The Search Is On: Googling “Barack Obama” and “Hillary Clinton” in the 2008 Democratic Primary

Abstract

The 2008 Democratic race between Barack Obama and Hillary Clinton was the longest primary campaign in history and one of the most intensive in recent memory. Besides following the campaign through traditional media like newspapers, radio, and television, more Americans turned to the Internet actively searching for information about the candidates. In this article, the author analyzes search traffic data for Obama and Clinton over the course of the primary to revisit the classic “minimal effects” hypothesis about campaigns. This hypothesis, first voiced in the 1940s and 1950s, argues that campaigns have only small marginal effects on citizens vote choices. The author tests a weaker version of this hypothesis, asking whether campaigns positively affect a kind of political engagement that he terms active online engagement. Using a statistical model, he compares online search traffic in the month before an election against the month after. He finds support for the hypothesis that campaigns spark political engagement by an average of about 50 percent over the general level of political interest in a given state. He also finds evidence contrary to the popular wisdom of the 2008 campaign that Hillary Clinton ran a poor race. In fact, the author finds that her campaign had more influence in engaging people online, although the overall levels of search traffic remained below that of Obama. Put simply: Hillary Clinton ran a solid campaign against an unusually popular opponent.

KEYWORDS:

• Barack Obama
• Google
• Hillary Clinton
• online search
• 2008 presidential election

Acknowledgments

The author acknowledges Eric Oliver, Betsy Sinclair, and Jong Hee Park for their comments on an earlier version of this article.

Notes

Note. Dependent variable: 4-week search average after state primary/caucus.

The Web site Congressional 140 (http://www.congressional140.com) aggregates the Twitter feeds of members with accounts.

Google, the search engine, began as a research project in 1996. Google, the company, was founded in 1998.

For example, the first page of hits from a search of “Joe Lieberman” in Google produces links to his Senate page, a Wikipedia entry, an official senate biography, a third-party site called liebermanmustgo.com, and a few popular articles from online media and blogs that use Lieberman's name in the headline.

Correctly remembering the spelling of a candidates name is a concern, though a minor one since most search engines now offer alternative correct spelling options if the user's spelling is close to the correct one.

In the unusual case that post-election search traffic matches pre-election search traffic, I can conclude that the Obama-Clinton race was not a state-by-state campaign, but a national online campaign where online searching is conditioned on closeness of the race, rather than on actual campaign activities. Although it may be tempting to associate this conclusion with the classic minimal effects thesis that campaigns only marginally interest voters, I strongly encourage readers to resist this urge. Campaigns do indeed have the ability to motivate interest among voters, but this interest quickly spills over via media coverage and social networks to affect interest elsewhere.

While the design shares some common traits with a regression discontinuity design, it would be inaccurate to call it one because the cutoff (election day) is not exogenous to the primary calendar.

“Barrack Obama” and “Hillary Clinton” were the most commonly searched incorrectly spelled names.

Google Trends does not have a separate “politics” category by which a filter searches data, plus candidate names cut across a number of potential categories like news, entertainment, society, and reference.

In the case of a general election, one common proxy is to use national party disbursements to states. Unfortunately, in a primary this measure is not available.

For instance, Barack Obama held his official announcement February 10, 2007.

Enzi is the Republican senator from Wyoming.

87 weeks × 52 states = 4,524 observations.

Alaska, Delaware, Vermont, Wyoming, North Dakota, South Dakota, and Montana are excluded from these calculations because of search traffic values of 0.