COPAFS
 

U.S. Census Bureau Conference On Population Estimates:
Meeting User Needs
Basic Underlying Principles

Background

The U.S. Census Bureau’s Population Estimates and Projections program is designed to fulfill the mandates of Title 13, Section 181, of the U.S. Code.

During the intervals between each census of population required under section 141 of this title, the Secretary, to the extent feasible, shall annually produce and publish for each State, county, and local unit of general purpose government which has a population of fifty thousand or more, current data on total population and population characteristics and, to the extent feasible, shall biennially produce and publish for other local units of general purpose government current data on total population. Such data shall be produced and published for each State, county, and other local unit of general purpose government for which data is compiled in the most recent census of population taken under section 141 of this title. Such data may be produced by means of sampling or other methods, which the Secretary determines will produce current, comprehensive, and reliable data.

To satisfy this mandate, the program of population estimates has grown over the years to produce the following products annually:

  1. Monthly estimates of the national population of the United States by age, sex, race, and Hispanic origin

  2. Annual estimates of the population of states by age, sex, race, and Hispanic origin

  3. Annual estimates of the population of counties by age, sex, race, and Hispanic origin

  4. Annual estimates of the total population of functioning governmental units

  5. Annual estimates of the number of housing units for states and counties.

In addition to meeting the mandates of Title 13, these estimate products are used for a variety of purposes, including the following:

  1. Controls for federally sponsored surveys, including the Current Population Survey (CPS) and the American Community Survey (ACS)

  2. Allocation of federal dollars totaling over $200 billion annually

  3. Denominators for various indicators, including vital statistics, per capita income, and cancer incidence rates

  4. Calculation of the number of clerks the Senate hires

  5. Requirements of the Federal Election Commission

  6. Denominators for poverty rate estimation at selected levels of geography

  7. Program planning by federal, state, local, and private entities

Implicit Assumptions

Implementation of the annual program of intercensal estimates is guided by several implicit assumptions.

Timely release of the annual products is critical

  1. The maximum lag time between estimate date and dissemination of last data product is 12 months.

  2. Annual national and state population totals must be released within 6 months of estimate date to meet requirements of IRS Bonding Authority.

  3. State estimates of the population aged 18 and older must be available within 6 months of estimate date to satisfy requirements of the Federal Election Commission.

  4. National and state population controls to be used for the new calendar year CPS must be available by late January of the new calendar year.

  5. Estimates of state and county characteristics must be available within 9 months to meet requirements for use as population controls for the American Community Survey.

  6. Estimates of functioning governmental units should be available within 12 months of estimate date for use by HUD in funds allocation.

Each annual production consists of a time series of estimates from the last decennial census date to the estimate date and is produced using the latest available data and the current approved methodology.

  1. Current-year data products contain revisions to the prior year’s estimates that are caused by incorporating:

    • Improved methodology.

    • New data inputs.

    • Revisions to prior year data inputs.

  2. The term “vintage” is used to refer to the reference date of an estimates cycle. Estimates released with a reference date of July 2005 are referred to as the “vintage 2005” set of population estimates and will include a consistent time series back to April 2000.

Within any vintage, all products use the same vintage of input data and must sum to the earlier released products of the same vintage for the same measurement.

  1. Since the national and state population totals are the first to be released, all subsequent estimate products must sum to the national and state totals that already appear for that vintage. This insures consistency within any vintage and means that the sum of the “parts” will always equal the previously released U.S., state, or county total.

  2. Since the national population estimates tabulated by characteristics are the first characteristics to be released, the sum of the state and county characteristics must equal the national characteristics of the same vintage.

Only one consistent set of products and related materials is developed within a vintage. That set of products is intended to serve all customers’ needs and uses.

  1. The methodology and data inputs used to develop the population estimates used as denominators for vital statistics rates are consistent with those used to develop the population controls for the CPS and ACS.

  2. Custom data products are consistent with the publicly released data products. For example, the annual race estimates for counties use a bridged race algorithm developed by NCHS. However, while the race data conform to the bridging algorithms developed by NCHS, the estimates of total populations and populations by age and sex generally agree with the publicly released data products.

The population estimates begin with the most recent decennial-census enumerated count updated to July 1 of each year, and as such, are based on the usual-residence concept used in the most recent decennial census.

  1. The population estimates base for each estimate date is updated to include Count Question Resolution (CQR) changes to the decennial census base as well as geographic updates due to annexation and other geographic program changes.v

  2. The components of population change used to update the most recent census will be consistent with the best set of components available. Ongoing evaluation indicates that the coverage and the consistency of vital statistics and other administrative records data differ from those of decennial census data. Therefore, in the annual estimates, the size of the population based mainly on administrative records data differ from the size based mainly on census data.

States, counties, and units of local government have the right to challenge the population estimates prepared by the Census Bureau under the provisions of Title 15, The Code of Federal Regulations, Part 90. The results of accepted challenges will be incorporated into the following year’s population estimates as long as the challenge is received by October 1 of the year in which the estimate was released.

Current Broad Methodological Assumptions

  • Prior to incorporating a new methodology or data set, it is desirable to thoroughly evaluate a set of estimates that use this new methodology or data set and compare it with the most recent decennial census results. When this is not possible, the methods are judged by the following criteria.

    1. Soundness: The method should be based on solid reasoning – i.e., the formulas that embody the method should be mathematically valid and respect the attributes of the input data as they relate to the estimation task.

    2. Integrity: A strategy that consistently applies the declared method is preferred to one that uses ad-hoc fixes to address particular challenges of the estimation task.

    3. Parsimony: A simpler strategy is preferred to a more complex one.

    4. Robustness: The method that produces the most reasonable estimates (defined below) across the full range of potential input-data values and in the presence of the random variation normally associated with those values while maintaining the orthodoxy and consistency of the estimates (also defined below) is preferred.

    5. Adaptability: A technique that can be applied more broadly (e.g., across geographic summary levels), thus promoting the integration of the Census Bureau’s estimates system, is preferred to a more product-specific remedy.

    6. Transparency: A strategy that is more readily understandable and replicable by external parties is preferred. Moreover, a strategy that provides some explanatory information (i.e., how did the size or distribution of the population come to be this way) is preferred over one that is merely predictive.

    7. Usability: The method must be executable along with all other current projects under current staffing levels in a way that allows the Census Bureau to meet current deadlines.

    8. Flexibility: The preferred method will allow the production of estimates when a specific instance of the input data normally required by the method is unavailable or deemed unsuitable.

As a final test, the method should produce output data that have the following qualities.

  1. Orthodoxy: The values of the population estimates should be appropriate (e.g., no negative population numbers, all population estimates in whole numbers).

  2. Consistency: The values of the population estimates for all universes (e.g., resident, civilian, civilian non-institutionalized), geographies (e.g., national, state, county), and characteristics (e.g., age, sex, race, Hispanic origin) should not contradict one another.

  3. Reasonableness: The values of the population estimates should approximate the real values as determined by the following assessments.

    • Post-Censal Change: The reasonableness of the total change in the population since the last decennial census.

    • Time-Series Change: The reasonableness of the annual change in the estimates since the last census.

    • Demographic Appropriateness: The values of the estimates and the demographic rates they imply fall within acceptable limits when evaluated by general demographic principles (e.g., the appropriateness of the sex ratios, age progression, implied family size, life expectancies, total fertility rates, etc.).

    • Comparability: The estimates appear realistic when compared with other indicators of the size and distribution of the population (e.g., Medicare enrollment, school enrollment, housing unit estimates, etc.).

A consistent method is used for entities at the same level of geographic aggregation.

  1. The method adopted for state totals must be used for all states.

  2. The method adopted for counties within a state must be used for all counties within that state.

The Census Bureau develops the basic estimates for the nation, states, and counties by disaggregated race groups in order to meet the various custom race aggregations needed by users.

The cohort-component method is the preferred method for development of the national, state, and county-level total population estimates and population estimates by characteristics.

The distributive housing-unit method is the preferred method for the development of the functioning subcounty governmental-unit-level estimates.

State total population estimates are not developed independently. National population estimates are first developed; then county total population estimates are developed and controlled to the national total population estimates. The state total population estimates are the sum of the “nationally controlled” county total population estimates for the state.

Data on vital statistics and group quarters provided by members of the Federal State Cooperative Program for Population Estimates (FSCPE) are included in the process of developing state and county population estimates.

Although state members of the FSCPE are provided the opportunity to review the state and county population totals prior to final production, they must follow strict criteria and provide objective evidence when requesting modifications.

Current Specified Methodologies

  • National level estimates will use the cohort-component technique applied to data from the latest decennial census as the base, data on births and deaths provided by the National Center for Health Statistics, and estimates of net international migration derived from data from the American Community Survey (ACS) See the url http://www.census.gov/popest/topics/methodology/v2005_nat_char_meth.html. For a detailed discussion of the methodology used to develop the most recent set of national population estimates by demographic characteristics.

  • State and county population estimates are developed using a demographic procedure called an "administrative records component of population change" method. A major assumption underlying this approach is that the components of population change are closely tracked by administrative data in a demographic change model. In order to apply the model, Census Bureau demographers estimate each component of population change separately. For the population residing in households, the components of population change are births, deaths, and net migration, including net international migration. For the non-household population, change is represented by the net change in the population living in group-quarters facilities.

    Each component in our model represents data that are symptomatic of an aspect of population change. For example, birth certificates indicate additions to the population resulting from births, so we use these data to estimate the birth component for a county. Other components are derived from death certificates, Internal Revenue Service data (IRS), Medicare enrollment records, Armed Forces data, group-quarters population data, and data from the American Community Survey.

    For a more detailed discussion of the development of county population totals see http://www.census.gov/popest/topics/methodology/2005_st_co_meth.html

  • State population characteristics are currently developed in a two-stage process. Estimates by age and sex are developed first using a cohort-component procedure whereby estimates of net migration are developed using school enrollment data. These estimates are controlled both to the national-level estimates by age and sex as well as the previously developed state population totals.

    The second step in the process distributes the state age and sex estimates into race by Hispanic origin categories. This is done by preparing an initial set of state estimates by age, sex, race, and Hispanic origin that are controlled to the state age and sex estimates prepared in the first step and to the previously developed national estimates by age, sex, race, and Hispanic origin.

    For a more detailed discussion of the development of the state population characteristics by age, sex, race, and Hispanic origin see http://www.census.gov/popest/topics/methodology/2004_st_char_meth.html

  • County population characteristics are developed using a proportional distribution method beginning with previously developed resident county population estimates by age (0-64 and 65+) and resident state population estimates by age, sex, race, and Hispanic origin. Then county-level estimates of age, sex, race, and Hispanic origin distributions are developed using information about post-censal change in the corresponding populations. Third, these distributions are applied to the original county estimates by age and state characteristics.

    A detailed discussion of this method is provided at http://www.census.gov/popest/topics/methodology/2004_co_char_meth.html

Enhancement Priorities

  • Improve estimates of net international migration

    1. Provide up-to-date, useful statistics and methodologies on the size, characteristics, and demographic impact of international migration to and from the United States for use in policy-making decisions and demographic and economic research.

    2. Goals of immigration research

      • Produce annual estimates of international migration

      • Improve current migration-related survey questions on the ACS.

      • Conduct extensive evaluations to determine the best method to incorporate ACS data into the population estimates.

    3. Activities

      • Evaluate reasonableness of estimates of annual change in the foreign-born data from ACS at the national level.

      • Produce revised estimates of net international migration at the national level.

      • Produce new demographic and geographic distributions for migrants.

      • Construct algorithms to estimate the migrant status of the foreign-born populations.

      • Produce estimates of international migrants by migrant status (legal migrants, temporary migrants, quasi-legal migrants, unauthorized migrants, and emigrants).

Improve Estimates of Internal Migration

  1. Improve the accuracy of the annual migration estimates by age, sex, race, and Hispanic origin for counties by maximizing the efficient use of available administrative data files, Census 2000 data, and the American Community Survey (ACS) data.

  2. The ultimate goal is to implement a person-based migration model incorporating administrative data from files such as the IRS 1040 and 1099 records, Medicare records, a derived person-characteristic file developed from the Social Security Administrative NUMIDENT file, and other administrative data that can be merged into the database. The database will enable analysts to match administrative data with Census 2000 (100% and sample data), CPS, and ACS data in order to develop models that correct possible demographic and geographic biases inherent in the use of an administrative records database when estimating migration rates for counties.

Develop a new methodology for estimating subnational population characteristics

  1. Replace the methodology that develops state estimates by age and sex based on school enrollment data with a method that is consistent with the best set of administrative data available and exploits the power of current computing capacity.

  2. Develop a method that addresses current deficiencies in the age distributions of the population in selected states and counties, especially the age distribution of the population aged 18 to 24.

  3. Develop a new method to estimate county population by age, sex, race, and Hispanic origin.

Develop procedures to systematically incorporate participation by State FSCPE Agencies in the production of state and county population estimates

  1. Address issues of consistency

  2. Establish criteria for incorporating state participation

Other Enhancements

Improve the distributive housing unit approach at the subcounty level.

  1. Develop procedures to update Census 2000 measures of vacancy and numbers of people per household (PPH or the Person Per Household measure) used in the estimates process.

  2. Improve estimates of housing units.

  3. Address inconsistencies between estimates developed using the distributive housing unit approach and those developed using the component approach.

    • Develop improved procedures to estimate housing unit loss.

    • Integrate enhancements from the Master Address File.

Address inconsistencies between data from the decennial census base and data on components of change from administrative records databases.

  1. Address inconsistencies between Census 2000 data and NCHS data on race and Hispanic-origin characteristics.

  2. Address unreasonable results from pairing NCHS mortality data with decennial census data and estimate results.

Administrative Constraints

  • The methods developed must be capable of being implemented with current resources and within the current time frame for estimate production.

  • Production of the complete set of estimates must continue during any development stages.

  • Methods must be transparent and reproducible.