The Downfall of Boris Johnson: Analysing Patterns of Opinion among Conservative Parliamentarians in the Confidence Motion and Mass Resignations

ABSTRACT

This article provides the first systematic analysis of the confidence motion against the Conservative Party leadership of Boris Johnson in June 2022 and the mass ministerial resignations in July. I have constructed a dataset of the parliamentary Conservative Party and used logistic regression to test hypotheses relating to Conservative parliamentarians’ behaviour in the confidence motion and ministerial resignations. My findings demonstrate that the long-running divide within the Conservative Party on Europe was statistically significant in the confidence motion, with MPs who voted leave in the 2016 EU referendum being more likely to have confidence in Johnson. Ministers’ university education was also significant in driving support for Johnson. My research findings contribute to existing academic debates as they show how Conservative MPs’ behaviour in the downfall of Boris Johnson compares with the selection and ejection of other Conservative leaders. I also contribute to emerging discussions on Johnson’s leadership of the Conservative Party within the context of governing degeneration.

KEYWORDS:

• Confidence motions
• British Conservative Party
• Prime Ministerial leadership
• Parliamentary behaviour
• Boris Johnson

Introduction

This paper contributes to our understanding of how British political leaders can be removed by formal procedural measures such as confidence motions, and by informal measures in the shape of internal party pressures. This contributes to the field of academic literature on leadership ejections which is a subset of the literature on British political leadership (see Heppell & Hill, 2008, 2010; Jeffery et al., 2018, 2022; McAnulla, 2010; Quinn, 2012; Roe-Crines et al., 2021). The focus of this study is twofold: first is the ballot held amongst Conservative parliamentarians in June 2022 asking whether they still had confidence in the incumbent party leader, Boris Johnson, and second is the mass ministerial resignations in July 2022. The pressure on Johnson came as a result of his poor handling of the Chris Pincher scandal, a Conservative MP who was the subject of allegations of sexual misconduct. The Pincher scandal followed allegations broken in the print media in November 2021 which documented parties attended by Johnson that contravened his own Covid lockdown rules. Discussions of these scandals have been effectively covered elsewhere (Stafford, 2022; see also Bale, 2023 and Seldon & Newell, 2023) and as such, it is not the intention of this paper to chart the sequence of events that defined the ‘partygate’ and Pincher scandals. I note that these scandals damaged the reputation of Johnson and his administration (Judge, 2022; Sanders, 2023). Walker (2023) has provided insights into the mass resignations, providing an ethically and individually driven explanation for why some ministers demanded Johnson’s resignation. This paper adds to Walker’s work by contextualising the wider parliamentary party that connects the mass resignations and confidence motion. This therefore leads to the research question for this paper: can we identify any ideological, personal, or electoral drivers that enable us to argue that clear patterns of opinion existed towards Johnson within the parliamentary Conservative Party (PCP)?

The confidence motion and ministerial resignations are the two key events that fatally destabilised the Prime Minister who had won an 80 seat majority less than three years earlier (Ford et al., 2021). Any attempt by some MPs to oust a sitting party leader is worthy of academic attention, particularly when that leader is also the sitting Prime Minister. Johnson won the confidence motion with the support of 211 MPs against 148 who did not have confidence in him, but the size of the rebellion against his leadership after a difficult period further destabilised his position. Despite his bluster as his ministerial colleagues abandoned him en masse, this loss of support left him no choice but to announce his resignation in July 2022.

This paper proceeds as follows. The first section offers an overview of the academic literature on Conservative Party leadership selection and ejection and on wider comparative perspectives on leadership change. The second section identifies the ideological, electoral, and personal hypotheses that have been selected and the reason for their selection, followed by data collation. The third section provides an overview of my findings, which feeds into my analysis and conclusions where I examine the significance of my findings and how our new appreciation of Johnson’s downfall deepens our understanding of Conservative parliamentarians and the Conservative leadership.

Existing Academic Literature on Conservative Party Leadership Change

The introduction of internal democracy for the selection of the leader of the Conservative Party in 1965 did not initially attract significant academic interest in leadership selection and ejection. Studies of Margaret Thatcher’s forced exit in 1990 (Alderman & Carter, 1991; Jesse, 1996) marks the start of a substantial body of literature dealing with Conservative leadership change. Subsequent papers on the election of her successor, John Major (Cowley, 1996; Cowley & Garry, 1998), stimulated reappraisals of her rise to the party leadership in 1975 (Cowley & Bailey, 2000; Wickham-Jones, 1997). The Conservatives’ propensity for leadership campaigns triggered a series of papers offering qualitative appraisals of Major’s re-election in 1995 (Alderman, 1996), William Hague’s victory in 1997 (Alderman, 1998), Iain Duncan Smith in 2001 (Alderman & Carter, 2002), David Cameron in 2005 (Denham & Dorey, 2006), and Theresa May in 2016 (Quinn, 2019). Running alongside these qualitative studies have been appraisals of the rules governing the selection and ejection of Conservative leaders. The significance of Hague’s reforms in 1998 that introduced membership participation and replaced the challenger provision with the confidence motion has been explored (Alderman, 1999; McSweeney, 1999; see also Quinn, 2005).

This paper is not positioned within either of these two traditions of summarising specific campaigns or assessing the rules by which they were conducted. Rather, it continues the tradition of quantitatively exploring the factors that drove Conservative MPs’ voting behaviour in leadership selections and ejections and seeking to identify patterns of behaviour. This tradition can be traced back to the work of Cowley and Bailey (2000) and Cowley and Garry (1998) who re-examined the 1975 and 1990 ballots of MPs respectively, looking for patterns of opinion amongst MPs. This methodological approach of constructing a dataset of the preferences of the PCP and testing hypotheses relating to ideological, personal, and electoral variables has been carried forwards in various studies of leadership elections since 1997. The elections of Hague and Duncan Smith have been studied by Heppell and Hill (2008, 2010). Jeffery et al. (2018, 2022, 2023) have studied the elections of Theresa May, Boris Johnson, and Liz Truss. Most recently, Booth et al. (2023) studied the election of Rishi Sunak. The quantitative approach has also been used by Roe-Crines et al. (2021) in their study of the confidence motion against May in 2018. These studies have demonstrated recurring drivers in terms of voting motivations. For example, ideology has been a constant driver of the vote, as we see an ongoing alignment between candidate positioning on European policy and support from likeminded backbenchers, whereas wider personal and electoral drivers have shown greater variabilities in terms of voting impacts (Cowley & Garry, 1998; Heppell & Hill, 2008, 2010; Jeffery et al., 2018, 2022, 2023; Roe-Crines et al., 2021; see also Heppell, 2020).

The body of literature that focuses on British examples of leadership selection and ejection feeds into wider comparative discussions about the merits of leadership change. Whilst much of political science has focused on political institutions, ideologies, structures, and processes, some work exists on the process of leadership (Hartley, 2010). Burns (1978) conceptualised leadership as a process, arguing that leadership is a series of acts that are used by individuals and groups to mobilise support. Work by ‘t Hart and Uhr (2011) contributes to understanding of leadership transition in government. So (2021) argues that challenges to leaders – if caused directly by challengers or confidence motions or indirectly by resignations – are damaging to governing political parties, demonstrating personal and ideological divides within the party and damaging their reputation for competence. This undermines their electoral appeal and chances of re-election (see also Bynander & ‘t Hart, 2008; Horiuchi et al., 2015; for challenges to this perspective see Fernandez-Vasquez & Somer-Topcu, 2019; Somer-Topcu, 2017).

This paper embraces the quantitively driven approach in order to determine whether any patterns of opinion can be identified in relation to those Conservative MPs who lost (or retained) confidence in Johnson. The research findings of the paper explain patterns of opinion in relation to confidence in Johnson and can be considered in relation to the research findings of similar studies of previous leaders. In doing so, this paper will reveal much about the contemporary Conservative Party, as it faces the challenges of governing degeneration that have engulfed long-serving post-war British administrations (see Heppell, 2008b; Roe-Crines, 2022).

Variables, Hypotheses and Data Collation

Below I provide details on the range of variables included in this analysis, the hypotheses I test (ideological, personal, and electoral), and the construction of my dataset. These hypotheses incorporate previous research in this field, but where necessary have been developed in recognition of the unique circumstances in June 2022. The hypothesis testing to establish what drove support for Johnson is done in two parts. First, the hypotheses are tested on backbenchers in the confidence motion. Frontbenchers have been excluded from this; as they were on the government payroll they were obliged to support Johnson (and they overwhelmingly did so) and so their inclusion would skew the results. Second, the hypotheses are tested on the frontbenchers in their slew of resignations.

Model 1: Ideological Hypotheses

The ideological divide on Europe within the Conservative Party has evolved from Europhiles vs. Eurosceptics to soft Eurosceptics vs. hard Eurosceptics (Garry, 1995; Heppell, 2013), although in 2022 there are still a number of Conservative MPs who do not subscribe to Johnson’s Euroscepticism. Johnson was a leading figure in the 2016 EU referendum campaign and made Brexit central to his premiership, a position that I hypothesise will see him win greater support from the leave wing of the PCP. This would mark a continuation of the significance of Brexit in Johnson’s victory in the 2019 leadership election (Jeffery et al., 2022), although in the 2018 confidence motion against Theresa May’s leadership, Roe-Crines et al. (2021) found that an MP’s position in the EU referendum was not significant. The European divide was statistically significant in the 1997, 2001, and 2016 leadership elections (Heppell & Hill, 2008, 2010; Jeffery et al., 2018). There are several options for defining an MP’s position on Brexit – e.g. votes in the 2016 referendum itself, voting in various stages of the EU withdrawal agreement, or public statements (for voting in the 2017–2019 parliament see Aidt et al., 2021; Hayton, 2022a; Woollen, 2022; and Xu & Lu, 2022). However, voting behaviour in the 2017–2019 parliament might have been tactical, rather than ideological, as MPs sought to control the agenda and navigate a series of indicative and meaningful debates (see Lynch et al., 2019 and Russell & James, 2023). I overcome this challenge by defining each MP’s position by how they voted in the referendum, presenting us with a binary choice that was a matter of principle. Therefore, the hypothesis for the Brexit divide is as follows:

H1 Brexit: MPs who voted leave in the 2016 referendum will be more likely to support Johnson.

I also include non-Brexit ideological variables that relate to some of the agendas that have defined Johnsonian Conservatism (see Gamble, 2021; Hayton, 2021; Peele, 2021). The variables here are positioned within the context of different groups within the PCP which feeds into emerging contemporary debates on factionalism within the Conservative Party (for a history of Conservative Party factionalism see Baker et al., 1999; Heppell, 2002; Rose, 1964; Seyd, 1972). The Blue Collar Conservatism group is comprised of MPs who seek to deliver policies that will benefit working class voters, a group that Johnson directed his efforts towards with the levelling up agenda and some socially conservative rhetoric. The Common Sense Group are a socially conservative ‘anti-woke’ group who are close to Johnson on cultural issues and rhetoric. The distinction between social liberals and conservatives was significant in the 1997, 2001, and 2016 leadership elections (Heppell & Hill, 2008, 2010; Jeffery et al., 2018). The European Research Group are the Praetorian Guard of Conservative Eurosceptics and have long supported Johnson’s attempts to leave the EU. I therefore hypothesise that the ideology of these three groups will affect their voting behaviour in the confidence motion in the following ways:

H2 Blue Collar Conservatism: MPs who are members of Blue Collar Conservatism will be more likely to support Johnson.

H3 Common Sense Group: MPs who are members of the Common Sense Group will be more likely to support Johnson.

H4 European Research Group: MPs who are members of the European Research Group will be more likely to support Johnson in the confidence motion.

Parliamentarians’ responses to the Covid measures introduced by Johnson’s government prompts another ideological hypothesis. There was some division on this indicator; sceptics towards the measures felt that they infringed on personal freedom (for wider literature on public opinion see Mellon et al., 2021). Many of these sceptics were aligned with the Covid Research Group but the lack of publicly available membership data leads us to 24 votes which saw rebellions. These votes, to varying extents, granted the government powers to introduce measures to manage the Covid pandemic and were central to Johnson’s thinking on Covid. This is a new divide within the PCP and I hypothesise that MPs’ opposition to Johnson’s Covid restrictions is a signal of a deeper ideological divide that will play out in the confidence motion:

H5 Covid restrictions: Backbenchers who showed greater rates of rebellion on Covid measures will be less likely to support Johnson.

Model 2: Personal Hypotheses

I include three hypotheses which are influenced by wider perceptions of Johnson as an Eton – and Oxford-educated elite and the criticisms of his macho persona (The Guardian, 2020; The Independent, 2021). The private and Oxbridge education variables were tested by Jeffery et al. (2018) in the 2016 leadership election, although they were not found to be significant, and gender was tested for the 2016 and 2019 leadership elections and 2018 confidence motion, and again not found to be significant (Jeffery et al., 2018, 2022; Roe-Crines et al., 2021). I contend that MPs who share Johnson’s elite educational background and gender will be more likely to have confidence in him.

H6 Private school: MPs who attended private schools will be more likely to support Johnson.

H7 Oxbridge: MPs who went to Oxbridge universities will be more likely to support Johnson.

H8 Gender: Female MPs will be less likely to support Johnson.

Model 3: Electoral Hypotheses

In light of Johnson’s unpopularity in the polls – he went from a net approval rating of +1% in May 2021 to −42% in May 2022 (YouGov, 2022) – I consider an MP’s electoral vulnerability and construct hypotheses relating to constituency marginality. MPs with larger majorities might be more comfortable with an unpopular leader as their seats are fairly safe, and so could be expected to be more inclined to back Johnson. Heppell and Hill (2008) tested electoral vulnerability in the 1997 leadership election but found that it was not significant, as did Jeffery et al. (2018) in the 2016 leadership election and Roe-Crines et al. (2021) in the 2018 confidence motion. I also test whether MPs who were first elected when Johnson was leader would be more loyal to him in the confidence motion – these MPs might attribute their success to Johnson. Similar cohort-style variables have been tested for the 1997 and 2016 leadership elections (Heppell & Hill, 2008; Jeffery et al., 2018). Finally, in the 2019 general election and throughout his premiership Johnson sought to exploit conflict over identity and values with particular effect in ‘Red Wall’ seats (Hayton, 2022b). MPs in these seats may feel a particular loyalty to Johnson or feel that his continuance as Conservative leader would bolster their own chances of re-election and therefore be more likely to have confidence in him. The concept of Red Wall seats emerged in the wake of the 2019 general election and refers to a group of 42 constituencies, in traditional Labour heartlands in the midlands and north of England, that were likely to vote Conservative for the first time ever, or for the first time in recent history (Kanagasooriam & Simon, 2021). This marks the hardening of a long trend in British electoral politics. The Red Wall variable was found to have some significance by Jeffery et al. (2023) and Booth et al. (2023) in the selection of Truss and Sunak respectively.

H9 Majority: MPs with larger majorities will be more likely to support Johnson.

H10 MPs elected under Johnson: MPs who gained their seat under Johnson will be more likely to support him.

H11 Red Wall MPs: MPs with Red Wall seats will be more likely to support Johnson.

Data Collation

365 Conservatives were elected at the 2019 general election, but 359 participated in the confidence motion in June 2022. Two of the MPs elected in December 2019, James Brokenshire and David Amess, passed away before the confidence motion in June 2022 and were replaced by Louie French and Anna Firth. After the death of Cheryl Gillan and the resignation of Owen Paterson two Conservative seats were lost to the Liberal Democrats in by-elections (which takes us down to 363 MPs). There was a gain from Labour after Jill Mortimer won a by-election (which takes us up to 364) but the resignations of Imran Ahmed Khan and Neil Parrish takes us down to 362. Christian Wakeford’s defection to Labour takes us down to 361. Two MPs, Rob Roberts and David Warburton, had the whip suspended and were not allowed to take part in the confidence motion. This gives us the 359 Conservative MPs who participated in the confidence motion. I ascertained the ministerial status of each MP, splitting the PCP into backbenchers and frontbenchers (numbering 212 and 147 respectively). The frontbench includes cabinet ministers, junior ministers, party officials, and parliamentary private secretaries. The Spectator (2022) provides details of the resignations from these ranks. The public nature of the frontbenchers’ decisions to resign (or not) means that I have ascertained the position of 100% of the ministerial team during the period of mass resignations.

I was able to construct a robust list of MPs’ positions in the confidence motion. For the purpose of this data analysis I focused on the actions of backbenchers. Whilst the result of the confidence motion was made public, the secret ballot presents challenges. The first MPs whose position I can ascertain with a reasonable degree of confidence are MPs who publicly stated that they had submitted a letter of no confidence in Johnson to Graham Brady. The letters of no confidence in the party leader to the Chairman of the 1922 Committee form part of the system of removing the party leader: when the threshold of 15% of the parliamentary party submitting letters is met, a vote of confidence is held in the leader (Norton, 2023). The next MPs whose position I can ascertain are those who did not call for a motion, but once one was announced publicly declared their intention one way or another. These were identified through the use of public statements in the media, on social media, on their own websites, or as reported in reputable newspapers (both national and local). This list was then cross-checked with the lists compiled by national media outlets – The Telegraph’s (2022) list was particularly useful. Through this method I have ascertained the position of 159 backbench MPs (75% of the backbench electorate). There are of course expedient reasons why an MP might not want to declare their lack of confidence in Johnson, particularly as it looked probable that he would win the motion. MPs on the government payroll were also bound by collective duty to state their public support for the leader, but what they did in the privacy of the ballot booth might contradict their espousals of loyalty. Of the backbench MPs whose position I was unable to ascertain with confidence 18 publicly refused to state their preference, many of whom served in party positions that were sensitive to the motion and so considered it improper to publicly declare their intentions (such as Graham Brady who was the Chair of the 1922 Committee). My success rate of 75% for the backbenchers’ positions during the confidence motion compares favourably with the 72.2% achieved by Jeffery et al.’s analysis of the 2019 leadership election (2022, p. 122) and is high enough to allow for rigorous testing of my hypotheses.

The independent variables were collated and coded as follows. Brexit positioning was ascertained through public statements in the media and on social media, and lists compiled by reputable newspapers and periodicals. The groups within the PCP included in this analysis have membership information online. Blue Collar Conservatism (2022) has its own website with a list of MPs who are supporters. Members of the European Research Group (ERG) have been identified through the expenses records published by the Independent Parliamentary Standards Authority (2023) cross-referenced with various sources: Open Democracy (2019), BBC News (2018), and an open letter from some ERG members to Donald Tusk (Tomlinson, 2016). The Common Sense Group has been identified through MPs’ websites and public statements. The Covid hypothesis is based on voting in 24 Covid-related divisions (see Appendix Table A1 for a list of the divisions used) where there were Conservative rebellions (data are taken from the UK Parliament website). An MP’s score for rebelliousness is calculated as a percentage of the 24 divisions in which they rebelled.

Schools data were broken into four categories (Eton, private, grammar, and state) with the information being found on MPs’ personal websites and interviews and reports in the media. MPs’ university type was broken down into four categories: no university, a non-Russell Group university, Russell Group university, and Oxbridge. The grading of universities might provoke some dissent from within the academic community – this is not intended to be a judgment on the merits of different universities. The aim here was to expose the potential for a social and educational hierarchy where graduates of the elite Oxbridge universities might be expected to vote differently to graduates of less prestigious universities. Where an MP attended more than one university, they were entered into the category of the highest prestige. Data for this was gathered from various news sources, MPs’ personal websites, and MPs’ LinkedIn profiles. The male/female variable is straightforward.

The marginality variable uses data taken from the 2019 general election, with the MP’s majority over the second-placed candidate being used. Some MPs in the dataset were elected at by-elections since the 2019 general election but the result from 2019 was used to maintain consistency in the variable definitions and avoid the odd results that by-elections, by virtue of their isolation in between electoral cycles, can produce. The only exception is the Hartlepool by-election of 2021 where Jill Mortimer took the seat from Labour who had won it in 2019. General election data provides all information regarding results and seats gained and held (House of Commons Library, 2020). Finally, the Red Wall seats have been identified from research by Kanagasooriam and Simon (2021).

Research Findings

My research findings are presented in six tables showing the output of logistic regression models with the coefficients (β) shown as log odds. To demonstrate how to interpret the coefficients in the second column of the tables it is helpful to give an example using the EU referendum variable in Table 1: ‘for every one-unit increase in the EU referendum (i.e. when an MP supported leave), the log odds (likelihood) of an MP supporting Johnson in the confidence motion increases by 0.961 when holding all other variables in the model constant’ (Table 2).

Table 1. Logistic regression output of Model 1 (ideological variables) – Backbenchers in the confidence motion.

N	129
Prob > chi^2	9.892
Pseudo R^2*	0.074			Collinearity statistics
Variable	Β	Standard error	P-value	Tolerance	VIF
EU Referendum	0.961	0.416	0.021**	0.812	1.232
Blue Collar	−0.369	0.401	0.357	0.859	1.164
Common Sense	−0.396	0.66	0.549	0.924	1.082
ERG	−0.037	0.584	0.949	0.922	1.085
Covid Restrictions	1.531	1.149	0.183	0.913	1.095
Constant	−0.939	1.439	0.514

*Cox & Snell.

**≤0.05.

Table 2. Logistic regression output of Model 1 (ideological variables) – Ministerial resignations.

N	134
Prob > chi^2	0.5331
Pseudo R^2*	0.039			Collinearity statistics
Variable	β	Standard error	P-value	Tolerance	VIF
EU Referendum	0.611	0.406	0.133	0.85	1.177
Blue Collar	−0.212	0.406	0.602	0.914	1.094
Common Sense	0.809	0.73	0.268	0.945	1.058
ERG	−0.579	0.608	0.341	0.968	1.033
Constant	−1.484	1.524	0.33

*Cox & Snell.

Model 1: Ideological Hypotheses

When it comes to backbench support for Johnson in the confidence motion, the logistic regression output shows that only significant ideological variable was an MP’s position in the 2016 EU referendum. MPs who supported leave in the referendum were more likely to have confidence in Johnson’s leadership which, perhaps unsurprisingly, supports H1 on Brexit. This reflects Johnson’s ongoing position as a figurehead of the leave side and his efforts to ‘Get Brexit done’ in parliament. The continued salience of Brexit within the PCP demonstrates the continued power of this issue for many Conservative MPs – even after the UK had left the EU Brexit attitudes continued to play an active part in Conservative leadership support and in support for the post-Brexit direction that Johnson wanted to move the party towards.

The other three variables in the ideological model all relate to an MP’s membership of various groups within the PCP: Blue Collar Conservatism, the Common Sense Group, and the European Research Group. There was no evidence to suggest that membership of these groups was significant in the 2022 confidence motion and so we cannot accept H2, H3, or H4. The lack of significance of these groups in the confidence motion might be explained in three ways. Firstly, these groups may not be very ideologically homogenous, and so the variation of their members might undermine cohesive voting behaviour in the confidence motion. These groups might not be an effective vehicle when it comes to forming a cohesive voting bloc in party matters, Secondly, using an MP’s membership of group(s) within the PCP as a signal of their underlying ideology might not be effective. Identifying an MP’s underlying ideological identity is challenging and the link between this and membership of these groups might be imperfect. Finally, and the most likely explanation for the insignificance of these groups in the confidence motion, is that these particular ideological issues were less important than other factors – voting was simply not along these particular ideological lines.

When it comes to behaviour in the ministerial resignations, none of the ideological variables were statistically significant and so we cannot accept H1, H2, H3, H4, or H5. Ideology does not appear to have been a motivating factor when a frontbencher considered their position amidst the troubles of Johnson’s government in spring 2022. Frontbenchers are chosen partly for their ideological affinity with the leadership, reducing the chance that ideology will drive a wedge between the leader and their top team.

For the three variables relating to the groups in this model, there is a possibility that some bias may be present. It may be the case that MPs who are less supportive of Johnson will have been kept on the backbenches and MPs who are more supportive promoted to the ministerial ranks. To test this I have created a new model that includes interaction terms for membership of Blue Collar Conservatives, the Common Sense Group, and the European Research Group, and backbench or frontbench status (see Table 3). The output suggests that whilst backbench or frontbench status had some effect here, this is not statistically significant. Therefore we do not need to be concerned about bias in this case.

Model 2: Personal Hypotheses

In Model 2 for the backbenchers’ behaviour in the confidence motion none of the variables (school, university, and gender) were statistically significant. We therefore cannot accept H6, H7, or H8. Johnosn’s elite Eton and Oxford educational background did not appear affect voting behaviour in the confidence motion. The hypermasculinity of Johnson’s leadership identified by Waylen (2021) and the hypermasculine leadership dynamics regarding the partygate scandal identified by Sanders and Richards (2022) do not appear to have played out in the confidence motion. Johnson’s masculine style of leadership does not seem to have resulted in a ‘women problem’ for him within the PCP.

In Model 2 for the frontbenchers’ behaviour in the ministerial resignations only university type was statistically significant. We therefore accept H7 and conclude that ministers who went to Oxbridge were more likely to support Johnson. The ‘old boys’ network’ effect (see Watters, 2016) might have come into play here, with well-connected graduates of Britain’s oldest universities protecting one of their own. School type and gender were not significant in explaining behaviour in the ministerial resignations and so we cannot accept H6 or H8 (Tables 4 and 5).

Table 4. Logistic regression output of Model 2 (personal variables) – Backbenchers in the confidence motion.

N	127
Prob > chi^2	3.27
Pseudo R^2*	0.025			Collinearity statistics
Variable	β	Standard error	P-value	Tolerance	VIF
School	−0.255	0.386	0.509	0.953	1.049
University	0.212	0.193	0.272	0.952	1.05
Gender	−0.727	0.559	0.193	0.999	1.001
Constant	0.227	0.936	0.808

*Cox & Snell.

Table 5. Logistic regression output of Model 2 (personal variables) – Ministerial resignations.

N	138
Prob > chi^2	6.573
Pseudo R^2*	0.047			Collinearity statistics
Variable	β	Standard error	P-value	Tolerance	VIF
School	−0.149	0.375	0.692	0.941	1.062
University	0.443	0.182	0.015**	0.937	1.067
Gender	0.13	0.383	0.735	0.991	1.009
Constant	−0.1729	0.844	0.04

*Cox & Snell.

**≤0.05.

Model 3: Electoral Hypotheses

None of the three electoral variables I tested were statistically significant in explaining voting behaviour in the 2022 confidence motion or ministerial resignations. We therefore fail to accept H9, H10, and H11 (concerning an MP’s majority at the 2019 general election, entry to parliament under Johnson, and Red Wall seats). The insignificance of an MP’s majority at the 2019 general election suggests that individual electoral concerns were not a priority in the decision to retain Johnson. MPs with more vulnerable seats (i.e. those with smaller majorities) were no more likely to withhold support from Johnson than their colleagues with safer seats. That is not to say that electoral concerns were not a factor – polling on the electoral chances of the party as a whole may have been in many MPs’ minds during the confidence motion and ministerial resignations – but individual electoral (in)security does not seem to have driven voting behaviour. The next two hypotheses on entry to parliament under Johnson and holding a Red Wall seat were based on the prediction that these MPs might feel some sense of loyalty to Johnson or believe that their electoral chances would be improved with him at the helm during the next general election. These hypotheses did not play out as expected. Loyalty is fickle in politics and for MPs who feel no personal connection to Johnson or indebtedness for their first electoral success under his leadership cannot be relied upon when it comes to safeguarding his leadership. The insignificance of the Red Wall variable can be explained in two ways. Firstly, MPs might not assess Johnson as a big vote winner in their Red Wall constituencies. Secondly, these MPs might not feel the obligation to reflect their constituents’ wishes in their voting at the confidence motion (Tables 6 and 7).

Table 6. Logistic regression output of Model 3 (electoral variables) – Backbenchers in the confidence motion.

N	140
Prob > chi^2	4.322
Pseudo R^2*	0.03			Collinearity statistics
Variable	β	Standard error	P-value	Tolerance	VIF
2019 Majority	0.0	0.0	0.481	0.808	1.238
Entry under Johnson	−0.412	0.461	0.372	0.725	1.379
Red Wall	−1.161	0.857	0.176	0.789	1.267
Constant	1.582	1.086	0.145

*Cox & Snell.

Table 7. Logistic regression output of Model 3 (electoral variables) – Ministerial resignations.

N	147
Prob > chi^2	2.141
Pseudo R^2*	0.014			Collinearity statistics
Variable	β	Standard error	P-value	Tolerance	VIF
2019 Majority	0.0	0.0	0.778	0.81	1.235
Entry under Johnson	0.555	0.408	0.174	0.768	1.302
Red Wall	−0.006	0.669	0.993	0.725	1.378
Constant	−1.413	0.968	0.144

*Cox & Snell.

Comparing the Models

Analysis of the dataset has revealed old and new divisions within the PCP, but none of the models produce a high pseudo R² value (a measure of the goodness of fit of each model). Model 1 (ideological variables) produces a pseudo R² of 0.074 for the confidence motion and 0.039 for the ministerial resignations; Model 2 (personal variables) produces a pseudo R² of 0.025 and 0.047; and Model 3 (electoral variables) produces a pseudo R² of 0.03 and 0.014. With the largest pseudo R² value it appears that Model 1 for the confidence motion is the best fit but it still cannot explain much of the variance in support for Johnson. The other values are all quite low. This reflects other factors that drove behaviour in Johnson’s downfall: the hidden relationships – both personal and professional – within the PCP, pressure from MPs’ local organisations, and MPs’ own assessments of Johnson’s performance. These factors are concealed from quantitative analysis. There are no concerns about multicollinearity in the models: the tolerances are sufficiently high and the variance inflation factors (VIFs) are sufficiently low (see Weisburd & Britt, 2014) (Tables 8 and 9).

Table 8. Logistic regression output of full model (including all variables) – Backbenchers in the confidence motion.

N	117
Prob > chi^2	15.357
Pseudo R^2*	0.123			Collinearity statistics
Variable	β	Standard error	P-value	Tolerance	VIF
EU Referendum	1.008	0.487	0.038**	0.702	1.425
Blue Collar	−0.814	0.467	0.081	0.753	1.327
Common Sense	−0.199	0.711	0.779	0.858	1.166
ERG	0.233	0.66	0.723	0.813	1.23
Covid Restrictions	1.692	1.287	0.189	0.801	1.248
School	−0.759	0.44	0.085	0.896	1.116
University	0.14	0.222	0.528	0.859	1.164
Gender	−0.693	0.697	0.32	0.899	1.112
2019 Majority	0.0	0.0	0.708	0.692	1.446
Entry under Johnson	0.51	0.646	0.43	0.589	1.698
Red Wall	−0.857	1.007	0.395	0.717	1.394
Constant	0.971	2.27	0.669

*Cox & Snell.

**p ≤ 0.05.

Table 9. Logistic regression output of full model (including all variables) – Ministerial resignations.

N	126
Prob > chi^2	12.659
Pseudo R^2*	0.096			Collinearity statistics
Variable	Β	Standard error	P-value	Tolerance	VIF
EU Referendum	0.763	0.443	0.085	0.794	1.26
Blue Collar	−0.213	0.476	0.654	0.754	1.327
Common Sense	0.918	0.862	0.287	0.816	1.225
ERG	−0.613	0.641	0.339	0.894	1.119
School	0.161	0.435	0.712	0.811	1.234
University	0.348	0.204	0.088	0.875	1.143
Gender	−0.221	0.448	0.622	0.924	1.083
2019 Majority	0.0	0.0	0.953	0.754	1.326
Entry under Johnson	0.725	0.506	0.152	0.685	1.46
Red Wall	−0.58	0.904	0.521	0.666	1.501
Constant	−3.01	2.0	0.132

*Cox & Snell.

The full model for the confidence motion reveals that only voting in the EU referendum was statistically significant. This full model has the highest pseudo R² of all the models, suggesting that it is best in explaining support for Johnson. In the full model for the ministerial resignations none of the variables were statistically significant.

Conclusions

The aims of this article were to identify how Conservative MPs voted in the June 2022 confidence motion against Boris Johnson’s leadership, to determine the factors that drove their voting behaviour, and to determine the factors that drove behaviour in the mass ministerial resignations in July. This was only the second time that the confidence motion procedure has been used against a sitting Conservative Prime Minister and was a step towards Johnson’s eventual ejection from office; his fate was sealed by the unprecedented resignation of over 50 of his ministerial team, including 6 Cabinet ministers. My research helps us to better understand the evolving composition and behaviour of the PCP and academic debates about Conservative leadership ejection in the following ways.

First, the European divide within the parliamentary party continues to be relevant. Support for leave in the 2016 referendum was significant in explaining backbench support for Johnson in the confidence motion, following the long Conservative tradition of Europe being a key ideological factor in the party leadership. Attitudes towards Europe partly drove the selection of John Major, William Hague, Iain Duncan Smith, Theresa May, and Johnson (Cowley & Garry, 1998; Heppell & Hill, 2008, 2010; Jeffery et al., 2018, 2022) and was relevant in the 2018 confidence motion against May (Roe-Crines et al., 2021). In spite of the UK’s withdrawal from the EU under Johnson we are still not in the post-Europe era of the Conservative Party (and may never be). Johnson’s position as a leading figure in the 2016 EU referendum and tenacity with the issue since has allowed him to take advantage of a base of support in the Conservative Party. As debates continue about the UK’s ongoing relationship with the EU – Rishi Sunak’s renegotiation of the Northern Ireland protocol for example (HM Government, 2023) – it would be premature to suggest that this issue will not continue to cause divisions within the Conservative Party as it has done for decades already.

Second, the perception of the Conservative Party as an elite-dominated party continues to hold weight as we see ministers who went to Oxbridge being more likely to support the Oxford-educated Johnson. This is an unusual phenomenon in the context of academic study of Conservative leadership selection and ejection; this variable was last found to be significant in the 1990 selection of Major (Cowley & Garry, 1998). Johnson served as President of the Oxford Union (a debating society) whilst an undergraduate, an organisation which has seen a number of senior Conservatives hold positions (Michael Gove, William Hague, and Jacob Rees-Mogg are just a few examples of Conservative politicians who held positions in the Oxford Union). The continued significance of an Oxbridge education speaks to debates around the limited extent to which the party has managed to widen recruitment to its parliamentary ranks and reduce the dominance of an elite educational background. Johnson did not enjoy the same support from Oxbridge-educated MPs in the October 2022 contest to decide Truss’s successor; Booth et al. (2023) found that Oxbridge-educated MPs were less likely to support a possible Johnson candidacy and more likely to support Sunak.

This research has tested ideological, personal, and electoral hypotheses. However, Johnson’s downfall was triggered by questions of morality and ethics in politics which cannot be captured by quantitative methods. This is not the first time that criticism of an ethical nature has brought a politician’s career to an abrupt end. Under John Major, allegations of sleaze in the wake of sex and financial scandals went wider than concerns about the behaviour of individuals and became associated with the Conservatives at a national level, contributing to Major’s downfall in 1997 (Farrell et al., 1998). The 2009 scandal over MPs’ expenses claims caused widespread anger against the political class (Pattie & Johnston, 2012). The partygate and Pincher scandals under Johnson are the latest demonstration of how unethical practice, this time at the heart of the leadership, can cause a backlash and bring a career to a premature halt. The confidence motion and ministerial resignations were fuelled (at least in part) by morality and standards. The quantitative models here aid us in our understanding of Johnson’s downfall but ethics must also form part of the explanation for the end of his leadership.

It is important to locate Johnson’s downfall within the wider context of a party that, by that point, had been in power for 12 years. Heppell (2008a) argued that long serving governments suffer from degenerative tendencies like those suffered by the Conservatives in 1963 and 1996. Roe-Crines (2022) picks up Heppell’s governing degeneration model to find that Johnson’s government showed signs of the malaise. This research strengthens the evidence of degeneration, particularly in the dimensions of leadership credibility, abuse of power, and party divides. Johnson’s handling of the partygate and Pincher scandals badly damaged his approval ratings and perceptions of his ability to either prevent crises or handle them when they occur. With the disastrous premiership of Liz Truss following Johnson, the Conservatives under Sunak must recover from this degeneration if they are to maintain power.

Table 3. Exploration of interaction effects between group membership and backbench/frontbench variable.

N	287
Prob > chi^2	7.205
Pseudo R^2*	0.025
Variable	B	Standard error	P-value
Blue Collar	−0.785	0.749	0.295
Common Sense	−0.484	1.149	0.673
ERG	0.988	0.965	0.306
Blue Collar × Bench	0.263	0.468	0.574
Common Sense × Bench	0.556	0.722	0.441
ERG × Bench	−1.016	0.628	0.106
Constant	0.29	0.703	0.68

*Cox & Snell.

Acknowledgements

The author wishes to thank Tim Heppell and Richard Hayton for their comments on earlier drafts.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Data Availability Statement

The authors confirm that the data supporting the findings of this study are available within the article [and/or] its supplementary materials.

Additional information

Notes on contributors

Oliver Booth

Oliver Booth is a PhD researcher at the University of Leeds.

Appendix

Table A1. List of divisions used for H5 (Covid restrictions).

Division title	Date of division	Number of Conservative rebels
The Health Protection (Coronavirus, Restrictions) (Blackburn with Darwen and Bradford) (Amendment) (No. 2) Regulations 2020	23rd September 2020	1
The Health Protection (Coronavirus) (Restrictions on Holding of Gatherings and Amendment) (England) Regulations 2020	23rd September 2020	2
The Health Protection (Coronavirus, Wearing of Face Coverings in a Relevant Place and on Public Transport) (England) (Amendment) Regulations 2020	23rd September 2020	2
The Health Protection (Coronavirus, Restrictions) (Blackburn with Darwen and Bradford) (Amendment) Regulations 2020	23rd September 2020	1
The Health Protection (Coronavirus, Wearing of Face Coverings in a Relevant Place) (England) (Amendment) (No. 2) Regulations 2020	23rd September 2020	2
The Health Protection (Coronavirus, Restrictions) (No. 2) (England) (Amendment) (No. 3) Regulations 2020	23rd September 2020	1
The Health Protection (Coronavirus, Wearing of Face Coverings in a Relevant Place) (England) (Amendment) Regulations 2020	23rd September 2020	2
Town and Country Planning (Permitted Development and Miscellaneous Amendments) (England) (Coronavirus) Regulations 2020 (S.I., 2020, No. 632) (Motion to revoke)	30th September 2020	1
Coronavirus Act 2020 (Review of Temporary Provisions) Motion	30th September 2020	7
Health Protection (Coronavirus, Restrictions) (No. 2) (England) (Amendment) (No.4) Regulations 2020 (S.I. 2020, No. 986)	6th October 2020	12
Motion to approve the Health Protection (Coronavirus, Restrictions) (No.2) (England) (Amendment) (No. 5) Regulations 2020 (S.I., 2020, No. 1029)	13th October 2020	42
Opposition Motion: Additional Covid-19 Restrictions: Fair Economic Support	21st October 2020	2
The Health Protection (Coronavirus, Restrictions) (North of England, North East and North West of England and Obligations of Undertakings (England) etc.) (Amendment) Regulations 2020	21st October 2020	3
The Health Protection (Coronavirus, Restrictions) (Self-Isolation) (England) Regulations 2020	21st October 2020	1
Health Protection (Coronavirus, Restrictions) (England) (No.4) Regulations 2020	4th November 2020	33
Health Protection (Coronavirus, Restrictions) (All Tiers) (England) Regulations 2020	1st December 2020	53
The Health Protection (Coronavirus, Restrictions) (All Tiers and Self-Isolation) (England) (Amendment) Regulations 2021 (SI 2021, No 97)	10th February 2021	24
Coronavirus Act 2020 (Review of Temporary Provisions) (No.2)	25th March 2021	35
The Health Protection (Coronavirus, Restrictions) (Steps and Other Provisions) (England) (Amendment) (No. 2) Regulations 2021	16th June 2021	49
Draft Health and Social Care Act 2008 (Regulated Activities) (Amendment) (Coronavirus) Regulations 2021	13th July 2021	31
Motion No. 4 – Health Protection (Coronavirus, restrictions) (Self-isolation)	30th November 2021	32
Health Protection (Coronavirus, wearing of face coverings) (England) regulations 2021	30th November 2021	20
Draft Health and Social Care Act 2008 (Regulated Activities) (Amendment) (Coronavirus) (No.2) Regulations 2021	14th December 2021	62
Health Protection (Entry to venues, etc.) SI No 1416	14th December 2021	99