By JENNIFER MERRITT
Payroll and unemployment data released Friday offered a snapshot of the labor market's health last month. But some researchers say they can get a read on such trends days or weeks ahead of the official numbers by studying Google searches, tweets and even queries at an online phone directory.
Economists painted a mostly positive picture of the latest government data, which showed the economy shed fewer jobs than expected in February and the jobless rate held steady. For people who look at early-warning indicators, the figures were no surprise. The Web-based data have been telling a similar story for at least a month: The job market is getting better—very slowly.
Development of the new techniques is still in its early stages, but researchers say the data could become a part of the traditional forecasting models—or replace some outdated predictive surveys and statistics.
Economists who analyze the job market using more conventional techniques are already paying attention. "The great benefit is that these [data] show people actually doing something," said Ed Leamer, director of Anderson Forecast, an independent economic-forecasting group at the University of California, Los Angeles.
Still, Mr. Leamer and others don't yet use such indicators in their official forecasts because the tools haven't been around long enough to prove their worth.
Beyond that, the habits and makeup of people who use the Internet are in flux, which can make it difficult to isolate the economic message sent by searches and comments.
A study of daily Twitter activity conducted for The Wall Street Journal by Crimson Hexagon, a firm that builds algorithms to look for key phrases that are indicative of consumer mood, shows a gradual decline in the number of people tweeting about being unemployed and looking for work over the six months ending Feb. 16. The study picked up tweets with phrases like "still looking for a new job" or "I just had an interview." The company also sees a gradual decline in those worried about losing their job in the past two months.
Hal Varian, who left a teaching post at the University of California, Berkeley, to become chief economist at Google Inc., conducted an analysis of searches using a Google tracking function called Insights that pointed to the trend in initial claims for unemployment benefits seven days before the government released them.
The same tool showed more Web searches were conducted over the past two months for "job interview questions" and "what to wear to an interview," phrases that likely mark the beginning of the hiring process. And Google can pinpoint the geographic areas producing the most searches. People typing in Minnesota, for instance, increased searches for job interview questions and answers more than any other state between late October and mid-February.
"I think of it as taking the pulse of the economy," said Mr. Varian, whose team is exploring whether Google wants to get further into forecasting. "We think [the data] definitely have predictive power."
Yellowbook.com, an online phone directory, has devised an indicator based on the types of businesses consumers search for the most. Yellowbook.com says an upturn in searches related to home improvement and remodeling began in October, presaging a fourth-quarter rise in such sales. In late January, Home Depot and Lowe's reported increased fourth-quarter sales.
Researchers also are looking to search patterns for clues to other aspects of the economy. Erik Brynjolfsson, of Massachusetts Institute of Technology's Sloan School of Management, has used Google search volume to forecast home sales. Mr. Brynjolfsson predicted two months in advance that National Association of Realtors' reading of home-sales volume would rise 6.4% in the third quarter, compared with a forecast of 8% by the NAR. The actual rise: 6.1%.
Write to Jennifer Merritt at jennifer.merritt@wsj.com
![Reblog this post [with Zemanta]](https://lh3.googleusercontent.com/blogger_img_proxy/AEn0k_toMAlKDVldCwLQb5ITJ9rMFLUVRDs5Gj7LbkOwREHyMq-4PQFAikNAQhFKejb0dK68xgxsiH5wu8OkZN2fQICuVIzMFsckYUS2hRoKtLo4mIj3XmfZ8IxHIsE3BfYvaYqrtMVV-h01JlWNg4inVek=s0-d)
No comments:
Post a Comment