Overcoming the challenges of forecasting

By Alexis Calla

At this time of the year, experts in the financial industry are busy trying to predict what will happen in the coming 12 months. Having a sense of how things will evolve is key to the decision-making of active investors.

The challenge is that people are not the unemotional, rational decision-makers often portrayed by economics, especially in an environment that is highly complex and probabilistic. We think too fast and take shortcuts. This is not our fault, but a result of the evolutionary process. For most of history, the number-one requirement was to survive – to secure access to sufficient food and not become something else’s fodder. To do that, one had to decide fast. Investing requires a less instinctive, more considered approach.

 

Too much information often means too much biased information

20 years ago, before the internet revolution, getting access to information was often difficult. Today, the world is producing exponentially more information than ever before. So one of today’s main challenges is working out how to process that information and make decisions. In trying to adapt to this new reality, our experience at Standard Chartered over the years has led us to believe the key to making better decisions is embedding diversity in the process – of input, perspectives and decision-makers.

Human evolution has created a brain that suffers from many behavioural biases. For instance, one of our weaknesses is the need to find a narrative when trying to explain why something has happened. Financial experts are asked every day to explain why the market has risen or fallen, and they build mental models, or shortcuts, to help them do it. The theories that result, in an always-changing, complex and probabilistic environment, are very difficult to test and definitively confirm or refute.

Confirmation bias reinforces experts’ mental models of the world as they unconsciously discard information that doesn’t fit their explanation and over-emphasise information that supports it. Therefore, we believe that individual experts, who are as biased as the rest of us, are far from being the most accurate forecasters. However, their views are invaluable when integrated into a diverse set of information.

 

How can diversity help?

Let’s take an example of how diverse perspectives may help a group of individuals to come up with a better forecast. Today, one of the biggest questions is whether the United States economy could go into recession in 2019. An expert economist may start to answer this by painting the picture of a “normal” seven-year cycle, which might involve a rebound from recession (stage 1), full-blown recovery (stage 2), inflationary growth (stage 3), and recession (stage 4).

The economist has then simplified the question to a matter of trying to determine what stage of the cycle we are in and when we might transition through the stages. For instance, many might argue we are currently in stage 2 and are thus waiting for inflationary pressures to build until we move into stage 3, at which point the Federal Reserve will tighten policy aggressively and push the economy into stage 4.

But this is just one way of looking at the problem. An expert on financial markets might prefer to focus on market-based indicators for predicting recessions. They do so on the premise that looking solely at economic data, which is released with a significant lag, has not delivered great forecasting results in the past.

Given that equities, corporate bonds and commodities often weaken before the economic data, such experts generate models based on causality they have experienced, which would have “predicted” the majority of recessions historically, in the hope that they will help predict future recessions. The problem with this method is it sometimes produces false positives. For example, in early 2016 many such models predicted a recession that did not eventuate.

Another, more recent approach is to avoid mental models entirely and “let the machine decide” by plugging data into a computer. Data-science experts can then let the computer identify which variables are the most important in predicting the probability of a recession. This approach is historically called “data-mining”.

An alternative approach is to cluster economic and financial-market variables and try to match the current environment with another period in history. This purely quantitative approach avoids relying on economic theory and lets the data determine what part of the cycle we are in.

However, this machine-learning approach comes with its own challenges. It would, for example, be unable to incorporate recent changes in the structure of the world economy or geopolitical order, which are particularly relevant today.

 

Decision-making in practice

The point of this analysis is not to say one approach is better than another, but that they all have some merit. We believe there is great value in being aware of all of them and trying to mitigate the individual and group biases of our investment committee. Here are three examples of how we try to improve the quality of our decision-making process on the basis of this understanding.

First, we try to break complex questions into several simpler ones. This has two benefits: Simpler questions are easier to answer, and errors made in answering different questions may offset each other. Both these factors are likely to make the answer to the complex question more accurate.

Second, we try to use anchoring bias – the human mind’s subconscious tendency to rely on the first piece of information given when making decisions – to our advantage in forecasting by trying to identify smart anchors as a starting point. These smart anchors, also called “the outside view”, generally come from a quantitative analysis of historical data, which is constantly refined and updated.

Third, we ensure investment-committee participants present and debate both sides of an argument to reduce the impact of confirmation bias.

As we worked on our own report on the outlook for 2019, we followed this nuanced approach to decision-making. Ahead of our committee discussions, we collected views and insights from as many diverse sources as possible. This is based on our core belief that diversity should be embraced, as it forces us to challenge our mental models and become better forecasters, both individually and collectively.

(The writer is Global Head of Investment Strategy, Advisory and Discretionary Portfolio Management at Standard Chartered Private Bank.)