Sunday 26 June 2016

Brexit predictions


One interesting feature of the Brexit (Leave) campaign was the role of experts. Expert opinion was almost uniformly in favour of Remain. The scientific community was particularly strong for Remain, fearing less collaboration between European scientists and permanent loss of funding.

Great. A year or two from now we will be able to test the accuracy of expert opinion. Or, perhaps, five years from now. Naturally, some commentators believe their judgments can be confirmed this weekend, on the basis of a sudden drop in the value of the pound, and a two day drop on the London Stock Exchange . Premature. The FTSE 100 finished the week up more than 2pc after recovering from a 8.7pc drop. Next week will be exciting, but not conclusive.

One bit of expertise we can check immediately is the accuracy of the public opinion polls. The consensus was Remain 52% Leave 48%, and the result was Remain 48% Leave 52%. Not brilliant, to say the least. What was the problem?

Brexit polls

Here is a typically poll result: Earlier today, Ipsos MORI published a poll for the London Evening Standard, which shows Remain in the lead on 52%. It was carried out from June 21-22 - right up to 9pm last night - and surveyed 1,592 people.

Can you spot the problem?

Here is another clue: A survey by ComRes for the Daily Mail and ITV News gave Remain a 6-point lead over Leave - on 48% compared to 42%. But the phone poll of 1,032 people on June 17-22 also underlined the confusion among voters as 11% of people said they were still undecided.

You may be worrying about the undecided voters. OK, third clue: The final YouGov survey for The Times, which quizzed a massive 3,766 people online on June 20-22, had each side in a dead heat with 45%. Yet far more of the 8% who were ‘don’t knows’ said they would probably end up picking Remain - meaning it gains the lead in the final result, 51%/49%.

Enough clues. Only 3 numbers should be in your mind: 1592, 1032 and 3766. The sample sizes are far too small. Poll companies are commercial enterprises, and do not want to sample the 16,000 persons required for reasonable accuracy, when they can get a fee and a headline by giving an up-to-date approximation based on a small sample. Yes, most of the result is captured by a small representative sample, so it is cost-effective to survey roughly 1500. Surveying 3,766 is not massive: it merely takes a step towards adequacy.

If I may quote myself, again and again: “Nobody gets round sampling theory, not even the Spanish Inquisition”.

Now to expert opinion on the other matters. Many professional bodies warned of the consequences of Brexit, mostly adverse economic consequences, but also adverse impacts on scientific research and cultural matters. In no particular order: Christine Lagarde,International Monetary Fund; Mark Carney, Governor of the Bank of England; Paul Krugman, economist and Nobel Laureate; Paul Johnson, Director of the Institute of Fiscal Studies; Ngaire Woods, Oxford professor of Economics; Paul Collier, Oxford professor of Economics; Richard Branson, entrepreneur; Lord Alan Sugar, entrepreneur; Warren Buffett, investor; Prof. Lord Darzi, Director of Institute of Global Health Innovation; Simon Stevens, NHS England Chief Executive; Peter Selby, professor of Cancer Medicine; Stephen Hawking, scientist; Rob Wainwright, Director of Europol; Barack Obama, US President; Hillary Clinton, US Presidential Candidate; Angela Merkel, German Chancellor; and many other world leaders whose names I am too bored to type out for you; the Russell group of top universities: 150 scientists writing to The Times; 200 doctors & Healthcare professionals;  300 historians; 9 out of 10 of 639 economists (I should check the number of economists, but that is the claimed total). You can get more of these notable opinions here:

In fact, so overwhelming were the numbers and status of the experts for Remain that the Leave side were reduced to saying “experts are part of the elite we are trying to get rid of”.

The betting markets got it wrong. No, I did not put any money on the referendum. The sum I would have bet I instead sent as a contribution to the Leave campaign.

Super-forecasters put the chance of Brexit at 35%. For once, the sensible weighting of the status quo let them down.

I intend no harm to astrologers, and have not read their prognostications, but if 50% of them got the outcome right, my commiserations to those who guessed the wrong way.

Now I ask my loyal readers to keep a close record of the  above expert predictions so that they can be tested against reality. Philip Tetlock will probably point out that many of these august bodies have made rather general statements which are not really capable of being tested, but I think that if GNP falls and science funding falls the experts will turn out to have given valid warnings.

For absolute validity we should wait 43 years before doing the Before and After comparison. If we were pollsters we would take snapshots of the financial markets next week, and extrapolate from those. In my view one year is the soonest we can get a preliminary measure, and five years would be more convincing, because most of the “uncoupling” disturbances would have run their course.

Please put a note in your diary for June 2021 when it will be possible to see if GNP and all the other economic indicators show a negative or positive effect of Brexit. My prediction is that if you plot out the 50 year data on the economy you will be hard pushed to notice any difference. I hope for a slight upward tick after a one year downturn, but that is just a hope. My prediction: no significant difference from the long-term trend line.

However, given the unanimity of the warnings about the severe consequences of Brexit, and the impeccable status and qualification of these experts, they are likely to be proved right very quickly, and there will be few in this benighted isle capable of decent research within 5 months, let alone 5 years.

Be kind to us you dear readers, who live in more favoured realms. When we fall into such ruin as to lack even parchment and quill, record  our demise for posterity so that all can learn from our error: the English voted the wrong way, and are no more.




  1. In five years, I fully expect the English to be reduced to a level close to that of the middle ages. Commerce will be done in gold coin. Travel will by horse. Football will be replaced by jousts. The plague will ravage major cities.

  2. Ah, back to normal! Just think how literature will flourish.

  3. "Expert opinion was almost uniformly in favour of Remain. The scientific community was particularly strong for Remain": but the scientists weren't speaking as experts, but rather as a vested interest. As a rule, one scientist speaking in public might be talking about science, but anything said by many scientists just translates as "Give us your money!"

    As for economists, in what are they expert? Macroeconomics has less predictive value than water-divining. As Her Majesty asked of the financial crisis "How come nobody saw it coming?"

    It was Keynes (an economist but nonetheless a very clever chap) who had the ambition that economists would eventually become as competent as dentists. We're still waiting, Maynard.

    1. The validity of a recommendation is inversely proportional to the number of people signing the letter.

  4. This comment has been removed by the author.

  5. The negotiations will be so difficult and the terms so unpalatable that they will be put to a second referendum and rejected. Then we will be sent to bed with a smack and told never to do that again.

    1. Seems a familiar story, as experienced by Ireland

  6. I worked in counterparty credit risk for 20 years. All the mainstream economists are wrong in predicting black swan events, those who call them correctly are outsiders or mavericks. You get to sit at the big table by telling people what they want to hear, not by being right.

    Await the failures of several major European banks before you cast judgement, I give it less than ten years.

  7. Also multiple subjective factors in polls. How to predict voting behaviour (tithing to vote) etc.

    Usually pollsters use earlier elections to calibrate the endless decisions involved. With a once off vote they have no prior data to work with. Hence, bad quality.

  8. Also multiple subjective factors in polls. How to predict voting behaviour (tithing to vote) etc.

    Usually pollsters use earlier elections to calibrate the endless decisions involved. With a once off vote they have no prior data to work with. Hence, bad quality.

  9. "The final YouGov survey for The Times, which quizzed a massive 3,766 people online..."

    Add a sampling frame with built-in selection bias, and a sample size of 10,000 would have likely given a misleading result.

    And of course, none of the experts have drafted a Plan B(rexit), so they're going be whingeing about Xenophobic Low Information Voters at least through Christmas. This Texan says thanks, England (and Wales)! This is much more fun than watching BBC dramas on our local Public Broadcast Stations.

  10. The best facts about Clixsense's GPT Click Program:
    1. Up to $0.02/click.
    2. 5 seconds lowest timer.
    3. Re-click every 24 hours.