Preferences versus Party Loyalty: Support for a Series of Free Votes Dealing with House of Lords Reform

Transcription

1 Preferences versus Party Loyalty: Support for a Series of Free Votes Dealing with House of Lords Reform Christopher D. Raymond Lecturer in Politics Queen s University Belfast C.Raymond@qub.ac.uk Abstract Research examining free votes shows party shapes voting behaviour even after the whips are withdrawn. While most research argues party cohesion persists due to shared preferences among members of the same party, this need not be the case. Building on recent research, I argue that party loyalties independent of shared preferences may divide members when issues are politicised by party leaders. Examining a series of free votes dealing with efforts to reform the House of Lords, the results show the effects of preferences on voting behaviour were limited at best. Instead, support was motivated primarily by lingering party loyalties. These findings force us to reconsider the role accorded to personal preferences as the primary determinant of MPs voting behaviour at a minimum, to revisit the conditions under which each type of effect will dominate.

2 There is growing consensus that the personal preferences of Members of Parliament (MPs) are important to understanding why political parties in most legislatures exhibit such high levels of party unity. While some MPs doubtlessly vote in line with their fellow party members due to fear of the party whips, the fact parties remain cohesive even on free votes (in which the whips have been withdrawn) suggests that pressure from the whips alone is insufficient to explain why parties exhibit such high levels of unity. Instead, previous research studying the voting behaviour of MPs on free votes concludes that the shared preferences of MPs belonging to the same party explains the sustained levels of party unity (e.g., Cowley and Stuart, 1997, 2010; Hibbing and Marsh, 1987; Marsh and Read, 1988; Mughan and Scully, 1997; Overby et al., 1998; Pattie et al., 1998; Plumb and Marsh, 2011; Plumb, 2015). While most studies conclude that preferences play an important role in explaining the high levels of party unity observed on free votes (and, presumably, on other divisions as well), there are reasons to suspect this may not be the case or at least not in all cases. In addition to the fact most studies do not account for the potential impact of constituency pressures or party loyalty both of which might sustain high levels of party unity in their own right most previous research has relied on indirect measures of MPs preferences, using measures that are either based on personal characteristics or MPs previous voting behaviour. Because such indirect measures of preferences are problematic, it remains to be seen whether the conclusions reached in previous research regarding the effects of personal preference are robust. Although it is likely MPs preferences go a long way towards understanding the high levels of party unity observed in previous research, it is possible too much emphasis has been placed on personal preferences, or that the impact of such variables needs to be qualified. In order to determine the robustness of personal preference-based effects on MPs voting

3 behaviour, this paper examines the impact of preferences relative to constituency pressures and party loyalties on several free votes cast in 2003 relating to proposals to reform the House of Lords. This issue is helpful for analysing the effects of MPs personal preferences because data measuring MPs individual preferences are available. Specifically, I use the British Representation Study 2001 (Norris and Lovenduski, 2001), which was a survey of candidates prior to the 2001 general election. This survey included questions measuring attitudes relevant to deciding how to vote on proposals to reform the House of Lords. This allows for direct measurement of MPs personal preferences that is rare in studies of legislative voting behaviour. Such direct measurement of preferences increases our confidence in our ability to estimate the effects of personal preferences on MPs voting behaviour. In the next section, I review the literature regarding the effects of personal preferences on MPs voting behaviour. Following that, I review the issue of House of Lords reform in greater detail in order to provide appropriate context for the analysis. Following that, I discuss the details of my research design before moving to a discussion of the results. A final section concludes with thoughts putting the findings into broader context, as well as a discussion of the implications of this study for future research on the analysis of legislative voting behaviour. Personal Preferences, Constituency Pressures, and Party Loyalties In studying the voting behaviour of MPs, party plays an influential role in most legislatures, with high levels of unity observed on nearly every division. Two explanations for this unity are given primacy. One is that MPs are kept disciplined by the whips, while a second holds that MPs of the same party share preferences on most divisions. Because most divisions are whipped, and because party unity is almost universally high, it is difficult to discern the impact of preference-based effects (in relation to or independently from discipline-based effects).

4 To control for the effects of party discipline in order to study the impact of preferences, many studies examining the impact of preference-based effects have relied on free votes, which are divisions on which the whips are relaxed and MPs are genuinely allowed to vote as they wish (Cowley, 1998; Richards, 1970). Most previous research examining MPs voting behaviour on free votes finds that MPs tend to coalesce along party lines even after the whips have been withdrawn (Cowley and Stuart, 1997, 2010; Hibbing and Marsh, 1987; Marsh and Read, 1988; Mughan and Scully, 1997; Overby et al., 1998, 2011; Pattie et al., 1994; Plumb, 2013, 2015; Plumb and Marsh, 2011). This suggests that discipline-based effects may not be the primary determinants of voting behaviour, and thus may not be the primary explanation of the high levels of party unity observed on most divisions (both free and whipped). Because party unity remains high on free votes, and because the variables used to measure the personal preferences of MPs explain much of the variance in voting behaviour on those particular divisions examined in previous studies, many have concluded that shared preferences explain the high levels of party unity. This supports other research, which finds that preference agreement is the most important determinant of party unity (e.g., van Vonno et al., 2014). Despite the primacy accorded to preference-based effects, it is not entirely clear that the shared preferences of MPs alone (or even primarily) explain the high levels of party unity observed on most divisions. One reason preventing us from concluding decisively that personal preferences dictate voting behaviour to the extent claimed in previous research is that most studies lack direct measures of MPs preferences. Some studies have attempted to operationalise preferences using proxies related to MPs personal characteristics (e.g., Hibbing and Marsh, 1987; Marsh and Read, 1988; Mughan and Scully, 1997; Overby et al., 1998, 2011; Pattie et al., 1998), which may be problematic because they do not directly measure MPs preferences. Others

5 have operationalised preferences using measures derived from MPs previous voting behaviour (e.g., Baughman, 2004; Plumb and Marsh, 2011; Plumb, 2015), which may be problematic given that MPs voting behaviour may not be reflective of actual preferences (Clinton, 2012). Even for those studies that do include more direct measures of preferences, there are other reasons for scepticism. For one, MPs have incentives to reflect the preferences of their constituents (André et al., 2014a, 2014b; Cain et al., 1987; Carey and Shugart, 1995). Because MPs have also been shown to express the interests of their constituents on divisive measures (Baughman, 2004; Overby et al., 2011; Pattie et al., 1994, 1998; Sieberer, 2015), what appear to be personal preference-based effects may instead be due to MPs articulating their constituents preferences. Additionally, most previous studies do not account for the possibility that MPs voting behaviour on free votes may be due in part to an underlying sense of loyalty to the party. For instance, some studies find that party unity remains high even when MPs preferences are not likewise unified (Andeweg and Thomassen, 2011) and when whips are absent (Norton, 2003), particularly when issues are turned into party issues (Plumb and Marsh, 2013; see also Dewan and Spirling, 2011). This is corroborated by other research, which shows that, while MPs may be willing to vote their consciences on the least consequential free vote divisions, they tend to revert to more party-line votes on those divisions with the greatest chances of impacting the final outcome, and thus the greatest consequences for their parties (Cowley and Stuart, 1997, 2010; Overby et al., 1998). Other research suggests MPs loyalty to the party (on both free and whipped votes) may be rooted in the party identification of MPs that leads them to behave in a partisan fashion similar to voters in the electorate (Raymond and Overby, 2014; Russell, 2014; see also van Vonno et al., 2014).

6 House of Lords Reform Because the review of the literature above suggests that the shared preferences of MPs belonging to the same party may not be the sole determinant of non-whipped voting behaviour, it is important to estimate the effects of preferences net of constituency pressures and party loyalties. To this end, this paper analyses the divisions regarding House of Lords reform pursued by the Blair government in Because MPs faced pressures from some groups of constituents, and these issues became embroiled in larger partisan conflicts, MPs personal preferences may not have been dispositive in deciding these issues. While the results of this analysis have implications for understanding this particular episode in the long history of efforts to reshape the House of Lords, the focus of this study is on isolating the effects of personal preferences after controlling for variables measuring constituency pressures and party loyalty. Readers looking for in-depth accounts of attempts at reforming the House of Lords (including this and other episodes) are instead directed to the excellent reviews and analyses conducted elsewhere (Ballinger, 2014; Raina, 2015; Russell, 2013; on this specific episode, see McLean et al., 2003). Like other recent attempts, this particular episode failed to produce any changes in the selection of members of the House of Lords. After the government s White Paper recommending that 20 per cent of the House of Lords be elected faced stiffer-than-anticipated criticism on its release in 2001, and after the Public Administration Select Committee unanimously called for a predominantly elected upper chamber, responsibility for proposing options regarding the shape of House of Lords reform was transferred to a Joint Committee on House of Lords Reform. The Committee presented a range of options for what a reformed upper chamber might look like, of which five were decided in the House of Commons as free votes on 4 February In the

7 order that they were decided: MPs voted on abolition of the House of Lords, 100 per cent appointed, 100 per cent elected, 80 per cent elected (20 per cent appointed), and 60 per cent elected (40 per cent appointed). None of the five proposals passed despite the fact that all three parties declared support for an elected upper house in their 2001 election manifestos. Though this study focuses on one specific chapter of the saga that is the effort to reform the House of Lords, which is but one set of divisions regarding one issue, this issue provides a critical test of the party-as-preference argument because reform divides opinion among MPs with strong preferences. In examining the breakdown of these votes, McLean et al. (2003) found evidence that MPs voted perhaps at times strategically according to their personal preferences, both regarding the election of members as well as the abolishment of the House of Lords altogether. Historically, many of the critiques of the House of Lords especially those calling for its abolishment have come from MPs on the left, while support for preserving the institution has often come from those on the right (Russell, 2013, pp ). All this suggests that support for House of Lords reform was determined to a large extent by MPs personal preferences. Moreover, this case is interesting because, though these strongly-held preferences produced considerable disunity, at least one or more parties demonstrated considerable party unity on each division. Though Labour were nearly evenly split on each issue, Liberal Democrats were highly cohesive in opposing both abolition and appointment (with roughly 94 per cent voting against both measures), and almost equally cohesive in supporting all three electionrelated measures (with support ranging from per cent in favour). Likewise, the Conservatives were highly unified in opposition to abolition (with 98.6 per cent voting against abolition); though they were roughly equally split on the 80 and 100 per cent elected measures,

8 nearly two-thirds voted against the 60 per cent elected measure. Thus, similar to other recent studies showing high levels of party unity on free votes (Cowley and Stuart, 1997, 2010; Hibbing and Marsh, 1987; Marsh and Read, 1988; Mughan and Scully, 1997; Overby et al., 1998, 2011; Pattie et al., 1994; Plumb, 2013, 2015; Plumb and Marsh, 2011), House of Lords reform in 2003 demonstrates evidence of party unity even after the whips were withdrawn. This, in turn, suggests that preferences may have played an important role in fostering cohesion and explaining who deviated from the rest of the party. Before one can conclude that MPs personal preferences explained the outcome, however, one must consider the possibility that constituency pressures may also have influenced MPs behaviour on these divisions. Mirroring the partisan divides among MPs, support for reforming the House of Lords divided along party lines among constituents as well. Like several key figures in the party, many Liberal Democrat supporters followed the party s manifesto in preferring an elected House of Lords (Henry, 2012). Though many Conservative supporters mirrored the stated position of the party in supporting an elected House of Lords, a sizeable share of Conservative supporters preferred to see the House of Lords remain unchanged (Henry, 2012). Recognising this, MPs faced incentives to voice the interests of their constituents. For instance, MPs with more Liberal Democrat supporters faced constituency pressures to support an elected House of Lords. Given the need of MPs to develop and maintain personal votes that help increase their chances of re-election (e.g., André et al., 2014a, 2014b; Cain et al., 1987), MPs from constituencies with high levels of Liberal Democrat support may have voted to represent their constituents in an effort to maintain their support among these voters. Though the Conservatives were split on the issue, some MPs may have responded to larger Conservative constituencies by voting against the proposed reforms to the House of Lords whilst other MPs

9 may have responded to Conservative constituency pressures by voting in favour of electing members of a reformed upper chamber. This suggests the possibility that any relationship between MPs preferences and support for House of Lords reform may have been due at least in part to constituency pressures that led MPs to vote the preferences of their constituents (rather than voting their personal preferences alone). Even after accounting for personal preferences and constituency pressures, MPs may have voted along party lines due to the partisan nature of some of these debates, as well as the partisan consequences of these divisions. Similar to the responses of pragmatic Conservatives to divisions with partisan consequences (Cowley and Stuart, 1997, 2010), MPs may have recognised the partisan implications of some of these divisions and, thus, may have been motivated to vote along party lines out of a sense of loyalty to the party rather than to express their personal preferences. This was particularly the case with Labour MPs: despite the fact that many of the most vocal supporters of electoral reform for the House of Lords came from Labour s ranks, a sizeable number voted in favour of an all-appointed House of Lords following the Prime Minister s expressed support for this option. Though this was a free vote, meaning that Labour MPs were genuinely free to express their preferences for an elected House of Lords, there is at least some evidence many voted in favour in response to the party-consequential circumstances in an effort to support Tony (McLean et al., 2003, pp ); similar loyaltytype effects may have been seen on the other divisions. Thus, it is possible MPs did not vote their preferences, but rather instead may have voted in their party s (and leader s) interests. Thus, there is need to examine the voting behaviour of MPs in order to confirm that preferences really do play the role ascribed to them in previous research. There is reason to believe that, similar to other cases examined in previous research, preferences played a definitive

10 role in explaining MPs behaviour on the issue of House of Lords reform. However, there is also reason to believe that constituency pressures and party loyalty may also have shaped the results. The next section outlines a research design for determining the relative impact of preferences versus constituency pressures and party loyalty on this issue. Research Design To estimate the effects of personal preferences on MPs voting behaviour, I analyse (separately) support for each of the five divisions regarding House of Lords reform. Those voting in favour of the proposed measure are coded one, while those voting against are coded zero. Because the dependent variables are binary, I use logistic regression to estimate support for each measure. These divisions are ideal for analysing the impact of personal preferences on voting behaviour because data measuring the personal preferences of MPs that might affect their voting behaviour are available. Specifically, I use data taken from the British Representation Study (BRS) 2001 (Norris and Lovenduski, 2001), which surveyed candidates attitudes regarding two key attitudes which may have been relevant to deciding House of Lords reform. Of 999 candidates surveyed, 296 were MPs. This produced a representative sample of MPs in terms of the outcomes on each division. As seen in Table 1, the BRS produces samples on each division that are broadly representative both of the final outcomes and of the party-specific vote shares for the three largest parties. 1 While this is less true for smaller parties, this is not problematic, as 1 One concern with analysing these divisions is that several MPs did not vote on every measure, and therefore may have abstained tactically in order to avoid hurting their prospects of reelection. While it is difficult to determine which absences were tactical and which were random, one way to determine whether these absences were tactical is to assume all absences were tactically-motivated and examine these absences as a separate choice. Using multinomial logistic regression to estimate support/opposition/absence as a robustness test, the results show that all

11 dropping these two Scottish National Party MPs from the analysis does not affect the results. One of those variables collected by the BRS regards MPs attitudes towards the efforts to reform the House of Lords. Specifically, I include a measure of MPs attitudes towards the election of members of the House of Lords. Using a five-point scale ranging from definitely should to definitely should not, candidates interviewed in the BRS responded to the following statement: Replace the House of Lords with an elected Second Chamber. Responses were recoded ranging from zero ( definitely should not ) to four ( definitely should ). 2 Descriptive statistics for this and each other predictor of MPs voting behaviour are presented in the Appendix. In addition to MPs attitudes towards an elected House of Lords, voting behaviour may also have been affected by MPs left-right ideological positions. Critiques and defences of the House of Lords have long divided along left/right lines, with many of the most vocal critics of the House of Lords calling for its reform or abolishment concentrated among the left, and the most stalwart defenders of an unreformed Lords concentrated among those on the right. To measure left-right ideology, I include a variable coded one (left) to 10 (right). In order to determine the robustness of individual preference-based effects on voting behaviour robust, I also include variables measuring the impact of party affiliation and but one of the significant coefficients in the models using the binary measure of the dependent variable remain significant in the models using the multinomial measure of the dependent variable. 2 The frequencies falling into each category are as follows: 49 (0 definitely should not ), 24 (1), 6 (2), 47 (3), 165 (4 definitely should ). Because the majority of MPs stated that they (strongly) agreed with an elected upper chamber, I also estimated models substituting the ordinal measure with a dummy variable coded one for MPs who strongly agreed with an elected upper chamber and zero otherwise; I also estimated models substituting this variable with one coded one for all MPs who agreed or strongly agreed for an elected chamber coded one, and zero otherwise. The results using these alternative measures are substantively equivalent to those using the ordinal measure.

12 constituency pressures. To control for the possibility that MPs voting behaviour is affected by lingering party loyalty-based effects, I include a variable coded one for Labour MPs, and zero otherwise. I focus on Labour MPs primarily because these MPs were particularly cross-pressured on these divisions: while the overwhelming majority of Labour MPs (95 per cent) preferred replacing the House of Lords with an elected upper chamber (with some preferring to abolish the upper chamber altogether), Labour MPs may have been rallied to support the party against party interests in two ways. For one, the fact that Prime Minister Blair came out in favour of an allappointed House of Lords forced Labour MPs either to vote their personal preferences or vote with the stated preferences of their party on the division calling for an all-appointed House of Lords. Second, the strong Labour showing on the division calling for the abolition of the House of Lords may been aided by party loyalties not captured by shared preferences (or common constituency pressures): because Conservative and Liberal Democrat MPs were so unified against abolition, some Labour MPs may have been motivated to support the division even if it went against their personal preferences. Given this, including a variable controlling for Labour MPs tests whether the influence of personal preferences was robust to the effects of lingering party loyalties. 3 I include two variables measuring the impact of constituency pressures: one variable measuring the percentage of the constituency-level vote won by the Liberal Democrats in 2001, and another measuring the constituency-level vote share won by the Conservatives. 4 Because 3 While it would have been interesting to explore the effects of lingering party loyalties among Conservative and Liberal Democrat voters, the smaller number of observations among these two parties inhibited me from doing so. 4 I also estimated models substituting the Conservative and Liberal Democrat vote share variables with Labour s constituency-level vote shares. This variable, however, fails to reach significance in any of the models, suggesting that MPs responded more to the pressures from

13 these two parties and their supporters possess distinct preferences regarding possible reforms to the House of Lords, and because MPs may be concerned with the implications of their voting behaviour for their personal re-election prospects, variation in the constituency-level vote shares won by these two parties might impact MPs voting behaviour on the issue. The Liberal Democrats position (dating back to the Liberals policy of the nineteenth century) has been to replace the House of Lords with a fully elected upper chamber. Because of this, MPs from constituencies with higher vote shares won by the Liberal Democrats regardless of their personal preferences may be more likely to vote against abolishing the House of Lords and retaining a fully appointed chamber, and instead more likely to vote in favour of all three options to elect members of the upper chamber, in order to increase their prospects of developing personal votes that may help them win re-election. In a similar vein, the size of the Conservative vote share might impact MPs voting behaviour. Because the party officially opposed abolition and reform of the House of Lords in its manifesto, MPs representing constituencies with higher Conservative vote shares may be more likely to oppose abolition and electing members of the House of Lords. However, many figures within the party actively supported an elected House of Lords (as did the party in its manifesto), potentially leading to a correlation between Conservative vote shares and support for measures that would make the House of Lords an elected body (in part or in full). These intra-party divisions might have put different pressures on different MPs looking to protect their re-election interests. For Conservative MPs, higher Conservative vote shares might put pressure on them to oppose all reform efforts in order to placate this constituency within the party for fear of recall or deselection by the constituency association. Similarly for non-conservative MPs, higher Conservative and Liberal Democrat constituents perhaps because their parties preferences were clearer than those of Labour.

14 Conservative vote shares might lead MPs to oppose efforts to reform the Lords for fear of alienating their base of personal votes and hurting their re-election prospects. However, for entrepreneurial Labour MPs looking to cultivate personal votes among moderate Conservative voters and the modernisers within the party, higher Conservative vote shares may lead them to support measures seeking to create an elected House of Lords and end the practice of appointment. To account for this possibility, I create an interaction between the Conservative vote share variable and the variable measuring Labour MPs. Results Parameter estimates for each model are presented in Table 2. One is immediately struck by the coefficients for the measures of personal preferences: though all of the coefficients are in the expected directions, none of these coefficients reaches statistical significance. Despite all of the emphasis in previous research on the importance of MPs personal preferences in deciding how to vote on un-whipped divisions, these results suggest that personal preferences alone need not be determinative of voting behaviour, at least not on every division. Instead, these findings suggest that voting behaviour was determined more by constituency pressures and lingering party loyalties than by personal preferences. In order to display the estimated effects of constituency pressures more easily especially given the conditional nature of the hypothesised effects relating to the Conservative vote share variable Table 3 presents predicted probabilities for both the Liberal Democrat and Conservative vote share variables. To account for differences in these variables estimated effects between Labour and all other MPs, I present predicted probabilities calculated for both sets of MPs holding all other variables at their median values.

15 Regarding the impact of Liberal Democrat vote shares on MPs voting behaviour, the results in Table 3 suggest that variation in Liberal Democrat vote shares significantly influenced MPs voting behaviour on all but the measure calling for the abolishment of the House of Lords. For both Labour and other MPs, higher Liberal Democrat vote shares are associated with increased support for three measures calling for the election of members of the second chamber (and decreased support for the measure calling for 100 per cent appointment). Though the increases (decreases) are particularly pronounced among Labour MPs (due to the lower estimated probabilities of voting in favour of each measure among Labour MPs), these estimated effects were statistically significant, as indicated by the coefficients for the Liberal Democrat vote share variable for these four models in Table 2. Much more modest are the decreases in probabilities of supporting the abolition division as Liberal Democrat vote shares increase, reflecting the statistically insignificant coefficient in Table 2. In keeping with the literature suggesting MPs will vote on free vote divisions to reflect constituency preferences as a means to increase their chances of re-election (Baughman, 2004; Overby et al., 2011; Pattie et al., 1994, 1998; Sieberer, 2015), these results suggest MPs in constituencies where the Liberal Democrats fared well in the 2001 elections were more likely to vote in favour of electing members of the House of Lords (and less likely to vote in favour of appointment) than those where the Liberal Democrats won fewer votes. Turning to the estimated effects of the Conservative vote share variable, the results suggest that constituency pressures coming from Conservative voters were conditional on the party of the MP. Beginning with the results for non-labour MPs, the results suggest that higher Conservative vote shares put significant pressure on MPs to support appointment and oppose efforts to elect the House of Lords. (Tests of joint significance suggest that the effect of

16 Conservative vote shares on support for abolition were not statistically significant.) Higher Conservative vote shares were associated with greater probabilities of support for full appointment of members of the House of Lords, while higher Conservative vote shares were associated with lower probabilities of supporting an elected House of Lords on the 100/80/60 per cent elected divisions. These findings suggest that MPs outside Labour s ranks felt pressures from Conservative voters that led them to oppose transforming the House of Lords into an elected body and instead to preserve the appointment of Lords. Consistent with the argument that Labour MPs recognised personal vote-building potential among moderate and modernising Conservative voters, the results in Table 3 suggest that the behaviour of Labour MPs may have been influenced by constituency pressures that led them to oppose abolition and to support the election of members of the House of Lords. On the measure to maintain the practice of appointing members of the House of Lords, the probability that Labour MPs would support the motion decreases as the Conservative Party vote share increases. On the three measures calling for the election of members of the House of Lords, the probability that Labour MPs would support the measure increases as the share of the vote going to the Conservatives increases (though this increase is significant only at the 0.10 level on the measure calling for 100 per cent of members to be elected). In order to see the estimated effects of lingering Labour party loyalties on voting behaviour, Figure 1 presents the predicted changes in probabilities associated with Labour Party affiliation (relative to all other MPs) on each division. Because the estimated effect of Labour MP affiliation is conditional on the Conservative Party vote share variable, I present the estimated changes in probabilities for MPs in constituencies at three levels of Conservative vote shares: the mean Conservative vote share among Labour MPs, one standard deviation below and

17 one standard deviation above the Labour MP mean. Beginning with the motion to abolish the House of Lords, the results show that Labour MPs in constituencies with levels of support for the Conservatives greater than the mean were significantly more likely to vote in favour of abolition than MPs from other parties. The fact that the estimated impact of Labour Party loyalty is strongest among MPs from constituencies with the highest levels of Conservative support suggests these MPs were responding to the partisan circumstances of the vote. Because the Conservatives (and Liberal Democrats) were highly unified in voting against abolition, this made abolition a partisan issue that, in turn, may have made several Labour MPs vote along partisan lines out of loyalty to the party. The fact that this was most pronounced among Labour MPs from constituencies with the highest levels of Conservative support is consistent with the notion that it was MPs in these constituencies who most needed party loyalty in order to vote in favour of abolition with others in the party, as these were the MPs most vulnerable to constituency pressures from Conservative voters who would have been most likely to oppose abolition. Though the results and the significant estimated effects of Labour affiliation cannot rule out the possibility that Labour MPs voting in favour of abolition did so because their first preference was to abolish the House of Lords, the fact that Labour party affiliation has significant effects on the other four divisions provides evidence to suggest that Labour MPs voted their party loyalties on the issue of House of Lords reform more than their personal preferences. On the three motions calling for the election of members of the House of Lords, Labour Party affiliation is negatively and significantly associated with support for each motion at the mean and lower levels of Conservative support, with this effect weakening at higher levels of Conservative support. This suggests that loyalty to the party led Labour MPs to support the Prime Minister s position at lower levels of Conservative support, but that the pressures coming from Conservative voters

18 won out and led many Labour MPs at higher levels of Conservative support to represent the voices of Conservative moderates and modernisers in their calls for an elected House of Lords. Finally, on the 100 per cent appointment motion, the results show that Labour affiliation leads to significant increases in support for appointment across all three levels of support for the Conservative Party. This suggests that Labour Party loyalty helped bring many in the party to support the position taken by the Prime Minister against their own personal preferences where constituency pressures from Conservative voters that would have brought MPs around to support the Prime Minister s position were absent. Conclusion The results presented above have several important ramifications. First, they suggest the emphasis placed on the role of personal preferences as determinative of voting behaviour may require some rethinking. Most previous studies have made the assumption, implicitly or explicitly, that personal preferences drive MPs voting behaviour on free votes. However, these studies have been limited by the fact that most have not possessed direct measures of MPs preferences, which limits confidence in such findings. One of the advantages of this study has been the availability of rare, direct measures of MPs personal preferences. With such measures, the fact that personal preferences regarding efforts to reform the House of Lords were not significantly related to voting behaviour on the issue of reforming the House of Lords provides some evidence to question the role of personal preferences in explaining voting behaviour, at least as to the level it has been assumed to shape voting behaviour. Second, the results show that, instead of personal preferences, MPs voted in favour of/against efforts to reform the House of Lords based more on constituency pressures and party

19 loyalties. While it may be difficult to generalise these results beyond this particular episode, and while it is doubtful that personal preferences are seldom important to explaining MPs voting behaviour, these results suggest that future research give greater consideration to the roles these other variables may play on other divisions. If constituency pressures and party loyalties play a greater role than previous research has accorded them, then this would mean that important explanations have been ignored. It may be that such omitted variables have led to erroneous conclusions that personal preferences dictate voting behaviour when in fact constituency pressures and party loyalties independent of MPs personal preferences explain much of the sustained party unity observed in previous studies. It will be important in future research to determine when and under what conditions personal preferences play a significant role in shaping MPs voting behaviour (and when they do not). Likewise, given the findings presented here regarding the effects of constituency pressures, future research will need to examine the tensions between MPs and their personal preferences on the one hand and the preferences of their constituents on the other. Finally, given the possible tensions between MPs personal preferences and the interests of their parties even on free votes as seen here future research will need to consider to a greater extent the conditions under which MPs may choose to side out of loyalty to their parties as well as the conditions under which MPs will vote their personal preferences.

20 References Andeweg, R.B. and Thomassen, J., Pathways to party unity: Sanctions, loyalty, homogeneity and division of labour in the Dutch parliament, Party Politics, 17(5), pp André, A., Freire, A., and Papp, Z., Electoral Rules and Legislators Personal Vote- Seeking, in Deschouwer, K. and Depauw, S. (eds.), Representing the People: A Survey Among Members of Statewide and Substate Parliaments. Oxford: Oxford University Press, pp André, A., Gallagher, M., and Sandri, G., Legislators Constituency Orientation, in Deschouwer, K. and Depauw, S. (eds.), Representing the People: A Survey Among Members of Statewide and Substate Parliaments. Oxford: Oxford University Press, pp. Ballinger, C., The House of Lords, : A century of non-reform. Oxford: Hart. Baughman, J., Party, constituency, and representation: Votes on abortion in the British House of Commons, Public Choice, 120(1-2), pp Cain, B. E., Ferejohn, J., and Fiorina, M., The Personal Vote: Constituency Service and Electoral Independence. Cambridge: Harvard University Press. Carey, J. M., and Shugart, M. S., Incentives to cultivate a personal vote: A rank ordering of electoral formulas, Electoral Studies, 14(4), Clinton, J. D., Using Roll Call Estimates to Test Models of Politics, Annual Review of Political Science, 15, pp Cowley, P. (ed.), Conscience and Parliament. London: Frank Cass. Cowley, P. and Stuart, M., Sodomy, Slaughter, Sunday Shopping and Seatbelts Free Votes in the House of Commons, 1979 to 1996, Party Politics, 3(1), pp Cowley, P. and Stuart, M., Party Rules, OK: Voting in the House of Commons on the Human Fertilisation and Embryology Bill, Parliamentary Affairs, 63(1), pp Dewan, T., and Spriling, A., Strategic Opposition and Government Cohesion in Westminster Democracies, American Political Science Review, 105(2), pp Henry, J., House of Lords - reform or retain?, Available at: (Accessed: 31 January 2016). Hibbing, J. R. and Marsh, D., Accounting for the Voting Patterns of British MPs on Free Votes. Legislative Studies Quarterly 12(2), pp Hug, S., Selection Effects in Roll Call Votes, British Journal of Political Science, 40(1), pp Marsh, D. and Read, M., Private Members Bills. New York: Cambridge University Press. McLean, I., Spirling, A., and Russell, M., None of the Above: The UK House of Commons Votes on Reforming the House of Lords, February 2003, Political Quarterly, 74(3), pp

21 Mughan, A. and Scully, R. M., Accounting for Change in Free Vote Outcomes in the House of Commons. British Journal of Political Science 27(4), pp Norris, P. and Lovenduski, J., The British Representation Study Data accessed at < Norton, P., Cohesion without discipline: Party voting in the House of Lords, Journal of Legislative Studies, 9(4), pp Overby, L.M., Tatalovich, R. and Studlar, D.T., Party and Free Votes in Canada Abortion in the House of Commons, Party Politics, 4(3), pp Overby, L.M., Raymond, C., and Taydas, Z., Free Votes, MPs, and Constituents: The Case of Same-Sex Marriage in Canada, American Review of Canadian Studies, 41(4), pp Plumb, A., Research Note: A Comparison of Free Vote Patterns in Westminster-Style Parliaments. Commonwealth & Comparative Politics 51(2), pp Plumb, A., How Do MPs in Westminster Democracies Vote When Unconstrained by Party Discipline? A Comparison of Free Vote Patterns on Marriage Equality Legislation. Parliamentary Affairs 68(3), pp Plumb, A. and Marsh, D., Beyond party discipline: UK Parliamentary voting on fox hunting, British Politics, 8(3), pp Raina, P., House of Lords reform: A history, Vol. 1 & 2. Frankfurt am Main: Peter Lang Edition. Raymond, C.D. and Overby, L.M., What s in a (Party) name? Examining preferences, discipline, and social identity in a parliamentary free vote, Party Politics, forthcoming. Richards, P. G., Parliament and Conscience. London: George Allen and Unwin. Russell, M., The contemporary House of Lords: Westminster bicameralism revived. Oxford: Oxford University Press. Russell, M., Parliamentary party cohesion: Some explanations from psychology, Party Politics, 20(5), pp Sieberer, U., Using MP statements to explain voting behaviour in the German Bundestag: An individual level test of the Competing Principals Theory, Party Politics, 21(2), pp van Vonno, C. M. C., Malka, R. I., Depauw, S., Hazan, R. Y., and Andeweg, R. B., Agreement, Loyalty, and Discipline, in Deschouwer, K. and Depauw, S. (eds.), Representing the People: A Survey Among Members of Statewide and Substate Parliaments. Oxford: Oxford University Press, pp

22 Table 1: Results of Five Free Vote Divisions Regarding House of Lords Reform in 2003 Division Party Abolition Appointment 100% Elected 80% Elected 60% Elected Labour 45.9% 49.0% 44.5% 43.8% 43.6% (45.2%/168) (48.6%/177) (44.3%/174) (45.1%/173) (41.5%/176) Conservative 1.4% 40.8% 42.5% 49.0% 33.8% (0%/62) (42.6%/61) (44.8%/58) (50.0%/64) (37.5%/64) Lib Dem 6.0% 6.1% 85.1% 93.9% 87.5% (0%/16) (0%/16) (86.7%/15) (100%/16) (93.8%/16) Other 45.0% 47.4% 71.4% 50.0% 35.0% (100%/2) (0%/2) (100%/2) (100%/2) (100%/2) Entries are the percentages of each party s MPs voting in favour of each respective amendment (with the percentage of each party s MPs voting in favour/total n in the BRS sample in parentheses).

23 Table 2: Determinants of Support for Each Division Division Predictor Abolition Appointment 100% Elected 80% Elected 60% Elected Support for Reform (0.21) (0.17) (0.16) (0.17) (0.18) Left-Right Scale (0.14) (0.11) (0.11) (0.11) (0.12) % Liberal Democrat * 0.04** 0.07** 0.08** Vote (0.03) (0.02) (0.02) (0.02) (0.02) % Conservative -0.22** 0.14** -0.09** -0.12** -0.11** Vote (0.11) (0.06) (0.05) (0.05) (0.05) Labour MPs -4.82* 7.66** -4.88** -6.52** -5.97** (2.90) (3.02) (2.02) (2.56) (2.25) % Conservative 0.22** -0.18** 0.13** 0.16** 0.16** Labour MPs (0.11) (0.07) (0.05) (0.06) (0.05) Constant 5.32* -5.82* (3.18) (3.13) (2.23) (2.66) (2.38) Wald Chi ** 27.71** 23.08** 36.80** 41.54** McFadden s R Observations * p < 0.10, ** p < 0.05, two-tailed tests. Entries are logistic regression coefficients with standard errors in parentheses.

24 Table 3: Predicted Probabilities Associated with the Liberal Democrat and Conservative Vote Share Variables Predicted Probabilities Division Lib Dem Conservative Abolition 1 σ Below 0.50 (0.13) 0.43 (0.67) Mean 0.42 (0.10) 0.44 (0.10) 1 σ Above 0.35 (0.08) 0.45 (0.01) Appointment 1 σ Below 0.49 (0.08) 0.54 (0.01) Mean 0.40 (0.06) 0.42 (0.06) 1 σ Above 0.31 (0.04) 0.32 (0.30) 100% Elected 1 σ Below 0.43 (0.66) 0.38 (0.89) Mean 0.53 (0.74) 0.51 (0.72) 1 σ Above 0.63 (0.81) 0.64 (0.44) 80% Elected 1 σ Below 0.40 (0.78) 0.40 (0.97) Mean 0.57 (0.88) 0.53 (0.85) 1 σ Above 0.72 (0.93) 0.65 (0.54) 60% Elected 1 σ Below 0.39 (0.67) 0.37 (0.94) Mean 0.56 (0.81) 0.52 (0.77) 1 σ Above 0.73 (0.90) 0.66 (0.44) Entries are the predicted probabilities of voting in favour of each division associated with changes in the Liberal Democrat and Conservative vote share variables at the mean, 1 standard deviation below and 1 standard deviation above the mean for each variable. Predicted probabilities simulate the likelihood of voting in favour of each measure holding all other variables at their median values. The predicted probabilities are for Labour MPs and all other MPs in parentheses.

25 Table 4: Predicted Change in Probabilities Due to Labour Party Affiliation Abolition 100% Appointed 100% Elected 80% Elected 60% Elected 1 sd below Mean 1 sd above 1 sd below Mean 1 sd above 1 sd below Mean 1 sd above 1 sd below Mean 1 sd above 1 sd below Mean 1 sd above Predicted Change in Probability of Voting for Reform Entries are the predicted changes in probability of voting in favour of each provision due to the variable measuring Labour versus all other MPs at different levels of the moderating variable (% Conservative vote). Conservative vote shares are held to the Labour MP-specific values (mean and standard deviation); all other variables held to their median values.

26 Appendix: Descriptive Statistics Predictors Mean/Std. Deviation Min/Max Support for Reform 2.88/1.56 0/4 Left-Right Scale 4.28/1.99 1/10 % Liberal Democrat Vote 16.80/ /60.18 % Conservative Vote 30.02/ /55.43 Labour MPs 0.70/0.46 0/1