Octavian Report: What do you mean by radical uncertainty, from an economic decisionmaking perspective?
Sir Mervyn King: What my co-author John Kay and I mean by radical uncertainty is, in essence, uncertainty that cannot sensibly be quantified.
We’ve all grown up with the idea that if you are intelligent, you think about uncertainty in terms of probabilities, and there are many people who will try to interpret any kind of future uncertainty in terms of some probability. I think this is a serious mistake and it detracts from good decisionmaking.
Two examples illustrate this very clearly. One is the financial crisis and the other is the COVID-19 crisis. In both cases, we knew something. We knew that banking crises could occur. We’ve seen many in the past. We knew that pandemics could occur. But there was no way that we could have said: “What’s the probability that there will be a virus emanating from Wuhan in China in December 2019?”. It didn’t make sense to ask that question.
What it made sense to do was to say: “We can’t anticipate precisely when or where, but we know that it’s quite possible that there will be a virus coming from a country and that will spread to the rest of the world. So what is the right basis for being prepared or ready for this kind of event?” In the financial crisis, I think what you saw very clearly was that people relied on what we call bogus quantification. Recall the famous statement coming out of Goldman Sachs that what we were seeing was 25 standard deviations several days in a row — literally virtually impossible. What we saw was that the model that they had been using was wrong. It just didn’t apply. We live in a world which is non-stationary, to use the technical phrase: things keep changing. It’s not simply changing in the sense that you toss a coin and you sometimes get heads and sometimes get tails and you can’t tell in advance which you’ll get, but you know that on average it’ll balance out at 50/50. This was something where you just didn’t know what would happen. To pretend that we can quantify it is potentially dangerous because it makes you feel that you can somehow manage these uncertainties in the future by pricing them if they’re financial or by making accurate quantitative predictions — when in fact we can’t.
OR: Given what the economic policy response to COVID needs to be, what do you think will be its ramifications? Will we be able to get out from under the debt? Do you see inflation coming as a way of getting out from under that? What do you think the future of money is in general?
King: I think the only honest answer to those questions is I don’t know. But I’ll make a few observations, which make it easier to think about it, I think. The first is that one of the arguments for being prudent about public finances in normal times is precisely to be able to allow national debt to rise very sharply and to use government borrowing to deal with an unexpected event that you can’t easily imagine and certainly can’t predict. This is the essence of resilience in managing the national finances. I think we’ve been quite good at that in much of the Western world. I think within the European Union there are serious challenges for those countries that have suffered under a monetary union, with Italy being the most obvious example, but Greece too.
For the U.S. and the U.K., it is perfectly sensible to deal with the pressure on the health services by saying: we will do whatever it takes to try and keep businesses still in operation so that when we get out of this, they will still then be able to function and the market economy can resume its normal job. That means just being prepared to borrow whatever you need to borrow it. I’m not worried about the scale of the increase in national debt as such because properly managed, we would then come out of this and say to ourselves: We have a much higher national debt now. We need to allow the ratio of national debt to national income to fall slowly over time so that we get a good run of years, and then it’ll be right back to where it was.
We don’t need to repay it straight away. We can just slowly let the ratio of debt to national income fall, and the best way to do that is through economic growth. That I think is all manageable. Whether or not we are able to do that will depend on two things. One is the ability of the measures that have been announced to work in practice on the ground. I don’t think we are short of high-level policy announcements or various schemes to help. What we need to worry about is whether they can actually be delivered.
If we look to the future and ask: will we get inflation? I think the answer to that does not depend on the particular types of measures that central banks will take. You often hear phrases like “helicopter money” or other phrases about expanding the practices that central banks have already engaged in through quantitative easing to print money. The method used to expand the amount of money in the economy is not the issue here. What is crucially important is who makes the decision about how much money gets printed. If we come out of this crisis in a situation where governments feel that central banks have been doing lots of things but their really important function here was to act as an agent for government to deliver many of these specialist schemes and therefore the government would like to have greater control over central banks, if they carry that through to having greater control over all central bank functions, then I think inflation will indeed be a genuine concern.
The jargon that economists use for this is fiscal dominance. That is: the government decides it would like to spend a lot of money, and it then tells the central bank that it has to print enough to finance that. I don’t think this is a problem today, but if we were to come out of this and if the independence of central banks in two or three years’ time in respect of setting interest rates and the amount of QE they engage in — if that is determined by pressure from government rather than by an independent central bank with a mandate for price stability, then I think we ought to be worried about a future inflation. If central banks can retain their independence, then I don’t think there’s any reason to suppose that we should find inflation a major problem.
OR: Do you worry about an emerging markets crisis with possible knock-on effects globally?
King: I think this is a real risk. Go back to the end of 2019, before the pandemic: I was certainly worried that we had not really solved the problem of the disequilibrium into which the world economy had tipped before, during, and indeed after the financial crisis, and that the amount of debt in the world as a whole was higher last year than it was in 2007. My concern was that there might be a small number of defaults, which in themselves were not very significant, but nevertheless were a signal to the market that there could be debt defaults. Then you would see sentiment deteriorating around the world, rather like what happened in September 2007, and that would immediately put pressure on the capital positions of many financial intermediaries and banks around the world and that could trigger another crisis.
Some countries in emerging market economies did not start this episode with comfortable levels of debt. Their levels of debt, especially in foreign currency, could have caused a problem in any event, and this will exacerbate it. That could be the trigger for the start of a wave of defaults, which would lead to a wider crisis. The banking system in the United States is a good deal more robust than it was before the financial crisis, and it’s also a good deal more robust than the banking system in most other parts of the world. Certainly in Europe, I would worry that the banking system is not in good shape and would find it difficult to withstand the significant number of defaults.
Countries in the developing world, which have for very understandable reasons weak health systems, are going to be in an extremely difficult position. The impact on their economies will mean they’re not in a good position to cope with significant expansions of national debt. I do think that many of these countries could suffer far more even than we in the West are suffering and they won’t be in a strong position to withstand it, and we will see many more defaults on debt than we would have expected to see only a few months ago. In turn, that could be a trigger for a very wide financial crisis.
But the whole spirit of what I’ve been saying, and what John Kay and I write about in our book, is that we need to avoid bogus quantification. Don’t predict things, be very honest about what you know and what you don’t know, and be prepared. Think about resilience of systems.
OR: Do you see governments issuing very long-term debt as a potential solution here? Wouldn’t that be a way of limiting risk going forward? Do you have a sense as to why that’s not being done?
King: I think it does make sense to issue much longer-term debt. I think one of the slightly odd features of national debt across the world is that for historical reasons, countries like the United Kingdom have a much higher duration of our national debt than most other developed countries. I think that’s a sensible position to be in because it minimizes the rollover risk of having to refinance debt at much higher interest rates than you can now — why not lock into these low interest rates? The other side of the coin, however, which may be affecting policymakers is the question of who’s going to buy this debt. One of the concerns that I was certainly worried about before this episode was that pension schemes which were buying long-term debt at very low interest rates would be viable only if they could charge much higher contributions than their members have been used to. We had seen the demise of defined-benefit pension schemes, and even for defined-contribution schemes. The amount of money that people had to put into their pension pot in order to have adequate pensions was becoming extremely high. Pension schemes as such were not looking terribly attractive at very low interest rates. What we do not want to see are pension funds, pension schemes, and indeed insurance companies ending up in a position where they too move towards insolvency because they can’t finance the liabilities they’ve incurred.
OR: What is the genesis of your (very timely) book?
King: John and I had written a book together 40 years ago on the British tax system, and our careers then went in rather different directions. John became a very successful business consultant. He was first head of the business school at Oxford and a well-known contributor to the Financial Times. I became an academic and I was for the first half of my career, and then I went to the Bank of England as chief economist and stayed on as deputy governor and then governor for 10 years. At the end of my time as governor, we had a conference at the Bank of England and John and I both made contributions to it. We discovered that quite independently, we had come to the view reflecting on the financial crisis and what people had said about the 1930’s.
The only way to make sense of these events was to do so in terms of what we call radical uncertainty. The idea that risk could be priced by the use of probabilities had led people into a false position in which they felt they could genuinely price all risk. It turned out in the financial crisis that it’s not just that the risk model is used by banks were useless because the models didn’t apply anymore, but the calculation of risk weights used by regulators of banks also turned out to be a very poor guide to which banks would be in serious trouble and which not. That’s because they had believed that they could use past data to capture and quantify the risk and price it — and they believed that somehow if you could price it, you could tame risk.
John and I felt this was quite wrong, and so, coming from different directions, we decided that we would join forces again and write another book on radical uncertainty. We learnt a great deal in writing that book. We give lots of examples in the book about the nature of uncertainty when you can’t easily quantify it. The example we like best — and we use it at the beginning of the book — is the challenge that President Obama faced when he was in the Situation Room and was asked to decide whether to send the Navy SEAL team in to the compound in Abbottabad in Pakistan for the capture of Osama Bin Laden. Everyone knew there was a person living in the compound. The question was: was that person Bin Laden?
Now, there are only two answers to that question, yes or no. Interestingly, after the Iraq War, Congress basically mandated the CIA when giving advice to the President to do so in probabilistic form. They had to give probabilities for these things. The different members of the CIA team when briefing the President said, “Well, I think it’s 90 percent certain that the man is Bin Laden.” Someone else said, “I think it’s only 60 percent likely.” Someone else said 70 percent and someone else said 40 percent. At the end of this — and we know this because President Obama gave an interview afterwards — President Obama said, “Look guys, this is 50/50. I can’t decide on the basis of your probabilities.”
Now, he did not mean by that that his own judgment was that it was 50 percent certain that it was Bin Laden. That only makes sense if you think that there are 100 episodes like this and 100 Bin Ladens to capture, and 50 percent of the time the guy turns out to be Bin Laden and 50 percent of the time he doesn’t. This was a one-off, unique situation. President Obama said that probabilities confused the situation rather than helped it because the only honest answer was: we don’t know. He had to make that decision knowing that there was no answer to the question. What he did was to probe the individuals from the CIA and asked them, “Why do you think it might be Bin Laden?” He tried to accumulate the evidence and then reach an overall judgment, but expressing it in terms of a probability did not in any way help him to reach the decision. We think, and we give many examples in the book, that lots of decisions in business and finance are exactly of that kind. They are unique decisions. They’re not simply a replay of past events when you can just carry over observed frequencies of what’s happened in the past to tell you what will happen in the future.
OR: Why do the ideas of quantification and of modeling based on it have such a hold on us?
King: I think it’s because if you believe literally in the model as a description of the world, you can turn what is in fact a mystery into a puzzle. Puzzles have well-defined solutions. Chess or Go or other games are puzzles, as are crossword puzzles. We know what the answer is at the end.
Most decisions are not like that. You may not know after you’ve made your decision, or even several years later, whether you did make the right decision. The answer is never fully revealed. I think there is this great temptation of decision-makers and politicians to believe that they must convey a degree of certainty far greater than that which they can possibly have — and also a matching wish on the part of the public to believe in that degree of certainty.
We all want to be certain about things. I think that’s a terrible mistake.
We have to recognize that some situations pose problems to us that are completely wrapped up with radical uncertainty. There is no simple answer, but there is an intelligent way to go about finding the information you may need to make a good decision.
One of my favorite examples in this is that, early on, during the spread of AIDS in southern Africa, the World Health Organization built large demographic models of the different countries in southern Africa and from these they built a very complex regional model. It was, in the end, almost a black-box model about how to predict the spread of AIDS in Southern Africa.
One of the variables in their model was the average number of sexual contacts per person per year, and they put a number in, turned the handle, and got out predictions for the spread of AIDS. The most sensible thing they did was to ask Bob May, our former chief government scientist here in the U.K. He said to them, “Look, stop turning the handle and thinking that you’ve got a model that predicts. You have a variable here — the average number of sexual contacts per person per year. Don’t you realize that it makes an enormous difference weather that is 100 contacts with the same person or 100 contacts with 100 different people? You get absolutely different answers.”
The use of the model was not to turn the handle and get a prediction, the use of model was: let’s go out and do some research on people’s sexual behavior because that’s what we need to understand in order to make a prediction about the spread of AIDS. Models can be very useful, but they’re not useful because they are literally a description of the world. They are useful for two reasons. One is they give you deep insights into what is going on. As with COVID-19, the general shape of an epidemic, the fact that it starts slowly, then accelerates rapidly, reaches a peak and comes down: it’s very important to understand that in formulating public policy.
The second way in which models are very useful is because they tell you what information really matters. For example, in COVID-19, we need to know how many people have been infected. If all you do is to say all that matters is the big black-box model and you stopped thinking about what’s the key number here, what really matters, then what’s going on here is the question that we keep coming back to time and time again in the book. Once you stop thinking like that and rely on the model as a literal description of the world, then you end up doing bogus quantification and making big mistakes.