New Estimates of Gaps and Transitions in Health Insurance

Abstract

Changes in individual or family circumstances cause many Americans to experience gaps and transitions in public and private health insurance. Using data from the 2004–2007 Survey of Income and Program Participation, this article updates earlier analyses of insurance gaps and transitions. Eighty-nine million people (one third of non-elderly Americans) were uninsured for at least one month during those four years. Approximately twenty-three million lost insurance more than once. The analyses call attention to the continuing instability and insecurity of health insurance, can inform implementation of national reforms, and establish a recent baseline that will be helpful in evaluating the reforms’ effects on coverage stability.

INTRODUCTION

The Congressional Budget Office (2011) has projected that the implementation of health insurance reforms in the Affordable Care Act (ACA) will reduce the number of uninsured Americans by thirty-three million in 2020, from fifty-six to twenty-three million people. Most of the coverage gains will come from expanding Medicaid to everyone below 133 percent of the poverty line (138 percent with income disregards), and from creating health insurance exchanges. In the exchanges, refundable federal tax credits will subsidize private health insurance for anyone below 400 percent of the poverty line who is not eligible for public coverage. The ACA also has new financial incentives for small employers to cover their workers.

CBO’s projections, which count the uninsured only at one point in time, offer a conservative picture of the number of Americans who would be uninsured without the reforms. Research describing changes in individuals’ health insurance over time has shown that gaps and transitions are common. About half of all spells without insurance end within six months, and short spells are far less likely than long spells to be counted in point-in-time estimates (Swartz & McBride, 1990; Swartz, Marcotte, & McBride, 1993; Swartz, 1994; Bhandari & Mills, 2003; Congressional Budget Office, 2003; Loveless & Tin, 2006). In an earlier article that documented the instability of health insurance in the late 1990’s, we reported that one out of three non-elderly Americans were uninsured for at least one month over four years (Short & Graefe, 2003). Moreover, fully one-third of those ever uninsured in four years lost health insurance more than once.

There was a major change in public health insurance for children in 1997, when most uninsured children with incomes below 200 percent of the poverty line who were not already eligible for Medicaid became eligible for the Children’s Health Insurance Program (CHIP). According to two recent studies that modeled the likelihood of a change in coverage from one month to the next, health insurance became increasingly unstable by the early 2000s compared to the mid-1980s and early 1990s, with more people who were initially insured becoming uninsured (Cutler & Gelber, 2009; Hill & Shaefer, 2011).

This article updates our earlier analyses of gaps and transitions in health insurance (Short & Graefe, 2003) with data from 2004 to 2007. Like the late 1990’s, the mid-2000s was a time of relative economic prosperity, when the unemployment rate hovered around 5 percent (U.S. Bureau of Labor Statistics, 2011). These are the most recent data available for four-year health insurance estimates prior to the “Great Recession,” which caused many people to lose employer-sponsored health insurance. As such, the data for 2004–2007 describe health insurance gaps and transitions during the last “normal” economic conditions that the U.S. will have experienced before the implementation of national health insurance reforms. This information will be important for assessing the reforms’ effects on coverage stability.

NEW CONTRIBUTION

In this paper, we examine longitudinal patterns of coverage over the four years of 2004–2007. In contrast to studies by Cutler and Gelber (2009) or Hill and Shaefer (2011), which considered coverage transitions in isolation from each other, our study tracks individuals’ coverage over the entirety of four years in order to describe the sequential patterns, including the multiple transitions that people often experience. Additionally, we report on patterns of coverage and transitions separately for adults and children by income categories. Information about gaps and transitions is important for state and federal officials as they consider ways of minimizing disruptions in the health insurance of people who experience changes in income, employment or family structure that could cause changes in eligibility for public or private coverage under the ACA.

Our goals in providing new estimates of gaps and transitions in health insurance are threefold: (1) to call attention to the continuing instability and insecurity of coverage that Americans would face – even in relatively good economic times – if the U.S. does not follow through in implementing the ACA; (2) to inform the implementation of the ACA reforms; and (3) to establish a recent baseline uncomplicated by the high unemployment rates that have characterized the U.S. economy since late 2007. When combined someday with estimates for the current period of high unemployment, this baseline will be critical in evaluating the effects of the ACA on long-term trends in coverage stability, abstracted from the cyclical instability associated with changing macroeconomic conditions.

DATA AND METHODS

Our earlier study examined gaps and transitions in coverage over the four years from 1996 to 1999, using monthly health insurance data from the 1996 panel of the Survey of Income and Program Participation (SIPP). Here we replicate that study’s methods and analyses with data from the most recently completed SIPP panel, covering 2004 through 2007. Additionally, we re-analyze the 1996–1999 data to highlight differences in trends for low-income adults and children associated with the implementation of CHIP. We estimate standard errors that account for SIPP’s complex design and note statistically significant differences (p<.05) in tables and figures, and we report only significant differences in the text.

DATA. SIPP is a nationally representative longitudinal survey that collects data for the same individuals every four months. Each month, a quarter of the households are interviewed about every household member’s status and experiences in the previous four months. As a consequence, the calendar dates for the 48-month reference period vary across the four quarters of the sample. For example, the first interviews for the 2004 panel were in February 2004 and collected data back to October 2003. The last interviews were in January 2008 and collected data through December 2007. The earliest data collected from the 1996 panel were for December 1995, while the last data were for February 2000.

SIPP respondents have a known tendency to report more changes in the “seams” between interviews than between months covered by the same interview (U.S. Census Bureau, 2001). Consequently, the total number of months that each person was insured or uninsured over the panel is often reported as a multiple of four. Seam bias particularly affects counts of insurance status by month, which is not a focus of our study. If SIPP respondents systematically under-report very short uninsured spells that begin and end between interviews, then our study may understate the number of people who were uninsured for at least one month in four years or experienced more than one uninsured spell.

POPULATION STUDIED

SIPP is a household survey that is representative of the civilian non-institutionalized population residing in the United States. The Census Bureau creates longitudinal survey weights that adjust for non-response and survey attrition. These weights match population totals on key dimensions (age, sex, ethnicity, race, household/family structure, employment, family income and assets, welfare receipt, education, and geographic location). As in our earlier study, we made an additional adjustment to the longitudinal weights to match weighted counts of coverage status by age and family income in the last month of the survey – using the monthly weights provided by the Census Bureau. This adjustment helps to guard against attrition bias in estimates of coverage over time, in case individuals who were more likely to become uninsured were also more likely to drop out of the survey because of moving or other life changes.¹

Our analyses describe health insurance gaps and transitions of the U.S. population who were under age sixty-five throughout the survey period. To standardize the opportunity for changes in health insurance over time, the analyses were further restricted to individuals continuously in the resident population for forty-eight months. Consequently, newborns (less than four years old at the end of the panel) were excluded, along with people who migrated in or out of the U.S., died, or were institutionalized during the four years. Approximately 225.6 million people in the U.S. population met the study eligibility requirements at the end of the 1996 panel, and 245.4 million people were eligible at the end of the 2004 panel. Unweighted sample sizes were 40,371 and 20,879 for the 1996 and 2004 panels, respectively.² The unweighted samples include 13,759 people ever uninsured during the 1996 panel and 6,966 people ever uninsured during the 2004 panel. In making separate estimates for adults and children, we defined children as younger than 19 years of age, in keeping with Medicaid eligibility rules.

COVERAGE PATTERNS OVER TIME

The longitudinal design of SIPP allowed us to follow the same people over time to study changes in health insurance and count months with and without insurance. We began by assigning individuals to a single coverage category for each month. Participants with multiple sources of coverage were assigned to the first applicable category in the following hierarchy: Medicaid or State Children’s Health Insurance Program (CHIP), Medicare, employer group, non-group private, and uninsured. The 1996 SIPP panel did not distinguish between Medicaid and CHIP; therefore, the two coverage sources were combined into one category. The employer-group category included military coverage such as CHAMPUS/CHAMPVA and union-based coverage.

We used the monthly coverage variables to assign everyone who was ever uninsured to one of seven mutually exclusive patterns of coverage over time. These patterns, which we used to describe coverage dynamics in the 1996 panel, are defined by the number and types of health insurance changes over four years. The patterns are depicted as different coverage timelines in Figure 1. They include (1) always uninsured, (2) transition into coverage, (3) transition out of coverage, (4) single gap in coverage, (5) temporary coverage, (6) frequent changes in coverage, and (7) repeatedly uninsured. The last two patterns are particularly unstable; they encompass all of the uninsured with three or more changes in coverage over four years.

Figure 1. Four-Year Coverage Patterns for Individuals Ever Uninsured, U.S. Population Under Age 65, 1996–1999 and 2004–2007.

Notes: Dashed lines denote uninsured periods; solid lines denote insured periods. Asterisks denote transitions experienced by some, but not all, of the people with the specified pattern. Total number of people ever uninsured: 89.0 million in 2004–2007; 84.8 million in 1996–1999. ^aSignificant change over time in percentage of the uninsured (p < .05). Standard errors in parentheses.

INCOME AS A PERCENTAGE OF POVERTY

We adopted a long-term measure of economic well-being in our calculation of income as a percentage of poverty. We summed monthly family income for each survey participant across the forty-eight months, defining this sum as long-term family income. Monthly poverty thresholds assigned to each person by the Census Bureau also were summed across the forty-eight months. The result of this summation is the aggregate amount of family income over four years that would have maintained each person exactly at the poverty standard every month, with monthly adjustments for changes in family size. We then assigned long-term income as a percentage of the federal poverty level (FPL) to each person by dividing long-term family income by the long-term poverty standard.

RESULTS

LONG-TERM RISK OF BEING UNINSURED

A total of eighty-nine million people, 36.3 percent of Americans ages 4 to 64, were uninsured for at least one month during the four years from 2004 to 2007 (Table 1). By comparison, nearly eighty-five million people were uninsured at some time during the four years from 1996 to 1999, when they represented 37.6 percent of the smaller population at that time. To put four-year estimates for these two time periods in perspective, point-in-time estimates from the National Health Interview Survey (NHIS) indicate a monthly average of 16.6–16.7 percent of Americans under age 65 were uninsured during each period (Cohen, Makuc, Bernstein, Bilheimer, & Powell-Griner, 2009). While NHIS estimates of the uninsured are similar for 1996–1999 and 2004–2007, employment-based insurance decreased between those time periods, while public insurance increased.

Table 1.

Number and Percentage of Children and Adults Ever Uninsured Over Four Years by Income, 1996–1999 and 2004–2007

	All ages		Children		Adults
	1996– 1999	2004– 2007	1996– 1999	2004– 2007	1996– 1999	2004– 2007
All incomes
Millions uninsured	84.8	89.0	25.5	26.6	59.3	62.5
Percent uninsured	37.6% (0.5)	36.3% (0.6)	42.0%b (0.7)	43.1%b (1.0)	36.0%b (0.5)	34.0%b (0.6)
Percent uninsured by income
<200% poverty	67.7% (0.8)	62.5% (1.3)	67.5%a (1.1)	59.5%a (1.7)	67.8% (0.8)	64.2% (1.3)
200–399% poverty	34.2% (0.6)	37.1% (1.0)	28.3%ab (0.9)	39.3%a (1.6)	36.3%b (0.6)	36.4% (1.1)
400+% poverty	14.7% (0.5)	15.0% (0.7)	13.1%a (0.9)	22.8%ab (1.7)	15.0% (0.5)	13.3%b (0.6)

Notes:

^aSignificant change for age group over time (p < .05).

^bSignificant difference between children and adults in the same time period (p < .05). Standard errors in parentheses.

Table 1 shows dramatic differences by income in the percentages ever uninsured in both time periods. In 2004–2007, 64.2 percent of adults and 59.5 percent of children with incomes below 200 percent of the FPL were uninsured for at least one month over four years. The rates were much lower for adults and children in higher income groups.

When we compared changes in ever-uninsured rates between 1996–1999 and 2004–2007 (Table 1), the changes differed between children and adults. For children, they also differed by income level. For adults, there was a small, but not statistically significant decrease in the uninsured rate below 200 percent of the FPL, and similarly little change at higher income levels. For children under 200 percent of the FPL, the income group targeted by CHIP, the uninsured rate dropped by 8 percentage points. In contrast, uninsured rates for children increased by 10 to 11 percentage points in each of the higher income categories. Thus, among children and adults below 200 percent of the FPL, children moved over time from parity with adults to having a substantially lower uninsured rate. In the next higher income category, children moved from a noticeably lower uninsured rate to parity with adults. In the highest income group, children moved from parity with adults in 1996–1999 to a much higher uninsured rate in 2004–2007, when 22.8 percent of children were uninsured for at least a month compared to 13.3 percent of adults.

FOUR-YEAR COVERAGE PATTERNS OF THE UNINSURED

Among the 89 million people who were ever-uninsured from 2004 through 2007, 12 million people (rightmost column of Figure 1) were uninsured throughout the four-year period. Just over 11 million transitioned into coverage and remained insured, while 11.5 million transitioned out of coverage and remained uninsured for the rest of four years. Fourteen million people experienced a single gap in coverage, while just over 6 million people had a single (temporary) spell of coverage and were otherwise uninsured (row 5 of Figure 1).

The two groups with the most unstable coverage patterns among the ever-uninsured, involving three or more changes over four years, are described in the bottom two rows of Figure 1. One group (amounting to about 11 million people) had one uninsured gap and changed coverage types at least once in four years. In our earlier article, we characterized this group as “scrambling for coverage.” The other group with particularly unstable coverage – almost 23 million people – lost health insurance more than once over four years. This was the most frequently occurring pattern among those ever uninsured over four years. It characterized the coverage of 25.6 percent of people who were ever uninsured.

Table 2 shows the total number of months that people experiencing each four-year pattern were uninsured over four years. In particular, as shown in the bottom row of the table, one-third of people who repeatedly lost insurance were uninsured for more than 24 months in total. About two-thirds were uninsured for more than 12 months in total. The majority of people experiencing three other uninsured patterns – transitions into coverage, transitions out of coverage, and temporary coverage – were also uninsured for more than 12 months in total. In contrast, the majority of people with a single gap in coverage over 48 months – with or without additional changes in type of coverage – were uninsured for 4 months or less.³

Table 2.

Percent Distribution of Individuals Ever Uninsured Over Four Years by Number of Months Uninsured, According to Coverage Patterns, U.S. Population Under Age 65, 2004–2007

Coverage pattern	Millions	Number of months uninsured
		All	1–4	5–12	13–24	25–48
		Percent distribution of the uninsured
Always uninsured	12.0	100.0%	NA	NA	NA	100.0% (0.0)
Transition into coverage	11.2	100.0%	26.2% (2.0)	23.3% (1.9)	20.9% (1.6)	29.6% (2.0)
Transition out of coverage	11.5	100.0%	15.0% (2.0)	18.0% (1.9)	20.1% (2.0)	46.8% (2.7)
Single gap in coverage	14.0	100.0%	52.1% (1.9)	28.8% (1.5)	13.9% (1.2)	5.1% (0.7)
Temporary coverage	6.3	100.0%	0.0% (0.0)	4.4% (1.3)	14.0% (2.4)	81.5% (2.6)
Frequent changes	11.2	100.0%	55.4% (1.9)	27.0% (1.9)	12.7% (1.2)	4.8% (0.8)
Repeatedly uninsured	22.8	100.0%	1.7% (0.3)	28.6% (1.4)	37.2% (1.7)	32.6% (1.5)

Notes: Standard errors in parentheses.

CHANGES IN UNINSURED COVERAGE PATTERNS BETWEEN TIME PERIODS

As can be seen in the last two columns of Figure 1, repeated lapses in coverage were more common than any other four-year pattern for people ever uninsured in both 1996–1999 and 2004–2007 (bottom row, “repeatedly uninsured”). In 1996–1999, 33.2 percent of the ever-uninsured were repeatedly uninsured, but the share dropped to 25.6 percent in 2004–2007. This reduction of 7.6 percentage points from 1996–1999 to 2004–2007 was the largest reduction in any coverage pattern. Single gaps in coverage also became significantly less common among the uninsured in the later time period. The largest increase between time periods was in the percentage of the uninsured who lost coverage and never regained it. Thus, there were significant decreases in two patterns where the uninsured returned to coverage over the four years, and a significant increase in a pattern where they did not. Additionally, relatively more of the ever-uninsured in 2004–2007 were in the category with frequent changes in coverage.

DIFFERENCES IN 2004–2007 UNINSURED COVERAGE PATTERNS BY AGE

Among people uninsured for at least one month in 2004–2007, children were much less likely than adults to be uninsured throughout the four years (5.4 percent compared to 16.9 percent, Table 3). At the same time, children were more likely than adults to have frequent changes in coverage, single or repeated gaps in coverage, and to transition out of coverage for the remainder of four years. Among uninsured adults, the percentage always uninsured increased with age, while the percentage repeatedly uninsured decreased with age.

Table 3.

Percent Distribution of Individuals Ever Uninsured by Four-Year Coverage Patterns, According to Age, U.S. Population Under Age 65, 2004–2007

	Age at end of four years
	Under 19	All adults	19–24	25–34	35–54	55–64
All uninsured
Millions	26.6	62.5	12.0	19.8	23.5	7.2
Percent of population	43.1%a (0.6)	34.0%a (1.0)	48.7% (1.5)	49.8% (1.3)	27.3% (0.7)	21.5% (0.9)
Percent distribution of age group
All uninsured	100.0%	100.0%	100.0%	100.0%	100.0%	100.0%
Always uninsured	5.4%a (0.9)	16.9%a (0.7)	9.2% (1.2)	14.5% (1.4)	20.9% (1.1)	23.0% (1.9)
Transition into coverage	7.0%a (0.8)	15.0%a (0.7)	8.1% (1.2)	18.1% (1.5)	15.1% (0.9)	17.8% (1.6)
Transition out of coverage	17.3%a (1.3)	11.1%a (0.6)	15.1% (1.5)	9.1% (1.0)	10.5% (0.9)	11.9% (1.6)
Single gap in coverage	21.3%a (1.3)	13.3%a (0.6)	13.4% (1.3)	12.1% (1.0)	14.7% (0.9)	12.3% (1.3)
Temporary coverage	3.2%a (0.6)	8.7%a (0.7)	8.1% (1.2)	10.1% (1.2)	8.0% (0.8)	8.1% (1.5)
Frequent changes	16.7%a (1.1)	10.8%a (0.5)	17.1% (1.4)	9.3% (0.9)	9.0% (0.7)	10.4% (1.3)
Repeatedly uninsured	29.0%a (1.5)	24.1%a (0.7)	28.9% (2.0)	26.8% (1.4)	21.8% (1.2)	16.5% (1.6)

Notes:

^aSignificant difference between children and adults (p < .05). Standard errors in parentheses.

The differences in four-year coverage patterns for uninsured children and adults in 2004–2007 were associated with differences in the types of coverage that each age group was most likely to gain or lose. Focusing on Medicaid/CHIP and employer insurance, the two most common sources of coverage in the population younger than 65, Table 4 indicates how often each of these sources was associated with any or all of the insured periods involved in each four-year coverage pattern. As shown in Table 4, Medicaid and CHIP played a much bigger role in the covered periods of uninsured children than uninsured adults. Approximately eighty-eight percent of children who were repeatedly uninsured were covered at some point by Medicaid or CHIP. Medicaid/CHIP was the only source of coverage over four years for 47.5 percent of children transitioning into coverage, 42.3 percent of children transitioning out of coverage, and 67.3 percent of children with temporary coverage.

Table 4.

Percent of Adults and Children Ever Uninsured who had Medicaid/CHIP or Employer-Sponsored Insurance in 2004–2007, According to Coverage Patterns, U.S. Population Under Age 65

	Percent of row with Medicaid/CHIP or employer coverage over four years
	Any Medicaid/ CHIP	Only Medicaid/CHIP	Any Employer	Only Employer
Ever-uninsured children
Always uninsured	N/A	N/A	N/A	N/A
Transition into coverage	64.5%a (5.4)	47.5%a (6.1)	43.3%a (5.8)	22.4%a (5.3)
Transition out of coverage	60.5%a (3.4)	42.3%a (3.5)	53.5%a (3.5)	30.3%a (3.2)
Single gap in coverage	47.2%a (3.3)	30.7%a (3.3)	67.0%a (3.4)	33.5%a (3.0)
Temporary coverage	67.3%a (7.0)	67.3%a (7.0)	25.0%a (7.1)	19.5%a (6.1)
Frequent changes	81.3%a (2.6)	0% (0.0)	88.4% (1.8)	0% (0.0)
Repeatedly uninsured	87.7%a (1.6)	28.2%a (2.3)	58.5%a (2.8)	3.5%a (0.7)
Ever-uninsured adults
Always uninsured	N/A	N/A	N/A	N/A
Transition into coverage	25.8%a (2.2)	18.0%a (1.8)	71.4%a (2.3)	54.6%a (2.4)
Transition out of coverage	30.8%a (3.0)	22.5%a (2.5)	71.7%a (2.7)	55.1%a (3.3)
Single gap in coverage	23.0%a (1.8)	11.6%a (1.3)	87.4%a (1.4)	62.2%a (2.0)
Temporary coverage	41.3%a (3.6)	39.8%a (3.6)	48.2%a (3.4)	43.5%a (3.2)
Frequent changes	46.1%a (2.6)	0% (0.0)	91.2% (1.4)	0.0 (0.0)
Repeatedly uninsured	48.4%a (1.6)	12.2%a (1.2)	81.6%a (1.4)	28.2%a (1.5)

Notes:

^aSignificant difference between children and adults (p < .05). Standard errors in parentheses.

By contrast, adults who were ever uninsured were most likely to gain or lose employer insurance. Indeed, the majority of adults with transitions into coverage, transitions out of coverage, or single gaps in coverage were covered only by employer insurance over the four years.

The general differences in four-year coverage patterns for uninsured adults and children discussed above (and shown in Table 3) were evident at both higher and lower income levels, as shown in Table 5. Within each income group, relatively fewer children were uninsured throughout four years. Uninsured children were more likely than adults to have frequent changes in coverage and single or repeated gaps in coverage, and to transition out of coverage for the remainder of four years. The only exception was that the higher percentage of children compared to adults who were repeatedly uninsured in 2004–2007 was evident only among individuals with incomes below 200 percent of the FPL (33.2 percent of children compared to 24.3 percent of adults, Table 5). At higher incomes, similar percentage of uninsured adults and children were repeatedly uninsured.

Table 5.

Percent Distribution of Adults and Children Ever Uninsured by Four-Year Coverage Patterns, According to Income, 1996–1999 and 2004–2007

	< 200% poverty		200+% poverty
	1996–1999	2004–2007	1996–1999	2004–2007
Ever-uninsured children	100.0%	100.0%	100.0%	100.0%
Always uninsured	7.5%b (0.8)	7.7%b (1.4)	5.0%b (1.0)	2.6%b (0.9)
Transition into coverage	7.5%ab (0.6)	8.0%ab (1.1)	10.0%b (1.0)	5.7%b 1.1)
Transition out of coverage	6.9%b (0.7)	17.9%b (1.7)	7.7%a (0.9)	16.6%a (1.6)
Single gap in coverage	17.4%b (1.1)	15.7%b (1.5)	32.7%b (1.5)	28.3%b (1.9)
Temporary coverage	4.4%b (0.6)	4.9%b (1.1)	2.7%b (0.6)	1.2% (0.3)
Frequent changes	11.1%b (0.8)	12.6%b (1.3)	12.9%a (1.1)	21.8%a (1.7)
Repeatedly uninsured	45.2%ab (1.3)	33.2%ab (1.9)	29.1% (1.7)	23.8% (2.1)
Ever-uninsured adults	100.0%	100.0%	100.0%	100.0%
Always uninsured	19.8%b (0.8)	23.3%b (1.2)	9.3%b (0.5)	11.0%b (0.9)
Transition into coverage	10.8%b (0.6)	13.6%b (0.9)	15.4%b (0.6)	16.3%b (0.9)
Transition out of coverage	10.3%b (0.6)	10.4%b (0.8)	8.2%a (0.5)	11.7%a (0.8)
Single gap in coverage	10.7%b (0.6)	8.6%b (0.6)	23.2%ab (0.8)	17.7%ab (1.0)
Temporary coverage	8.1%ab (0.6)	11.3%ab (1.0)	4.9%b (0.4)	6.3% (0.9)
Frequent changes	7.3%b (0.5)	8.4%b (0.7)	11.1% (0.5)	13.0% (0.7)
Repeatedly uninsured	33.0%ab (1.0)	24.3%ab (1.1)	27.8% (0.8)	24.0% (1.1)

Notes:

^aSignificant change for age group over time (p < .05).

^bSignificant difference between children and adults in the same time period (p < .05). Standard errors in parentheses.

CHANGES BETWEEN TIME PERIODS, BY AGE GROUP AND INCOME

The positive impact of CHIP was evident in the 8 percentage-point decline in the percentage of low-income children ever uninsured from 1996–1999 to 2004–2007, shown above in Table 1. Table 5 examines changes over time in four-year coverage patterns of the ever-uninsured separately for low-income children and other subgroups defined by age and income. The percentage of low-income children who were repeatedly uninsured was lower in 2004–2007 (33.2 percent, compared to 45.2 percent in 1996–1999). However, the percentage of low-income uninsured children who transitioned out of coverage and remained uninsured increased over time (from 6.9 percent to 17.9 percent). Thus, much of the reduction in the percentage repeatedly uninsured was offset by the increase in the percentage who lost coverage that they never regained. Eighty-eight percent of low-income children who transitioned out of coverage and remained uninsured in 2004–2007 were previously covered by Medicaid or CHIP; 29 percent were previously covered by employers. (Data on employer and Medicaid/CHIP coverage specifically for uninsured low-income children are not shown; the percentages of those transitioning out of coverage who had each type of insurance sum to more than 100 percent due to children moving between Medicaid/CHIP and employer insurance before becoming uninsured.)

As we noted earlier, at higher incomes at or above 200 percent of the FPL, a significantly higher percentage of children were uninsured for at least a month in 2004–2007 compared to 1996–1999. In the later period, relatively fewer uninsured children in higher-income families experienced patterns involving any transitions into coverage (i.e., one-time transitions into coverage, single gaps, and repeated gaps), and there was a large increase (from 7.7 percent to 16.6 percent) in the share of children who transitioned out of coverage for the rest of four years. At the same time, the fraction of higher-income children experiencing frequent changes in coverage jumped from 12.9 percent in 1996–1999 to 21.8 percent in 2004–2007.

Compared to the changes in coverage patterns experienced by children between the two time periods, changes in four-year coverage patterns were modest for ever-uninsured adults (Table 5). In both income groups, the percentage of ever-uninsured adults with repeated gaps in coverage declined from 1996–1999 to 2004–2007. Relatively fewer adults in the higher income group had single gaps in 2004–2007. Among higher-income adults, relatively more had transitions out of coverage in 2004–2007, and among low-income adults, relatively more were temporarily insured.

DISCUSSION

Large numbers of Americans continue to move in and out of coverage over time, even when the economy is growing, as it was in 2004–2007. Nearly twenty-three million people had more than one gap in health insurance coverage between 2004 and 2007, a powerful indication of the instability and insecurity that characterizes health insurance in the United States. Churning repeatedly in and out of coverage was particularly a problem for low-income children: three out of five were uninsured at some time in four years, and one out of five was repeatedly uninsured.⁴ Our updated estimates indicate that transitions in and out of Medicaid and CHIP have become a major factor in the coverage dynamics of uninsured children, as children’s overall enrollment in public insurance has increased.

Although low-income children remained at high risk of cycling in and out of insurance in 2004–2007, our analysis of changes in coverage dynamics suggests that the implementation of CHIP had a stabilizing effect on their health insurance. Not only did the percentage of low-income children continuously insured over four years increase by 8 percentage points, but the percentage of all low-income children who were repeatedly uninsured fell by 11 percentage points. Unfortunately, the percentage of all low-income children who transitioned out of coverage and remained uninsured for the rest of the four years also increased by 6 percentage points. Our conclusions regarding trends in the stability of children’s health insurance are consistent with those of Hill & Shaefer (2011), who found that public coverage for children became less stable in the late 1990s, after implementation of CHIP, and then private insurance coverage became less stable for children in the early 2000s. Consistent with their findings related to private health insurance of children, we found that higher-income children were less likely to be continuously insured in the mid-2000s.

Promoting stability and minimizing uninsured gaps should be high priorities as federal and state officials proceed with the implementation of national reforms. In addition to eliminating the financial risks associated with being uninsured, promoting continuity of coverage will facilitate continuity of care and reduce health insurance costs. If people are forced to move from one plan to another, even without becoming uninsured, they may have to change providers or pay more to stay with providers who are out of network in the new plan. They may also have to start new annual deductibles part way through the year. In addition, minimizing turnover can reduce the costs of both public and private insurers by eliminating some of the administrative expenses associated with the enrollment and disenrollment of plan members.

There is no question that avoiding gaps and awkward transitions in the new system created by the Affordable Care Act will be challenging. Families’ incomes are quite variable around the threshold that will divide eligibility for Medicaid from eligibility for subsidized insurance from the exchanges (133 percent of the poverty line), and many people are likely to transition between those two forms of assistance (Short, Swartz, & Uberoi, 2011; Sommers & Rosenbaum, 2011). Above the Medicaid eligibility limit, the tax credits that subsidize insurance from the exchanges will change if income or family size changes during the tax year. Taxpayers with significant increases in income could be required to pay back thousands of dollars for tax credits advanced during the year.⁵ Uncertainty about future income and the ultimate amount of premium subsidies could discourage people from using the credits to buy insurance. In addition, after the ACA is implemented, employment changes could force people to move between the exchanges and large-employer plans, as well as between individual and small-employer exchanges.

Policy analyses have suggested a variety of ways to address these challenges in the course of implementing national reforms (Short et al., 2011; Sommers & Rosenbaum, 2011; Bachrach, Boozang, & Dutton, 2011). Coordinated determination of eligibility for premium credits, Medicaid, and CHIP in each state will be particularly important. Making the same plans and provider networks available through the exchanges and public insurance programs also will help minimize changes in care providers. Transitional coverage from Medicaid and CHIP could be extended to the start of the next calendar year, when plan selections made during the exchanges’ annual open season go into effect. Premium tax credits could be made available for continuation coverage purchased from employers under COBRA, so there would be less need to fall back on the exchanges to fill short-term gaps.

With one-third of non-elderly Americans experiencing gaps in coverage over four years, many citizens have a personal stake in the success of national reforms. Even at high income levels during relatively good economic times, no one’s health insurance is entirely safe. Among people with incomes above 400 percent of the FPL, one out of ten adults had a gap in health insurance during 2004–2007, and the coverage of high-income children seemed to be eroding. The data presented here underscore the importance of designing national reforms to stabilize health insurance over time and will serve as part of the baseline for evaluating progress towards that goal.