From the TUC

Another view on the referendum results: Unity – not division – against the metropolitan economy

20 Jul 2016, by in Economics

The Resolution Foundation (RF) has led the way in interpreting referendum results according to economic and other factors. However their latest contribution “The importance of place“, seems to me to underplay the importance of place.

Outside London and other metropolitan centres, the vote to leave was remarkable for its uniformity. A relation with income is evident throughout, but the effect primarily operates between regions, and most obviously between London and the rest of the country. Antony Hilton in the (London) Evening Standard saw this immediately: ‘the country has voted against the London economy’. Earlier this week in the FT Gordon Brown set ‘cosmopolitan cities’ apart from the rest of the country.

In not drawing out this effect, my fear is that the RF are overplaying other explanations for the results and underplaying the basic story of the economy that is not delivering for great parts of the country.

The analysis here is based on a series of charts from different regional views, which I hope are of interest in their own right.

Leave vote by main region/nation

Central to understanding the voting is the aggregate (NUTS1) regional perspective (nine English regions and Scotland, Wales and Northern Ireland). On the basis of the share voting to remain or leave in each region/nation (as on the left): eight English regions voted to leave, excepting London; Scotland voted most decisively to remain; Northern Ireland also voted remain; Wales voted to leave.

Looking at the share of local authorities within each region returning a majority remain, the results are thrown into a different and starker relief. In Scotland all local authorities had a majority for remain. In the West Midlands it was only one. (NB Northern Ireland is omitted here because there are no results on a LA basis.)

unity3_regions alone

Of all English local authorities outside London, only 18 per cent voted to remain in the EU. As will be seen in the detail below, the ‘metropolitan’ factor is well proxied by those authorities with Russell Group (i.e. major) universities, for example the West Midland’s one ‘remain’ LA was Warwick. If these are excluded the share falls to 15 per cent.

Income and the leave vote by region / nation

The Resolution Foundation’s initial analysis pointed out the generalised relation between the leave vote and income. Torsten Bell showed a scatter plot of earnings against the share voting to leave by local authority (I’ll come back to “R2=0.3289”).


At this stage the RF did not address the regional dimension. The most obvious approach is to compare the aggregate regional results with regional prosperity, according to GDP per head (strictly, for regional data, GVA – gross value added).

Leave vote against GVA per head


Three points are obvious:

  • London is on its own in terms of both prosperity and the scale of the remain vote
  • across London and all other English regions, the leave vote is negatively related to income – the lower income, the higher the remain vote
  • there is a very different relationship for the nations (excuse the shorthand), with a stronger remain vote than might be expected on the basis of income alone

The standard measure of association is ‘R2’, which is zero when there is no relationship and one when the relationship is perfect. Bell reports for all LA datapoints a R2 of 0.33. Taking just the English Regions on the chart above, R2 is 0.91 – absurdly strong. Because London is so far removed on this basis, statistically it could be treated as an outlier and omitted from calculations. R2 is then 0.57, still a strong association.

Detail by local authority / nation

Aggregating results in this way abstracts from action in the detail. The chart below adds a regional dimension to Bell’s scatter plot, distinguishing: a. Scotland Wales and Northern Ireland (red), b. London (black), and c. everywhere else in England (green).

Weekly earnings v % leave vote, by local authority


(NB I have reverted to ASHE/annual survey of hours and earnings data on pay, as GVA is not calculated for local authorities. The RF have used hourly pay, I have used weekly.)

The scatter bears a strong relation to the earlier ‘NUTS1’ chart. The majority of the remain vote is still seen to be LAs in London and the nations, with a smattering of others (for instance the green dot with the strongest remain vote is Cambridge; conversely the strong London leave is Havering). I have put a box around the main cluster of leave votes to help put subsequent charts into perspective. Looking at just England outside of London, the R2 Is only 0.13; excluding metropolitan, it falls to .09.

The next chart shows London by local authority, some of which are labelled. Remain LAs were across all incomes. While these remain votes were at incomes generally associated with leave through the rest of England, incomes in all London LAs are also towards the top part of the distribution outside London. Any leave vote in London was in lower paid LAs (though not the lowest).


Within London there is a stronger relation with income than over the rest of England, with a R2 of 0.26.

The next two charts are example English regions – the rest are in an annex.


For the West Midlands everywhere is in the box, except for Warwick. The R2 is 0.23 including Warwick and .08 excluding Warwick. In the North West, of those outside the box, seven LAs voted for remain – including Manchester and Liverpool (university cities) and others with lower incomes. Again, excluding university cities gives an R2 of 0.0 (though Copeland and Fylde are oddities as well). The six other English regions are in an annex. A negative relation is basically evident throughout (except Yorkshire and the Humber), but R2 is generally below 0.1 and the association is therefore weak – figures for all regions are also in the annex.

Scotland is unique in returning remain in all LAs, in spite of earnings consistent with a leave vote in England. A quarter of Welsh LAs returned remain – a higher proportion than all English regions but London. However in both cases, after the ‘metropolitan adjustment’, both have a very faint positive association between the leave vote and income.


At this detailed level, therefore, the association with income is therefore evident, but only faintly, and there are exceptions. This is especially the case when metropolitan centres are omitted from comparisons. London is the single region in the UK where higher income is more clearly associated with a lower leave vote. The difference is really London/metropolitan on the one hand, and everywhere else on the other.

Lastly however it is worth stressing that in most LAs the difference between the share for remain and leave was not huge, it was not e.g. 80 per cent leave and 20 remain. The median (i.e. middle) LA was (54 leave, 46 remain). 80 per cent of all LAs, were between (65 leave, 35 remain) and (40 leave, 60 remain). A fuller distribution is shown below:



According to the above, the biggest driver of the leave vote was place. And the most plausible driver of this is the stark difference in prosperity between London and the rest of the country. Given these regional disparities are known to have existed for decades, the inference is that the vote reflects a deep-rooted and long-standing disaffection with economic outcomes.

The RF illustrate this by showing that while the leave vote is proportionate to income, it is unrelated to changes in income over the past 5 years. I am not sure it is this straightforward, given regional disparities have increased over recent years – GDP and employment in London have grown significantly faster than over the rest of the UK (see also Haldane’s speech last week, drawing attention to significant widening of inequality on a number of different measures). Disaffection is likely to be deep rooted, but intensified over recent years.

The RF however emphasise that economics is “by no means the only consideration”. Their results show associations between the leave vote and various other characteristics, notably age and education. Here for example is the relation with education, the strongest they record with a R2 of 0.63.


However this relation follows very closely the generalised relation with income. The grouping ‘higher paying, low leave’ identified in dark blue is a group of London LAs, with Oxbridge and a couple of Scottish LAs. But this is still likely to be primarily a result of the regional factor. We know more generally that the better qualified are better paid. If the better paid are concentrated in London then the strength of the relationship with education may simply be a side effect of the London/income effect.

The same seems also true of age. The RF results here are weak, with a R2 of only 0.08.


I suspect if you dropped the dark blue figures, the R2 would be close to zero. Young people gravitate to the prosperous parts of the country. The average age in London is 33, compared to 39-42 elsewhere.  So again the relation with age is likely to be distorted by the strength of the London vote.

RF also finds for example that LAs with a high numbers of migrants are associated with a low leave vote. Again this is likely to be economic. Migration flows to high income areas – that is where the work is. Conversely RF finds that home ownership is associated with a higher leave vote – surely reflecting the inaccessibility of the housing market in London.

Finally a caveat. Obviously it is very difficult to draw clear cut conclusions when actual individual motivations one way or the other are not known. But my sense is that this London/ metropolitan factor trumps all others, though I am anxious that the RF see it so differently. (Perhaps when you correct for everything you inevitably see no London effect, because London is where everything impacts?) I should stress that age and education may also play a role – I have not explicitly explored this – but my point is that without correcting for London/metropolitan the effects are likely to be exaggerated, potentially to a significant extent.  There has been a lot of talk of division – but on my reading the defining feature of the referendum results is unity (*).


For me the scale of the leave vote is bound up with widespread disaffection around economic conditions and prospects. The failed economic outcomes in the wake of de-industrialisation is hardly news, as are the extraordinary imbalances associated with the bloating of the financial sector. Perhaps from metropolitan England this has been too easily ignored or downplayed for too long. As I have repeatedly stressed, even the IMF have now recognised that the neoliberal or financial globalization agenda of the past 40 years or so may have been ‘oversold’. Gordon Brown has begun to address these issues– this debate must now be joined in a serious way.


(*) For this blog I have to thank Mrs Tily, who was worried by the constant emphasis on division. As with everything else, she can’t be held responsible if I messed it up.

ANNEX: Other English regions



3 Responses to Another view on the referendum results: Unity – not division – against the metropolitan economy

  1. P
    Jul 20th 2016, 7:27 pm


  2. John Papadachi
    Jul 27th 2016, 5:19 pm

    You seem not to have used residence-based median weekly earnings. St. Albans sticks out like a sore thumb in terms of the difference between pay of commuters and those working in the district and was very big for Remain. So it seems the relationship between earnings and voting is underplayed in that district at least. If I am right you no doubt had good reason to use the ‘wrong’ dataset. But I would have preferred it if you had admitted this.

  3. George Krimpas
    Aug 10th 2016, 11:16 am