:::: MENU ::::

Campaign Media, Elections, Internet

Predicting U.S. elections through search volume in Google Trends



A growing number of studies have examined how search engine and social media activity can be used to document current social trends and predict future patterns. Indeed, research literature has found that activity registered in Google Trends and Twitter chat can help predict various kinds of consumer and social data, including consumer goods marketability, the success of movies pending release and disease outbreaks.

In a 2011 study “On the Predictability of the U.S. Elections Through Search Volume Activity,” researchers from Wellesley College examined if Google Trends volume — which reflects the number of search queries around keywords — was predictive of 2008 and 2010 U.S. Congressional elections. The study’s authors separated electoral races into two categories: those with trend data recorded for one candidate; and those with trend data recorded for both candidates.

The study’s findings include:

  • For political races where there were Google Trends data for only one candidate, the data predicted the outcome 70% of the time in 2008 and 52% in 2010.
  • For highly contested races in which data were available for only one candidate, the success rate for predictions was 81% in 2008 and 39% in 2010.
  • In races where there were data for both candidates, the data successfully predicted the outcome in 43% of cases in 2008 and 40% in 2010.
  • In highly contested races where data for both candidates were available, the data successfully predicted the outcome in 33.3% of cases in 2008 and 39% in 2010.

The authors conclude that, compared to the traditional methods of election forecasting, incumbency and New York Times polls, and even in comparison with random chance, Google Trends did not prove to be a good predictor of either the 2008 or 2010 elections. The study suggests that “the sentiment of a user’s query” may be an omitted variable that could affect Google Trends’ accuracy as a tool for predicting elections.

Tags: technology

    Writer: | Last updated: April 5, 2011

    Citation: Lui, Catherine; et al. "On the Predictability of the U.S. Elections Through Search Volume Activity", Department of Computer Science, Wellesley College, 2011, PDF.

    Analysis assignments

    Read the issue-related Politico article "Next Big Thing in Social Media and Politics."

    1. If you were to rewrite the article based on knowledge of the study, what key changes would you make?

    Read the full Wellesley College study " On the Predictability of the U.S. Elections Through Search Volume Activity" (PDF).

    1. Summarize the study in fewer than 40 words.
    2. Express the study's key term(s) in language a lay audience can understand.
    3. Evaluate the study's limitations. (For example: Do the results conflict with those of other reliable studies? Are there weaknesses in the study's data or research design?)

    Newswriting assignments

    1. Write a lead (or headline or nut graph) based on the study.
    2. Spend 60 minutes exploring the issue by accessing sources of information other than the study. Write a lead (or headline or nut graph) based on the study but informed by the new information. Does the new information significantly change what one would write based on the study alone?
    3. Interview two sources with a stake in or knowledge of the issue. Be prepared to provide them with a short summary of the study in order to get their response to it. Write a 400-word article about the study incorporating material from the interviews.
    4. Spend additional time exploring the issue and then write a 1,200-word background article, focusing on major aspects of the issue.

    1 comment

    You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.

    More than feeling lucky? Using Google trends to predict the stock market Oct 18, 2012 13:59

    […] developing nations are more focused on the past. (Interestingly, Google search data appears to be a poor predictor of U.S. Congressional […]

    Leave a comment

    Please prove you’re not a robot * Time limit is exhausted. Please reload CAPTCHA.