in Household Data
Griffith Feeney
The following list shows a small illustrative sample of person records, grouped by household, from a population census. Abbreviations and codes are given below, followed by notes on the data and the matching process.
HHID PID LNM AGE SEX RHH MS MCB FCB MCS FCS Own Children History 67 1 0 27 1 1 2 0 0 0 0 67 2 0 25 2 2 2 1 0 1 0 25 1 6 (0) 67 3 2 6 1 3 0 0 0 0 0 25 1 0 53 2 1 3 1 0 1 0 53 1 (1) 84 1 0 34 1 1 2 0 0 0 0 35 1 0 58 1 1 3 0 0 0 0 35 2 0 27 1 3 2 0 0 0 0 35 3 0 27 2 3 2 1 0 1 0 27 1 2 (0) 35 4 3 2 1 4 0 0 0 0 0 69 1 0 36 2 1 2 1 0 1 0 69 2 0 37 1 2 2 0 0 0 0 37 1 9 (0) 69 3 1 9 1 3 0 0 0 0 0 65 1 0 66 1 1 2 0 0 0 0 65 2 0 62 2 2 2 6 0 6 0 62 6 26 20 (4) 65 3 2 26 1 3 2 0 0 0 0 65 4 2 20 1 3 2 0 0 0 0 64 1 0 52 2 1 2 1 1 1 1 64 2 0 58 1 2 2 0 0 0 0 64 3 0 26 2 3 2 1 0 1 0 26 1 1 (0) 64 4 3 1 1 4 0 0 0 0 0 64 5 0 80 2 5 3 0 0 0 0 54 1 0 71 2 1 3 0 0 0 0 54 2 0 45 1 3 2 0 0 0 0 54 3 0 48 2 3 2 2 0 2 0 48 2 15 13 (0) 54 4 3 15 1 4 1 0 0 0 0 54 5 3 13 1 4 0 0 0 0 0 44 1 0 73 1 1 2 0 0 0 0 44 2 0 72 2 2 2 0 0 0 0 44 3 0 32 2 3 2 1 0 1 0 32 5 (0) 44 4 3 5 1 4 0 0 0 0 0 44 5 ? 9 2 4 0 0 0 0 0 50 1 0 64 2 1 2 0 1 0 1 50 2 0 69 1 2 2 0 0 0 0 50 3 0 25 2 3 2 0 1 0 1 25 3 (0) 50 4 3 3 2 4 0 0 0 0 0 Abbreviations and codes are as follows: Field Headings HHID: Household ID number PID: Person ID number LNM: Line number of mother (result of matching, not given by data) AGE: Age in completed years at time of census/survey SEX: 1-male, 2-female RHH: Relation to head of household, codes below MS: Marital status, codes below MCB: Number of male children born (women aged 15-64 only) FCB: Number of female children born (women aged 15-64 only) MCS: Number of male children surviving (women aged 15-64 only) FCS: Number of female children surviving (women aged 15-64 only) Relation to Head Codes 1-head 2-spouse 3-child/spouse of child 4-grandchild/spouse of grandchild 5-parent/spouse of parent 6-grandparent/spouse of grandparent 7-other relative 8-nonrelative Note: `Grandchild' includes grandchild, great grandchild, and any higher orders. `Grandparent' includes grandparent, great grandparent, and any higher orders. Marital Status Codes 0-not applicable (too young for marital status question=never married) 1-never married 2-currently married 3-widowed 4-divorced Own children histories are expressed as an ordered series of numbers between 0 and 99 giving the following information, where [...] indicates that the information ... is optional: age of woman number of children ever born [age of youngest own child] [age of next youngest own child] . . . [age of oldest own child] number of children to be imputed (0, 1, ...)
Examples
25 1 6 (0) denotes a 25 year old woman with one child ever born and a matched child aged 6, with no children to be imputed 53 1 (1) denotes a 53 year old woman with one child ever born and one child to be imputed
Reading Person Record Information. Each line shows information about one person, including an identification number for the household in which this person was enumerated (HHID), a person identification number (PID) obtained by serial numbering of all persons in a given household (observe the pattern of these numbers within households), the age (AGE) and sex (SEX) of the person, and so on. Blank lines are placed between household groups improve readability.
Information in the records must be read in conjunction with the abbreviation and codebook information on the preceding page. In the household whose ID number is 65, for example, the first person is a 66 year old man (SEX=1) who is the head of the household (RHH=1) and currently married (MS=2). Since this person is male, and the children ever born and surviving questions are asked only of women aged 15-64, the values in the remainder of the record are zero.
The second record in the same household is that of a 62 year old currently married woman, the spouse of the head of household, with six male children ever born (MCEB=6), all surviving (MCS=6). Observe that we know from the relation to head codes that these first two persons are man and wife.
The third record in the same household is for a 26 year old man who is either a child or a child-in-law of the head of household. This man is currently married, but there is no one in the household who could be this person's wife. Similar observations apply to the fourth record in the household.
The Matching Process The object of the match process in birth history reconstruction is to determine, for every person, whether or not this person's mother is present in the household in which this person was enumerated, and if so, which person in the household is this person's mother. This information may be coded in a single two column field `line number of mother'. Note that LNM is not part of the original census record. The process of matching is best illustrated by example (see below).
Following long established usage developed in connection with the own-children method of fertility estimation, a person living in the same household as his or her mother is referred to as an own child. All other persons are referred to as non-own children. Children born to a woman thus divide into three categories, own children, non-own children, and deceased children.
Household 67. This household includes three persons, a male head aged 27, his spouse aged 25, and a child of the head aged 6. Note the redundancy in this information. The relation to head code tells us that person number 2 is the spouse of person number one, whence both person 1 and person 2 should be (and are) coded as currently married (MS=2).
The relation to head code tells us that person number 3 is a child or child-in-law of person number 1. Since person number 3 is only 6 years old, however, the child-in-law possibility is effectively ruled out and we conclude that person number 3 is the child of person number 1. While this does not strictly speaking imply that person number 3 is also the child of person number 2, we observe that person number 2 has one male child ever born, consistent with person number 3 being the son of person number 2.
There are no persons in household 67 who could be the mother of the head or his spouse, whence line number of mother for these persons is coded `0' signifying `mother not present'. For person number 3, however, we code line number of mother `2'. While it is conceivable that this match is incorrect, it is most unlikely in the Chinese context. Making such a match in many thousands of cases will certainly result in occasional mismatches, but the information gained by the correct matches outweighs the error introduced by mismatches.
The identification of person number 3 in the household as a child of the woman who is person number 2 gives us a complete birth history for this woman. We know that it is complete because the children ever born information shows that the woman has only one child ever born, a male.
Form of Birth Histories. Because we use age at census to infer period of birth, the birth history information available from the census is limited to the `census year' of birth, where `census year' denotes a year beginning and ending at the census reference date. Census schedules sometimes include year and month of birth, of course, and using this information would provide more detailed birth history information. This has not been done in applications to date.
Though birth histories refer by definition to the dates/periods at/during which events occur, it is convenient in this context to represent birth histories in terms of age at census. For deceased persons, age at census refers to the age the person in question would be if alive at the time of the census. The birth history for the woman in household 67 may thus be expressed as
a 25 year old woman with one child ever born, a six year old own child, and (therefore) no deceased or non-own children.
Household 65. This household illustrates the problems resulting from the 1990 census relation to head code's confounding of biological and affine relationships. The first two persons in the household are a husband, age 66, and his wife, aged 62. The wife has six male and no female children ever born. The relation to head codes do not distinguish between child and child-in-law, however, and it is possible that either or both of these persons are in fact sons-in-law. This is unlikely, because the woman has so many sons, because co-residence of sons is culturally prescribed and statistically more common than co-residence of sons-in-law, and because the wives of the putative sons are not present. These are very particularistic reasons, however, and the case illustrates the sorts of complex judgment that matching requires.
Household 64. The 80 year old woman (person 5) is either the grandparent or the spouse of the grandparent of the head of the household, the 52 year old woman who is person 1. If the 80 year old woman is the head's mother, we would put LNM=5 for the head and LNM=0 for the spouse of the head. If the 80 year old woman is the head's mother-in-law, however, we would put LNM=0 for the head and LNM=5 for the spouse. Because we cannot distinguish these cases, we do not make a match and code LNM=0 for both the head and the spouse.
The 26 year old woman who is person 3 in household 64 is coded as a child or child-in-law of the head. The head is shown as having one male and one female child ever born, so it is possible that person number 3 is the head's daughter, and not a daughter-in-law, but there is no way to decide this and we put LNM=0 for person number 3. Person number 4 in household 63 is fairly clearly the daughter of person number 3, however, whence we put LNM=3 for this person.
Household 54. The head of the household is a 71 year old widowed woman. Because the children ever born questions were addressed only to women aged 15-64, her record provides no information on her numbers of children ever born. The 45 year old man and 48 year old woman, persons 2 and 3 in the household, are both coded as her children or children-in-law. These two persons are presumably husband and wife. We note that the woman has two male and no female children ever born, and that the last two persons in this household, males aged 15 and 13 are coded grandchild/spouse of grandchild, whence that these are probably the woman's two sons. We have no way of know which of the married couple is the child and which the spouse of child of the head.
Ambiguity of the 1990 Relation to Head Codes. The failure of the 1990 census relation to head codes to distinguish between offspring and spouses of offspring has a simple implication for matching. If a person too young to be married is coded child/spouse of child, we know that they are in fact a child of the head. If they are old enough to be married, however, we cannot identify children from the relation to head code.
<gfeeney@gfeeney.com>
