Another Explanation for the Productivity/Employment Split

June 12th, 2011 at 12:28 pm

Recall the discussion and pictures from this post.  I’ve updated the graph, btw, swapping out total employment for private (leaving out government) since that’s comparable to the coverage in the productivity series.

Source: BLS

There’s another reason why productivity may be diverging from employment: mismeasurement of productivity growth of a particular type that would show up as both higher productivity and less employment growth.

Economist Susan Houseman has been writing about this for years, and while I don’t think it explains the whole gap, it likely explains part of it.

The problem has to do with outsourcing.  When US firms buy cheaper goods or employ cheaper labor from abroad, that can show up as faster productivity growth, even as it displaces domestic workers.  Basically, to measure productivity accurately, you have to net out the costs of goods that go into the process (“intermediate goods”) and the labor it took to produce the good or service.

Globalization, at least the way we currently account for in our national spreadsheets, leads to a) an underestimate of how much imports are contributing to value added, and b) doesn’t deal well with the substitution of foreign for direct labor.

Houseman and Mandel give this example:

“Suppose a US automaker imports one million parts from a Japan-based supplier at $10 per part, for a total import bill of $10 million. Consider two scenarios:

Scenario 1: The US automaker improves its production process in its domestic factories, so it only needs half as many components. The import bill goes down to $5 million.

Scenario 2: The US automaker switches to a China-based supplier that only charges $5 per part. The import bill goes down to $5 million.

Surprisingly, these two scenarios are indistinguishable in the US economic statistics. In both scenarios, the import bill goes down to $5 million. The value-added of the US auto company goes up (sales minus the cost of materials), as does its profitability (sales minus cost of labor and materials) and measured productivity (value-added per worker).

Note that even though value-added per domestic factory worker goes up in the second scenario, the impact on domestic real wages and employment is ambiguous. For example, if value-added per worker is rising because of an improved ability to identify new sourcing opportunities, that same capability could be easily used to replace domestic workers… In other words, a measured productivity gain from increased efficiency in the supply chain doesn’t necessarily improve the real wages or employment of US workers in auto factories.”

Two additional wrinkles here.  First, this mis-measurement wouldn’t explain the acceleration of productivity in the above figure unless offshoring to lower-cost platforms was a growing part of our manufacturers’ strategy.  But of course, there is increasingly more of precisely this type of offshoring.  Second, this wouldn’t affect the productivity trend in the chart if the offshoring were domestic—one domestic industry’s gain would be another’s loss.  So this is really a globalization story.

Houseman et al guess that if we were measuring this correctly, we might shave as much as half a point off of annual productivity growth, 1997-2007.  If so, this would close the gap you see towards the end of the series above by about a third.  So again, not the whole story, but not trivial either.

And this ain’t just the stuff of academic measurement.  There are many clear benefits to globalization, but from the perspective of those whose wages and jobs are lost, this is a cost—one to which we are not paying enough attention.


Print Friendly, PDF & Email

21 comments in reply to "Another Explanation for the Productivity/Employment Split"

  1. Sandwichman says:

    It seems to me that this explanation, while plausible, still leaves a gap in the traditional “productivity creates more jobs than it destroys” story — which has its free trade version going back to the early 18th century. The assumption is that however the proverbial US automaker saves the $5 million dollars, it will plow the money back into employing more people in some other capacity.

  2. RandyH says:

    I used to work for a company as an IT engineer. When they outsourced my job to an Indian company, they replaced me with two much lower paid engineers over there. (Their combined incomes were about 1/3 of my salary.) But now they have removed me from the domestic headcount and they just count the overseas workers (who are technically employed by another company) as a service expense. This makes the remaining domestic employes at the company look considerably more productive, no?

    Maybe I have it all wrong, but I remember the executives boasting about the huge productivity gains that they could claim (at bonus time) after they eliminated half of the domestic employees from the payroll because more work would be getting done by fewer domestic employees and for significantly less money.

  3. Elwood Anderson says:

    Productivity can go up simply because employers lay off employees and require those retained to do their work. In the last decade, particularly since financial crisis, that’s what’s happened.

    • AMM says:

      This was the same point that I was going to make. When firms are forced to downsize of course they are going to keep their most productive employees. The other side of the business will also contribute, when the economy is booming and the labor market is tight, firms are going to scrape the bottom of the barrel just to get warm bodies.

  4. Paul J says:

    It is interesting to compare the above chart to Chart 1 in the BLS publication “The compensation-productivity gap: a visual essay” (, which looks at compensation rather than employment compared to productivity. The chart above shows that productivity and employment started to diverge in around 1957, but size of the gap was more or less constant for a long time, with the gap narrowing somewhat during the Clinton years but thereafter greatly widening. The chart in the above-cited publican shows compensation tracking productivity fairly closely until around 1973, when it started to diverge. Chart 2 of that publication shows that the gap in the 1990’s, i.e., mostly Clinton Administration, was narrower than in the periods before (Reagan, Bush 41) and after (Bush 43).

    What is going on? Did the Clinton Administration deliberately do something to narrow the gap, or did the Republican Administrations deliberately do things to widen the gap, or is it just a coincidence?

    • Kevin Rica says:


      Isn’t that your “Jaws of the Snake?”

      When did you first do that?

      • Jared Bernstein says:

        Similar but not the same. You’re thinking of productivity growth plotted against real median wages or real median family income. Those gaps opened earlier than this one.

  5. Michael says:

    In a standard game theoretic negotiation model, the surplus from a given transaction is split in proportion to the discount rates of the negotiators. If the buyer can afford to wait longer than the seller, the case for most employers and employees, then buyer gets more of the surplus.

    Unions decrease the discount rates of the sellers (by smoothing consumption of members) and increase the discount rates of the buyers (by threat of strike). Thus, wouldn’t unionization levels be the very first place one would look if one saw a divergence between productivity growth and wage growth? Seems like the buyers got more power and the sellers ended up with less, so the new pie got split increasingly unequally.

    Why bother with all these 2nd and 3rd order bank shots? We have decreased unionization and a split between wage growth and productivity growth. It even tracks how pro-union the President’s party is. Case closed, unless some truly compelling evidence presents itself.

  6. Virgil Bierschwale says:

    you’re getting close to smoking tobacco instead of that wacky weed ya’ll smoke in the White House..

    Send me an email address to and I’ll send you a chart that shows that our businesses sent a lot of work offshore in 2002 and a much larger amount in 2007 which caused the subprime mortgage crisis in 2008.

    After all, if you don’t have a job, you can’t pay your mortgage which creates a crisis, doesn’t it…

    And washington continues to snooze…

  7. BarbD says:

    Not sure what impact this has (perhaps too small to measure), but I figure I and other retirees may play some role in the improved productivity of my former employer (who paid me generously to retire early 3 years ago).

    On a contract basis, I now handle a piece of my former job judged important but not strategic, and I’m about to take on another chunk of work for the same reason. This both frees up employees to focus on other pressing work — and I’m guessing shows as a productivity improvement for the company overall.

    I’m also noting an increase in projects listed with YourEncore (, which specifically targets retired scientists, marketers and others for project work. Again, probably too small a blip to register but I think a positive way to both help people who aren’t ready to retire working and companies be more productive.

  8. giulio says:

    Macroeconomic theory predicts a relationship between the rate of wage growth and that of labour productivity. I do not think, it makes particular predictions about the relationship between the latter and employment growth in isolation.

    On the other hand, if wages grow at a lower rate than that of labour productivity one would expect the employment rate to grow.

    • Michael says:

      You’d expect that, but we don’t have it. So we must have violated one of the assumptions of macro theory. Macro theory generally assumes full employment…

      • giulio says:

        I was not being dogmatic about macro theory and not all of it assumes full employment all the time.

        I was just stating that, while a statement about wage growth, productivity growth and employment growth together was a meaningful one, I am not sure the point of this post is well posed.

  9. Keith Roberts says:

    Re BLS employment data, as a non-economist I am curious about a claim that is stridently made in one of the investment newsletters: namely, that BLS employment figures include an “adjustment” for construction employment that adds or subtracts a significant number of jobs based on a model that relates construction employment to other employment numbers. The reason for the model instead of sample data is that the BLS cannot collect reliable construction employment information. The problem, according to the newsletter, is that the model is based on pre-recession data, and therefore falsely adds a huge number of nonexistent construction labor whenever other employment figures go up. What do you think?

  10. Kenny Liao says:

    I was wondering if there is any way to get the source data for the “productivity and employment: what happened” graph? I would like to re-create it for a presentation (sourcing this article, of course).

  11. Thomas Aubrey says:

    Hi Jared, I am unable to replicate your chart. Please can you clarify which series you are using. The rate of productivity growth (output her hour) according to Alex Field fell not rose between 2003 and 2007 whilst private sector employment rose to all time highs.

  12. Thomas Aubrey says:

    Thanks Jared. I can replicate now. One further question, in your NYT piece on where have all the jobs gone you stated
    “a growing number of economists believe that our very high levels of inequality are not just whacking the incomes of the “have-nots” but are slowing job growth as well. Part of this works through the demand channel: with so much spending power in the hands of so few, consumer demand is becoming bifurcated. Walmart will do well on one end, Neiman Marcus on the other, with too little in between. Another part works through misallocation: too much economic activity devoted to “innovative” finance and too little to sectors more germane to middle-class jobs and incomes.”

    Are there specific references to work done in both of these areas?

    • Jared Bernstein says:

      These points are made all over the place these days…see Stiglitz’s most recent book, Price of Inequality, eg.

  13. Thomas Aubrey says:

    I didn’t find much empirical analysis in Stiglitz on AD but this ILO stuff I think backs up your point.—ed_protect/—protrav/—travail/documents/publication/wcms_192121.pdf