Not what it looks like

The Office for National Statistics put out a big new article yesterday (12 April) titled “International immigration and the labour market, UK: 2016”. As that’s one of my many interests I eagerly clicked on the link only to find that I’d been baited as soon as I saw the subtitle ….

While I’m rather bored of saying it, anyone interested in the actual outcomes in the labour market of international immigration needs to look at country of birth rather than nationality. Looking at things only by nationality means that people from abroad disappear from sight in the data when they gain UK nationality and render many questions about the impact of immigration on the labour market incapable of any meaningful answer. Take a simple question like “Do many people from Pakistan drive taxis in the UK?” or its converse “Are many taxi-drivers in the UK from Pakistan?”. If all the people from Pakistan who have gained UK nationality are taken out of the equation then the answer is a clear no. Of Pakistan-born workers in the UK Labour Force Survey in the final quarter of 2016, fully 65% of employees and 75% of the self-employed were by then UK nationals. So looking at things by nationality means overlooking two-thirds of Pakistan-born employees and three-quarters of the Pakistan-born self-employed.

So why would the ONS do this? Well the Article says

Hmm … the “therefore” suggests that there were reasons given for this recommendation. Let’s look at what they were. Here is the entirety of the relevant section of the evidence given to the Sub-Committee.

In any analysis of usual resident populations, the most common attribute used for identifying migrants is country of birth. This is a measure that does not change throughout life and indicates that a person at some point did arrive in the UK having lived abroad. Country of birth is particularly useful when analysing changes over time, because while a person’s nationality can change, country of birth remains constant throughout life.
It is important to note that a person may be born outside the UK, but have British nationality from birth. Many children of British Armed Forces based overseas would fall into this category. This can noticeably boost numbers born overseas, for example this is particularly evident for those born in Germany.
In 2015, 40% of the 8.6 million UK residents born outside the UK were British nationals. The period of time that a person has been a permanent resident in the UK is not indicated by country of birth (e.g. most Irish born migrants arrived in the UK more than 30 years ago).
In many cases therefore, nationality is the preferable measure to use when seeking to understand the interactions of migrants with, for example the labour market, the benefit system, housing, education and health. With such a large proportion of those who are born abroad having now acquired British nationality it is not recommended that country of birth be used as a headline measure in analysis of such data.

OK. The only reasons appear to be that using country of birth would — correctly though presumably in the eyes of ONS misleadingly — identify British ‘army brats’ and Irish oldsters as migrants. That might well be, but how significant is this to any analysis of the labour market? If you take all the people working in the UK who were born in Germany but who have UK nationality (which will include but go well beyond just army brats) and then add in all workers born in Ireland regardless of their present nationality you come up with about 300,000 people. That’s less than 1% of the workforce and barely 5% of the the number of workers born abroad. On the other hand, there are over 2,000,000 workers born abroad who now have UK nationality. That’s nearly 7% of the workforce and nearly 40% of the number of workers born abroad.

So it looks as though to avoid the possibly misleading classification of up to 5% of workers born abroad as ‘immigrants’ it has been decided instead to classify 40% of workers born abroad as UK workers. To my mind that is simply bizarre. It’s great that the ONS have churned out all the tables and charts showing all sorts of compositional breakdowns of the workforce by skill level, education, type of work etc etc but this methodological decision strips a lot of the meaning from them because of overall distortion and because of inter-group distortion.

For overall distortion let’s look at the chart illustrating the distribution by age within the working-age population of UK and non-UK people. The ONS narrative accompanying it points out that “more than half (estimated 53%)” of the non-UK nationals are aged between 25 and 39. Certainly those bars stretch bigly far beyond much older and younger age groups!

Unfortunately I have to digress from the main theme at this point. Anything strike you as odd in the picture? If 16 to 19 year olds really do form such a tiny proportion of the 16 to 64 population regardless of nationality then we really are in line for a demographic squeeze and an unprecedentedly tight labour market! ONS say that they have used the Annual Population Survey for January 2016 to December 2016 and having poked around a bit in that it looks as though they have got their numbers wrong by using the numbers only for 18 and 19 year olds. 16 to 19 year olds actually comprise something over 7% of the 16 to 64 population [But see Update below].

Anyway, back to the theme. Below I’ve super-imposed bars on the ONS chart showing the distribution by country of birth rather than nationality (and taking account of everyone aged 16 to 64). Using country of birth, the proportion of non-UK people aged from 25 to 39 shrinks from over half to well under half, and the proportion aged from 60 to 64 more than doubles as older people who have acquired UK nationality reappear from out of the data.

Casting our minds back to the ONS evidence to the House of Lords, this does make me wonder about how nationality can be “the preferable measure to use when seeking to understand the interactions of migrants with, for example … the benefit system, housing … and health” when, to take state pensions and health costs associated with the elderly as examples, migrants increasingly fail to appear in the data as migrants on a nationality basis as they age. That is a key aspect of the overall distortion, to add to the difficulty with observing who is doing what in the labour market as per the taxi-drivers above.

This overall distortion is bad enough in static observations i.e. what is happening at any point in time. It is far worse for observations over time. As I’ve said over and over again, things like the changing number of people in work (or in any particular kind of work) by nationality can be completely meaningless. For example if our Pakistan-born taxi-driver gets UK nationality there is an increase in the number of UK nationals in work and a decrease in the number of non-UK nationals in work. Or as a headline might put it “More Brits working!!”. In their evidence, ONS did concede “Country of birth is particularly useful when analysing changes over time” but their money is in a different place from their mouth on this issue as in their routine quarterly report on employment by nationality and country of birth, they prioritise and make a more extensive analysis by nationality than by country of birth.

The inter-group distortion is perhaps worse. The reason for this is that there are very big differences in the proportion of people in the different country groups used by the ONS who have UK nationality. Of people in the EU2 and EU8 groups, very few have UK nationality as until recently there wouldn’t have seemed to be much point in going through the bother of applying for it when their status as EU nationals gave them the right to stay in the UK for as long as they liked. At the other end of the spectrum, a clear majority of people from non-EU countries have acquired UK nationality. So when comparing for example the labour market outcomes of immigration from Romania and Bulgaria with those of immigration from Pakistan and Bangladesh on a nationality basis, these are not properly observed. The scale of this can be seen in the picture below. The number of workers in the UK born in Romania and Bulgaria is a little less than the number born in Pakistan and Bangladesh. But the ONS analysis counts as migrants only the small proportion of the latter who are still Pakistani and Bangladeshi nationals and so will say that far more migrants from Romania and Bulgaria are working in the UK.

Looking at the country grouping used by the ONS in their new analysis the differences in acquisition of UK nationality are stark. The Home Office publishes data on grants of UK citizenship and grants to EU nationals are minuscule in comparison with those to non-EU nationals.

Grants to EU nationals are barely 10,000 a year from a population of near 3 million. Even allowing for their generally more recent arrival in this period (1990–2015) it is clear that there has simply not been anything like the same take up of citizenship. This means that observations of things like pay levels or the skill levels of their jobs are simply not comparing like with like. While post-referendum (let alone Brexit itself) this might change, that rather proves the point, as future observations will be further confounded by the changing rate at which people would disappear from the data as EU nationals and appear instead as UK nationals.

On an initial look, for these reasons it seems to me that the ONS have missed a trick. It’s not that what could have been a really interesting piece of work is un-illuminating, but worse than that, it throws light on quite the wrong things and throws up all sorts to barriers to a proper understanding of the possible impacts of immigration in the labour market.


On the ONS picture of age distribution ….

This is introduced in the ONS publication by the passage

“In 2016, within the UK, the household population aged 16 to 64 was estimated to be 41.0 million2 (+/-0.3 million).
Figure 1 shows that immigrants tend to be younger. More than half (estimated 53%) of non-UK nationals aged 16 to 64 years old were between 25 and 39 years old, reflecting the concentration of working migrants in younger age groups.”

The natural reading of this is that within that household population, immigrants tend to be younger, and the figure illustrates the extent to which this is so. Obviously, as originally blogged, something then appears wrong for the 16–19 age group. Looking at the APS it seemed that this was because only 18–19 year olds had been included (as they account for 3.7% of the 16–64 UK-national population, matching the number on the ONS figure)

However, looking at the accompanying ONS dataset it appears that the reason is rather different, and that the figure actually shows the age distribution of people in employment. Hence the very small percentage of 16–19 year-olds. But in that case the introductory passage and figure heading don’t properly describe what it is.

More importantly, even if properly described, the age distribution of people in work does not seem at all meaningful in the context of the publication because it is affected by the different rates of people in employment i.e. as rates of inactivity (including retirement) or unemployment vary significantly both by age and by migrant status. It is thus uninformative both about the migrant component of the workforce and about the relative age distribution of the workforce for both migrant and UK groups (whether on the basis of nationality or country of birth). My superimposed bars continue to show the more informative matter of the age distribution of migrants and non-migrants in their respective parts of the workforce!

Update on update:

I’ve been in touch with ONS directly about their age distribution figure. To give credit where due, they did look into this quickly and confirmed that it is actually a picture of the age distribution of those in employment and that they would “amend the title and the accompanying text so that the Figure is accurately described”. On the point of whether it is a meaningful illustration (which I also raised) they thanked me for my suggestion.