3109.0 - Demography Working Paper 1996/1 - Evaluation of Administrative Data Sources for Use in Quarterly Estimation of Interstate Migration Between 1996 and 2001, 1996
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 21/09/1999
|Page tools: Print Page Print All|
Demography Working Paper 96/1
TABLE OF CONTENTS
1 MAJOR FINDINGS
1 Australia produces high quality postcensal population estimates with accuracies which compare favourably with those prepared by Statistics Canada, the United States Census Bureau, Statistics New Zealand and the Office for National Statistics (England and Wales).
2 Australia's relative expertise in postcensal population estimation is a reflection of the frequency of Australian population censuses (five yearly) and the availability of administrative data pertaining directly to births, deaths and international migration. As a consequence, the accuracy of Australia's postcensal population estimates principally reflect the accuracy of postcensal interstate migration estimation. By contrast, Canada, USA and England and Wales are required to estimate both international migration (or parts thereof) and internal migration because of lack of administrative data that relates directly (and or completely) to these population activities. The comparable accuracy of Canadian postcensal estimates to Australian accuracies is therefore, particularly noteworthy.
3 At the national and State level, the quality of Australia's postcensal population estimates improved between 1981 and 1991. These improvements principally reflected reductions in errors associated with estimating interstate migration between 1981 and 1986. Considerable variability still exists in the quality of postcensal interstate migration estimates between the States and Territories. Since 1986, estimation of postcensal interstate migration in Australia has been based on Health Insurance Commission (Medicare) transfers data relating to persons aged 1-14 years.
4 This paper provides results from a comparison of Medicare data with eight alternative administrative data sources as well as estimates of interstate migration from the ABS Monthly Population (Labour Force) Survey. In brief, population coverage and data timeliness considerations indicate Medicare data remains the best administrative data source available for preparing postcensal interstate migration estimates. Discussions with a number of senior officers from the Health Insurance Commission indicate that while no-one is sure what policy changes will be introduced by the present Government, speculated changes are unlikely to adversely effect Medicare data with regard to population estimation during this current term.
5 A number of recommendations are proposed to improve the estimates being produced from Medicare data. It is recommended the ABS develop a closer working relationship with the Health Insurance Commission for the perceived mutual benefit of both organisations. In particular, the Health Insurance Commission is encouraged to statistically verify why numbers of Medicare registrants grew at approximately twice the annual number experienced prior to the card reissuance program in 1991. Preliminary advice from the Health Insurance Commission suggests that use of Medicare by non Australian residents possibly increased significantly during the early 1990s and changes in activities designed to maintain the Medicare register during this period possibly resulted in the deregistration of some eligible persons, requiring them to re-register. This latter suggestion in particular, raises questions about the success of the ongoing reissuance program to maintain the Medicare registrar. Consequently research to quantify the impact of alterations on the Medicare register since 1992 is expected to benefit both the Health Insurance Commission and the ABS. In addition, the Health Insurance Commission is encouraged to obtain deaths data from each State and Territory Registrar to update the Medicare register. The Medicare register is currently updated using deaths data from only some State Registrars because of data access cost considerations. This has a potentially biasing effect on Medicare data, especially for use in population distribution purposes.
6 Based on international experiences, it is recommended that the ABS look to augment its present policy of preparing postcensal population estimates from one administrative data source. Both Canada and the USA have found that estimates of postcensal internal migration can be improved by preparing estimates from more than one data source. Two principle benefits are perceived from using multiple data sources. The first advantage is the potential to validate estimates prepared from separate data sources as part of the estimation process. The second advantage is that multiple sources can aid estimation of specific population subgroups. For example, the US Census Bureau uses pension data to augment migration estimates initially prepared from other data sources for persons aged over 65 years. Another potential advantage for Australia of using multiple datasets, is that an alternative data source and or methodology will already be in development if future policy changes severely affect Medicare data.
7 To determine whether the ABS could benefit from augmenting Medicare data with alternative age specific administrative data sources, it is recommended a comparison of age related interstate migration estimates be undertaken using 1996 Census data.
8 A number of recommendations have been made which are specific to the alternative data sources evaluated as part of this research. For brevity purposes, these recommendations are not contained in this Major Findings section but can be found in the final subsection of each section relating to these alternative data sources. Data timeliness considerations prevent a number of these alternative data sources from being superior to Medicare data for use in preparing quarterly postcensal estimates of interstate migration (i.e. many are annual collections). This does not however, necessarily limit their potential value in validating quarterly postcensal estimates prepared from Medicare data.
9 It is difficult to recommend which alternative administrative data source be used to replace Medicare data if it is abandoned or altered significantly after 1998. One of the principal conclusions that can be drawn from this research is that changes to administrative data definitions are virtually continuous. Some of these changes reflect technological advancements which impact service provision and or database capacities. Other changes reflect policy decisions which either directly or indirectly impact specific administrative collections. As a consequence, it is imperative that administrative data collections used in postcensal population estimation be monitored regularly and good working relationships with data suppliers be nurtured to encourage the systematic discussion of issues of mutual interest.
10 The Australian Bureau of Statistics (ABS) is required by law to prepare quarterly estimates of the population for each Australian State and Territory. (Unless specified otherwise, for the remainder of this paper the term State refers to Australia's States and Territories.) Additional to this legal requirement, the ABS prepares annual population estimates by Local Government Area (LGA) and Statistical Local Area (SLA) and population projections at the State level every two to three years.
11 The population bases on which these estimates and projections are prepared are Australia's quinquennial Censuses of Population and Housing. Preparation of postcensal estimates and projections have been based on component (administrative records) methods principally because of ready access to quality births, deaths and overseas migration data. Australia does not keep administrative records relating directly to interstate population migration. Over time, the ABS has used a number of indirect administrative data sources to estimate quarterly interstate migration, including electoral rol1 registrations, family allowance payments and presently Health Insurance Commission (Medicare) data. Thus, the quality of Australia's postcensal population estimates at the State level, has principally reflected the quality of indirect estimation of interstate migration.
12 The ABS has undertaken regular evaluations into the quality of their postcensal interstate migration estimates. Such evaluations have periodically involved assessing alternative administrative data sources and infrequently recommended that an existing data source be substituted because its coverage had deteriorated or a superior data source had been newly identified. Current investigations therefore, update considerable past research undertaken by the ABS. Most of the past research remains unpublished with a notable exception being the work of O. B. Di Iulio in 1984, published as Postcensal Interstate Migration Estimates 1966-1981.(O.B. Di Iulio, ABS Occasional Paper 1984/2). A comprehensive description of changes to ABS practices in interstate migration estimation since then is provided in the 1995 publication Population Estimates: Concepts, Sources and Methods. (ABS Cat. No. 3228.0).
13 This report contains details about investigations into Medicare data and alternative administrative data sources. The research has focussed on identifying alternative data sources to Medicare for quarterly estimation of interstate migration from 1996 to 2001. However, some consideration has also been given to the potential of these data sources in relation to annual sub-State population estimation. The characteristics of each data source have been assessed against criteria which define ideal interstate migration data characteristics and are detailed in separate sections of this report. In addition, a comparison of postcensal estimation techniques employed by the US Census Bureau, Statistics Canada, Statistics New Zealand and the Office for National Statistics (created from the former Office of Population Censuses and Surveys and the former Central Statistical Office) was also completed as part of this evaluation. These international comparisons allow past international research to be assessed with regard to Australian requirements and provide an opportunity to compare the relative effectiveness of postcensal population estimation methods used by different nations. To assist readers, this report also contains a section which highlights the major findings from this research.
14 The knowledge and co-operation of a large number of private and public sector employees, both within Australia and overseas, has been sought during the preparation of this paper. The views expressed in this report however, are principally the author's and do not necessarily reflect those of the ABS or the other agencies contacted.
3 EVALUATION CRITERIA
15 Administrative data with the following characteristics are expected to provide ideal symptomatic indicators of postcensal interstate migration. The purpose in defining these ideal characteristics is to create criteria against which Medicare data and alternative data sources can be evaluated.
4 INTERNATIONAL COMPARISONS
16 A comparison of the accuracy of international postcensal population estimates is presented in this section. This comparison was undertaken to provide an independent assessment of the relative success (or otherwise) of Medicare data in Australia's postcensal population estimates. Since international agencies use different measures of population data accuracy, some care has been required in undertaking this exercise. Furthermore, careful consideration is needed to interpret the results of international comparisons because the extent of estimation required to prepare postcensal estimates, varies between international agencies. For example, Statistics Canada, the United States Census Bureau and the Office for National Statistics are required to estimate overseas migration (or parts thereof) as well as internal migration. Conversely the ABS and Statistics New Zealand can rely on comprehensive administrative data for overseas migration estimates.
17 The United States Census Bureau and Statistics New Zealand use errors of closure as a measure of population accuracy. Errors of closure are the difference between population estimates produced prior to a census and corresponding census count estimates. A closely related term is intercensal discrepancy. Intercensal discrepancies are the difference between population estimates prepared prior to a census and census count estimates adjusted for those inadvertently missed or overcounted by the census (i.e. net census undercounts). Census based estimates may also be adjusted to reflect a reference date other than the census date. As a consequence of adjustments, intercensal discrepancies are a more robust indicator of data quality than errors of closure. The ABS uses intercensal discrepancies to measure the quality of Australian census based population estimates and these measures can also be used to assess the quality of its postcensal population estimates and compare them internationally.
18 In general, variations in a nation's intercensal discrepancies reflect changes in census non-sampling errors (e.g. variations in collection practices, net census undercount estimation and processing techniques) and at the national level, changing accuracies in estimates of births, deaths and international migration. National level intercensal discrepancies provide benchmarks against which subnational intercensal discrepancies can be compared. Variations between national and subnational intercensal discrepancies provide a rough guide to the quality of postcensal subnational (i.e. internal) migration estimation. Such comparisons will provide only rough guides as variations between subnational and national intercensal discrepancies are expected to also reflect the greater difficulty associated with estimating smaller populations. Since 1986, variations between national and State intercensal discrepancies provide a broad guide to the ability of Medicare data to estimate interstate migration in Australia.
19 Rough comparisons between errors of closure and intercensal discrepancies are possible if estimates of net census undercounts are also available. The table below provides data sought from Statistics Canada, the US Census Bureau, Statistics New Zealand and the Office for National Statistics to allow these comparisons to occur. (Note: Statistics New Zealand was unable to provide census undercounts for the years requested because the agency did not commence estimating census undercounts until 1996.)
Comparison of Error of Closures and Intercensal Discrepancies
(a) Absolute average error in percentage
20 Valid conclusions about the relative accuracy of international postcensal estimates require comparisons between intercensal discrepancies rather than errors of closure. For example, when intercensal discrepancies are approximated for the United States (i.e. errors of closure less net undercounts), these figures exceed Australia's in absolute terms (i.e. -0.75 and 0.21 for 1990 and 1991, respectively). Comparing these figures leads to the conclusion that between 1980 and 1990, US population change was overestimated via US postcensal estimation techniques and by a rate greater than that by which Australia's population growth was underestimated between 1986 and 1991, via Australian postcensal estimation techniques.
21 The figures above indicate that Australia's postcensal population estimation methodologies compare favourably with those used by other international statistical agencies. This finding is not surprising since Australia is required to estimate only internal migration from indirect data sources while the US, Canada and England and Wales are required to estimate international migration (or parts thereof) as well as internal migration from indirect data sources. Thus the similarity in accuracies between Canadian and Australian postcensal estimates is particularly noteworthy given Canada's greater reliance on indirect data estimation.
22 At the national and subnational levels, Canada's intercensal discrepancies have been of a similar absolute magnitude to Australia's for the past three quinquennial censuses (i.e. 1981, 1986 and 1991).
23 The magnitude of the Canadian net census undercount estimates stands out from the other national undercount estimates (e.g. 2.82 per cent in 1991). This may be a consequence of their estimation by the Reverse Records Check method, which differs substantially from the Post Enumeration Surveys used by most other national statistical agencies. Even the undercount for the 1991 Census of England and Wales (coinciding with the introduction of the unpopular poll tax) is estimated below 2.82 per cent. With regard to England and Wales, it is interesting to note that the average accuracy of postcensal estimates for the constituent countries is not greatly dissimilar to Australia's (i.e. 0.27 and 0.21 respectively, for 1991). The relatively large intercensal discrepancy for England and Wales combined (0.38) is expected to reflect errors associated with estimating England's international migration. Statistics Canada is currently evaluating the relative merits of census coverage estimation methods being used by selected international statistical agencies.
24 Statistics Canada prepares preliminary quarterly estimates of total interprovincial migration using Child Tax Benefit (CTB) transfers data. Child endowment style data has been used since 1956 because it is timely and until 1993, gave 100 per cent coverage of address details relating to children aged under 18 years. The introduction of eligibility criteria in 1993 is believed to have reduced coverage to about 80 per cent. Since 1976, Statistics Canada has used personal income tax data to produce finalised estimates of interprovincial migration, disagregated by age and sex, on an annual basis. Although such finalised estimates are not capable of identifying multiple migrations within a year, personal income tax data is considered superior to CTB data because the increased population coverage of the former dataset (i.e. persons of working age and their identified dependents who submit taxation returns in two consecutive years).
25 The US Census Bureau prepares annual estimates of postcensal interstate migration. Taxation data is used to estimate migration for persons aged under 65 years. Estimates of child migration are prepared from school enrolments data and social security data is used to provide sample estimates of migration by the elderly. As a separate exercise, a variety of symptomatic indicators such as housing units, vehicle registrations etc. are combined to prepare change in stock estimates of interstate migration. The separately prepared sets of estimates are then combined to produce composite estimates of postcensal interstate migration. The obvious disadvantage to this method is the need to be able to access a variety of data sources in a timely manner and the complex manner of estimate derivation. As a consequence, the US Census Bureau is currently looking to create an integrated and rationalised model for population estimation. However the US Census Bureau perceives that a considerable advantage in using multiple data sources (and or methods) is that migration for specific populations (e.g. the aged) can be estimated from specialised datasets (e.g. pensions data) and these estimates can provide postcensal cross checks. For example, unexpected estimates from one data source can be compared to see whether they reflect aberrations in that one data source or are replicated in estimates from other data sources. Obviously the benefit of regular validation activity increases as the frequency of census collection decreases. This does not imply however, that the benefit of such activity is reduced as the frequency of census collection increases.
26 In the absence of accurate administrative record data, Statistics New Zealand have used a variety of symptomatic indicators to prepare annual estimates of net internal migration by sex, five year age group and at Territorial Local Authority level since the 1970s. The symptomatic indicator data are converted to net internal migration estimates by informal and subjective estimation practises. Consequently, estimation success is not only dependent on symptomatic data coverage considerations but also on staff skills and local knowledge. Because of these limitations, Statistics New Zealand is looking to develop models using recently available taxation and or school enrolment data to estimate internal migration post 1996.
27 The main data source used to estimate internal migration in England and Wales is patient transfers details (excluding patient address) from the National Health Service Central Register. This data is used in conjunction with electoral registration transfers data to estimate local authority migration.
28 Although Australian estimates of internal migration compare favourably with those produced by other leading international statistical agencies, the experiences of Statistics Canada and the US Census Bureau suggest multiple estimation practices can provide better quality estimates than those produced from single data sources. A comparison of age related intercensal discrepancy data or 1996 Census interstate migration data is expected to reveal the potential benefit of augmenting Medicare data with alternative age specific data sets.
29 The postcensal estimates which generate the most user contention are those prepared annually for use in allocating Commonwealth Grants between the States and Territories. Consequently, if validation of Medicare estimates by alternative datasets is adopted (refer paragraph 28), the ability to validate the Grants Commission estimates (i.e. year ended December estimates) is the recommended priority. Prefered alternative datasets will therefore need to relate to this (or a similar) reference period and be available to the ABS within 4 months after December 31.
5 MEDICARE DATA
30 A preliminary evaluation of Medicare data was undertaken via extensive discussions with a number of Health Insurance Commission staff and by examining Medicare registrations data as presented in the table below.
* The Commission cancelled the Medicare cards of persons whose cards had not been used for five or more years; hence the significant decrease in the figures.
SOURCE: Health Insurance Commission, Annual Report 1994-1995.
31 Resource constraints within the Health Insurance Commission have prevented them from statistically verifying why numbers of Medicare registrants grew by approximately twice the annual number experienced prior to the card reissuance program in 1991. This finding is contradictory to what had been anticipated, since a primary purpose of card reissuance was to identify and delete persons who'd died or were resident overseas during this program. The Health Insurance Commission has proposed that the growth could be explained by an increase in temporary (visitor) registrations. Specifically, it is suggested that increased temporary registrations could have resulted from extensions to the reciprocal Health Care arrangements for international visitors and changes to the New Zealand Health Care Service in 1992, which may have facilitated greater use of Australia's Medicare system by visiting New Zealanders. Alternatively, the Health Insurance Commission suggest the growth in registrations post 1992 could reflect re-registrations by eligible persons who were cancelled from the Medicare system as a consequence of the reissuance program. This latter suggestion in particular, raises questions about the success of the reissuance program to maintain the Medicare registrar, and the possible deleterious effects the ongoing program may have on the register and possibly population estimation.
Evaluation against ideal internal migration data criteria (refer Section 3)
32 In theory the Medicare population base is all Australian usual residents plus non Australian residents granted temporary registration. In practise however, a variety of Australian usual residents who are eligible to use Medicare are unlikely to do so consistently because of access to alternative health services (for example aborigines, defence force personnel, prisoners and persons eligible for Department of Veteran's Affairs Health Services). Conversely Medicare's population base is expected to exceed the ideal population base (i.e. all Australian usual residents) with regard to registrations relating to international visitors and former Australian usual residents who have died or are now permanently resident overseas but have not been deleted from the register. The extent to which the Medicare register is deficit/exceeds the ideal population base is not currently quantified.
33 The Health Insurance Commission anticipate coverage of children by the Medicare register, especially the recently born and toddlers, has been further enhanced by the introduction of the National Childhood Immunisation Register in January 1996. The National Child Immunisation Register is administered by the Health Insurance Commission as a subset of the Medicare register.
34 Medicare data meets all ideal data criteria with regard to data content, timeliness, historical availability and electronic access. Changes in register maintenance activities and reciprocated health care arrangements are expected to have altered data consistency during the 1990's. The impact these changes may have had on postcensal interstate migration estimation, has yet to be quantified.
35 With regard to future changes, discussions with a number of senior officers from the Health Insurance Commission indicate that while no one is sure what policy changes will be introduced by the present Government, speculated changes are unlikely to adversely effect Medicare data with regard to population estimation during this current term. For example, Medicare is not expected to be abandoned between 1996 and 1998 and eligibility for Medicare registration is not expected to become means/assets tested during this period either.
36 Medicare transfers data for all age groups is obtained from the Health Insurance Commission. (All age group data has been used in previous investigations of possible alternative interstate migration models and sub-State population distribution models.) Presently, the ABS interstate migration model uses Medicare transfers data relating to persons aged 1 to 14 years. The present model therefore contains transfers associated with short-term (visitor) registrants as well as permanent registrants aged 1 to 14 years. It is recommended an evaluation of the impact of short-term (visitor) transfers be undertaken with the objective of assessing whether such data should be excluded from Medicare transfers data used in interstate migration estimation in the future. Numbers of short-term registrants are expected to rise in the future and therefore so is the possibility that Medicare transfers data will capture details about movements involving short-term registrants. Short-term registrations are expected to rise in the future as a result of continued extensions in reciprocal arrangements for international health care and increased international tourism, due to the introduction of the electronic travel authority (ETA) system and the forthcoming Olympics.
37 It is recommended additional research be undertaken by the Health Insurance Commission to statistically verify why numbers of Medicare registrants grew at approximately twice the annual number experienced prior to the card reissuance program in 1991. The Health Insurance Commission is encouraged to undertake such research for its own benefit as it may raise questions about the success of the ongoing reissuance program to maintain the Medicare registrar.
38 With regard to maintenance of the Medicare register, the Commission is also encouraged to obtain deaths data from each State and Territory Registrar. The Medicare register is currently updated using deaths data from only some State Registrars because of data access cost considerations. This has a potentially biasing effect on Medicare data, impacting especially its use in sub-State population estimation/distribution models. In addition, with the Medicare card reissuance program being extended from a five year to a seven cycle, the accuracy of the Medicare register at any point in time is likely to be reduced.
39 The existence of linked child adult records offers the potential for a closer examination of child/parent migration patterns following adult separation.
40 It was recommended that a comparison of age related intercensal discrepancy data or 1996 Census interstate migration data be undertaken to reveal the potential benefit of augmenting Medicare data with alternative age specific data sets (refer paragraph 28). Similar US Census Bureau research suggests that if this recommendation is adopted, intercensal discrepancies for persons aged 65 years and over will exceed those for other age groups. The US research indicates that although migration by the elderly is currently small, it is rising and is quite dissimilar from migration which involves children. Australia's ageing population therefore, suggests present estimation methods could benefit from further investigation. Research to date however, indicates Australia lacks a good quality administrative data source other than Medicare for estimating migration by this age group. For example, age pension data has significant population coverage deficiencies (refer paragraph 80).
6 MONTHLY POPULATION SURVEY (LABOUR FORCE) DATA
41 Unlike other alternative data sources being investigated, estimates of internal migration from Monthly Population Survey are not a by product of an administrative data source. Details about the Monthly Population Survey and its historic and current use in capturing estimates of internal migration are provided below.
Monthly Population Survey and internal migration estimation
42 The Monthly Population Survey is a multistage area sample, principally designed to estimate the number of employed; unemployed; and those not in the labour force; for the civilian population aged 15 years and over at State and national levels (i.e. the Labour Force Survey). The Monthly Population Survey is also used to estimate other population based characteristics, such as internal migration. Data relating to additional topics can be collected either as supplements to the monthly Labour Force Survey or as Special Population Surveys. Differences in population sizes between the States are reflected in the Monthly Population Survey sample design by differential sampling fractions. Regional estimates are produced from this survey by weighting regionally stratified estimates using Capital City and Balance of State weights. Balance of State sample sizes tend to be relatively small especially in States with highly centralised populations (e.g. South Australia). Consequently regional estimates tend to have relatively high standard errors. Redesign and reselection of the sample after each Population Census ensures that changes in Australia's population distribution are incorporated into the Monthly Population Survey at five yearly intervals. The Monthly Population Survey also has the capacity to reflect significant regionalised growth between reselections but is less flexible with regard to population declines.
43 Between 1966 and 1987, the ABS conducted an Internal Migration Survey as part of the Monthly Labour Force Supplementary Survey Program. Amongst other uses, results from this survey were expected to be used to validate quarterly interstate migration estimates produced from administrative data sources, specifically Electoral Roll and Family Allowance transfers data. The Internal Migration Survey was discontinued after May 1987 because the quality of the estimates produced from this survey were considered substandard. The Internal Migration Survey estimates had high standard errors and contradicted estimates produced from administrative sources. In discontinuing the Internal Migration Survey, the ABS responded to user criticisms about the practice of publishing inconsistent and potentially confusing estimates.
44 In 1991, a single question relating to usual residence one year ago was introduced into the February Labour Force Experience Supplementary Survey. The question does not capture details about possible multiple migrations within that year. The incoming rotation group is excluded from answering this question, consequently it is asked of only seven eighths of the sample. Because of population growth, the size of the Monthly Population Survey sample is expected to grow from around 60,000 to exceed 65,000 persons between the 1992 and 1997 sample reselections.
45 In essence, internal migration can be viewed as a regional activity even when State borders are crossed since it involves movement from one area to another. Such movement is often disproportionate to the base population of the States involved and consequently differentials in the State sampling fraction will not necessarily reflect migration differentials between or within States. The ability of the Monthly Population Survey to capture and estimate population movements is expected to require special sample design considerations not inherent in the basic design. In particular, capture of movements outside capital city areas are expected to be especially limited by the current design. Sample design changes capable of affecting the Labour Force Survey will justifiably need to be treated with caution. The importance of this socio-economic indicator, the value of time series comparisons and the relative stability of the survey design since its inception in 1960, suggest processing modifications which impact internal migration estimates alone, may be more appropriate.
46 Table 1 below compares Monthly Population Survey estimates of interstate migration (year ended February 1991) with Medicare estimates revised after the 1991 Census (i.e. intercensal estimates) of interstate migration (year ended March 1991). In general the Monthly Population Survey underestimated all movements, resulting in significant overestimation of net movements. Some of this underestimation will reflect the inability of the Monthly Population Survey to estimate multiple migrations within a year. The variation in State estimates is however, less readily explained. The number of respondents found to be involved in migration via the Monthly Population Survey was around 2.0 per cent per annum in 1991.
Table 1: Estimated Interstate Migration of Persons aged 15 years and over, 1991.
1: Source: Monthly Population Survey, estimate of persons aged 15 years and over who moved within Australia during the twelve months ending February 1991.
2: Source: Revised Medicare estimates following the 1991 Census of Population and Housing, estimate of persons aged 15 years and over who moved within Australia during the 12 months ending 31 March 1991.
47 Comparison of Monthly Population Survey estimates and Medicare estimates of interstate migration for 1994 and 1995 are provided in Table 2 and Table 3, respectively. Whilst these later comparisons show significant differences in estimates, neither series can be confirmed until after the 1996 Census of Population and Housing has been conducted. It is anticipated however that the Monthly Population estimates will prove to be less accurate than the Medicare estimates of internal migration, as the conclusions are expected to replicate those found from the 1991 data comparisons.
Table 2: Estimated Interstate Migration of Persons aged 15 years and over, 1994.
1: Source: Monthly Population Survey , estimate of persons aged 15 years and over who moved within Australia during the twelve months ending February 1994.
2: Source: Medicare based estimate of persons aged 15 years and over who moved within Australia during the 12 months ending 31 March 1994.
Table 3: Estimated Interstate Migration of Persons aged 15 years and over, 1995.
1: Source: Monthly Population Survey , estimate of persons aged 15 years and over who moved within Australia during the twelve months ending February 1995.
2: Source: Medicare based estimate of persons aged 15 years and over who moved within Australia during the 12 months ending 31March 1995.
Evaluation against ideal internal migration data criteria (refer Section 3)
48 The Monthly Population Survey population base is less than the ideal population base as persons aged less than fifteen years and members of the Defence Forces are excluded from the survey. Details on internal migration are captured only annually from the Labour Force Experience Supplementary Survey and for the year ended February. The captured data meets ideal data content requirements except that movements are not referenced to specific quarters in a survey year. Results from the Labour Force Supplementary Survey tend to be available 6 weeks after the date of the survey but can not be considered timely as they are annual survey results. Monthly Population Survey data meet the ideal data criteria with regard to historical availability and electronic access. With regard to data consistency, changes to the Monthly Population Survey design have not been sufficient to ensure it is a source of good quality annual estimates of interstate migration (refer paragraph 46).
49 The data comparisons presented in this section suggest that Monthly Population Survey estimates can not currently be relied upon to validate annual estimates of interstate migration. The ABS is investigating what changes would be needed to allow reliable estimates of interstate migration to be produced from the Monthly Population Survey. If the required sample design or processing changes are impracticable, it is recommended collection of internal migration data from the Monthly Labour Force survey be discontinued.
50 As part of their response to the 1995 ABS Demography Review, the NSW Department of Urban Affairs and Planning recommended the scope and size of the Labour Force Survey be expanded to provide an alternate source to Medicare data. For the reasons specified above, increasing the sample size alone is not expected to produce greatly improved estimates of internal migration.
7 TAXATION DATA
51 Change of address details from Personal Income Tax files are currently used by Statistics Canada, the US Census Bureau and is soon to be used by Statistics New Zealand, to produce annual estimates of internal migration. This data source can not be used to produce quarterly estimates of internal migration as data is only available annually. Statistics Canada uses taxation derived annual estimates of interprovincial migration in preference to quarterly estimates of interprovincial migration prepared from child endowment style data. Taxation data is valued for its estimation of migration relating to persons aged 15 to 65 years in particular. Statistics Canada commenced using taxation data in 1976, at a time when their child endowment data gave 100 per cent coverage of address details relating to children aged under 18 years, and despite the inability of annual taxation data to estimate multiple migration within a year.
52 The ABS undertook an extensive evaluation of taxation data as part of research into alternative administrative data sources in 1990. A number of limitations associated with taxation data were identified from this 1990 research and many were consistent with limitations reported by Statistics Canada and the US Census Bureau. The principal limitation was lack of data timeliness. Specifically, data would not become available from the ATO until 10 months after capture. Consequently the reference period of interstate migration estimates would be lagged by more than one year.
53 Recent discussions with the ATO focused on identifying any changes in collection or policy practises which may have altered the 1990 findings. In summary, whilst a number of changes have occurred since then, the principal result is that access to taxation data will be even less timely although data is electronically captured for more Australian taxpayers than in 1990.
Evaluation against ideal internal migration data criteria (refer Section 3)
54 The population available to be covered from personal income tax records is all persons who submit tax returns to the Australian government in two consecutive years and their identified spouses/partners. Dependant child details are not captured on tax returns but estimates of numbers of dependant children may be possible from rebate details. This population base differs from the ideal population base (i.e. all Australian usual residents) for those Australian usual residents who do not submit income tax returns in two consecutive years and for those persons who submit returns but are not usual Australian residents. In particular, coverage for persons aged over 65 years tends to be particularly inconsistent.
55 Data entry policy changes and the introduction of direct electronic lodgement since 1990 have resulted in the ATO maintaining details on all submitted returns not just those considered 'taxable'. Individual returns (tax-pack forms) are required to be lodged by October of the referenced financial year unless submitted via an agent. Agent submitted returns are not required to be lodged until March of the year following the reference year. In 1993/94, agent submitted returns accounted for 73 per cent of all (9.3 million) returns. To ensure unimpeded processing of returns, the ATO currently prohibits use of their databases for non tax purposes until 12 months after the reference financial year. This means the reference period of interstate migration estimates would be lagged well in excess of one year.
56 The captured data is expected to meet ideal data content requirements except that movements will not be referenced to specific quarters in a survey year. Taxation data is also expected to be able to meet ideal historical and data consistency requirements.
57 Taxation data is unsuitable for use in quarterly estimation of interstate migration because it is collected annually. It's potential for validating or augmenting quarterly estimates is impaired by timeliness issues and problems associated with comparing quarterly estimates and annual estimates of migration because of the possibility of multiple migration within a year. However since international experiences indicate taxation data is a good source for estimating internal migration involving persons aged 15 to 65 years, it is recommended that a comparison of age related interstate migration data from the 1996 Census be undertaken to reveal the potential benefit of augmenting Medicare data with age specific data sets. If the results of this analysis are positive, it is recommended discussions be held with the ATO regarding the possibility of improving timely access to taxation data.
8 FINANCIAL INSTITUTIONS DATA
58 Investigations into the likelihood of sourcing quality change of address data from financial institutions commenced with discussions with officers from ABS Finance Section in Canberra. As Finance Section was due to start a series of formal consultations with Australia's major banks, it was hoped preliminary details about problems the major banks envisages in supplying change of residential address data could be sought on Demography's behalf. During the May 1996 consultations, Finance Section officers decided against pursuing such information because of the undue burden such discussions would add to meetings that were already extremely demanding. The principal purpose of these consultations was to advance data requirement changes which had not been well received when first proposed at a Bankers Association meeting held in Sydney on 13 March, 1996.
59 To be of value in estimating interstate migration, it is anticipated financial institutions will need to provide selected data relating to bank accounts in which a residential address change occurred during the reference quarter. Bank account holdings rather than credit card holdings are expected to cover a greater proportion of the Australian usual resident population since segments of the population (e.g. low income earners, children etc.) are expected to be ineligible for credit card accounts. Details on name, age (date of birth), sex, old residential postcode and new residential postcode will be required. Name data is considered necessary because of the possibility of persons having an account with more than one financial institution or having more than one account with one institution.
60 If undertaken, extraction of migration information from financial institution data is expected to require the development and maintenance of an extensive database. Record matching will be required to avoid the possibility of double (multiple) counting occurring as a result of persons having a bank account with more than one financial institution. The success of detecting replicated address changes will depend, amongst other factors, on customers notifying more than one institution of an address change during the same quarter and that each institution processes this information during the same quarter.
Evaluation against ideal internal migration data criteria (refer Section 3)
61 Data relating to bank account holdings are expected to be greater than the ideal population base (i.e. all Australian usual residents) as non Australian usual residents are also expected to have bank accounts in Australia. Address details may allow accounts held by persons usually resident overseas to be identified and excluded from coverage. Data relating to name, age, sex, old residential postcode and new residential postcode are expected to meet data content requirements. Information specific to anticipated data delivery times, historical availability, and data consistency are not currently known. Initial discussions with ABS officers in Finance Section however, suggest national data may not be readily available from some banks because of operational size considerations and or deliberate regionalisation policies.
62 Preliminary discussions with ABS officers working with the financial sector indicate this sector currently finds ABS data demands extremely burdensome. Furthermore there is some contention as to whether all financial institutions would be able to provide national change of address data. In particular, the Commonwealth and ANZ banks are expected to find this difficult because of operational size considerations and regionalisation policies, respectively.
63 Whilst financial institutions may capture data which would assist with interstate migration estimation, it is expected to be unduly expensive in terms of reporting load considerations and data storage and processing costs, to obtain and extract the necessary information. Consequently, it is recommended financial institutions data not be pursued further at this point.
9 AUSTRALIA POST DATA
64 In addition to maintaining a database on national residential delivery points (i.e. house and unit addresses), Australia Post operates a commercial address change service. Details relating to permanent address changes are held on an address change database. The database captures old and new address details along with the name/s of the household member/s whose mail is to be redirected from this address.
65 Australia Post believe their address change facility is currently used by about 70 per cent of households which move within Australia (i.e. either intrastate or interstate). The data comparisons below suggest these household estimates translate to approximately 45 per cent of persons who move interstate. Substantial regional variations in usage of this facility are apparent even at State level.
Comparison of Estimated Interstate Migration data with Australia Post data,
by States and Territories of Arrival and Departure, September Quarter 1995.
1: Source: Australia Post, estimate of persons who moved within Australia during September Quarter 1995 based on change of address data.
2: Source: Australian Demographic Statistics, September Quarter 1995, Cat. No: 3101.0, estimate of persons who moved within Australia during September Quarter 1995 based on Medicare data, preliminary estimates.
Evaluation against ideal internal migration data criteria (refer Section 3)
66 The population base of Australia Post's Change of Address database is all persons (usual resident of Australia or otherwise) with an active permanent redirection notice and others who are nominated on these notices. It is difficult to assess how well this population base meets the ideal population base (i.e. all Australian usual residents). An evaluation of Australia Post data in 1994 suggests service usage is inversely related to distance moved. Definitely, since overseas redirections are not provided by this service, it appears unlikely that emigrants or overseas visitors would use it unless looking to redirect mail to another Australian address for subsequent private redirection overseas. However, there are a number of other possible factors which may also influence service usage. For example, income, intended duration of migration and whether or not all household members will be involved in the move. In addition, age and or English literacy skill may dictate which household members are nominated on redirection forms. In essence, Australia Post's change of address population base is unlikely to be a random sample of Australia's usual residents. Consequently, unless specific characteristics can be identified from within the Australia Post change of address population base, unbiased estimates of interstate migration are not possible from this data source.
67 With regard to the data content criterion, the absence of age details is a significant data deficiency. Furthermore, Australia Post reference dates will relate to the date the redirection notice becomes active and this date may or may not coincide with date of migration. For example, variations in arrival and departure rates (refer table above) may reflect significant regional differences in submission lead/lag times (refer NT and Qld). Since submission of change of address notifications depends on knowing a destination address, variabilities can be expected because of factors such as differences in regional housing vacancies and access to temporary accommodation arrangements (e.g. staying with friends or family). An assessment of submission lead/lag times is needed before Australia Post reference date data can be used with confidence.
68 Australia Post change of address data is expected to be able to meet the ideal data criteria with regard to data timeliness, historical availability and electronic access.
69 In theory, Australia Post change of address data has considerable potential to assist interstate migration estimation. Current lack of information about the characteristics of users of this address change service however, prevent this data source from being able to be used to prepare unbiased estimates of interstate migration. Consequently, at present Australia Post change of address data is not considered a good alternative administrative data source to Medicare data.
70 Details about the propensity for Australia's usual residents to use Australia Post's change of address services is required. Investigations will need to identify the characteristics of direct users and indirect users (i.e. other nominated household members). Details about submission practices in relation to date of migration also need to be determined. It is recommended the ABS encourage Australia Post to undertake these investigations for their own benefit. Australia Post is aware of the market potential for electronic address change data.
71 Although now dated and restricted to only two Queensland LGAs, the result of the 1994 analysis of Australia Post data indicates a potential for change of address data to assist with sub-State population distribution. Further data investigations will be required if this option is to be pursued.
10 TELEPHONE CONNECTIONS DATA
72 At present, telephone connections data is a household based administrative data collection. Whilst this does not prevent the data source from being used to estimate interstate migration, it would require substantive changes to be made to the estimation methodology presently used by the ABS. Furthermore, although the latest estimates indicate 94.4 % of households have at least one telephone connection (ABS, Household Telephone Connections Survey, August 1991. Catalogue No: 4110.0), Telstra is unable to assure change of address details except for phone transfers (i.e. short distance migrations). Thus telephone connections data is not presently recommended as an alternative to Medicare data or even as a data source to augment Medicare data.
73 Telephone connections data is expected to increasingly become a person based data source. Telstra is presently trialing personal numbers (based on mobile phone technology) in Melbourne. Known as Telepath, this service is designed to allow phone users to have personal phone numbers, eliminating the need for separate home, business, mobile and pager numbers. Personal numbers based on mobile phone technology are however, subject to the range limitations associated with mobile phones. Consequently, Telstra is also developing phone number portability from fixed network based technology. Significant technological developments are required to current telecommunications systems to allow personal numbers to be unique throughout Australia. This is necessary before phone number portability is possible from fixed network based technology, consequently this service is not expected to be possible before 1998.
74 The introduction of personal phone numbers is expected to require significant changes to telephone administrative data. Details on changes to billing addresses for portable personal phone numbers are expected to offer an excellent potential source of internal migration data. It is recommended the ABS initiate high level discussions with Telstra and Optus with a view to negotiating quarterly access to these future data sources. Discussions with Telstra to date, indicate successful negotiations will involve protracted consultations on issues including data confidentiality. Consequently, it is recommended discussions commence shortly with both agencies to maximize the potential for ABS advice to be implemented with regard to database development.
11 SOCIAL SECURITY DATA
75 All Social Security payments are subject to means testing (i.e. income and assets tested, with some assets considered to have income potential according to deeming rules) of individuals as well as their spouses/partners. At present, immigrants are required to wait 6 months before becoming eligible for Social Security payments (except Family Allowance payments). Media reports indicate the Government is currently looking at extending this waiting period to two years and Family Payment eligibility may not be exempt. Details specific to Unemployment, Family Payment and Age Pension data are provided separately below.
76 With regard to unemployment data, recipients are required to lodge declarations fortnightly. Consequently residential address data is potentially current within a two week time frame. The reliability of this address data is expected to have declined however, since unemployment allowances began being paid directly into bank accounts rather than being posted to recipients. Change of address data is not specifically captured by the unemployment database but it is expected to be available by comparison of database snapshots at different points in time. Population coverage issues are expected to limit the value of unemployment data. Means tests prevent all unemployed persons from being eligible to receive benefits while some low income employed persons are eligible for partial payments. As a consequence, it will be difficult to find appropriate benchmarks against which unemployment transfers data can be weighted to produce estimates of internal migration. To further exacerbate this issue, migration amongst the unemployed is unlikely to reflect general migration patterns.
Family Allowance Data
77 Social Security holds details about the number of children eligible for benefit. Change of address details are not expected to be updated more frequently than annually as payment of Family Allowances is made directly and electronically into bank accounts. Payments are assessed against the previous year's taxable income. Such details are requested each year before payments continue. It is expected change of address data could be obtained by comparing separate snapshots of this file at different points in time.
78 An examination of Family Allowance data revealed the number of recipients had declined significantly over the past five years. The following is a listing of policy changes to Family Allowance eligibility since 1985, which are expected to have impacted on data comparability over this period.
1985 Dependent children aged 18-24 years excluded from Basic Family payment (extended to exclude dependent children aged 18-24 of pension recipients in 1990)
1987 Children who have never lived in Australia, excluded from all Family payments
1987 Introduction of Income test for all Family payments
1989 Children who have lived overseas for 3 or more years, excluded from all Family payments
1992 Introduction of Assets tests for Basic Family payment
1994 Introduction of Income and Assets tests for all Family payments, including deeming rules in relation to specific assets.
Age Pension Data
79 Eligibility for an age pension allowance is subject to means testing of the individual and the spouse. Change of address details are not expected to be updated more frequently than annually as payment of Age Pension Allowances is made directly into bank accounts.
80 An extensive assessment of age pension data and its value in estimating interstate migration for persons aged 60 years and over, was undertaken as part of investigations into alternative data sources to Medicare in 1990. The summary conclusion from this investigation was that whilst age pension transfers would probably be adequate to estimate the total number of interstate movers for Australia as a whole in the 60 plus age range, coverage would be insufficient to estimate net interstate gains or losses for individual States and Territories.
Evaluation against ideal internal migration data criteria (refer Section 3)
81 As a consequence of the eligibility conditions which apply to all Social Security payments, the population bases associated with the types of Social Security data outline above are less than the ideal population base (i.e. all Australian usual residents). The introduction of significant policy changes to eligibility conditions, especially with regard to Family Allowance Payments, in the past have impaired the historical consistency of these datasets. Notification of address changes are required only annually for Age Pension and Family Allowance payments and in general, payment by direct electronic transfer, limits the ability of Social Security data to be used with confidence to estimate internal migration within Australia. Past and expected changes to Social Security eligibilities are expected to impair data consistency. Social Security data is expected to meet ideal data criteria with regard to time-lags and electronic availability.
82 Changes (both past and expected) to Social Security Allowance policies limit population coverage of associated payment datasets and have impaired their historically comparability. Furthermore population coverage limitations associated with these datasets are more extensive than Medicare. In addition, since address details of Age Pension and Family Allowance recipients are updated annually, these datasets can not be used in quarterly estimation of interstate migration. Consequently, it is recommended Social Security data not be considered a viable replacement for Medicare data at present.
12 SCHOOL ENROLMENT DATA
83 A comprehensive assessment of school enrolment data for use in estimation of interstate migration was undertaken by the ABS as part of investigations into alternative administrative data sources to Medicare in 1990. The conclusion was that interstate migration data were not available for school students and could not be derived from existing administrative records at the State or national level. However, the data were generally captured (though not necessarily in the required format) by schools for new enrolments and could conceivably be collected for all enrolments and collated as part of the National Schools Statistics Collection (NSSC), results of which are published by the ABS.
84 Data comparisons undertaken during the research in 1990 revealed a high correlation between head counts and ERP at 30 June. This suggested the potential for developing estimates of interstate migration for children aged 6 to 14 years from school enrolment data.
85 Recent discussions with ABS personnel indicate that comprehensive data relating to student transfers may be kept by schools but such data are not collected as part of the annual NSSC. In addition, school censuses are conducted more frequently than annually by the States and Territories. Consequently, it is recommended State and Territory Education departments be encouraged to capture change of address details as school enrolment transfers data offer great potential as a validating data source of interstate migration for children aged 6 to 14 years. In addition, such data are expected to assist research into student retention rates.
86 The ABS is presently investigating the possibility of obtaining student interstate transfers data from State and Territory Education departments. If preliminary enquires look promising the issue may be pursued via the Ministerial Council on Education, Employment, Training and Youth Affairs (MCEETYA) Task Force on School Statistics, of which the ABS is a member.
13 OTHER ADMINISTRATIVE DATA
87 A number of other commonly used datasets were excluded from this investigation. The excluded datasets included electricity connections, residential building permit applications, electoral roll registrations and motor vehicle registrations. The decision to exclude these other data sources was based on the results of previous investigations by the ABS or other international organisations. In general, past investigations revealed criteria deficiencies which exceed those associated with Medicare data and which were considered unlikely to have altered significantly since last evaluated. For example, electricity connections transfers data is not available nationally and connections data (like building permit data) is subject to fluctuations in the building industry as well as population change. Similarly, past research, most recently in New Zealand and Canada, reconfirms motor vehicle registrations data (like drivers licence and electoral roll data) is problematic because of timeliness considerations. This is despite the legal requirement for people to notify authorities of address changes within a fixed time period following migration. These conclusions do not necessarily limit the value of these datasets in sub-State population estimation but such evaluations are outside the scope of this paper.
Australian Bureau of Statistics (C.Y. Choi),  - Population Estimates in Australia - a discussion paper. Australian Bureau of Statistics: Demography and Social Branch, R-Series, Demog-1.
Australian Bureau of Statistics,  - Methods and procedures in the compilation of estimated resident population 1981 and in the construction of the 1971-81 time series. Australian Bureau of Statistics: Catalogue No. 3103.0
Australian Bureau of Statistics (O.B. Di Iulio),  - Postcensal Interstate Migration Estimates 1966-1981. Australian Bureau of Statistics: Occasional Paper 1984/2
Australian Bureau of Statistics (C.Y. Choi),  - Report on an Overseas Visit to the Statistical Offices of the UK, Netherlands, USA and Canada, 7 November to 7 December 1996. Unpublished working paper.
Australian Bureau of Statistics,  - Internal Migration, Australia 12 months ended 31 May 1987. Australian Bureau of Statistics: Catalogue No. 3408.0
Australian Bureau of Statistics, [c1990] - Alternative Estimates of Interstate Migration. Unpublished working paper.
Australian Bureau of Statistics,  - Household Telephone Connections, Australia, August 1991. Australian Bureau of Statistics: Catalogue No: 4110.0
Australian Bureau of Statistics,  - 1991 Census Data Quality - Undercount. Australian Bureau of Statistics: Catalogue No: 2940.0
Australian Bureau of Statistics (P. Trickett),  - Postcensal estimates of Interstate migration - an evaluation 1986-91 estimates and proposed method for 1991-96. Unpublished working paper.
Australian Bureau of Statistics,  - Population Growth and Distribution in Australia. Australian Bureau of Statistics: Catalogue No. 2822.0, 1991 Census.
Australian Bureau of Statistics - unpublished Labour Force estimates of interstate migration, 1990-1994.
Australian Bureau of Statistics - unpublished Medicare estimates of interstate migration, 1990-1994.
Australian Bureau of Statistics,  - Population Estimates: concepts, sources and methods. Australian Bureau of Statistics: Catalogue No. 3228.0
Australian Bureau of Statistics,  - Labour Force Survey Sample. Australian Bureau of Statistics: Catalogue No. 6269.0
Australian Bureau of Statistics,  - Australian Demographic Statistics, September Quarter 1995. Australian Bureau of Statistics: Catalogue No. 3101.0
Australian Bureau of Statistics (W. McLennan),  - Taxation data and economic statistics.(speech notes) Address to National Taxation Liaison Group Meeting, 5 March 1996.
Australia Post,  - Customer List Interface Subscriber Guide. Address Post: Version 1.6
Committee on Population Statistics,  - Migration Statistics in the United States. The Carolina Population Center, University of North Carolina at Chapel Hill.
Department of International Economic and Social Affairs,  - Preparing Migration Data for Subnational Population Projections. United Nations, New York.
Department of Social Security (J. Robinson),  - Data requirements of the Department of Social Security : an information paper on the use of statistics, prepared for the Australian Statistics Advisory Council. Unpublished working paper.
Health Insurance Commission,  - Annual Report 1987- 88. Health Insurance Commission.
Health Insurance Commission,  - Annual Report 1994 - 1995. Health Insurance Commission.
Office for National Statistics (A. Teague),  - 2001 Census coverage validation Diffusion, No 13, June 1996.
Office of Population Censuses & Surveys,  - Patterns of migration from the 1981 Census. OPCS Population Trends 35: Spring 1984
Office of Population Censuses & Surveys,  - A Review of the 1981 Census of Population in England and Wales. OPCS Population Trends 36: Summer 1984
Office of Population Censuses & Surveys,  - First results from the quality check element of the 1991 Census Validation Survey. OPCS Monitor: SS 94/2
Office of Population Censuses & Surveys,  - Undercoverage in Great Britain. OPCS, General Register Office of Scotland: 1991 Census User Guide 58.
Office of Population Censuses & Surveys,  - 1981 Census Evaluation Program in England and Wales. OPCS Population Trends 40: Summer 1995
Shryock, H.S. and Siegel, J.S.,  - The methods and materials of demography, condensed edition by Edward G Stockwell. Academic Press, New York.
Statistics Canada (Verma, Basavarajappa & Bender),  - New approaches to methods of estimating the population of census divisions. Unpublished working paper.
Statistics Canada (Norris and Standish),  - A technical report on the development of migration data from taxation records. Unpublished working paper.
Statistics Canada,  - Population estimation methods, Canada. Statistics Canada: Catalogue 91-528E.
Statistics Canada,  - Report on the Demographic Situation in Canada 1994 , current demographic analysis. Statistics Canada: Catalogue 91-209E Annual.
Statistics Canada,  - Annual Demographic Statistics, 1994. Statistics Canada: Catalogue 91-213 Annual.
Statistics Canada (R. Bender and M. Bedard),  - Issues in estimating the population of Canadian municipalities. Unpublished seminar paper.
Statistics New Zealand (B. Boddington),  - New Zealand Population Estimation. Unpublished working paper.
Steel, D., Vella J., and Harrington, P,  - Quality issues in telephone surveys: coverage, non-response and quota sampling. Australian Journal of Statistics, Vol 38, No. 1 pp 15 - 34.
US Bureau of the Census (D. L. Ward),  - The Census Bureau approach for allocating internal migration to states, counties and places: 1981 - 1991. Technical Working Paper No. 1.
US Bureau of the Census (J. F. Long),  - Postcensal population estimates: states, counties and places. Technical Working Paper No. 3.
US Bureau of the Census (S.T. Davis),  - Evaluation of postcensal county estimates for the 1980s. Technical Working Paper No. 5.
Verma, Ravi B.P. and Parent, Pierre,  - An overview of the strengths and weaknesses of the selected administrative data files. Journal of Statistics Canada, Vol 11, No. 2.
See also Demography Working Paper 99/2 - Estimating Interstate Migration, 1996 - 2001
These documents will be presented in a new window.