Comparative Politics and the Comparative Method Comparative Perspectives Fall 2021 Marek Rybář, PhD. An overview of today‘s lecture –Comparison and its goals –Differences between social and natural sciences and the role of comparison –Methods of comparison –Pitfalls and problems of comparison –Strategies of comparison – Comparison and its goals –Comparing is a natural human activity –Prices of cellphones, courses at college, job offers, income, etc. –What is the difference between such everyday comparison and scientific comparison? Q&A – –Why do we compare in comparative politics? –What comparisons have you already carried out? Comparison and its goals –The two differ in their goals: comparison of states, political systems, regimes etc. has these four basic goals: –description –classification –testing of hypotheses –prediction Description 1/2 –A systematic scientific exploration of a subject needs a good description of the phenomena under investigation –Description of political phenomena and events in one or several countries –Sometimes referred to as “old/traditional” comparison, in contrast to more scientific “new comparison” –Almond: "evidence without inference“ –Lijphart: atheoretical case study Description 2/2 –The author describes a considerable and interesting „story“ without more general inferences and generalizations –Specific events, important personalities who played a role in decision-making etc. –Potentially important information, data for case studies and comparisons –General political phenomena (e.g., the emergence of social movements, military dictatorships etc.) Classification 1/2 –Helps categorize (classify) cases into several groups on the basis of a few similar features –Simple dichotomy (democracy vs. non-democracy) as well as more complex schemes (1 party, 2 parties, several parties) –Classifications simplify the real world and outline differences among classes --> a basis for comparative inquiry Classification 2/2 –Inductive and deductive reasoning: Blondel vs Aristotle –Blondel: one, two, two and a half, multiparty with a dominant party, multiparty without a dominant party –Aristotle: number of rulers and the character of their government –One, several, many // good, bad –Typology: monarchy, aristocracy, politeia, tyranny, oligarchy, democracy Hypotheses Testing –Comparisons help to assess several competing explanations and to eliminate those that are not supported by the evidence: –1. Identify the key variables –Specify the relations among them –When comparing empirical evidence, we generate hypotheses about the relations between variables that are subsequently tested on several/many cases Predictions 1/3 –A logical extension of testing –Predictions about development in the cases that were not included in the original set of cases –Predictions in comparative analysis are probabilistic, [ceteris paribus], e.g.: –Incumbents are more likely to be re-elected than their challengers Predictions 2/3 –OR: countries that use the PR electoral systems are more likely to have more relevant political parties than countries with a single member plurality electoral system –We can thus predict the effects of electoral system change from plurality to PR –HOWEVER: It does not mean we can predict the results in a specific country Predictions 3/3 –Prediction are less common in comparative politics than a few decades ago –A well-know “recent” prediction is Huntington’s assertion that conflicts are most likely to take place along civilizational “borders” –Huntington believed his prediction was more accurate than any other competing explanation Differences between social and natural sciences 1/2 –The four goals of comparative politics (description, classification, testing of hypotheses and prediction) are also shared by natural sciences –Newton’s gravitation theory was originally formulated on the basis of empirical evidence that led to generalization and predictions –gravity (as well as other concepts) cannot be observed directly, we can only observe its consequences: it is an intellectual construct that was verified in repeated experiments; only after that a theory was formulated Differences between social and natural sciences 2/2 –Experiments are nearly impossible in comparative politics but are typical for most natural sciences –The importance of “counterfactuals”, i.e. thought experiments in which analysts imagine the absence of particular variables in their cases –i.e. they imagine an alternative course of events (one variable would be different) in the case under investigation –Democratic transition in Spain in 1975: parliamentarism vs. presidentialism Comparison instead of experiments –When we emphasize the importance of an explanatory variable, we always implicitly work with counterfactuals –To say that single member plurality electoral system tends to produce bipartism involves considering a counterfactual situation in which a country would not have a two-party system without single member plurality electoral system –In comparative analysis, we use a real-world case(s) to replace counterfactuals: comparison substitutes experiments Question – – – –Do you know any political science laws? Comparative Politics is not strong in producing “laws” –(However, there are some exceptions): –Duverger’s law –Michels’ Iron law of oligarchy –Democratic peace –Too few cases/too few observations –Instead of laws, CP produces understanding and explanation of phenomena about which we have “a lot” of observations and our level of certainty is considerably high How do we compare? –Case studies –Small-N comparisons –Large-N comparisons –Differences rest in the level of abstraction of our conclusions –The fewer cases we have, the less opportunity for generalizations Case studies 1/2 –What is comparative about single case studies? –We can work with concepts that can be used in other cases (contexts) –We can formulate conclusions about the more general aspects of our case –We can supply a good description of the relevant context –We can supply new classifications and generate hypotheses for subsequent comparative studies –We can support/reject theories or explain deviant cases Case studies 2/2 –When analyzing one case (e.g., one country) we can increase the number of observations –CASE is not OBSERVATION –Analyze several elections –Analyze several regions –Italy and the civic culture –India and the role of protestant missionaries in democratic development Small-N Comparisons (2 - 20) –We deliberately choose several cases from the entire population of cases –Search for similarities and differences –Contrasting similarities and differences can reveal possible explanations of our research puzzles Large-N Comparisons 1/2 –Closest to the logic of experimental methods of natural sciences –Advantages: ability to statistically control and eliminate alternative explanations –Covers cases/countries across space and time –Law-like generalizations Large-N Comparisons 2/2 –Risks and pitfalls : –Validity of measurement is questionable –Not suitable in analyzing processes where complex causal mechanisms are at play –Not suitable for analyzing phenomena whose meaning is strongly linked to local (i.e. unique) context Problems of comparison –1) Too few cases, too many variables –2) Questionable equivalence –3) Selection bias –4) Spuriousness –5) Ecological and individual fallacies – Too few cases, too many variables 1/4 –when there is more potential explanations than cases to test them –Possible solutions: –1) increase the number of cases or observations Too few cases, too many variables 2/4 –Lijphart (1970) suggests: –increase the number of cases geographical and temporal strategy –reduce the number of variables by merging some of them –reduce the number of variables by focusing on the relevant variables (guidance offered by an existing theory) – Too few cases, too many variables 3/4 –2) use the most similar systems design (MSSD) –eliminate the variables that are the same across cases and focus on those variables that are different and thus potentially cause the observed outcome –Unfortunately, when using the MSSD, we will never be able to eliminate many alternative explanations (variables) Too few cases, too many variables 4/4 –3) minimize the number of relevant variables by employing the most different systems design (MDSD) –We compare totally different cases with similar outcomes, focus is on the different variables across cases that potentially lead to the similar outcomes – – Equivalence –Different understanding of the key concepts may lead to different (non-comparable) ways of measurement –It is important to specify what the equivalent concepts could be –Concepts must be modified to take into account cultural specificity of each case –Best if applied to cases that are well-known to the researchers Selection bias –Comparison is a substitution for experiments, however, it is an imperfect substitution –Experiments select cases randomly, while in CP we choose among cases deliberately –The most visible selection bias emerges when we use only those cases that support our argument – Selection bias –Less visible selection bias exists when we choose cases on the dependent variable: –E.g. when we only work with cases with a particular outcome: where a revolution did take place –If there is no variation on the dependent variable, we may reach conclusions that overestimate the importance of some of our independent variables Spuriousness –Exists when we omit the key variable that influences both our dependent and independent variable –ice-cream and murder –There is no perfect solution to the problem! The most similar systems design (MSSD) – – –We identify the key characteristics that are different in otherwise similar cases –we thus expect that these different features lead to/explain the outcomes – CASE 1 CASE 2 VARIABLES a a b b c c X Non-X OUTCOMES Y Non-Y Variables Togo Ghana Similarities: Climate High Temperatures High Temperatures Per capita income Low Low Ethnicity Heterogeneous Heterogeneous Dominant Religion Christianity Christianity Other religions Islam, traditional tribal Islam, traditional tribal COLONIZING POWER France United Kingdom Outcome Regime Type Authoritarian Democratic The most different systems design –Cases that are totally different, have only a few shared similarities –They also share the same outcome – CASE 1 CASE 2 VARIABLES a d b f c m X X OUTCOMES Y Y France 1780-1790 China 1940-1945 Differences Geography Europe Asia Population < 30 mil. > 500 mil. Century 18. 20. Regime Monarchy One party state XXXXX X X Outcome Social Revolution yes yes