One problem that confounds economists and policymakers has been what to do about job polarization: the phenomenon, over the last three decades, of a declining number of "middle-skill" jobs and their replacement with growth in high- and low-skill occupations. While 59 percent of those employed had middle-skill jobs in 1983, only 45 percent did by 2012. This trend increases income inequality and leaves workers without the education or ability to perform high-skill work with fewer, and grimmer, job opportunities.

Source: Federal Reserve Bank of Kansas City
Source: Federal Reserve Bank of Kansas City

Many observers (including me) have tended to assume this phenomenon was driven mainly by sectoral changes in the economy: Manufacturing and related industries tend to require a lot of middle-skill workers, while service industries are more dependent on high- and low-skill workers. That assumption also shows up in the political discussion of job polarization. President Barack Obama's administration has focused specifically on expanding manufacturing employment as part of an effort to create more middle-class jobs.

But a new paper from Didem Tüzemen and Jonathan Willis, economists at the Federal Reserve Bank of Kansas City, shows that assumption is wrong. They find that only a third of the last 30 years’ shift away from middle-skill employment was due to sectoral changes. Two-thirds of the shift can be attributed to changes within sectors. Manufacturing employment didn’t just decline; middle-skill work also declined as a share of employment within manufacturing, and also in every other sector of the economy, as shown in the following chart from their paper.

Source: Federal Reserve Bank of Kansas City
Source: Federal Reserve Bank of Kansas City

The classic example of a white-collar, middle-skill job is administrative assistant, and service-sector companies employ a lot fewer than they used to because computers have reduced demand for secretarial help. In 1983, 31 percent of working women had jobs classified as "office and administrative support"; by 2012, the figure was just 20 percent. Relatedly, the share of jobs in the education and health sectors that were middle-skill fell from almost a quarter in 1983 to about 15 percent in 2012.

The dominance of intra-sector shifts calls into question the ability of government to reduce inequality and unemployment by supporting job growth in the manufacturing sector. A better option might be to try to adjust how we train the workforce, so that more workers are suitable for high-skill jobs.

Tüzemen and Willis note that the labor force has actually been quite responsive to the last 30 years’ change in the kind of labor companies are looking for. Normally, if companies start to demand more high-skilled workers and fewer middle-skilled workers, you would expect high-skill job wages to rise and middle-skill job wages to fall. But the wage differential between high-, middle- and low-skill work has actually stayed pretty constant over the last 30 years. They conclude that “decline in labor demand has been approximately matched by a decline in labor supply for these occupations.”

Such shifts are to the good -– if they reflect the fact that people who under other circumstances would have been middle-skill workers have gotten the necessary training to become high-skill workers. That is part of the story, and is particularly true of women workers, who have greatly increased their educational attainment. But the labor force has also shifted to meet changing labor demand in less promising ways.

Some young people are getting more education in order to be able to get high-skill jobs, and some high-skill people who are near retirement age are working longer. But some young people who in previous decades might have gone into middle-skill work are jumping right into the workforce and taking low-skill jobs; and some middle-skill workers in upper middle-age are retiring early because their job skills are no longer in demand. Both of those are undesirable methods of labor supply adjustment.

You can see these effects in the chart below: In all age cohorts, the share of workers in middle-skill jobs has declined by about 14 percent since 1983. Among older workers, all of the shift has been toward high-skill employment, as more skilled workers retire later and less-skilled workers retire earlier. And among younger workers, the bulk of the shift has been toward low-skill employment, as some young people delay workforce entry and others go directly into low-paying jobs.

Source: Federal Reserve Bank of Kansas City
Source: Federal Reserve Bank of Kansas City

This raises two key policy challenges: Can we make educational opportunities more suitable and affordable for young people who are currently being shunted into low-skill work? And can we re-train older middle-skill workers so they don't feel compelled to drop out of the labor force? The latter, in particular, is an important barrier to cracking the problem of long-term unemployment.

(Josh Barro is lead writer for the Ticker. Follow him on Twitter.)