Winner of the New Statesman SPERI Prize in Political Economy 2016

Tuesday 23 April 2024

The Bernanke Review of Bank of England Forecasting


My guess is that the world is divided into two groups of people: those who are really interested in the process by which central banks decide on monetary policy, and those who are largely disinterested. If I’m right, and you are part of the second group, it may be wise to skip this blog.

As the job of central banks is to set interest rates to keep inflation close to target, and inflation over the recent past has been well above target, then a typical reaction might be that central banks have slipped up in some way. The Bank of England is a typical central bank in this respect, and it commissioned Ben Bernanke - Nobel prize winner and former US central bank (the Fed) chair - to review their forecasting procedures. His published report is here. This post is informed by that report, but really just represents some of my own personal reactions.

My first point is rather fundamental and crucial. The fact that inflation got way above target last year does not necessarily imply the Monetary Policy Committee (MPC) of the Bank or the forecasters at the Bank made any avoidable mistakes, as Tony Yates has repeatedly pointed out. As we should all know by now, the burst of inflation we saw after the pandemic was led by higher commodity prices (helped by an unforecastable war) and was exacerbated by supply chain problems that were not a reflection of domestic overheating in the major economies, including the UK. To use the jargon, both represented short term supply shocks that would lead to only a temporary increase in inflation. Other things being equal, the best monetary policy reaction to such shocks might well be to do nothing, which is pretty well what central banks in the US, UK and EZ decided to do in 2021.

For those who understand this argument, please skip this paragraph. For those who don’t, imagine if central banks had instead raised rates aggressively through 2021. This would have had only a minor impact on oil and gas prices, or supply side bottlenecks, but might well have severely weakened the major economies as they recovered from the pandemic. In a nutshell an earlier central bank reaction would have had little impact on the upward path of inflation, but might have damaged the real economy at a very sensitive moment. It would have had no impact whatsoever on the cost of living squeeze, because wage inflation would have been lower. In some situations it is best for central bankers to do nothing, and 2021 was one of those situations.

By 2022 it became clear that wage inflation was also rising, reflecting a quite tight labour market, and this could threaten to lead to inflation being above target for more than a couple of years. As a result central banks did begin to increase rates pretty rapidly. Here the jury is still out on whether central banks were too slow, too fast or went too far in doing this, so it would be premature to say that central banks in general, and the Bank of England in particular, had a problem it needed to resolve.

Which all means that Bernanke was not looking for a smoking gun which would reveal why the Bank had got things so wrong over the last few years, because it is not obvious that the Bank (and other central banks) have got anything very wrong. Instead his review should be seen as a useful periodic check by an eminent outsider on what the Bank does. (Here is a short blog post based on an earlier review.) However there is a more subtle sense in which recent events do point to a weakness at the Bank, which I will come to at the end.

There is potentially a great deal to talk about from the Review, so here I want to limit myself to just three issues. The first issue, and one where I have some historical expertise, is the forecasting model the Bank uses. The second is about interest rate assumptions in the forecast, and the third is about openness.

Reading between the lines of the Bernanke Review, it is clear that the Bank’s core macromodel, called Compass, is increasingly unhelpful in producing the Bank’s forecast. Chris Giles asks “how on earth did the BoE’s management and governance arrangements allow its modelling to get into such a bad mess?” I think part of the answer reflects a problem with academic macroeconomics, about which I used to write a lot in this blog many years ago.

The current model, Compass, is a pretty basic DSGE model. For those unfamiliar with what that means, can I suggest my own taxonomy of macromodel types here. DSGE models are dominant within academia, and they put internal theoretical consistency above matching the data. An alternative style of model, which goes by many names and which Barnanke calls ‘semi-structural’, puts more emphasis on matching the past data and as a result sacrifices being certain about internal theoretical consistency.

The predecessor at the Bank to the Compass model, called BEQM, was an intriguing and ambitious attempt to combine a DSGE model with a type of semi-structural model. I was the external advisor on that project. I think it’s fair to say that this attempt failed because it was just too complex for the forecasters and decision makers to use and understand.

The Bank’s reaction, to resort to a simpler DSGE model, was in my view the wrong choice. Instead it should have built a modern semi-structural model. The problem with DSGE models is that they don’t help forecasters very much, and are more difficult to revise to take into account new data and issues, as I explain in detail here. I cannot say why they made the wrong choice, but the view that a lot of academic macroeconomists hold that DSGE models are the only way to do structural modelling may have been decisive. My own view, shared by Olivier Blanchard, is rather different, as I explain here. In short, the Bank needs to build a new semi-structural model to replace its current DSGE model, and ensure it devotes sufficient resources to continuously updating and improving this model.

The second issue is about what interest rate assumptions the Bank should use in its forecast. At present it uses both constant interest rates and market expectations, but both of these are unsatisfactory. It is, to put it simply, just daft that the Bank doesn’t forecast the one variable it actually sets. It should follow best practice elsewhere, as I and others have long advocated. It is a shame Bernanke dodged making this a strong recommendation.

What Bernanke did recommend is ditching fan charts showing forecast uncertainty, and instead look at alternative scenarios to the main forecast. How useful this will be remains to be seen, but I think it points to a more fundamental problem. Chris Giles talks about the need to examine past forecast errors, but I think an equally big gap is the lack of Bank analysis of the impact of monetary policy itself.

If you look at the Bank of England’s website you will find plenty of interesting research papers, many of which look at a particular aspect of the ‘transmission mechanism’ between monetary policy instruments and key economic variables. A few, like this, examine the whole thing. However there is a world of difference between a piece of research using a particular estimation method or theoretical approach and a considered Bank view based on looking at the totality of evidence. It is the latter which is missing.

For example I started this post by suggesting that the Bank couldn’t have done much to prevent recent inflation without harming the economy, but this should be something the Bank itself should address by looking at what the impact of alternative past interest rate profiles might have been, and publishing the results. More generally, the Bank should publish and regularly update its own assessment of how changes in interest rates impact the economy over time. The Bank must do such analysis internally, so why not make this public? Having a core model that it could trust would of course make this much easier.

No comments:

Post a Comment

Unfortunately because of spam with embedded links (which then flag up warnings about the whole site on some browsers), I have to personally moderate all comments. As a result, your comment may not appear for some time. In addition, I cannot publish comments with links to websites because it takes too much time to check whether these sites are legitimate.