Evan Soltas
Mar 3, 2015

Causing the Great Compression

This blog post mostly comes from some longer stuff I'm working on, but I think it's interesting and blog-worthy, so here it goes.

The greatest decline in income inequality in the history of the United States happened between 1930 and 1950. The nation went from Great Gatsby to Ozzie and Harriet. Economists call it the "Great Compression." What caused it?

First, some quick background. Broadly speaking, there are two schools of thought in economics research on inequality.

One school, associated with economists like David Autor, Alan Krueger, and Lawrence Katz, likes to think about inequality as being determined by relative supply and demand for different categories of labor -- such as skilled versus unskilled, college-educated versus not, cognitive white-collar versus repetitive manual.

Another school, associated with economists like David Card, David Lee, and John DiNardo, puts its focus on institutions that affect labor markets -- such as the decline in unionization and in the real minimum wage, the rise of international trade and immigration, social norms about executive compensation, and financial deregulation.

That's a crude and incomplete picture of how economists think about inequality, and the two schools are mostly complementary. But it sets up an intuitive contrast when we turn to look at the Great Compression: To what extent was it caused by supply-and-demand, and to what extent was it caused by institutions?

Economists don't really know, or at least, that's my assessment of the state of knowledge on this topic. Claudia Goldin and Robert Margo, the economists who came up with the term "Great Compression," point to both supply-and-demand and institutions. On the supply-and-demand front, for instance, the GI Bill increased the relative supply of skilled and educated workers in the 1950s and the rise of manufacturing increased demand for unskilled labor. For institutions, they look to the National Industrial Recovery Act and the National War Labor Board.

Here's where I come in.

Using data from the NBER's Macrohistory database, I was able to construct monthly estimates of the gap in earnings between skilled and unskilled workers in construction for the 1930s and 1940s. I also found the gap in hours worked between skilled and unskilled workers in manufacturing for the same period. "Skilled" is defined by a trade -- in construction, for instance, it's a bricklayer, carpenter, or ironworker.

The advantage with my data is that I have more frequent observations, so I can see more finely what is going on in wages and hours than what's been written on the Great Compression before me. So let's look at the wage premium, interpreting the scale as a percentage difference (e.g., 0.8 = 80%) in wages between skilled and unskilled workers:


The wage gap in construction shrunk 10 percentage points in the months immediately following the passage of the National Industrial Recovery Act in 1933, a law which Goldin and Margo suggest should be important to understanding the Great Compression. The change in relative wages resulted from a sharp increase in the unskilled construction wage in 1933, with relatively little change in the skilled wage over this period. Then the gap remained stable from 1934 until 1939. And then, from 1940 to 1950, it declined 25 percentage points.

Let's now take a look at hours, again interpreting the scale as a percentage difference between skilled and unskilled workers:

What we should notice is that at exactly the same time as the wage gap collapses, the hours gap goes from -0.06 to +0.04. That is, in the fall of 1933, skilled workers' hours rose 10 percentage points relative to unskilled workers' hours. Then, once the wage gap starts declining again in 1940, the hours gap starts to rise.

Why do hours matter to this story? Simply put, hours allow us to distinguish between the causes of the compression, between supply-and-demand and institutions.

A change in institutions towards ones that reduce the wage premium should also increase the hours premium. For intuition, if you're basically being forced to pay unskilled workers more than you would have, you might respond by reducing unskilled workers' hours. Another related explanation are labor rules that simultaneously restrict maximum pay and maximum hours, which would tend to produce the same result as before.

Contrastingly, a change in supply-and-demand that favors unskilled labor over skilled labor should reduce the wage premium but decrease the hours premium. For intuition, if the rise of the assembly line makes unskilled workers more productive, you'd want to get as many hours as you can out of them.

So we have the predictions for hours going in opposite directions. Up would suggest institutions, down would suggest supply-and-demand. And, of course, we see that relative hours of skilled workers rose. Moreover, that rise occurs in the shadow of the National Industrial Recovery Act -- which, we've already said, should point us towards institutions.

Second, we can look at the much more gradual compression in the 1940s. Again, skilled workers gain hours over unskilled workers. What's going on in this period? As Goldin and Margo explain, the National War Labor Board (see earlier poster) is regulating wage increases, resulting in a compression.

Both periods of compression in the 1930s and the 1940s, then, seem to fit the pattern of institutions rather than supply-and-demand. If relative supply-and-demand drove the compression, we'd expect to see hours of unskilled workers gaining on skilled workers' hours. Instead, we see the reverse. When the wage gap between skilled and unskilled workers dwindled, firms shifted towards skilled workers -- a phenomenon that makes sense, really, only if the wage compression is imposed on firms, and firms respond through hours.

There's a lot more to be done on this, for sure. But this is, I think, some neat new evidence that helps to pin down the critical role of institutions in the Great Compression.