This tweet from Laura Kuenssberg, the BBC’s Political Editor, was the latest example of how misunderstandings about the nature of forecasting have clouded the debate about the costs and benefits of Brexit.

Screen Shot 2018-02-08 at 21.42.50

Today [Feb8[ the BoE released its latest inflation forecast.   Unsurprisingly, there has been economic news, and the forecast has changed a bit.  Doesn’t that mean that this forecasting business, including estimating the impact of Brexit, is a bit fishy?  Can’t we therefore just set aside that economics as undecided, and focus on the cultural and political implications of Brexit?  This was the clear implication of Laura’s tweet.


The BoE’s forecast update is best thought of as a comparison between two unconditional forecasts.  Each one is conditioned on the path for market interest rates.  But that is also changing from quarter to quarter, as liquidity, the price of risk, and expectations of future rates evolve.

The Brexit impact studies are a comparison of two conditional forecasts.  In one case the forecast is conditioned on ‘stay the same’.  In the other the conditioning assumption is ‘erect trade barriers to implement Brexit’.  ‘Conditioning’ means ‘calculating what can happen under the particular assumption that one thing does actually happen.

The classic example of the difference between conditional and unconditional forecasting, whose recent ancestry is traced back to Chris Giles, is as follows.  You will find it difficult to forecast my weight in 10 years, based on information I give you about my lifestyle.  In particular because you will have trouble forecasting how my lifestyle evolves – how much craft beer and quality crisps I eat, for example.  However, you can with much greater confidence forecast how much my weight will differ, holding other things equal, comparing the cases when my beer and crisp consumption stays as it is, with the assumption [totally implausible if you knew me] that I were to give up both.

We can now update this old analogy to discuss why Laura Kuenssberg’s reference to the updated BoE inflation forecast was misplaced.

Over time, as information about my lifestyle and its evolution evolved, you would update your 10 year forecast of my weight.  And this might move around a lot.  How could you have anticipated that a craft beer shop would open on my doorstep?  Or that there would be an explosion of artisanal crisps?

But that update would not invalidate the activity of doing the forecast of the weight-reducing benefits of giving up crisps and beer, a calculation that would not change by much at all.

This seems unfathomably difficult for people to grasp.  Andrew Neil had difficulty with it in this interview with me on his Daily Politics show.  How on earth can that be an answer to my point?  He seemed to be thinking.  Iain Martin was also at it today:

Screen Shot 2018-02-08 at 22.21.13

This issue has come to be one that seems to divide Remainers from Leavers;  but this is really an issue of Leaver non-economists impugning the vast majority of economists who stand behind the impact assessments of Brexit [that deduce that it would be costly].

The very small coterie of economists who actually think Brexit would be beneficial are also articulating their beliefs [as logically they must] in the form of conditional forecasts.  It’s just that those conditional forecast comparisons are different.

For them the conditioning alternative of ‘Leave’ will involve a consequential liberation of trade with non-EU countries, and deregulation of UK business that will generate more prosperity.  For the overwhelming majority of fellow conditional forecasters, these notions are absurd.  But they are members of the same family of forecast objects that anti-Brexit economists are themselves producing.