Since the early 1990s, a handful of stock characteristics have explained a surprising share of the differences in returns across equities. Cheap stocks have tended to beat expensive ones. Recent winners have tended to keep winning, at least for a while. Profitable firms have outpaced unprofitable ones. Under a strict reading of market efficiency, patterns this simple should not survive once they are widely known.

Yet these patterns, now called factors, have been documented across decades, countries, and asset classes. They anchor a large and growing smart beta industry and form the backbone of modern quantitative investing. The open question is not whether they appear in historical data. It is why they exist, and whether investors can actually capture them.

This article examines the evidence behind value, momentum, quality, and size. It also covers the long stretches when each has disappointed and the practical choices involved in building a factor tilt you can stick with. The answers say as much about investor behavior as about the factors themselves.

The modern factor framework began as a challenge to the Capital Asset Pricing Model. CAPM held that a stock's expected return depended solely on its sensitivity to the overall market. In 1992, Eugene Fama and Kenneth French showed that two other characteristics did a far better job of explaining the cross-section of returns: company size and the ratio of book value to market value. Small caps and cheap stocks earned more than CAPM predicted.

Momentum followed soon after. In 1993, Narasimhan Jegadeesh and Sheridan Titman documented that stocks with strong returns over the prior three to twelve months tended to keep outperforming in the months that followed. Quality entered the canon later. Researchers defined it broadly as profitability, stable earnings, and conservative balance sheets. Robert Novy-Marx's 2013 work on gross profitability and Fama and French's 2015 five-factor model gave it formal standing.

Two explanations compete for why these factors pay. The risk-based view holds that factors compensate investors for bearing genuine risk. Value companies are often distressed. They may perform worst precisely when the economy deteriorates and investors most need their wealth. The behavioral view argues that systematic errors create the premiums. Investors extrapolate past growth too far and overpay for glamour stocks. They also underreact to new information, which leaves room for momentum to build.

The distinction matters. A risk premium should persist, because someone must be paid to bear the risk. A behavioral mispricing may fade as capital arbitrages it away. The evidence suggests both forces are at work, in proportions that differ by factor. Momentum looks largely behavioral, while value carries features of both. Size has weakened notably since Rolf Banz identified it in 1981, and much of its historical premium sits in illiquid microcaps. The most credible factors share one trait: they reappear in markets and periods their discoverers never examined.

Takeaway

Ask why a factor should pay before asking how much it has paid. A premium with a durable economic reason is far more likely to survive a crowded, well-informed market.

If factors deliver premiums over the long run, the obvious next step is to own them only when they are about to work. This is where theory and practice part ways. Factor returns are noisy, and their drawdowns can last long enough to test any conviction.

Value offers the starkest example. Measured by the Fama-French HML factor, value underperformed growth for much of the period from 2007 to 2020, a stretch that included its deepest drawdown on record. Many investors abandoned the strategy. Others concluded that intangible assets had permanently broken book value as a measure. Value then rebounded sharply in 2021 and 2022. Those who held on were rewarded, while those who left kept the losses and missed the recovery.

Momentum fails in a different way. It tends to deliver steady gains punctuated by sudden crashes, usually when markets reverse sharply after a severe decline. In 2009, beaten-down stocks rallied off the financial crisis lows. Within a few months, momentum strategies lost gains that had taken years to accumulate. Kent Daniel and Tobias Moskowitz showed that such crashes cluster in panicked, high-volatility markets after steep declines, yet that knowledge does not make them painless.

Attempts to time factors have produced weak and unstable results. These attempts have relied on valuation spreads, macroeconomic signals, and recent performance. Cliff Asness of AQR has argued that factor timing is deceptively difficult, with valuation signals offering at best a modest edge. Research by David McLean and Jeffrey Pontiff adds another headwind. They found that anomaly returns fell by more than half on average after academic publication, suggesting that discovery itself invites the arbitrage that erodes a premium. For most investors, the sounder approach is to choose factors deliberately and hold them through discomfort.

Takeaway

Underperformance is not a flaw in the premium but the price of it. If factors never disappointed, there would be nothing left to earn.

Investors have two broad routes to factor exposure. The first is selecting individual securities based on factor characteristics. That might mean screening for low price-to-book or price-to-earnings ratios, strong trailing returns, high return on equity, or low leverage. This route offers full control and transparency. It also demands discipline, especially the frequent rebalancing momentum requires. Trading costs deserve close attention, because they can quietly consume a premium.

The second route is factor funds, often marketed as smart beta. These range from inexpensive single-factor ETFs to complex multi-factor strategies, and the labels hide significant differences. Two value funds may define value with different metrics, weight holdings differently, and rebalance on different schedules, producing materially different results. A fund's name matters less than three things: its methodology, its turnover, and its measured loadings on the factors it claims to target.

Diversifying across factors is one of the more robust practical findings. Value and momentum have historically been negatively correlated, so a combined portfolio tends to have shallower drawdowns than either factor alone. Quality pairs naturally with value. It helps filter out stocks that are cheap for good reason, the so-called value traps. Investors also face a design choice. They can hold a mix of separate factor sleeves, or use an integrated approach that ranks each stock on several factors at once.

Implementation costs deserve as much attention as expected premiums. Long-only funds capture only part of the academic long-short return. Expense ratios, turnover, and taxes subtract further. Crowding is another concern, because heavy flows into popular factors can stretch valuations and compress future returns. Perhaps the most important variable is behavioral. A modest tilt an investor can hold through a decade of lagging the index will likely outperform an aggressive one that gets abandoned at the bottom.

Takeaway

The best factor portfolio is not the one with the highest backtested return. It is the one you will still own after five years of trailing the market.

Factor investing sits where efficient markets meet behavioral finance. The major factors have survived decades of scrutiny, new data, and out-of-sample tests. That track record suggests they reflect something real, whether compensation for risk, persistent investor error, or both.

But the same features that let premiums persist also make them hard to harvest. Long droughts, sudden crashes, and post-publication decay ensure that the edge never feels comfortable. If it did, it would likely be arbitraged away.

The practical lesson is modest. Choose factors with a clear economic rationale, diversify across them, keep costs low, and size the tilt to your tolerance for tracking error. The systematic edge belongs less to the cleverest investor than to the most patient.