The recession drumbeat grows louder. This is not unexpected. Most forecasters have an asymmetric loss function; the cost of being wrong by missing a recession exceeds the cost of being wrong on a recession call. Hence economists tend to over-predict recessions. Eight of the last four recessions or so the joke goes. And while I don't believe a recession is imminent, there are perfectly good reasons to be wary that a recession will bear down on the economy in the not-so-distant future. Historically, when the Fed begins a tightening cycle, the clock is ticking for the expansion. By that time, the economy is typically in a late-mid to late-stage expansion, and you are looking at two to three years before the cycle turns, four at the outside.
Of course there are some not so good reasons for worrying about a recession. Like listening to an investor talking their book. Or someone who needs to whip up a never ending stream of apocalyptic visions to hawk gold.
So what I am looking for when it comes to a recession? It's not a recession until you see it economy wide in the labor markets. When it's there, you will see it everywhere. Clearly, we weren't seeing it in the final quarter of last year. But, you say, employment is a lagging indicator, so last quarter tells you nothing. Not nothing, I would say, but a fair point nonetheless. One would need to look for the leading indicators within the employment data.
First, since the manufacturing sector is the proximate cause of these recession concerns, we would look to leading indicators in that sector. One I watch is hours worked:
Hours worked are off their peak, just as prior to the 1900 and 2001 recessions, but not the 2007 recession (lagging indicator that time). But hours also dropped in 1994, 1998, 2002, and 2005. And that would be an extra four recessions that didn't happen. To add a bit more confusion, hours works are coming off a peak not seen since, sit down for this, World War II:
That caught me by surprise; I am thinking the surge in hours worked was not sustainable in any event. Overtime hours worked holds a bit more promise:
OK, not much more promise. Best as a leading indicator ahead of 2001, not counting 1994 and and 1998. Not particularly useful for 1990 and somewhat useful ahead of 2007. On balance, I would say manufacturing hours worked data is necessary but not sufficient for a recession call.
Perhaps the JOLTS data offers something more:
Unfortunately we a working with only two cycles here, and then only barely so. But it seems reasonable that manufacturing hires might be a coincident indicator (maybe leading by the few data points ahead of the 2001 recession) and layoffs/discharges a lagging indicator. But if a manufacturing "recession" were underway, then we would expect hiring to drop off quickly here.
Quits, however looks like a leading indicator:
Looks like quits in manufacturing dropped sharply ahead of 2001, modestly during 2007, but were still rising at the end of 2015. If quit rates aren't dropping among those at the front lines, the pain can't be reaching recessionary levels just yet.
But manufacturing is just one sector of the economy - just 8.8% of employment. The real hypothesis the recessionists are proposing is that manufacturing is an indicator of an economy wide shock. Here I would say the JOLTS data is less supportive:
If we are entering a recession, firms are a minimum should be pulling back on the pace of hiring. We are not seeing that yet. And workers should be wary of quitting:
Again, the workers are on the front lines of the economy. If the economy is in trouble, they know it, and quit rates start declining. Not there yet.
I also have a soft spot for the temporary help series as it as rolled over twelve months or more ahead of the last two recessions:
So if we were to see temporary help roll over now, we would still not see recession until 2017.
And finally, there is initial jobless claims, which typically lead a recession by six to twelve months:
Not seeing it. If claims started rising now, and continued rising for six months, then the probability of recession would rise sharply, and if they rose continuously for twelve months, the probability of recession would approach 1. But now? Nothing to fear.
Bottom Line: From a labor market perspective, I am not seeing conclusive evidence of an impending recession in manufacturing, let alone the overall economy. Might be at the tip of one, but even that will take a year to evolve. I have more sympathy for the view that the economy has evolved into a mid-late to late stage of the cycle, and the transition and associated uncertainty results in some not-surprising volatility in financial markets.
Update: Greg Ip calls me out on the direction of forecast accuracy:
I was thinking of the frequency of recession scares - 1995, 1998, Y2K, 2012, now - on top of actual recessions. In 1998 I remember a forecast from JPMorgan calling for a recession due to a trade blow from the Asian Financial Crisis. I remember a particular dark forecast from Deutsche Bank for 2000 due to a Y2K. Then the ECRI 2012 recession call. Calculated Risk describes the last six years as an "an endless parade of incorrect recession calls."
That said, if you were to tell me that these recession callers were not on average "economists" I would concede the point. And forecasts from official agencies such as the IMF and the Federal Reserve always miss recessions (a point Larry Summers makes with regard to the IMF here.) So Ip's point is well taken.
Related:
