Current Population Survey Estimates Research Articles

Errors in Survey Reporting and Imputation and Their Effects on Estimates of Food Stamp Program Participation

Benefit receipt in major household surveys is often underreported. This misreporting leads to biased estimates of the economic circumstances of disadvantaged populations, program takeup, and the distributional effects of government programs, and other program effects. We use administrative data on Food Stamp Program (FSP) participation matched to American Community Survey (ACS) and Current Population Survey (CPS) household data. We show that nearly thirty-five percent of true recipient households do not report receipt in the ACS and fifty percent do not report receipt in the CPS. Misreporting, both false negatives and false positives, varies with individual characteristics, leading to complicated biases in FSP analyses. We then directly examine the determinants of program receipt using our combined administrative and survey data. The combined data allow us to examine accurate participation using individual characteristics missing in administrative data. Our results differ from conventional estimates using only survey data, as such estimates understate participation by single parents, non-whites, low income households, and other groups. To evaluate the use of Census Bureau imputed ACS and CPS data, we also examine whether our estimates using survey data alone are closer to those using the accurate combined data when imputed survey observations are excluded. Interestingly, excluding the imputed observations leads to worse ACS estimates, but has less effect on the CPS estimates.

Regional Views on the Role of Immigrant Labor on U.S. and Southern Dairies

Although immigration is a controversial issue, it is also of major importance to the United States and to agriculture. Immigration policy has been discussed in the recent presidential debates and will likely be debated again in Congress at some point in the near future. Agricultural producer organizations, commodity associations, and lobbying groups have been at the forefront of this issue for many years. Our profession certainly has a role to play by informing constituent groups and the public with objective analytical results. Immigrant labor is an important component of many, if not most, U.S. agricultural enterprises. The Current Population Survey estimates that in 2010, 57.2% of the agricultural hired labor force was foreign-born (U.S. Census Bureau, 2010). Approximately 62% of those foreign employees worked in crop production, while the remainder worked in livestock operations (U.S. Department of Agriculture—Economic Research Service, 2012). Recent evidence suggests the role of immigrant labor on U.S. dairies is significant. In Wisconsin, for example, it was estimated that immigrant labor accounted for 40% of the dairy farm workforce in 2008, compared with almost no immigrant labor before 2000 (Harrison, Lloyd, and O’Kane, 2009). Immigrant labor is particularly important to dairies in the Southwest. Our results indicate that those farms employ many more foreign workers than farms in other regions of the country. With increased enforcement of U.S. immigration regulations, lack of congressional action to resolve immigration policy issues, and economic recovery of the U.S. economy, concerns have been raised about the prospects of reduced labor availability in many sectors of agriculture. Dairy farming is no exception.

Datapoints: Reported Mental Health Specialty Care in the 2010 HealthStyles Survey

O measure of progress in reducing negative attitudes toward people with mental illness is the extent to which members of the general public who report having a mental health condition seek specialist care. To examine this issue more closely, we analyzed adult responses to the 2010 edition of the HealthStyles Survey. The survey is a large, private proprietary, national mail panel marketing survey conducted annually to assess attitudes and beliefs about chronic and infectious disease and behaviors, exposure to health information and to health communication campaigns, and self-reported symptoms, risk factors, diseases, and disorders (1). The survey is designed to be representative of the U.S. population, and the 2010 survey data were weighted to match the 2009 Current Population Survey estimates for age, sex, marital status, education, and race-ethnicity. The 2010 survey was mailed to 6,255 respondents; the response rate was 67% (4,184 individuals age 18 and over). Most respondents agreed with statements that a person with mental illness would improve if given treatment and support (68%) and that treatment can help people with mental illness lead a normal life (65%). A subset of respondents (24%) self-reported having one or more of eight listed mental health conditions in the past year. Among those with a self-reported mental health condition, approximately 42% reported having more than one mental health condition. For those with either a single condition or multiple conditions, depression and anxiety were the most prevalent, accounting for almost 80% of all conditions in both groups. Figure 1 presents data about respondents with one or more self-reported mental health conditions who reported visiting a mental health specialty provider in the previous year and the extent of these visits. Only 17% of those with one self-reported condition and 46% of those with multiple self-reported conditions saw a mental health provider during the previous year. Further, among those who visited a mental health provider, most made fewer than six visits. Among those with a single condition who sought treatment, 79% reported six or fewer visits. Among those with multiple conditions, 58% reported six or fewer visits. Reported Mental Health Specialty Care in the 2010 HealthStyles Survey

Errors in Survey Reporting and Imputation and Their Effects on Estimates of Food Stamp Program Participation

Benefit receipt in major household surveys is often underreported. This misreporting leads to biased estimates of the economic circumstances of disadvantaged populations, program takeup, and the distributional effects of government programs, and other program effects. We use administrative data on Food Stamp Program (FSP) participation matched to American Community Survey (ACS) and Current Population Survey (CPS) household data. We show that nearly thirty-five percent of true recipient households do not report receipt in the ACS and fifty percent do not report receipt in the CPS. Misreporting, both false negatives and false positives, varies with individual characteristics, leading to complicated biases in FSP analyses. We then directly examine the determinants of program receipt using our combined administrative and survey data. The combined data allow us to examine accurate participation using individual characteristics missing in administrative data. Our results differ from conventional estimates using only survey data, as such estimates understate participation by single parents, non-whites, low income households, and other groups. To evaluate the use of Census Bureau imputed ACS and CPS data, we also examine whether our estimates using survey data alone are closer to those using the accurate combined data when imputed survey observations are excluded. Interestingly, excluding the imputed observations leads to worse ACS estimates, but has less effect on the CPS estimates.

Comparison of Physician Workforce Estimates and Supply Projections

Estimates of physician supply in the United States have been based on data that may overestimate the number of older physicians in the workforce. To compare physician workforce estimates and supply projections using the American Medical Association Physician Masterfile (Masterfile) data with estimates and projections using data from the US Census Bureau Current Population Survey (CPS). Parallel retrospective cohort analyses of employment trends of the number of active physicians by age and sex using annual data from the Masterfile and the CPS between 1979 and 2008. Recent workforce trends were used to project future physician supply by age. Annual number of physicians working at least 20 hours per week in 10-year age categories. In an average year in the sample period, the CPS estimated 67,000 (10%) fewer active physicians than did the Masterfile (95% confidence interval [CI], 57,000-78,000; P < .001), almost entirely due to fewer active physicians aged 55 years or older. The CPS estimated more young physicians (ages 25-34 years) than did the Masterfile, with the difference increasing to an average of 17,000 (12%) during the final 15 years (95% CI, 13,000-22,000; P < .001). The CPS estimates of more young physicians were consistent with historical growth observed in the number of first-year residents, and the CPS estimates of fewer older physicians were consistent with lower Medicare billing by older physicians. Projections based on both the CPS and the Masterfile data indicate that the number of active physicians will increase by approximately 20% between 2005 and 2020. However, projections for 2020 using CPS data estimate nearly 100,000 (9%) fewer active physicians than projections using the Masterfile data (957,000 vs 1,050,000), and estimate that a smaller proportion of active physicians will be 65 years or older (9% vs 18%). The increasing proportion of female physicians had little effect on physician supply projections because, unlike male physicians, female physicians were found to maintain their work activity after age 55 years. Compared with the Masterfile data, estimates using the CPS data found more young physicians entering the workforce and fewer older physicians remaining active, resulting in estimates of a smaller and younger physician workforce now and in the future.

Does the Current Population Survey Overstate the Number of Uninsured Americans

Several researchers have argued that the Current Population Survey (CPS) undercounts the number of individuals on public insurance and as a result overstates the number of individuals who are uninsured. This paper compares survey estimates from the CPS to estimates from the Medical Expenditures Panel (MEPS). The CPS estimates for both the size of the Medicaid and uninsured populations are above the full-year MEPS estimate but below the at-least-part-year MEPS estimate. The CPS does not appear to overstate the number of uninsured because CPS uninsured estimates are only slightly higher than MEPS full-year uninsured estimates and MEPS data clearly demonstrate that many people are uninsured for only part of the year.

Estimates Of Health Insurance Coverage: Comparing State Surveys With The Current Population Survey

The Census Bureau produces annual state-level estimates of health insurance coverage using the Current Population Survey (CPS) Annual Social and Economic Supplement. Many states also conduct their own population surveys of health insurance status; in most cases, the state survey estimates of uninsurance are lower than the estimates produced by the CPS. This discrepancy fuels debate about the true count of uninsured Americans and changes in that number over time. This paper compares state survey and CPS estimates of uninsurance, highlights key reasons for these differences, and discusses the policy implications of this persistent discrepancy.

The Contribution of Medicaid Managed Care to the Increasing Undercount of Medicaid Beneficiaries in the Current Population Survey

The Current Population Survey (CPS) is an important source of data for comparing beneficiaries across insurance groups. However, the CPS routinely underestimates the Medicaid population, and for unexplained reasons these underestimates have been increasing over time. We sought to determine whether the penetration of Medicaid managed care is associated with the magnitude of the underestimate of Medicaid beneficiaries in the CPS using a pooled cross-sectional comparison of survey and administrative databases on all California Medicaid beneficiaries younger than 65 years of age. We compared the CPS estimates of Medicaid beneficiaries in California from 1995 to 1999 with the gold-standard number derived from the Medicaid eligibility file for the same time period and examined the association between the CPS underestimate and penetration of managed care in the beneficiary's county. The CPS underestimated the Medicaid population by approximately a third. At the county level, errors in estimated numbers of Medicaid beneficiaries in the CPS increased in association with the penetration of Medicaid managed care. Each percentage point increase in the penetration of managed care was associated with an underestimate in the CPS of 0.4 percentage points. A substantial portion of the increase in the underestimates of the number of Medicaid beneficiaries in the CPS can be explained by the growth of Medicaid managed care. Steps must be taken to improve the CPS if this survey is to remain useful for making accurate estimates of Americans' health insurance status.

Under-Reporting of Medicaid and Welfare in the Current Population Survey

Conventional estimates of the number of uninsured Californians are derived from the Current Population Survey (CPS). Unfortunately, CPS estimates of the number of people receiving Medi-Cal and welfare (AFDC/CalWORKs) are well below the numbers implied by official Medi-Cal records, suggesting that the conventional estimates of the number of uninsured Californians (and their characteristics) are seriously flawed. To improve understanding of these issues, the California HealthCare Foundation (through its then separate the Medi-Cal Policy Institute-MCPI) and the U.S. Department of Health and Human Services, Administration for Children and Families (DHHS-ACF) funded RAND to match CPS data to individual-level administrative data for the Medi-Cal program. With the cooperation of the California Department of Health Services (CDHS), the U.S. Bureau of the Census, and the California Census Research Data Center (CCRDC), that match was performed. This document describes the findings of the analysis of those matched data.

Empirical Bayes estimation of median income of four-person families by state using time series and cross-sectional data

The Department of Health and Human Services uses estimates of the median income of four-person families for all the fifty states and the District of Columbia to formulate its energy assistance program for low income families. Such estimates are provided by the US Census Bureau on an annual basis. A hierarchical time series model is considered to combine information from three relevant sources: (a) Current Population Survey (CPS), (b) Decennial Censuses and (c) Bureau of Economic Analysis. An empirical Bayes (EB) method is used to smooth the CPS estimates of the median income of four-person families for the states. The proposed method is an advancement over the EB method currently used by the US Bureau of the Census in the sense that it uses a more realistic model, provides maximum likelihood and residual maximum likelihood method of variance components estimation and provides a valid measure of uncertainty of the proposed estimates which captures all different sources of variations. Compared to the corresponding hierarchical Bayes estimation, the method is very easy to implement and saves a tremendous amount of computer time. The proposed EB method is compared with rival estimators using the 1989 four-person median income figures obtained from the 1990 Census.

Hierarchical Bayes Estimation of Unemployment Rates for the States of the U.S.

Abstract Under a federal-state cooperative program, the U.S. Bureau of Labor Statistics (BLS) publishes monthly unemployment rate estimates for its 50 states and the District of Columbia. The primary source of data for this estimation problem is the Current Population Survey (CPS). However, the CPS state unemployment rate estimates are unreliable, because the survey provides relatively few observations per state. Various federal agencies use state-level unemployment rate estimates for policy making and fund allocation. Thus it is important to improve on the CPS estimates. For this, we propose a hierarchical Bayes (HB) method using a time series generalization of a widely used cross-sectional model in small-area estimation. The proposed method is compared in detail with the corresponding HB method, which uses the HB analog of the well-known Fay-Herriot cross-sectional model. A third model based on a time series approach to repetitive surveys is found to be very hard to implement for these data; the resulting estimates are very unstable and not meaningful. If we ignore some important factors from this model, then the reduced model can be fit, but the resulting model is found to be less than adequate. Gibbs sampling is used to obtain the posterior means and variances of the state unemployment rates. Based on some diagnostic tools recently developed for hierarchical models, our proposed model emerges as the best. The coefficients of variation of the proposed HB estimates are considerably lower than those of the rival estimates.

Hierarchical Bayes Estimation of Unemployment Rates for the States of the U.S.

Under a federal-state cooperative program, the U.S. Bureau of Labor Statistics (BLS) publishes monthly unemployment rate estimates for its 50 states and the District of Columbia. The primary source of data for this estimation problem is the Current Population Survey (CPS). However, the CPS state unemployment rate estimates are unreliable, because the survey provides relatively few observations per state. Various federal agencies use state-level unemployment rate estimates for policy making and fund allocation. Thus it is important to improve on the CPS estimates. For this, we propose a hierarchical Bayes (HB) method using a time series generalization of a widely used cross-sectional model in small-area estimation. The proposed method is compared in detail with the corresponding HB method, which uses the HB analog of the well-known Fay-Herriot cross-sectional model. A third model based on a time series approach to repetitive surveys is found to be very hard to implement for these data; the resulting estimates are very unstable and not meaningful. If we ignore some important factors from this model, then the reduced model can be fit, but the resulting model is found to be less than adequate. Gibbs sampling is used to obtain the posterior means and variances of the state unemployment rates. Based on some diagnostic tools recently developed for hierarchical models, our proposed model emerges as the best. The coefficients of variation of the proposed HB estimates are considerably lower than those of the rival estimates.

Assessing the devolution revolution: how accurate are state-level estimates from the Current Population Survey?

The devolution of many social policy responsibilities from the Federal government to states has prompted increased interest in state-level measures of need. One data source that could be used to provide more state-level information on a variety of topics is the Current Population Survey (CPS). During the past ten years the CPS has been used to produce state- level estimates on a variety of measures. However, there has been little systematic evaluation of these data. This paper provides measures of accuracy for several state-level estimates derived from the CPS. These include standard errors for single-year estimates, three-year averages, and five-year averages of the March CPS measures; standard errors for three-year averages of 12-month CPS files; and comparison of CPS-based estimates to data from the Decennial Census. The paper also examines the relative accuracy of CPS estimates based on states' size. The information in this study will help analysts better understand the tradeoffs between timeliness and accuracy to be considered when using state-level estimates derived from the CPS.

State smoking prevalence estimates: a comparison of the Behavioral Risk Factor Surveillance System and current population surveys.

This study examined whether there are systematic differences between the Behavioral Risk Factor Surveillance System (BRFSS) and the Current Population Survey (CPS) for state cigarette smoking prevalence estimates. BRFSS telephone survey estimates were compared with estimates from the US Census CPS tobacco-use supplements (the CPS sample frame includes persons in households without telephones). Weighted overall and sex- and race-specific BRFSS and CPS state estimates of adults smoking were analyzed for 1985, 1989, and 1992/1993. Overall estimates of smoking prevalence from the BRFSS were slightly lower than estimates from CPS (median difference: -2.0 percentage points in 1985, -0.7 in 1989, and -1.9 in 1992/1993; P < .05 for all comparisons), but there was variation among states. Differences between BRFSS and CPS estimates were larger among men than among women and larger among Blacks than among Hispanics or Whites; for most states, these differences were not significant. The BRFSS generally provides state estimates of smoking prevalence similar to those obtained from CPS, and these are appropriate for ongoing state surveillance of smoking prevalence.

Data Watch: The Redesigned Current Population Survey

The Current Population Survey (CPS), a national survey of 50,000 households, is a major source of information about the American labor market. In January 1994, the CPS underwent a major redesign both in the wording of the questionnaire and the methodology used to collect the data. This article reviews the motivation for the redesign, compares several key CPS estimates before and after the implementation of the new survey, and explains some of the new data collected in the redesigned CPS.

Assessing post-census state poverty estimates

While the decennial census provides poverty figures for states and other subnational geographic units every ten years, their utility declines over the course of a decade. Consequently, there is growing interest in producing post-census estimates for a variety of indicators. This study extends recent efforts to estimate post-census poverty figures for states by producing such estimates using a multiple regression approach. The accuracy of the multiple regression estimates along with recently produced estimates from the Current Population Survey (CPS) are evaluated relative to the decennial census. The mean absolute percentage point error (MAPPE) using the ratio-correlation technique (1.56 percentage points) was somewhat higher than the MAPPE of 1989 CPS (1.37 percentage points) and an average of 1988–1990 CPS data (1.15 percentage points). However, a simple regression technique using data from 1979 to estimate poverty in 1989 produced a set of estimates where the MAPPE (1.37 percentage points) is nearly as accurate as the single-year CPS estimates. Estimates which average regression estimates and CPS-based estimates are more accurate than either regression or CPS estimates used alone. Several suggestions are offered for improving regression estimates.

How Accurate are Estimates of State and Local Unemployment?

Local and area unemployment has increased interest in and the importance of state and local unemployment estimates. This study examines the accuracy of such estimates by comparing those provided by the 1960 Census of Population with those of the Bureau of Employment Security of the United States Department of Labor (for April 1960). Additional comparisons are also made with national and regional Current Population Survey estimates for the same month. The study shows that there was considerable divergence between the BES and Census estimates of state and local unemployment, thus raising the question of the adequacy of these estimates for purposes of resting policy decisions on them. (Author's abstract courtesy EBSCO.)