How to Compare Countries Lecture 2

1
How to Compare CountriesLecture 2

• Michaelmas Term 2004
• Dr. David Rueda

2
Practical Stuff

• The reading list and the outline for the first
  lecture is now available at
• http//users.ox.ac.uk/polf0050/
• (Go to Teaching and Oxford University)

3
Today

• Main Points from Last Weeks Lecture (Small N
  analysis).
• Theory-Driven Small N Analysis.
• Initial Thoughts.
• Challenges.
• Choosing Cases in Theory-Driven Small N Analysis.
• Mills Methods for Small N Analysis.
• Most Similar Systems Design.
• Most Different Systems Design.
• Problems of Most Different and Most Similar
  Systems Design.
• Advantages of Most Different and Most Similar
  Systems Design.
• Next?

4
Main Points from Last Weeks Lecture Goals of
Different Methods.

• All methods share the same goals, they attempt to
  solve the same problems in different ways
• We want to demonstrate general theoretical
  propositions by
• (1) establishing a general relationship among the
  variables,
• (2) controlling for the effects of other factors
  that may distort the analysis.
• More statistically oriented and more
  qualitatively oriented comparative research
  designs try to approximate the ideal control of
  exogenous variables accomplished by the
  experimental method.

5
Main Points from Last Weeks Lecture Small N
Analysis (1)

• Advantages
• Depth of knowledge
• Extensive dialogue between data and theory.
• Ability for theory corroboration or theory
  rejection?
• Good for theory building?
• Causality and process tracing.
• The need for thick description winking versus
  twitching (Geertz, The Interpretation of
  Cultures).
• Some topics can only be analyzed with small N
  analysis (example everyday resistance, James
  Scotts Weapons of the Weak).

6
Main Points from Last Weeks Lecture Small N
Analysis (2)

• Disadvantages
• Clear theoretical goals?
• Systematic comparison?
• Theoretical generality?
• Systematic analysis?
• Public procedure, easy (or possible)
  reproduction?
• The dialogue between ideas and evidence can
  distort the evidence.
• Selection bias.

7
Initial Thoughts on Theory-Driven Small N
Analysis

• Przeworski and Teune, 1970, The Logic of
  Comparative Social Inquiry
• The role of comparative research in the process
  of theory-building and theory-testing consists of
  replacing proper names of social systems with the
  relevant variables.
• Why is this important?
• An example not from Political Science
• Boiling water in Oxford, boiling water in Madrid
  Different temperatures.
• Conclusion Oxford and Madrid are different? The
  relevant explanatory variable is location?
• Replacing proper names with the relevant
  variables atmospheric pressure.
• Examples from Political Science?

8
Challenges to Theory-Driven Small N Analysis (1)

• Usually the problem is not finding variables
  correlated to the outcome (there are often many).
• The problem is knowing which variables are
  theoretically and empirically important.
• There are two general approaches
  variable-oriented and case-oriented.

9
Challenges to Theory-Driven Small N Analysis (2)

• Variable-oriented approach
• From last week nomothetic approach.
• Przeworski and Teune. A reaction to
  historicist approaches different countries
  must be studied as a whole and therefore
• Comparison is impossible apples and oranges.
• Sartori apples and oranges are only incomparable
  until we have the concept of a fruit then
  they are.
• Variable-oriented approach
• From last week idiographic approach.
• A reaction to Przeworski and Teunes
  reductionist approach.
• Cases should be preserved as meaningful wholes
  and complex and unique sociohistorical
  configurations (Skocpol).
• Are these two approaches contradictory?

10
Challenges to Theory-Driven Small N Analysis (3)

• If really theory-driven, both variable-oriented
  and case-oriented approaches employ the same
  underlying logic.
• If really theory-driven, all research is
  variable-oriented (even when using case studies)
  comparing means choosing properties or
  characteristics (in other words, variables).
• The controversy is really about whether to
  analyze few cases or many cases.
• The research design should be dictated by the
  question we are interested in and whether the
  focus is on complexity (small N) or generality
  (large N).

11
Choosing Cases in Theory-Driven Small N Analysis

• The nature of Small N comparison
• Comparing means choosing variables are cases
  comparable with respect to which properties or
  characteristics? or incomparable with respect to
  which properties or characteristics?
• Our cases will be similar in some respects and
  different in others.
• If we could manipulate variables at will, we
  would do experimental method.
• We cannot, so we try to take advantage of the
  similarities and differences we see in nature.
• Two main approaches most-similar and
  most-different designs.

12
Mills Methods for Small N Analysis (1)

• The Method of Agreement
• If, within the systems we are comparing, the
  phenomenon we are interested in explaining have
  only one of several possible causal circumstances
  in common, then the circumstance in which all the
  instances agree is the cause of the phenomenon.
• Graphically

13
(No Transcript)
14
Mills Methods for Small N Analysis (2)

• The Method of Difference
• If, within the systems we are comparing, there is
  an occurrence and a non-occurrence of the
  phenomenon, and the circumstances in which these
  are observed are the same in all factors save
  one, then that one is the cause of the
  occurrence.
• Graphically

15
(No Transcript)
16
Most Similar Systems Design

• Semantic issue
• This is in fact Mills Method of Difference.
• Also known as most similar systems design
  (Przeworski and Teune) or comparable cases
  strategy (Lijphart 1971 and 1975).
• Example
• We are interested in the determinants of suicide
  rates.
• We look at the Scandinavian countries they share
  many economic, cultural and political
  characteristics.
• We identify one crucial difference as explaining
  the different outcomes (high versus low suicide
  rates).

17
Most Different Systems Design

• Semantic issue
• This is in fact Mills Method of Agreement.
• Przeworski and Teune believe the logic of the
  most different systems design to be clearly
  superior.
• Most similar systems design unrealistically
  assumes that characteristics of a group of
  systems (such as the Nordic countries) can be
  removed individually in a semi-experimental way.
  Most different systems design eliminates factors
  as a package (system and conditional causality
  advantages).
• Famous example
• Skocpols States and Social Revolutions.
• Historical analysis of the revolutions in France,
  Russia and China.
• Skocpols question what was common among these
  different systems to produce political events
  that were essentially similar?

18
Problems of Most Different and Most Similar
Systems Design (1)

• General issues
• All potential causal factors need to be
  identified and included in the analysis.
• Generality problem unknown representativeness of
  the cases chosen.
• The dichotomous nature of variables mean a loss
  of information.
• Problems with multiple causation (even
  interaction effects are difficult to measure).
• Absence of probabilistic assessment.
• Number of causes and number of cases must be
  small (or method becomes unmanageable).
• Causal connection?
• I will not emphasize (1) but will analyze the
  rest in more detail.

19
Problems of Most Different and Most Similar
Systems Design (2)

• Generality problem unknown representativeness of
  the cases chosen.
• Lijphart many variables, small number of
  cases. Example how representative is Skocpols
  analysis of France, Russia and China?
• The small N is associated with selection bias
  (King, Keohane, and Verba 1994 and Collier 1995).
  Example criticism of Skocpol in Geddes (1991).
• The dichotomous nature of variables means a loss
  of information
• It virtually eliminates the possibility of
  analyzing anything but the limited phenomena that
  can be defined in terms of the existence or
  inexistence of a quality. Example Skocpols
  revolutions (but how about degree of
  international threat, the power of landed
  classes, etc?)
• How about growth, inequality, etc?

20
Problems of Most Different and Most Similar
Systems Design (3)

• Problems with multiple causation
• It cannot seriously consider multiple causation
  (either A C or B D cause E).
• Absence of probabilistic assessment
• Not knowing the frequency of a particular
  combination of causes and outcomes can give the
  same analytical weight to extremely unlikely
  events. If the goal is to discover theoretically
  relevant patterns, the Millian disregard for the
  probability of the factors seems
  counter-intuitive.
• Number of causes and number of cases must be
  small
• Ragin (1987) Mills method is extremely
  complicated even with an only slightly large
  number of cases (the number of combinations for
  causal conditions gets out of hand very fast).
• Causal connection?
• Mills method only address correlation. The
  historical analysis can resolve this (like in
  Skocpol), but causation is left to the
  case-study. Mills techniques do not really help.

21
Advantages of Most Different and Most Similar
Systems Design

• Qualitative advantages
• Depth of knowledge
• Extensive dialogue between data and theory.
• Ability for theory corroboration or theory
  rejection?
• Good for theory building?
• Causality and process tracing.
• The need for thick description winking versus
  twitching (Geertz, The Interpretation of
  Cultures).
• Some topics can only be analyzed with small N
  analysis (example everyday resistance, James
  Scotts Weapons of the Weak).
• Conceptual advantage once you have decided that
  a qualitative analysis is the best way to address
  your theoretical claims, how do you think about
  your design?

22
Next?

• Week 7 Boolean method. Historical development of
  comparative methods. More problems of
  comparative research design.
• Week 8 New approaches. Triangulation of
  methods. Summary.