A joint Home Office and Department for Business, Innovation and Skills (BIS) report looking at the impacts of migration on employment of the UK-born population has been published today.
This report builds on a previous report produced by the UK Migration Advisory Committee (MAC). A much publicised result from the MAC research, which was highlighted in speeches by the Home Secretary, Theresa May, was that “an extra 100 non-EU working-age migrants are initially associated with 23 fewer native people employed.”
This conclusion comes from a statistical model estimating that, for the period 1995-2010, the ‘coefficient’ representing the association of UK-born employment rate with the ratio of non-EU-born migrants to UK-born individuals was -0.230, with other relevant factors held constant. (A ‘coefficient’ in a statistical regression shows how much change in one variable is associated with a single unit change in another variable.)
But this finding is carefully caveated. In the report, the MAC presents the results as ‘tentative’ and presents other variations of the estimations in which the results do not hold – for earlier time periods, for periods of economic strength (a positive ‘output gap’ for the UK), for EU-born migrants, and for other technical changes to statistical models.
The results from the MAC differed from those of other studies published around the same period and the Migration Observatory explained differences in the methods used in those studies. However, given the importance of this discussion for policy regarding migration, the Home Office and BIS produced today’s new report, which summarises the existing evidence on this topic and explores the robustness of the MAC results. BBC’s Newsnight and other media seized on internal emails about this report prior to publication, and making claims that the new study challenged the results of the MAC report, found that the relationship between immigration and UK-born employment rates was smaller than what the UK government claimed.
Now the report has been published. Are the conclusions of the Home Office/BIS report very different to those from the MAC?
In short, the new report confirms the MAC’s initial analysis and shows they were ‘robust’ (i.e. held firm) in several alternate models; however, the report adds new results that went further than the MAC, especially in highlighting that the initial findings were very dependent on the inclusion of the economic downturn in the time period being examined.
The Home Office/BIS report tests the robustness of the MAC results in four key areas:
- The labour market variable of interest: it is possible to look at the impact of immigration on different labour market variables. The possibilities include employment rates, unemployment rates, and economic ‘inactivity’. These variables measure different things and are all valid indicators of labour market conditions, but results may change as the estimation looks at one or other of these variables. The original MAC report focused on the employment rate as the key variable of interest. The Home Office/BIS report also provides estimates in which the variable of interest is the UK-born unemployment rate, which refers to people who are out of work but looking for work. Employment rate is not simply the flip side of the unemployment rate because official statistics include a third category: the economically inactive, who are neither working at paid employment nor looking for work. The Home Office/BIS report finds that estimates of the relationship to non-EU migration change depending on the variable used in the analysis. In particular, the association between immigration (EU and non-EU) and the UK-born unemployment rate is not statistically significant. Verdict: MAC estimates are not robust to looking at unemployment rate, but impact on the employment rate is also a valid way of looking at labour market effects.
- The period of analysis: it is possible to conduct the analysis during different periods to explore the robustness of the results and changes in the results across time. The original MAC analysis was conducted for the period 1975-2010 and for the sub-periods 1975-1994 and 1995-2010. The Home Office/BIS added analyses for the periods 1995-2007, 1995-2008, 1995-2009 and 1995-2012. The results show that the MAC findings are very sensitive to the period of analysis. In particular, analyses that exclude the years 2009 and 2010 do not show a statistically significant relationship between non-EU migration and employment rates of the UK-born. The analysis for the period 1995-2009 finds a significant but smaller impact (i.e. the coefficient for the impact was -0.156, versus -0.230 in the original estimation). Likewise, the analysis for the period 1995-2012 suggests that the impact of immigration on native employment is negative but a bit smaller than the original MAC estimates (i.e. coefficient for the impact was -0.210 versus -0.230 in the original estimation). Verdict: MAC findings depend on inclusion of economic downturn years. It is worth noting that, in a less widely reported result, the MAC found this as well, showing that the association between non-EU migration and employment rates was only statistically significant during years with a positive ‘output gap’ indicating strong economic performance for the UK.
- The definition of a migrant: The two most common definitions of a migrant in economic analysis are by country of birth and by nationality. The MAC report defined migrants as foreign-born individuals. The Home Office/BIS report also provides results in which migrants are defined as foreign nationals. This change do not alter the results much (i.e. coefficient for the impact was -0.221 versus -0.230 in the original estimation). Verdict: MAC results are robust to alternate definitions of a migrant.
- Geographical location: Much of the economic activity of the UK takes place in London. At the same time London is home to about 40% of the foreign-born population of the UK. The Home Office/BIS report presents an estimation in which London is omitted from the analysis. Removing London from the analysis actually results in a slightly stronger negative impact of non-EU migration on native employment (i.e. coefficient for the impact was -0.242 versus -0.230 in MAC estimation). Verdict: MAC results are robust to the inclusion or exclusion of London from the analysis.
Table 1 summarises the findings from the Home Office/BIS report.
Table 1 – Home Office/BIS report findings
|Model||Impact of non-EU/native ratio on native unemployment rate|
|MAC original model (1995-2010)||Significant association with coefficient of -0.230|
|London omitted (1995-2010)||Significant association with coefficient of -0.242|
|Migrants defined as foreign nationals (1995-2010)||Significant association with coefficient of -0.221|
|Period 1995-2012||Significant association with coefficient of -0.210|
|Period 1995-2009||Significant association with coefficient of -0.156|
|Period 1995-2008||Insignificant association|
|Period 1995-2007||Insignificant association|
|Variable of interest is unemployment rate (1995-2010)||Insignificant association|
In addition, the re-examination shows that there is not a contradiction between the MAC’s results and the Home Office/BIS results. The latter report builds on the former, showing that the MAC results, given the formulation of their data set, depend critically on the inclusion of a period of economic downturn while remaining robust to other changes in statistical modelling.
In general, the main message that one could take from this set of reports is that there is no one universal relationship (holding at all times and for all migrants and all parts of the labour market) between immigration and employment rates among the native-born. The relationships seem to vary substantially with time, and particularly with economic circumstances, with strong evidence for a relationship only during times of weaker economic performance for the UK. On the other hand, it is not impossible that some displacement occurs at other times. Although there is not evidence for it in these results, a statistically insignificant result (i.e. lack of evidence) is not evidence that there is no real effect. As the saying goes, the ‘absence of evidence’ is not the same as the ‘evidence of absence’. (And, equally, a statistically significant result has a small chance of occurring by random variation.)
- Home Office and Department for Business, Innovation & Skills – Impacts of migration on UK native employment: an analytical review of the evidence
- Migration Advisory Committee – Analysis of the Impacts of Migration
- Migration Observatory commentary – Migrant workers: Taking our jobs – or not?