Midas Oracle .ORG

Eric Zitzewitz December 13th, 2006

Here is the presentation I am about to give at the Yahoo confab. In my 10 minutes, I’m going to describe some recent work with Erik Snowberg and Justin Wolfers on financial market responses to the 2004 and 2006 elections. We use a method we call a Prediction Market Event Study — which takes advantage of a prediction market tracking the probability of the event in question. Prediction markets help solve some of the traditional problems with event studies (calibrating the extent to which the event was a surprise, identifying the event window, controlling for other information released). Also, especially when we get big swings in probabilities in short time windows like we did in 2004 and 2006, they allow for much more precise estimates than the traditional method.

In the two papers, we find that equities are expected to be worth slightly more under Republican governments (and Treasuries worth slight less), but that the effects are fairly modest. Bush’s reelection in 2004 raised the value of the S&P 500 by about 2 percent, as compared with a Kerry presidency. At the same time, it increased the yields of both 2 and 10 year T-bonds by 10-12 basis points. We find consistent results for Bush v. Gore in 2000, and use Koleman Strumpf and Paul Rhode’s data on historical Presidential markets to look at elections going back to the 1880s. The stock market’s preference for a GOP winner seems remarkably consistent. Of course, we shouldn’t necessarily equate the stock market effect and welfare.

In the more recent paper, we examine market reactions to the House and Senate control shifts in 2006, the Senate shift in 2002, and the 1994 shifts. We can get an especially good estimate for the Senate in 2006, where the probably of the GOP keeping it fell from 90 percent to 10 in an hour as the final vote tallies for Missouri and Virginia reversed early GOP leads. We find that a “control” of a house of Congress is about 10-30 percent as important as the Presidency in affecting economic policy, across a number of indicators. Contrary to some pontificating, we do not find any evidence that the stock market prefers divided government.

I’m going to wrap up with a little pontificating of my own. I have thought for a long time that the most interesting and underexplored set of corporate prediction market applications were the financial ones.

One idea is to run public markets on events that affect one’s stock market value. Eli Lilly did this for public policy events with Newsfutures in 2003. But as far as I know, no one has expanded on this idea or taken it outside of public policy (to earnings or product launch success, for example).

Doing so could have several benefits. First, by understanding what the market is thinking, you can know to correct misimpressions when it is advantageous to do so. Second, by understanding how the market thinks a decision would affect your value, you might make better decisions.

Third and potentially most significant, by aggregating information on one source of uncertainty about your value, you reduce asymmetric information, and potentially improve the liquidity of your shares. Kathy Yuan at Michigan had a nice paper a few years back (now joint with Bob Dittmar) showing that when government bonds were introduced in emerging markets, the liquidity of corporate bonds improved, and spreads fell. The reason is that the government bonds aggregated information about country risk, so corporate bond traders only had to worry about asymmetric information about company risk. By running markets that capture various sources of uncertainty about its value, a firm could get that same effect working for it. It’s like the positive effect of getting analyst coverage, but the analysts are prediction market traders.

Anyway, see many of you there.

• All Best Posts Ever , All Guest Authors's Posts , Analysis (Data) , Analysis (Market Proposals) , Analysis (Meta) , Exchanges & Markets , Finance
• Comments(2)