There is considerable geographic variation in cross-sectional estimates of uninsurance in the United States, ranging from less than 5 percent of the nonelderly in Massachusetts to well over 25 percent in Texas (U.S. Census Bureau 2011).  It is also well known that spells without coverage are quite varied in terms of their length:  between 2001 and 2004, 40 percent of adult spells concluded within 4 months, while one quarter lasted longer than 2 years (Cutler and Gelber 2009). 

What is less well understood is whether cross-sectional variation in coverage is explained by differences in the incidence or the duration of uninsured spells. That is, while we know that 16 percent of Oregonians and 11 percent of Rhode Islanders lacked coverage in 2010, we do not know whether the uninsured in Oregon remain without coverage for longer periods than those in Rhode Island. 

Understanding the implications of geographic variation in spell dynamics is important because the targeting, cost, and retention of enrollment in the insurance coverage expansions envisioned by the Affordable Care Act (ACA) will depend critically on the underlying factors currently generating periods without coverage.  For instance, if most spells in a given state last for short durations as people change jobs, then policymakers should expect a greater amount of transitional “churning” between ESI and programs such as Medicaid and the subsidized Exchanges.  By contrast, states with a higher prevalence of long spells among individuals ineligible for ESI could see greater retention in subsidized coverage, but higher costs as these individuals may be less likely to experience “triggering” events -- such as shifts in income due to new employment -- that affect their eligibility. 

This study utilizes survey data from the 2008 Survey of Income and Program Participation, coupled with calibration data from the 2009 and 2010 American Community Survey, to produce the first state-specific estimates of coverage dynamics among the nondelderly U.S. population.  To address small sample sizes observed for many states, I utilize a novel Poisson regression-based reweighting algorithm for small-area estimation developed by Zaslavsky and Schirm (1997, 1998, 2002) to produce state-specific estimates that “borrow strength” from similar neighboring areas.  The results show considerable geographic heterogeneity in spell durations, with most of the variation occurring after 12 months.  Preliminary results show that one half of the uninsured spells in Minnesota conclude within 9 months, compared to 8 months in Florida.  But among the remainder, nonelderly adults in Florida remain uninsured for much longer:  one quarter are uninsured for longer than 32 months, while the comparable figure (i.e., the 75^th percentile) in Minnesota is 19 months. 

            The effectiveness, efficiency and cost of covering the uninsured in the U.S. requires better understanding the dynamic factors that inform periods without coverage.  This study provides the first such estimates at the state-level, and provides key lessons for state and federal policymakers as they move forward with targeting the coverage expansions of the ACA.