29. This section covers the data sources each NSO uses to estimate emigration applying the UN definition. All 11 NSOs surveyed follow the UN definition of a (long-term) migrant to estimate emigration which, by definition, excludes short-term movements.

30. Of the 11 country responses, a variety of data sources were used, with 5 responding that they use an ‘Other administrative database’ (Table 1). However, each country uses multiple data sources (Table 2).

Data source	Country
Population Census	Albania, Italy, Poland, Mexico, United States
Household sample survey	Albania, Mexico, United States
Border sample survey	United Kingdom
Population register	Italy, Latvia, Poland
Permits of stay register, aliens register	Albania, Hungary, Poland
Passenger cards system, flight data	Australia, United Kingdom
Other administrative database	Canada, Hungary, Italy, New Zealand, United Kingdom

31. The respondents were asked to provide a detailed description of the data that they use.

32. Albania uses a population census to produce an estimate of the number of individuals who have emigrated between the two censuses. A household sample survey is used which includes data collected from the migration section of the labour force survey combined with population projection rates. Data on Asylum Seekers and the acquisition or loss of citizenship in Albania are provided by the Department for Border and Migration at the General Directorate of State Police, Directorate of Policies for Asylum, Foreigners and Citizenship at the Ministry of Interior.

33. Australia uses administrative data supplied by the Department of Home Affairs. This data contains unit record data of all border crossings and are sequenced for each person to determine durations in and out of Australia. The 12 months threshold to be considered resident does not have to be continuous and is measured over a 16-month period.

34. To estimate emigration from Canada to the United States, data from the Office of Immigration Statistics of the US Department of Homeland Security (DHS) provides information on Canadians who acquire long-term residency rights in the United States. For estimating emigration to countries other than the United States, Canada relies on tax data.

35. Hungary uses data from the National Health Insurance Fund to estimate the emigration of Hungarian citizens. Estimates of emigration of foreign citizens are based on data from the National Directorate-General for Aliens Policing (NDGAP). This includes registration certificates and settlement permits of foreign citizens from EU and EEA countries, as well as third-country citizens.

Country	Order of importance
	1	2	3
Albania	Population Census	Household surveys	Permits of stay register, aliens register
Australia	Passenger cards system, flight data	Medicare data (secondary source for state or territory of residence)
Canada	Tax data from Canada Child Benefit (CCB) program from Canada Revenue Agency (CRA)	T1 Family File (T1FF) tax data from CRA	Immigration Statistics of U.S. Department of Homeland Security (DHS)
Hungary	Register of National Directorate-General for Aliens Policing	National Health Insurance Fund
Italy	Population Register	Population Census	Other administrative database
Latvia	Population estimation	Population register	Mirror statistics
Mexico	Population Census	Household sample survey
New Zealand	Border crossings passport data	Other country (Australia) data on migrant arrivals (from New Zealand)
Poland	Population Register	Population Census	Permits of stay register; aliens register
United Kingdom	Passenger cards system, flight data	Other administrative database (tax data)	Border sample survey
United States	Household Survey (American Community Survey)	Population Census

36. Italy uses a population register to count the number of immigrants and emigrants in the reference year. The data used to determine migration flows are derived from the change of residence of the resident population. Data on residence permits refer only to non-EU citizens, so are used as a signal of life to capture the over- or under-coverage component of the census population.

37. Latvia uses the population register to count the number of immigrants and emigrants in the reference year. For migration adjustments, if the country is not known, mirror statistics are used.

38. Mexico uses the Population and Housing Census, or specialized surveys such as the National Survey of Demographic Dynamics (ENADID); using specific questions that allow the identification of people who have changed their place of residence. Topics about Recent International Emigration ask whether any household members migrated to another country in a reference period (e.g., in the last 5 years).

39. New Zealand receives a daily supply of all border crossings into and out of New Zealand and applies a 12/16-month rule, meaning the UN 12-month threshold does not need to be a consecutive 12 months, but can be cumulative over a 16-month window. Australian migrant arrival data is used to estimate migrant departures from New Zealand to Australia (historically, the main country of destination for emigrants).

40. Poland uses as one of the main sources a permit of stay, aliens and population register to estimate emigration.

41. The United Kingdom uses passenger cards system/flight data to link an individual’s visa to their travel into and out of the UK. They also use a database of everyone in the UK who has a National Insurance number which details their interactions with the tax and benefits system. A border sample survey is also used but becoming decreasingly used given its limitations.

42. United States use an indirect (residual) method, so emigration is not directly derived from a single survey question/section.

43. Respondents to the questionnaire did not include countries that produce emigration statistics using comprehensive population register data. While such countries (e.g. Denmark, Finland and Sweden) also use additional administrative data sources to identify persons who may have left the country without submitting a change of address notification, they do not need to rely on combining multiple data sources, statistical modelling or data obtained from destination countries.

44. Of the 11 country responses, 5 responded that they have considered alternative or new data sources to estimate emigration, with 4 of these considering tax data (Table 3).

Country	Data Source
Canada	Population Census Household sample survey (Census Undercoverage Study (CUS)) Passenger cards system, flight data
Hungary	Population register Permits of stay register, aliens register Other administrative database (Education Office, National Tax and Customs Office)
Poland	Other administrative database (Tax data)
United Kingdom	Other administrative database (Tax data)
United States	Other administrative database (Tax data)

45. Australia indicated that Medicare data is linked to the travel data as a secondary source for state or territory of residence through probabilistic linking using name, date of birth and sex.

46. New Zealand and Poland indicated that they use linked data sources to estimate emigration.

47. New Zealand uses border-crossings passport data – linking arrivals and departures to ascertain precise duration of stay and absence – with the key linking variables being passport number, date of birth, and name (although all 3 are not needed to enable a link).

48. Poland used data received from main destination countries to estimate Polish emigrant stocks. These were linked on macro level using aggregated data on mirror statistics from destination countries.

49. Based on the available data sources, various countries output different variables at different detail levels. This section summarises both the variables, detail levels, and the desired expansions to those details.

50. Various breakdowns of emigration statistics are produced by countries, but all would like to expand the available breakdowns (Table 4). Where countries currently produce a breakdown and have indicated they also want to produce that breakdown, we have interpreted that to mean that they currently produce a limited version of that variable but would like to expand on that (i.e., more detail).

Breakdown	Currently produce	Do not currently produce but want to (or want to expand breakdown)
Demographics (age/sex)	Australia, Canada, Hungary, Italy, Latvia, Mexico, New Zealand, Poland, United Kingdom	Albania, United States
Country of destination	Hungary, Italy, Latvia, Mexico, New Zealand, Poland	Albania, Australia, Canada, Hungary, New Zealand, United Kingdom, United States
Country of birth	Australia, Hungary, Italy, Latvia, Mexico, Poland	Albania, Canada, New Zealand, United Kingdom, United States
Citizenship	Australia, Hungary, Italy, Latvia, New Zealand, Poland	Albania, Canada, Mexico, United States
Nationality	Australia, Italy, United Kingdom, United States	Albania, Hungary, Mexico, United States
Skill type (occupation or education level)	Italy	Albania, Hungary, Latvia, Mexico, New Zealand, Poland
Reason for emigration	Mexico, United Kingdom	Albania, Hungary, Poland

Note: In some countries ‘nationality’ and ‘citizenship’ are synonymous, but in others they can be different. For example, people may be citizens of the United Kingdom but separately identify as English, Scottish, Welsh, etc.

51. Australia, New Zealand and United Kingdom indicated that they use additional data sources to provide the above breakdowns.

52. In New Zealand, citizenship and age/sex come from passport data collected at the New Zealand border, while the country of destination (Australia) relies on data collected at the Australian border (supplied by the Australian Bureau of Statistics). This is the only country of destination that New Zealand currently produces estimates for, and therefore they would like to be able to produce estimates for other countries of destination.

53. In the United Kingdom, nationality and age/sex come from passenger cards/flight data, while reason for emigration relies on data from the IPS (a border sample survey).

54. In Mexico, the demographic variables, country of birth and destination, as well as the reasons for emigration are recorded through the population census and the National Survey of Demographic Dynamics (ENADID). Mexico is the only country that collects information on reasons for emigration, although still is not possible to integrate in these sources of data information about citizenship and nationality.

55. The time lags in the production of initial estimates from a given reference month ranged from 1 to 5 months (Table 5). They ranged from 5 months to 3 years to produce final estimates.

Country	Initial estimates produced	Final estimates produced
Albania	Not produced	Unknown
Australia	6 months	18 months
Canada	3 months	3 years
Hungary	3 months	7 months
Italy	3 months	9 months
Latvia	1 month	5 months
Mexico	6–8 months	1 year
New Zealand	1.5 months	17.5 months
Poland	Not produced	1 year
United Kingdom	5 months	2 years
United States	Not produced	Unknown

56. Of the 11 country responses, 5 responded that they “combine data sources using statistical modelling”, namely Albania, Canada, Latvia, New Zealand and Poland. The other 6 responded that they “use data sources individually” (Australia, Hungary, Italy, Mexico, United Kingdom and United States).

57. However, as a point of clarification, Australia, Canada, New Zealand and United Kingdom use statistical modelling to produce initial emigration estimates. Of these countries, only Canada uses statistical modelling in the production of final emigration estimates.

58. In Australia estimates of overseas migration are required less than six months after the reference quarter to produce quarterly Estimated Resident Population (ERP) statistics. At that time, complete traveller histories for the 16 months following a reference quarter are not available. To obtain the required estimates in the absence of complete duration information, preliminary overseas migration estimates are initially modelled using past traveller migration propensities. See Measuring Net Overseas Migration in Australia.

59. The number of long-term emigrants in Canada is estimated using three different data sources:

(a) Office of Immigration Statistics (U.S. Department of Homeland Security);

(b) Canada Child Benefit (CCB);

(c) T1 Family File (T1FF).

60. The first source is used to directly estimate emigrants to the United States.

61. The number of emigrants moving to countries others than the United States is based on CCB data. CCB data must be adjusted using four factors because it is not universal and does not provide direct information on the number of adult emigrants.

62. For more information see ‘Data quality, concepts and methodology’ and ‘Emigrants’.

63. Although CCB data is very timely, its level of completeness data is initially significantly lower. As a result, to avoid underestimating emigration, we assume a constant emigration rate, derived from the most recent final emigrant estimates (which are usually 2 years old). This method assumes that emigration patterns remain stable over time, allowing data from the previous year to inform the preliminary estimates for subsequent periods.

64. In New Zealand, the migrant status of people crossing the border is not immediately known, so a machine-learning model is used to probabilistically classify border-crossings and produce provisional migration estimates. See ‘An outcomes-based approach to measuring migration – method overview’.

65. In the United Kingdom, historical tax data is used to estimate how many people living in the UK (and interacting with the tax system) will subsequently depart long-term. Where flight data is used, visa expiry dates give an indication of the proportion of departures that will return to the UK within 12 months versus those who will depart long-term (emigrate).

66. Measurement of emigration relies on a variety of sources, with each of them having their own strengths and limitations. All countries surveyed used multiple data sources to estimate and improve their emigration statistics, which use a combination of both conventional and non-conventional data sources.

67. For alternative data sources, countries are mainly exploring the use of other administrative databases. Administrative data have not been collected for statistical purposes, and there are often coverage and measurement issues. However, estimates using administrative data can often produce more timely estimates. Nowcasting techniques using ‘signal’ data can also meet the ambition for even timelier estimates.

68. The benefits of using administrative data include being able to link data sources together. Linked data helps identify new patterns and trends and can help develop further insight into migration statistics. This is an area where some countries are beginning to explore.

69. Other non-conventional data sources, such as commercial data sources, were not listed as currently being explored to estimate emigration.

70. Various breakdowns of emigration statistics are produced by countries, but all would like to expand the available breakdowns. Demographic information, including age and sex, are produced by most countries. No responding countries currently produce estimates by reason for emigration and over half the countries would like to produce country of destination, highlighting a data gap.

71. Producing timely estimates of emigration using the UN definition can be challenging. Definitionally, 12 months or more need to pass before being able to classify someone as emigrating long-term. Statistical modelling can help countries produce provisional estimates of emigration. However, when travel and migration patterns change, it is difficult to reliably and timely estimate international migration using past trends alone (e.g., as experienced by multiple countries during and after the COVID-19 pandemic).

72. The task force has made the following recommendations:

(a) For a better understanding of migration movements, it is recommended to use multiple data sources to estimate emigration. Integration of data from different sources, such as the population census, administrative data, including border records, and mirror statistics, can enhance the richness of data and reduce coverage or accuracy problems.

(b) Explore the use of alternate or new data sources. New non-conventional data sources can be useful for producing migration statistics when used in combination with conventional sources. Countries should explore the use of non-conventional data sources such as mobile phone, credit card, and social network data and share their findings.

(c) Linked data can provide further insight into emigration patterns. When using administrative data sources, countries should explore how these could be used by linking to other sources and how this could improve the quality and richness of breakdowns produced.

(d) Reach out to further countries to understand if any produce a breakdown by reason for emigration. Investigate alternative or new data sources that may be able to provide this data that could be modelled or linked to current emigration statistics.

(e) Timely migration statistics are important for policy and planning. Countries should explore how modelling can improve the timeliness of estimates while still ensuring accuracy and learning from countries who already model their estimates.