One piece of evidence comes from the relationship between job openings and unemployment. Normally these two numbers move inversely: the more job openings, the fewer Americans out of work. And this traditional relationship remains true if we look at short-term unemployment. But as William Dickens and Rand Ghayad of Northeastern University recently showed, the relationship has broken down for the long-term unemployed: a rising number of job openings doesn’t seem to do much to reduce their numbers. It’s as if employers don’t even bother looking at anyone who has been out of work for a long time.

To test this hypothesis, Mr. Ghayad then did an experiment, sending out résumés describing the qualifications and employment history of 4,800 fictitious workers. Who got called back? The answer was that workers who reported having been unemployed for six months or more got very few callbacks, even when all their other qualifications were better than those of workers who did attract employer interest.