Week 23: Factor Investing — Value, Momentum, Quality, Low-Vol, Size

---

1. Why This Is Important

For half a century the Capital Asset Pricing Model (CAPM) told a single story: stocks beat bonds because stocks carry more market risk, and the only thing the market rewards is bearing that one risk. One factor. One reward. End of equation. Then in the early 1990s Eugene Fama and Kenneth French sat down with sixty years of CRSP data and discovered that the equation was missing terms. Small companies beat big ones by more than their beta. Cheap (high book-to-market) stocks beat expensive ones. Once they added a size factor and a value factor the residual "alpha" of most US mutual funds collapsed — what looked like stock-picking skill was, on the data, just systematic exposure to factors the manager had never named. By 2015 Fama and French had added profitability (RMW) and investment (CMA), and Mark Carhart had bolted on momentum (UMD) into a six-factor framework that today explains roughly 90% of the cross-section of US equity returns.

This matters for four reasons.

It tells you what your portfolio actually is. Whatever you

own — a single mutual fund, a 60/40 mix, your spouse's "growth stock" picks — has factor loadings whether you measure them or not. Two portfolios with the same beta and the same expected return can behave very differently in the next regime if one is heavy on value-and-size and the other is heavy on quality-and-low-vol. Factors are how you read what's under the hood.

It separates structural alpha from luck. When a manager beats

the S&P by 3% over a decade, the right question isn't "are they good?" — it's "what factor was that?" If their excess return loads onto small-value, you can replicate it with VBR for 7 bps of fee. If it loads onto momentum, you can replicate it with MTUM. The moment a putative "skill" fits inside an academic factor it stops being skill.

It is one of the durable structural alpha lanes. Among the

handful of structural mispricings worth chasing — liquidity, sector rotation, long-term trends, buying what passive flows have abandoned, and the academic factor premia we're discussing this week — factors are the most systematic and the most crowded — which means they have the lowest barrier to entry and the most compressed forward expected return. They still pay; they just pay less than they did before everyone owned VLUE and MTUM.

It frames the 2007 quant quake we covered in Week 14. When

AQR, Renaissance, and a hundred lesser stat-arb shops were all running the same factor signals on the same universe, a single mid-August deleveraging event vaporised three years of returns in four days. Factor investing with leverage was the canonical case study in crowded-trade risk. Retail factor ETFs in 2026 are unleveraged, so the blow-up math is gentler — but the compression of premiums is the same disease.

---

2. What You Need to Know

2.1 What a "factor" actually is

A factor is a long-short portfolio built from a measurable stock characteristic. Take book-to-market: rank every US stock by B/M, go long the top decile, short the bottom decile, weight the legs equally, rebalance monthly. The monthly P&L of that portfolio is the *value factor* (HML — High Minus Low book-to-market). Every factor in the academic literature has the same shape: rank, long top, short bottom, hedge out the market.

That construction matters, because the long-short structure is what strips out the market. A pure value mutual fund is roughly *0.9 of market + 0.4 of HML + noise; the academic HML is the 0.4 of HML* piece on its own, with the market dialled to zero. The factor premium is the average return of that long-short portfolio. Six factors dominate the modern literature: market (MKT-RF), size (SMB — Small Minus Big), value (HML), profitability (RMW — Robust Minus Weak), investment (CMA — Conservative Minus Aggressive), and momentum (UMD — Up Minus Down). The first five are Fama-French 5; adding UMD gives you the FF5 + momentum framework that almost every sell-side risk model in 2026 starts from.

2.2 The historical premia, in one breath

From July 1963 to December 2024, the realised annualised premia for the US factors are roughly: MKT-RF ~6.6%, UMD ~7.5%, HML ~3.8%, CMA ~3.3%, RMW ~3.0%, SMB ~2.4%. Add a low-volatility factor (not in the official Fama-French set but built the same way: long bottom-quintile beta, short top-quintile beta) and you get roughly 1-2% of premium with materially lower standard deviation — which is why low-vol shows up well on a Sharpe-ratio basis even though the raw premium looks small.

A few orderings on this list are surprising. Momentum, the only factor that didn't come out of Fama-French's stable, has the largest premium of any of them — and the worst single drawdown. Size, the factor that started the whole literature, is the smallest premium and arguably indistinguishable from zero post-2000 once you adjust for microcap illiquidity. Value, the factor your wealth manager probably mentions first, is middle-of-the-pack and has *flatlined for the last seventeen years*.

2.3 The factor-decay problem since 2003

Pull up the rolling 10-year HML premium and you'll see something ugly: it averaged about +5% per year from 1963 to roughly 2002, then fell off a cliff. The rolling 10-year ending in 2020 was negative. Between 2008 and 2020, the academic value factor — the long-short construction, not value mutual funds — lost money on average. UMD has its own version of the same disease, with a different shape: a long mean of 7-8% interrupted by the 2009 momentum crash, when the factor drew down roughly 45% in calendar 2009 (and ~83% peak-to-trough) as the market V-bottomed and previously losing junk stocks ripped higher off the lows.

There are three competing explanations for why the premia have compressed, and they're probably all true. First, *publication arbitrage*: every academic factor that gets written up in the JoF attracts capital, and that capital bids up the long leg and sells down the short leg until the premium shrinks. The peer-reviewed finding is the obituary of the trade. Second, *intangibles mis-measurement*: book-to-market punishes asset-light tech firms that show low book equity because their value is in non-capitalised R&D and brand. The HML construction reads them as "expensive" when on a corrected basis they're not. Third, *post-GFC central-bank liquidity*: zero rates compress dispersion across stocks. Everything trades together; nothing trades on fundamentals; nothing is cheap or expensive any more. Pick whichever theory you like — the empirical fact is that the premium has roughly halved versus its pre-2003 mean.

2.4 The 2007 quant quake — what factor crowding actually feels like

The week of August 6, 2007 is the textbook example of factor crowding, and we covered it in detail in Week 14 from the pair-trading angle. Here's the factor-investing version. Every quantitative long-short shop in 2007 was running essentially the same factor portfolio: long value, long momentum, long quality, short the opposite, leveraged 5-8x for risk-target purposes. On August 6 something — most plausibly a multi-strat fund forced to deleverage to meet redemptions in its credit book — started unwinding the standard factor portfolio in size. Because *every other quant fund held the same book*, the unwind didn't find counterparties in the way an idiosyncratic position would; it found a hundred funds all running the same risk model and all stopping out at roughly the same loss threshold, all at the same time. By Friday August 10 the standard factor portfolio had drawn down about 25% — three years of realised premium gone in four trading days. The market itself moved about 1%. The retail tape never noticed.

Two lessons survive from that week. First, *factor exposure is risk even when the factor's beta to the market is zero*. The risk you took wasn't market risk; it was crowded-trade risk. Second, *unleveraged factor exposure is a different animal*. Retail factor ETFs in 2026 have no leverage and no daily VaR-driven stop-out. They can sit through a quant unwind and just take the temporary mark. The blow-up math is mostly a leverage-and-redemption problem, not a factor problem per se.

2.5 The retail factor ETF menu in 2026

The factor literature got cheap and scalable. iShares and Vanguard between them now run the standard menu of single-factor ETFs at expense ratios between 8 and 25 bps:

• VLUE (iShares MSCI USA Value Factor) — value tilt, 0.08% ER.
• MTUM (iShares MSCI USA Momentum Factor) — momentum tilt, 0.15% ER.
• QUAL (iShares MSCI USA Quality Factor) — RMW-flavoured quality

tilt, 0.15% ER. Largest of the bunch by AUM.

• USMV (iShares MSCI USA Min Vol Factor) — low-volatility tilt,

0.15% ER.

• IWM / IJR (iShares Russell 2000 / S&P SmallCap 600) — size

tilt via small-cap exposure, 0.19% / 0.06% ER.

• AVUV / AVDV (Avantis US / International Small Value) —

combined size+value+profitability tilt, 0.25% ER. The 2020s flagbearer of the academic-factor school.

You can build a respectable multi-factor blend with four of these tickers, total expense ratio under 0.20%, and you've replicated what hedge funds were charging 2-and-20 for in 2003. The premium has compressed, but so has the cost of harvesting it — and the net premium (gross minus fee minus tax drag) is roughly the same as it was for the leveraged hedge funds twenty years ago, because the fee compression has roughly matched the premium compression. Cheap beta ate expensive alpha.

2.6 Where this fits in the bigger picture

Factor investing is the most structural of the available alpha lanes — and the least idiosyncratic. It scales infinitely (any size of capital can run it), it requires no special information, and the academic literature documenting it is freely available on SSRN. That's also exactly why the premium has compressed: anything anyone can run, eventually everyone runs. Compare that to "buying what passive flows have abandoned" (the contrarian sector trade): that requires discomfort, requires being early, requires holding through a period when nobody else does. The factor trade requires only that you click VLUE and forget about it. The market is efficient at arbitraging away premium that requires no discomfort to capture.

The implication for your own book: factors belong in the *passive core (the foundational, lowest-active tranche), not in the active sleeve*. A modest tilt — 10-20% of your equity allocation in QUAL or AVUV versus a market-cap-weighted core — gets you most of the diversification benefit without taking concentrated factor-blow-up risk. Putting 100% of your equity in MTUM is taking the 2009-style tail risk; putting 10% in MTUM and 90% in VTI gets you 80% of the long-run mean and 20% of the tail.

---

3. Common Misconceptions

"Factor investing is the same as value investing." Value is

one factor of six. The Fama-French framework is a system: each factor is a separate long-short portfolio, and a "factor investor" usually owns several of them. Calling factor investing "value" is like calling exercise "running."

"The size factor has a big premium." SMB has the smallest

realised premium of the headline factors, around 2.4% annualised. Most of the gap shows up only in the bottom microcap quintile, which has serious illiquidity issues. The "small-cap effect" you hear about is mostly a small-value effect — value, not size, doing the work.

**"Momentum is just the trend-following you read about in the

1990s."* The academic UMD factor is cross-sectional*: rank stocks by their past 12-month return (skip the most recent month), go long the winners, short the losers, monthly rebalance. Trend-following in commodities is time-series: long if the asset itself is up, short if it's down. Different construction, different correlation, different drawdown profile.

"Factors are dead — they don't work any more." The premia have

compressed, not disappeared. Even at half their pre-2003 means, a four-factor multi-factor blend has positive expected excess return net of fees in 2026. What's dead is the leveraged-quant version from 2007 — the unleveraged retail-ETF version is alive but smaller.

"More factors is always better." The literature has identified

over 400 "factors" in published papers, most of which fail to replicate out of sample. The Fama-French 5 plus momentum is the conservative consensus. Anyone selling you exposure to a 20-factor model is selling overfit.

"Low-vol works because low-vol stocks are mispriced." Maybe

partly, but mostly low-vol works because of leverage aversion: most institutional investors can't lever, so to get equity-like returns they crowd into high-beta stocks, bidding their prices up and their forward returns down. Low-beta stocks get under-bid as a result. Take away leverage constraints and the premium would shrink — which is partly why hedge-fund "betting against beta" strategies have done well.

"Quality and low-vol are the same factor." They overlap (quality

companies tend to be lower-vol) but diverge in stress: in 2008-09, USMV-style low-vol got hammered along with everything else as correlations went to 1, while QUAL-style quality (high ROE, low debt) outperformed. Different definitions, different risk shapes.

**"I should rebalance my factor blend monthly to capture the

premium."* Monthly rebalancing for a single* factor (the academic construction) is correct. For a multi-factor blend held in retail ETFs, monthly rebalancing has a tax cost that often exceeds the rebalancing benefit. Annual rebalance is a defensible compromise for taxable accounts.

"The 2007 quant quake means factor investing is dangerous."

The 2007 event was about leverage and *redemption-driven deleveraging*, not factors per se. An unleveraged retail factor blend in a taxable account doesn't have the same blow-up topology. The risk is forgone return (factors compress further), not catastrophic loss.

"If everyone knows about factors, the premium should be zero."

The premium has halved since 2003, which is the strong-form version of "everyone knows." It hasn't gone to zero because (a) a lot of capital still cares more about benchmark tracking than about factor exposure, and (b) some of the premium is risk-compensation that doesn't go away just because the risk is well-understood. The remaining premium is the residual after the arbitrageable portion has been arbed out.

---

4. Q&A Section

Q1: Should I replace my S&P 500 fund with a multi-factor ETF? No — keep the cap-weighted core for cost, tax, and tracking-error reasons, then tilt with 10-30% of your equity allocation toward factors. The cap-weighted core gets you the market premium for almost no cost; the factor tilt gets you the (compressed) factor premium on top. Replacing the core with a factor product takes on tracking-error risk you usually don't need.

Q2: Which single factor is best? Long-run, momentum has the highest gross premium and the highest volatility — so it has the worst tail (2009) and the best mean. Quality has the best Sharpe ratio. Value has the best valuation case in 2026 because it's the most beaten-down. Pick by your tolerance for which kind of regret: missing-the-rally (avoid momentum) or the-thing-keeps-getting-cheaper (avoid value).

Q3: How long should a factor go through a drawdown before I quit? Long enough that the answer should be "I won't quit." HML drew down for thirteen years from 2007 to 2020 before the 2021-22 value rally. If you can't sit through a decade-plus drawdown of the factor you tilted to, you don't have the conviction to harvest it; you'll sell at the bottom and miss the rebound. If 13 years sounds intolerable, don't tilt — own the cap-weighted index.

Q4: Why is small-value (AVUV) so popular if size is a weak factor? Because the interaction of size and value (and profitability) is where the literature still finds robust premium. Small junk stocks underperform; small value-with-profitability stocks have a materially better historical profile than either size or value alone. AVUV is built around that interaction.

Q5: Are factors correlated with each other? Some are highly correlated (HML and CMA), some are negatively correlated (HML and UMD historically — value and momentum are partial hedges for each other), and some are weakly correlated (SMB with everything). A multi-factor blend gets diversification benefit because the factors aren't perfectly correlated. The HML/UMD negative correlation is the single most useful pairing for a retail blend: when one is in a drawdown the other is often having a good year.

Q6: Does international factor investing work? The academic premia replicate internationally (developed and EM) with similar magnitudes and similar 2003-2020 compression. We stay US-only for investability reasons (FX, custody, tax, information disadvantage), but the academic story is broadly international, not US-specific.

**Q7: How do I tell if my actively-managed mutual fund is just running a factor tilt?** Run a regression of the fund's monthly excess returns against the six factors (you can pull the data free from Ken French's website). If the R-squared is over 0.90 and four or more factor loadings are statistically significant, the fund's performance is mostly factor exposure, and you should compare its fee against the factor-replication ETF cost. Morningstar Direct and Portfolio Visualizer both run this regression for free; there's no reason to pay an active manager for what amounts to a buyable factor blend.

Q8: What about the "low-vol anomaly" — isn't it a free lunch? It's the closest thing the equity market has to a free lunch on a Sharpe basis, but it's not free. Low-vol underperforms in melt-up regimes (think 1999, 2020-21) and tracks-error materially against the cap-weighted index. The factor pays for itself in risk-adjusted terms over a full cycle, but the absolute return drag versus the S&P during a tech-led rally can be 10-15% over two years.

Q9: Did the 2009 momentum crash kill momentum forever? No — UMD has had positive returns in most years since 2009, and the post-2010 mean is roughly half the long-run mean (still positive, just smaller). The 2009 crash is a permanent feature of momentum's return distribution, not a one-off — there were similar (smaller) momentum crashes in 1932, 2002, and a mini one in March 2020. Owning momentum means accepting that one out of every twenty years it loses 30%+.

Q10: Should I time my factor exposures based on valuation? Empirically yes — Cliff Asness's "Value of Everything" research shows that the valuation spread of a factor (cheap-leg P/B vs expensive-leg P/B) predicts the factor's forward 5-year return. In 2026 the value factor is at the cheap end of its history (so the forward expected premium is above the long-run mean) and momentum is near the expensive end. But this is a slow-moving overlay, not a trading rule — you adjust 10-20% of your tilt every few years, not every quarter.

Q11: How does factor investing fit with the four-tranche framework? Factors live in tranche one (the passive core) and tranche two (the slightly-active tilt). They do not belong in tranche three (the active concentrated book) or tranche four (the convex tail). A factor blend isn't trying to compound 30% — it's trying to add 50-100 bps a year to your passive return at modest tracking error. That's the tranche-one-and-two job description.

Q12: What's a reasonable multi-factor blend for 2026? A defensible starter: 60% VTI (cap-weighted core), 10% AVUV (small-value-quality), 10% MTUM (momentum), 10% QUAL (quality), 10% USMV (low-vol). Total ER under 12 bps. This gives you all six factor tilts at modest weights, with cap-weighted core dominating. Annual rebalance. If you can hold this through a 30% momentum drawdown without selling, you'll harvest most of what's left of the academic factor premium.