The Econophysics Blog

This blog is dedicated to exploring the application of quantiative tools from mathematics, physics, and other natural sciences to issues in finance, economics, and the social sciences. The focus of this blog will be on tools, methodology, and logic. This blog will also occasionally delve into philosophical issues surrounding quantitative finance and quantitative social science.

Sunday, January 13, 2008

Google vs. New Hampshire: Prediction Markets & Networks

As an American ex-pat living in the UK, I seem to be following the Presidential primary process with a greater degree of interest than I would have had I still been in the States. So when I read some articles recently about Google and a separate set of articles on how the pundits were surprised by Hillary Clinton's surprise victory over Barack Obama despite polls to the contrary, I thought there could be an interesting 'mash up' between these two seemingly unrelated issues. Oddly enough, the last week or so has shown that there are surprising commonalities between one of Google's business practises and how predictions of an Obama landslide failed to materialise.

Google encourages it's employees to use a prediction market to place bets on forecasts of events such as 'Will Google open a Russia office?' or 'How many users will Gmail have at the end of the quarter?' The reward for making correct predictions? Goobles (rhymes with rubles) which can be converted into prizes. Economists, Justin Wolfers (who has been featured on several blog posts on this site -- do a Google search of this blog), Eric Zitzewitz, and Bo Cogwill, wrote up a research report, Using Prediction Markets to Track Information Flows: Evidence From Google, that reaches an interesting conclusion: Information -- in the form of correlation in betting -- is "shared most easily and effectively among office neighbors, even at an Internet company where instant messaging and e-mail are generally preferred to face-to-face discussion," and that this type of information or social network -- based on "microgeography" (i.e., how people are [literally in this case] located in relation to each other) -- outweighs even friendship as a factor in information transmission within Google (as a business).

Although the use of prediction markets within Google is an interesting topic in it of itself, it got me thinking about how the findings from the research on Google's prediction markets might relate to making predictions about the electoral process -- especially, after the mini-debacle in New Hampshire. New York Times columnist, John Tierney wrote in his blog that prediction markets -- specifically Intratrade -- was the idea venue for coming up with accurate predictions in elections. Unfortunately for Mr. Tierney (and perhaps Intratrade as well), the results of the Democratic primary in New Hampshire made it, in John Tierney's words, "the wrong day to tout the Intratrade futures market." That's an understatement! (Although in fairness to Mr. Tierney and Intratrade, the tide started to turn toward Hillary at Intratrade a lot faster than it did in the news media.)

A less sanguine blog post was written by another New York Times columnist, David Leonhardt. In his post, Primary Predictions Popped, Mr. Leonhardt likened what happened on prediction markets, like Intratrade, where Barack Obama was given a greater than 95% chance of winning, to the bursting of "their version of the dot-com bubble." Part of the problem is that, as Mr. Leonhardt points out, the markets for these prediction markets are "thin" -- in stockmarket-speak, 'illiquid' -- which makes these markets especially vulnerable to bubbles and volatility.

Although I agree with Mr. Leonhardt, I think there is something else going on here -- which is precisely what the research shows on Google's prediction market -- the importance of the old real estate addage: 'Location, location, location.' In other words, the 'microgeography' -- or how people in a business or social environment interact with each other both by physical proximity and institutional strictures -- of how politicos, whether they be political operatives, journalists, pundits or 'experts', etc. -- can bias the information flow. Such swarming or coalescing of information flow can lead to bubbles in both political prediction (whether or not it takes place in a formal marketplace) and in the stockmarkets (this was one of the points that Didier Sornette made in his book, Why Stock Markets Crash: Critical Events in Complex Financial Systems [you can read my review of that book here]).

The New York Times article, Analyzing the New Hampshire Surprise (Jan. 10, 2008), by Jacques Steinberg and Janet Elder, seems to provide some evidence for my hypothesis. The article suggests that some political journalists and pundits were getting on the Barrack Obama bandwagon too soon (a particularly pointed example was Newsweek all-but annointing him on its front cover). This type of 'information cascade' may have at least indirectly contributed to incorrectly predicting the outcome of the Democratic primary in New Hampshire.

Perhaps the anecdote to these types of predictive mishaps is what Google does -- presumably based, in part, on its research into its own prediction markets -- with its employees: Move them around frequently. Political pundits, journalists, and operatives need to get out and about more often with people who are dis-similar to them rather than always hanging out with people who do more or less the same thing they do: making politics cynical and downright silly.

Labels: , , , , , , , ,


Post a Comment

<< Home