Factor Investing Is Very Much Alive

“Factor investing is dead” makes for catchy headlines, but it is difficult to reconcile with the recent record of AQR Style Premia Alternative QSPRX, with about $3 billion in assets under management. I wrote about the fund in November 2014, recommending investors consider it, and I have owned it almost since inception. The fund is not a long-only factor tilt offering that can be overwhelmed by an equity bull market; it is designed to harvest long–short style premiums across markets. Its results are not proof that every factor, every implementation, or every manager will succeed. They are, however, meaningful evidence against the much broader claim that publication and investor flows have eliminated factor premiums.

The Claim: Publication and Flows Killed Factor Investing

A common critique holds that once academics document an anomaly and investment products package it for investors, capital flows into the trade, valuations adjust, and the premium is arbitraged away. In its strongest version, the claim is that factor investing has failed—or is “dead”—because publication and smart-beta flows have destroyed the opportunity set.

There is real evidence of postpublication decay in some individual anomalies. Research summarized in the “factor zoo” literature reports a roughly 58% decline in average anomaly returns after publication in US evidence. But a lower realized premium after publication is not the same as evidence that the premium has been eliminated. Nor does it establish that all factors have been arbitraged away across all markets and implementations. In fact, if a factor has a risk-based explanation for a premium, the underlying risk cannot simply be arbitraged away—though premiums can, of course, be time-varying (just as is the market beta premium). Investor flows can also affect valuations and, therefore, the near-term size of expected premiums. Note that even with behavioral-based explanations for a premium, while it is reasonable to expect that it will be arbitraged down, limits to arbitrage (trading costs, borrowing fees, and risk of unlimited losses) can allow it to persist at a lower level. (For those interested, there is an excellent 2015 article by AQR’s Cliff Asness on this subject.)

Moreover, there is an important difference between the evidence and the headline. The evidence supports a more modest conclusion: Some reported signals weaken after discovery, especially where trading is easy and the original result may have reflected data-mining, a behavioral mispricing with limited impediments to arbitrage—as can be the case with large-cap stocks or an unusually favorable sample. The headline “factor investing is dead” requires far more evidence: that diversified, economically grounded, investable long–short factor portfolios no longer earn positive expected returns.

A Useful Empirical Counterexample

AQR Style Premia Alternative offers a useful real-world test. The fund seeks exposure to four widely studied styles—value, momentum, carry, and defensive (including quality-related characteristics)—and implements them through long and short positions across equities, fixed income and rates, commodities, and currencies. That breadth matters: It is designed as a multi-asset, market-neutral or low-beta alternative-risk-premiums portfolio rather than as a long-only equity product whose outcome is dominated by the stock market’s direction.

The following calendar-year returns for AQR Style Premia Alternative, based on Morningstar data, stand in sharp contrast to the declaration that factor investing is dead: The fund returned 22% per year over the period:

2021: 25.0%

2022: 30.8%

2023: 12.8%

2024: 21.2%

2025: 14.9%

2026 through Aug. 13: 17.0%

It is important to note that these returns were achieved with virtually no exposure to market beta—the fund provided diversification benefits, and returns were not just the result of a bull market.

That is not the return path one would expect if the underlying premiums had been competitively eliminated. Over the six observations in the table—including the partial 2026 result—the fund posted positive returns in every period. These are net returns. The R6 share class of the fund had an adjusted expense ratio of 1.42%, illustrating that the recent performance was not merely a gross return result unavailable to investors. While discussing the fund’s expenses, investors should keep in mind that a long-short fund is very different from a long-only fund. In the case of AQR Style Premia Alternative, investors are incurring 71 basis points for the long positions and 71 basis points for the short positions, with each side contributing to the fund’s performance.

Of course, six consecutive positive calendar-year observations do not prove that factor premiums are permanent, nor can one fund establish the efficacy of every factor strategy. But they are difficult to square with the categorical assertion that factor investing has been arbitraged away.

In discussing the performance of AQR Style Premia Alternative, it is important to note that, while publication of findings can lead to reduced premiums, the fund continues to benefit from ongoing research into these factors. That research has led to enhancements within the original four core themes, which themselves have remained unchanged. New research allows predictability to persist even as markets learn.

Why Predictability Persists Even as Markets Learn

As we discussed, a common objection to factor investing is that publication should erode premiums: Once a pattern is public, arbitragers trade it away. However, this doesn’t mean the well runs dry. Researchers continue to identify new sources of return predictability, partly because the tools for finding them keep improving.

Large language models are the latest example. Traditional factor research relies on researchers hypothesizing a relationship, then testing it with linear regression—a method that can only capture patterns humans think to look for and that fit a specified functional form. LLMs flip this approach. For example, the authors of the 2023 paper “Expected Returns and Large Language Models” used LLMs to extract signals from financial news and found that they predict cross-sectional stock returns. The 2023 study “Can ChatGPT Forecast Stock Price Movements? Return Predictability and Large Language Models” went further, documenting that GPT-4 could predict stock market reactions directly from news headlines without any finance-specific training, achieving roughly 90% hit rates on next-day portfolio direction and also predicting subsequent price drift, particularly for small stocks and negative news.

The mechanism matters for the efficiency debate. The 2026 study “Option Return Predictability via Large Language Models” on LLM-generated option-return factors found that the machine-discovered signals showed little self-correlation—evidence of genuine new information rather than repackaged old factors—while remaining grounded in market microstructure and behavioral finance logic, a combination that the authors argue traditional machine-learning approaches struggle to match.

The implication for investors isn’t that today’s known premiums are permanent. It’s that market efficiency is a moving target: As one source of predictability is arbitraged away, better tools uncover others. This argues for humility on both sides—skepticism toward the durability of any single anomaly, but also skepticism toward claims that markets have become too efficient for further discovery.

Why Structure Matters

AQR Style Premia Alternative is relevant precisely because it is closer to the economic object discussed in the academic literature: a portfolio that is long relatively attractive assets and short relatively unattractive ones. A typical long-only value, quality, low-volatility, or momentum mutual fund is not that object.

A long-only value fund, for example, can underperform because equities decline, because value stocks have lower market betas, or because its sector and country exposures differ from the benchmark. Conversely, it may look successful simply because the equity market rose. Those outcomes do not cleanly identify the return to the value spread itself.

A long–short multi-asset implementation is not immune to risk, drawdowns, model error, crowding, trading costs, or manager-specific decisions. AQR Style Premia Alternative itself had meaningful negative years before this recent run: 2018 (negative 12.3%), 2019 (negative 8.1%), and 2020 (negative 21.9%). But that is exactly the point. Factor premiums are risky and time-varying; they are not supposed to produce a smooth, guaranteed return stream. A difficult stretch does not establish their demise, just as a good stretch does not establish permanence.

Looking across regimes, Morningstar reports that the 10-year annualized return of the fund was 7.8%, which was about 5.5 percentage points above the 2.3% return on one-month Treasury bills. Using Portfolio Visualizer’s backtest tool, we find that AQR Style Premia Alternative produced a Sharpe ratio of 0.46. Its correlation with stocks has been close to zero, and its correlation with bonds has been low, characteristics that can provide meaningful diversification benefits in a broader portfolio.

With that said, it’s important to understand that assessing a diversifier by its stand-alone performance understates its portfolio benefit. With that in mind, I examined how the addition of AQR Style Premia Alternative affected the risk, return, and efficiency (risk relative to return) of the portfolio. To provide the longest data series possible, I used the institutional share class of the fund (QSPIX), which has a slightly higher expense ratio (1.53%). That allowed me to expand the investment period to November 2013 through July 2026. Note that investors with access to the lower-cost share class would have slightly higher returns.

I began with a standard 60% equity/40% bond portfolio. I allocated 60% of the equities to Vanguard Morningstar Total Stock Market ETF
VTI
and the remaining 40% to Vanguard Total International ETF VXUS, approximating the global equity market cap. The bond portion went to Vanguard Total Bond Market ETF BND. Portfolio 2 allocates 10% to the AQR Style Premia Alternative institutional share class, and Portfolio 3 allocates 20% to it, with its allocation coming roughly proportionally from the other three funds. The table below shows the results.

As you can see, adding the AQR Style Premia Alternative institutional share class to the portfolio not only improved returns, but it also lowered the volatility of the portfolio and significantly reduced the maximum drawdown. The result was a meaningful increase in the Sharpe ratio.

What the Evidence Supports

A defensible conclusion is not that factor investing is dead, but that naive factor investing may be inadequate. Investors should distinguish among:

• A backtested anomaly selected from a very large factor zoo.
• A long-only product with substantial unintended market and sector exposures.
• A concentrated single-factor allocation.
• A diversified, carefully implemented, cost-aware, long–short portfolio of economically motivated premiums.

The last category is what AQR Style Premia Alternative approximates. Its recent returns do not disprove the possibility that some published anomalies have decayed. They provide substantial counterevidence to the sweeping claim that publication and investor cash flows have destroyed factor investing. At a minimum, the record suggests that value, momentum, carry, and defensive premiums—combined across equities, bonds, commodities, and currencies, and implemented on both the long and short sides—have remained very much alive.

The death of factor investing has been declared repeatedly. What the evidence more plausibly shows is the death of simplistic extrapolation: Individual signals can decay, long-only funds can be poor proxies for factors, and diversification and implementation remain central to turning robust economic ideas into realized returns.