Humans are prediction machines. We have to be. Every decision we make is based on predicting something. Whether it is the sun rising tomorrow morning, our car slowing down when we press on the brake pedal or the bank we use remaining solvent. Every choice we make is built on assumptions about the future. Despite this, I spend plenty of time annoying people by talking about how pointless (or worse) most financial market predictions are. So, are they essential or worthless? Well, it depends.

It depends because the types of prediction that investors make are wildly different – some are fiendishly complex and dynamic, whilst others are simple and stable. Before comparing some examples, let’s think about the factors that are likely to moderate the challenge posed by an investment forecast.

Our basic approach to any prediction should be Bayesian. This means that we have a set of potential outcomes that we apply probabilities to based on some starting assumptions or prior beliefs. Crucially, when we receive new and relevant information we update our priors and probabilities.

Let’s take a simplified example. I believe that there is a 70% chance that US equities will outperform European equities over the next twelve months because of stronger earnings growth (this is not true, I have no idea). Now, if US company earnings are surprisingly weak in the next quarter, I may revise down my probability.

Sounds simple, but there is a problem. While a Bayesian method is the best way to approach this scenario, it doesn’t mean it leads to good or helpful answers from an investment perspective. This is due to the forecast itself being just too difficult.

What makes a forecast hard? There are three key questions:

1) Can we define the variables that matter to our forecast? We need to be able to identify the factors that will influence the outcome of the thing we are forecasting.

The list of variables that could impact the relative performance of US equities and European equities over the next twelve months is huge and includes things that we know (such as earnings) and things that we don’t (unexpected events).

2) Is the group of variables that matter stable? The more the factors that matter change, the harder it is to make good predictions.

It is not just that the list of variables that matter is long and unknowable, but the relative importance of them is not constant. Maybe earnings will matter this year, maybe a pandemic the next.

3) How predictable are those variables? It is one thing identifying what the variables of importance are, but we still need to be able to predict them with some level of accuracy.

We don’t just need to define the factors that will influence the relative performance, we need to be able to predict them. Even if earnings growth was the most important variable in any given year, we would still need to forecast it well.

This type of forecasting is exceptionally tough, even if we take a sensible, measured approach.

All our investment decisions are a type of forecast – does this mean they are all equally problematic? No, they are not. Let’s take another example that seems similar at face value, but is an entirely different proposition.

Imagine we have thirty years to retirement and invest the majority of our long-term savings into equities. We are making a forecast here – that investing in the stock market is the best way to grow our wealth over time. What assumptions are we making in this instance?

– That economies will grow in real terms.

– That corporate earnings growth will be closely linked to economic fundamentals.

– That shareholder rights will be upheld.

Now, we might want to expand this list a little or be more nuanced, but in basic terms these are the elements that will impact our forecast. Over the long-run these are likely to be the aspects that matter, they are unlikely to change and we can predict them with reasonable confidence.

There are, of course, no guarantees. There is by no means a 100% probability that equities are the right place to be even over thirty-years – there are all sorts of remote but not impossible adverse scenarios – but the likelihood we should attach to this being accurate is far, far higher than our one-year market conjecture.

—-

As much as I dislike admitting in, we are constantly making forecasts about financial markets – it is inescapable. That doesn’t mean, however, that we cannot differentiate between predictions that are necessary and reasonable, and those that are impossibly difficult and almost certainly damaging.

–

My first book has been published. The Intelligent Fund Investor explores the beliefs and behaviours that lead investors astray, and shows how we can make better decisions. You can get a copy here (UK) or here (US).