Case Illustration: Interpretivist Analysis

Understanding Policies (10700193-90)

Yue Hu

Tsinghua University

Administration: Guide for Proposal presentation

• Upload the slides to Web Learning before 13:30, 2026-06-02
• 8-10 min presentation (randomly ordered)
• 1 question (I asked)

Content

• Objective to study
• Justifications
  • Valuable puzzle to be solved
  • Proper design
  • Valid data and method

Not required

• Data
• Analysis
• Results

The doability and value of your project!

答辩

Overview

• large N
• Small N
  • Case study
  • Interview 🌟
    • Focus group
  • Process tracing
  • Content analysis

Interpretivist approach

• Interview methods
  • Interview
  • Focus group
• Discourse methods
  • Content analysis
• Causal inference
  • Process tracing

Roadmap for today. Left column = the method map we built across the term (large-N vs. small-N). Right column = where we land now: the interpretivist toolkit. Three families, ~3 slides each — interview methods, then causal inference via process tracing, then discourse methods in the bonus if time allows.

Opening question (30 s): “Which of these have you already planned to use in your proposal?” — tells you which family to slow down on.

Interview Methods

Why being interpretivist

要克服自弹自唱、自娱自乐、隔空喊话、封闭运行的倾向，进万家门、访万家情、结万家亲，经常同群众进行面对面、手拉手、心贴心的零距离接触，增进对群众的真挚感情。要克服以主观想象代替群众真实需求的倾向，把握群众所需所急所盼，少搞一些强加于群众的活动，多提供一些对路的服务。(习近平，《在中央党的群团工作会议上的讲话》（2015年7月6日）， 中共中央文献研究室 2017, 197)

The epistemological hook. “面对面、手拉手、心贴心” (face-to-face, hand-in-hand, heart-to-heart) is the interpretivist premise: meaning lives in people’s lived experience, not in aggregate numbers. “以主观想象代替群众真实需求” names the positivist trap — imposing the researcher’s own categories on respondents.

Discussion (1 min): “Name something a survey would miss that an interview would catch.” Collect 2-3 answers, then move on.

What’s an Interview

Working definition: a guided conversation to elicit a respondent’s own framing, language, and reasoning — not just facts. Contrast with a questionnaire, where the categories are fixed in advance. The interview’s strength is surfacing the why and how, in the respondent’s words.

How to organize an interview

Brexit: The Uncivil War

Benedict Cumberbatch as Dominic Cummings, main political strategist for Vote Leave,

Watch for (tell students before playing): how Cummings listens more than he talks, follows the respondent’s lead, and lets a contradictory, emotional answer stand instead of correcting it. Afterward ask: “What did he learn that a yes/no poll item would have flattened?”

Prepare for an interview

Techniques

• Structured interview
• Semi-structured interview
• Unstructured interview

Outcome

• Open answers
• Messy details
• Contradictory points

Conditions

• Safe environment
• Design the interview strategy & pilots
• More listening, less talking

Dominic Cummings

Pacing: reveal Techniques → Outcome (left) before Conditions (right) so students see what you get before how to get it good data.

Key line: the “messy, contradictory” outcomes are not failures — they are the data. Tie back to the previous clip. The three conditions are what separate usable mess from noise.

Discussion: “Which interview type fits your proposal, and why?” — 2-3 hands.

Psychological Mindset

The interviewer’s discipline: you are not there to be agreed with or to win. Suspend judgment, tolerate silence, and resist “fixing” the respondent’s view. The gif is the wrong reflex — visible disapproval shuts people down. One sentence, then move on.

Focus Group: An Economic and Societal Version of Interview

Brexit: The Uncivil War

Rory Kinnear as Craig Oliver, the former Director of Politics and Communications for British prime minister David Cameron

How to setup a focus group

1. Size: 5–15
2. Subjects read or watch something.
3. Researcher ask the questions.
4. Subjects express the feelings and opinions.

Focus group vs. interviews

Pros:

1. Real-life (conversational) environment
2. Flexibility
3. Low cost (< case studies)

Cons:

1. Non-privacy-protection
2. Who leads the talk?
3. Selecting interviewees

Craig Oliver

The trade-off in one line: focus groups buy you a conversational, lifelike setting at the cost of control and privacy. The “who leads the talk?” con is the big one — one dominant voice can capture the room, which is exactly the data-quality risk to flag.

Discussion: “Interview or focus group for studying how families decide on a child’s school — and why?”

Take-Home Points

The toolkit

• Interview
  • Structured → semi → unstructured
• Focus group
  • Lifelike, cheap, flexible
  • …but watch who leads & privacy

The principles

1. Meaning lives in the respondent’s own words — don’t impose your categories
2. Match technique to question: depth vs. comparability
3. Messy, contradictory answers are the data
4. Good data = safe setting + piloted design + listen > talk

Interpretivism: understand the why and how, in their words.

Land the four principles, not the bullet list — these are what should survive into their proposals. If time allowed the bonus, add the fifth: process tracing tests rival explanations against a within-case causal sequence.

Bridge to the proposal slide we opened with: “If you’re using interviews or a focus group, which of these four will be hardest to satisfy with the access you actually have?”

Reference

Van Evera, Stephen. 1997. Guide to Methods for Students of Political Science. 1 edition. Cornell University Press.

中共中央文献研究室. 2017. 习近平关于社会主义政治建设论述摘编. 中央文献出版社.

Bonus: Discourse Methods

Process Tracing: Causal Interpretation

Tenet

• Goal: Theory testing
• Hypothesis: Explanation rivalry

信条，2020年上映英国与美国合拍的科幻动作惊悚片，克里斯托弗·诺兰编剧和执导

Core idea to land: process tracing opens the black box between X and Y in a single case. Unlike large-N (which estimates average effects across cases), here we test rival explanations against the within-case causal sequence. The Tenet gif = following the chain of events forward and backward in time.

Example

Why was English recognized as an official language in Singapore?

Only 21% of the population spoke this language when the country was independent.

Run this as a hook. Pose the question, show the language map, and ask students to guess why English before you advance. Let them commit (most will say “colonial legacy” or “it was the majority language”). THEN reveal the 21% callout — the puzzle is that English was a minority tongue, so the obvious answers fail. That tension is what process tracing exists to resolve, and it sets up the six steps that follow.

Process tracing procedure

Signpost slide: the next six slides walk these steps in order, using the Singapore-English puzzle as the running case. Tell students to hold the 21% puzzle in mind — we resolve it at Step 6.

Step 1: Identify Hypotheses

State (institutional) capacity

vs.

Prioritization of the education ministry

Goh Keng Swee (吴庆瑞): 财政部长、内政及国防部长、副总理, 奠定了新加坡教育的精英主义政策, 取得伦敦政经学院经济学博士学位。

The Goh Report identified and set out recommendations to address three shortcomings: high education wastage, low literacy, and ineffective bilingualism. English was taught to maximize proficiency

Step 2: Establish Time Line

1. The very second step right after determining the DV
2. Before the evidence collection

Step 3: Construct Causal Graph

Causal process of X → Y
Used to identify relevant moments and actors that could affect the results.

Step 4, 5: Identify Alternative Choice and Counterfactual Outcomes

Counterfactual 1

Weak State → language of the dominant population.

Counterfactual 2

Lee did not assign Goh as the MOE → the status quo.

A reason that choices might make differences
If there’s no comparative cases, then conterfactual outcomes.

Step 6: Find Evidence for Primary Hypothesis

The primary hypothesis failed the hoop test: when recognition happened, the state was not a strong state

Close the loop: recall the Singapore puzzle — the strong-state hypothesis failed the hoop test at Step 6, so prioritization of education (Goh) survives as the better explanation. That is process tracing doing its job: not “more data,” but rival explanations tested against the causal sequence.

Bridge to their proposals: which of these six steps is hardest to do with the evidence they can actually get?

Content analysis

Cover only if time remains — this section is self-contained and safe to assign as reading. If teaching it live, the through-line is the same three steps every time: sample → code → analyze. Anchor it as “systematic close reading.”

Step 1: Content Sampling

• Terminology:
  • Content analysis: 100 and below
  • Text analysis: above 100 (usually above 1,000)
  • Unit of analysis?
• Sampling technique
  • Convenient sample
  • Stratification
  • Clustering

Step 2: Coding

• Manifest content
• Latent content
• Online assistance: CATMA or eMargin

Step 3: Analysis

1. Context
2. Transparency
3. Transferability

Analytic Tools

Coding

Open coding

Axial coding

Selective coding

Lab Notes: Keep everything in record!!

• Code notes
• Theoretical notes
• Operational notes

Content analysis vs. text analysis

Pros:

• Economic
• Allowing correction
• Overtime survey
• Third party

Cons:

• Limited content
• Limited scale
• Subjective consistency

Understanding Evidence

• When A is a sufficient condition of B, then once A happens, B must happen.
• When A is a necessary condition of B, then there must be A for B to happen.

This is the engine behind the evidence tests two slides later. Quick mnemonic: sufficient = “A guarantees B”; necessary = “no A, no B.” Draw the two arrows on the chalkboard (B) if students look unsure before sending them into the practice.

Practice

• Being 2, 4, 6, or 8 is __ for being an even natural integer.

• Being 2, 4, 6, or 8 is __ for being an even natural integer <10.

• Being highly intelligent is __ for being a student of Tsinghua.

• Showing up to class is __ for getting an A.

• Being mortal is __ for being human.

• Being a rational animal is __ for being human.

Answer key (have students call them out, then confirm):

1. Sufficient, not necessary — these are even, but 10, 12, … are even too.
2. Necessary and sufficient — the even naturals below 10 are exactly {2,4,6,8}. (Edge case to needle them with: does 0 count as a natural number?)
3. Necessary, not sufficient — admission needs it, but many intelligent people aren’t here.
4. Neither — you can get an A without perfect attendance, and attending doesn’t guarantee it.
5. Necessary, not sufficient — all humans are mortal, but so are dogs.
6. Necessary and sufficient — the Aristotelian definition; this is what defines “human.”

Tie-back: the same logic underlies the four evidence tests on the next slide (hoop = necessary, smoking-gun = sufficient).

Type of evidence (Van Evera 1997)

N/S	Yes	No
Yes	Double-Decisive	Hoop Test
No	Smoking Gun	Straw-in-the-Wind

柯南：幸级呲 哇 一次摸 hi 拖呲

1. Hoop test: no alibi, the accused in the state when murder happens? If not innocent, but passing does not mean being guilty. 《顶楼》中女孩儿死时最上面女主在现场，但并不代表就是她造成的女孩儿的死亡（虽然其实是）
2. Smoking-gun test: a smoking gun seen in a suspect’s hand moments after a shooting is quite conclusive proof of guilt, but a suspect not seen with a smoking gun is not proven innocent.《狂飙》中高起强是白手套，京海市市长兼市委副书记赵立冬才是幕后保护伞
3. Double-decisive test: hoop + smoking-gun, like a camera record 柯南说是就是
4. Straw-in-the-wind test: Richard Liu seemed friendly with the girl. 之前女主挽着Richard上电梯，并不代表之后不存在sexual harassment或rape