The 2005 and 2006 Current Population Survey (CPS) Annual Social and Economic Supplement (ASEC) data have been revised to improve the consistency of estimates for the insured and uninsured as part of ongoing efforts to improve the quality of Census Bureau data. The CPS asks about health insurance coverage in the previous year (for example, the 2006 survey asked about coverage in 2005).
Revised calendar-year coverage estimates for 2004 and 2005 reflect the results of an enhancement to the process that assigns coverage to dependents. The revision was necessary to better reflect the information that respondents were providing during the interview on health care coverage.
The instrument used to administer the Annual Social and Economic Supplement (ASEC) to the Current Population Survey (CPS) has been undergoing a conversion to a more modern operating system. Every question and question path was examined for accuracy and consistency.
During this process we found that, under certain circumstances, information provided by respondents was not fully recognized by the editing system. The questionnaire asks which household members had an insurance policy (either through an employer/union or a privately purchased plan) in their own name. If a plan is reported, questions then ask whether anyone else was covered by this plan, and if so, which other household members were covered.
The survey allows two ways to report that everyone else in their family or household was covered by a policy. Interviewers can either report, person by person, each other person that was covered or they could simply make an indication that “all” other household members were covered. In original form, the process always accepted respondents who reported each other person covered by a plan; it did not, however, recognize the “all other household members were covered” response. Instead, those cases were imputed coverage.
Effects of Imputation
In most cases, the imputations resulted in the same answers as if the “all other household members were covered” designation had been accepted, an accurate reflection of the household’s responses. However, in a small percentage of cases, people were imputed as “not covered” when in fact coverage had been reported for them.
Specifically, 3.7 percent of people for whom employer or union coverage was reported in the “all other household members covered” response were allocated as “not covered.” Similarly, 6.0 percent of people for whom privately purchased coverage was reported in the “all other household members covered” response were allocated as “not covered.”
New Process Improves Health Insurance Coverage Data
The new process allows us to produce more accurate coverage data. The effect was to reduce the uninsured rate by .6 percentage points for calendar-year 2005 and by a similar percentage in 2004. Tables 1 (2004) and 2 (2005) below show the results of the revision for various population characteristics.
In August 2006, when the Census Bureau first released its 2005 health insurance estimates, we reported that there was an increase in the percentage of persons without health insurance between 2004 and 2005, from 15.6 to 15.9 percent. As shown in tables 1 and 2, while the numbers of persons without health insurance are somewhat lower, the revised numbers still show a comparable increase in the uninsured rate, from 14.9 to 15.3 percent.
Results for calendar year 2006, which are scheduled for release in August 2007, will reflect this revision. At that time, the Census Bureau will release time series for 1995 to 2006 reflecting the more accurate health insurance data resulting from this improvement to the process.