Justin Fox at Bloomberg put up a few articles (here, here) recently digging into trends in productivity in the US manufacturing sector. The overarching theme of the two articles is that claims regarding the success of the manufacturing sector in the US (he cites Binyamin Applebaum’s article specifically) are overstated. Fox’s first article is titled “No, U.S. Manufacturing isn’t Really Booming”.

I’m not going to tell you that manufacturing is booming. I’m also not going to tell you that Fox is right about it not booming. I am going to tell you that making claims about the business or financial success of manufacturing based on things like multi-factor productivity (MFP) or labor productivity, as Fox does, is using these measures incorrectly. The “health” or “success” or “boominess” of manfucturing are not things one can infer from productivity statistics.

I’ll pick on Fox’s articles here because they were recent, not because he’s done something especially egregious. They are two of many examples of how measures like MFP and labor productivity get misused to make points about financial or business performance.

The message is that you all need to just let productivity be productivity, you know?

MFP is about value-added, not numbers of goods

Fox takes issue with the measures of manufacuring multi-factor productivity because they involve quality adjustments made the BLS.

… the numbers are skewed by huge gains in real output in computer and electronics manufacturing that mainly reflect quality adjustments made by government statisticians, not increases in real-world sales.

There are couple of problems with this. One reflects some ambiguity. When Fox says “increase in real-world sales”, is he asking for a measure of business performance? I think, although I’m not sure, he means “increases in the number of physical things shipped”. And if he means increases in the number of physical things, then this is precisely why the BLS makes these quality adjustments.

Multi-factor productivity is simply a ratio of value-added to an index of inputs. Value-added in the manufacturing sector is a mesure of the economic value of all the goods produced. Not the physical number, the economic value. And hence manufacturing MFP is a measure of how much economic value that sector produces - not the physical number of goods - per unit of input used. I know this paragraph was redundant, but it is crucial to keep the distinction between value-added and the actual number of things clear.

If all we were doing was counting units, then no quality adjustments would be necessary. But we are trying to measure value-added, so they are. You cannot just add the 50 or so million iPhone 7’s that will be sold this quarter to the 20 million or so iPhone 6’s that will be sold to get 70 million phones. That ignores the fact that these items are different goods. And this problem gets exacerbated when I try to compare value-added over time. How do I compare the 50 million iPhone 7’s to the 5 million IBM PC’s produced in 1986? The BLS makes these adjustments to try and make the measure of value-added - and hence the measure of MFP - comparable over time.

You could quibble with their exact method - I think an iPhone 7 is roughly equivalent to 400 IBM PC’s, but the BLS thinks it’s equivalent to 500 - but without these quality adjustments, MFP would be uniquely scaled in each year. It would be impossible to compare MFP in 2016 to MFP in 2015, and hence to know if it grew or not. While imperfect, you can’t skip the quality adjustment if you want to make any comparisons over time.

The other sectors are not “more real” than computer equipment because they have smaller adjustment factors. You could easily argue that the other sectors are “less real” because they have improper quality adjustments. The BLS could be missing all the wonderful innovations we’ve made in refrigerator construction, for example, and understating MFP growth in that sub-sector.

Ignoring computer equipment in MFP does not give you a better read on what is “really happening” in the manufacturing sector. MFP is what is really happening to …. MFP. If you mean “are manufacturing firms profitable or successful or expanding or exporting”, then you mean something different than MFP.

MFP has nothing to do with corporate accounting

The prior section hints at how people make the mistake of evaluating the economy as if it were a business. The following quote shows what I mean.

… but the adjustments do raise questions about what exactly the productivity numbers signify. They don’t necessarily mean, for example, that computer and electronics manufacturers in the U.S. are actually growing, or succeeding in international competition.

No, they don’t. They’re not intended to. MFP is not a measure of whether firms or industries are “succeeding”, in any corporate or financial sense of that word. MFP is a measure of how much value-added the manufacturing sector produced for a given set of inputs. That’s it.

Value-added is not profits, in any financial accounting sense. The closest you can get to an analog in financial accounts is to take revenues minus cost of goods sold. And revenues minus cost of goods sold is not profits. Why? Because the firm still has to pay wages, interest costs, taxes, and dividends, among other things. So firms could easily be losing money, either in the sense of having negative net profits or a negative cash flow, even though their value-added is positive. Or maybe they are making money, and then good for them. But it is irrelevant to MFP whether the firms are making money or not. MFP and firm or industry “success” are fundemantally different things. One implies nothing about the other.

To get back to the prior section, the BLS is not trying to produce a quality adjustment so that their numbers reflect the financial accounting position of manufacturing firms. They are producing a quality adjustment so that MFP reflects a measure of how much value-added we get per unit of inputs. If their MFP number doesn’t mesh with your intuition or knowledge of the financial position of manufacturing as a sector, then … so? It’s not intended to.

MFP is not about competitiveness

Fox quotes Michael Mandel, of the Progressive Policy Institute, on why one should pay attention to MFP as opposed to labor productivity,

Because it includes a wider scope of inputs and costs, multifactor productivity is a better indicator of how innovative and competitive an industry really is.

No, it’s not. It has precisely zero to do with how competitive an industry is. First of all, “competitive” is a word devoid of meaning. “Competitive” could mean “cheap”, like textile firms in Southeast Asia, which have lower MFP than US firms. “Competitive” could mean “very low profits can be made” because entry of competitors is easy, like gas stations or coffee shops. That has nothing to do with whether MFP is high is low in those industries. “Competitive” could mean “produces a product that has an export market”, but that could just mean that the industry holds patents or other IP that makes it hard for others to replicate their product. And having intellectual property rights over a product doesn’t mean your MFP is large or small.

What Mandel probably means is “profitable” or “growing in revenues”, but I don’t know for sure. If his meaning is along those lines, then MFP doesn’t tell us anything about that, for the reasons I outline in the prior section. MFP is not a measure of profitability or market size.

It also isn’t a measure of how innovative an industry is, because as I’ve mentioned before on this blog, MFP is not a pure measure of technology or innovation. MFP is a garbage dump for everything we don’t know how to measure, including technology. So MFP can be stagnant even if an industry is innovative, or vice versa. If there is any imperfect competition in the industry so that markups exist, then MFP is affected by the growth rate of labor and capital, and the direction of this effect is indeterminate without knowing more about input growth. If there are re-allocations of inputs across firms within industries, then this influences MFP, and again the effect is indeterminate until you know which firm is shedding inputs and which one is gaining them.

MFP is a ratio, not a story

Because they are ratios, both MFP and labor productivity can rise because the numerator (value added) rises or because the denominator (all inputs or just labor) falls. It doesn’t matter which one moves.

On the whole, the U.S. manufacturing sector does not appear to have been a hotbed of innovation and productivity gains in recent years. It has gotten by with fewer workers, but in many cases that seems to have been more about managing decline through layoffs and plant shutdowns than boldly leaping into the automated future.

As we’ve learned, you cannot say anything about “innovation” from looking at MFP. But this is accurate that manufacturing has not been a hotbed of MFP or labor productivity growth in recent years. But then there are a couple things to unpack in the last sentence of this quote.

If MFP is stagnant but the manufacturing sector has shed workers, then this means either value added in manufacturing has fallen (it has not) or the other inputs like capital have risen (they must have). This indicates that labor productivity, value-added divided by number of workers only, must be rising.

But that’s it. We’re done. It is irrelevant to MFP or labor productivity what the story is behind the decline in labor and the increase in capital. The fact that you cannot attribute the decline of labor to lasers or robots or AI or hologram technology doesn’t change the fact that labor use declined, labor productivity went up, and MFP stayed stagnant. Closing a plant reduces labor in manufacturing, and if value-added doesn’t fall by enough, labor productivity in manufacturing just went up. It’s irrelevant for productivity that this is layoffs or shutdowns versus automated whatever.

Just let productivity be productivity. It measures value-added - not profits, not sales, not cash flow - relative to inputs, and therefore tells us how effectively the economy - not businesses - turns inputs into value-added. Slower MFP growth means slower MFP growth, not anything about competitiveness or the future business success/failure of US manufacturing. Heck, it doesn’t even tell us anything about the past business success or failures of US manufacturing.

Coda: Manufacturing is a relatively small sector

This last point isn’t directly related to the main post, but I like to harp on this, and its my blog, so you’re kind of stuck with it. Manufacturing represents a small slice of economic activity in the United States. And there is no reason why we should be particularly worried or concerned with productivity in this sector relative to other sectors, many of which employ more people and are responsible for larger portions of value-added.

Manufacturing used about 9% of the hours worked in the US in 2015 (all numbers from the BLS). That makes it about as important, in employment terms, to retail (10%) and the recreation, accommodation, food service sector (9%). It employs less than education and health (15%), government (15%), and profession and business services (14%). But no one seems to write articles worrying about the MFP numbers in any of those other sectors.

Drilling down, durable goods manufaturing - like, you know, “real” economic output - employed about 6% of hours worked in 2015. That means it is as large as … administrative and waste services (6%) and profession and scientific services (6%), and slightly ahead of the “other retail” sector (5%) and construction (5%). It’s behind accomodation and food (8%) and state and local government (12%). The vast majority of hours, and workers, are employed doing anything except putting together durable goods.

In terms of value-added, manufacturing accounts for about 11% of the aggregate, comparable to profession and business services (13%) and government (12%), and well less than finance and real estate (20%).

Manufacturing is not the dominant origin of employment or value-added. Which means that any action in labor productivity or MFP in manufacturing has a small effect on aggregate labor productivity of MFP. For labor productivity, you’d use the share of hours worked to weight manufacturing, and for MFP, you’d use the share of value-added. So as a rule of thumb, the reported change in manufacturing only counts for about 1/10 of that in the aggregate. Labor productivity rises by 3% in manufacturing? Then this adds about 0.3% to aggregate labor productivity. Manufacturing MFP falls by 1%? Then aggregate MFP falls by about 0.1%. There is nothing magic or special about manufacturing that priviledges its hours or value-added above other sectors in the aggregate calculations. It’s a relatively small part of the pie.