Case Illustration:
Case Study

Understanding Policies (10700193-90)

Yue Hu

Tsinghua University

Overview

Where we are in the toolbox

• Large N (statistics, surveys)
• Small N — today’s focus
  • Case study ← today
  • Process tracing
  • Interview
  • Focus group
  • Content analysis

What you’ll be able to do

1. Define what counts as a case
2. Judge whether a case study is scientific
3. Design a comparison across cases

综合性决策部署要考虑方方面面的差异性，专项性决策部署要找准在全局中的合理定位，这些都需要充分熟悉情况、深入分析论证、科学把握尺度 (中共中央党史和文献研究院 and 中央学习贯彻习近平新时代中国特色社会主义思想主题教育领导小组办公室 2023, 103)。

What’s a case study

Definition

An intensive study of a single unit for the purpose of understanding a larger class of (similar) units (Gerring 2004, 342).

What’s a case

A unit within a boundary

• Biological boundaries (a person, a species)
• Physical boundaries (a city, a year)
• Social boundaries (a party, a policy)

Types of cases

1. 1 unit over time (e.g., China, 1978–today)
2. N subunits at one time (e.g., 31 provinces in 2020)
3. N subunits over time (e.g., 31 provinces, 2000–2020)

Variance is the key — no variation, no comparison.

Observation, Variable, or Case?

For each goal below, what is the case, the variable, and the observation?

1. To study regional differences
2. To study party differences

Observed	Party	Region
1	Party Member	Educational
2	Non-PM	Educational
3	Party Member	Educational
4	Non-PM	Educational
5	Party Member	Dependents
6	Non-PM	Dependents
7	Party Member	Dependents
8	Non-PM	Dependents

The point: “case” is defined by your research question, not by the row of data. Same dataset, different cases.

1. Goal = regional differences → cases are regions; party membership is a variable within each case.
2. Goal = party differences → cases are party / non-party groups; region is the variable.

Scientific Case Study

Goal of Case Studies

Small-N quality criteria

• Credibility — believable account
• Transferability — useful elsewhere?
• Dependability — traceable process
• Confirmability — findings tied to evidence

• Credibility: data and implications presented credibly, based on enough description and background
• Transferability: a result of detailed, illustrative description — responsibility shifts from the researcher (sender) to individuals in other settings (receivers)
• Dependability: what kinds of data, what period of time
• Confirmability: findings clearly linked to analysis, data, and the research site

A case study aims for…

• Deep rather than broad — bounded scope
• Compare with rather than represent other cases
• Mechanism (the how) rather than effect (the how much)
• Deterministic rather than probabilistic relations
• Exploring rather than confirmatory testing

Case Selection

Is random selection a good method? Why/why not?

With small N, randomness gives no guarantee of representativeness — and we usually don’t want an average case anyway. We want a case that earns its keep.

Six strategies for picking this case (next slides):

1. Relevance
2. Policy Relevance
3. Comparison
4. Deviance
5. Illustration
6. Historical Significance

Representativeness is a large-N idea; for small N we choose cases by their analytical leverage, not by sampling probability.

Relevance

The case represents a significant or typical empirical instance of a phenomenon, providing rich data.

• Examining the 2016 US presidential election provides empirical data crucial for understanding contemporary electoral behavior and polarization.

• The lessons from Sweden’s welfare state can be generalized to other advanced economies facing similar social challenges.

Policy Relevance

The case has significant policy implications, offering practical lessons for policymakers.

• The case of Germany’s Energiewende (energy transition) provides valuable insights for countries aiming to transition to renewable energy.

• The study of urban planning in Copenhagen offers policy-relevant lessons for sustainable urban development worldwide.

Comparison

The case offers a unique point of comparison that can help understand broader patterns or anomalies.

• Comparing the political systems of Singapore and Malaysia allows us to explore the divergent paths of development under similar initial conditions.

• Analyzing healthcare systems in Canada and the US offers comparative leverage to understand the impact of different health policy approaches.

Deviance

The case deviates from expected patterns, offering insights into exceptions and prompting theory refinement.

• China’s sustained economic growth without Western democratization challenges conventional theories of development and democracy.

• Singapore’s economic success without liberal democracy is a deviant case that challenges established theories of development.

Illustration

The case vividly illustrates a particular process or mechanism, making abstract concepts more concrete.

• The case study of Rwanda’s post-genocide reconciliation provides a clear illustration of the mechanisms of peacebuilding and transitional justice.

• The example of Brazil’s participatory budgeting process illustrates the practical application and outcomes of participatory governance.

Historical Significance

The case has historical importance, helping to understand key events and their long-term impacts.

• The study of the fall of the Berlin Wall provides critical insights into the end of the Cold War and the dynamics of political change in Eastern Europe.

• Analyzing the Cuban Missile Crisis provides historical insights into Cold War diplomacy and crisis management.

Two styles for a Case Study

Descriptive

• Goal: show the mechanism in rich detail
• Question: How does X work in this case?
• Output: a careful narrative

Analytic

• Goal: test an inference by comparing
• Question: Does X cause Y here, but not there?
• Output: a structured comparison

Let’s see how they work…

Descriptive approach: An Illustration

• What’s the target problem?
• What type of case does she choose?
• How did she do it?
  • What logic did she apply?
  • What empirics did she provide?
    • How do they relate to the problem?
• What conclusions did she draw?
• Are you convincing, and why specifically?
  • Can you do better?

Descriptive Case Study

• Objective: surface the details and mechanism of one case
• Approach:
  1. Pose a clear question
  2. Build a detailed narrative
  3. Call back to the question — does the evidence answer it?
  4. Extend — what does this teach us beyond the question?
• Quality priority: Credibility > Confirmability > Transferability > Dependability

Rule of thumb: a reader who knows nothing about your case should leave understanding both what happened and why it matters.

Analytic approach: An Illustration

• What’s the target problem?
• What type of case does she choose?
• How did she do it?
  • What logic did he apply?
  • What empirics did he provide?
    • How do they relate to the problem?
• What conclusions did he draw?
• Are you convincing, and why specifically?
  • Can you do better?

Analytic Case Study

Objective: testing inferences through comparison

Analytic

• Clear question
• Deliberate case selection
• Designed comparison
• General conclusion

Descriptive

• Clear question
• Detailed narrative
• Calling back to the question
• Extension beyond the initial question

What’s the key difference? → A phase of comparison.

• Quality priority: Credibility = Confirmability = Dependability > Transferability

Core difference: analytic adds a comparison phase, which brings:

• Hypothesis test
• Completion of the inference
• Different quality balance

Take-Home Points

1. What’s a case study
   • A case = a unit within a boundary; it’s defined by your question
   • You must defend why this case (Relevance, Policy, Comparison, Deviance, Illustration, Historical)
2. Logic of a scientific case study
   • Descriptive — depth and mechanism (credibility first)
   • Analytic — designed comparison to test an inference
3. Variance is the key — no variation, no leverage

Reference

Gerring, John. 2004. “What Is a Case Study and What Is It Good For?” American Political Science Review 98 (2): 341–54.

Mill, John Stuart. (1843) 2002. A System of Logic: Ratiocinative and Inductive. University Press of the Pacific.

中共中央党史和文献研究院, and 中央学习贯彻习近平新时代中国特色社会主义思想主题教育领导小组办公室. 2023. 习近平关于调查研究论述摘编. 党建读物出版社.

Bonus: Effective Case Comparison

Show Your Logic

The most classic case-comparison logic:

Mill ([1843] 2002)’s Five Canons (mnemonic: DARC + Joint)

1. Method of Difference — Most similar systems (MSS)
2. Method of Agreement — Most different systems (MDS)
3. Joint Method of Agreement and Difference
4. Method of Residues
5. Method of Concomitant Variations

Two questions every canon answers: What stays the same? and What changes?

Originally published in 1843.

MDS — Most Different Systems

If two or more instances of a phenomenon share only one condition in common, then that shared condition is the cause (or effect) of the phenomenon.

Plain English: cases look different in almost every way, yet the outcome is the same. The one thing they also share must be the cause.

Example: Three cities with very different cultures, economies, and politics all see a drop in crime. The one factor they share? They all expanded street lighting. → Lighting is the candidate cause.

MSS — Most Similar Systems

If an instance in which the phenomenon occurs and one in which it does not have every circumstance in common except one, that one differing circumstance is the cause (or effect).

Plain English: cases look similar in almost every way, but the outcome differs. The one factor that also differs must be the cause.

Example: Two neighboring provinces — same culture, economy, demographics. One adopts a new policy, the other doesn’t. The first sees a drop in pollution; the second doesn’t. → Policy is the candidate cause.

MDS vs. MSS in Practice

	MDS (Most Different)	MSS (Most Similar)
Explanatory variable	Similarity across very different cases	Difference across very similar cases
Control variables	The many differences (assumed irrelevant)	The many similarities (assumed to cancel out)
Key assumption	Systemic factors don’t drive the outcome	Shared traits can be safely “controlled away”
Best for	Finding a universal cause	Pinpointing a specific cause

Residue

Subtract from any phenomenon the part known by previous inductions to be the effect of certain antecedents; the residue is the effect of the remaining antecedents.

Plain English: rule out what you can explain. Whatever is left points to the unknown cause.

Example — diagnosing a rare disease: doctors rule out common conditions one by one; what remains is the candidate diagnosis. (Sherlock’s logic.)

1. Forensic Investigation

Example: The Murder of O.J. Simpson’s Ex-Wife, Nicole Brown Simpson

• Context: In the investigation of Nicole Brown Simpson’s murder, detectives gathered evidence from the crime scene and examined various suspects.
• Process: Investigators eliminated suspects who had alibis or lacked motive, opportunity, or means to commit the crime.
• Result: Despite being highly improbable to some, the remaining evidence pointed to O.J. Simpson as the prime suspect, leading to his arrest and trial. This case highlighted the principle of deductive reasoning in forensic science.

2. Medical Diagnosis

Example: Diagnosing Rare Diseases

• Context: Patients with rare diseases often experience a long diagnostic journey where common diseases are initially suspected.
• Process: Doctors conduct tests to rule out common conditions that fit the symptoms.
• Result: When common diseases are ruled out, doctors consider rare or less likely conditions. For example, diagnosing a rare genetic disorder like Ehlers-Danlos syndrome (EDS) often involves ruling out other connective tissue disorders first.

3. Scientific Discovery

Example: Discovery of the Higgs Boson

• Context: Physicists at CERN sought to confirm the existence of the Higgs boson, a fundamental particle predicted by the Standard Model of particle physics.
• Process: Through experiments at the Large Hadron Collider (LHC), scientists ruled out other particles and interactions that could explain the observed data.
• Result: After eliminating other possibilities, the evidence pointed to the existence of the Higgs boson, leading to its discovery in 2012, despite its elusive nature.

4. Aviation Accident Investigation

Example: Malaysian Airlines Flight MH370

• Context: In the search for the missing MH370 flight, investigators considered various scenarios including mechanical failure, pilot error, hijacking, and terrorism.
• Process: Through extensive analysis of radar, satellite data, and debris, many possibilities were ruled out.
• Result: While the exact cause remains uncertain, eliminating other explanations narrowed the focus to scenarios involving deliberate human action or catastrophic failure, guiding ongoing search and investigation efforts.

5. Historical Investigation

Example: Identifying the Cause of the Black Death

• Context: Historians and scientists debated the cause of the Black Death, which devastated Europe in the 14th century.
• Process: Various theories included bubonic plague, anthrax, and other diseases. Modern research involving genetic analysis of remains ruled out some of these diseases.
• Result: The identification of Yersinia pestis DNA in ancient remains confirmed the bubonic plague as the primary cause, despite initial skepticism due to the speed and mortality of the epidemic.

6. Legal Trials

Example: The Innocence Project

• Context: The Innocence Project works to exonerate wrongfully convicted individuals through DNA testing and reevaluation of evidence.
• Process: By eliminating the possibility that the convicted individual was the perpetrator through new evidence, they uncover the truth.
• Result: Many individuals, such as Kirk Bloodsworth, have been exonerated after eliminating the impossibility of their guilt with DNA evidence, revealing the truth of their innocence.

Concomitant Variation

If two phenomena vary together — one rises as the other rises, or falls as it falls — they are likely causally related.

Plain English: when X moves, Y moves with it. The pattern of co-movement is the clue.

Example: more screen time at night → less sleep; less screen time → more sleep. The systematic co-movement suggests a causal link.

Caution: correlation ≠ causation. Always ask whether a third factor drives both.

1. Studying and Grades

Example: Study Time and Test Scores

• Observation: When you spend more time studying for a test, your test scores tend to be higher. When you study less, your test scores tend to be lower.
• Conclusion: There is a relationship between the amount of time spent studying and the grades you get. More studying generally leads to better grades.

2. Rain and Umbrella Sales

Example: Rainy Days and Umbrella Sales

• Observation: On days when it rains, more umbrellas are sold at stores. On sunny days, fewer umbrellas are sold.
• Conclusion: There is a relationship between the weather and umbrella sales. Rainy weather increases the sale of umbrellas.

3. Ice Cream Sales and Temperature

Example: Ice Cream Sales and Hot Weather

• Observation: When the temperature goes up, ice cream sales increase. When the temperature goes down, ice cream sales decrease.
• Conclusion: There is a relationship between the temperature and ice cream sales. Hot weather leads to more people buying ice cream.

4. Physical Activity and Fitness

Example: Exercise and Fitness Levels

• Observation: Kids who play sports and exercise regularly tend to be fitter and have more energy. Kids who don’t exercise as much tend to be less fit.
• Conclusion: There is a relationship between physical activity and fitness. More exercise leads to better fitness levels.

5. Screen Time and Sleep

Example: Screen Time and Amount of Sleep

• Observation: When you spend a lot of time on screens (like watching TV or playing video games) before bed, you have a harder time falling asleep and get less sleep. When you spend less time on screens, you fall asleep more easily and get more sleep.
• Conclusion: There is a relationship between screen time and sleep. More screen time before bed can lead to less sleep.

6. Practice and Skill Improvement

Example: Playing a Musical Instrument

• Observation: The more you practice playing the piano, the better you get at it. When you don’t practice, your skills don’t improve as much.
• Conclusion: There is a relationship between practice and skill improvement. More practice leads to better piano playing skills.

7. Healthy Eating and Energy Levels

Example: Eating Vegetables and Feeling Energetic

• Observation: When you eat more vegetables and healthy foods, you have more energy throughout the day. When you eat junk food, you feel tired and sluggish.
• Conclusion: There is a relationship between what you eat and how much energy you have. Healthy eating leads to higher energy levels.

8. Homework Completion and Free Time

Example: Finishing Homework and Having Free Time

• Observation: When you finish your homework quickly, you have more free time to play. When you take a long time to do your homework, you have less free time.
• Conclusion: There is a relationship between homework completion and free time. Finishing homework faster leads to more time to play.