Stanley Jungleib

Revised: 20121118

THE DEFINITIONS

Everyone knows about Nate Silver’s projections. And everyone is going to have to learn about prediction or ‘futures’ markets, which are argued to potentially be highly efficient ways to quickly ‘aggregate’ the truth across disparate sources.

Statistics and predictive markets are entirely different things and should not be directly compared. They are designed to operate completely independently. Statistical projections are numbers, of course, derived from historical trends, and current polls—but which remain an a priori calculation of current behavior. In contrast, predictive markets are a posteriori betting pools where people place hard money where their mouths are: they have skin in the game. You can also think of the difference as in vitro test-tube experimental medicine compared to real in vivo bodily results. Forecasts and polls are grounded in numerate statistics, whereas predictive markets are grounded in money-backed, diversely informed, crowd psychology.

The data of crowd psychology is expected to diverge from the pundits: that is the point. But as well, one can reasonably expect an approximate correlation if not convergence as statistics increasingly  support emerging reality. For the U.S. Presidential Election 2012 that didn’t happen, providing abundant ground for well-supported suspicions about GOP manipulation of Intrade.com, the world’s largest independent predictive betting site.

Background on Prediction Markets:

Video: Riz Khan:Prediction Markets

Idea Futures – The Concept

Policy Analysis Market

Iowa Electronic Markets

Book: The Wisdom of Crowds

THE STRATEGY

November 1

With Nate suggesting 80% I noticed a great bargain at the prediction site intrade.com, which was showing the President in the low 60%s. Votamatic also maintained great confidence in its scenario—which, by the way it nailed.

Believing in the numbers and The President, I was able to wire money barely in time to average-in for Obama at these 2:1 odds, also shorting Romney at his 35%, thus planning to make a 50% return in a few days. I separately bet on Ohio, as well, as that was key. The goal was to make back my contributions to the Democratic campaign. I was also hypnotized by daily talk of ‘doubling-down,” thus entrained for the opportunity to join in.

THE WEIRD

I then sat fairly perplexed as over the next few days Silver’s numbers tightened along with many others’ but Romney’s Intrade percentage sat stuck up at about 35%, unaffected by the stats. This was a little disconcerting, but I’m new to it, so I watch. And by this point I’m all in, so watch is all I can do.

November 5

Basically, Intrade still hovers around 65/35 throughout the day. Remembering yet again that by design there is no tight linkage between forecasts and crowds, Nate has tightened to 85, so there remains something arguably inexplicable about such a 20-point Obama bargain, or Romney overvaluation, at Intrade:

The other notable oddity is that everyone knew that Ohio was key. Yet Intrade’s Ohio market consistently ran 5 points higher for the President e.g. 71/29 than the overall election. There was a gap between confidence on Romney in Ohio and nationally. That is, if you believed Romney would win with Ohio, you would make more money by placing a bet in Ohio with odds at 29% than nationally at 35%. And it seemed to me quite odd that there would be a GOP segment betting that Ohio was not key for Romney.

November 6

In the morning, Silver reports 90/10% for the President. Intrade still 65/35%. At least this motivates me to make a few dozen GOTV calls to New Hampshire and North Carolina! (My favorite OGM, from a ‘little old lady’: “I’m out testing the Ferrari.”)

On election night, I bring up Intrade, three vote reporting windows, and Excel. (CBS soon proved too conservative in their state calls or useless, as you prefer.) I started charting the President’s Intrade % [Blue] with his averaged Electoral College  returns [Red]. The time range is 4:30-7:00 Pacific = 7:30-10:00 Eastern.

Again, the first question one brings is to the situation is, why is Romney at 32% on intrade, not 10%. Or again Obama not 90% but fluctuating from 66-68%. Apparently “the crowd” is still unconvinced, and with an international pool widely subject to Fox perhaps this could be expected, as well. My serious impression at the time was that the market was bearish out of fear of wide reports of GOP voter suppression efforts, potential voting machine tampering, and the likelihood of protracted litigation complicating and delaying Intrade’s final call.

At point 9 5:13/8:13 one sees an Obama dip as the early southern Red states reported—but those were all expected. A rational, let alone predictive, market should not really react that way. But as the EC totals amassed, Romney’s and the President’s probabilities remained strangely static, still hovering at 70/30 as if this great prediction market was actually severely lagging reality. This pattern persisted until an hour later POTUS hit about 158 EC votes at point 32 6:22/9:22, after which Obama continued upwards with Intrade nevertheless still lagging.

My data for election night shows that the price support denying Obama Intrade market momentum persisted at least while Ohio was still officially open. Manipulation also accounts for the graph betraying a persistent lag on Intrade even after the decision was clear: you can’t dump a huge position at once without setting-off alarms. You have to transact realistic share quantities at plausible prices. This buffered Romney’s Intrade descent throughout the month and night.

Let’s review Romney’s month at Intrade. Dark blue is the daily close, brown is the running average. The record seems strangely resilient and resistant to change, admittedly with hindsight and in the face of the downward trend noted by 23 of 24 polls in Silver’s final report.

Romney-National

But consider, as the statisticians and pundits gave up on Romney over the month, he lost debates 2 & 3, then Sandy hit him, after the Nov 1 dip shown, why did Intrade recover stiffly instead of continuing downwards? Sure, some divergence from the result is explicable by mere speculation and day-trading. Yet, on both the Romney and Obama Intrade charts note the sharp, unrealistic discontinuity at Nov 6. Why would a diverse market of opinions behave this way? How can one learn about this putative predictive ‘wisdom of crowds’ arena without some expert opinion on why they seem to have so lagged reality at this point?

Obama-National

I sent my data around for opinion by some smart folk (Thank you Drew Linzer of Votamatic.com) and they agreed it was rather inexplicable. As well, they cautioned these are new areas where the psychology of the marketplace is not yet well-known.

November 7
Someone else noticed! A shoe loudly drops: Overcoming Bias presents its evidence that Romney’s value had indeed been artificially supported.

Was Intrade being manipulated over the last month?

According to Overcoming Bias, someone thought it was so important to control the most-watched independent index, that they entirely distorted its operation with as little as $1M. Within a campaign that reportedly spent $2B, this would seemingly be a tiny piece of the cost of doing business.

THE PRECEDENTS

Wild theory. Unbelievable, you say.

Admittedly there is not a lot of comparative data, and what exists may well be compromised as well. First, Intrade’s Ohio charts indeed show a slightly softer trajectory; but that market was only 12 – 14K shares, compared to  2 – 2.4M for the National. The interesting question remains, if an agent believed Romney winning Ohio were crucial and wanted to maximize their return, why would they not invest in Ohio at 30%, as opposed to placing money into the national at 35%? This suggests that profit might not have been their primary motive. In my case I was not confident about Ohio, but at 70% compared to 65% nationally, the opposite choice gave me the 5% advantage.

Romney-Ohio

Obama-Ohio

It Takes Practice

But, here comes the real, matching other shoe. In 2008 the New York Times reported similar but more successful manipulation of Intrade, for McCain:

Trading Variance in Election Predictions Raises
Questions

As to the question of motivation, this year expert research appeared exposing direct GOP voting manipulation already in the 2012 primaries:

Video: Retired NSA Analyst Proves GOP Is Stealing Elections

Republican Primary Election 2012 Results:Amazing Statistical Anomalies

THE LESSONS

Now you see the source of their Nate Hate. His numbers were a continuous refutation not only of their narrative, but their entire methodology and echo-chamber. The disparity I noted between Ohio and the national is easily explained: they simply focused on the most-watched indicators. That covers the oddities deserving report.

Yes, I made 54%, roughly matching my contributions to the President and Democrats and to the extent this analysis is correct, largely from the GOP’s own money! But aside from noticing an oddity, my result was from dumb luck and with great help by GOP operatives, not any particular skill on my part. Keep that in mind before placing your next bet in a public predictive market. As one writer hinted: Democrats might want to watch Intrade for similar opportunities in forthcoming elections. On the other hand, it may well be some time before such as good a bet as President Barack Obama with David Axelrod again appears.

Prediction markets are here to stay. Under the right conditions, they can find the truth of a question better than experts. Want to know what your staff really thinks about the new product design? Offer an anonymous betting pool, if you have the guts for the answer. Predictors within government may decide whether you become a Person of Interest. So it is keenly important for everyone to know both that these methods are in use privately, and that until the public markets become too large to buy, any prediction market may well be subject to corruption. These systems only work correctly under well-defined conditions, which include transparency, anonymity, a free flow of information, and independent judgment of a diverse pool.

To be clear, I do not see this as a failing of Intrade, as any attempt to regulate the market would negate the theoretical advantage that prediction methodology contributes. For example, it would be easy to evade limits on particular accounts by just opening a plethora of them.

Nevertheless, though active for at least a decade, public prediction by crowds subject to this degree of market influence demonstrably classifies it as at best an immature, if promising industry. The whole affair should give interested economists pause.

ORCA, OVERCONFIDENCE, INCOMPETENCE

Initially, the GOP’s ORCA project seemed to be a secret weapon campaign control system, perhaps from which the alleged manipulation of market appearances may have been integrated. Now, according to a range of reports from GOP sources, ORCA was a disaster on several levels. Unlike that corruption affecting my little corner of the economic world, the subsequent GOP blame and devolution is being well covered.

Voter suppression was obviously ineffective. And I’m not aware of reports of electronic tampering for which Anonymous warned it was monitoring GOP servers:

October 22
Video: Anon2Rove

What remains of the grand conspiracy at this point is a trove of hated Republican state officials, and a link collection recording the sad progression of inevitable overconfidence, incompetence, defeat, and betrayal. After reviewing the journalism below, it is difficult to avoid the conclusion that Romney and Rove were defeated by staffs operating under exactly the selfish ideals they embodied and espoused. Rove and others fed Romney inaccurate, unrealistic, self-interested data which he displayed the poorest of judgment in trusting. Rove is suspect for deliberate manipulation of an important nascent public market, alleged ballot machine tampering, and collapse of the Orca system which so thoroughly embarrassed him on Fox. Finally, Norquist dismisses the entire ‘rolling calamity’ (famously, per Peggy Noonan)  with playground talk. Imagine this reckless, hapless team within the Oval Office. The nation truly dodged a bullet of nuclear scale.

November 1
Overconfidence: Mitt Romney’s Son Went to Russia With a Secret Message for Putin

A Romney Travels to Russia, but on Strictly Friendly Terms

November 5
“Orca will deploy 34,000 operatives in key precincts tomorrow, linked by a smartphone “killer app” Romney Campaign Enlists Help of ‘Killer Whale’ Project to Get Out the Vote

November 6
Rove said on Fox he was on the phone getting updated exit polls from his field. He so believed in his model that he could not even accept Fox’s call, and running his own inaccurate numbers about Ohio as far as he could take them: this is the meltdown against Fox’s own experts that you witnessed.

Then comes Romney’s inexcusable delay and curt concession speech.

November 7
Romney’s Delay Due to Failure ORCA Exit Poll Analysis

Romney’s concession speech was not gracious

They All Got it Right: Polls, Markets, and Models

November 8
“the end result was that 30,000+ of the most active and fired-up volunteers were wandering around confused and frustrated ” The Unmitigated Disaster Known As Project ORCA

Adviser: Romney “shellshocked” by loss

November 9
“If you’re in the con game and you don’t know who the mark is … you’re the mark.” Mitt Romney had no idea what was coming.” Campaign Sources: The Romney Campaign was a Consultant Con Job

“I like firing people.” Mitt Romney cuts off campaign workers credit cards, leaves workers stranded

“When so much data started flowing into the facility, it was perceived to be a hack and rejected” Mitt Romney’s ORCA program couldn’t stay afloat

Epic Whale: Romney volunteers say ‘Orca’ was debacle

Inside Team Romney’s whale of an IT meltdown

“This didn’t materially change the course of the election.” Romney digital director: Orca wasn’t a loss

“If ORCA had no relation to outcome, then tell me, what the hell was the point of the project? ” Team Romney: Actually Project ORCA Was Pretty Darn Successful

November 10
Karl Rove Under Fire After Election Failure

Rethinking “shellshocked” Romney

November 11
Pollsters Helped the Republicans

November 12
Orca was no fail whale, says Romney’s digital director

Grover Norquist: Mitt Romney painted as ‘poopy head’. But you told us that is exactly what you wanted to buy; a sheisskopf with a pen.

November 16
I thought we were done. But Anonymous has checked back in with a completely different take on Orca: Anonymous Claims it Stopped Karl Rove from Hacking the Election by Hacking ORCA, We Think I leave it to the press to sort this out.

Thank you to my facebook interlocutors who contributed to this reporting and linkage.