APPENDIX 1 INTERNATIONAL STUDIES OF INTERNAL MIGRATION ESTIMATION - METHODS AND SOURCES


New Zealand
United States of America
Canada
United Kingdom


New Zealand

Statistics New Zealand has published The Internal Migration Report (Statistics NZ, 2006) which provides a statistical overview of internal migration in New Zealand. It is based on information from the Census of Population and Dwellings and the Survey of Dynamics and Motivations for Migration in New Zealand. All data for this report are derived from the 1981 to 2006 Censuses, so the data are available every five years - specifically the questions on usual residence at the time of the Census and five years before the Census. The data are published for 16 major regions and includes arrivals, departures and net migration rates.

The proportion of respondents for whom data on previous residence are missing or incomplete has been rising from Census to Census, and reached 8 percent in 2006. To account for this, Statistics New Zealand adjusts the Census migration counts by assuming that (i) people who did not give any response had the same distribution by previous residence (including residence overseas) as other people living in their region, and (ii) people who stated that they were in New Zealand but did not clearly specify a region had the same distribution by previous residence as other people living in their region whose previous residence was in New Zealand.

Subnational resident population estimates are derived by Statistics New Zealand by updating the estimated resident population of each area at 30 June for births, deaths, net international migration, and net internal migration. Administrative data sources (e.g. residential building consents, electoral enrolments, Inland Revenue Department (IRD) data, school rolls data, university students by university location; offenders by prison location and armed forces by living location) are used to estimate the internal migration component. Estimates are available at regional council area, territorial authority area and urban area boundaries and are produced quarterly (Statistics NZ, 2009).

Continued improvements in the maintenance of the IRD administrative system to reflect the population of interest are necessary to make the IRD data series a valuable input to the estimation of internal migration. Changes in the school attended, school zoning and other factors limit the usefulness of school rolls as indicators of population change.

Statistics New Zealand notes that population estimates for small geographic areas (e.g. area units) are more uncertain than estimates released at broader geographic levels. This added uncertainty arises because information on external and internal migration is not readily available at the area unit level and localised developments (e.g. greenfield subdivisions) can have a significant impact on the net migration patterns and population structure of area units, but this impact can be difficult to discern.

United States of America

The US Census Bureau estimates internal migration for states and also for counties (US Census Bureau 2009), the primary legal divisions of most states.

To measure internal migration in the United States, the US Census Bureau uses data from Federal income tax returns and Medicare and makes separate estimates for the population aged under 65, and 65 and over. These estimates are combined to arrive at migration estimates for the total population.

To estimate internal migration for the population aged under 65, two years of tax returns are matched and the addresses are compared to identify the number of movers from one county to another between tax filings. For persons who did not fill out a tax return, migration rates by county are assumed to be the same as those who filed a return.

To measure inter-county migration for the age 65 and over population, Medicare enrolment data are used. It is assumed that the year-to-year change in enrolment represents the total change in the age 65 and over population in each county. Independent estimates of the deaths and international migration of the age 65 and over population are benchmarked to the Medicare data, with the population change not attributable to either deaths or international migration assumed to be due to net domestic migration. A Medicare-based net migration rate is calculated for each county by dividing the net internal migration estimate by the total number of Medicare enrollees at the beginning of the time period - this accounts for those who do not receive Medicare.

Most administrative record data sources lag the current estimate year (by as much as two years), therefore, they project the data for the current year based on past years’ data. As updated data become available, the US Census Bureau revises the projected input data so that each year’s estimates are always based on the most recent data available.

These estimates of internal migration are used to calculate annual population estimates for each US state and county, where population change is estimated by adding natural increase, net international migration and net change in group quarters population with the internal migration estimates.

Canada

An annual series of internal migration estimates is produced by Statistics Canada (Statistics Canada, 2008). These estimates are prepared for Census Divisions (available since 1976-77) and Census Metropolitan Areas (since 1992-93).

These migration estimates are derived from a comparison of addresses from individual income tax returns for two consecutive years. The development of the data then involves four main steps:

• geocoding of tax records;
• estimation of non-filing dependants of taxfilers, by age group and sex;
• identification of the number, age and sex of migrant taxfilers; and
• adjustment for the population not covered by the taxation system.

The postal code on each tax record is matched to a geography basefile to pick up additional geographic information such as Census Division and Census Metropolitan Area. However, the use of postal codes can result in incongruity between postal code service areas and Census Divisions. The problem may arise in rural areas where a postal code service area is large and crosses one or more Census Division boundaries. Since the postal code is assigned to only one Census Division, a certain amount of misallocation can occur. There may be positive and negative misallocation that cancels out in some Census Divisions, but geographic misallocation may result in net under or over-coverage. Provincial totals are not affected by this problem.

The system estimates taxfilers' dependants by linking all filing family members together and estimating non-filing members from information on the taxfilers' returns, based on information as for dependants.

After accounting for the time taken to submit tax forms and processing, estimates are generally available 15 to 18 months after the income tax deadline.

In addition to the estimates prepared using annual tax data, estimates of interprovincial migration are also prepared based on monthly Child Tax Benefit records.

United Kingdom

The Office of National Statistics (ONS) produces estimates of internal migration within the UK at the country level and at lower levels, such as Government Office Regions and Local Authorities (ONS, 2011). These estimates are derived from two administrative data sources, namely the National Health Service Central Register (NHSCR) and the Patient Register Data System (PRDS). These sources have different properties in terms of coverage and content. To estimate movements between local authorities, the NHSCR figures are combined with the more detailed PRDS. Comparing records in one year with those of the previous year by linking based on NHS number enables identification of people who change their postcode.

Every NHS authority holds a record of NHS patients registered with GPs in their area comprising NHS number, sex, date of birth, date of acceptance and patient’s address. ONS receive an annual mid-year download from each authority’s register which, when combined, forms a total patient register for England and Wales. Comparing registers from consecutive years, linked by NHS number, ONS identifies all individuals whose postcode differs between the two years. These counts are then constrained to NHSCR Health Authority (HAs) level data to produce internal migration estimates.

Data from the NHSCR system are limited to movements between the former health authorities. These former health authorities no longer exist as administrative entities but continue to be used due to technical constraints of the NHSCR processing system. PRDS data are also inadequate as a stand alone source as a number of within year moves are not captured.

Limitations of estimates derived from a combination of NHSCR and PRDS include: variation in the delay between a person moving and registering with a new doctor; some moves may not result in a GP re-registration and therefore will not be recorded, and individuals may move and re-register more than once in a single year but remain within the same health authority

Despite some limitations, research has shown that there is no other source that has as good coverage and quality as the NHS administrative data (NHSCR and PRDS) for estimating internal migration within the UK.

These estimates, along with births, deaths and international migration, form the key components of the annual population estimates produced by the ONS.