Will the second decade of the twentieth century be the decade of "Manufacture, Baby, Manufacture"?
It seems increasingly that way -- I had referenced the unusual recovery in manufacturing employment since the 2008 recession in an offhand way in my post on industry-level signaling, and that sparked my curiosity to see to what extent this recovery is cyclical, and to what extent is it structural.
The answer is that I cannot explain the recovery of manufacturing employment with a cyclical story. In short, what's going on in manufacturing employment represents, to a large extent, a turnaround at the industry level, rather than broader macroeconomic gains.
I'll show this in two different ways. The first is quite simple: looking at annual percent change in nonfarm and manufacturing payrolls, manufacturing has always grown slower than nonfarm as a percent change (or it has shrunk while payroll has grown) at all times in the last twenty years. That changed in November 2010, when manufacturing employment began growing at a faster year-over-year percent change than nonfarm. In April 2012, manufacturing payrolls grew 2 percent year-over-year, as compared to nonfarm growth of 1.4 percent.
That gap of 0.6 percent year-over-year may seem small. But historically the American economy has not seen anything like this in a long, long time -- from 1992 to 2010, nonfarm payrolls grew on average 3 percent faster than manufacturing payrolls year-over-year.
Another way to examine the change and discern whether there is a structural component is to look at a scatter plot of the year-over-year percent change in manufacturing vs. nonfarm payrolls. What you see is a positive-slope line running through the chart -- the relationship between manufacturing and nonfarm payrolls is strong, with a coefficient of determination of 0.894. (This means that macroeconomic variables explain 89.4 percent of the change in manufacturing payrolls since 1940, when my dataset begins.)
But the intercept of this positive line is changing -- it has been rising since June 2007. What this means is that nonfarm payroll growth of X percent year-over-year used to induce manufacturing payroll growth of Y percent year-over-year, and now it induced manufacturing payroll growth of Y + b percent year-over-year, where b is a non-negligible positive number.
In fact, using the last year's worth of data points as a sample, I determined the intercept of this linear relationship -- what this means is if nonfarm payrolls were to grow at 0 percent year-over-year, at what year-over-year percent rate one should expect manufacturing payrolls to grow?
Over the past two decades, the intercept had been largely in negative territory -- normally between -2 and -5, with the severe manufacturing contraction during the 2000 recession seeing a moment in which the intercept hit -15 for the year.Now, however, the intercept is positive, and increasingly so, as more and more of the data comes to reflect manufacturing payroll growth in excess of nonfarm payroll growth. Moreover the change in intercept looks visibly different from nearly every other instance we have on record of a positive intercept -- a sustained and gradual shift rather than a momentary shock. So this manufacturing recovery is for real.