We obsess about the aggregated prices that emerge from markets, whether it is oil, the Dow Jones, or the prediction market contract on who will be the next president of the United States. The price is a reflection of the subjective beliefs of individual traders, and we spend too little time considering the individual traders’ expectations, strategies, and motivations that combine to create that price. Rajiv Sethi of Barnard College and I were very lucky to examine a unique dataset this summer, which allowed us to learn more about how individual traders behave in markets; specifically, we examined trade-level data for all trades that occurred in the final two weeks of the 2012 election for either Obama or Romney to win on Intrade, the largest political prediction market in 2012. Our main academic finding is that traders are surprisingly one directional, almost always buying contracts favoring one of the two candidates. A secondary finding that has garnered popular attention is that one trader heavily influenced the price of these contracts, possibly for potential political gain, by investing nearly $4 million in Romney positions over two weeks.
The academic question our paper examines is the trading strategies and motivations of the traders. And our finding is that most traders are either performing arbitrages (i.e., buying and selling contracts that guarantee them a small return) or trading in one direction (i.e., only going long on Obama or Romney). The archetypical informational trader gets new information and goes long on whichever side it favors. What we see is that traders get new information and use it to keep trading for their chosen candidate. An example would be a situation where there was bad news for the Romney campaign. Obama traders would go long for Obama and push up the price for Obama, but after a little while, the Romney traders would push back when they thought the price has moved too far towards for Obama. Everyone gets to keep going long for their candidate, but the new price still reflects the new information. Rajiv goes into this in more detail in an earlier blog post.
But, you are probably not reading this article to learn about trading strategies, so I will pivot to the possible market manipulation.
Many observers noted in real-time that a peculiar wedge opened up between Intrade and another prediction market, Betfair. Intrade was consistently 5-10 percentage points more bullish on Romney (i.e., a contract that paid out $1.00 if Obama won could trade for $0.70 on Intrade and $0.75 or $0.80 on Betfair). Our new dataset shows that one trader was making that happen by providing massive amounts of liquidity to the Romney side of all trades (accounting for roughly 1/3 of all action favoring Romney) and creating “fire-walls” of several hundred thousand dollars to keep the Romney price static at times of high information flow (e.g., if the price for an Obama contract was $0.70 the trader would note on the order book the willingness to sell $100,000 or more of Obama to win at $0.70 so that Obama traders could keep buying at that $0.70 price indefinitely without the price going up). Note that we have no personal information on this trader, just the trader’s trades.
We provide three possible explanations for the trader’s strategy: (1) the trader was convinced that Romney was underpriced throughout the period and was expressing a price view, (2) the trader was hedging an exposure held elsewhere, or (3) the trader was attempting to distort prices in the market for some other purpose.
(1) Simply going long Romney is unlikely because at any point the trader could have gone to Betfair and purchased the same contract for less money. There are many costs to utilizing Betfair: it trades in British Pounds, it blocks U.S. ip addresses, etc. But, a trader of this size could have overcome those costs at less money than the trader’s loss by utilizing Intrade versus Betfair.
(2) Hedging is a more plausible explanation for the trader’s behavior, but still unlikely. Earlier academic literature shows that some market indexes and the likelihood of the election outcome can be correlated, but we could not find similar patterns in 2012 (e.g., historically, sudden increases in the likelihood of a Democratic victory are adversely correlated with the S&P, despite historical correlations of stronger stock markets under Democratic presidencies). We cannot eliminate the possibility that the trader was hedging some more detailed securities like specific energy contracts that traders may have concluded would be more tightly impacted by the election outcome (e.g., the trader could have been going long Romney to cover potential losses in renewable energy contacts should Romney have won the election).
(3) Distorting the price for some other purpose, possibility political, is the most likely motivation of the trader. Placing hundreds of thousands of dollars on the order book, precisely at times of a lot of new information, is a strategy that maximizes the impact of the investment, but not the return. One of the trader’s most active periods was between 7:30 PM and 9:00 PM ET on Election Day, when new information was arriving by the second. At that time the trader essentially placed so many potential trades at about $0.70 per $1.00 for Obama that the trader was telling the market that s/he would match anyone who wanted to buy Obama at that price. This froze the price for the crucial 1.5 hours between the first major reports of election returns and the last swing state poll closings. Ultimately, the trader spent about $375,000 during that 1.5 hours and, as soon as the trader left the market at 9:00 PM, Obama shot up past $0.90 per $1.00 contract.
If this trader were attempting to manipulate the market, it is a three step process to make it successful: (1) change price, (2) convince people it is real, and (3) have people change behavior because of it.
(1) One trader was able to control the price of a liquid market with a massive wall of limit orders for two reasons. First, markets, by design, move with large quantities of money regardless of whether it is 1,000 people with $100 or 1 person with $100,000. A sea of traders could move a stock price or Warren Buffett alone could move a stock price, as the market does not care about the motive or quantity of traders. Second, the government’s harassment of Intrade, and other online markets, made it difficult and risky to join and keep money in Intrade, which limited participants and readily available money. Thus, even if it was tempting to buy Obama long at $0.70 per $1.00 contract on Election Day, it is likely that few traders had the money sitting in the market necessary to buy up all of the contracts that the Romney trader was willing to sell. You would need to already be a trader and have tens or hundreds of thousands of dollars sitting idle in your account.
(2) This is the hardest step; it is not as easy to convince people that the price level is real, but maybe people who want evidence will appreciate any data source that validates them. Sites that present prediction market data frequently aggregated Intrade with other markets to ameliorate the concern that any one data source can be wrong. Overall, prediction market data was very successful in providing accurate and timely predictions of the 2012 election. Yet, if someone was looking for a reason to be hopeful about Romney, Intrade’s price provided a solid piece of data for them.
(3) If people are convinced it is real, the impact on the campaign is going to happen. There is a cascading effect to being a viable candidate; the more viable a candidate appears, the more money and volunteers, support, and turnout the candidate receives. Thus, the more viable the candidate appears, the more viable the candidate becomes. When it comes to Election Day, one piece of positive news may be the validation someone needs to stop off at the polling place on the way home from a long day at work.
If manipulation could be successful, it would be worth it. If a few million dollars could boost fundraising and morale, than it would be a good investment next to one more television advertisement in a flooded Ohio, Florida, or Virginia market. Roughly $28 million was spent on TV advertising in just one state, Ohio, in the last week of the election alone.
We cannot say for sure if this market was manipulated, but someone definitely shifted the price heavily towards Romney and maintained that price imbalance until 9 PM on Election Day, when the polls closed in the last swing state and the election was finally up to the vote counters. Yet, despite this trader’s efforts, most observers, even if they were following just prediction markets, still received a very accurate forecast of the election.