Via Noah Smith for Bloomberg
There’s a divergence between “soft” measures of sentiment and “hard” data. Partisan leanings offer an explanation.
A strange thing seems to be happening to the U.S. economy. On surveys, businesspeople and consumers say the future looks bright. But recent economic activity hasn’t appeared very robust.
Andrew Ross Sorkin of the New York Times noted this in a recent article about mergers and acquisitions. A number of surveys have been reporting that chief executive officers are highly optimistic. For example, the website Chief Executive and the Wall Street Journal/Vistage Small Business CEO Survey both report a surge in CEO confidence since the 2016 election, while Business Roundtable’s CEO Economic Outlook Surveyfinds an average level of confidence.
But as Sorkin reports, M&A activity is at its lowest level since 2013, and has fallen 40 percent in the past two years. Share buybacks have also slowed. Those “hard” numbers indicate that whatever CEOs are saying on paper, they aren’t taking actions that signal confidence in the future of their businesses. Capacity usage, which fell slightly in May, is another indicator of that true business sentiment is far from giddy.
Another example is consumption. The University of Michigan’s Surveys of Consumers show confidence at the highest levels they’ve been since before the crisis.
But again, some hard numbers tell a different story. Retail sales fell in May, and have been relatively lackluster for the entire year so far.
Auto sales are falling as well. Since cars are expensive, long-term purchases, consumers often signal lack of optimism by holding back on the purchase of a new car, choosing instead to drive their old model for a little while longer. So this is another data point that belies rosy consumer confidence numbers. Pending home sales provide a third spot of weakness.
Employment isn’t particularly strong either. Nonfarm payrolls expanded by only 138,000 in May, lower than the 185,000 that had been forecast. That follows another middling month in April and a dismal 79,000 in March.
Why the divergence between the “soft” numbers of confidence surveys and the “hard” numbers of the real economy? One possibility is that this is just a momentary spot of economic weakness, and the numbers that measure sentiment point to better days in the near future. But survey numbers have been rosy for a half-year now, so if these surveys were doing their job of forecasting the real economy, it seems like the good times they predict would have started to show up in the data by now.
It’s also possible that the hard numbers are just very noisy and full of error. That’s always a danger with up-to-the-minute analysis of the latest economic statistics. For example, capacity utilization rose substantially in April, so the slight fall in May might be due to the correction of a random mismeasurement the previous month.
Another possibility is that the seemingly weak “hard” numbers are cherry-picked by me and others. No one knows exactly which economic numbers to trust at any given moment in time. The Conference Board’s index of leading economic indicators was actually up slightly in May. But Morgan Stanley reports a “record gap” between hard and soft numbers in recent months, so the phenomenon seems real.
A third possibility — and one I personally find likely — is that surveys of consumer and business confidence have systematic errors that make them unreliable in certain economic and political climates but not in others.
Surveys have some predictive power, but only a bit. The Index of Consumer Sentiment is believed to explain only 13 percent to 26 percent of the variations in economic output. Similarly, financial surveys have a slight bit of ability to predict the stock market, but only when many are used in conjunction — and even then, the signal is a weak one.
With all this noise, error and uncertainty, there’s plenty of room for systematic errors to arise. Surveys may be very predictive in certain conditions but terrible in others. If so, the U.S. could be in the middle of one of the latter periods.
One big source of systematic error is partisanship. It’s well known that the party of the president has a dramatic effect on people’s views of the economy. Before the election in November, Democrats held a much more favorable view of the economy than did Republicans; after Trump won, that gap promptly flipped. Republicans now claim to be extremely confident about the economy, while Democrats report pessimism.
If CEOs are more likely to be Republican, then partisanship could explain the weirdly high CEO confidence numbers. The bias in the Surveys of Consumers is harder to assess, but the emotionally charged election seems like it could have easily had an effect on survey response rates.
So it’s possible that survey numbers are registering Trump supporters’ euphoria rather than a real hard-nosed forecast of better times ahead.
Sentiment surveys aren’t useless, but they probably warrant less attention than the economy’s hard data.