It turns out the news media isn’t fake after all.
Financial news headlines are proving a useful measure of broad economic sentiment for forecasters seeking to deepen their arsenal of data, according to a new report from the Federal Reserve Bank of San Francisco.
“Researchers have recently developed ways to extract data on sentiment from news articles using text analysis and machine learning techniques,” write Adam Hale Shapiro and Daniel Wilson, research advisors at the San Francisco Fed.
“These measures of news sentiment track current economic conditions quite well. In fact, they often do a better job than standard consumer sentiment surveys at forecasting future economic conditions.
“The results corroborate previous studies showing that sentiment has predictive power for future economic activity,” the authors write. “Sentiment extracted from newspaper articles correlates with both contemporaneous and future key business cycle indicators. These methods of sentiment text analysis hold great promise for improving our understanding of news sentiment shocks and how they affect the economy.”
The authors describe the difference between “hard” indicators of economic activity like growth and employment and “soft” ones, such as measures of consumer sentiment based on surveys. Indicators that use increasingly complex verbal analysis tools fall into the latter category.
“Now more than ever, information about the public’s perceptions of the labour market, prices, interest rates, and other economic conditions is readily available in digital format online,” the authors say. “At the same time, recent advances in computational techniques for quickly analysing big data — particularly large volumes of text — have made it possible to measure and summarize public perceptions of economic conditions in a way that could be useful to policymakers and analysts trying to track and forecast the economy.”
Still, “soft” data can still be very useful.
“Newspaper articles and editorials about the economy do more than just report on official data releases. They also often convey how the journalist and those interviewed feel about the economy,” the paper adds. “Using the news sentiment indexes enhanced a model’s forecasting power in many cases … [and] provide information that can complement existing consumer sentiment indexes.”
Financial journalists, it seems, are providing an additional public service without even knowing it. You’re welcome, Mr. President.