ONS has conducted a review of the quality of Long-Term International Migration (LTIM) estimates over the decade from 2001 to 2011. These estimates are predominantly produced from the International Passenger Survey (IPS).
The review follows research conducted in light of the results of the 2011 Census for England and Wales, which found that the Census-based mid-year population estimate was 464,000 higher than the mid-year population estimates rolled-forward from the 2001 Census base. Several possible causes for the difference were cited but it was considered that the ‘largest single cause is most likely to be underestimation of long-term immigration from central and eastern Europe in the middle part of the decade’ ( ONS, 2012 (171.1 Kb Pdf) ). The review extends this work to compare LTIM estimates to further data from the 2011 Census, as well as a range of other data sources across the decade from 2001 to 2011.
Aims of the review
This review of the quality of LTIM estimates aimed to address the following questions:
What evidence is there to suggest that long-term immigration was underestimated during the middle part of the decade between 2001 and 2011?
This includes an examination of alternative data sources and looks for evidence that immigration was inaccurately estimated.
To what extent was any underestimation of immigration due to inadequate coverage of the IPS during 2004 to 2008?
This examines the coverage of the IPS from 2001 to 2011 and identifies evidence that shows the impact of improvements to the IPS that were fully implemented from 2009.
Were the adjustments made to calculate LTIM, for example to estimate the number of visitors who switched to become long-term migrants, adequate during this time?
This examines the methodology of calculating LTIM and assesses if the underlying assumptions remain fit for purpose in effectively measuring migration.
What information about the quality of estimates of emigration can be deduced from comparisons with international migration data?
The quality of emigration estimates has been included for completeness, as it is used in the calculation of net migration.
Key findings of the review
There is evidence that shows the IPS missed a substantial amount of immigration of EU8 citizens that occurred between 2004 and 2008, prior to IPS improvements from 2009. This is evident from comparisons of IPS data with a number of other data sources related to immigration. The EU8 migrants were missed due to IPS interviewing being concentrated at the time at principal airports, such as London Heathrow, London Gatwick and Manchester. During this time, many migrants from the EU8 countries were travelling on the increasing number of routes connecting their countries with the UK regional airports (the number of routes connecting UK airports with airports in EU8 countries increased from 30 in 2001 to a peak of 190 in 2007). Many of these routes were not covered, or not fully covered, by the IPS for migration purposes prior to 2009.
The IPS has underestimated the migration of children. There is evidence from comparisons with other data sources that estimates from the IPS of children under 15 years old are too low. Investigations have shown that this is not due to the weighting of the IPS, and clear instructions are given to interviewers that when children are sampled responses should be provided on behalf of the child, and not on behalf of any accompanying adult. Field procedures have already been improved, but the impact of this action may not be fully resolved until e-Borders data are available which will allow the direct comparison of passenger numbers by age with IPS data.
The IPS improvements have both reduced the relative error around the IPS estimates, as well as improving the balance of the sample improvements to the IPS. Starting in 2008, more regional airports were included in the IPS and there were an increased number of migration interviews at key regional airports such as Luton and Stansted. These improvements reduced the skew towards particular migrant groups (typically non-EU) who predominately travel through the main airports (mainly London Heathrow). The outcome of these improvements is that the IPS sample is much more balanced towards all groups of migrants.
Comparisons between IPS data and other data sources showed that in the years since the IPS improvements, the trends in the IPS series more closely track those seen in other data sources.
Data from the Civil Aviation Authority has shown that by the time the improvements to the IPS were implemented, the expansion of EU8 passenger numbers and routes had already begun to level off. This suggests that the IPS improvements were too late to capture the main wave of increased migration following EU accession in 2004, and explains why more long-term migrants from the EU8 were identified on the 2011 Census than would have been expected based on LTIM estimates.
There is no evidence to suggest that the current methodology used in LTIM calculations needs adjusting. An adjustment is made as part of the LTIM calculations to account for a proportion of people who stay longer in the UK than originally intended. These people originally arrived as visitors to the UK and switched to becoming migrants. The methodology groups people according to their citizenship. It has been suggested that a greater proportion of migrants from EU8 countries may have switched from being visitors to long-term migrants than was accounted for by the visitor switcher methodology ( ONS, 2012 (171.1 Kb Pdf) ). However, analysis of visitor switcher data suggested that EU8 migrants were no more likely than other EU migrants to switch.
Comparisons between LTIM estimates for the year ending March 2011 and implied migration flows from the 2011 Census demonstrate the improved quality of LTIM estimates following the improvements made to the IPS. There is, however, evidence that LTIM estimates of immigration of EU-born migrants are still lower than those implied by Census, although notably LTIM estimates are very close to Census estimates for EU8-born immigrants. By contrast, LTIM estimates were higher than implied Census estimates for immigration of New Commonwealth-born citizens.
Outcomes from the review
Revision to net migration estimates
It seems that the underestimation of immigration between 2004 and 2008 occurred principally due to an inadequate sampling design and coverage of the IPS prior to 2009. As a result ONS has published within this review a revised set of net migration estimates for this period for the United Kingdom. These estimates, which are shown in Table 1 below, give an indication of what ONS now considers the magnitude of net migration to have been between 2001 and 2011.
Table 1: Revised net long-term international migration series for United Kingdom, calendar year, 2001-2011
|Calendar year||Revised Net migration estimates||Original Net migration estimates||Difference|
|2001||+ 179||+ 171||+ 8|
|2002||+ 172||+ 153||+ 19|
|2003||+ 185||+ 148||+ 37|
|2004||+ 268||+ 245||+ 23|
|2005||+ 267||+ 206||+ 61|
|2006||+ 265||+ 198||+ 67|
|2007||+ 273||+ 233||+ 40|
|2008||+ 229||+ 163||+ 66|
|2009||+ 229||+ 198||+ 31|
|2010||+ 256||+ 252||+ 4|
|2011||+ 205||+ 215||- 10|
Table source: Office for National Statistics
The adjustments applied increase the estimate of net migration across the decade from 2001 to 2011, but most particularly in 2005 to 2008, when the evidence suggests that the majority of migrants who were missed by the IPS immigrated to the United Kingdom.
Users who wish to see a more detailed breakdown of inflows and outflows of long-term international migrants between 2001 and 2011 by variables such as reason for migration, age and sex, citizenship and country of birth should continue to use the existing LTIM and IPS 1, 2 and 3 series tables, but should bear in mind the caveat that the headline net migration estimates have now been revised as outlined above. Please see Section 5 of the full report for additional guidance.
Continuing improvements to the IPS
ONS will address on-going issues with the quality of the LTIM for particular sub-groups of the population, for example children under the age of 15. These issues arise because migration estimates are still based on a relatively small annual sample of 4,000-5,000 migrants identified by the IPS. The quality of these estimates would be improved by further increasing the sample size of the IPS, but this would have substantial cost implications. It has been estimated that to halve the size of the confidence intervals around LTIM estimates would require a four-fold increase in the IPS sample size, and an accompanying four-fold increase in the cost of the survey from the current £5million per annum.
ONS are already exploring whether e-Borders data could be used to improve international migration and population estimates. The results of exploratory analysis on an early set of e-Borders data was published at the end of the Migration Statistics Improvement Programme (MSIP) in March 2012. Current research is investigating whether the data could be used to identify the travel history of migrants, and how the data could be used to improve the quality of IPS estimates, for example by providing an age-sex breakdown of passenger flows which could feed into the weighting of the IPS. This analysis should resolve issues identified within this report of apparent underestimation of migrants under the age of 15.
The design of the IPS needs to continue to be responsive to changing migration trends. Although it is difficult to anticipate routes that migrants will favour in advance, it is possible to monitor new routes and passenger numbers and respond accordingly. For example, the number of boats sampled on the Dover-Dunkirk route has been increased from 2014 in order to improve the robustness of the sample on this route, and to potentially boost the sample of EU2 migrants, following the lifting of transitional controls in January 2014.
Read the full version of the Quality of LTIM estimates report (1.04 Mb Pdf) .