|Page tools: Print Page Print All RSS Search this Product|
Demography Working Paper 96/3
ESTIMATING FAMILIES: AN EXAMINATION OF METHODOLOGICAL ISSUES, DATA SOURCES AND PROVISIONAL ESTIMATES
HOW DOES THE ABS DEFINE 'FAMILY'?
1 A family is operationally defined by the ABS as two or more persons, one of whom is at least 15 years of age, who are related by blood, marriage (registered or defacto), adoption, step or fostering, and who are usually resident in the same household. The basis of a family is formed by identifying the presence of a couple relationship, parent-child relationship or other blood relationship. Some households will therefore contain more than one family (ABS, 1995).
WHAT ARE FAMILY ESTIMATES?
2 Family estimates are statistics showing the number and demographic characteristics of people living in families, the relationship between family members and the types of families they live in. The statistics relate to given time periods and are based on the usual resident population of Australia. They therefore reflect the family type and family relationship characteristics that exist in Australia among the usual resident population.
3 The estimates are conceptually analogous to the estimated resident population of Australia and differ from counts of families obtained from the Censuses of Population and Housing.
4 Although the ABS recognises the existence and importance of family members who do not live together in the same household, the current estimates refer only to the ‘residential family’ or families whose members live together in the same household.
USES OF FAMILY ESTIMATES
5 Family estimates have a variety of uses. They could potentially be used by the ABS as benchmarks of the number, type and characteristics of families and persons living in families. The benchmarks could either be used in preparing a sampling scheme for sample surveys of families and/or in preparing weights for sample survey estimates of families and persons living in families. As the ABS conducts several household and family sample surveys the need for family benchmarks cannot be overemphasised.
6 Family estimates may also be used by Commonwealth, State and Territory Governments in budgeting and preparing estimates and projections of the cost of providing various types of family assistance and social security payments to families. Family estimates could serve as a valuable input into government policies connected with taxation, education, health care and housing. The effect of various government measures on families could also be best analysed by studying estimates of families and the changes taking place in families.
7 Community organisations that provide welfare and other types of assistance to families need estimates of families. Family estimates may help them to monitor and evaluate changes taking place in families as well as to anticipate future family needs.
8 Because in most cases the ‘residential’ family coincides with the basic economic and consumption unit of society, that is the household, family estimates have commercial applications. Organisations that sell goods and services to families and households need information on the number of families, the average number and the age and sex composition and geographical distribution of people living in families and the changes taking place in families.
9 Family estimates are also important from the point of view of research into the family. Family research provides insight into the structure of Australian society and the changes taking place in the types, composition and growth of families in Australia. This makes possible a comparative analysis of Australian families over time and across countries.
WHY PREPARE A NEW STATISTICAL SERIES OF FAMILY ESTIMATES?
10 These family estimates will form a new statistical series to be produced by the ABS. They will complement estimates of the resident population published regularly by the ABS and the new statistical series on household estimates. They will form the second stage of the ABS plans to prepare estimates and projections of households and families.
11 This will not be the first time, however, that the ABS will publish estimates of families. The ABS collects and publishes census information on families and the household relationship and personal characteristics of persons living in families. A number of regular and occasional surveys, including the Labour Force Survey and Family Survey, also collect information on families.
12 The census, for instance, is a complete enumeration and is the only data source that provides near-complete data on the count and characteristics of families and persons resident in families in Australia. Census data on families, however, pertain only to the count of families and characteristics of family members as enumerated on census day and may not, ideally, be suitable as benchmarks. They exclude Australian residents who were temporarily overseas at the time of the census, persons who spent census night in Australia in non-private dwellings and persons who spent census night in Australia but were missed during the enumeration itself.
13 In the 1991 Census, for instance, 312,000 usual residents of Australia were missed (ABS, 3101.0). Household and family data for these people were therefore not available. About 224,000 Australian residents were also overseas at the time of the 1991 Census. Persons who spent census night away from their place of usual residence were not coded back to their households of usual residence (ABS, 2901.0). They were considered in household and family coding but their actual characteristics were not available at their households of usual residence.
14 Although estimates and characteristics of families obtained from various sample surveys are based on the resident population and are representative of families throughout Australia they vary from sample to sample because of sampling errors. They also pertain to only the reference date of the survey. The number and type of households and families obtained from sample surveys may also be affected by missing households, the omission of persons from within enumerated households as well as by omissions caused by the various scope and coverage rules of exclusion of the respective surveys. As such, they have not usually been used as benchmarks.
15 The purpose of this development work is, therefore, to develop a method that will enable the ABS to prepare regular estimates of families that pertain to the total usual resident population of Australia and which can be used as family benchmarks.
POTENTIAL SOURCES OF DATA AND THEIR LIMITATIONS
16 There are potentially four main sources of data for this project. These are:
18 The base population for estimating the population usually resident in households is the ERP The ERP does not, however, distinguish between persons resident in private and non-private dwellings. Since household and family estimates are concerned exclusively with the population usually resident in private dwellings, the ERP for each period has been adjusted by removing from it persons usually resident in non-private dwellings using correction factors estimated from the census (see para 55-62 and also ABS 3229.0).
CENSUS OF POPULATION AND HOUSING
19 A census of population and housing is held every five years. It produces a count of households and families and the relationship of household members to a household reference person. The census data are compiled at three levels - the person, family and dwelling levels. Prior to the 1986 Census, occupied houseboats in marinas and caravans, tents and cabins in caravan parks and roadside parking areas were treated as non-private dwellings. In the 1986 and 1991 Censuses, the classification of these dwellings was changed to private dwellings and household and family data were collected for them.
20 The census provides two sets of data for preparing family estimates. These are the proportions of the total usual resident count of people that live in households (private dwellings) and their ‘relationship in household’ characteristics. The latter is obtained from responses to the census household question on the relationship of each member of the household to Person 1 or Person 2.
Limitations of the census dataset for preparing family estimates
21 The census data will not, however, be used to produce relationships in household data because of limitations outlined below:
22 Firstly, the census question used to obtain household relationships data solicits information on only one level of household relationships. This may lead to a distortion and underestimation of the number of families, relationships within households and family type. For instance, to obtain information on relationships within households, the census asks the following question:
23 This question returns information on only one level of household relationships, that of each person in the household to Person 1 or Person 2. In situations where household members are related to each other but not to either Person 1 or Person 2, then that second level of relationships is lost.
24 Consider, for instance, two brothers and a sister boarding with an elderly married couple to whom they are not related. Responses to the census question will show that the only family relationship in that household is that between the elderly couple themselves. That between the two brothers and their sister is not captured in the solicited response because the census question does not ask if other members of the household are related to other persons in the household other than to Person 1 or Person 2. Thus, whereas two families exist in this household, only one is reported in the census.
25 Some ABS surveys have additional questions that capture these additional relationships. For instance, the 1992 Family Survey asked:
Are there any (other) step, de facto or in-law relationships within the household?
26 In addition to asking of the relationship of each person in the household to the head of the household, the current LFS questionnaire also asks of other relationships within the household.
EXTRACT FROM CURRENT LFS HOUSEHOLD FORM
27 However, the design of the household form and the nature of the question do not allow for exhaustive identification and listing of all household relationships, including de facto relationships. There are also indications that this part of the form is not uniformly and consistently filled in by interviewers.
28 The computer assisted interview (CAI) questionnaire of the LFS which is now under development, however, plans to ask the following additional question:
Is (Name of person) more closely related to someone else in the household?
If the answer is Yes the following additional questions are asked:
30 As a result of issues raised above regarding how household relationships and family data are collected in the census compared to other enumerations, data on households and families from various collections are not nearly the same. For instance, a comparison between the census, Labour Force Survey and Family Survey counts and estimates of households and families shows that the census undercounts the total number of households, especially family households and one person households and overestimates the number of group households.
31 Similar observations about an undercount of households in the 1986 and 1991 Censuses compared to estimates from the Labour Force Survey at the same reference points are also made in Household Estimates, Australia: 1986, 1991-94 (ABS, 3229.0, pp 3944). For instance, the 1986 and 1991 Census counts of households were lower than comparable LFS estimates by about 364,000 and 321,000 households, respectively. Evaluation work carried out in connection with publication number 3229.0 also indicates that the census tends to undercount one person households.
32 The table below shows that in the same month in 1991, the census count of families was lower than the LFS estimate by about 194,000 families and lower than the April 1992 Family Survey by about 476,000 families.
NUMBER OF FAMILIES: COMPARISON OF CENSUS COUNTS, LABOUR FORCE SURVEY AND FAMILY SURVEY ESTIMATES
a: The census counts are based on usual residents living in households
b: The August 1991 LFS estimates are those prepared by the Labour Force
Section and published in 8203.0. Note that these estimates are different from the current
provisional estimates on page 13. Estimates of one-parent families from publication
6203.0 do not include one parent families where there are no dependent children.
c: The Family Survey estimates are published in 4418.0 and include all usual residents.
33 It is not only that the census underestimates the number and types of persons living in households and families in Australia but also that estimates from the census, LFS and Family Survey are all different. For instance, the number of persons living in couple families was higher in the April 1992 Family Survey (13,453,800) than even in the June 1993 LFS (13,281,200) and much higher than in the August 1991 Census (11,963,500). Similarly, the number of persons in one-parent families was 1,673,100 in the Family Survey, 1,113,000 in the LFS and 1,488,900 in the 1991 Census. Further comparisons between the census, LFS and Family Survey estimates of families and relationships within households are carried out on pages 13-15.
34 Because of the better design of the relationship in household questions in the LFS and Family Survey, estimates of families, persons living in households and families and family types are expected to be more reliable than in the census. The more exhaustive identification of defacto, step and in-law relationships in the Family Survey compared to the LFS, however, ensures that these relationships are more accurately identified in the Family Survey than in the LFS.
LABOUR FORCE SURVEY
35 The LFS is based on a multi-stage area sample of private dwellings and a list sample of non-private dwellings and covers about 0.5% of all persons aged 15 years and over who are usual residents of Australia. It is a household-based sample survey.
36 The LFS excludes from scope members of the permanent Australian defence force, non-Australian diplomats, non-Australian diplomatic staff and non-Australian members of their families and members of non-Australian defence forces and their dependants. The sample is divided into eight panels or groups of households called rotation groups. Each rotation group is surveyed each month for eight months and is then dropped from the sample. The outgoing rotation group is replaced by a new rotation group which then remains in the sample for eight continuous months.
37 Among other data items, information is collected on the age and sex characteristics of household members as well as the relationship of each member of the household to a household reference person. The relationship information is used to classify household members into those living in families and those not living in families.
38 Apart from limitations of the LFS form design affecting the collection of relationship in household data, the LFS dataset also has a number of limitations which have implications for preparing the proposed household and family estimates.
39 Firstly, children aged 0-14 years are not directly enumerated in the LFS. The LFS dataset does not, therefore, have individual records for children. It does not also have children’s weights with which to weight and estimate the number of children at the population level.
40 Without creating child records and calculating child weights for these records, the number of children in the population may be estimated indirectly by weighting the reported number of children in each household by the person weights of the adults in the household or by household weights (if they are available or estimable). The number of children can also be estimated directly by calculating direct weights from dummy records for children residing in the household based on age, sex and part of State in the same way as for adult weights.
41 Secondly, members of the permanent defence forces are out of scope of the LFS and axe thus excluded from data compiled for their households. Their households are normally excluded from family statistics.42 Thirdly, although persons who are away from their usual residence for 6 weeks or less at the time of interview may be enumerated at their usual residence, with relevant information being obtained from usual residents present at the time of the survey, this is not the case for persons who are away for more than 6 weeks. Data on these people are not included in their households of usual residence even if they axe enumerated elsewhere in Australia. This may affect the characteristics of households and families containing long-term absent members.
43 Fourthly, the LFS constructs family characteristics for only families where information was obtained for every member. Families data available for the preparation of family estimates may therefore be affected by the degree of mobility of family members or by non-response and non-contact of certain family members.
THE FAMILY SURVEY
44 The 1992 Survey of Families in Australia (Family Survey) was conducted between March and May 1992 as part of the Special Supplementary Survey (SSS) program. It covered approximately 34,000 people in 17,798 private dwellings and special dwelling units.
45 As described above, the Family Survey questionnaire was designed to obtain accurate information about relationships between individuals living in households. Specifically, relationship questions were designed to not only obtain primary level relationships between each household member and the head of the household but also step, de facto and in-law relationships among other household members. Because of this, the Family Survey data on relationships within households and types of families are of extremely good quality.
46 A Family Survey was also conducted in 1975 and 1982. However, although there are currently proposals for another SSS family survey, there are no firm plans to conduct another one in the near future. Thus, while SSS surveys on families may provide accurate information, they do not provide regular family data from which to derive demographic estimates.
47 In April 1997 the Family Characteristics Survey (ECS) will be included in the Monthly Population Survey MPS program. This will be the first monthly supplementary survey on family structure and composition run in conjunction with the LFS. It will be subject to the same scope and coverage exclusions as the LFS, use the same household form and incorporate the same sampling and weighting methodology. While the FCS will clarify some characteristics of family composition, such as step relationships, estimates of the number and type of families are not expected to differ significantly from LFS estimates.
48 With the implementation of changes to the LFS questionnaire design it is expected that the quality of LFS family data will improve.
49 In view of these improvements and the irregularity of other sources of family data, the LFS is likely to be the most reliable and readily available source of regular data for family estimates.
50 The main source of data for the proposed family estimates will therefore be the LFS and the ERP The latter will be converted into the population usually resident in households using census proportions of persons enumerated in private and non-private dwellings.
METHOD OF ESTIMATION51 The method to be used to produce the family estimates is a modification of the household size propensity method used by the ABS to produce the estimates of households, in Household Estimates, Australia: 1986, 1991-94 (ABS 3229.0). The basic feature of the household size propensity method is the estimation of the propensity or probability of belonging to a household of a given size aAcC, that is, the number of adults, A and number of children, C.
52 For family estimates, the method will be adapted to produce estimates of the propensity of possessing a certain ‘relationship in household’ characteristic and living in a given family type. Some of the relationship in household characteristics are husband, wife or partner in a couple family, a lone-parent, a dependent or non-dependent child, a related individual living with relatives, a non-family member living with non-relatives or a person living alone.
53 An important aspect of the method is to estimate the household population or the resident population that usually lives in households and to use a reliable data source to determine the probability or propensity of possessing a certain ‘relationship in household’ characteristic. LFS data are available for the calculation of these propensities for each month.
54 Assuming that the estimated relationship in household propensities of household members are accurately determined, then they can be applied to the estimated household population at given reference points to yield estimates, at those reference points, of the number of persons living in households who have those characteristics as well as the distribution of these persons into family and non-family households and into different types of families.
BASIC STEPS IN PREPARING FAMILY ESTIMATES
55 Estimating families involves three basic steps;
Estimation of the population that usually lives in households by age, sex and part of State, at given reference dates;
Estimation, from a reliable data source, of relationship in household propensities by age, sex and part of State, at given reference points; and
Application of the relationship in household propensities at given reference points to the distribution of the household population by age, sex and part of State, to yield estimates of the number of household members with given relationship in household characteristics and the distribution of household members into
families by family type.
Step 1: Estimation of the resident population living in households
56 The first step in preparing family estimates is to prepare a distribution of the household population by age, sex and part of State. The household population is the estimated resident population ERP living in private dwellings less overseas visitors (see ABS 3229.0).
57 The household population comprises two subsets, namely, a subset that lives in family households (the family household population) and a subset that lives in non-family households such as group households and one-person households. A family household is a household with one or more families present.
58 The family household population is made up of: (a) family members living with family members and (b) unrelated persons such as friends, lodgers or boarders living within family households.
59 Because the ERP from which the household population is estimated is not classified into persons living in private and non-private dwellings the distribution of the household population by age, sex and part of State is estimated from the latest census. This is done by applying to the ERP the proportions of the census usual resident household population’ by age, sex and part of State.
60 The census data used to produce the proportions of people usually living in households are carefully selected so as to approach as closely as possible to the concept of the estimated resident population. They include only usual residents of private dwellings. They exclude overseas visitors but do not include Australian residents temporarily overseas at the time of the census because those data are not available.
61 The household population on 30 June 1986 and 1991, for instance, could be estimated by applying the estimated 1986 and 1991 census proportions of persons living in households by age, sex and part of State to the ERP at the same reference dates. Because the 1991 Census took place on 6 August 1991 (rather than 30 June as was the case m 1986), the proportions of the ERP usually living in households as at 30 June 1991 is estimated by linearly interpolating between the proportions of the resident population that usually lives in private dwellings as observed from the June 1986 and August 1991 Censuses.
62 This interpolation may be unnecessary because the difference between the 1986 and 1991 Census proportions of people living in private dwellings is very small. Any expected differences in the proportions between 30 June 1991 and 6 August 1991 will be even smaller. The interpolation may however be carried out for the sake of methodological rigour.
63 Estimates of the number of persons living in households for any intercensal reference date could also be prepared by applying to the ERP at the specified date interpolated estimates of the proportions of persons living in private dwellings in two successive censuses. For example, if the focus of interest is in yearly estimates of the household population between the 1986 and 1991 Censuses, then a linear interpolation between the 1986 and 1991 Census estimates of the proportions of people living in households by age, sex and part of State, would provide the needed proportions to be applied to the Fall’ at the required intercensal reference dates.
64 Estimates for periods beyond the latest census are estimated in a slightly different fashion. For example, as the latest census was in 1991, estimates of the household population in June 1992, 1993,1994, 1995 and 1996 are prepared by a linear projection of the proportion of the 1991 Census usual resident population living in households, by age, sex and part of State, using the avenge annual change in the proportion of persons living in households between the 1986 and 1991 Censuses (see ABS, 3229.0, page 36).
65 The estimated proportions of persons living in households at any given reference point after the 1991 Census are then applied to the respective ERPs to yield estimates of the household population for those reference points.
Step 2: Estimation of relationship in household propensities
The LFS dataset classification of family type and of persons living in households relationship in household
66 The propensity of possessing a given relationship in household characteristic and living in a given family type is estimated by the proportion of people of a given relationship in household category living in a given family type. Because the LFS only contains records for persons aged 15 years and over, all sample statistics are based on propensities estimated for persons aged 15 years and over.
67 The LFS has two datafiles for estimating families. The first is a person datafile containing information on the personal and household characteristics of all persons included in the survey. Household members are also assigned a family number to differentiate between household members living in different families within the same household. There is one record for each person aged 15 years and over. There is also a family dataflie containing summary information on each family identified in the survey. There is one record for each family. The family datafile contains variables showing the presence or absence of a husband, wife or partner, dependent and non-dependent children, relatives, and non-family members. These variables make it possible to identify and class ify families into the family types listed below. Common household numbers on the two files allow the two files to be merged, thereby making it possible for individuals on the person datafile to be assigned a family type.
68 For purposes of estimating households and families, data are pooled over a five-month period centred on the month for which estimates are required. For June estimates, for instance, all eight rotation groups in June are pooled, along with the two outgoing rotation groups in April and May and the two incoming rotation groups in July and August. The pooling increases the sample size by about 50% and strengthens the stability of the estimates. The method of pooling the outgoing and incoming rotation groups ensures that all families have only one chance of being included in the pooled dataset.
Classification of family type and of persons living in households
69 The following relationship in household categories will be used. They conform with the ABS Directory of Concepts and Standards for Social Labour and Vein Demographic Statistics: Volume 1, April 1995.
70 All persons living in families will be identified as living in one of these family types.
72 Thus the presence of unrelated persons within family households do not affect the classification of those family households into family types.
73 The classification scheme used to prepare the propensities from which the family types are derived classifies each household member as possessing a given relationship in household characteristic and living in a given family type. For instance, a 20 year-old fall-time student maybe classified as a dependent child living in a couple family containing dependent and non-dependent children or in a one parent family containing dependent children only, while a wife may be classified as a wife living in a couple family with dependent children only or in a couple only family containing no children. This classification is equivalent to the ABS family type categories at the major group level, with children under 15 years and dependent students being grouped together into the category ‘dependent children’.
74 This major group level classification is adopted because of the small size of the sample, the small number of observations in some cells when minor groupings are adopted and the potential for high relative standard errors when the number of entries falls below a certain number.
Step 3: Estimation of the number and type of families
75 The estimated propensities are then applied to the estimated number and type of household population by age, sex and part of State to yield estimates,
families at the population level, of the distribution of the household population into relationship in household and family type categories. This is done through weighting the sample by a new set of person weights.
76 The new set of weights are necessary in order to adjust for changes in the age, sex and part of State/Territory distribution of the sample due to non-response, non-contact and the exclusion of non-household dwellers and visitors from the sample used to prepare the family estimates. The new set of weights also take into account the pooling of 5 months data and an increase in the final sample size by about 50%.
77 The weights are prepared by comparing the number of adults or persons aged 15 years and over in the estimated household population by age, sex and part of State/Territory with the number of adults in the final pooled dataset, by age, sex and part of State or Territory. The weights are therefore benchmarked to the ERP less the number of persons in non-private dwellings.
78 Apart from using person weights one can also use the harmonic mean household weight, geometric mean, arithmetric mean and weight of head to estimate households and families. The harmonic mean household weight tends to underestimate the number of income units, households and families, while using the weight of the head of the household tends to underestimate the number of these units. Using individual person weights, as is done for these estimates, trends to give estimates between these two extremes (Landt et at 1994). However, regardless of the weight used, the number of children is usually overestimated.
79 A criticism of using person weights is that they give accurate estimates of person characteristics at the population level but not of aggregate social units such as households and families. This is because although the sample design may yield a representative area sample of households, the absence of benchmarks means that appropriate household weights are not easily constructed.
80 Although an internal ABS investigation recommended the use of harmonic weights (Niedorfer, 1982) the harmonic mean method of nonperson weighting has, since 1994, been dropped from most surveys run off the parallel sample, in favour of using the new household estimates as benchmarks with an integrated weighting option, or linear weight on its own - which approaches could be best described collectively as ‘calibrated weighting’.
81 In these estimates, however, re-estimated person weights are used, with families being estimated indirectly via the family type and household relationship characteristics of the household population. Application of the person weights to the sample estimates will yield estimates of the number of people in each relationship category in each family type at the population level. No matter an individual's relationship in household, that individual can only belong to one family type. Thus, in a couple family with dependent and non-dependent children, all members belong to that family type irrespective of their relationship in household characteristic. If the data are sorted by family number and the number of adults in each family is summed up, then division of the total number of adults in each family type by the average number of adults in each family type will yield estimates of the number of families by type
82 For instance, if there are 25,000 adults living in couple only families containing no children, and if the avenge number of adults in families of that type is 2, then the total number of couple only families without children is 12,500. Similarly, if there are 350,000 adults living in one parent families with dependent and non-dependent children, and if the average -number of adults in that family type is 2.4, then the total number of families of that type is 140,000. This estimation procedure can be carried out by age, sex and part of State.
83 The basis of the construction of the family types is the presence of a parent-child relationship. However, each family type may also contain other relatives and other persons such as friends, boarders or lodgers, even though it is classified as containing only a certain category of children. Because of this categorisation, there are no separate categories for families containing only parents and children and categories containing other relatives and persons.
84 Some preliminary family estimates are presented below and compared with estimates from the 1992 Family Survey and counts of families from the 1991 Census of Population and Housing. Anticipated changes in the definition of aspects of relationship in household have been incorporated into these estimates.
RELATIONSHIP IN HOUSEHOLD: COMPARISON OF 1991 CENSUS COUNTS, PROVISIONAL AUGUST 1991 LFS AND APRIL 1992 FS ESTIMATES
86 The above table shows that the provisional LFS estimates match those from the Family Survey, although the latter are for a period one year later than the US estimates. The LFS and Family Survey estimates of the total usual resident population living in households are also similar, with the census estimates being about 1.8 million persons less than in the Family Survey.
87 However both the Census and the LFS tend to underestimate the number of couples and single parents relative to the Family Survey. The higher number of persons categorised by the LFS as other related individuals and unrelated persons probably points to inadequacy of the LFS form in properly identifying all levels of household relationships.
88 The census figures are lower probably because whereas the census figures are counts of persons as enumerated on census night, and do not include missing persons and residents temporarily overseas at the time of the census (RTOs), the LFS and Family Survey estimates are adjusted for both missing persons and RT0s However, as the sum of missing persons and RTOs is about 600,000, the large difference between the census figures and estimates for the other two data sources might be due to issues associated with the design of the census question on relationship in household.
89 The table below examines the distribution of families by family type for the three datasets. Once again, the Family Survey estimates and provisional estimates from the LFS are similar and much higher than the census counts of families.
FAMILY TYPE: COMPARISON OF AUGUST 1991 CENSUS, AUGUST 1991 LABOUR FORCE SURVEY AND APRIL 1992 FAMILY SURVEY ESTIMATES
90 Inspite of the better design of the Family Survey questionnaire compared to that of the LFS for obtaining relationship in household data, the closeness of the Family Survey and the provisional estimates from the LFS is very encouraging. It is apparent, however, that the census has underestimated most couple and one parent family types. The number of families classified as 'other’ is higher in the census and LFS than in the Family Survey probably because the Family Survey from design allows it to more successfully identify and classify families into appropriate types rather than to classify them as ‘other’ as appears to be the case with the census and LFS.
91 To better compare the provisional estimates from the LFS and those from the Family Survey, estimates at the same point in time, April 1992, are compared. April 1992 is taken as the mid-point of the Family Survey which took place between March and May 1992.
RELATIONSHIP IN HOUSEHOLD: COMPARISON OF APRIL1992 LABOUR FORCE SURVEY AND APRIL1992 FAMILY SURVEY ESTIMATES
92 The table shows differences between the two data sources. For instance, the number of persons living in households is slightly higher in the provisional estimates than in the FS estimates. On the whole, estimates of the number and proportion of partners in couple households and persons living alone are lower in the provisional estimates than in the FS estimates. The number of persons classified as non-dependent children, other family person, unrelated person in family household and unrelated person in group household are larger compared to estimates from the Family Survey.
93 The family type table (below) shows that estimates of the total number and types of families in the 1992 Family Survey and the current provisional estimates are similar. Differences observed are a slightly higher number of couple families with non-dependent children, one parent families with dependent children and one parent families with non-dependent children in the provisional estimates than in the Family Survey, and a slightly lower number of couple families without children in the provisional estimates than in the Family Survey.
94 Some of these differences may be due to the different designs of the questionnaires of the two collections as well as to the method of weighting the sample statistics. It is expected that when the new LFS form and computer assisted interviewing CAI method are implemented, the LFS dataset will be comparable to that of the Family Survey.
FAMILY TYPE: COMPARISON OF APRIL 1992 LABOUR FORCE SURVEY AND APRIL 1992 FAMILY SURVEY ESTIMATES
95 In the pages that follow some provisional estimates of families by family type and relationship in household are presented. Analysis and description of the statistics on families are not included in this document. The final estimates will contain a statement and estimates of standard errors.
96 These provisional estimates will serve as the basis of comments and evaluation towards the later production of family estimates.
97 The provisional estimates show that plausible estimates of families can be prepared using the LFS as a data source, by the estimation method described. One advantage of the LFS dataset is the frequency of the collection.
98 The feasibility of redeveloping LFS collection forms and derivation of family data is being considered. It is possible that these considerations may result in more complete estimates of Relationships in Household, Family Type and other family data; and in closer comparability between LFS and the Family Survey in terms of scope, concepts and classifications. (For further information about LFS developments, contact Geoff Neideck on 6252 6753).
99 In spite of concerns expressed about relationship in household data from the census, census data can still be used to prepare family estimates for small areas such as SLAs. The small sample size of the LFS means that estimates for small geographical areas based on the LFS dataset will have unacceptably high standard errors.
REFERENCES AND BIBLIOGRAPHY
Australian Bureau of Statistics (1992): Australia’s Families - selected findings from the Survey of Families in Australia, Canberra.
Catalogue No. 44180.0.
Australian Bureau of Statistics (1994): Information Paper: Population Estimates: Concepts, Sources and Methods. Canberra.
Australian Bureau of Statistics (1995): Directory of Concepts and Standards for Social Labour and Demographic Statistics, vol. 1,
second edition, October 1995.
Australian Bureau of Statistics (1996): Household Estimates, Australia: 1986,1991-1994. Canberra. Catalogue No. 2710.0.
Ironmonger, D.S. & Lloyd-Smith C.W. 1992: Projections of Households and Household Populations by Household Size Propensities’, Journal of the Australian Population Association, vol. 9 (2).
Landt, J A Harding, K Perceival K Sadkowski (1994): Reweighting a Base Population for a Microsimulation Model’, NATSEM Discussion Paper Series, University of Canberra, Discussion Paper No.3.
Niedorfer, G. (1982): ‘Investigation into Household Weighting’, Unpublished Paper, Australian Bureau of Statistics, Canberra.
Peter McDonald (1994) ‘Household Projections’, personal communication.
LIST OF TABLES
1A RELATIONSHIP IN HOUSEHOLD, AUSTRALIA
1B RELATIONSHIP IN HOUSEHOLD, AUSTRALIA
2A FAMILY TYPE, AUSTRALIA
2B FAMILY TYPE, AUSTRALIA
3 RELATIONSHIP IN HOUSEHOLD, STATES AND TERRITORIES: JUNE 1991
4 RELATIONSHIP IN HOUSEHOLD, STATES AND TERRITORIES: JUNE 1992
5 RELATIONSHIP IN HOUSEHOLD, STATES AND TERRITORIES: JUNE 1993
6 RELATIONSHIP IN HOUSEHOLD, STATES AND TERRITORIES: JUNE 1994
7 RELATIONSHIP IN HOUSEHOLD, STATES AND TERRITORIES: JUNE 1995
8 FAMILY TYPE, STATES AND TERRITORIES: JUNE 1991
9 FAMILY TYPE, STATES AND TERRITORIES: JUNE 1992
10 FAMILY TYPE, STATES AND TERRITORIES: JUNE 1993
11 FAMILY TYPE, STATES AND TERRITORIES: JUNE 1994
12 FAMILY TYPE, STATES AND TERRITORIES: JUNE 1995
13 RELATIONSHIP IN HOUSEHOLD, BY SEX: JUNE 1991 AND JUNE 1992
14 RELATIONSHIP IN HOUSEHOLD, BY SEX: JUNE 1993 AND JUNE 1994
15 RELATIONSHIP IN HOUSEHOLD, BY SEX: JUNE 1995
16A RELATIONSHIP IN HOUSEHOLD, BY AGE GROUP: JUNE 1991
16B RELATIONSHIP IN HOUSEHOLD, BY AGE GROUP: JUNE 1991
17A RELATIONSHIP IN HOUSEHOLD, BY AGE GROUP: JUNE 1992
18A RELATIONSHIP IN HOUSEHOLD, BY AGE GROUP: JUNE 1993
18B RELATIONSHIP IN HOUSEHOLD, BY AGE GROUP: JUNE 1993
19A RELATIONSHIP IN HOUSEHOLD, BY AGE GROUP: JUNE 1994
19B RELATIONSHIP IN HOUSEHOLD, BY AGE GROUP: JUNE 1994
20A RELATIONSHIP IN HOUSEHOLD, BY AGE GROUP: JUNE 1995
20B RELATIONSHIP IN HOUSEHOLD, BY AGE GROUP: JUNE 1995
21 AVERAGE FAMILY SIZE BY FAMILY TYPE, STATES AND TERRITORIES: JUNE 1991
These documents will be presented in a new window.