Factor Investing

Decades of research found that certain measurable characteristics of stocks have been associated with different long-run returns. This guide covers the five factors with the strongest evidence, what could explain them, what implementation actually costs, and why patience is the entire price of admission.

By Karanvir Singh, founder of Quantix Invest

The core idea

A total market index fund gives every stock a weight proportional to its market value. That is a defensible default: it is what the average investor holds, it costs almost nothing, and it requires no forecast about anything.

Factor investing asks a narrower question. Rather than sorting stocks by size alone, sort them by some measurable characteristic (cheapness, recent price trend, profitability, price volatility) and check whether the groups have earned systematically different returns over long periods. Where they have, and where the difference survives reasonable scrutiny, that characteristic is called a factor, and the return difference is called a premium.

Two things follow, and both matter.

First, factors explain a great deal of what looks like manager skill. If a fund manager beat the market for a decade while systematically holding cheap, small, or highly profitable stocks, most of that excess return may be exposure to a known factor rather than stock-picking ability. Factor models are how the industry separates the two. That is arguably the more important use of the research: as a diagnostic, not a strategy.

Second, if a premium is real and persistent, it can be harvested cheaply and systematically by a rules-based fund. That is what factor funds (often marketed as "smart beta") try to do.

Where factors came from

The original asset pricing model, the Capital Asset Pricing Model developed in the 1960s, said one thing determined a stock's expected return: its sensitivity to the overall market, called beta. Higher beta, higher expected return. Everything else was noise.

Empirically, that turned out to be incomplete. By the 1980s researchers had documented several patterns the model could not account for. Small companies had outperformed large ones. Cheap stocks, measured by book value relative to price, had outperformed expensive ones. In 1992 and 1993, Eugene Fama and Kenneth French formalized this into a three-factor model: market, size, and value. Later work added momentum (Jegadeesh and Titman documented the effect in 1993, and Mark Carhart incorporated it into a four-factor model in 1997), and in 2015 Fama and French extended their own model to five factors by adding profitability and investment.

These are not fringe ideas. Fama shared the 2013 Nobel Prize in Economics, and the factor framework is standard in academic finance and institutional portfolio construction. That is a genuine point in the evidence's favor, and also, as we will see, a reason to expect the premiums to be smaller now than in the data that discovered them.

Value

The rule: favor stocks that are cheap relative to a fundamental anchor, such as book value, earnings, cash flow, sales, or dividends.

The evidence: in long-run US data going back to the 1920s and in most developed markets studied since, portfolios of the cheapest stocks outperformed the most expensive ones over multi-decade horizons. The gap in the long-run academic series has often been quoted in the range of roughly 3 to 5 percentage points a year, though the figure depends heavily on the period, the definition, and whether it is measured long-only or as a long-short spread.

The recent experience: value had a brutal stretch. From roughly 2007 through 2020, the classic price-to-book value factor in US large caps delivered one of the worst drawdowns in its recorded history, underperforming growth for well over a decade. This was not a mild disappointment; it was long enough for a large share of investors who had committed to value to abandon it, and long enough for serious people to argue the premium had been arbitraged away or had never been robust to begin with. Value staged a meaningful recovery starting in late 2020, which is roughly when many investors had given up.

The definitional problem: the classic value measure is price-to-book, and as discussed in the valuation guide, book value has become a weaker economic anchor. Research and development, software, and brand investment are expensed rather than capitalized, so the most valuable assets of modern firms do not appear in book value. Multiple research groups have shown that value definitions using earnings, cash flow, or an intangible-adjusted book value performed considerably better than raw price-to-book over the past two decades. Whether that is a genuine improvement or a retrofit to recent data is a fair question, and the honest answer is: partly both.

Size

The rule: favor smaller companies over larger ones.

The evidence: weaker than the others, and it deserves the demotion. The size effect was documented in 1981, and in the decades after publication it largely disappeared in US large-cap-adjacent data. Much of the historical premium concentrated in January (the "January effect"), in the very smallest and most illiquid microcaps where trading costs are highest, and in stocks that would be hard or expensive for a real fund to own in size.

The nuance that survived: later research, particularly work associated with AQR, argued that size looks much stronger once you control for quality. Small stocks are a mix of two very different populations: small profitable firms with real businesses, and small unprofitable speculative firms. The junk half dragged the average down. Small plus quality has a considerably better record than small alone.

Practically, size is best treated as something to combine with another factor, not as a standalone bet. A plain small-cap index fund is a reasonable diversifier, but the case that it carries a reliable standalone premium is the weakest of the five.

Momentum

The rule: favor stocks that have outperformed over the past 6 to 12 months (typically skipping the most recent month, which tends to reverse). Hold for a few months, then re-sort.

The evidence: momentum is, in raw statistical terms, the most consistently documented factor across asset classes and geographies. It has been found in US and international stocks, industries, bonds, currencies, and commodities, and in data reaching back to the 1800s. It is also the most awkward for efficient-market theory, because it is a pure price pattern with no fundamental anchor at all.

The catch, and it is a serious one: momentum crashes. The strategy earns steady positive returns most of the time and then loses enormously at sharp market turning points, when the losers that momentum is short (or underweight) rebound violently. In 1932 and again in 2009, momentum strategies suffered devastating reversals within months. The return profile resembles selling insurance: many small gains, occasional catastrophic losses.

And it is the most expensive to run. Momentum requires high turnover by construction, often well above 100% a year, which generates trading costs and, in a taxable account, short-term capital gains taxed at ordinary income rates. A meaningful share of the paper premium can be consumed by implementation. Momentum is the factor where the gap between the academic long-short spread and what an investor actually keeps is widest.

Quality and profitability

The rule: favor companies with high and stable profitability, strong returns on capital, low leverage, low earnings volatility, and conservative investment (that is, not aggressively growing their asset base).

The evidence: Robert Novy-Marx's 2013 work on gross profitability showed that profitable firms outperformed unprofitable ones, and Fama and French's 2015 five-factor model added profitability and investment as formal factors. Quality is intuitively strange at first: if these are the better businesses, why should they also earn higher returns? A risk-based story struggles here, which is why quality is more often explained as a persistent underreaction to how durable good businesses actually are.

Why it is popular: quality has behaved well as a portfolio component. It has tended to hold up better in downturns than the market, and it pairs unusually well with value. Buying cheap stocks is dangerous precisely because cheap often means deteriorating; screening for quality filters out a meaningful share of value traps. The combination has a better record than either alone in most studies, which is the single most useful practical result in the whole factor literature.

The definitional weakness: quality is the least standardized factor. There is broad agreement on value (price relative to fundamentals) and momentum (past return). There is no agreed definition of quality. Different index providers use different metrics with substantially different results, which means "quality fund" tells you far less about what you own than "value fund" does. Read the methodology.

Low volatility

The rule: favor stocks with low price volatility or low market beta.

The evidence: low-volatility stocks have historically delivered returns similar to or better than the broad market with meaningfully lower volatility, which means better risk-adjusted returns. This directly contradicts the textbook prediction that higher risk earns higher return, and it has been observed across many markets, which is why it is sometimes called the low-volatility anomaly rather than a premium.

The leading explanation: leverage constraints. Many investors cannot or will not borrow to amplify returns, so an investor who wants higher returns bids up high-beta stocks instead. That demand pushes high-volatility prices up and their expected returns down. Add the behavioral pull of lottery-like stocks, where investors overpay for small chances of large gains, and you get persistent overpricing at the volatile end.

The practical caveats. Low-volatility strategies carry large, unintended sector bets: they load heavily on utilities, consumer staples, and real estate. They also behave like a bond proxy and have historically suffered when interest rates rise sharply. And by 2016 to 2020 the strategy had attracted enough money that low-volatility stocks themselves became expensive relative to their own history, which is a reminder that a factor can be crowded out of its own premium. Buying any factor at an unusually high relative valuation is a worse proposition than buying it at a normal one.

Why would a premium exist at all?

If a rule reliably produced extra return, why would anyone leave it lying there? Three families of explanation, and which one you believe determines how much confidence the premium deserves going forward.

1. Compensation for risk. The factor earns more because it carries a risk that matters to real investors. Value stocks may be genuinely more fragile: they are often distressed, operationally leveraged, and most likely to fail in a recession, exactly when investors can least afford losses. Under this story the premium is not free money; it is a fee you are paid for holding something genuinely uncomfortable. It should persist indefinitely, because the discomfort is real and does not go away when it gets published.

2. Behavioral error. Investors systematically overpay for exciting, fast-growing, story-driven companies and underpay for dull ones. They underreact to news (which produces momentum) and overextrapolate recent growth (which produces the value premium). Under this story the premium exists because of human wiring, which changes slowly. It should persist to the degree the bias does, but it can shrink as the trade gets crowded.

3. Structural and institutional limits. Leverage constraints support low volatility. Benchmark-hugging career risk means professional managers cannot afford to look wrong for years, so they avoid the trades that require it. Short-selling constraints allow overpricing to persist at the expensive end. These are real frictions, but they can be relieved by product innovation and capital flows.

Note the tension. Risk-based explanations imply persistence but also imply that you must actually bear the risk, meaning long stretches of pain are the mechanism, not a malfunction. Behavioral and structural explanations imply the premium can shrink as more money chases it. Both stories predict that the easy version is over.

The skeptical case

Take this section seriously. The strongest arguments against factor investing come from within the research community itself.

• The factor zoo. Researchers have published hundreds of claimed factors. Campbell Harvey and coauthors surveyed this literature and argued that with so many tests run on the same data, conventional statistical significance thresholds are far too lenient, and that a large share of published factors are likely false discoveries. Their proposed remedy was raising the bar substantially. Most published factors do not clear it.
• Replication problems. Independent attempts to reproduce published anomalies on fresh data have found that a substantial fraction weaken sharply or vanish. Estimates vary by methodology, but the direction is consistent: out-of-sample performance is materially worse than in-sample.
• Post-publication decay. Research by R. David McLean and Jeffrey Pontiff found that anomaly returns fell by roughly a third to a half after publication, consistent with investors trading them away once they were public. Every factor you have heard of has been public for years or decades.
• Data mining is nearly unavoidable. There is one history of financial markets. Every researcher works on it. A pattern that survives thousands of collective attempts to find something may have survived by construction rather than by truth.
• The gap between paper and portfolio. Long-short, cost-free, tax-free academic spreads are not investable. After fund fees, turnover costs, market impact at scale, taxes, and the dilution of long-only implementation, a meaningful chunk of any premium disappears before it reaches you.
• Crowding. Money flowed into factor products in enormous size after 2010. If a premium depends on other investors making a mistake, and the mistake becomes an industry, the premium narrows.

Where does that leave a reasonable person? Roughly here: the market factor is well established. Value, momentum, quality, and low volatility have the strongest and most independently replicated evidence of the remainder, and they are supported by plausible economic stories rather than statistics alone. But expected premiums going forward should be assumed to be smaller than the historical numbers, possibly much smaller, and the uncertainty around them is wide enough that a diversified low-cost total market fund remains a completely defensible choice for anyone. Nothing in this literature makes factor tilting necessary.

Patience is the price of admission

This is the part that ends most factor investing in practice, and the numbers deserve to be spelled out.

Worked example: how long is a long time? Suppose a factor genuinely earns a 2% annual premium over the market, with a tracking error (the annual standard deviation of the difference between the factor's return and the market's) of 8%. Those are realistic figures for a long-only factor fund.

The signal-to-noise ratio is 2 / 8 = 0.25. To be reasonably confident the premium is real rather than luck, you would want the cumulative outperformance to exceed roughly two standard errors. The standard error of the average annual excess return over N years is 8 / the square root of N. Setting 2 x (8 / square root of N) equal to 2 gives square root of N = 8, so N = 64 years.

Sixty-four years of data to statistically distinguish a genuine 2% premium from noise. That is longer than most investing lifetimes. Over any single decade, the noise dominates the signal completely. And that calculation assumes you already know the true premium is 2%, which you do not.

Worked example: the probability of looking wrong. With the same assumptions, in any given year the excess return is centered on plus 2% with a standard deviation of 8%. The chance of a negative year is the probability that a draw falls below zero, which is the probability of being more than 0.25 standard deviations below the mean: roughly 40%.

Over a five-year window, the average annual excess return has a standard deviation of 8 / square root of 5 = 3.58%. The chance the five-year average is negative is the probability of being more than 2 / 3.58 = 0.56 standard deviations below the mean, roughly 29%.

Over ten years: standard deviation of 8 / square root of 10 = 2.53%, so the chance of a negative decade is the probability of being 0.79 standard deviations below, roughly 21%.

Read that again. Even with a factor that genuinely works, there is about a one in five chance that a full decade of your life produces underperformance. Roughly two years in five will be losing years. And this is the optimistic scenario where the premium is real and you have the discipline to hold.

Now add the human element. What actually happens is that an investor tilts to a factor, watches it lag for three or four years, reads increasingly persuasive articles about why this time the factor is broken, and switches into whatever has been working. That behavior converts a positive-expected-return strategy into a reliably negative one, because you have systematically sold each factor after it got cheap and bought each after it got expensive. The 2010s value drawdown did exactly this to a very large number of investors, many of whom exited shortly before the 2020 to 2022 recovery.

Implementation and costs

Costs are where paper premiums go to die, and they are the part of the analysis you can actually control.

Worked example: what survives. Assume a value factor with a true forward-looking long-short premium of 3%. A long-only fund captures perhaps half of a long-short spread, since it can underweight expensive stocks but not short them: call it 1.5%. Subtract an expense ratio differential of 0.22% (0.25% versus 0.03% for the index alternative), trading costs of 0.30%, and, in a taxable account, a tax drag of 0.40% from higher turnover and distributions.

Remaining: 1.5 minus 0.22 minus 0.30 minus 0.40 = 0.58% per year.

That is the realistic expected reward for accepting a strategy that will underperform in roughly two years out of five and may underperform for a full decade. It might still be worth it. But the honest number is well under 1%, not the 4% that appears in the research summaries, and it goes to zero or negative if the premium has decayed further than assumed or if you pay more than 0.30% in fees.

Two structural implications. First, hold factor tilts in tax-advantaged accounts where possible, particularly momentum. Second, fee discipline is not optional: a factor fund charging 0.60% has consumed most of the realistic premium before it starts.

How to read a factor fund

Factor funds vary enormously despite similar names. Before buying, answer these:

• What is the actual rule? Read the index methodology document, not the marketing page. How is the factor defined, how many stocks are held, how often does it rebalance?
• How concentrated is the tilt? Many "value" funds hold hundreds of stocks with weights barely different from the market. A weak tilt gives you index-like returns at above-index fees, which is the worst combination available. Look at the fund's factor loadings if published, or compare its top holdings against the broad index.
• What is the turnover? Above roughly 50% a year, costs and tax consequences become a first-order concern.
• What unintended bets came along? Value funds tend to overweight financials and energy. Low-volatility funds overweight utilities and staples. Quality funds can drift toward large-cap growth. You may end up with a sector bet you never chose.
• Is it multifactor? Combining value, momentum, and quality in one fund can smooth results, since the factors have historically been imperfectly correlated (value and momentum in particular have often been negatively correlated). But a poorly built multifactor fund can also net its own signals out into an expensive index fund. Check that the tilts are meaningful.
• What does it cost relative to 0.03%? That is the alternative, and the comparison is the whole decision.
• How long has this specific fund existed? Backtested index history is not live performance, and the gap between the two is systematically unfavorable.

Building a sensible allocation

If, after all of the above, you still want factor exposure, some principles that follow from the evidence rather than from marketing:

1. Start from a total market core. Factor tilts are a modification of a broad, low-cost portfolio, never a replacement for one. A common structure is a large majority in a total market index with a modest tilt around the edges.
2. Keep the tilt modest. A tilt large enough to seriously change your outcome is also large enough to seriously test your discipline. Something in the range of 10% to 30% of equities is enough to matter without being able to wreck the plan.
3. Diversify across factors. Value and momentum have historically been negatively correlated, so holding both smooths the ride considerably. Quality pairs naturally with value by filtering out traps. A single-factor bet concentrates exactly the risk you are least able to sit through.
4. Prefer low turnover and low fees. These are the only inputs you control with certainty.
5. Use tax-advantaged space. Especially for momentum or any strategy turning over more than half its portfolio annually.
6. Write the plan down before you buy. Include the specific sentence that you will not change the allocation for at least ten years regardless of relative performance. Then reread it during the drawdown, because there will be one.
7. Rebalance mechanically. Do not let the tilt drift with performance, which is how you end up overweight whatever just worked.

And the alternative that deserves equal standing: do none of this. A three-fund portfolio of total US stock, total international stock, and total bond market, held at rock-bottom cost with disciplined rebalancing, captures the market return and requires no belief about factor premiums at all. Over any realistic horizon it will land close to the factor-tilted version, with less to go wrong and far less to second-guess. The evidence supporting factors is real but noisy; the evidence that costs and behavior dominate outcomes is overwhelming.

This is education, not individualized financial advice. Whether any tilt suits your situation depends on your horizon, taxes, and temperament, and temperament is the input people most consistently overestimate in themselves.

Common mistakes

• Chasing the factor that just worked. Buying a factor after five strong years, when it is expensive relative to its own history, is the reverse of the strategy. Factor returns mean-revert enough that recent performance is a poor entry signal.
• Abandoning during the inevitable drawdown. The single most costly mistake, and the most common. Every real factor has multi-year losing stretches by construction.
• Paying too much. A 0.60% factor fund needs to beat the market by 0.60% just to draw even with a 0.03% index fund. That is a large fraction of any realistic premium.
• Confusing a factor tilt with a market-beating system. Factor investing at best shifts expected returns modestly while increasing tracking error substantially. It is not an edge, it is a different set of risks.
• Treating the historical premium as the expected premium. Post-publication decay and crowding both point the same direction: assume less.
• Holding high-turnover factors in a taxable account. Momentum in particular can lose a large share of its premium to short-term capital gains.
• Stacking too many tilts. Owning six factor funds plus a total market fund often nets back to the market at three times the cost.
• Ignoring what the fund actually holds. Names are marketing. Methodology documents are the product.