Electoral and Parliamentary Incentives for Position Taking. Exploring Absence in the House of Commons, ÕÉÉß - óþÕþ∗

Zoltan Fazekas Martin Ejnar Hansen

Syddansk Universitet Brunel University [email protected] [email protected]

May óä, óþÕ¦

Abstract e ability to hold an MP accountable for his or her actions is one of the cornerstones of modern representative democracy. Voters are informed by their MPs when they, or their party, take a position on an issue, i.e. cast a vote. We argue that it is important for MPs to send signals to their constituents about their record and to their party about their willingness to support the party. In this respect it is interesting that a large number of MPs are absent for votes. In this paper we analyse the absence rates of MPs in the United Kingdom from ÕÉÉß to óþÕþ. We link absence rates to both parliamentary and constituency characteristics. Our ndings suggest that in terms of scheduling and work organization, government or committee service hinders the participation in votes, but more importantly we nd systematic association between the size of the constituency majority and vote participation. Likewise we nd a clear pattern in that MPs who will not run for re-election take an “early retirement” at least from participating in votes. e paper shows that absence is a more complex than previously thought and the ndings have impact not only for studies of the UK Parliament, but for studies of parliamentary voting more broadly.

∗Paper prepared for CEU Polberg Meeting, May óþÕ¦. Working paper.

Õ Introduction

When an election is held with a low turnout debate oŸen follows lamenting the decision of many voters to abstain from voting. When the happen in a parliamentary setting the outcry is much less clear, and where elections only happen infrequently absence from voting in parliaments is a daily occurrence when the parliament is sitting. Today, when a voter by chance stumbles across televised sessions of the deliberations of their parliament they will wonder where all their representatives are. In most chambers today only the number required to maintain a quorum will be present at the debate, oŸen those members interested in the topic. When a vote is called for a portion of the remaining members appear from around the parliamentary buildings to cast their vote for or against the proposal on the žoor. Even then it is very rare that all members of the parliament turn up; indeed, it might be seen as an anomaly if that was to happen, usually signalling a very important vote where the majority of the government would be in jeopardy. us most votes and hence, most bills, are passed with an oŸen times high proportion of the legislators absent. While this might seem a trivial parliamentary irregularity it is far from the case; the implications of parliamentary absence essential, both from a representation perspective and a more practical parliamentary perspective. First, the strong direct relationship between voters and representatives entails that voters elect representatives to represent their views and interests and punish those who act against these preferences (Ansolabehere and Jones óþÕþ, ¢˜ì). Evidence from the United States (e.g. Canes-Wrone, Brady, and Cogan óþþó) lends support to the notion that representatives who deviate from the preferences of their constituents will lose o›ce. Newer evidence suggests that legislators use their legislative action to claim credit in order to cultivate a personal vote (Grimmer, Messing, and Westwood óþÕó). In order for this to happen it is necessary that the constituents have information about the behaviour of their representatives in parliament and that the representatives have taken a stand on the issues before parliament. Second, absence from votes is a normal occurrence in any modern day parliament and is a much used part of the toolbox of a modern day legislator, not least due to the various other tasks they are given besides voting in the chamber. Despite the oŸ occurrence of absence there is still preciously little research on the topic, especially in parliamentary systems. It is treated as insignicant if an MP is not present at a vote or assumed to occur randomly or simply assumed that the absent member is paired (Benedetto and Hix óþþß, ßߘ). Despite the fact that one of the most important goals for at least the governing party is to pass its policies and in order to do so they need their MPs to vote regularly (R. Johnston et al. óþþó). However, we do not know if absence actually does happen by chance or whether there are patterns to absence. What we do know is that legislative voting is a necessity for enacting laws, and political history has seen dramatic incidents where the absence of a single legislator has had serious impact on the outcome.Õ In the latter cases it is oŸen times very clear who is to blame for the result by his absence. However, absence is most frequent in non-controversial votes, where it does not change the outcome. Indeed, for certain members – for instance the case of ministers – it would be surprising if they had a high attendance rates. e existing studies of absence use either bill characteristics to explain when legislators are absent or legislator or electoral characteristics. We argue that to fully understand absence it is necessary to include both sets of characteristics in a model to take into account the spe-

ÕFor instance the vote which caused the fall of the Callaghan government in ÕÉßÉ could have had a dišerent outcome if only one absent MP had voted in favour of the government.

ó cic dynamics of the legislative institution. In this paper we explore what might explain the decision of a member of the British House of Commons to be absent from a vote. We argue that this is not a random decision, but in fact governed by both bill and legislator character- istics. We have created a dataset of all parliamentary votes cast from ÕÉÉß-óþÕþ by all MPs. In total our data includes nearly ó.¦ million individual votes cast by ˜¢¦ dišerent MPs over Õó˜ó days. Our ndings suggests that some of the factors inžuencing absence in the U.S. Congress can also be found in the United Kingdom, whereas other variables are inžuential as we would expect them to be given the institutional setup of the House of Commons. e paper con- tinues with discussions of the role of MPs in the House of Commons and on the existing literature on absence and voting. From this we present our data and model before analysing the ndings.

e House of Commons

Members to the UK House of Commons are elected in single-member districts with the can- didate achieving the most votes winning. is means a strong link between representative and electorate who can reward or punish their representative more directly than had it been a pro- portional . Voting in Westminster style parliament resembles a government- opposition dimension (Hix and Noury óþÕÕ) which ts well with one of the key assumption of the tasks of a parliament; to vote in a government and to vote it out if necessary (Laver óþþä). e Government controls the agenda of the House of Commons and while deliberations on opposition initiatives are guaranteed it is very rare that they get accepted. Indeed, when mem- bers are voting it is in most cases on bills and on amendments to the bills, amendments which can come from both government and opposition. However, there is another possibility to avoid providing a preference on a vote. In the day to day running of the parliament pairing is an oŸen occurring event; for some votes the parties agree to hold back some of their members in order to keep the power balance in the chamber, freeing some members to be present at other functions, committees, or taking care of their ministerial o›ces. e institution of pairing is perhaps one of the few white spots on the map of legislative studies. As pairing is a voluntary agreement between parties there are no o›cial statistics on who is paired with whom, when and why, making it impossible to directly control for this in any empirical analysis. Indirectly it is possible, for instance through including a variable for cabinet membership. In the UK House of Commons the whips of the parties normally distribute a list to its legislators with the items of the agenda underlined with one, two or three lines. ree line votes cannot be paired, thus enabling the legislators to negotiate in between themselves which of the non-three line votes to be paired. It should also be noted, that as pairing is an agreement between the parties of the government and the opposition it can be terminated at any point in time. It would oŸen be the case if the opposition saw an opportunity to defeat the government in a salient vote. Pairing is also not an arrangement recognised by the rules of the House of Commons.ó For this particular study having the information on pairing in the just one of the three parliaments in question would be nice but as this is not the case we must make do with assumption regarding pairing, such as ministers being more absent, i.e. more paired than non-ministers. In sum, whether an MP is absent due to illness, foreign travel, strategically, or through pairing, he is not voting, and thus not taking a position in the votes.

óFactsheet PÉ: Division, House of Commons Information O›ce (version August óþÕþ).

ì Absence can also be seen as a cousin of the much more researched legislative action of rebellion. A rebellion takes place where a number of MPs votes against the party line on an issue. e history of the House of Commons has seen a number of such rebellions either bringing down governments or undermining the government of the day. Rebellions are oŸen given much attention in media and in recent years a considerable scholarly interest has also build up on this topic. (e.g. Benedetto and Hix óþþß; Cowley óþþó; Cowley and Norton ÕÉÉÉ; Norton ÕÉߢ; Pattie, Fieldhouse, and Johnston ÕÉɦ). e outcome of a rebellion can lead to the MP losing the either for a limited time or permanently. It is, bar crossing the žoor, the strongest expression of unhappiness with party policy that an MP can use. Considering situations where an MP might not favour party policy on an issue but at the same time does not attach high salience to the issue, then the costs associated with rebelling might be too high. Here absence becomes a procedure which can help the MP avoid supporting a proposal he or she is against, without risking the full wrath of the party leadership and party whips. MPs rebelling can oŸen be distinguished between those only rebelling on a few issues and those more seasoned. In the former case the rebellion is a sign given on issues of high saliency for the MP,while this is also the case for the latter group there might also be a tendency for those to be more frequently against the party. However, being a member of a party also means that consistent rebellion might not be tolerated leaving the rebellious members with a need for other options to signal discontent on low saliency issues. Here absence might be a useful remedy for the seasoned rebels. In this respect it is remarkable that so little attention has been devoted to absence in the study of voting in the House of Commons.

Absence in Legislatures

e lack of attention given to decisions by legislators to turnout or not is not unique to studies of the UK Parliament, they are barely mentioned in many studies of legislator vote choice (Forgette and Sala ÕÉÉÉ, ¦äÉ). However, a few exceptions do exist. Most of those exceptions are based on data from the U.S. Congress, although in recent years there has be- come more focus on absenteeism in other legislative settings. In the literature on the U.S. Congress the turnout decision is usually explained by either using legislator or bill charac- teristics. Lott (Õɘß; ÕÉÉþ) and Zupan (ÕÉÉÕ) are examples of the use of the former, while Cohen and Noll (ÕÉÉÕ) and Poole and Rosenthal (ÕÉÉß) represents the use of the latter. For- gette and Sala (ÕÉÉÉ) also use bill characteristics and nds that the more salient a vote is for a party the more likely it is that their members will turnout. e ešect of constituency on the voting behaviour of members of Congress has been analysed (Mayhew óþþ¦; Arnold ÕÉÉó; Ansolabehere and Jones óþÕþ). Others have also used electoral factors to explain by the same members (e.g. Cohen and Noll ÕÉÉÕ; Rothenberg and Sanders ÕÉÉÉ; Rothenberg and Sanders óþþþ). Hart and Munger (ÕɘÉ) argue that transactional costs like the distance from constituency to capitol have an impact on absence rates. Jones (óþþì) presents a model of strategic position avoidance by U.S. Senators, in which he nds that position avoidance is not a random phenomenon but has a signicant link to electoral contexts. Poole and Rosenthal (ÕÉÉß) argue that absence is, among other things, based on indišer- ence between the choices presented, the cost of voting and closeness in the votes. Rothenberg and Sanders (óþþþ) analyses the number of abstentions on each roll call by using a negative binomial model. ey nd that turnout are predictable, though it cannot be explained by the standard assumption of benets connected to the votes, nor by the likelihood of a legislator

¦ being pivotal in a vote, which is contrary to previous studies. Instead the decision to abstain is guided by a choice between running for re-election and legislating. However, the results rest on votes from only one Congress, thus opening up for critique, especially if this Congress were an anomaly to others. Arnold (ÕÉÉó) uses the voting behaviour of congressmen to ar- gue that while the Congress votes in a large number of divisions they mostly do so on trivial issues, such as incrementally reducing or increasing a budget with the more important deci- sions being made behind closed doors. is constitutes a problem for the electorate to hold the legislators accountable for their votes (Arnold ÕÉÉó, ó䢖óßó). Mayhew (óþþ¦) suggests that it is the incentive for re-election that drives the members to take and avoid positions on issues. is is corroborated by Cohen and Noll (ÕÉÉÕ) who nd that prospects for re-election matters a great deal for the vote choice of legislators. ey also nd that abstention and absence are not random events but is oŸen use purposely by legislators in order to enhance inžuence and pay oš especially when they perceive the result to become very close. However, the generalization of the ndings by Cohen and Noll are critiqued by Rothenberg & Sanders (ÕÉÉÉ, ìÕ¦) who argue that since Cohen and Noll only looks at one time point they are excluding the possibility of variation across the election cycle, and since they only presents results based on votes on a single issue, though important, the ndings must be treated with caution. Absence, or as it is sometimes termed in the literature, position avoidance, is a possible strategy for members of the U.S. Congress. It is also a valid strategy in other legislative set- tings, for instance at the state level in the US where Riker (Riker Õɘä, ¢˜ß:Õþì–Õþ¢) claims as an example of strategic absence, a vote from the Virginia state legislature where a vote was won by not voting. While absence remains somewhat under-researched in the case of the United States it is nothing in comparison with the almost non-existent studies on absence in multi-party parliamentary systems. ere are but a few studies on absence in other leg- islative setups than the US Congress. Among the exceptions are Longley (óþþì) on a single controversial bill in the Canadian House of Commons, Noury (óþþ¦) on abstention in the European Parliament and Johnston, Cowley, Pattie and Stuart (óþþó) on Labour MPs in the UK House of Commons. Scully (ÕÉÉß) also works on the European Parliament, and shows that in areas where the EP has the most inžuence the highest number of MEPs will partici- pate in the vote. e limited focus on MP absence is also seen in one of the newest works on comparative party cohesiveness by Kam (óþþÉ) who presents a comparative study of dissent Canadian, Australian, British and parliaments and ignore absent MPs. e key assumption in most of the studies is that a legislator can choose his position on bills independently of parties, and this respect that the MP has an ideal position which is dif- ferent from his or her party. While this has been included in a number of studies of absence in the U.S., e.g. Lott (Õɘß); Rothenberg and Sanders (ÕÉÉÉ; óþþþ; óþþó), it is not common to use individual positions for the study of MP behaviour in the case of the United Kingdom. One reason for this is that expert surveys or self-reporting of policy positions are rarely included in survey and if included only a small sample reply. In their study of rebellion among Labour MPs Benedetto and Hix (óþþß) use a sample of Labour MP positions obtained from a candi- date study. Another possibility would be to utilise the votes to derive ideal point estimates for British MPs, however, as shown by Spirling and McLean (óþþß) and later also Spirling and Quinn (óþÕþ) the impact of makes it di›cult to distinguish MPs from one another allowing only for the discovery of voting blocs. In other words, the assumption that individual MPs have distinct ideal positions dišerent from that of their party is problematic when working with most .

¢ An important question thus becomes, how can a member avoid going against his own party and at the same time not provide a preference which contradicts his own positions? If for some reason an MP wishes not to divulge his preference on a vote he is leŸ with little alternative than to be absent. One such alternative would be to seek a pair from the party whip allowing the MP to be absent with the permission of the party. is would leave the MP in the clear with the party, but not with the electors of the MP who would not have that MP to hold responsible for the decision made by the parliament. Avoiding divulging a preference might be useful in several cases. It could prevent problems with the constituency; it could be based in an ideological stand; it could be harmful later in the members’ career, as we should not forget that every issue is a potential campaign issue (D. R. Jones óþþì, ˜¢ì). Pairing is for the UK House of Commons something slightly dišerent than in other parliaments and there are specic institutional issues which warrant a closer look on the House of Commons of the United Kingdom before the data and methodology used are presented. Using literature based on the U.S. Congress it is important not to translate the theoretical expectations directly as they are based on a very dišerent, and unique, institutional setup (Gamm and Huber óþþó). is entails formulating our expectations based both on the literature on absence, which is predominantly American and the particular institutional setup of the House of Commons. is is important for our expectations with regards to ndings, as presented below. In his study of absence in the European Parliament Abdul Noury (óþþ¦) argued that the day of the week was important for the decision by the MEP to be absent or not. e logic applies also to the decision facing MPs in the House of Commons where votes can be held on all day of the week. Noury also argued that the closer the vote the more likely MEPs were to turnout. Similar arguments were given by Rothenberg and Sanders (ÕÉÉÉ; óþþó). For the present we use these insights to hypothesize that: Day-HÕ. Absence will be dependent on what weekday the vote is scheduled. Vote-Hó. e closer the vote the less likely it is that an MP will be absent. For these the rst hypothesis, we consider an additional aspect related to the particular day of the vote: how many bills are voted on that day. Accordingly, our focus here is to purge articial dišerences that are associated with the fact that on some days, the House of Commons has Õ-ó votes, whereas on other days, this can be as high as Õ¦-Õß votes. e decision to be present on a particular day can be inžuenced by the number of votes the MP: although not considered here directly, if there are more votes on a particular day, odds are there will be a bill discussed and voted on that is salient for the MP, but this might also mean that, if already there, the MP will be present for other votes as well. Or, if absence is deemed undesirable by ones party, absenting on a day with many votes is more costly for the MP compared to absenting on days with very low voting activity. Yet again, as we will discuss it in the next section, we consider this aspect of the “day” as a necessary control, not directly a feature that where we ošer a detailed hypothesis. Lott (Õɘß; ÕÉÉþ) found that member turnout in the U.S. Congress tended to decrease in the session before voluntary retirement, see also Jones (óþþì). Given that elections in the UK happen with much larger intervals than in the U.S. House we believe that the decision taken by a member to not run again can, if taken early, result in a much lower participation rate. us, we hypothesize that: MP-HÕ. Absence will be higher for members who have decided to retire at the next election.

ä In a study of member turnout over ìþ years in the U.S. Zupan (ÕÉÉÕ) found that Democrats turned out less for votes than Republicans due to a greater focus by Democrats on con- stituency services. In this paper it is argued that it is not party belonging per se but government- opposition that is important for a turnout. In connection with this is the hypothesized rela- tionship by Longley (óþþì) that absence is positively connected to seniority, i.e. the more seniority the more likely absence, electoral support, i.e. the higher the majority an MP holds the higher the likelihood of absence, and nally that MPs who serve in the cabinet will also be more absent. ese expectations are all valid for the UK case; although we argue that membership of the shadow cabinet should also be taken into consideration when speaking of absence among MPs. If the logic is that absence is due to other responsibilities that take time from the MPs, the shadow ministerial role should be taken into consideration as well. We hypothesize the following relationships:

MP-Hó. MPs for opposition parties will be more likely to be absent.

MP-Hì. Absence will be higher for cabinet and shadow cabinet members.

MP-H¦. e more seniority held by an MP the more likely absence is.

MP-H¢. e higher the majority an MP enjoys the more likely the MP is to be absent.

However, as we have discussed the motivation behind absence as a possibility to avoid position taking that might go against the party line, when analyzing the relationships between absence and party or the vote majority (Vote-Hó), we add a qualitative dišerentiation whether MPs are absent at bills introduced by their own side, or by the other block in the House of Commons. If the goal is to avoid rebellion in a case when MPs have divergent opinions from that of their party, absence should be higher on bills that were introduced by their own party, but the expected outcome (in terms of vote majority) should factor into this decision: a bill deemed to pass either way (higher majority on the proposing side) a vote against that bill might trigger less punishment within the party. As seen here, we do not hypothesize a specic relationship here, instead we argue that this aspect of the bill should be taken into consideration when drawing inferences about dišerent patterns of absence of opposition and government party members. Hart and Munger (ÕɘÉ) and Rothenberg and Sanders (ÕÉÉÉ) include a distance measure in their studies of absence measuring the distance between the mid-point of a congressional district to Washington D.C. While the distances in the UK are much smaller than in the US it is not hard to envisage that distance may inžuence the decision to be absent. Members for constituencies that are distant from Westminster face a long journey aŸer the closing of parlia- mentary session, though most MPs also hold residence in London allowing for the commute to constituency not being a daily occurrence. On the other hand MPs serving constituencies close to Westminster might not have second homes in London, or even if they do nd it more di›cult to say no to appear in constituency events on weekdays. Additionally, other duties the MP has outside voting in the House of Commons can factor into the decision to be in London or not. Depending on other scheduled activities with her party group, presence in the House of Commons can be only a partial reason for staying in London. Overall, a large number of factors can overwrite or reinforce the previously reported positive relationship between constituency distance and absence, conditioned also on other roles that anchor the

ß MP to London. In this sense, the rst place to start considering heterogeneity is whether the MP is member of the party in Government or Opposition. Not least since it is the role of the Government to keep its majoirty in the chamber and the committees, which should lead to opposition MPs have a less constrained role with regards to being present for votes.

MP-Hä: e larger the distance from Westminster to an MP’s constituency the higher the likelihood of absence, but this ešect can vary depending whether the MP is member of an Opposition or Government party.

We now turn to the presentation of the data and methodology used to test our hypotheses.

Data and methodology

Our analysis covers three legislatures – ÕÉÉß-óþþÕ, óþþÕ-óþþ¢, and óþþ¢-óþÕþ – including all bills voted on in the House of Commons. e votes were centralized and made available by e Public Whip project (ePublicWhip óþÕþ) and our outcome variable is coded as Õ if a particular MP was absent for the vote on a particular bill. When present, independent of the vote choice, the variable is coded to be þ. Accordingly, each MPs can be present or absent for each bill voted on in the House of Commons. In this sense, our data has a hierarchical structure: absence is a bill specic quantity that is nested within MPs. However, a further nestation level needs to be considered, namely the day of the vote. Each day, multiple bills are discussed and voted on in the House of Commons. But if an MP is not in London for a particular day, it also implies that will be absent from all votes during that particular day, simply suggesting that votes held on the same day are more similar to each other (with regards to our outcome variable) than to votes held on other days. Also, our rst hypothesis is related to which weekday bills were voted on, we use the date as a second nestation level as well. Consequently, absence is nested within MP and the date of the vote, and these nestation levels are equal in terms of hierarchy, resulting in a cross- classied model with two independent second-level groupings. is specication allows us to handle the possible non-independence issues. In practical terms, data handling terms, this simply means that the bill × MP original matrix is transformed from wide to long format. Finally, as our outcome variable is dichotomous, we use a logit link function in the analyses, modeling directly the probability of absence vs. presence. We introduce below the opera- tionalization of the quantities described in the theoretical section. Bill specic predictors: At the lowest level of the hierarchy (level-Õ), we include two predictors. First, we dišerentiate between bills proposed by the Government and Opposition bills, as heterogeneity in terms of motivation can appear depending on who proposed the bill in question (coded as dichotomous variable taking the value Õ for bills proposed by the Government). Second, we account for dišerences in how close the vote was in terms of vote TotalYes−TotalNo majority, operationalized as: abs( TotalYes+TotalNo ) . In this case, þ would režect a perfect tie and Õ would indicated that all present MPs voted in the same way (for or against the bill). e interaction between these two variables gives the combined information of how far oš was an Opposition proposed bill from passing, or with how large of a margin did a bill proposed by the Government pass.ì Yet again, we need this information because our argument discussed

ìAn alternative operationalization would be to keep both negative and positive margin scores (where nega- tive scores režect bills that did not pass) and model a non-linear relationship. Our results are substantively the same if that operationalization is chosen.

˜ possible dišerences in how MPs will approach bills that were proposed by their own side on one hand, and depending on the “chances” of the bill passing. As discussed below, this motivates the inclusion of a three-way interaction between the MPs party membership and the two quantities described here. Date specic predictors: We include three predictors at this level. First, as a režection of our rst hypothesis, we use a nominal variable capturing which day of the week the bill was voted on, using Wednesday as the baseline category. Secondly, we add a control variable that measures the closeness of the next national election. Although not hypothesized directly, given the pre-campaign period and its necessary activities (community meetings, campaign- ing, fund-raising, etc.) we deem it important to include a control that marks at which stage of the legislature the specic bills were voted on. is is a continuous variable expressed as number of days till the next election, and rescaled as discussed above. Finally, as per the dis- cussion in the previous section, the total number of bills voted on for each day is included in the model. MP specic predictors: All but two MP specic variables are from “e British Par- liamentary Constituency Database, ÕÉÉó-óþþ¢, Release Õ.ì” (Norris óþþ¢). roughout the whole period under consideration, Labour was in government. Accordingly, we use a Gov- ernment variable that takes the value Õ if the MP is member of the Labour party, þ otherwise. Retirement was coded as a dichotomous variable (Õ for retiring next election, þ otherwise). ose MPs who re-ran for o›ce but lost are not coded as retiring. Hence, the condition was that the MP did not run in the next elections. Our data does not include cases where the MP did not run in the óþþÕ elections, but returned to politics and ran for o›ce in óþþ¢. Based on o›cial records of governmental composition and parliamentary records, min- isterial role was coded as a dichotomous variable (Õ for ministers and shadow ministers, þ otherwise). Constituency majority režects the winning margin in the previous elections, ex- pressed in terms of percentages. It režects the dišerence in the percentage of popular votes between the winner and the runner-up in a constituency. e last two variables are based on our own coding. Seniority has been coded as the rst year an MP entered Parliament based on o›cial MP biographies. In most cases MP senior- ity entails an unbroken length of service. Where there was a break of service, the rst year of entry is used as seniority value.¦ Constituency distance from Westminster is measured in kilometers and režects driving distance. We have used the largest local authority in a par- ticular constituency as a starting point and Westminster as an endpoint in the Google Maps API to retrieve distances. We chose the starting point because even if MPs take the train, it is likely that they will do it from the larger towns or cities. Furthermore, it is not sure whether MPs drive or take the train to Westminster, but even if there are minor dišerences between driving distance and distance for the train, we believe this is not a systematic bias depending on constituency. For the current iteration of the paper, we excluded three types of MPs from our empirical analysis: Õ) MPs from Northern Ireland, ó) MPs who switched parties, and ì) MPs who leŸ the House of Commons for various reasons, and accordingly, there were by-elections held in their constituency. In the case of MPs from Northern Ireland we are looking at a subset of MPs who have extremely high absence rates, but are also located very far away from Westminster. Our results related to constituency distance would have been biased upwards if we had kept

¦For example, Sir Peter Tapsell who represented Nottingham West from ÕÉ¢É until losing his seat at the ÕÉä¦ election. He re-entered parliament at the ÕÉää election for what is now Louth and Horncastle. In this case, his seniority would be taken as ÕÉ¢É.

É them in the data, and more importantly, Northern Ireland has a completely dišerent party system which actors’ rarely play a signicant role in the politics of Great Britain. e last two types dropped are mostly due to the fact that those MPs who change parties are suspect of having very dišerent motivations and parliamentary behaviour, already possi- bly before the party switch. Similarly, in the case of involved in by-elections, some of which are due to illness and death which in some cases are due to long-time illness and hence, pro- longed absence. is also means that MPs who join as a result of a by-election are not included until then have entered parliament aŸer a general election. e nal dataset analyzed (with all predictors having valid values) is comprised by ˜¢¦ MPs covering Õó˜ó days of parliamentary activity. e full vote-MP data matrix has ó,ìßÕ,ìì˜ observations (MP x bill pairings). If the MP was a member of more legislatures, we consider it as one individual, also because we control for the date of the parliamentary (in)activity. Table Õ presents descriptive statistics for all of the introduced predictors in their original form. In the multivariate models, all continuous variables were recoded by subtracting the grand-mean and dividing by two standard deviations, this making it possible to get directly comparable coe›cients with those for dichotomous predictors (Gelman and Hill óþþß).

Table Õ: Descriptive statistics, ÕÉÉß-óþÕþ

Continuous variables Mean SD Distance from Westminster ó¢ß.¢ó óÕþ.¦ó Constituency majority óó.þóß Õä.¢˜ Seniority ÕÉÉó.äó Õþ.þÕ Number of bills/day ¦.ìä ó.ߢ Days till election -˜óþ.þ¢ Õììþ.ßó Vote majority þ.¦óß þ.ó¢¦ Indicator variables Õ Government party äÕ.ÉÛ Retiring óì.¢Û Minister óó.óÛ Government proposal ¦É.˜Û

In the next section we present the results from the multivariate analysis. First, we discuss the MP and date specic averages estimated in the baseline hierarchical model, and then discuss the ndings from the model including all the predictors of interest.

Results

e rst two gures display estimated average absence probabilities based on the baseline model including no predictors. As apparent from these gures, we do have quite substantial variation both between MPs and also between days. e overall average absence probability is þ.ì¢, and as displayed in Figure Õ, most MPs indeed have below þ.¢ probability of absence. is simply means that our baseline prediction suggests participation, however, denitely not full participation at each an every vote in the House of Commons. More importantly, we see that some particular MPs deviate a lot from the average absence rates, being almost inexistent in terms of votes cast in the House of Commons. Simply eyeballing some of the MPs with above þ.˜ probability of absence, we can spot cabinet members and and MPs like Judith Church who

Õþ succeeded former Labour shadow minister Bryan Gould in Dagenham in ÕÉɦ, but retired already in óþþÕ, or Andrew Welsh the SNP MP for various Angus constituencies from ÕÉߦ-ßÉ and ÕÉ˜ß until his retirement in óþþÕ.

Figure Õ: Predicted absence probability of MPs.

Model based predicted average absence probability for each M with names added for MPs with estimated probability of absence above þ.˜, ÕÉÉß-óþÕþ. Dashed red line represents the mean probability across all days and all MPs.

Figure ó: Predicted absence probability for each day.

Model based predicted average daily absence probability, ÕÉÉß-óþÕþ. Dashed red line represents the mean probability across all days and all MPs.

e daily averages – displayed in Figure ó – reveal an interesting pattern. As expected, the averages estimated for Fridays are the highest ones, clearly deviating from the average absence rates. Additionally, it is not only that absence is very high on Fridays, it is also ap- parent that it is markedly dišerent from all other days, when looking for systematic patterns.

ÕÕ ere are other weekdays that in some cases register high absence rates, but this happens only sporadically. e second column of Table ó displays the results from our full model with predictors at all levels. Given the sample size and the nature of the population, we will focus on the magnitude of the estimated ešects.¢ In line with our expectations, we can see that there are systematic dišerences in the es- timated absence probabilities depending on which day of the week the bills are voted on. Everything else held constant at their meansä, the average absence probability on a Wednes- day is þ.¦Õ, essentially identical for Tuesdays as well. ere are the most frequented days, and close to them we nd Monday (þ.¦¦), and at substantively higher absence probabilities ursday (þ.¢¢) and especially Friday, with a frightening þ.Éì predicted absence probability. In simple terms, whereas for the rst three days of the week, our model predicts presence in the House of Commons, for the last two days absence becomes more likely. Unsurprisingly though, as displayed in Table ì, the lowest number of votes take place on Fridays, and it is more likely that Fridays host bills proposed by the Opposition. Conversely, Wednesday is the example day of many votes held, with slightly higher percentage of bills proposed by the Government. Continuing with day specic predictors, we nd that when more bills are voted on during a day, absence probability is substantively lower. For example, the estimated probability for a Wednesday with three votes is þ.¦ó, and this drops to þ.ìä if twelve votes are held. Similarly, when there are roughly three years until the next election, absence probability is estimated at þ.ì˜, but at þ.¦¦ when only Õþ months are leŸ until the next elections. e comfort of previous success in the elections substantially inžuences the likelihood of the absence. MPs with strong support režected by large winning margins in their constituen- cies are less present in the House of Commons: those who won their constituency with a ìÛ margin have an absence probability of þ.ì¦, whereas those who won with a ìþÛ margin are substantively more absent (þ.¦¦). MPs holding cabinet positions are more likely to be absent than their colleagues with no such responsibilities, but the dišerences are very small: for example, a Labour MPs with no ministerial duties has an estimated probability of absence of þ.óþ, compared to þ.óÕ for her colleague with ministerial duties. Similarly, when retirement is considered aŸer the end of the current legislative period, the probability of absence is slightly higher (þ.¦Õ vs. þ.¦¢). Finally, and contrary to our expectations, in terms of MP specic predictors, we nd no evidence for systematic dišerences as a function of seniority. For a better display of our results involving variables included in interactions, Figures ì and ¦ display the change in predicted absence probability, conditional on the terms included in the interactions. Undoubtedly, as displayed by Figure ì, Labour MPs is more present at votes in the House of Commons. However, when the vote margin is very low – namely we are looking at extremely close votes, there is no real dišerence depending on the Government vs. Opposition dichotomy: Opposition MPs mobilize as well, registering a very low rate of absence. However, we also see that, for Labour MPs, absence probability is substantively un- changed or independent of the vote majority. e opposite is found for Opposition MPs,

¢As reported in the table, when compared to the baseline model, our full model ts the data signicantly better, and we can also see reduction in terms of unexplained variance on the second level. äUnless otherwise stated, the dišerences presented as examples in this section are values predicted when all continuous variables held constant at their means and we are looking at an opposition MP with no ministerial position, no plans of retiring, being present or absent for a vote on an Opposition proposal.

Õó Table ó: Model results, logit coe›cients, ÕÉÉß-óþÕþ.

Absence Baseline Full model Friday ó.ÉÉþ∗∗∗ (þ.þßÕ) Monday þ.ÕÕß∗∗∗ (þ.þ¦¦) ursday þ.¢¢˜∗∗∗ (þ.þ¢¦) Tuesday −þ.þÕ¢ (þ.þ¦ó) Days till next elections (óSD) −þ.óߢ∗∗∗ (þ.þìó) No. bills voted that day (óSD) −þ.Õóä∗∗∗ (þ.þ¦¢) Minister þ.þìþ∗∗∗ (þ.þþ˜) Retiring next election þ.óþþ∗∗∗ (þ.þþß) Majority size within constituency (óSD) þ.¦¢ß∗∗∗ (þ.þþÉ) Absolute vote majority (óSD) Õ.äÉÕ∗∗∗ (þ.þÕþ) Member of party in government −þ.ÉÉÉ∗∗∗ (þ.þìÕ) Government proposal þ.þ¦ß∗∗∗ (þ.þþä) Distance from Westminster (óSD) þ.ÕÉÕ∗∗∗ (þ.þ¦þ) Seniority (óSD) þ.þ¢Õ (þ.þ¦Õ) Vote majority × Government party −Õ.¦¢ì∗∗∗ (þ.þÕÕ) Vote majority × Government prop. −þ.Õäó∗∗∗ (þ.þÕó) Government party × Government prop. þ.þɦ∗∗∗ (þ.þþä) Government party × Distance from Westminster −þ.ߦÕ∗∗∗ (þ.þ¢Õ) Vote majority × Government Prop. × Government party þ.Õä¢∗∗∗ (þ.þÕ¦) Constant −þ.ìäó∗∗∗ (þ.þ¦ä) MP level variance þ.ߦ þ.ßþ Day level variance Õ.óó þ.ìó N ó,ìßÕ,ìì˜ ó,ìßÕ,ìì˜ MPs ˜¢¦ ˜¢¦ Days Õó˜ó Õó˜ó Log Likelihood −Õ,óßä,¢˜ä.þþþ −Õ,ó¦Õ,ìÉì.þþþ AIC ó,¢¢ì,Õߘ.þþþ ó,¦˜ó,˜ìþ.þþþ BIC ó,¢¢ì,óÕä.þþþ ó,¦˜ì,Õþ˜.þþþ Note:∗p<þ.Õ; ∗∗p<þ.þ¢; ∗∗∗p<þ.þÕ Baseline: member of opposition parties; Wednesday. (óSD) = ó SD standardization, see variable description.

Õì Table ì: Number and type of bills for weekdays Day Number of bills Proportion of government proposals Friday ó.¦ì þ.ìÉ Monday ¦.ó¢ þ.¢ó ursday ì.ää þ.¦É Tuesday ¦.¦ì þ.¦ä Wednesday ¦.˜ì þ.¢ì who are substantively more likely to be absent if there is a strong voting majority for a partic- ular bill. In both cases, it does not matter too much whether the MPs own party (or group) proposed the legislation or the opposite group.

Figure ì: Vote majority and probability of absence.

Model based predicted ešect of vote majority from Westminster plotted separately for government vs. op- position depending on the proponent of the bill, ÕÉÉß-óþÕþ.

Finally, constituency distance raised several issues in terms of theoretical expectations, and the results reported in Figure ¦ suggest that its association with absence is very much conditional on whether we are looking at an MP from the Government party or from one of the Opposition parties. We nd that probability of absence is substantively higher for MPs from more remote constituencies, but only in case of Opposition party members. Further inspection of the data revealed that this dišerence is driven mostly by the MPs who are not from the three largest parties (Labour, Conservatives, and Liberal Democrats). When party specic heterogeneity is considered, there is a weak negative relationship for the three major parties, and a strong positive relationship for MPs of other parties. Furthermore, there is a systematic relationship between the average constituency distance and party, Conservatives being on average closest to London (Õ¢þ km), then Labour (óߘ km) and Liberal Democrats (ììþ km), and then all other parties (¢ó¦ km). constituency distance did not have the expected relationship with absence. is might be due to the support for purchasing or renting second homes that MPs have, or at least did have during the parliaments under examination, which

Õ¦ Figure ¦: Consituency distance and probability of absence.

Model based predicted ešect of constituency distance from Westminster plotted separately for government vs. opposition, ÕÉÉß-óþÕþ. is before the infamous expenses scandal which hit the headlines in óþþÉ. It remains to be seen whether the same relationship can be found in the coming parliaments where the rules for second homes have been tightened.

Discussion

Clearly, absence is an important behaviour for political scientists studying political decision- making, accountability, and political competition within legislatures. However, should the average voter worry if her MP is more absent than present in the House of Commons? In order to answer this question, we argued that we need to understand why MPs are absent. is distinction is important. While an MP can be absent from the House of Commons, in the meantime they might be doing constituency work or taking meetings with possible future investors in their constituencies. Or, they can simply be absent because they have other political duties, such as ministerial delegations or state level meetings. is points to one important limitation of our current study: we cannot assess what the MPs are doing while not in the House of Commons, but theories of MP behaviour and motivation joined together with focus on the institutional constraints can help in further explicating the possible implications. For example, as elections are closer, many MPs start already campaigning to reinforce their chances of re-election in their constituencies. is work takes them away from the House of Commons, and we nd higher absence rates in election-proximate periods. But besides this example of electoral motivation, MPs knowing that they do not seek re-election anymore, have less persuasion work to do both in terms of electorate and within-party com- petition. As they do not plan to continue their work as an MP, the range of possible electoral or party level sanctions is very much reduced. Instead, they might be planning their future careers outside the parliament, and consequently registering higher levels of absence. Ad-

Õ¢ ditionally, costs of attending votes might be very dišerent for MPs who live in London and those who have to embark on a longer journey. Traveling to London and then staying for those days where many votes are held (Tuesday, Wednesday), but leaving for days where the parliamentary activity is lower can be a rational decision. And yet again, if there are few non- salient bills on the agenda and many colleagues are either way absent from the vote, MPs can use that time for other activities without too many costs. is indeed raises the question whether accountability is damaged by no position taking, but this question also boils down to e›cient organization of parliamentary activity, and ultimately a rationalization of public spending that nds a balance between nancing MP costs for travel or second residence and reduces logistic obstacles for presence in the House of Commons. Yet again, in this mixture of institutional, electoral, and individual level determinants, we need to vote specic aspects as well. In many cases, if chances of passing a particular bill are high, the possible penalties or implications for absence are lower. Although our analy- sis considered bill specic predictors, we clearly need to consider the topic of the bill voted on, where dišerences across policy domains, or simply dišerences between procedural and substantive bills can arise, a missing feature that constitutes an important limitation of our study. A further limitation which regrettably cannot be rectied in a later iteration of this paper is the black box of pairing. As discussed previously some MPs are given carte blanche to be absent on a particular vote or a particular day as one of the members for the opposing side agree to also being absent. ere is a lack of research in this area, not least due to the problem of getting data on pairings. Pairing does involve ministers and shadow ministers, but also ordinary MPs are at times paired. Whether the ndings presented in this paper are exacerbated by pairing is hard to predict. What we can say is that pairing is also absence, and it entails a cognitive decision being made about partaking in this activity, and is sanctioned by the whips. We also cannot rule out that there is in fact no dišerence between MPs who are paired absent and those who are absent with pair, or absent without leave. In this paper we have shown that absence in the House of Commons is an empirical re- ality that follows some well dened patterns. Although it might be argued that absence is a normatively undesirable behaviour, our focus was to build a case that in order for such claims, we need to understand the underlying factors that inžuence the probability of absence. Ac- cordingly, our goal was to ošer a detailed analysis of which MPs are more likely to be absent from the votes in the House of Commons, and using this systematic information is the rst step in order to discuss the implications of absence. ere are a myriad of future avenues for both extending the scope of analysis or further substantiating the claims made in this paper. As outlined in the theoretical section pairing is a unique and important feature of the MP behavior in the House of Commons. Future research ought to devise possibilities of measur- ing or inferring pairing within the House of Commons, and extending the early steps of the present work by dišerentiating between absence that is “covered” or absence that is outside the realms of pairing.

References

Ansolabehere, Stephen, and Philip Edward Jones. óþÕþ. “Constituents’ Responses to Con- gressional Roll-Call Voting.” American Journal of Political Science ¢¦ (ì): ¢˜ì–¢Éß.

Õä Arnold, R Douglas. ÕÉÉó. e Logic of Congressional Action. Yale University Press.

Benedetto, Giacomo, and Simon Hix. óþþß. “e Rejected, the Ejected, and the Dejected: Ex- plaining Government Rebels in the óþþÕ-óþþ¢ British House of Commons.” Comparative Political Studies ¦þ (ß): ߢ¢–ߘÕ.

Canes-Wrone, Brandice, David W Brady, and John F Cogan. óþþó. “Out of Step, Out of O›ce: Electoral Accountability and House Members’ Voting.” American Political Science Review Éä (þÕ): Õóß–Õ¦þ.

Cohen, Linda R, and Roger G Noll. ÕÉÉÕ. “How to Vote, Whether to Vote: Strategies for Voting and Abstaining on Congressional Roll Calls.” Political Behavior Õì (ó): Éß–Õóß.

Cowley, Philip. óþþó. Revolts and Rebellions: Parliamentary Voting Under Blair. Politico’s.

Cowley, Philip, and Philip Norton. ÕÉÉÉ. “Rebels and Rebellions: Conservative MPs in the ÕÉÉó Parliament.” e British Journal of Politics & International Relations Õ (Õ): ˜¦–Õþ¢.

Forgette, Richard, and Brian R Sala. ÕÉÉÉ. “Conditional Party Government and Member Turnout on Senate Recorded Votes, ՘ßìÕÉì¢.” e Journal of Politics äÕ (þó): ¦äß–¦˜¦.

Gamm, Gerald, and John Huber. óþþó. “Legislatures as Political Institutions: beyond the Contemporary Congress.” Political Science: State of the Discipline: ìÕì–¦ì.

Gelman, Andrew, and Jennifer Hill. óþþß. Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge University Press.

Grimmer, Justin, Solomon Messing, and Sean Westwood. óþÕó. “How Words and Money Cultivate a Personal Vote: e Ešect of Legislator Credit Claiming on Constituent Credit Allocation.” American Political Science Review Õþä (¦): ßþì–ßÕÉ.

Hart, Douglas B, and Michael C Munger. ÕɘÉ. “Declining Electoral Competitiveness in the House of Representatives: e Dišerential Impact of Improved Transportation Technol- ogy.” Public Choice äÕ (ì): óÕß–óó˜.

Hix, Simon, and Abdul Noury. óþÕÕ. “Government-Opposition or LeŸ-Right? e Institu- tional Determinants of Voting in Legislatures.” Unpublished Working Paper.

Johnston, Ron, Philip Cowley, Charles Pattie, and Mark Stuart. óþþó. “Voting in the House or Wooing the Voters at Home: Labour MPs and the óþþÕ General Election Campaign.” Journal of Legislative Studies ˜ (ó): É–óó.

Jones, David R. óþþì. “Position Taking and Position Avoidance in the US Senate.” Journal of Politics ä¢ (ì): ˜¢Õ–˜äì.

Kam, Christopher J. óþþÉ. Party Discipline and Parliamentary Politics. Cambridge University

Õß Press.

Laver, Michael. óþþä. “Legislatures or Parliaments in Comparative Context?” e Oxford Handbook of Political Economy ß: ÕóÕ.

Longley, Neil. óþþì. “Modelling the Legislator as an Agent for the Party: e Ešects of Strict Party Discipline on Legislator Voting Behaviour.” Contemporary Economic Policy óÕ: ¦Éþ– ¦ÉÉ.

Lott, John R. Õɘß. “Political Cheating.” Public Choice ¢ó (ó): ÕäÉ–Õ˜ä.

———. ÕÉÉþ. “Attendance Rates, Political Shirking, and the Ešect of Post-Elective O›ce Employment.” Economic Inquiry ó˜ (Õ): Õìì–Õ¢þ.

Mayhew, David R. óþþ¦. Congress: e Electoral Connection. Vol. óä. Yale University Press.

Norris, Pippa. óþþ¢. “e British Parliamentary Constituency Database, ÕÉÉó-óþþ¢, Release Õ.ì.”

Norton, Philip. ÕÉߢ. Dissension in the House of Commons: intra-Party Dissent in the House of Commons’ Division Lobbies, Õɦ¢-ÕÉߦ. Macmillan.

Noury, Abdul G. óþþ¦. “Abstention in Daylight: Strategic Calculus of Voting in the European Parliament.” Public Choice ÕóÕ (Õ-ó): ÕßÉ–óÕÕ.

Pattie, Charles, Edward Fieldhouse, and RJ Johnston. ÕÉɦ. “e Price of Conscience: e Electoral Correlates and Consequences of Free Votes and Rebellions in the British House of Commons, Õɘß-Éó.” British Journal of Political Science ó¦ (ì): ì¢É–ì˜þ.

Poole, Keith T, and Howard Rosenthal. ÕÉÉß. Congress: A Political-Economic History of Roll Call Voting. Oxford University Press.

Riker, William H. Õɘä. e Art of Political Manipulation. Vol. ¢˜ß. Yale University Press.

Rothenberg, Lawrence S, and Mitchell S Sanders. ÕÉÉÉ. “Rational Abstention and the Con- gressional Vote Choice.” Economics & Politics ÕÕ (ì): ìÕÕ–ì¦þ.

———. óþþþ. “Legislator Turnout and the Calculus of Voting: e Determinants of Absten- tion in the US Congress.” Public Choice Õþì (ì-¦): ó¢É–óßþ.

———. óþþó. “Modeling Legislator Decision Making a Historical Perspective.” American Politics Research ìþ (ì): ó좖óä¦.

Scully, Roger M. ÕÉÉß. “Policy Inžuence and Participation in the European Parliament.” Leg- islative Studies Quarterly: óìì–ó¢ó.

Spirling, Arthur, and Iain McLean. óþþß. “UK OC OK? Interpreting Optimal Classication

՘ Scores for the UK House of Commons.” Political Analysis Õ¢ (Õ): ˜¢–Éä.

Spirling, Arthur, and Kevin Quinn. óþÕþ. “Identifying Intraparty Voting Blocs in the UK House of Commons.” Journal of the American Statistical Association Õþ¢ (¦Éþ): ¦¦ß–¦¢ß.

ePublicWhip. óþÕþ. “Votematrix for House of Commons, ÕÉÉß-óþÕþ.”

Zupan, Mark A. ÕÉÉÕ. “Local Benet-Seeking and National Policymaking: Democrats Vs. Republicans in the Legislature.” Public Choice ä˜ (Õ-ì): 󦢖ó¢˜.

ÕÉ