Monday, June 27, 2016

Brexit Dreams And Nightmares


The option "Leave" beat the option "Remain" in Britain's vote about whether it should stay in the European Union  (EU) or leave it.  And the floodgates opened.

If I were a wiser writer I'd stop right there, because so far I've uttered no opinions unsupported by any evidence:  the kind of analysis I've far too often read about the nightmare or the dream that is Brexit. 

Sure, there are factual articles, too, telling us how Cameron got into this political mess in the first place, what Britain pays to the EU and what Britain receives in return,  and what the various options Britain now has might be. But one reason why so many articles about the Brexit are opinion pieces is that the kind of data we would need for strong conclusions about the vote are very hard to find.

Take the information we might get from exit polls:  To find out the demographics of those who voted Leave and those who voted Remain.  But the Brexit vote didn't have official exit polls.  The ones which exist are private polls.

The private poll I saw was on the Lord Ashcroft site.  Google Lord Ashcroft and you will find that he is a British conservative, domiciled in Belize (to avoid paying UK taxes?), with rather Trumpian characteristics, though without the financial inheritance Trump has.  He also tells us on his site that he was for Leave himself.

This doesn't necessarily make the polls biased or inaccurate, because I doubt that it is Lord Ashcroft himself who carries out the polling.   In any case, his site offers the most comprehensive demographic data on voting patterns that I have been able to find and it is that data I wish to discuss here, with some fairly serious criticisms.


One final caveat:  I couldn't find a description of the sampling procedure on the Lord Ashcroft site.  The flavor I get from the survey results is that they were not based on a convenience sample but some type of random sampling.  Still, I cannot vouch for that.

Demographic Results

The number of people the survey interviewed was 12,369.  This table gives the main demographic results of the survey:

Note that there was no average difference between men and women, and that the older the voter, the more likely she or he was to vote Leave.  Much was made out of that in the social media (including some pretty nasty ageist stuff which is unfortunate when it comes from people who usually stress human rights and respect towards all in their work), but I saw very few attempts to analyze if the benefits and costs of belonging to the EU in fact might fall differently on different age groups.  As an aside, it's odd how very little data of that kind seems readily available.

The last four rows in the above table are about the social classes in Britain.  The groups A and B are the upper middle and middle classes, the group C1 the lower middle class, the group C2 is the upper part of the working class, consisting of skilled manual workers, while the group D is the lower part of the working class, consisting of semi-skilled and unskilled manual workers.   Finally, the group E gathers together everyone else: Casual or lowest grade workers, the retired and others who depend on the state for their income.

It's unfortunate that the group E lumps together those who have retired from the labor force altogether with those who are of the age to be in the labor force but are not, for various reasons.  The frustrations these individuals might feel about the EU are likely to be different, and their income-determined social class could also widely differ.

Then to the results:  There is a social class correlation in the Brexit votes:  The higher one's social class, the more likely the vote to stay in the EU.

Now contrast this result with the same survey's findings about how people belonging to different political parties voted:

A majority of those who backed the Conservative in 2015 voted to leave the EU (58%), as did more than 19 out of 20 UKIP supporters. Nearly two thirds of Labour and SNP voters (63% and 64%), seven in ten Liberal Democrats and three quarters of Greens, voted to remain.
It looks to me that the traditional tie between the Labor Party and the working classes has broken. 

But the survey also found that a majority of those working either full-time or part-time voted to remain in the EU.  I couldn't find the social class data for only those who work (though it may exist somewhere in the 304 pages of raw data tables), which makes the interpretation of all this more difficult. 

It also matters when evaluating the assertion that the Brexit was a working class revolt.  It could have been that, but it could also have been a revolt by those who are poor and/or on fixed incomes.  The Lord Ashcroft survey found that two-thirds of council house and housing association tenants, all living in forms of subsidized housing, voted for Brexit.  It looks like the poor were the ones most likely to wish to leave the EU.

Voters who listed their race or ethnicity as other than white tended to prefer the Remain alternative, and so did those who listed their religion as Islam:

White voters voted to leave the EU by 53% to 47%. Two thirds (67%) of those describing themselves as Asian voted to remain, as did three quarters (73%) of black voters. Nearly six in ten (58%) of those describing themselves as Christian voted to leave; seven in ten Muslims voted to remain. 
The Leave option won, because whites are the dominant majority in the population.  These results reflect the fact that concerns about immigration were a significant motivation among those who voted to leave the EU, and that would mostly be people who are not recent immigrants or whose parents weren't immigrants. 

What those immigration concerns might be is the crucial question.  It could be that Leave-voters were motivated by racism* or nationalism or xenophobia or anti-Muslim bigotry, and any time spent on reading the comments on British right-wing sites tells us that those sentiments certainly thrive

At the same time, very rapid immigration does stress the system of local public services (in the UK schools, health care services and subsidized housing), and these stresses are much greater in the poorer areas,  because those are the areas where immigrants are most likely to initially settle. 

Upper middle class and middle class individuals, on the other hand, are unlikely to find greater difficulties in getting an appointment with their doctors or dentists or suddenly larger class sizes in their children's schools or longer waiting for a state-subsidized apartment or a house.  Instead, they are more likely to benefit from the increased numbers of, say,  carpenters, landscape workers and nannies offering their services at prices which fall with the increased supply**.

The UK austerity policies have not exactly helped.  When more doctors, dentists and hospitals are needed, the UK government wants to cut health care spending.

Different parts of the United Kingdom had different percentages voting for Leave.  As a gross simplification, the Leave-option triumphed in England, with the exception of the capital, London,  and also triumphed in Wales, while the Remain-option did so in Scotland and Northern Ireland.  London strongly voted to remain in the EU, and some opinion-writers see the results in the light of London reaping most benefits from the EU membership while many other areas bear the costs.

So What Are The Brexit Dreams and Nightmares? 

 I chose the title to make you read a fairly boring post about statistics.

My particular nightmare is that the survey might be a convenience survey, after all, in which case all my work has gone wasted, because we can't make statistical generalizations about a convenience sample.  Still, the BBC article linked to it, so if I have egg on my face I share it with the august writers of the BBC.

The title also reflects my incessant reading of large numbers of opinion articles on the Brexit results.  Many of those told me much more about the dreams and nightmares of the individuals who wrote them than about the final results, simply because we are so very close to the actual election that more objective information is hard to get.  Imagine trying to look at a fern frond, say, by sticking it right next to your eyeball.  Any description of it you would then manage would reflect your ideas about fern fronds, not what you actually can see.

One nightmarish aspect of the results might already be noticeable, however:  Even those who voted to leave the EU expected the Remain-option to win.  Thus, the Leave-vote may have been a protest voice, a scream of frustration, aimed at all the faceless powers in London and in Brussels, not a decision that was carefully considered as a possible outcome of the vote.

One Final Statistical Point

To finish this post, I want to show you how statistical tables can be deceptive and how those statistical tables can then be used to draw false conclusions.

Let's begin by looking at this table from the Lord Ashcroft site:

And a very interesting table it is!  The sentence on the top is a question from the survey (I'm fairly sure of that), and the answers to that  question ( "Do you think each of the following as being a force for good, a force for ill, or a mixed-blessing?") are then given on a zero to ten Likert-type scale, where ten would mean that someone thinks a particular concept is very much a force for good and zero would mean that the same concept is seen as very much a force for ill.

Now in the data tables (Table 26 on page 108) for this survey we find out stuff like that.  We also find out that the people who might have answered "a mixed-blessing" are those who chose the five in the middle of the scale.  Then we compare that to the table Lord Ashcroft chose to show us and we find out something else:  All the people who chose to answer "mixed-blessing" are omitted from it.  Gasp.

Why does this matter?  Take, as an example, the question about whether feminism in the opinion of the respondents has been a force for good or a force for ill or a mixed-blessing.  For everyone who participated in this survey, whatever their final Brexit votes were, the percentage choosing "a force for good" (which means answers from 10 to 6 on the scale) was 57%.  An additional 27% chose the "mixed-blessing" alternative (score 5 on the scale), while only 16% picked the "force for ill" alternative.

Thus, what the left column of the above table tells us about the Brexit votes of those who think that feminism has been a "force for ill" applies to only 16% of those who participated in the survey.  Similar problems crop up with most of the "force for ill" answers:  They apply to  minorities among all survey-takers.  The one exception is the question about immigration, where 40% chose the "force for ill" option, 25% the "mixed-blessing" option and 35% the "force for good" option.

By omitting all the people who answered "mixed-blessing" on the various items the table manages to exaggerate the correlation which does exist in the underlying population.  I see no statistical reason to omit that muddy-middle, because it is a large group of people, ranging in size  from a high of  34% of the total sample on the impact of globalization to a low of 20% of the total sample on the impact of the Internet.

It must be this table*** that Charles Payne on Fox News used to state that the Brexit vote was a reaction to all sorts of things (at about 2:25 into the video):

Listen, this was a vote against immigration, establishment, multi-culturalism, feminism, a lot of the movements that have really swept western Europe and now America, being rebuffed.

The Lord Ashcroft survey questions don't cover the "establishment,'" but they do cover globalization.  Payne didn't mention the globalization results which suggest that those who consider globalization "a force for ill" were also much more likely to vote to leave the EU.

What Payne does in that quote is a further misuse of a table which has already been constructed to exaggerate differences as much as possible.  It's as if he studied some similar-looking table where the left column gave data on the beliefs of the individuals who voted to leave the EU and the right column on the beliefs of the individuals who voted to remain in the EU.  But that's not what the above table does.

* This post-Brexit article describing nasty harassment of Poles in London calls the incidents racist.  It's odd how "racism" has expanded to cover what one might call xenophobia or perhaps tribalism or just the hatred of the "other". 

I'm not sure that this linguistic development is altogether a healthy one.  It may water down the meaning of racism when its original meaning is really needed to defend groups who are treated abysmally simply because of superficial differences in features of the type that are attributed to the concept of race.  Besides, we already have the term "xenophobia."

** Recent immigrants are also much more likely to compete for jobs with British working-class individuals than with upper middle class or middle class individuals, so any downward pressure on wages/salaries is unlikely to affect the latter group which often has jobs protected by country-specific licensing or education requirements.  Here is one example about physicians.

*** Unless another study exists which asks so similar questions.  I doubt that.