The Value Premium

The value premium is the historical tendency for stocks that are cheap relative to a measure of fundamental value to earn higher returns than stocks that are expensive by the same measure. It is the most studied pattern in empirical asset pricing, the subject of the paper that made Fama and French's reputation in this area, and the source of the longest-running argument in quantitative investing: whether the premium is compensation for bearing real risk, a persistent error in how investors price companies, or an artefact that has already been arbitraged away.

Why book-to-market became the standard measure

Cheapness has to be measured against something. The earliest work used earnings: Basu (1977) showed that stocks with low price-to-earnings ratios subsequently outperformed high P/E stocks, a result that was awkward for the efficient markets view then dominant. Rosenberg, Reid and Lanstein (1985) found the same pattern using book value, and Fama and French (1992) established book-to-market as the measure that best captured the effect in the cross-section of US stocks.

Book-to-market won out for practical rather than theoretical reasons. Book equity is reported consistently, it is available for essentially every listed company, and unlike earnings it is rarely negative and does not swing wildly from year to year. Earnings collapse in recessions, which makes P/E-based sorts unstable at exactly the moments when the value signal matters most. That practical advantage is also the source of the measure's most serious modern criticism, discussed below.

The HML factor formalises the sort. Fama and French rank stocks by book-to-market, form portfolios of the cheapest and most expensive groups within size buckets, and take the return difference. HML stands for High Minus Low: long high book-to-market (value), short low book-to-market (growth). Because it is a long-short construction, HML measures the spread between the two groups rather than the return of value stocks on their own, which is why the factor can be negative in a year when value stocks rose.

The risk explanation

Fama and French have consistently argued that the value premium is compensation for risk. In this account, companies trade at low book-to-market ratios because they are genuinely more fragile: they tend to carry more financial leverage, have more of their value tied up in physical assets that cannot be redeployed cheaply, and show earnings that are more sensitive to the business cycle. Investors demand a higher expected return to hold them, and that higher expected return is what shows up as the premium.

Zhang (2005) gave this a formal mechanism through costly reversibility. Value firms are typically asset-heavy and find it expensive to shrink when demand falls, so they are stuck with unproductive capacity in downturns. Growth firms hold more of their value in options to expand, which they can simply decline to exercise. The asymmetry means value stocks suffer disproportionately in bad times, which is exactly when investors least want losses, and a rational market prices that in.

The behavioural explanation

Lakonishok, Shleifer and Vishny (1994) argued the opposite: that value stocks are not riskier, merely unloved. Investors extrapolate recent performance too far into the future, becoming excessively pessimistic about companies that have disappointed and excessively optimistic about those that have grown. Prices overshoot in both directions, and the premium is the correction as results turn out closer to average than expected.

The evidence they marshalled was that value portfolios did not appear riskier on conventional measures, and that value stocks tended to hold up relatively well rather than badly in several severe market declines. Later work added that analyst forecasts are systematically too optimistic for glamour stocks, which is what the over-extrapolation story predicts.

Both explanations have survived thirty years of testing, which is itself informative. The debate has not been settled because the two accounts are difficult to separate using return data alone. Any pattern that looks like mispricing to one economist looks like an unmeasured risk factor to another.

The drawdown that changed the conversation

From roughly the global financial crisis onward, value strategies as measured by book-to-market went through the deepest and longest underperformance in the recorded history of the factor. This was not a quiet period of mild lag. It lasted long enough that a generation of investors entered the profession having never seen the premium work, and it prompted serious researchers to ask publicly whether the factor was broken.

Three responses emerged, and they are not mutually exclusive. The first is that this is simply what risk feels like: a premium that never disappoints would not be compensation for anything, and drawdowns of this length are within what the historical distribution admits. The second is that publication and the growth of systematic investing compressed the premium, as would be expected if it were partly mispricing. The third is the most specific and, to many, the most persuasive.

The intangibles problem

Book value is an accounting construct, and accounting rules require most spending on research, development, software, and brand building to be expensed immediately rather than capitalised. When Fama and French settled on book-to-market in the early 1990s, the largest companies in the index were manufacturers whose value genuinely sat in factories and inventory that appeared on the balance sheet. Today a far larger share of corporate value sits in intangible assets that accounting rules keep off it.

The consequence is mechanical. A company that spends heavily on research reports lower book equity precisely because it is investing, so it is classified as a growth stock by construction, regardless of whether it is expensive. If book value has become a progressively worse proxy for fundamental value, then the measured decline in the value premium may partly reflect a broken measuring instrument rather than a vanished phenomenon. Researchers testing this have generally found that capitalising intangibles improves the value signal, though it does not erase the drawdown.

Reading the current numbers on this site

This dashboard tracks several value spread definitions side by side for exactly this reason. HML from the three- and five-factor models sorts on book-to-market, while the E/P, CF/P and D/P spreads sort on earnings, cash flow and dividends. When these measures disagree, the disagreement is usually informative about whether an apparent value signal is an artefact of one particular accounting definition.

A caution about interpretation. These series are monthly factor returns, not valuation levels. A percentile reading near the bottom of the historical range means the recent return spread has been unusually poor, not that value stocks are unusually cheap. Those two statements are often correlated, they are not the same, and conflating them is the most frequent error made with this kind of data. The historical time series on the dashboard shows how persistent the current reading has been, which is more informative than any single month.

None of this is investment advice, and a low reading is not a recommendation to act on it. The factor timing article in this section explains why that inference is far weaker than it appears.

Fama, E. F., & French, K. R. (1992). "The Cross-Section of Expected Stock Returns." Journal of Finance, 47(2), 427-465.

Lakonishok, J., Shleifer, A., & Vishny, R. W. (1994). "Contrarian Investment, Extrapolation, and Risk." Journal of Finance, 49(5), 1541-1578.

Zhang, L. (2005). "The Value Premium." Journal of Finance, 60(1), 67-103.