Winner of the New Statesman SPERI Prize in Political Economy 2016


Thursday 19 October 2017

Forecast errors compared

And a coda defends experts against Aditya Chakrabortty

A recent conversation got me thinking about different types of macroeconomic forecast error, and what implications they might have for macroeconomics. I’ll take three, from a UK perspective although the implications go well beyond. The errors are the financial crisis, the lack of a downturn immediately after Brexit, and flat UK productivity.

The immediate cause of the Global Financial Crisis (GFC) was the US housing market crash, but that alone should have caused some kind of downturn in the US, with limited implications for the rest of the world. What caused the GFC was the lack of resilience of banks around the world to a shock of this kind.

Were there any indications of this lack of resilience? Here is an OECD series for banking sector leverage in the UK: the ratio of bank assets to capital. The higher the number, the more fragile banks are becoming.

UK Banking sector leverage: Source OECD

The first and perhaps most important problem with forecasting the financial crisis was that macro forecasters were not looking at data like this. For most it was not on their radar, because banks, let alone bank leverage, played no role in their models. It was a sin of omission, a big gap in our macro understanding. (Whether, if forecasters had been having to forecast this data, they would have predicted a crisis is improbable, but some would have at least noted it as an issue.)

Moving on to the second mistake, it is often said (correctly) that forecasters are very bad at predicting turning points or dramatic changes. But many did predict such a change immediately following the Brexit vote: a sharp and immediate slowdown in demand caused by the uncertainty of Brexit. It didn’t happen. The main reason was consumption, which held up by more than people were expecting, given the fall in real incomes that was likely to come from the Brexit depreciation. There are two and a half obvious explanations for this. First, because of Leave propaganda half the population thought Brexit would make them at least no worse off. Second, those who did anticipate the rise in import prices may have taken the opportunity to buy consumer durables made overseas to beat the prospective price increase. The half is that the Bank cut interest rates a bit.

None of these effects are very new. They may not have been incorporated into the forecasters’ models, but they could in principle have been incorporated using the forecaster’s judgement, although getting the quantification right would have been very difficult. In the end we got the slowdown, but delayed until the first half of this year, as Leavers began to face reality and the higher import prices came through, so it was an error of timing more than anything else (although it was apparently enough to make MP Liz Truss change her mind and support Brexit!). You could describe it as an unchallenging error, because it could easily be explained using existing ideas. It is the kind of error that forecasters make all the time, and which makes forecasting so inaccurate.

The third error was UK productivity, which I talked about at length here. Until the GFC, macro forecasters in the UK had not had to think about technical progress and how it became embodied in improvements in labour productivity, because the trend seemed remarkably stable. So when UK productivity growth appeared to come to a halt after the GFC, forecasters were largely in the dark. What many like the OBR did, which is to assume that previous trend growth would quickly resume, was not the extreme that some people suggest. It was instead a compromise between continuing no growth and reverting to the previous trend line, the second being what had happened in previous recessions.

My point of writing about this again is that I think this third error is much more like the GFC mistake than the post-Brexit vote mistake. In both cases something important that forecasters were used to taking for granted started behaving in a way that had not happened since WWII. Standard models were used to treating technical progress as an unpredictable random process. Now it is just possible that this is still the case, and the absence of technical progress in the UK and to a lesser extent elsewhere is just one of those things that will never be explained. But for the UK at least the coincidence with the GFC, austerity and now Brexit seems too great. As as I showed in the earlier post growth has not been exactly zero but has oscillated in a way that could be related to macro events.

If there is some connection, both in the UK and elsewhere, between the decline in economic productivity growth and macroeconomic developments, then this suggests an important missing element in macromodels. And like the financial sector, there is an existing body of research that economists can draw on, which is endogenous growth theory. There are examples of that happening already.

But I want to end with a plea. After the financial crisis too many people who should have know better said that failing to predict the financial crisis meant that all existing mainstream macroeconomics was flawed. It was rubbish, but such attitudes did not help when some of us were arguing against austerity on the basis of standard macroeconomic ideas and evidence. Now with UK productivity, we have Aditya Chakrabortty saying that experts at the OBR “are guilty of a similar un-realism and they have proven just as impervious to criticism” as people like Boris Johnson or Liam Fox. Not content with this nonsense, he says “This age of impossibilism is partly their creation”.

This is just wrong. Look at the elements of neoliberal overreach. Economists didn’t start calling for tight immigration controls and using immigrants as a scapegoat for almost everything. Most academic economists did not call for austerity. Almost all economists did not want to get rid of our trade agreements with the EU. Even if economists had warned about the financial crisis they would have been ignored because of the political power of finance. If all economists had thought productivity would continue flat we would have just had more austerity. [1] And in making this basic mistake, it is ironically Aditya Chakrabortty who has joined Michael Gove and other Brexiteers in having had enough of experts.



[1] Less expected productivity growth means lower future output which means lower future tax receipts which means, given the government’s austerity policy, more cuts in public spending.

6 comments:

  1. I think with Brexit you have to add that PM committed (before vote) to staying in and triggering Art50 very quickly. As he didn't, we were in a state of "not left yet and not even moved towards doing so" for some months, which again complicates the timing.

    ReplyDelete
  2. 1] Less expected productivity growth means lower future output which means lower future tax receipts which means, given the government’s austerity policy, more cuts in public spending.

    yes production is the only way to increase wealth and increase real value of tax receipts

    but without inflation at all with increased production we would not have nominal tax receipts be expected to increase, just real tax receipts

    ReplyDelete
  3. «macro forecasters were not looking at data like this. For most it was not on their radar, because banks, let alone bank leverage, played no role in their models. It was a sin of omission, a big gap in our macro understanding. (Whether, if forecasters had been having to forecast this data, they would have predicted a crisis is improbable, but some would have at least noted it as an issue.)»

    Especially huge sin of omissions as JM Keynes, H Minsky, R Kindleberger and many other influential political economists had made compelling arguments a long time ago as to the central importance of the financial system to the macro level too.

    «After the financial crisis too many people who should have know better said that failing to predict the financial crisis meant that all existing mainstream macroeconomics was flawed. It was rubbish, but such attitudes did not help when some of us were arguing against austerity on the basis of standard macroeconomic ideas and evidence.»

    But arguing against misdescribed "austerity" “on the basis of standard macroeconomic ideas and evidence” is really your own problem by choice isn't it? Why should people who think that JM Keynes and H Minsky and others had valuable insights pretty much pretend otherwise just because people who ignores those insights wanted to use arguments built “on the basis of standard macroeconomic ideas and evidence”?

    As to "productivity", “standard macroeconomic ideas and evidence” basically assume it away, by having microfoundations that give as exogenous mythological fantasies like the aggregate "production function" and "demand and supply schedule", and worse, like the expunging of "land" from microfoundations achieved by JB Clark, which amalgamated the fertility of land into the productivity of labour, which is ridiculous.

    I am sympathetic as always to the plight of our blogger, but he made his own awkwardly microfounded bed (by necessity as he tells in a previous story), and now he has got to lie in it. :-)

    Note: I don't disagree that microfounded macro would be a nice thing to do, but not all microfoundations are equally fruitful, and regrettably “current standard macroeconomic ideas and evidence” have pretty demented microfoundations.

    ReplyDelete
  4. "So when UK productivity growth appeared to come to a halt after the GFC,"

    What's that being caused by in your view? It seems to be a unique British problem. I'm rather skeptical about reports that tell a story of the British worker not getting as much done as German or French workers in one hour due to an education or skills gap...

    ReplyDelete
  5. These days there is so much data, no individual can expect to have a finger on every pulse. Dividing responsibility between teams or organisations inevitably creates cracks, and hides elements of the big picture, as models/data are simplified/filtered for presentation at higher levels of authority.
    (The Albanese et al article in VoxEU shows value of analysing base data as well as summary)

    Presumably historically someone somewhere (or could be identified) had/has primary responsibility for monitoring each of these economic indicators, like bank ratios, productivity.
    Whether regulators, Treasury, Central banks, they have had databases into which these data are collected and consolidated.
    Do they not, could/should they not have data alerts (a database construct for the last 20+ years) configured when an indicator goes out of normal variation or tolerance? And presumably some investigation has followed.
    The mere variation won't tell why, or cause, but like all data, it raises a question, which requires further deeper investigation.
    Have the necessary resources been available, or prioritised, to do adequate real time investigation?

    Going forward, one question is how best to execute that investigation in the shortest time, in order to effect remedy?

    As an ageing techie, it occurs, as big data becomes central in economics, whether/where the rapidly developing field of AI learning (sadly not my field) offers some potential to identifying these unpredicted, unprecedented, risks?

    Short of some sort of Armageddon, economies are only going to get more complicated, sophisticated, generating ever more data. It seems essential that regulators and policy makers will need to consolidate and analyse ever greater amounts of data, right down to the original transactions.
    Hypothesis: how long will humans be able to analyse the unimaginable quantities of economic data better than AI?

    As a footnote, tech innovation has a long cycle from science research, thru product development. I have seen little evidence that innovation is slowing. UK long extant problem converting science to profit is surely separate from exploitation/investment of available tech?

    ReplyDelete
  6. So, essentially Chakrabortty dribble?

    ReplyDelete

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.