Why MigrationWatch is wrong — a plea for a more robust debate on immigration
If MigrationWatch wants to be taken seriously, it needs to go back to the statistics textbooks.
Last week, a series of media headlines suggested that immigrants were taking jobs away from British people as the economy enters recovery. These stories were based in part on new ONS statistics, bolstered in some of the papers by reference to a report from MigrationWatch which purported to show that recent immigration to the UK has caused higher unemployment.
I wrote here about why these stories were misleading, and explained that the MigrationWatch report had failed to demonstrate the causal link claimed by the headline on its press release ("Immigration has damaged employment prospects for British workers", MigrationWatch press release, 12 August).
Perhaps it's to be expected that MigrationWatch might play a bit fast and loose with the evidence -- it is a campaign group, after all. But this seems to be becoming a habit: yesterday it published another report, using an even more flawed methodology, to make very strong claims about immigration and youth unemployment. This was duly picked up by the right-wing media -- the Express today repeats this study's claims under the headline "Migrants rob young Britons of jobs", and the Telegraph goes for "13 per cent rise in Neets 'linked to immigration' ".
So here, for the record, are three important methodological reasons why MigrationWatch's claims don't stand up.
1. It is looking at the wrong variables (part 1)
Yesterday's MigrationWatch report on immigration and youth employment made the basic error of comparing the absolute numbers of migrants and young unemployed people between local authorities, rather than migration or youth unemployment as a percentage of local populations. This makes its findings more or less worthless.
It means that, in all likelihood, all it has demonstrated is that different local authorities have different populations (for example, in the 50 local authority areas the report analyses, Manchester has a population of roughly 500,000, while Cambridge has about 120,000 only), and that local authority areas with higher populations have both more migrants and more young unemployed people.
This probably also explains why MigrationWatch finds a much weaker relationship between migration and youth unemployment in London: London boroughs have more evenly-sized populations than the areas that the organisation's analysis focuses on.
2. It is looking at the wrong variables (part 2)
Last week's MigrationWatch report did look at migration and unemployment rates as percentages of local populations, rather than absolute numbers. But it plotted net migration between 2003 and 2009 against levels of unemployment in 2008/2009 -- any basic descriptive analysis should have looked at the change in unemployment in the period in question.
MigrationWatch would like us to think that immigration has caused unemployment to rise, but perhaps migrants have moved into areas that have always had high levels of unemployment, and immigration has had absolutely no impact at all on this. In fact, perhaps the migrants have been part of an economic renaissance in those areas and unemployment has fallen. The MigrationWatch report can tell us nothing about any of this, because it looks only at unemployment levels in 2008/2009, not changes in unemployment.
In fact, plot MigrationWatch's own 2003-2009 immigration measure in "high immigration" areas against changes in unemployment between 2003/2004 and 2008/2009, and you see that the relationship it claims to find between migration and unemployment disappears. In fact, we might even suggest that it is reversed -- the chart below (analogous to Figure 3 in last week's MigrationWatch report) could be taken to show that higher immigration is associated with lower increases in unemployment (coefficient = -0.2).
If I was working with the MigrationWatch model, I would now issue a press release saying: "Immigration reduces unemployment: for every 1,000 immigrants moving into these areas, unemployment was reduced by 200, on average. The government should therefore increase immigration in order to tackle worklessness."
I'm not doing this, because the chart above shows no such thing. If we excluded a couple of outliers it would show no relationship at all; and at best it shows correlation, not causation. This is MigrationWatch's third major methodological mistake.
3. The reports show (if anything) correlation, not causation
Both MigrationWatch reports show, at best, correlation, not causation. But the organisation makes strong causal claims on the basis of their analysis -- that immigration causes unemployment, and that reducing immigration would help to reduce unemployment. Last week Andrew Green used one of the reports to claim that
. . . this demonstrates that the "open-door" policies of the past decade have had a damaging effect on the employment, and therefore the standard of living, of UK-born workers in the areas most affected.
And yesterday, he said that: "Many factors contribute to youth unemployment but this research suggests that immigration is a significant factor in areas of high immigration. The case for getting immigration down to sensible levels, as the government has promised, gets stronger by the day."
Oddly, MigrationWatch now seems aware of this limitation of its methodology (perhaps someone attached to the organisation has read my paper, or Monday's column in the Guardian from Gary Younge), and indeed it admitted as much in yesterday's report (and associated press release) when it said:
These findings are based upon correlation analysis which does not, by itself, provide evidence of any causal relationship . . .
But it can't stop itself making unfounded claims. The report continues directly:
. . . but the findings nonetheless are highly indicative of the relationships between youth unemployment and migration.
Overall, the key findings are that:
- The relationship between immigration and youth unemployment is positive and significant in the 50 local authorities in England with the highest rates of migration in the period 2003-2009, and in London.
- The relationship becomes stronger and more adverse the higher the rate of immigration. In a sample of those local authorities outside London with the highest rate of immigration, the relationship is very strong (correlation of 0.9), and shows that, on average, for every 1,000 immigrants into these areas, the number of young unemployed rises by around 900.
The relationship MigrationWatch showed in yesterday's report may be positive (though, as noted above, it is looking at the wrong variables, which makes the results more or less irrelevant), but it isn't significant (in the statistical sense), and the group certainly hasn't demonstrated that an additional 1,000 immigrants arriving in any area caused 900 more young people to be unemployed.
I set out some more detail on this point, and some alternative interpretations of the findings from last week, in a paper published this week. To demonstrate that immigration has caused unemployment, MigrationWatch would have needed to conduct some econometric modelling, controlling for a range of other factors and establishing statistically significant results.
In fact, a number of studies have already done this. They are summed up in a paper by my colleagues Maria Latorre and Howard Reed:
In short, the best available UK microeconomic evidence on the effects of migration on employment finds either no effect at all, or very small negative effects.
This conclusion is also supported by a wide range of research in other OECD countries.
None of this is to say that immigration has never had an impact on employment -- indeed, it seems likely that it has had impacts in some areas of the UK, and perhaps for young people in particular. In general, however, the claim that migration causes increased unemployment is not supported by the evidence, and is definitely not proved by either of the MigrationWatch reports.
Poor-quality evidence harms the debate
If MigrationWatch wants to engage seriously on the question of migration and employment, it needs to go back to its statistics textbooks. I don't agree with it, but I'd like to engage in debate about its arguments, rather than its research methodologies.
This is, in effect, a plea for MigrationWatch to up its game -- if it finds any real evidence that migration has harmed employment in the UK, we all need to know about it. It is also worth remembering that this isn't just some intellectual or rhetorical exercise; important questions of government policy (and the lives of real people) are affected by the public mood that MigrationWatch helps to generate.
But this article is written much more in sorrow than in anger. Nick Clegg said this week that it is important to make sure that the immigration system "has people's confidence and trust and they know it works". The first step towards this objective is to have an open and robust public debate about immigration; but that debate can happen only on the basis of a shared understanding of certain basic facts, and of what constitutes valid evidence.
Sadly, MigrationWatch's contributions in the past week confuse, rather than enlighten.
Sarah Mulley is a senior research fellow at the Institute for Public Policy Research (IPPR).