There's widespread agreement at this point that wage inequality has been growing at a remarkable pace since the 1970s. The data on that much are clear:

But what's causing this is a matter of dispute. One explanation, initiated by Harvard's Lawrence Katz and Claudia Goldin and elaborated on by MIT's David Autor and Daron Acemoglu, is that technological innovations are making easily automated jobs less common. But these jobs tend to be low to medium-skill, high-paying occupations, such as working on a factory assembly line. The result is a growth in occupations that are hard to automate, which tend to either be very menial and low-paying (such as janitorial labor) or high-paying but requiring considerable skills (like computer programming). So as the middle of the distribution gets carved out, the low and top ends grow.

At the 2013 meeting of the Allied Social Sciences Associations (ASSA), EPI's Lawrence Mishel, John Schmitt, and Heidi Shierholz presented a paper arguing that the theory, especially as promoted by Autor, is mistaken. A key part of their argument is the following chart. Autor determined the skill level of various occupations by ranking them by average wage, and devised "skill percentiles" to see where each fell. Then he found out how much each profession's share of employment changed between 1989 and 2000. Here are all his data points:

The X axis represents where the people fall on the skill distribution, and the Y axis represents the log of how much that occupation's share of employment changed. There's definitely a relationship: higher skilled occupations grew more than low skilled occupations, which grew more than medium-skilled occupations. Even a simple regression shows that 17.6 percent of the change in the groups' shares is due to technological change, and Autor uses more sophisticated techniques that find an even stronger relationship.

But Mishel tells me that he thinks the above graph is just noise, as the widely ranging data points show. If you just know what skill level a given occupation requires, that's not enough to tell you how much it'll grow in the future. There's no evidence, the paper concludes, to blame technological growth for wage inequality. Political forces, like the stagnant minimum wage, de-unionization, trade liberalization, and deregulation of key industries, Mishel says, deserve the brunt of the blame.

Autor says this is specious reasoning. "The paper is, of course, attacking a caricature of my work," he told me over the phone. "I would never claim that technical change is the only thing that matter for the labor market. What that paper claims is that technical change is the only thing that doesn’t matter." The first problem, he says, is that "the graph looks like noise" isn't a substitute for statistical analysis. "That’s why we used statistical analysis," he says. "You can’t just draw data points and draw inferences."

Secondly, he argues that you should expect to see a wide range in where the occupations fall just due to measuring error. Just splitting people into occupations is a process rife with opportunities to miscode people, and thus produce strange estimates of occupations' average wages. "An occupation is not stamped on your forehead. It’s not in your DNA," Autor says. "There’s tremendous observation error there. All that contributes to making the thing noisy." That's why it's important, he argues, to look at overall trends, not individual data points.

Another key piece of evidence Mishel and his coauthors cite is the fact that most of the growth in inequality has happened within professions, not between them. If technical change is causing inequality, then it shouldn't cause inequality among plumbers, but between plumbers and, say, bankers. But Mishel finds that between-occupation inequality hasn't grown much since the 1989, leaving little room for explanations that rely on technological change:

1995-2000 is left off because that period actually saw wage inequality decline, so dividing up the decade's non-existent increase in inequality into between and within-occupation numbers is nonsensical.

Autor responds by pointing (a) again, that his contention isn't that technology explains all of the increase and (b) technological change could cause within-occupation inequality to increase. "[The paper] claims that everyone who works in an occupation is identical," he tells me. "It’s kind of risible, right? No one thinks that. We have secretaries now and we had secretaries 34 years ago, and they do very different things. They did a lot of filing and typing back then, but now they’re problem-solvers…The notion that there would be no [skill] variability within occupations is totally incorrect."

Mostly, Autor thinks this fight is a distraction. "Larry and people in that group hate technical change as an explanation of anything. My opinion about why they hate it that much is that it’s not amenable to policy," he speculates. "All these other things you can say, Congress can change this or that. You can’t say Congress could reshape the trajectory of technological change." But that's not a good enough reason, Autor thinks: "Your diagnosis of the cause of something should be independent of what you want to do about it."