Monday, August 24, 2015

The sectoral shift in labor

I have been looking at wages by sector, and I noticed some interesting patterns.  I don't know, maybe everyone already knows this.  But, I thought it might be worth sharing.  (I should note that there is one short-cut here.  Employment shares are based on all private employees, but wages and hours are based on non-supervisory and production workers, because a longer history of data is available.)

First, there has been a shift, which we can date to about 1984, where sectoral weights in employment have been pretty stable, except for three categories.  Manufacturing has declined from 19% of private employment to about 9%.  About 7% of this shift has been replaced by Health & Education (H&E)employment and about 3% by Leisure & Hospitality (L&H).  Manufacturing employment has been falling much longer than that, but during this period, the shift has been contained within these three sectors while the remaining sectors have remained fairly stable.

I should note that this is private employment, so the Health & Education sector is dominated by health care employment.

The first interesting finding to note is that H&E (purple) has seen steadily higher hourly wage growth compared to the other sectors.  In 1984, average hourly wages in H&E were about 10% below average, and now are about 5% above average.  M (blue) wages have declined from about 7% above average to 5% below average over time.  L&H (red) is a significant outlier.  Average wages there have run at about 60% of the average for the entire period.

If we aggregate hourly wages using the weights of sector employment in 1984 (orange), we see that the shift from M to L&H and H&E has had no material effect on hourly wage levels.  (The black line representing actual hours worked is mostly hidden by the orange line representing the hours worked at fixed 1984 sector weights.)

When we look at weekly hours, we see that the M sector is an outlier, with much higher average hours than other sectors.  H&E is slightly lower than average, and L&H is again way below average.  By sector, average hours has declined slightly in L&H, but the other sectors are pretty flat over time.  Here, if we aggregate hours worked by sector using 1984 weights, we see that aggregate hours worked also would remain fairly level.  In other words, the drop in average hours worked of a little more than 1 hour per week (about a 4% decline) is almost entirely related to a shift in sectors.

So, when we look at weekly earnings, we see that L&H earnings are less than half of the other sectors.  The combination of slightly higher average wages and slightly lower hours puts H&E earnings right at the average.  And, the high average hours worked leads to higher weekly earnings for M.

So, there has been a 4% decline in average earnings over 30 years due to sectoral shifts, due to lower hours worked, not due to lower hourly wages.  But, within these averages, there has been an increase in L&H employment, from 8% to nearly 11%, which is characterized by much lower average earnings.  It seems as though this would lead to a sort of hump at the lower end of the wage scale, relative to 1984.

Outside of the L&H sector, if we compare all sectors except L&H (light blue line) to earnings at the 1984 sector weights (orange line) what we generally see is a substitution of fewer hours for higher pay, with the net effect being a slight increase in average weekly earnings compared to sector weights of 1984 for workers outside the L&H sector.

I don't know if there is any actionable information here, but maybe there is something here to add to a perspective of changes over time in the labor market.

One impression I want to be careful not to give is that there is some sort of fatalism in these trends.  There are surely a mixture of supply and demand factors in the changing context of job characteristics.  These shifts may reflect changes in worker preferences as much as changes in employer demands.  For instance, the past 30 years has seen the remaining shift toward a gender-neutral labor force.  The labor force was more than 70% male after WW II.  By 1984 it was down to 56%, and is now around 53%.  Some of the shift in American production may be a reaction to a more feminine labor force.  Possibly, some of the movement in manufacturing employment out of the country is a reaction to American preferences for shorter work weeks.

I think we have a tendency to falsely treat labor markets as if they are imposed on workers.  Politically, the left naturally does this as an outgrowth of the progressive tendency to view economics through the lens of perceived power imbalances.  But, the right has this tendency too, as we saw in the expectation that Obamacare would lead to a massive shift from full time to part time work, as if employers could single-handedly determine the terms of employment.  If avoidance of the mandate was valuable, it seems as though some Coasean bargain could have been struck.  If it was, it appears that the bargain did not include a widespread reduction in hours.  There have been anecdotal reports of firms switching positions from full time to part time.  But, the lack of any significant aggregate shifts suggests that, on net, the workers filling these new part time positions were being pulled from other employers, not shifted from full time positions.  Possibly, if there was added demand for part time workers from some employers, there was a resulting shift among other employers to move more to full time workers.  In the big, messy world of voluntary associations, it is difficult to imagine all of the possible ramifications.  What does seem clear is that there has been no clearly discernible deviation in employment trends from what historical experience might have led us to expect.  Like with many regulatory issues, much of the effect of Obamacare, pro or con, will be in unmeasurable changes in quality of life that will not be easily confirmed with readings of employment or GDP growth.

Ironically, while the right-wing expectation for a surge in part time employment failed to support criticism of Obamacare, that same failure also punctures the progressive assumption that labor lacks the ability to influence the terms of employment.  The demands of labor have an enormous effect on the shape of the labor market.  It doesn't seem that way to us, because the labor market is an emergent phenomenon.  It appears to be presented to us, as individuals, in a sort of take-it-or-leave-it proposition.  But, it is just as true that each of us, individually, enters, say, a Macy's with a sort of take-it-or-leave-it proposition.  Most of us were not asked what color oxfords we would prefer to wear this year or how wide we would like our neckties to be.  At the point of entering their store, Macy's does offer us a take-it-or-leave-it proposition.  But, is there any question that the demands of their potential shoppers were the primary focus of every decision that involved the stocking of their stores?

Unfortunate predispositions about the nature of employers and firms keeps this understanding from gaining a foothold in our universally accepted presumptions about the economic world.


  1. Kevin, off topic, I thought you'd "enjoy" this article:

    "The S&P/Case Shiller composite index of 20 metropolitan areas in June gained 5.0 percent year over year... suggesting resilience in the housing sector..."

    Translation: "thanks to 3% shelter inflation, 1% everything-else-inflation, supply constraints in housing, and zero interest rates as far as the eye can see, home prices are rising as fast as they possibly can given frozen home credit markets..."

    "While a series of Fed rate increases and a "full blown bear market" for stocks would hurt the housing market, one rate hike and a stock market correction will unlikely damage the housing sector"

    Translation: "Monetary tightening that strangled the economy would be bad for asset prices. Current monetary policy, including the rate hike that everyone already expects, is already priced into the market. Imagine that."

    How did I do, professor?

    I feel like I learned enough macro from Scott Sumner to demolish almost any lay reporting in the area, and enough housing finance from you to do the same in that area. If only this information could be transferred... (Yes, I realize Jason means something else by that phrase :-)


    1. Sounds about right. That article wasn't so bad, though. Interesting that Dallas is up there.