Here’s a correlation conundrum - does immigration affect domestic unemployment? According to one highly-credible, academic study, it has no impact at all. But another, equally credible study both identified an impact and quantified it. So how can such opposing views emerge from analyses of the same problem?
The issue of a correlation of these two activities is important because it is a highly political (and politicised) subject which many pressure groups focus on. Elections can even be swung according to the perception of whether new workers entering the UK are displacing UK citizens from jobs. So establishing the truth plays a key role in proving whether the perceptions and reality are aligned or not.
The National Institute of Economic and Social Research (NIESR) used national insurance numbers issued to migrant workers in 2010/11 as its core data set. This assumes that immigrants are about to enter the UK workforce (or plan to) and also provides a large data set, rather than relying on samples. Spatial correlation with local unemployment benefit claimant counts was then examined. After modelling and regression, NIESR found an insignificant correlation between the two factors.
By stark contrast, the Migration Advisory Committee concluded that for every 100 non-European immigrants, 23 UK workers are displaced. That is a sizeable impact and one likely to fuel arguments against admitting workers in to the country at a time of recession. This study looked at migration and employment data from 1975 to 2010 and concluded that since 2005, 160,000 British workers had lost their jobs as a result of immigration.
How can these two findings be so completely opposite? As data analysts know only too well, it all depends on the data set you use and what modelling technique is applied. The report which did find a correlation made a virtue of its use of 25 years’ worth of historical data to allow it to identify a true trend. By contrast, NIESR argued that such an approach assumes all immigrant workers are direct substitutes for existing British employees. Their model allowed for flexibility in the labour market, job creation and even internal migration.
If there is no correlation between migrant workers and domestic unemployment, then the fact that benefit claimant numbers have risen points to underlying skills and training problems in the UK workforce. That requires long-term strategies based around education and support for employers running in-job training programmes.
If there is a correlation, the solution could be as simple as stopping non-EU worker migration into the UK in order to protect the domestic workforce. That is a quicker fix and also has the advantage of reducing how much the government pays out in benefits.
It is not hard, then, to see which insight is more likely to be adopted and gain traction in the political sphere. Whether that is right is a matter of individual bias rather than technical proof - both studies were carried out with great care and consideration for the data.
What this proves is that analytics does not always provide simple answers, but can lead to more questions. Deciding which insights to adopt and act on is often just a matter of choice.