Not for the first time in his life, Alan Budd has blurted out a truth:
1. The OBR shouldn’t provide point forecasts at all, but simply probability ranges. So, for example, it would say: “there is a 50-50 chance of PSNB being between £134bn and £164bn in 2010-11, and a 25% chance of it being less than £134bn and of being more than £164bn.”
This would, of course, require the OBR to teach journalists about the fact that uncertainty is a large and ineradicable fact about the future. Having seen Mervyn King bash his head against this brick wall for years, I wouldn’t fancy taking this job.
2. It could take the ego out of forecasts. Rather than present a single forecast, it could present a range of forecasts that arise from competing models of the economy, and discuss why they vary. This is the sort of job the ESRC’s macroeconomic modelling bureau did, until it was shut down.
The problem here, though, is that one important model would present government borrowing as (largely) a passive response to private sector saving and investment decisions. Would the government really want the OBR to draw attention to this possibility?
3. Fiscal policy could become less dependent upon forecasts. One possibility here is to make the automatic stabilizers more powerful. For example, if marginal tax rates were higher, taxes would fall more in a slump and rise more in a boom, so fiscal policy would stabilize activity without the use of government discretion. This though, might run into the problem, that there is a trade-off between short-run stabilization and promoting long-run growth.
An alternative is to have a state-dependent rule for tax rates - of the form, “tax rates will rise by x percentage points if GDP growth is above y, and fall by x percentage points if growth is below z.” Of course, this means fiscal policy would only respond to the reality of booms and slumps, rather than to expectations thereof, but it’s not obvious how much worse this would be than existing arrangements.
The big problem with this is rather that it takes (some) discretionary fiscal policy out of the hands of government.
I don’t know if there is a “politically acceptable” solution here. But let’s be grateful to Alan Budd for raising an important issue - that, in fiscal policy, there is something inadequate about a “predict and control” mentality, when predictions are so prone to go badly awry.
I of course wasn't in my right mind when I took this job. No-one in their right mind would take a job where your success is going to be judged on your success in producing fiscal forecasts.This raises the question. What can be done to ameliorate this problem? I can think of three possibilities.
1. The OBR shouldn’t provide point forecasts at all, but simply probability ranges. So, for example, it would say: “there is a 50-50 chance of PSNB being between £134bn and £164bn in 2010-11, and a 25% chance of it being less than £134bn and of being more than £164bn.”
This would, of course, require the OBR to teach journalists about the fact that uncertainty is a large and ineradicable fact about the future. Having seen Mervyn King bash his head against this brick wall for years, I wouldn’t fancy taking this job.
2. It could take the ego out of forecasts. Rather than present a single forecast, it could present a range of forecasts that arise from competing models of the economy, and discuss why they vary. This is the sort of job the ESRC’s macroeconomic modelling bureau did, until it was shut down.
The problem here, though, is that one important model would present government borrowing as (largely) a passive response to private sector saving and investment decisions. Would the government really want the OBR to draw attention to this possibility?
3. Fiscal policy could become less dependent upon forecasts. One possibility here is to make the automatic stabilizers more powerful. For example, if marginal tax rates were higher, taxes would fall more in a slump and rise more in a boom, so fiscal policy would stabilize activity without the use of government discretion. This though, might run into the problem, that there is a trade-off between short-run stabilization and promoting long-run growth.
An alternative is to have a state-dependent rule for tax rates - of the form, “tax rates will rise by x percentage points if GDP growth is above y, and fall by x percentage points if growth is below z.” Of course, this means fiscal policy would only respond to the reality of booms and slumps, rather than to expectations thereof, but it’s not obvious how much worse this would be than existing arrangements.
The big problem with this is rather that it takes (some) discretionary fiscal policy out of the hands of government.
I don’t know if there is a “politically acceptable” solution here. But let’s be grateful to Alan Budd for raising an important issue - that, in fiscal policy, there is something inadequate about a “predict and control” mentality, when predictions are so prone to go badly awry.
"that, in fiscal policy, there is something inadequate about a “predict and control” mentality, when predictions are so prone to go badly awry."
Something of a death knell for detailed Keynesian demand management then really, isn't it?
Posted by: Tim Worstall | July 21, 2010 at 04:59 PM
What is the point of giving ranges and percentage probabilities? It's useless. I could do it. Because whatever the outcome I will have assigned a possiblity to it, and am therefore covered.
What will GDP growth be next year? I predict a 50% chance of it being in the range 1.75%-2.25%, with a 25% chance of being lower, and 25% chance of being higher. So on a one off basis (which each year is) I'm never wrong.
Now if the same year could be repeated a number of times, my prediction could be tested. If GDP growth for 2011 could be run 100 times, and 50 times it was in my range, 40 times lower, and 10 times higher, then I'd have been wrong. The outcomes did not match my predictions.
But on a one off basis it is useless.
Posted by: Jim | July 21, 2010 at 10:18 PM