book would result in some improvement in coverage and in keeping biases in sample selection at a low level, even with the sample designation being done during the census canvass.

Callback and closeout procedures. --The callback procedure in the 1950 censuses depended a great deal on the discretion of the individual enumerator. The instruction in the 1950 enumerator's manual was the following:

"If, after several calls, you have not been able to get in touch with a suitable respondent, return to the dwelling unit and get as much information as possible from the best sources available. However, in preference to obtaining information from unreliable sources, it is advisable to leave an ICR Individual Census Report form for each member of the household with the request that each be filled out, sealed, and left with a neighbor."

For the 1960 censuses, explicit instructions were given the enumerators regarding the number of callbacks to be made and the action to be taken when no respondent could be found at the household after the prescribed number of callbacks. The 1960 census enumerator for the 100-percent phase of the work was instructed to obtain information from neighbors after three calls, but for the sample questionnaire the 1960 enumerator was instructed to obtain most information only from acceptable household members and, except for a few specified items, to leave the sample items on the schedule blank after the designated number of calls rather than to obtain the information from neighbors or other sources. The intent was to insure coverage for the 100-percent items by using neighbors or other sources where necessary, but to discourage obtaining the more complex sample information from neighbors.

The callback problem has been an important source of enumerator discontent and of difficulty in maintaining control over costs. The increased number of women in the labor market, the increased freedom from household chores arising from the introduction of labor-saving devices, and the increase in off-farm work by farmers and their families all result in greater difficulty in finding people at home. Determining the maximum number of callbacks to be made for any one household and giving precise instructions for closeout procedures were an effort to get the callback problem under control. The closeout procedures provided for hourly-rate enumerators to work out of the field office on some of the more difficult followup work instead of having the enumerators paid on a piece-rate basis followup indefinitely.

Quality control.--For the first time in a major U.S. census, a formal program of quality control was instituted for the field enumeration. (This program is described in chapter 7, The Enumeration.) It provided supervisory personnel with definite procedures for detecting and, when necessary, rejecting unacceptable work.

Other procedures considered.--A number of other changes in enumeration procedures were seriously considered. Two methods were under active consideration for increasing the completeness of coverage of housing units in the censuses: (1) the use of post-office personnel to examine the listings made by enumerators and to identify housing units that were omitted from the listings, or (2) the use of crew leaders to make an independent listing of all housing units in every enumeration district under their supervision, to compare later with listings made by the enumerators. Both methods were rejected because of budgetary considerations. The post-office method, however, was tested in a probability sample of areas throughout the United States as part of the 1960 evaluation program (see chapter 10, Evaluation and Research Program).

Processing and Publication Procedures

The 1960 population census was the first for which the final 100-percent counts required to be submitted to the President of the United States within 8 months after the census date were obtained by machine processing instead of hand counting.

As noted above, a major reduction in data-processing time and costs was achieved by collecting on a sample basis the information regarding those items which would require a separate clerical coding operation. Nonetheless, the coding of the sample information was the largest clerical operation involved in processing the data from the 1960 censuses. It had been discovered in processing the 1950 censuses that verifiers examining the codes assigned by clerks missed or failed to report a substantial proportion of the errors. Therefore, a marked change in concept and procedures for the quality control program of the coding work was introduced for the 1960 censuses. The system used for quality control of the 1960 coding required that samples of the schedules be coded independently by different clerks, and then a comparison was made of the codes assigned. For a description of this system, see the section on "Quality Control" in chapter 8, Processing the Data.

The other major innovations in procedures for processing and publishing the data were accomplished primarily with high-speed electronic equipment. All of them were adopted to reduce the time and cost of making the census statistics available, below what they would have been if 1950 census procedures had been used again, and also to achieve improvements in accuracy through a reduction in the

number of errors previously introduced into the data during the clerical operations. The innovations in data-processing equipment used were the following:

FOSDIC.--The use of the FOSDIC equipment required that the data be recorded on schedules designed for the purpose. These were microfilmed, and FOSDIC then transferred the information from the microfilmed schedules to magnetic tape for input to the computers. (See chapter 8, Processing the Data.) The microfilm-FOSDIC process eliminated the necessity for a card-punching operation which would have involved the rental or purchase of large quantities of equipment and the hiring and training of many punchcard operators. There was thus a considerable saving of both time and money involved. In addition, census pretest evidence indicated that the FOSDIC equipment would transcribe the data more accurately than would key punch operators. The output of the FOSDIC was magnetic tape that contained the information recorded by the census enumerators (and coders, in the case of the sample questionnaires), ready for direct high-speed input to the computers.

Computers.-- Electronic computers are much more versatile, more accurate, and faster than the traditional punchcard equipment. Their potential for making consistency checks on the reported data and for assigning values where no answers were recorded was an important consideration in their use. (As a worthwhile byproduct, the use of computers for this work made it possible to inform users of the statistics in the published reports of the extent to which inconsistencies were corrected and imputations for missing data were made.) The electronic equipment provided much tighter control over the reliability of the tabulation processes than processing by punchcard equipment, as well as higher speed processing at lower costs.

High-speed printers.--The use of high-speed printing equipment as an integral part of the electronic computer system also contributed to reducing costs and reducing the time between taking the censuses and issuing the results. The high-speed printer was programed and adjusted to print out final tables, which, after certain manual steps, were then used as publication copy for photo-offset printing of census reports.

DETERMINING THE DATA TO BE COLLECTED

General Method

Soon after the results of the 1950 censuses became available, the staffs of the Population and Housing Divisions as well as other parts of the Bureau began to study the use of the data from those and previous censuses and the requests for new information. They examined not only the extent to which the data were used but also the shortcomings revealed when the figures were applied to specific problems. New questions for the 1960 censuses were evaluated in terms of the need for information and the problems and costs that would be involved in getting it. Tests were made to see if respondents could provide the new information with reasonable accuracy.

Criteria for Inclusion of Items

A nationwide census is a large-scale, complex, and expensive operation. It involves sending a large body of temporary personnel, which has had only brief training in its duties, into every place where people may be living, to interview persons of widely varying educational, cultural, and language backgrounds. Its successful completion depends to a great extent upon keeping the questions simple enough that they can be understood by both the enumerators and the respondents, and few enough that the schedule can be completed in a reasonably short interview. On the other hand, a great many agencies and groups, both governmental and private, are interested in obtaining a wide variety of census data as a basis for their planning and policy for the next several years. Thus there are conflicting pressures on a census, both to keep the list of items short enough that the census operation is feasible and to expand it to meet as many needs as possible at the time that the interviewers are in the field.

Some of the general criteria for the selection of items of inquiry for the census, which have been adopted on the basis of more than a century and a half of experience in taking censuses and with the advice of many government, business, and professional groups, are the following:

1. The data are needed for public policy

2. The data serve the general interest by enabling a wide segment of private industry or the public to plan more effectively

3. The information can be obtained by one or two or at most a very few questions

4. The questions should be simple and unambiguous, so that both enumerator and respondent understand what information is wanted

It is preferable that there be enough variation in the information obtained on an item so that the same response is not given in almost every instance. For example, the question on literacy was dropped after the 1930 census because such a small proportion of the population was illiterate, and a question on educational attainment was substituted. If an overwhelming proportion of the answers are

the same, then the question essentially obtains information about the few exceptions. Also, it has been observed that if the enumerators receive the same response in almost all cases, as the enumeration progresses some of them may begin to enter that response on the schedule without asking the question.

The information should not be of a type such that much of the public would object strongly to giving responses, partly because the data obtained would be invalid. However, any question may be objectionable to some respondents, and the importance of the information has some bearing on this consideration. For example, the age composition of a population is so important for public planning and other purposes that a question on age is included even though difficulties are sometimes involved, and a body of knowledge has grown up about tendencies of respondents to exaggerate or understate age at various ages and in various situations.

Items of information will be of little use if the situations which they describe are greatly changed by the time the data are published or soon after. Small-scale sample surveys which can be taken and processed in a very short time, and can be repeated as frequently as necessary, are preferred for rapidly changing subjects.

Considerations of cost, anticipated reliability, and the effect on timing of the census program are also important in the selection of items.

Each item included in the censuses is chosen as a result of a balancing of the various factors involved and a determination based finally on judgment as to its need, suitability, cost, etc. There is no firm yardstick by which to measure the many complex considerations.

Population Items

Most of the questions asked in the 1960 Census of Population were essentially the same as in the 1950 census. There were, however, some important changes. A list of the population items in the 1960 census is shown in table 1. The principal differences from the 1950 items are summarized below.

100-percent and sample items.--The only population items collected for 100 percent of the population in 1960 were on relationship to head of household, sex, race, date of birth, and marital status. All the other population items were collected on a 25-percent sample basis.

New and revised items.--Because of interest in commuting patterns, two new questions were asked, on place of work and on means of transportation used in getting to work.

In the field of population mobility, a new question on length of residence was asked, on when the person moved into his present dwelling, and the question on migration was in terms of place of residence 5 years earlier, as it had been in the 1940 census, rather than 1 year earlier as in 1950. The item on place of residence was refined by the addition of a new question, for persons who had lived in a different house in 1955 and who named a city or town as their place of residence in that year, on whether the person had lived inside the city limits.

The data on educational characteristics were expanded by an additional question, for persons enrolled in school, on whether the school was public or private. Provision was made for the first time for separately identifying those persons who had attended 6 or more years of college. Information on enrollment was recorded for persons 5 to 34 years old in 1960 instead of persons 5 to 29 years old as in 1950; the extended age coverage for publication of enrollment data reflects the increasing number of persons in their early thirties who have been attending regular colleges and universities in recent years.

In the field of employment statistics, the coverage of the questions on occupation, industry, and class of worker was broadened to include persons not in the labor force at the time of the census who had worked at any time in the preceding 10 years. The addition of a new question, "for whom did he work," made possible the use of "Company Name Lists" by the coders. These lists included the names of all manufacturing employers of 50 or more employees and all businesses having 100 or more employees and the associated industry code for each, thus making the industry coding faster and more precise. A question on "year last worked" was included in the census for the first time in 1960, primarily to provide a means of evaluating the current applicability and significance of the inventory of occupational skills for those persons not currently in the labor force and of presenting basic data for analysis of the labor reserve. It provided some indication of the duration of unemployment for persons currently seeking jobs, at least for those who were regular members of the labor force. Finally, it served as a basis for editing the entries for some persons who were erroneously reported as "having a job but not at work" even though they had not actually worked for some time.

Data on mother tongue of the foreign-born were obtained in 1960. Data on mother tongue were collected in 1940 and in several preceding censuses but not in 1950. This question was included again to supplement the data on country of origin and to provide some information on language facility.

In order to help improve the quality of the age statistics, the 1950 question on age at last birthday was replaced by one on month and year of birth. It was believed that the use of self-enumeration coupled with the wording of the question in terms of date of birth would result in fewer errors in age reporting.

A question on date of first marriage was substituted for the 1950 question on duration of current marital status.

In the 1950 census, members of a household were classified in the farm population if the householder answered affirmatively the question "Do you live on a farm?" In 1960, the farm population included the people who lived on farms as defined by the twin criteria, developed for the census of agriculture, of acreage and value of sales of farm products. The questions used to determine whether a household was on a farm were in the housing portion of the schedule.

Essentially unchanged items.--Inquiries which were much the same or the same in 1960 as in 1950 were the following: relationship to head of household, sex, color or race, marital status, whether married more than once, place of birth, country of birth of parents, number of children ever born to women ever married, hours worked in preceding week, occupation, industry, class of worker, weeks worked in preceding year, income, and veteran status.

1950 items omitted in 1960.--The question on citizenship of the foreign-born was dropped. However, citizenship data were obtained in New York State, at the expense of the State, to meet State constitutional requirements for State legislative apportionment. The question on duration of unemployment was omitted in 1960, primarily because of the problem of timeliness of the data, and partly because of the high nonresponse rate in 1950. The question on whether a migrant previously had lived on a farm was dropped, partly because of problems in defining farm residence 5 years earlier.

New items considered but not included in 1960.--A great many new population items were suggested. The most seriously considered of those which were omitted from the 1960 census was a question on religion. The inquiry on religion was omitted because of the opposition of some persons and groups, as a matter of principle, to the inclusion of such a question in a decennial census for which replies are mandatory, and the possibility of delays to the whole census operation from court cases, or lack of cooperation in some geographic areas, or requests that answers to this and other questions be made voluntary.

A number of education items were proposed but not accepted for inclusion in the census. These included enrollment in adult education programs and in correspondence courses, academic degrees ever received and fields of study, and academic attainment in other than regular schools and colleges. These subjects were rejected as census items largely because pretests and other information had shown that there would be special problems in definition and classification, in recalling such information, and in reporting these types of data for other persons.

A proposal to obtain separate data on income from farming was omitted largely because the cost would have been more than seemed justified by the utility of the item. Also, the 1960 data do show the total income of persons living on farms, though this includes income from other sources.

A question on secondary occupations was omitted because of the very high expense that would have been involved since the responses would have had to be clerically coded into the detailed occupational classification if the data were to be of much use for analysis.

A question on medical insurance was omitted because of the high cost of obtaining and classifying the information; it was expected that many people would answer by naming a company or program or by saying their office carried some type of insurance for them but they did not know just what was covered.

An item on wealth was rejected because of problems of clarifying the concept enough to obtain the information with simple questions.

A question on blood type, desired particularly by health and civil defense authorities, was omitted because most people would not have known their blood type.

Separate identification of migratory farm workers was not attempted in the census because of severe problems in identifying them.

Housing Items

A list of the items in the 1960 Census of Housing, and the percent of housing units for which each item of information was collected, is shown in table 2, "Housing Items of Data Collected."