Representation in Wales: An empirical analysis of policy divisions between voters and candidates

Politics in Wales is often portrayed as being relatively consensual and enjoying healthy levels of trust between voters and elites when compared with the rest of the United Kingdom. Recent events, like the decision of Welsh voters to reject membership of the EU against the advice of most of its political establishment, are however calling into question this perception. Using the 2016 Welsh Candidate Study and 2016 Welsh Election Study data, this article evaluates the extent of policy divisions between voters and candidates in Wales. I find that candidates hold more liberal policy positions and are less likely than voters to think of immigration as the most important policy priority. In addition, they tend to favour a different approach to parliamentary representation, deeming it more acceptable for Assembly Members to discard the views of their voters in favour of their own views or those of their party.


Introduction
We live in an era where the majority of people in Britain do not trust politicians, believe that they do not care about what ordinary people think, and that they prefer playing party political games over furthering public interests (Fieldhouse et al., 2016). In fact, politicians are now less trusted than estate agents or bankers, with less than a quarter of people expecting politicians to tell voters the truth (Ipsos MORI, 2016). While sentiments like these are of course not unique to Britain (World Economic Forum, 2016), they nonetheless raise concerns about the current state of British democracy and the ability of elites to engage effectively with voters.
Against this backdrop of discontent, the political environment in Wales is often portrayed as being relatively harmonious. Although certainly not without its own divisions, Wales has shown that cooperation between the main political parties is possible, in a way that cooperation between the Labour Party and the Conservative Party in Westminster seems improbable; that a broadly supported vision for the pace and extent of devolution can be reached so that debates on independence are not quite as partisan and divisive as those in Scotland; and that a formal cross-community power-sharing agreement, like in Northern Ireland, is not necessary to allow unionists and nationalists to work together within the devolved system. Instead, political debates in Wales are often seen to play out on a slightly narrower ideological spectrum, characterised by soft-nationalist cultural politics, devolution-maximising constitutional reform, and a social democratic policy agenda (Jones and Scully, 2008;Moon, 2013Moon, , 2016. Welsh politicians also have a more positive reputation among voters. In contrast to the broader trend of disillusionment with politicians, most Welsh voters trust their Assembly Members and believe in their integrity (Scully and Jones, 2015b).
Recent events, however, provide some evidence that the political environment in Wales is not particularly consensual after all. Although the 2017 general election in Wales saw a return to traditional two-party politics, 1 one does not have to look hard to find instances of disconnect between voters and elites. In contrast to the widespread cross-party campaign in Wales, backing membership of the European Union, 2 52.5% of Welsh voters opted instead to 'take back control'. The Abolish the Welsh Assembly Party and the UK Independence Party (UKIP) Wales surprised many with their good performances at the 2016 Assembly election. The political environment in Wales appears to be in flux as voters seem increasingly willing to challenge the status quo. Policy divisions between voters and elites in Wales may, in fact, be more substantial than often portrayed.
This article looks at the extent to which policy divisions exist between voters and candidates in Wales. It does so by using data from the 2016 Welsh Candidate Study and the 2016 Welsh Election Study to compare voters' and candidates' views on the policy agenda, policy positions on a range of issues, and on whose interests Assembly Members ought to prioritise when carrying out their duties in the National Assembly for Wales. As such, this article provides a complex account of voter-candidate congruence (or the lack thereof) at different points of the policy-making process.
The analysis uncovers strong evidence that the relationship between voters and candidates in Wales is not particularly harmonious. Significant differences exist between them in all three aspects of the policy-making process included in this article. First, voters are more inclined than candidates to believe that immigration is the most important issue facing Wales and less likely to say the same of the economy. Second, voters tend to adopt more authoritarian policy positions than candidates. Third, voters and candidates differ in their beliefs about whose interests Assembly Members ought to prioritise when carrying out their duties in the Assembly. Candidates find it considerably more acceptable for Assembly Members to discard the views of their voters in favour of their own views or those of their party. Taken together, these differences suggest that the political environment in Wales is not particularly consensual after all.
The article is organised as follows. First, it describes the different elements of consensus that have been associated with Wales and the recent events that appear to challenge this consensus. Second, it outlines the different aspects of the policy-making process to be analysed and discussed in this paper. Third and fourth, it illustrates the data and measures used in the analysis and presents the empirical findings. Finally, the paper concludes with a discussion of the implications that emerge from the empirical evidence.

Political environment in Wales
The perception of a less divisive politics in Wales is usually based on the idea that a degree of common understanding exists over politics, policy, and implementation, which together lead to a more positive relationship between voters and elites. First, the political climate in Wales has been characterised by stability and cooperation. On one hand, the Labour Party has been the cornerstone of Welsh politics. Welsh Labour has been so central to the political life of the nation, having won the most votes in Wales at every general election since 1922, and having led the Welsh Government since it was first formed in 1999, that the state of one-party politics has earned Wales the nickname 'LabourLand' (Morgan and Mungham, 2000). On the other hand, there are examples of cooperation between the main parties in Wales. In instances where Welsh Labour has fallen short of a majority in the Assembly, it has found willing partners to form coalition governments in Plaid Cymru (2007)(2008)(2009)(2010)(2011) and the Welsh Liberal Democrats (2000 to date). The dominance of Welsh Labour, together with the willingness of other parties to support an administration led by it, highlights a rare degree of political stability and collaborative will. No other nation in the United Kingdom has witnessed the same level of consistency and dominance of a single party over such a long and uninterrupted period of time.
Second, policy debates in Wales are often perceived to play out on a quite narrow ideological spectrum. This practice has been labelled the Welshminster consensus, embodying (1) devolution-maximising constitutional reform; (2) social democratic policy agenda; and (3) soft-nationalist cultural politics (Moon, 2013(Moon, , 2016. In practice, this consensus manifests in broad acceptance of interventionist government programmes; willingness to extend the welfare state; asking for additional devolved powers but not independence; and the continuing commitment to Welsh language and heritage. Public opinion data suggest that these broad preferences are not only shared by elites but also enjoy strong support among the Welsh public (Scully, 2017;Scully and Jones, 2015a). Wales does not seem to play host to issues quite as divisive as the independence debate in Scotland or the unionist-nationalist clash in Northern Ireland.
Third, scholars point to emerging evidence of a 'made in Wales' approach to administering public policies. A well-documented example of this approach concerns youth justice which tends to be implemented less punitively in Wales than in England (Haines, 2010;St.Denny, 2016). In fact, the Welsh approach can be seen as part of a broader trend towards prioritising prevention over punishment and the balancing of short-term and long-term needs, as set out in the Well-Being of Future Generations (Wales) Act 2015. The emergence of a distinctive 'made in Wales' approach to administering public policy and delivering public services helps to shape the rules of the game and specifies acceptable frameworks for policy implementation. In doing so, the 'made in Wales' approach restricts the extent to which policy divisions are likely to emerge, as potential disagreements over policy implementation should be more confined.
Recent events, however, challenge the idea of congruence between the attitudes of voters and elites in Wales. Perhaps, the most striking example of this divergence in views was the decision of Welsh voters to reject membership of the European Union by 52.5% to 47.5% in 2016, despite the prominent campaign by the Welsh political establishment for Remain. The Remain campaign also had the economy on its side as Wales is widely accepted to benefit from the European Union budget, with the net annual gain estimated at around £245 million in 2014 (Wales Governance Centre, 2016). Despite the strong Remain messaging from elites, the majority of voters opted for a Leave vote, signalling a lack of trust in the judgement of political elites. The Leave vote is, of course, not the only recent challenge to the idea of harmonious politics in Wales. The 2016 National Assembly election saw the Abolish the Welsh Assembly Party win 4.4% of the regional vote despite it being publicly launched only in late 2015, and UKIP Wales gained seven Assembly Members with 13% of the regional vote. Their successes indicate a stronger desire among voters for right of centre politics and a slower pace of devolution than conventional ideas of consensus suggest. Taken together, it appears that the idea of 'business as usual' is being challenged in Wales. It does not, of course, necessarily mean that there is a fundamental disconnect between voters and elites in Wales, but there are reasons to suggest that the extent of such a disconnect might be greater than is often portrayed.

From policy positions to policy process
There is nothing new in studying voter-elite congruence. In particular, considerable attention has been given to assessing how well elites represent public opinion on a range of policy issues. Scholars do so by typically utilising one of two approaches. Some choose to focus on the comparison of voters' and elites' positions one issue at a time, even if they later aggregate the observed trends (e.g. André and Depauw, 2017;Krimmel et al., 2016;Lax and Phillips, 2012;McAllister, 1991); while others opt for comparing aggregate 'ideology scores' such as placements on the left-right scale (e.g. Belchior, 2013;Blais and Bodet, 2006;Golder and Stramski, 2010;Kim et al., 2010). This second body of literature has reached a broadly accepted understanding that there are three main issue dimensions-the left-right dimension associated with economic issues, the cultural dimension, and the European integration dimension-that capture political contestation, with congruence generally high on the economic dimension but weak on the cultural dimension. These insights are informative, but they are, nonetheless, limited to the narrow comparison of policy positions.
There is more to the policy process and understanding policy outcomes than can be gleaned from policy positions alone. Policy agenda and principles that guide parliamentary behaviour also matter (e.g. Howlett et al., 2009;Knill and Tosun, 2012;Sabatier, 1991). For example, it is conceivable that voters and elites may share similar policy positions but prioritise different issues, which could lead to contrasting preferences on the allocation of resources and a sense of dissatisfaction among the public with policy outcomes. At the same time, if there is a shared understanding among voters and elites about whose views parliamentarians should prioritise when carrying out their legislative duties, the impact that differences in their policy agenda or policy positions has on policy outcomes is mitigated. Given that policy positions tell only part of the story about how policies are enacted and whether voters are likely to feel represented at the end of the process, it is important to think of voter-elite congruence in broader terms. It is necessary to compare not only voters' and elites' policy positions but also their policy agenda and preferred approach to parliamentary representation to offer a truly multifaceted account of policy divisions between voters and elites.
Existing insights on voter-elite congruence over policy agenda and approach to parliamentary representation, however, remain limited. In terms of policy agenda, some studies have looked at the priorities of voters and elites (e.g. Bevan and Jennings, 2014;Hobolt and Klemmensen, 2008;Jennings and John, 2009;Spoon and Klüver, 2014), but their underlying aim has been to assess policy responsiveness by comparing elite issue attention at T with public opinion at T-1. The study by Reher (2014) is a notable exception as it compares voters' and candidates' policy priorities at virtually the same time, but even that focuses predominantly on evaluating how issue congruence influences turnout and not on the nature of this congruence. In terms of how parliamentarians ought to approach representation, existing studies tend to focus on the perceptions of voters (e.g. Bengtsson Wass, 2011;Bowler, 2017;Carman, 2007;Doherty, 2013) or elites (e.g. André et al., 2016;Heitshusen et al., 2004;Önnudóttir, 2014;Rush, 2001), but do not compare these. Those that do, explore the effect of congruence on satisfaction with democracy and do not focus on the nature of congruence (André and Depauw, 2017) or reveal mixed insights. While Andeweg and Thomassen (2005) and Méndez-Lago and Martínez (2002) uncover considerable differences between voters' and elites' visions of parliamentary representation, Campbell and Lovenduski (2015) and von Schoultz and Wass (2016) find that levels of congruence between voters' and elites' preferences for the focus of representation are rather strong. Existing insights on voter-elite congruence for policy agenda and approach to parliamentary representation remain limited and have raised as many questions as they provide answers.
It is important to expand the study of voter-elite congruence from the conventional focus on policy positions alone to compare simultaneously voters' and elites' views on policy agenda, policy positions, and approach to parliamentary representation. This comparison should focus on voters and elites within the same political context and at the same point in time.

Data and methods
The analyses presented here rely on compatible survey data from the 2016 Welsh Candidate Study and the 2016 Welsh Election Study. These surveys are ideally suited for comparing the views of candidates and voters in Wales as they were conducted within a couple of months of each other, include compatible questions on the different aspects of the policy process, and have an extensive reach in terms of the number of respondents. 3 The 2016 Welsh Candidate Study is a survey of candidates who stood for election to the National Assembly for Wales. 4 It was carried out immediately after the 2016 election and uses a mixed-mode design whereby candidates could return their responses online or via post. The survey was conducted in English and Welsh. The final sample includes 159 candidates (35% of all candidates) 5 and is highly representative of the general population of candidates. When using the Duncan index of dissimilarity on the distributions of two major characteristics-partisanship and candidacy type-within the sample and the full population of candidates, it yields values of 0.07 and 0.01, respectively (Duncan and Duncan, 1955; see Appeendix A). Moreover, the percentage of women among all candidates and those in the sample is similar at 34% versus 37%, as is the percentage of successful candidates at 13% versus 12%.
The 2016 Welsh Election Study is used to capture the views of voters as it includes a threewave survey of a representative sample of the Welsh electorate. 6 The analyses presented here rely on data from the pre-election wave of the voter survey as it has the greatest overlap with the 2016 Welsh Candidate Study questionnaire. This wave was carried out by YouGov in March 2016, for Cardiff University, and the final sample includes 3272 respondents.

Policy agenda
The comparison between voters and candidates on policy agenda is based on the question, 'As far as you are concerned, what is the single most important issue facing Wales at the present time?' Both sets of respondents were invited to provide open-ended answers, and these were later coded into the following broader categories: (1) economy; (2) living standards; (3) health; (4) education; (5) immigration; (6) European Union; and (7) devolution. 7 For each category, the respondent is coded 1 if she considers the policy area to top her policy agenda and 0 if not.
Following this, seven parallel difference of means tests are conducted to evaluate the comparative likelihood of voters and candidates to prioritise each of these seven policy areas. 8

Policy positions
The 2016 Welsh Candidate Study and the 2016 Welsh Election Study include questions that explore respondents' views on three issue dimensions: (1) European integration; (2) economy; and (3) authoritarian versus liberal values. To account for the possibility that voter-candidate congruence may vary across and within the different issue dimensions, their policy positions are compared separately across each available survey item. 9 Attitudes towards European integration are measured as the self-placement of respondents on a scale from 0 ('European integration has gone too far') to 10 ('European integration should be pushed further'). Economic positions are captured through the respondent's level of agreement with three statements: (1) government should redistribute income from the better-off to those who are less well-off; (2) big business benefits owners at expense of workers; and (3) ordinary working people do not get their fair share of the nation's wealth. Authoritarian versus liberal attitudes are inferred from the respondent's level of agreement with another three statements: (1) young people today do not have enough respect for traditional British values; (2) people who break law should receive stiffer sentences; and (3) schools should teach children to obey authority. Responses to the policy statements range from 1 ('strongly disagree') to 5 ('strongly agree').
The extent to which voters' and candidates' policy positions differ is addressed through seven parallel difference of means tests, each focusing on a particular policy position.

Approach to parliamentary representation
People can also have different ideas about the principles that ought to guide the behaviour of elected representatives. Some believe that elected representatives should be party agents and prioritise the interests of their party, while others believe that they should be constituency parliamentarians and prioritise the interests of their voters or that they should be entirely independent-minded instead.
The manner in which voters and candidates believe that Assembly Members should approach parliamentary representation is captured through survey questions that ask respondents about how Assembly Members should act if the views of certain stakeholders are in conflict. These conflicts are as follows: (1) party versus voters; (2) Assembly Member versus voters; and (3) Assembly Member versus party. For each choice, the respondent is coded 1 if she believes that Assembly Members should prioritise the views of the former and 0 if the views of the latter.
Following this, three separate difference of means tests are conducted to compare how likely voters and candidates are to believe that Assembly Members should: (1) prioritise the views of their party over those of their voters; (2) prioritise their own views over those of their voters; and (3) prioritise their own views over those of their party.

Policy divisions between voters and candidates
I start by comparing the percentage of voters and candidates who prioritise each of the seven policy areas. These are presented in Table 1. The differences between voters and candidates do not appear overly great, but some discrepancies do stand out. The negative score of −10.3% (22.3% vs 32.6%) for the economy indicates that voters are less likely than candidates to consider it the most important issue; whereas the positive score of 5.5% (6.9% vs 1.4%) for immigration shows that they are more likely than candidates to think that immigration is the most important policy issue. Both differences are statistically significant at p < 0.05 level, with the respective t-statistics being 2.88 and −2.59. The percentage of voters who prioritise living standards is also higher than the corresponding percentage of candidates (29.4% vs 22.2%), even if not statistically significant. In fact, living standards is the most frequently mentioned issue by voters, whereas economy topped the list for candidates. Taken together, these differences do not reveal a fundamental disconnect between voters and candidates. Instead of telling a story of contrasting visions about the most important issues facing Wales, they tell a story of limited discrepancies. Candidates are modestly more inclined to prioritise structural, 'big picture' issues, while voters tend to attach greatest importance to issues that have more immediate personal implications. 10 Considering the varying attitudes of voters and candidates toward the economy and immigration, we need to understand whether these differences are present across the ideological spectrum or limited to candidates and voters of a certain ideological persuasion. Table 2 shows the percentage of voters and candidates, by party, who consider the economy and immigration to be the most important issues. 11 Note first that differences in voters' and candidates' likelihood of prioritising the economy are most salient within the more leftist parties. The percentage of Plaid Cymru candidates who consider the economy to be the most important issue is more than double the corresponding percentage for Plaid Cymru voters (52.6% vs 24.6%). It is rather telling that economic issues are mentioned by Plaid Cymru candidates more often than everything else combined, whereas economic issues are mentioned by less than a quarter of its voters. This disconnect is weaker, but still salient, within Welsh Labour (43.8% vs 22.3%) and the Welsh Liberal Democrats (47.4% vs 27%). Interestingly, however, this pattern is not present when looking at the Welsh Conservative Party and UKIP Wales. Candidates of both these parties are in fact less likely than their voters to prioritise the economy, but the differences here are small. The disconnect between voters and candidates over whether economic issues should take priority is driven by differences within the more leftist parties.
A closer look at the salience given to immigration reveals an even starker discrepancy. While immigration is not considered to be the most important issue by any candidate of the five parties represented in the National Assembly for Wales, following the 2016 Assembly election, it is by some voters of all these parties. The percentage of voters who prioritise immigration as the most important issue is highest among those who support UKIP Wales at 27.3% and notable among those who support the Welsh Conservative Party at 7.2%, while dropping below 5% for the more leftist parties. There is variation in how likely voters of different parties are to prioritise immigration, but the salience of the issue is still evident right across the ideological spectrum. Voters of all main parties in Wales are more inclined to believe that immigration is the most important issue than the corresponding candidates.
Next, I compare voters' and candidates' policy positions. Table 3 presents their responses to seven policy statements which, taken together, capture attitudes on European integration, the economic left-right dimension, and the cultural dimension. Perhaps somewhat surprisingly, I do not find meaningful differences between voters' and candidates' attitudes towards European integration. The difference of 0.4 (4.0 vs 3.6) is small, given the scale of the measure, and does not reach conventional levels of statistical significance. In fact, both an average voter and an average candidate feature a strong dose of Euroscepticism as they are in the 'integration has gone too far' half of the spectrum. The former is of course unsurprising as Wales voted for the United Kingdom to leave the European Union by 52.5% to 47.5% in 2016; but the latter is somewhat unexpected since most politicians were thought to have campaigned for a Remain vote in the lead up to the referendum. What these data suggest is that the Remain stance was a grudging acceptance of European Union membership rather than an expression of one's Europhile nature for more Welsh politicians than perhaps is often perceived. Moving on to voters' and candidates' economic positions, two of the three policy statements reveal statistically significant differences. On average, candidates are more inclined than voters to favour governmental policies that redistribute income (3.9 vs 3.5) and less likely to believe that big business benefits at the expense of workers (3.7 vs 4.0). Both differences are statistically significant at p < 0.01 level. However, they are of relatively minor scope and do not reveal contrasting economic visions. Voters and candidates are both, on balance, positive about income redistribution and critical about big business. The differences in voters' and candidates' views on economic issues are not substantively meaningful.
Salient and consistent disconnect in voters' and candidates' policy positions is, however, very much evident when focusing on issues on the cultural dimension. Voters tend to hold notably more authoritarian attitudes than candidates. They are more likely than candidates to agree that young people have no respect for British values (3.6 vs 2.9), that breaking the law should lead to stiffer sentences (3.8 vs 3.0), and that schools should teach children to obey authority (3.8 vs 2.9). These differences are statistically significant (p < 0.01) and large in scope, but they also reveal contrasting beliefs. Whereas the average scores for candidates are consistently in the middle of the scale, revealing a neutral collective stance in response to these statements, an average voter tends to agree with all three statements. This finding suggests that voters in Wales hold considerably stronger authoritarian views than candidates. 12 Finally, I compare how voters and candidates believe that Assembly Members should behave when carrying out their duties in the National Assembly for Wales (Table 4). Note first that they are almost equally likely to suggest that Assembly Members should prioritise their own views over those of their party in case of a conflict (62.8% vs 59.2%). However, when asked about whether Assembly Members should discard the preferences of their voters in favour of their party position or own views, voters and candidates have different visions of how representation ought to be carried out. Voters are considerably more inclined than candidates to suggest that Assembly Members should prioritise their voters' views over their party position in case of a conflict (83.7% vs 59%). Voters are also more likely than candidates to suggest that Assembly Members should prioritise their voters' views over their own opinion if in conflict (75.3% vs 45%). These differences are statistically significant at p < 0.01 level and reveal a telling mismatch between voters' and candidates' visions of representation. Whereas a clear majority of voters believe that Assembly Members should prioritise voters' views over their party line or own views, candidates are almost evenly split in their belief as to whether voters' views should dictate parliamentary behaviour or not. 13 The greater tendency of candidates to suggest that discarding voters' views, in favour of party line or Assembly Members' own views, is acceptable is also particularly relevant in the current political climate. It re-enforces the perception that there is a substantial disconnect between voters and elites, with the views of the former being readily cast aside.
Overall, it is apparent that there are salient policy divisions between voters and candidates in Wales. These differences are by no means big enough to suggest that the political system is fundamentally unrepresentative, but they do exist and should be addressed. This need to address divisions is especially the case in the current political environment where the levels of public trust in politicians and politics more broadly are uncomfortably low.

Conclusion
Elites in the United Kingdom, and beyond, are increasingly seen as being out of touch with voters, with the growing disconnect between them seen as a salient feature of the current political climate. It has been argued that public disillusionment with politics has not just led to growing levels of electoral volatility (Denver et al., 2012) but has also helped to fuel the successes of populist parties who have made it a key element of their political identity (Ford and Goodwin, 2014). In few places has this pattern played out more clearly in recent years than in Wales where voters rejected membership of the European Union, against the almost unanimous advice of the political establishment, and delivered notable election success for UKIP Wales at the 2016 Assembly election.
In this study, the current extent of voter-candidate congruence in Wales is explored. Using data from the 2016 Welsh Candidate Study and the 2016 Welsh Election Study, it compares voters' and candidates' views on policy agenda, policy positions on a range of issues, and whose interests they believe Assembly Members should prioritise when carrying out their duties in the National Assembly for Wales. The analysis reveals some salient differences between the attitudes of voters and candidates. Voters are notably more likely than candidates to think about immigration as the most important policy priority, whereas they are less likely than candidates to think that about the economy. In addition, voters tend to hold considerably more authoritarian attitudes than candidates. The most fundamental difference between the two, however, is in their views on whose interests Assembly Members should prioritise. Voters find it significantly less acceptable than do candidates for Assembly Members to discard their voters' views in favour of their own views or those of their party. There are some salient policy divisions between voters and candidates in Wales.
There are three broader points arising from this study. First, these findings support a growing body of evidence that the political environment in Wales is not particularly consensual after all and that the relationship between voters and elites is characterised by considerable disconnect. Welsh voters are not just willing to go against the advice of the political establishment, as was shown by the European Union referendum in 2016, but they also hold somewhat different attitudes than candidates. They are more likely to think of immigration as the most important policy priority, hold considerably more authoritarian policy positions, and believe in a more voter-centred style of representation. Rather than telling a story of shared attitudes, the dynamic of voter-candidate congruence in Wales emphasises a degree of disconnect.
Second, the presence of significant policy divisions between voters and candidates in Wales suggests that elites need to intensify their efforts to engage with the public. It has been shown that higher levels of voter-elite congruence are linked to greater satisfaction with democracy (e.g. André and Depauw, 2017;Brandenburg and Johns, 2014;Mayne and Hakhverdian, 2017) as well as electoral turnout (Heath, 2016), both of which are seen to contribute to the health of a democracy. Not only do they encourage a strong civic culture and compliance with public policies, but they also empower and legitimise political institutions. In contrast, disconnect between voters and elites implies that counting on permissive consensus is insufficient. It is evident that elites need to engage in greater dialogue with voters over the different aspects of the policy-making process to counter the perception of being out of touch. Third, this study contributes to the broader debates on how the idea of elite-voter congruence should be thought of conceptually and studied empirically. The presence of salient divisions between voters and candidates with regard to all three aspects of the policy-making process analysed here suggests that the conventional reliance on policy positions alone to capture congruence is insufficient. The level of disconnect between voters and candidates can, and does, vary across the different aspects of the policymaking process and this possibility needs to be accounted for. It is vital to take a more nuanced approach to evaluating the degree of voter-elite congruence and compare not simply policy positions but also policy priorities and preferred approaches to parliamentary representation. Ideally, this kind of analysis should be done by focusing on voters and elites in the same political context and at the same point in time.

Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes
1. The return of two-party politics was characteristic of the 2017 general election, in general, as the Labour Party and the Conservative Party won a combined 82.4% of the popular vote across the United Kingdom. 2. Notable exceptions were the UK Independence Party (UKIP) Wales, which was the only major party in Wales to campaign for a Leave vote, and Andrew R.T. Davies, leader of the Welsh Conservatives in the National Assembly for Wales, who campaigned for a Leave vote in contrast to most of his colleagues. 3. The exact number of candidates and voters included in the analyses of their comparative policy agenda, policy positions, and approach to parliamentary representation varies slightly, as only those who answered the relevant survey question are included in the corresponding analysis. These subsamples, however, remain representative of candidates and voters in general. 4. Focusing on candidates' views is a widespread practice in representation and congruence literature to capture elite attitudes (e.g. Belchior, 2013;Costello et al., 2012;Leimgruber et al., 2010;Reher, 2015).
Other data have of course been used as well, as some scholars have instead opted for party or government positions (e.g. Golder and Stramski, 2010;Hakhverdian, 2010;Huber and Powell, 1994;Powell, 2009) and others for the positions of elected parliamentarians (e.g. Hanretty et al., 2017;Karyotis et al., 2014;Vasilopoulou and Gattermann, 2013). Reliance on candidates' views, however, has some useful advantages. Whereas focusing on party or government positions, by treating these as unitary actors, does not account for what elites actually think on individual level, and the reliance on parliamentarians' positions fails to account for the broader political class and full range of political cues that voters receive; candidate survey data capture attitudes that are exogenous to parliamentary and party institutions, like whipping and log rolls, as well as the views of a wider range of political actors. The trade-off here is that the focus on candidates, as opposed to governments or elected representatives, restricts the analysis to the comparison of attitudes and views without capturing their effect on policy outcomes. 5. In total, 457 candidates stood for election to the National Assembly for Wales. This includes 153 constituency candidates, 208 regional list candidates, and 96 dual candidates. 6. For further information on the 2016 Welsh Election Study, see its project page on the Research Councils UK website (Research Councils UK, 2017). 7. Responses that did not fall under these categories were omitted from the analysis. They constituted less than 10% of the total responses.
8. There is no formal baseline standard established in the representation literature to classify the differences in policy agenda, policy positions, and approach to parliamentary representation as small, large, and so on. As such, the analysis presents not only the extent of the differences but also the underlying average scores for voters and candidates so the substantive meaning of the observed differences is easier to understand. 9. It is important to compare voters' and candidates' policy positions separately across the different survey items as a growing body of literature suggests that voter-elite congruence can vary according to the issue at stake (e.g. Costello et al., 2012;Freire and Belchior, 2013;Walczak and van der Brug, 2012). 10. Although there is little evidence from other Western democracies regarding voter-elite congruence on policy agenda at a specific time as existing studies tend to compare elites' issue attention at T with public opinion at T-1, there is some indication that the differences observed here are not unique to Wales. Lindeboom (2012) shows that Dutch voters were more likely than elites to prioritise issues surrounding immigration between 1981and 2006, and Reher (2014 finds that German voters were more likely than candidates to consider immigration and labour issues as the most important ones in 2009 but less likely to think that of economy. 11. Table 2 is limited to parties that won seats in the National Assembly for Wales at the 2016 devolved election. Information on candidates and voters of other parties is available upon request. 12. These patterns of congruence in policy positions are common in Western democracies. It has been frequently shown that voter-elite congruence in Europe is high in terms of the left-right dimension and specific economic policies, but elites tend to hold more liberal views than voters (e.g. Belchior and Freire, 2012;Costello et al., 2012;Dolný and Baboš, 2015;Vasilopoulou and Gattermann, 2013). 13. As highlighted earlier, existing evidence from other Western democracies is mixed. While some studies find considerable differences in voters' and elites' visions of parliamentary representation (Andeweg and Thomassen, 2005;Méndez-Lago and Martínez, 2002), others find high levels of congruence between the two (Campbell and Lovenduski, 2015;Von Schoultz Wass, 2016). 14. Information on Assembly Members' parliamentary contributions is obtained from the website of the National Assembly for Wales (National Assembly for Wales, 2017a, 2017b).
while the comparison of candidacy type has a value of 0.01. The survey sample is highly representative of the general population of candidates on these two key characteristics.

Successful candidates' views and parliamentary activity
This appendix provides a supplementary analysis of whether candidates' views are evident in their parliamentary behaviour. Although most candidates do not get elected, the 2016 Welsh Candidate Study includes 15 candidates who did and whose responses can be compared with their parliamentary behaviour in the first year of the 2016-2021 Assembly term. This analysis remains exploratory, given the small-n, but offers a useful preliminary indication as to whether a connection between the views and parliamentary activity of successful candidates exists. I start by looking at the extent to which successful candidates' contributions in the National Assembly for Wales focus on policy areas they considered most important in the 2016 Welsh Candidate Study. 14 First, the comparison of their committee membership and policy priorities shows that nearly two-thirds of these candidates (64.3%) became members of a committee with a remit that covers the policy area they considered the most important in the first year of the 2016-2021 Assembly term. This rises to 92.9% when considering membership in committees with remits that cover policy areas of either primary, secondary, or tertiary importance. Second, successful candidates are asking questions related to the policy areas they consider important with regularity. On average, 33.2% of the oral questions they tabled in the same 12-month period related to the policy area they considered the most important. This rises to 67.4% for questions related to any of the three most important policy areas. This evidence does seem to suggest that the focus of successful candidates' parliamentary behaviour, in terms of policy agenda, is consistent with the views they voiced in the 2016 Welsh Candidate Study.
Moving on, I look at whether those successful candidates who attached greater importance to following party positions in the 2016 Welsh Candidate Study were, in fact, less likely to defect from the party line during the first year of the 2016-2021 Assembly term. There are 14 successful candidates who revealed their views on whether an Assembly Member should vote according to her party position if her voters have a different opinion or if she has a different opinion herself; seven saying 'no' in both instances and seven saying 'yes' in at least one of these instances. When comparing the roll call voting records of these Assembly Members, it is evident that the former not only hold less partisan views but also act in a less partisan manner in the National Assembly for Wales. Although the level of party loyalty-measured as the percentage of roll call voting occasions when Assembly Members cast the same vote as the plurality of their fellow partisans-is high among both groups, as one would expect given the strong whipping practice in the Assembly, it is higher among Assembly Members who believe in a more partisan approach to parliamentary representation (99.7% vs. 98.8%). The difference is small in absolute terms, but Assembly Members who hold a less partisan view of representation are still four times (1.2% vs. 0.3%) more likely to defect from their party line than Assembly Members who hold a more partisan view. There does seem to be a link between successful candidates' views on how representation ought to be carried out and their subsequent parliamentary behaviour.
Although no causal claims can be drawn from this, the initial insights suggest that there is a connection between successful candidates' views and parliamentary activity. Their behaviour as Assembly Members, in terms of what is analysed here, appears largely compatible with the views they voiced in the 2016 Welsh Candidate Study.