‘Party competition and dual accountability in multi-level systems’ the independence echo: the rise of the constitutional question in Scottish election manifestos and voter behaviour

ABSTRACT

Since the formation of the Scottish Parliament, the idea of Scottish independence increased in salience and popularity among Scottish voters to such an extent that it now constitutes the country’s defining political cleavage. Given that Scottish politics is increasingly organized around this constitutional question, support for either side of the debate among voters and elites drives political engagement, election turnout and public attitudes to other major issues. Although much popular and academic work has sought to explain the rise of support for independence, few scholars have explored changes in elite behaviour or its consequences for public opinion. From an elite-driven perspective, the increased salience of independence may be but an echo of elite and partisan attention. Developing hypotheses from this approach, we predict that voters identifying with parties developed stronger views on independence following increased attention in parties’ campaigns. We examine these hypotheses by performing computer assisted, unsupervised content analysis of Scottish Parties’ election manifestos. We then use estimates from a structural topic model to predict change in voter support for independence from the British Election Study. The theory and results suggest that increasing salience on alternative dimensions of politics likely closely relates to elite-driven choices in their election campaigns.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed at https://doi.org/10.1080/17457289.2020.1727486.

Notes on contributors

Zachary Greene is a Senior Lecturer in Political Science at the University of Strathclyde, Glasgow. His research focused on intraparty politics, elections and quantitative text analysis has appeared in journals such as the British Journal of Political Science, European Journal of Political Research, the Journal of European Public Policy, Electoral Studies, European Political Science Review and Party Politics. For additional information on this article and ongoing projects see http://www.zacgreene.com.

Fraser McMillan is a postdoctoral Research Associate at the University of Glasgow. He completed his doctoral thesis on the topic of election pledges and party competition at the University of Strathclyde, Glasgow in 2019, and has previously published-related work in British Politics. For more information about Fraser and his current projects visit https://fraser-mcmillan.com/

Notes

1 Although distinct concepts, issue salience and position relate. A party ignoring an issue (zero salience) has an unclear position on it. Taking up issues necessitates taking positions. Voters may infer parties’ positions on latent issues, but ultimately parties choose whether and how to address them, even those they deem unimportant. Most measures of issue position from textual sources exacerbate the problem as position is a count of opposing issue statements and salience is the count of total statements.

2 We present parties’ left-right positions, measured by the RMP, in the Appendix (Figure A2 and A3).

3 We present qualitative examples of this rhetoric from Conservative and Lib Dem manifestos in the STM discussion in the Appendix.

4 We are limited in potential controls as variables across waves of the Scottish Election Study are inconsistent.

5 We selected terms on face validity, based on a preliminary reading of the manifestos and observation of Scottish campaigns. We measure feature occurrences using Quanteda in R and include wildcards to capture variants on word stems - “nation” (“national”, “nations”).

6 We searched the 2016 Conservative manifesto for “referendum” to ensure mentions referenced independence. 16 of 17 instances do so.

7 SNP manifestos were checked to ensure that “nation” reflected more than mentions to the party’s full name, which is used sparingly.

8 We followed common pre-processing techniques, including converting all words to lower case, reducing words to stems and removing stop words, punctuation, symbols and numbers. We also removed words that did not occur in at least two manifestos and removed a large number of additional terms that did not have obvious substantive meaning. We created a list of collocations (multiple word pairs) which frequently co-occurred within the texts to aid substantive interpretation of the results (e.g. “Scottish Independence”).

9 Topic models require the researcher select k, the number of topics. Few substantive guidelines direct this choice as most approaches measure how quickly the computer estimates the model or the coherence/exclusivity of topics. Models including more topics tend to estimate finer grained results but may mask substantively connected issues, while those containing only a small number of topics may unnecessarily conflate issues using similar words. We find that models using smaller and larger numbers of topics lead to substantively similar results. We present the results of a 10 topic model which succinctly captures three frames used to discuss CIs.

10 See the Appendix for additional qualitative Topic validation.

11 “Exclusivity” measures the extent to which top words in a topic do not appear as top words in any other topic, whereas “coherence” refers to the co-occurrence of top words in the same documents.

12 Topic 1 positively correlates with the salience of the RMP centre-periphery dimension (SNP, 0.63, Con, 0.341. Lab, 0.97), but other topics correlate less so (except the Tories, Topic 3: 0.51). The RMP Independence scale correlates less consistently (SNP .05, Con .51, and Labour .97) suggesting that the RMP’s single issue coding scheme cannot easily account for diverse framing of issues using an independence framework (Gómez, Alonso, and Cabeza 2018).

13 For a breakdown of voter perceptions by partisan identifiers, see Figure A4 in the Appendix.