APT vs CAPM: the key differences
CAPM is a special case of APT with exactly one factor: the market portfolio. The theoretical foundation of CAPM requires assumptions that are difficult to defend: all investors hold the same efficient mean-variance portfolio, all assets are tradeable, there are no taxes or transaction costs, and so on. APT requires far weaker assumptions — only that there are no arbitrage opportunities (a condition that must hold approximately in any well-functioning market) and that returns can be decomposed into factor exposures plus idiosyncratic noise. APT does not require all investors to be mean-variance optimisers. The trade-off: CAPM specifies the factor (market beta) and derives it from equilibrium. APT leaves factor selection to the empiricist — it tells you the structure of expected returns but not which factors to use.
Common APT factor models in practice
The most influential empirical realisation of APT is the Fama-French three-factor model, which adds size (SMB — small minus big) and value (HML — high book-to-market minus low) to the market factor. Carhart later added momentum (UMD — up minus down) as a fourth factor. Today’s multi-factor models used by risk systems like Barra (now MSCI) and Axioma include dozens of factors: market, size, value, momentum, quality, low volatility, growth, earnings yield, and sector exposures. Practitioners also use macroeconomic APT models, where factors are surprises in GDP growth, inflation, credit spreads, and term spreads (the Chen, Roll, Ross 1986 model). These macro factors carry an economic interpretation — they represent the undiversifiable macroeconomic risks that rational investors require compensation to bear.
“There are three kinds of lies: lies, damned lies, and factor models.” — Finance academic paraphrase
What this means for you
APT is the theoretical foundation for factor investing and smart beta. When an ETF provider claims their "quality factor" strategy will outperform, they are implicitly making an APT argument: quality exposure commands a risk premium because quality stocks outperform in ways that market beta alone cannot explain. The challenge is that identifying true APT factors (compensated for systematic risk) versus data-mined spurious correlations is extremely difficult. The practical test: does the factor have a plausible economic story for why it represents undiversifiable risk? Value has one (distress risk). Momentum has a weaker one. The fifth decimal place of some obscure accruals ratio probably does not. When evaluating factor-based strategies, focus on turnover, capacity, and out-of-sample evidence — not in-sample factor significance.