A janitor walks past a MARC train waiting room. Sarah L. Voisin / The Washington Post

Talk of "two Americas" feels like 2004 election kitsch at this point, but for some economists, it's a going concern. The Great Recession, according to a popular theory known as "job polarization," has led to a loss of middle-income jobs and created a new reality in which the only jobs are either lucrative and highly skilled (computer programming, for instance) or else menial and poorly remunerated (e.g. janitorial work).

Some, like Duke’s Nir Jaimovich and the University of British Columbia’s Henry E. Siu, credit this change to technological developments that have made middle-income careers like manufacturing and sales obsolete. If a robot arms mean you need fewer man-hours to build a car, and Amazon and Staples mean you need fewer flesh-and-blood paper salesmen in Scranton, employment in those sectors is going to fall, or so the logic goes.

But a new paper by Josh Mitchell and Austin Nichols at the Urban Institute challenges this picture. Mitchell and Nichols look at the Survey of Income and Program Participation (SIPP), a Census survey that tracks groups (or "panels") of respondents over the course of four years. That allows analysts to compare how the same people's economic lives changed over an extended period. To test the job polarization story, Mitchell and Nichols compare the 2004 panel (which tracked workers from the start of 2004 through the end of 2007) to the 2008 panel (which covered 2008 through 2011).

Of course, one would expect that one's risk of becoming unemployment would increase between 2004 and 2008, no matter where one falls on the income spectrum. But if job polarization is speeding up during the recovery, you'd expect that risk to shoot up more for middle-income people than for the poor of the affluent. But Mitchell and Nichols find that the risk of unemployment grew across the board, and not just for those in the middle:

Source: Mitchell and Nichols

This belies the conclusion of articles like that of Jaimovich and Siu. Everyone sees their odds of becoming unemployed shoot up, and the biggest jump is among low earners, rather than among middle income people.

That said, Mitchell and Nichols do find two areas for concern. While middle-income people were no likelier to lose their jobs, those that did become unemployed saw bigger pay cuts once they got another job than people at the top or bottom of the income scale:

Source: Mitchell and Nichols

In 2004, if you were a middle-earner who lost your job, you were likely to take a 10 percent pay cut once you got back into the workforce. That was pretty bad, but better than the nearly 25 percent cut faced by high earners. But in 2008, the fates of the two groups converged. Both saw pay cuts of about 20 percent, on average, while the poorest actually got a raise, typically. You see a similar phenomenon in health insurance coverage. Mitchell and Nichols find that middle-income people were the likeliest group to lose health insurance in the 2008 panel, a reversal from 2004, when the poor were the likeliest to lose coverage:

Source: Mitchell and Nichols

So job polarization probably isn't preventing people in the middle from keeping their jobs. But the fact that middle earners' compensation — both in terms of money and in terms of benefits — is taking a harder hit than that of other groups' suggests that the recession may be upsetting class dynamics in an important, enduring way.