There is no statistical evidence of job lock even prior to government welfare programs.

Anecdotal evidence confirms that job lock is a serious problem for some people. Nevertheless, we find no statistically significant increases in employment durations or decreases in wages for those with employer-provided health insurance and with health problems in the family as would be expected if job lock were pervasive. We test for job lock using a hazard model framework exploiting variation in employment durations by employer-provided health insurance status and the actual health conditions of family members. Thus, we obtain a very direct test of the job lock phenomenon, instead of relying on proxies for family health such as other health insurance coverage or family size.

Other researchers have found that worker mobility is affected by family size or the presence of other insurance, and they have interpreted these results as evidence of job lock. When we estimate duration models that depend on the presence of other insurance or family size, we do get similar results to those obtained earlier. We find that employment durations increase for those in large families with employer-provided health insurance, and that employment durations decrease for those having both employer-provided health insurance and health insurance from another source. On the other hand, using the measures based on the interaction of employer-provided health insurance and the actual health conditions of family members, we find no evidence of job lock in employment durations. Some would argue that this constitutes mixed evidence in favor of job lock on employment durations.

But if job lock is a pervasive phenomenon that is distorting labor market decisions, we should see lower wages for those workers who are potentially locked into their jobs. We do not, whether we use the more direct measures of job lock (i.e., the interaction of employer-provided health coverage and health conditions) or measures using dual coverage and family size. In fact, for men we find exactly the opposite of what we would expect if job lock were driving the results. Male workers with large families have higher wages, and male workers with dual health insurance coverage have lower wages. Thus, we consistently find no evidence of job lock using the measures based on the actual health conditions of family members, whether we look at the outcome of employment duration or wages. When using the measures of family size or dual coverage, the employment duration and wage results yield opposite conclusions with respect to job lock. One potential explanation for this inconsistency is that measures such as family size and dual coverage may be more likely to be correlated with unobserved factors such as ability and motivation. For example, those with large families may be more likely to have longer employment durations and higher wages, which could explain the inconsistent results. On the other hand, job lock tests based on the health status of family members are less likely to suffer from this potential problem.

What are we left to conclude? Despite anecdotal evidence of job lock, our results based on the actual health conditions of family members do not suggest that this phenomenon is pervasive in the U.S. economy. So why does this all matter? Earlier research on job lock led to estimates that put the number of job-locked workers in the millions. The perception that millions of workers were job locked figured into the legislative cost-benefit analysis that led to the passage of HIPAA in 1996. If HIPAA has no downside, then overestimating the number of job-locked workers is not important. But the guaranteed issue requirements and limitations on preexisting condition exclusions in HIPAA have raised the costs of supplying health insurance in the small group market (U.S. General Accounting Office 1998), and that may be contributing to the continuing rise in the number of Americans who lack employer-provided health insurance.

The Employee Benefits Research Institute (1998) survey found that a considerable number of workers, when asked about job lock, indicated that they somehow feel less mobile because of employer-provided health insurance. Does this mean that job lock is a problem that indeed required a legislative fix, or do these survey responses merely indicate that employer-provided health insurance, along with any number of other job-connected or personal factors, may cause workers to think twice before looking for alternative employment? Our results indicate that the latter interpretation may be closer to the truth. Even in the pre-HIPAA environment, we were able to detect no statistically significant decrease in the mobility or wages of workers with employer-provided health insurance and a family member with an adverse health condition.