Brexit, Predictions, Polls and Bias

June 16, 2016

Story of the week, and after 6/23 could be the story of the year is the looming referendum in the UK to decide if it will stay in the European Union. 

A surface level understanding of the situation teaches that the UK, after Germany is the largest “net” contributor into the EU financial partnership. The balance of payments to and from the EU systems means that the UK ends up with the second shortest straw as it comes to the measurable financial benefits of the economic union. (

Greece by contrast is one of the biggest beneficiaries receiving nearly four times what I contributes in the balance of payments. 

This single factor is helpful to understand much of the unrest over the last seven years and the pending referendum and ensuing global markets uncertainty for the outcome. 

I’m not intending to discuss the depths of the issues facing the EU, not am I qualified. I have, however, been sucked into watching the daily polling data via a variety of sources including Bloomberg’s Brexit Tracker ( as well as PredictIt’s poll-wagering system. 

For full disclosure I have a modest wager that the vote next week will be for the UK to remain in the EU. This is based on immaterial  facts and more of a social common sense than any good on-the-ground intelligence. The level of change and uncertainty that will result from the UK’s exit from the European block simply seems too palpable and I cannot imaging 51% of the voting population will want to risk damage to the UK economy and ignite conversations that could lead to the unwinding of the economic block. 

What has facinated me is the run up in polls from a probability in the 20% range to nearly 50% range is less than two weeks, based heavily on polling data. 

I have learned that if you had to split the deongraphics, the older generations within the UK would opt to remain in the EU while younger generations would opt to exit. 

There have been some article talking about the polling proces which in this day and age is largely digital. It inflates my curiosity to think that there could be a strong adverse selection bias in the recent polls, particularly as many are driven by technology which is assumed to be far more prolific among younger voters. 

In addition as prediction markets have re-proliferated in recent years, I followed my curiously to track the odds on PredictIt over the last week or so. 

The gambling bug got the best of me, so I signed up and put about $100 bet on a “No” vote for next week. 

After I entered my order to buy about 160 contacts for about $100 a bet that would pay me $160 if I’m right and leave me with 0 if I’m wrong (not great odds, but to me seemingly easy money)  I figured well if I like the long “No” vote, I would probably also like a short “Yes” vote. 

So without proper research I simply attempted to do what I though was sell a “Yes” but the resulting transaction simply sold my “No” contacts for about a $3 loss (the spread at that time).  After a few minutes and not being able to find my two open positions, I realize my follie, and I immediately rebought my “No” votes. All in I probable paid about .60 cents (worth about .53 cents today). 

After a couple of days and seeing the trade go against me, but seemingly more swiftly than even the polls would suggest it occurred to me that in a market where the only option is to buy a No or a Yes that distortions are possible. 

Conceivably the sum of the prices for No+Yes  should equal $1.00.  This is not completely true as Predict it takes about a 3 cent spread on the bid/ask. This is the balancing mechanism in place to ensure all bets between rational actors can settle for $1. If Yes gets bid up the next buy of No should be paying less for it. 

However here is my concern:  Those who are betting on a Yes vote are one part speculator but it’s conceivable (even in this micro-marketplace) that there are some bonafide hedgers. Bonafide hedgers would pay more than the probability  up to an offsetting price against their hedged position.  It is unclear if shorting the GBP or EUR or going long the Yen or USD will truly hedge an adverse outcome from the referendum. Although those are the cheapest ways for large institutions to do it. However, PredictIt provides direct protection which is worth a premium if the trade size and volume is sufficient.  

Also, since registered participants are not allowed to sell-short either the Yes or No bet, there is no true counter-balancing trade and it’s very possible that prediction markets are creating an adverse feedback loop for a “Yes” vote as the insurance becomes over priced to the probability of the outcome. 

That said these are binary trades, and either you will be 100% right or 100% wrong. And in a world bubbling with populism anything is quite possible. 

It’s surprising to me that few pundits are talking about the self fulfilling reinforcement of excess polling and prediction markets.  If so, such big data endeavors me become a form of campaigning that can alter outcomes in our post-modern world. 



June 7, 2016

All this talk of Trump.  All this aggrandized media attention and even the smart people have missed that THEY are feeding the machine.

So many of us are “afraid”, “frustrated”, “surprised”, or “concerned” by the dominance Trump has had in the delegate race.

However it is mostly “us” who actually read and write the news that have fed his insatiable desire to remain in the spotlight. When will we all realize that the best way to squash his power will be to turn the news cycle past him and not let him back in.

When we stop reading about Trump, then we will remove his megaphone.

The power remains squarely in the hands of the press.  If they want this guy as the GOP nominee, then there is nothing we can do but hope that Hillary can beat him in the general election.

When the GOP fails to endorse him, however, it will only further his appeal and power base.

They Say a Picture is Worth 1,000 Words

June 1, 2016