LOSING SUPPORT: THE ROLE OF ISSUES IN PARTY EXIT
Nuffield College, Oxford
It is commonly assumed that the supporters of a political party
have incentives to pull their party towards their own ideological
positions. This paper challenges these assumptions and presents an
alternative issue theory of party loyalty and exit. Using British
Election Study panel data between 1992 and 2001, I illustrate that
the stronger the party identifier the more likely they are to believe
their views are closest to their party, regardless of where they are
ideologically located. The result is that in a model of Conservative
party identification exit between 1992 and 2001, issues matter little,
particularly when controlling for party identification strength. The
implication is that parties are free to move but their main objective
should be to strengthen the partisanship of the growing proportions
of weaker identifiers.
Paper presented in part to the American Political Science Association Annual Meeting,
thWashington DC, 1 – 4 September, 2005, and the Elections, Public Opinion and Parties
thconference, Essex University, 9 – 11 September, 2005.
I am grateful to Jeffrey Freyman, Bonnie Meguid, Mark Andreas Kayser and Maria Sobolewska for
their helpful comments on an earlier version of this paper.
DRAFT: PLEASE DO NOT QUOTE WITHOUT AUTHOR’S PERMISSION 1
“If political parties are primarily concerned with ensuring that their ambitious office
seekers obtain power from the electorate, and if the ambitious office seekers depend
on the electorate to continue to realize their ambitions, then the place to begin to
1understand contemporary partisan politics is in that electorate”.
To understand political party competition we assume that parties compete based on how they think their existing and target voters will respond. The theoretical orthodoxy is that parties have to balance the motives of their existing voters with appeals to new voters. The party base acts as an anchor of stability but also a drag against change. This paper challenges these assumptions and presents a counter-intuitive explanation for why core voters do not restrain a party from positioning itself wherever it wants to.
This paper also links for the first time empirical observations about partisanship to correct misconceptions about core voters. The anomaly is resolved by applying the implications of the Michigan School theory of party identification (Campbell et al, 1960) – that selective
perception, projection effects, or evaluation bias influences the propensity of party supporters to align the position of the party to their own position (or visa versa). According to my argument, parties cannot effectively move away from their strongest core supporters. These supporters will adjust their evaluations to position the party in close proximity to themselves regardless. The stronger the party identifier the smaller the perceived distance will be between individual and party. Consequently, issue distance should not matter when predicting the likelihood of exiting party identification, particularly among the strongest partisans.
1 Aldrich (1995:164)
DRAFT: PLEASE DO NOT QUOTE WITHOUT AUTHOR’S PERMISSION 2
Party members and activists are commonly described as ideologically polarized and motivated by policy oriented goals (for example, see May, 1973; Kitschelt, 1994; Fiorina, 1999). As argued by McLean (1982), why else incur the costs of political participation? However, when we think about the party base it is commonly assumed that core voters are also polarized. For example, the predominant explanation for the Conservative Party‟s predicament subsequent to 1997 has been that an appeal to its core vote prevents an appeal to the centre (Butler and Kavanagh, 2002; Cooper, 2002; Norris and Lovenduski, 2004). This is problematic for two reasons. Firstly, it may be inaccurate to assume that party members are ideologically motivated and divergent in the first place (Whiteley et al, 1994; Seyd et al, 1996; Webb and Farrell, 1999; Whiteley and Seyd, 2002). Secondly, it is not apparent why, compared to members and activists, the Conservative core vote should be divergent to swing voters or even partisans of other parties (Green, 2005).
However, theoretical explanations for party divergence maintain that the ideological location of a party‟s existing vote is an important determinant of party position (Key, 1966; McLean, 1982; Aldrich, 1995; Adams, 2001). This is based on the strongly held assumption that if a party moves too far from traditional positions then its core vote will resist such a change. Supporters will either „exit‟ in the case of less loyal supporters, or use „voice‟ in the case of more loyal supporters, in order to constrain the movement of the party away from their own views (Hirschman, 1970). This becomes particularly damaging to a party unpopular among the wider electorate. In this case the risk of losing existing supporters outweighs the potential gains of attracting new ones, and so weaker parties are forced back onto the less representative positions of their existing voters (Robertson, 1976). “Parties making no effort to break out of their electoral „ghettos‟ can suffer serious decline” (Ware, 1987: 159).
DRAFT: PLEASE DO NOT QUOTE WITHOUT AUTHOR’S PERMISSION 3
Due to the correlation between ideological position and party identification (Heath et al, 1985), it follows that the strongest identifiers will be the most likely to resist a move towards the centre. These individuals are most likely to hold views towards the far ends of, for example, the left-right spectrum, and therefore the distance between partisan and party would
2be the greatest if a party moves. Among more weakly identifying partisans, and overall, we
have witnessed a growing consensus between partisans (Green, 2005a) but where there is divergence, this is most likely to be evident among the most strongly identifying. Although the proportion of strongly identifying partisans is in decline (Särlvik and Crewe, 1983; Dalton and Wattenberg, 2000; Seyd and Whiteley, 2002) a party may be reliant on these voters due to its electoral position, or because they are more likely to turn-out. However, this assumes that the strongest identifiers who are more polarized are also most likely to resist change. I now argue to the contrary.
The „expressive‟ party identification school of thought was put forward by the Michigan School (Belknap and Campbell, 1952; Campbell et al, 1960). Briefly summarized, this theory argued that party identification is a stable characteristic based on enduring strong group attachment and social identity. This is in challenge to later arguments that party identification is a combination of more short-term performance evaluations (see Fiorina, 1981). Much debate has passed between these opposing views (see Franklin, 1992 and Green et al, 2002), particularly regarding the stability of partisanship (Schickler and Green, 1995; Franklin, 1992; Franklin and Jackson, 1983).
If party identification follows the Michigan model, and this paper lends weight to this view, then an important implication follows. This is that perceptions of politics are influenced by the centrality of one‟s political group or social identity, and partisans are more likely to judge
their party favourably (Campbell et al, 1960; Himmelweit et al, 1981; Heath et al, 1985;
2 Note that this polarization may be due to correlations which also exist between strength of party identification and age, political interest and political knowledge.
DRAFT: PLEASE DO NOT QUOTE WITHOUT AUTHOR’S PERMISSION 4
Aldrich, 1995). Importantly, the causal direction is predominantly from partisanship to a positive evaluation. It does not necessitate that the causation can not be in the opposite direction – from positive evaluations to partisanship, but the assumption is that the former direction is more pervasive as an explanation of political attitudes and behaviour. Therefore important vote predictors such as issue distance, leader evaluations and economic evaluations are endogenous to party support, rather than purely exogenous. “It is as though individuals
were looking at the issues and the parties‟ policies through tinted spectacles with the effect of making the party one sympathises with look a bit better and the opposite party a bit worse” (Särlvik and Crewe 1983:220). The stated mechanisms are ones related to a need for psychological consistency or selective perception in what an individual chooses to be exposed to, because of psychological bias resulting from the pervasiveness and importance of the relationship between party identification and, for example, social networks (Zuckerman, 2005). Party identifiers are therefore more likely to demonstrate projection effects, distancing oneself from less favoured parties, minimizing the distance between themselves and their own party or simply being more likely to follow party cues (see Zaller, 1992).
If core voters are more likely to bring positive evaluations to bear, specifically minimising the difference between their own position and their party, then we should expect the strongest identifiers to be the group most likely to do so. It logically follows that if partisanship influences perceptions, more committed partisans will exhibit this effect most strongly. My argument now has clear implications for party behaviour and presents a paradox. We can hypothesise that the strongest identifiers are the most likely to positively bias the evaluation of their party, minimising the difference between their own position and that of the party. However, the strongest identifiers are also the most likely to hold views furthest from the centre ground. Therefore, precisely due to their heightened sense of affiliation to the party, stronger partisans are less likely to constrain a party from shifting its ideological position, because they are most likely to place the party closest to their own views regardless.
DRAFT: PLEASE DO NOT QUOTE WITHOUT AUTHOR’S PERMISSION 5
The implication of minimizing individual and party distance is that, wherever a party locates, we should expect the strongest identifiers to be least likely to exit the party on the basis of issues. The smaller the distance between respondent and party, the less such a variable should influence the propensity to exit.
For example, let‟s suppose, consistent with existing theories, that the decision to identify with
a party depends on the perceived policy consistency between the individual and the party. That is, an individual i, supports party j, on policy k, which generates the highest utility, U,
3the closer the party is perceived to be to that individual‟s own preferences. An individual will
be less likely to exit a party the greater the value of U. Let‟s also assume, that the position of
individual, i, on policy k, is a function of demographic factors, where A, B, C, and D iiii
represent age, gender, social class and educational attainment of individual, i. We can then
write equations for the influences on U. ijk
The first equation captures the effects of demographic factors upon issue position.
(1) W = f (A, B, C, D,) ikiiii
Where Wis the position of individual i on policy k. ik
(2) U = f (W – W) ijkikijk
Where W is the perceived position of party j on issue k. ijk
However, the influence of partisanship strength, P, of individual i, is to minimize W – W, ikijk
3 I use the codification developed by Heath et al (2001:160).
DRAFT: PLEASE DO NOT QUOTE WITHOUT AUTHOR’S PERMISSION 6
-1(3) U = f (( P (W – W)) ijkiikijk
In a statistical model based on U = f (W – W) we should find, firstly, that among party ijkikijk
identifiers issue distance effects are minimal (because among party identifiers we should already find U to be larger than among other respondents in a sample, and secondly, in a ijk
-1model based on U = f (( P (W – W)), controlling for Pwill minimize the significance ijkiikijki
of any issue distance effects upon party exit. This is because the smaller the perceived distance the higher the utility to the individual of identifying with the party, thus minimizing
4the likelihood of exit.
The hypotheses generated by the theoretical arguments proposed are as follows:
1H = More strongly identifying partisans will be more likely to position their party in the
same ideological location as themselves, regardless of their own position
2H = When controlling for strength of partisanship in a predictive model of party
identification exit, the effects of issue distance should be reduced.
The null hypotheses of this paper would suggest that more strongly identifying partisans may be more likely to position their party in the same ideological location as themselves, but only if each group has the same policy view. That is, they align with their party but only where this is reflective of their actual policy positions, rather than partisan bias. The second null hypothesis would conclude that equation 2 above is equal to equation 3: party identification strength does not alter the utility calculation outcome based upon perceived issue distance.
4 If we were to include other predictive factors in the model, such as the likelihood that party j would
implement policy k, or that party j is seen as competent on policy k (as argued by Heath et al) I would
also hypothesize that partisan bias also influences these evaluations in the same direction to minimize party exit.
DRAFT: PLEASE DO NOT QUOTE WITHOUT AUTHOR’S PERMISSION 7
The sample used reflects political context and theoretical orthodoxy. The British Conservative Party has been most commonly described as having a problem of core vote constraint. It is also a party recently reduced for its support to its core voters. During the 2001 British Election Conservative Party identification was almost exactly equal to Conservative Party
5vote intention, whereas Labour had more potential voters than party identifiers. The
Conservative Party has also undergone the most dramatic exodus of party identifiers between 1992 and 2001. As can be seen in Figure 1, even in Labour‟s election of greatest unpopularity (1983), the proportion of respondents in British Election Surveys (cross-sections) identifying with Labour was greater than those identifying with the Conservative Party by 1997 –
plummeting still further by 2001. Therefore if we are to test the predictors of party identification exit, the Conservatives provide the ideal case.
- Figure 2 about here –
Underpinning my argument is the prerequisite that the Conservative Party is perceived to have moved away from its supporters. A centrist move would be consistent with observations (for example, Bara and Budge, 2001). Among Conservative identifiers there was a perceived shift of the Conservative party among Conservative identifiers between 1992 and 2001 on the left-right scale towards a more centrist position, and a shift right-ward on the EU scale in the
6same period. There was also a right-ward shift among politician‟s self-placement between
71992 and 1997, measured using the British Candidate Study.
5 Source: British Election Study rolling thunder campaign panel, 2001. 6 Evaluations of the Conservative position on the left-right index shifted from mean 6.93 to 7.21 among Conservative identifiers in 1992; and from 6.93 to 7.01 among Conservative identifiers in each wave. This suggests that Conservative stayers thought the Conservative party was less right-wing than baseline identifiers in 1992. A further left-ward shift is evident between 1997 and 2001. Conservatives in 1997 placed the party at 6.66 and 6.35 respectively, and Conservative stayers placed the party at 6.66 in 1993 and 6.29 in 2001. Conversely, evaluations of the Conservative position on the EU scale moved from mean 6.03 to 6.79 among Conservative identifiers in 1992; and from 6.12 to 6.88 among
DRAFT: PLEASE DO NOT QUOTE WITHOUT AUTHOR’S PERMISSION 8
However, also integral to my argument is the implication that perceived distance between an individual and the party will be subjective to partisanship. Therefore it is only a difference in distance between identifier and party that would influence the propensity to exit the party.
Figure 2 presents a comparison of party proximity on the left-right index and the EU scale for Conservative identifiers between 1992 and 2001. Due to a loss of Conservative partisan support over the course of both panels, the graph depicts the perceived mean distance among respondents who were Conservative identifiers in the first waves (dashed line) and respondents who said they were Conservative party identifiers in each wave (bold line). Therefore in the first wave identifiers may not have exited, but each point of the bold line represents individuals who have maintained their party identification in that year. Positive numbers denote that the Conservative Party is to the right of respondents, and negative numbers denote that the Conservative Party is to the left of respondents, because;
Issue distance = respondent‟s perceived location of party – respondent‟s self-placement
- Figure 2 about here -
We can see from Figure 2 that rather than a move to the centre being a move away from supporters on the left-right index, the perceived movement of the Conservatives to the left cited above was a move towards its core. Likewise, a move towards the right of the EU scale was a perceived shift towards the more right-wing core voters. This also effectively means that the stronger the identification, the closer individuals would be to the Conservative Party on the left-right index and the further away they will be on the EU scale. However, it should be borne in mind that these evaluations may simply reflect a minimizing of projection effects
Conservative identifiers in each wave. This suggests that Conservative stayers thought the Conservative party was more right-wing than baseline identifiers in 1992. A further right-ward shift is evident between 1997 and 2001. Conservatives in 1997 placed the party at mean 6.81 and 7.71 respectively, and Conservative stayers placed the party at 6.49 in 1998 and 7.84 in 2001. 7 www.pippanorris.com
DRAFT: PLEASE DO NOT QUOTE WITHOUT AUTHOR’S PERMISSION 9
or selective perception. No comparable objective measure exists, and so any perceived location is biased by partisanship – positive or negative.
It is evident from Figure 2 that on the left-right index respondents exiting Conservative Party identification between 1992 and 1997 (the dashed line) were more likely to think they were further from the party, particularly between 1995 and 1996. However rather than conclude that this is the reason these individuals no longer identified with the Conservatives, the identification itself could have lead these individuals to project the party further from their own position. There is no difference between the groups between 1997 and 2001, possibly because projection effects were minimized among this more loyal group of identifiers.
An opposite trend is evident on the EU scale. Here respondents who chose to exit Conservative Party identification were more likely to think the party closest to themselves, whereas loyal partisans were actually more likely to position the Conservative party further from their own view (the bold line). In both cases there are grounds to predict that on this issue the Conservative Party will not lose its supporters – it is the only issue, due to the
distribution of preferences evident here, where the party‟s position does not compete with
another major party. The Conservative Party may lose votes to minor euro-sceptic parties, but the Labour and Liberal Democrat parties are further away from the Conservative Party‟s core vote than the Conservative Party is.
DRAFT: PLEASE DO NOT QUOTE WITHOUT AUTHOR’S PERMISSION 10