30 January, 2005: You cannot hope to bribe or twist...

[ Home page | Web log ]

So, then, that panic over immigration, eh? Whether you're a crazed xenophobe or have more-or-less sensible views on the matter, it can't have escaped your attention that lots of people seem to be very concerned about the topic. Anthony Wells recently wrote a summary of recent polls on immigration (that's on his old weblog -- he's now moved to a new one on his splendid Polling Report site), which led me to ponder again whether the current concern about immigration is real, or whether it's a self-fuelling media scare story.

That there is substantial public concern isn't in doubt. MORI does regular polls in which they ask the questions,

What would you say is the most important issue facing Britain today?

and

What do you see as other important issues facing Britain today?

-- and as you can see, over the past few years the fraction of people responding with concerns related to immigration has been rising pretty steadily:

MORI polls on immigration

(That shows the combined percentage of people who named any of the categories `racism', `immigration' or `immigrants' in their answers to either question. The points in the plot are individual poll results; the smooth red curve is derived from those results by a filtering procedure; the dotted lines show the approximate scale of the margin of error. Poll results are quite noisy and you can see that approximately the right fraction fall outside the c. ±3% margin of error for 1,000-person polls at 95%.)

However, it turns out that much the same rise -- and a later fall -- occur for lots of the issues which MORI track. Here, for instance, is a similar plot for respondents concerned about Europe (specifically, `the Common Market', `the EU', `Europe', and `the Single European Currency'):

MORI polls on Europe

and if you align the two curves, concern about each of the two issues can be seen to rise at about the same rate -- and, in the case of concern about the EU -- fall afterwards on about the same timescale:

Comparing EU and immigration issues

And this despite the fact that objectively (to borrow an Oliver-Kamm-ism) there is probably more cause for public concern now about the EU -- over the Euro, the proposed Constitution and Single European Arrest Warrant -- than there was in 1997-99. Similarly, it's not clear that there's any particularly good reason for people to be very concerned about immigration now rather than at any other time.

So, this leads to the slightly cynical hypothesis that transient public concern about issues -- like immigration and the EU -- which have little tangible effect on people's everyday lives is actually a product of reporting in the newspapers. Journalists, stuck for a story on a slow news day, make up some rubbish about asylum seekers stealing our jobs or the EU criminalising bananas or whatever, other journalists nick their good idea, and soon enough everybody's talking about it. A couple of years later, we're bored of the whole topic, editors across the land cry out, ``Not another feature about bloody bananas/asylum seekers/whatever! Can't you come up with something different for once?'', and so the cycle begins anew.

A nice story, but is it true?

Certainly, if you look at mentions of `immigration' or related terms in the tabloids (I searched for ```immigration' or `immigrant' or `asylum''', on the grounds that the Great British Public doesn't understand the difference between asylum and other routes to immigration, and the papers seem agressively to pander to our ignorance on this), you see a rise in the number of related stories over much the same timescale as the MORI polls show increasing concern over immigration: (I got these from the Infotrac database of newspaper stories, which is generously provided to me by Cambridgeshire libraries; there are some issues with the data quality there -- certainly some of the stories are repeated -- but I don't think this invalidates the whole exercise)

Mentions of immigration in the papers

Note that the paper with the largest overall number of stories on immigration topics is -- probably to nobody's great surprise -- the Daily Mail -- but the paper which has shown the largest relative increase (compared to the level in 1995/6-99) in the number of such stories is the Sun:

Mentions of immigration in the papers relative to `background' level

And if you plot up the two variables (showing, for clarity, only one newspaper), the two curves coincide quite nicely:

Mentions of immigration in the 'Sun', vs. poll results

... but this doesn't tell us whether the newspapers caused the panic about immigration, responded to it, or an unholy mixture of the two.

The classic way to investigate this is to difference the data (showing, in this case, the change on the previous month's values), and then to see whether there is any evidence for causation (i.e. rises in the number of mentions of immigration in the newspaper precede rises in public concern). Here's a typical plot (showing mentions in the Daily Mail against poll data):

Differenced data

-- it's not abundantly clear, but basically there's no evidence here that changes in what the Daily Mail says affect later public opinion on immigration, and similarly for the other papers.

We have to be a bit cautious, though. If MORI's polls were consistently conducted at the end of the month, for instance, then they might be affected by what the papers say during that month; if they were consistently conducted at the beginning of that month, they might affect what the papers say later on. In fact they seem to be conducted at different times each month, but I haven't checked this completely.

Effectively, there would have to be a lag larger than a month in the effect of the papers on public opinion (or vice-versa) for this technique to show it up, and that seems to be a bit unlikely -- after all, the tabloids change their editorial line on subjects much more quickly than that, suggesting that they don't believe that their readers' memories are that long.

With more data about when, specifically, each poll was conducted, it might be possible to be a bit more accurate -- in particular, stories published immediately before and during the fieldwork for each poll -- usually two or three days -- could be counted, rather than those for the whole month. But there doesn't seem to be a convenient tabulation of that data to do the comparison.


Copyright (c) 2005 Chris Lightfoot; available under a Creative Commons License.