Comstock Partners, Inc.May 31, 2003
A Simple Calculation
Price divided by earnings: What could be easier? By CHARLIE MINTER and MARTY WEINER THE READERS OF MOST financial publications must be somewhat confused by the different P/E ratios that are used by the so-called experts on Wall Street. They read or hear a well-respected analyst or strategist eloquently making a very compelling case as to why the market is more undervalued now than almost any time in history. After hearing this, they must feel more secure about their investment portfolio. Then, the very next day, another pundit (it could be either of us) comes on the same show, or is quoted in the same newspaper, making the case that the stock market has never in all history been so overvalued! The difference between the two different cases is very confusing, but we will try to make it as simple as possible. The difference between those who assert that P/Es are high and those who believe P/Es are low comes down to three items. First, in making the P/E calculation we can use "reported" earnings, which correspond to GAAP (Generally Accepted Accounting Principles) or we could use "operating" earnings, which exclude "write offs." Second, we can use either trailing actual earnings for the last four quarters, or estimated 2003 earnings. Third, we can adjust the calculated P/E for current interest rates and inflation, or decide that no adjustment is necessary. The so-called Fed model compares the forward operating P/E to the 10-year Treasury-bond yield. Let's first examine the logic of the pundits who believe the market is undervalued. They use "operating" earnings, which exclude "write offs," because they believe a legitimate argument can be made to exclude expenses that are not recurring. They also believe the earnings estimates for 2003 or even 2004 should be used since the stock market is "forward looking". The last point is whether to adjust for interest rates and inflation. They make the case that when interest rates are low it is sensible to adjust the P/E to higher levels since alternative investments offer lower yields. They also believe that discounted future earnings are worth more when the inflation rate is low. We, on the other hand, believe that the only important data to use for the P/E valuation metric are "reported" trailing twelve-month earnings, not adjusted for interest rates. We believe this is the best way of determining whether the stock market is overvalued or undervalued. History shows what investors paid at market peaks and at market troughs. These are the only earnings measures that can be viewed over very long periods. Buyers of the Standard & Poor's 500 Index are presently paying over three times the P/E they normally have paid at the end of bear markets. Our problems with the way most pundits on Wall Street look at valuation follow: First, there are no accounting standards to determine operating earnings. What expenses are included or excluded is purely discretionary and therefore inconsistent from one company to another. In addition, write-offs often occur year after year and are not truly nonrecurring. If companies and their accountants determine, with no disciplined method, that something should be written off and not included in the "operating earnings" it is excluded from the expenses. Remember, operating earnings only came into existence about 15 years ago and proliferated during the bubble, when no one cared what was being used for earnings as long as everyone was making money. Prior to the mid-1980s there was only one generally recognized earnings number, and that was reported -- GAAP -- earnings. Second, we don't like the idea of using forward earnings because, for the most part, earnings are impossible to predict. The estimate is almost always wrong, and almost always wrong on the high side. Who would have been able to predict that "reported" earnings (GAAP) would have dropped from $50 in 2000 to $25 in 2001? Even more importantly, forward-looking earnings are almost always higher -- usually much higher -- than trailing earnings. Therefore, the historical average P/E on forward-looking earnings (if we had the numbers) would be much lower than the average historical P/E of reported earnings. In other words, the average P/E over the past 75 years of trailing twelve-month earnings is 15.5, but if you were to use forward-looking earnings the average P/E would probably be closer to 12. Third, it's misleading to adjust P/E ratios for the low inflation and interest rates that exist presently. Most present-valuation models go back only a few decades and the models would not have worked if they covered a longer period of time. Take the case of the widely used Fed model, which relates the P/E ratio of the S&P 500 to the 10-year Treasury-bond yield. This model is based on data going back to about 1970. The problem is that the 10-year T-bond rate is now lower than at any point on the model, meaning that the model doesn't contain even a single data point that covers the present situation. In fact, prior to the late 1960s, interest rates were lower for one hundred years than they were from 1970 through 1997. Yet never during this time did the market sell at more than 23 times earnings. In our view, this is a flawed model. The hazards of using current but fleeting interest and inflation rates to determine reasonable P/E ratios were clearly illustrated in 1982. At that time interest and inflation rates were well into double digits, leading most experts to the erroneous belief that the market was not undervalued at seven times earnings. It would be just as erroneous today to assume that the market is not overvalued at the current high P/E multiple of 31.4. In response to the confusion, Standard & Poor's has proposed a new figure called "core earnings." The core earnings for the last 12 months came in at $23.34, which is much lower than both the "operating" and "reported" numbers. Perhaps this would make our case even stronger if we were to use these earnings, but there is no published history for comparing these core earnings to past peaks and troughs in the stock market. We will continue using the same earnings that have been used year after year for the past 100 years. In doing our calculation, we divide the present S&P 500 of 955 by $30.42 of earnings estimated for the four quarters ended March 31, 2003 and come up with a price-earnings multiple of 31.4, compared to a long-term historical average of 15.5. Many of the pundits use estimated 2003 operating earnings of $53.12 and get a P/E of 18. The $53.12 number is forward operating earnings and is most likely far too high. In our view, this method is betting on an ephemeral outcome with souped-up numbers. Keep in mind, too, that had forward operating earnings been used throughout history, the historical average P/E would be far lower than 15.5. In sum, we believe that using estimates of forward operating earnings and adjusting for interest rates produce valuations that are highly misleading and are likely to result in further losses for stock-market investors.