Not all approaches are created equal, and there's no single forecasting tool to rule them all
Whether to assess the likelihood of interest-rate cuts or to determine when it’s time to create a more defensive investment strategy, most if not all financial professionals would benefit from recession warnings. But given the lack of consensus among analysts and models, there has to be a process to separate useful signals from distracting noise.
That’s what James Picerno, editor of the well-respected finance blog CapitalSpectator.com, contended in a recent piece. He explained that recession analytics, which informs the discourse behind gloom-and-doom news, can be generally categorized into backward-looking summaries, estimates of the current state of the economy, and projections of the future.
“[A]s you move from 1 to 3, your skepticism should rise,” Picerno said. “Too often, however, all three are treated as equivalent.” He also pointed out that the availability of published data — or lack thereof — influences the reliability of a given recession estimate.
Citing one recent headline, he said that Bank of America Merrill Lynch survey of U.S. credit investors was presented as proof that “recession fears [in the U.S.] have hit an all-time high.” A deeper look at the details shows that “the perceived probability of a recession in the next year spiked to 25%,” but the data set that the headline is based on goes back just three years.
In contrast, a recent update to the Philadelphia Fed’s ADS Index showed a backward-looking measure of economic activity, based on hard data, showed a moderate advance to slightly better-than-average conditions relative to its benchmark’s multi-decade history. The latest reading, estimated at +0.139 as of September 7, was well above the tipping point of -0.8 that signals the start of a recession, according to one San Francisco Fed paper that sets out recommendations for analyzing ADS data.
Turning to forecasting, Picerno noted that such forward-looking projections come in many varieties, resulting in wide variation in reliability and accuracy. “It’s a safe bet that the weakest results will be found in the forecasts that look far ahead, as several studies confirm,” he said. Citing research published in the International Journal of Forecasting in 2013, he said that forecasts of U.S. GDP were considerably more accurate for short-term horizons compared to those made for longer terms.
To arrive at the best decision, he advised consumers of business-cycle analytics to not rely on just one indicator — an inversion of the yield curve, for example — but to triangulate their moves based on a mix of the different types of estimates.
“All three are useful, assuming they’re designed intelligently,” he said. “[C]arefully aggregating data from all three disciplines will likely offer a higher level of reliability and timeliness compared with looking at any one set of metrics in isolation.”