Arbitrage Pricing Theory (APT)

Arbitrage pricing theory (APT) is a model of asset pricing that explains an asset's expected return as the risk-free rate plus a set of premiums for exposure to several systematic risk factors, such as unexpected changes in inflation, industrial production or interest rates. It was developed by the economist Stephen Ross in a 1976 paper and rests on the idea that, in well-functioning markets, mispricings would be removed by arbitrage.

APT is the main multi-factor alternative to the capital asset pricing model, which uses a single factor, the market portfolio. It is used by portfolio managers, risk analysts and corporate finance teams to estimate a required rate of return, understand what drives a portfolio's returns and build portfolios with chosen factor exposures.

Imagine the price of an ice-cream shop's shares. Hot weather, the price of milk and how much money people have to spend all move its profits. Each of those forces is a risk factor, and the shop is more sensitive to some than others. APT says that the return investors expect from the shop should reflect how sensitive it is to each force, and how much investors demand to be paid for bearing that kind of risk.

If two investments have the same sensitivities but different expected returns, traders could buy one and sell the other and lock in a profit. APT assumes such opportunities get competed away, which is what keeps prices in line with risk.

The APT relationship for asset $i$ is:

• $E(R_i)$ is the expected return on the asset.
• $R_f$ is the risk-free rate.
• $\beta_{ij}$ is the asset's sensitivity to factor $j$, estimated by regression analysis.
• $\lambda_j$ is the risk premium investors demand per unit of exposure to factor $j$.
• $k$ is the number of factors.

Suppose the risk-free rate is 3% and a share has three factor exposures: inflation ($\beta$ = 0.8, premium 2%), GDP growth ($\beta$ = 1.2, premium 3%) and interest rates ($\beta$ = 0.5, premium 1%). Expected return is 3% + (0.8 × 2%) + (1.2 × 3%) + (0.5 × 1%) = 3% + 1.6% + 3.6% + 0.5% = 8.7%. If the share's price implies an expected return well above 8.7%, APT suggests it is cheap relative to its risks.

Ross's theory does not say which factors to use, which is both its strength and its weakness. A widely cited 1986 study by Nai-Fu Chen, Richard Roll and Ross used macroeconomic variables, including industrial production, unexpected inflation, changes in expected inflation, the gap between low-grade and high-grade bond yields, and the slope of the yield curve. Practitioners also use statistical factors derived from return data, and fundamental factors such as company size and value, as in the Fama-French models of the 1990s. Commercial risk models used by asset managers apply dozens of industry, country and style factors in the same spirit.

Both models agree that only systematic risk should be rewarded, because company-specific risk can be removed through diversification.

APT-style models are used to estimate the cost of capital, explain why a fund outperformed or underperformed, hedge exposure to particular factors and build factor-tilted portfolios. Researchers have applied multi-factor models to crypto assets, testing factors such as the overall crypto market, size, momentum and network activity. Results have been mixed, partly because crypto return histories are short and markets have changed quickly.

APT's flexibility means two analysts can choose different factors and reach different answers, and factor premiums estimated from history may not hold in future. Estimating sensitivities requires long, stable data, which many assets lack. The assumption that arbitrage keeps prices in line can fail when trading is costly or risky, as studies of the limits to arbitrage show. Despite these weaknesses, the multi-factor view introduced by APT is now standard in investment practice.