18
No class this Thursday.

No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

No class this

Thursday.

Page 2: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Homework #4 is due

Friday by 5pm.

Homework #5 is due Friday,

March 14th, by 5pm.

Page 3: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Political Science 15

Lecture 16:

Small-N Methods

Page 4: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Qualitative vs. Quantitative

Research “Large-N” studies look for patterns in a large

number of cases. Random selection of cases.

“Small-N” studies examine a small number of

cases in depth. Deliberate selection of cases.

Usually large-N approaches have better external

validity, while small-N approaches have better

internal validity and measurement validity.

Probabilistic versus deterministic causation.

Page 5: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Types of Small-N Methods

The comparative method. Based on J.S. Mill’s

System of Logic (1843).

Most-similar method (“method of difference”).

Most-different method (“method of agreement”).

Qualitative comparative analysis (QCA).

Case study methods (when N = 1).

Page 6: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

The Comparative Method:

Most-Similar Design

The most-similar method is the most common

approach to small-N research problems in political

science.

Examine a handful of cases that are as similar as

possible, except on the outcome of interest (the

dependent variable).

Similarity of cases means we control for many

alternative explanations.

If one factor is different between cases, and outcome is

different, this is our probable cause for the outcome.

Page 7: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Example: Causes of Revolution

Potential Causes England France

Repressive Monarchy Yes Yes

Nonpropertied Agrarian

Proletariat

Yes Yes

Expensive Foreign Wars Yes Yes

Stagnant Standard of

Living

No Yes

Outcome Stability Revolution

Page 8: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Example: Dreze and Sen

Potential Causes India China

Disadvantaged Beginning Yes Yes

Moderate Economic

Growth

Yes Yes

Adequate Calories Per

Person

Yes Yes

System to Distribute

Public Resources

No Yes

Life Expectancy 57 69

Page 9: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Problems with the

Most-Similar Method

We generally treat the independent variables as

something simple (yes/no, for instance). The

more complicated the operationalization, the

harder this method is to do.

Deterministic causality.

Multiple causal factors and causal complexity are

hard or impossible to determine.

External validity is low.

Page 10: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

The Comparative Method:

Most-Different Design

This is the opposite of the most-similar method.

Examine a handful of cases that are as different

as possible, except on the outcome of interest

(the dependent variable), which is the same.

Difference of cases means we control for many

alternative explanations.

If one factor is the same between cases, and

outcome is the same, this is our probable cause

for the outcome.

Page 11: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Example: Causes of Revolution

Potential Causes China (1927) France (1789)

Repressive Monarchy No Yes

Nonpropertied Agrarian

Proletariat

Yes Yes

Expensive Foreign Wars No Yes

Stagnant Standard of

Living

No Yes

Outcome Revolution Revolution

Page 12: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Problems with the

Most-Different Method

As with the most-similar method, we can’t use complicated variable codings, multiple causal factors are hard or impossible to determine, and external validity is low.

Deterministic causality.

Case selection on the dependent variable – without variation on the dependent variable determining causality is extremely difficult.

This method is more useful for ruling out “necessary” causes than determining causality.

Page 13: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Qualitative Comparative Analysis

(QCA)

A medium-N method (N between a few and 50 or so). A middle ground.

Independent variables are coded as binary (yes/no).

Sequences of these variables are entered into a “truth table.”

Reach conclusion through Boolean logic – which combinations of factors produce which outcomes?

Page 14: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Example: Military Coups

Cases

Internal

Military

Conflict

Death of

Dictator

CIA Involvement

Coup?

9 0 0 0 0

2 1 0 0 1

3 0 1 0 1

1 0 0 1 0

2 1 1 0 1

1 1 0 1 1

1 0 1 1 1

3 1 1 1 1

Page 15: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Problems with QCA

As with small-N methods, we can’t use

complicated variable codings.

Deterministic causality.

However, multiple causal factors/causal

complexity can possibly be distinguished, and

external validity is higher.

A tradeoff between large-N and small-N

positive traits.

Page 16: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Case Studies

These are methods of examining a single case.

In a sense we have an N of 1, but we must make some kind of comparison in order to make a causal inference, either within or across cases. This comparison can be implied.

Internal and measurement validity very high, external validity very low – we know a lot about one case, but very little about how our observations will generalize.

Page 17: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Case Study Methods

Extreme Case: A clear example of a hard-to-measure concept: Nazi Germany for Fascism, North Korea for isolated state.

Typical Case: Examine a representative or average case in depth.

Crucial Case: Classic examples of a concept, or most/least likely cases.

Counterfactual Case: Consider causality if an independent variable had a different value.

Can combine case studies to examine a range of cases (e.g., two extreme cases, one from each end of the spectrum).

Page 18: No class this Thursday. - UCSBglasgow.faculty.polsci.ucsb.edu/ps15/ps15lect16.pdf · Most-Different Design This is the opposite of the most-similar method. Examine a handful of cases

Small- versus Large-N Methods

All methods are a tradeoff between

internal, external, and measurement

validity.

Small-N lets us go more in depth into our

cases, but generalizing is harder.

No matter the method, our goal is the

same. We want to make a causal inference

and learn how the world works.