The Tories’ missing half million voters

The Conservative party’s well-documented failure to attract ethnic minority voters may not be set in stone, find Richard Webber and Trevor Phillips.

With the next general election just a year away, informed opinion is increasingly anticipating a close result and another hung parliament. It seems likely that the shape of the next government will depend on the precise share of seats between the three (or possibly four) major UK parties.

As the poll approaches, attention will focus first on the likely behaviour of electors in marginal seats, and in particular, those whose choices are likely to defy easy prediction. A vital handful of votes could mean the difference between success or failure locally, and the keys to Number 10 Downing Street nationally.

In this context, analysts worry about the effect of class (or more precisely, occupational background), geography and age; and much has been said and written about David Cameron’s ‘woman problem’ elsewhere. However, in recent years, less attention has been focused on the preferences of ethno-cultural minorities. The principal reason for this neglect has, of course, been the consistent historic support for Labour amongst visible ethnic minority voters in the UK.

But the high degree of uncertainty about the outcome of the next election, coupled with the recent appointment of the UK’s first Secretary of State from an Asian background, invites some re-examination of this aspect of politics. Will ethno-cultural background influence the outcome of the 2015 general election; and if it does, who will benefit, and under what conditions?

We have studied some past and current voting trends amongst minority voters and our analysis suggests that in 2015 we should definitely count ethno-cultural identity as a critical determinant of voting behaviour. In particular, we have found evidence that recent demographic changes in the distribution and the degree of clustering of visible minority ethnic groups is likely to lead to material change in their voting intentions.

In this essay we use the term visible minority interchangeably with ‘non-white Britons’ or ‘non-white minorities’; we distinguish their behaviour from more recently arrived white minority groups such as Poles and Americans.

The state of play

The most recent study, commissioned by the Conservative peer Lord Ashcroft, suggests that there has been little long-run change in the voting patterns of visible minorities. In fact, Ashcroft found that the single best predictor of not voting Conservative is ‘not being white’; just 16 per cent of non-whites voted for David Cameron’s party in 2010. A further study for British Future suggests that the shortfall between this share and the Conservatives’ overall share of 36 per cent represents around half a million voters. Both studies conclude that it will become increasingly hard for Conservatives to assemble a majority of any kind from an electorate as ethno-culturally diverse as the UK’s.

Yet there is a great deal of evidence to suggest that any emerging volatility amongst the UK’s five million or so visible minority ethnic electors could produce unexpected outcomes; in fact, changes in minority voting behaviour could turn out to be at least as significant as the effect of the three million electors who tell pollsters that they might support UKIP in the forthcoming election.

A glance across the Atlantic gives us a hint why seemingly insignificant trends may be vital to the result of the next election. The impact of racial preferences on the last US Presidential election has caused many observers to challenge the conventional wisdom that the predictability of the ethnic minority vote makes it politically unimportant.

The white/non-white split between the President and his challenger was indeed foreseeable; but Mitt Romney’s 59 per cent to 39 per cent lead amongst whites was comfortably offset in securing electoral college votes by Barack Obama’s crushing lead – 80 per cent to 20 per cent – amongst black and Hispanic voters, despite the latter constituting just 28 per cent of the electorate. Why the difference between the popular vote and the electoral college? What counted most in 2012 was not the Democrats’ net advantage amongst minority voters, but two other factors: turnout and distribution.

A study for the Brookings Institution by the demographer William H Frey suggests that it was, in large part, differential turnout by ethnic groupings which made Obama’s second victory possible. Had the turnout numbers been more in line with historical figures, (that is, had whites been more enthusiastic about voting, and blacks and Hispanics less so) we could have seen a very different outcome. Frey estimates that on a 2004 turnout, 2012 would have been either a dead heat, or a Romney victory.

However, the impact of differential turnout would have been minimal if all that had taken place was that black and Hispanic votes simply piled up in Democrat held electoral districts. But it was made far more significant by the rapidly-changing demographic distribution of minorities. As new voters have joined the rolls in previously non-diverse swing states, such as Nevada, they neutralised the Republicans’ historic advantage amongst whites.

Nevada has been carried by the eventual winner in every Presidential contest but one, for a century. In 2012, Obama owed his victory to a huge turnout by Hispanic voters, who now constitute 19 per cent of the state’s electorate – up from 10 per cent in 2004; he took three quarters of the Latino vote. This story was repeated elsewhere; thus a tiny 50.5 per cent to 48.0 per cent Obama margin in the popular vote nationally translated into a substantial 332 to 206 vote victory in the electoral college.

In the complex arithmetic of the UK’s parliamentary elections, it’s not hard to imagine a similar, relatively small shift in the distribution and turnout of minority voters having an amplified impact. This volatility effect could be further intensified in years to come by three demographic factors particular to the UK.

First, the minority population in the UK is growing both absolutely and relatively. There are now proportionately fewer white Britons in the electorate than ten years ago – a decline that is likely to continue given that one in four children in British schools is from an ethnic minority background. The direction in which minority voters shift will be of increasing significance in UK elections as a whole; and it is already demonstrably decisive in London and cities such as Leicester. On current trends, this growth should be good news for Labour.

However there are two other factors that might favour the Conservatives. One is generational. The orthodox view is that ‘integration’ (or more precisely, convergence of behaviour between different ethno-cultural groups) will naturally occur as time passes. The expectation would be that succeeding generations of minority voters may drift away from their parents’ reflex support for Labour – or at the very least, might not inherit the first generation immigrant’s hostility to Conservatives.

The other factor that could boost the Tory share of seats is a steady out-movement of ethnic minority families from the inner city neighbourhoods in which they first settled. This process is accelerating in London in particular due to the combination of spiralling inner-city housing costs and changes to the benefit system. Whilst it’s true that some of this movement is to areas dominated by other visible minority (and Labour-voting) groups, a significant proportion is into what were previously traditional white middle-class seats such as Streatham, Harrow East and Harrow West. These seats experienced some of the highest net swings to Labour between 1992 and 2010.

The centrifugal movement of visible minority populations to the outer suburbs takes two forms. In some seats, Streatham, Hendon and Finchley being examples, residents in formerly white middle class neighbourhoods have been replaced by lower income groups, many of whom are members of visible minority communities. In other seats, such as Harrow East and Harrow West, upwardly mobile members of visible minority communities have collectively colonised and replaced a white population economically less successful than they are. Some of these trends have been described in Issue 1 of Demos Quarterly by Eric Kaufmann.

If the type of middle class minority voter such as the ones who have colonised the Harrow seats were to persist in voting Labour, it would reinforce Lord Ashcroft’s gloomy message to the Tories and increase Mr Miliband’s chances of taking some formerly marginal seats which, on the basis of white British preferences alone, he would have little expectation of winning. If on the other hand upwardly mobile members of minority communities adopt the party preferences of the people whose homes they have taken over the horizon would look rather brighter for Mr Cameron.

History suggests that the former scenario – a persistence of preference for the left – is more likely. A superficial study of historical voting patterns amongst ethno-cultural groups suggests that, even if these patterns do not start out as a conscious cultural preference, they can come to resemble one. For over a century both Jewish voters in the USA and Irish voters in the UK have persistently leant leftwards notwithstanding their upward social mobility. More recently, unpublished work by Bobby Duffy at Ipsos MORI has demonstrated clearly that London’s electorate has been effectively colour coded, in Mayoral elections at least: in 2012, whites, including those in ‘working class’ postcodes went heavily for Boris Johnson, whilst non-whites tilted to Ken Livingstone.

So what should we expect from minority voters in 2015? Is there the beginning of a drift away from Labour; or will the historic split persist?


To find out we examined the postcodes of 2,468 members of minority ethnic communities surveyed by YouGov between 1st October and 29th November 2013. The sample is not exactly representative of the UK’s minority population – it will probably, for example, contain an untypically high number of better-educated individuals. Our results, therefore, must be considered indicative rather than definitive.

We were able to determine the ethno-cultural composition of the postcodes in which respondents live using Origins, a software package which uses algorithms derived from a database of over 1.2 billion individual records and over 3 million different personal and family names to infer a person’s ethno-cultural origins.

Broadly speaking these algorithms simply codify historical patterns of settlement and marriage, for example associating names starting with ‘Mac’ with a Scottish origin, or ‘Singh’ with a Sikh heritage. In some cases the associations result from more subtle factors; for example, whilst most people of African Caribbean descent have ‘English’ or ‘Scottish’ surnames, the fact that most of the population of the islands were compelled to take the names of plantation owners – a very small pool of names indeed – allows our programmes to ascribe a high probability of Caribbean associations to a name such as ‘Brathwaite’, for example. The size of our database means that our level of accuracy is at or above 95 per cent in any given list of names.

Our findings derived from YouGov’s sample are shown below.


YouGov asked respondents if they had any particular party identification. Broadly speaking, compared to white British voters, Figure 1 shows the 2,468 visible ethnic minority voters have a strong antipathy to the Conservatives and a moderate disposition to support Labour:

Figure 1: Party identification, Non-white British and White British

The 17 per cent who identify with the Conservatives is slightly higher than Lord Ashcroft’s finding of 16 per cent. The main beneficiary of this anti-Tory tilt is clearly the Labour Party. This is reflected both in the intentions of those intending to vote today (as of November 2013) and in reported voting preferences in May 2010:

Figure 2: Voting intention November 2013 compared to 2010 General Election, visible minority voters

As with other voters, the major change since May 2010 has been a dramatic fall in support for the Liberal Democrats, with most of their supporters going to Labour or to smaller parties.

However, notwithstanding this long-term historical pattern, a rather more significant change amongst visible minorities over recent years has been a gradual rise in Conservative support. Compare the picture above with research conducted following the 2005 General Election by Roger Mortimore and Kully Kaur-Ballagan of Ipsos MORI. Whilst the figures are not directly comparable with those in our YouGov sample, the Labour dominance is familiar. However, it is evident that the share attracted by the Conservatives has risen since 2005, largely at the expense of the Liberal Democrats:

Table 1: Voting Shares, Visible Minorities, May 2005

All African Caribbean African Indian Bangladeshi Pakistani Mixed/Other
Conservative 10 3 2 11 9 11 13
Labour 58 80 79 56 41 50 47
Liberal Democrat 16 5 11 14 16 25 22
Other 4 2 1 1 21 8 5
Refused/DK 12 11 7 17 13 7 12

Whatever the cause, there clearly is some movement of visible minority voters towards the Conservatives.

We analysed these figures further to try to discover whether a spatial demographic movement of electors from different ethnic minorities accounted in any way for this rise in the Conservative share. Were their voting intentions affected by whether they lived in a postcode with a high representation of ethnic minority electors; and did the effect depend on the proportion of other electors in the postcode where they lived who belonged to their own community?

In the figure below we compare the voting intention of members of visible minority communities according to whether they live in predominantly white British – ‘white’ neighbourhoods; or in areas with a significant ethnic minority population – ‘clusters’. We have defined a ‘cluster’ as an area in which 31 per cent or more of those eligible to vote are from visible minorities.

Figure 3: Differences in voting intention between non-white British neighbourhoods and white British neighbourhoods

Whilst the concentration of ethno-cultural minorities seems to have very little effect on the Conservative or Liberal Democrat shares of minority votes, it has a dramatic impact on Labour’s fortunes. Outside ‘cluster’ neighbourhoods Labour’s share of the visible minority vote falls by a fifth.

However this isn’t necessarily good news for the Labour’s main opponents. Looking at the detailed breakdown of ‘Other’ parties, it appears that the single largest beneficiaries of the Labour defection are UKIP and the Green Party, each of whom adds three percentage points in less mixed areas.

Overall, based on the analysis above, there seems little evidence to support the hypothesis that Conservative-voting visible minority electors change their preferences according to whether they live in ethno-cultural clusters or not. It may be that members of visible minority populations who live in mixed neighbourhoods tend to be members of communities who are already sympathetic to the Conservative party whilst those who live in multi-ethnic clusters are traditionally more strongly aligned to Labour. In short the predisposition to vote Conservative may well precede the move to a more mixed area, rather than follow it.

The same does not appear to be true of Labour voters. It seems clear that a significant proportion of visible minority electors who vote Labour in ‘cluster’ areas are ready to desert when they move to white areas. We can therefore infer that whilst there is a cluster effect in minority voting of some significance, at present, the principal impact is on the Labour vote. This cluster effect could clearly have some impact on the outcome of the contest in 2015. However, the detailed picture on the ground, will matter enormously.

We are not, in this paper, examining individual constituencies, but we believe that it may be useful to develop a general picture of the way that different groups of minority voters might behave.

The Cluster Effect: Race or Class?

One central question is whether the cluster effect is a result of social mobility, or of ethno-cultural factors. If all minorities who move from clusters to mixed areas behave in roughly the same fashion then we can conclude that any cluster effect is primarily a socio-economic phenomenon. If, on the other hand, there are significant variations in the change of voting intentions between different minority groups then a more likely explanation of the cluster effect would be rooted in ethno-cultural factors.

It is generally supposed that the change in the Labour vote illustrated in Figure 3 above probably reflects a change in socio-economic status associated with movement to a more middle-class community. If Labour polls less well among minority groups who are more likely to live in white neighbourhoods, it would be reasonable to infer that any weakening of traditional party alignments must be associated with social mobility. So does a comparison of groups by degree of clustering show a clear correspondence with voting preferences?

Table 2: Concentration of people by ethnic background in multi ethnic cluster neighbourhoods

Ethnic background n Proportion (%) of group in multi-ethnic clusters Proportion (%) intending to vote Conservative
Any other Asian background 180 13.9 37.2
Chinese 232 10.8 30.2
Indian 471 29.1 22.9
Average 18.1 18.0
Bangladeshi 95 38.9 11.6
Black African 235 19.6 11.1
Pakistani 203 36.0 9.9
Black Caribbean 180 21.1 8.3

The answer is ‘not necessarily’. But though there is not an exact correspondence, Table 2 does suggest that the visible minority groups which are least likely to live in multi-ethnic cluster neighbourhoods are the ones most likely to support the Conservative party.

Does this lead is to the conclusion that social mobility factors account for the fall off in Labour support in more mixed areas? No, because further analysis suggests that the size of the ‘cluster effect’ may itself vary between different visible minority groups.

In Table 3 we compare the Labour lead over the Conservatives among members of minorities according to the dominant minority in the postcode. It is clear that Labour polls much better among minorities living in neighbourhoods dominated by South Asian Muslim and Black African populations than in neighbourhoods dominated by Hindus.

Table 3: Voting intention according to dominant minority in the neighbourhood

Type of neighbourhood n Conservative Labour Labour lead
All neighbourhoods 2648 24.1 50.2 26.1
White 2167 24.4 48.1 23.7
Non white 481 22.9 60.1 37.1
Bangladeshi 30 3.7 74.1 70.4
Black African 32 11.1 70.4 59.3
Pakistani 140 15.1 69.9 54.8
Sikh 61 24.4 57.8 33.3
Hindu 115 35.3 48.2 12.9

Sample sizes limit some more detailed analyses but it is possible to compare the voting preferences of different groups within the South Asian community according to whether or not they live in a multi-ethnic cluster. This reveals the influence of a factor associated with, but not identical to, ethnic background: religion.

Table 4 confirms this. Labour polls very much less well among people with a Indian background who live in white areas (where its lead is cut from 30.9 per cent to 13.3 per cent) whereas among Muslims of Bangladeshi or Pakistani origin voting intentions are almost entirely correlated to religious background and hardly at all to the ethnic composition of their immediate neighbourhood.

Table 4: Voting intention among South Asian communities according to whether or not they live in cluster neighbourhoods

South Asians n Conservative Labour Labour lead
All in cluster areas 247 22.5 62.9 40.4
Indian 137 27.7 58.5 30.9
Pakistani 73 16.4 65.5 49.1
Bangladeshi 37 17.2 72.4 55.2
All in white areas 522 25.1 52.4 27.3
Indian 334 32.0 45.3 13.3
Pakistani 130 11.5 65.6 54.2
Bangladeshi 58 14.0 65.1 51.2
Total 769 24.3 55.7 31.4

Conclusions: which seats are at risk?

Our analysis reveals that in a tight election, political parties would be unwise to ignore the impact of patterns of voting amongst visible minority electors.

We should emphasise once again that our sample is not an exact representation of the ethnic minority population of the UK; and that some sub-sets we have identified are relatively small. Nonetheless we believe that our study of a sample of YouGov respondents offers some unmistakable indications that the behaviour of visible minorities could be one of the factors that tips the outcome one way or the other; and our conclusions challenge some commonly held beliefs amongst political journalists and activists.

First, our analysis shows that for visible minority voters, their ethno-cultural origin is a better predictor of voting intention than any other factor, including socio-economic background. Whilst most voters will assert that they do not vote on racial lines, for all practical purposes the behaviour of visible minority voters belies this assertion.

Second, the common supposition that over time the political preferences of visible minorities become more ‘integrated’ and less distinguishable from the average is not wholly borne out by the evidence. Over the past decade there has been an increase in the Conservative vote, but principally at the expense of the Liberal Democrats rather than Labour.

Third, the drift towards the Conservatives is clearly associated with demographic shifts among some visible minority groups from the original areas of settlement to more mixed areas.

Fourth, amongst those visible minority voters who are most likely to change their behaviour, the single largest movement is amongst Labour deserters who have moved out of their areas of original settlement.

Fifth, the largest group of Labour deserters are of Indian origin, and of the Hindu religion. In practice, the most telling drivers for the ‘cluster effect’ appear to be religious and cultural rather than socio-economic or racial.

As interesting as these conclusions might be in general terms, they are only significant if the movement we have described may influence the electoral outcome. A rigorous examination of marginal seats would need to be the subject of a further study. However we have taken an initial look at the 100 most marginal Parliamentary seats and developed a typology to show which kinds of seats may be at risk from the cluster effect:

  • Seats with small Conservative majorities and large numbers of Indian voters. These seats should be vulnerable to Labour, but may be less likely to change hands because of Labour desertion by upwardly mobile Indian-heritage – and Tory-voting – electors. Examples are Hendon and Wolverhampton South West.
  • Seats with small Conservative majorities and large numbers of non-Indian voters. These seats should be vulnerable to Labour, and may indeed be more likely to change hands because of the inward migration of upwardly mobile black or Pakistani – and Labour-voting – electors. Good examples are Enfield North and Dewsbury.
  • Seats sensitive to differential turnout. These are seats where the level of turnout by different ethnic groups in the constituencies may be decisive. The three-way hyper-marginal of Hampstead and Kilburn is a spectacular example.

Elsewhere, outcomes in other – apparently less marginal – seats, such as Harrow East and Harrow West, Derby North and Tooting will very likely turn on other kinds of behaviour amongst minority voters. This will, we hope, be the subject of further studies.