Social Science Statistics Blog

« Bayesian Propensity Score Matching | Main | Garip presenting "How Do Network Externalities Lead to Intergroup Inequality?" »

15 February 2009

How many people could improved COBRA cover?

One feature of the recently-passed stimulus legislation is a temporary federal subsidy program for individuals to purchase transitional health insurance coverage in the event they lose their job. This coverage -- called COBRA after the legislation that created the program in the 1980's -- is generally available to displaced workers in firms with more than 20 employees, although some states have also adopted policies that allow employees of smaller firms to buy into transitional coverage. The catch, however, is that workers must pay the full premium amount plus 2 percent for administrative expenses.

The new federal program allows for subsidies of up to 65 percent of the cost of health insurance, which is aimed to provide a boost to individuals trying to make ends meet and remain insured while they're unemployed. But the question is, how many people could this program potentially cover in a given year?

Below I've created a population flow diagram using R's diagram package. The underlying data are drawn from the 2006 Medical Expenditure Panel Survey (MEPS) -- a nationally-representative survey of American households conducted each year by the Agency for Health Care Research and Quality. To construct the diagram I took the health insurance coverage status of non-elderly adults and children in January 2006, and compared this to their coverage status in December of that year.

Clearly, the potential for reducing the number of uninsured is large -- according to these estimates, 6.4 million adults and 1.4 million children lost employer-based group insurance between January and December 2006.* If just one-half of these individuals took up a subsidy and were able to continue that group coverage, the number of uninsured in 2006 could have been reduced by about 4 million. Moreover, this is almost certainly an underestimate, since there are also individuals who lost employer-based coverage prior to January, as well as individuals who were intermittently uninsured during the year, who may not show up in the diagrammed flows but could also have benefited from access to subsidized transitional coverage. Finally, I would also note the potential for spillover and cost-offsetting effects within Medicaid, as the estimated 1.2 million who went from group coverage to public coverage could have also retained their employer-based insurance, saving states and the federal government the costs of these extra Medicaid/SCHIP enrollees.

* Note, however, that I have made no attempt to produce confidence intervals around these figures, though in principle this would be straightforward to do with R's survey package. If anyone knows how to easily input these into the figures (I couldn't figure out a way), please let me know!

Posted by John Graves at February 15, 2009 4:51 PM