Economic Reformism vs Sociocultural Conservativism: Parties’ Programmes, Voters’ Attitudes and Territorial Features in the UK General Elections 2019

Abstract

This study explored the determinants of the massive vote shift that characterised the UK’s 2019 elections. To do so, this study firstly investigated opinions expressed by pro-Labour social groups about the hot issues of the political campaign; secondly, it conducted a cross-constituency analysis on the 2019 vote. The present findings reveal that the issues emphasised by the Conservative manifesto tallied with the opinions of traditional left-wing social groups. The cross-constituency statistical analysis confirmed this point and found a significant association between Labour’s losses and the territorial distribution of the abovementioned social groups. These findings suggest that the crucial aspect of the vote shift was the Conservatives’ appeal to the sociocultural conservativism of part of Labour’s traditional voters.

1. Preliminary Considerations

The United Kingdom’s general elections of 12 December 2019 have been regarded as a major political event. The elections resulted in a massive victory for the Conservatives, who won 56.2% of the seats, making a net gain of 48 seats compared to the previous general elections held only two years earlier (Figure 1). Labour won only 31.1% of the seats, losing 60 of them, among which seats—in working-class bastions—it had never lost before. The Figure 1 maps, showing the distribution of the vote variations by constituency, reveal not only the size of the changeover but also large differences and a spatial dependence in the electoral results, suggesting the territorial features’ relevance in the vote shift.

Such a vast political changeover hints at a collapse of voters’ partisan preferences. Political scientists have attributed to partisanship—the attachment to a party—a near-deterministic role in voters’ decisions (Campbell et al. 1960; Schaffner and Streb 2002; Weinschenk 2013). However, partisan loyalty inevitably varies over time and has weakened in the last years (Bafumi and Shapiro 2009; Mummolo et al. 2021). Partisanship endurance, moreover, would depend on the partisanship nature. Some scholars have regarded this nature as instrumental and, therefore, reactive to the party’s stances and performances (Downs 1957; Abramowitz and Saunders 2006; Garzia 2013). Other scholars have recently regarded partisanship as a visceral social identity comparable to religion and, therefore, not primarily reactive to ongoing events. Thus, such partisanship would affect voter stances rather than depend on party positions (Green et al. 2004; Huddy et al. 2018; Mason and Wronski 2018). However, the large and rapid shift characterizing the UK 2019 elections can hardly be ascribed to scarcely reactive partisanship.

More realistically, such a shift could be ascribed to an instrumental partisanship reacting to specific events. Was the shift, then, the result of the voters’ dissatisfaction with the government’s performance? The referendum-voting model (Simon 1989) considers all elections as ‘referendums’ on the ruling party’s performance. Economic growth plays a pivotal role within this model (Weschle 2014). If the economy thrives, voters tend to support the government in office; otherwise, they will vote for the opposition (Remmer and Gélineau 2003). However, it would have been difficult to regard the UK’s economics preceding the 2019 elections as entirely satisfactory or, vice versa, utterly disappointing. The UK’s yearly economic growth was ~1.4%, which was lower than the EU’s average figure but still not a negligible rate. The unemployment rate was slightly decreasing. Therefore, the massive vote shift characterizing the 2019 elections could hardly be traced back to the UK’s economic state.

The shift in question would better suit the paradigm of reactive partisanship responding to changes in the parties’ stances. According to the classical proximity model, voters tend to choose the party with stances closest to their own (Black 1948; Downs 1957; Joesten and Stone 2014). At the same time, when parties stress certain stances, their impact on voters and the probability of a shifting vote are expected to be greater (directional model: Rabinowitz and Macdonald 1989; Cho and Endersby 2003; Nicolet and Sciarini 2006). Moreover, parties may hope to attract voters insofar as they put aside issues shared by rival parties and privilege issues that they ‘own’ (Petrocik 1996; Klüver and Sagarzazu 2016; Seeberg 2017). That being the case, to explain the outcome of the UK’s elections, we intend to focus on themes presented as pressing and differently treated in the main parties’ programmes and on stances close to voters’ opinions.

2. Differences in the Programmes

The 2019 Conservative programme emphasised five objectives: to implement Brexit; enforce controls on foreign immigration; strengthen the security system; reduce pollution while implementing green policies; and improve the National Health Service (Conservative and Unionist Party 2019).

Labour, in turn, emphasised the following objectives: to negotiate a new deal on Brexit and then put it to a referendum; maintain the free movement of people within the EU; protect the environment; spend more on welfare including health services; support equality while reducing poverty; nationalise services, such as gas, electricity, water and transportation; and prevent crime (Labour Party 2019; Allen and Bara 2021).

As for the most important nonregional minor party, the Liberal Democrats, their key proposal was to cancel Brexit, as was made abundantly clear by the title of their manifesto: Stop Brexit.

We notice some overlapping between the Conservatives’ programme and Labour’s. Both parties highlighted the environmental protection issue. Both agreed on spending more on public health. Law-and-order was repeatedly mentioned in both manifestos; however, the Conservative manifesto mentioned law-and-order positively 12.5 times for each negative mention, while in the Labour manifesto, the ratio was 3.7 to 1 (Burst et al. 2021). Basically, the Conservative Party advocated a tougher criminal policy, and Labour a crime prevention strategy by helping young people stay away from crime. The Conservative frontman, Boris Johnson, expressed his determination to oppose early releases from prison, support a system of stiffer sentences, and prevent foreign offenders from entering the country.

At the 2017 elections, a stricter criminal policy was not top of the Conservatives’ manifesto, where crime was mentioned less often than in Labour’s manifesto (Allen and Bara 2021). Therefore, a stricter criminal policy was a distinctive feature of the 2019 Conservative campaign.

As for the remaining issues, the divergence between the main parties’ 2019 programmes was even more marked. The Conservative manifesto gave totemic relevance to Brexit, addressing this topic from an issue competition perspective: it repeated ‘Brexit’ 61 times, negatively described the EU 10 times, and positively mentioned the national-way-of-life 83 times (Burst et al. 2021). In turn, Labour’s message about Brexit had a certain ambiguity. During the Brexit campaign, the party’s leader, Jeremy Corbyn, had said that his support for the EU was ‘seven to seven and a half out of 10’. His support remained lukewarm throughout the 2019 campaign, leaving the monopoly of ‘let’s cancel Brexit’ to the Liberal Democrats. The Labour manifesto mentioned the EU positively, but not many times and without emphasis.

As for immigration control, Conservatives politicised this issue by highlighting it and taking a distinctive position on it. The Conservative manifesto contained 10 negative references to immigration, and the Labour manifesto contained none (Burst et al. 2021). The 2019 Conservative campaign promised an Australian-style, points-based system, intended to ‘end freedom of movement’, reduce the overall number of immigrants, and ‘ensure that the British people are always in control’ of their borders (Conservative and Unionist Party 2019).

Therefore, we shall focus on the objectives about which there was a substantive divergence between the two main parties’ 2019 electoral programmes: that is, the realization of Brexit, the enforcement of controls on immigration, and a stricter criminal policy.

3. Issues and Hypotheses

These political divergences are relevant to the implicit messages sent to the voter. Traditionally, Labour has been more sensitive to the working-class community than to supranational integration. Even if unenthusiastic, Labour’s support for the EU can be ascribed to the party’s proletarian internationalism (catchphrase: ‘international solidarity’, Labour Party 2019) as the antonym of bourgeois nationalism. This support can also be ascribed to the party’s perception of the EU as an institution regulating capitalism rather than a free-market-supporting one. In turn, Labour’s softer approach to treating criminals is in tune with the socialist tenet that crime is a by-product of capitalism. While Labour’s programme of expanding the welfare system and nationalizing public services can be regarded as a project of income redistribution under a state economy. Briefly, Labour relied on a traditional left/right dimension by being interventionist in economics and concerned with wealth redistribution (Marks et al. 2006). Labour chose a positional competition (Downs 1957; Green-Pedersen 2007), presenting itself as the paladin of positions stemming from a structural watershed, of which class-related divisions are the most obvious facet. Overall, Labour’s 2019 programme was more left-wing than all its previous programmes since the 1990s (Burst et al. 2021).

The Conservatives’ 2019 manifesto, in turn, presented some messages appealing to its traditional supporters through shibboleths such as national identity and independence (e.g., ‘We will stand up for our country and our values around the world’, Conservative and Unionist Party 2019). These messages would seem to be run-of-the-mill for the Conservatives, and therefore, they would not explain the vote shift. Stability cannot explain change. However, the Conservative messages of the 2019 campaign do not revolve around the regulation of the economy, namely the most traditional left/right divide. Instead, they fit the concept of ‘new politics’. In particular, they tally with the social paradigm described by the TAN acronym—Traditional, Authoritarian, Nationalist (Hooghe et al. 2002)—which includes frowning upon immigration, multiculturalism, and supranational agencies jeopardizing sovereignty.

Within this ‘new politics’ dimension, Conservatives rearranged their messages around three hot issues: national oneness vs supranational integration, immigration regulation, and security. The choice of these three issues is significant because all of them surfaced only recently in party manifestos (Budge et al. 2001; Burst et al. 2021). These issues are, in some respects, ideologically loaded and therefore not at odds with the positional competition. However, they can become part of an issue competition (Green-Pedersen 2007; De Sio and Lachat 2020; Fumarola 2021) as far as they are presented, innovatively, as hot issues associated with widely shared and desired policies rather than with class-related positions.

This study, therefore, contributes to the debate about the contraposition between positional and issues competition, between a traditional, spatial—i.e., along a left-right scale—competition and a new one with thematic emphasis, targeting policy problems and policy sectors (Guinaudeau and Persico 2014). However, we suggest that behind the shift in the 2019 UK elections, there was something else as well.

This additional factor is the shift of the working-class vote towards right-wing parties. A declining relevance of working-class support for left-wing parties represents a phenomenon identified in Western countries during the last decades. It has been usually ascribed to structural reasons: from the erosion of factory jobs by the services sector to rising affluence and education and to the declining share of representatives coming from working-class backgrounds (Franklin 1992; Heath 2015; Cutts et al. 2020). Other authors pointed to ideological convergence between parties, owing to a shift to the centre by the left-wing parties (Evans and Tilley 2017, Chp. 6). Still others ascribed the decline of class–party association to new attitudes in the working class, such as a backlash of left-behind people against multiculturalism, cosmopolitanism, and political correctness (Inglehart and Norris 2016; Gidron and Hall 2017).

We believe, however, that the working-class shift to the right can be better understood by focusing on the appeal exerted on traditional Labour voters by the Conservatives’ new hot issues.

Our primary and general hypothesis is the following:

H1. 

The British working class has been sharing conservative stances on nationalism, foreign immigration control, and the punishment of delinquents, and has been doing it for a long time.

From this hypothesis, we develop further, more specific hypotheses:

H2. 

From social surveys, it is possible to infer that traditionally pro-Labour groups had manifested, long before the 2019 elections, opinions and attitudes in tune with the distinctive points of the 2019 Conservative programme.

H3. 

Post-election survey data provide evidence that pro-Labour social groups sharing those opinions and attitudes tended to shift their vote from Labour to the Conservatives.

H4. 

Cross-constituency regression models of the vote shift provide a crucial statistical confirmation of the above: we expect the territorial distribution of the abovementioned social groups to predict the vote shift in the 2019 elections.

H5. 

Some macro features of political relevance because connected with the main divisive points of the 2019 electoral campaign (e.g., foreign immigrants’ share and crime rates) statistically contributed to that vote shift.

H6. 

A spatial test would prove that the vote shift was not spatially independent, i.e., that close-by territorial units presented similar electoral outcomes.

4. Methods

To investigate the reason for vote shifts, one could use micro data from post-election surveys. These data would say much about the relationship between individual characteristics and voting. Still, they would ignore the role of territorial features, such as urbanization, immigrant share, and crime rates. A macro analysis, instead, would take into account these features. A typical macro analysis of election or referendum outcomes (e.g., Becker et al. 2017; Johnston et al. 2018) would follow a two-step procedure. First, it would trace back the cross-territory votes for a party to the territories’ features and the local voters’ demo-socioeconomic traits. Second, it would infer from this association the motivation for those voters’ choices. However, while one could challenge the first step only in terms of statistical reliability, the second step would produce an inference easily challenged in terms of validity. This is because behind the voters’ choices there might have been a motivation utterly different from that inferred.

To reduce these validity problems, we intend to anchor the inference about the vote determinants to voters’ opinions rather than only to their demo-socioeconomic characteristics. The recourse to voters’ opinions can also help us reduce the risk of spurious correlations between cross-territory electoral results and the inhabitants’ characteristics. In this paper, we will regard as plausible only those cross-territory correlations indirectly confirmed by the opinions of people sharing the characteristics of interest.

In short, we intend to organise our analysis along three steps. First, we will hypothesise—with the help of the literature—which social groups would most likely agree with the main policies advocated by major parties. Second, we will test these hypotheses by identifying social groups who expressed opinions in tune with those policies. Because the loss of votes suffered by Labour was the momentous event of the 2019 elections, we will identify issues about which people close to this party expressed, before those elections, opinions more in tune with the Conservative programme (or, subordinately, with that of the Liberal Democrats) than with Labour’s. We expect that those people would be enticed to shed their former political allegiance and cast a vote for the party presenting stances closer to their opinions. We will use the post-election survey to find confirmation of the above. Finally, we will analyse the association between those groups and the election outcomes. If a share of those social groups set aside its former political allegiance, Labour’s electoral losses would emerge as correlated with the groups’ cross-territory distributions. Therefore, we intend to test the following Equation (1):

$$\Delta y_i=\alpha +{\beta }_1{x}_i+{\beta }_2{d}_i+\lambda {\sum }_{j=1}^N{w}_{i,j}\Delta y_{j,j\ne i}+{\epsilon }_i$$

(1)

where $\Delta y_i$ is the Labour’s vote 2017–2019 variation in the territorial unit i, $\alpha$ a constant, $x$ a set of explanatory variables relating to territory characteristics (social groups’ share, further macro features and past vote), $d$ a regional dummy, $\lambda$ a spatially weighted average of vote variations in the other places (which would measure the spatial dependence of those variations), and $\epsilon$ the error.

5. Data

To identify citizen’s opinions before the 2019 elections, we used the British Social Attitudes surveys’ micro data (hereafter BSA) (UK Office for National Statistics 1983, 1990, 2003, 2016, 2017, 2018, as well, owing to some differences in the questionnaires over time) and the British Election Study (hereafter BES) (UK British Election Study 2016). Additionally, to check the impact of individual attitudes on the 2019 vote, we used the 2019 Post-Election Survey (hereafter PES) (UK British Election Study 2021), although this provides narrower information compared to BSA.

To test the paper’s hypotheses, we operationalised them by translating them into explanatory and response variables deriving from aggregated data. These data relate to the UK electoral constituencies, each electing a single member of the House of Commons by the first-past-the-post system.

We used cross-constituency votes for the main parties to measure the electoral outcome. In particular, we considered Labour’s share in the 2017 and 2019 elections and the variations between the two elections. To this, we added a dummy identifying the constituencies where Labour won in 2017. We also calculated the variations in the Conservatives’ and Liberal Democrats’ votes to understand where the shifting vote went. Because in Scotland and Northern Ireland other political parties hold the field, we limited our analysis to England and Wales: each of their 573 constituencies had, on average, 73,675 electors on 264 square kilometres.

Statistical analyses of electoral votes may present problems too often disregarded (Alaimo and Solivetti 2019). We eliminated two strong outliers from the vote variations between the two elections, due to the main parties’ local absence in one election. We checked heteroskedasticity and collinearity and found them to be within normal limits.

Next, we collected many potential predictors of the outcomes. To measure the impact of the Brexit vote, we considered the cross-constituency vote shares for ‘Leave’ and the same shares squared, plus the ‘Labour 2017 win and “Leave” win’ dummy. We considered basic demographic features as well: the population density; a measure of urbanization that we calculated as the share of people living in an area regarded as belonging to a city; age groups; local population by ethnicity (white, mixed, etc.) and by religion (Christians, Muslims, etc.); population born in the UK, in England and Wales, in the EU, elsewhere in Europe, etc.; foreign population (people holding only a foreign passport); native population (people born and residing in England or Wales); and people regarding themselves as having British identity. We created some interactions as well: for instance, ‘natives × Christians’.

As for occupations, we considered people employed in professional jobs, in ‘process, plant and machine jobs’, in ‘caring, leisure and other services’, in ‘sales and customer services’, and in ‘elementary occupations’. Finally, as for activity, we selected the unemployment rate.

We also considered educational qualifications (BSA), from ‘no qualification’ to level 1 (1–4 GCSEs), apprenticeship, level 2 (five or more GCSEs), level 3 (two or more A-levels), and level 4 (corresponding to a first or higher degree). In addition, we created interactions such as ‘process-plant-machine-jobs × no-qualification’.

As for socioeconomic status, we considered the average income of the constituencies’ population, their median weekly wages, the wages divided by the unemployment rate, the 20th-percentile wages, and the ‘universal credit and job seekers allowance claimants’.

As for crime, we considered the total crime rates and the rates of some offences particularly common or serious, such as antisocial behaviour, shop-lifting, ‘criminal damage and arson’, and ‘violence and sexual offences’.

Lastly, we used dummies to take into account regional identities.

Table A2. List of the variables and summary statistics.

Variables	Obs.	Mean	Std. Dev.	Min.	Max.
Labour’s 2015 votes share (%)	572	33.189	16.790	5.433	81.301
Labour’s 2017 votes share (%)	572	43.386	17.614	8.135	85.729
Labour ‘s 2019 votes share (%)	571	35.387	17.276	4.350	84.681
Labour’s 2015–17 vote variation (%)	571	10.197	4.139	−3.229	30.273
Labour’s 2017–19 vote variation (%)	571	−7.978	4.276	−24.904	11.548
Conservatives’ 2017 votes share (%)	571	44.104	14.955	7.301	69.918
Conservatives’ 2019 votes share (%)	571	46.018	15.861	7.818	76.724
Conservatives’ 2017–19 vote variation (%)	571	1.914	4.446	−9.706	19.053
Lib Dems’ 2017 votes share (%)	571	7.242	8.682	0.000	52.752
Lib Dems’ 2019 votes share (%)	571	11.285	10.073	0.000	56.069
Lib Dems’ 2017–19 vote variation (%)	571	4.044	4.812	−18.107	28.863
Labour 2017 win (dummy)	571	0.445	0.497	0	1
Labour 2017 win and ‘Leave’ win (dummy)	571	0.277	0.448	0	1
‘Leave’ 2016 share (%)	571	53.490	10.826	20.480	75.650
(‘Leave’ 2016 share)2	571	2978.2	1075.6	419.4	5722.9
Population density (people per sq. km.)	571	2284.7	2918.3	23.000	16,802.2
Population living in a city (fraction)	571	0.207	0.396	0	1
Population 18 to 24 years old (%)	571	8.442	3.793	4.902	31.154
Population 65 years or older (%)	571	18.818	5.159	5.868	34.734
Native population (%)	571	83.543	12.213	39.847	96.835
Muslims (fraction)	571	0.043	0.073	0.001	0.521
Christians (fraction)	571	0.599	0.095	0.242	0.815
Natives × Christians	571	50.812	13.100	13.048	78.878
White ethnicity population (%)	571	87.180	15.415	23.086	98.979
Black ethnicity population (%)	571	2.957	5.314	0.057	37.380
Mixed ethnicity population (%)	571	2.066	1.583	0.396	7.827
Population with British identity (%)	571	91.686	8.046	58.044	98.891
Population born in the UK (%)	571	87.623	11.637	40.728	98.018
Population born in England or Wales (%)	571	85.925	11.613	40.123	97.460
Population born in the EU (%)	571	3.409	2.774	0.585	16.924
Population born elsewhere in Europe (%)	571	0.455	0.771	0.054	8.089
Population born in North Africa (%)	571	0.211	0.289	0.015	3.139
Population born abroad (%)	571	12.377	11.637	1.982	59.272
Population with only foreign passport (%)	571	6.829	6.874	0.790	36.340
Educational qualification none (%)	571	22.945	5.685	9.566	39.227
Educational qualification 1 (%)	571	13.391	2.272	5.680	19.204
Educational qualification 2 (%)	571	15.387	2.209	7.257	18.549
Educational qualification 3 (%)	571	12.322	2.414	8.337	27.653
Educational qualification 4 (%)	571	26.806	8.366	12.065	57.391
Educational qualification apprenticeship (%)	571	3.659	1.179	0.732	8.480
Pop. in administrative and secretarial services (%)	571	11.454	1.649	7.747	18.795
Pop. in sales and customer services (%)	571	8.543	1.723	3.826	14.970
Pop. in elementary occupations (%)	571	11.282	2.845	4.212	19.131
Pop. in skilled trades (%)	571	11.699	2.749	3.317	21.233
Pop. in process, plant and machine jobs (%)	571	7.437	2.638	1.485	16.407
Pop. in process, etc., jobs × no educ. qualification	571	183.51	102.69	14.790	586.31
Pop. in professional occupations (%)	571	17.018	4.902	8.634	37.420
Unemployment rate (%)	571	4.114	1.380	1.800	10.700
Universal credit & job seekers allowance claimants (%)	571	2.861	1.347	0.800	8.300
ln(total income) (GBP average annual value)	571	10.072	0.138	9.804	10.667
Wages (median) 2019 (GBP per week)	571	588.20	84.971	420.00	890.00
Wages (median)/unemployment rate	571	160.55	61.335	45.098	416.67
Wages of the 20th percentile (GBP per week)	571	389.89	44.652	318.60	568.10
ln(shop-lifting) (per 100K pop.) *	571	4.849	0.603	2.890	6.992
ln(antisocial behaviour) (per 100K pop.) *	571	6.259	0.516	4.291	7.718
ln(criminal damage and arson) (per 100K pop.) *	571	5.387	0.391	3.550	6.684
ln(violence and sex offences) (per 100K pop.) *	571	6.612	0.383	5.514	7.638
ln(all crimes) (per 100K pop.) *	571	7.856	0.369	6.926	9.822
London Region (dummy)	571	0.128	0.334	0	1
North East Region (dummy)	571	0.051	0.220	0	1
North West Region (dummy)	571	0.130	0.336	0	1
Yorkshire and Humber Region (dummy)	571	0.095	0.293	0	1

* Crimes rates concern the last three months (April to June) available for all the constituencies before the 2019 elections.

provides the list of the variables and their summary statistics.

Our data come from the UK House of Commons Library (2020) and other national statistics sources, including the UK 2011 Census (UK Office for National Statistics 2011). Crime data come from the UK Police records (UK Police 2019) as figures by Lower Layer Super Output Areas; we upscaled them to wards and finally to constituencies.

6. People’s Socioeconomic Characteristics and Their Opinions about Political Issues

From the 1970s onward, political science literature has identified the characteristics of people backing supranational integration or opposing it. By and large, people favouring supranational integration are better off, better educated, and with higher, competitive skills, making them potential winners in the international arena (Kriesi et al. 2006; Goodwin and Heath 2016; Hobolt 2016); and vice versa. Age, in turn, seems to be as good as class when it comes to predicting the attitudes towards internationalism and supranational integration (Sloam et al. 2018): young people being much more in favour of them.

As for, in particular, the UK citizens’ opinions, we know that supranational integration was much more frowned upon than loved. While 54.7% of citizens agreed that the ‘EU undermines Britain’s right to be an independent country’, only 24.9% disagreed with that (BSA 2018).

Moreover, confirming the literature, BSA surveys show that hostility towards supranational integration is correlated with SES. Since those possessing higher, competitive skills tend to gather in cities, the interviewees thinking Britain should leave the EU were 23.9% among city dwellers and 41.1% among people living in a village or farm (BSA 2017). As for age groups, 55.6% of those aged 65 years or older chose ‘leave the EU’, while the share dropped to 24.0% among those 18 to 25 years old. Among Labour’s supporters, ‘leave’ was the choice of 40.4% of those aged 65 years or older and only 10.2% of those 18 to 25 years old. Additionally, while those in favour of ‘leave’ were 25.2% among ‘professionals’, they were 55.0% among ‘manual skilled workers’. ‘Leave’ represented the choice of a high share (59.7%) of unemployed people as well. If people regarded themselves as supporters of, or close to, Labour, that percentage remained large (48.3%). Concomitantly, the share of unemployed people close to Labour was high (62.7%), while only a few of them (16.7%) were close to the Conservatives (BSA 2017). This could have been relevant to the election outcomes: likely, among the ~33% of Labour’s supporters who had voted ‘Leave’ in the Brexit referendum (BES 2016), some later cast a vote for the party pressing for the Brexit implementation. The post-election survey confirms the above: 70.6% of the voters who shifted from Labour to the Conservatives had voted ‘Leave’; among those who stayed true to Labour, the percentage was 19.8 (PES 2021). On the other hand, hostility to supranational integration has long been deep-rooted among Labour’s supporters. BSA provides repeated evidence about it. In 1983, for example, ‘leave the EEC’ was the answer of 53.8% of all manual workers and 59.6% of the unskilled ones. These shares were even higher (62.1% and 60.8%) when the interviewees were close to Labour (BSA 1983).

As for foreign immigration control, it partly overlaps the Brexit issue because the free circulation of people has been a top feature of the project underlying Europe’s integration. However, the public demand for stricter controls on immigration is more likely a cause—rather than a side effect—of the aversion to supranational integration. The current literature has identified a few motivations behind this demand. One motivation derives from competition in the labour market: low-skilled native workers blame immigrants of similar skills for lowering their wages, taking their jobs away or profiting from welfare benefits (Dancygier and Donnelly 2013). Another motivation derives from the natives’ devotion to their cultural identity and from their hostility towards foreign culture and customs (Helbling and Traunmüller 2016). The opposition to immigrants and multiculturalism among poorly educated people is widespread and intense (Zick et al. 2011; European Union 2018).

Voters’ origin, as well, is expected to influence their attitudes towards immigration (Pantoja et al. 2001). Native people, that is, people living where they were born, tend to be less favourable to immigration than both foreign and domestic immigrants.

Whatever the reason for hostility to foreign immigrants, findings revealed that the foreign population’s size is directly prognosticative of such hostility (Hainmueller and Hangartner 2013; Hangartner et al. 2019).

BSA data provide information on UK citizens’ opinions about foreign immigration. When asked what to do about immigration into Britain, those answering ‘it should be reduced a lot’ were significantly more numerous (27.8%) among the natives than among non-natives (15.6%). Moreover, ‘reduced a lot’ was the answer of 37.0% of the skilled manual workers, 37.8% of the unskilled ones, and only 10.8% of the professionals. When the interviewees were close to Labour, the share among skilled workers slightly decreased (31.0%), but it increased to 38.2% among the unskilled ones and dropped to 3.7% among professionals. As for education, ‘reduced a lot’ was the answer of 40.2% of those without any qualification and only 10.1% of those holding a university degree. Concerning those close to Labour, the percentage for the least educated ones was slightly lower (37.8%), while the degree-holders’ percentage fell to 6.4% (BSA 2017). These attitudes of the working class towards immigration are nothing new. In BSA 2003, for example, ‘reduced a lot’ was the answer of 64.4% of skilled manual workers and 70.6% of the unskilled ones.

Because Labour was by far the preferred party of ‘no qualification’ people and both skilled and unskilled workers (BSA 2017), shifting voters among them would have been relevant to the electoral outcome in 2019. What is sure is that 64.7% of the voters who shifted from Labour to the Conservatives thought ‘too many immigrants [were] let into this country’. Among those who stayed true to Labour, the percentage was only 28.1 (PES 2021).

The political relevance of the security issue, in turn, could be caused by more than one factor. One factor could be the fear of foreign terrorism and immigrant crime. In this case, people favouring securitization would be those holding anti-immigrant attitudes. However, the security issue appeal could also derive from a more wide-ranging worry about crime (Johnson 2009; Armborst 2017) and from dissatisfaction with lenient criminal policies. Apropos of this, we notice that the vast majority (67.5%) of those living in Great Britain think that ‘who break[s] the law should be given stiffer sentences’ (BSA 2018). If the interviewees were close to Labour, the share would be lower, but only slightly (58.4%). Among 65-year or older people, those in favour of stiffer sentences rose to 72.5%.

In addition, lower social classes—Labour’s traditional vote bank—tend to be much more in favour of the extreme penalty than higher classes. In 2018, those who thought that ‘for some crimes, the death penalty is the most appropriate measure’ (BSA 2018) were only 28.1% of professional people but represented the large majority (64.9%) of those in manual skilled occupations, that is, those who could be regarded as the working-class aristocracy. However, again, these attitudes are nothing new. Among skilled manual workers, those supporting the death penalty had been 68.6% (BSA 2003) and even 78.7% (BSA 1990).

Interestingly, 23.1% of people close to the Conservatives were worried about crime in their local areas; the share rose to 24.0% among people close to Labour (BSA 2016). Of course, concern about crime is the child of many fathers: it is the outcome of mass communication, panic waves, and the relevance of concurrent concerns. However, to deny a link between concern about crime and crime diffusion and victimization is contrary to logic and figures. Victim status is predictive of worry about crime (for example, Jansson 2007; Flatley 2017). About this, we notice that the UK North East is the region with the highest rates of sexual offences, stalking and harassment, shop-lifting, and criminal damage and arson. In turn, Yorkshire and Humber fractionally outdoes the North East in terms of violence against the person, theft, total recorded crime, and ‘miscellaneous crimes against society’ (UK Office for National Statistics 2020). Correspondently, the North East is the region with the highest percentage of people who think that ‘who break[s] the law should be given stiffer sentences’. It is also the region with by far the highest percentage of people in favour of the death penalty (BSA 2017). In turn, Yorkshire and Humber is the region with by far the highest rate (33.1%) of people worried about crime in their local area (BSA 2016).

Empirical findings revealed that crime rates influence voters’ political choices (Hagerty 2006; Cummins 2009; Drago et al. 2020). Therefore, owing to the increased relevance of the security issue in the 2019 Conservative programme, a shift in partisan alignment may have occurred in areas where crime was rife and the share of those worrying about crime larger. The post-election survey showed indeed that those advocating the death penalty were much more numerous among the shifting voters (56.5%) than among those who stayed true to Labour (25.5%) (PES 2021).

7. Results of the Statistical Analysis

Labour’s votes of the past can be traced back to the electoral turf of a left-wing party relying on positional competition. Labour’s bond with the working class was not as close as it used to be (Heath 2015; Cutts et al. 2020;

Table A1. Pearson correlations between Labour’s 2015–2017–2019 vote, 2017–2019 vote variation in, respectively, the Labour, Conservative and Liberal Democrats parties and explanatory variables; England and Wales’ constituencies.

Explanatory Variables	Labour’s 2015 Vote %	Labour’s 2017 Vote %	Labour’s 2019 Vote %	Labour’s 2015–17 Vote Var.	Labour’s 2017–19 Vote Var.	Tories’ 2017–19 Vote Var.	Lib Dems’ 2017–19 Vote Var.
Labour’s 2017 votes share	0.972	1.000	0.970	0.312	−0.202	−0.009	−0.260
Labour 2017 win	0.834	0.845	0.801	0.213	−0.249	−0.059	−0.270
Labour 2017 win and ‘Leave’ win	0.533	0.492	0.389	−0.068	−0.459	0.196	−0.267
‘Leave’ 2016 share	−0.128	−0.186	−0.290	−0.275	−0.404	0.672	−0.269
(‘Leave’ 2016 share)2	−0.090	−0.153	−0.264	−0.284	−0.436	0.700	−0.280
Population density (people per sq. km)	0.524	0.551	0.599	0.218	0.153	−0.233	0.021
Population living in a city	0.420	0.437	0.498	0.158	0.211	−0.278	0.040
Population 18 to 24 years old	0.313	0.379	0.428	0.342	0.171	−0.101	−0.188
Population 65 years or older	−0.619	−0.651	−0.679	−0.260	−0.062	0.205	−0.003
Native population	−0.312	−0.350	−0.419	−0.226	−0.252	0.306	−0.110
Muslims	0.503	0.526	0.583	0.199	0.188	−0.153	−0.058
Christians	−0.340	−0.399	−0.462	−0.319	−0.223	0.188	0.012
Natives × Christians	−0.323	−0.383	−0.458	−0.319	−0.273	0.267	−0.047
White ethnicity population	−0.519	−0.538	−0.593	−0.182	−0.179	0.195	−0.020
Black ethnicity population	0.462	0.481	0.528	0.172	0.151	−0.176	0.012
Mixed ethnicity population	0.405	0.444	0.511	0.249	0.234	−0.309	0.102
Population with British identity	−0.344	−0.380	−0.444	−0.221	−0.227	0.325	−0.149
Population born in the UK	−0.383	−0.408	−0.469	−0.185	−0.212	0.321	−0.167
Population born in England or Wales	−0.365	−0.392	−0.455	−0.188	−0.223	0.335	−0.180
Population born elsewhere in the EU	0.218	0.252	0.310	0.189	0.212	−0.320	0.212
Population born elsewhere in Europe	0.237	0.260	0.305	0.147	0.160	−0.280	0.167
Population born in North Africa	0.329	0.359	0.404	0.191	0.155	−0.314	0.168
Population born abroad	0.383	0.408	0.469	0.185	0.212	−0.321	0.167
Population with only foreign passport	0.354	0.383	0.443	0.195	0.208	−0.334	0.182
Educational qualification none	0.396	0.340	0.245	−0.157	−0.414	0.617	−0.435
Educational qualification 1	−0.045	−0.080	−0.151	−0.156	−0.282	0.564	−0.255
Educational qualification 2	−0.391	−0.422	−0.481	−0.209	−0.207	0.433	−0.144
Educational qualification 3	−0.019	0.046	0.100	0.274	0.214	−0.076	−0.180
Educational qualification 4	−0.217	−0.180	−0.095	0.117	0.355	−0.684	0.461
Educational qualification apprenticeship	−0.305	−0.335	−0.406	−0.188	−0.258	0.386	−0.191
Administrative and secretarial services	0.010	−0.006	−0.014	−0.069	−0.028	0.065	0.041
Sales and customer services	0.565	0.561	0.520	0.096	−0.210	0.367	−0.424
Elementary occupations	0.509	0.493	0.442	0.032	−0.247	0.490	−0.451
Skilled trades	−0.311	−0.343	−0.393	−0.199	−0.175	0.534	−0.297
Process, plant and machine jobs	0.340	0.284	0.183	−0.172	−0.429	0.637	−0.403
Process, etc., jobs × no educ. qualification	0.398	0.334	0.228	−0.192	−0.456	0.613	−0.391
Professional occupations	−0.103	−0.057	0.028	0.174	0.349	−0.672	0.393
Unemployment rate	0.651	0.637	0.602	0.073	−0.196	0.186	−0.280
Un. credit and job seekers allowance claimants	0.700	0.680	0.642	0.057	−0.209	0.281	−0.362
ln(total income) average annual value	−0.315	−0.302	−0.250	−0.007	0.233	−0.596	0.546
Wages (median) 2019	−0.265	−0.252	−0.199	−0.001	0.234	−0.557	0.519
Wages (median)/unemployment rate	−0.638	−0.628	−0.586	−0.084	0.221	−0.412	0.479
Wages of the 20th percentile	−0.211	−0.193	−0.139	0.035	0.233	−0.544	0.475
ln(shop-lifting)	0.311	0.333	0.293	0.153	−0.186	0.134	−0.153
ln(antisocial behaviour)	0.408	0.421	0.397	0.134	−0.132	0.136	−0.167
ln(criminal damage and arson)	0.384	0.386	0.323	0.086	−0.286	0.299	−0.283
ln(violence and sex offences)	0.504	0.521	0.487	0.172	−0.180	0.349	−0.354
ln(all crimes)	0.582	0.602	0.578	0.203	−0.147	0.148	−0.222
London Region	0.246	0.241	0.284	0.027	0.155	−0.259	0.156
North East Region	0.194	0.162	0.098	−0.098	−0.271	0.091	−0.089
North West Region	0.275	0.259	0.267	−0.015	0.012	−0.002	−0.133
Yorkshire and Humber Region	0.138	0.124	0.079	−0.033	−0.190	0.061	−0.083

: Labour’s vote in 2017 vs 2015). However, Labour’s votes in the 2017 elections revealed a robust spatial dependence (Moran test chi² = 1172, p = 0.000, ID matrix) and they were still territorially associated with higher rates of unemployment and benefit claimants, working-class and lower-education people, immigrants, urbanization, a larger presence of youth, and higher crime rates. Labour’s votes exhibited a regional differentiation as well, England’s northern regions representing the traditional Labour bastions. In both the 2017 and 2019 elections, the ‘Remain’ share—‘Remain’ and ‘Leave’ shares were complementary to each other—came and joined those traditional correlates of Labour’s vote (Table A1).

However, as to the 2017–2019 variations in Labour’s votes, while the spatial dependence remained relevant (Moran test chi² = 231, p = 0.000, ID matrix), the territorial determinants were different (Table A1). Although Labour’s votes in both the 2017 and 2019 elections were negatively correlated with ‘Leave’ votes in the Brexit referendum, that correlation was closer in the 2019 elections. Therefore, variations in Labour’s votes were negatively correlated with the ‘Leave’ votes: a 1% cross-constituency increase in the ‘Leave’ vote’s squared share corresponded to a 0.63% decrease in Labour’s 2019 vote. Liberal Democrats as well lost votes in proportion to the ‘Leave’ share, while Conservatives gained votes. Losses and, respectively, gains in the 2019 elections were higher when we considered the ‘Leave’ squared share in place of the simple share. This means that Labour made more than proportional losses where ‘Leavers’ had reached the highest shares. Moreover, Labour’s losses were higher where the party had won the seat in the previous elections.

However, Conservative Party supporters who had voted ‘Remain’ in 2016 could have voted for Labour in the 2019 elections. Indeed, Labour’s votes increased in constituencies with larger shares of professionals, degree-holders, and young people: categories favouring supranational integration. In the 2015 elections and earlier ones, the professionals’ and degree-holders’ shares were negatively correlated with the Labour’s vote. Instead, this correlation became positive, namely to Labour’s benefit, in the 2017 elections—concomitantly with the Brexit issue—and became even more momentous in the 2019 ones. Concurrently, in 2019, Conservatives scored significant losses in constituencies with many professionals and degree-holders. However, those who gained more in those constituencies were the Liberal Democrats, not the Labour Party (Table A1).

To continue with our determinants, we notice that the 65-year or older age group was negatively correlated with Labour’s vote and positively with the increase in the Conservatives’ vote. The white-ethnicity population’s share had a significant negative impact on Labour’s vote and a positive one, approximately to the same extent, on the Conservatives’ vote. What happened with the white-ethnicity population was replicated in the case of people born in England and Wales, the native population, Christians, ‘natives × Christians’, and people regarding themselves as having British identity. The opposite occurred with Muslims, people born in the EU, and mixed ethnicity population: all contributing to a positive variation in Labour’s votes.

As for education, we already noticed the correlation between Labour’s vote and people without an educational qualification. In 2019, however, that correlation became weaker. Therefore, where the share of people without qualification was higher, Labour lost votes: a 1% cross-constituency increase in people without qualification corresponded to a 0.89% decrease in Labour’s votes. Conservatives, concurrently, increased theirs. Something similar occurred where unemployed people, claimants for universal credit and job seekers’ allowance, people occupied in elementary occupations, and process-plant-machine workers were more numerous. In the constituencies’ top decile in process-plant-machine workers’ share, Labour lost 11.8% of its votes; it lost only 6.4% in the bottom decile.

Labour gained, instead, where median wages and total income were higher. There, the correlation with Labour’s vote had used to be negative, and so it remained in the 2019 elections, but Labour’s share became relatively larger and the Conservatives’ smaller. However, as in the case of professionals and degree-holders, those who gained more where wages and income were higher were the Liberal Democrats, not Labour.

As for crime, we already noticed the lasting territorial correlation between Labour’s vote and crime rates. This correlation concerned the total crime rate and both minor offences, such as antisocial behaviour, and serious ones such as ‘violence and sex offences’. However, in high-crime constituencies, Labour (and the Liberal Democrats) significantly lost votes in the 2019 elections, and the shift benefitted Conservatives (Table A1).

We can check the reliability of these correlations by employing spatial multiple regression models, which allow us to analyse coeteris paribus the impact of each determinant and the spatial factor on the election outcomes. Model 0 (

Table 1. Generalised spatial two-stage least squares regression models of statistical determinants of Labour’s 2017 vote and 2017–2019 vote variation by England and Wales’ constituency: coefficients and ‘t’ values.

Explanatory Variables	Model 0		Model 1		Model 2		Model 3		Model 4
	Previous Labour’s Vote Determinants		Hostility to EU and Labour’s Vote		Hostility to Immigration and Labour’s Vote		Crime Control Policy and Labour’s Vote		Summarising Model: Labour’s Vote
	Labour’s 2017 Vote %		Labour’s 2017–2019 Vote Var.		Labour’s 2017–2019 Vote Var.		Labour’s 2017–2019 Vote Var.		Labour’s 2017–2019 Vote Var.
Labour’s 2017 votes share									−0.066	−4.52
(‘Leave’ 2016 share)2	−0.009	−14.46	−0.001	−3.69					−0.001	−3.24
Labour 2017 win and ‘Leave’ win									−1.644	−3.85
Population density	0.001	1.46
Population living in a city			2.758	4.66	2.361	4.22			1.836	2.94
Population 18 to 24 years old	0.217	1.76	0.076	1.56	0.104	2.35			0.095	1.95
Population 65 years or older	−0.791	−5.93					−0.322	−6.04	−0.101	−2.09
Black ethnicity population	0.402	3.80
Muslims					23.226	7.29			21.596	7.76
Natives × Christians					−0.032	−1.40
Pop. born elsewhere in Europe							0.613	2.40
Pop. with only foreign passport					−0.120	−2.56			−0.180	−4.01
Educational qualification none	1.358	9.40	−0.149	−2.56
Edu. qualification apprenticeship	2.616	4.86
Elementary occupations							−0.433	−4.56
Process, plant and machine jobs	1.230	4.44
Process, etc., jobs × no educ. qualification					−0.017	−8.23			−0.099	−3.34
Sales and customer services	1.823	5.85
Unemployment rate			−0.477	−2.80	−0.666	−4.11
Wages (median)/unemployment rate	−0.026	−2.90
Wages of the 20th percentile			−0.091	−1.56			−0.009	−1.42	−0.018	−3.24
ln(antisocial behaviour)	2.105	2.95
ln(criminal damage and arson)							−2.091	−3.78
London Region	−0.544	−0.29							1.673	1.88
North East Region	7.751	4.78					−4.593	−5.97	−3.721	−5.73
North West Region	6.904	5.81
Yorkshire and Humber Region	4.686	4.01					−1.687	−3.01	−1.795	−3.77
Constant	−3.631	−0.46	5.543	1.76	1.433	0.96	23.340	4.09	11.394	3.24
W (inverted distance matrix) dependent var.	0.269	4.83	0.439	3.74	0.586	5.01	0.850	6.66	0.506	4.10
(Pseudo)R2	0.82		0.24		0.33		0.22		0.46
N	572		571		571		571		571
Variance Inflation Factor (average)	3.31		2.70		2.65		2.05		3.38

) provides a concise picture of the time-honoured correlates of the pro-Labour vote: in particular, it confirms the support traditionally given to the party by working-class and lower-education people, ethnic minorities, and England’s northern regions. Models 1–4 (Table 1) concern Labour’s 2017–2019 vote variations. Model 1 shows the impact of variables that we have traced back to attitudes towards supranational integration. While the over-time variations in Labour’s vote show a positive correlation with the degree of urbanization, those variations are negatively correlated with the share of citizens who voted for Brexit and the unemployment rate.

Model 2 shows the impact of variables that we have traced back to attitudes towards foreign immigrants. We notice, first, a close correlation between the share of ‘process-plant-machine-jobs × no-qualification’ people and the decrease in Labour’s votes. The presence of a foreign population is also correlated with a reduction in Labour’s share.

As for the impact of crime, Model 3 confirms that high-crime regions—namely the North East and Yorkshire and Humber—were associated with negative variations in Labour’s votes. Model 3 also shows the significance of ‘criminal damage and arson’. These negative variations in Labour’s votes were also associated with the 65-year or older age group: the group most favouring stiffer sentences. These results hold at parity of variables considered both the correlates of Labour’s vote and the determinants of common crime, such as elementary occupations and the wages of lower-income workers.

Model 4 summarises the factors contributing to variations in Labour’s votes. In this model, we notice the negative statistical contributions made by the ‘Leave’ share, Labour’s share in the 2017 elections, and ‘Labour 2017 win and “Leave” win’. In Model 4, we also find the negative contributions made by the share of ‘process-plant-machine-jobs × no-qualification’ people, the foreign population’s size, and England’s northern regions. Therefore, Labour’s losses were hefty where the share of poorly educated, working-class people was higher and where foreign immigrants were more numerous. Moreover, Labour’s losses were heavier where this party used to get its largest shares of votes and had won the previous elections despite local voters being predominantly in favour of ‘Leave’.

Regardless of the explanatory variables used, the spatial lag is significant in all models. Therefore, the spatial regression models provide a more accurate prediction of the electoral outcomes.

8. Discussion

After the 2019 elections, some political pundits imputed Labour’s disastrous defeat to the Brexiteers’ vote. In the 2017 election, however, despite the Brexit thorn, Labour had obtained an honourable defeat, increasing its vote share by 9.6% compared to the 2015 elections. In the 2019 elections, when the Brexit implementation became dominant in the Conservative agenda, the association between Brexiteers and Labour’s losses became closer. Moreover, those losses were higher where groups that had been traditional supporters of Labour but had manifested an aversion to supranational integration—such as manual and elementary occupations workers and unemployed people—were larger.

True enough, Labour’s losses could have been compensated by the votes of groups not particularly close to this party in the past but supportive of supranational integration. However, the shifting vote of professionals and degree-holders benefitted more the Liberal Democrats, whose 2019 manifesto was staunchly in favour of cancelling Brexit, than Labour, which had supported ‘Remain’ rather lukewarmly.

The scenario associated with the immigration issue mirrors the above. As we gathered from BSA data, the opposition to foreign immigrants was widespread among poorly educated people and manual and semi-skilled workers. Accordingly, in the 2019 elections, Labour’s losses were correlated with the shares of groups largely opposed to immigration: the party registered its highest losses in constituencies with higher shares of ‘process-plant-machine-jobs × no-qualification’ people.

Significant losses for Labour—and corresponding gains for Conservatives—were registered as well in constituencies with higher shares of people less favourable to multiculturalism, such as natives and white-ethnicity people regarding themselves as having British identity. As a confirmation of this, there was, concurrently, an inverse—though less closely correlated—flow of votes to Labour’s benefit (and to the Conservatives’ disadvantage) from constituencies with higher shares of Muslims, black-ethnicity population, and born-abroad people. To complete the picture, the regression analyses found that, at parity of other factors, a larger share of foreigners increased Labour’s losses, as if voters resented the immigrant presence. This point suggests that aversion to immigration played in the shifting vote a role not overlapping with hostility to supranational integration.

As to the third issue differentiating the parties during the electoral campaign, criminal policy, we already noticed the usual association between Labour’s vote and crime rates. This association is hardly surprising: Labour has traditionally obtained larger shares of votes in disadvantaged areas, where the educational and income levels are lower, the share of benefit claimants higher, and common crime endemic.

However, the correlation between crime rates and Labour’s votes does not imply that Labour’s electors were happy about crime: those worried about crime in their areas were more numerous among Labour’s electors than Conservatives’. Concurrently, most of Labour’s electors called for stiffer sentences for the lawbreaker. In the 2019 elections, when the security issue rose in importance in the Conservatives’ manifesto, Labour lost votes to Conservatives in high-crime constituencies. Meaningfully, the high-crime areas coincided with those where people were most worried about crime and inclined to harsher penalties against delinquents. All this suggests that voters in high-crime constituencies got attracted to the party promising a stricter policy on crime.

Ultimately, we can say that to attribute Labour’s defeat to its stance on Brexit would be reductive. Other factors as well have surfaced in this study as influential regarding these losses.

Going back to the hypotheses advanced earlier in this paper, we can summarise the present findings as follows. First, there is extensive evidence that social groups traditionally pro-Labour had long shared stances in tune with the new hot issues introduced and highlighted in the 2019 Conservative manifesto. People sharing those stances tended to shed their former political alliance and cast a vote for the party that took a firm position in favour of those stances. Regression models revealed a significant statistical association between the territorial distribution of social groups sharing the abovementioned stances and the vote shift from Labour to the Conservatives. Further macro features of political relevance (foreign immigrants’ share and crime rates) were also associated with the vote shift. Finally, the spatial dependence of the vote shift provided additional evidence that the electoral outcomes were located social facts, namely facts significantly depending upon territorial determinants.

All this translated into an overall electoral outcome that presented paradoxical sides. The Labour Party’s programme for the 2019 elections was far left compared with its previous programmes since the 1990s. Labour’s proposals to nationalise services, such as gas, electricity, water and transportation, spend more on welfare, and fight poverty were expected to be highly appealing to the party’s traditional supporters. However, our results show that Labour unexpectedly scored its larger losses in constituencies where social groups regarded as left-wingers were more numerous: particularly where the share of poorly educated, working-class people was higher. Labour’s 2019 losses (and Conservatives’ gains) were heftier in traditional Labour bastions, where the party used to win and obtain its highest shares of votes.

The UK 2019 elections provide significant scope for reconsidering some general issues of the political science debate. Firstly, the large and rapid shift of traditional left-wing voters is at odds with the idea of partisanship as social identity. That shift can be traced back to instrumental, reactive partisanship. However, the referendum-voting model—voters’ reactions to government performances—does not seem to have been momentous in the vote shift. Instead, more momentous seem to have been the voters’ reactions to new political stances, particularly stances implying the overcoming of the traditional left/right contraposition, primarily centred on economics.

In the 2019 elections, Labour’s markedly left-wing programme bet on the expected economic reformism of the working classes. Economic reformism is a topic on which Labour has enjoyed a monopoly; however, it is also a topic with a class-bound character. In contrast, Conservatives chose an issue competition. While keeping a hold on political independence and law-and-order topics, near and dear to the party’s habitual supporters, Conservatives rebranded them as hot issues. Moreover, for each of their hot issues belonging to the ‘new politics’ dimension—national oneness vs supranational integration, immigration regulation, and security—they suggested they had policies desirable for most voters, no matter their partisan affiliation. By doing this, the Conservative Party succeeded in getting across to a larger electorate while hanging onto its vote bank.

The dramatic swing of working-class support from the long-established left-wing party cannot be attributed to either structural or cultural changes affecting that class. Structural changes can explain a long-term trend, but they fit with neither the ups and downs we noticed in the left-wing parties’ vote, nor the large and rapid shift recorded in the UK 2019 elections. Moreover, it is hard to identify any major socioeconomic change between 2017 and 2019: for instance, the cross-constituency distribution of unemployment was substantially stable, and its median value decreased from 4.1% to 3.9%. Cultural changes, too, are at odds with, and unable to explain, an abrupt vote shift over a two-year period. The findings of this research suggest that the lower classes’ support for the workers’ party ebbed away because those classes, while being reformist as to the economy, often hold, as shown here, more conservative attitudes than the middle and higher classes on issues such as internationalism/nationalism, the acceptance of foreign cultures and people, and the punishment of delinquents. These conservative attitudes are not new: they have characterised those classes since long ago. Therefore, the political change did not concern the demand-side: it concerned the supply-side. The new hot issues highlighted by Conservatives in the 2019 campaign resonated well with those long-lasting conservative attitudes of the classes representing Labour’s traditional supporters.

The crucial aspect of the elections was the Conservatives’ successful appeal to the sociocultural conservativism of a relevant part of Labour’s traditional electorate. Such an appeal prevailed over Labour’s concurrent appeal to the economic reformism of its supporters.

9. Limitations and Scope for Further Research

As mentioned in the Methods section, in this research we intended to go beyond typical analyses of election outcomes based on macro data. Macro data often lead to inferences affected by ecological fallacies, namely spurious associations between electoral outcomes and territorial features. We tried to avoid this trap by combining the findings provided by macro analyses with the information from individual data, which, in turn, unfortunately, ignore the territorial features’ impact on electoral outcomes. Our macro and micro data, however, came from different datasets. This fact made all the analyses more difficult and posed limitations. A better understanding of the determinants of electoral outcomes could be achieved in the future when there may be integrated datasets: namely, datasets containing the voters’ characteristics, opinions, and political choices, as well as the essential features—electoral results included—of the territorial units to which voters belong. Integrated datasets would allow the use of multilevel statistical models, which can simultaneously handle individual and territorial data to predict political choices.