10.20.10

Election Oracle Methodology

The Election Oracle uses WiseWindow’s proprietary Mass Opinion Business Intelligence technology to uncover millions of online consumer opinions on every high-profile race. The Oracle combs through more than 40,000 blog sites, message boards and forums, along with Twitter and other social media, on a 24/7 basis. The Oracle counts the total opinions registered for each candidate, and rates each as positive, negative, neutral or mixed.

The Oracle then calculates the plus/minus ration for each candidate daily, by taking the number of positive comments for a candidate and dividing them by the number of negative comments. The two candidates are then compared head-to-head, based on this data. (After sorting the comments by topic, based on numerous keywords, a similar process is used to determine which candidate is winning the debate on specific issues.) This data is then tracked using a 10-day moving average, to assure that sudden changes in changes do not unduly skew results.

In terms of daily predictions, the Oracle blends this plus/minus factor with the each candidate’s share of overall opinion—a key indicator of momentum and interest. And these results are then mixed with traditional polling results—again using a 10-day moving average, which mutes the effect of any one poll, insulating the Oracle from dramatic shifts based on one late result—to form the final judgment, which is presented in deciles.

This is the first time that Internet sentiment has been used in a significant way to help predict elections. After this fall's midterms, we will back-test the results to continue to enhance the Oracle's utility and reliability. We welcome your feedback at oracle@thedailybeast.com