By all standard definitions, the U.S. economy is at “full employment,” which loosely means that everyone who wants a job has one.

The unemployment rate declined to 4.2 percent in September, and the number of unemployed persons declined by 331,000. At last count (August), there were more than six million job openings across the United States, up more than 590,000 from a year prior and the highest ever recorded.

Ray Perryman

A small amount of unemployment is desirable from an economic perspective and necessary for the proper functioning of the economy. As people enter the workforce, a lag in finding a job is to be expected. Similarly, if individuals relocate, some time may be needed to find employment. This type of friction is to be expected. Some areas have even lower rates for periods of time, but they are generally not sustainable.

Even with this good news, however, nearly seven million Americans are unemployed, about 25 percent of them for six months or more. The labor force participation rate has changed little over the year and now stands at more than 63 percent, which is very low by historical standards. More than five million persons are employed part time for economic reasons (sometimes referred to as involuntary part-time workers). These are people who would rather be working full time, but who were either unable to find full-time work or their hours had been cut.

The point is that the overall unemployment rate can conceal significant problems, even when it’s low enough to be called “full employment.” Although labor market conditions have improved drastically since the recession, there are nonetheless millions of people who can’t find work (the seven million unemployed), can’t find enough work (the five million more working part time who would rather be working full time), or have given up entirely (the millions more who are no longer even counted because they’ve stopped looking and, thus, are not participating). From a societal perspective, this pattern is clearly undesirable.

One problem is geography, in that there is significant disparity in job availability in various areas. Some states have unemployment rates over six percent, for example, while others are less than three percent. Even within Texas, rates vary significantly by city. Although workers have become more mobile over time, there are still barriers to moving to find work, and many people are simply unwilling or unable to do so. For some areas, the current “full employment” doesn’t mean plenty of jobs. This problem was particularly acute during the Great Recession as upside-down mortgages limited the ability of many workers to relocate.

Another issue is skills mismatches. Many of the unemployed are not qualified for the jobs that are currently available. The Bureau of Labor Statistics (BLS) analyzed employment trends by educational requirements before, during, and after the recession. Between May 2007 and May 2010 as the recession took a toll on the job market, nearly 7.4 million U.S. jobs were lost in occupations that typically require a high school diploma or no formal educational credential for entry. By May 2016, although overall employment exceeded the May 2007 level, the number of jobs in positions with no formal educational credential required for entry remained nearly 1.3 million lower. In other words, a disproportionate number of the jobs that permanently disappeared during the recession were concentrated in occupations that typically require only high school or less.

The skills mismatch goes beyond general level of education. Currently, many job openings are in industries that tend to require specialized training. In health care and social assistance, for instance, the rate of job openings is 5.3 percent. Without the training required for the field, however, a job seeker would never be hired. Similarly, some technology-oriented businesses are finding it difficult to locate workers with the necessary tech skills.

Industries are shifting and jobs are changing, and the trend will continue. Machine learning and automation may eliminate the vast majority of positions in some occupations. Not only will these changes come in the form of robots in manufacturing facilities or self-driving trucks, but also in sophisticated softwares which can reduce the need for highly trained persons ranging from accountants to attorneys to software developers who are doing repetitive tasks which machines can learn to do (thereby shrinking the need to hire humans). The transformation to online and self-checkout retail is also a major factor for lower skilled workers.

Looking ahead, BLS projects that the fastest growing occupational categories will be those requiring the highest levels of education or postsecondary nondegree awards. While the fastest growth doesn’t mean the largest numbers of new jobs given differences in the sizes of the categories, it is nonetheless an indication of an underlying shift. The correlation between education and income and the likelihood of having a job are well documented, and it is only going to intensify over time. Encouraging and facilitating educational attainment is essential to ongoing prosperity.

Another requirement is more and better training and retraining for those in categories of occupations which are shrinking to better prepare the affected workers for jobs that are available. Millions of Americans are looking for work, and if we could equip them better, they could fill some of the millions of jobs now available. Supporting existing public and private education and training programs is essential, and other ways to get people the skills they need should be explored.

Economic growth is necessary, but not sufficient, to eliminate undesirable joblessness. Even with the economy at “full employment” and millions of job openings, there are millions of people who want more or better work. Long-term individual and societal prosperity depend on finding ways to chip away at this mismatch problem.