Introduction

A common requirement of many social programs and issues is the number of people in a population of interest. New priorities for revenue sharing and block grants require current estimates of the population for extremely small areas. In addition to information about the number of people, data are needed on the geographic distribution of the population; its characteristics-age, sex, race, ethnic origin and other demographic and social attributes-as well as its income, wealth and other economic variables. The requirements for population estimates and projections are not limited to the population of the United States. Our national interests and security require that we have estimates and projections for most other nations as well.

Data are also needed for families and households. The number of families, their size, composition and age structure are important variables for demographic analysis. These kinds of family data are required to plan for and to evaluate various income maintenance, health, education, and similar programs.

Producing Agencies

There are a limited number of agencies involved in the production of basic population data. The greatest responsibility falls on the Bureau of the Census. The Census Bureau is responsible for the constitutionally mandated decennial count as well as the new middecade census requirements. Census is also responsible for providing the official intercensal population estimates. The Department of Commerce directive 13 (formerly OMB Circular A-46 exhibit I) requires that all Federal assistance programs which utilize population as an element in a grant formula must use the most current estimate from the Census Bureau, except where other measures, such as the decennial census, are required by law. In addition the Bureau produces domestic population projections as well as foreign estimates and projections. The Bureau of Economic Analysis also produces population projections which are widely used by Federal programs. Census provides regular data on the

characteristics of the population through the monthly current population survey and its supplements.

It would not be possible for Census to develop accurate estimates of the population without good vital statistics data-births and deaths. The collection of these vital records is the responsibility of the National Center for Health Statistics (NCHS) through a cooperative effort with the States. NCHS also conducts the survey which provides the data for the World Fertility Survey.

The General Revenue Sharing Program requires Census to develop estimates of the population and per capita income for about 39,000 units of government. In order to provide such small area data, Census has had to rely on information from the Internal Revenue Service. Periodically, questions have been included on personal income tax returns (SF-1040) to ascertain the actual political jurisdiction in which the taxpayer lives. These data provide a measure of mobility needed to develop estimates for extremely small areas.

The Center for Population Research of the National Institutes of Health is actively involved in supporting research which sheds light on the dynamics of demographic change. While this research has less direct impact on population statistics, it is useful in anticipating changes in population. It also provides insights into the causes and correlates of population change which frequently has direct policy relevance.

The Immigration and Naturalization Service (INS) of the Department of Justice is developing a program to estimate the number and distribution of illegal aliens in the United States.

Users

The users of domestic population data include virtually all Federal agencies, especially those involved in Federal domestic assistance programs. Most of these programs distribute funds based on a simple population formula or on a formula which uses population as the denominator for some ratio.

Many States also distribute funds to local governments and agencies based on population counts, estimates or projections.

The original reason for the collection of population statistics is the constitutional requirement that the House of Representatives be apportioned according to population. The Constitution requires that a census of the population be taken no less frequently than every 10 years. The States similarly (though without the Federal constitutional mandate) use census counts for redistricting and other political purposes. As important as the governmental uses are the uses within the private sector. The business community depends very heavily on data from the decennial census and current surveys, estimates and projections. Decisions concerning the production and marketing of products and services are determined by the demographic characteristics of potential markets. Similarly, the academic community depends heavily on population data to in almost all of their research in the human behavioral sciences.

The Agency for International Development (AID) has the responsibility (along with the Central Intelligence Agency (CIA), for analyzing the demographic changes in nations of interest to the United States. These estimates are currently produced by the Bureau of the Census. Because of the need for comparability, and because of the high level of technical skill required, it is appropriate that Census continue in the role.

Advisory Committees

There is no shortage of advisory committees whose major functions involve population data. There are several committees advising the Bureau of the Census alone. In addition to those committees at the Census Bureau, there is the United States National Committee on Vital and Health Statistics, the Federal Agency Council on Demographic Censuses, and others which impact on the collection, development and analysis of population data.

The Decennial Census

The major program involving population statistics is the decennial census of population and housing. Every 10 years the Census Bureau counts the population and identifies its demographic, social, and economic characteristics. The content and the methods of collection of census data have changed radically over the past 2 centuries.'

See Population and Housing Inquiries in U.S. Decennial Censuses, 1790-1970, Working Paper 39, U.S. Department of Commerce, Bureau of the Census, Washington, D.C., 1973.

Title 13 United States Code gives the Secretary of Commerce and the Census Bureau wide Bureau, wide discretion in determining the content and procedures which are used for the census. The Bureau makes an extensive effort to ascertain the needs of various users and potential users through a series of programs. One of these efforts for the the 1980 Census was a set of "town meetings," in which State and local planners, the academic community, and other interested groups were asked to describe their needs. The Federal agencies' interests are transmitted to the Bureau through the Federal Agency Council for the 1980 Census. The Council is organized and chaired by the Office of Federal Statistical Policy and Standards in cooperation with the Census Bureau. While the Council is in no way a board of directors, it does provide a formal mechanism through which agencies' interests are channeled. Since a mid-decade census has now been authorized, the Council has been a permanent entity, and has been renamed "the Federal Agency Council on Demographic Censuses."

Concern for the amount of data collected and the increasing reluctance of the population to provide data has prompted a serious review of the amount of detail in the Census questionnaire. This is not a new issue. In 1960 the format of the census was changed so that the majority of the population had to answer only a limited number of questions. A sample of the population, on the other hand, had to answer a more detailed set of questions. The Census Bureau made the decision at the beginning of the planning cycle for the 1980 census that the burden on respondents receiving the "long form" would not exceed the corresponding form in 1970. During the past year significant changes have been made in the overall design of the census to reduce burdens even further. For example, an earlier decision to include an income question on the 100% short form was reversed and now income will appear only on the sample form. However, in order to provide reliable data for small jurisdictions the sample design was modified significantly. The current plan is to have a 1:6 sample in places with a population of 5,000 or more and 50% sample in places with a population of 5,000 or less. These changes will also provide reliable data for all of the sample items for small areas.

Any interest in further reducing the size of the census questionnaires would conflict sharply with the need for additional data for the design and management of Federal programs as well as decisions concerning the allocation and use of State and local revenue sharing and block grant funds. A significant reduction in the amount of data collected would also conflict with the current effort to develop a strategy for the 1980's, which seeks to increase the data

available for planning, monitoring, and evaluation while reducing the size of existing intercensal data collection activities.

A number of legislative and other changes have made an impact on census operations since the 1970 Census. The White Runnels Act, for example, required the Census Bureau to provide data for geographic subdivisions defined by the States following guidelines laid down by Census. Census is also required to provide data for legally recognized communities. As mentioned above, legislation was passed in 1976 authorizing for the first time a mid-decade

census.

Several other bills were also introduced which had some impact on the Census Bureau's operation. The most significant of these was H.R. 8871 "The Census Reform Act" and its successors HR 9623 and HR 10346. In their initial conception this bill would have radically altered Census Bureau activities as well as the overall collection of Federal Statistics. The Bureau and the Department of Commerce made a number of procedural changes largely as a result of the introduction of the legislation.

Another piece of legislation which caused great concern was one which involved an agreement made between the Census Bureau and the National Archives and Records Service (NARS). At the center of the problem is the agreement between the Census Bureau and NARS for Census to turn over Census schedules to Archives which in turn will make them available to medical researchers, historians and genealogists 72 years after the census year. This agreement was made pursuant to the Federal Records Act, which gives the National Archives the responsibility for making available important historical records of the United States. This statute appears to conflict with title 13 U.S.C and the Federal Reports Act which seem to require that the census records be kept confidential in perpetuity.

One solution to this problem would be the development of an amendment to title 13 which would:

1. make the data confidential either in perpetuity or at least for a period of time beyond the lifespan of any individual, say 125 years, and

2. require Census to develop methods of meeting reasonable demands for data from historians, researchers, and genealogists.

Capabilities and Problems

As indicated above, the basic ongoing programs involved in measuring the population are the decennial

census, the vital statistics system, and the current population survey, particularly the annual March supplement which provides detailed information on the characteristics of the population.

The Bureau of the Census has developed a capability of conducting major sample surveys which provide regular or ad hoc population data. A recent example of a major exploitation of this capability is the Survey of Income and Education (SIE). The SIE, which was required by Public Law 93-380 (the 1974 Elementary and Secondary Education Amendments), collected data on the number of families in poverty, the number of children of such families, and the number of school-aged individuals who have physical or mental disabilities or who have difficulties with English. This survey constitutes the richest set of demographic data which will be available until the publication of the 1980 census.

The National Center for Health Statistics also has significant capacity to expand on information routinely available from regular sources. They periodically conduct "follow-back" surveys based on samples drawn from vital certificates to provide on a national basis supplementary information on births and deaths. Data developed from these surveys provides explanatory material to help interpret vital statistics.

Although population statistics are among the more sophisticated of the data now generated, problems still exist. Some of these are credibility problems which arise because of the new administrative uses of population data for revenue sharing, block grants and related purposes. When the distribution of substantial amounts of money depends on "estimates," persons and units of Government are not easily convinced that the "estimates" are correct.

The Undercount

All of the currently used estimates are based on the decennial census. The infrequency of the census has, in the past, constituted a major problem. While the decennial measures may have been adequate to apportion the House of Representatives, a 10-year measure is not reasonable for the distribution of funds, particularly to areas with highly mobile populations. The legislation passed in the 94th Congress authorizing a mid-decade census may eliminate the problem of infrequency. More frequent census counts alone will not solve all of the problems. The census, as currently conducted, fails to identify and count a significant proportion of the population. It can be argued that this undercount is the responsibility of those being counted and their failure to "stand up" only results in a diminution of their advantages. The

fact remains that those who are not counted-"the undercount"-still require services from every level of Government and thus represent a disproportionate burden on others within the jurisdiction in which they live. In addition, they provide an undeserved benefit to those in jurisdictions which do not contain sizable numbers of the uncounted.

Although the Census Bureau has done a great deal of research on the problem of the undercount, it has not developed a comprehensive set of estimators for adjusting its estimates and projections for States and local units of Government. In late 1977 the Bureau did publish an important work concerning the estimation of undercoverage for states.' The report described alternative methods for determining undercoverage at the State level and identified problems with those methods. This is a significant step forward in the state of the art but it still does not solve the problems of correcting the undercount at the State level and does not address the problem at the substate level.

Some recommendations have been made by statisticians within government and by outside organizations. However, although the Census Bureau has various methods for adjusting the census counts and subsequent estimates for non-interviews, item non-response and other factors the Bureau argues that recommendations for adjusting for the undercount are technically inferior to the other estimators currently in use, therefore they cannot be adopted. High priority must be given by the Bureau to the development and utilization of a reasonable method of adjusting for the "undercount."

There are other serious problems which relate to the Census Bureau program of subnational estimates and projections. The first concerns migration back and forth from Puerto Rico. The Northeastern States in particular have large Puerto Rican populations which move freely to and from Puerto Rico. While people from Puerto Rico are United States citizens, they are not required by law to file income tax forms when they reside in Puerto Rico. Thus, the Census Bureau is deprived of its primary source of migration data, the IRS 1040, for a significant segment of the population. The Census Bureau must continue to search for methods of estimating this migration.

Another problem which is particularly acute with the Hispanic population is the poorly estimated number of undocumented workers or illegal aliens. Large numbers of persons illegally enter the United

? Developmental Estimates of the Coverage of the Population of States in the 1970 Census: Demographic Analysis Current Population Reports Series P 23 No. 63, U.S. Department of Commerce, Bureau of the Census, Washington, D.C., 1977.

States to seek work. Estimates of the number of such illegal immigrants varies by several orders of magnitude, depending on the assumptions of those making the estimates.

The problem of illegal aliens has many other implications. For example, there is a clear privacy and confidentiality problem. While the Census Bureau is required by law to maintain the confidentiality of the data collected regardless of the legal status of the respondent, knowledge of that illegal status can present a problem to the Bureau. There have been unhappy coincidences (for example, in Austin in 1976 and in San Diego in 1970) between the collection of data by Census and deportation "raids" by the Immigration and Naturalization Service (INS). The coincidence of such activities is perceived in some quarters as a violation of Census Bureau confidentiality and thus could conceivably damage the continued collection of valid data particularly through the decennial census.

Finally, at present, it is not possible to examine births and deaths by ethnic origin. Racial identification has historically been included on vital certificates (although the validity of these data is open to question). Ethnic background, on the other hand, has not generally been requested. To the extent that we need estimates and projections for specific ethnic groups, particularly the Spanish-origin population, ethnicity will have to be included on vital records. It should not be assumed that this is an easy task. Problems resulting from a multiethnic background, recorders not having physical contact with the person of interest, and so forth, must yet be faced.

It must be pointed out that the registration of births and deaths is not primarily a Federal function. Although the NCHS promulgates standard certificates, each State is free to accept or reject all or part of the certificate. While NCHS has made a serious effort to encourage States to include Hispanic ethnicity, the distribution of the Hispanic population within relatively few States reduces the motivation of other States to require these data. However, by 1980 it is estimated that Hispanic ethnicity will be included on the vital certificates in States containing about 85% of the population. This is an excellent start but efforts must be maintained to include the balance of that ethnic group.

Unlike the problem of Puerto Rican migration and illegal aliens, identification of ethnic origin on the vital certificates does not affect any intercensal estimates or projections of the total population, but only the characteristics of that population.

There is a natural tendency for people to want information about groups to which they belong so

that they can examine changes and compare their group status and performance with that of other groups. This tendency has been demonstrated recently by the increased demand for data about various ethnic groups within the population. Within the past few years several resolutions have been introduced in the Congress which would require improving and expanding social and economic data on these groups. One such resolution on the Hispanic population has already passed and been signed into law (Public Law 94-311). Similar resolutions have been introduced for Americans of Balto-Slavic origin, Americans of Philippine origin, and East Asian and Pacific Island Americans.

In response to the perceived need illustrated by these resolutions, the Census Bureau has agreed to produce data from the 1980 census for virtually all ethnic groups. This will not, however, satisfy the intent of these resolutions to obtain current data. In order to produce such current data a manifold increase in the size of several sample surveys would be required. For example, to estimate the annual average unemployment rate of the Philippino population (assumed to be approximately 500,000, homogeneously distributed throughout the population) with reliability of a relatively modest 10% coefficient of variation would require a monthly CPS sample of 100,000 households compared with 50,000 households currently interviewed. Similar increases would be required for other surveys. It is difficult to see how the costs involved in such increases could be justified.

This entire issue should be studied and a national policy developed with respect to the provision of separate statistical data for ethnic minorities in the United States.

A similar serious problem involves the identification and count of American Indians. Many Federal programs currently identify Indians differently. These program differences make it impossible to provide denominators for Indian programs. A major effort is needed to standardize the programmatic and statistical definitions of Indians. There is a concern with the statistical systems being able to do this because of different tribal definitions and the legal implications of modifying definitions.

A resolution has also been introduced recently concerning data for women in the society. Much labor force participation data is not published for women because of low participation rates in various occupations and the resulting lack of reliability in the data. Census should prepare an annual report on women in the society using annual average data much as they prepare an annual report on the status and progress of the black and Hispanic populations.

New Requirements

Given both the initiatives of the Administration and the Congress to decentralize the authority and responsibility for domestic programs, timely, valid data are needed on the population of the country. The President's new urban initiative announced in March, 1978, for example levies heavy responsibilities on States and municipalities to develop urban programs. In order to accomplish this, detailed data will be needed for small areas. These data can only be developed through a program of estimates, since annual censuses are clearly inappropriate. The increased reliance on estimates over the years has led to the awareness that estimates need to be benchmarked more frequently than every 10 years.

This need for more frequent data led, in 1976, to the enactment of a law authorizing a mid-decade census. Although the mid-decade effort is referred to in the legislation as a census, the legislation clearly states that statistical sampling can be used. A variable sample survey could satisfy the needs of the Federal agencies.

Various programs have vastly different reliability requirements that should be reflected in the sample design, and those requirements have major implications on resource requirements. For example, a 1 percent coefficient of variation (CV) on the total population for all places would require a sample in excess of 42% of the entire population. On the other hand, a 10% CV level of reliability on total population for all places would require only a sample of 13% of the total population. (See table.)

The levels of reliability needed would have to depend on the requirements which will become obvious in the early part of the 1980's. Another way in which the burden and cost of the mid-decade effort can be minimized while meeting the needs of the agencies is through the utilization of "nested samples." A fuller description of this technique can be found in the Framework chapter on standdards for statistical methodology.

In order for a mid-decade census to eliminate other data collection efforts, especially for small area data, an intensive and extensive program to develop methods of generating small area estimates in a number of subject matter fields is needed. For example, if multivariate estimation procedures could be developed for housing quality, some of the Annual Housing Survey in several Standard Metropolitan Statistical Areas, conducted for the Department of Housing and Urban Development, could be eliminated. This effort should not be limited to housing but extended to other fields such as health, crime and education.