Long vs. Short-Term Stock Market P/E Ratios
Beware of misleading year-over-year earnings-per-share comparisons
by Robert E. Bronson, III, Bronson Capital Markets Research. April 19, 2010
So, Bunky, do you feel better by simply replacing the loss of your loved one?
How would you feel if a loved one of yours and others died in a crash one year ago, and now people are telling you that it wasn't such an Armageddon wipe-out because you can simply substitute living people for the deceased? This is just what Standard & Poor's and other indexers did when they erased the complete wipe-out of corporate earnings per share (EPS) by replacing 15 to 20 of the biggest losers in the S&P 500 index with much more profitable companies.
But we agree it wasn't Armageddon, even though it was the biggest loss in over 140 years of EPS history — in fact, 69% worse than during The Great Depression, even after the replacements. The EPS loss was no true love lost for even the most fundamentally-based investors, who claim to love EPS more than anything else for valuing the stock market. This is because they are periodically very moody and always trend-following momentum investors, who simply want to at least get back to break even and so are predictably "irrationally complacent" in the ongoing echo-mania bust.
In any case, corporate earnings reporting season for the first quarter of this year started on Monday, April 12. But before we explain why, how and to what degree investors' current consensus expectations are excessive, as reflected by the current valuation level of the stock market, some background will be helpful on the pros and cons of various concepts of the coming, sure-to-be-hyped reports of year-over-year (YoY) rate of change (RoC) comparisons. It will help explain why we expect investors will come to change their minds significantly on the attractiveness of the stock market, especially at its current level and as the first of the coming, after-shock, double double-dip recessions gets underway. (See As Forecasted — A 12-Year Retrospective, pages 5-8).
Many Different Ways to Calculate RoC and Their Advantages
Rate-of-change is usually calculated point-to-point (PtP), using just two data points, but the ratio involved can be an average of the latest data points (as the numerator) to an average of older data points (as the denominator). Or it can be the slope (like the first derivative in calculus) of an all-data best-fit line, which can be linear or curvilinear using either arithmetic or logarithmic data. Or it can even be the slope of the line that best fits just cyclically-extreme data — i.e., just cycle peaks or troughs. For example, we use both of these types of best-fit trendlines in our P/E high-low volatility channel, as well as with weekly initial unemployment claims, as is illustrated in the two-panel chart below.
In general, RoC calculations as a calculus of motion tool, more commonly understood as velocity and acceleration, or first, second and higher order derivatives (or differences), are useful as cyclically-leading indicators. As such, these RoC calculations very rarely exhibit false negatives, but that's at the expense of generating many false positives because they are so increasingly cyclically-sensitive the higher the order of differencing. Just like you had better know what you are doing when you use an extremely sharp knife, you had better understand the mathematics and know the pros and cons of RoC calculations. See the following explanation from a commentary written by one of our investment adviser clients, describing how we've used RoC analysis to anticipate the end of the "green shoot" downtrend in initial unemployment claims. The analysis in the chart was prepared a week before today's report of "Weekly jobless claims jump 24,000 to 484,000" (Wall Street Journal), which dramatically reinforced our forecast.
"Following our April Commentary, more employment and unemployment data was reported that confirmed the analysis and expectations we discussed in our previous commentaries. The new data, especially viewed in its historical context, as seen in the chart below, clearly demonstrate that the media-hyped assertion of improvement in the unemployment situation has been flat wrong. The obvious end of the downtrend in initial unemployment claims, the precursor to their flat-lining, if not rising again, is part of the continuous confirmation from numerous economic data series of the imminent dip in the economy, consistent with Bob Bronson's forecast of after-shock, double double-dip recessions in this echo-mania bust.
"In addition to forming classical Growth Cycle completion patterns for the downtrend, technically indicating that it's over, the rate of change in the four-week average (blue dotted lines in the both the upper and lower chart panels) has now doubly re-confirmed the end of the downtrend with a "more bad" (a direct repudiation of the overreaching "less bad" hype in recent months), "red shoot" (in repudiation of the over-hyped "green shoots" that did not indicate sustainable economic improvement) break-out to the upside in the rate of change of the rate of change in the raw initial unemployment claims data. The second derivative (or second rate of change, or acceleration) breakout (red circle and arrows in the lower chart panel), along with an associated 4th-degree polynomial (three cycle turns) upturn, confirms that the 45-week decline in initial unemployment claims from the March 28, 2009 high of 651,000 claims ended nine weeks ago on February 6.
"All of this is a leading indicator for both worsening employment and for the first of the after-shock, double double-dip recessions in the core, or private-sector, business cycle, even though the data also includes government unemployment claims. Also note that the stock market, which is a widely accepted leading indicator, and thus should be coincident with at least the trend in initial unemployment claims, has so far erred by ignoring this data, since it is well above its February 6 level. This is consistent with our previous explanation of how the stock market actually lags, rather than leads, the business cycle in a Supercycle Bear Market Period, and especially in a deflationary economic Supercycle Winter, like currently."
Click for a larger image for an extensive view
Like the weekly changes in initial unemployment claims and the monthly changes in payroll data, GDP is another example of RoC more commonly referred to in its first derivative form. Although GDP raw data is measured in trillions of dollars, which is almost always in an uptrend, the headline reports are about its growth rate, or its (first) rate of change, which is equivalently its first derivative or velocity.
Recently the concept of "green shoots" became popular (until "red shoots" appeared), which referred to the (slowing?) rate of change in the (declining?) rate of change, or the rate of change of the growth rate, or the second derivative, or the cyclical trend in these.
To clarify further, here is a table of some equivalent terminologies:
Seasonal Adjusting with PtP RoC
Although analysis using just two data is usually not very scientific, trivially simple PtP statistics are often required by law, like by the IRS for tax reporting, as is PtP RoC, like by the SEC for investment performance reporting. Most PtP RoC calculations are calendar-based ones like month-over-month (MoM), quarter-over-quarter (QoQ) and especially year-over-year (YoY), since the latter also provides adjustment for the annual weather- and holiday-seasonal cycle. However, while seasonal adjustment is very common and necessary, it doesn't always work for data like retail sales, which can be significantly affected by floating holidays like Easter. This, along with extraordinarily easy comparisons, favorably distorted the latest headline report on the YoY PtP RoC in same-store retail sales, which misleadingly hyped the stock market last Friday. More about this later.
We have examined the various seasonality software programs used by the government in reporting economic data. Although they remove almost all subjective judgment, which is appropriate for the general acceptance of bureaucratic-based analysis, we have found it wanting, even with their latest modifications of ARIMA X-13. This is because, among other problems, it arbitrarily limits the number of historically-averaged periods used in deriving a seasonal effect, and it cannot ferret out one-time or extraordinary shocks, like hurricanes, for example, which we've reported before. We prefer to graphically overlay many seasonal, or otherwise quasi-periodic, cycles to first visually identify, and then mathematically ferret out any anomalies and/or otherwise exogenously-impacted data, as well as analyze any changing trends that may be occurring. For example, this is how we discovered the start of the morphing of the 40-month Kitchin inventory cycle with the 48-month election cycle that occurred during the 1950s, which is explained in footnotes 4 and 14 here: A Forecasting Model That Integrates Multiple Business and Stock Market Cycles.
Beware of Distorting Denominators
It also can be very misleading when the older data used for comparison (the denominator of the ratio) is anomalously too high or too low, as is currently the case with corporate EPS (see our commentary below), the popular technical indicator of 52-week new highs and lows, and same-store retail sales, all of whose denominator distortions have confused the current, "irrationally complacent" Market Mind during the echo-mania.
Well-trained, experienced analysts know the calculation problems associated with extraordinarily small denominators, but many of them nevertheless opportunistically use the favorable distortion from extremely small denominators for excessively easy comparisons to advance their permabull and/or new-bull agendas and related spin.
Although all of this reminds one of the well-known Mark Twain expression, "Figures don't lie, but liars figure," the broader truth of the matter is better captured by "Lies, damned lies and statistics," or even better, "Anything can be proved with statistics — even the truth."
Analysis with Diminishing Denominators
Unbiased, scientifically-based RoC analysis requires both understanding and compensating for the problem of a zero, or negative, denominator, which cannot simply be done by transposing data into logarithms, since no power function equals zero. Here's a good book describing the history of the number zero and how it has puzzled and plagued even the most learned mathematicians, scientists, philosophers and theologians over the millennia: Zero: The Biography of a Dangerous Idea.
But unlike in RoC calculations, zero denominators are not always statistically fatal. For example, to correct for them in our multi-algorithm forecasting machine, which we call the Growth Cycle Trend Projector (GCTP), we use the The Calculus of Infinitesimals since zero in the denominators of these algorithms are removable singularities. GCTP operates on extremely small datasets and essentially extends the scope of traditional statistics, which is otherwise based only on the law of large numbers. GCTP generates the most likely follow-on, price-time geometry consequence out of all possible technical chart patterns (swings) over all time horizons, ranging from minutes and hours to Supercycles [1] and centuries. We believe it is one of the most powerful technical analysis tools ever developed. We are developing an easy-to-use application for its controlled public release.
But for now, what follows here is why, how and to what degree investors' current, consensus expectations are excessively bullish, as reflected in the current level of the stock market.
Expectations for EPS Are Excessive
The bullishly-biased financial media continue to tout supposedly "strong corporate earnings" to justify investors buying stocks at ever more expensive prices. Investors are not reminded, however, that the key factor driving both the magnitude and timing of this rebound is that it is coming off last year's extraordinarily low EPS base — in fact, the lowest base in over 140 years of U.S. corporate history! The two recent quarters of the rebound in EPS came only after nine quarters of decline that amounted to the biggest drop in EPS since The Great Depression, as measured by the companies of the Standard & Poor's 500 index and its previous and subsequent indexes that we use in our database, described here: Revealing BAAC Supercycles.
Moreover, the plunge in EPS would have been the worst ever recorded (see the steep, nearly vertical drop in blue dotted line in the chart below) if Standard & Poor's had not removed 15 to 20 companies reporting huge losses (primarily financial-sector companies) from the index and replaced them with much more profitable companies. (Note the uneven-handed treatment of outlying EPS results: when the financial companies, with their stratospheric EPS from their red-hot derivatives business, were the only companies keeping S&P 500 EPS in the black a year ago, those companies were kept in the index; but as soon as their massive losses, the result of the collapse of that very same derivatives business, would have plunged S&P 500 EPS into the red for the first time in history, those companies were removed.) Even with these "timely" substitutions, the drop in corporate EPS was still 69% greater than during The Great Depression, as seen in the magnitude of the near-vertical drop in the solid blue line in the chart below.
And EPS even in this newly constituted S&P 500 index have not recovered to anywhere near the previous peak in trailing four-quarter As Reported (GAAP) EPS of $84.92 in the second quarter of 2007, as seen in the solid blue line in the chart above. The current, consensus estimate of bottom-up analysts for the trailing four-quarter EPS through the now-completed first quarter is still just $59.26, while full-year consensus estimates, which assume a strengthening economy throughout 2010, are $62.09 for bottom-up analysts and $67.61 for top-down analysts, still well below their previous peaks.
Keep this in mind when you hear talk — completely out of context — about the "blow-out" 70% growth (ranging up to110% growth: S&P bottom-up operating GAAP EPS of $17.16 vs. $10.11 YoY and GAAP EPS of $15.81 vs. $7.52) that is the consensus expectation for the first-quarter EPS comparison. By the way, the Q1 growth rate of somewhere around 70% will be the peak rate for GAAP EPS on a YoY basis, after which it will rapidly cut in half — twice - over the following two quarters, Q2 and Q3, since there will no longer be the extraordinarily low base from a year earlier for such exaggerated comparisons.
More importantly, factors in our forecasting models unequivocally point to the economy weakening, not strengthening, throughout the rest of 2010. Furthermore, corporate profits to date have been driven more by cost-cutting — especially job elimination — than by revenue growth. Firing people is the most effective means of lowering expenses, since labor costs are on average about two-thirds of a company's cost structure. However, with consumers' and businesses' persistent reluctance to spend, it will become increasingly difficult for companies to expand their profitability through additional cost-cutting measures. It is easy to see, therefore, that both analysts' and investors' current, overly-hyped expectations are far too overly optimistic, the coming realization of which will translate into the selling of equities that will drive the stock market, as well as P/E ratios properly measured on a longer-term basis (as discussed below), down to the substantially lower levels we have been anticipating.
The Stock Market Is Extremely Overvalued — Again
Our Supercycle valuation indicators all consistently agree that the stock market has not reached the end of the current deflationary economic Supercycle Bear Market Period. That the market could ever have risen to recent levels at all serves as testament to the revived "irrational complacency" of investors as they piled in, hoping despite all data to the contrary that a sustainable economic recovery and, therefore, a sustainable bull market might be underway. To justify buying stocks at their current lofty levels, permabulls are purposely and misleadingly playing down the fact that EPS are rebounding from all-time record lows (which is further obscured by the apples-to-oranges comparisons that followed S&P's removal of the worst-performing companies from the S&P 500 index), and are shifting attention instead to the YoY EPS growth rate bounce (from extraordinarily low year-ago levels) and to short-term P/E ratios, which they argue are not out of line.
With today's intraday high in our capitalization-weighted index, which is our unique price measure of the stock market Revealing BAAC Supercycles (equivalent today to 1234 in the S&P 500 index), the stock market has a historically very high short-term P/E ratio of 20.0, using S&P's latest (April 7) estimate for trailing four-quarter GAAP EPS through Q1 of $59.26. During the past 139 years there have been 84 monthly periods, or about 5% of the time, in which the P/E ratio has been within a 5% range of the current P/E ratio (i.e., 21.0 max and 19.0 min). The subsequent, median six- and 12-month price-only returns were 56% and 0% less, respectively, than the median returns of all 1,667 monthly periods, which were +3.2% and +6.1%, respectively. However there was a huge dispersion of performance comprising these medians, which included big stock market declines following, most notably: the six- and 12-month periods following Sep 2007 (the month before the last bear market top); Apr, May and Sep 1987; Dec 1968; Feb, Mar and Aug 1946; Oct-Dec 1938; and various months during 1929-31. Today's stock market environment has similarities to several of these overvalued, over-owned and overbought periods.
Our historical study of stock market P/E ratios (described in the next section below) demonstrates that using future four-quarter GAAP EPS in a walk-forward model of historical data, which proxies for what would have been perfect one-year forecasts of GAAP EPS, did not improve the performance results. Thus, despite the common practice of research analysts using short-term P/Es to value individual stocks, industries and sometimes even sectors, their use for predicting the whole stock market — even with perfect earnings foresight — has not proven to be useful. This is because even with perfect foresight on aggregate earnings, GAAP EPS, EBITDA or any other cash flow-type measure, it does not overcome the problem that the stock market is composed of both cyclical- and growth-company stocks, whose individual P/E ratios move in opposite directions throughout the business cycle and, thus, are extremely confounding in their aggregate ever-changing mix, especially over shorter periods of less than several years with an also ever-changing business cycle. See Supercycle Corporate Earnings and Market P/E Ratios.
So even if the next four-quarters' GAAP EPS were to be as high as the $84.92 peak in Q2 2007, which is virtually impossible, the current P/E of that future GAAP EPS would generally be considered to be a more reasonable 14.5, or slightly above the historical average of 14.3. However, during the 100 previous times that such a future-EPS P/E ranged between 14.1 and 15.0, the subsequent median six- and 12-month price-only returns, were 14% more and 28% less, respectively, than the median returns, which were +3.7% and +4.4%, respectively. There was also a huge dispersion of performance comprising these medians, which included big stock market declines following, most notably for example: May-Jul 1973 (which was the "irrationally complacent" echo-mania high during that Supercycle Bear Market, which we think is similar to today's stock market); and May 1937 and Nov 1939 (also similar to the 2007 stock market top and to the current stock market).
We believe it cannot reasonably be argued that today's stock market is fairly valued, much less undervalued, based on past or future earnings, however measured (e.g., GAAP, operating, EBITDA or any cash-flow type), or the related short-term P/E ratio, or their relationship to bond yields (interest rates), the latter as explained here: Quantifying and Forecasting an Equity Risk Factor.
The Importance of the Stock Market's Long-Term P/E Ratio
More important than the impact on short-term P/E valuations, the stock market's speculative,13-month, 64% retracement of its previous 57.7% decline, which has been on persistently low trading volume concentrated in the most financially distressed companies, has brought the stock market P/E, when more properly measured on a longer-term basis, into nosebleed territory once again to 29.1. This is illustrated in the chart below of monthly P/E data through March 2010. Keep in mind that this ridiculously high valuation follows immediately on the heels of the biggest drop in GAAP EPS in over 140 years of U.S. corporate history and the most severe recession since The Great Depression!
In our P/E Predictor Study I (expanding on research initially done by Yale economist Robert Shiller and Harvard economist John Campbell) - available upon request - we demonstrate that the best predictor of future stock market performance using P/E ratios is obtained by using a very-long-term averaging of EPS for the E component. Our work demonstrates that E is optimally averaged using a monthly exponential weighting factor of 94.6%, which is both smoothing- and lag-equivalent to a three-year, or 12-quarter, moving average of EPS. Using such a very-long-term EPS average, the resulting market P/E ratio explains about 50% (the R-squared) of the stock market's performance over the subsequent 12 to 20 years. One hundred forty years of U.S. stock market history shows that a Supercycle Period is exactly the minimum time horizon for which such optimally computed stock market P/Es have that maximum predictability for the stock market. For shorter time horizons, the R-squared for, and thus the usefulness of, using the stock market's P/E ratio for forecasting its future performance, no matter how the P/E is computed, falls off dramatically.
In our BAAC Supercycles report (available upon request), we demonstrated that the market P/E ratio is highly cyclical, rising to an extreme high by the end of Supercycle Bull Market Periods, then falling to an extreme low by the end of Supercycle Bear Market Periods, as investors' appetite for the risk of owning stocks waxes and wanes over time, as you can see in the chart above. The fact that the current market P/E ratio, optimally computed and correctly analyzed on a longer-term basis, is still so close to its all-time record high is yet another, strong indication that the stock market has to decline substantially to bring the P/E to its necessary and sufficient, extremely low level.
Even using short-term calculations of P/E ratios, the S&P bottom-up analysts' consensus estimate through the first quarter of 2010 for trailing four-quarter GAAP EPS of $59.26 still results in a market P/E ratio of a relatively high 20.5, as compared with a long-term historical average of around 14. The current, forward-looking consensus estimate among S&P analysts of $62.09 GAAP EPS for all of 2010 still results in a market P/E ratio of 20.0. Most permabulls are expecting full-year GAAP EPS to exceed $70, with many stretching to justify the recent stock market advance by hoping for an absurd $80 or so, or back to the level of the previous 2007 highs, which would require another 26% to 37% extension of GAAP EPS that simply is not going to occur before the after-shock, double double-dip recessions start taking their toll on GAAP EPS this year. See As Forecasted — A 12-Year Retrospective.
Only either an even more enormous rise in GAAP EPS — which not even permabulls or new bulls are anticipating — or a much greater than 50% decline in stock prices can quickly bring the current, extremely overvalued, long-term stock market P/E ratio of 29.1 down to the extremely undervalued level that would mark the end of the current Supercycle Bear Market Period. But more realistic than the linear projections described in the callout boxes on the right-hand side of the chart below, a mean-reversion target area of 8 (idealized) to 10 (minimum) can be realized over the next 4.5 years ending in October 2014. From a timing point of view, that would be an idealized end of the Supercycle Bear Market Period, as we forecasted some 12 years ago, which we anticipated would result from the combination of the typically negative impacts at that time of the four-year stock market cycle, the four-year Presidential Election Year cycle (it's a mid-term Congressional election year) and the five- to six-month weak portion of the annual cycle. Over the next 4.5 years, this target area can be reached with a combined path of the stock market declining 50%, then rallying 50% and then declining back to test its Supercycle Period low, along with GAAP EPS eventually reaching its previous Q2 '07 all-time high of $84.92 and then declining back to this year-end's estimated level of $62.09. The combination of both of these paths can vary by 20 percentage points and still end within the 8 to 10 P/E target area, so it's still a reasonable working model that we will benchmark as the price and various EPS data comes in over time.
The previous two charts show the fundamental overvaluation of the stock market in relation to GAAP EPS. The following chart shows the market's overvaluation in comparison to the nation's economic activity as measured by GDP, which is largely independent of corporate EPS. Again, the stock market will have to decline substantially to bring this important valuation indicator, the ratio of the stock market's price to GDP — Warren Buffett's favorite valuation metric — to the extremely low level that indicates an end to the current Supercycle Bear Market Period.
Meanwhile, the tremendous overvaluation currently priced into stocks has not been overlooked by corporate insiders, who continue to massively sell shares of their companies' stock, as they have done continuously for months. While net selling-to-buying by corporate insiders typically runs as high as about 10:1 on a dollar basis, it has been running at far higher levels for months. Recently, for example, the dollar-based selling by insiders was an incredible 194:1 in companies in the technology sector and 201:1 in the consumer-services sector, both of which sectors would be key to any self-sustaining economic recovery, in anticipation of which insiders would be buying. That they have been on balance unloading their shares tells a different story.
These typically multi-millionaire corporate officers, directors, and major shareholders, who are in the best position to have an inside track both on the future profits outlook for their companies and on the absolute and relative attractiveness of their companies' stock prices, are unmistakably confirming in the aggregate, but through their independent, bottom-up decision-making, that the stock market is markedly overvalued and headed for a major decline.
[1] Supercycles (more formally, Bronson Asset Allocation Cycle, or BAAC, Supercycles) in the stock market are made up of a Supercycle Bull Market Period and a Supercycle Bear Market Period. Such Supercycle Periods are alternating secular periods (12 to 20 years, averaging about 16 years) of over- and under-performance in total return relative to both price inflation and risk-free (Treasury bills and money market mutual fund) returns, especially when downside-volatility-risk is taken into account. During Supercycle Bear Market Periods (the five pinkareas in the chart with the ABC down-up-down, zigzag stock market patterns), both economic recessions (usually three to four) and the bear markets in equities that anticipate them are twice as frequent and twice as severe (magnitude times duration) as Supercycle Bull Market Periods (the five green areas). One hundred forty years of U.S. stock market history shows that a Supercycle Period is exactly the minimum time horizon for which optimally computed, long-term stock market P/Es have maximum predictability for the stock market, which is a maximum R-squared of about 50%. Our Stock Market and Economic Cycle Timing model, or SMECT, puts economic and stock bear market cycles in historical perspective, and is available on request. Supercycles in various asset classes (e.g., equities, bonds, real estate, commodities, and currencies) and investment styles (e.g., growth vs. value, large vs. small cap, domestic vs. foreign) are all interrelated, as we explain in our BAAC Supercycle report and other related material, which are also available on request.
© 2010 Bob Bronson
Contact Information
Bob Bronson | Bronson Capital Markets Research | Email
