Trusting Truss: Conservative MPs’ Voting Preferences in the (First) British Conservative Party Leadership Election of 2022

ABSTRACT This paper identifies the candidate preferences of Conservative parliamentarians in the final parliamentary ballot and the membership ballot of the Conservative Party leadership election of 2022. We code each parliamentarians’ candidate preference within a wider dataset on the parliamentary Conservative Party (PCP) and test a range of hypotheses covering the following: a). personal factors relating to gender, sexuality, ethnicity and education b). political factors, relating to constituency marginality and type, as well as prior support for Johnson, and c). ideological factors, covering for example, attitudes towards Brexit, cultural debates, levelling up, economic liberalism, net zero and Covid restrictions. Our research findings demonstrate that despite common media narratives at the time of the party leadership election, variables capturing loyalty or opposition to Johnson did not structure patterns of support for candidates in either round. A parliamentarian's position on levelling up, as well as the UK's membership of, and relationship with, the European Union (based on whether they were remainers, leavers, or a member of the European Research Group) were more important in structuring support. However broader measures of ideology, as evidenced through the membership of extra-parliamentary groups, was found not to be important.


Introduction
This paper offers the first academic account of the first of the two Conservative Party leadership elections of 2022, which saw the foreign secretary, Liz Truss, defeat the former chancellor, Rishi Sunak, in the contest to replace Boris Johnson as party leader and prime minister.Truss' premiership was the most short-lived in British political history.Her time in office was marked by a disastrous mini-budget and a farcical response to a Labour Party parliamentary motion to ban fracking.The next day, following a meeting with Graham Brady, the chairman of the 1922 Committee representing Conservative backbenchers, Truss offered her resignation.
That it took Conservative MPs such little time to lose confidence in Truss begs the question of why they shortlisted her as one of two candidates to be put forward to the membership ballot just three months previously.Criticisms that she was unsuited to the premiership due to her personality and political views were raised by her Conservative parliamentary colleagues during the leadership contest, yet still she succeeded.In light of this, our research objective is to identify the parliamentary bases of support for Truss and her rivals, Sunak and third-placed Penny Mordaunt, to better understand the motivations of Conservative MPs in shortlisting Truss.
The leadership election was required following the enforced resignation of Johnson in July 2022 due to the 'party-gate' scandal.The print media ran stories from late 2021 about how Johnson had repeatedly breached the very Covid lockdown restrictions his government had implemented.A police investigation resulted in Johnson being issued a fixed penalty notice for having broken the law in respect of the lockdown guidelines, and he was accused of lying to Parliament when he denied having attended any of these events (BBC, 2022).Although a confidence motion failed to remove Johnson in early June 2022he won the ballot by 211-148 votesmass ministerial resignations in early July 2022 made his position untenable and he was forced to resign.One of the first ministers to resign was Sunak, whilst Truss remained loyal (Walker, 2023).
The leadership election comprised of two stages.The first consisted of a series of eliminative ballots amongst Conservative MPs.The final two candidates proceeded to the second state, a one-member, one-vote ballot of Conservative party members.To enter the contest, potential candidates needed the nominations of 20 members (5.6 per cent) of the 358-strong parliamentary Conservative Party (PCP).Of the eight who met this threshold, four were ministers: Truss was foreign secretary; Mordaunt was international trade minister; Nadhim Zahawi was the chancellor of the exchequer; and Suella Braverman was the attorney-general.Two were former ministers who had resigned their positions in an effort to unseat Johnson: Kemi Badenoch was ex-equalities minister, and Sunak was the erstwhile chancellor of the exchequer.The remaining two candidates were Jeremy Hunt, the former foreign secretary whom Johnson defeated in the final round of the 2019 leadership election, and Tom Tugendhat, chair of the Foreign Affairs Select Committee and the only candidate without any ministerial experience (Maidment & Penna, 2022).
Table 1 summarises the eliminative parliamentary ballot stages, identifying who was eliminated in each round, either as the last placed candidate or because they failed to pass pre-determined performance thresholds, which resulted in Truss and Sunak being presented to the Conservative membership.After a six-week campaign, Truss won the election with 80,326 votes to Sunak 's 60,399 (57.4-42.6 per cent, on a turnout of 82.6 per cent) (Johnston, 2022).
Using an original database covering a range of personal, political and ideological variables, we find that, contrary to media narratives, perceived loyalty to Johnson had no effect on MPs' support for the candidates vying to replace him.Instead, the results show evidence of both old and new ideological dividing lines within the PCP structuring support.The issue of Europe continues to be salient, with remainers backing Sunak, leavers backing Mordaunt and hard-line leaversmeasured by membership of the European Research Groupbacking Truss, again highlighting that Eurosceptic Conservative MPs cannot be considered a monolithic bloc (Jeffery et al., 2022), and that there are important and persistent differences between different leave-supporting MPs.We also find evidence that social liberalism and support for Johnson's flagship policy of levelling up structure bases of support for candidates within the PCP.Alongside these ideological factors, we find that much of the variance in MPs' support for Truss, Sunak and Mordaunt is unexplained, emphasising the importance of personal relationships and judgements in motivating support for different leadership candidates.
Our paper opens with our research design and methods section, where we identify why we selected our personal, political and ideological variables of interest and outline how we constructed our dataset.We then present our research findings and position these within the wider academic literature on political parties and political leadership in our analysis and conclusion section.

Research Design and Methods
The first Conservative Party leadership election took place in 1965.Before then, the leader typically 'emerged' based on a consensus among the 'magic circle' of senior Conservative figures.Since the first contest, which saw Edward Heath defeat Reginald Maudling and Enoch Powell, analysis of MPs' voting behaviour in these contests has been an important area of study for scholars of the Conservative Party.Generally, ideological divisions have significantly affected Conservative MPs' support for leadership candidates (Cowley & Bailey, 2000;Heppell & Hill, 2008;2009;2010;Jeffery et al., 2018;2022) with other demographic variables proving significant when there has been a stark contrast between the demographics of the main leadership contenders (Cowley & Garry, 1998).
Given the importance of ideology in structuring support in leadership contests, we wanted to increase the emphasis on the attitudes that Conservative parliamentarians held as opposed to their behaviour, which is normally measured through their voting records.Voting records are problematic due to the whipping system which pressurises backbenchers to vote for the government position, and the requirement of collective responsibility that has the same effect on ministers.Our means of measuring attitudes was to track the membership of extra-parliamentary groupsa method that was key to Norton's study of the PCP under Thatcher (Norton, 1990).An increased focus on extra-parliamentary group membership also enables us to capture emerging dividing lines within contemporary British Conservatism.As a result, we can extend the focus of intra-party divides to go beyond the classic dividing lines of economics, European policy, and social, sexual and moral issues (Heppell, 2002) to also include the attitudes that defined the Johnson era, namely Brexit, cultural debates, levelling up and the environment.
We approached our analysis by including a wide range of plausible demographic, political and ideological variables in our models to predict MPs' support for the main leadership candidates.We subsequently carried out a stepwise regression, which maximises model fit (using AIC) while minimising the number of predictor variables, and hence allowed us to see which variables best explain support for each candidate.This is important because, as outlined above, the significant ideological divides within the Conservative Party have changed over timefor instance, the disappearance of the wet-dry divide after 1997.In removing variables which do not contribute to model fit, stepwise regression allows us to identify variables that do contribute to that fit, whether statistically significant or not.In contrast to simply presenting full models, this stepwise approach thus allows us to say something about the important ideological divides that exist within the party today, and the variables that do (and, indeed, do not) structure these divides.
For demographics, we included gender as a variable, given the numerous critiques of Johnson's masculine approach to political competition (Smith, 2021, pp. 450-1).We also noted the role of sexuality and ethnicity.In terms of LGBT + issues, the issue of femaleonly spaces was a key divide during the parliamentary ballot stage, with Mordaunt being attacked by other leadership candidates for her alleged support for trans self-identification.Given Cowley and Garry's (1998) findings that demographic variables can be significant when there are stark contrasts between leadership candidates, we also include ethnicity in our study, as of the eight MPs who made it to the first round of ballots, four were ethnic minorities, one of whom, Sunak, made it to the final round.
Next, we considered political variables.Constituency marginality has been used in prior studies of voting motivations of Conservative parliamentarians in leadership selection ballots, and we assumed that those in marginal constituencies would note the opinion polling evidence which suggested that 37 per cent of voters thought Sunak would make a 'good' prime minister, whereas Truss (24 per cent) and Mordaunt (21 per cent) secured less favourable feedback (Ipsos-Mori, 2022).Given the fall of the 'red wall' to the Conservatives at the 2019 general election, we decided to include a dummy variable for whether an MP represented a red wall constituency or not (Kanagasooriam & Simon, 2021).Here we noted that because Truss' economic offer was the furthest away from the government's levelling up agenda (Grierson & Quinn, 2022), and Sunak had been caught 'claiming he worked to divert money from "deprived urban areas" when chancellor' (Brown, 2022), neither candidate would appeal to MPs in red wall constituencies.We also wanted to test the validity of the media narrative that those who had been loyal to Johnson formed the basis of Truss' support as she was one of the senior ministers who did not resign from Johnson's cabinet before his resignation (Boycott-Owen & Penna, 2022).Therefore, we included the public declarations of voting preferences in the vote of confidence in Johnson's leadership in June 2022.We also included a measure of whether a seat was won or held in 2019, on the basis that MPs with newly-won seats would be more loyal to Johnson due to his role in the 2019 general election, and therefore more likely to support Truss.
We also wanted to consider various European-related influences on MPs' voting behaviour, noting the pragmatism inherent within Truss' career (Forrest, 2022).When she was climbing the ministerial ladder to cabinet in the Cameron administration, her career status seemed best served by advocating remain in the 2016 European Union referendum, given the widely-held expectation that remain would win.In contrast, both Mordaunt (at the time a junior minister) and Sunak (a backbencher, having only entered Parliament one year earlier), decided to vote to leave (Grierson & Quinn, 2022;Riley-Smith, 2022), with Mordaunt being relatively high-profile during the campaign.Media coverage emphasised the support that existed for Truss among some high-profile Johnson and Brexit loyalists, whilst despite his support for Brexit Sunak was seen as the preferred candidate of Conservative MPs who had backed remain.
We consider three different European-related influences: the strength of the leave vote in an MP's constituency, an MP's position in the EU referendum and whether an MP was a member of the European Research Group of strongly pro-Brexit MPs.This overwhelmingly leave-backing group of MPs were seen as hardliners during the EU withdrawal negotiation process under both May and Johnson.The ERG was split between whether to back Truss or Braverman and ultimately agreed to back whoever made it furthest in the competition, in this case Truss.
Finally, we considered ideological factors related to social conservatism and social liberalism, economic positions vis-à-vis levelling up, economic intervention, free trade, and support for the environment.To do this, we collected lists of members or parliamentary supporters for specific extra-parliamentary groups.There are some issues with the use of these lists.Not all groups have published membership lists, and so members have to be garnered through a variety of other sources, such as group publications, whilst others do not allow ministers to be affiliated (as noted below, where relevant).However, these groups remain a good proxy for ideology since affiliation is voluntary on the part of the MPeither through an invite from the organisation or by the MP approaching the organisationand thus reflects how an MP would like to publicly position themselves.As a result, we can expect it to be a good measure of ideological placement on certain policies.This is preferable to voting lists which have typically been used in these kinds of studies, which are structured by collective responsibility and the whipping system.
For our measure of social conservatism we included whether Conservative parliamentarians were associated with the Common Sense Group or not, either as contributors to the book Common Sense: Conservative Thinking for a Post-Liberal Age (Common Sense Group, 2021) or as having signed a letter on behalf of the group to the Telegraph (Leigh, 2020).This group is concerned with political correctness and the 'woke' agendacapturing debates around free speech, no-platforming, and how we discuss identity and equality issues (Duffy et al., 2021) and present themselves as the defenders of traditional British values in the face of these developments.Given the disagreements over gender selfidentification raised during the campaign -Mordaunt was criticised by social conservatives for previously adopting a more pro-trans stance (Boycott-Owen, 2022)we felt that this was a useful variable to consider.To capture the socially liberal wing of the party, we used Bright Blue's parliamentary supporter list and the Tory Reform Group's patron list to measure whether an MP could be classed as socially liberal, given both groups describe themselves as supporters of 'liberal conservatism' (Bright Blue, 2023;Tory Reform Group, 2023).
To measure support for the levelling up agenda we used membership of the Onward taskforce for levelling up (2020).For whether an MP supports a more interventionist economic policy we used the parliamentary caucus list of the Blue Collar Conservatism group (2021), a 'grassroots' campaigning force for 'practical conservatism' to support 'working people', and the signatory list of a letter from the Northern Research Group, which campaigned for increased spending for the north of England (O'Carroll, 2020;Toal, 2020).Contrastingly, for our measure of supporting economic liberalism, including further deregulation of the economy and lowering of taxation alongside an associated drive to control public expenditure, we used the list of parliamentary supporters of the Institute for Economic Affairs' Free Market Forum (2021) and Conservatives for CANZUK (2021) as measures. 1 Finally, to capture the importance of the environment and the divides over net zero, we used the parliamentary caucus list (including the 'alumni' who were members but had to leave due to becoming a minister) of the Conservative Environment Network, which backs the net zero campaign, recognises man-made climate change, and seeks marketorientated solutions to environmental and climate issues.Given the language used by the three candidates, Mordaunt was the most enthusiastic about the net zero policy (Fisher, 2022).
To test which variables were most significant in predicting MPs' behaviour in the leadership contest, we established the voting preference of Conservative parliamentarians in both the final parliamentary ballot (round five of the contest) and the period leading up to the subsequent membership ballot (round six of the contest).Although parliamentarians are not required to reveal how they voted at any stage in the contest, previous studies on parliamentary ballots in Conservative Party leadership elections have been effective at establishing the preferences of around 80 per cent of all parliamentarians (Cowley & Garry, 1998;Heppell & Hill, 2008;Jeffery et al., 2018;2022, although the latter study only identified 72 per cent of preferences).Our approach to establishing candidate preferences replicated the methods adopted in these prior studies, namely cross-referencing a range of declared supporter lists within various newspaper outlets and blogs, which was then verified and supplemented by an exhaustive tracking of the social media posts of all Conservative parliamentarians.
In total, we identified the candidate preferences of 171 Conservative parliamentarians for the final parliamentary ballot: 77 of the 135 Sunak backers, 43 of the 113 Truss backers, and 51 of the 105 Mordaunt backers.This represented 47 per cent of the PCP having their voting preference identified, much lower than in prior studiesthis was largely due to the timings of the contest, as the previous round, which eliminated Badenoch, took place the day before and thus MPs did not have long to make up their mind whom to back in the final round, let alone declare publicly.However, our performance for the membership ballot period was significantly better, whereby we identified the preferences of 299 MPs (84 per cent of the PCP): 159 backers for Truss and 140 for Sunak.
We began constructing our dataset by downloading a full list of Conservative MPs as of 5 September 2022 from the UK Parliament's Members' Name Information Service (MNIS) API, using the R package Parlitools (Odell, 2022).This also included a variable for gender.We gathered data on the sexuality of MPs from the website LGBT + Members of Parliament (Peart, 2022) and information on the ethnic background of MPs from a House of Commons Library report (Uberoi & Tunnicliffe, 2021, p. 16).
For constituency marginality we used data from the British Election Study election results dataset, again accessed via the Parlitools package.We coded red wall constituency status according to the work of Kanagasooriam and Simon (2021).For how an MP voted in the confidence motion we gathered data from several sources, either crowdsourced (Election Maps UK, 2022; Singh, 2022) or provided by a media outlet (The Spectator, 2022).This information was then verified via MPs' social media accounts and public statements via other channels.
For our European variables, we used the Parlitools package to access Hanretty's (2017) estimated constituency leave vote.An MP's position in the EU membership referendum was based on datasets by Heppell et al. (2017) and Roe-Crines et al. (2021) for the 2015-17 and 2017-19 PCPs, and for those first elected in 2019 we determined their Brexit positioning from the research compiled by Lynch (2020).For ERG membership, and group membership used to denote ideological positions, we drew on publicly available membership lists.

Research Findings
Table 2 contains the descriptive statistics of MPs' public declarations of support in the fifth round of ballots.In terms of our demographic variables, there is no clear evidence that female MPs were more likely to back female candidates: 20 per cent of female MPs backed Mordaunt, but just 12 per cent backed Truss, and 18 per cent backed Sunak.Although Mordaunt received the most support from LGBT MPs, she was closely followed by Sunak.Ethnic minority MPs were not more likely to support Sunakin fact, they were overrepresented in Truss' support base: she won 24 per cent of the ethnic minority MP vote, compared to her share of 12 per cent of the PCP as a whole.Overall, there is no clear evidence that MPs were voting for leadership candidates based on demographics.
Moving to our political variables, we see little in the way of difference between average majority size and support for each candidate, and this is confirmed with a Kruskal-Wallis rank sum test showing no statistically significant difference between groups (p = 0.888).We also find no difference in the number of red wall MPs backing each candidatethree apiece, with 12 undeclared.Mordaunt did, however, perform more strongly among those who had gained their seat from another party in 2019 (a group that includes red wall MPs), with 12 MPs out of 56 supporting her, compared to half that number for both Truss and Sunak.
There are interesting factors surrounding the confidence vote variable.Whereas Rishi Sunak won the support of 20 per cent of MPs who backed Johnson, and 20 per cent of MPs who opposed him, support for the other two candidates was not so evenly balanced: Truss won 16 per cent of backers compared to 7 per cent who publicly did not, and Mordaunt won 25 per cent of MPs who publicly did not have confidence, to 9 per cent of those who did.
We also find evidence in line with our expectations for all three of our European variables: the average constituency leave vote among Sunak's backers was lower than Truss' or Mordaunt's.A Kruskal-Wallis rank sum test shows there was a statistically significant difference between the mean value for each group (p = 0.011) and a pairwise Wilcox rank sum test shows the differences in consistency leave vote share between Sunak and all other groups were statistically significant (vs.Truss p = 0.048, vs. Mordaunt p = 0.012, and vs. Unknown p = 0.033), but the differences between other groups were not.Most leave-backing MPs had not publicly declared their position (61 per cent) and support for each candidate was at broadly similar levels: 9 per cent for Sunak, compared to 14 per cent for Truss and 15 per cent for Mordaunt.However, remain MPs broke for Sunak (39 per cent), over Truss and Mordaunt (10 per cent each).Although leavebacking MPs largely remained on the fence, more than twice as many ERG members backed Truss (21 per cent) compared to her rivals (10 per cent each).
For our other group variables, the findings are more complex.Socially conservative MPs were not willing to back any candidateof the 30 MPs labelled as such, 27 did not make their position publicwhereas socially liberal MPs made up a much greater share of Sunak's support (29 per cent) than Mordaunt's or Truss' (13 and 10 per cent respectively).In a complete reversal of our expected relationships, MPs who were part of pro-levelling up groups were more likely to support Sunak than Truss or Mordaunt, whilst MPs who were part of the pro-economic intervention groups were equally likely to support all three candidates.Pro-free trade MPs were more likely to back Sunak (20 per cent) than Truss or Mordaunt (15 and 11 per cent respectively), whereas members of the CEN were generally split between Sunak and Mordaunt (at 25 and 21 per cent respectively), and just three per cent backed Truss.
Table 3 shows the output from a series of Firth's logistic regression models where candidate support was given as a binary variable, with 1 representing support for that candidate.The results of the full models are presented alongside the stepwise models predicting support for each candidate.We use Firth's logistic regression due to its superior ability to deal with small samples and complete separation compared to classic logistic regression (Firth, 1993;Puhr et al., 2017).
We find limited evidence that our demographic variables are important.Female MPs were more likely to support Mordaunt, but not Truss (nor Sunak), whilst neither MPs' sexuality nor ethnicity was a significant factor in predicting support for a given candidate.
In terms of our political variables, we find no relationship between majority size and support for a given candidate, but we do find that red wall MPs were less likely to support Mordaunt than non-red wall MPs, although this statistically significant relationship disappears in the stepwise model.We also find that MPs who did not have confidence in Johnson, relative to those who did, were less likely to support Sunak (but this relationship did not make it into the stepwise model) and more likely to support Mordaunt in both models.
Moving to our European variables, we find that MPs representing constituencies with a larger leave vote were more likely to support Mordaunt in both the full and stepwise models.Interestingly, we find that while the EU referendum vote does not structure support for Truss, remain MPs were much more likely to support Sunak and less likely to support Mordaunt or not make a public declaration, relative to MPs who backed leave.Truss' support is, however, structured by membership of the ERGmembers were much more likely to support Truss than not and were less likely to back Mordaunt (in the stepwise model).Thus we find evidence of a more complicated Brexit divide -Sunak won the support of remainers, Mordaunt the support of moderate or pragmatic leavers, and Truss the support of the hard-line leave-backing ERG.This echoes a study of the 2016 leadership elections, where remainers backed May, but socially liberal leavers backed Gove and socially conservative leavers backed Leadsom (Jeffery et al., 2018), and highlights the importance of not treating leave-voting Conservative MPs as a monolithic bloc: there are significant, and seemingly persistent, divides among Eurosceptics.
In terms of the extra-parliamentary group variables, we find little support for a socially-conservative or socially-liberal bloc vote, albeit socially conservative MPs were more likely to have an unknown position than notwhich is not surprising given their favoured candidate, Badenoch, was eliminated the day before this ballot took place.Groups representing certain economic positions also did not structure support in statistically significant ways, but membership of the Conservative Environment Network was a significant driver in opposition to Truss, perhaps linked to her rhetoric around wind power and fracking during the leadership campaign.
As noted in the previous section, a significant number of MPs had not made their declarations public for the final parliamentary ballot round.By the end of the membership ballot, however, many more MPs had made a public declaration of support for either candidate (MPs with an unknown position fell from 187 to just 59), thus it is also worth analysing patterns of support in the final round of the contest (Table 4).remainer Truss, and a majority of remain-backing MPs (57 per cent) supported the Brexiteer Sunak.In this instance, those with an unknown referendum position acted more like leave supporters and were much more likely to back Truss than Sunak.Relatedly, we see Truss consolidated her lead among the ERG too, winning 62 per cent of the group (to 23 per cent for Sunak and 15 per cent for unknown).
Finally, in terms of our group variables, we see that Truss won majority support from socially conservative MPs (70 per cent), pro-free trade MPs (52 per cent), and, ironically, pro-economic intervention MPs (53 per cent).Sunak won a majority among socially liberal MPs (51 per cent) and pro-levelling up MPs (54 per cent) and lead among members of the Conservative Environment Network (47 per cent to Truss' 36 per cent).
In order to analyse these patterns in more detail, we return to our Firth's logistic regression models, shown in Table 5.The first model is based on choosing Truss over Sunak, and the second on choosing to keep one's position unknown over publically supporting either candidate.
Female MPs were more likely to support Truss over Sunak, but other demographic variables did not produce a statistically significant relationship.We do find that MPs who held their seat prior to 2019 were less likely to back Truss, but only in the stepwise model (in the full model this is significant at the p = 0.1 level).Relative to those who publicly had confidence in Johnson, those who had an unknown positionand therefore were presumably more likely to be against Johnsonwere less likely to support Truss and more likely to support Sunak.Additionally, in the stepwise model, those who publically did not have confidence in Johnson were also less likely to support Truss, but this again is only at the p = 0.1 level.Unlike in the previous round, we find that an MP's constituency leave vote was positively related to their supporting Truss over Sunak, and that their EU referendum positionalthough not statistically significant in these modelsdoes make it into the stepwise model, showing it was an important factor in improving model fit.In the stepwise model, MPs who supported remain were less likely to support Truss over Sunak, but this relationship is only statistically significant at the p = 0.1 level.As expected, and as with the previous round, members of the ERG were more likely to support Truss and less likely to support Sunak.
Of our ideological variables, we only find support for levelling up as a statistically significant dividing line within the party.However, social conservatism also makes it into the stepwise model, and the variable narrowly misses out on statistical significance at the 0.05 level (p = 0.07 in both the full and stepwise model).As such, one can argue that social conservatism should be considered a pertinent dividing line within the parliamentary party.Despite Truss' clearly stated free-market agenda, we found no increase in support from MPs classed as pro-free trade in this round, nor a decrease in support among those who supported increased economic intervention.Membership of the CEN was no longer a statistically significant predictor of opposition to Truss.
Finally, when it came to MPs who had not made a public declaration by the end of the contest, we find that no single variable had statistically-significant explanatory power in either the full model nor the stepwise model (although the socially liberal variable was significant at the p = 0.1 level in the stepwise model).The poor explanatory power of these models may be because MPs who had decided to come off the fence had done so to increase their chances of a ministerial post, rather than any ideological or demographic reason that could be picked up in these models.
Linking back to the fifth round of nominations, in a regression model not presented here we included two dummy variables in the final round model to account for an MP's position in round 5: one for support for Mordaunt and one for holding an unknown position.We find both Mordaunt supporters and unknowns were less likely to back Sunak than to not back him, and were much more likely to remain quiet about their position than make a public declaration of support.This relationship holds up in the full model and the stepwise model.There was no statistically significant relationship with supporting Truss, suggesting that she did not win over significantly more Mordaunt supporters or more waverers than one would expect, despite her momentum in the final round.

Analysis and Conclusions
This analysis provides a challenge to dominant media narratives around the first Conservative Party leadership election of 2022.In an election framed around the dramatic disintegration of the Johnson government, and questions of loyalty to the outgoing leader, variables associated with support for Johnson played out in interesting, and changing, ways.In the fifth round, Mordaunt's support was driven by MPs who did not have confidence in Johnson, whereas Sunakone of the first ministers to resigndid not initially benefit from a first-mover advantage: in fact, our regression shows that MPs who did not have confidence in Johnson were less likely to back Sunak.However, whilst support for Johnson was not statistically significant in determining MPs' support for Truss or Sunak during the membership round, whether an MP kept their position in the June confidence vote secret was a driver of opposition to Truss and support for Sunak.Given that, numerically, most of these MPs with unknown positions must have voted against Johnson in the confidence vote, and the fact that being anti-Johnson only narrowly missed out on statistical significance at the 0.05 level, we can suggest that Sunak did consolidate the anti-Johnson base within the PCP.
Furthermore, support for Johnsonismas measured via support for his flagship economic policy, levelling upwas also an important factor in structuring support.MPs who were in favour of levelling up supported Sunak and were less likely to support the economically liberal Truss.However, despite her clear positions on economic intervention and free trade, support for Truss was not determined along these lines.Whilst the day-to-day personally-driven accounts of the Johnson government's disintegration may have made for interesting copy, the role of ideologyand, in particular, of levelling up as a new dividing line in the Conservative's post-Thatcherite political economyin structuring today's Conservative Party needs to be re-emphasised.
Further evidence of this is that the much-vaunted Brexit divide has evolved beyond simply a leave-remain dichotomy.In the final ballot of MPs, Mordaunt was the Brexiteers' choice and Sunak the choice of the remainers.Truss, however, won the support of the ERGand so while her Damascene conversion to the leave cause may have isolated remainers and moderate leavers, it won over the hardliners.When the contest progressed to the membership vote, support for leave or remain was not statistically significant in driving support for Truss over Sunak (although in the stepwise model backing remain relative to leave came close to statistical significance at p = 0.06).However, Truss maintained the support of the ERG, which at just under one-third of the parliamentary party was a significant bloc of support, and the ERG's views carried significant heft among Conservative Party members.
These results simultaneously demonstrate the continuing importance of the European issue in determining MPs' support for leadership candidates alongside the ways in which this European divide is mutating (i.e. between remainers, moderate leavers, and hard-line leavers/ERG members).The results also highlight the fact that, despite Truss' own ideological priorities being most defined by her economic positions, it was her positioning on Europe that did more to attract support from her parliamentary colleagues.As such, the base of support within the PCP for Trussonomics was perhaps not as solid as media narratives may have made it seem, and helps explain why the PCP lost confidence in Truss so quickly and changed the party's economic course in the subsequent leadership election (Booth et al., 2023).
Given the long-running importance of the leave-remain divide to structuring Conservative Party leadership elections, and the role the ERG played in this too, it is surprising that associations with other extra-parliamentary groupings do not seem to have had the same impact.In the fifth round of the contest, only membership of the Conservative Environment Network was statistically significant in structuring opposition to Truss.In the final round, however, we see the emergence of a divide along economic policy linesas mentioned above pro-levelling up MPs were more likely to back Sunak over Trussand, although not significant at the 95% level, social conservatism makes it into both the full and stepwise model (p = 0.07), whilst socially liberal MPs were more likely to remain unknown, again at the 90% significance level.This hints at a long-term ideological divide within the PCP: attitudes on social issues also affected the contests in 1997, 2001, 2005and 2016(Heppell & Hill, 2008, p. 2009;2010;Jeffery et al., 2018).It is also worth noting that not all wings of the Conservative Party are of equal strength: Truss' relative strength among the ERG (107 MPs) was more consequential than, say, Sunak's strength among the pro-levelling up group (37 MPs).Similarly, socially conservative MPs make up under 10 per cent of the PCP.
Finally, we identify a cohort effect.Truss' performed more strongly among those first elected in 2019 compared to those elected before then.This may partly explain why she succeeded despite concerns about her suitability for the top job (that turned out to be well-founded)while colleagues who had known her for longer were less likely to back her for the top job, many new MPs did not have this first-hand experience of her on the personal level.
Whilst our analysis of an original dataset has revealed the old and new ideological dividing lines between Conservative MPs, none of our models produce a pseudo R-squared of over 0.35.The fact that models combining demographic, political, European, and ideological variables explain no more than a third of the variance in support for a given candidate reveals that Conservative MPs' support for leadership candidates is also structured around other factors.These may include relationships (both personal and professional, i.e. having worked for a candidate as a minister, being promised a position in a future government), their own assessment of a candidate's performance in government, the views of an MP's local membership, and an MP's own perception of who is most likely to win the contest (and reward them with a ministerial position), or win the next general election.
Despite this warning, it is clear that today's Conservative Party is divided along ideological lines.Our contribution to the literature on the Conservative Party is to provide clear evidence that MPs' positions on social liberalism, political economy and, of course, Europe, can and do structure support in leadership elections and that these divides can be measured through membership of extra-parliamentary groupings.

Note
1. Conservatives for CANZUK back deeper social and economic ties between the nations of Canada, Australia, New Zealand and the UK, including free movement of people and free trade.

Table 1 .
Parliamentary Eliminative Ballots.Failed to meet the progression threshold of 30 voters + Tobias Ellwood had the whip temporarily removed at the time of the fourth ballot for failing to return to Parliament in time to vote with the Johnson government in the confidence motion tabled by the Labour opposition, but this was restored in time for the fifth and final parliamentary ballot.

Table 2 .
Descriptive statistics of MPs' public declarations of support in the fifth round of ballots.

Table 3 .
Firth's logistic regression model output for MP support in round 5 of the parliamentary ballot.

Table 5 .
Firth's logistic regression model output for MP support in the membership round.