Since the referendum in 2016, the claim that the Leave victory was delivered by the ‘left behind’ has become commonplace. But the prominence of the claim is not matched by its precision: ‘left behind’ is used as shorthand for relative economic decline, for political neglect and policy failures, or for liberal cultural shifts that have alienated socially conservative voters.
Clear empirical evidence in support of the claim is hard to come by. Where evidence is presented, it is based either on the demographic characteristics of geographic areas or on polling that aims to identify the ‘values’ of voters.
It is assumed that places where the population tends to be whiter, less well-educated and more likely to be working in manual occupations are more ‘left behind’. Those who hold authoritarian views and oppose immigration are likewise given the label.
In our opinion, precise use of language is necessary for successful analysis. In a recent working paper, therefore, we compared a precisely-defined measure of geographic deprivation with Leave vote shares in the 2016 referendum.
We found that the relationship between the two is neither strong nor straightforward, and varies significantly between different types of deprivation.
To perform the exercise, we compared Leave vote shares to deprivation rankings in English electoral wards. We used the English Index of Multiple Deprivation (IMD), which is a weighted average of seven separate ‘domains’ of deprivation: income, employment, education, health and disability, housing and services, crime, and living environment.
For each, several indicators are combined to construct an index of deprivation in that domain. For example, the index of living environment deprivation combines indicators of housing in poor condition, houses without central heating, air quality and road traffic accidents to produce a single index.
These domain indices are then combined – using quite a complex procedure – to produce a single headline ranking of multiple deprivation.
Alongside the IMD, we built the most detailed dataset of Brexit voting data yet available, using unofficial ward-level voting data obtained by BBC journalists Martin Rosenbaum and George Greenwood.
Although the dataset does not cover the whole of England, it provides substantial coverage at much higher resolution than the official data.
We performed two analyses with this data. First, we assessed the correlation with Leave voting using headline IMD rankings; and second, we used a modified ranking that adjusts the weights applied to the separate types of deprivation.
Using the headline IMD ranks, we found only a weak positive relationship between deprivation and strength of Leave voting, as shown in the figure below.
The correlation of 0.17 between the IMD and the Leave share represents a 59% probability that any pair of geographic areas will have the same relative ranking when using the IMD as when using Leave vote strength. This is not substantially greater than the 50% probability that would be expected if there were no correlation. 
As well as examining correlations between the IMD and Leave vote shares, we also investigated the relative importance of individual types of deprivation by applying weights to these that maximises the association between each one and Leave voting.
This exercise confirmed that education is the most important type of deprivation in this context. Further, if we control for education, as in the figure below, the relationship between deprivation and Leave voting is reversed: less deprived places are more likely to vote Leave.
Thus, while there does appear to be a positive relationship between geographic deprivation and Leave voting, the association is weak, varies significantly between different types of deprivation, and doesn’t fare well when we include other variables.
The latter problem, in particular, casts some doubt on the existence of a causal relationship between deprivation and Leave voting, but the data employed in our study do not allow us to arrive at sharp conclusions.
More generally, our results neither support nor undermine the ‘left behind’ hypothesis, for the simple reason that it is too imprecisely defined to be tested formally. In our opinion, it is time that the term is retired from use.
The conflation of economic issues with cultural effects – and the lack of precision concerning either mechanism – obfuscates rather than enlightens. At its least sophisticated, it is simply a commitment of the nominal fallacy: the belief that labelling something is equivalent to explaining it.
This leads to a danger that the term can become all things to all people: academic use of the term could be taken as confirmation of a wide range of pre-existing views.
Moreover, lack of precision and critical awareness has led to a perhaps surprising range of statements being accepted as credible – take, for example, the statistically dubious claim that the ‘working class’ have abandoned the Labour party.
In some manifestations, the ‘left behind’ explanation converges with views which privilege the ‘white working class’ as a social group, despite there being no basis for treating this demographic as holding distinct economic interests.
Instead of uncritical acceptance of imprecise labels, we should recognise that the drivers of Brexit – and, indeed, of populism more generally – are complex and multidimensional. It is hard to draw firm conclusions; reasonable people will continue to disagree over the relative strengths of competing explanations.
We hope that the evidence in our paper will serve as further stimulus for these ongoing debates.
By Rob Calvert Jump, Research Fellow, Institute of Political Economy, Governance, Finance and Accountability at the University of Greenwich, and Jo Michell, Associate Professor of Economics, UWE Bristol.
