CHENG-SHAN (FRANK) LIU
INSTITUTE OF POLITICAL SCIENCE NATIONAL SUN YAT-SEN UNIVERSITY
WPSA 2020 @ Virtual Conference 2020.5.21
Say No to Chinese Identity?
Identification (mis)Measurement and
Overinterpretation in Taiwan
Taiwan’s official South
China Sea Map (2015)
One China
One Taiwan ?
https://cms.qz.com/wp-content/uploads/2019/06/RTX6THCP.jpg?quality=75&strip=all&w=1380&h=920
https://www.6parknews.com/newspark/view.php?app=news&act=view&nid=393789
https://www.taiwannews.com.tw/en/news/3549160
Taiwan voters have multiple views about
• nation ( 民族) : Taiwanese and/or Chinese?
• state ( 國家) : ROC? Taiwan? Unification/Independent?
Chinese---Both---Taiwanese
“In today’s society, some think of themselves first as Taiwanese. Others may think of
themselves first as Chinese, or as both Taiwanese and Chinese. Do you consider yourself as Taiwanese, Chinese, or both?”
https://esc.nccu.edu.tw/course/news.php?Sn=166
https://esc.nccu.edu.tw/course/news.php?Sn=166
Unification---Status Quo---Independence
“Concerning the future Taiwan-mainland China relationship, some think that Taiwan should be independent, while others think we should unify with mainland China. Which comes closer to your view?”
https://esc.nccu.edu.tw/course/news.php?Sn=166
https://esc.nccu.edu.tw/course/news.php?Sn=167#
We have discussed too little about the validity of Identification
measures
And we haven’t had evidence about if they are really trustworthy.
Research
Question How valid are the most commonly used survey questions
• Unification/Independence/status quo
• Chinese/Taiwanese/Both
when they are used to measuring
state/nation identification?
Methodology and Method
• Methodology: the Exploratory Data Analysis (EDA) approach
• Method: Multiple Correspondence Analysis (MCA) : put survey
questions into their context and examine how the “options/choices”
related to each other.
• Explore patterns emerging from a set of categorical variables that are mostly used for studying political identity
• Double check patterns with multiple datasets over the past 7 years
Data
Representative Samples
• F2F Survey: Taiwan Social Change Survey 2013 (TSCS, n=1,952)
• CATI Telephone survey 2015 (n=1,100) Convenient Samples
• Web panel 2015-2016 (n=468)
• Web panel 2019-2020 (n=504)
Main Findings
• Neither Chinese/Taiwan/Both nor
“unification/independence/status quo” has internally consistency in all data sets
• U/I cannot fit into any major (latent) concepts.
• The patterns confirmed in NMF simulation
F2F 2013
Phone 2015
Web panel
2015-2016
Web panel
2019-2020
Web panel
2019-2020
Chinese <- ! -> Both <- !-> Taiwanese
Unification <- ! -> Status Quo <- ! -> Independence
Conclusion & Discussion
• This paper serves as the first piece of empirical evidence that echoes the suspicion about the validity of political identity measures.
1. Rejecting unification does not necessarily equal to favouring independence, while pro-independence cannot be interpreted as anti-unification.
2. Chinese identifies may hold reasoning that is not opposite to those rejecting the Chinese identity.
3. voters who chose “status quo” cannot be interpreted as the set of their
political attitudes stand in between.
Future Studies
• Continue to use EDA to discover multi-layers of reasoning behind the survey questions regarding identification and belonging.
• Provide deeper and insightful interpretation without the presumption of linearity or spectrum for targeted survey questions.
• Re-examine current explanation about the Taiwan voters’ will and
• Provide solutions to avoid accelerated crisis of social polarization along
one’s belonging and misinterpretation of the will of the ”opposite” side.
Comments and
contact Chengshan (Frank) LIU
csliu@mail.nsysu.edu.tw
Figure 7: NMF Analysis of TSCS2013 (k=10)
Figure 8: NMF Analysis of Telephone Survey Data 2015 (k=10)
Figure 9: NMF Analysis of Web Survey Data 2016 (k=10)
