By LARRY SWEDROE
One of the great debates in finance has been whether the source of the value premium is risk or behavioural errors leading to mispricing—or perhaps it is some of both, with limits to arbitrage (especially in the more-expensive-to-trade small stocks) preventing sophisticated arbitrageurs from fully correcting those errors. The study “Expectations and the Cross-Section of Stock Returns” by Rafael La Porta, published in the December 1996 issue of The Journal of Finance, demonstrated that returns on stocks with the most optimistic analysts’ long‐term earnings growth forecasts are lower than those on stocks with the most pessimistic forecasts. La Porta concluded that “investment strategies that seek to exploit errors in analysts’ forecasts earn superior returns because expectations about future growth in earnings are too extreme.”
La Porta found that “the one-year post-formation raw return for stocks with low expected growth rates is 20 percent higher on average than the return for stocks with high expected growth rates. Furthermore, in the year following formation, analysts revise their expectations sharply for both high and low expected growth stocks in the direction and magnitude predicted by the errors-in-expectations hypothesis. In addition, the behaviour of excess returns around quarterly earnings announcement dates strongly supports the errors-in-expectations hypothesis.” He interpreted his findings “as evidence that analysts, as well as investors who follow them or think like them, extrapolate and make systematic errors of excessive optimism for stocks with rapidly growing earnings, and conversely for stocks with deteriorating earnings.”
Pedro Bordalo, Nicola Gennaioli, La Porta and Andrei Shleifer, authors of the study “Diagnostic Expectations and Stock Returns,” published in the December 2019 issue of the Journal of Finance, sought to determine if La Porta’s finding still held true, or if markets had become more efficient, eliminating the mispricing errors. In December of each year between 1981 and 2014, they formed decile portfolios based on ranked analysts’ expected growth in earnings per share (LTG) and reported the average one-year return over the subsequent calendar year for equally weighted portfolios. The LLTG portfolio is the 10 percent of stocks with the most pessimistic forecasts (low growth), and the HLTG portfolio is the 10 percent of stocks with the most optimistic forecasts (high growth). Their findings were consistent with La Porta’s, well after publication of the original research:
- HLTG stocks exhibit fast past earnings growth, which slows down going forward, while earnings of LLTG firms recover during the post-formation period.
- The forecasted growth rate in earnings per share ranges from 4 percent for the lowest LTG decile (LLTG) to 38 percent for the highest decile (HLTG), an enormous difference.
- Forecasts of future earnings growth of HLTG stocks are excessively optimistic (as if analysts overreact to good news) and are systematically revised downward later, while those of LLTG stocks are revised moderately upward.
- Three years after portfolio formation, earnings of HLTG firms are still expected to grow faster than those of LLTG firms, but the spread in expected growth rates of earnings has narrowed considerably.
- The overestimation of earnings growth for HLTG firms is economically large. By year four, actual earnings are a small fraction of what analysts forecast: Earnings per share grow from 0.16 upon formation to 0.21 compared to the prediction of 0.70 based on LTG at formation.
- HLTG stocks consistently disappoint analysts and investors in the post-formation period. The converse holds for LLTG stocks, but in a much milder form.
- Industries with stronger measured overreaction exhibit a larger return differential between LLTG and HLTG firms.
- Inflated expectations mean revert even without bad news.
- LLTG stocks have lower operating margins to asset ratios than HLTG stocks but higher return on equity (4 percent vs. -6 percent). In fact, 36 percent of HLTG firms have negative earnings per share (eps) versus only 16 percent of LLTG stocks. The high incidence of negative eps companies in the HLTG portfolio underscores the importance of the definition of LTG in terms of annual earnings growth over a full business cycle. Current negative earnings do not hinder these firms’ future prospects.
- HLTG stocks exhibit good past returns, but their average returns going forward are low. The reverse is true for LLTG stocks, though the results are less extreme.
- For equal-weighted portfolios, the LLTG portfolio (decile 1) earned a geometric return of 15 percent in the year after formation, while the HLTG portfolio (decile 10) earned just 3 percent. Deciles 2 through 6 earned between 13 and 14 percent annualised, while deciles 7, 8 and 9 earned 12 percent, 10 percent and 7 percent, respectively.
- Deepening the puzzle, the HLTG portfolio has higher market beta (1.51) than the LLTG portfolio (.78) and performs much worse in market downturns. This finding is consistent with the research showing the highest beta stocks have the worst returns.
Bordalo, Gennaioli, La Porta and Shleifer concluded: “Over the past 35 years, betting against extreme analyst optimism has been on average a good idea.” Their interpretation is that “beliefs are forward looking just as with rational expectations, but distorted by representativeness, which biases the interpretation of the news. Specifically, analysts update excessively in the direction of states of the world whose objective likelihood rises the most in light of the news.” This leads to extrapolation and overreaction to news. For example, they explained, “After exceptionally high earnings growth, analysts think a stock is a Google, but they imagine too many Googles relative to reality.”
Bordalo, Gennaioli, La Porta and Shleifer explain that the key property of diagnostic expectations is what is referred to as “the kernel of truth”: Distortions in beliefs exaggerate true patterns in the data. For example, they wrote, “Googles are overweighed in the HLTG portfolio because they occur much more often there than elsewhere.”
An interesting finding was that the HLTG group has a fatter right tail of strong future performers than do non-HLTG firms. The result, they surmised, is that “these exceptional performers are thus representative of the HLTG group, even though they are unlikely in absolute terms”—there are just not enough of them to justify their valuations. They added: “Because strong past eps growth is diagnostic of future strong growth, analysts become excessively optimistic about HLTG firms, driving up prices and generating negative forecast errors. Returns are low post formation as analysts correct their inflated forecasts.”
Bordalo, Gennaioli, La Porta and Shleifer showed empirically that securities whose long-term earnings growth analysts are most optimistic about earn low returns going forward. They then provided a theory of belief formation that delivers this finding. The central feature of their theory is that investors are forward looking, in the sense that they react to news. However, their reaction is distorted by representativeness, the fundamental psychological principle that people put too much probability weight on states of the world that the news they receive is most favourable to. (This is known as the kernel of truth hypothesis: People react to information in the right direction, but too strongly). They called such belief formation “diagnostic expectations” and showed that a theory of security prices based on this model of beliefs can explain not just previously documented return anomalies but also the joint evolution of fundamentals, expectations and returns.
Over the past few years, we have seen value dramatically underperform growth. Perhaps this might lead you to conclude that the authors’ findings no longer hold. Before you draw that conclusion, consider the following.
Déjà vu all over again
There are similarities between the most recent decade and the five-year period ending 1999. (note that the value factor experienced its largest premium in the eight succeeding years, 2000-07). Over the long term, companies with negative cash flows have produced negative returns. Recall the finding of Bordalo, Gennaioli, La Porta and Shleifer that 36 percent of HLTG firms have negative earnings per share (eps) versus only 16 percent of LLTG stocks. Data from AQR Capital Management covering the 24-year period 1995-2018 shows that there were just five years when companies with negative cash flows produced positive returns, and in only two of those were the returns above a few percent. Those two years were 1999 and 2018. In 1999, companies with negative cash flows returned 19 percent, and in 2018 they returned 6 percent. Similarly, companies with negative earnings produced negative returns in all but seven of those years.
In 1999, companies with negative earnings returned 27 percent, and in 2018 they returned 8 percent. In both cases, these abnormal/unexpected outcomes contributed to the negative value premium. However, over the long term, companies with these traits do not make for higher-returning investments. In fact, the AQR data shows that over the 24-year period ending 2018, they have produced negative returns—about -3 percent for companies with negative earnings and about -5 percent for companies with negative cash flow. The performance of companies with negative earnings and cash flow has contributed to the recent underperformance of value stocks. However, just as trees don’t grow to the sky, the historical evidence shows that such anomalies don’t persist. Bubbles eventually burst. We have seen evidence of the anomaly coming to an end with the recent performance of such stocks as Uber Technologies (the stock is now trading at about 31, down from its 52-week high of about 47, a drop of about 34 percent) and Slack Technologies (the stock is now trading at about 22, down from its 52-week high of about 42, a drop of about 48 percent).
Second, in both periods we experienced a bubble in private equity. In most periods we see private companies trading at a discount of about 25 to 30 percent relative to public companies. That discount reflects a liquidity premium. Recently, we have seen several companies that have gone public, or tried to, experience substantial markdowns from their private equity valuations. The most recent example is WeWork. A $2 billion investment from SoftBank in January 2019 valued the co-working company at $47 billion. The company’s attempted initial public offering (IPO) failed, even at a much lower valuation. Private equity firm SoftBank announced that it would accelerate a $1.5 billion investment into the company it had committed to making next year. In addition, it would buy up to $3 billion in shares from other investors in WeWork. In return, it would own 80 percent of the company. Note that the historical evidence is that public companies with high investment and poor profitability have had, on average, very poor returns.
Third, Robert Arnott, Campbell Harvey, Vitali Kalesnik and Juhani Linnainmaa examined the cause of value’s recent underperformance in their December 2019 study, “Reports of Value’s Death May Be Greatly Exaggerated.” They found that more than 100 percent of the underperformance of value since 2007 is due to it becoming considerably cheaper relative to growth. Value cheapened by more than 4 percent per year, creating a performance shortfall of just over 3 percent per year. Note that in 2007, the valuation spread was narrow (22nd percentile); by July 2019, it had widened to the 97th percentile. They also found that there is little difference in overall profitability of growth and value over the two periods. The average return on equity difference between value and growth is almost the same before and after 2007: -11 percent versus -12 percent. This evidence also helps show that the argument that the value premium has disappeared due to overcrowding is demonstrably false. If there was overcrowding, the spread would have narrowed, not widened. Internationally and in emerging markets, the valuation spread is at similarly wide historical levels.