Risk management appeared as a new area of knowledge 25 years ago, causing profound changes around the world. Nowadays, universities have professors and courses specialised in risk, companies have risk management areas, and risk is regularly discussed by boards of directors. Central banks and securities and exchange commissions have risk areas and a slew of recent regulation is based on risk. The purpose of risk management is to deal with future uncertainties in order to minimise severe potential losses. How did the art of risk management reach this status?
The first step on this journey came when individuals’ attitudes changed regarding future uncertainties. A situation is uncertain when it has different possible outcomes. If the gods determine the future, there is no uncertainty to be managed. It took thousands of years to exchange oracles’ answers for assessments made by the individual themself. Sumerians, Greeks and Mayans did not take this step. The new approach emerged in Renaissance Europe.
Renaissance hearts and minds were open to accepting uncertainties and capturing them through the development of probabilities. Games of chance posed the first challenges. What is the probability of two sixes from two dice throws? In this intellectual endeavour, the Italians Luca Paccioli, Girolamo Cardano and Galileo Galilei were the beacons.
In the middle of the 15th century, probability left the gaming rooms when Frenchmen Piérre Fermat and Blaise Pascal built the pillars of the systematic method for calculating probabilities. Out of this already available theory, demographics and the insurance sector arose in the 1700s.
Brilliant mathematicians such as Pierre-Simon Laplace and Carl Friedrich Gauss continued to develop probabilistic theory. Several types of probabilistic models were drawn up to determine rules of association between events and their likelihood of materialisation, based on the frequency of their occurrence in the past. By the 1900s, the financial market had become the object of probabilistic models. Management of the future was based on the hypothesis that the future would repeat the past exactly. At the end of the 19th century, the Victorian world was perfect, stable and predictable. It was believed that any probability could be measured.
World War I buried those certainties. Albert Einstein and Sigmund Freud were not the only bringers of radical changes. In the 1920s-30s, two economists, Frank Knight and John Maynard Keynes, turned down the hypothesis that the future always repeats the past exactly. This hypothesis, which had prevailed for some centuries, was valid for roulette but not for economic dynamics, whose agents are always looking ahead and trying to adapt unprecedented conditions. The probability theory is useful but not sufficient: the decision-making process needs to count with subjective assessments. The world had become quantum and unstable. The measurement of probabilities in the social sciences proved to be more difficult than supposed.
The first economic model to include uncertainty in decision-making was proposed in 1944. Some years later the risk appetite concept was developed. To this point, uncertainty was treated in an abstract way, with no figures. In 1952, Harry Markowitz and Arthur Roy were the first to add figures to risk, the harmful part of uncertainty. They proposed different formulae. Markowitz defined risk as volatility, the variability of frequent losses and gains. Roy defined risk as severe and rare loss. The concept of volatility prevailed for nearly 60 years thanks to its simplicity.
A group of economists then made a significant advance in the 1980s. Although it is not possible to foresee the return on an asset tomorrow, they built probabilistic models to foresee the interval within which tomorrow’s return would be based on the recent history of returns.
The theory described above resulted from the work of six Nobel Prizewinners in Economics and of John von Neumann, a colleague of Einstein’s and Kurt Godel’s at Princeton’s Institute for Advanced Study.
In 1993, a manual entitled RiskMetrics set out a detailed methodology for predicting the volatility of a large and complex portfolio of financial securities based on a probabilistic model. The manual contained no new theory. It did, however, combine several of the ideas described above in a very practical and implementable way. Risk management was born.
The predictive ability of RiskMetrics proved to be so good that it caused a global microeconomic revolution. Financial firms and regulators across the globe rushed to adopt it. It did not take long for non-financials to join the race. At the start of 1997, the methodology was extended to credit risk portfolios. It appeared that risk had been tamed.
In the end of 1997, the Asian Crisis made it clear that risk had not been reined in. Losses substantially surpassed the methodology’s predictions. Banks and companies went bust. The fatal error had been to choose volatility as the risk measure. Broadly speaking, the usage of volatility in risk monitoring assumes that crises and extreme returns do not exist. Roy was right: risk is severe and rare loss.
The immediate solution for coping with severe losses was to develop stress tests – methodologies that seek to generate substantial losses without a probabilistic stamp. Extreme returns on a single asset were successfully modelled at the start of the millennium. Meanwhile, extreme returns on a portfolio of assets were modelled from the lessons learned from the 2008 crisis. To price real estate derivatives in a simple way, a sophisticated model had been restricted by the assumption that different debt issuers would never default simultaneously. This unrealistic simplification was one of the sparks in the early phase of the crisis. This simplification was discarded and the subsequent model was popularised. The risk measurement of large and complex portfolios henceforth contained models capable of dealing with simultaneous losses of several of their assets.
There has been a rich learning curve over the past 25 years. Governance is essential. Risk appetite must be unequivocally determined. Risk management must be based on the subjective judgment of experts with a foundation in historical knowledge, useful data and probabilistic models. It must consider a past that may be repeated similarly in the future, a past that might have happened but did not, and one which never happened but which plausibly might.