Equal Weight Factor Exposure: More Value Than Size

Key takeaways

• Over February 1967 to December 2009, equal-weighted portfolios of S&P 500 stocks returned 13.19% a year against 10.48% for value-weighted ones, a gap of 271 basis points.
• The value tilt supplied 139 of those basis points and the size tilt only 81, on the study's own factor loadings and premia.
• Momentum exposure ran the other way and cost 108 basis points a year, because holding every stock at the same weight means selling whatever just went up.
• The remaining 115 basis points was alpha, and it fell from 175 basis points to 80 as the rebalancing interval stretched from one month to twelve.
• Over the ten years to 31 December 2025 the S&P 500 Equal Weight Index returned 11.71% a year against 14.82% for the S&P 500, trailing by 3.11 points.

Equal weight factor exposure, decomposed: 81 basis points of size, 139 of value

You've heard that equal weight beats cap weight, and that it's really a small-company bet wearing a large-company name. The first claim was historically true. The second is about a third right.

The most careful decomposition on the record is by Yuliya Plyakha, Raman Uppal and Grigory Vilkov, in a paper published by EDHEC in March 2012. They took the constituents of the S&P 500 and ran them from February 1967 to December 2009. For each weighting rule they built 1,000 portfolios of 100 stocks drawn at random, then averaged the results. Equal weighting returned 13.19% a year. Value weighting, which is what a cap-weighted index does, returned 10.48%. Price weighting returned 12.07%.

The 271-basis-point gap between equal and value weighting is the figure everyone quotes. What matters is what sits inside it. The authors split the gap into a systematic part, which is payment for factor exposure, and alpha, which is not. The systematic part came to 156 basis points. Alpha accounted for the other 115. In their words, "42% comes from the difference in alpha and 58% from the excess systematic component."

So more than half of the gap was paid for by risk the reader already owns. Equal weight's factor exposure is real, then. The interesting question is which factors it actually buys.

Four betas, four premia, and one of them works against you

The systematic 156 points can be taken apart further, because the paper publishes every input. It reports the four-factor loadings for each weighting rule. It also reports the premium each factor earned over the same sample. Multiply the difference in loading by the premium and you get that factor's contribution.

Those four contributions sum to 157 basis points. The paper's own figure is 156, so the arithmetic closes to a rounding error, and it says something specific.

The size tilt is worth 81 basis points, which is 30% of the whole 271-point gap. The value tilt is worth 139, or 51%. Equal weight factor exposure is more a value bet than a size bet, by close to seven to four. That is not how the strategy is usually described, and it is not a small distinction, because the size premium and the value premium have each spent whole decades negative since these numbers were measured.

The momentum leg is the one nobody mentions. An equal-weight rule sells whatever rose and buys whatever fell, every time it rebalances. That is a negative momentum premium exposure by construction. The paper puts it at -0.1379 for equal weight against -0.0130 for value weight. Set against a momentum premium of 8.61% a year over that sample, the difference cost 108 basis points. Had momentum exposure been neutral, the gap would have been near 379 basis points rather than 271. The chart above shows all five components, the four factor legs and the alpha.

The other 42% is not a factor bet at all, and it decays

Alpha is what the four-factor model cannot explain. Equal weighting earned 175 basis points of it a year. Value weighting earned 60, and price weighting 67.

The authors then did the experiment that makes the paper worth reading. They slowed the rebalancing down. Nothing else changed: same stocks, same equal starting weights, same sample. The annual alpha "drops from 175 basis points to 117 basis points and then to 80 basis points" as the interval went from one month to six to twelve. At twelve months the difference against value weighting was no longer statistically significant, with a p-value of 0.96.

Their conclusion is blunt. Alpha "depends only on the monthly rebalancing and not on the choice of initial weights." Equal weighting is not the source of that 115 basis points. Trading back to equal weights is. The mechanism they identify is short-horizon reversal, and they attribute 11% of the alpha to a formal reversal factor. Buying what fell last month is a contrarian trade, and over this sample the contrarian trade paid.

That has an awkward implication for anyone reading equal weight as a factor story. The half of the outperformance that is a factor story is the half that can, and did, reverse. The half that isn't a factor story only shows up at a rebalancing frequency almost nobody runs.

The real equal weight index rebalances four times a year, not twelve

S&P Dow Jones Indices assigns each company in the S&P 500 Equal Weight Index "a fixed weight of 0.20% at each quarterly rebalancing", after the close on the third Friday of the quarter-ending month. That rule is set out in the June 2014 methodology document for the index. Four rebalances a year, not twelve. On the paper's own frequency curve that sits between the 117 basis points it measured at six months and the 175 at one month.

The index provider is not coy about the rest. Its methodology says the index lets an investor "make size, style, and sector bets relative to the S&P 500." That is the factor exposure stated on the tin by the people who build it. Calling it accidental is generous: what's accidental is that most people who buy the 500 leading companies in equal measure have never priced the four bets they picked up.

One date is worth holding on to. The index launched on 8 January 2003. Its published history runs back to 29 December 1989, so everything before 2003 is backfilled rather than lived.

The size premium by decade, which is the part that can go negative

Fama and French build SMB, the size factor, as "the average return on the three small portfolios minus the average return on the three big portfolios". Their annual series runs from 1927 and they give it away. Compounded decade by decade, with the value factor beside it, it looks like this.

The last row covers six years, not ten. Five of the ten rows are negative for size. Four are negative for value. The last two rows are negative for both at the same time, which is the single worst configuration for an equal-weight index, since those are the two legs carrying 220 of its 271 historical basis points.

These are realised premia over finished periods. They are a record of what happened, not a forecast of what comes next, and a decade is a small sample for a series this volatile. What the table does establish is range. A factor leg that returned 9.23% in one decade and -4.16% in another is not a constant you can quietly assume.

The last ten years are what a losing factor bet looks like

You don't have to model this. It already ran. Invesco's summary prospectus for its equal-weight fund, filed in August 2026, reports the two index records side by side for the periods ended 31 December 2025.

The S&P 500 Equal Weight Index returned 11.71% a year over ten years. The S&P 500 returned 14.82%. That is a shortfall of 3.11 points a year, compounding, over a decade. Over five years the equal weight index returned 10.48% against 14.42%, a gap of 3.94 points.

Read that next to the decade table and it stops being mysterious. The size leg compounded at -0.76% through the 2010s and -4.16% from 2020 to 2025. The value leg did worse. Two of the three positive legs in the 1967 to 2009 decomposition went negative at once, and the rebalancing alpha, which was 80 basis points at annual frequency in the original study, was never large enough to cover that. The equal weight vs market cap record over the full half-century is a separate question with a separate answer; this is what one decade of the factor legs going the wrong way costs.

The strongest objection: alpha in a factor model is not money

Here is the case against the rebalancing story, and it is a real one.

Alpha is a residual. It is what's left after four factor exposures are priced at their sample premia, which means it is an accounting entry, not a cash return. A strategy can carry positive alpha and still lose to the index, if its factor legs go against it. That is precisely what the last ten years did.

Ken French's own data shows how thin the raw effect is. His library publishes both a value-weighted and an equal-weighted return series for the same ten size deciles, built from June market equity and NYSE breakpoints. Run them annually from 1927 to 2025. In the smallest decile, equal weighting compounded at 15.96% against 12.18%, an advantage of 3.78 points. In the largest decile, the one that most resembles an S&P 500 universe, equal weighting compounded at 9.59% against 9.95%. It lost by 0.36 points a year over 99 years.

So the equal-weight advantage in raw returns is concentrated where the stocks are smallest, which is another way of saying it is a size effect after all, not a free rebalancing bonus. Among large companies the same rebalancing rule, applied for a century, produced no advantage at all. That doesn't refute the alpha finding, because alpha is risk-adjusted and these returns are not. It does mean the two results answer different questions, and only one of them is the question a reader with money in a fund is asking.

Cost belongs in the same paragraph. Invesco's equal-weight fund turned over 27% of its portfolio in its most recent fiscal year and charges 0.20% a year. The iShares cap-weighted S&P 500 fund turned over 3% and charges 0.03%. Nine times the turnover and 0.17 points more in fee is the entry price for the rebalancing leg, before any tax. The EDHEC paper put equal weight's own trading cost at 0.41% a year against 0.07% for value weight, on a 50-basis-point assumption. Equal weighting also ran hotter: annualised volatility of 0.1790 against 0.1583.

What this evidence cannot tell you

Start with the sample. The decomposition is US large companies over February 1967 to December 2009. It stops before the mega-cap concentration of the 2010s and 2020s, which is the single biggest thing that has happened to cap-weighted indices since. Applying a 1967 to 2009 loading to a 2026 portfolio assumes a stability nobody has demonstrated.

The premia used in the table are estimated on that same sample. Reusing a sample's own factor returns to explain that sample's outperformance is standard practice and it is also circular in a way worth naming. Change the window and every contribution in that table changes with it.

It is a single study, published by EDHEC in March 2012, and its portfolios are resampled 100-stock baskets rather than the actual S&P 500 Equal Weight Index. The index itself, as noted, is backfilled before 2003. The decade table is a backtest of a factor series, and factor series are reconstructions of what a costless long-short portfolio would have earned. No investor earned SMB. Factor ETF returns and the paper records they are built on are not the same thing.

Everything here is also US. The decomposition has not been run on any other market in this piece, so nothing above should be read as a claim about a global equal-weight index.

What would change the conclusion

The cleanest falsification is a sustained positive run in both SMB and HML. Two of the four legs of equal weight factor exposure are those factors, and together they carried 220 of the 271 basis points. A decade like the 2000s, where size compounded at 4.51% and value at 7.20%, would put an equal-weight index back in front without anything about the rebalancing mechanism changing at all. The case here is arithmetic about exposures, not a claim that those exposures are finished.

The second is concentration. The momentum drag of 108 basis points is the cost of systematically trimming winners. In a market where a handful of companies keep compounding faster than everything else, that drag widens. If index concentration falls back, it narrows, and the leg that has hurt equal weight most turns into the leg that helps it.

The third is frequency. If an equal-weight product moved from four rebalances a year to monthly, the paper's own curve says the alpha roughly doubles, from around 80 basis points to 175. It also says the turnover bill goes up with it. Nobody has run that experiment in a live fund with real spreads and real tax, which is why the 115-basis-point alpha in the table above is the least reliable number in this piece rather than the most exciting one.

The thing to watch isn't the equal-weight product. It's what your own holdings weigh against what you think they weigh, and which of the four legs above you've taken without meaning to. LedgerTouch shows the weights; the factor legs you have to reason about yourself.