Talking Manufacturing and Its Wage Premium

May 21st, 2014 at 5:10 pm

Spoke on a Brookings manufacturing panel this AM and since I actually did some number crunching for this one, allow me to briefly recount my argument and show some new results on the manufacturing earnings premium.

Numerous speakers argued that, as with agriculture, productivity in the factory sector will continue to chip away at manufacturing employment.  If there was a consensus, lightly held, I’d say it was pretty pessimistic about the sector’s prospects.

While I don’t buy into the notion of some ongoing “manufacturing renaissance,” neither am I convinced that the sector’s future is necessarily bleak:

–Manufacturing has the potential to grow beyond its current size and employ more people in decent jobs, but we’ve got to get the policy right.

–As per Susan Houseman’s work, manufacturing productivity is significantly overstated for a number of reasons (this paper is a must read if you’re in this debate), implying that the agriculture analogy is incorrect.  For one, it’s largely driven by one relatively small sub-sector: the production of computers and electronic products.  Houseman: “Over the 15-year period [1997-2012], the quantity produced at U.S. factories increased by 38 percent in official statistics; without computer and electronic products manufacturing, output increased by a mere 7 percent.”

–Second, Houseman points to two problems with the way we count inputs that also biases up estimates of manufacturing productivity.  First, we’re undercounting the quantity of imported intermediate inputs and thus overcounting value added.  Second, when manufacturers employ workers from outside the sector, say from Manpower-type staffing services, that biases down measured sectoral labor inputs and further biases up value added.

–Even in the highly productive computer sector, productivity-induced job loss is not obvious.  Houseman again: “Productivity growth also may reflect improvements to product design that result from research and development activities. In fact, research and development, not automation, largely underlies the declines in price deflators and, correspondingly, the rapid increases in measured real value-added and productivity for computers and semiconductors. Computers today may be far more powerful than previous models, but this fact, by itself, does not mean that it takes fewer workers to assemble a computer today than in the past. Within manufacturing, the computer and electronic products industry has sustained above-average employment losses since 2000, yet a recent report concluded that these losses are entirely attributable to the shift in production to Asia. The reason employment in electronic computer manufacturing has declined by 41 percent since 2000…is not because the assembly process has been automated but because most computer assembly has moved to Asia.”

–Despite rumors to the contrary, manufacturing still pays better than other jobs for workers with similar characteristics.  More on that in a sec.

–Current policies, however, are helpful but insufficient.  Basically, we provide a bunch of scattershot tax advantages of dubious merit, and, from the White House, an export strategy, and the IMIs (Innovation Manufacturing Institutes—read this to learn about them).  I think the IMIs make a lot of sense (Germany has about 60 up and running) though I don’t think anyone can speak to the magnitude of their impact.  An export strategy that’s not a “net export” strategy seems woefully insufficient.  If we double our manufacturing exports while our trading partners triple their exports to us, we’ll still be exporting too many factory jobs.

–No one knows the true potential of the manufacturing sector.  The pessimists are (not unreasonably) looking backwards and (less reasonably) are convinced that productivity and automation will permanently constrain employment growth.  The optimists are conflating exports with net exports and seeing a renaissance in a flattening trend after the terrible 2000s.

–But I can tell you this: if we continue to largely ignore the currency management in which our competitors engage to give their exports a price edge, I don’t see how US manufacturers can have a chance to realize their potential.  See here for ideas to fight back (I particularly like the “Levin bill” and reciprocity).  Somewhat to my surprise, in a politically balanced panel, all panelists agreed with this point (though some weighted it less than others).

Okay, can I show off my number crunching now?

If you go to the monthly BLS jobs report, Table B-3, you’ll see that average hourly earnings in manufacturing (just the wage part, no benefits) was $24.70 in April 2014, compared to $24.31 for the overall economy’s wage.  Hard to see much of a manufacturing premium there.  The weekly differential is larger ($1,008 versus $839) but that’s clearly a function of more hours worked in the factory sector.

That’s led some to argue that the old claim that “manufacturing jobs are better than average jobs” is no longer correct.

I disagree.  You have to compare compensation, not just wages, and to really measure a wage premium, you have to control for other wage determinants.  The table shows hourly compensation values from the most recent published data for key sectors, along with manufacturing’s benefit share of hourly comp, and the wage premium in manufacturing from an hourly (log) wage regression with the controls you see listed.  Re the regression output, unfortunately, there are no publically available microdata with compensation beyond wages.  But if there were, I strongly suspect that given the higher likelihood of benefits in the factory sector relative to say, services, all else equal, the relative manufacturing premium would be larger.

This premium is probably smaller than it used to be, but it’s still statistically significant and economically meaningful.


Sources: BLS ECEC data and my analysis of CPS ASEC data for 2013 (h/t: AS)

Print Friendly, PDF & Email

4 comments in reply to "Talking Manufacturing and Its Wage Premium"

  1. Robert Buttons says:

    If we read between the lines, we can see Jared is talking Say’s Law. I must be finally turning him into an Austrian!!

  2. Brett says:

    It’s an 11% difference on average. That’s not small, per se (it’s an extra $6000 a year), but it’s not really huge. And I’ll bet the farm that it will be automated away in the US, since most of the automation seems to be focused at pushing productivity in factories higher.

  3. Guest says:

    Don’t forget a huge implicit subsidy is from local government’s zoning codes. There are large swaths of high-in-demand land that are zoned for manufacturing only in many coastal cities. This provides them with cheap land that would otherwise be used on different sectors (commercial, residential etc).

  4. Spencer says:

    I suspect that the Fed’s methodology for calculating industrial production actually causes the official data to significantly understate its growth..

    Each year the Fed re-weights the index according to its nominal weight —-price times real output.

    This is a reasonable method most of the time. But when you have one sector like information technology (IT) experiencing massive and rapid price declines it distorts the data.

    As you pointed out IT output growth has massive outpaced all other sectors growth.. Normally you would expect that the faster growing sector would have a greater weight in the next year. But because of the Fed’s using nominal weights and the sharp drop in IT prices, it’s share or weight in the industrial production index actually declines every year. It now only has a weight of about 7.5% in the index.

    Because of this I suspect that US industrial production growth is biased downward.