How can AI help to solve the problems with MPT?

Modern portfolio theory (MPT) is a hypothesis about investment theory that Harry Markowitz published in 1952. Since that time, Markowitz’s theory has been one of the most influential forces in finance for both academics and practitioners. Markowitz asserted that risk-averse investors could construct portfolios of assets that maximize return for a given level of risk. The application of MPT allows investors to create an optimal portfolio of assets for any particular level of risk. Depending on the individual’s risk tolerance, they should invest in the return-maximizing set of assets. This set of choices for each level of risk creates the efficient frontier, and investors should never choose a portfolio of assets that does not fall along that frontier. In addition, MPT quantified the benefits of asset diversification through risk reduction.

What are the assumptions of MPT?

Like any theory, MPT is based on a number of assumptions. Without these assumptions, the model does not hold true. So, the more that the assumptions deviate from actual observations, the weaker the theory becomes. It is always good to know the assumptions underlying any model as well as its limitations. While MPT is the standard for investment and financial planning decisions, some investment professionals have noted that the market does not always behave according to modern portfolio theory. Specifically, the model assumptions sometimes break down and don’t match actual market behavior.

In order for Markowitz’s model of efficient frontiers and risk-return payoffs to hold true, the following market assumptions must also be true:

· Investors must consider each investment alternative as a probability distribution of expected returns over an investment horizon.

Unfortunately, investors are driven by recent returns and short-term expected returns. Even if there is a high probability of long-term return, investors have a hard time sticking with an asset through a period of negative returns.

· The risk of a portfolio is measured as the variance of expected return.

The problem with these measurements is that they utilize historical data that may not reflect future risk potential. In addition, upside variance is weighted the same as downside variance. In reality, investors care far more about unexpected loss than they do about unexpected gain.

· Expected returns follow a normal probability distribution.

Experience proves that stock returns don’t follow a normal probability distribution. Stock returns have a probability distribution that includes fat tails, which the normal distribution does not have. If returns were normal, market crashes would not exist.

· Investor utility curves are only a function of risk and return.

While it is true that investors are motivated by risk and return, other motivations also factor into their investment choices. For instance, personal preferences and social beliefs can impact investments. The existence of mutual funds dedicated to social and environmental issues proves this point.

· Investors are rational.

Investors are humans, and humans are not always completely rational beings. Investors are prone to overreacting (or irrational exuberance) and a variety of other choices that may not be described as rational. The growth of behavioral finance research is an attempt to understand investor actions that do not follow rational models. For example, Daniel Kahneman’s work on Prospect Theory and Richard Thaler’s work on behavioral economics show that people don’t make rational decisions. People make decisions based on a variety of personal biases and are primarily motivated by the short-run rather than the optimal final outcome.

How can AI help to solve these shortcomings of MPT?

Recognizing the shortcomings of MPT and its assumptions doesn’t mean that investors should forget the model entirely. Today’s technology simply offers investors a new way to calculate the inputs and evaluate the data. Financial technology and AI provide an opportunity to improve theories about risk, return, investor behavior, and asset allocation.

· By observing and learning from past behavior, AI forecasts future market behavior. Rather than using historical data as a proxy for future market returns and volatility, AI uses this data to understand the market and create a predictive model of returns and volatility.

· AI not only predicts market movements but also how investors will react to those movements. Even when investor behaviors don’t seem to exhibit rational expectations, there are still patterns in behavior that AI can recognize and predict.

· Joel Greenblatt’s method of formula investing is just one of the ways to overcome human emotions and the propensity to make decisions based on short-run biases. AI is a more advanced way to accomplish this goal. This technology overcomes the problem of poor investment decisions that result from fear, greed, and other emotions.

Just because there are some problems with the assumptions of MPT doesn’t mean it is time to throw away the theory. Leveraging AI insights perfectly complements MPT and is the ideal way to use it. TrueRisk Labs employs AI technology to predict stock prices and volatility. Incorporating this data into asset allocation models can be part of the process for truly modernizing Modern Portfolio Theory.