Tuesday, July 26, 2011

Automation or Offshoring? Evidence from General Motors

Yesterday, I read an editorial from last year in BusinessWeek, written by the President of the Robotics Industries Association -- Jeff Burnstein -- which went as follows:
If you work in an American manufacturing company today, you should be worried about your job. I live in Michigan and have witnessed the destruction caused by shuttered factories and jobs shipped overseas. When plants close, whole communities suffer.

With unemployment at about 14 percent or higher in Michigan, it's not surprising some workers are afraid of robots capable of working seven days a week, 24 hours a day with great accuracy and reliability, capable of performing many tasks better than people.

That fear, so prevalent in the early days of robotics, today is misplaced. What should really give workers pause is when their companies won't use robots and other automated technologies to become stronger global competitors.

U.S. technology and business innovators recognize that robots in factories have the potential to save and create more jobs than they eliminate. Robots help companies turn out higher-quality and lower-cost goods to compete with those made in China, Mexico, India, or other low-wage nations. They remove people from dangerous and boring jobs they shouldn't have been doing in the first place, and put them in higher-skilled, higher-paying positions.
Obviously, Mr Burnstein is doing his job - advocating for robotics and trying to minimize the downsides.   But is he telling the truth?  How can we determine how much of the manufacturing job loss of recent decades is due to offshoring versus automation?

It occurred to me that one way to get a rough idea is to look at the auto industry.  In that industry, to a first approximation, domestic cars are still domestically produced and so offshoring can be measured by loss of market share of the domestic auto producers.  Meanwhile, we can get statistics of employment and car production and use those to estimate how much job loss is due to automation of the domestic industry.

I picked on General Motors as the largest and most visible domestic car company.  I'm using 1955 and 2009 as my comparison years because I happened to be able to find convenient statistics for those years.

I found GM's overall market share over time here.

I estimate GM's market share in 1955 at 47% - which was probably slightly before their peak in market share.  By 2009 it had fallen to 19.5%, only 41% as large as it was in 1955.  So we can attribute a reduction in jobs by a factor 0.4 to to be due to loss of market share (relative to the 2009 situation had there been no loss of market share).  Presumably, that can mainly be viewed as due to competition from imports (ie the equivalent of offshoring).

I found an extended excerpt from GM's 1955 annual report here.  In that year, GM made 4,477,000 cars and trucks in US factories.  To do so, it employed about 555,000 people domestically.  Thus GM was producing 8 cars/employee in 1955.  In 2009, according to its annual report, GM sold 2,084,000 cars in the US.  To accomplish that required 77,000 US employees.  Thus GM was domestically producing 27 cars per employee.  (I checked and it doesn't make much difference to this ratio if you include Mexico and Canada).  The ratio 8/27 is 30%.  Thus we can attribute a reduction of jobs by a factor 0.3 to be due to increased productivity which we can presumably mainly put down to automation.

Of course, these two factors are multiplicative.  If GM had had no loss of market share but somehow was managing to do this at 1955 levels of productivity, it would employ 1/(0.3*0.4) = 8.3 times as many people.  Of the two factors, both are roughly equal in importance, but productivity improvements (automation) appear to be slightly more important than loss of market share (offshoring) -- though I wouldn't place too much stress on that difference given the uncertainties of this rough calculation.

Thus when Mr Burnstein claims "U.S. technology and business innovators recognize that robots in factories have the potential to save and create more jobs than they eliminate", I don't think he's telling us the truth.


Unknown said...

"Of the two factors, both are roughly equal in importance, but productivity improvements (automation) appear to be slightly more important than loss of market share (offshoring) -- though I wouldn't place too much stress on that difference given the uncertainties of this rough calculation."

Loss of market share does not imply offshoring. Most Japanese cars sold in the US nowadays are made *in* the US. Same is increasingly true of Korean makes. The only way to do this calculation would be to look at the total number of US auto workers over two or more years, and compare it to the number of US-made cars over the same years. Imports are irrelevant because autos made in another nation say absolutely nothing about productivity and automation within US borders.

Stuart Staniford said...

Unknown - it's true that quite a lot of Japanese cars are assembled in the US, but my impression is that a lot of the parts and the majority of the value added are still coming from Japan.

Seth said...

I think the 'higher productivity is always good' argument depends on a transfer of labor into other sectors that remain more labor intensive (if only in relative terms).

It isn't that auto-making robots would generate more demand for auto-making workers. Rather, the cheaper cars would save consumers a lot of money which they would instead spend on other products, increasing demand in sectors which would hire the displaced auto-workers.

It obviously works some of the time. But the market ideology requires us all to believe that it ALWAYS works. (And bite your tongue rather than speak of other possible outcomes ...)

Greg said...

Mr. Burnstein is implicitly relying on what he learned from economists - that the more-highly-paid controllers of the robots will demand new products and services, and that new and existing industries supplying the new products and services will take on all of the 'released' workers, and more. So "more jobs" are created indirectly, elsewhere.

However, factory-floor wages in the auto industry (for example) are trending down despite the demand for greater skill, so supply cannot create its own demand.

Also, it's unlikely that an auto assembler can straight away get work as a nurse, yoga instructor or personal coach, so any adjustment will require several other people to change jobs as well. The new jobs could take decades to appear.

A weaker argument, that using robots stops firms going out of business altogether and so saves more jobs than not using robots, is valid but not very inspiring.

Lars-Eric Bjerke said...


You have made an interesting try to look at the effects of automation by looking at the number of employees in GM and the number of cars produced at two different times 1955 and 2009. However the comparison is difficult. When looking at automation we often think about the assembly line, which is a small part of the cost of the car. Today it takes about 10 to 20 man-hours to assembly a car. The manufacturing the last 60 years has changed, today a much bigger portion of the parts are purchased from sub suppliers and a lot of the services are outsourced. The cars also contain a lot more equipment and technology. Car manufacturing today I guess is mainly research and development, design, purchasing, logistics and marketing and not so much manufacturing.

porsena said...

Stuart, I made a long reply with links about this but Blogger wiped it out. Basic points:
Auto manufacturing has changed enormously since 1955. Aside from body stampings, some engines and maybe some parts of the drive train, the GM of 2009 was basically a systems integrator. Design, engineering and production of the vast majority of parts was contracted out.

The auto industry has been driven to a least-cost model that encourages offshoring.

GM's management made a decision in the early 200's to focus US production on trucks and SUVs, because that's where the money was. By the time the 2008 crunch rolled around, nothing smaller than a Malibu was assembled by GM in an American GM plant. The Pontiac Vibe, basically a rebadged Toyota Matrix, was made alongside the Matrix at the NUMMI joint venture plant in California. GM market share but not GM jobs. Likewise, the Aveo was imported from Korea.

I don't think your offhand dismissal of foreign manufacturers is well founded. For the third year in a row this year, the US-made vehicle with the largest content of American jobs per unit is the Toyota Camry. Honda uses over 600 North American parts suppliers for its vehicles assembled in the US and Canada.

As Lars-Eric said just above, the auto manufacturing industry has changed enormously since 1955. Then, GM made in-house most of the parts in its cars and trucks. In 2009, the industry had moved to a least-cost model, with the design, engineering and production of most of its vehicles outsourced to who-knows-where. Some of these contractors are American, others aren't.

porsena said...

And I meant to add that, as a consequence of GM's 2010 withdrawal from the NUMMI joint venture, Toyota will be opening a new plant in Mississippi this fall to make the Corolla, which used to be made alongside the Matrix/Vibe. This new plant will have 2,000 employees and will make 150,000 Corollas a year for the North American market. That's 75 units per employee per year. At the other end of the scale, Fiat operates a much reviled plant in Italy where the corresponding figure last year was 6. That difference is not all about automation, it's mostly about trade union contracts.