Age at First Marriage (Volume II, Part IV - Subject Reports)
Collection and Processing of Data
Collection of Data
Several enumeration forms were used to collect the information for the 1960 Census of Population. A few days before the census date, the Post Office Department delivered an Advance Census Report (ACR) to households on postal delivery routes. This form contained questions which were to be answered for every person and every housing unit. Household members were requested to fill the ACR and have it ready for the enumerator. The census enumerator recorded this information on a. form specially designed for electronic data processing by FOSDIC (Film Optical Sensing Device for Input to Computer). The information was either transcribed from the ACR to the complete-count FOSDIC schedule or entered on this schedule during direct interview.
In the densely populated areas, the enumerator left a Household Questionnaire to be completed by each household (or person) in the sample and mailed to the local census office. The population and housing information was transcribed from the Household Questionnaire to a sample FOSDIC schedule. When the Household Questionnaire was not returned or was returned without having been completed, the enumerator collected the missing information by personal visit or by telephone and entered it directly on the sample FOSDIC schedule. In the remaining areas, when the enumerator picked up the ACR, he obtained all the information by direct Interview and recorded it directly on the sample FOSDIC schedule.
Soon after the enumerator started work, his schedules were examined in a formal field review. This operation was designed to assure at an early stage of the work that the enumerator was performing his duties properly and had corrected any errors he had made.
More detailed descriptions of the 1960 procedures in the collection and processing data are given in reports entitled. United States Censuses of Population and Housing, 1960: Principal Collection Forms and Procedures, 1961; and Processing the Data, 1962, U.S. Government Printing Office, Washington, D.C., 20402.
After the FOSDIC forms had been checked for completeness in the field, they were sent to a central processing office for Manual Editing and Coding and for microfilming. Except where some special problems arose, there was no manual coding of the FOSDIC forms for complete-count data. On the sample forms, the manual operation was .limited to those items where coding .required the reading of written entries and therefore could not be done effectively by machine. The coding clerks converted the written entries to codes by marking the appropriate circles on the FOSDIC schedules and at the same time were able to correct obviously wrong entries and sometimes supply missing Information.
After the enumerators and coders recorded the information by marking the appropriate circles, the schedules were microfilmed. The information on the microfilm was then read by FOSDIC, which converted the markings to signals on magnetic tape. The tape, in turn, was processed in an electronic computer, which was used extensively to edit and tabulate the data and to produce the publication tables.
For a majority of items, nonresponses and inconsistencies were eliminated by using the computer to assign entries and correct inconsistencies. In general, few assignments or corrections were required, although the amount varied by subject and by enumerator.
The assignment of an acceptable entry by machine was based on related information reported for the person or on information reported for a similar person in the immediate neighborhood. For example, in the assignment of age in the complete-count tabulations, the computer stored reported ages of persons by sex, color or race, household relationship, and marital status, each stored age was retained in the computer only until a succeeding person having the same characteristics and having age reported was processed through the computer; this stored age was assigned to the next person whose age was unknown and who otherwise had the same characteristics. This procedure insured that the distribution of ages assigned by the computer for persons of a given set of characteristics would correspond closely to the reported age distribution of such persons as obtained in the current census.
The extent of the allocations for nonresponse and inconsistency for most characteristics is shown for the United States in 1960 Census of Population, Volume I, Characteristics of the Population, Part 1, United States Summary, tables B-1, C-2, and D-1.
Specific tolerances were established for the number of computer allocations acceptable for a given area. If the number was beyond tolerance, the data were rejected and the original schedules were reexamined to determine the source of the error. Correction and reprocessing were undertaken as necessary and feasible.
The information on date of first marriage was transferred from the microfilmed stage II FOSDIC schedule to the computer tape without prior manual Editing. No manual coding was necessary because the information was precoded on the schedules. A computer edit was used to eliminate inconsistencies and to allocate blanks. There were three consistency checks. In the first check, the computer inspected the entries for each ever-married person to insure that the recorded date of first marriage was at least 14 years later than the person's date of birth and that it was not later than March 1960. All dates of marriage failing to meet this requirement were treated as if they were blank. The other two consistency checks related only to married couples. First, if each partner had been married only once, the dates of first marriage had to be the same. If they were not, the earlier of the two dates was changed to agree with the later one. Second, the date of first marriage for a spouse married more than once had to be earlier than the date of first marriage for a spouse married only once. If it was not, the date of first marriage for the person married more than once was treated as if it were blank.
The procedure for allocating date of first marriage when the person and his spouse had been married only once was to assign to the person the date of first marriage of the spouse. For the remaining allocations, three matrices were used. One matrix was used to allocate date of first marriage for married persons, spouse present, where one partner was married once and the other more than once, and where the spouse's date of first marriage was known. In this matrix the difference in years between the dates of first marriages of previously recorded spouses was stored in 10 cells according to the age of the husband. If the person was married more than once, his date of first marriage was obtained by subtracting this difference from the spouse's marriage date. If the person had been married only once, a date of marriage was obtained for him by adding this difference to the spouse's date of first marriage.
A second matrix, of six cells, was used to allocate the date of first marriage of ever-married women 14 to 44 years old with one or more children in the household. The difference between the number of years a previously recorded woman in this category had been married and the age of her oldest child was stored in this matrix according to the woman's own age. When a date of first marriage was missing, the last stored value was subtracted from the date of birth of the woman's oldest child to obtain, a date of first marriage to allocate. If the woman was married, husband present, and the date of the husband's first marriage was also unknown, and if both husband and wife had been married only once, the husband was assigned the same date.
A third matrix of 104 cells was used to allocate date of first marriage for all other persons. The number of years married was stored by age, sex, and whether married more than once. The last stored value was subtracted from 1960 to obtain a date to allocate for the person's year of first marriage. The quarter of the year in which the marriage occurred was then assigned by random allocation. If the person was a married woman whose husband's date of first marriage was also unknown, and if both, she and her husband had been married only once, her husband was assigned the same date.
It would have been possible under this system for the allocated date of first marriage to be inconsistent with the person's age or the date of the census, i.e., either less than 14 years later than the person's date of birth or later than March 1960. The principal means used to eliminate the first type of inconsistent allocation was to advance the date of first marriage beyond the date of the person's fourteenth birthday by the j amount that the date originally allocated was less than the date of the person's fourteenth birthday; where this yielded a date later than March 1960, a date of first marriage of January to March 1960 was assigned. The principal means used to eliminate the second type of inconsistent allocation was to subtract from January-March 1960 the amount by which the originally allocated date of first marriage was later than January-March 1960.
Despite the precautions taken to eliminate unacceptable dates of first marriage, a relatively small number remained after the edits described above. These were reallocated before tabulation of the data in the present report and those in report PC(2)-4E, Marital Status. Dates of first marriage were reallocated for about 316,000 persons, or three-tenths of one percent of all persons ever married. The persons with date of first marriage reallocated are included in the total number of persons with an allocated date of first marriage shown in tables A to D .and in appendix tables A-1 to A-3, and constitute about U.8 percent of that total. Because these reallocations were made after tabulation of data for reports PC(2)-3A, women by number of children ever born, and PC(2)-4A, Families, there are some minor differences between the statistics In those reports and the statistics in the present report and in PC(2)-4E, Marital Status.
The extent of allocation for nonresponse on date of first marriage is described in the section on "Quality of the Data," above. The figures shown there do not include the 0.2 percent of the population for whom data, were replicated because there were no sample data for them; for these persons, the characteristics shown are those for persons in the substitute households.
Human and mechanical errors occur in any mass statistical operation such as a decennial census. Such errors include failure to obtain required information from respondents, obtaining inconsistent information, recording Information in the wrong place or incorrectly, or otherwise producing inconsistencies between entries on interrelated items on the field documents. Sampling biases occur because sortie of the enumerators fail to follow the sampling instructions. Clerical coding and editing errors occur, as well as errors in the Electronic Processing operation.
Careful efforts are made in every census to keep the errors in each step at an acceptably low level. Review of the enumerator's work, verification of manual coding and Editing, checking of tabulated figures, and Ratio Estimation of sample data to control totals from the complete count reduce the effects of the errors in the census data.
Very minor differences between tables in this report or between corresponding data in this report and PC(2)-4B, Marital Status, and other reports based on the 5-percent sample, result from imperfections in the electronic equipment. No attempt has been made to reconcile these minor discrepancies.
Some innovations in the 1960 Censuses reduced errors in processing and others produced a more consistent quality of Editing. The elimination of the card-punching operation removed one important source of error. The extensive use of electronic equipment insured a more uniform and more flexible edit than could have been accomplished, manually or by less intricate mechanical equipment. It is believed that the use of electronic equipment in the 1960 Censuses has improved the quality of the Editing compared with that of earlier censuses but, at the same time, it has introduced an element of difference in the statistics.
A group of reports designated Evaluation and Research Program of the U.S. Censuses of Population and Housing, 1960, deals with the methods, results, and interpretation of a group of evaluation and research studies of the 1960 Censuses of Population and Housing. A report entitled The Post-Enumeration Survey: 1950, Technical Paper No. 4, presents evaluative material on the 1950 Census.