Population Census and Population Estimates
U.S. Census Bureau
The Decennial Census of Population has been held in the United States every 10 years since 1790. Since 1930, it has enumerated the resident population as of April 1 of the census year. Data on sex, race, Hispanic origin, age, and marital status are collected from 100% of the enumerated population.
Race Data on the 1990 Census
The question on race on the 1990 census was based on the Office of Management and Budget’s (OMB) 1977 “Race and Ethnic Standards for Federal Statistics and Administrative Reporting” (Statistical Policy Directive 15). This document specified rules for the collection, tabulation, and reporting of racial and ethnic data within the federal statistical system. The 1977 standards required federal agencies to report race-specific tabulations using four single-race categories: American Indian or Alaskan Native, Asian or Pacific Islander, Black, and White. Under the 1977 standards, race and ethnicity were considered to be two separate and distinct concepts. For this reason, people of Hispanic origin may be of any race.
Race Data on the 2000 Census
The question on race on the 2000 census was based on OMB’s 1997 “Revisions to the Standards for the Classification of Federal Data on Race and Ethnicity” (see: Federal Register 62(210):58782–90. 1997.). (Also see Sources and Definitions, Race.) The 1997 standards incorporated two major changes in the collection, tabulation, and presentation of race data. First, the 1997 standards increased the minimum set of categories to be used by federal agencies for identification of race from four to five: American Indian or Alaska Native, Asian, Black or African American, Native Hawaiian or Other Pacific Islander, and White. Second, the 1997 standards included the requirement that federal data collection programs allow respondents to select one or more race categories when responding to a query on their racial identity. This provision means that there are potentially 31 race groups, depending on whether a person selects one, two, three, four, or all five of the race categories. The 1997 standards continue to call for use, when possible, of a separate question on Hispanic or Latino ethnicity and specify that the ethnicity question should appear before the question on race. As a result, under the 1997 standards, as under the 1977 standards, people of Hispanic origin may be of any race.
Race Data on the 2010 Census
Similar to race data on the 2000 census, the question on race on the 2010 census was based on OMB’s 1997 “Revisions to the Standards for the Classification of Federal Data on Race and Ethnicity” (see: Federal Register 62(210):58782–90. 1997.). (Also see Sources and Definitions, Race.) The 1997 standards require a minimum set of five categories to be used by federal agencies for identification of race: American Indian or Alaska Native, Asian, Black or African American, Native Hawaiian or Other Pacific Islander, and White, and require that federal data collection programs allow respondents to select one or more race categories when responding to a query on their racial identity. Like the 1977 standards, the 1997 standards continue to call for use, when possible, of a separate question on Hispanic or Latino ethnicity and specify that the ethnicity question should appear before the question on race. Consequently, under the 1997 standards, as under the 1977 standards, people of Hispanic origin may be of any race.
Modified Decennial Census Files
For several decades, the U.S. Census Bureau has produced modified decennial census files. These modified files incorporate adjustments to the 100% April 1 count data for a) errors in the census data discovered subsequent to publication, b) misreported age data, and c) nonspecified race.
For the 1990 census, the Census Bureau modified the age, race, and sex data on the census and produced the modified age-race-sex (MARS) file. The differences between the population counts in the original census file and the MARS file are primarily due to modification of the race data. Of the 248.7 million people enumerated in 1990, 9.8 million did not specify their race (more than 95% were of Hispanic origin). For the 1990 MARS file, these people were assigned the race reported by a nearby person with an identical response to the Hispanic-origin question.
For the 2000 and 2010 censuses, the Census Bureau modified the race data and produced the modified race data summary files. For these files, people who did not report a race (reporting only the category “Some other race”) as part of their race response were assigned by imputation to one of the 31 race groups, which are the single- and multiple-race combinations of the five race categories specified in the 1997 OMB race and ethnicity standards. In the 2000 census, 97% of the 15.4 million people who did not report a race were of Hispanic origin. Because a large proportion of those identifying their race as “Some other race” are Hispanic, a new instruction was added to the 2010 census specifying that, for the census, Hispanic origins are not races. In the 2010 census, 97% of the 19.1 million people who did not report a race (reporting only the category “Some other race”) were of Hispanic origin.
Postcensal Population Estimates
Postcensal population estimates are estimates made for the years following a census, before the next census is taken. Postcensal population estimates are derived annually by updating the resident population enumerated in the decennial census using a components-of-population-change approach. Each annual series includes estimates for the current data year and revised estimates for the earlier years in the decade. The following formula is used to derive national estimates for a given year from those for the previous year, starting with the decennial census enumerated resident population as the base:
- Resident population estimate
+ births to U.S. resident women
– deaths to U.S. residents
+ net international migration
The postcensal estimates are consistent with official decennial census figures and do not reflect estimated decennial census underenumeration.
Estimates for the earlier years in a given series are revised to reflect changes in the components-of-population-change data sets (for example, births to U.S. resident women from a preliminary natality file are replaced with counts from a final natality file). To help users keep track of which postcensal estimate is being used, each annual series is referred to as “vintage,” and the last year in the series is used to name the series. For example, both the vintage 2011 and the vintage 2012 postcensal series have revised estimates for July 1, 2011, but the estimates for July 1, 2011, from the vintage 2011 and vintage 2012 postcensal series differ.
The Census Bureau also produces postcensal estimates of the resident population of states and counties, using the components-of-population-change method. An additional component of population change—net internal migration—is involved.
Intercensal Population Estimates
Intercensal population estimates are estimates made for the years between two decennial censuses and are produced once the census at the end of the decade has been completed. They replace the postcensal estimates produced before the completion of the census at the end of the decade. Intercensal estimates are more accurate than postcensal estimates because they are based on both the census at the beginning and the census at the end of the decade. They are derived by adjusting the final postcensal estimates for the decade to correct for the error of closure (the difference between the estimated population at the end of the decade and the census count for that date). The patterns of population change observed over the decade are preserved. The intercensal estimates for the 1990s were produced using the same methodology used to generate the intercensal estimates for the 1980s. The revised intercensal population estimates for 2000–2009 were produced using a modified version of the methodology used previously. Vital rates calculated using postcensal population estimates are routinely revised when intercensal estimates become available.
Bridged-race Population Estimates
During the transition to full implementation of the 1997 OMB standards on race and ethnicity, race data on the 2000 and 2010 censuses were not comparable with race data on other data systems that were continuing to collect data using the 1977 OMB standards on race and ethnicity. For example, states implemented the revised birth and death certificates—which have race and ethnicity items that are compliant with the 1997 OMB standards—at different times, and some states still used the 1989 death certificates that collect race and ethnicity data according to the 1977 OMB standards. As a result, population estimates for 1990 and beyond with race categories comparable with the 1977 OMB categories were needed so that race-specific birth and death rates could be calculated. To meet this need, the National Center for Health Statistics (NCHS), in collaboration with the Census Bureau, developed methodology to bridge the 31 race groups in U.S. Census 2000 and 2010 to the four single-race categories specified under the 1977 OMB standards. Beginning in 2016, all states and the District of Columbia (D.C.), in addition to Guam, Northern Mariana Islands, Puerto Rico, and U.S. Virgin Islands, used the 2003 revision of the U.S. Standard Certificate of Live Birth. Beginning in 2018, all states and D.C. reported deaths using the 2003 revision of the U.S. Standard Certificate of Death. Starting with 2016 birth and 2018 death data, Health, United States has reported race according to the 1997 revised OMB standards. However, to provide longer-term trends for specific groups found only in the 1977 OMB standards, selected estimates using bridged single-race categories continue to be presented in Health, United States.
The bridging methodology was developed using information from the 1997 through 2000 National Health Interview Survey (NHIS). NHIS provides a unique opportunity to investigate multiple-race groups because, since 1982, it has allowed respondents to choose more than one race but has also asked respondents reporting multiple races to choose a primary race. The bridging methodology developed by NCHS involved application of regression models relating person- and county-level covariates to the selection of a particular primary race by the multiple-race respondents. The bridging proportions derived from these models have been applied by the Census Bureau to various unbridged resident population files. These applications have resulted in bridged-race population estimates for each of the four single-race categories: American Indian or Alaska Native, Asian or Pacific Islander, Black, and White.
In Health, United States, vital rates for 1991–1999 were calculated using the July 1, 1991–July 1, 1999, bridged-race intercensal estimates. Vital rates for 2000 were calculated using the bridged-race April 1, 2000, census counts, and those for 2010 were calculated using bridged-race April 1, 2010, census counts. Starting with Health, United States, 2012, vital rates for 2001–2009 have been recalculated using the July 1, 2001–July 1, 2009, revised intercensal bridged-race population estimates. Vital rates for 2011 and beyond are calculated using bridged-race estimates of the July 1 population from the corresponding postcensal vintage.
- Ingram DD, Parker JD, Schenker N, Weed JA, Hamilton B, Arias E, Madans JH. United States Census 2000 population with bridged race categories. National Center for Health Statistics. Vital Health Stat 2(135). 2003. Available from: https://www.cdc.gov/nchs/data/series/sr_02/sr02_135.pdf.
For more information, see the U.S. Census Bureau website at: https://www.census.gov, and the NCHS website for U.S. Census populations with bridged-race categories at: https://www.cdc.gov/nchs/nvss/bridged_race.htm.