In a new Yahoo column, the gracious Wharton professor Jeremy Siegel takes on his critics (including us) and defends his argument that stocks are “dirt cheap.”
The background: A few weeks ago, Professor Siegel argued that fair value on the S&P 500 is 1380, or 50% above current levels. He based this estimate on the assertion that “trend” earnings for the S&P 500 are $92 a share and that the appropriate P/E to apply to these earnings is 15X. These assumptions (and others) produced howls of protest here and elsewhere, from those who believe fair value on the S&P 500 is about 1000.
Professor Siegel’s key assumptions, most of which we challenged, included the following:
- Using “operating earnings” instead of GAAP earnings. Another way to describe “operating earnings” is “earnings before bad news.” It is OK to use operating earnings, as long as you apply the appropriate P/E to them, but in our opinion Prof. Siegel didn’t.
- Using only 18 years of earnings history instead of a century’s worth. Earnings have grown faster in the past two decades than in previous history, so this skews the data.
- Using a P/E that is too high. Using operating earnings would be defensible if one also used an average P/E applied to operating earnings, but Prof. Siegel uses a GAAP earnings P/E (and a higher-than-average one, at that).
But now Professor Siegel has fought back. In his new column, he defends his assumptions. We don’t buy his logic (see below), but, as usual, he is charming and articulate:
In last month’s article, my fair-value estimate for the S&P 500 Index was derived from a “normalized” earnings of $92 a share for the S&P 500 Index times a 15 average price-to-earnings ratio for the stock market. “Normalized” earnings are earnings stripping away the effects of business cycle, so that normalized earnings are higher than actual earnings in recessions, such as now, and lower at the top of booms. Currently, Standard and Poor’s is projecting that operating earnings for the S&P 500 Index in 2008 will be $64.14 while reported earnings, which contains very large write-offs from a few firms, are at $49 per share.
My estimate of $92 for normalized earnings drew considerable criticism, partially because it was based only on earnings over the past 18 years. If one looks at considerably longer-term time spans, some believe the trends forecast much lower earnings.
The farthest back we have historical earnings is 1872, based on valuable data that Prof. Robert Shiller brought to light in his book Market Volatility, published in 1989. These historical data are based on splicing together data from Standard and Poor’s that goes back to 1928 to earlier data compiled by Cowles Foundation researchers.
The graph below depicts those earnings per share data for the US stock market from 1872 through the present, corrected for inflation. Looking at this graph, it certainly appears that earnings over most of the past 20 years have been far above average and that these earnings may now correct to the mean of only $56.40 per share, not far from reported earnings projections this year. If we apply the historical P-E ratio of 15 to $56.40, we get a projected level of the S&P 500 Index of only 840.
Flaw in Analysis
But there is an important flaw in using a single trend line to project the future. Doing so assumes that there has been a constant growth rate of real per share earnings over the entire period, an assumption that depends, among other factors, on an unchanging dividend policy of firms.
Yet dividend policy of firms has changed dramatically. The ratio of dividends paid to earnings, called the payout ratio, has dropped significantly over the past 30 years. The decline in the payout ratio has substantially boosted the growth rate of earnings per share. Failure to take this into account will result in a serious underestimate of trend earnings.
The table below confirms that the fall in the payout ratio has boosted earnings growth. The shift in dividend policy took place in the early 1980s when firms, because of liberalized rules for share repurchases, taxes favouring capital gains, and the proliferation of management stock options, reduced the amount of earnings paid out as dividends. The average payout ratio before 1982 was 64.7%, about 40% higher than the 46.6% payout ratio after 1982. As the payout ratio declined, the dividend yield declined and the growth of earnings accelerated.
Finance theory predicts, and historical data confirms, that if a firm pays a lower proportion of its earnings as dividends, then these unpaid earnings must be used to either repurchase shares, lower debt, or invest in capital. In any of these cases, per share earnings growth will increase.
Table 1. Selected Stock Market Variables
Financial Variable 1871-2007 1871-1981 1982-2007 Payout Ratio 61.2% 64.7% 46.6% Dividend Yield 4.52% 4.96% 2.66% EPS Growth 1.56% 1.41% 4.5%A higher rate of earnings growth from lowering the dividend does not imply that investors obtain a higher return on their investment when management lowers the dividend payout. The return to shareholders is the sum of the dividend yield and price appreciation. On average, a dollar of retained earnings will yield investors a return that compensates the investor for the lost dividend income. But dividend policy will impact earnings growth.
A higher rate of earnings growth also means that using “smoothed” earnings, derived by averaging 10 years’ of past data as Prof. Shiller and others advocate, will yield a distorted valuation of the market. If earnings growth has accelerated, the average of the past 10 years will result in a greater underestimate of earnings in more recent decades.
Look back at the long-term chart again. A new trend line has been drawn covering the period of faster earnings growth that occurs after 1981. According to the new trend, the normalized level of 2009 earnings is a far higher $87.66 than predicted by the single trend line. This earnings figure is slightly below the $92.00 that I estimated in my previous article using analysis over the past 18 years, but it is not appreciably different.
If we apply a 15 P-E ratio to $87.66 level of earnings, we get a fair value of the S&P 500 Index at 1315, almost 50% above the level the market closed on the Wednesday before Thanksgiving. Furthermore, with the 10 year government bond rate falling below 3%, we are in the lowest interest rate environment we have had over the past 50 years. Normally, when interest rates are this low, stocks sell for a higher-than-average P-E ratio. To obtain 1380 fair value of the market that I estimated last month, the P-E ratio would need to be less than one point above its long run average of 15.
Why this is wrong:
Prof. Siegel’s “earnings are growing faster now” argument is just a variant of the familiar “it’s different this time.” As Prof. Siegel wisely pointed out in his excellent book Stocks For The Long Run, this argument shouldn’t be dismissed out of hand, because sometimes it is different this time. But in this case, it almost certainly isn’t.
Prof. Siegel’s “earnings are growing faster now” logic rests on the dividend payout ratio, which is currently lower than it has been historically. Even if one buys the argument that a lower dividend payout ratio should produce higher earnings growth (which many experts don’t), this assumes that the payout ratio will stay depressed forever. We doubt it. If the stock market continues to perform poorly, we expect that companies will begin to increase their dividend payouts to appease angry shareholders. If this happens, and Prof. Siegel is right that the low payout has accelerated earnings, this effect will actually work in reverse: Earnings growth will be artificially SLOW until the payout ratio stabilizes.
Prof Siegel may be right that something has caused earnings to grow more rapidly than normal in the past couple of decades–although this “faster” growth could also simply have resulted from starting his measurement at the trough. What do we mean? Look where Prof. Siegel’s pink “trend” line starts above. If you went back and drew a similar pink line between 1921 and 1951, you’d probably conclude that earnings were “growing faster now” then, too.
If earnings really have grown faster than usual in the past two decades (measured from midpoint to midpoint, not trough to peak), we suspect this has more to do with accounting, debt (which also funds share buybacks), tax policy, and the expansion of profit margins, all of which TEMPORARILY boosted earnings growth. It’s certainly possible that the dividend-payout ratio played a role. But that doesn’t mean it will continue to.
“Operating Earnings” and P/E Assumption
Prof. Siegel also defends his decision to use operating earnings and a 15X P/E:
Others may object to my use of “operating earnings” rather than reported earnings, in making my projection. Operating earnings exclude some of the big write-offs that arise from restructuring, pension revaluations, impairment charges, and portfolio losses, although they do include many of the recent write-offs of financial firms.
Although operating earnings are only reported back to 1988, the earnings data before that date are, according to sources I have contacted at Standard and Poor’s, closer to the concept of operating earnings than reported earnings. This is because FASB has sharply increased the number and type of charges that firms must write off and this has depressed reported income, especially during recessions.
The increasing number of large write-offs has also led to a distorted look at the earnings levels for the S&P 500 Index. In another article, I have discussed how there are major distortions in the official earnings numbers provided by index providers like S&P. Six firms (AIG, Wachovia, Sprint, GM, Merrill Lynch, and Citigroup) with over $170 billion dollars in losses over the past year, lowered aggregate earnings on the S&P 500 Index by nearly $20 per share. Yet these firms are tiny and represent less than 1% of the S&P 500 by weight. That is to say over 99% of the S&P 500 had earnings that were closer to $70 per share when the officially reported results are more like $50 per share.
With the 10 year government bond rate falling below 3%, we are in the lowest interest rate environment we have had over the past 50 years. Normally, when interest rates are this low, stocks sell for a higher-than-average P-E ratio. To obtain 1380 fair value of the market that I estimated last month, the P-E ratio would need to be less than one point above its long run average of 15.
Why this is wrong:
Prof. Siegel’s argument here is that we only have “operating earnings” data back through 1988 and that GAAP earnings before that date are very close to “operating earnings,” so the distinction is irrelevant. This is an interesting argument, one we haven’t heard before. For now, therefore, we’ll simple say that others violently disagree.
Professor John Hussman, for example, whose fair value estimate is about 1100, has also looked at the “operating earnings” question. He suggests that the appropriate average P/E multiple to use on operating earnings is about 11X-12X, not the 15X Prof. Siegel uses.
Lastly, Prof. Siegel bases his decision to use a higher-than-normal PE on interest rates. This is defensible given current interest rates, which are lower than average. But, again, it assumes that interest rates won’t change, which they almost certainly will.
The point of doing a “fair value” analysis is to try to determine the average fair value over time, taking into account many constantly changing variables. By the time the rest of the market comes around to Prof. Siegel’s conclusion that the market P/E should be higher on account of low interest rates (a consensus, we might add, that hasn’t stopped it from plummeting over the past year), interest rates may well be higher. At which point, Prof. Siegel will have to lower his fair value estimate.
The only variable that changes the fair value estimates of Robert Shiller, Andrew Smithers, Jeremy Grantham, John Hussman and others, meanwhile, is the gradual growth of trend earnings. Which, as we’ve already noted, these experts place far closer to the $56 Prof. Siegel cites above than the $92 based on the assertion that it’s different this time.