholders. Another important ratio is the price-to-book (P/B) ratio, as it takes into account all the factors (external and internal) influencing the price, giving an idea of the company’s health to investors. In fact, the P/B ratio is a sort of an extension of ROE and also is correlated strongly to P/E ratio.
Generally, the P/E ratio relationship to earnings is countertrend, as earnings mostly peak before prices; thus, earnings being reported quarterly is a problem. This is solved by the P/B ratio, which has different ratios for different market sectors and business types. Hence, an indicator for market sentiment is created that indicates whether people are ready to invest more in stocks in the immediate future.
Study shows that relationship of the P/B ratio is similar to earnings:
- An expanding P/B ratio is bullish.
- A contracting P/B ratio is bearish.
Our P/B data begins March 31, 1988 and is available only for the Dow Jones Industrial Average, so the test will be limited and stick to the simpler methodology — that is, SP500_EarningSimple (MkLen,IntLen). While the P/B ratio data is for the Dow, it is applicable enough to monthly moves in the S&P 500 to serve as a proof of concept. The test runs through Nov. 30, 2011. During this period, the S&P 500 made 822.77 points. We used a six-month moving average for both S&P500 and the P/B ratio.
The model went short on June 30, 2008, but did not reverse and cover up until June 2010. The market rallied almost 30% before we exited this short position. Some of the logic we used in our more complex earnings model would have helped this system, but unfortunately the limited data set doesn’t allow the degrees of freedom needed for a more complex approach to be valid.
History points the way
Understanding and using history to build a long-term market model is a ripe area of research. Perception affects prices. An expanding P/B ratio means people are willing to pay a high multiple for assets. Further, it means people perceive economic improvements — similar to how corporate bond spreads narrowing can precede market rallies.
Alas, these models are not necessarily viable for trading. The holding periods are too long, masking the risks and uncertainties of short-term fluctuations when leverage is employed. However, as long-term business or personal financial planning tools, they can provide invaluable guidance. In addition, combined with the right timing logic, they may offer filtering benefits for shorter-term systems.
Murray A. Ruggiero Jr. is the author of “Cybernetic Trading Strategies” (Wiley). E-mail him at ruggieroassoc@aol.com.
