Modern portfolio theory (MPT) emphasizes that investors can diversify away the risk of investment loss by reducing the correlation between the returns from the select securities in their portfolio. The goal is to optimize expected return against a certain level of risk. According to the modern portfolio theorist, investors should measure the correlation coefficients between the returns of different assets and strategically select assets that are less likely to lose value at the same time.

Study of Correlation in Modern Portfolio Theory

MPT looks for correlation between expected returns and expected volatility of different investments. This expected risk-reward relationship was titled "the efficient frontier" by Chicago-school economist Harry Markowitz. The efficient frontier is the optimal correlation between risk and return in MPT.

Correlation is measured on a scale of -1.0 to +1.0. If two assets have an expected return correlation of 1.0, that means that they are perfectly correlated. When one gains 5%, the other gains 5%; when one drops 10%, so does the other. A perfectly negative correlation (-1.0) implies that one asset's gain is proportionally matched by the other asset's loss. A zero correlation has no predictive relationship. MPT stresses that investors should look for a consistently uncorrelated (near zero) pool of assets to limit risk.

Criticisms of Modern Portfolio Theory's Use of Correlation

One of the major critiques of Markowitz's initial MPT was the assumption that the correlation between assets are fixed and predictable. The systematic relationships between different assets do not remain constant in the real world, which means that MPT becomes less and less useful during times of uncertainty – exactly when investors need the most protection from volatility.

Others assert that the variables used to measure correlation coefficients are themselves faulty and the actual risk level of an asset can be mispriced. Expected values are really mathematical expressions about the implied covariance of future returns and not actually historical measurements of real return.