Why the Middle Class Is Dying

Summary:  The slow growth of employment (far slower than GDP or corporate profits) continues to puzzle economists.  Perhaps it’s just a short-term oddity.  Perhaps something has changed in the overall economy.  Slowly economists see, albeit dimly, the coming robot revolution (no, it’s not funny).  This is another in a series about the possibility of radical and severe evolution of employment in the 21st century (aka the robot revolution); links to other chapters appear at the end.

From the brilliant (albeit dogmatic leftist) Brad DeLong (Prof Economics, Berkeley):

What are we all going to do in the future as machines replace more and more of our jobs?

This has worried economists ever since the eighteenth-century French physiocrats tried to figure out how an economy could avoid mass unemployment if the agricultural share of the labor force ever fell below two-thirds. What would all those extra people do? Everybody knew that non-agricultural workers were not creators of net wealth but simply ways of transforming wealth–food and raw materials go into a non-agricultural sector and useful stuff comes out but, the physiocrats said, the non-agricultural workers transform value, they don’t create value. Only the agriculturalists create value. And should the agricultural share of the labor force drop below fifty percent, there won’t be enough useful ways to transform the agriculture-created value to employ everybody.

The physiocrats were, of course, wrong. We found lots of useful things that people could do not just to transform but to create value as the agricultural share of the labor force headed down to its current 2% or so share. But what happens next as hardware robots take over manufacturing, mining, and transportation and as software ‘bots take over the routine paper shuffling?

From the always insightful Paul Krugman – when he’s an economist, not channeling Spiro Agnew (i.e., a partisan attack dog) — in the New York Times, 6 March 2011:

{T}he influential analysis of Autor, Levy, and Murnane … argued that the crucial difference in terms of possible replacement of humans by machines was one of routine versus non-routine, rather than white-collar versus blue-collar, and that computerization was if anything likely to increase demand for some “low-skill” occupations and reduce demand for some traditionally well-paying white-collar jobs …

Autor pursued this research in a new paper:  “Skills, Tasks and Technologies: Implications for Employment and Earnings“, Daron Acemoglu and David Autor (both Profs of Economics, MIT), National Bureau of Economics Research, June 2010.  Well worth reading, with some interesting insights about the labor dynamics of the 21st century.  Here we look at two of the most important graphs.

(1)  How has American prosperity been shared since 1973? 

Let’s look at wages of men, as the job composition of women workers has changed too greatly for useful comparisons over time.  There are many factors at work in this graph.  Demographics, women entering the paid workforce, technology, immigration.   The result is grim.

  • HSD — high school dropout,
  • HSG — high school graduate,
  • SMC — some college
  • CLG – college level graduate,
  • GTC – greater than college


For an even clearer demonstration, see the Economic Policy Institute’s graphs about The State of Working America.  Especially this interactive graphic:  When income grows who gains?  which shows how iocome gains were shared between the 10% and the rest of us, looking at any start and end dates.  Between 1973 and 2007 the  average income in the US grew $15,238 — the bottom 90% got 2% of that.  Including 2008, a recession year (in which the richest both 10% were hurt the least) gives an even darker picture.

(2)  Looking at current trends, who will benefit in the future?  Who will relatively suffer?

Krugman notes something even more disturbing in this paper.  Figure 10 below shows the death throes of the American middle class.  With employment gains in the low-skills segment (with falling real wages) and high-skills segment (with rising wages).  Each year with less and less in between.  Unfavorable circumstances for our Republic; good news for those seeking to re-impose rule by plutocracy (one that is, like most such, largely an inherited rich ruling class).  Krugman:

In the 1980s, the higher the skill required for an occupation, the bigger the employment gains. In the 1990s, there was “hollowing out”, with the middle-skill occupations losing relative to both ends. And most recently, the hollowing seems to have spread further up the scale. This is real, and it calls some of our favorite platitudes into question.


The author’s explanation:

Figure 10 provides a starting point for the discussion of job polarization by plotting the change over each of the last three decades in the share of U.S. employment accounted for by 318 detailed occupations encompassing all of U.S. employment. These occupations are ranked on the x -axis by their skill level from lowest to highest, where an occupation’s skill rank is approximated by the average wage of workers in the occupation in 1980. The y-axis of the figure corresponds to the change in employment at each occupational percentile as a share of total U.S. employment during the decade. Since the sum of shares must equal one in each decade, the change in these shares across decades must total zero. Thus, the height at each skill percentile measures the growth in each occupation’s employment relative to the whole.

The figure reveals a pronounced ‘twisting’ of the distribution of employment across occupations over three decades, which becomes more pronounced in each period. During the 1980s (1979-1989), employment growth by occupation was nearly monotone in occupational skill; occupations below the median skill level declined as a share of employment and occupations above the median increased. In the subsequent decade, this monotone relationship gave way to a distinct pattern of polarization. Relative employment growth was most rapid at high percentiles, but it was also modestly positive at low percentiles (10th percentile and down) and modestly negative at intermediate percentiles.

In contrast, during the most recent decade for which Census/ACS data are available, 1999-2007, employment growth was heavily concentrated among the lowest three deciles of occupations. In deciles four through nine, the change in employment shares was negative, while in the highest decile, almost no change is evident. Thus, the disproportionate growth of low-education, low-wage occupations became evident in the 1990s and accelerated thereafter.  This pattern of employment polarization is not unique to the United States …

For more information

Other posts about employment

  1. Important:  Globalization and free trade – wonders of a past era, now enemies of America, 16 March 2011
  2. America passes a milestone!, 20 January 2010 — More jobs in government than manufacturing
  3. Yes, it is a “mancession”, with men losing more jobs than women. Just like all recessions., 5 October 2009
  4. Update on the “mancession”, 2 December 2009
  5. A look at the engines of American job creation, 12 January 2010
  6. An ominous trend: number of Americans working for the government vs. those making things, 5 March 2010 — Update to the Oct 2009 post.
  7. The coming big increase in structural unemployment, 7 August 2010
  8. Important:  The coming Robotic Nation, 28 August 2010 — Part 1 of 2
  9. Important:  The coming of the robots, reshaping our society in ways difficult to foresee, 22 September 2010 — Part 2 of 2
  10. Economists grapple with the first stage of the robot revolution, 23 September 2010
  11. Arithmetic of decline: America’s lost decade for jobs, 27 November 2010
  12. A snapshot of the recovery so far, looking at the most important factor: employment, 4 March 2011

Originally published at Fabius Maximus and reproduced here with permission.