Gaming the Electoral College

From the Archive: A model uses game theory to predict how changes to the electoral system could shift campaign strategies and ad spending — and alter election results.
October 29, 2010 | Event Highlights
Print this page

Only three times in US history has the electoral vote trumped the popular vote in a presidential election, but the particularly contentious results in 2000 rekindled the debate about whether to abolish the winner-take-all electoral college in favor of a direct national vote. Since then, Hawaii, Illinois, Maryland, and New Jersey have passed legislation pledging their electoral votes to the winner of the popular vote, and several other states have considered following suit.

Perhaps the greatest criticism of the electoral college is the way it influences candidates’ campaign strategies, creating battleground states that attract a disproportionate amount of attention from candidates in terms of advertising dollars, fundraising, and candidate appearances relative to the number of popular votes those states represent. California, New York, and Texas — large states with huge numbers of electoral votes — barely register in presidential campaigns. These blockbuster-sized states are viewed as noncompetitive because the margins of victory for the dominant party in each are so big that campaigns view their election results as predetermined.

Consider the 2004 presidential campaign. The Kerry campaign advertised very little in New York, probably on the calculation that George W. Bush was unlikely to advertise much in the heavily Democratic state. Likewise, the Bush campaign most likely passed over advertising in Texas because it anticipated that the Kerry campaign would not invest much in ad efforts in the Republican stronghold (and Bush home state). But both campaigns showered swing state Ohio with advertising, partially because each believed the other would do the same.

In the 2000 race, it was battleground state Florida that brought these criticisms most sharply into relief, prompting the most recent calls for electoral college reform. “A lot of people — safe to say mostly Democrats — think back to Gore and Bush in 2000,” Professor Brett Gordon says, “and conclude that if we’d had a direct national vote instead of using the electoral college, Gore would have won and so many things would have been different.”

“But the simple truth is that it may not be true,” Gordon contends. “If you change the way the system works, then you change the way candidates will campaign.” After all, it would be political suicide if candidates didn’t also shift their campaign strategies and continued to ignore large markets like Los Angeles, New York City, or Houston. “Even though the larger stronghold states lean strongly one way or the other, there may be many people in the middle whose votes could be swayed one direction or another. These areas would become much more competitive in attracting candidates’ attention.”

Gordon, working with Wesley Hartmann of Stanford University, created an econometric model of the electoral system and combined it with advertising data from the Wisconsin Advertising Project, a research initiative documenting US presidential and congressional campaign advertising between 1998 and 2004. The model combines game theory with methods from economics, marketing, and political science to predict how changes in the electoral system would prompt changes in campaign ad spending, and how those changes would in turn affect voter decisions and election outcomes.

The results so far suggest that, without the electoral college, advertising would indeed shift away from battleground states and toward the large states where candidates, at present, tend to do little advertising. Interestingly, a direct national vote might make for cheaper elections — preliminary results show total spending on advertising going down. (The jury is still out on exactly how advertising affects voters. Political scientists seem more convinced now that advertising persuades voters who to vote for, but has a negligible effect on whether or not a person decides to vote.)

Gordon and Hartmann also ran their model as if there had been no advertising in the 2000 presidential election (with the electoral college in place). While the scenario itself is unrealistic to say the least, researchers run such counterfactuals as checks on the soundness of their underlying methods and math. If results seem too far out — say, either candidate winning by 400 electoral votes — that’s a sign the model or methods are probably flawed. In this alternate reality, Gore picked up Florida and New Hampshire for a total of 29 electoral votes, while New Mexico, Oregon, and Wisconsin shifted to Bush for a total 23 votes, putting the outcome at 273 to 265, in Gore’s favor. “These states all had thin margins of victory in the actual election,” Gordon says, “so it’s not surprising to see them switch. But without the model, we wouldn’t have known whether eliminating advertising would have tilted a given state one way or another.”

There are a number of other questions Gordon plans to address using the model, mainly to evaluate the different ways candidates would have to choose to allocate limited resources to the most effective markets. For example, how might spending limits imposed by changes in campaign finance alter election outcomes?

“But we are particularly interested in the electoral college first and primarily because of the vocal criticism that it disenfranchises voters in non-battleground states and deprives them of attention from candidates, who might position their own policies to appeal to voters in battleground states rather than more broadly,” he says. “For these voters, the electoral college seems unfair.

“Should we encourage a system where as many people are involved in the process as possible? How do we value spreading out an election?” Gordon asks. “There are real policy questions, and our work can make predictions about the likely changes that would happen under a new system and inform policy discussion.”

Brett Gordon is assistant professor of marketing at Columbia Business School.

Brett Gordon

Brett Gordon was a Columbia Business School faculty member from 2007 to 2014.