In recent years, there has been plenty of discussion about looking at Internet search data to map and predict trends. Now, economists are saying say that Internet search data may even now be a better predictor of economic trends than quantitative surveys.
Economists have relied on market survey data for years. Housing starts, household employment surveys, business confidence indexes, purchasing management surveys, and regional manufacturing surveys all are closely watched for indications of growth or decline ahead.
But there may be enough data now generated from the Internet to paint an even more expansive picture of what’s around the corner, some economists say.
Sendhil Mullainathan, behavioral economist at Harvard University, recently spoke with NPR’s Kai Ryssdal, says Internet searches are emerging as a leading economic indicator. “Once the sales data come in from Hionda, it’s too late,” he says.
As Ryssdal put it: “Think about all the Google searches you’ve done today. Multiply that by 365 days in a year, multiply that by pretty much everybody around you, and what you have is piles and piles of data. Data that’s valuable not only for advertisers, but for economists as well who are mining it for all kinds of things they can learn about us.”
Rebecca Hellerstein and Menno Middeldorp, writing in the newsletter of the New York Federal Reserve Bank, report that “Internet search counts possess useful information, not available in other variables, to now-cast or forecast the trajectory of some financial market data. While this predictive power is by no means universal, the basic message is of a useful addition to the economist’s toolkit.”
That’s because “a key challenge for economists has been to identify indicators that provide timely information about the release before it comes out,” they add. Counts of Internet search terms may predict trends in financial markets.
Some closer-to-the-ground measures may also reveal themselves in search term counts as well, Mullainathan says. “Look closely at phrases like ‘foreclosures’ or ‘pawnshops,’ it becomes a leading economic indicator. So when these things start to go up, well it looks like the economy is starting to tank,” he says. Conversely, he says, “we look and see if whether people are searching for things like car purchases, ’cause those are pretty good leading indicators that indicate people are starting to feel times are good.”
In a report published last year, Nick McLaren and Rachana Shanbhogue also concur that Internet search data is a viable economic indicator. Consider employment trends, they write: “It is now likely that people who are unemployed, or fear they may soon lose their job, will search on the internet to find out about [unemployment benefits] and to search for new jobs,” th economists say. “So Internet search terms may be useful for monitoring the labor market.”
Housing prices are another area that search terms can capture before they start moving one way or another. “For house prices, the results are somewhat stronger,” the report adds. “Search term variables can outperform some existing indicators over the period since 2004. There is also evidence that these data may be used to provide additional insight on a wider range of issues which traditional business surveys might not cover.”
In addition, McLaren and Shanbhogue add, people interested in both buying and selling properties “make use of the internet to monitor market conditions and advertise their properties,” thus reflecting a variety of conditions in the housing market.
The economists add that mining search data offers a range of advantages over traditional surveys. “These applications treated the search data in a similar manner to existing surveys in conducting standard regression analysis,” they point out, adding that search data also has the potential to answer different sorts of questions that surveys can’t answer:
Internet search data “can help analyze issues that arise unexpectedly. Whereas survey data must be consciously collected based on pre-determined questions, internet data are collected based on behavior at the time, and a backrun will be available provided the term was searched widely on the internet.”
(Photo by Joe McKendrick.)