Radical Uncertainty with Professor John Kay CBE

Eminent British economists Professor John Kay CBE and former Governor of the Bank of England Lord Mervyn King argue in their timely book Radical Uncertainty that the future is largely unknowable and we have unwarranted confidence in our ability to predict the future. Recently, Adept Economics Director Gene Tunny discussed Radical Uncertainty with Professor John Kay on his Economics Explained podcast:

Radical Uncertainty with Prof. John Kay CBE

For the transcript of the conversation, read on.

Professor John Kay Podcast Transcript

Gene Tunny  0:08  

Welcome to the Economics Explained podcast, my name is Gene Tunny. Radical uncertainty is the title of a new book from two of Britain’s most eminent economists. I’m privileged to be interviewing one of these eminent economists this episode. My guest is Professor John Kay CBE, FRSE, FBA and Fellow in Economics at St. John’s College, Oxford. Professor Kay was appointed Commander of the Order of the British Empire CBE in the 2014 New Year’s Honours for Services to Economics. His other books include Obliquity, The Truth About Markets and Other People’s Money. His latest book, Radical Uncertainty: Decision Making for an Unknowable Future, was co-authored by the former Bank of England governor Mervyn King. Professor John Kay, welcome to the podcast.

Professor John Kay  1:06  

Good to be with you, Gene.

Gene Tunny  1:09  

Very good. John, I’m really grateful for your time. When I saw this book in Dymocks here in Brisbane, I grabbed it immediately because I’ve read books by both yourself and also Mervyn King in the past. It was so timely too because it was just at the time that we were learning about Coronavirus, but we hadn’t introduced the social distancing measures yet, so we hadn’t had the full economic impact. And I thought, well, this is a very important book, a very timely book, and I know you did make one comment about pandemics in your book, which you’ve pointed out to in a recent blog post on your site. So, I suppose you probably didn’t expect that your book would be introduced at this time in history, but given that it has been, it’s great and I think that it could help us get through this time. I’d like to explore that this episode. But to start with, it would be great if you could please explain what you mean by radical uncertainty and how it is different from what economists might call risk or what you label as resolvable uncertainty.

Professor John Kay  2:34  

If we go back a century, economists, notably Keynes and Frank Knight, made a distinction between risk and uncertainty, and risk was what they thought could be described probabilistically and uncertainty was what couldn’t. And that distinction between risk and uncertainty got elided by economists in financial markets over the next 50 years so that in the last part of the 20th century, people treated risk, uncertainty, and indeed volatility as if they all meant the same thing. Our argument is that you can’t do that. It’s certainly unhelpful to do that. And we need to go back and rethink our categories in this sense. Now, the way we talk about it, is we say that uncertainty arises because our information is imperfect. And that’s what makes the world uncertain. We sometimes don’t know what the present situation is, we certainly don’t know what the future situation will be. Now, there are two ways you can try and resolve uncertainty. One is if you don’t have enough information, you can go out and get more. People may be uncertain what the capital of Philadelphia is, but if you are uncertain, you can look it up and find out. So a lot of uncertainty you can reduce, if not resolve, by getting more information. And then there’s the kind of uncertainty that arises when you have something random that arises from a probability distribution. That’s where people talking about probabilities started; for games of chance. If you toss a coin, or you draw some cards from a pack, or you spin a roulette wheel, then there’s a known probability distribution and a known list of outcomes. That’s resolvable uncertainty in the sense in which we mean it. And you could extend that to talk about things like mortality, things like motor accidents and so on, some of the risks that are easily insurable, you can describe them probabilistically. But what people have tried to do in the last 50 years, is to extend this kind of probabilistic reasoning to absolutely everything. And you can’t do that. Now radical uncertainty isn’t Black Swans, and some people have talked about Coronavirus as if it were what Nassim Taleb calls a black swan and something that you couldn’t have anticipated. It’s not. As you politely mentioned in the book, we talk about pandemics. And we talk about them exactly as an example of radical uncertainty. We knew, or we had high confidence in believing, that there would be a pandemic of this kind, sometime, someplace, but to say it’s going to break out in Wu Han in December 2019, is not something anyone could have anticipated. We know something about pandemics, but we don’t know enough to make predictions. And a lot of the world is actually like that.

Gene Tunny  5:55  

Okay. So that’s a good place to go off wrong. I think that’s a good point. A lot of the world really is like that. The economy and business, I know that those are examples you gave in your talk at LSE, a couple of months ago. So you’re saying that this fundamental uncertainty or this radical uncertainty, which applies to pandemics because we just don’t know just how bad it could be. So, I remember, I think was it last month, before we had the restrictions everyone was starting to worry and we had everyone stockpiling here in Australia. I don’t know, you probably had the same thing in Britain where everyone’s rushing out and buying toilet paper and pasta, and we just didn’t know how bad it would be and it must be very challenging for politicians. So, from what I understand, you’re saying that this uncertainty applies across the economy as well across business. We can’t nicely define probabilities and it’s not as if we’re taking these bets on outcomes and we know what the odds are – it’s not that at all. It’s a lot more uncertain than that.

Professor John Kay  7:10  

We start the book by contrasting the achievement of, say, NASA who were able to launch a space probe into space to orbit mercury. It took that probe seven years to reach mercury. They had got it to rotate the earth, then rotate around Venus, rotate several times around mercury before it moved into the position. After seven years, it went into orbit more or less exactly the position they’d predicted seven years earlier. Now, why could they do that? They could do that because we know what the equations of planetary motion and rocket motion are. We know the system, we know the model. Secondly, the model remains unchanged over time. And thirdly, the model isn’t affected by what we think about it. In people, the transit of Venus isn’t affected by your opinion or my opinion about what it is. And these things are true of a lot of physical processes. They’re not true, however, of most of the processes we talk about in business. But that’s why prediction in economics is a very limited value and will remain so.

Gene Tunny  8:40  

Yeah. I was just thinking that with the Coronavirus, one of the things that has been unpredictable is just how the public will respond. In Australia, we had people still going to beaches in large numbers until they closed the beaches. So even though the politicians were saying, “You need to take this seriously”, and, “We need to socially distance”,  people were still a bit blase about it all. So, whenever you’ve got humans involved, you can get a great deal of unpredictability and uncertainty. I’d like to ask about the question that you see as fundamental, which I think is a good question, but it’s deceptively simple. Or maybe I’m thinking it’s deceptively simple. It’s this question about what is going on here. Why is that so important? And the follow up to that is do you think our policymakers were slow to figure out what was going on with the Coronavirus at least in the countries that have been most badly affected such as Italy and the US and possibly to a lesser extent than the US, of course, the UK

Professor John Kay  9:54  

You know, you’re right that “what is going on here” sounds like an extremely banal kind of question. And yet as we thought about these issues we came to see that was more and more important. But let’s take an example. It was one of the examples that prompted Mervyn to be interested in writing this book. Before 2008. What you had was this massive growth of trade among banks in securitized products. What was going on there? Why was there this explosion of trade? And if you ask why people trade in risky assets, there are roughly two reasons. One is they may be trading because one party can bear a risk more effectively than another. And that’s why we take out insurance policies and things like that: we are passing risks to people who are better able to take them. And the other is that you can pass risks to people who understand less about the risks that you do. You’re essentially passing them off onto other people. Asking the question, what was going on here, in that run up to 2008? Why is there this growth in trade? Would have been fundamental to asking, will this process end well or badly? Is it a way of reducing and minimising the costs of risk-bearing? Or is it a way in which you’re concentrating risk among people who don’t know much about what they’re doing? And in 2008, we learned it was the latter. Now, for Coronavirus, we need to ask what is going on here. And we need to understand what is going on here better than we do. Now, we can build quite good epidemiological models. What we don’t know in these models is the values of the key parameters. The two key parameters are if someone has the virus, how many people do they pass it on to, and if someone contracts the virus, what is the likelihood that they will die or become seriously ill as a result. Now, we still don’t know, even roughly, what these parameters are, and we don’t know much about what the incidence of infection is in the underlying population. Now, interestingly, just in the last couple of days, we have plans in the UK to do some random testing over the next year of people to try and establish what these parameters actually are. And if you do that random testing, combined with a kind of contact tracing that was done in Singapore and Korea, then you can get a pretty good picture of what is going on here. The interesting thing is that the costs of obtaining additional information are pretty small, relative to the cost of making policy on the basis of poor information. That’s why asking and understanding what is going on here is so important and key to determining sensible policies.

Gene Tunny  13:11  

Absolutely. I like that example you gave about the lead up to the financial crisis. I was in the Treasury here in Australia in the lead up to that crisis and then through part of that crisis. And I remember that when the crisis hit, arguably, because no one was expecting across that magnitude in financial markets, we had come to believe this story that these financial market participants are all consenting adults, they all knew what they were doing, it was all about diversifying risk. It was a bit of a blind spot. And so we had to scramble to just try and figure out what on earth was going on. And yeah, just what the problems were. And you know what those solutions could be, so I think that’s a really good example. Okay.

Professor John Kay  14:07  

Right. And it’s certainly what made Mervyn realise the importance of these kinds of questions, and asking at an early stage this, “What is going on here?” kind of question didn’t really happen in the run-up to the crisis. And it was only gradually during the crisis that people started to figure it out and start and make, well, we could hardly make sense of it, but to the extent that you could understand it, you could begin to make sense of it.

Gene Tunny  14:39  

Yes. I want to ask now about one of the core concepts in your book. This is the idea of non-stationary processes. Now, am I correct in saying that, if we look at what’s been happening in the US with the additional 26 million people filing for unemployment insurance, is that an example of a non-stationary process? Is that a good example of what you’re talking about there, John?

Professor John Kay  15:12  

It certainly is, but it’s an extreme one. The stationary process classic is something we were talking about earlier, something like the motion of the planets. We know what these equations are, and they are the same today as they were 500 years ago. Now, the underlying equations that might describe a model of the economy, even if you set aside Coronavirus, they aren’t the same this week as they were last week. The nature of the underlying processes is constantly changing. And that’s what makes it so often misleading to try and represent these things. probabilistically. Obviously, what you describe in terms of the rise in the US unemployment figures, is, in a sense, an example of this because if you were to write down a probability distribution of unemployment data, this is so far out as to be, you know, one of these 25 standard deviation events. And of course, there aren’t 25 standard deviation events. What these extremes tell you is the kind of distribution was just not a very useful way of thinking about this kind of event. 

Gene Tunny  16:28  

Absolutely. And that’s the sort of thing you would never be able to forecast or project from a model I am guessing. I’d like to go to your discussion in the book about modelling, and this is something that’s close to my heart because as a former Treasury economist, and as someone who does consulting work I have built and do build a lot of economic models and what you write in the book, and I’ll quote from page 261, is that “you cannot derive a probability or a forecast or a policy recommendation from a model”. What does this mean for all the economists who are trying to understand an economic problem? They’re trying to forecast the impacts of a policy or an investment project? What does this mean for them? Should we just give up? Or is there a way forward? Could you please elaborate on this, John?

Professor John Kay  17:30  

There is a way forward, and let me describe what it is. And I should say, I spent the first part of my life running a consulting business where we made most of our money selling models to people. It was quite important that I came to realise that in the main, people didn’t use these models actually to make decisions, they used these models in order to justify decisions they had already made. Either to other people within their organisation to external agencies or regulators. And that is actually right because a lot of this kind of economic modelling is saying, we are in a world of radical uncertainty, we don’t have all the information we need, so we will make lots of information up. And that’s what enables you to fill in all the cells in your spreadsheet and arrive at an answer. But that answer doesn’t tell you anything about the real world. And we use in the book, the example of the 25 standard deviation event. This was David Viniar, of Goldman Sachs, who said in 2007, as the crisis was breaking, “We’ve experienced 25 standard deviation events several days in a row”. Now, of course, that wasn’t what had happened. It’s simply impossible, if you know anything about statistics, to have 25 standard deviation events several days in a row. In order to talk about a probability in the world, you have to multiply the probability derived by the model by the probability that the model is true. But you don’t know what the probability that the model is true is, you just know that it’s really very low. It’s a misunderstanding of the way in which you use models, I’m still very much in favour of using models, but you use models not to make statements about the real world, but to get insights into the world. So, the great thought experiments like prisoner’s dilemma, or the lemons model, none of these things are as it were literally true. What they are, they’re ideas, they’re insights, which once you’ve understood them, help people worldwide in a range of practicums. That’s, to my mind, what makes a good economic model.

Gene Tunny  20:06  

Okay, but we should still try to estimate the impact on households of a tax policy change or the impact on different industries of climate change policies. You’re not saying we should give up on that, we should just be realistic about the believability or the actual accuracy of those predictions.

Professor John Kay  20:32  

They’re not predictions or forecasts, they’re illustrations. Take the model of the moment, these epidemiological models of Coronavirus. As I said earlier, they’re critically dependent on basically three parameters: what proportion of the population are infected, how many other people does an infected person infect, and finally, what’s the incidence of serious complications when people are infected. As I said, we don’t know any of these parameters. We should, therefore, do two things: one is we should devote quite a lot of effort to finding out better information about these parameters, and the second is we should use hypothetical numbers in these calculations to see what the range of possible outcomes might be. So the models say, if we did this, what kind of effect might it have? It’s very likely that doing that kind of exercise, you will learn this policy might help, this policy is not likely to help.

Gene Tunny  21:38  

Yes, yes, that’s, that’s good, John. I might quote from you again, you made a great point in that LSE seminar from a couple of months ago. You said that we need to think of economic models as being parables rather than as true models of the world and I think that’s what you were essentially saying then. The other thing I’ll point out, I’ll put it in the show notes too, I think you’ve got some great lessons about using models appropriately in your chapter, The Use and Misuse of Models. And you have first: deploy simple models, which I like because you can spend a lot of time building a hugely complicated model, and, I mean, there are diminishing returns to it out there. And in fact, you could end up with a worse model.

Professor John Kay  22:28  

Yeah, absolutely.

Gene Tunny  22:30  

Absolutely. Yes.

Professor John Kay  22:32  

We’re talking about the gigantic cost-benefit model, which is used in the UK to appraise transport projects. And if you look from the web, that model will give you answers to questions like, what will be the average number of passengers in a car one afternoon in 2036? Or what will the British growth rate be in 2080? Exercises, making up numbers like that, are just ridiculous. And what becomes worse is when people do as they sometimes do these kinds of Monte Carlo simulations, in which you make up a different set of numbers or lots of different sets of numbers, and you tell people that the distribution of outcomes of all of these different estimates is a probability distribution of what will happen. Of course, it isn’t. And the extraordinary thing is that, to my mind, anyone would ever think that it was.

Gene Tunny  23:38  

Yes, yes. Okay. So, I’d like to go on to, just finally, what all this means for decision making and how we design our systems. And if I’m reading your book correctly, what I think you’re writing in that book is that when we’re making decisions, use the models, but don’t use them to tell us what to do, they help us make the decision. Ultimately, we’ve got to form an understanding, a narrative, of what’s going on here and use that to guide the decisions.

Professor John Kay  24:20  

Yep. So we talk about formulating a reference narrative, which is in broad qualitative terms, what you’re aiming to make happen. And you want to deal with radical uncertainty by formulating that kind of reference narrative, and then ensuring that your reference narrative is robust and resilient to things you are not going to be able to predict. Again, it takes us back to this Coronavirus issue. How do we make complex systems robust and resilient? Typically what engineers do in complex systems like that is they build in modularity, which means that you reduce the interdependencies in the system, so that part of the system can fail without the whole system screwing up as a result of it. That’s modularity. Redundancy is when you make everything a bit stronger than it needs to be to cope with these unanticipated contingencies. Now, that’s interesting when one looks either at the full crisis, or at the way business more broadly has behaved in the last couple of decades, because these kinds of things, redundancy and modularity, have been regarded as a bad thing, signs of inefficiency, but actually, they make people’s businesses more robust to events they can’t anticipate.

Gene Tunny  26:12  

Yes, yes. Very good point. And I should ask as I’ve been asking a lot of my guests on this podcast about climate change, which until Coronavirus was the big policy issue to discuss. And I think I picked this up from the seminar that you and Mervyn had at LSE: If we think about responding to climate change, we should be thinking about buying options because there’s this massive fundamental radical uncertainty and it could be beneficial to give ourselves the option of responding early, and then if it turns out we don’t need to respond, then we can wind back that policy setting. I think I’ve mangled that up. But is that one of the lessons from your book?

Professor John Kay  27:14  

Again, it’s about robustness and resilience. Not a huge amount of effort has gone into the climate change area. It’s developing models, which are examples of what we think of as all these made-up numbers to fill in all the spreads cells of your spreadsheet. We can’t forecast what’s going to happen to the temperature in 2100. Well, what we can do is understand better these processes, and what we most of all, we need to go back to these words of robustness and resilience to ensure that within a plausible range of what might happen, given that we can’t anticipate where it’s going to be, our policies and our economies are robust and resilient to alternative developments. And for me, I think we should be spending a lot more money on underlying ecological research, both about how we can reduce the greenhouse gas problem and how we can find methods, particularly of transmission and storage of energy, which enable us to reduce the carbon content of our economies by substantial amounts.

Gene Tunny  28:34  

Yes, yes. That’s a great point to end on. Professor John Kay, I really appreciate your time. I’m going to be recommending everyone buys a copy of your book. I’ve been mentioning it to friends and colleagues of mine, I think it’s very timely. It’s a great review of all the theory and evidence relating to decision making over at least the last hundred years. You mentioned Frank Knight, you mentioned Maynard Keynes, Savage, and Friedman. And yes, you talk about the financial crisis and I’m sure if it came out a bit later you would have had a big section on Coronavirus but maybe that’ll be in the revised edition. I’m sure you’re being asked to comment a lot on this issue right at this time. So, again, I really appreciate your time, john, thank you.

Professor John Kay  29:40  

I enjoyed talking with you. Thanks.

Gene Tunny  29:41  

Thank you.

Free Economic Updates

Subscribe to receive our monthly economic e-newsletter, packed with the latest insights, hints and tips from our leading economists.
  • This field is for validation purposes and should be left unchanged.