What Migration Statistics are needed for Public and Policy Debates about Migration?

Migration Observatory Response to the ONS Migration Statistics Engagement Report

By Madeleine Sumption

Context

The Migration Observatory at the University of Oxford produces impartial analysis of migration in the UK, making use of the full range of ONS and Home Office statistics. Our research aims to help policymakers, civil society and the media understand the implications of data and research about migration—including questions such as how migration is changing, how policies affect migration and its impacts, and what are the social and economic effects of migration in the UK. As part of this work, we participate in frequent discussions with policymakers and practitioners about what information they need. Our response to this ONS engagement report thus draws both on our own experience using migration statistics and on our understanding of what practitioners are looking for.

The Migration Observatory welcomes the move to put administrative data at the core of migration statistics. In addition to improving the accuracy of the headline statistics, we believe that the transformation programme brings excellent opportunities to make available data more relevant to policymakers and practitioners.

The suggestions in this document can be summarized as follows:

• ONS migration statistics have traditionally focused on overall levels of migration and net migration, but tell us less about migrants’ characteristics and legal status, which is important for policy purposes. Linking ONS migration-related statistics to Home Office records on migrants’ immigration status and to information on migrants’ outcomes (e.g. employment, industry, occupation, individual earnings and family income) would have a transformative effect on the ability to understand the impacts of past and future policy choices.
• For the purpose of evaluating the impacts of policy, the most useful way of disaggregating migration flows data would be based on the legal route through which people come – i.e. whether they have been given visas as students, family, workers and refugees/asylum seekers.
• For population estimates or stocks, it would be valuable to have information on both legal status at entry and current status, including whether migrants have been granted indefinite leave to remain (ILR) or equivalent.
• Analysing migrant populations/stocks by their current nationality can be misleading because 1) nationality is less important than ILR in determining a person’s entitlements and how they interact with institutions like the labour market; and 2) nationality changes over time for some groups but not others, making it unclear exactly what is being measured.
• Short-term migrants are set to become more important in the UK’s labour migration system after Brexit, making it useful to understand the movements and characteristics of those moving for less than a year. It would be helpful if these figures were published together with long-term migration estimates, using a methodology that allowed them to be added or subtracted as needed.
• There will probably be a continued need for survey data to study migration, to understand subjective variables that cannot be captured in administrative data. It would be worth systematically reviewing the migration related variables in the LFS/APS and other surveys to make improvements and/or identify some variables that could be included periodically (e.g. parents’ country of birth).
• There is a risk that the welcome move to make better use of administrative data will mean that non-government researchers have more restricted access to core migration data sources. It would be useful to consider options for anonymised microdata samples to be made available under End User License agreements.

Definitions: who counts as a migrant?

The main ONS statistical products classify migration flows by self-reported duration of stay, using the UN definitions of short-term and long-term migrants (less than or at least 12 months). These data can be broken down by reason for migration, which among other things enables us to (crudely) distinguish overseas visitors from short-term workers or students. For population estimates, figures are broken down primarily by nationality or country of birth. Self-reported reason for migration is available but not included in the published data tables.

While ONS produces both short-term and long-term international migration estimates, the long-term ones receive more attention because they are included in every MSQR, are described in the main bulletin, and can be produced more quickly than STIM estimates. (Many casual users of migration statistics are in fact entirely unaware that STIM figures exist.)

Classifying migrants by legal status

Whether or not someone’s expected length of stay is at least a year is arguably not the key distinction of policy interest, as it is not the main variable affecting how they interact with institutions, people and communities in the UK.[1] To see why, consider some of the key questions that policy-focused users of migration statistics are interested in:

• The role of migrants in the labour market, including what jobs they do and how heavily employers depend on them;
• Migrants’ expected use of public services, which is of interest to those planning service provision and understanding the impacts of migration policies;
• Social impacts of migration, including impacts on communities (e.g. due to ‘churn’ and pace of change);
• Impacts of migration policies on the number of migrants admitted and the composition (e.g. by skill levels or country of origin).

For many of these questions, people staying less than one year are quite important. Non-UK citizens who are legally resident in the UK have full access to public services such as the NHS and schools immediately, and also play an important role in the labour market, for example. The policy significance of short-term migrants is expected to increase with the implementation of government proposals that labour demand in low-wage jobs should be met primarily through a short-term work route lasting less than one year (HM Government, 2018).

While it does make sense to be able to consider long-term and short-term migrants separately, arguably the most important factor to consider when analysing the role of migration and the impacts of migration policy is therefore the type of legal status a person entered with and the status they currently have. Legal status matters because it affects people’s rights and entitlements, such as their rights to work, whether they have full access to benefits, and whether they are likely to be on a pathway towards indefinite leave to remain (ILR).

Broadly speaking, there are three main categories of interest from a policy perspective:

• Visitors and others who are not considered migrants and thus need to be identified and removed from migration-related statistics (these people generally have very limited entitlements or access to public services while they are visiting);
• Temporary residents whose stay here is still subject to conditions and whose rights are constrained in different ways depending on their visa type; for this group it is also important to understand whether they have come as students, workers, family members or refugees because this will affect their entitlements; and
• People with a permanent status (i.e. ILR or citizenship), who have very extensive rights. (See below for a discussion of ILR vs. citizenship.)

While these legal distinctions can only currently be made for non-EU citizens, government proposals for the post-Brexit immigration system are expected to bring EU citizens into the same framework, enabling a more consistent approach across all countries of origin.

Flows data: immigration and emigration

The International Passenger Survey (IPS) already broadly captures most of the groups outlined above, but with the significant limitation that key distinctions are self-reported and may not match the person’s actual legal status. The Transformation Programme provides an opportunity to move away from self-reported measures to ones based on legal status. This would make it easier to evaluate the impacts of policy choices, since analysts would have more confidence that the category being analysed matches the relevant policy route. It would also make it possible to disaggregate more finely within routes (e.g. identify people on different types of work visas).

In particular, flows data based on the type of permission a person enters with (student, family, work, other) would more closely match policy uses of the data than self-reported motivation. Individuals will usually have several motivations for moving, but should only hold one immigration status at once. For emigration statistics, individuals would need to be classified by their visa type at entry (i.e. the equivalent of the IPS data on original main reason for migration), in order to understand the propensity of particular groups of migrants to stay vs. emigrate after a period of time. In other words, developing administrative statistics using an approach conceptually similar to that used in the ONS (2018) research report on international migration data sources (notably the Home Office data displayed in Figure 1 of that report) would be valuable.

While short-term migrants remain important from a policy perspective, it may be helpful for statistical purposes to continue to sort long-term and short-term migrants by expected duration of stay, as long as statistics on both are published—ideally in the same publication and using a methodology that allows them to be added and subtracted more easily.

It may be possible to drop intentions-based data entirely. Self-reported duration of stay from the IPS data has been useful for creating timely statistics (because we do not need to wait to see how long a person actually stays), but it is also likely to be unreliable given that migration intentions can easily change.[2] It would therefore be useful to test whether more reliable information could be produced by replacing self-reported intention with the duration of the visa on which the person enters. (This too would be provisional given that we know people often do not stay for the full length of their visa, but has the benefit of representing a legal ‘fact’ rather than a subjective judgment).

Stocks data

Like flows data, population estimates would ideally also classify people depending on their type of legal status. The key additional variable that is needed in stocks estimates is whether someone has a permanent status (either ILR or citizenship – but not citizenship alone, as discussed below), since acquisition of permanent status is the key point at which entitlements change. This would enable analysts to understand how many residents at any one time held time-limited statuses (for different purposes), and start to estimate their impacts, for example in the UK labour market.

When drawing on administrative data sources, it will also presumably become necessary to distinguish between part-year and full-year residents. Current population estimates largely do not face this problem because surveys can ask people whether they are usually resident at a given point in time. When using administrative data, a challenge will be distinguishing between circular or temporary migrants moving on a regular basis, and people whose ordinary residence is in the UK but who have significant temporary absences (i.e. they spend significant time abroad but are not circular migrants because they do not change their ordinary residence). There will not be a single optimal rule for distinguishing between such cases, although relevant considerations for classifying someone as a full-time resident could include 1) whether the person is absent from the UK for no more than 180 days per year, since this is the residence requirement for settlement; or 2) whether they are on a long-term visa; or 3) whether they meet both of these criteria.[3]

Nationality vs permanent status

When considering inflows, a person’s nationality is of some interest as it enables us to sort people into groups by country of origin. Historically, the breakdown between EU and non-EU nationality has also been important because it as associated with quite different legal rights at the moment a person first enters the country.

When considering the composition of the population living in the UK, however, nationality becomes less useful and can even be misleading.

Nationality is not the main factor that affects a person’s rights and how they interact with institutions in the UK such as the labour market or public services. As noted above, the type of leave a person has to be in the UK (different categories of temporary leave vs. ILR) is crucially important in determining their entitlements. The additional rights a person receives when moving from ILR to citizenship are relatively small. Settled residents already have full access to the public services, benefits and the labour market (with minor exceptions such as government jobs that can only be held by some nationals). The main additional rights that come with citizenship do not hugely influence analysis of the impacts or role of migration, and include voting in national elections (except for Commonwealth citizens, who can vote immediately), and greater protections against deportation for people who commit serious criminal offences.

Research on the impacts of migration and of policy choices generally uses country of birth instead of nationality to identify migrants. Country of birth provides a proxy for nationality at first entry, although the latter is more useful when available (e.g. in the United States, survey data identify people born abroad to American parents so that they can be removed from the ‘migrant’ population). There are two reasons for preferring country of birth in analysis of stocks. First, the impacts of migration do not cease to be of interest after the 6+ year period after which non-UK nationals generally start to be eligible to naturalise. In the case of fiscal impacts, for example, existing evidence suggests that the impacts evolve considerably over the entire lifecycle (Oxford Economics, 2018). Second, some non-UK nationals are much more likely to naturalise than others, which means that the composition of the naturalised and non-naturalised populations are quite different. EU nationals, in particular, have traditionally had low naturalisation rates and many stay in the UK for decades without naturalising.

Since nationality does not indicate clearly what rights people have and is not expected to substantially affect how people interact with the labour market or other institutions, it is difficult to justify removing naturalised citizens from the analysis other than for symbolic reasons.

As a result, it would be useful to ensure, to the extent possible, that a person’s original nationality when entering the UK (or their country of birth, as a proxy) is available in any new statistics that will be used to analyse the composition of the resident population. If nationality at the time of registration for a national insurance number (NINo) is to be used more widely as a result of reliance on administrative data, it would be useful to conduct some analysis to understand whether a meaningful share of migrants naturalise before applying for a NINo. If a measure of a person’s current status can feasibly be included in the new administrative data sources (e.g. by linking to Home Office data), this should identify people with permanent status/ILR or citizenship, and not just those with UK citizenship.

What do we need to know about migrants’ outcomes and impacts?

Labour market characteristics and impacts

A key barrier to understanding and evaluating migration policies is that very little is known about the labour market outcomes and characteristics of people who enter the UK under different visa categories. This information is important because it crucially affects how one evaluates the economic and social impacts of migration policy choices. The Home Office can produce data (occasionally accessed by others, such as the Migration Advisory Committee) from sponsored visa applications – e.g. the proposed occupation and salary of a skilled worker coming under Tier 2 of the Points Based System. However, almost nothing is known about the economic activities of those on visas that do not require sponsorship, including many categories within Tier 5.

This will become much more important after Brexit, since the government has proposed that there will be greater reliance on self-sponsored short-term work visas. The White Paper says that the proposed short-term work route will be evaluated around 2025. This will require statistics addressing questions such as what people in this route do in the UK, what they earn, how long they stay, and the extent to which they move on to other routes in the immigration system (e.g. other family or other work visas).

Variables of particular interest include:

• Industry of the employer;
• Occupation of the individual (this is important as it allows us to understand the skill level of the job)
• Earnings from employment and self-employment;
• Employment, unemployment and use of benefits;
• Household or family income;
• Demographic variables such as age and gender;
• Place of residence (region/local authority) within the UK.

Home Office data provide useful information on migrants’ journeys through the visa system, from initial entry to settlement (Home Office, 2018). Linking this data to sources of information on people’s economic activity and characteristics would have an enormous impact on our understanding of the impacts of migration policy choices. This would enable us to answer questions such as:

• How do migrants admitted on different routes fare in the labour market several years later? For example, are Tier 1 migrants (the ‘most highly skilled’) actually in skilled employment or business activity? Do Tier 2 migrants perform better in the labour market if they have previously been international students (and thus to what extent should policy favour former international students over other applicants)?
• Does the visa system have similar or different impacts across the nations and regions of the UK, or in different types of local areas (e.g. urban vs. rural)? How does the distribution of family migrants, asylum seekers, and workers change over time after they arrive in the UK?
• What are the long-term outcomes of people admitted on family visas or as refugees? Do they attain reasonable standards of living, including after achieving settlement?

Integration

As the government implements its Integrated Communities Action Plan (MHCLG, 2019), it plans to develop quantitative metrics to monitor integration. Currently, the most important data sources for understanding migrants’ integration and wellbeing in the UK come from surveys, primarily the LFS/APS but also others such as the Family Resources Survey and Understanding Society. Our expectation is that this will continue to be the case even despite the development of new administrative data sources, as surveys will be needed to produce variable that are generally self-reported, such as language proficiency.

The LFS/APS is arguably the most important source because the sample size of foreign-born people is relatively large. It provides a good overview of labour market outcomes, but is less well equipped to address other integration outcomes. For example, The LFS language variable is only asked every 3 years and only refers to language difficulties in work or education—perhaps surprisingly given that language is arguably the most widely recognised single indicator of integration. Asking about language difficulties is not particularly useful as it does not indicate the level of proficiency, is not comparable to other data sources (e.g. the Census) and is not well suited to analysing some of the groups of particular interest to the government in the green paper strategy, e.g. non-working women.

Because of the likely continued importance of the LFS/APS for research on migration, it would be worth reviewing the migration-related variables to assess whether improvements can be made. For example:

• Current official migration data sources do not accurately capture dual citizenships. The LFS/APS only record one nationality, while the Census asks about passports rather than citizenship. Collecting data on passports is helpful for avoiding respondent confusion about what they are being asked, but also raises problems: many dual citizens will only have one non-expired passport and some people have no passport at all, making it difficult to identify their nationality. Asking about passports previously held in addition to those currently held (and potentially allowing those with no passport to specify a national identity document), could help to remedy this problem.
• The LFS/APS could pilot a question on whether people have been granted a permanent immigration status (i.e. indefinite leave to remain) by the government. This question is included in equivalent surveys in other high-migration countries such as Australia and Canada. As noted earlier, it would be more useful for policy purposes to understand whether someone has permanent status than to know their current nationality. This variable would allow more nuanced analysis of the impacts of temporary migration and integration outcomes of migrants with different profiles and rights/entitlements.
• Given that pressure on survey length is an important concern, it would be worth exploring possibilities for additional questions to be asked on an occasional basis, either through regularly scheduled questions (e.g. the current language variable is asked one quarter every 3 years) or through ad-hoc modules. This would allow, for example, collection of information on variables such as:
  • Parents’ country of birth, which allows analysis of the long-term and intergenerational impacts of migration and integration; currently the children of migrants are ‘lost’ from survey data once they leave their parents’ home;
  • Visa status and history, rather than reason for migration (that is, what was the purpose of the visa the person was issued); this variable could be used in conjunction with a variable on permanent status, discussed above, to create proxies for irregular migration status.
  • Integration-related variables across a range of domains, such as demand for language instruction or perceived discrimination.

In other cases (e.g. the Health Survey for England, the National Pupil Database and the Community Life Survey), rich data are collected but without migration-related demographic variables (i.e. country of birth), which means that analysis can only be conducted by ethnicity.

How should migration statistics be communicated and made available outside of government?

The quality of evidence on migration depends heavily on research conducted outside of government, even if government policymakers are the ultimate beneficiaries. It is therefore important to think about the implications of the transformation programme for the accessibility of migration statistics to practitioners and researchers across the UK.

What regular tables are published for use by practitioners?

While the aggregate migration figures (e.g. total net migration) receive most attention in political debate and the media, practitioners are typically looking for more detailed statistics for purposes such as understanding target groups of a policy intervention in a given local area or identifying the outcomes and impacts of particular groups of migrants. The practical benefits of improved migration statistics drawing on administrative data will be greatly improved if the outputs can be published in a form that allows practitioners to drill down in to the data and extract information about specific groups as they need—particularly at the local authority level or even below if feasible.

For this purpose, it would be worth examining the options for a flexible dissemination system along the lines proposed for the 2021 Census, to enable practitioners to create customised tabulations. Key variables of interest include residents’ countries of origin, duration of stay in the UK and status (e.g. students, family, work or asylum migrants), as well as age and gender. If a flexible dissemination system is not feasible, it would be useful to provide detailed tables with as much information as is possible (e.g. by local area and the demographic characteristics mentioned).

There is also high demand for data on the internal migration of existing migrants, and ‘churn’ – that is, movement of people both within and between local authorities – that could be used to inform service planning.

Access to data for research

The government holds an enormous amount of information about migration and migrants in the UK, most of which is not published as official statistics. Published tables from the LFS and APS are limited to breakdowns by nationality/country of birth and employment status (in the MSQR and LMS respectively). Non-government analysts such as the Migration Observatory and many others address this gap by regularly publishing their own analysis. In the medium term, external analysis forms the knowledge base for government policy. By way of example, the recent Migration Advisory Committee (2018) report on EEA migration draws heavily on research conducted using ONS microdata by academics across the UK.

One risk of the move towards administrative data is that it could become much harder for researchers to conduct analysis using core migration data sources under development. This is because access to administrative data tends to be heavily restricted and available only in conditions that many researchers cannot use (e.g. access only via secure labs or in person in government facilities). With administrative data, data owners may need to sign off on particular uses of data or agree to outputs before they are released, which makes them subject to political considerations and not just the quality of the research.

It will therefore be important to consider how different levels of data access can be arranged to facilitate research. For example, providing anonymised microdata samples that can be downloaded under an End User License agreement would probably have a greater impact on future research than providing richer data under restrictive conditions.

References

• HM Government. 2018. The UK’s future skills-based immigration system. Available online.
• Home Office. 2018. Statistics on changes in migrants’ visa and leave status: 2016. Available online.
• Lemaitre, Georges, Thomas Liebig, Cécile Thoreau and Pauline Fron. Standardised statistics on immigrant inflows results, sources and methods. Paris: OECD. Available online.
• MHCLG. 2019. Integrated Communities Action Plan. Available online.
• Migration Advisory Committee. 2018. Migration Advisory Committee (MAC) report: EEA migration. Available online.
• OECD. 2018. International Migration Outlook 2018. Paris: OECD. Available online.
• ONS. 2018. Report on international migration data sources: July 2018. Available online.
• Oxford Economics. 2018. The fiscal impact of immigration on the UK. Available online.
• Stats NZ. 2017. Outcomes versus intentions: Measuring migration based on travel histories. Available online.

 

[1] Note that the UN definition is actually not particularly helpful for facilitating international comparisons. While the UN has recommended that countries produce statistics on long-term migrants using a 12-month definition, in practice a large number of countries do not actually use it. Out of 36 countries reporting international migration inflow data to the Organisation for Economic Cooperation and Development (OECD) for its International Migration Outlook (2018) publication, only 11 did so based on a 12-month definition. Most either supplied data on grants of temporary or permanent status, or entries onto the population register for an actual or intended duration of at least 3-6 months (ibid, p315). Instead of using the UN definition to harmonise statistics across countries, OECD prefers a standardized measure of ‘permanent’ or ‘permanent-type’ migration, defined as those receiving long-term or permanent residence permits (see Lemaitre et al, 2008). As noted above, permanent status is the key ‘moment’ at which the rights of migrants change, making it arguably the more important measure to collect.

[2] New Zealand, for example, which has traditionally collected intentions data but now constructs actual travel histories using entry and exit records, found that people consistently moved for longer than expected both when immigrating and when emigrating; they also found that the accuracy of intentions-based measures varied by nationality and visa type (Stats NZ, 2017).

[3] There may be ambiguous cases where a person on a long-term visa does change their ordinary residence several times on the same visa. For example, a person on a Tier 2 (intracompany transfer) visa which allows full-time work for several years, but who in practice only comes for periods of a few months at a time when assigned to a UK-based project.