Moving from a decennial census to a continuous measurement survey: factors affecting nonresponse at the neighborhood level

The Census Bureau is testing a continuous measurement program, known as the American Community Survey (ACS), which will provide census “long form” data annually, though with slightly higher levels of sampling variability. This paper focuses on the 1999–2001 ACS in the Bronx, 1 of 31 ACS test sites. It examines whether the quality of ACS data in the Bronx varies across neighborhoods, focusing specifically on how neighborhood sociodemographic factors influence nonresponse, as measured by mail return and allocation rates. It also examines whether these neighborhood factors have a differential impact on nonresponse in the ACS and the 2000 decennial census, and discusses reasons why this may be so.The ACS mail return rates are not only lower than those of the census, but are highly sensitive to race and socioeconomic distress. Despite this initial disadvantage, the ACS has lower levels of allocation on key variables, relative to the 2000 Census. Moreover, the effect of neighborhood socioeconomic distress on allocation rates in the ACS was minimal, compared to its effect on census allocation. We find that the overall quality of ACS data in the Bronx is superior to that of the decennial census. Our analysis of Bronx data suggests that the proposed elimination of the decennial long form and its replacement with the ACS is a reasonable tradeoff for users of small area data.     

An Application of Bayesian Methods to Small Area Poverty Rate Estimates

Efforts to estimate various sociodemographic variables in small geographic areas are proving difficult with the replacement of the Census long-form with the American Community Survey (ACS). Researchers interested in subnational demographic processes have previously relied on Census 2000 long-form data products in order to answer research questions. ACS data products promise to begin providing up-to-date profiles of the nation’s population and economy; however, unit- and item-level nonresponse in the ACS have left researchers with gaps in subnational coverage resulting in unstable and unreliable estimates for basic demographic measures. Borrowing information from neighboring areas and across time with a spatiotemporal smoothing process based on Bayesian statistical methods, it is possible to generate more stable and accurate estimates of rates for geographic areas not represented in the ACS. This research evaluates this spatiotemporal smoothing process in its ability to derive estimates of poverty rates at the county level for the contiguous United States. These estimates are then compared to more traditional estimates produced by the US Census Bureau, and comparisons between the two methods of estimation are carried out to evaluate the practical application of this smoothing method. Our findings suggest that by using available data from the ACS only, we are able to recreate temporal and spatial patterns of poverty in US counties even in years where data are sparse. Results show that the Bayesian methodology strongly agrees with the estimates produced by the SAIPE program, even in years with little data. This methodology can be expanded to other demographic and socioeconomic data with ease.     

An Evaluation of Persons per Household (PPH) Estimates Generated by the American Community Survey: A Demographic Perspective

The American Community Survey (ACS) is a U.S. Census Bureau product designed to provide accurate and timely demographic and economic indicators on an annual basis for both large and small geographic areas within the United States. Operational plans call for ACS to serve not only as a substitute for the decennial census long-form, but as a means of providing annual data at the national, state, county, and subcounty levels. In addition to being highly ambitious, this approach represents a major change in how data are collected and interpreted. Two of the major questions facing the ACS are its functionality and usability. This paper explores the latter of these two questions by examining “persons per household (PPH),” a variable of high interest to demographers and others preparing regular post-censal population estimates. The data used in this exploration are taken from 18 of the counties that formed the set of 1999 ACS test sites. The examination proceeds by first comparing 1-year ACS PPH estimates to Census 2010 PPH values along with extrapolated estimates generated using a geometric model based on PPH change between the 1990 and 2000 census counts. Both sets of estimates are then compared to annual 2001–2009 PPH interpolated estimates generated by a geometric model based on PPH from the 2000 census to the 2010 census. The ACS PPH estimates represent what could be called the “statistical perspective” because variations in the estimates of specific variables over time and space are viewed largely by statisticians with an eye toward sample error. The model-based PPH estimates represent a “demographic perspective” because PPH estimates are largely viewed by demographers as varying systematically and changing relatively slowly over time, an orientation stemming from theory and empirical evidence that PPH estimates respond to demographic and related determinants. The comparisons suggest that the ACS PPH estimates exhibit too much “noisy” variation for a given area over time to be usable by demographers and others preparing post-censal population estimates. These findings should be confirmed through further analysis and suggestions are provided for the directions this research could take. We conclude by noting that the statistical and demographic perspectives are not incompatible and that one of the aims of our paper is to encourage the U.S. Census Bureau to consider ways to improve the usability of the 1-year ACS PPH estimates.     

American Community Survey overview and the role of external evaluations

In 2010 the U.S. Census Bureau will achieve its goal of eliminating the long form sample from the decennial census and will produce its first set of five-year data products from the full sample American Community Survey (ACS). This paper provides an overview of the call for change that prompted the Census Bureau to pursue the development of a new approach to collecting socioeconomic and housing data. The paper details the evolution of the ACS from its earliest origins to its current design and describes that design in detail. The current design has benefited from external debate and consultation. Work such as that described later in this journal exemplifies the key role that external users and advisors have played, and will continue to play, in the evolution of the ACS. Over the past 10 years, the Census Bureau has undertaken research and testing to demonstrate operational feasibility and to assess survey quality. Research has also compared ACS and Census 2000 data. ACS staff are involved in survey improvement efforts and continue to confront survey challenges. In the next few years the ACS will give priority to developing user tools to aid all users in the correct interpretation of multi-year estimates. The ultimate validation of the ACS is, however, in the hands of users. Continued input from the people who are responsible for administering and evaluating programs, identifying local needs, and planning for the future will allow the ACS to grow in value and utility.     

An evaluation of the American Community Survey: results from the Oregon test site

The American Community Survey (ACS) is a Census Bureau product designed to provide accurate and timely demographic and economic indicators on an annual basis for both large and small geographic areas within the United States. Operational plans for Census 2010 call for ACS to replace the decennial census long form (Census LF), pending the results of evaluation studies. This plan represents a major change in that variables that traditionally have been collected on a “snapshot” basis once every 10 years would be collected on a “rolling” annual basis. Using a loss function analysis and other tools, this paper reports preliminary findings from a comparison of ACS and Census 2000 results in Multnomah County, Oregon, one of five national “local expert” test sites set up to compare ACS data collected at the time of Census 2000. The preliminary findings suggest that there are notable differences between some of the corresponding variables found in the ACS and Census LF that require more detailed examination. For example, the loss function analysis reveals notable differences for race and disability variables. In other comparisons of corresponding variables between ACS and Census 2000, differences are found within each of the four major areas of interest: (1) demographic characteristics, (2) social characteristics, (3) economic characteristics, and (4) housing characteristics, with housing characteristics showing the least similarity overall. These results also suggest that more detailed examinations are needed to understand differences between corresponding variables collected by ACS and the Census LF.     

The American Community Survey in counties with “seasonal” populations

The U.S. Census Bureau designed the American Community Survey (ACS) to provide annual estimates of social and economic characteristics for states, counties, municipalities, census tracts, and block groups. Because of its April 1 reference date, in northern nonmetropolitan counties with substantial seasonal population fluctuations the decennial census provides a statistical representation of the demographic and social characteristics of the population at a time when the population is close to its annual minimum. The year-round monthly ACS sample survey has the potential to provide local communities with an unprecedented understanding of the average population characteristics over the course of a year. In the future, the ACS even has the potential for providing social and economic characteristics of the population by season. This paper examines four ACS pilot data collection counties, Oneida and Vilas Counties in northern Wisconsin, and Lake and Flathead Counties in northwest Montana. We hypothesize that the ACS will reflect a resident population over the course of the year that is different from the traditional April 1 decennial census population. While the ACS holds much promise, our research uncovered some sampling problems that are not yet fully resolved. In addition, our analysis was not able to examine ACS estimates for minor civil divisions (MCDs), which are functioning governmental units in many states. The fact that these MCDs often have very small populations, together with the fact that estimated standard errors at the much larger census tract level in these counties are disconcertingly large, raises (currently unanswerable) questions concerning the eventual statistical quality of ACS estimates for small MCDs. Consequently, the adequacy of the ACS as a replacement for the census long form may depend on the ability of the Census Bureau to effectively address the concerns presented in this analysis.     

Comparison of Census 2000 and American Community Survey 1999–2001 Estimates: San Francisco and Tulare Counties, California

Data from the United States 2000 decennial census long form sample is compared to the U.S. Census Bureau’s fledgling American Community Survey (ACS) that was designed to replace the census long form in 2010. This article concentrates on two California counties, San Francisco and Tulare, which were part of the demonstration phase of the ACS. These counties are described and an overall comparison of the demographic, social, economic, and housing variables is presented. The project data and measures of census and survey quality such as self-response rates and nonresponse rates are displayed and discussed. Differences in the census and survey results are noted in the context of statistically significant and meaningful differences. Finally, strategies for analyzing and using ACS data are suggested.     

State-Level 2010 Census Coverage Rates for Young Children

The Census Bureau’s demographic analysis (DA) shows that the net undercount rate for children aged 0–4 was 4.6 percent in the 2010 U.S. Decennial Census while adults (age 18 and older) had a net overcount rate of 0.7 percent. For the population aged 0–4, DA estimates are seen as more accurate than the U.S. Decennial Census because the estimates for this young population rely heavily on highly accurate birth certificate data. Given the relatively high net undercount rate for young children, it would be useful to examine census coverage rates for this population in subnational geographic units. In this study, the 2010 U.S. Decennial Census counts of children aged 0–4 are compared to the corresponding figures from the Census Bureau’s Vintage 2010 Population Estimates in each state. Differences between the 2010 U.S. Decennial Census count and the Vintage 2010 Population Estimates for the population aged 0–4 range from an estimated net undercount of 10.2 percent in Arizona to an estimated net overcount of 2.1 percent in North Dakota. Larger states tended to have higher net undercounts than smaller states. The ten largest states account for about 70 percent of the national net undercount of the population aged 0–4. Of all the factors examined here, the relative size of the Blacks Alone or in Combination plus Hispanics population is most highly correlated with the estimated net undercount of the population aged 0–4. Other measures that were highly correlated with net undercount rates for the population aged 0–4 were linguistic isolation, percent of adults without a high school degree, and the unemployment rate. In general, characteristics of people are more highly correlated with the net undercount rates of young children than the characteristics of housing units.     

Recent net alien immigration to the United States: Its impact on population growth and native fertility

Estimates of the size and structure of recent alien immigration to the United States are made. Substituting these revised estimates in the Series II projections of the U.S. Bureau of the Census implies a future U.S. population smaller than that implied by the Census Bureau’s estimates of immigration. The analysis of Coale (1972)—which calculates the decline in native-born fertility required to accommodate immigration and, at the same time, maintain a stationary population—is replicated, using both the Census Bureau’s estimates and the revised estimates reported here. The revised estimates indicate a smaller reduction in native fertility and a smaller ultimate size of the stationary population than are implied by the Census Bureau’s immigration estimates. The importance of age structure in all of these calculations is demonstrated.     

Research Note: US Census Same-Sex Couple Data: Adjustments to Reduce Measurement Error and Empirical Implications

US Census same-sex couple data represent one of the richest and most frequently used data resources for studying the LGBT population. Recently, the Census Bureau conducted an analysis of a serious measurement problem in these data, finding that as many as 40 % of same-sex couples tabulated in Census 2000 and 28 % of those tabulated in Census 2010 were likely misclassified different-sex couples (O’Connell and Feliz, Bureau of the Census, 2011). As a result, the Census Bureau released new state-level “preferred” estimates for the number of same-sex couples in these years, as well as previously unavailable information regarding the error rate of sex misclassification among different-sex married and unmarried couples by state and year. Researchers can use this information to adjust same-sex couple tabulations for geographic areas below the state level. Using these resources, this study: (1) considers in greater detail how the properties of the same-sex couple error might affect statistical inference, (2) offers a method for developing sub-state estimates of same-sex couples, and (3) demonstrates how using adjusted estimates can improve inference in analyses that rely on understanding the distribution of same-sex couples. In order to accomplish the third task, we replicate an analysis by McVeigh and Diaz (American Sociological Review 74: 891–915, 2009) that used county level Census 2000 unadjusted same-sex couple data, substitute our adjusted same-sex couple estimate, and examine the way in which this substitution affects findings. Our results demonstrate the improved accuracy of the adjusted measure and provide the formula that researchers can use to adjust the same-sex couple distribution in future analyses.     

The Quality of Male Fertility Data in Major U.S. Surveys

Researchers continue to question fathers’ willingness to report their biological children in surveys and the ability of surveys to adequately represent fathers. To address these concerns, this study evaluates the quality of men’s fertility data in the 1979 and 1997 cohorts of the National Longitudinal Survey of Youth (NLSY79 and NLSY97) and in the 2002 National Survey of Family Growth (NSFG). Comparing fertility rates in each survey with population rates based on data from Vital Statistics and the U.S. Census Bureau, we document how the incomplete reporting of births in different surveys varies according to men’s characteristics, including their age, race, marital status, and birth cohort. In addition, we use Monte Carlo simulations based on the NSFG data to demonstrate how birth underreporting biases associations between early parenthood and its antecedents. We find that in the NSFG, roughly four out of five early births were reported; but in the NLSY79 and NLSY97, almost nine-tenths of early births were reported. In all three surveys, incomplete reporting was especially pronounced for nonmarital births. Our results suggest that the quality of male fertility data is strongly linked to survey design and that it has implications for models of early male fertility.     

Assessing the accuracy of Australia’s small-area population estimates

The Australian Bureau of Statistics provides population estimates of Statistical Local Areas annually. The accuracy of these estimates can be assessed after population estimates are rebased after each quinquennial Census of Population and Housing, however there appears to be no straightforward method of assessing these estimates. Errors that occur with population estimates can be attributed to several factors, both broad and specific to individual areas. These factors include inherent characteristics of the region, such as population size and growth rate; changes in the geographic boundaries; quality of input data; estimation method; and adjustments to control totals (state populations).     

Selective Out-Migration from Florida

I test if selective out-migration of unhealthy seniors explains why disability rates are so much lower for Florida, as compared to the national average. This particular area of research is timely given the significant demographic changes relating to aging. Moreover, this study contributes to the body of literature examining migration with respect to disability and widowhood. Using State Federal Information Processing Standard (FIPS) and Public Use Microdata Areas (PUMA), I create national maps showing disability rates for the following age-groups: 50–59, 60–69, and 70+. After creating maps in ARCGIS and conducting univariate and clustering analysis on mobility disability and personal care limitation, I employ multinomial logit (MNL) analysis to test if individuals with disability are more likely to out-migrate from Florida. The regression analyses lend support to the relaxed Litwak and Longino (The Gerontologist, 27(3): 266–272, 1987) second-move hypothesis, which claims individuals with progressively worse health are more likely to undertake another move to be closer to family and friends. I state “relaxed” because the data does not allow one to determine the reason for migration—only that migration occurred during the past year. This research informs policy-makers to recognize that elderly in better health may migrate to places such as Arizona and Florida due to amenity-seeking behavior, but unhealthy elderly are more likely to leave these states due to assistance-seeking behavior. This out-migration can place excess demand on health services for the incoming regions, which requires state and local government to ensure resources are in place. Also noteworthy, my results are less likely to be flawed by erroneous age and sex data in the public use microdata samples (IPUMS) since I stack the 2006 and 2007 American Community Survey (ACS). A recent working studies by Alexander et al. (Inaccurate age and sex data in the Census PUMS files: Evidence and implications. Munich: CESifo, 2010) shows inaccuracies in the IPUMS for the 1 and 5% 2000 Census, the 2003–2006 ACS, the 2005–2007 3-year ACS, and the 2004–2009 current population survey (CPS) files.     

The business of demographics

The emergence of "demographics" in the past 15 years is a vital tool for American business research and planning. Tracing demographic trends became important for businesses when traditional consumer markets splintered with the enormous changes since the 1960s in US population growth, age structure, geographic distribution, income, education, living arrangements, and life-styles. The mass of reliable, small-area demographic data needed for market estimates and projections became available with the electronic census--public release of Census Bureau census and survey data on computer tape, beginning with the 1970 census. Census Bureau tapes as well as printed reports and microfiche are now widely accessible at low cost through summary tape processing centers designated by the bureau and its 12 regional offices and State Data Center Program. Data accessibility, plummeting computer costs, and businessess' unfamiliarity with demographics spawned the private data industry. By 1984, 70 private companies were offering demographic services to business clients--customized information repackaged from public data or drawn from proprietary data bases created from such data. Critics protest the for-profit use of public data by companies able to afford expensive mainframe computer technology. Business people defend their rights to public data as taxpaying ceitzens, but they must ensure that the data are indeed used for the public good. They must also question the quality of demographic data generated by private companies. Business' demographic expertise will improve when business schools offer training in demography, as few now do, though 40 of 88 graduate-level demographic programs now include business-oriented courses. Lower cost, easier access to business demographics is growing as more census data become available on microcomputer diskettes and through on-line linkages with large data bases--from private data companies and the Census Bureau itself. A directory of private and public demographic resources is appended, including forecasting, consulting and research services available.     

Reconstructing the size of the African American population by age and sex, 1930–1990

We estimate the size of the African American population in five-year age groups at census dates from 1930 to 1990 using a three-part strategy. For cohorts born after 1935, we follow the U.S. Census Bureau in using classical demographic analysis. To estimate the size of cohorts born before 1895, we use extinct-generation estimates. For remaining cohorts, we implement an age/period/cohort model of census counts. All approaches are applied to a data set in which the age distribution of deaths has been corrected for age misreporting. Results provide strong confirmation of the basic validity of Census Bureau estimates of census undercounts for African Americans while extending estimates to new cohorts and periods. Our estimates are less consistent with an historical series prepared by Coale and Rives (1973).     

The Impact of Local Labor Market Conditions on Work and Welfare Decisions: Revisiting an Old Question Using New Data

Using Maryland administrative data between 1996 and 2005, this paper examines the impact of local labor market conditions on work and welfare use among single mothers. Our estimates rely on the new Census Bureau Quarterly Workforce Indicators database, which provides county-level economic indicators filtered by industry, gender, and age-group. We specify a multinomial choice model to estimate the effects of these local labor market variables on the full set of work–welfare combinations. The results indicate that lower unemployment rates and increased new hires and new hires’ earnings in key industries increase the likelihood that women choose alternatives that include work. African American women and those with fewer years of education respond differently to changing economic conditions. Our results are robust to controls for fixed effects, county-specific time trends, and endogenous migration.     

Population Trends in Metropolitan and Nonmetropolitan America: Selective Deconcentration and the Rural Rebound

This study uses revised annual population estimates that incorporate adjustments from the 2000 Census to backcast demographic change for U.S. counties during the 1990s. These data are supplemented with new post-censal population estimates for 2001–2003. We use these data to examine demographic trends in the late 1990s and first years of the new century. Our findings are consistent with a model suggesting that a selective deconcentration of the U.S. population is underway. Our findings also confirm the occurrence of the rural rebound in the first half of the 1990s and a waning of this rebound in the late 1990s. Post-censal data also suggest a modest upturn in nonmetropolitan population growth rates in 2001–2003.     

Misclassifying New York’s Hidden Units as Vacant in 2010: Lessons Gleaned for the 2020 Census

The Local Update of Census Addresses (LUCA) program allowed local governments to include hard-to-find units in the Census Bureau’s Master Address File (MAF), which is the cornerstone of the mailout/mailback decennial census. These improvements have allowed the Census Bureau to penetrate the more marginal parts of the housing stock, where units are often not formally labeled, and where their very existence can be difficult to determine. In New York City, where address updating included two rounds of LUCA, the Census Bureau acknowledged an increase of 170,000 housing units between 2000 and 2010. However, there was a dramatic growth in vacant units, equivalent to almost one-half of the total increase in housing units. The increase in vacant units was disproportionately concentrated in 2 of the 18 local census offices in New York City. The paper uses local administrative data on new construction, property foreclosures, and property values; data from the United States Postal Service; as well as survey data from the New York City Housing and Vacancy Survey and the American Community Survey to show why this concentrated increase in vacant units is untenable. From the standpoint of the enumeration, units added in LUCA would challenge the best enumerator, but these hurdles were largely overcome, but for the two local census offices. The paper goes on to discuss how the Census Bureau can adopt measures in 2020 to ensure that housing units and their occupancy status are accurately enumerated in New York and across the nation.     

Using administrative data to estimate population displacement and resettlement following a catastrophic U.S. disaster

After a large scale evacuation, authorities need to know the new and frequently changing population distributions in order to meet needs for housing, schools, health care, and other services. This paper reviews literature from the fields of demography and other disciplines to identify available administrative data sets that can form the basis of sound, relevant, and timely county-level population estimates following a catastrophic U.S. event. The most appropriate data to estimate population in damaged counties will be disaster-specific data such as housing damage estimates and FEMA applicant counts initially, and later electric accounts and USPS active residences. In heavily damaged counties, data on electric accounts and USPS active residences may not be consistently collected for many months, during which time sample surveys may be needed. For counties that receive an influx of population, school enrollment data provide the most appropriate basis for population estimates. Population estimates for large, heavily damaged counties are highly uncertain. Sensitivity analysis when using estimates for planning in these areas is recommended. The Census Bureau can build on this research by codifying recommendations to local authorities for developing frequent post-disaster population estimates.     

Coverage of the population in Census 2000: Results from Demographic Analysis

Since the 1960 Census, Demographic Analysis (DA) has been used by the Census Bureau to evaluate the coverage of the population. Administrative statistics on births, deaths, immigration and Medicare enrollments as well as estimates of legal emigration and net undocumented immigration are used to produce demographic analysis estimates of the population for the census date. These estimates are compared to the Census 2000 data to evaluate coverage in the census. The results are also compared to measures of undercount obtained from dual system estimation. The DA measures substantial reduction in net undercount in Census 2000 compared to 1990. The reductions occur among all demographic categories: all broad age groups, males and females, Blacks and Non-Blacks.