If and when an insolvency event occurs, it should be a risk you didn’t see coming. Perhaps it’s my insurance and pension background but I find it absolutely extraordinary that regulators can allow a situation to happen where a blatantly obvious risk like interest rate mismatching threatens the solvency of a bank.
Is it a privileged position to be in to be able to say I never forecast, and I think all forecasting is a mugs game? Surely we can accept forecasts are absolutely necessary as a basis for decision making about the future?
As an actuary, all I do all day in work is forecast. In general, the wise strategy is to minimise the use of your own judgement where possible. You don’t want your neck on the line, and to be in a position of having to defend your own judgement. So what do you do? Where possible you rely on market data. And when there’s none, you rely on the expert judgement of others, and hopefully a variety of others to establish a range against which to benchmark. And with the remaining areas of uncertainty, where you have to choose something sensible yourself, you sensitivity test to show what the forecast might look like if you’re wrong. And ultimately, to some degree, all sources will be in some way wrong.
But what can you do in the absence of a forecast? Base decisions on a hunch, or on the assumptions that the future will be exactly like the past?
I often heard during the pandemic armchair commentators bemoan how wrong the COVID models were. A model developed at speed using very little data in an environment where both the virus and human behaviour was continuously changing, making past data less and less relevant. Yet nobody slagging off the modellers could tell me what was a better substitute for the model - other than their own opinion of what they themselves would prefer be done (which differed from person to person).
In the case of US regulation by SVB, two things seemed to have happened. The regulators were restricted in terms of what they could regulate - Trump era rolling back of Dodd-Frank laws. That rolling back was inspired by lobbying, including, famously, by SVB itself. Second, the regulator didn’t use the limited powers that it had available to it. ‘Asleep at the wheel’ is a phrase I have seen used.
Your remarks about forecasting are well made. They actually raise all sorts of hard questions mostly ignored by economists, statisticians and econometricians. Keynes tried to wrestle with all this in his ‘Treatise on probability’, which was more philosophy than maths. How to make decisions under conditions of uncertainty? When we pretend to divine the future we can get into trouble. Of course, we need to make assumptions about the future to enable decisions to be made today. That should lead us towards decision rules that give ‘good enough’ outcomes over a range of plausible scenarios or assumptions about the future. And stress testing for when our plausible scenarios turn out to be arseways.
It’s really about intellectual honesty. Often, the most precise we can be is to say ‘I believe x is more likely to occur than y’. Too often, we then start enumerating, making up probabilities when we don’t have anything like the necessary knowledge to do so.
Modelling is incredibly important because it forces us to make explicit our beliefs about how the world works. Epidemiology models are based on models first developed in the 1920s. So-called SIR models. They’ve gotten a bit more sophisticated in recent years, but they aren’t that difficult. Playing around with them enables understanding of just how sensitive disease spread is to the key parameters. As with all models, particularly non-linear ones..
All models are wrong. But they are all we have got. My beef with economic forecasting is that too much attention is given to things like “GDP to grow by 3%”. With no acknowledgement that the most likely outcome is not 3%, or work done to provide contingency plans for when the forecast turns out to be wrong -which is the most likely outcome.
If and when an insolvency event occurs, it should be a risk you didn’t see coming. Perhaps it’s my insurance and pension background but I find it absolutely extraordinary that regulators can allow a situation to happen where a blatantly obvious risk like interest rate mismatching threatens the solvency of a bank.
Is it a privileged position to be in to be able to say I never forecast, and I think all forecasting is a mugs game? Surely we can accept forecasts are absolutely necessary as a basis for decision making about the future?
As an actuary, all I do all day in work is forecast. In general, the wise strategy is to minimise the use of your own judgement where possible. You don’t want your neck on the line, and to be in a position of having to defend your own judgement. So what do you do? Where possible you rely on market data. And when there’s none, you rely on the expert judgement of others, and hopefully a variety of others to establish a range against which to benchmark. And with the remaining areas of uncertainty, where you have to choose something sensible yourself, you sensitivity test to show what the forecast might look like if you’re wrong. And ultimately, to some degree, all sources will be in some way wrong.
But what can you do in the absence of a forecast? Base decisions on a hunch, or on the assumptions that the future will be exactly like the past?
I often heard during the pandemic armchair commentators bemoan how wrong the COVID models were. A model developed at speed using very little data in an environment where both the virus and human behaviour was continuously changing, making past data less and less relevant. Yet nobody slagging off the modellers could tell me what was a better substitute for the model - other than their own opinion of what they themselves would prefer be done (which differed from person to person).
In the case of US regulation by SVB, two things seemed to have happened. The regulators were restricted in terms of what they could regulate - Trump era rolling back of Dodd-Frank laws. That rolling back was inspired by lobbying, including, famously, by SVB itself. Second, the regulator didn’t use the limited powers that it had available to it. ‘Asleep at the wheel’ is a phrase I have seen used.
Your remarks about forecasting are well made. They actually raise all sorts of hard questions mostly ignored by economists, statisticians and econometricians. Keynes tried to wrestle with all this in his ‘Treatise on probability’, which was more philosophy than maths. How to make decisions under conditions of uncertainty? When we pretend to divine the future we can get into trouble. Of course, we need to make assumptions about the future to enable decisions to be made today. That should lead us towards decision rules that give ‘good enough’ outcomes over a range of plausible scenarios or assumptions about the future. And stress testing for when our plausible scenarios turn out to be arseways.
It’s really about intellectual honesty. Often, the most precise we can be is to say ‘I believe x is more likely to occur than y’. Too often, we then start enumerating, making up probabilities when we don’t have anything like the necessary knowledge to do so.
Modelling is incredibly important because it forces us to make explicit our beliefs about how the world works. Epidemiology models are based on models first developed in the 1920s. So-called SIR models. They’ve gotten a bit more sophisticated in recent years, but they aren’t that difficult. Playing around with them enables understanding of just how sensitive disease spread is to the key parameters. As with all models, particularly non-linear ones..
All models are wrong. But they are all we have got. My beef with economic forecasting is that too much attention is given to things like “GDP to grow by 3%”. With no acknowledgement that the most likely outcome is not 3%, or work done to provide contingency plans for when the forecast turns out to be wrong -which is the most likely outcome.