Unboxed: More Data Can Mean Less Guessing About the Economy

Other than their owners, most small businesses have no employees. His is one of the 4.3 million that do have them, and that employ fewer than 20 people each. And while these companies collectively produce roughly 15 percent of the nation’s economic output, their activities aren’t captured by the official numbers in a timely or detailed way.

Yet this measurement shortfall in the small-business sector, and a series of other information gaps in the economy, may be overcome by what experts say is an emerging data revolution — Big Data, in the current catchphrase. The ever-expanding universe of digital signals of behavior, from browsing and buying on the Web to cellphone location data, is grist for potential breakthroughs in economic measurement. It could produce more accurate forecasting and more informed policy-making — more science and less guesswork.

“We’re seeing the emergence of new sets of data and knowledge that we’ve never had before,” says James Poterba, president of the National Bureau of Economic Research and a professor at the Massachusetts Institute of Technology. “It’s a real opportunity for policy makers.”

In the small-business case, Mr. Tabor’s company, American Home Inspectors and Engineering Assessments Inc., is one of more than 200,000 that have allowed Intuit, the software maker, to gather data on their use of Intuit’s online payroll or online accounting products for research on employment and sales trends. The data are stripped of identifying information, and, in asking permission, Intuit also emphasizes that it uses the data to improve its products.

Mr. Tabor said he had no qualms about contributing his anonymized data. “I’m happy they use the data for research and to make their services better,” he says.

Intuit began its research in 2004 on small businesses and has expanded its scope, including many more companies and becoming more fine-grained in its data tracking as it has added products, services and customers.

Researchers at the Bureau of Economic Analysis, the government’s statistical scorekeeper of economic activity, are now experimenting with the Intuit data, seeking to tap it to improve the official estimates.

“The promise of new data sources, like Intuit and others, is more accurate, more timely and cheaper data for monitoring the economy,” says Steve Landefeld, director of the bureau. “That could be a really big step forward.”

National income accounting emerged in the Great Depression, in an effort to bridge the economic information gap of its day. In June 1930, based on scattered reports available to him, President Herbert Hoover declared, “The Depression is over,” when in reality conditions were quickly worsening.

The main tool for government statistics remains telephone and in-person surveys of households and businesses — surveys that are costly and time-consuming.

Tracking behavior online can pull in far more data, more quickly — so that governments should be able to see signs of inflation, deflation and employment trends sooner and adjust policy faster.

For example, the Intuit monthly employment data, based on online monitoring, is current. That is eight months to a year ahead of the government’s best statistical look at the health of small business, which is culled from quarterly surveys, state unemployment records and tax returns. The monthly Intuit survey data have also proved accurate, almost mirroring the government results when they are finally reported, according to Susan Woodward, a consulting economist for Intuit.

“Whatever is happening, it is better to know sooner,” Ms. Woodward says.

Yet whether more data, collected faster, will improve economic forecasting is uncertain. So far, the results are mixed. An encouraging study, begun in 2009 and repeatedly updated, has used Google searches to predict home sales and prices three months into the future. In the study, the higher the frequency of search terms like “house prices,” “real estate agent” and “mortgage rates,” the more likely the national housing market would heat up.

The results of the study, “The Future of Prediction,” by Lynn Wu, an assistant professor at the Wharton School of the University of Pennsylvania, and Erik Brynjolfsson, a professor at the M.I.T. Sloan School of Management, have held up over time. In the most recent version, their model using search data predicted future home sales 24 percent more accurately than the forecasts by experts from the National Association of Realtors.

But another major project using Google search terms points to the limits of such techniques: Google Flu Trends uses the same methods as Big Data-style economic prediction, although it focuses on public health.

The service monitors flu-related search terms and seeks to predict the incidence of flu, ahead of official statistics based on doctors’ reports to the Centers for Disease Control and Prevention. In 2009, with the outbreak of H1N1 flu, Google Flu Trends was prescient.