How “consensus expectations” harm your financial health

Overview

It’s “an article of faith among Wall Street research departments,” wrote David Dreman in Contrarian Investment Strategy (1979): “nothing is as important in the practice of security analysis as estimating the earnings outlook.” For market indexes as well as companies, “forecasting (earnings 12 months hence) is the heart of most security analysis as it is practiced today.” Almost 50 years later, in Australia, the U.S. and elsewhere, Dreman’s assessment remains apt.

That’s deeply unfortunate. The foundations of the mainstream’s approach – “consensus earnings estimates” and “forward earnings multiples” – have long been unsound. Specifically, they contain three fatal flaws.

Firstly, the strength and direction of the influence of stocks’ earnings upon their returns are so erratic that, in the short term (that is, periods of 12 months), they average zero. In Do earnings drive stocks’ returns? (22 January 2024), I demonstrated this crucial point with American data; in Everything the Mainstream Says about Earnings Is Wrong (13 March 2024), I corroborated it with Australian data.

Secondly, analysts don’t know – or if they do, decline to admit publicly – that in effect they’re striving in vain to predict the unpredictable. In Stop kidding yourself: Nobody can “time the market” (30 June) I demonstrated that stocks’ short-term (rolling 12-month) total returns are almost perfectly randomly distributed – and thus almost always unforeseeable.

In this article, I demonstrate that the same point applies to consensus forecasts of stocks’ earnings. These forecasts are, in effect, random guesses which masquerade as prophesies of the unforeseeable!

The insuperable problem is that analysts are human beings, and human beings are innately poor – yet overconfident – forecasters. They’re usuallt wrong, but they’re seldom in doubt. That’s particularly so when they’ve been fooled by randomness; that is, fail to comprehend that they’re trying vainly to foresee random fluctuations. Everybody occasionally gets lucky, but nobody can consistently anticipate random changes such as companies’ short-term earnings, stocks’ prices, markets’ returns, etc.

A third fatal flaw, which this article also demonstrates, undermines the mainstream’s credibility: “analysts” are actually – I assume unwittingly – shills and propagandists. Their assessment of the near future is heavily biased: in particular, it’s overly optimistic.

As a result, and as I’ll show, predictions of stocks’ earnings almost invariably – and usually greatly – exaggerate actual earnings. Similarly, “forward earnings multiples” persistently underestimate the risk of overvaluation. On that flawed basis, “fair value” is what the overconfident consensus deems rather than what a conservative analysis concludes.

That’s why bear markets, panics and crises – including the GFC – almost inevitably surprise and shock “analysts” and other “experts.” “Nobody could have foreseen this!” is their bewildered response to such events – which indignantly disclaims any responsibility for their occurrence.

Analysts concoct various after-the-fact explanations in the wakes of bear markets, etc. Their own exuberant myopia, which helped to inflate expectations and thus dulled prudence, seemingly NEVER figures among their rationalisations. Hence analysts remain free to wreak further damage and cause more losses.

In light of the mainstream’s glaring failings, what can investors do? Leithner & Co’s approach is three-fold. Firstly, we’re acutely aware of analysts’ systematic tendency to “mispredict” (see, for example Stop Obsessing about the RBA, 14 February 2025).

Accordingly, and except in order to track their biases and errors, particularly when they become egregious, we ignore the consensus and its estimates of forward earnings, PEs and the like. Instead, we take our own counsel, conduct our own analyses and make our own decisions.

Secondly, our analyses disregard short-term earnings and emphasise medium-term dividends (see in particular Dividends Aren’t a Bane – They’re a Boon (20 November 2023). More generally, and unlike the mainstream, we analyse financial statements – and our analysis downplays P&Ls and emphasises balance sheets and cashflow statements. In sharp contrast, mainstream analysts are obsessing about what’s relatively insignificant and ignoring what’s actually important.

As Robert Shiller concluded (Irrational Exuberance (3^rd ed., 2015), “the reliable return attributable to dividends, not the less predictable portion arising from (earnings and) capital gains, is the main reason why stocks have on average been such good investments historically.”

Thirdly, our conclusion – namely that essentially everything the mainstream asserts about earnings is false – directs our focus to justifiable measures of valuation such as cyclically-adjusted PE ratios. CAPE is a long-term valuation tool, and not a short-term market-timing tool; nonetheless, our orientation has on several occasions enabled us to anticipate and take advantage of downturns, bear markets and the like. Unfortunately, however, those who heed the consensus, its earnings estimates and forward PEs have – like the supposed experts who comprise it – repeatedly been caught unawares.

As a result, in the past they’ve endured – and in the future will likely continue to suffer – significant losses.

Actual (“Trailing”) versus Prospective (“Forward”) Earnings

It’s vital to distinguish a company’s or index’s actual (“trailing”) earnings over a past period from its prospective (“forward”) earnings for a future period. Trailing earnings during a given month are (1) actual earnings over the previous 12 months. As such, they’re (2) consistently and (3) strictly defined. In short, they conform to Generally Accepted Accounting Principles. Colloquially, GAAP earnings include “all the bad stuff.”

Actual earnings are usually of little interest to analysts. Instead, “consensus forward earnings” and the correspondence or otherwise of expected to actual earning obsess them. A “consensus forward earnings” estimate during a given month is the average of analysts’ estimates of a company’s or market index’s earnings during the next 12 months. This consensus doesn’t attempt to foresee GAAP earnings: forward estimates routinely incorporate “good stuff” (such as one-off gains) and exclude “bad stuff” (such as allegedly temporary losses, write-downs and write-offs of assets, etc.).

Forward estimates, in short, are inconsistent over time; they’re also transient (that is, the estimate for a particular point in time is revised repeatedly as time passes) and subjective at all times.

Which Consensus Estimate Counts?

What does it mean to beat, hit or miss the consensus estimate of earnings? The uninitiated might think that it’s a straightforward matter, but it isn’t; Koller et al. (“Avoiding the Consensus-Earnings Trap,” mckinseyquarterly.com, January 2013) clarify it.

Suppose that on 15 February 2024 X Ltd reported its actual (trailing) earnings of $2.00 per share for the year ended 31 December 2023. Also on 15 February 2024, the consensus – that is, the average of the brokers covering the stock – forward estimate of earnings for the year to 31 December 2024 was $2.10. By 15 February of this year – well after the conclusion of the period in question, taking into consideration X’s half-year results but before the release of its full-year earnings – the consensus forward estimate for 31 December 2024 had fallen to $1.98. Finally, suppose that on 15 February X reports actual earnings of $2.00 per share for the year ended 31 December 2024.

Did X Ltd’s actual earnings for CY24 exceed the consensus forward estimate? It’s vital to understand: analysts constantly revise their estimates of forward earnings for a given interval; and according to conventional practice, a company (or market) has beaten the consensus estimate if its actual earnings are greater than the final consensus estimate – which almost always appears AFTER the period in question has ended!

Accordingly, in this example X has beaten the consensus estimate of forward earnings, which fell 5.7% during the year, even though its trailing earnings were 4.8% less than the consensus estimate at the beginning of the year!

Results

Preliminaries

In Stop kidding yourself: Nobody can “time the market” (30 June) I demonstrated that stocks’ and market indexes’ CPI-adjusted, short-term (rolling 12-month) total – that is, including dividends – returns are almost perfectly randomly distributed; accordingly, they’re almost always unforeseeable. Except under rare circumstances, that is, when returns are extremely low or high, market-timers can’t consistently anticipate them.

That’s because these fluctuations mostly reflect mere chance. Random movements are irregular variations around a mean. They have no identifiable cause and are trendless; hence they’re unpredictable. A successful speculation is merely a lucky guess; as such, these “successes” are ephemeral.

Market-timing fails because it ignores a fundamental, obvious and insuperable difficulty: stocks’ and markets’ returns are random; by definition, random phenomena are unforeseeable – and nobody can predict what’s inherently unpredictable.

I’ve computed the S&P 500’s CPI-adjusted total 12-month returns (that is, including dividends) for January-1979-January 1980, February 1979-February 1980, ... and June 2024-June 2025 (the reason that I’ve commenced the series in January 1980 will shortly become clear). The closer these returns’ resemble a normal (that is, bell-shaped) distribution, the more random they are – and the more unpredictable their future 12-month returns will be.

A Quantile-Quantile (“QQ”) plot provides one means to assess the extent to which a variable follows a normal distribution. (A quantile is a value below which a given percentage of data falls.) A QQ plot compares the quantiles of the S&P 500’s 12-month returns to those of a normal distribution whose mean and standard deviation equal the returns’. The more normally the returns are distributed – that is, the greater the extent to which they form a normal (bell-shaped) distribution – the closer the points on the QQ plot will fall along a straight line (whose maximum measure of best-fit (R²) is 1.0).

Figure 1 shows that the S&P 500’s CPI-adjusted, 12-month total returns since January 1980 have been almost perfectly normally distributed (R² = 0.98). The distribution’s lower “tail,” however, is truncated: the Index has sustained severe losses less frequently than if short-term returns were completely random. Yet the implication is clear – and crucial: analysts, via their “consensus forward earnings,” are mostly attempting to foresee the unforeseeable.

Figure 1: Quantile-Quantile Plot, CPI-Adjusted Total 12-Month Return, S&P 500 Index, January 1980-June 2025

From a mainstream point of view, consensus forward earnings estimates don’t consistently predict stock returns; they do, however, influence them. In particular, changes of these estimates over time are a widely-accepted gauge of a company’s or market’s prospective earnings. Accordingly, these estimates guide many investors and speculators, and the financial press.

My results produuce a very different interpretation: consensus estimates of forward earnings, and percentage changes thereto, are indistinguishable from random guesses. How do “analysts” attempt to predict returns’ random – and thus unpredictable – fluctuations? In effect, if not by intention, by randomly guessing forward earnings. Random garbage in, random garbage out!

Figure 2 is a QQ plot of the CPI-adjusted, 12-month percentage change of the Index’s consensus forward earnings since January 1980. In that month, the estimate was $15.85; in January 1981, it was $16.84; hence its percentage change over these 12 months was ($16.84 - $15.85) ÷ $15.85 = 6.2%, and so on for the subsequent months to June 2025.

Figure 2: Quantile-Quantile Plot, S&P 500 Index’s CPI-Adjusted Consensus Forward Earnings, 12-Month Percentage Change, January 1980-June 2025

Apart from a “skinny” lower tail (that is, fewer sharp 12-month decreases of consensus estimates than would be expected by a perfectly normal distribution) and a “fat” upper tail (a greater number of sharp increases) the distribution is almost perfectly (R² = 0.97) normally distributed.

How Accurate Are Consensus Estimates of Earnings?

It’s a question which investors – never mind “analysts” – never seem to ask. Yet it’s easy to answer. Using Standard & Poor’s collation of analysts’ consensus forward estimates of the S&P 500’s earnings since January 1980, for each month I’ve paired the consensus forward estimate to the Index’s actual earnings 12 months hence.

The consensus estimate in January 1980 (CPI-adjusted $15.84), for example, attempted to predict actual earnings in January 1981 ($14.74). I’ve paired these two earnings and dated them January 1981, and so on for subsequent months to June 2025. Figure 3 plots the results.

Figure 3: CPI-Adjusted Actual and Consensus “Forward” Earnings,S&P 500 Index, January 1981-June 2025

Two are paramount: firstly, consensus forward estimates almost invariably exceed the actual earnings they attempt to predict; secondly, this disparity is particularly marked during recessions (such as the early-1980s, early-1990s and early-2000s) and crises such as the GFC and COVID-19 pandemic.

Figure 4: Changes of the S&P 500’s CPI Adjusted, 12-Month Earnings over Six Intervals

Figure 4 quantifies this latter point. During the “double-dip” recessions of the early-1980s, the Index’s actual earnings fell 24%. Consensus estimates which purportedly predicted these earnings, however, fell just 1%. The disparity during the GFC was greatest: actual earnings collapsed an astounding 92%, but consensus estimates fell just 6%. Most recently, actual earnings in 2022-2023 slumped 18% – but the consensus estimates which attempted to predict them soared 32%.

Crucially, consensus forward estimates don’t merely fail to anticipate recessions, panics, crises, etc.: they remain oblivious to them even after they’ve commenced!

By How Much Do Consensus Forward Estimates Exceed Actual Earnings?

Figure 5 plots the deviation, in percentage terms, of consensus forward earnings from actual earnings. (In order to render it more readable, I’ve restricted its vertical axis to a maximum value of 100%; that’s a small fraction of the consensus’ overestimate of actual earnings during the Dot Com Bust and GFC).

Figure 5: Percentage Variation, CPI-Adjusted Forward versus Actual Earnings,S&P 500 Index, January 1981-June 2025

The CPI-adjusted consensus forward estimate for January 1981, for example, was $15.84; actual earnings in that month, which the consensus estimate was trying to foresee, were $14.74; hence the percentage variation of consensus forward to actual earnings was ($15.84 - $14.74) ÷ $14.74 = 6.3%. The more the consensus overestimates actual earnings, the further above 0% the deviation rises; the more the consensus underestimates actual earnings, the further below 0% the variation falls.

Only occasionally – most notably during the immediate aftermath of the COVID-19 panic – have consensus forward estimates underestimated actual earnings. Including the Dot Com Bubble and GFC, the estimate has overestimated actual earnings by an average of 47%; excluding them, the forward estimate has exceeded actual earnings by an average of 32%.

Long-Term Average Earnings versus the Most Recent – and Next Year’s – Earnings

In Chapter 12 of The Intelligent Investor, Benjamin Graham urged investors (as opposed to speculators): “don’t take a single year’s earnings seriously.” “In former times,” he elaborated, they “paid considerable attention to the average earnings over a fairly long period in the past – usually from seven to ten years. This ‘mean figure’ was useful for ironing out the frequent ups and downs of the business cycle, and it was thought to give a better idea of the company’s earning power than the results of the latest year alone. One important advantage of such an averaging process is that it will solve the problem of what to do about nearly all the special charges and credits (detailed in that chapter).”

“If such (long-term average) figures are used in conjunction with (measures of) growth and stability of earnings during the same period,” Graham concludes, “they could give a really informing picture of the company’s past performance.”

How, asked Graham, can “the careful investor ... be reasonably sure in advance that he is not committing the typical Wall Street error of over-enthusiasm for good performance in earnings and in the stock market?” In a footnote to Chapter 12, Jason Zweig answered: “Graham insists on calculating the price/earnings ratio based on a multiyear average of past earnings. That way, you lower the odds that you will overestimate a company’s value based on a temporarily high burst of profitability.”

“Imagine,” Zweig continued, “that a company earned $3.00 per share over the past 12 months, but an average of only $0.50 per share over the previous six years. Which number – the sudden $3 or the steady fifty cents – is more likely to represent a sustainable trend? At 25 times the $3 it earned in the most recent year, the stock would be priced at $75. But at 25 times the average earnings of the past seven years ($6 in total earnings, divided by seven, equals $0.857 per share in average annual earnings), the stock would be priced at only $21.43. Which number you pick makes a big difference.”

“Finally,” wrote Zweig, “it’s worth noting that the prevailing method on Wall Street today – basing price/earnings ratios primarily on “next year’s earnings” – would be anathema to Graham. How can you value a company based on earnings it hasn’t even generated yet? That’s like setting house prices based on a rumor that Cinderella will be building her new castle right around the corner.”

How Well Do “Forward PEs” Anticipate Future Returns? What about CAPEs?

Starting with Graham’s insights, in the 1980s Robert Shiller, an academic at Yale University, and his colleague, John Campbell of Harvard, began to devise and refine what’s become known as the cyclically-adjusted PE (CAPE) ratio. They presented their findings to the Federal Reserve in 1996; their article, “Valuation Ratios and the Long-Run Stock Market Outlook: An Update” (NBER Working Paper 8221), followed in 1998.

The standard price-to-earnings (“PE”) ratio has a numerator and a denominator. Its numerator is the price of a company’s shares or a market index’s level at a given point in time; its denominator is the company’s earnings per share, or the market’s earnings, at that time. If, for example, X Ltd’s shares sell for $10 and during the preceding 12 months it earned $1 per share, then its PE ratio is $10 ÷ $1 = 10.

CAPE’s basic structure resembles the standard PE’s. Its numerator is a company’s CPI-adjusted price (or a market’s CPI-adjusted level) at a given point in time; its denominator is the company’s or market’s average of CPI-adjusted earnings over the previous 10 years. As with the standard PE, so too with CAPE and the forward PE: the higher is the ratio, the greater is the price paid per $1 of (CPI-adjusted) earnings – and thus the dearer is the stock or index.

It bears repetition: CAPE is a long-term valuation tool – NOT a short-term market-timing tool. Nonetheless, high ratios typically precede major selloffs.

Figure 6 plots these three measures of the S&P 500 Index’s valuation since January 1980. (In order to maximise its legibility, I’ve restricted the vertical axis to a maximum of 50; during the GFC, the Index’s earnings collapsed and thus the trailing PE skyrocketed into triple figures.)

Figure 6: Three Measures of Valuation, S&P 500 Index, January 1980-June 2025

Not surprisingly, given that “consensus forward earnings” almost invariably exceed the actual earnings they attempt to predict, the “forward PE” ratio’s denominator is larger than the trailing PE’s; as a result, without exception the forward PE is smaller than the trailing PE.

It’s also unsurprising: given that earnings tend to grow over long intervals, trailing earnings at a given point typically exceed their CPI-adjusted average over the preceding ten years. The trailing PE’s denominator is thus usually larger than the CAPE’s; as a result, with some notable exceptions the trailing PE is smaller than the CAPE. Since 1980, therefore, CAPE has typically been the largest – and the most conservative – measure of the Index’s valuation, and the forward PE the smallest and least conservative.

The worst flaw of the “forward” PE ratio is that, since the GFC, it’s lulled investors and speculators into a false sense of confidence. Unlike CAPE, a higher ratio hasn’t generated lower subsequent returns. The “forward” PE, in other words, often emits unduly confident sifnals.

For each month since January 1980, I computed the S&P 500’s CPI-adjusted total return (expressed as a compound annual growth rate (CAGR)) during the preceding five years and during the next five years; sorted these data by each month’s CAPE ratio, divided the dataset into quintiles (that is, five subsets with equal (net of rounding) numbers of observations) and then computed the average preceding and subsequent CAGR each quintile. I then repeated the process with the standard (trailing) PE ratio and the forward PE ratio. Table 1 summarises the results.

Table1: S&P 500 Index, Five-Year Total Returns (CAGRs), by Quintile and Type of PE Ratio, January 1980-June 2020

Three results are paramount. Firstly, reading down the columns, from one quintile to the next (that is, as each ratio rises) trailing CAGRs increase. Since 1980, for example, in the lowest (ranked by CAPE) 20% of observations, the previous five year’s CAGR averages 1.2% per year; in the highest quintile, it averages 14.8% per year. Similarly, as the standard and forward PEs rise, so do corresponding trailing CAGRs.

Secondly, as each ratio rises subsequent CAGRs decrease. Since 1980, in the lowest (ranked by CAPE) 20% of observations, the next five year’s CAGR averages 9.5% per year; in the highest quintile, it averages -0.1%. Similarly, as the standard and forward PEs rise, the corresponding CAGRs during the next five years fall. The higher is the valuation (as measured by CAPE, standard and forward PEs) at a given point in time, the lower is the return over the next five years.

Thirdly, during and since the GFC the efficacy of CAPE and the standard PE as gauges of (over)valuation has weakened – and the forward PE’s has largely disappeared.

Since 2008, as each of three ratio rises, trailing CAGRs continue to increase. And as CAPE and the standard PE rise, subsequent five-year returns continue to decrease – but not as much as they have since 1980 (CAPE’s R² value decreases to 0.35). But as the forward PE rises, subsequent returns no longer fall: they remain roughly constant. Hence the R² has decreased to 0.13.

Figure 7: CPI-Adjusted Prospective Five Year Total Return (CAGR), S&P 500 Index, by Forward PE Ratio, January 1980-June 2020

Figure 7 and Figure 8 detail this crucial development. Blue circles denote observations before the GFC; red ones denote observations since 2008. Since the GFC, forward PEs (Figure 7) have clustered within a relatively tight range (from 10 to 20); so have subsequent CAGRs (5% to 15% per year). As a result, as the forward PE rises, the subsequent CAGR decreases only marginally. CAPEs (Figure 8), however, have varied more since the GFC (from ca. 12.5 to 35); as a result, as CAPE rises the CAGR continues to fall.

Figure 8: CPI-Adjusted Prospective Five Year Total Return (CAGR), S&P 500 Index, by CAPE Ratio, January 1980-June 2020

Crucially, CAPE is a more reliable – and the “forward” PE is a less reliable – measure of valuation and precursor of future medium-term returns.

Have Consensus Forward Estimates Degraded the Quality of Earnings?

What explains the results in Figure 7 and Figure 8? One possibility is that consensus forward earnings don’t just exaggerate the quantum of actual earnings; since the GFC they’ve also overstated their quality – and indeed, helped to worsen them.

Bulls ubiquitously chant the phrase “earnings quality,” but they rarely bother to define it. Yet it has a precise meaning: according to Jim Sepe, et al. (Intermediate Accounting: IFRS Global Edition (McGraw Hill, 7^th ed., 2012), the higher is the quality of a company’s reported net profit after tax, the better they predict its cash flow from operations. Hence cash flow relative to NPAT is an – but not the sole – indicator of earnings quality. The higher is the correlation, the better is the quality. More generally, the higher is the quality of earnings, the more they faithfully reflect the results of the company’s operations.

In short, quality earnings are conservative; that attribute, however, doesn’t necessary make them predictable.

Is the quality of corporate earnings improving or deteriorating? The sensible but vague and messy answer is that it can and does vary significantly between companies and industries; hence nobody can answer “yes” or “no” definitively. On the one hand, regulations clearly require that companies provide detailed and accurate financial information. Equally evidently, companies face even greater pressure from analysts, shareholders and others to boost their earnings.

Consequently, in response to this pressure some companies adopt aggressive accounting practices to meet earnings targets – which can seemingly improve but actually worsen the quality of earnings.

Jason Zweig, in his Commentary to Chapter 12 of The Intelligent Investor (2008), observes that “even Graham would have been startled by the extent to which companies and their accountants pushed the limits of propriety in the past few years. Compensated heavily through stock options, top executives realized that they could become fabulously rich merely by increasing their company’s earnings for just a few years running. Hundreds of companies violated the spirit, if not the letter, of accounting principles—turning their financial reports into gibberish, tarting up ugly results with cosmetic fixes, cloaking expenses, or manufacturing earnings out of thin air.”

Zweig detailed these “unsavoury practices.” “Pro forma earnings,” for example, “enable companies to show how well they might have done if they hadn’t done as badly as they did. As an intelligent investor, the only thing you should do with pro forma earnings is ignore them.”

Yet aren’t “consensus forward earnings” merely a prospective form of “pro forma” earnings”?

One survey of the CFOs of major U.S. public companies found that a substantial portion believe that the manipulation of earnings is commonplace – and that a startling 20% of them admit anonymously that they’ve intentionally manipulated their company’s earnings – even while conforming to GAAP (see Ilia Dichev, et al., “The Misrepresentation of Earnings,” Financial Analysts Journal, vol. 72, no. 1, 2016).

We’re not talking about massive, Enron-style fraud: we’re talking about subtle – and legal – interpretations of accounting standards which “manage” earnings in order to present companies’ results in the best possible light. In the words of The Wall Street Journal (“Earnings Wizardry,” 1 October 2012), “it’s lightly searing the books rather than cooking them.”

By their own admission, CFOs employ various techniques to manipulate earnings. These include tweaking activities (such as accounts payable and receivable, etc.) to meet short-term targets, obfuscating unfavorable information, and above all exploiting the flexibility of accounting standards. The imperative of meeting the company’s earnings guidance and analysts’ expectations, avoiding negative reactions and achieving internal performance targets may incentivise companies to “massage” earnings. Although some acknowledge that the manipulation of earnings exists, it can be difficult to detect, especially if companies are generally adhering to GAAP.

One CFO expressed this point especially well: “if you build expectations, then you have to live by those expectations. That’s an art because you are looking into a crystal ball, whereas closing the books is a science. So you are trying to marry an art and science.”

Robert Howell, an accounting academic at the Tuck School of Business at Dartmouth College, was more explicit. He told WSJ: “the quality of earnings is inversely correlated with whether a company makes earnings estimates.”

Forget Short-Term Earnings: Focus upon Medium-Term Dividends

Thus far, two main conclusions have emerged from my analysis:

1. “consensus forward estimates” aren’t just random; they’re consistently and severely biased. As such, they incentivise companies to impair the quality of earnings. Specifically, analysts consistently and considerably exaggerate eventual and actual earnings – and during bear markets, recessions, panics and crises, the consensus egregiously overestimates actual earnings.
2. as measures of (over)valuation, CAPE ratios have long been superior to standard and “forward” PEs. Moreover, during and since the GFC the forward PE has largely ceased to act as a gauge of overvaluation.

In light of these results, how can a value investor justifiably assess companies and evaluate markets? Omitting a huge number of details, here’s the gist: we weight companies’ balance sheets more heavily than their profit-and loss statements. Specifically, as a value investor, Leithner & Co is a long-term investor; we therefore greatly downplay (indeed, mostly ignore) short-term earnings – and emphases medium-term dividends.

For each month since January 1876, I’ve computed the CAGRs of the S&P 500’s dividends during the preceding and subsequent five years. I’ve then subtracted the latter from the former. How well do past dividends anticipate future dividends. Figure 9 and Figure 10 summarise my results. Dividends’ CAGR over the preceding five years have on average been long been very accurate proxies of their CAGR over the next five years.

Figure 9: Differential, Dividends’ Past and Future Five-Year CAGRs, S&P 500 Index, 1876-2020

The average error since 1876 has been a miniscule 0.1 percentage points. That compares very favourably, to put it mildly, to the error rate of consensus forward versus actual earnings – which since 1980 has averaged 47% (including the Dot Com Bust and GFC) and 32% (excluding them, recall Figure 5).

Errors, of course, occur. Yet over the past 150 years they’ve never exceeded 20%; moreover, since 2000 (Figure 10) the average error has been 0.3 percentage points and the maximum error less than 15%.

Figure 10: Differential, Dividends’ Past and Future Five-Year CAGRs, S&P 500 Index, 2000-2020

Conclusions

“Consensus forward estimates” can’t – and thus don’t – predict short-term earnings. That’s not just because these estimates are random guesses; it’s because short-term earnings, too, fluctuate randomly and are thus also unpredictable. Further, short-term earnings don’t foretell short-term returns – which also fluctuate randomly and are thus unpredictable.

Charitably speaking, consensus earnings estimates are merely random noise; trenchantly but truthfully, much mainstream “analysis” of stocks is nothing more than overconfident and overoptimistic “garbage in, garbage out.”

Robert Shiller agrees. In Irrational Exuberance (Princeton University Press, 3^rd. ed., 2016) he wrote: “unknown to most investors is the troubling lack of credibility in the quality of research being done on the stock market (by “analysts” at major financial institutions) ... Some of this so-called research often seems no more rigorous than the reading of the tea leaves.”

Forward PEs encapsulate the fatal flaws of consensus forward estimates. As a result, they’re less accurate gauges of valuation than CAPE ratios; indeed, since the GFC they’ve become relatively unreliable. 

In their 2017 paper, “The Many Colours of CAPE” (Yale ICF Working Paper No. 2018-22), Shiller and Farouk Jivraj of Imperial College London demonstrated that CAPE is a better measure of valuation any of the alternatives – including those advocated by its critics – they considered.

That, I suspect, is why mainstream analysts and journalists mostly ignore it – and fixate upon and even obsess about consensus forward estimates and forward PEs; it’s also why they dismiss and ignore actual earnings. By overestimating forward earnings, in their minds they minimise the possibility of overvaluation.

No matter how high valuations increase, consensus forward earnings rise even faster – and suppress the forward PE. By this fatally flawed measure, markets are never overvalued!

Why don’t forward PEs “work” as a measure of value? As I’ve demonstrated elsewhere (see, for example Stop Obsessing about the RBA, 14 February 2025), analysts can’t – because nobody can – predict interest rates, stock market’s returns, etc., over the coming 12 months. Analysts are also unable to predict earnings. They’re so unable, in fact, that the CPI-adjusted average of trailing earnings over the past ten years is a much better predictor of returns over the next five years than is the consensus of earnings over the next 12 months.

That’s why CAPE ratios do a better job at selecting value stocks than forward PEs. And although I haven’t detailed the relevant results in this article, it’s also why dividends’ trailing CAGRs identify value stocks even more accurately than CAPEs.

What explains consensus forward earnings’ lack of accuracy? It’s mostly, but more than, the innate human incapacity to see the future: despite this inability, analysts are overconfident and thus overly optimistic. Since the early-1980s, and excluding the Dot Com Bust and GFC – which, as far as consensus forward estimates were concerned, didn’t occur – CPI-adjusted forward earnings have been, on average, approximately one-third higher than subsequently realised earnings.

This excess of optimism isn’t stable over time or across all stocks. Paul Hribar and John McInnis have demonstrated that analysts’ over-optimism – particularly regarding so-called “growth” stocks – rises when the sentiment of market participants becomes buoyant (see “Investor Sentiment and Analysts' Earnings Forecast Errors,” Management Science, vol. 58, No. 2, 2011).

Moreover, although analysts are on average too optimistic, those affiliated with an underwriter of a company’s securities tend to be “strategically overoptimistic” – and thus systematically distort their forecasts (see Ulrike Malmendier and Devin Shanthikumar, “Do Security Analysts Speak in Two Tongues?” The Review of Financial Studies, vol. 27, no. 5, 2014).

Ben Graham knew all this 75 years ago. In The Intelligent Investor he observed: “while enthusiasm may be necessary for great accomplishments elsewhere, on Wall Street it almost invariably leads to disaster.”

Implications

In Contrarian Investment Strategies (2011), Dreman finds that analysts’ ability to predict earnings is so poor that it’s “impossible to distinguish growth stocks ... from average companies ... or even from also-rans.” He therefore asks: if earnings estimates “are not (accurate) enough to weed out the also-rans from the real growth stocks, ... why (would) anyone pay enormous (earnings) premiums” for so-called “growth” stocks? His conclusion is sensible (if understated): the consensus of expected earnings “should be viewed with some suspicion.”

Given the results of my analysis, as well as Dreman’s, investors should ignore companies’ earnings guidance, analysts’ consensus of forward earnings – and the journalists who parrot them. The consensus is biased and overconfident; hence it creates (and abets managements’) unrealistic expectations – and eventually, induces sizable losses (see also How experts’ “systematic mispredictions” improve our returns, 6 August 2024).

Jason Zweig is correct: “history – and a mountain of financial research – (has) shown that the market is unkindest to rapidly growing companies that suddenly (fail to meet earnings expectations). More moderate and stable growers, ... tend to suffer somewhat milder stock declines if they report disappointing earnings. Great expectations lead to great disappointment if they are not met; a failure to meet moderate expectations leads to a much milder reaction.”

“Thus,” Zweig cautions, “one of the biggest risks in owning ‘growth stocks’ is not that their growth will stop, but merely that it will slow down. And in the long run, that is not merely a risk, but a virtual certainty.”

Those who obsess about forward earnings don’t merely err almost continuously and sometimes enormously: they also blunder systematically. The consensus of forward earnings’ level and trend is unable to foresee downdraughts, etc; but the sudden realisation that expectations have enormously outpaced reality can trigger them.

Conservative contrarian investors like Leithner & Co thus mostly (except to monitor their errors, etc.) ignore consensus forward earnings and forward PEs; equally, we bear in mind the excesses of tipsters’ emotions. Analysts are systematically overoptimistic; as a group, they’re therefore dangerously overconfident.

Moreover, in order to promote and protect their own popularity they tip what according to the consensus – that is, forward PEs – are popular stocks. They thereby tend to generate underperformance and losses (see in particular Why you’re probably overconfident – and what you can do about it, 14 February 2022).

It thus bears repetition: you must either ignore or take advantage of the consensus’ bullishness; you’ll harm your financial well-being if you adopt it.

“Mr Market is there to serve you, not to guide you,” wrote Warren Buffett in his letter to Berkshire Hathaway’s shareholders in 1987. “If he (is) ... in a particularly foolish mood, you are free to either ignore him or to take advantage of him, but it will be disastrous if you fall under his influence. Indeed, if you aren’t certain that you understand ... Mr Market, you don’t belong in the game. As they say in poker, ‘If you’ve been in the game 30 minutes and you don’t know who the patsy is, you're the patsy’” (italics in the original).

If you heed the consensus (including its forward estimates and forward PEs), its overconfidence and its tips – like lottery tickets, they tacitly overpromise but eventually under-deliver – you’re speculating rather than investing.

And if you’re speculating, you’re gambling – that is, playing a game whose odds are stacked overwhelmingly against you (see in particular Stock tips are for patsies – are you a patsy? 12 February 2024).

Click LIKE so that Livewire knows that you want more of this type of content. Click FOLLOW for notification when my next wire appears.