Adults who Remained Uninsured at the End of 2014
2014 Kaiser Survey of Low-Income Americans and the ACA: Methods
The 2014 Kaiser Survey of Low-Income Americans and the ACA, conducted by the Kaiser Family Foundation (KFF) in Fall 2014, examines health insurance coverage, health care use and barriers to care, and financial security among insured and uninsured adults across the income spectrum, with a focus on populations targeted for coverage expansions under the Affordable Care Act (ACA). The survey captures experiences a year after open enrollment began for health coverage through the ACA and builds on a similar baseline survey conducted in summer 2013.1 The 2014 survey includes a national sample as well as two state-specific samples in California (conducted with support from the Blue Shield of California Foundation (BSCF)) and Missouri (conducted with support from the Missouri Foundation for Health (MFH)). Costs for the national survey were paid by KFF.
The survey was designed and analyzed by researchers at KFF, with feedback on the California and Missouri state-specific components from BSCF and MFH, respectively. Social Science Research Solutions (SSRS) collaborated with KFF researchers on sample design and weighting; SSRS also supervised the fieldwork.
The survey was conducted by telephone from September 2 through December 15, 2014 from representative random samples of California and Missouri residents between the ages of 19-64, along with respondents from the remaining 48 states and the District of Columbia. In total, 10,502 interviews were completed; of these, 4,555 were with respondents living in California, 1,864 with respondents in Missouri, and 4,083 with respondents from other states. Computer-assisted telephone interviews (CATI) conducted by landline (5,105) and cell phone (5,397) were carried out in English and Spanish by SSRS.
The study was designed to oversample lower- and middle-income populations in order to have more statistical power in describing the views and experiences of these groups. To efficiently reach lower-income respondents, the sample in each state was stratified based on the estimated income level of geographic areas within California and Missouri as well as within the remaining states. This process was done separately for the landline and cell phone sampling frames. For the landline sample, strata were defined based on the median income within telephone exchanges; for the cell phone sample, strata were defined based on the household income associated with the billing rate-center to which the cell phone number is linked. The exact criteria for distinguishing between the strata in the cell phone sample varied from state to state. In addition, 481 interviews (217 on landline and 264 on cell phone) were conducted with respondents who were previously interviewed by SSRS as part of omnibus surveys of the general public and indicated they were ages 19-64, resided in the appropriate geography for the sample (if part of one of the state samples), and reported annual income of less than $25,000. These previous surveys were conducted with nationally representative, random-digit-dial landline and cell phone samples. Landline and cell phone samples were provided by Marketing Systems Group.
Screening for the survey involved verifying that the respondent (or another member of the household for the landline sample) met the criteria of: 1) being 19-64 years old; and 2) providing income information that allowed them to be classified by family income. People who did not know or refused to report their income or family size were excluded from the survey. Respondents were classified by family income as a share of the federal poverty level (FPL) based on their family size and total annual gross income.2 Poverty level groups included income < 138% of FPL (the income range for the Medicaid expansion), income of 139-400% FPL (the income range for Marketplace tax credits), and income above 400% of FPL (eligible only for unsubsidized coverage). For the landline sample, if two or more people met the criteria, a respondent was randomly selected by the CATI program. Selected respondents were asked to confirm their state of residence.
A multi-stage weighting approach was applied to ensure an accurate representation of the various income groups ages 19 to 64. The weighting process involved corrections for sample design as well as sample weighting to match known demographics of the target populations in order to correct for systematic non-response along these parameters. The base weight accounted for the oversamples used in the sample design, as well as the likelihood of non-response for the re-contact sample, number of eligible household members for the landline sample, and a correction to account for the fact that respondents with both a landline and cell phone have a higher probability of selection. Demographic weighting parameters were based on population estimates for the 19-64 year old poverty-level population in each state based on the U.S. Census Bureau’s 2013 American Community Survey (ACS). The weighting parameters for each poverty-level group within the two state-specific samples and the remaining national sample were: age, education, race/ethnicity, presence of own child in the household, marital status, and region. All statistical tests of significance account for the effect of weighting.
The margin of sampling error (including the design effect) for national estimates, state estimates and state-by-poverty-level estimates are shown in Table A. For the national sample, the margin of sampling error is plus or minus 2 percentage points for both the low- and moderate-income groups. For the remaining uninsured, the margin of sampling error is plus or minus 4 percentage points. For results based on other subgroups, the margin of sampling error may be higher. Sample sizes and margin of sampling errors for other subgroups are available by request. In reporting results, any estimate with a relative standard error (standard error divided by the point estimate) greater than 30 percent or based on a sample less than 100 is considered unreliable and not reported. Note that sampling error is only one of many potential sources of error in this or any other survey.
In analyzing results, we often categorize respondents according to insurance coverage or eligibility for insurance coverage. We classify anyone who indicated that they did not have any form of health insurance or health plan at the time of the interview as “Uninsured.” People who indicate coverage through their own employer, a spouse’s employer, or a parent’s employer are classified as having employer coverage. People who indicate Medicaid coverage, either alone or in conjunction with Medicare, are classified as having Medicaid. People who indicate that they purchased their coverage through their state Marketplace or healthcare.gov are classified as having Marketplace coverage. People who indicate that they purchase their coverage directly from an insurance company but did not purchase coverage through their state Marketplace or healthcare.gov are classified as “Private Nongroup.” People with other sources of coverage, including Medicare, VA, school-based coverage, or an unidentifiable source are classified as “Other.” In asking about both Medicaid and Marketplace coverage, state-specific program names were used, corresponding to the respondent’s state of residence. In some cases, we recoded coverage type based on verbatim responses, other information in the survey, or call backs to confirm type of coverage.
We use information on when coverage began to classify people into categories of “newly insured” or “continuously insured.” Newly insured individuals are those who indicate that they have insurance coverage, that their coverage started on or after January 2014, and that they were uninsured before that coverage started. Continuously insured people are those who indicate that they have insurance coverage and had insurance coverage since before January 2014.
Last, we assess uninsured respondents’ likely eligibility for coverage under the ACA based on family income as a share of poverty, state of residence, immigration status and length of time in the United States, parent status, and availability of coverage through an employer. We define undocumented immigrants as those who reported 1) they were born outside the United States, 2) are not a citizen, 3) did not have a green card when they arrived in the United States, and 4) have not received a green card or become a permanent resident since arriving. This measure may be subject to error in several ways. First, it relies on self-reporting, and respondents have an incentive not to reveal unlawful immigration status. Second, those that did not answer all questions in the series of immigration status items (84 respondents) were not able to be categorized as undocumented and were therefore included; if they are in fact undocumented, then the results may differ slightly. Third, a small number of people may have a legal status besides permanent residency or green card (such as refugees, asylees or other humanitarian immigrants). Unfortunately, due to time constraints, the survey was not able to fully explore all of these immigration pathways.
|Table A: Number of Respondents and Margin of Sampling Error for National and State-Specific Samples|
|N||Margin of Sampling Error|
|U.S. Total||10,502||+/- 2 percentage points|
|U.S. ≤ 138% FPL||4,295||+/- 3 percentage points|
|U.S. 139%-400% FPL||4,826||+/- 3percentage points|
|U.S. >400%||1,381||+/- 4 percentage points|
|California Total||4,555||+/- 2 percentage points|
|CA ≤ 138% FPL||2,044||+/- 3 percentage points|
|CA 139% – 400% FPL||2,003||+/- 3 percentage points|
|CA >400%||508||+/- 5 percentage points|
|Missouri Total||1,864||+/- 3 percentage points|
|MO ≤ 138% FPL||751||+/- 5 percentage points|
|MO 139% – 400% FPL||801||+/- 5 percentage points|
|MO >400%||312||+/- 7 percentage points|