SHA DAC S TATE H EALTH A CCESS A Health Data Resource for States March 2001/Issue 2 D ATA A SSISTANCE C ENTER Impact of Changes to the Current Population Survey (CPS) on State Health Insurance Coverage Estimation Congress recently provided specific To reduce the amount of sampling error and appropriations to the U.S. Census Bureau increase the precision of the estimates used in to improve the Current Population Survey the SCHIP formula, three-year rolling averages (CPS) for purposes of estimating health are used rather than single-year estimates. insurance coverage within states.1 The However, the three-year averages still have legislation specifically required the Census a large amount of sampling error, making it Bureau to increase the CPS sample size in difficult to precisely measure changes in the order to more accurately estimate the number number of uninsured children at the state of uninsured low-income children who are level. In order to remedy this problem, potentially eligible for the State Children’s Congress appropriated $10 million annually Health Insurance Program (SCHIP). This issue to produce more precise annual state estimates brief describes these changes, the time frame of the number of low-income children who for adoption, and the potential impact on do not have health insurance coverage. future CPS state coverage estimates. Increase in CPS Sample Size Background The Census Bureau plans to achieve more SCHIP funds are allocated to states using a precise estimates of the number of low-income formula that incorporates two factors: uninsured children in two ways: 1) by increasing 1) a “state cost” factor based on annual average the overall CPS sample size, and 2) by wages in the health services industry, and increasing the number of targeted CPS 2) a “number of children” factor based on respondents to the CPS March Supplement. the number of low-income children and the Through both strategies, the number of number of low-income, uninsured children in households that are sampled for the March a state. (Low-income children are defined as Supplement will increase from approximately those 18 years of age or younger living in 64,990 in 2000 to 98,990 in 2001. University of Minnesota families below 200 percent of the federal School of Public Health Table 1 shows the sample size for each state poverty level.)2 The data for calculating the for the March Supplement prior to and after “number of children” factor for each state Sponsored by a grant the planned sample expansion, the number comes from the CPS March Supplement, from The Robert Wood of additional households, the percent of the which includes health insurance questions. Johnson Foundation. sample increase within the state, and estimated percentage decrease in the standard error. The increased uninsurance for all states, with the greatest impact on CPS sample will be devoted disproportionately to states states with small and/or diverse populations (i.e., people where the CPS currently estimates uninsurance rates with with a broad range of socio-demographic characteristics the least precision (i.e., states with the largest variation live within the state). In general, the amount of error in the estimates.) This variation is a function of the size associated with a state estimate of coverage will of the state population, and its diversity in terms of decrease approximately 10-30 percent with the CPS demographic factors that are likely to influence health expansion. For example, the largest reduction in insurance characteristics. Most states will receive between error will occur in Connecticut from .80 to .49, a 38 one and two percent of the total increased sample size. percent reduction. The Census Bureau plans to administer the March The CPS sample expansion will allow states to Supplement to additional targeted CPS sample households calculate more accurate state level estimates of that are not in the March sample but are in the February uninsurance than in previous years. The sample and April samples. The targeted households include expansion will also result in less random variation those with a member of a minority group or with children. from year to year in state estimates of health insurance and uninsurance. Sample Expansion from Existing Primary Sampling Units Single year estimates based on the expanded sample will be available starting with 2001 data; three year The 2001 expansion of the CPS sample will be drawn rolling average estimates using the expanded sample from the existing CPS Primary Sampling Units (PSUs). will be possible when the 2003 data is available. The CPS divides states into PSUs composed of a metropolitan area, a large county, or a group of contiguous Summary smaller counties. Some of the PSUs are then randomly The CPS March supplement continues to be the primary selected for inclusion into the CPS sample. The goal of source of data to monitor health insurance coverage selecting PSUs is to ensure that the PSUs selected into across states. The CPS is getting more attention as the CPS sample are representative of the other PSUs that programs to increase access to health insurance coverage, are not selected. PSUs consisting of major metropolitan such as SCHIP, are being developed, implemented and areas are always included in the CPS sample. evaluated at the state level. In addition, data generated The CPS undergoes a major redesign every ten years by the CPS are used in the SCHIP funding formula and using data collected in the decennial census. The PSUs have a direct impact on the amount of funds states are evaluated to reflect changes in population, labor receive under this new federal-state program. force characteristics, and county Metropolitan Statistical Congress provided new funds to increase the sample Area (MSA) status. New PSUs will be added and some size for the CPS March Supplement in an effort to will be taken out of the CPS once the current redesign improve the accuracy of the state-level estimates of is complete in 2003. insurance coverage. The expansion of the CPS sample Impact of Sample Expansion on State size will result in more precise state estimates of health Health Insurance Coverage Estimates insurance coverage which in turn will provide the ability to observe more subtle changes in state coverage rates The increase in the sample size of the CPS will result in more precise state estimates of health insurance and year. While the increased state sample will not be The increase in the CPS sample is a positive step toward enough to do all the sub-state analysis of interest to increasing the accuracy of the CPS estimates of health states (e.g., coverage estimates by county, or by minority insurance coverage. The changes reduce the standard groups) it will provide a more accurate account of error by a considerable margin and should increase the insurance coverage rates from year to year. usefulness of the CPS for state and national policy makers. Table 1. Current Population Survey March Supplement sample size before and after planned expansion States Number of Number of CPS Sample Standard Error Estimated Estimated Households in Households in Increase for 1999 Annual Standard Error Percent Decrease CPS Before CPS After Within the Coverage for 2000 Annual in Standard Expansion Expansion State Estimate Estimate* Error Alabama 920 1600 74% 0.80 0.61 24% Alaska 900 1590 77% 0.90 0.68 25% Arizona 1100 1320 20% 0.90 0.82 9% Arkansas 930 1190 28% 0.80 0.71 12% California 5230 6570 26% 0.40 0.36 11% Colorado 930 1890 103% 0.80 0.56 30% Connecticut 630 1650 162% 0.80 0.49 38% DC 720 1270 76% 0.80 0.60 25% Delaware 810 1390 72% 0.90 0.69 24% Florida 3560 4320 21% 0.50 0.45 9% Georgia 1000 1390 39% 0.70 0.59 15% Hawaii 560 1320 136% 0.80 0.52 35% Idaho 1100 1370 25% 0.90 0.81 10% Illinois 2500 3220 29% 0.50 0.44 12% Indiana 850 1740 105% 0.70 0.49 30% Iowa 800 1500 88% 0.60 0.44 27% Kansas 900 1750 94% 0.80 0.57 28% Kentucky 830 1320 59% 0.80 0.63 21% Louisiana 870 1210 39% 0.90 0.76 15% Maine 780 1760 126% 0.80 0.53 33% Maryland 790 1720 118% 0.80 0.54 32% Massachusetts 1400 1740 24% 0.50 0.45 10% Michigan 2030 2700 33% 0.40 0.35 13% Minnesota 870 1710 97% 0.60 0.43 29% Mississippi 780 1040 33% 0.80 0.69 13% Missouri 800 1500 88% 0.70 0.51 27% Montana 920 1160 26% 0.90 0.80 11% Nebraska 820 1520 85% 0.70 0.51 27% Nevada 990 1790 81% 0.90 0.67 26% New Hampshire 670 1710 155% 0.80 0.50 37% Table continues on next page SHA DAC S TATE H EALTH A CCESS D ATA A SSISTANCE C ENTER Table 1. Current Population Survey March Supplement sample size before and after planned expansion (Continued from Page 3) States Number of Number of CPS Sample Standard Error Estimated Estimated Households in Households in Increase for 1999 Annual Standard Error Percent Decrease CPS Before CPS After Within the Coverage for 2000 Annual in Standard Expansion Expansion State Estimate Estimate* Error New Jersey 1840 2370 29% 0.50 0.44 12% New Mexico 1110 1320 19% 1.00 0.92 8% New York 4140 5240 27% 0.40 0.36 11% North Carolina 1690 2190 30% 0.60 0.53 12% North Dakota 890 1580 78% 0.80 0.60 25% Ohio 2200 2930 33% 0.40 0.35 13% Oklahoma 990 1440 45% 0.80 0.66 17% Oregon 790 1480 87% 0.80 0.58 27% Pennsylvania 2660 3380 27% 0.40 0.35 11% Rhode Island 640 1620 153% 0.70 0.44 37% South Carolina 720 1330 85% 0.90 0.66 26% South Dakota 880 1640 86% 0.70 0.51 27% Tennessee 860 1230 43% 0.70 0.59 16% Texas 3530 4360 24% 0.50 0.45 10% Utah 730 1090 49% 0.70 0.57 18% Vermont 770 1730 125% 0.80 0.53 33% Virginia 980 1640 67% 0.80 0.62 23% Washington 770 1670 117% 0.90 0.61 32% West Virginia 970 1490 54% 0.80 0.65 19% Wisconsin 910 1850 103% 0.70 0.49 30% Wyoming 930 1450 56% 0.90 0.72 20% Total 64,990 98,990 52% 0.10 0.08 19% Source: 3 US Census Bureau data available from SHADAC.3 * The 2001 standard errors were estimated by multiplying the 1999 standard errors by the square root of the old CPS sample size divided by the new CPS sample size. Notes 1 University of Minnesota P.L. 106-113, The Medicare, Medicaid, and SCHIP Balanced Budget Act of 1999. Division of Health Services 2 Health Care Financing Administration. “Final SCHIP Allotments to States, Commonwealths and Research and Policy Territories for Fiscal Year 2000” (HCFA-2067-N). Federal Register, May 24, 2000. Available at URL: www.hcfa.gov/init/children.htm. The number of children factor for FY 2000 uses 75% of the 2221 University Avenue low-income, uninsured children in a state and 25% of the low-income children in a state. For FY Suite 345 2001-2007, the formula uses 50% of the low-income, uninsured children in a state and 50% of the Minneapolis, MN 55414 low-income children in a state (HCFA, 2000). Phone 612-624-4802 3 Fax 612-624-1493 US Census Bureau. “U.S. Census Bureau Plans for Expanding the March Current Population Survey (CPS) Annual Demographic Supplement Sample.” Unpublished data available from SHADAC. www.shadac.org IB-02-0301