Using Intrade.com to Teach Campaign Strategies in the 2008 U.S. Presidential Election

Abstract

During the six weeks before the 2008 elections, I conducted a contest for the 72 students enrolled in my upper-division course Campaigns and Elections. Using contract prices posted by Intrade.com, an electronic gaming market based in Dublin, I asked students to choose among 10 political outcomes. The “contracts” earned by each choice were determined by the Intrade “bid” prices as of September 24, 2008, the day the contest began. The contest helped teach students about campaign strategies, the way electoral rules affect electoral outcomes, provided a reference point to discuss the campaign, and was designed to stimulate interest in the election.

United States presidential elections are won by winning electoral votes, and campaign strategists focus on gaining 270, an electoral-vote majority. Any serious attempt to win will focus on this target, although strategists want a margin of safety to carry them beyond this minimum. Political scientists know that campaigns spend minimal resources on states they have little hope of winning and that they ignore states they are confident of winning. Most campaigns focus resources on competitive states with a relatively large number of electoral votes, even though in extremely close contests, such as in 2000, switching the smallest state from the winner to the loser will change the outcome.¹

These facts are not obvious to many undergraduates, even those with the good fortune to major in political science. Having taught a course called Campaigns and Elections at Michigan State University for over two decades, I wanted to develop an additional method to teach students about campaign strategies and the ways that electoral rules affect political outcomes. And although many students were very interested in the 2008 presidential election, I wanted to generate even greater attention. In addition, I wanted students to learn more about the strategies required to amass 270 electoral votes.

I conducted a contest in which students could choose among election outcomes, offering a cash prize of $200. John H. Aldrich, David W. Rohde, and I (Abramson, Aldrich, and Rohde 2007, 312–20) had used contract prices posted by Tradesports.com, a gaming site based in Dublin, to predict potential Democratic and Republican presidential candidates, and I decided to use this site for my contest. By 2008, however, Tradesport.com was listing contracts on politics and current events on a sister site called Intrade.com. Intrade offered a major advantage over the well-known Iowa Electronic Markets (IEM). For the general election the IEM offers only two political contracts, one on the share of the popular vote for president and one winner-take-all contest for the party to win the most popular votes. But the popular-vote winner is not necessarily elected, as the 2000 election revealed. And while candidates doubtless want to win the most popular votes, and although in 1996 Bill Clinton set a goal (which he missed) of winning a popular-vote majority, campaigns focus mainly on winning an electoral-vote majority.

Intrade offered opportunities to buy and sell contracts for many political outcomes. Granted, there were not actual opportunities to buy contracts on every posted outcome. As Intrade acts as a clearinghouse, contracts can be bought and sold only if someone is willing to buy and to sell them. And if the “bid” price and the “ask” price are too far apart, no contract can be made. Each Intrade contract is worth $10 and the minimum bid is $0.01. Aldrich, Rohde, and I view the bid price as a measure of the “subjective probability” that a specified outcome will occur. For example, on December 24, 2006, the subjective probability that Hillary Rodham Clinton would win the Democratic presidential nomination was .525, whereas the probability that Barack Obama would win was .207 (Abramson, Aldrich, and Rohde 2007, 313). On December 24, 2006, the subjective probability of John McCain winning the Republican nomination was .520 (Abramson, Aldrich, and Rohde 2007, 319).

The rules for my contest are in appendix A and a sample entry form is in appendix B. As can be seen, I did not charge students to enter, and whether or not they entered and how well they did had no bearing on their grades.

There is controversy about the relative predictive ability of electronic markets compared with public opinion polls, although most of the controversy focuses on the IEM. Robert S. Erikson and Christopher Wlezien (2008a) present an excellent summary of these controversies, and conclude that the IEM is not superior. Moreover, some pundits claimed there were biases in Intrade prices, maintaining that someone was attempting to push up contract prices for McCain (Rogin 2008; Silver 2008).² I believed it was unlikely that such biases, if they existed, would affect contract prices for all 10 of the events used for my contest. As can be seen from the contract prices as of September 24, 2008 (see appendix B), the Democrats were favored to win the presidency and to win in Colorado and Michigan, with Virginia a virtual toss-up.³ The Republicans were seen as clear favorites in Florida and as slight favorites in Ohio. On the other hand, contract prices suggested that it was an even bet for the Democrats to win 270 or more electoral votes and for the Republicans to win 260 or more. The Democrats were seen as heavily favored to retain control of the House and Senate.

As Election Day approached, the subjective probabilities came closer and closer to the actual outcome. The Intrade favorites as of 5:04 p.m. on Election Day won 49 of the 51 state (and D.C.) contests, missing only by making the Republicans the narrow favorite in Indiana and Democrats the narrow favorite in Missouri.⁴

On the first day of class, I announced that I would hold a contest, making this clear with both an in-class announcement and through e-mails that reached all the enrolled students. In fact, I sent an attachment with the subjective probabilities for a Democratic presidential victory, a Republican presidential victory, the number of electoral votes the Democratic presidential candidate would receive, the number of electoral votes the Republican presidential candidate would receive, the party to win the presidential contest in each of the 50 states and D.C., the party to control the U.S. House, the number of Democratic House seats, the party to control the U.S. Senate, the number of Democratic Senate seats, and the party to win each of the 35 Senate contests. These probabilities were forwarded to students a week before class began and at least once a week through the day before the election.

Late September was not a fortuitous time to begin the contest since Lehman Brothers filed for bankruptcy protection on September 15, contributing to a financial meltdown on Wall Street. Five days after I launched my contest, Congress was debating a bailout package, McCain urged that his first debate with Obama be postponed, and when the House first voted down the bailout package the Dow fell 778 points. Indeed, by October 27 the Dow would close at 8,176, even though it recovered somewhat by the day before the election. If I had the prescience to foresee these economic changes I would have chosen to include different states that turned out to be more competitive shortly before the election. For as the economic meltdown occurred, McCain's chances progressively diminished and the day before the election the probability of a Republican presidential victory had dropped to .113, while the IEM probability for a GOP popular-vote majority was only .112. And by the day before the election, the Democrats were favored to win all five of the contest states.

Having sent the students the contest rules and the contest entry form via e-mail on Wednesday, September 24, I spent 40 minutes during the next class, Monday, September 29, explaining the rules.

First, I explained how contract prices are derived as a result of bettors being willing to bid real money to buy a contract and of other bettors being willing to put up real money to sell one. The price listed by Intrade is the cost of a $10 contract. The more likely an outcome is seen to occur, the higher the price. I pointed out that actually betting online from the U.S. was probably illegal, but that despite this, Intrade contract prices were widely reported, especially on online sources discussing the election.⁵

I argued that the contract price can be seen as a measure of the subjective probability of an outcome occurring. Thus, as appendix B shows, on September 24 some bettors were willing to offer $5.42 to buy a contract for the Democrats to win the presidency. Some bettors were willing to offer $4.46 for a contract for the Republican candidate to win. And some were willing to offer $0.02 to buy a contract that neither party would win. As I explained, people who buy a contract want to pay the lowest possible price and people selling a contract want the highest price, so in principle the market will drive the contract value. The contest included contract prices for partisan control of the U.S. House and the U.S. Senate, because this allowed me to introduce material on congressional elections. In order to make my task manageable, students were required to allocate their money equally across all 10 events and to make only one choice per event. I rounded so that the winnings would be in whole contracts. The student with the most total contracts would be the winner.

I explained the best strategies to win and the way in which presidential campaigns allocate their resources.

First, I pointed out that as the tiebreaker was the time and date of entry, it was advisable to enter the contest early. This was especially true since students were able to send in a new entry if they changed their minds.

Second, I suggested that choices among events should be consistent. For example, it was very unlikely that a Democrat would win without carrying Michigan. More importantly, as I pointed out, the Republicans had never won the presidency without winning Ohio.

Third, I emphasized that it would be useful to watch projections from several sources, including updated contract prices that I sent them from Intrade and from the IEM. To help students track these projections I sent them an attachment through which they could link directly to Intrade, the IEM, pollyvote.com, fivethirtyeight.com, realclearpolitics.com, pollingreport.com, pollster.com, and cqpolitics.com. The last site is especially useful since it classifies the competitiveness of all 435 House districts and provided an opportunity to discuss two highly competitive races in Michigan, both of which ended in the defeat of Republican incumbents.

The contest had mixed success. Twenty-seven of 72 students entered, and 11 of them waited until November. Some of the late entrants were either extremely optimistic Republicans or had paid very little attention to the race. Moreover, some students made inconsistent choices, such as predicting the Democrats to win the presidency but choosing a cutoff for Democrats winning that was fewer than 270 electoral votes. And only one student submitted a revised entry, although this was a reasonable strategy given the way the candidates' chances shifted. Toward the end of the contest, an informed student would have chosen a Democratic presidential victory, the Democrats to win all five states, and to control both the House and Senate. And the four most successful students made all these picks. The winner was ultimately determined by the best choice for the number of electoral votes that the Democratic presidential candidate would win. The winner selected the Democrats to win 350 or more electoral votes, and thus earned a total of 55 contracts. Employing this contest provided many opportunities to discuss election strategies.

These discussions made clear that McCain had an increasingly narrow path to attain 270 electoral votes. By continually providing students with updated contract prices, I discussed the ways in which the probabilities of Republican success were diminishing even in states such as Virginia that had voted consistently Republican for president from 1968 through 2004. I also pointed out how Republican chances in Ohio were slipping. In discussing McCain's chances I drew upon the concept of retrospective voting (Fiorina 1981), explaining how George W. Bush's low approval ratings were damaging McCain's prospects. This allowed me to present projections based upon several academic forecast models that included approval as a component (see Abramowitz 2008; Erikson and Wlezien 2008b; Lewis-Beck and Tien 2008).

The contest also provided a handle for class discussion questions that I e-mailed to students after major political events, such as the three presidential debates and the vice-presidential debate. These questions often asked students to evaluate the campaign strategies of both campaigns, especially in regards to their attempts to muster 270 electoral votes. Unfortunately, as the campaign developed it became increasingly clear that it would be very difficult for McCain to win an electoral-vote majority.

I plan to use a similar contest again, although I will make several changes to try to increase participation. For example, several political science majors told our undergraduate advisor that they were too busy to enter. Participation might have been higher if I had asked students to fill out their entry form on a Web application. And I will also award multiple prizes, although this will require more money if the first prize is to remain attractive. And perhaps I should ask my chair to assign me to teach smaller classes. For example, Charles Tien adapted my contest for a graduate class called Voting and Elections at the Graduate Center at CUNY. It had nine students and, even though students paid a modest fee ($5) for entering, all nine participated.

Moreover, the contest could have been more pedagogically effective by turning students' attention to the interaction between campaign tactics and changes in the opinions of the electorate. For example, students could have been asked to discuss why contract prices were changing over time by discussing news stories in the media and by examining the changing campaign tactics of the candidates. Likewise, the contest could have been used more effectively to link campaign events to voter decision making; although of course I did discuss major events over the course of the campaign and the way they were systematically reducing Republican chances over the period between mid-September and the election.

Finally, it should be noted that other academics have used Intrade and its predecessors as a teaching device. Indeed, a fairly large number of academics have used similar methods and even encouraged students to back up their predictions with their own money. To the best of my knowledge, however, this is the first systematic attempt to explain how a betting Web site can be used to teach students about U.S. elections. I hope that this article will encourage other scholars to share their experiences and to suggest other methods that may accomplish this goal.