A picture of four small children in a row
target imageTarget Population image AHRQ User Liaison Program
  Skip Navigation | Home Target Population  National Data for State Est. Enhancing CPS

previous page Previous  -  Nextnext page
top of nav bar space

Target Population Link Outreach Enrollment link Benefit Design and Service Delivery link Cost Sharing Link Crowd Out Link Monitoring and Evaluation link
Space Saver Image Site map link link legend link questions and comments acronyms list workshop materials link how to use this site link

Enhancing CPS Data to Obtain More Accurate State Estimates

Although there are significant limitations in the utilization of national data to obtain State-specific information, there are methods to enhance the data to obtain more accurate State-specific estimates. James Verdier suggests two methods for statistically enhancing Current Population Survey (CPS) data in order to increase the reliability of State-level estimates:

  • Reduce the margin of error by increasing sample sizes. Utilizing 3-year averages can help to smooth out year-to-year variations in annual survey results, but significant margins of error may remain. For example, based on 3-year averages from the October 1997 CPS data, the total number of uninsured in California ranged from 1,466,000 to 1,886,000 and the total number in Delaware ranged from 12,000 to 32,000.
    • Pat Purcell (1997) utilizes the 3-year average technique - Purcell, Patrick. "Health Insurance: Uninsured Children by State, 1994-1996." October 3, 1997. Congressional Research Service.
  • Increase the sample size in one State by "borrowing" similar households from other States to increase the reliability of information regarding subpopulations within the national sample. Several rules should be followed when employing this method:
    • Borrowed households must be similar with respect to age, race, income, family type, and employment status.
    • Relationships between employment characteristics and insurance coverage must be similar in the borrowing and lending States. For example, insurance coverage in a State with small employers and non-unionized employees is very different from coverage in a State that has mostly large heavily unionized employers.

Mr. Verdier cautioned that while statistical enhancements of the CPS can provide better information for program planning and design, and provide better estimates of year-to-year changes in number of low-income uninsured children than unenhanced data, it involves complex analyses and the large margins of error and methodological limitations of the CPS will remain.

Conclusions Regarding the Use of CPS

  • The annual March Current Population Survey (CPS) is not reliable for generating State-specific estimates.
  • State-specific surveys may be more reliable than CPS, but will not be comparable across all States.
  • Statistically enhancing CPS data can produce more reliable State estimates, but probably not reliable enough to track the impact of SCHIP on the number of uninsured children. The precise State-by-State impact of SCHIP on the number of uninsured children will likely remain unknown.
  • Implications for States
    • Review and utilize CPS, Urban Institute, the Community Tracking Survey, and other surveys.
    • Carefully consider what accountability requirements exist Federally and within your own State when planning for data collection.

Related Questions

Show me other SCHIP Workshop presenter-suggested strategies

What other issues are talked about in this section?

What led to this?

What other methodological issues are discussed?

Who presented this material?

arrow up