Migration statistics

Written evidence submitted by Westminster City Council (5STATS 10)

The City Council would also like to offer speakers to give oral evidence to the committee.

Westminster City Council is pleased to respond to the request from the Public Administration Select Committee to set out our thoughts in relation to the measurement of migration.

International migration levels to Westminster have been traditionally significant, and published ONS figures, covering the years 2001-2011 show Westminster as having the highest volume of international migration per 1,000 population in England & Wales [1] . Westminster is uniquely affected therefore by the accuracy and volatility in the measurement of migration.

Westminster remains concerned that migration estimates based on survey data and modelling of incomplete administrative data may not reflect complex urban environments well enough and have corollary detrimental effects on funding and service planning. It should be noted that although there are direct impacts of migration to local authority areas such as the pressures to the private rented sector, overcrowded accommodation, public health, community safety, rough sleeping, increased cleansing and refuse collection, the indirect impacts are greater still. Indirect impacts are those caused by the lack of accurate government management data. In the case of population estimates this is extremely urgent and will cause severe detrimental impacts in the future grant settlements and needs based assessment of funding.

The ONS in its attempts to improve migration estimates has brought forward numerous changes to methodology resulting in uncertainty and volatility that has undermined confidence in official estimates at LA level –Westminster’s population estimates have varied considerably due to the numerous methodology changes in measuring migration over the last 10 years. In 2012 for example 5% of Westminster’s estimated population was simply removed by a change of methodology.

This volatility creates impossible service planning conundrums. Helping Government meet priorities around worklessness, community safety, public health and most critically from a financial perspective the care of the vulnerable adults are challenging enough, without the absence of a secure population baseline. Population uncertainty makes a mockery of evidenced based planning. For example - In the 2011 census estimates there to be only 3,100 over 85s in Westminster when some 6,800 claim a pension here.

We hope this inquiry will help in the process of improvement of migration statistics.


a) The ONS Report of the Interdepartmental Task Force on Migration Statistics [2] identified barriers to the accurate measurement of migration and set a clear vision for the future:

The aspiration is to move to a situation where we have timely, accurate estimates of the number of people coming into, going out of and present in the country for different durations and reasons (short and long term residence, seeking refuge, study, etc). Sufficient detail (e.g. geographic breakdowns) would be available to meet key requirements for population and migration statistics. This would be accompanied by the capacity to interpret trends and anticipate how changes and new uses will impact on the requirement for statistics.

b) We recognise that a significant work programme (IMPS) was put in place following concerns raised in 2004 but despite this work which ended in July this year we still have significant concerns about the validity of migration estimates both at national cohort level and local authority level.

c) The use of survey data was cited by the Statistics Commission in 2004 [3] as unlikely to provide the answer to the measurement of migration yet a methodology based mainly on surveys for the national cohort is still being used. Also the use of modelling migration using administrative data to distribute this national cohort has not improved estimates in complex hyper-diverse urban areas such as Westminster.

d) We believe the International Passenger Survey (IPS) remains a questionable methodology of estimating TIM and should be strengthened by the full use of e-borders scheme or replaced altogether.

e) The new ONS methodology using a small sample of DWP administrative data and social surveys for the distribution of migration from national cohort to LA level has not improved the measurement of migration for places like Westminster.

The continued use of GP Register Flag Four data to estimate internal migration within E&W was highlighted as a weakness in previous methodology but remains a core determinant of migration movement.

Although the 2011 census may have worked well for most of the country in areas like Westminster there remains significant doubt about the outputs in Westminster and a small number of other authorities. Basing future migration estimates on the 2011 Census therefore could lock significant flaws into estimates for places such as Westminster for the next 10 years.

f) The level of undocumented migrants, short-term migrants and illegal migrants remain essentially unknown. For other areas of the country this would not be a problem but for Westminster where an estimated 24,000 people remain uncounted it makes up a substantial ‘hidden’ population.

g) Despite the Treasury Select Committee of 2007/8 considering the measurement of short term migration (staying in the England and Wales for between 3-12 months) as urgent there are still no official estimates for STMs which we believe is an unacceptable timeframe for measuring an estimated 195,000 STMs [4] nationally equating to a town the size of York.

h) Migration estimates should not be used without an understanding of their ‘uncertainty’ at a local as well as national level. We recommend that estimates should carry a flag of uncertainty where appropriate and published alongside them with both percentages confidence intervals and population ranges.

i) ONS has significantly improved its contact with users of statistics and we welcome the time the organisation has spent holding seminars and with us individually. However, ONS consultation with users is mainly focussed on explaining the methodology it has already developed rather than meaningful engagement in the creation or challenge of it. We believe that it more flexibility in the current one-size-fits-all approach to measuring migration should be adopted and that ONS needs to be open to challenge that an approach that suits some areas will not work well in others.

j) Furthermore Government Agencies need to work much more collaboratively to support ONS. Measuring migration is fundamental requirement, yet for example the DWP samples that are provided through the L2 database covers is just 1%. Why introduce uncertainty into local estimates though restrictions on data that ONS require?

k) Current ‘consultative’ bodies such as the CLIP and SWG should be strengthened with representatives from LA groupings such as Central London Forward or National Association of Local Councils to make sure the different challenges of counting population in different areas are recognised fully.

l) There remains no official challenge process to migration or population estimates. Despite attempts by LAs to set up a resolution process before the publication of the 2011 census this did not happen. We recommend that a body such as the UKSA should set up an independent review where concern exists between the ONS and users of population statistics.

m) The general public remain under the illusion that migrants are physically counted in and out of the country at border checks. This is simply not the case with national estimates of migrant sub groups being scaled up from very small IPS sample sizes. The public and policy makers should be aware of the uncertainty and limitations of assessing migration using surveys and modelling administrative data.


1. Do the published migration statistics – at the national, regional and local levels – meet the full range of their users’ needs namely:

a. Are they easily discoverable and accessible to all users?

b. Are they easy to use and understand?

c. Do they provide an appropriate level of detail?

d. Are they effectively summarised?

1.1 At a national level we support the quarterly reports (MSQR) on migration introduced in 2011 bringing together approximately 40 different reports on migration and administrative data to give a more rounded picture and context.

1.2 However, care should be taken that the migration statistics are not published without an explanation of uncertainty of the estimates and that the wrong inferences are not drawn from the reports by emphasising certain aspects over the whole picture. Although the MSQR does publish an explanatory guide to the use and reliability of the data published this should be made much more obvious in the summary and press releases issued with the reports.

1.3 Standard error percentages and confidence intervals should be published together and in the headline summary as that is likely to be the most widely read part of the report. The reports should make absolutely clear what should be reported as statistically significant and what should be treated with caution.

1.4 The reports seem to be written with headlines in mind and for a media audience rather than from the users of statistics perspective. Perhaps publishing a standardised table of gross immigration/emigration and net immigration/emigration with confidence intervals alongside comparative administrative data such as NiNo would help objectify commentary.

1.5 Emigration should be treated with care as IPS sample sizes are very low in scaling this aspect of migration.

1.6 The publishing of National Insurance Numbers (NiNo) should be used to give context to the migration figures. However, there is no comparison between administrative data and estimates of net immigration. For example there is no explanation why 601,000 NiNos were allocated to non-UK nationals in the year to March 2012 but inward migration stood at 536,000.

1.7 At a local or regional level we don’t believe the publication of migration estimates is anywhere near adequate for the planning of services or the distribution of funding. As a local authority we would welcome more granular level information published for each council quarterly. This should take the form of a standardised table with LA’s net/gross immigration/ emigration published alongside confidence intervals. A breakdown of ethnicity or nationality should be provided on a quarterly basis alongside the national report to enable each council to understand the flow of migration and for the public to gain a better understanding of what is taking place within its area.

1.8 A Migration Observatory report [5] seems to underpin the view of the former Statistics Commission that survey data such as the Labour Force Survey and the Annual Population Survey carry significant uncertainty but this is especially so in complex areas like Westminster. These surveys have specific weaknesses when conducted in complex and diverse urban areas.

2. How well have producers of migration statistics engaged with users? How responsive have they been to feedback from users of the statistics?

2.1 As a local authority we welcome the improvement in ONS the provision of information concerning changes to migration estimates holding regular seminars, working groups and presentations. We also welcome the time that ONS has spent with the Council at technical meetings concerning migration estimates and the 2011 census.

2.2 However, this much improved contact is used mainly to explain decisions already taken over the creation of methodology rather than to solicit thoughts on the co-creation of methodology through meaningful consultation.

2.3 We understand the need for a nationally consistent set of population data but the ONS seems welded to a one-size-fits-all approach to migration measurement that does not take into consideration significant local differences such as using Flag 4 GP registration data tracking patient registration when in central London a significant amount of the local migrants population will use walk in centres or A&E services and not register with GPs.

3. Do the migration statistics which are published enable members of the public to gain a better understanding of the issues?

a. Are the right migration statistics being collected?

3.1 We don’t believe that the publication of the current reports benefit the public’s understanding of migration. At the moment the public is informed by the media reporting of the national picture taken directly from the summary of the quarterly and annual reports. The headlines generated by this coverage will in some cases not reflect the experience or perceptions of the public in their local communities.

3.2 Without more geographic specific information being published by the ONS on a quarterly basis the public are not being served well by these reports.

3.3 We believe it will be extremely difficult for the ONS to produce quarterly LA estimates of migration at any reliable level as sample sizes from the IPS and Flag 4 data used from GP Registration are so small as to make the estimates unreliable. We can only determine that the measurement of migration from the perspective of a LA user and as reliable information of our residents is failing.

4. Is the degree of uncertainty surrounding estimates of migration properly reported and widely understood?

a. Is the degree of uncertainty surrounding estimates of migration acceptable or should it be reduced? If so how could it be reduced?

4.1 As above we think that the uncertainty surrounding migration measurement is not well reported and care should be taken to explain this in summary reports and press releases accompanying them.

5. Are the migration statistics adequate for measuring the Government’s progress against its net migration target?

5.1 Care should be taken the reporting of the progress towards reducing migration to the government’s target of 10s of thousands within this term of Parliament. As outlined by the Migration Observatory in August 2012 the margins of error on migration statistics makes a definitive statement on progress difficult. We would also suggest that methodological weaknesses in the measurement of migration should be taken into consideration.

6. What could be done to improve the quality of migration statistics?

a. Should data from other sources, such as e-Borders, be incorporated?

6.1 Although we recognise that the improvements made by the ONS in its IMPS programme we do not think the results have overcome the problems of providing accurate and reliable LA level data on international and internal migration.

6.2 Our concerns rest in two distinct areas:

1) the national cohort of migration measured by the International Passenger Survey (IPS)

2) the distribution of international in-migrants within the UK using ONS modelling

National cohort

6.3 International migration estimates, which include only people who move to a country other than that of their usual place of residence for a period of at least a year, are currently based on estimates of Total International Migration (TIM) allocated on the basis of the International Passenger Survey (IPS).

6.4 The IPS is the bedrock of migration measurement and by far the largest component of TIM. The IPS has been the subject of much criticism [6] over the years and we recognise the work undertaken by the ONS to improve the survey. However, we remain concerned that the survey is still inadequate as a means to determine migration levels.

6.5 Primarily our concerns with the IPS are:

· Although the much-used public line is that the IPS is a survey of over 319,000 people this is misleading. This may be the number of people asked to take part but the actual sample of those who say they are migrants coming to the UK for more than one year in 2010 was only 2,990. From this small sample the national migration cohort is modelled.

· This small sample size doesn’t seem to have grown substantially since 2004 when only 2,810 identified LTMs took part.

· Are people telling interviewers that they are not migrating to the UK when in fact they are? From a total of 1,342 Pakistani’s interviewed only 260 identified themselves as LTMs. Of 1,981 Chinese interviewed 164 identified as LTMs. 8,294 Polish interviewed with only 103 identified as migrants.

· ‘Each of the passengers contacted through application of the first-stage sampling rate is screened to assess whether or not he or she is migrating (if the contact is migrating, he/she will then be asked migration-specific questions, but if not migrating no further questions will be asked).’ ONS IPS methodology [7] . Are migrants more likely to give a false answer to this question if they are approached by officialdom in this way?

· Those travelling at night are not sampled in the survey and the overall response rate for the 2010 survey was 81 per cent.

· Despite holding the survey at more ports of entry following the ONS review it still seems that this is essentially a Heathrow centric survey. For example the difference between sample sizes at major airports feeding London in 2010 is marked. Heathrow (2,267); Gatwick (39); Stansted (115); Luton (118). There seems to be an assumption that inward migration patterns will be the same at each airport.

· Also is the way the survey is conducted robust enough? This is a section of the IPS handbook for ONS staff: "There may be times when, owing to a particular flood of passengers, you just cannot keep an accurate count. Do not panic if this happens but keep counting as best you can." http://www.theyworkforyou.com/wrans/?id=2012-10-22b.123460.h

· Modelling of total inward migration from different countries is based on very small numbers of surveyed migrants. The number of Iraqis interviewed in 2010 was 2; Jamaicans 3; Portugal 16; Romanians 36; Lebanese 2; Libya 5; Poland 103; UAE 3.

· As below there is a discrepancy between the IPS and records collected by the DWP L2 data base which records all migrant economic activity i.e. NiNo, self employed and benefits. In 2009/10 the difference is quite wide between the total number of long term migrants entering the country according to the IPS and those that are economically active in that year according to the DWP. One or both of the data sets must be wrong.

Distribution of international migration

6.6 Whilst the Census 2011 estimate of 219,400 represents an increase in population since the 2001 census (181,300) in Westminster, it also represents a large reduction of 21,800 (9%) from the previous Revised 2010 Mid Year Estimate [8] of 241,100 and a 43,500 less than the 267,000 (2008 MYE projections) used in the last government funding formula. If population estimates are to be rolled forward from a questionable base such as the 2011 census in Westminster we may never be able to accurately estimate the level of migration.

6.7 In an attempt to improve migration estimates at a LA level the ONS published revised methodology in November 2011. The new methodology altered how the national cohort was distributed around the country with some regions losing out and others gaining large amounts of migrants. The IPS is still used to determine the national cohort of migration but administrative data such as the L2 data (a sample from the DWP database) which holds NI numbers/ PAYE and HESA student data is then used to distribute migration within the UK.

6.8 The use of administrative data for the modelling of migration to local authority levels should be treated with great caution. There is much to commend the ONS and government departments in their attempts to share data and link records across government databases. However, the modelling methodology using this data should be reviewed in detail as we believe administrative data can have just as significant weaknesses as the social survey data they replace.

6.9 For example, data cut from the Department of Work and Pensions (DWP) is a 1% sample of all records over a 30 year period. The proportion of migrants per local authority who are long term are taken from this sample which is determined by the length of economic activity i.e. if a NiNo is used for more than 12 months. There will be many flaws in data extracted in this way – for example sample sizes of the data by LA could be very low and suffer from weighting the data which will be constrained to national figures.

6.10 Also LAs ethnographic research suggests that some migrant communities clone or share NiNos which illustrates how a statistical analysis of available data alone may not fully appreciate the social behaviours or movement they seek to explain.

6.11 The statistical solution to counting migration that has been pursued by the ONS is limited to the use the available data or social surveys. However, we believe that in complex urban environments data extracted in this way will not provide accurate assessment of migration. Estimates of migration complied by statisticians working alone with data should be treated with some scepticism as they do not understand the complex social interaction between migrants and officialdom.

6.12 Tony Travers’ comment on the 2011 census captures this well: "This [the census] is a good example of the ‘official’ world coming into contact with worlds of which it knows nothing. Behind the doors of households, what goes on is a complicated thing… imagine 15 or 20 people living behind a ‘front door’, constituting neither a household nor a series of households because no one has responsibility for all of the others. The front door, should be understood as a gateway from an ordered world to a disordered world, and we are therefore the worst people to be trying to understand it." [9]

January 2013

[1] At the time of this response ONS had not yet published revised migration flows from 2001 to 2006 post 2011 census results.

[2] 15 th December 2006

[3] Our 2003 report The 2001 Census in Westminster contained a recommendation that "the quality of migration data should be addressed with urgency" and noted that the solution "might involve major changes in administrative practice" and that "statistical surveys are unlikely to be adequate". This position was repeated in subsequent reports. http://www.publications.parliament.uk/pa/ld200708/ldselect/ldeconaf/82/82we34.htm

[4] http://www.google.co.uk/#hl=en&tbo=d&sclient=psy-ab&q=195%2C074+short+term+migrants&oq=195%2C074+short+term+migrants&gs_l=hp.3...847.6301.0.6514.,or.r_gc.r_pw.r_cp.r_qf.&fp=417da4c150e265d0&bpcl=39468505&biw=1394&bih=642

[5] http://www.migrationobservatory.ox.ac.uk/top-ten/5-local-area-statistics

[6] In its report, the Treasury Select Committee (2008) noted that:

[6] Society is becoming more mobile and the information held electronically about events, persons and services by government agencies and other bodies has substantially increased.

[6] The International Passenger Survey was designed to provide data primarily for tourism and business travel purposes. It is now called upon to play a central role in estimating international migration. It is clear from the evidence.... that the Survey is not fit for this new purpose

[7] http://www.ons.gov.uk/ons/guide-method/method-quality/specific/travel-and-transport-methodology/international-passenger-survey/index.html

[8] The MYE 2010 figures were revised by ONS and reduced the Westminster population from 253k to 241k through a change to migration methodology

[9] Tony Travers, Visiting Professor LSE London Department of Government, Census Coverage Survey and Imputation: Deliberative Event. 07/10/10

Prepared 4th February 2013