09 July 2012

Overcomplicating Debates Over Jobs

Writing in the NY Times, Thomas Edsall has an essay exploring whether technological innovation is leading us to a permanent era of declining employment possibilities:
The issue of the disappearing middle is not new, but credible economists have added a more threatening twist to the argument: the possibility that a well-functioning, efficient modern market economy, driven by exponential growth in the rate of technological innovation, can simultaneously produce economic growth and eliminate millions of middle-class jobs.
Support for this argument in Edsall's piece comes from Andrew McAfee at MIT who provides this graph.
McAfee explains the graph in ominous terms, focusing in on the red curve in the graph, which shows a ratio of employment to population (you can click on the graph to embiggen it):
Since the Great Recession officially ended in June of 2009 G.D.P., equipment investment, and total corporate profits have rebounded, and are now at their all-time highs. The employment ratio, meanwhile, has only shrunk and is now at its lowest level since the early 1980s when women had not yet entered the workforce in significant numbers. So current labor force woes are not because the economy isn’t growing, and they’re not because companies aren’t making money or spending money on equipment. They’re because these trends have become increasingly decoupled from hiring — from needing more human workers. As computers race ahead, acquiring more and more skills in pattern matching, communication, perception, and so on, I expect that this decoupling will continue, and maybe even accelerate.
McAfee tells Edsall:
“In my dystopian vision of the future, that red line (in the chart) keeps falling down – or suddenly drops off a cliff”
Sounds scary. But what is it that the red curve is actually showing?

The graph below shows the annual change in unemployment and the annual change in the employment/population ratio, using the same data presented by McAfee and over the same time period.
The graph shows that the complicated metric of employment/population is virtually identical to the more conventional metric of unemployment. Thus, the debate we should be having is not about a "red curve falling off a cliff" but rather, where do jobs come from?

Now, my revealing that the red curve in the McAfee graph is just a fancy way to convey unemployment does not tell us whether or not we are entering a dystopian era of machines run amok. But being clear in what we are talking about is a helpful first step.


  1. As we approach a productivity singularity, including one engendered by technological innovation, it is likely that we, as a society, will, once again, recharacterize individual valuation. Perhaps this is the stage we are in now, but I remain skeptical to the extent of its effect.

  2. In the discussion of the NY Times article, Hamilton questioned the finding that the machines would soon make people obsolete and pointed out that the real culprit was the loss of jobs in housing and construction. Before 2006 this was a pretty major part of the economy and it dropped back to only a quarter of where it had been. As they say, that's gotta hurt. The other part of the equation no one talks about are all the expenses for employees that do not show up in their pay checks but employers pay. The health care benefit alone for a family currently exceeds annual earnings for someone making minimum wage. The you add in unemployment insurance, workers comp, along with other costs and pretty soon a hourly wage earner's paycheck has less than half the cost to a business that employs them. I think of these as "headcount" costs and with these being very high, you try to pay a decent wage to a few productive people and rely on overtime when demand gets heavy. If you want more jobs for people at the low end of the wage scale, reduce the costs to employ those folks. And when I say reduce costs, I'm not talking about reducing wages, I talking about all the fees and taxes the state imposes on employers for the privledge of having employees.

  3. One thing many of these discussions are missing is the concept that demand does have limits. How much food can you eat? How many cars can you drive? Even if you give it away, there will only be so many takers. Ask yourself what happens when you can make everything we reasonably need and want with just 10% of the population? How does the other 90% participate in the economy if the economy doesn't need them? Our economic system is built on the proposition that the economy will grow to absorb the excess workers, but it cannot grow once it has produced all that needs to be produced.

    Studies have shown that well-being seems to plateau at $50k-75k in income, the idea being that we can buy everything we reasonably desire. That is also the wage level that savings rates start to skyrocket. Median household income in the US was $51,413 in 2011. Have we reached the limits of demand in the US?

  4. @Sean (#2)

    Health care costs are a factor - but the other costs you note are actually secondary. I say this because these costs are a function of wages, and wages are pumped up by the FIRE economy (Finance, Insurance, Real Estate, and Education).

    FIRE means the cost of living is high, which means wages must be high, which in turn means Social Security and other costs - both for employers and employees - are high in absolute terms.

    Fix the health care problem, defined as bringing US costs at least somewhat in line relative to GDP with the entire rest of the 1st and 2nd world.

    Fix the FIRE problem: return mortgages to the realm of 20%, re-regulate banks, perhaps even a mortgage writedown.

  5. Question: what exactly is the denominator of blue line employment ratio? Is it total population, or workforce, or something else?

    And the vertical axis says "change from one year ago". Does that mean that it is the slope/derivative of McAfee's curve?

  6. Hi,

    According to my calculations on my blog, the the number of human brain equivalents added by computers will be:

    1 million in 2015 (should not be noticeable)

    1 billion by 2025 (should definitely be noticeable)

    1 trillion by 2033 (should consistently deliver higher growth per year than ever observed before)

    1 quadrillion by 2040 (Singularity territory).


    So even if the effects of computers on employment aren't discernible in 2012, they definitely should be in the next 10-20 years.

  7. -5-Dean

    Here is the source,

    BLS says aged 16 and over

    The change is year on year change. Thanks

  8. Hi Roger,

    If you haven't already, I think you should let Andrew McAfee know about your blog post, and invite him to comment.


  9. -8-Mark Bahner

    Good idea! Just done ...

  10. Hi Roger,

    You wrote, "Now, my revealing that the red curve in the McAfee graph is just a fancy way to convey unemployment..."

    I don't agree with that characterization at all. My understanding of the red curve in the McAfee graph is that it conveys the percentage of adults who are employed. Further, this percentage oscillated between approximately 62 to 64 percent from 1995 to 2008. But it dropped below 60 percent in approximately November 2008 and is now at about 58 percent.

    Because it is the percentage of adults who are employed, I think his curve can and will be influenced by the retirement of the Baby Boomer generation (born 1946 to 1964). In the next 10-15 years, more than half of that (my) g-g-generation will be retired. And that red curve will drop and continue to drop as the Baby Boomer generation retires. This will happen regardless of what happens to the unemployment rate, or how computers affect anything.

  11. Oops. I should have checked the definition first:

    "Definition of 'Employment-To-Population Ratio'
    A macroeconomic statistic that takes the ratio of the total working age of the labor force currently employed to the total working age population of a region, municipality or country. It is calculated by:

    Labor Force Employed/Total Population

    The working force and population only include individuals within the working age.

    Read more:


    But even though I was wrong, I don't agree that your characterization is accurate. The employment-to-population ratio includes such people as stay-at-home "non-working" parents, and people who have given up looking for work. So it's not just "a fancy way to convey unemployment."

  12. Roger - I'm no 'on' Twitter, but I just followed a link from your home page left margin to a Jeffrey Sachs column in the FT. Was Sachs' 'increased drought, floods and heat waves' posted with your approval?

  13. -12-Mark B.

    Sachs hasn't yet contacted me to approve his op-eds, though perhaps he should ;-)