Expressive voting and two-dimensional political competition: an application to law and order policy by New Labour in the UK

There has been much debate regarding the electoral strategy adopted by New Labour in the lead-up to and then during their time in government. This paper addresses the issue from the perspective of left/right and liberal/authoritarian considerations by examining data on individual attitudes from the British Social Attitudes survey between 1986 and 2009. The analysis indicates that New Labour’s move towards the right on economic and public policy was the main driver towards attracting new centrist voters and could thus be labelled ‘broadly’ populist. The move towards a tougher stance on law and order was more ‘narrowly’ populist in that it was used more to minimise the reduction in support from Labour’s traditional base on the left than to attract new votes. The evidence presented provides support for an expressive theory of voting in that law and order policy was arguably used to counter alienation amongst traditional, left-wing Labour supporters.


Introduction
A distinctive feature of the New Labour governments that held power in the UK between 1997 and 2010 was the focus on a tougher approach to law and order. They signalled their intent 5 years before coming to power with the famous slogan 'tough on crime, tough on the causes of crime'. Once in power, New Labour introduced the 1998 Crime and Disorder Act and their taste for tougher policies could be seen across a range of policy measures; tougher sentencing, the introduction of antisocial behaviour orders (ASBOs), increased surveillance and, notably after 9/11, anti-terror legislation. In this paper, we do not aim to discuss the merits of this approach in terms of its efficacy or any normative judgement regarding the trade-off between freedom and security. Rather, we wish to explore the extent to which this approach was part of Labour's electoral strategy. 1 A common accusation was that Labour's approach to law and order was nakedly populist. 2 Law and order could be viewed as being bundled with the party's shift from the left in terms of economic and public policy as a general package of policies aimed at appealing to the middle ground of British public opinion. 3 However, Labour's approach to law and order is more complex than for economic and public policy. It is not immediately clear that the approach taken by Labour was actually in the centre ground. For example, while very few Conservatives would find themselves arguing that the Labour approach to economics and public policy was too right-wing, a strand of Conservatives (most notably David Davis and Kenneth Clarke) argued that the Labour approach to law and order was too authoritarian. 4 The criminologist Ian Loader in a Guardian article in 2008 also questioned the view that tough attitudes to law and order dominated the centre ground, 'there is also evidence that the majority of people have little experience of crime, rarely think about it and, when prompted to do so, express ambivalent feelings about the proper response to it '. 5 In response it could be argued that there is a difference between the actuality of a policy and its perception. For Labour to be perceived in the centre ground of law and order they may actually have to be tougher than the Conservatives to draw 1 Labour and New Labour should be treated as synonymous labels for the party and we will only use the latter term if we wish to emphasise policy changes that were particularly associated with Tony Blair's leadership of the party from 1994 on. 2 See for example, Grayling (2009) and Timothy Garton Ash, 'Liberty in Britain is facing death by a thousand cuts. We can fight back ', The Guardian, 19 February, 2009. For an analysis of the origins of penal populism, see Garland (2000). 3 In terms of moving economic policy to the centre, before coming to power Tony Blair successfully fought to revise Clause IV in the Labour party constitution which called for state ownership of the means of production, distribution and exchange. On coming to power in 1998, the Bank of England was made independent and income tax was not increased. In subsequent years much greater sympathy was given to the use of market forces in the provision of key public services such as education and health. 4 David Davis, the then shadow home secretary, resigned his seat in 2008 and fought for re-election to draw attention to what he perceived as the erosion in civil liberties in Britain. On the election of the Conservative Liberal Democrat coalition in May 2010 Kenneth Clarke became Secretary of State for Justice and argued for a reduction in prison numbers. 5 'Straw's embrace of penal excess ignores public will', The Guardian, 28 October, 2008. attention to how they have changed. This seems somewhat borne out by the fact that, according to MORI (Market and Opinion Research International) polls, as late as the 2005 election the Conservatives had an issue advantage over Labour on crime even though Labour were dominant in most other areas.
Even if a tough approach to law and order may possess more general appeal than a soft approach, there may be an alternative more powerful explanation for Labour's approach which focuses more narrowly on popularity within their traditional base rather than appealing to the public at large, who were clearly the target when moving to the centre ground on the dominant issue of economic and public policy. This will rest on the claim that law and order could be used as a tool to retain core support in the event that they lose the battle on the centre-ground of politics which is concerned primarily with economics and public policy.
The idea here is that traditional Labour support is left-wing but that this support may hold highly differing views along an authoritarian/liberal axis. As Labour necessarily moved to the centre ground on the left-right dimension in order to be electorally competitive they risked alienating their traditional supporters. From a purely electoral perspective, this would not matter if the sacrifice of the base led to victory in the elections due to the greater weight of newly acquired floating voters motivated primarily by Labour's shift to the centre on left-right issues. The longterm risk is that as political battle loses its ideological edge, Labour would lose centrist voters to the Conservatives and also lose their traditional support by moving too far from their left-wing preferences.
Therefore, an electoral strategy for Labour was to move to the centre-ground on the key left-right dimension, but ask whether there was something they could do on the authoritarian-liberal dimension that would keep them close enough to their traditional base so that it keeps voting for them. The proof of narrow populism in this case would be that more votes are retained in taking a tough rather than a more liberal stance on law and order. That is, there would be more votes to be retained by appealing to authoritarian left-wingers than liberal left-wingers. Indeed, it is claimed that law and order only became an electoral issue because Labour under Tony Blair made it into one. 6 We would like to consider the idea that the tough approach was not developed so much to appeal to the general public, but to what they considered a key constituency within their traditional core support who in other respects may have felt abandoned by the other policies pursued by the Labour party. Our hypothesis is that if the approach of Labour was purely electorally motivated it was two-pronged; the primary strategy was to move to the right to attract floating voters and a secondary strategy was to move towards authoritarianism to hang on to a specific constituency of their traditional left-wing base.
We analyse data from the British Social Attitudes survey to test our hypothesis that if the New Labour approach to law and order was populist, it was populist in a different sense to their economic and public policies in that it was not a policy 6 'Anti-Social Behaviour: It's Back', The Economist, October 1st, 2009. Morgan (2002, 2007) provide a historical context for this claim. They point out that law and order only started to appear in party manifestos in the 1960s and it was only with Thatcher that the Conservatives used it as a political stick with which to beat Labour. Labour did not contest an area which they correctly viewed as a weakness until the 1990s when Blair tried to wrestle law and order policy from the Conservatives.
Expressive voting and two-dimensional political… 81 primarily aimed at winning new voters in the centre ground of politics, but rather retaining as many votes as possible on the left. We label the former 'broad' populism and the latter 'narrow' populism.
The empirical findings bear out this argument, however, we are aware that we cannot rule out the possibility that the policy on law and order was conducted for reasons unrelated to popularity. First, an alternative explanation is that they succumbed to the eternal temptation to strengthen the power of government regardless of the electoral consequences. A second explanation is that they were statesmanlike and took a tougher stance on law and order because they thought it was the right thing to do regardless of its electoral consequences. 7 This would fit with the common retort to the left-liberal constituency that it easy to hold liberal views on law and order (and to dismiss those who hold tough views as ill informed) because they come in the main from a largely privileged background and are thus unaffected by crime. This point is made repeatedly by Tony Blair in his autobiography (Blair 2010). Populism and good policy are not by definition in conflict, although much of the commentary on the Labour approach suggests that the alleged populism was, in fact, in serious conflict with good policy on law and order. After discussing related literature, we discuss the theoretical background. These two sections will lay out the idea that our theory concerning Labour's approach to law and order is also an expressive account of political support (or nonsupport). We then introduce the data to be analysed and consider some empirical evidence before making some concluding comments.

Related literature
There are a small number of surveys of New Labour's approach to law and order. Morgan (2002, 2007) discuss Labour's law and order policy within a broad historical context. Chalmers and Leverick (2013) analyse the enormous expansion of new criminal offences over the period of Labour government from 1997 to 2010. Saward (2006) provides a discussion of civil liberties in the post 9/11 world and covers in detail anti-terror legislation and ASBOs. Beloff (2007) discusses the relationship between Tony Blair and the law and judiciary and provides another example of the accusation of populism when he writes, 'he displayed no particular appetite for engaging with legal issues, apart from using populist philosophy' (p. 292). In the same book, Newburn and Reiner (2007) discuss crime and penal policy and with reference to the 1998 Crime and Disorder Act argue that 'Blatcherism was Butskellism in reverse' (p. 324). They catalogue the sheer quantity and speed of legislation and increases in incarceration and conclude that Majoritarian electoral systems do require centre-left parties to capture a substantial section of the middle-class vote. Arguably they must allay middleearners' fears about redistribution at their expense by tough leadership espousing middle-class values against the left of the party. Sun-worshipping law and order policies and tight central control follow from this logic. (p. 339-340).
This quotation is at odds with the one by Ian Loader used earlier which suggests that the majority view on the issue is not so clear, so it would be unclear why a tough stance on law and order would compensate for fears about a future lurch to the left by Labour on economic and public policy. Also, and an even stronger point, given the effort by New Labour in the years before coming to power and the policies they initiated on taking power, relates to how realistic is the idea that the middle-class would have feared a lurch to the left. Our thesis is that the tougher stance is taken to allay the sense of alienation felt by those with left-authoritarian preferences due to Labour's shift towards the middle ground on economic and public policy.
The idea of a voter feeling alienated and not voting, or 'wasting' a vote on a minor party is commonplace, but the logic does not follow easily from standard models of voting. Probably the most famous model of voting stems from Downs' (1957) depiction of political competition. Voters are considered as having heterogeneous preferences along issue dimensions, they vote instrumentally (as an indirect means to achieve favourable outcomes) and parties compete to win power by choosing a political position in the issue space. When there is only one dimension, usually thought of as left-right, the prediction is that if there are two parties and they are purely concerned with winning they will converge at the preference of the median voter. Even if we relax this extreme prediction, convergence towards the middle ground is likely. But in this model, as Brennan and Hamlin (1998) point out, the citizens with greatest incentive to vote are those on the extremes because the stakes for them are higher than for those with preferences close to where the parties have set their policies. In the Downs environment, alienation should not play a role. If some do not vote it must be because their preferences lie somewhere between the largely converged positions of the parties and they are indifferent to the outcome.
An alternative instrumental theory of voting originates with Stokes (1963) and removes the idea of heterogeneity of preferences over policy in the issue space. Instead, this theory considers voting to be determined by valence issues. Voters agree on the policy, such as a well-managed economy, and they will vote for whichever party they believe will do best at delivering that. There is a literature that argues that UK elections have shifted from being Downsian to being valence dominated (Sanders and Brynin 1999;Whiteley et al. 2005;Green 2007;Green and Hobolt 2008).
To a large extent, the Downsian and valence models are not competing models of voting, but rather that if the Downsian prediction of policy convergence is borne out we would expect valence to play a large role in determining votes because there is less disagreement on the content of policy between the main parties. The public will then vote for which party they believe will be best in delivering the agreed policy. Again, there would appear to be no role for alienation. If valence issues are dominant then political competition is dominated by issues on which voters agree, Expressive voting and two-dimensional political… 83 so citizens who are more extreme should have no less incentive to vote than for more centrist citizens. 8 The third model of voting (expressive voting) makes sense of the idea of alienated voters. There is no single agreed theory of expressive voting, but the logic of expressive voting is well understood, see Hamlin and Jennings (2011) for a definition and a survey of the literature. As the number of voters becomes large, the probability of an individual being decisive in determining the outcome of elections becomes very small. As a result, the indirect instrumental benefit of choosing one party over another becomes very small. This means that if there is a direct expressive benefit associated with voting, then this benefit must weigh more heavily in voting calculus than other domains of decision-making where the individual is more likely to be decisive, for example in market choice. Brennan and Hamlin (1998) use this insight to provide an explanation for alienation. They agree largely with a purely instrumental Downsian model of voting that there will be a centripetal force pulling political positions towards the centre. However, this will lead them to occupy political positions that lie some distance from citizens towards the extremes. These citizens will receive lower expressive benefits from voting for centrist parties and given the very small instrumental benefit of voting may decide not to vote at all, or vote for a party located close to them although it may have no realistic chance of gaining power. 9 If this model of expressive voting is accepted, it would follow that as Labour moved to the right they risked alienating traditional left-wing voters. The choice of policy on the liberalauthoritarian axis could then be selected to retain the greatest amount of this potentially alienated support and it could be for this, more narrowly populist, reason that they choose to become more authoritarian.
Interestingly, Brennan (2008) has written on crime and punishment from an expressive perspective. He argues that the very high levels of incarceration that has been witnessed in the US and Australia bear witness to an expressive explanation for electoral behaviour. Calculations of whether the cost of incarceration is justified by the benefits it provides are not important for voters because they are so unlikely to determine the outcome of elections. As a result, they choose a policy that expressively appeals to them. Brennan argues that increased incarceration is expressively appealing for many voters.
In a detailed study of the British National Party, Ford and Goodwin (2010) argue that the BNP mainly attracted ex-Labour voters. Our theory is not that a tough stance on law and order increases support for Labour, but rather that it would stem the loss of support from a constituency such as this, as they are likely to support a tough stance on law and order. Later we will argue that the strategy worked reasonably well until 2001. Evans and Chzhen (2013) find that immigration became an increasingly salient issue and that this plays a key role in defection from Labour from 2005 to 2010. Given that those that hold strongly anti-immigrant views are more likely to support a tough stance on law and order (Chirumbolo et al. 2004), the salience of this issue may well have begun to undermine law and order policy by the mid-2000s. 10 Throughout this paper, we take the position that the most critical issue determining the voting decision is economic policy and its execution broadly defined. However, an eyebrow might be raised when one considers how an issue such as crime seems to consistently rank so highly in polls as to 'the most important issue'. Johns (2010) demonstrates that we should not leap to conclude from these polls that issues considered most important are the issues that actually determine voting. Johns found that 'there was little evidence of issue voting among those naming asylum and crime' (p. 156). Therefore, if the approach of Labour was purely electorally motivated we argue that it was two-pronged; the primary strategy was to move to the right to attract floating voters and a secondary strategy was to move towards authoritarianism to hang on to a specific constituency of their traditional left-wing base.

Theoretical background
We depict a 2-dimensional setting with the horizontal axis being left-right and the vertical axis liberal-authoritarian with increasing authoritarianism as we move up the vertical axis as shown in Fig. 1.
An individual x has preferences in two dimensions given by x 1 ; x 2 ð Þ where x 1 depicts left-right and x 2 depicts liberal-authoritarian preferences. Their voting choice will be determined, first, by which party is closest to them and second (capturing the possibility of expressive alienation) that the closest party is still within a distance c of x 1 ; x 2 ð Þ. If both parties are further than c, x will prefer to abstain or vote for a minor party. The distance from x's preference to a policy profile y 1 ; y 2 ð Þ could be depicted as the simple Euclidean distance ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi . This would mean that the indifference curves are circles. However, we are assuming that left-right is the dominant policy dimension. As such we should consider weighted Euclidean distance given by ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi where 0 A 1. This means that the indifference curves are ellipses, which given that the left/right dimension is on the horizontal axis the ellipses are crunched in from the sides, as shown in Fig. 1. This figure indicates that 10 One of the main groups that voted for the UK to leave the European Union in the June 2016 referendum was traditional working-class communities. One might argue that as immigration surpassed law and order in salience by the mid-2000s, tough law and order policy could be viewed as the last issue on which the modern Labour party connected with their traditional base. When the party leadership in 2016 recommended voting for remain in the referendum, this constituency no longer listened and voted leave instead.
Expressive voting and two-dimensional political… 85 a small movement away from a citizen's ideal position on the left-right axis would require a larger movement on liberal-authoritarian axis to compensate. So in order to win a vote, two conditions need to be met. First, the distance must be less than the rival party. Second, the distance must be within distance c. These are given by ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi In a model where there is no expressive voting and all voters vote in two dimensions where political competition is dominated by two strategic parties there would not be an equilibrium, but we would expect parties to offer roughly similar policies somewhere in the region of the median of the two axes. If the dominant policy dimension is very dominant, reflected in a low value of A, then we might expect the location of policy on the liberal-authoritarian axis to be largely irrelevant so politics would be dominated by the left-right axis. The reason is that left-wing voters will keep voting for the more left-wing party even if it is not very left-wing at all. This fits with the Brennan and Hamlin analysis-in a purely instrumental model those with the strongest preferences on left and right have the most incentive to vote even if the left and right parties are moderate. In an expressive model, this changes because these voters will drop out/vote elsewhere due to alienation. In a purely instrumental approach, the conclusion may be that policy on a minor dimension is only marginally significant. In an expressive approach, policy on the minor dimension could become much more important because it may be a means by which voters can be retained.
In terms of conditions (1) and (2), in a purely instrumental model dominated by two main parties condition (2) is irrelevant. So long as Labour is to the left of the Conservatives (which can easily be understood as rooted in the need to satisfy, to at least some extent, the traditional core support within the two parties) left-wing voters will vote Labour. In an expressive model, the distance from the party position should not be too great in terms of (2) to justify voting Labour. So although the leftwing voter would prefer Labour to the Conservatives they will either not vote or vote for a minor party. It is within the structure of this model that it may make sense to manipulate policy (by setting y 2 very close to x 2 ) on law and order to ensure that policy distance is kept below c. If the charge of narrow populism is correct, then we should find evidence that shows this approach retained a larger proportion of leftwing support than was lost in the further alienation of left-wing voters who were also liberal.
New Labour's tough approach to law and order providing evidence for expressive voting thus rests on the distance Labour had moved to the economic right from its traditional base on the economic left. An instrumental account of voting would predict continued left-wing voting for Labour as they more closely represent left-wing interests than the Conservatives (although the distance between Labour and the Conservatives has closed). Under this theory, so long as Labour is to the left of the Conservatives then that is all Labour needs to do to retain all of their left-wing support. An expressive theory would predict instead, that votes will leak from the left for Labour as these citizens feel alienated by New Labour's economic policy. However, given that expressive voting is based on a feeling of identity with a party, Labour realised that law and order policy could be used to maintain sufficient identity for some of these voters who would otherwise be lost to non-voting or voting for minor parties. As we shall see there are more authoritarian left-wingers than liberal left-wingers and for this reason we argue that policy was set so as to expressively appeal to the former and minimise the loss of support on the left for Labour.

Data
The data source used in this paper is the British Social Attitudes (BSA) survey. This is a representative survey of adults aged 18 and over living in private households in Great Britain, which contains information on a wide range of attitudes. 11 Areas north of the Caledonian Canal are excluded because of their dispersed population, whilst a separate survey is carried out in Northern Ireland. The survey has been carried out annually since 1983, apart from in 1988 and 1992, with some variations in the sample size over time. We analyse information from surveys that have taken place since 1986 since these contain information on the left-right and liberalauthoritarian variables, as well as on party identification.
In order to classify different political views, we construct a categorical variable that combines information from the five left-right variables and six liberalauthoritarian variables that have been included in the BSA survey. These have been asked in each survey since 1986. 12 The items used to construct the left-right variable are the following: • Government should redistribute income from the better off to those who are less well off. • Big business benefits owners at the expense of workers.
• Ordinary people do not get their fair share of the nation's wealth.
• There is one law for the rich and one for the poor.
• Management will always try to get the better of employees if it gets the chance.
Respondents are asked to choose how strongly they agreed or disagreed with each of the statements using a 5-point Likert scale. The left-right variable was then created by grouping (by taking an average of) the responses to the items. Therefore, if a person strongly agreed with each of the above statements then they were assigned a value of 1 (left) and if they strongly disagreed with each of the statements they were assigned a value of 5 (right) on the left-right scale. Further details on the construction of this measure can be found in Heath et al. (1994) and Evans et al. (1996). Similarly, the items used to construct the libertarian-authoritarian scale (which we relabel as liberal-authoritarian) are the following: • Young people today don't have enough respect for traditional British values.
• For some crimes, the death penalty is the most appropriate sentence.
• Schools should teach children to obey authority.
• The law should always be obeyed even if a particular law is wrong.
• Censorship of films and magazines is necessary to uphold moral standards.
• People who break the law should be given stiffer sentences.
Therefore a value of 1 on this scale would indicate an extreme liberal and a value of 5 an extreme authoritarian. Again see Heath et al. (1994) and Evans et al. (1996) for a detailed discussion of this measure.
The political attitudes categories that we use in the empirical analysis have been created by combining information from both the left-right and liberal-authoritarian variables. In particular, the categories are a combination of three groups of left-right indicators (left, centre and right) and three groups of liberal-authoritarian indicators (liberal, moderate and authoritarian). These two sets of groups have been defined in the following way: where LR i is the value on the left-right scale for respondent i.
Liberal: 1 LA i \3:5 Moderate: 3:5 LA i 4 Authoritarian: 4\LA i 5 where LA i is the value on the liberal-authoritarian variable for respondent i.
The upper and lower boundaries for each group were selected to ensure an adequate distribution across the three groups over time, especially since neither the left-right or the liberal-authoritarian variable is truly continuous since they have been constructed using either five or six integers and as a result there is some clustering around particular values. A single variable containing nine categories was then created using both sets of groups to summarise a person's political outlook. These categories are shown in the first column of Table 1.
The analysis that follows uses information from annual BSA surveys that have been pooled together from 1986 until 2009, with five time periods identified during this interval. These are 1986-1991, 1993-1997, 1998-2001, 2002-2005 and 2006-2009. These have been chosen to achieve reasonable sample sizes in each period but also to roughly accord with recent periods in Labour party history: opposition, lead-up to government, first, second and third administrations. The number of observations in these time periods varies according to the total number of respondents and on whether the attitudinal questions were asked to all of the individuals that were interviewed or just a proportion of them who were interviewed in each year. In total, the full sample includes over 57,000 observations, with the highest number appearing in the final time period and the lowest number in the first. Table 1 presents the distribution of political attitudes across the nine categories over the five time periods. It shows that there has been a rightwards shift in political  1986-1991 1993-1997 1998-2001 2002-2005 2006-2009 1986-2009 Left-liberal 6. attitudes since the late-1980s. As a result, attitudes have become more concentrated in the middle, with the percentage accounted for by the centre-authoritarian category increasing from 12. 2 % in 1986-1991 to 16.3 % in 1998-2001, and remaining around this percentage in the final period. A slightly smaller percentagepoint increase was observed in the centre-moderate category over the five periods.
In contrast, the percentage in the left-liberal category more than halved in size and falls of 2-4 % points were also observed in the left-moderate and left-authoritarian categories. There were only relatively small fluctuations in the percentage accounted for by the three right categories. The largest changes were observed for the right-authoritarian category, which fell by almost 3 % points over the first two periods before increasing slightly over the final three periods.

Analysis of changing political attitudes
In this section, the distribution of political categories presented in Table 1 are examined with respect to the political party that the individual identifies themselves with. Party identification occurs at three different levels: individuals reporting that they support a certain party; those stating that they are sympathetic to a party and finally people who say that they are closest to a party. In our analysis, the three groups identifying themselves with a particular party are analysed collectively. Table 2 reports the percentage of Labour party identifiers within each political category over the five periods. This table also indicates whether the percentage of Labour identifiers in each category during periods 2-5 is significantly different to the first period. The table shows that support amongst what might be considered to be traditional core Labour voters (i.e. those with left-leaning attitudes) declined after the mid-1990s. In particular, the percentage of Labour party identifiers in the left-liberal category fell from just over 70 % in 1986-1991 and 1993-1997 to under 1986-1991 1993-1997 1998-2001 2002-2005 2006-2009 1986-2009  Right-liberal 14.7 24.0*** 33.6*** 32.5*** 27.9*** 27.7 Right-moderate 5.9 13.6*** 27.1*** 24.6*** 20.5*** 19.1 Right-authoritarian 3.0 9.1*** 20.3*** 19.4*** 16.6*** 13.6 The stars indicate a significance difference in the percentage of Labour identifiers within each political category relative to the 1986-1991 period. * Significance at the 10 % level (using a two tailed test); ** at the 5 % level; *** at the 1 % level 48 % in 2006-2009. The largest falls occurred between periods 2 and 3 and periods 4 and 5. The percentage of Labour identifiers in the left-moderate and the leftauthoritarian categories also declined by around 21 and 17 % points respectively between the second and fifth periods, with the highest falls observed between periods 3 and 4 and periods 4 and 5, as also indicated by the tests of significance. The loss of support was least for the most populated (authoritarian) category on the left. This is in line with our hypothesis. However, the loss of support across the whole sample period misses the more nuanced aspect to the theory. By 2001, Labour had established itself as a centre party, tough on law and order and this is reflected by the big fall in left-liberal support from period 1 to 3 of 9 % points relative to the fall of 3 % points for left-authoritarian support and, indeed, the increase in support of about 1 % point for left-moderates. We would argue that the big falls in support after 2001 were related to the general fall in support for Labour which could be seen across all nine categories and, as mentioned earlier, the increasing salience of immigration as identified by Evans and Chzhen (2013). The period up to 2001 captures the two trade-offs we wish to investigate. First, there is the loss of left support in order to make gains in the centre and on the right. Second, minimising the loss from the most populated left (authoritarian) category by leaking more support in the least populated left (liberal) category. The percentage of Labour party identifiers in each of the three centre categories peaks in the middle period (1998)(1999)(2000)(2001). The largest increase up to this point was seen in the centre-authoritarian category, since 42 % of this category were Labour party identifiers in 1998-2001, compared to 27 % in 1986-1991. This 15 % point change compares to increases of 6 and 11 % points in the other two centre categories. The majority of the increase in Labour identifiers within the three centre categories occurred between the first and second period. Fairly similar percentage point declines were seen for all three centre categories between the third and fifth periods, with falls of around 14, 11 and 11 % points being observed. This would appear to provide confirmation of the importance of moving economic/public policy to the centre ground in the early years of new Labour.
The incidence of Labour party identifiers within the three right categories showed a similar pattern. This is particularly the case for the increase in Labour identifiers up to the third period, with the decline after this being less pronounced than that observed for the three centre categories. For example, the percentage of Labour party identifiers in the right-authoritarian category increased from 3 % in 1986-1991 to 20 % in 1998-2001 and still remained at an appreciable 17 % in 2006-2009. The importance of the centre-moderate and centre-authoritarian categories is further demonstrated in Table 3. It indicates that these two categories accounted for 34 % of all Labour identifiers in the first period before rising to around 40 % in the second period and 43 % of all Labour identifiers in the third period. It further increased to 44.6 and 46.5 % in the final two periods. In contrast, the percentage of Labour identifiers accounted for by the three left categories fell across the five periods, from 12 to 4 % for left-liberals and from 15 to 9 % for left-authoritarians.
The key point here is that support in the left-authoritarian category was more numerous than the left-liberal category. While the proportion of overall Labour support declined across the five periods in all three categories on the left, the proportion of support coming from the left-authoritarian category remained very important. There was a fairly steady rise from the initial small proportion of Labour identifiers in the three right categories. By the final period the three right categories accounted for almost 16 % of all Labour identifiers, up from under 5 % in the first period, with a relatively large increase observed in the right-moderate category, especially between the second and third periods.
We now further explore how political allegiance changed for what might be thought to be the traditional Labour support base by using data from the BSA survey to establish where non-Labour identifiers from each of the three leftist political categories were distributed across the five periods. In particular, Table 4 reports how the distribution of individuals in the three left political categories varies for non-Labour identifiers across the five periods. The table shows that the percentage of individuals with leftist political views generally increased for all four other parties/groups. The only real exception to this was the significant decline observed in the percentage of Liberal Democrat identifiers amongst the left-authoritarian category. The largest increases occurred amongst individuals with leftist political views identifying with other parties and those not identifying with any party at all. The increase in support for other parties is most noticeable amongst the left-liberal category, which grew by over 10 % points between the first and final periods. Further analysis of this change reveals that the main party to experience a rise in support was the Green party. In particular, the percentage of Green party identifiers amongst this category increased from 2.5 % in the first period to 9.1 % in the final period. There was also a significant rise in the percentage of left-liberals identifying themselves as Liberal Democrat in the later periods.
A rise in support for some minority right-wing parties was more evident amongst people in the left-authoritarian category, with UKIP/BNP identifiers increasing to 4.9 % in period 5, compared to 1.3 % for the Green party. However, the most Table 3 Percentage distribution across the political categories for labour identifiers 1986-1991 1993-1997 1998-2001 2002-2005 2006-2009 1986-2009  important change for individuals in this category occurred for those not identifying with any political party, since the percentage increased from just over 8 % to almost 20 % of the total sample in this category across the five periods. A fairly similar percentage point increase can also be observed in the left-moderate category, whereas the change in the percentage of no-party identifiers in the left-liberal category was far smaller.
A key point is that up to 2001, leftists with authoritarian preferences increase primarily in the non-identifying category, while they actually drop for the Conservatives and Liberal Democrats and increase only slightly for other parties. In the same periods, left-liberal support increases for the Conservatives, Liberal Democrats and other parties, although these differences tend not to be significant. This is in line with our hypothesis that Labour's law and order policy protected support among left-authoritarians (at least against capture by other parties), at the expense of losing support (to other parties) amongst left-liberals. After 2001, however, there were some significant increases in left-authoritarians supporting other parties, as well as non-identifying. 13 This suggests the use of law and order to  1986-1991 1993-1997 1998-2001 2002-2005 2006-2009 1986-2009  The table reports the percentage of individuals from each of the three leftist political categories reporting that they were Conservative identifiers, Liberal Democrat identifiers, other party identifiers or did not identify with any party The stars indicate a significance difference in the percentage of such identifiers within each political category relative to the 1986-1991 period. * Significance at the 10 % level (using a two tailed test); ** at the 5 % level; *** at the 1 % level 13 There was also a significantly higher percentage of Conservative identifiers in this category in the final period.
Expressive voting and two-dimensional political… 93 hold on to the flank was beginning to crack after 2001, but was fairly successful up to that point. The analysis presented would appear to provide evidence in support of the primacy of economic and public policy over law and order. In particular, the BSA data in Table 2 indicates that Labour support amongst centre and right liberals increased up to 1998-2001 at a time when it was declining for left-liberals. This would seem to indicate that the move to the centre in economic and public policy was of greater consequence than a more authoritarian approach towards law and order. The evidence indicates strong support for our hypothesis regarding two types of populism. It suggests that economic and public policy was the main electoral issue and the move to the right was used by Labour to attract new votes and was thus broadly populist. A more authoritarian law and order approach could not prevent a loss of support on the left, but it stemmed the flow from the most populated left-wing category, especially up to the end of the first administration in 2001. In terms of the theoretical background discussed earlier, it is really only the expressive voting model that can make sense of this. Valence voting would eliminate differences as a factor and instrumental voting would imply that leftists should be the most motivated to continue supporting Labour, so long as Labour was still somewhat to the left of the Conservatives. Leftist support was leaking prior to 2001, when it was increasing for centrists and rightists, and we take this as very suggestive evidence for alienation and the need to minimise alienation's negative electoral effect through the use of tough policy on law and order.

Conclusions
This paper considers some of the large changes that have occurred in the recent UK political landscape by examining how the support base for the Labour party has evolved along both left-right and liberal-authoritarian dimensions since the mid-1980s. This has mainly been achieved through analysing consecutive annual data from the BSA survey. The analysis reveals that there has been a shift in Labour party identifiers from the traditional base amongst left-leaning voters towards a more centrist political stance but there is also evidence that voters with more authoritarian attitudes were targeted. This is consistent with the harder stance that was taken on law and order. Our thesis is that the tough approach to law and order was not intended as a broadly populist vote-winner. This role was played (extremely well) by economic and public policy shifts to the centre. Rather, it was a narrowly populist policy to retain as large a share of the reduced support on the left as possible. Economic and public policy was used to win votes in the centre and right and law and order policy was used to minimise the loss of votes on the left. The evidence suggests this worked quite well up to 2001 but appeared to become less effective after that.
We set aside the question as to whether the tough approach to law and order was also 'cynical' populism where policy is simply implemented for its popularity regardless of whether it is good policy or not. We believe that those commentators who view the new Labour approach to law and order as a broadly populist policy in the same mould as their economic and public policy are missing a subtle distinction between the two approaches. We argue that this finding can only be properly understood, theoretically, as driven by expressive voting, as it is only expressive voting that can properly address the role of alienation in voting. Finally, we believe the analysis conducted here is not just of recent historical significance. The use of tough law and order policy worked reasonably well until the early 2000s as a connection with the traditional Labour base. It failed after that and this disconnection between the party and its traditional supporters would, we argue, eventually appear as a significant factor in the 2016 UK referendum on European Union membership.