Can Big Data Help Measure Inflation?

A new database tracking online purchases and prices might provide insight beyond traditional government statistics.

Keith Bedford / Reuters

There are two economic acronyms that should ring a bell even with those who aren’t experts: GDP and CPI. They’re two of the most important government statistics regarding the economy: The first, the gross domestic product, measures a country’s economic output, while the second, the consumer price index, records the rises and falls of the prices of everyday goods—which, when combined together, describe inflation. Accurate measurements of GDP and CPI are hugely important for central bankers, because they use those determine monetary policy.

In the last decade, though, the government has had a harder time measuring CPI. Their method is usually to go around from store to store, taking stock of prices around the country. But e-commerce now accounts for around 7 percent of U.S. GDP, which means online spending is an important component of the CPI. As more and more people are shopping online, calculating this index has gotten more difficult, because there haven’t been any great ways of recording prices from the sites disparate retailers.

Data shared by retailers and compiled by the technology firm Adobe might help close this gap. The company is perhaps known best for its visual software, including Photoshop, but the company has also become a provider of software and analytics for online retailers. Adobe is now aggregating the sales data that flows through their software for its Digital Price Index (DPI) project, an initiative that’s meant to answer some of the questions that have been dogging researchers now that e-commerce is such a big part of the economy.

The project, which tracks billions of online transactions and the prices of over a million products, was developed with the help of the economists Austan Goolsbee, the former chairman of Obama’s Council of Economic Advisors and a professor at the University of Chicago’s Booth School of Business, and Peter Klenow, a professor at Stanford University. “We've been excited to help them set up various measures of the digital economy, and of prices, and also to see what the Adobe data can teach us about some of the questions that everybody's had about the CPI," says Goolsbee. "People are asking questions like ‘How price sensitive is online commerce?’ ‘How much is it growing?’ ‘How substitutable is it for non-electronic commerce?’ A lot issues you can address with this in a way that we haven't really been able to do before." These are some questions that the DPI has the potential to answer.

While Goolsbee says that the CPI is still the gold standard, he believes that the Bureau of Labor Statistics (BLS), the Fed, and other analysts who use price data would be “extremely interested in having real-time measures of prices." One notable finding of Adobe’s DPI, for instance, is what has happened to the prices of electronics in the past year. While the CPI reports 7.1 percent deflation for computers and 14.4 percent for TVs over that time period, the DPI found 13.1 percent and 19.4 percent.

Another advantage of the Adobe data, according to Goolsbee and Klenow, is that it gives a sense of how many units of any given product are being sold, which helps economists identify instances in which consumers substitute one product for another. Some substitute goods, such as Coke and Pepsi, are obvious, but others, such as a bike and a car, are less intuitive. “That's a basic economic concept, and we don't have great data because we don't have great quantity data in seeing how substitutable things are,” says Klenow. “Being able to see quantities as the prices are changing is critical to that.” Klenow says that seeing what share of online shoppers’ spending any given product accounts for is also very helpful in figuring out the impacts of price fluctuations.

While this new trove of data will certainly be helpful to economists and analysts looking at inflation, it surely won’t replace the CPI. Currently, the government sends out hundreds of BLS employees to stores around the country to collect price data. Online pricing is a small part of the BLS calculation, which is incorporated into its methodology as people increasingly report shopping from retailers online, but there’s a significant time lag. While it’s unlikely that the BLS would incorporate private sources of data into its inflation calculations, as e-commerce grows they might look to improve the way they include online prices. Still, economists are optimistic about the potential of Adobe’s DPI. “I don't think we know the digital economy as well as we should,” says Klenow, “and this data can help us eventually nail that better.”