Running a Nuclear Plant While Misunderstanding the Instruments Can be Hazardous

…the same is true of establishing policy for a national economy and while misunderstanding the relevant economic indicators.

It has often been asserted, by economists and others, that the decline in US manufacturing employment is largely a result of great strides in automation-based productivity, and that offshoring and imports have had relatively little effect…some have gone so far as to say that the offshoring/import effect on jobs has been practically irrelevant compared with that of automation.

I was quite willing to believe that there have been great strides in manufacturing productivity, but the idea that offshoring & import effect on jobs was unimportant never sat very well with me…it seemed clear that the tens of millions of workers producing for export, in China and elsewhere, must have had a very material impact of American jobs, even given the greatly superior productivity of American factories to those located in most other countries.

But now it seems that even the assumption of a broad-based productivity improvement in American manufacturing must be questioned.  Susan Houseman, an economist at the Upjohn Institute, has done some interesting work in unpacking the productivity numbers.  Her analysis indicates that a very high proportion of the measured growth in US manufacturing productivity actually reflects productivity growth in a single sector:  computers and electronic products.  Excluding this sector reduces to overall productivity growth for US manufacturing reduces annual productivity growth from about 3% to a little less than 1%.  Moreover:  Houseman argues that the productivity growth in that computers & electronics segment is less a result of automation-driven manufacturing productivity than it is a result of (a) better product design, and (b) the way the price deflators are calculated to turn nominal into real numbers.  And in all segments, the handling of imported intermediate goods (parts, subassemblies, and materials) changes the productivity estimates in ways that may be questionable:

An article summarizing Houseman’s work, and an interview: Don’t blame the robots.

Also, direct links to some of her work:

2011:  Offshoring Bias in US Manufacturing

2014:  Measuring Manufacturing–How the computer and semiconductor industries affect the numbers and perceptions

2016:  Is American manufacturing in decline?

I learned about this work via Marginal Revolution…a few relevant comments at the link.


2 thoughts on “Running a Nuclear Plant While Misunderstanding the Instruments Can be Hazardous”

  1. From the Quartz article (first link):

    “One reason why Houseman’s revelation is so important is that the myth of automation continues to have a strong grip on the minds of American policymakers and pundits. The lessons of the populist backlash during the 2016 presidential election didn’t seem to take. As the US gears up for mid-term elections this year, the Democrats have no vision for how to reverse the industrial backslide.

    Ironically, that criticism applies to Trump, too. His campaign ignited a vitally important national conversation on the relationship between US trade policies and manufacturing’s decline. Since he took office, however, Trump has paid minimal attention to boosting US manufacturing. Instead, he’s favored counterproductive protectionism and ignored currency manipulation, preferring the punitive over the constructive.”

    I don’t think that assertion about Trump is true at all. Some of the tax-policy changes, specifically the expensing of capital investment, will be particularly beneficial to ‘stuff’ businesses.

  2. My observation of robots is that they are Malthusian in the following way. As degrees of freedom increase, their capabilities increase arithmetically but the things that can go wrong increase exponentially. There’s a narrow window of optimization that limits their use.

    The other day someone sent me a video of Boston Dynamics dog robot walking up and down stairs with warnings that the robots will soon take over, the machines are rising, etc. The question I had was, this is the video they released where the robot accomplished its task, but how many times did it fail? Did it succeed 9 out of 10 times, 19 out of 20 times, 99 out of 100 times? To be suitable for six sigma mass production it has to do it 999,996 out of a million.

Comments are closed.