Research in Population Problems, on loan to the Bureau of the Census.*

Relation to Standard Metropolitan Areas

The combination of counties into State economic areas has been made for the entire country, and in this process the larger standard metropolitan areas (those with a population of 100,000 or more in 1940) have been recognized as metropolitan State economic areas. When a standard metropolitan area is located in two or more States, each State part becomes a metropolitan State economic area. In New England this correspondence does not exist because State economic areas are composed of counties whereas standard metropolitan areas are built up from towns. Here, counties with more than half their population in a standard metropolitan area are classified as metropolitan. Likewise, because standard metropolitan areas were set up on the basis of the 1950 population and State economic areas on the basis of the 1940 population, there are some standard metropolitan areas which are not recognized in the State-economic-area classification and one metropolitan State economic area (Michigan C) is not recognized in the standard-metropolitan-area classification.

Uses

In the establishment of State economic areas, factors in addition to industrial and commercial activities were taken into account. Demographic, climatic, physiographic, and cultural factors, as well as factors pertaining more directly to the production and exchange of agricultural and nonagricultural goods, were considered. The net result then is a set of areas, intermediate in size between States, on the one hand, and counties, on the other, which are relatively homogeneous with respect to a large number of characteristics. Areas of this type are well adapted for use in a wide variety of studies in which State data are neither sufficiently refined nor homogeneous and in which the manipulation of county data presents real difficulty. Moreover, a standard set of areas, such as these, makes possible studies in widely different fields on a comparable area basis.

Economic Subregions

These areas represent combinations of State economic areas. By this combination, the 501 State economic areas are consolidated into a set of 119 areas which cut across State lines but which preserve to a great extent the homogeneous character of the State economic areas. The economic subregions are perhaps best adapted to those analyses of the geographic distribution of characteristics of the population within the country as a whole in which there is no need for the recognition of State boundaries and in which the greater refinement permitted by the larger number of areas is desirable. The publication of data from the 1950 Census in several fields has been planned for economic subregions.

For further discussion and materials on State economic areas and their uses, see U. S. Bureau of the Census, State Economic Areas, U. 8. Government Printing Office, Washington, D. C., 1951.

Figures on the total population of economic subregions and their component State economic areas are presented in table 31. A map showing the boundaries of State economic areas and economic subregions appears on page 32. In table 19, which presents statistics for counties, each county is identified by the number or letter designation of the State economic area into which it falls. CENSUS TRACTS

Definition

Census tracts are small areas, having a population generally between 3,000 and 6,000, into which certain large cities (and in some cases their adjacent areas) have been subdivided for statistical and local administrative purposes, through cooperation with a local committee in each case. The tract areas are established with a view to approximate uniformity in population, with some consideration of uniformity in size of area, and with due regard for physical features. Each tract is designed to include an area fairly homogeneous in population characteristics. In cities where ward lines are infrequently changed, the tracts may form sub divisions of wards; but the tracts are usually laid out without regard to ward boundaries. The tracts are intended to remain unchanged from census to census and thus to make possible studies of changes in social and economic characteristics of the population within small sections of the city.

Areas Tracted in 1950

There are 12,633 tracts in the 69 tracted areas for which 1950 Census data are available on this basis. Tract data were tabulated for 8 cities in 1910 and 1920, 18 cities in 1930, and 60 areas in 1940. Figures on the total population by tracts have been published for each of the 1950 areas in Series PC-10, Advance Reports. The characteristics of the population and housing of census tracts will be published as Volume III, Census Tract Statistics, for all but a few of the 69 areas. Table 33 presents the number of tracts in the city and in the adjacent part of each tracted area.

SPECIAL CENSUSES

The Bureau of the Census has an established procedure for taking a special census at the request and expense of a local com. munity. Generally, the areas for which special censuses are taken are those which have experienced an unusual increase in population either because of changes in political boundaries or because of relatively high in-migration. The areas in which special censuses were conducted by the Bureau of the Census between April 1, 1940, and April 1, 1950, are shown in table 32; more than 400 separate special censuses were conducted during the decade 1940 to 1950.

The Bureau of the Census has published separately the results of these special censuses in varying detail in Current Population Reports, Series P-SC and P-28.

1 For a further discussion of census tract data and their uses, see U, S. Bureau of the Census, Census Tract Manual, 3d edition, January 1947.

Statistics on the characteristics of the population of the United States are presented in Chapters B and C of this volume. The following text summarizes the available evidence on the quality of the census information and provides, in graphic form, in figures 15 to 33, the major summary data based on Chapter B tables. Definitions of the pertinent concepts used in the 1950 Census are also given. Several of these definitions differ from those used in 1940. The changes were made after consultation with users of census data in order to improve the statistics, even though it was recognized that comparability would be adversely affected. In many cases the new definitions were tested in connection with the Current Population Survey; and, where feasible, measures of the

The schedule used in enumerating the population in the 1950 Census is reproduced in the Appendix. The Enumerator's Reference Manual supplied the census enumerators with instructions on the method of filling out the schedule and on the determination of what persons should be enumerated in their respective districts. The major part of these instructions is also reproduced in the Appendix.

QUALITY OF DATA ON CHARACTERISTICS

The 1950 Census is the first for which there is any substantial body of evidence on the quality of the statistics on characteristics of the population. This evidence is based on the Post-Enumera

Figure 14.-ECONOMIC SUBREGIONS AND STATE ECONOMIC AREAS: 1950

available these indications of quality are cited in connection with the discussion of the specific characteristic in the following pages. As a guide to users of the data, the text emphasizes the limitations of the statistics and the specific problems that might affect interpretation in certain types of analysis. In general, however, it appears that census enumerators were successful in obtaining sufficiently adequate reports concerning the characteristics of the population for most purposes for which census results are used. Many of the difficulties arose from failure to report accurately on unusual or complicated cases or to obtain information on minor activities.

The Post-Enumeration Survey provides measurements of total error in the census statistics on characteristics of the population that are made up of several parts: (1) Errors arising from faulty coverage of persons in the census; (2) errors arising from misreporting of a particular characteristic; and, for some subjects, (3) errors arising from misreporting of age when it is used as a control upon the reporting or tabulating of the given characteristic. The Post-Enumeration Survey, like other quality checks based on reinterviews, reveals considerable variability of response as between one interview and another. In general, the results suggest that the original census report frequently differed from the check enumeration report but for most characteristics these differences tend to cancel each other. As a consequence, the distributions as shown by the census data and by the PostEnumeration Survey are similar for most subjects. The percentage of either net or gross errors depends upon the number of class intervals or other categories chosen for comparing the distributions. Analysis of the effect of the errors on statistics showing cross-classifications is still incomplete, but will appear in the full report on the accuracy of the 1950 Census.

Several factors affect the interpretation of the measurement of differences between the census and the Post-Enumeration Survey. As stated earlier, the available Post-Enumeration Survey data did not include information for persons in hotels, in institutions, and other large nondwelling unit quarters. Such persons are typically difficult to interview and generally differ from persons in private households with respect to age, income, occupation, and other characteristics. Because of this factor, the degree of accuracy in census data indicated by the check survey may be somewhat overstated.

For various reasons the Post-Enumeration Survey did not start until the end of July 1950. The lapse of time between the original census and the check interview was of little consequence in evaluating the accuracy of such items as age, nativity, and education, but may affect the comparisons of the original and second reports on the job held during the census week or the recollection of income received during 1949. For example, some of the differences in descriptions of occupation and industry undoubtedly reflect actual differences in jobs and not variations in the descriptions of the same job.

Finally, the Post-Enumeration Survey estimates are based on a small sample, and in some cases, have relatively large sampling errors which do not permit detailed analysis of the data.

INTERNAL CONSISTENCY OF DATA

In the 1950 Censuses of Population and Housing, sampling procedures were used both in the enumeration, to supplement the information obtained from the total population, and in tabulation of the results. For a description of the sample, estimates of the sampling variability, and method of obtaining improved estimates, see the section on "Reliability of sample data." The statistics in Chapter B were based on complete counts of the population, except for those characteristics, such as school enrollment and income,

• For a further discussion of the plan of the Post-Enumeration Survey, see Marks, Mauldin, and Nisselson, "The Post-Enumeration Survey of the 1950 Censuses. A Case History in Survey Design," Journal of American Statistical Association, June 1953,

that were reported for only a 20-percent sample of the population. For Chapter C, only the tabulations relating to occupation and industry are based on complete counts. All the other tabulations are confined to the 20-percent sample. Accordingly, because of sampling variability and certain small biases, described in the sec tion on "Reliability of sample data," differences may be expected between figures obtained from the complete count and the corre sponding figures based on the 20-percent sample.

Some differences between figures for corresponding items in different tables are caused by errors in the tabulation processes. These errors include machine failure, loss of punch cards, and other types. (The net effect is a tendency toward slightly smaller counts of the same item in successive tabulations.) Experience has shown that in mass operations two tabulations of a set of punch cards are not likely to yield precisely identical results. Therefore, tolerance limits allowing for insignificant variations were established in advance for each tabulation. If the differences between the results of two tabulations fell within these limits, nothing was done to bring them into exact agreement with each other. This procedure was adopted in order to provide a greater volume of data within the limits of time and resources available. In earlier censuses, however, the results of different tabulations were adjusted to bring them into exact agreement.

The Chapter B tabulations for four States (Alabama, Iowa, Louisiana, and Virginia) were carried out on electronic equipment that differed from the conventional machines. Adaptation of the standard procedures for these States resulted in different treatment of certain types of cases. For example, some cards were erroneously included in the sample tabulations on education and on residence in 1949 and were classified with the group for whom information was not reported.

MEDIANS

Medians are presented in connection with the data on age, education, and income which appear in this bulletin. The median is the value which divides the distribution into two equal parts-onehalf of the cases falling below this value and one-half of the cases exceeding this value. In the computation of medians, cases for which the information was not reported are omitted. The median income for families and unrelated individuals is based on the total number reporting, including those reporting no income.

FARM POPULATION-RURAL AND URBAN

Definitions

The farm population for 1950 includes all persons living on farms, as determined by the question, "Is this house on a farm (or ranch)?" The enumerators were also instructed to classify persons living on what might have been considered farm land as nonfarm if they paid cash rent for their homes and yards only. Persons in institutions, summer camps, "motels," and tourist camps were classified as nonfarm residents.

Farm residence is therefore determined without regard to occupation. The classification depends upon the respondent's conception of what is meant by the word "farm," and consequently reflects local usage rather than the uniform application of an objective definition. For this reason, there is considerable variability of response among families living in areas where farm operation is part time or incidental to other activities.

The population living on farms has never, of course, been identical with the population dependent on agriculture; but there is some evidence that the number of persons dependent on agriculture has been decreasing more rapidly than the total number of farm residents. Some workers living on farms have both an agricultural and a nonagricultural job; others shift from agriculture to nonagricultural work with the season. Some farm households contain both an agricultural and a nonagricultural worker. Finally, some farm households do not contain any agricultural

farmers and farm laborers, do not live on farms. According to the 1950 Population Census, 27 percent of employed workers living on rural farms were working in nonagricultural industries. On the other hand, 18 percent of the workers employed in agriculture did not live on rural farms. These data on industrial attachment referred to only the job the worker held (or to the one at which he worked the greatest number of hours) during the week preceding his enumeration, as explained in the section on "Occupation, industry, and class of worker." Further evidence on this point is provided by data on major source of earnings over the period of a year. According to the Current Population Survey, about 35 percent of the families living on rural farms in April 1951 reported that nonfarm work was the major source of their earnings in 1950. Apparently the distinctions between the farm and nonfarm populations have become somewhat blurred, but large differences in way of life still persist and a majority of farm residents are dependent on agriculture for their livelihood.

In this volume, most of the data are presented for the ruralfarm population rather than for the total farm population, since virtually all the farm population is located in rural areas. Only 1.2 percent of the farm population lived in urban areas in 1950.

Comparability

In the 1940 and 1930 Censuses of Population, as in 1950, the farm population was defined to include all persons living on farms, without regard to occupation. In 1940, the decision as to whether the household lived on a farm was left to the enumerator and no specific definition given him. In 1930, the instruction to the enumerator read, "If the family lives on a farm, that is, a place for which a farm schedule is made out and which is also locally regarded as a farm, the answer should be 'Yes,' even though no member of the family works on the farm. It is a question here

there was no specific exclusion of persons paying cash rent for their homes and yards only, nor of persons living in institutions, summer camps, etc. Moreover, in the 1950 Census, but not in earlier censuses, the enumerators in rural areas were specifically instructed to base the farm-nonfarm classification on the respondent's answer to the question, "Is this house on a farm?" For the United States as a whole, there is evidence from the Current Population Survey that the farm population in 1950 would have been somewhat larger had the 1940 procedure been used, but the change in procedure accounts for only a minor part of the indicated change in the size of the farm population over the decade. Hence the 1950 and 1940 Census levels are only roughly comparable, and the indicated declines are probably somewhat exaggerated in most areas.

Estimates of the population living on farms have been provided by the Current Population Survey in the years following the 1940 Census. For April 1950, the revised estimate from the Current Population Survey was about 1,700,000 higher than the final 1950 Census count, although the same definition was specified for both enumerations. This difference was well in excess of the sampling error of the estimate. Examination of the returns for identical households in the two enumerations revealed that the largest part of the discrepancy (approximately 11⁄2 million) was due t differences in classification. Although there is no conclusive evidence on the relative validity of the Current Population Survey as compared with the census classification of particular households as farm or nonfarm, investigations in other subject-matter fields have demonstrated that the more experienced and bettertrained Current Population Survey enumerators were more likely to treat marginal cases in a correct fashion.

The extensive revisions required in the farm population series over the years and the large Current Population Survey-census difference reflect the serious nature of the measurement problem

the relative vagueness of the various definitions of farm residence which have been used in censuses and in enumerative surveys. The definition adopted for the 1950 Census was initially and may still be regarded as a significant improvement over those used previously. Regardless of its theoretical merits, however, it apparently does not embody the objectivity required to obtain uniform and consistent measurements. The whole question of the comparability of the census and Current Population Survey figures is dealt with in considerable detail in: U. S. Bureau of the Census and U. S. Bureau of Agricultural Economics, Farm Population, Series Census-BAE, No. 16, "Revised Estimates of the Farm Population of the United States: 1910 to 1950," March 9, 1953.

RURAL-NONFARM POPULATION

The rural-nonfarm population includes all persons living outside urban areas who do not live on farms. In 1940 and earlier, persons living in the suburbs of cities constituted a large proportion of the rural-nonfarm population. The effect of the new urbanrural definition has been to change the classification of a considerable number of such persons to urban. The rural-nonfarm population is, therefore, somewhat more homogeneous than under the old definition. It still comprises, however, persons living in a variety of types of residences, such as isolated nonfarm homes in the open country, villages and hamlets of fewer than 2,500 inhabitants, and some of the fringe areas surrounding the smaller incorporated places.

RACE AND COLOR Definitions

The concept of race as it has been used by the Bureau of the Census is derived from that which is commonly accepted by the general public. It does not, therefore, reflect clear-cut definitions

tionalities. Although it lacks scientific precision, it is doubtful whether efforts toward a more scientifically acceptable definition would be appreciably productive, given the conditions under which census enumerations are carried out. The information on race is ordinarily not based on a reply to questions asked by the enumerator but rather is obtained by observation. Enumerators were instructed to ask a question when they were in doubt. Experience has shown that reasonably adequate identification of the smaller "racial" groups is made in areas where they are relatively numerous but that representatives of such groups may be misclassified in areas where they are rare.

Color. The term "color" refers to the division of population into two groups, white and nonwhite. The group designated as "nonwhite" consists of Negroes, Indians, Japanese, Chinese, and other nonwhite races. Persons of Mexican birth or ancestry who were not definitely Indian or of other nonwhite race were classified as white in 1950 and 1940. In the 1930 publications, Mexicans were included in the group "Other races," but the 1930 data published in this report have been revised to include Mexicans in the white population.

Negro. In addition to full-blooded Negroes, this classification includes persons of mixed white and Negro parentage and persons of mixed Indian and Negro parentage unless the Indian blood very definitely predominates or uniess the individual is accepted in the community as an Indian.

Other races. This category includes Indians, Japanese, Chinese, and other nonwhite races.

Mixed Parentage

Persons of mixed parentage are classified according to the race of the nonwhite parent and mixtures of nonwhite races are gener