Random Responses to Random Questions
Where Did You Get the Idea to Develop this Model?
A co-worker — In 2004, I ran a non-monetary office pool on the presidential election. The winner was a generally apolitical economics professor who was working for us at the time. Puzzled, I inquired how he’d made his picks (only missed one state — Wisconsin). He said he’d used the “political futures” markets to make his picks.
My dad — While reading Fooled by Randomness by Taleb, he called to ask a question about an example in the book. Without going to the book to find the example, he was asking about the math involved when you own four stocks with probabilities of 25%, 30%, 15%, and 20% of doubling in the next year in calculating the probability that at least one of the four will double. Lot’s of other factors involved, but it led me to question whether some type of portfolio analysis approach could be used to evaluate likely electoral outcomes.
The Wisdom of Crowds: Why the Many Are Smarter Than the Few and How Collective Wisdom Shapes Business, Economies, Societies and Nations by James Surowiecki. Surowiecki points out that the collective wisdom and intelligence of large groups can prove better than that of smaller groups of elites.
InTrade’s Prices Don’t Seem to Match Your Probabilities. Why ?
As I have developed the model since 2006, I observed that the “Last Trade” price isn’t necessarily the best indicator. Why? Because the “Last Trade” price is in the past. I am interested in the price of the next trade. As a result, I have developed a number of ways to generate implied prices based on the current “Bid” (what someone is offering to pay to buy the contract) and the “Ask” (what someone is willing to sell the contract for).
How do your Political Views Impact the Analysis?
I am a conservative. I believe limited government and free enterprise are the best path to economic prosperity for our country. So I pull for conservatives to win elections.
I am extremely critical of "Keynesian" economic models because they produce results that are effectively pre-determined by the assumptions the model itself makes. As a result, I have been extremely careful to avoid building a model that generates results based on its assumptions and methods.
Past results suggest those views don’t impact the results produced by the models I use. 2006 was a year of "hope" as I watched the earlier version of the model suggest that Republicans had slightly better than a 50% probability of holding the Senate, but also spit out the most frequently observed result saying that Republicans would lose Missouri and Virginia while winning in Tennessee. In 2008 (a year mainly of "despair"), the most frequently observed electoral vote outcome was off by one electoral vote — the Nebraska split electoral vote. Many of my Republican friends asked me to keep my model’s projections to myself as we entered the last two weeks of the campaign.
Will the Model Change as the Election Approaches?
Yes. In October, the model parameters are adjusted to generate a number of different projections based upon possible scenarios. For instance, Nate Silver of FiveThirtyEight.com raised the question of whether Democrats were building a “pacific firewall” to hold the Senate. I will evaluate probabilities of various scenarios — for example, suppose the data suggests that Senate control hinges on Democrats needing to win two out of three states (California, Nevada, and Washington). I will evaluate the likely outcomes when focusing on those three specific states.
How Often Do You Run Your Model(s)?
I run the models daily? In the week before the election, I will run them twice a day or more. And on election day I will run (or have someone else run ) the model on an hourly basis throughout most of the day. However, I will not do a full analysis of the data on an hourly basis on election day. I will blog or tweet significant updates, but most election day data collection is geared towards collecting data to assist me with future improvements in the model and a more detailed post-election analysis. For instance, how do reports of turnout, etc. impact the market perceptions of the likely outcome.
Have a Question or Comment? Email us at ElectionMarkets@gmail.com
