Party images in Northern Ireland: evidence from a new dataset

ABSTRACT The literature on belief systems in mass publics shows that survey respondents typically have difficulty in describing their images of political parties; only about half offer a meaningful description of how they see individual parties. This paper investigates what people in Northern Ireland think that parties stand for in their home jurisdiction, in Great Britain and in the Republic of Ireland, using open-ended questions in a survey of 1,008 Northern Ireland residents. Northern Ireland respondents resemble those elsewhere, in that only about half seem able to offer a politically meaningful description of what local parties stand for. Among the more politically sophisticated, the Northern Ireland parties are described in ethnonational terms, the British parties are placed in socio-economic (social class and left-right) categories, but few respondents know how to describe the parties in the Republic of Ireland. There is an intriguing asymmetry in the characterization of Northern Ireland’s unionist and nationalist parties: the DUP emerges as only marginally more ‘hard-line’ than the UUP, whereas a great gulf exists between the SDLP and Sinn Féin, the former being perceived as much more moderate. Notwithstanding high levels of electoral stability in Northern Ireland, our findings show that party supporters vary greatly in their levels of political sophistication, perhaps allowing elites greater freedom of action than if all voters were highly politically informed.


Introduction
The very first issue of this journal contained an important article on 'party images' regarding Northern Ireland policy among parliamentarians in the Republic (Sinnott, 1986). More than three decades later, we report here on a new survey that looks at party images from a different perspective, that of the Northern Ireland electorate. This dataset permits us to link empirical findings regarding Northern Ireland to a broader literature on the belief systems of mass publics, in particular that of party images. Public perceptions of parties in Northern Ireland may now be probed in a fuller way than has been possible since the landmark study of the older party system in 1968 (Rose, 1971). The topic is important, not least because people's images of the parties constrain party leaders in their efforts 'to structure political conflict' (Sanders, 1988, p. 583).
We begin by outlining the research domain in which empirical findings on party images may be situated. The general literature on party images in mass publics facilitates the derivation of expectations of how respondents are likely to view the parties. Our methods and principal results are then presented, focusing on the ideological lens through which Northern Ireland respondents appear to view local parties within Northern Ireland, but also on how they view the main parties in Great Britain and in the Republic of Ireland. The conclusion reflects on the implications for our understanding of the formation of party images in contemporary Northern Ireland, but also for the literature on party images more generally.
(A) ideologues, who rely on 'a relatively abstract and far-reaching conceptual dimension' in evaluating political objects, grouped with near-ideologues, who appear to recognize such a dimension but place little reliance on it; (B) 'group interest' respondents, who show no consciousness of any such dimension but evaluate candidates or parties by reference to 'their expected favorable or unfavorable treatment of different social groupings'; (C) 'nature of the times' respondents, who show little appreciation of any policy dimension but associate parties or candidates with particular good or bad times, or evaluate them by reference to purely personal matters; and (D) respondents displaying 'no issue content', who show no policy consciousness whatsoever, who have no idea what a particular party stands for, or who simply do not respond.
We have investigated the extent to which, in evaluating their parties, Northern Ireland respondents may be classified in respect of this typology. The institutionalization of parties in Northern Ireland since the Belfast or Good Friday Agreement of 1998 gives this question particular importance, and raises further questions related to the deeply divided status of the region, the decades of civil unrest it has experienced, and the volatile implementation over the past two decades of a consociational system. According to its critics, the consociational institutions reinforce the underlying division between unionists and nationalists; according to its supporters, this settlement recognizes and accommodates deep divisions that are historically entrenched (see Taylor, 2009, for the various positions, and O'Leary, 2019, for a detailed treatment). There are also reasons to suspect that other political themes may play a significant role in contemporary Northern Ireland, and may even signal an emergent 'normalization' of politics. Economic left-right issues have sparked significant political debate, especially in the early twenty-first century, including tensions over welfare 'reform' (code for austerity according to critics). Moral-religious issuesfor example, laws relating to gay rights and abortionhave also been salient (Garry, Matthews, & Wheatley, 2017;Tonge & Evans, 2015). Strikingly, it was ostensibly over a 'bread and butter' administrative question, a renewable heating initiative that raised questions of corruption and competence, that the power-sharing executive collapsed in January 2017 (McGuinness, 2017).
In assessing the level of ideological sophistication and the extent to which other policy dimensions may cut across the ethnonational cleavage, we asked respondents in a representative survey what they thought, in their own words, each party 'stands for'. This could have elicited several responses. Would they simply describe the parties in ethnonational terms, with the unionist parties characterized as standing for British unionism (and cultural Protestantism), and the nationalist parties standing for Irish nationalism (and cultural Catholicism)? Insofar as people do not characterize the parties in an ethnonational manner, would they view them through an economic or a moral-religious prism? Would many respondents perhaps have no idea what the parties stand for, as the American and comparative research cited above suggests may be the case for a large segment of respondents in contemporary democracies? And how would respondents evaluate the parties in the neighbouring jurisdictions of Great Britain and the Republic of Ireland?
Our analysis is novel in two respects. First, we made no initial assumption about the sophistication of respondents and the dominance of ethnonationalist values, but simply allowed our sample to express, in open-ended form, what they thought each of the parties in the system stood for (to our knowledge, the sole occasion on which this was previously done in Northern Ireland, was in 1968; see Rose, 1971). We avoided presuming that the political system is Northern Ireland is overwhelmingly ethnonational. When given a chance to express the images they have of the parties, if respondents characterized them in non-ethnonational terms (for example, by reference to the economic, moral-religious or other dimensions) that would provide evidence of the development of what is oftenalbeit misleadinglyreferred to as 'normal' politics. Second, as well as assessing respondents' characterizations of Northern Ireland parties we asked them to characterize parties from the associated polities of Great Britain and the Republic of Ireland. This extension allowed us to probe the attitudes of Northern Ireland respondents further: if they described Northern Ireland parties in ethnonational terms, would they describe the British and Irish parties through the same thematic lens, or by reference to other criteria, such as economic or moral-religious categorizations? To what extent would their images of these 'external' parties be diffuse, unclear or perhaps even non-existent?

Perspectives on parties
Northern Ireland's parties are typically characterized as competing in distinct blocs (Mitchell, 1991). In the unionist or Protestant bloc, observers generally see the Democratic Unionist Party (DUP) as more adamantly unionist than the Ulster Unionist Party (UUP). The latter's leaders played a central role in negotiating the 1998 Good Friday/Belfast Agreement which established power sharing, while until 2007 the former was formally opposed to that agreement and to sharing power with republicans. Since the 2003 Assembly election the DUP has overshadowed the UUP as the electorally dominant unionist party, and, shortly after the St Andrews Agreement of 2006, the party's founder, the Reverend Ian Paisley, brought the DUP into government with Sinn Féin, sharing the dual premiership with former IRA commander Martin McGuinness. In the nationalist bloc the more hard-line party also displaced the more moderate party from group leadership in the 2003 Assembly election, completing the ascendancy of tribune parties in both blocs (Mitchell, Evans, & O'Leary, 2009). The Social Democratic and Labour Party (SDLP) had a long track-record of supporting power-sharing, though Sinn Féin has moved a long way from its former support for the IRA, and insistence on a complete British withdrawal from Northern Ireland. It embraced the 1998 Agreement and pragmatically acknowledged Northern Ireland as part of the UK (as long as a majority consent to that status).
Some might have expected these ideological trajectories to have a big impact on perceptions of the parties. Insofar as voters characterize parties in ethnonational terms, the DUP and UUP would be expected to be seen as representing the interests of Protestants and pursuing a unionist agenda, with the former likely to be characterized as more forceful than the latter. Likewise, we would expect that Sinn Féin and the SDLP would be seen as representing Catholics and pursuing a nationalist ideological agenda, with the former seen as more vigorously nationalist.
In contrast to the main unionist and nationalist parties, the Alliance Party characterizes itself as a cross-community party, and its Assembly members do not self-designate as either 'nationalist' or 'unionist'the labelling introduced in 1998 to facilitate consociational voting arrangements in the Assembly. Alliance MLAs designate as 'other'. 1 Thus, we might have expected that our respondents would see Alliance as rejecting the politics of the two main blocs and as seeking to represent Catholics and Protestants equally, steering a middle path between nationalism and unionismin short, as a party that defines itself by reference to its 'centrist position' on the ethnonational dimension.
If social class and economic ideology play a role in determining respondents' perceptions of the parties, it was reasonable to expect, especially given the 'outsider' origins of the party, that the DUP would be seen as more likely to represent the working class, with the UUP more likely to be seen as middle-class. Similarly, research has previously shown that the SDLP has a more middle-class support base than Sinn Féin. 2 At the level of economic ideas, a candidate study in 2003 by Gilland-Lutz and Farrington (2006), and an expert survey by Benoit and Laver (2006) focused on the same period, painted a fairly similar picture of Northern Ireland parties' economic policy positions: Sinn Féin is the most 'left-wing' and the UUP the most 'right-wing, with the three remaining partiesthe DUP, the SDLP and the Alliance Partyclustered around the centre ground.
It was possible that in recent years the parties' positions (and particularly that of the DUP) had both adjusted to new institutional realities and had become more evident to our respondents. Most notably, since 2012 the Northern Ireland Executive and Assembly have been clearly divided over the implementation of welfare cuts introduced by the British government. Notwithstanding its origins as an anti-establishment party, the DUP has adopted a firm 'balance the books' stance, while Sinn Féin has advocated a staunchly left-wing position, refusing to implement retrenchment and vetoing the budget. The main parties, along with the British and Irish governments, have participated in several high-profile initiatives to resolve their differences, the most substantial of which resulted in the 2014 Stormont House Agreement (Coakley, 2018, pp. 341-342). The question of welfare and the parties' differing stances on how best to address benefits have been consistent and dominant features of political news coverage since 2012.
Moral issues have also dominated the media at times in recent years, especially relating to Northern Ireland's exceptional legislative stance on abortion compared to other constituent parts of the UK, and on issues related to gay rights, notably that of same-sex marriage (see Tonge & Evans, 2015, pp. 128-131). Previous studies of party positions by Gilland-Lutz and Farrington (2006) and Benoit and Laver (2006) showed that on moral-religious issues the local parties occupy distinct positions: Sinn Féin and Alliance are the two most liberal parties on these questions, the UUP and SDLP occupy the middle ground, and the DUP is the most conservative on moral-religious matters.
As well as the parties which contest elections to the Assembly, Northern Ireland voters operate within a wider UK framework, participating in elections to the Westminster parliament, so the positions of the main British parties are also of relevance. The British Labour party has traditionally refused to contest Northern Ireland elections, arguing that Labour is already represented there by its 'sister' party in the Socialist International, the SDLP, whose MPs normally accepted the Labour whip at Westminster. By contrast, the Conservatives had a close link with the Ulster Unionist Party dating back to the nineteenth century, a relationship that broke down in 1972 when the Conservative government suspended Northern Ireland's original devolved (but entirely unionist controlled) institutions. The links ended formally only in 1986, when the UUP cut its final ties in vehement opposition to the 1985 Anglo-Irish Agreement, and the Conservatives began to organize in Northern Ireland (O'Leary & McGarry, 1993, pp. 255-256). The party subsequently contested elections in Northern Ireland, but so far has not won a seat in the Assembly (see Garry, 2016, pp. 64-83). 3 Since constitutional debate in Northern Ireland focuses not just on the region's current membership of the United Kingdom but also its possible future membership of a united Ireland, and given that Irish governments retain institutional roles in the workings of the 1998 Agreement, the local evaluation of the main Irish parties also needs to be considered. While a number of smaller parties, including People before Profit, operate on an all-Ireland basis, only one major party, Sinn Féin, does so. As a party whose primary stated goal is Irish unity, this is not, perhaps, surprising; but it is significant that the other three main Irish parties, each of which formally supports Irish unity, only contest elections on the southern side of the border. Fine Gael has never intervened electorally in Northern Ireland. The Irish Labour party has been reluctant to compete in Northern Irish elections, notwithstanding some mid-twentieth century interventions in Belfast. Fianna Fáil has declared its intention to compete in Northern Ireland elections from 2019, but has not put forward candidates in elections there at any level since 1933 (see Garry, 2016, pp. 64-83). 4 It was therefore reasonable to expect that respondents in Northern Ireland would have opinions on those external British or Irish parties which intervene, or might intervene, in elections there. The traditional association of Conservatives with unionism and of Labour as more likely to empathize with nationalism is well known (Cunningham, 2001), and might have been expected to feature in perceptions of these parties, alongside the very obvious divisions between them on matters of socio-economic policy. Unlike the main British parties, the largest Irish parties are not generally regarded as particularly ideologically distinct from one another (Cunningham & Marsh, 2018;Weeks, 2018). Even well-informed Irish voters have difficulty in discerning major differences on economic policy, though there are reasonably clear differences on moralreligious matters between the typically quite liberal Labour party and the traditionally more conservative Fianna Fáil and Fine Gael parties. For this reason, we might expect Northern Ireland survey respondents to have even greater difficulty in distinguishing between these parties.
Comparison across polities enables us to address whether there is a generic or conditional characterization of parties in ethnonational terms. Is the ethnonational division so deeply embedded in Northern Ireland that parties there and in linked polities are characterized in either unionist or nationalist terms? Or perhaps ethnonational factors have pre-eminence in Northern Ireland, but Northern Ireland respondents are fully able to discern the primarily non-ethnonational nature of party stances in Great Britain and the Republic of Ireland? Another possibility is that ethnonational factors are largely irrelevant in linked polities, and that they are only partially relevant in Northern Ireland, with other themes playing a strong role in framing how respondents see the parties. Our aim was to confront these different possible interpretations with reliable and relevant data. Let us explain first how we generated the data.

Measuring party images
When survey-based investigation of party images was at its height in the 1950s and 1960s, researchers used two broad approaches: an open-ended one, where respondents were asked what they liked and disliked about specific parties, or what they saw those parties as standing for, and a closed approach, where they were asked to rate parties regarding specific issue dimensions (Denver, Carman, & Johns, 2012, p. 131). Since our object was to cast our net wide and to collect respondents' descriptions of parties in an open-ended, unframed and unprompted manner, the open-ended approach was the more appropriate: respondents were invited to describe in their own words their images not only of their own local parties but also of those in Great Britain and in the Republic of Ireland.
The Northern Ireland party system offers a particularly attractive laboratory for examining the role played by party images. When the most recent civil unrest was in its infancy, the founding father of survey research across Northern Ireland, Richard Rose (1971, p. 218), pointed out that it had 'a society in which party loyalties reinforce discord rather than allegiance to the regime'. 5 His path-breaking 'Loyalty survey' in mid-1968, of exceptional value as a baseline of public attitudes before the outbreak of sustained conflict, investigated images of the three main parties at the time. Responses to the question 'what would you say the Unionist Party stands for?' saw respondents identify the party with support for the union with Britain (47 per cent) and defence of Protestant interests (16 per cent). In the case of the Nationalist Party (the main voice of the Catholic community), the two most cited images were support for a united Ireland (52 per cent) and defence of Catholic interests (13 per cent). Respondents saw the Northern Ireland Labour Party (a left-leaning, pro-union party that sought to avoid discussion of the 'national question') as a party of welfare socialism and the working class (39 per cent) (Rose, 1971, pp. 223-233).
The context of Rose's survey, conducted in an era of Protestant dominance that was about to be challenged, and a stable dominant party system facing two weak challengers, is strikingly different from the current political world, in which there is a much more even balance in the political resources available to Catholics and Protestants and a multi-party party system with provisions for power sharing. Nevertheless, there was a case for replicating Rose's questions about party images, and we did so in a survey of a representative sample of the adult Northern Ireland population carried out in late 2015 and early 2016. 6 Our questions took the following form: 7 Responses as recorded by the interviewers varied greatly in length, from an abrupt 'no' and a succinct 'crap' to a bitter 214-character critique of the DUP's fundamentalism and an enthusiastic 340-character commentary on the British Labour Party. The minimum length of a response was two characters, the maximum 340 characters, with a median of 12 and a mean of 21. The total length of responses, regarding all 10 parties, ranged from a minimum of 20, with several respondents simply responding 'no' to all 10 parties, to a maximum of 1,339 characters provided by one loquacious respondent (median for all 10 responses, 166; mean, 213). The brevity of these responses, and the often cryptic character of others, made coding difficult, but reflected the reality of respondents' authentic reactions, calling for a number of rather general descriptive categories.
Although respondents were prompted to give additional reactions, as indicated in the question wording, it was possible in almost all cases to code responses under the heading of a single descriptor. In only a few cases did responses appear to imply a second perspective on the party, and in very few was a third position identified. On further analysis, it appeared that taking the second and third positions into account made virtually no difference to the broad pattern. The analysis that follows is therefore based on the primary position with which each party is identified (ignoring those coded as second and third positions). 8 In devising a coding frame to assess systematically how respondents characterize the parties, we wished to distinguish between our three themes of ethnonational, economic and moral-religious. Additionally, we sought to identify those respondents who did not give any substantive response (indicating that they did not know what the party stood for) and also those respondents who indicated a substantive response that does not fall into any of the three major domains mentioned above. Furthermore, to enable more detailed analysis, we distinguishedas did Rose, and the authors of the American Voterbetween the ideological or policy aims of parties on the one hand and how parties may be seen as representing the interests of certain groups on the other. For example, a respondent may describe a party in ethnonational terms and the response could either fall into the ideological category (favours united Ireland / favours the union with the UK) or the group representation one (represents Catholics / Protestants). Similarly, a respondent may describe a party in economic terms, and the coding could fall under either the ideological category (socialist beliefs / right wing free market beliefs) or the group category (looks after the working class / middle class). Additional nuanced measurement in our coding frame facilitates breaking down the ideological characteristics further by reference to strength of belief (for example, very strong or fairly strong nationalist beliefs) and to group representationthe particular way in which the group was described (for example, Catholics, nationalists, republicans, or paramilitaries).
Devising a coding frame for the data was an iterative process. It included: (1) initial examination of the data to provide a preliminary overview of the responses, (2) devising a provisional coding frame capturing different themes and the distinction between groups and ideas, (3) generation of the full set of nuanced categories, (4) categorization of all 1,008 respondents' descriptions of the 10 parties, resulting in 10,080 coded units of analysis, (5) synthesizing categories to produce a more coherent list of responses, and (6) validation of our coding system by a blind parallel review by an independent coder.
Ultimately, the large number of initial categories was reduced to 48, but as some of these distinctions were unnecessary for purposes of analysis our effective number of categories was reduced to 35 (the full coding scheme is described in the appendix). 9 For the most part, classification of responses was unproblematic, but three categories require particular comment. First, an enormous number of respondents expressed no view at all on what the parties stood foran outcome entirely compatible with what we have learned from the American Voter literature (see Figure 1 and discussion below), but surely a source of surprise and dismay to democratic theorists. Typical responses in this category included 'no comment', 'baffles me', 'haven't a clue', 'heard of them but don't know', Note: See appendix for coding system. Northern Ireland data refer to classification of respondents regarding their attitudes towards the five Northern Ireland parties (N = 1,008). For the relationship with class, chisquare = 86, p < .001; for the relationship with voter versus non-voter, chi-square = 99, p < .001; for the relationship with community background chi-square is not significant.
'Dr Henry used to be my doctor', 'all of those are names, all these political goals, life goes on', but also included were such responses as an assessment of the DUP as 'they'll not do a pile either; I'm not much into politics', of the UUP as 'hard to say because they don't know themselves', and of the Alliance Party as 'insurance?' (one of two responses that appeared to confuse the Alliance Party with the insurance company Allianz).
Second, we placed a considerable number of respondents in the 'general' category: people who gave very diverse but unspecific (and usually quite negative) responses, such as 'load of rubbish', 'bunch of idiots', 'gangsters, all about money and power', 'less said the better', 'none of them are any good', 'self-serving', 'party in disarray', 'they've gone to the dogs', and other less printable judgements. A related general category was that in which respondents associated a party with a leader rather than with a policy position, such as 'Hume sold them out', or 'Ian Paisley and that crowd'.
Third, we had anticipated that the moral-religious category would be of considerable significance, but it proved very difficult to distinguish between this category and related ethnonational positions. Thus, it is clear that such assessments as 'they go in for Christian things', 'good values regarding abortion' or 'very religious' fall into the moral-religious category, but we have classified such comments as 'staunchly Protestant rights', 'another Protestant party' or 'Catholics' under the ethnonational heading, even though conceivably in some cases the respondent had in mind religious principles rather than community identity. In adhering here to the principle of consistency in coding, we may thus have been attributing too few responses to the moral-religious category, but we are satisfied that this is likely to have had only a marginal effect, and in any case we have little alternative.
Unfortunately, we were unable to code responses as 'negative' or 'positive' comprehensively. It was easy to identify about 2 per cent of responses as positive (for example, 'helping people' or 'they are alright'), and about 11 per cent as negative (for example, 'bunch of twits' or 'dinosaurs that live; Jurassic Park III'). In most cases, though, such classification was impossible (for example, a description of a party as standing for 'big business' might be either positive or negative, as might even the description 'middle of the road'). We were, however, struck by the fact that it was so much easier to find negative rather than positive descriptions, and the vehemence of the distaste for parties shared by manya pattern in line with broader European trends (Dalton & Weldon, 2005;Mair, 2013, pp. 2-7).

Comparing perceptions of parties
Before even looking at such ideological images of parties as emerge in the words of respondents, it is important to assess their overall reaction to the parties and their capacity, or willingness, to categorize them. Figure 1 presents a summary of the overall level of conceptualization used by respondents in assessing parties, basing this as far as possible on the measurement instrument used by the American Voter group. There are big differences between the two approaches; the American project began with questions about what respondents liked and disliked about the main parties, rather than the more open-ended question on which our analysis is based. This important difference, together with the very different socio-economic and political cultural profile of Northern Ireland and the USA, makes the similarity in findings between the two cases all the more remarkable.
The first three bars in Figure 1 report the position at successive points in time as reported in American studies, showing that the balance between two more conceptually sophisticated groups as described above (ideologues and near-ideologues, and those assessing parties by reference to group interest) amount consistently to about half of the sample, with the remainder divided between those showing little consciousness of any political matters and those entirely unwilling or unable to provide a substantive response. The fourth bar indicates the distribution of our Northern Ireland sample, which shows a remarkable similarity to its contemporary US counterpart. Here, respondents' descriptions of each of the five Northern Ireland parties were first grouped into a four-point scale running from 0 (no issue content) to 3 (ideologue or near-ideologue; see appendix); these were summed, to form a 16-point 'conceptualisation' scale (running from 0 to 15); and this scale was in turn recoded to produce four categories roughly matching those of the American Voter. 10 The remaining bars in Figure 1 show important contrasts between selected groups regarding their levels of conceptualization. First, class matters: 66 per cent of middle-class respondents compared with 40 per cent of working-class respondents fall into the two higher levels of conceptualization, a finding compatible with research in the USA, where the important variable is not class as such, but level of education, a variable not available to us. Second, level of political commitment is important: 65 per cent of those naming a particular party for which they would vote in a future election fell into the two higher categories, as opposed to 35 per cent of those saying that they would probably not vote, that they were undecided, or who simply refused to answeranother outcome in line with American findings. 11 Finally, the last two bars consider Northern Ireland's major source of division, but suggest that there is little difference between those of Protestant and Catholic backgrounds, with 55 per cent of the former and 52 per cent of the latter falling into the two highest conceptualization categories. Overall, these findings suggest that Northern Ireland is by no means unusual from a comparative perspective when it comes to levels of conceptualization of party images.
We now turn to one of our central concerns: the extent to which party images, where they offer meaningful judgements on the parties at all, reflect the three substantive policy orientations introduced earlier: the ethnonational, socio-economic and moral-religious dimensions. We begin with a comparison across the three polities of Northern Ireland, Great Britain and the Republic of Ireland in exploring system-level differences in party characterization.
The three sets of bars in Figure 2 refer in the first place to three jurisdictions: images of five parties in Northern Ireland, two in Great Britain and three in the Republic (we assume that Northern Ireland respondents' images of Sinn Féin, though it is organized on an all-Ireland basis, are conditioned mainly by its role in Northern Ireland). The several categories are grouped as described in the appendix, with substantive responses, whether falling into the ideological or group interest categories, recategorized into three groups of images: ethnonational, socio-economic and moral-religious. Since it is important to test whether religious background matters, each of the three clusters is broken into two bars, one based on Protestant respondents, the other on Catholics. The first and most visible point to emerge from Figure 2 is the enormous difference in perceptions of parties in the three jurisdictions. Respondents had great difficulty making sense of the southern Irish parties, but were able to describe their images of the two main British parties almost as well as they were in the case of the Northern Ireland parties. This contrast is not very surprising. After all, Northern Ireland operates within the UK system, is exposed to British institutions and media, and participates in UK-wide elections. Furthermore, coalitions are relatively rare in the UK, so party brands are likely to be clearer. As well as the prevalence of coalitions in the Republic of Ireland, there has traditionally been limited ideological difference between the parties, so discerning what exactly they stand for is a challenging task. In fact, when a similar question was asked in the Republic 45 percent of respondents were unable to say what Fine Gael, a major governing party, stood for (Marsh, Sinnott, Garry, & Kennedy, 2008, p. 46).
Second, there is a huge difference between perceptions of the political stance of the Northern Ireland parties and their British and Irish counterparts: Northern Ireland parties are overwhelmingly seen as relating to the ethnonational dimension, British ones to the socio-economic dimension, and Irish parties, to the extent that respondents can present any image of them, as socio-economic, but with some ethnonational colouring. In respect of the British parties, it is easy to distinguish between group-centred and policyoriented positions, as indicated by such descriptors as 'working class' and 'socialist', or 'middle class' and 'right-wing'. Not surprisingly, the Conservatives are overwhelmingly seen as middle class and of the right, with Labour seen almost as overwhelmingly as working class and of the left. As well as the more general descriptors (such as 'for poorer people', or 'more socialist'), socio-economic responses for the Labour Party included 'nationalise everything', 'equality between the classes', and 'get off the rich and give it to the poor'. For the Conservative Party, responses were often bitter and dismissive, including 'posh boys', 'cuts, cuts, cuts', 'for the rich and keeping the poor down', 'more for the more well-off people', and 'capitalism unfettered'.
Irish parties, as we have seen, present a particular difficulty. One respondent exclaimed 'I don't even follow the ones in Northern Ireland never mind the ones down south'. Fianna Fáil was predictably puzzling, attracting such responses as 'unprincipled; no kind words', 'party of freedom'; 'a pack of crooks', 'for the people of Ireland', and 'they are running scared of Sinn Féin'. Fine Gael fared a little better; few were able to categorize it at all, but those who did were less critical than in the case of Fianna Fáil: 'more conservative, Catholic orientated', 'moderate nationalism, greater empathy with UK and NI' or 'middle of the road nationalist'. Only the smallest of the three parties, Labour, evoked a more clear-cut ideological response, with many seeing it as representing working class or left-wing interests, though often vaguely (such as 'not sure but probably working-class values', 'similar to the British one', or 'ordinary man'). Very few respondents directly mentioned moral-religious issues (such as abortion, or gay rights) when describing the parties in any of the three politiesa surprising outcome in the Northern Ireland case, to which we will return.
Finally, since Figure 2 breaks down the data by religious background of the respondent we have material to compare the position of the two communities. In fact, there are few differences, but there is one striking exception: the tendency for Protestants to be much more likely than Catholics to respond 'don't know' regarding Irish parties, suggesting greater Catholic interest in and knowledge of the politics of the Republic. By contrast, Catholics are only a little more likely than Protestants to respond 'don't know' to Northern Irish and British parties.

Images of Northern Ireland parties
We now focus in more detail on how the Northern Ireland parties compare with each other in respect of their public images. We move from the more general to the more specific, using three tables. First, Table 1 reorganizes the data presented in Figure 2 in relation to the five main Northern Ireland parties. The two most conceptually advanced categoriesideological and near-ideological, and 'group interest' responsesare merged, and then broken down into three policy dimensions: ethnonational, socio-economic and moral-religious. The 'not policy related' and 'position not described' categories refer to images that correspond with the 'nature of the times' and 'no issue content' respondents.
Some fascinating points emerge immediately from Table 1. The first is the barely perceptible impact of the moral-religious dimension (though, as indicated above, it is possible that this dimension is slightly under-reported). This finding is unexpected, given the long-recognized importance of religion in public life in Northern Ireland and its role not just in the formation of ethnonational identity but also in promoting certain types of politico-religious activism (Ganiel & Dixon, 2008;Mitchell, 2006, pp. 91-132). Yet only 3 per cent of images of the DUP were classified as falling into this category, and virtually none in the case of the other parties. Second, there is considerable asymmetry in respondents' characterizations of the main unionist and nationalist parties. Observers of Northern Ireland typically assume that the DUP and Sinn Féin are the most 'hard-line' representatives of unionism and nationalism respectively, with the UUP and the SDLP adopting more moderate positions within their respective blocs. No such clear pattern emerges from our respondents' assessments: the DUP and UUP are characterized in very similar ways, whereas Sinn Féin and the SDLP are seen as very different. As the table shows, there is just a three percentage point difference between the DUP and UUP regarding the proportion of party characterizations that are ethnonational; by contrast, just over one third of SDLP but fully half of Sinn Féin descriptions fall into this category. When broken down by religious background, these differences remain: 38 per cent of Protestants and 33 per cent of Catholics categorize the SDLP as ethnonational, while the respective proportions for Sinn Féin are 54 per cent and 48 per cent.
Third, notwithstanding its vintage statusit was founded in 1970and high media profile, the Alliance Party seems to have presented a particular challenge to the respondents, with 45 per cent not offering any description of the party at all. In addition to more blunt expressions of indifference or absence of knowledge, such agnostic responses included some that were relatively positive (such as 'best of a bad job', 'trying to help the public' or 'well intentioned'), some negative ('change their mind every two minutes; don't know what they do', 'they fight among themselves', 'I find it hard to believe anything that they say'), or simply puzzled ('are they like the Greens?', 'Chinese? Ethnics, not just Chinese, but all of them?', 'Going by the name it is a united party').
We explore further the significance of the dominant 'ethnonational' label in Table 2. This breaks down the first row of Table 1 into four categories, two relating to the two main communities and their associated ideologies, an intermediate 'cross-community' category (where a party, even if it stood for transcending the ethnonational division, was defined by reference to that), and a residual 'general ethnonational' category. The last of these includes diverse judgements, such as 'bigoted', 'enemy of Ulster', 'noise, shouting, tribal politics', and 'sold out Ulster', and was particularly prominent in judgements of the DUP. The general implications of the table are predictable, with one exception. The DUP and the Ulster Unionists are obviously associated with Protestant or unionist people and ideology, Sinn Féin with Catholic or nationalist people and ideology, and Alliance with an inclusive, bridge-building role. But the SDLP's position is unusual, and it departs sharply from Sinn Féin: only 64 per cent of our respondents saw it as identified with one community, while 31 per cent (36 per cent of Catholics and 27 per cent of Protestants) saw it as essentially cross-community, a remarkable result. Table 3 examines the kinds of ethnonational labels that have been grouped in the discussion above more closely; it includes the first three rows of Table 2, dropping the 'general ethnonational' one. Two of these three rows, referring to the two main communities and their associated ideologies, are further  Notes: The table refers only to the first three categories in Table 2. Each subtotal is broken down into the percentages in italics. Source: NIAES: PIS, 2015-2016 subdivided, first on the basis of whether the images appear to be group-or policy related, and then regarding more refined terminology. The most interesting contrasts are between the top left and the bottom right quadrants. There is hardly any perceived association of the unionist parties with paramilitary groups, and the label 'loyalist' is sparingly used; by contrast, there is a stronger association of Sinn Féin with the IRA (5 per cent) and with the label 'republican' (19 per cent). The Ulster Unionist Party is much more likely to be described as 'moderate' (10 per cent) than the DUP (0 per cent), and this is matched on the nationalist side, where the SDLP is described as 'moderate' (18 per cent) but Sinn Féin is not (less than 1 per cent). However, the fact that so many descriptions of the SDLP fall into the 'cross-community' category highlights the asymmetry in the relationship between the two large parties within either bloc. There is no evidence here, though, of anything analogous to the fascinating contrast uncovered in American research between 'ideological republicans and group interest democrats' (Grossman & Hopkins, 2015).

Conclusion
Our general findings from the dataset may now be summarized. At the most fundamental level, Northern Ireland respondents reacted to questions about what parties stand for in much the same way as respondents elsewhere, and particularly in the USA, where most research in this area has been concentrated. About half do not describe the parties in any politically relevant way, whether because they are entirely unwilling or unable to do so, or because their responses show little or no recognition of the parties as political entities. Among those who provide meaningful assessments of party positions, whether regarding policy or by relation to the groups with which they are seen to be associated, Northern Ireland respondents characterize their own parties almost entirely in ethnonational terms, and British parties in socioeconomic terms. However, theyand particularly Protestants among them have considerable difficulty in describing parties in the Republic at all.
There is also an intriguing asymmetry with respect to how respondents describe the Northern Ireland parties. Contrary to the commonplace characterization of the DUP and Sinn Féin as 'hard-line' or 'extreme' and the Ulster Unionist Party and SDLP as 'moderate', there appear to be only marginal differences in how the two main unionist parties are described, but a much sharper contrast between the images of the two nationalist parties, perhaps a legacy of the conflict, in which Sinn Féin was was the political partner of the IRA. Furthermore, ideas appear to be more important than groups when respondents characterize what parties stand for. There are elements of symmetry between the two main political blocs, but with an important exception: the SDLP emerges as being different from the DUP, the Ulster Unionists and Sinn Féin, in that it is much more likely to be characterized as a 'cross-community' party.
These findings, based on a systematic quantitative and qualitative analysis of a representative sample, provide novel insights into how Northern Ireland respondents mentally navigate and characterize the core parties in their system. Northern Ireland respondents have the capacity to assess the parties in Great Britain through a non-ethnonational lens, and do not universally view matters through an ethnonational prism (though this was dominant in their assessments of the Northern Ireland parties). They are not greatly different from their counterparts elsewhere in their capacity to offer a meaningful categorization of the political positions of the partiesthat is, about half appear to show no capacity to do so. But there may be a positive side to such apparent indifference to party images: party leaders may have a freer hand to conduct political business, since a large section of the electorate effectively opts out of judging them. But, over time, there is little evidence of fundamental change. A half-century after the civil unrest began, the conclusion reached about party images on the basis of the 1968 loyalty survey remains valid: 'collectively, Protestants and Catholics agree about what the major parties stand for' (Rose, 1971, p. 234).

Notes
1. In 2001, three Alliance Party members of the Assembly briefly self-designated as 'unionist' for tactical reasons. 2. These differences are more pronounced on the Catholic side (Coakley, 2008, pp. 776-784). For brief overviews of the origins of the parties see Evans & Duffy (1997) and McAllister (1983). 3. In fact, the 2017 UK general election saw the end of SDLP and Ulster Unionist representation in the House of Commons, where Northern Ireland representation is currently confined to the DUP, Sinn Féin and one independent unionist. 4. In one earlier intervention, Fianna Fáil leader Éamon de Valera was elected to the Northern Ireland House of Commons in 1933 on an abstentionist platform (Elliott, 1973, p. 72 (1973). 6. In all, 1,008 interviews were conducted with a representative sample of the population of Northern Ireland aged 16 and over, using two-stage quotabased sampling (selection of geographical points, and selection of respondents within geographical points). Forty-five sampling points were chosen at random from 285 electoral wards, and quotas were set to ensure that the sample was representative in respect of age, gender and social class based on the latest census estimates. All interviewing was conducted face-to-face using Computer Assisted Personal Interviewing (CAPI). Interviewing was conducted in-home, between 17 November 2015 and 18 January 2016. 7. Despite an exhaustive bibliographical and data search and extensive queries to colleagues, we were unable to find any examples of this kind of wording in any other major survey, apart from the Irish National Election Study of 2002 (Marsh et al., 2008). Richard Rose reports that Philip Converse had failed to get this question included in the 1968 American National Election Survey, and in a similar French survey, suggesting a path-dependent explanation for its exclusion then and later (email from Richard Rose to John Coakley, 7 November 2016). 8. Since there were 1,008 respondents, each of whom was questioned about 10 parties, and we saw non-responses, 'don't knows' and refusals as substantively significant, we recorded a total of 10,080 primary positions, but only 457 second positions (4.5%) and 35 third positions (0.3%). We tried to take account of these (1) by adding the second and third positions to the first positions (to give us a total of 10,572 party images) and (2) by similarly adding these, but adjusting them by a weighting factor (1 for single responses, 0.5 for double responses and 0.33 for triple responses). The resulting distribution of labels made almost no difference; the Pearson's correlation coefficient between each of these amended distributions and the original was 1.000. 9. We began by devising a coding frame on the basis of intensive discussion, and then attempting to apply this as consistently as possible. To test the robustness of our coding decisions we invited a post-doctoral researcher specialising in Northern Ireland politics to familiarise himself with our coding scheme and its implementation, illustrating this by our practice in respect of 1,000 randomly selected responses. We then selected a further 1,000 of our 10,080 responses at random and presented these to the new coder (this time without indicating our coding decisions), and he coded them blind. There was a very high level of agreement between the new coder and the original group in respect of the 35 effective categories (the coder's categorization was identical to ours in 88.7% of cases). We tested the relationship between the new and original codes using Krippendorff's Alpha, which yielded a coefficient of 0.84, comfortably above the level of 0.80 that is conventionally seen as the cut-off for a high level of reliability (see Hayes & Krippendorff, 2007). 10. The 16-point scale was recoded by grouping categories 0-2, 3-7, 8-12 and 13-15. While a large number of respondents showed consistency in their level of conceptualization across parties (for example, all without issue content, or all at ideological level), many of the categories reported here are of course hybrid: a respondent originally coded 0 in respect of two parties, 2 in respect of two more, and 1 in respect of the fifth would be coded as 1 ('nature of the times'). 11. American research shows the great importance of education and degree of political knowledge and interest in determining level of conceptual sophistication (Knight, 1985, p. 838). We have treated social grade as an approximate surrogate of education, a variable that was not available to us. Instead of political interest and knowledge, we have used disposition to vote (1 = will vote for named party at next election, 0 = will not vote, other response). Regression analysis involving these and a range of demographic variables suggests that position on the 16point 'conceptualisation' scale can be explained to some degree by a number of demographic variables: social grade (high), age (old), sex (male), region (greater Belfast) and home ownership (owned, not rented); R 2 = 0.22. When account is taken of these variables, the intention to vote variable is not a significant predictor. 507-ideological general (centrist) 37 700-general 1,466 509-general 1,332 600-associated with leader/individual (past/present) 134 Total 10,080 Grouping of data Figure 1 No issue content 000-115 Nature of the times' 509, 600 Group interest [202][203][204][205][206][207][208][209][210][211][227][228][229][230][231][232][233][234][235][236][237][302][303][315][316][317][318][319][241][242][243][244][245][246][247][248][249][250][251][307][308][309][310][311][312][313][314] Figure 2, Table 1 Ethnonational 202-251 Socio-economic 302-324, 500-507 Moral-religious 400-409 Not policy related 509-600 Position not described 000-115  Table 3 Based on values 202-249