November 8, 2012
Beware Of Zombie Recession Forecasts
With the election behind us and the fiscal cliff approaching, recession forecasting is in full swing again, and so it's time once more to roll out the standard caveat—not all predictions are created equal. In fact, quite a lot of the opinions are of poor quality, largely because one or more of the following applies: 1) the predictions are driven by emotion; 2) the analysis relies on cherry-picking the data; 3) the analyst is generally misreading and abusing the economic signals and models; 4) the analysis is overly focused on recent data that's probably infected with short-term/seasonal distortion; 5) the analyst has another agenda to promote that conflicts with objective macro analysis.
For example, some pundits claim that there's a clean, direct link between the business cycle and policy debates in Washington related to decisions that may or may not happen in the future. For example, Steve Forbes predicted yesterday that "we will have a recession." Yes, that's a reliable forecast--now and forever. There's always another recession lurking. Timing, however, is a complicating factor. That's a reminder that we must consider the source--the model--for any recession prediction.
On that score, Forbes' analysis looks wobbly: "Raising taxes on capital, raising taxes on small businesses, which we will likely get now, particularly since the Republicans did so badly in the Senate races, that is going to pose a real burden."
Sounds plausible, in a warm and fuzzy way. But if you spend any time analyzing the business cycle, you'll quickly discover that the link between macro fluctuations and tax rates in the short-to-medium term is clear as mud. As a tool for deciding if recession risk is high or low, rising or falling, this approach is worse than useless. If it were otherwise, your first source for predicting recessions would be listening to debates in Congress and press conferences at the White House. Good luck with that.
Fortunately, there's a better way, although it doesn't lend itself to quotable commentary in 30-second sound bites: Analyzing and monitoring a broad set of economic and financial indicators. Aggregating and tracking the broad changes in the data is the idea behind The Capital Spectator Economic Trend Index (CS-ETI). It's not perfect—nothing is—but it's performed admirably since I rolled it out publicly in the summer (I tested it privately for nearly a year before that). There are pretty good odds that CS-ETI, after a long period of testing and tweaking, will continue to provide reliable real-time signals on recession risk in something close to real time.
On that note, not much has changed since October 17, when I last published a CS-ETI update. In other words, recession risk is still quite low, based on data through September. A few more data points for September have been released since October 17—personal income and spending—and the signals remain positive. As usual, I'll continue to monitor the incoming data from various sources and run the analysis for any statistical hints of trouble. That includes looking ahead several months via econometric techniques for projecting where CS-ETI appears to be headed and keeping an eye on the vintage data (see the third chart in the link above for a recent example). I also translate CS-ETI into probabilities via a probit model for another take on how the business cycle is faring. Crunching the data on GDP nowcasts adds another dimension to the analysis. (I'll have updates for all of these reports later in the month.)
The basic message: it's essential to look at the data from multiple angles, regularly, in an econometrically intelligent way. It's not rocket science, but you can't whip up an intelligent review in a few minutes either. What could go wrong? The main risk is that a bolt from the blue renders the latest data points irrelevant. That's a risk for every model, including CS-ETI. For example, new war in the Middle East that quickly sends oil prices to the moon, or a failure in Washington to resolve the fiscal cliff threat by January, could trigger a recession—risks that aren't reflected in the current numbers.
Keep in mind, however, that if the economy does start tipping over the edge, it'll be obvious in the data, and relatively soon. That is, over the course of two, three or four months, history suggests that clear signs of fatal macro deterioration, for any reason, will be increasingly conspicuous. But isn't that too late as a practical matter? Not necessarily. Most folks, and even many professional economists, don't recognize the early stages of rising recession risk until it's blindingly obvious. How could they miss it? I think one reason is that they're not routinely monitoring a broad set of numbers. Business cycle analysis that's focused on calling recessions in real time is a highly specialized field, and most of the time the analysis isn't practical for the simple reason that recessions are relatively rare events in the grand scheme of macro. No wonder that contractions and the associated signals mislead so many people.
That leaves a fair amount of opportunity for identifying high levels of recession risk several months ahead of the crowd. Sure, it'd be better if we could identify the exact moment when a recession begins, but that's impossible because a high-confidence assessment requires a wide array of data—most of which arrives with a lag.
The best-case scenario that's both practical and relatively reliable is looking for a high-probability signal that a recession has recently started—and identifying such a tipping point as early as possible. The good news is that this type of analysis is feasible, although it requires a fair amount of vigilance in tracking the numbers.
It also helps quite a lot if the analysis is transparent, replicable, and uses publicly available data. All of those points apply to CS-ETI, but that's hardly the norm. It's one thing to claim to have a great business cycle model, but if the understanding the mechanics requires a Ph.D. and deep insight into highly complex calculations, the result for most folks is that they're looking at a black box that relies on the kindness of strangers.
On that note, I'll have an update of CS-ETI soon—in a week or so, once more October numbers are released. Meantime, the early reports for last month look modestly encouraging for expecting that October will go into the history books as another month of growth for the U.S. economy. Monday's news on the services industry, for instance, looks decent. Ditto for last week's economic updates.
Meanwhile, don't let the economic porn that spews regularly from the usual suspects distract you from the necessity of a hard, objective look at the facts when it comes to business cycle analysis. There are lots of things that could go wrong (or right) in the months ahead, but there's a big risk of declaring a new recession before the data supports such a call. Ignoring the telltale signs of a new downturn is no trivial risk. But the same is true for yelling fire in the macro theater without sufficient cause. The goal is finding a reasonable compromise, and that takes hard work—day in, day out. Some pundits would have you believe otherwise. Fortunately, that's one risk that's easily avoided.
Posted by jp at November 8, 2012 7:22 AM
Spot on post.
Posted by: Orbea at November 8, 2012 8:09 AM