This is one of a series of columns that were produced for Moneyweb Investor in which Stuart Theobald explores the intersection of philosophy of science and finance. This followed an earlier series for Business Day Investors Monthly on the same theme. This column was first published in August 2017.
Since its launch in The Investor, this column has been a space to connect ideas from the academic world to the real world of investors. The topics have ranged from behavioural finance to studies of the impact of media on share prices. Of course, the academic world of finance was given a bloody nose by the financial crisis, and it is still trying to recover its reputation. But I believe it remains very important to continue interrogating ideas in academia from the perspective of practice and in my concluding column for this magazine, I want to provide something of a defence for academic finance.
The financial crisis was in part the result of the inappropriate use of ideas from academia. For example, academics had created several models to illustrate how financial returns are a function of risk. One important and typical one is the Capital Asset Pricing Model, which shows that particular instruments have to be priced such that the returns to owners justify the risk that the particular instrument brings to a portfolio. There is solid reasoning to the idea – it seems clear that no rational investor would want to own assets that increase the risk of a portfolio without also believing that it will increase the returns. The Capital Asset Pricing Model provides an elegant mathematical depiction of this relationship, by seeing risk as volatility in returns.
This is not unlike many other models in economics. The Hotelling Location Model, for instance, explains why companies tend to cluster in particular areas. It argues that consumers consider the price of goods and the transportation costs. Companies maximise profits by being just a little closer to the consumer than a competitor. Because every company thinks the same way, they tend to end up being in almost the same place. Usually this model is illustrated by imagining a single road. If there are four consumers spread equally along the road, who will go to whichever firm is nearest, each firm has an incentive to move closer to the other firm to eat up its market share, until both are in the middle of the length of road.
Now, no one thinks that the single road interpretation of the Hotelling Model, which also has some elegant mathematics, is a good formula for companies to use in deciding where to open their businesses, largely because cities don’t consist of single roads. It does, though, get at something important about the reason competitor retailers end up clustering together and may well support a rule of thumb for businesses in that they should open up where their competitors are. The lesson they provide is about a general tendency at work in the complex real world. In the same way, the Capital Asset Pricing Model also gets at something important about prices in financial markets, in that they must somehow reflect investors’ beliefs about the contribution to overall risks that they face. But just as real cities are not a single road, risk is not volatility. Volatility gets at something about the nature of risk, but it is not the whole story. It is useful shorthand, an abstraction, that makes the maths tractable. It helps the model do the academic job of explaining something about the messy world in an un-messy way.
The problem is the leap from academic models to practice where they become a calculating device to make decisions or to use in designing institutions. In those circumstances, the abstractness of the model is forgotten. We suddenly behave as if we live in a world where we have one-road cities and where volatility is a sufficient notion of risk. So we ended up using computers to calculate the volatility of complex portfolios held by banks and using those results to manage their risk. Fund managers used volatility and back-testing to convince themselves they have a low-risk strategy. We treat the model as if it is a faithful representation of the world, rather than an abstraction designed to illustrate something about it. We also use the model as a blue print for institutions, hardwiring them to deliver returns appropriate to the risk, when risk is measured as volatility. In doing that, the logic of the models in the academic world is turned into something quite different. In academia the models aim to fit the world, but in practice we were busy making the world fit the models. Of course, models are useful in practice too, but they are usually quite different. When we engineer new aircraft or bridges we test them in wind tunnels. We build prototypes and subject them to all kinds of stresses and strains to see how they’ll work. We often build in levels of redundancy so they’ll survive far worse than what we throw at them. We don’t take the abstract drawings and decide that because they are neat and tidy they’ll work in the real world.
We have in part learned this lesson. We now “stress test” banks in way that is vaguely analogous to testing scale models. We also sometimes run experiments on designs of new institutions to see how they’ll work, though it is difficult to get such experiments right. We still, however, confuse the logic of the academic setting. We fail to appreciate that its aim is to explain tendencies in the world, not to give us designs for how it should work. In that we remain at risk of getting lost in translation, of embracing academic work in finance but doing it in an entirely inappropriate way, leaving us with vulnerable institutions and weak risk management.