Measured as the return of small-cap portfolios minus large-cap portfolios, the size premium was one of the earliest documented anomalies. Later work found much of it concentrated in the smallest, least liquid and least profitable names, and in January returns, which weakened the case considerably.
Small caps carry real costs that averages hide: wider spreads, thinner depth, higher borrow costs, and greater sensitivity to funding conditions. A backtest computed on closing prices can overstate an achievable premium by a large margin once slippage is charged.
Size still matters as a portfolio descriptor even where its premium is doubted, because it drives much of the difference between a cap-weighted and an equal-weighted-index, and it interacts strongly with quality-factor.
Related: factor-investing, equal-weighted-index, quality-factor, market-cap, slippage, fama-french-three-factor