SAN FRANCISCO — Margo Sugarman spent months last year searching on Google for the appliances to complete her dream kitchen, scouring the Internet for information on the latest double ovens and low-noise mixers. SAN FRANCISCO — Margo Sugarman spent months
SAN FRANCISCO — Margo Sugarman spent months last year searching on Google for the appliances to complete her dream kitchen, scouring the Internet for information on the latest double ovens and low-noise mixers.
Not only did those queries guide the Tel Mond, Israel, resident to the best deals for her $17,680 renovation, they also helped the Bank of Israel, which looks to searches like Sugarman’s to assess the state of the nation’s $243 billion economy.
The central bank stands at the forefront of the world’s hunt for new economic indicators, analyzing keyword counts for everything from aerobics classes to refrigerators — reported by Google almost as soon as the queries take place — to gauge consumer demand before official statistics are released. The Federal Reserve and the central banks of England, Italy, Spain and Chile have followed up with their own studies to see if search volumes track trends in the economies they oversee.
At stake is the ability of the guardians to deploy nimbler policy responses. Greater foresight could make the difference between a slowdown and a recession, a recovery and an inflation- stoking boom, according to Erik Brynjolfsson, a member of the Federal Reserve Bank of Boston’s Academic Advisory Council.
“When central bankers were looking at traditional data, they were essentially looking out the rear-view mirror,” said Brynjolfsson, a professor at the Massachusetts Institute of Technology in Cambridge. The December 2009 study he co-authored on predicting U.S. home sales using search volumes was cited by three of the central bank studies.
“If the Fed had had access to this information, they would have been able to make better forecasts of what was happening to the housing market and known more quickly the depth of the problem” during the 2007-2009 recession, he said.
• • •
It all started with a hunch in Mountain View, Calif. On the heels of developing a new website reporting how often users searched for certain keywords, Hal Varian, Google Inc.’s chief economist, said he wondered whether this data could foreshadow what traditional economic reports would show later. So he ran the numbers.
“The ‘aha moment’ was, gee, this actually works,” Varian said in an interview.
The result was a 23-page paper he co-wrote in April 2009, demonstrating how data reported on the Google Trends service improved forecasts of auto and home sales and retail spending in the U.S.
Tanya Suhoy, a senior economist at the Bank of Israel, released a paper three months later finding that the new data helped predict slowdowns and slumps in Israel.
Economists from the Central Bank of Chile to the Bank of England soon began to ask similar questions: Did more people browsing for cars predict an increase in auto sales? Was a jump in research on unemployment benefits a hint that people were losing jobs?
Neither Varian nor MIT’s Brynjolfsson say they are surprised central bankers have been particularly interested in the technique: One of the challenges of policymaking has been having to set interest rates based on delayed economic information.
Google makes its data available one to three days after users perform searches. The U.S. Commerce Department typically publishes its monthly report on retail sales two weeks into the following month.
“With monetary policy, you can turn the spigot left and right in a few seconds,” said Varian, whose former students from his more than three-decade teaching career include Fed Chairman Ben Bernanke. “If you’re going to be more responsive, you need to have up-to-date information.”
That’s the philosophy underpinning the Bank of Israel’s analysis, spearheaded by Suhoy, who began studying Google’s data at Governor Stanley Fischer’s suggestion.
Using query volumes in place of government statistics that are not yet available, the staff computes a monthly index that reflects the current health of the economy. The figures then are presented to Fischer and his policy committee before they set the nation’s benchmark interest rate.
Margo Sugarman in Tel Mond exemplifies why this works. Like a growing number of consumers around the world, the 45-year-old communications consultant turns first to the Internet whenever she’s considering buying something. That makes her queries a precursor to what often results in a purchase.
When she wanted a mixer for her three children’s favorite cream-cheese brownies, Sugarman scrutinized the reviews she found through Google to decide between a model from Kenwood and one from KitchenAid. She ended up buying the KitchenAid appliance at her local electronics store.
• • •
To be sure, the research connecting economic forecasting with Google’s search counts — totaling 119 billion worldwide in June, according to Internet research company ComScore Inc. — is still in its early stages. Even the biggest proponents of the tool cite causes for reservation. The figures go back only to 2004, constraining comparisons with statistics that have a longer track record. And by limiting its sample to Internet users, the search volumes may not reflect the purchases of those who spend less time online: the elderly and the less affluent.
“Potentially, using Google could be interesting, but at the moment its forecasts of macroeconomic variables aren’t reliable,” said Lucrezia Reichlin, a former head of research at the European Central Bank and now a professor at the London Business School. “Google is sexy and something may come of it, but more research is needed.”
Even with those limitations, the Bank of England is another central bank that tracks Google’s searches. The popularity of search terms like “JSA,” short for Britain’s jobseeker’s allowance, helped predict unemployment data, according to a June 2011 study authored by BOE researchers Nick McLaren and Rachana Shanbhogue.
Search counts “are likely to become an increasingly useful source of information about economic behavior,” they wrote. Chris Shadforth, a spokesman, declined to comment further.
At the Bank of Spain, Concha Artola and Enrique Galan analyzed travel-related queries in Britain in a paper released in March. Their conclusion: Searches predicted the inflow of British tourists into Spain with a lead of almost one month.
At the New York Fed, Rebecca Hellerstein and Menno Middeldorp compared the popularity of the phrase “mortgage refinance” with an index tracking the number of refinancing applications filed by homeowners. In Italy, Francesco D’Amuri at the Rome-based central bank found that Internet searches generated jobless-rate estimates that were more accurate than ones relying on traditional leading indicators of the labor market.
Even in a developing economy like Chile, where less than half the population goes online, data on car-related searches helped predict auto sales, according to a paper from Central Bank of Chile economists Yan Carriere-Swallow and Felipe Labbe.
The central banks of Spain, Italy and Chile don’t use Google data to assess their economies at the moment, while Spain is exploring future uses, and Chile believes the results of research on the tool worldwide are promising, according to spokesmen and spokeswomen at the respective banks who declined to be named because of bank policy. New York Fed spokeswoman Andrea Priest declined to comment.
Investors hoping for an edge in the markets also are canvassing the trove of data left behind by Internet users. Brian Jacobsen, chief portfolio strategist at Wells Fargo Advantage Funds, uses search counts to examine what consumers intend to buy even before they actually make their purchases.
“The biggest benefit of this data is that you get an imperfect look at the mind of individuals,” said Jacobsen, who helps oversee $211 billion in assets in Menomonee Falls, Wisc. “Oftentimes, we can look at incomes, interest rates and past behavior and try to determine what individuals are going to do. This gives us a peek at what people are interested in at that very moment.”
Beyond forecasting, Google’s statistics are helping economists determine the effectiveness of policy. In an August 2011 paper, Mark Spiegel, vice president of international research at the San Francisco Fed, looked at search volumes of words such as “freeze” and “crisis.” He examined changes in perceptions of various countries’ default risk during the 2008 credit crunch, when the Fed was establishing swap lines with other central banks to ease global liquidity shortages.
“This is all in its infancy, but it’s fascinating,” San Francisco Fed President John Williams told reporters in March, referring to what he sees as a “data revolution.” It’s “an enormous amount of information” that will better help “us understand in very real-time what’s going on.”
With assistance from Simon Kennedy in London, Randy Woods in Santiago, Emma Ross-Thomas in Madrid and Lorenzo Totaro in Rome.
