Having worked for three decades in both health care and finance, I have noticed several parallels where the shortcomings of one field are reflected in the other. Doctors are criticized for treating the symptoms of an illness rather than the underlying disease. Similarly, investors tend to treat the symptoms of portfolio risk by focusing on the historic returns of the underlying assets. With the advent of big data and artificial intelligence, a new field of diagnosis has emerged called systems medicine, where a wide range of factors are simultaneously analyzed to make more informed decisions. Systems economics aims to apply a similar concept to finance.
A fundamental tenet of investing is that stock prices are typically driven by the earnings and revenue growth of the company. It should follow that the fundamentals of companies with similar business models will experience similar trends and shocks (e.g., automobile manufacturers are exposed to the price of the raw materials, such as steel or aluminum and their customers are largely exposed to the price of oil). In other words, they share related business risks, and since earnings ultimately drive prices, their returns should be correlated. That is to say, correlation is largely a symptom of exposure to a shared business risk.
Managing risk is a cornerstone of institutional money management and institutional investors use sophisticated risk models such as Barra or Axioma to calculate ex-ante predictions of value at risk or find optimized weights based on risk budgets and other assumptions. These systems almost always require predictions for the correlation among the assets involved. Traditionally such forward looking correlation forecasts are derived from historical stock returns.
Though this is the industry standard, it is dangerous for many reasons. As any financial disclaimer will tell you, past returns are not a guide to future returns. Though this warning is clearly designed to highlight the danger of extrapolating trends when calculating an expected profit, it extends to other uses of historic returns, like predicting volatilities or correlations. As with the medical industry, these risk models do not take into account the cause of the correlation. They are merely addressing the symptom (the covariance of realized returns) instead of the cause (related business risk).
In systems medicine, a wide variety of data is analyzed simultaneously by computers in order to make more accurate diagnoses. This approach is powered by a robust classification system for grouping together similar patients based on genetic, demographic and environmental data. But, until recently, economics has lacked a methodology to classify business risk. Systems economics is a relatively new field of finance that is gaining traction with several asset managers. Analogous with systems medicine, it seeks to identify and control the root cause (bias) to correct the whole system, rather than merely treating the symptoms of the problem (e.g., diversifying weights based on correlations and volatilities).
The key to this approach is a standardized method for classifying the various business risks (biases) similar to the approach used in systems medicine. One such approach is FIS, or functional information system, which uses functional tags to describe the customer groups, supply chains and product types for all of a company's business lines. For example, according to the most widely used Global Industry Classification Standard, Amazon is considered a consumer discretionary stock. However, Amazon also has significant technology exposure, which comes with an entirely different set of investment risks. FIS tags both Amazon's technological function (as an online platform) and its consumer retail function (as a retailer of consumer goods), allowing users to view both aspects of Amazon's (related business risk) business model (Table 1).
