If you want to know where the economy is headed, new research from the San Francisco Fed suggests you should just keep reading newspapers.
Two San Francisco Fed economists, Adam Shapiro and Daniel Wilson, used computer linguistic techniques to extract the “emotional content” from articles in 16 major U.S. newspapers over a 35-year period ended in 2015. They then built new monthly news sentiment indexes.
Computers can be trained to find emotions from groups of words after being trained using data from a social-network website. Researchers developed indexes constructed from the model for “negativity.”
Think you would win an argument with a robot? Watch this.(1:46)
Osaka University and NTT Laboratories recently created the first "discussion dialogue" androids – the only robots that have human-like independent cognitive abilities, as opposed to scripted speech.
When the news was negative, employment, production, stocks and spending were all sinking, the study found.
Using standard measures of job growth, stock prices, consumer spending and factory output, the research found that the new sentiment indexes tracked current economic conditions quite well.
But the news sentiment indexes also had “significant power to predict these variables up to 12 months ahead,” the researchers said.
The news sentiment indexes were not without flaws. Researchers said that the indexes were poor forecasters of future inflation trends.