David M Rothschild on Posted on

When considering the myriad of predictions before Brexit, when it comes to polling, financial markets, and prediction markets, there are certainly some bright spots and some concern:

First, polling did pretty well, depending on what polls you cherry picked. The final Pollster numbers were 45.8 Remain and 45.3 Leave; the final votes were about 52% Leave and 48% Remain. The trick was that Leave was leading the overall average for most of the last few weeks and Leave lead the average on online polls (where Remain lead the average for telephone polls).

Second, financial markets did really bad at predicting. Working with Glen Weyl, Nikete, and others, we posted a week ago that the Pound would trade at about 1.5 to the Dollar if Brexit failed and about 1.3 to the Dollar if Brexit succeeded. At the time of the writing the Pound as about 1.40 to the Dollar and the prediction markets were about 65% for not leaving. The currency exchange hit 1.50 at 10:45 PM local time or about 45 minutes after the polls closed. It reached a low of 1.32 at 10:30 AM local today. Financial markets went all in Brexit would fail. Dozens of other financial instruments were doing the same and not just on the night of the vote. The Pound was trading at 1.47 or 1.48 for the entire week; numbers that translate into nearly 90% for failure.

Third, the goal of prediction markets is to aggregate available data efficiently and add in additional idiosyncratic and dispersed information; a consistent 75% for Brexit to fail over the last week was a reasonable, but unexciting, prediction for prediction markets. (1) Prediction markets correctly anticipated that the polls, leaning Leave, would break towards Stay. (2) Prediction markets were more conservative than the financial markets in their prediction for Stay.

Predicting something at 25% and it happening is not an “epic fail“, as in Brexit; predicting something at <1% and it happening is an “epic fail“, as in nothing prediction market did! But, it is certainly better to be on the winning side. Where did prediction markets go wrong?

First, as Glen, Nikete, and I point out in the post from last week, the prediction market were trading too high for no Brexit, because the traders priced in that they would all lose money if Brexit passed. I did not incorporate this into the PredictWise prediction. Obviously, with the passage of Brexit I wish I did, but I did not have precise enough estimate and I try to keep PredictWise pretty clean in its data and predictions.

Second, market did not pick up on enough idiosyncratic data in the field. Maybe this is because traders do not have the pulse of working masses? Possible. But, I go back to the first point as probably more important. This morning I have a new line of research that I am obsessed with: political impact on financial markets is under-explored and fascinating. I suspect that we underestimated the effect that the volatility of the underlying currency had on the prediction and financial markets.

Experimental polling with Pollfish was perfect, but I do not publish multiple predictions, I just bury them in blog posts! Seriously, as I put in the Huffington Post on Wednesday, there were two key lessons from our experimental polling data. First, the stay voters were terrified of the economic outcome of Brexit, but the leave voters thought it would be slightly beneficial. Second, the undecided voters looked exactly like the leave voters. The leave voters did not vote for economic turmoil, they believed that this vote would have little impact on the British economy. The leave voters did think it would cut back on immigrants; the stay voters did not believe them on that.

This one-shot succession election in the United Kingdom does not substantially alter anything regarding the prediction methodology of the regularly occurring presidential election in the United States. The predictions for the presidential election are done with more data types, more history, and more sets of outcomes. That being said, the outcome certainly will affect the US. First, it has already sent the S&P 500 futures down 5%. Second, there will be ongoing turmoil in the United Kingdom, one of our closest allies. Third, the breakup of trade agreements and anti-immigrant sentiment that fueled the Brexit vote will surely influence the discussion in the United States.

Beyond the predictions for the Brexit vote, there are now a plethora of things to consider moving forward. Expect predictions on these in the near future:

First, there will be a new British Prime Minister. Most likely Boris Johnson, but that is not confirmed yet.

Second, the new Prime Minster will have to start negotiating a trade deal with the European Union. S/he could go hardline to satisfy the sentiment from the Brexit vote. Or, s/he, could be so worried about the crashing financial markets, the very real threat of Eurocentric Scotland leaving the United Kingdom, and and the Irish in Northern Ireland getting angry (not to mention poor Gibraltar, which may get cut off again!), that the deal does not substantially change much.