The future is not like the present. I don’t mean this in the trivial sense that things will change. I mean it because we should think of the future in a different way to how we think about the present.
The present is a collection of facts, some of which we know. For example, as I write this, the FTSE 100 is at 6547; latest ONS data show that GDP didn’t grow in the last quarter; and so on. The present comprises one state of the world.
The future, however, is not a single state of the world. It is instead a collection of probability distributions. (Note that I’m writing about the future in the present tense: this is because it already exists, as a mental state.) When we think of a future date (say next December) we should not think that the FTSE 100 will be at (say) 7100 or inflation at (say) 1.7 per cent, but rather that both will be within ranges – and quite large ones. Forecasts should be distributions, not precise numbers.
Thinking of the future in this way serves some useful functions.
One is that it helps explain stock market volatility of the sort we saw last week. To simplify, imagine there are just two possible future states of the world. In one, the economy is doing okay and the FTSE 100 is at 7800. In the other we have a disaster in which the FTSE 100 is at 5000. The current value of the index should be equal to the probability-weighted average of these two scenarios. So if the probabilities are 90 and 10 per cent, respectively, this gives us an index of 7520 – which is around where it was in mid-January.
Since then, however, two things have happened. One, obviously, is that the high-probability scenario has deteriorated. Measures to contain the coronavirus, such as the quarantining of some Italian and Chinese towns and closure of Japanese schools will depress global output and profits, justifying a lower level of share prices.
But something else has happened. The probability of an economic disaster has risen: it’s possible that the coronavirus will last a long time and so do a lot of economic damage. So, say we now face a central scenario of FTSE being at 7200 with a 70 per cent probability and a 30 per cent probability of it being at 5000, we get a current level of the FTSE 100 of a bit over 6500 – which is where it is as I write.
Now, I’ve given a simplified and stylised example. But it captures some key facts. One is that market volatility can be driven not just by changes to the central scenario, but also by heightened chances of disaster. Another is that we can see big falls in prices, even if a benign outcome (in this case of FTSE being at 7200) remains most likely. And a third is that there is not necessarily anything irrational about big swings in prices. Cardiff Business School’s David Meenagh and colleagues have shown that past stock market volatility is consistent with investors attaching changing but reasonable probabilities to things that might have happened but did not.
This also helps explain why we should hold cash and bonds even though they offer losses in real terms in the central scenario. It’s because this scenario isn’t all that matters. Even in today’s depressed stock markets there is a chance of increased fears of recession and a further loss of appetite for risk. In such scenarios, bonds and cash would hold up well.
Our portfolios must be resilient not just to the central scenario, but to the entire probability distribution – and in particular to the small probability of disaster. Hence the case for cash and bonds.
Now, in saying that the future is a probability distribution I do not mean to say that it is a normal distribution. It’s not. The FTSE 100 fell by 11.1 per cent last week. Based on weekly changes since 1988, this is a five standard deviation event. If returns were normally distributed, that’s the sort of thing we should see only one week in around 70,000 years. But it’s not even the worst drop in the past 12 years – that came in October 2008. Instead, as Harvard University’s Xavier Gabaix has shown, extreme returns are distributed as a cubic power law. This says that last week’s drop was the sort of thing we should see around one week in every 13 years.
Nor am I saying that probabilities can be known. They often cannot be. They are instead hunches or gut feelings. It is accurate – if not precise – to say that the market fell last week because investors’ gut feel that something nasty is possible intensified.
And herein lies a danger. As Richard Bookstaber showed in The End of Theory, we sometimes face “radical uncertainty”, in which we just cannot know whether the past is a guide to the future.
This matters, because we have known ever since Daniel Ellsberg’s work in 1961 that people hate unknown probabilities more than they hate known risks.
Which is another reason for last week’s falls. Because we cannot know the spread or duration of the coronavirus, or the size and duration of its economic impact, radical uncertainty is just what we face. Because investors hate this, they have dumped assets that are sensitive to it. True, this means equities will bounce back if or when such uncertainty recedes. But nobody knows if or when this will happen.
Does all this mean futurology is useless? Not necessarily. It’s a way of testing hypotheses. For example, when I forecast earlier this year that the All-Share index would rise by around 4 per cent this year, it was a test of the hypothesis that lead indicators give us some predictability of returns over longer periods.
For many other purposes, however, futurology is redundant and sometimes downright misleading. Investors should use it as little as possible. What matters instead is that we have portfolios that can withstand the shocks we will inevitably face.
