No information was obtained on the characteristics of the population abroad such as is available from the schedules employed in the 1950 Census. In this bulletin the term "United States" when used without qualification refers to the 48 States and the District of Columbia and excludes outlying Territories, possessions, etc. Sometimes, however, the United States in this sense is referred to as "continental United States." The Census of 1890 was the first at which a complete enumeration was made of the area now comprised within the boundaries of the 48 States and the District of Columbia. Indians living in Indian Territory or on reservations were not included in the population count until 1890, and at earlier censuses large tracts of unorganized and sparsely settled territory were not canvassed by the enumerators. Thus, the sum of the areas enumerated was not always identical with the area included within the legal boundaries of the United States at the respective dates, nor was it always possible to indicate the exact boundaries of the enumerated areas. In the earlier censuses not all of a State or territory was covered by the enumerators but only that part up to the "frontier line" and any large isolated settlements beyond. For example, Iowa Territory in 1840 included all of what is now Iowa and most of what is now Minnesota, but within the Territory the only substantial settlements were in the southeastern corner of what is now Iowa, and hence only this part was covered by the Census of 1840. It is not feasible to make a more exact statement than that the area of what is now Iowa was added to the area of enumeration in 1840. The western part of what is now Minnesota, however, was not included until later. The Census of 1790 covered areas now embraced in the District of Columbia and the following States: Maine, New Hampshire, Vermont, Massachusetts, Rhode Island, Connecticut, New York, New Jersey, Pennsylvania, Delaware, Maryland, Virginia, West Virginia, North Carolina, South Carolina, Georgia, Kentucky, and Tennessee. Large areas in some of these States, however, were not covered in the enumeration. Only about one-fourth of the area of Georgia, for example, was enumerated.1 The area added at each census to the area of enumeration within the boundaries of continental United States may be briefly indicated as follows: 1800. The area now constituting the States of Ohio, Indiana, Illinois, Michigan, Wisconsin, and the south central parts of Alabama and Mississippi. In that year the area now within the States of Illinois and Wisconsin and a part of the present area of Michigan were included in the Territory of Indiana; and three years later, when Ohio was admitted to the Union as a State, the remainder of the present area of Michigan was added to Indiana Territory. The population shown for Indiana Territory in 1800 was substantially that residing within the present limits of Indiana, Illinois, Michigan, and Wisconsin. The population shown for Mississippi and Alabama in 1800 was that residing within Mississippi Territory as then constituted, which embraced the area now forming the south central parts of the States of Mississippi and Alabama. 1810. The area now constituting Arkansas, the northern parts of Mississippi and Alabama, and all but the southwestern part of Louisiana and the northwestern part of Missouri. (The remainder of the Louisiana Purchase of 1803 was not enumerated in 1810.) The population shown for Mississippi and Alabama for 1810 included that residing within Mississippi Territory as then constituted. 1820.-The extreme southern parts of Alabama and Mississippi, and the southwestern part of Louisiana. Florida was purchased in 1819, but was not enumerated in 1820. 1 For maps showing the distribution of the population at each census from 1790 to 1910, see U. S. Bureau of Census, Statistical Atlas of the United States, U. S. Government 1830.-Florida. 1840.-Iowa, northwestern Missouri, and northeastern Minne sota. 1850.-Texas, Utah, California, that part of New Mexico Territory now constituting the State of New Mexico with the exception of a small portion of the Gadsden Purchase of 1853, and that part of the Territory of Oregon now constituting the States of Oregon and Washington. 1860.-Dakota Territory (organized in 1861 from the area now embraced within the States of North and South Dakota and those parts of Montana and Wyoming lying east of the crest of the Rocky Mountains and north of the forty-third parallel), the remainder of Minnesota, Nebraska (then including that part of the area now constituting Wyoming which lay south of the forty-third parallel and east of the Rocky Mountains), Kansas, Colorado, Nevada, that part of Washington Territory now constituting Idaho and those portions of Montana and Wyoming lying west of the Rocky Mountains, that part of New Mexico Territory now constituting the State of Arizona (including the greater portion of the Gadsden Purchase of 1853), and that part of the Gadsden Purchase which now forms the southwestern part of New Mexico. The population shown for Washington Territory for 1860 was that residing within the limits of the Territory as then constituted, which embraced the area of the present States of Washington, Idaho, and western Montana and Wyoming. 1870 and 1880.-No change. 1890.-Indian Territory and Oklahoma Territory (later combined to form the State of Oklahoma) and Indian reservations. 1900-1950.-No change. Alaska was first included in a Federal decennial census in 1880, Hawaii in 1900, Puerto Rico in 1910, and American Samoa, Guam, and the Canal Zone in 1920; but a special census of Puerto Rico had been taken in 1899 under the direction of the War Department, and a special census of the Canal Zone, had been taken in 1912 by the Department of Civil Administration of the Isthmian Canal Commission. The Virgin Islands of the United States were first enumerated in a regular decennial census in 1930. A special census, however, had been taken as of November 1, 1917, immediately after purchase of the islands by the United States. Since the Republic of the Philippines was established as an independent country in 1946, the islands were not covered in the 1950 Census. The Philippine Islands had never been enumerated at a decennial census. A special census of the archipelago was taken in 1903 by the Philippine Commission and censuses were taken in 1919 (as of December 31, 1918) and in 1939 (as of January 1, 1939) by the Philippine government. COMPLETENESS OF ENUMERATION The degree of completeness of enumeration has always been a matter of deep concern to the Bureau of the Census; and, in the course of its history, a number of devices have been developed to aid in securing adequate coverage. These devices include the special procedures for the enumeration of transients and infants, urging notifications from persons who believed that they may not have been enumerated, and the early announcement of population counts in local areas to make possible the thorough investigation of complaints as to the accuracy of the count. In the 1950 Census earlier procedures were strengthened and additional procedures were introduced. Adequate handling of the problem of underenumeration involves not only the development of techniques in order to insure satisfactory coverage but also methods of measuring the completeness of coverage. Prior to 1950, no method had been devised to give an over-all direct measure of the completeness of enumeration of the total population. For the most part, discussion in census reports was Figure 3.-MAJOR ACQUISITIONS OF TERRITORY BY THE UNITED STATES [For other areas under the jurisdiction of the United States, see table 11 evidence. In the 1950 Census, the population of all ages was re-enumerated on a sample basis in a carefully conducted postenumeration survey; thus a direct check on a case-by-case basis of the actual enumeration was afforded. The results of this survey indicate a net underenumeration in the census count of the total population of the United States of about 2,100,000, or 1.4 percent. Procedures to Improve Coverage From the earlier discussion of usual residence and date of enumeration, some of the difficulties involved in obtaining a complete and unduplicated count of the population should be clear. It may safely be said that no national census, either in the United States or abroad, has ever represented an absolutely accurate count. Experience had shown that many devices might be used to improve the completeness of coverage. The major ones used in the 1950 Census of Population were: 1. A longer and better planned period of training was provided for enumerators. The enumerators were paid while taking a three-day course of training that emphasized the importance of an accurate count, the kinds of people who tended to be missed, and how to discover them. A Training Guide for the instructor, film strips, records, and practice enumeration were among the devices used. 2. Each enumerator was furnished with a map of his enumeration district, showing the boundaries of the area for which he was responsible. 3. An "Infant Card" had to be filled for each baby born after January 1, 1950, since experience had shown that babies are easily missed. Enumerators received 7 cents extra for each infant card filled. 4. A crew leader was assigned to supervise each group of approximately 15 enumerators. His duties included helping enumerators with problem cases and spot-checking a sample of the dwelling units assigned to them. 5. A special enumeration of persons in hotels, tourist courts, and other places where transients usually pay for quarters was made the night of April 11. When transients claimed a usual place of residence elsewhere, records were compared to ensure that they were counted once and only once. 6. "Missed Person" forms were published in the newspaper at the end of the field canvass so that persons who thought they had been missed could fill them out and mail them to the district supervisor. 7. District supervisors made preliminary announcements of the population counted so that any complaints or criticisms concerning the completeness of the enumeration could be submitted before field offices were closed. If the evidence, usually in the form of lists of names and addresses of people believed missed, seemed to indicate appreciable underenumeration, a re-enumeration was made of the affected area. In this country, the length of the enumeration period, the high degree of population mobility, the difficulty of finding many dwelling units, the living habits of apartment dwellers and lodgers in our metropolitan centers, and the inexperience of most of the enumerators, all represent relatively great problems. In some foreign countries, the canvass is completed in a day or so by means of a radically different organization of the field work. The existence of a continuous population register, the use of selfenumeration, and the use of permanent Government employees as enumerators are factors that may make a quick canvass possible. In some foreign censuses, everyone must remain at home until the entire enumeration is completed or may move about on the streets only with some form of identification to prove that he has been counted. Even with such drastic interference with normal activities, some persons are missed. Of course, there are considerable differences among censuses with respect to completeness of enumeration, and these differences are due partly to differences in procedures. Accuracy in a census can be increased by using better procedures, but some procedures are so expensive that the improvement would not be worth the Indirect Methods of Evaluating Completeness of Coverage One of the simplest types of evaluation is obtained from the examination of rates of changes for a series of several censuses with respect to their consistency and reasonableness. For example, a comparison of figures for the Southern States among the Censuses of 1860, 1870, and 1880 shows unreasonably low rates of increase for the decade 1860 to 1870 and abnormally high rates of increase for the decade 1870 to 1880. These differences are of such a magnitude that it appears evident that the enumeration of 1870 in these areas was seriously incomplete, undoubtedly as a result of the unsettled conditions of the reconstruction period. In terms of the total population for the United States as a whole, the number initially enumerated was 38,558,371; whereas a later revised figure, taking into account the underenumeration in the Southern States, put the total population of the United States at 39,818,449. For the portion of the United States outside the South, the rate of increase for the decade 1860 to 1870 was almost exactly the same as for the decade 1870 to 1880. Therefore, the figure for the South for 1870 was revised on the assumption that the rate of increase during these two decades was the same. Another method of estimating the comparative completeness of successive censuses involves the use of vital statistics and immigration statistics in conjunction with census data. Since the population at a given census should represent the population at the previous census plus births and immigration and minus deaths and emigration in the intervening period, it is possible, given the necessary statistics, to calculate the expected population on a given census date and to compare it with the enumerated population. If this comparison shows that the expected population exceeds the enumerated population, it may be inferred that the net amount of underenumeration in the current census exceeded that in the previous census; if, on the other hand, the enumerated population exceeds the expected population the inference is that the current census is the more complete one. These inferences, of course, rest on the assumptions that the error in census counts is always in the direction of net underenumeration and that errors in the measurement of births, deaths, immigration, and emigration are small in relation to the amounts of comparative underenumeration. The application of this method and assumptions to the decade 1940 to 1950 results in an estimate that the 1950 count was more complete by some 100,000 than that of 1940. For the decade 1930 to 1940, application of the method suggests that the total net number of persons missed in 1940 may have been about 1,300,000 more than that missed in 1930. The components of population change were probably estimated more accurately for the forties than for the thirties because not all States were in the birth and death registration areas until 1933 and because the registration of vital statistics within these areas has been increasing in completeness. Allowances were made for these factors in the case of births and in the case of infant deaths, but these estimates may be still subject to considerable error. No adjustment was made for underregistration of deaths, other than infant deaths, although some deaths of older persons were not registered. International migration was the smallest component of population change in these decades, but the figures were probably subject to the greatest relative error. In view of these considerations, the result of 100,000 is so close to zero that we cannot be sure whether coverage was more adequate în 1940 or in 1950. The comparison of the expected with the enumerated population provides figures only on the difference between the amounts of total net underenumeration at two censuses. If, however, one of these totals can be estimated, then it is possible to specify total net underenumeration for each census linked together by this method. As described more fully below, the Post-Enumeration percent. On this basis, the percentage of net underenumeration would be 1.6 in 1940, and 0.7 in 1930. Since the expected population under 10 years of age at a given census can be derived from the number of births in the preceding decade, a comparison of the expected and enumerated population gives a direct measure of total net understatement. On this basis, it is estimated that the total net amount of understatement, including both underenumeration and the misreporting of age, in the age group under 10 years was about 1,300,000 in the 1950 Census and about 1,500,000 in the 1940 Census. The corresponding percentages are 4.3 and 6.7. Despite the possible errors in the estimates of births, deaths, and migration used in determining the expected population, it is felt that the indicated difference between net understatement in 1940 and 1950 is in the true direction. Comparisons of census data with independent counts of corresponding segments of the population are sometimes possible in the case of certain other age-sex groups. For example, there have been several studies for both World War I and World War II relating figures for males of military age from the census to registration figures. Here again, however, interpretation of the differences is complicated by the fact that there are no adequate measures of the accuracy of the Selective Service figures. There is, in fact, some evidence of overreporting in these figures, which were compiled by local boards with little statistical supervision. Nonetheless, these studies do suggest an appreciable underenumeration of males in the appropriate age groups in the Censuses of 1920 and 1940, particularly among Negroes. Post-Enumeration Survey A particularly important and useful method of checking the adequacy of enumeration is a direct check on a case-by-case basis of the actual enumeration. A procedure of this type was used in the Post-Enumeration Survey of the 1950 Census, in which a re-enumeration on a sample basis was undertaken. To check for entire households erroneously omitted from the census, a probability sample of about 3,500 small areas was recanvassed and the relistings carefully compared with the original census listings. In addition to the check for erroneously omitted households, a sample of about 22,000 households was reinterviewed to determine the number of persons erroneously omitted in cases where the household had been included. This sample of households was also used to determine the number of persons erroneously included in the census listings and the accuracy of the reports obtained on the characteristics of enumerated persons. Available preliminary evidence of the quality of the census data on characteristics of the population, as revealed by the Post-Enumeration Survey, is presented in connection with the discussion of the specific characteristics in the sections which follow. Great The Post-Enumeration Survey interviewers were carefully selected and were given intensive training and supervision. efforts were made to limit respondents to the person who was presumably best informed regarding the information desired-usually, the person himself. These precautions resulted in an expenditure per case in the Post-Enumeration Survey many times that of the original enumeration-an expenditure which was feasible only because the study was done on a sample basis. A full description of the procedure and results of this Post-Enumeration Survey will be published at a later date. As indicated in table A, the net underenumeration in the census oint of the total population of the United States is estimated at 1.4 percent (with a standard error of 0.2 percent). The net underenumeration is the difference between the erroneous omissions and the erroneous inclusions. The figures shown in table A represent those errors in the census count which were detected by the Post-Enumeration Survey. Errors not reflected in these Szures may have arisen because of the following factors, among 1. In the check for erroneously omitted persons, large nondwelling-unit quarters (i. e., those where 35 or more persons had been enumerated), such as hotels and other accommodations for transients, were excluded. A separate check on those accommodations was undertaken, but the results of this study are not yet available. 2. Identifying all errors in the census coverage is extremely difficult. Although some of the errors in the census listings came from carelessness or ineptness of the enumerators, many of them are a result of the intrinsic difficulty of enumerating certain types of persons—for example, persons with no fixed place of residence. The Post-Enumeration Survey interviewers did succeed in locating many of the persons who were missed or erroneously included in the census, but they could not identify all such cases. A smallscale field check on the Post-Enumeration Survey results indicates that the Post-Enumeration Survey errors were, in gereral, in the direction of underestimating the number of erroneously omitted persons. This conclusion is also supported by examining the Post-Enumeration Survey figures in the light of other evidence on errors. For example, estimates of children under 5 based on records of births, deaths, and migration point to a shortage in the census figure for this age group of considerably greater magnitude than that reported by the Post-Enumeration Survey; again, although the Post-Enumeration Survey indicates, as had been expected, a greater error in the enumeration of the nonwhite population than the white population, it shows less error for the nonwhite population in the age group 15 to 24 than for other nonwhite age groups, a difference which might possibly be valid but is more likely attributable to the difficulty experienced by both the census and the Post-Enumeration Survey in listing the most mobile sectors of the population. 3. The reliability of these estimates, as in all statistical surveys, is also affected by errors in the application of sampling and other procedures. Those errors in the Post-Enumeration Survey which could be identified were almost all in the direction of underestimating the number of persons erroneously omitted from, or erroneously included in, the census, with probably more erroneous omissions than inclusions. These and other considerations suggest that the estimated net underenumeration of 1.4 percent in 1950 is a minimum estimate. As indicated in table A, there is some variation with residence in the coverage error of the census. In general, the net underenumeration was somewhat greater in rural than in urban areas and it was somewhat greater in the South and West than in the other regions of the United States. Among urban areas the error |