Fairer Council Tax: consultation analysis

Footnotes

1. The Commission on Local Tax Reform. Volume 1 – Just Change: A New Approach to Local Taxation. Commission on Local Tax Reform, 2015. https://web.archive.org/web/20220119190235/https:/localtaxcommission.scot/download-our-final-report/

2. Scottish Government. Consultation on a Fairer Council Tax. Scottish Government, 2023. https://www.gov.scot/publications/consultation-fairer-council-tax/pages/2/

3. Scottish Assessors Association. “Council Tax Bands.” Accessed 1 December 2023. https://www.saa.gov.uk/council-tax/council-tax-bands/

4. The Council Tax Reduction Scheme is a means-tested scheme to reduce or eliminate Council Tax liability for lower income households, depending on household circumstances and ability to pay. Local authorities have responsibility for administering the Council Tax Reduction Scheme.

5. Phillips, D. “Scottish council tax proposals are a small step in the right direction but duck the biggest issue: revaluation.” Institute for Fiscal Studies, 2023. Accessed 1 December 2023. https://ifs.org.uk/articles/scottish-council-tax-proposals-are-small-step-right-direction-duck-biggest-issue

6. For example, “Not Answered”, “No comment”, “Ditto” or “See above”.

7. After filtering the dataset for respondents who selected “Organisation” as their respondent type and filled in the input field “Full name or organisation’s name” with the name of a council, we manually reviewed all responses that remained. Responses that mentioned “I am a councilor” or similar were removed.

8. This total is less than the 15,628 total respondents in Table 1 as not all respondents answered this question.

9. In other words, respondents who answered “Yes” to the closed-format component of Question 1.

10. For example, one respondent expressed their belief that that new builds were assigned to higher property bands that did not reflect the value of the property.

11. “Respondents who agreed with a proposed tax increase” refers to respondents who answered “Yes” to the closed-format component of Question 1 (“Do you think that Council Tax in Scotland should be changed to apply increases to the tax on properties in Bands E, F, G, and H?”)

12. Although the text of question 3 prefaced that the question should only be answered by respondents who replied “No” to question 2, the consultation did not include survey routing, so all consultation respondents could answer this question (regardless of their response to question 2).

13. The same caveat applies here as with question 3: even though the question text stated that only people who selected “Other” should respond, all respondents could answer this question as the consultation did not include survey routing. Out of the 10,700 responses to this question, 9,765 selected “Other” and 804 selected one of the defined options to question 4 (the remaining 131 respondents did not provide an answer to the closed-format component of question 4).

14. “Budget Council tax” refers to a 3% increase across all Council Tax bands – this was the maximum increase allowable by the Scottish Government for 2022/23.

15. This could be due to many respondents not feeling qualified to make a judgment on this question (because they did not know how to answer or they felt it would not affect them as non-island dwellers).

16. UK Government. “Consultation principles: guidance.” 2018. Accessed 1 December 2023. https://www.gov.uk/government/publications/consultation-principles-guidance

17. Scottish Government. “Consultations in the Scottish Government: Guidance.” 2022. Accessed 1 December 2023. https://www.gov.scot/publications/consultations-in-the-scottish-government-guidance/

18. This hybrid approach is often used in academic fields such as sociology whose research methods include analyses of large-scale text data, for example: Li, Zhuofan, Daniel Dohan, and Corey M. Abramson. 2021. ‘Qualitative Coding in the Computational Era: A Hybrid Approach to Improve Reliability and Reduce Effort for Coding Ethnographic Interviews’. Socius 7 (January): 23780231211062345. https://doi.org/10.1177/23780231211062345.

19. Abbasi, Ahmed, Suprateek Sarker, and Roger Chiang. 2016. ‘Big Data Research in Information Systems: Toward an Inclusive Research Agenda’. Journal of the Association for Information Systems 17 (2). https://doi.org/10.17705/1jais.00423.

20. Nelson, Laura K., Derek Burk, Marcel Knudsen, and Leslie McCall. 2021. ‘The Future of Coding: A Comparison of Hand-Coding and Three Types of Computer-Assisted Text Analysis Methods’. Sociological Methods & Research 50 (1): 202–37. https://doi.org/10.1177/0049124118769114.

21. Chen, Nan-Chen, Margaret Drouhard, Rafal Kocielnik, Jina Suh, and Cecilia R. Aragon. 2018. ‘Using Machine Learning to Support Qualitative Coding in Social Science: Shifting the Focus to Ambiguity’. ACM Transactions on Interactive Intelligent Systems 8 (2): 9:1-9:20. https://doi.org/10.1145/3185515.

22. Grandeit, Philipp, Carolyn Haberkern, Maximiliane Lang, Jens Albrecht, and Robert Lehmann. 2020. ‘Using BERT for Qualitative Content Analysis in Psychosocial Online Counseling’. In Proceedings of the Fourth Workshop on Natural Language Processing and Computational Social Science, edited by David Bamman, Dirk Hovy, David Jurgens, Brendan O’Connor, and Svitlana Volkova, 11–23. Online: Association for Computational Linguistics. https://doi.org/10.18653/v1/2020.nlpcss-1.2.

23. “Trusting Your Data with Google Cloud.” Google Cloud Whitepaper, December 2022. https://services.google.com/fh/files/misc/072022_google_cloud_trust_whitepaper.pdf. In particular, “Confidential Computing allows you to encrypt data in the cloud while it’s being processed. With the confidential execution environments provided by Confidential VM and AMD SEV, Google Cloud keeps customers’ sensitive code and other data encrypted in memory during processing. Encryption keys are ephemeral, generated on chip and are non-exportable based on the CPU-based encryption engine that transparently encrypts and decrypts the data in memory. Encryption keys are kept hidden from untrusted parts of the platform and most importantly non-extractable by software. Google does not have access to these encryption keys.”

24. These included Comhairle nan Eilean Siar, Argyll and Bute Council, Orkney Islands Council and Shetland Islands Council.

25. Fereday, J. and Muir-Cochrane, E. (2006) ‘Demonstrating Rigor Using Thematic Analysis: A Hybrid Approach of Inductive and Deductive Coding and Theme Development’, International Journal of Qualitative Methods, pp. 80–92. DOI: 10.1177/160940690600500107.

26. “Multilabel” refers to the fact that more than one theme can be assigned to an individual response.

27. Labels refer to any characteristic that the researcher wants to predict: this includes the key themes and ideas in consultation responses but could also include other characteristics depending on the nature of the dataset (for example, a researcher might want to predict the subject of a news article for a large dataset of news articles).

28. GPT-2 is an earlier version of the current LLM underlying ChatGPT, GPT-4. The key difference between GPT-2 and ChatGPT is that GPT-2 is not a cloud service (unlike ChatGPT) and exists completely separately from OpenAI servers, as the model is finetuned and deployed by Alma Economics directly. This means that OpenAI cannot access or view any data passed to the GPT-2 model. In addition, GPT-2 is smaller in size and has been trained on less text data than ChatGPT/GPT-4, and as a result underperforms GPT-4 on specialised tasks such as creating music or storytelling. Unlike ChatGPT/GPT-4, the source code for GPT-2 has been made publicly available by the OpenAI team.

29. Edwards, Aleksandra, Asahi Ushio, Jose Camacho-Collados, Hélène de Ribaupierre, and Alun Preece. 2023. ‘Guiding Generative Language Models for Data Augmentation in Few-Shot Text Classification’. arXiv. https://doi.org/10.48550/arXiv.2111.09064.

30. Guo, Zhen, Peiqi Wang, Yanwei Wang, and Shangdi Yu. 2023. ‘Improving Small Language Models on PubMedQA via Generative Data Augmentation’. arXiv. https://doi.org/10.48550/arXiv.2305.07804.

31. Largeron, Christine, Christophe Moulin, and Mathias Géry. 2012. ‘MCut: A Thresholding Strategy for Multi-Label Classification’. In Advances in Intelligent Data Analysis XI, edited by Jaakko Hollmén, Frank Klawonn, and Allan Tucker, 172–83. Lecture Notes in Computer Science.