Saturday, February 11, 2017

Strikes in the 20th Century, Macro Data.

This blog post was tipped off by a throwaway comment by JW Mason about a graph that made the rounds on "lefty Facebook" (for lack of a better term) that showed declining strike activity from the immediate post WWII era until today.  His comment was that the graph did not have much information because the graph only showed work stoppages that affected 1000 or more workers.

This version  I stole from Doug Henwood:

Since I need to pay attention to this stuff anyway,  I figured double checking Josh's skepticism would be a useful exercise, though I apologize to the reader that my interest is really mostly just in the immediate post-WWII period.

For the most part, I think that Henwood's graph captures the overall trend fine.  We actually have richer data than what is available at the BLS thanks to the Historical Statistics of the United States, Millennial  Edition (HS).

Here is my graph:
Figure 1

HS has two series, one that is just the "1000+" data set Henwood used.  The other series comes form a more comprehensive attempt to estimate strike activity by the BLS that was discontinued in 1981.  The more comprehensive data set goes back to 1881.  I started my series at 1916 because there are holes in the series before 1916. My series is the number of workers "made idle" by strikes or lockouts, not the number of strikes/lockouts.  To account for a growing workforce I also scaled my data by total workers employed, another series I got from HS.  This is different than Henwood's graph which just  captures the number of strikes/lockouts each year.  The information conveyed is essentially the same.

There are two things I wanted to get out of the graph.  First, to speak to Josh's point: as you can see, the 1000+ measure shows the trend in strike activity adequately, though it obviously underestimates it (by about 600k workers idle on average for the period we have overlapping data).  

The other thing is more for my own benefit.  You can see the new measure of strike activity captures the basic pattern shown in Henwood's graph.  However,  if you extend the series back before 1950 you can capture the moments where labor dramatically assets itself in the US.  The two big strike activity peaks are in the years immediately after WWI (1919) and WWII (1946).  In both cases, once the mixture of wartime cooperation and wartime control evaporated simmering issues over earnings exploded.  I think it is different specific issues about earnings that is driving the two strike explosions, but I don't feel like I can adequately address those subtleties here.

What is really interesting about that graph is that you can see the effect of the New Deal labor reforms.  Post WWI strike activity is dramatic for a couple of years and then disappears through the 1920s.  While strike activity does not reach the 1946 peak, strike activity remains a a sustained, elevated level for at least a decade or two after the war.  Of course, the graph also shows the gradual decline of the New Deal labor period.

The HS also have data on strikes in different industries.  This graph shows strike activity in manufacturing and non-manufacturing:

Figure 2
I have added a measure of how important manufacturing is to the economy because I think it is important to interpreting Figure 2.  You can see that in the 1960s workers idle due to non-manufacturing strikes/lockouts starts to dominates manufacturing strikes/lockouts.  This seems largely driven by the shrinking importance of manufacturing  to the economy.  The other thing that is interesting about this graph is that up until the early 1950s the two series move together.  Then, say, after the Korean War the series are no longer tightly tethered.

Finally, we have a little more detail about strikes in "non-manufacturing" industries:

Figure 3
I want to apologize for this graph.  This is an instance of my not being able to edit and the result is an ugly graph.  First and foremost, you should notice right away that wholesale and retail workers do not strike very much.  Construction workers seem to have a higher 'ambient" level of strike activity, but I only really see one big spike for that industry (1952).  Most of the variation in non-manufacturing comes from mining and the unfortunately too broad category that covers transportation, communication, electric and gas industries).  it is striking how mining strike activity drops off dramatically after the Korean War, with a diminished revival in the 70s.  

Anyway, I don't have many new conclusions to draw about the decline of labor.  I would say what I find interesting here offers some interesting contours to the story we are all familiar with.  First,  I see inflation as a primary "macro" driving force of strike activity.  You see it in the decade after WWII and you see it more weakly in response to the inflation of the late 60s/70s.  The other thing I think you see here is the role played by the way secular changed in production on union strike activity.  For instance, the too broad category of "transportation, communication, electric and gas" covers up the fact that in 1946 the spike in that category was driven by a railroad strike, while the 70-71 spike is best characterized by a postal strike in 1970 and a strike by workers of the Bell System in 1971. 

No comments:

Post a Comment