A primer on predictive markets. What they are, how they work and why they're important

[Godhumor]

Last edited Wed Feb 24, 2016, 02:23 PM - Edit history (3)

My initial disclaimer, as always. In real life I am a statistical analyst and regularly work with data sets that follow the same rules as polling. I am also a Clinton supporter, but this post is intended to be nonpartisan. I write these for educational purposes and any examples I give using actual candidates are intended to only be examples and not commentary.

There are a few of us, myself especially, that regularly post updates from a site called Predictwise (http://www.predictwise.com). Predictwise is an aggregation site, similar to Pollster and RCP, but for predictive markets

A predictive market is similar to a stock market, except investors make trades on potential real life events. These can involve sports, referendums, votes, etc but the most widely followed events are political elections. Investors buy and sell "stock" in who will win--this isn't abstract or theoretical, investors put their own money on the line in these markets.

In case it is not immediately clear predictive markets differ from polling in that they don't measure sentiment (I.e. who would you vote for, who would you like to win) but go right to the who will win side of the question. As a result polling next February could show that 55% of people would vote for Clinton to 45% for Sanders, but investors might translate that spread as a 95% chance that Clinton emerges with the nomination.

Predictive markets also have an inherent contradiction in that they're both stable and volatile, at the same time. The stability makes them very appealing as a tool for measuring how a race responds to events and information. The theory is simple because predictive markets are investor based, investors have baked in certain allowances for current events in their buys. In other words, a poorly received speech or a closer than expected poll result may not budge the predictive markets feelings on the state of the race. 2016 is actually a good example of this. During the Bernie surge in the summer, predictive markets were very slow to move away from Clinton and never dropped below a 65% chance or so at the nomination. The reason given is that investors expected a primary challenger to Clinton to emerge and had already accounted for it in the "stock" they held on the event. By the same token, some sort of faux scandal (email, Benghazi) was also expected due to her long tenure and visibility with the public. As a result, it is believed that predictive markets are the best predictor as to whether an event or speech or whatever will change the perception of the race.

On the other hand, these markets can be volatile as investors are trying to make money. As a result people may buy or sell based on financial interest instead of whether they believe an event will occur. Right now, Clinton is shifting between 89% and 91% for two reasons. First, investors who bought at her cheapest are selling because the difference between 90% and 100% is small (And she could still technically collapse). They may believe she will win, but they're divesting shares to pocket guaranteed earnings. On the other hand, she is also fluctuating up because more people are certain she will win and are buying at 90% going for a modest profit when she secures the nomination.

So predictive markets are different than polling, but their monitoring has become a fairly normal part of political race analysis. For me, I like them as a tool to see how stable a race is perceived to be, but there are also legitimate criticisms for how they work and what the represent.

Let me know if you have further questions. Thanks.

[brooklynite]

1. I "predict" you'll get the following responses...

This is the same as the stock market, which is a ponzi scheme

these people aren't investing, they're "gambling"

like the "corporate polls", the investors are ignoring the "millions of disaffected voters", hiding below the surface, who will rise up to elect Bernie.

[whatchamacallit]

4. And those responses would be valid

Thanks

[Electric Monk]

6. Those people aren't investing, they're gambling. Fact. nt

[NobodyHere]

13. I was gonna say political fantasy football

[hootinholler]

15. Yep. That said Hillary no is a good buy right now on predictit

She's running around $0.19/share which is $0.02 under Bernie yes.

[CincyDem]

2. It's a Presidential "Futures" market

As I understand futures markets, they don't tell you what something is worth...they tell you the collective judgement of what it will be worth on some future predefined date.

The sum total and monetization of all participating opinions about the present and the future. Wisdom of crowds measured in willingness to invest in a candidates future.

I like the differentiation vs. polls. First time I've thought of it that way. Polls measure hope and, as they say in the futures markets, hope is not an investment strategy.

thanks for the read.

[Godhumor]

14. Yeah, emotion versus measurement is one of the biggest issues people have with analysis

And analysts. It is never comforting to see your desires reduced down to a probability statement.

[PowerToThePeople]

3. How many working class and poor participate in these "markets"?

eom

[whatchamacallit]

5. "Oh... them

who cares?"

[Godhumor]

7. Again, not polling.

It is an analog state of will person A win, yes or no? It is not about sentiment or who should win or who a demographic supports.

[PowerToThePeople]

8. So you don't think the Koch bros

Would sacrifice some minor cash in these markets to get the numbers looking the way they want? The external results will pay greater dividends than these predictive markets.

Seems too easy to game.

[Godhumor]

9. Market manipulation is what shut down intrade, so it is a threat

But if the Koch brothers decide to do so, it really doesn't matter if working class people are investing out not; they will easily be able to move the needle.

[PowerToThePeople]

10. So you accept that they may not reflect reality.

You know what reflects reality?

This.

https://www.google.com/search?q=bernie+rally+pic&tbm=isch

[Godhumor]

12. No, you didn't lay out an it is not A ergo it is B supposition n/t

[jeff47]

11. And if predictive markets were accurate, there would be no 2008 collapse. (nt)

[Bernie-2016]

16. It's called speculative market for a reason

It's just speculation.

[riderinthestorm]

17. Interesting. Thanks for posting nt

[Gothmog]

18. I appreciate and agree with your explanation

I miss the old Intrade market which had its problems but was very accurate. The CFTC shut down Intrade because it was allowing too many US investors and was being played or manipulated without the protections of a US regulated market. Predictwise is an aggregate of the predictive markets which means that you have to set up an overseas account to participate in one of the markets that is used by the Predictwise system

Towards the end, Intrade had people trying to play the market with last second closing bids and high opening bids to affect some of the option pricing. These tricks were methods that are outlawed on US exchanges but were allowed on Intrade because is was not regulated by the CFTC. There was some strange volatility and people who were taking large positions to try to control the market.

I suspect that these problems still exist for the underlying markets used by Predictwise which is why one has to take these positions with a grain of salt. Overseas (mainly Irish) markets have different regulators than US markets and practices that are illegal in the US can be used to some degree on these markets.

[Bread and Circus]

19. We didn't need this. Hillary is very likely to win the nomination.

Many of us concede that.

But can you guys please act like Hillary is a real front runner and show some sportsmanship and class?.

[Godhumor]

20. ? It was a primer for how predictive markets work

I've also done other primers on polling, definitions, aggregation and smoothing. I made a very specific point of being nonpartisan in the post. So, whatever caught your attention, I'm sorry and, if youpoint it out to me, I will change the section to be more neutral.

[Gothmog]

21. Predictwise is the successor to Intrade which had a great track record

The free market system works and people who invest their money in positions on election outcomes are very very careful and use a number of different tools. Intrade was very accurate in predicting election results https://www.quora.com/How-accurate-has-Intrade-been-at-predicting-the-result-of-U-S-elections

In the last election it was 90% for all elections (non presidential) the last Presidential election they were correct on all of them, except Missouri but that didn't get called until two weeks after the election, the earlier elections were in excess of 88% but were in the very early days of Intrade.

The interesting part about the Intrade data is that you can make accurate predictions a week in advance of the election,

Here is the actual results of the intrade prediction for the 2008 election vs the actual results http://electoralmap.net/2012/2008_election.php

2008 Electoral Map - Election Results
Shown immediately below is the electoral map depicting the results of the 2008 presidential election in which Barack Obama won with 365 electoral votes to John McCain's 173. Below, the Intrade results are shown. Further down you will find the 2008 pollster report card.


2008 Electoral Map - Intrade Forecast
Shown immediately below is the Nov 4, 2008 election day forecast from the Intrade prediction market. Intrade did not predict Nebraska splitting its votes, and it was the first time in state history that this happened. Missouri and Indiana were also reversed in the forecast, but both having eleven electoral votes resulted in a nearly dead-on electoral vote count.

Predictwise is the successor to Intrade with features built in to make it harder for US investors to place bets. I would not discount the accuracy of these systems

[Gothmog]

22. Professional Bookies Predict Elections With 91% Accuracy – They Say The Result Will Be…

This is a great article on the predictive markets http://www.addictinginfo.org/2015/12/19/professional-bookies-predict-elections-with-91-accuracy-they-say-the-gop-nominee-will-be/

Here’s something everyone needs to know: professional gamblers are far more accurate than polls are. This has been proven time and time again. The Huffington Post calls professional bookies “History’s Most Accurate Election Forecasters.” And, there’s a reason for this. Where presidential polls have been wrong in the past, even so-called reliable exit polls, gamblers have gotten it right with 91% accuracy.

While it’s not legal to bet on U.S. elections in the United States, that hasn’t stopped popular websites like Paddy Power, or Predict Wise, from placing their bets. And, yes, they do pride themselves on their history of accuracy in predicting elections.

Here’s what Koleman Strumpf, a University of Kansas economics professor who tracks betting trends, has to say on why he thinks gamblers are more trustworthy than polls:

“Relative to the polls, the betting markets have to think hard about what they’re saying since they are putting their money at stake. Also polls tend to reflect what people are thinking at a given moment, versus a forecast of what will happen on election day — post-convention bounces, for instance.”

Here are some screen shots of the latest Predictwise odds

?zoom=1.5&resize=630%2C551

?zoom=1.5&resize=630%2C586

?zoom=1.5&resize=630%2C372

?zoom=1.5&resize=630%2C388