#Personal vs #Party: A comparative study of candidates’ new media campaigning in Japan and the United Kingdom

This paper examines how new media is affecting candidate personal vote seeking behaviour, at the constituency level by comparing data from general elections in two countries with different styles of campaigning – party-centred campaigning in the United Kingdom (2015 & 2017) and candidate-centred campaigning Japan (2014 & 2017). By utilising both content analysis of candidates’ use of new media platforms and in-person interviews with candidates, this study gives a deep description of how individual candidates are using online campaigning and to what degree they are pursuing a personal, rather than party vote. This study confirms that that Japanese candidates use new media to run more candidate centred campaigns, replicating traditional campaign styles, but also finds that other factors, namely candidates’ levels of experience and the strength of the national party, play a role in how candidates utilise new media. Now more so than ever, candidates are relying on personal image promotion as a major element of their campaigns.


Introduction
With an increasing level of internet diffusion and the adoption of online campaigning by political actors, the advent of new media on campaigning raises questions about what effect it may have on candidate behaviour. One question which arises out of the ability for political candidates to take control of their own campaign narrative is 'Does new media encourage constituency level candidates to pursue a more personalised vote?' This paper aims to explain what factors encourage candidates to pursue personal vote seeking strategies online by comparing candidates in two countries with very different campaign traditions -the United Kingdom and Japan. By comparing a country with a party-centred campaigning style, the UK, with a country with a traditionally more candidate-centred approach, Japan, this paper will show that that campaign traditions, seen in country level differences, continue to be the most dominant factor, although other well-established factors, such as incumbency and level of candidate experience, continue to influence the degree to which candidates pursue the personal vote. In general, new media is not transforming the way in which candidates campaign at the constituency level, but we are seeing examples of highly personal vote seeking behaviour even in systems where the focus of campaigning has traditionally been party-centred. To this extent, we can infer that the new social media are indeed political tools that facilitate personalised campaigning and vote-seeking. *Corresponding author: Sean Vincent, Meiji Gakuin University, Yokohama, Japan, E-mail: tsve001@gmail.com

The personal vote
In examining the personal vote, this paper adheres to the definition given by Cain, Ferejohn and Fiorina (1984, pg. 111): "that portion of a candidate's electoral support which originates in his or her personal qualities, qualifications, activities and record". For candidates, there are a number of reasons why pursuing a personal vote may pay electoral dividends. Firstly, the electoral system used will affect the degree to which candidates are able to and required to pursue the personal vote. Carey and Shugart's (1995) attempt to provide a model based on electoral systems highlights that candidates with the greatest incentive to pursue a personal vote will be those contesting constituencies where votes are non-transferable (Single Member District seats) and have a high district magnitude (for example in Japan, before electoral reform in 1993).
Voters, particularly in rural areas, expect their local representative to be in touch with the local community, making appearances at events and showing support for local community initiatives. Less than being about policy, representatives and electoral candidates must be visible to the public (Miller & Stokes, 1966;Fenno, 1978;Evans, 2011). Representatives must also be seen to be working on behalf of their constituents, actively engaged in casework and representing their concerns in the national parliament (Norton & Wood, 1993).
National party performance also affects the extent to which candidates are willing to pursue the personal vote. Candidates trying to avoid being dragged down by an unpopular national party brand have an incentive to campaign on their own personal strengths and this is apparent even in countries where the party brand is traditionally seen as the key determining factor behind voter choice (Burnham, 1975;Carey & Shugart, 1995;Hellweg, 2011).
Finally, the growing trend towards personalisation of political candidates at both the national and constituency level has created a greater need for candidates not just to be more visible but to be more relatable and accessible. There is evidence, albeit limited, that personalisation is becoming more common amongst constituency level candidates both in countries where there is a tradition of candidate centred-voting such as Japan and the US and in party-centred systems like the UK and Australia (Reed, 1994;Herrera & Yawn, 1999;McAllister, 2015).

New media and the personal vote
Recent studies of how new media affect the personal vote focus on how the internet allows candidates to create/expand a personalised image and how voters are now coming to expect more relatable and accessible politicians (Comer, 2003;Gulati, 2004;Stanyer, 2008). Livak et al (2011: p.459) conclude that websites allow candidates to express their own opinions, which may even run contrary to the party line, and to fund local campaign activities: "It seems that adopting personal websites…is another step in the in the disassociation of the individual politician from his party".
In perhaps the most relevant study on constituency level campaigning so far Zittel and Gschwend (2008) find that the majority of candidates in Single Member District (SMD) seats in Germany still campaigned on a party platform in the 2005 election. While there was evidence of candidates individualising campaigns by producing their own campaign material and highlighting issues not raised by the central party, this was found to be much more prevalent in marginal seats. New technology points to a potential change in the party-centred status quo: "A candidate actively seeks personal votes on the basis of candidate-centred local campaign organizations, candidate centred campaign agendas and candidate centred means of online campaigning" (2008, p. 299).
Other studies have focused on how candidates' Internet use is creating a greater degree of personalisation. Hermans and Veerger's (2013) cross-national study of the personalisation of candidate websites in the 2009 European election assesses how candidates use websites to appeal to voters -whether they keep their content professional or use personal information, such as details about their family or personal likes/dislikes to create a bond with the public. In two separate studies Thamm and Bleier (2013) and Kaczmirek et al (2014) examine the use of Twitter by legislators to both promote themselves and personally engage with followers. The theme of personalisation is taken further by Margaretten and Gaber (2014), and Goldbeck et al (2010), looking at Scottish and US legislators respectively, researching legislators' Twitter use and finding that it is not just used as a platform for professional communication but also offers followers insights into the legislator's personal life.
Facebook's popularity, with over 850 million visitors every day, the ability to upload photos and video and ability to create online networks, also makes it an attractive campaign tool. In a study of candidate Facebook pages in the 2006 and 2008 US congressional elections Williams and Gulati (2013) found that those candidates using Facebook extensively were more likely to be engaged in marginal races. Miyawaki's (2014) study of the 2013 Upper House elections in Japan went further, by measuring post frequency and how engaged candidates became in their replies to comments posted on their pages. In Germany Kaczmirek (2014) found an increasing level of interaction between politicians and members of the public posting on their Facebook pages. What these examples demonstrate is that new media acts as an agent for greater interaction between the candidate and voter, allowing users to see politicians in a more authentic light. This interaction is increasingly direct and personal which implies a much different relationship between candidates and voters than seen through traditional, top-down campaigning.
By combining aspects of both personalisation and organisation, and engaging in a cross-country analysis based upon different campaign styles this paper provides an original insight into understanding how technological changes are influencing both the conduct of campaigns and candidates' pursuit of the personal vote. Moreover, this study combines candidate use of both websites and the two most popular social media platforms, Twitter and Facebook. By taking data from multiple online sources, this study will be able to draw more detailed conclusions about candidates' overall new media campaign strategies than previous studies which have only examined a single platform.

Research design and methodology
It can be expected that systemic differences in campaign traditions are a key determinant of the degree to which candidates use new media to campaign independently. If this is the case, then traditional candidate campaign behaviour can be said to have transferred over to new media. In addition, studies into the personal vote have established a number of factors which affect both the desire and the capability of candidates to actively court the personal vote. As such these factors will also be taken into account in the following hypotheses. H1: Candidates in a candidate centred campaign system (Japan) will use new media to pursue a personal vote to a greater degree than those in a party centred campaign system (the UK).
Candidates from the UK and Japan have been chosen as subjects for this study because of their relatively similar electoral systems which in practice produce very different campaign styles. Both countries utilise first past the post elections in SMD seats. According to Carey and Shugart (1995) candidates contesting SMD seats are amongst the most likely to pursue a personal vote. However, Japan and the UK have campaign styles which have been described as candidatecentred and party-centred respectively. In Japan, high levels of personal voting are a consequence of the SNTV voting system used in the post-war period up until 1993, which often saw candidates from the same party running against each other in multi-member constituencies (Krauss & Pekkanen, 2011). Under SNTV candidates built up strong local organisations, koenkai, which operated on behalf of the candidate rather than the local party branches (Ramsayer & Rosenbluth, 1997). Electoral success relied upon candidates having a high degree of visibility. Successful independent legislatures have been actively recruited by political parties or even welcomed back with open arms after leaving the party (Reed, 1994). While the elections of 2005 and 2009 showed how influential party leadership and brand value could be (Maeda, 2007;Jou & Masahisa, 2015), Japanese campaigning at the constituency level still revolves around the candidate more so than the party.
There are several reasons why UK election campaigning has been described as party-centred. With political parties able to reach mass audiences through television starting from the 1960's party leader's personal image and perceived capabilities, became the focus of both party and media attention (Heffernan & Webb, 2005). Both the running of campaigns, which became more professionalised and thus required the expertise and financial muscle which only the national party could provide, and candidate selection became increasingly centralised (Farrell & Webb, 2002;Low, 2014). The major UK political parties have their foundations linked to class cleavage representation. While these cleavages have weakened over the years and parties have become more ideologically similar, parties in the UK have traditionally stood for something concrete and voters have found it much easier to identify with a party and nationally promoted policies. While they don't have the same ideological basis, parties are often seen by both themselves and by the public as brands (Scammell, 2007). As Stanyer (2008 p. 421) writes, "at a general election, voters still very much vote for the MP because he or she is the representative of a party".
Overall, then, while both Japan and the UK produce exceptions to their general types (Maeda, 2007;Pempel, 2008;Norton & Wood, 1990;Johnston, et al., 2012), classifying their respective styles of campaigning as candidate-centred and party-centred is broadly accurate.
H1 represents a country level hypothesis. It can be expected that the organisation and policy promotion potential of new media will particularly appeal to politicians in a candidate-centred system. This would be evident in Japan as an extension of existing campaign methods which rely on candidates being able to raise considerable funds through supporter groups, mobilise volunteers and voters through personal networks and promotion of local/personal policies. To demonstrate a systemic difference, candidates in the UK would be expected to have less organisational capacity evident in their new media platforms and be less likely to promote personal/local policies and their own personal brand. Evidence of high levels of personal vote seeking behaviour regardless of systemic difference would imply potentially significant changes in political campaigning and candidate-party relations. These changes would be so general as to overcome country or party-specific traditions.
H2: Incentive for candidates to cultivate a personal vote will relate to national party popularity. The lower the party popularity, the more incentive for candidates to use new media to pursue the personal vote. This hypothesis represents party-level analysis. Gibson (2013) has showed that nationally competitive parties in the UK, e.g. the Liberal Democrats, which have less resources in comparison with their competition, are more likely to encourage "citizen initiated campaigning", effectively decentralising some aspects of campaign organisation in the Obama/hybrid mould. It seems logical that this would also apply to constituency candidates. In an era of declining party membership, candidate self-sufficiency may be born out of necessity, even in countries where party-centred/ controlled campaigning is the norm. This also applies to party popularity. Candidates representing parties which are unpopular on a national level may make a rational choice to differentiate their policy stances or attempt to build a more personalised identity in an attempt to counter a weak national party image (Burnham, 1975). Unpopular national parties can have a direct negative effect on the vote of even relatively safe incumbents. (Fisher, et al., 2016). Smaller national parties may also lack the resources to support their candidates in all constituencies, with candidates left to fend for themselves both in the lead-up to and during a campaign (Primo & Snyder, 2010). For the purposes of this study parties have been combined from both countries into three groups: "Ruling party" (Conservative/Liberal Democratic), "Main opposition party" (Labour/Democratic) and "Third party" (Liberal Democrat/Japanese Restoration).
H3a: Candidates fighting in marginal seats will have more incentive to use new media to pursue the personal vote. H3b: Incumbent candidates will have more incentive to use new media to pursue the personal vote.
These hypotheses represent constituency level analysis. Candidates fighting marginal seats are more likely to promote localized policy platforms and be more pragmatic, appealing directly to the voters who will get them elected, rather than be ideologically dogmatic along party lines (Zittel, 2015). In addition, marginal seats are also likely to feature higher levels of campaign intensity, with marginal constituencies seeing a greater degree of volunteer activism and evidence of higher levels of campaign spending (Fisher & Denver, 2009). While the resources for higher levels of campaign intensity may come from the national party, or private donors, greater marginal campaigns can "boost local campaign efforts and promotion of the (constituency) candidate" (Cutts et al, 2012 pg. 361). If marginality is not shown to be a significant predictor of personal vote seeking behaviour, then we may conclude that reliance on party image remains the most significant influence on candidate behaviour. Incumbent candidates, with an already established local name and infrastructure are also expected to have a stronger personal vote already in place, which they will actively pursue (Miller & Stokes, 1966;Curtice & Steed, 1980;McAllister, 2015). The performance of the national party does have a significant effect on candidate performance at the constituency level (Fisher & Denver, 2009). It is intuitive that where an incumbent can remove the potential effect of party performance and strengthen their own position not just for the present election but also for subsequent contests, that opportunity will be taken. Candidates contesting marginal seats and incumbents have either more of an incentive or more of an opportunity to engage in more individualistic behaviour, such as producing their own campaign material or seeking localized forms of additional funding (Pekkanen et al, 2006;Zittel & Gschwend, 2008;Adams, et al., 2010). H4a: Older candidates will be more likely to use new media to pursue the personal vote. H4b: Candidates who have already been elected multiple times before will be more likely to use new media to pursue the personal vote.
These hypotheses represent indivdual level analysis. Existing studies have shown that new media use is expected to be greater amongst younger candidates who are more in tune with new media and open to using less traditional campaign methods (Hooghe & Vissers, 2009, Williams & Gulati, 2013, Miyawaki, 2014. However, younger candidates are also more likely to be beholden to party leadership, have less in the way of a committed local support base and therefore are likely to have less of a personal vote to draw on, whereas older and more experienced candidates are more likely to use their established position within their constituency to rely on a personal vote. Candidates who have served multiple terms in office are more entrenched within their local community and should already have a strong personal vote to draw on (Burnham, 1975). Multiple time incumbents will most likely have greater financial resources and/or a network of local activists which have been built-up and battle hardened over multiple election campaigns (Fisher & Denver, 2009). As with H3a and H3b, there is a rational choice assumption; where there is an opportunity for greater independence it will be taken by the candidate.

Case selection
In total 343 candidates from across two lower house elections in Japan (2014 and 2017)  6 Data and measurement

Candidate websites
Recent studies have found that websites are becoming more geared towards encouraging activism, establishing online networks for voters to both discuss issues and organise offline support for election campaigns, rather than just passively consume information content (Strandberg, 2008;Vaccari, 2009;Cardenal, 2013;Greffert, 2013).
A measure of candidate independence has been created by awarding points relating to the functions listed below. Both fundraising and volunteer sign-up (organisational functions) have been given primacy over personalised functions promoting candidate image or personalised policies. With a trend of decline for parties both in financial support and human/volunteer resources, the ability for candidates to personally attract both donations and volunteers is becoming increasingly important (Carty, 2004). Online campaigning has not yet replaced traditional campaigning as candidates' primary focus on campaigns. Money and volunteers remain the most important resources in a constituency candidates' arsenal (Fisher & Denver, 2009;Cutts et al 2012). Candidates less reliant on the national party have a much greater opportunity to run independent, candidate focused campaigns (Zittel & Gschwend, 2008). In the UK for example, financing at the constituency level consistently fails to meet the needs of local parties, with this being especially true in untargeted seats at election time (Pattie & Johnston, 2016). The same is true in Japan, where the national party is not always viewed as being capable of organising effectively at the local level (Krauss & Pekkanen, 2011). These features of candidate websites have been coded as follows: The above criteria are used to rank candidate websites on a 7-point additive scale, with 7 being the maximum and showing candidates using websites to pursue the personal vote at a high level. This is referred to hereafter as the Personalised Website Index (PWI).

Social media
While the way in which social media allows candidates to interact with members of the public and present a more personalised image is an important part of pursuing the personal vote, it is also of interest to see what policies candidates are promoting, especially those which are constituency-specific, as these can be regarded as signs of autonomy from the central party. In order to operationalize the concept of candidate independence through, content analysis will focus on what ways candidates are pursuing the personal vote i.e. a vote for themselves rather than the party they represent.
For the purposes of this study candidate tweets/retweets and Facebook posts/shares, all signifying a conscious endorsement by the candidate, were categorised into six areas: The number of these posts from both Twitter and Facebook falling into the category of local policy promotion, personal and campaign activity are combined and the percentage of all posts made up of these categories was used to determine each candidate's overall personal vote content from social media (SMPV score). There are two key reasons as to why campaign activity posts have been classified as related to personal vote. Firstly, posts which are focused chiefly on the candidate and their activities can be seen as a form of microblogging (Orkibi 2015). As these posts are often being made in real time and show the candidate in a less managed environment, in comparison to prearranged event appearances and campaign literature, their intention is to project an air of authenticity. Secondly, these posts help to document a candidate's campaign. They can be designed to advertise upcoming campaign apperances as well as create a record of where the candidate has been active during a campaign. An important aspect of cultivating the personal vote is being visible to one's constituents (Miller & Stokes, 1966, Fenno, 1978. In trying to project an active and dynamic campaign on the ground, candidates use campaign activity posts to show voters both a connection with the local area they wish to represent and potray a campaign led by an hardworking and dynamic candidate.
In addition to the above data, a number of interviews were conducted with candidates from both the UK and Japan. Material from these interviews provides context that helps when interpreting results form the content analysis; this better justifies conclusions that are drawn.
After detailing the descriptive statistics pertaining to candidates' use of websites and social media, the paper will continue by combining data from both website and social media use to create a Personal Vote Index (PVI) score and apply multiple classification analysis in order to statistically test the candidates' overall new media strategy. Each candidate is given a PVI based upon the addition of standardised scores derived from their Website Index and SMPV scores (PWI Z score + SMPV Z score = Personal Vote Index (PVI) score.  Table 1 shows the averages of PWI scores based on the variables under study. Based on a 7-point scale, the mean scores show little difference in the extent to which candidates from the UK (3.42) and Japan (3.49) are using websites to strengthen their own campaign organisation or their personal image. Table 1 also shows candidates from weaker parties are more likely to use personal vote features and there is a clear distinction between the three party groups studied, although not at a statistically significant level. PWI averages were highest amongst "third party" candidates (4.17), with a steady decline in average scores of opposition parties (3.50) and the ruling parties (3.13) providing some evidence to support H2, which is to say that candidates from weaker parties are using websites to in some way compensate for their national party's shortcomings. Amongst the two constituency level variables there is a differing degree of support for H3a and H3b. There is a slight difference in the utilisation of websites between candidates based on incumbency, with incumbent candidates using more personal vote features. A more substantial difference can be seen between marginal and non-marginal candidates. At the individual level, in the case of both age and experience there is no clear evidence to show that older and more experienced candidates use more personal vote features. In terms of website use the data points to what might be described as "mid-career" politicians using more personal vote features. It is the second to oldest group (aged 53-58) which has the highest PWI score (3.93) and candidates with between 3 and 4 previous terms served (4.00). While this does little to prove the validity of H4a and H4b, it does show that the opposite is certainly not true, namely that younger and less experienced candidates are using more personal vote features. A more detailed look at the different ways in which candidates use websites provides a more useful assessment of website utilisation.

Candidate Websites
There is a qualitative difference in the way in which candidates from Japan and the UK use personal vote features on their websites. Figure 1 shows that the use of candidate biographies was standard amongst all candidates. It should be noted that Japanese candidate biographies were very often much more detailed, for example describing the candidate's entire life, from their schooldays, rather than just their professional histories. All Japanese candidate websites featured a unique design as opposed to the vast majority of UK candidate websites, which were built from standard templates shared by members of the same party. Japanese candidate websites generally featured much more extra information, such as monthly newsletters and campaign posters available to download and videos of candidate activities either in Parliament or in local politics. This accounts for their much higher scoring in the other personal vote feature category. The other area in which Japanese candidates had a higher frequency was in donations. 47.7% of Japanese candidates had a donate function compared with 39.10% of UK candidates' websites1. The volunteer function was used to a much higher degree by UK candidates (48.1%), especially those representing Labour and the Liberal Democrats. The same also applies to policy. 50% of UK candidates had a local manifesto or specifically mentioned local policies, a much higher proportion than in Japan (21.2%), where mentions of policy overall were much lower (see Figures 2 and 3).
"I have had a website for about ten years and the current design for the last two years. The design and maintenance is all down to me, I have no input from the party…The most important function of a website, for me, is that it helps to build my profile and gives people an access point to contact me" -DPJ Candidate and former MP Volunteers play a much more important role in UK elections and are a vital resource for any candidate wishing to make contact with potential voters. It should be noted there have been instances of candidates attracting volunteers and donations through websites but this is still at an early stage and has yet to play an important part in campaign organisation: "My experience of fundraising is that it had to be me going and making a personal contact….In terms of volunteering, we would occasionally get people coming through (the website), but physically going out and saying can you help us, that's how it works" -Liberal Democrat candidate In a content analysis of candidate websites it becomes apparent that at the country level there is a stark difference in the way candidates use websites. The focus of Japanese websites is almost always on the candidate personally with any party symbolism, where it is found, coming from images of the candidate with the party leader. In the UK, some organisational functions (i.e. volunteer) and promotion of individualised manifestos are being used to a greater degree although this is balanced out by the strong identification that each website has with the national party represented. In regard to the personalisation of websites there is consistent support for H1, with organisational features determined by the respective importance placed on financing by Japanese candidates and activist mobilisation by UK candidates (see Fig. 1).

Social media
The SMPV percentages from Table 1 show a considerable difference between Japanese and UK candidates in their use of social media. Total output from the respective election campaign periods shows that Japanese candidates (68.03%) used output which could be categorised as seeking a personal vote to a much higher degree than UK candidates (46.97%) At face value this is a clear statistical difference which firmly supports the hypothesis set out in H1. As with candidate websites it is important to examine the qualitative differences between how websites are used by candidates in both countries. Figure 2 shows the breakdown of social media use by candidates from both countries by content. For candidates from both countries social media output classed as campaign activity makes up the biggest single category of social media activity. In the case of Japanese candidates (63.42%), well over half of their posts related to campaign activity, significantly higher than those of UK candidates (37.70%). Campaign activity posts were utilised in a number of different ways. Most commonly, candidates reported what they had been doing on the campaign trail and which places they had visited in their constituency.
The quote below, from the campaign manager of a Conservative MP in a marginal seat helps explain the benefit of using social media to help candidates build a local profile, something which would be more difficult for challenger candidates to do through traditional media.
"Without constraints (from traditional media) on how much is published, we were able to convey how busy (the candidate) was without having to say it: she had already proved it….we didn't need to say (the candidate) was working hard for 'X' ward or 'Y' community because there was proof" This quote also brings to the fore the sense of candidates building a "narrative". Through social media, a whole campaign can be documented online. People can see candidates working hard in their constituency, which builds and promotes the personal image of the candidate as hard working and concerned with local issues. Campaign activity posts also mobilise supporters and followers and act as a tool for sending out campaign information. It was common amongst Japanese candidates in particular, for upcoming itineraries for the next several days campaigning to be published. Networks created through social media also allow candidates to reach voters who do not follow them but are linked to people who do: "Social media is a good way to create networks between friends and people who know each other and communicate what is going on in a campaign. This is similar to the way in which koenkai work…." DPJ Lower House Member Figure 2: Content of social media output by candidates from UK and Japan.

Note: All figures are percentages
The final way in which campaign activity posts relate to the personal vote and were utilised by candidates relates directly to organisational impact. This took the form of direct appeals for donations or volunteers and by using the kind of positive campaign imagery which may encourage latent activists or previously inactive members of the public to become a part of the campaign. The predominance of image amongst Japanese candidates is reinforced by the lack of posts based on policy which make up only a combined 10.69% of all social media output, compared to overall policy posts totalling 23.71% amongst UK candidates'.
In breaking down the other variables in Table 1 party factors show the Ruling parties, unexpectedly, have the highest SMPV scores, although the difference between all three parties is negligible. The two constituency level variables show expected results, especially between marginal and non-marginal candidates, where there is a statistically significant difference in personal vote seeking behaviour, lending strong support to H3a. Social media use at the personal level tells a similar story to that of candidate websites i.e. less experienced candidates using personal vote posts the least.

Modelling candidates' overall use of new media
In this section, websites and social media are combined into a single measure, Candidate PVI score, which is utilised as the dependent variable in a multivariate model. In order to more robustly test hypotheses.
Candidate PVI scores range between 3.49 for the candidate using a personal vote new media strategy to the greatest degree and -3.29 for the least. Multiple classification analysis (MCA) was then conducted using candidates PVI score as the dependent variables and each of the independent variables under study in order to determine which played a significant factor in candidates overall new media strategy. Multiple classification analysis is an additive modelling technique which is appropriate for interval-level dependent variables and nominal-level independent variables, which is the case here. The Beta coefficients express the explained variance in the dependent variable as a proportion of the total variance, while controlling for the effects of other independent variables (Retherford & Choe, 1993). Results from the MCA can be seen in Table 2 along with the mean PVI scores for each independent variable.
As expected the country variable shows the most significant difference between candidate pursuit of the personal vote. Japanese candidates averaged a PVI score of 0.4497 compared to -0.4008 for UK candidates showing firm support for H1 and proving that thus far, at least in a comparative perspective, campaign traditions have transferred over into the online sphere. At the party level it is clear that membership of the ruling party naturally leads candidates to run to a greater degree on the strengths of their party's record in government and conversely candidates from weaker parties are more likely to pursue personal vote strategies -in direct support of H2. In interviews with candidates, there were concerns from candidates from both the Liberal Democrats and the JIP (the "Third") parties about the effect of their national parties' popularity. For Liberal Democrats, the low popularity of the national party after their time in coalition government had turned many former supporters against them. For JIP candidates, especially those outside of the party's Osaka stronghold, candidates felt that voters had trouble identifying what the party's core polices were. It was more common for candidates from both those parties to be newcomers to the constituency they were running in and lack established support On average both incumbent (0.1102) and marginal (0.2340) candidates were more likely to use new media to pursue the personal vote although only the latter of these differences could be described as statistically significant; candidates in marginal seats were more likely to have a high PVI score than challengers, resulting in firm support of H3a, identifying marginality as a significant factor (β = 0.106).
The other key independent variable is previous terms served, which again shows that it is the least experienced candidates which are using personal vote features the least (-0.5551), providing some support for H4b. Results relating to candidate age show no clear pattern and H4a can safely be rejected. This ties in with the above explanatory factors i.e. challenger candidates, and especially those in non-marginal constituencies, are more likely to use new media to promote themselves under the party label by promoting the party manifesto, talking about national issues and focusing much of their social media output on what party leadership are doing or saying.

Conclusion
The way in which political campaigns are being conducted is changing, inevitably, in line with technological changes within society. What this paper shows, firstly, is that campaign traditions continue to shape candidate behaviour on new media. There is a clear country level difference, providing strong evidence that campaign styles are transferring over onto online platforms. Secondly, moreover, the other factors under study, party, incumbency and experience contribute to how candidates use new media. Candidates representing opposition parties or with greater public profiles and more experience in office, are more likely to campaign on the personal vote, showing that new media campaigning is mirroring conventional theories of personal vote behaviour.
Beyond these findings, however, there is also clear evidence that the promotion of candidate image is a key part of most candidates' online campaigning, regardless of country. This can be expected in Japan but it is perhaps surprising how much of candidate social media output is candidate centred in the UK and implies that regardless of campaign tradition, candidate image promotion is something which is being encouraged by online campaigning platforms. Candidates are no longer so constrained by resources in terms of generating publicity and are using this extra outlet to focus on themselves to a greater degree. Thus far it is clear that new media enhances the ability for independently minded candidates to campaign autonomously and brings into question the influence that parties can exert over their candidates in future campaigns.