Methods of Data Collection in Taiwan Political Research

These are some notes taken from Jonathan Sullivan's talk at the 2011 Taiwan Studies Summer School at SOAS.

1. Where does data come from? It is empirical information about the world.
2. Which methods are Taiwan scholars using?
3. How should we go about choosing methods and data? This should be driven by the research questions and research design.
4. What empirical evidence will prove/ falsify our hypotheses?
5. Existing data can be found using Inter-University Consortium for Political and Social Research, TEDS Barometer, WVS etc.
6. With research on elites, text is an underutilised source, a bi-product of strategic political activity. It can be written or spoken - manifestoes, articles, campaign ads, speeches, discussion transcripts etc. Taiwan sources are, for instance: The Mainland Affairs Council and The Legislative Yuan websites.

How to cut text down to be able to generate data from source to do a text or a content analysis?

What does the text reaveal about the attitudes, sincerity, authenticity and behaviour of the speaker through, for instance, the frequency of certain lexical items and the context or cotext through corpus analysis using  concordancing software and automated methods?

The Political Compass, spin doctors and prepared texts.

What is the purpose of content analysis applications? Ideal point estimation, salience, framing, sentiment analysis and conflict scores. Automated methods preferable to hand coding.

Example:

Research on Chen Shuibian's independence discourse and strategic considerations related to audience are here and here.  While the content shows a discourse of independence, independence, independence and the style is increasingly provocative and extremist, ambiguous and contradictory on issues like independence, identity and sovereignty, when one looks at the audience he is speaking to the discourse takes on a different significance.

Process for text analysis:

1. Download text.
2. Build dictionary using Yoshikoder
3. Apply dictionary to text.
4. Determine audiences.
5. Determine key events.
6  Aggregate audiences.

Using statistical analysis like regress mentions and speech proportions, attitudes, sentiment, sincerity and attitude can be determined.

Don't select data based on what you're hoping to prove.