Size and the Small-Cap Premium: The Evidence after Removing Junk and the Case for Small-Cap Value

What you will learn

• How "small" is defined in research and in indices.
• Why the stand-alone small-cap premium looks weak after the 1980s and what "controlling for junk" changes.
• Why small growth has been the worst corner of the size–value grid and small value the strongest.
• The extra costs and risks of small-cap exposure.

Explanation

The size factor is the historical tendency of small companies, measured by market capitalisation, to earn higher average returns than large ones. Rolf Banz documented it for US shares in 1981, and Fama and French built it into SMB in 1993. Of all the factors in this course, size has the most disputed evidence.

First, definitions. In the Fama–French data, "small" means below the median market capitalisation of New York Stock Exchange companies, which captures far more listed companies than market value. Commercial indices cut by cumulative free-float capitalisation: as of 2026, the MSCI investable market indices treat roughly the top 85 percent of each market as standard (large and mid) and roughly the next 14 percent as small cap; check the current methodology because thresholds change. A "small-cap" fund therefore holds companies worth hundreds of millions or billions of euros; the micro-caps that drive many academic results are mostly excluded.

Second, the evidence. After publication the raw size premium weakened considerably; in many samples from the mid-1980s onward it is statistically indistinguishable from zero, and it was concentrated in January and in the smallest, least liquid stocks. Several researchers concluded that size was not a robust factor at all.

Asness, Frazzini, Israel, Moskowitz and Pedersen responded in "Size Matters, If You Control Your Junk" (2018). Their argument: the small-cap universe contains many low-quality companies, unprofitable, highly leveraged, with volatile earnings, which they call junk. Junk earns low returns in every size group and is over-represented among small caps, which drags the small-cap average down and hides the size effect. Comparing small and large companies of similar quality, the size premium became larger, more stable across decades and countries, and no longer concentrated in January. In short: small and high-quality has historically beaten large and high-quality; small junk has not beaten anything.

The same logic explains the small-cap value result. Sorting stocks by size and value into a grid, the small-value corner has historically shown the highest average return and the small-growth corner the lowest. Small growth is where speculative, loss-making, lottery-like companies cluster: the junk again. Small value avoids most of them. This is why many practitioners recommend a small-cap value or small-cap quality tilt rather than a plain small-cap index, and why a plain small-cap fund can disappoint for years even if the size effect is real.

Third, costs and risks. Small caps have wider bid-ask spreads, lower volume and higher market impact. They fall more in recessions and credit crunches because they depend more on bank financing. Small-cap indices also have higher turnover from companies graduating to mid cap or being delisted. These are reasons the premium might exist, but they are real costs whether or not the premium arrives.

Worked example

Hypothetical example: compare the spread cost of a large-cap and a small-cap fund. Assume an average bid-ask spread of 0.05% for the large-cap basket and 0.40% for the small-cap basket, one-way turnover of 30% per year for both, and that each trade costs half the spread, so a round trip costs one spread. Commissions, impact, fees and taxes are excluded.

• Large cap: 30% × 2 (sell and buy) × 0.025% half-spread = 0.015% per year.
• Small cap: 30% × 2 × 0.20% = 0.12% per year, eight times more.

Now add quality. Suppose an invented sample shows these average annual returns before costs: large quality 9%, large junk 7%, small quality 12%, small junk 5%. A plain small-cap index with 60% quality and 40% junk by weight shows 0.6 × 12% + 0.4 × 5% = 9.2%, barely above large quality. Like with like, small quality beats large quality by 3 points while small junk trails large junk by 2 points: the size premium is visible only after junk is held constant. These numbers are invented to show the mechanics.

Common mistake

Practice activity

A hypothetical small-cap fund holds 1,000 companies; the unprofitable ones make up 25% of its weight. In an invented back-test, profitable small caps returned 10% per year and unprofitable ones 3%, both before costs. (a) Calculate the fund's weighted return. (b) Calculate the return of a version that excludes the unprofitable companies. (c) Name one cost or risk the exclusion does not remove.

Key takeaways

• "Small" means different things in research and in index funds; commercial small-cap indices exclude most micro-caps.
• The stand-alone size premium weakened after publication; controlling for quality (removing junk) restores a larger, more stable premium.
• Small value has the strongest record in the size–value grid; small growth the weakest.
• Small caps carry real costs and risks whether or not a premium shows up.

Next step

The next lesson looks at low volatility and minimum variance: how such portfolios are built, why they behave a little like bonds, and why lottery-seeking investors may be the reason they exist.