The science of predictions

6/3/2013 - Living in the age of big data, real-time information pours into the coffers of private enterprises, including retailers, financial services corporations, supermarket scanners and search engines. Meanwhile, government agencies continue their carefully constructed historical series with delayed releases.

If integrated, both worlds of mounting data would better enable economists to create more informed predictions, Google Chief Economist Hal Varian said.

“This is a challenge for economists … but a challenge worth undertaking,” Varian told a crowd of more than 200 guests on May 22 at the James L. Allen Center.

Varian’s hour-long program, “Predicting the present with search engine data,” was the 2013 Nancy L. Schwartz Memorial Lecture. The Schwartz Lecture, a 30-year tradition at Kellogg, explores issues of fundamental importance in current economic theory and honors the life of the late Nancy Schwartz, the Morrison Professor of Decision Sciences at Kellogg.

Automating discovery
While Varian began his presentation with humorous looks at Google search engine data related to hangovers (a search term that peaks every Sunday and on January 1), the Civil War (which sees rising interest during term-paper season) and gift ideas for spouses, his presentation soon turned to more important topics, such as the economic recovery.

Varian called the recent recession “a sentiment-driven recession,” which made Google search terms all the more telling and relevant.

Introducing Google Trends and Google Correlate, Varian said Google’s data tools could be useful for businesses as well as the social sciences to produce short-term predictions that drive decision-making.

“And, kids, you can try this at home,” Varian said, noting that both Google Trends and Good Correlate remain readily accessible tools.

The trick, Varian informed, is to overcome particular challenges related to seasonality or trend and variable selections to produce a final forecast weighted to the average of many models rather than one giant model.

Ultimately, Varian acknowledged, “Our goal … is to automate this kind of discovery.”

A valuable tool for marketers
Varian also touted Google Consumer Surveys, a recent addition to Google’s analytics line-up, as a swift, accurate and affordable way for businesses to run quantitative market research online.

With Google Consumer Surveys, organizations can pose straightforward questions to web users looking to access premium content. Businesses and marketers then receive aggregated data that provides telling consumer insights.

“Here at a business school, you can imagine how attractive this is to marketing people because you can get brand recognition, design choices and all sorts of data literally in a matter of hours,” Varian said.