Testing the social identity relative deprivation (SIRD) model of social change: The political rise of Scottish nationalism


Peter R. Grant, Psychology Department, University of Saskatchewan, 9 Campus Drive, Saskatoon SK, S7N 5A5, Canada (e-mail: peter.grant@usask.ca), or Dominic Abrams, Centre for the Study of Group Processes, Department of Psychology, University of Kent, Canterbury CT2 7NP, UK (e-mail: D.Abrams@kent.ac.uk). Both authors made equal contributions to this paper and share senior authorship. Earlier versions of this research were presented at the July 2009 EASP meeting held in Groningen, The Netherlands on ‘Collective Action and Social Change’ and at the EAESP General Meeting in Budapest, 1990.


We tested a social-identity relative deprivation (SIRD) model predicting Scottish nationalist beliefs and intention to vote for the separatist Scottish Nationalist Party (SNP). Data were from a survey of a large and representative sample of Scottish teenagers administered in the late 1980s. The SIRD model distinguishes effects of group-based and personal relative deprivation, which should be independent of one another. Importantly, social change beliefs should mediate the effects of both collective relative deprivation and group identification on protest intentions (in this case intention to vote for the SNP). Egoistic relative deprivation should be the strongest predictor of feelings of depression. Using structural equation modelling, the results strongly support this model and replicate in two different cohorts.


How do social identity and relative deprivation affect support for social change and personal well-being? This broad question has substantial relevance to any country facing a general election, as well as less-formalized changes in which collectives mobilize to challenge the prevailing political structure. In many instances, there are political parties whose aim is to gain independent governance or sovereignty for a particular disadvantaged region within a country. Individuals in such regions typically face both personal and collective deprivation, so the question of whether they become personally distressed and demoralized or become motivated to achieve social change by supporting separatist movements is of fundamental importance. This article proposes a social identity-relative deprivation (SIRD) model that distinguishes the impact of collective deprivation from personal deprivation, and specifies a route through which social identity and collective relative deprivation (CRD) promote a social change belief structure. This structure, in turn, mobilizes support for separatist political movements, including politically separatist voting intentions. In contrast, personal deprivation leads to a decline in personal well-being, including feelings of depression.

Van Zomeren, Postmes, and Spears’ (2008) meta-analysed the relationship between disadvantage, group identification and collective action. Among over 60 articles and over 90 hypothesis tests, only a small subset involved situations of structural disadvantage and even fewer focused on political or ethnic identity. For political movements, only three studies (all by the same research group) examined the disadvantage – action intentions link (fat acceptance, gay people, older people) and only two published studies examined the political identity-action link (Stürmer & Simon, 2004). Surprisingly, social psychology has not yet addressed the question of how social identity and relative deprivation contribute to political support for separatism.

The present research uses a unique data set to investigate this issue for the first time, using the context of the nationalist attitudes and voting intentions of a substantial sample of young people in a specific constituency in Scotland during the politically turbulent period of the late 1980s and early 1990s. We test predictions from the SIRD model, which specifies how strength of group identity and feelings of deprivation should relate to separatist political beliefs and how these, in turn, relate to intentions to vote for a separatist political party, the Scottish Nationalist party (SNP) in this case. Further, the SIRD model emphasizes that egoistic (personal) and collective (group) relative deprivation have independent and different effects. Therefore, testing the model provides a direct examination of this key theoretical distinction. We also test the specificity of the model, showing the distinctiveness of its predictions from those of a more general theory of general political engagement (Duncan & Stewart, 2007).

Regional and historical context of the research

The data for the present research were collected in 1988 from a representative sample of Scottish youth from the Kirkcaldy region, an area within Scotland that experienced relatively high unemployment and economic and social deprivation, and coincidentally, the parliamentary constituency represented by Gordon Brown, (Prime Minister of Great Britain 2007–2010). During the 1980s, Scottish youth regarded the English as a salient outgroup (Abrams & Hogg, 1987). Moreover, these youths are part of an important cohort in Scotland's political history because they are among the voters who, 10 years later, voted by a margin of three to one in favour of reconvening a Scottish parliament that had been adjourned in 1707 (McCrone, 2004). This reflected a process that Reicher and Hopkins (2001) consider a clear example of the construction and political expression of Scottish identity. At the time the present data were collected, Scotland was a predominantly left wing country but was governed by the Conservative UK government in Westminster (London). Scottish resources, such as North Sea oil reserves, were used as a source of revenue by the UK government, and overall poverty in Scotland was substantially higher than in England. The SNP was a minority party but support for its separatist aims was beginning to increase. Now, the SNP has a majority in the Scottish parliament. These data offer, therefore, a unique opportunity to test our model with participants at what is often regarded to be a critical period of political socialization (Campbell, 2006), and who were approaching their first opportunity to exercise their political muscle; that is, at or approaching the age at which they would first be entitled to vote and, in so doing, deliberately choose or reject a politically viable separatist option.

The social identity relative deprivation model

In the SIRD model, we integrate social identity theory with relative deprivation theory to argue for a crucial double mediation role of social change beliefs when predicting intentions to take part in political protest actions – in this instance, intentions to vote for a separatist party. The model also specifies the distinct paths and effects of collective (group) and egoistic (personal) relative deprivation; namely that it is only the former which results in the development of political protest intentions, whereas the latter results in a lack of well-being, in this instance, depressive feelings.

Social identity, social change beliefs, and relative deprivation – a double mediation process

For social identity theory (Hogg & Abrams, 1988; Tajfel, 1978; Tajfel & Turner, 1979), a key question is why members of disadvantaged groups sustain commitment to their group when they perceive that it has an illegitimately low status relative to other groups. Social identity theory holds that, even when an ingroup's status is perceived to be low, strong group identification can act as a psychological barrier that prevents group members from seriously considering leaving or from trying to ‘pass’ into the dominant group (Abrams, Hinkle, & Tomlins 1999; Ellemers, Spears, & Doosje, 1997). More positively, people who feel more strongly bonded to their group should be more likely to care about its illegitimately low social status and to support political movements that seek to address the injustice. Thus, social identity motivates an ‘imagined future’ (Klein, Spears, & Reicher, 2007) through which minority groups who feel unjustly treated by society develop a ‘politicized identity’ that can motivate separatist political intentions (Abrams & Emler, 1992; Kelly & Breinlinger, 1996; Simon & Klandermans, 2001; Stürmer & Simon, 2004).

Tajfel's original formulation of social identity theory (SIT) explicitly proposed that a particular belief structure should mediate between social identity and deprivation on the one hand, and collective action on the other. This belief structure envisions that, when the current social system is under the control of a dominant outgroup, the ingroup's illegitimately low, but stable, status can only be improved if there is a major societal change. Tajfel (1978) proposed that ‘the development of a “social change” structure of beliefs must be accompanied by a great deal of social creativity, i.e., it must be related to the development of new ideologies and attitudes’ (p. 54) which, in turn, motivate support for separatist movements. Thus, the present research offers a distinct and neglected perspective derived from SIT, which helps elaborate on a social psychological process through which social identity and relative deprivation influence political support for separatism.

The context of the present research provided ideal social conditions within which to test our first hypothesis; namely that members of a disadvantaged group who identify more strongly with that group (in this case, the nation, Scotland) will be likely to embrace a social change belief structure (in this case, beliefs in support of political independence), which in turn will be associated with stronger separatist intentions (in this case, intentions to vote for the SNP).

Relative deprivation theory (RDT) addresses both the personal (egoistic) and the group (fraternal or collective) levels of analysis (Runciman, 1966). Egoistic relative deprivation (ERD) results when a person feels that he or she is personally unjustly deprived relative to other individuals and it causes stress and lower life satisfaction as, for example, when a person makes an interpersonal comparison and feels personally dissatisfied because his or her rate of pay is lower than others with the same training and experience (Crosby, 1976). In contrast, CRD results from an intergroup social comparison in which a person feels that his or her group has been unjustly deprived relative to other groups. The cognitive component (cog CRD) is the belief that an expectation (e.g., of wage equality between groups) has been unjustly violated. Cog CRD then leads to an affective reaction (aff CRD) of frustration and dissatisfaction about the injustice. Research confirms that aff CRD is the proximal cause of separatist attitudes and engagement in collective protest actions, which fully mediates the more distal effects of cog CRD (Dube-Simard & Guimond, 1986; Grant, 2008; Grant & Brown, 1995; Kawakami & Dion, 1995; Pettigrew, 2002). However, RDT does not invoke any explanation of the belief structure that might be stimulated by aff CRD or specify the content of subsequent action.

Tajfel (1978) wrote that much of the theoretical development of SIT ‘can be seen as an attempt to articulate some of the social psychological processes which are responsible for the genesis and functioning of relative deprivation’ (Tajfel, 1978, p. 67). An extrapolation from social identity theory that is directly relevant to RDT is that group identification should feed stronger dissatisfaction about injustices against the group (Van Zomeren, Spears, & Leach, 2008).

Taking this one step further, we argue that those with a strong sense of dissatisfaction about injustices against their group will be more likely to embrace a social change belief structure. Our second hypothesis is, therefore, that social change beliefs mediate between aff CRD and stronger separatist intentions (in this case, intentions to vote for the SNP). Surprisingly, this hypothesis has never been directly tested within RD research (Smith, Pettigrew, Pippin, & Bialosiewicz, 2011); Smith & Ortiz, 2002). Figure 1 (above the dotted line) shows this causal path when applied to the motivation to vote for the SNP. Beliefs that Scottish people are deprived relative to the English (cog CRD) should predict feelings of frustration and dissatisfaction at this situation (aff CRD), which should predict Scottish nationalist (social change) beliefs, which should predict intentions to vote for the SNP (intentions to join others in separatist political actions).

Figure 1.

Predicting the effects of collective relative deprivation, strength of Scottish identity, social change beliefs, and political engagement on the intention to vote for the Scottish Nationalist Party.
Note: All the relationships shown are positive. For example, respondents with a stronger Scottish identity were predicted to hold stronger social change beliefs. Dashed arrows indicate causal links that should not be found according to RD theory. Measured variables are shown as rectangles and latent variables as ovals. When two path coefficients are shown, the first is the standardized path coefficient estimated by EQS when testing the full relative deprivation model shown in Figure 1 and the second is the standardized path coefficient estimated when testing the model above the dotted line. Correlations among the exogenous variables in the full model were estimated, but are not shown as they are small (r= .04 to .12). The exogenous variables were assumed to be independent in the model above the dotted line. The insert shows how general voting intentions and political action intentions rather than intentions to vote for the SNP are predicted more strongly by political engagement for the model above the dotted line. *p < .05; ***p < .001.

Applying both Hypotheses 1 and 2 to the case of support for the SNP, we argue that identification with Scotland should affect social change beliefs both directly and indirectly through aff CRD. Further, we predict that both identification and the affective component of CRD should influence the intention to vote for the SNP through social change beliefs (a double mediator). That is, the SIRD model identifies an integrative social psychological process through which identification and collective deprivation will result in members of a disadvantaged group developing strong separatist intentions. Our study directly examines the plausibility of this hypothesized process for the first time through an overall test of the model shown in Figure 1 and through specific tests of the extent to which social change beliefs act as a double mediator.

Independent and different effects of ERD and CRD

The SIRD model follows RD and SIT theories in positing that CRD should have distinctive consequences from ERD. Only CRD should promote collective responses, whereas only ERD should result in personal responses. Although separate lines of existing evidence are consistent with the overall model, there are extremely few direct tests of the independence of the effects of CRD and ERD (see Smith & Ortiz, 2002) and none of these tests are based in real-world politics or with people at the very start of their political participation. The SIRD model makes this distinction explicit and it is tested directly using the present data. Our third hypothesis is that the effects of CRD, which are at the group level of analysis, are distinct from and independent of the effects of ERD, which are at the personal level of analysis. Specifically, aff CRD should relate positively to social change (Scottish nationalist) beliefs and ERD should relate positively to depression, but there should be weaker, and ideally non-significant, relationships between aff CRD and depression and between ERD and social change beliefs. If supported, the comprehensive model will confirm the key theoretical distinction between collective and egoistic RD that is so central to RDT, but that has so rarely been tested directly.

Specificity of the model

As well as testing the overall fit of the SIRD model to the data, it is also important to demonstrate its specificity. The role of social identity is to provide a particular focus for social action. Therefore, social change beliefs should predict SNP voting intentions more strongly than does general political engagement or interest. In contrast, Duncan and Stewart (2007) hypothesize that political engagement underpins political activity generally without implying a specific political goal. Political engagement, therefore, should be a better predictor of intention to vote per se than social change beliefs. If this pattern of results is obtained, the specific role of identity relevant social change beliefs in predicting separatist political intentions will be illustrated.


Sampling frame and procedure

The data were collected during May 1988 as part of the Economic and Social Research Council's multidisciplinary “16–19 initiative” research programme on the political and economic socialization of young people, conducted in four regions of the UK, one of which was Kirkcaldy, Scotland (Banks et al., 1992). Representative random samples were drawn from local educational authority lists of two cohorts of the students enrolled in all of the schools in the parliamentary constituency of Kirkcaldy, thereby ensuring that all respondents’ voting intentions were formulated in relation to the same potential political candidates and the same social and political context.

The first sample was a cohort of 15 and 16 years old in their final year of compulsory schooling during the 1986–1987 school year, while the second cohort was drawn from a cohort of 17 and 18 years old who had been at the same point in their education 2 years earlier. The sample, (N= 911) showed a very high response rate of 72% for the younger cohort and only slightly lower (66%) for the older cohort, attributable to non-tracked changes of home addresses from those in the school records 2 years earlier.


The sample was predominantly self-defined as ‘Scottish’ (91.6%) and single (97.1%), 51.1% were male, 53.7% lived in a family-owned home and 42.6% in public housing. In line with census-based social class estimates, average weekly earnings were low and the majority described themselves as working class (63.2%) rather than middle class (36.6%).


Except where otherwise noted, the measures were responded to on a 5-point Likert scale ranging from ‘strongly agree’ to ‘strongly disagree’.

Relative deprivation

These measures were single items modelled after those used by Guimond and Dube-Simard (1983). cog CRD was measured with the item, ‘People in Scotland generally earn (more, the same, less) than people in England’ while aff CRD was measured with the item, ‘I feel frustrated and dissatisfied about the amount people earn in Scotland compared to people in England’. The measure of ERD was, ‘comparing myself with people in Scotland I am satisfied with the way my life is just now’.


Identification with Scotland was measured using three items: ‘I have a strong sense of belonging to Scotland’, ‘I am not very proud to be seen as Scottish (reversed)’, and ‘It is better to live here than in any other part of Britain’.

Social change beliefs

Social change beliefs were measured by three items, as specified by social identity theory. Specifically Tajfel (1981) wrote, ‘in the case of ethnic or national groups – the exit option is fought for in the form of a separatist movement’ (p. 303) and this option was measured by the item ‘Scotland could get along quite well without the rest of Britain’. Similarly, the items, ‘It is important for Scotland to have control and manage its own resources’ and ‘People in Scotland will only get a fair deal if there is a separate Scottish government’ measure the belief that only fundamental change in the structure of society (separatism) can improve the conditions of all the Scottish people because gradual social change within the current social system is very difficult, if not impossible (Tajfel, 1978). Note that, these belief measures do not specifically imply a voting intention (e.g., there is no presumption that separatism is a personal goal or priority). Indeed our measures allow us to distinguish between specific intentions to vote for each party and general political engagement.

Voting intentions

Participants viewed a list of political parties and indicated either their intentions to vote for one of these ‘if there was a general election tomorrow’, their uncertainty regarding whether to vote, or their intentions not to vote at all. Separate dichotomous variables (coded 1 for the relevant party, 0 otherwise) were created to indicate whether respondents said they would vote for the SNP (15.1%), the official opposition (Labour Party, 32.9%) or the governing party (Conservative Party, 10.4%).

Also a general intentions to vote variable was constructed by coding respondents who indicated that they did not intend to vote, or who were uncertain as ‘0’ (37.6%), while the remainder who named a political party for which they intended to vote, including various small ‘fringe’ parties (3.9%), were coded as ‘1’.

General political involvement

Desire to be politically engaged was measured by the question, ‘How interested are you in politics?’

Intentions to take part in common political protest actions was measured by asking participants whether they ‘would’, ‘might’, or ‘would never’ take part in a boycott, attend a political meeting or a rally, picket, help organize a political meeting, or hand out political, union, or campaign leaflets in the future.

Depressed feelings

Depressed feelings were measured using the following six items, ‘If I could, I would be a very different person from the one I am now’, I feel unsure of most things in my life’, ‘I am happy to be the person I am’ (reversed), ‘I sometimes cannot help but wonder if anything is worthwhile’, ‘I feel that I am as worthwhile as anybody else’ (reversed), and ‘I am often troubled by the emptiness in my life’.


Table 1 shows the descriptive statistics, reliabilities, and correlations among variables included in the structural equation modelling analyses. Respondents identified quite strongly with Scotland (M= 3.80) and agreed with Scottish nationalist sentiments (M= 3.65). They felt quite frustrated and dissatisfied about the wage disparity between Scottish and English workers (M= 3.41). They were only moderately interested in politics (M= 2.23) and, on average, were ambivalent (might do) about whether they would participate in the political activities listed in the survey in the next year (M= 0.76).

Table 1.  Correlations, means, standard deviations, and reliabilities of the major variables
Variable name1.
  1. Note. A high score in this table always indicates more of the construct in question. Respondents indicated whether Scottish people were paid more (1), the same (2), or less (3) than the English (the cognitive component of CRD), and whether they would vote for the SNP if an election were held tomorrow (0 =‘no’, 1 =‘yes’). Political engagement was measured on a 4-point scale ranging from ‘not at all interested’ to ‘very interested’. All other items were answered on a 5-point Likert scale from 1 =‘strongly agree’ to 5 =‘strongly disagree’.

  2. *p < .05; **p < .01; ***p < .001 (two tailed).

1. Cog CRD         
2. Aff CRD.38***        
3. Scottish identity.02.16***      0.49
4. Social change beliefs.08*.31***.39***     0.78
5. SNP voting intention.02.12***.13***.32***     
6. Egoistic RD.11**.17***–.11**–.01–.01    
7. Depression.14***.18***–.13***.04–.03.30***  0.75
8. Political engagement–.06.06.05–.03.15***–.03–.09**  
Standard deviation0.421.040.690.840.360.860.680.84 
N 876903909908893883909901 

Tests of the SIRD model

EQS 6.1 (Bentler & Wu, 2002) was used to test the SIRD model shown in Figure 1. Mardi's coefficient showed that some of the variables were not normally distributed. We used the scaled maximum likelihood estimation procedure with adjusted standard errors, therefore, and we report the Satorra-Bentler scaled χ2S-B and corresponding adjusted comparative fit index (CFI) and statistical significance of the path coefficients (Tabachnick & Fidell, 2007). A model is a good fit if the standardized root mean square residual (sRMR) < .08 and CFI > .95 (Bentler, 2006). The complete model shown in Figure 1 is supported as it is a reasonably good fit to the data; CFI = .91, sRMR = .052; χ2S-B (109, N= 803) = 327.88, p < .001.1

Independence of ERD and CRD

As Figure 1 shows, the analysis supports our third hypothesis, based on the critical prediction from RDT that the effects of collective and egoistic RD are distinct and largely independent of one another. ERD had a strong effect on depression (β= .34, p < .001) and no effect at all on social change beliefs (β= .01, ns), while aff CRD had only a small influence on feelings of depression (β= .13, p < .001).

CRD, identification, and social change

Given the largely independent effects of ERD and CRD, we tested our first and second hypotheses by examining only the relationships that should affect SNP voting intention (the model above the dotted line in Figure 1). This is a very good fit to the data, CFI = .97, sRMR = .030; χ2S-B (33, N= 837) = 76.49, p < .001, with no standardized residual greater than .10.2 To test our (second) hypothesis that the relationship between aff CRD and SNP voting intention should be fully mediated by social change beliefs (Scottish nationalism) we added a direct path from aff CRD to SNP voting intentions. In line with the hypothesis, this non-significant path (β=– .03) did not improve the fit of the model, χ2S-B change (1, N= 833) = 0.79, ns.

The data also supported our first hypothesis, which was that there should be a strong effect of identification on social change beliefs (β= .57, p < .001) which, in turn should strengthen intentions to vote for the SNP (β= .37, p < .001). Adding a direct path from Scottish identity to SNP voting intentions (β=– .08, ns) did not improve the model fit (Figure 1); χ2S-B change (1, N= 837) = 2.18, ns. That is, social change beliefs fully mediated the relationship between identification and intentions to vote for the SNP, providing strong support for the SIRD model.

As expected (see Figure 1), the effect of Scottish identification on social change beliefs is partially mediated by aff CRD. Respondents who identified more strongly with Scotland felt more frustrated and dissatisfied that Scottish people earn less than English people (β= .23, p < .001). In turn, this emotional reaction fed stronger nationalist social change beliefs (β= .21, p < .001), which strengthened SNP voting intentions.

Because social change beliefs were hypothesized to be a double mediator, we compared the fully mediated SIRD model shown in Figure 1 with a model, which added both of the direct paths from identity and from aff CRD to voting intentions. The addition of these two non-significant paths (βs=– .08; –.03, respectively) did not improve the model fit, χ2S-B change (2, N= 837) = 3.02, ns. This analysis, which directly examines the goodness-of-fit of one model nested within the other, directly supports the hypothesis that social change beliefs mediate the effects of both identity and relative deprivation on voting intentions and confirms that no additional direct paths would improve the model.

In sum, these analyses confirm our three key hypotheses. Social change beliefs fully mediate between identity and aff CRD on the one hand and intentions to vote for the SNP on the other, while ERD shows a distinct path to depressive feelings and not to social change beliefs.

Specificity of the model

As Figure 1 shows, political engagement (general interest in politics) does predict intentions to vote SNP. However, in line with our model, the influence of general political engagement is much smaller than that of nationalist social change beliefs (β= .16 vs. β= .37, Zdiff= 4.18, p < .001).3

If our theoretical model is sufficiently well specified, then nationalist social change beliefs should be less predictive of general intention to vote than is political engagement per se. This latter prediction was also confirmed; β= .14 versus β= .35, Zdiff= 4.25, p < .001 one tailed, see insert in Figure 1. Note that the model still has a reasonable goodness-of-fit; CFI = .97, sRMR = .033; χ2S-B (33, N= 837) = 77.68, p < .001.

In addition, we used five variables, such as organizing a political meeting or attending a political rally to index general intentions to engage in political actions for a (unspecified) political party. Some respondents did not complete these measures but the model remained a reasonably good fit; CFI = .94, sRMR = .042; χ2S-B (74, N= 678) = 172.63, p < .001, with no notable changes in the path coefficients to the left of social change beliefs shown in Figure 1. Similar to the findings for general voting intentions, political engagement predicted intentions to take part in these political actions much more strongly than did social change beliefs; β= .48 versus β= .10, Zdiff= 5.41, p < .001, see insert in Figure 1.

Finally, we explored how well the CRD model shown in Figure 1 above the dotted line predicted intentions to vote for the mainstream Conservative and Labour political parties. For the former, the model still fitted the data well, CFI = .96, sRMR = .032; χ2S-B (33, N= 833) = 78.07, p < .001. The only substantial change was the paths from social change beliefs to voting intentions – the less respondents endorsed Scottish nationalist beliefs the more they intended to vote for the ruling Conservative Party (β=– .31, p < .001), somewhat mirroring the path from social change beliefs to intentions to vote for the SNP. In addition, political engagement predicted intentions to vote for the Conservative party (β= .20, p < .001). Neither social change beliefs nor political engagement predicted intentions to vote for the Labour Party very well (β= .07, ns and β= .08, p < .05, respectively).

Reliability of the model

We tested whether the model fit was replicable by treating the two cohorts as independent samples. The model fit equally well within both cohorts, the same pairwise relationships were significant within both and no different paths were indicated by the modification indices. In a multiple group analysis, we constrained the model to be equal for both cohorts in terms of (1) the path coefficients shown in Figure 1; (2) the factor loadings for the identity and social change factors; and (3) the variances of the exogenous variables, including the variances for the errors and disturbances of the endogenous variables. Comparison of the fit of this constrained multiple group model with the same model with these constraints lifted revealed that constraints did not decrease the goodness-of-fit of the model; χ2S-B change (22, Nyounger= 397, Nolder= 363) = 15.41, ns. This confirms that the parameter estimates for the theoretical model shown in Figure 1 obtained from the younger cohort sample are indistinguishable from the parameter estimates obtained from the older cohort sample.


We developed the SIRD model to provide a more comprehensive account of the way relative deprivation and social identity combine to influence collective and personal responses to disadvantage. No previous studies have provided a complete analysis involving all these variables, but, as noted in the introduction, the model integrates other research that provides convergent evidence for the separate paths in the model. The SIRD model is the first to set out explicitly the mediating role of a social change belief structure in the context of personal and collective deprivation, as proposed by social identity theory.

In line with Tajfel and Turner's (1979) presentation of social identity theory, we find that among members of a disadvantaged group, the route from identifying with that group to developing intentions to change the status quo (in this case, voting for a separatist political party) is mediated by social change beliefs (Hypothesis 1). The SIRD model is also novel in specifying that social change beliefs should (and do in this case) mediate between aff CRD and voting intention (Hypothesis 2).This mediated path has not been tested to date because most studies examine either the relationship between aff CRD and militant attitudes, or the relationship between aff CRD and protest intentions or behaviours (Smith & Ortiz, 2002; van Zomeren, Postmes, & Spears, 2008). Most importantly, the model proposes that both aff CRD and social identification become articulated through social change beliefs, which mediate the impact of both variables on specific intentions to change the status quo. By showing that, with social change beliefs in the model, there are no direct paths from either identification or aff CRD to voting intention we strongly confirm the theoretical framework derived from social identity theory.

Distinctive consequences of CRD and ERD

Guimond and Dubé-Simard (1983) highlighted the conceptual and measurement distinction between ERD and CRD but did not fully test their distinctive effects, a gap that remains unfilled, particularly within the same sample and context (Smith & Ortiz, 2002). The SIRD model specifies that, because they relate to different levels of identity, there will be separate and different consequences of ERD and CRD (Hypothesis 3). This hypothesis was clearly supported. Smith and Ortiz's (2002) meta-analysis showed a strong effect of negative interpersonal social comparison on self-esteem, stress, and depression. Consistent with this, we found that ERD, which is based on interpersonal or intragroup comparisons, strongly predicted feelings of depression. In contrast, aff CRD, which is based on intergroup comparisons, predicted Scottish nationalist social change beliefs, which in turn, predicted SNP voting intentions. This reinforces other evidence that aff CRD is the proximal cause of militant attitudes and engagement in protest action and that aff CRD fully mediates the more distal effects of cog CRD (Dube-Simard &Guimond, 1986; Grant, 2008; Grant & Brown, 1995; Kawakami & Dion, 1995; Pettigrew, 2002).

More recently, intergroup emotions theory has emphasized the role of intergroup emotions in collective action and highlights that group and personal emotions have different antecedents and consequences (Smith, Seger, & Mackie, 2007).The present work is compatible with IET because the inclusion of aff RD, as well as the two outcomes – political intentions on the one hand and feelings of depression on the other – can be seen as relating to problem-focused and emotion-focused responses to the situation (see also van Zomeren, Spears, & Leach, 2008).

We also considered the specificity of the model. Might social identification or social change beliefs predict more general outcomes? Might general political engagement and activity be just as, or more, predictive of SNP voting intentions? The data showed clearly that a specific social identity, a specific intergroup comparison, and a specific social change belief structure were associated with a specific voting intention, irrespective of people's general engagement in politics. In contrast, general political engagement more strongly predicted general voting intentions and general political action intentions than did social change beliefs. These findings demonstrate the discriminant validity of our model. Identification with Scotland only motivated political actions that were specifically relevant to the improvement of the comparatively unfair conditions experienced by Scottish people as a whole.

Strengths, limitations, and future directions

Mummendey, Kessler, Klink, and Mielke (1999, p. 230) observed that, ‘studies of real groups in a particular social and historic context are especially rare’. Such studies give us a much better handle on the external validity of theory as well as opportunities to test proposed causal pathways to see if they fit the data. Indeed, the small amount of previous research that studied both national identity and relative deprivation (e.g., Kessler & Mummendey, 2002; Mummendey et al., 1999) has not included measures of political intentions and was conducted post-event (German re-unification). We tested the model using two representative cohorts of young people from a single political constituency and who were approaching their first opportunity to vote for a political representative. Our sample captured the full range of variance in the population of potential activists, rather than the more restricted ranges that might be observed in self-selected samples of activitists, or samples of psychology students participating in studies for course credit. Moreover, the identity and issues in question had longstanding social significance rather than being transitory or having been fabricated for the purposes of the study. Thus, the external and ecological validity of the study are extremely high compared to many others in the field (see van Zomeren, Postmes, & Spears, 2008). Moreover, as our respondents had not yet had an opportunity to vote, it was possible for us to be sure that their intentions were not simply self-reports of prior action or habit and that prior measures in the model were not simply post hoc justifications (cf. Manstead, Proffitt, & Smart, 1983).

We acknowledge limitations in the present data. In common with many multidisciplinary large-scale surveys, even though each survey item was subjected to extensive pilot work and scrutiny, the research required the use of a single or a small number of items to measure each construct. Unusually, however, in the present case the items were designed specifically to test social identity and relative deprivation theories, and the relevant items were drawn from previously established measures in the research literature, enabling us to test the model quite comprehensively. Overall, we feel that the clear support for the theoretical model shown in Figure 1, and the absence of non-predicted relationships among variables, suggests that the measurement quality was high.

Nonetheless, two specific measurement issues deserve mention. First, we are conscious that the measure of ERD, which refers to ‘satisfaction’, has semantic overlap with the depression measures even if the referent is different. Ideally, future research will employ multiple item measures for relative deprivation and, if relevant, will focus more precisely on the same domains (e.g., income) for each type of measure. Second, the reliability of the Scottish identity measure is poor. We suspect that this is because two of the three items are skewed as most respondents identify strongly with Scotland. We felt, therefore, that we were justified in using this scale and then repeating the analysis using the single ‘belonging’ item as a measure of identity because belonging is a key component of many multi-item identity scales (see Jackson & Smith, 1999). Note that the structural equation modelling results confirm the plausiblity of the measurement model for our identity measure as well as strong support for our theoretical model. That is the evidence suggests that our identity measure has validity because it relates to other variables in ways that were expected, a priori, according to theory.

Other aspects of the data merit some comment. The negative correlation between social identification and depression is consistent with the idea that a collective identity can buffer against other threats to the self-image (Major, Quinton, & McCoy, 2002). Importantly, however, this relationship is not substantially mediated by aff RD or social change beliefs, suggesting that it revolves around overall feelings of self-worth rather than the motivation to act on behalf of the collective. This is also consistent with the idea that being embedded in a strong collective identity and community can promote personal well-being (Abrams, 1992; Layard, 2005; Putnam, 2000).

Based on social identity theory and our own previous research, in which the path from identity to aff CRD was significant, whereas the reciprocal path was not (Grant, 2008), we assumed that identification should promote stronger aff CRD. However, the causal relationship could be reversed. The rejection identification model (Branscombe, Schmitt, & Harvey, 1999) would propose that respondents’ emotional reaction to disadvantage should strengthen their identification with Scotland. In fact, reversing the path between identification and aff RD did not change the model fit or reduce the strength of paths from either aff RD or identification to social change beliefs. Thus, the data are compatible with the idea that the causal relationship between identification and feelings of aff CRD are reciprocal as was suggested by van Zomeren, Postmes, and Spears (2008). Our current belief, therefore, is that the causal direction between these two variables may be tilted one way or the other by external contextual factors; these require investigation in future research. Moreover, the causal direction between identification and aff CRD does not undermine the doubly mediating role of social change beliefs for both variables proposed by the SIRD model.

A further limitation of the present research is that we did not have a measure of collective efficacy. Recently, this variable has been highlighted as contributing importantly to protest actions (e.g. Abrams & Randsley de Moura, 2002; Kelly & Breinlinger, 1996; van Zomeren, Postmes, & Spears, 2008). Moreover, just as we have highlighted distinct pathways for ERD and CRD, we would propose that personal efficacy should have distinct effects from collective efficacy, for example, the former having stronger effects on intentions to leave the group (a social mobility belief structure; Abrams et al., 1999). This perspective is generally compatible with the idea that social identities guide our political actions towards imagined futures, given situational constraints generated by existing political realities (Klein et al., 2007; Reicher & Hopkins, 2001). Therefore, a future development of the SIRD model will need to embrace the role of personal and collective efficacy.


Our model suggests that a strong social identity together with intense dissatisfaction arising from social injustices perpetrated against one's minority group is likely to spawn social change beliefs and intentions to take actions designed to right these injustices. Voting is probably the single most important and comprehensive means by which large proportions of a population are directly involved in political change. Our evidence emphasizes the value of a SIRD model to understand and predict intentions to vote for political separatist movements. Beyond voting, however, recent collective protests and other socio-political schisms, such as those in the Balkans, China/Tibet, former Soviet territories, and more recently Tunisia, Egypt, and Libya, suggest that a social change belief structure, especially once it becomes explicitly articulated and shared, may be necessary for people to channel their sense of identification with their group and frustration about its situation into coordinated political action. It remains for future research to test the generality of the SIRD model in different political contexts and with a wider range of intentions and outcomes, such as more costly or non-normative protest actions by members of a disadvantaged minority as they seek to achieve greater social equality.


  • 1

    This model is described as a reasonably good fit because it meets the stricter criterion for the standardized root mean square residual (sRMR < .055), which essentially shows that correlations among the measured variables are estimated to within .055 on average. Hu and Bentler (1999) have shown that models with a standardized RMR < .055 are a good fit even when CFI < .95 provided that N≥ 1,000. All the other models described in the results section meet the conventional criteria for a good fit; namely, standardized RMR < .08, CFI > .95.

  • 2

    Because the internal consistency of the identity items was somewhat low, we retested the model shown in Figure 1 above the dotted line using only one item of obvious face validity as the index of Scottish identity, ‘I have a strong sense of belonging to Scotland’. The model was a good fit to the data and the estimated path coefficients were similar to those shown in Figure 1; CFI = .97, standardized RMR = .032; χ2S-B (19, N= 840) = 59.88, p < .001, with no standardized residual greater than .10.

  • 3

    The Zdiff statistic was calculated by subtracting the relevant unstandardized path coefficients from one another and dividing by the square root of the sum of the squared, Satorra-Bentler corrected standard errors for these path coefficients.