The Value Factor

The value factor is the oldest and most philosophically contested premium in finance: buy stocks that are cheap relative to fundamentals, short those that are expensive, and, historically, earn a positive return. It is the empirical heart of the The Fama-French Factor Models models (the HML factor), the intellectual descendant of Graham and Dodd, and the subject of the most consequential live debate in factor investing after its brutal 2007–2020 drawdown.

The signal

Value ranks stocks by a valuation ratio, price scaled by a fundamental anchor. The Fama-French workhorse is book-to-market, $\text{B/M} = \text{book equity}/\text{market equity}$; high B/M means cheap. Other common measures are earnings-to-price (E/P), cash-flow-to-price, sales-to-enterprise-value, and dividend yield. The point of dividing by a fundamental is to make firms comparable: a $200 stock is not "expensive," a stock priced at 8× its book value is. The cross-sectional value signal for stock$i$ is a standardized ratio, e.g.

$v_i = \frac{\text{B/M}_i - \overline{\text{B/M}}}{\sigma(\text{B/M})},$

and the factor is long high-$v$, short low-$v$ (see Signal Construction).

Why value earns a premium: two stories

The value premium's existence is well documented across markets and centuries; its cause is the flagship instance of the Market Efficiency (The EMH) joint-hypothesis problem.

The risk story (Fama-French). Value firms are cheap because they are distressed, high leverage, declining earnings, poor prospects. Their cash flows are procyclical and their fortunes fall hardest exactly in recessions when investors' marginal utility is highest. So the value premium is rational compensation for bearing distress/recession risk you cannot diversify away. In this view the premium is a permanent feature of equilibrium and should survive being widely known.

The behavioral story (Lakonishok-Shleifer-Vishny). Investors overextrapolate: they get too excited about glamorous growth stocks with great recent performance and too pessimistic about beaten-down value names. Prices overshoot, and value earns its premium as those expectations mean-revert to reality, a form of cross-sectional Mean Reversion. LSV's evidence: value's outperformance is not concentrated in bad states (undercutting the risk story) and value firms' realized fundamentals are far better than their prices implied.

Both stories predict a positive average return; they diverge on durability. The risk story says value persists forever; the behavioral story says it persists only until arbitrageurs and cheap indexing compete it away. The 2010s were, in effect, a live test.

The link to expected returns

There is a clean accounting reason valuation predicts returns. The Gordon/dividend-discount identity says price equals discounted expected cash flows:

$P = \frac{\mathbb{E}[D]}{r - g}.$

Holding expected cash flows fixed, a low price mechanically implies a high discount rate $r$, i.e., a high expected return. Rearranging, book-to-market is a noisy proxy for the market's required return: cheap stocks are cheap because the market demands more return to hold them. This is why Fama and French can call value a "risk" factor without specifying the risk, the valuation ratio is the expected-return signal, whatever its ultimate source. It is also why value and The Quality Factor are complementary: value buys high discount rates, quality ensures the low price is not simply justified by deteriorating fundamentals (a "value trap").

Worked example: HML construction and a tilt

Recall HML is long the cheap third, short the expensive third, size-balanced (see The Fama-French Factor Models). Suppose over a sample HML earns $6\%$/year with volatility $12\%$, a Sharpe of $0.5$. A portfolio with value loading $h = 0.5$ inherits roughly $0.5 \times 6\% = 3\%$/year of value premium, before costs. The catch is that value has negative correlation with Momentum (cheap stocks are usually recent losers), which is why the two are almost always run together: a 50/50 value-momentum book historically had a higher Sharpe than either alone because their drawdowns rarely coincide. "Value and momentum everywhere" is the canonical demonstration that this diversification holds across equities, bonds, currencies, and commodities.

Value's lost decade

From roughly 2007 to 2020, value suffered its worst and longest drawdown on record, HML was flat-to-negative for over a decade while expensive growth stocks (mega-cap tech) dominated. This was an existential stress test, and the post-mortem taught several lessons:

• The premium can go missing for a decade. A 0.5-Sharpe factor has enormous variance; ten-year windows with negative returns are statistically unsurprising, which is exactly why factor timing and conviction are so hard.
• Cheapness of the value trade itself. By 2020 the valuation spread (how cheap value was versus growth) had blown out to dot-com extremes, an argument that the premium was intact and merely un-realized, which the sharp 2020–2022 value rebound partly vindicated.
• Intangibles and accounting. Book value poorly captures R&D and brand for modern firms, so mechanical B/M increasingly misclassified good growth companies as "expensive." Many practitioners now adjust book equity for intangibles or blend multiple value measures.

Failure modes

• Value traps. A stock can be cheap because it is permanently impaired; value works on average but is riddled with individual disasters, hence combining with quality/profitability screens.
• Crowding and compression. When value is popular the spread compresses and forward returns fall; the premium is state-dependent on how cheap value already is.
• Definition and accounting drift. B/M's meaning has degraded as balance sheets shifted from tangible to intangible capital.
• Sector and macro bets in disguise. Naive value loads heavily on financials and energy and is short duration; unhedged, it is partly a rates/sector bet, not pure "cheapness."

In interviews

Define value precisely (a valuation ratio like book-to-market, long cheap / short expensive) and give both explanations for the premium, distress risk (Fama-French, permanent) versus overextrapolation (LSV, decays), noting the joint-hypothesis problem means data cannot fully adjudicate. Strong candidates connect value to the discount-rate identity (low price ⇒ high required return) and volunteer the value-momentum diversification. Expect a question on the lost decade: "Is value dead?", the honest answer is that a 0.5-Sharpe factor routinely has decade-long droughts, the 2018–2020 valuation spread reached extremes, and intangibles distort book value, so the premium is better viewed as impaired-in-measurement and cyclical than dead. See The Quality Factor for the natural complement that screens out value traps.