The raw data table on PredictWise shows a very clear gap opening up between Betfair and PredictIt for the winning party of the 2016 presidential election. Translating the Betfair data into dollar contracts, one can sell the Democratic nominee to be president for $0.699, while one can go onto PredictIt and buy that exact same contract for $0.590. This is not a clear arbitrage because PredictIt charges a 5% withdrawal fee on total wealth and 10% on winnings, while Betfair charges between 2-5% on winnings. And, beyond other fixed and marginal costs, Betfair trades in British Pounds and PredictIt in USA Dollars.
But, even if it is not an arbitrage, it is unusual to see two contracts for the same thing selling at a different price. There should be informational sharing. But, we saw the same divide between Intrade and Betfair in 2012 and 2008. In 2012 saw evidence of manipulation, but do not see that today. In 2012 we saw massive firewalls of orders, in the order book, for Romney at salient values. First, that is not evident now on PredictIt. Second, the $850 per trader per contract cap would make that really hard to do. So, what are my theories:
1) There are two almost distinct groups of traders. It is very hard for USA-based traders to enter into Betfair. And, PredictIt is currently restricted to USA-based traders. It is possible that European traders are more skeptical of Trump actually winning the election.
2) The CFTC as imposed two restrictions on PredictIt: 5,000 traders per contract and $850 investment per trader, per contract. Since building an exchange is loaded with fixed costs and PredictIt cannot make up for fees with whales (i.e., traders with huge deposits), they charge a high fee of 5% on withdrawal and 10% on winnings. These restricts should favor longshots over favorites.
Imagine a contract that is trading at $0.90, but you think has a 99% likelihood of happening on Election Day. Without fees that is about a 10% return over 5.5 months (i.e., ($0.09*99%-$0.90*1%)/$0.90). But, you can only invest $850, so the max investment is 944 shares that could net $94. The fee would be $9.40 for winning, so $84.60. But, this contract is not very liquid, so you are holding all of your money there for the length of the election. Now, you cannot make that money work in other ways. If you made this a single investment with your deposit, you lose 5% of the base + winnings, so you are only clearing $37.87 for your trouble. That is probably lower than the fixed cost of investing in a new contract.
That 5% withdrawal fee was not as much an issue in the primary, where there are a continuous flow of elections. The traders can keep reinvesting the same money. But, as traders look towards the general election, when all contract end nearly simultaneously, it is going to loom very large. It is a penalty on long-term hold positions on favorites, pushing traders to short-term positions and long-shots, with the ability to rotate the deposit between outcomes and th possibility of a big payoff.
Thus, it is not surprising to see PredictIt trailing Betfair on Clinton, as it is likely influenced by a bias against the favorite. That does not mean one is right or wrong, but I would expect PredictIt to be systemically more biased to long-shots, especially earlier in the cycle.
The 5,000 trader limit is another huge problem as more and more traders are in a long-term hold. Imagine that 10 traders a day buy the Democratic candidate for president and hold the position. These are traders out to just make a small, fun investment. Now imagine that this pace starts at 1/1/2016. By Election Day 3,120 contact spaces would be held by long-term traders who are not interested in buying and selling on new information. This will restrict the ability of information traders to move the price efficiently on new information.
This will likely not be a binding concern for the state-by-state contracts, which require a more varsity level understanding of the election. Thus, I will follow the state-by-state contract relationship with the headline overall winner market very closely. Any discrepancy would indicate that the CFTC regulations are causing a big problem. State-by-state prediction market data will be on PredictWise by the End of the Day today, extending the predictions launched in February!