Introduction to Qualitative Comparative Analysis (QCA)

Transcription

1 Introduction to Qualitative Comparative Analysis (QCA) Vaidas Morkevičius Policy and Public Administration Institute Kaunas University of Technology 2012 November 11, Riga

2 Lecture 1 Basic introduction to 'macro'-comparative social research approaches An example comparing results of quantitative statistical and qualitative comparative research

3 Basic introduction Comparative analysis in social science is often applied at the 'macro'-level Political systems, societies, civilizations Different aspects are observed of the whole system At this level, the number of cases to be examined is of necessity limited For example, at most 200 states world wide

4 Basic introduction Comparative analysis in social science is often applied at the 'macro'-level Political systems, societies, civilizations Different aspects are observed of the whole system At this level, the number of cases to be examined is of necessity limited Usual situation not many useful cases on which sufficient materials are available

5 Basic introduction At the same time, these 'macro'-level systems exhibit a high level of complexity The classic many variables small N dilemma Notwithstanding, two dominant camps or schools now can be observed in 'macro'- social research: 'Macro-quantitative' 'Macro-qualitative'

6 Basic introduction Macro-quantitative methods: Behavioral revolution in social sciences in s: 'Micro' level research using individual survey data Preference for statistical analysis (as a result of a large number of cases) This 'scientistic' position dominated macro analyzes too macro-quantitative methods and comparative aggregate data analyzes

7 Basic introduction Macro-quantitative methods: In political science, comprehensive data handbooks have been compiled since the 1960 s, which together with official and unofficial (primarily economic) statistics formed the starting point for numerous macroquantitative analyzes. The largest possible number of cases (usually states) with comparable data usually formed the foundation of such studies

8 Basic introduction Macro-quantitative methods: Problems with this approach: Lack of information Frequent data problems Relatively limited level of basic similarities No random sampling and 'normal distributions' the basis for representative interpretations of the survey results on the micro level 'Inferential' statistics are, strictly speaking, out of the question... However, such data be used for descriptive analysis or serve as a basis for presumably universal explanations based on the actual number of cases

9 Basic introduction Macro-quantitative methods: Problems with this approach: Specific characteristics of individual cases not taken into account If they differ too much in the observed distributions dismissed and neglected as outliers Limited number of cases increases the possibility that including one or a few deviating cases can drastically change the results Certain economic deterministic bias Initial data were available from the UN organizations, the World Bank, the OECD, the national statistical offices etc.

10 Basic introduction Macro-quantitative methods: Problems with this approach: No doubt one reason why so much attention has been given to the relationship between regime and socio-economic level is simply that reasonably acceptable (if by no means wholly satisfactory) 'hard data' are available from which to construct indicators. This is a perfect example of how the availability of data may bias the emphasis of theory (Dahl 1971: 206)

11 Basic introduction Macro-quantitative methods: Problems with this approach: In contrast, differentiated socio-cultural or political data in a more specific sense are much more difficult to obtain and seldom collected on a regular basis Result misspecifications on the basis of limited data and prejudiced selection of variables (in light of the general requirement of quantitative social science that the choice of variables should be guided by careful theoretical analysis and specification of hypotheses)

12 Basic introduction Macro-quantitative methods: Problems with this approach: Causal relationships are 'probabilistic' usually based on correlations between a dependent and one or more independent variables Can be 'spurious' (may have occurred due to a third factor which has not been taken into account) Direction of a causal relationship is also not always clear (which came first? what depends on what?) Causality is 'universal' in nature (average of the observed totality) However, 'inferential' generalisations are not allowed since selection of cases is not representative

13 Basic introduction Macro-quantitative methods: Problems with this approach: Causal relationships are 'probabilistic' usually based on correlations between a dependent and one or more independent variables Can be 'spurious' (may have occurred due to a third factor which has not been taken into account) Direction of a causal relationship is also not always clear (which came first? what depends on what?) Causality is 'universal' in nature (average of the observed totality) 'Conjunctural' causalities (based on differing combinations of variables, Mill 1843) must also be discarded

14 Basic introduction Macro-quantitative methods: Problems with this approach: Little of long-term value to the social sciences has emerged from the hundreds of studies conducted during the last few decades that have run statistical analyses including most of the world s nation states (Tilly 1984: 76)

15 Basic introduction Macro-quantitative methods: Solutions: Traditional statistics is 'robust' enough? Small sample estimation procedures? New and more 'robust' statistics? Something 'different'? Macro-qualitative?

16 Basic introduction Macro-qualitative methods: Intensive expansion of 'macro-qualitative' and 'diversity-orientated' methods since 1970s: Systematic 'most different' and 'most similar' research designs (Przeworski & Teune 1970) Historically oriented studies (Skocpol 1979, 1984) New conceptual, methodological and technological developments resulting in introduction of Qualitative Comparative Analysis (see RAGIN 1987, 2000, 2008)

17 Basic introduction Macro-qualitative methods: Definition of 'macro-qualitative' analysis of presence or absence of characteristics specific to examined cases at the 'macro' level Should not be confused with qualitative methods at the micro-level (such as participant observations in ethnology) or with qualitative interpretative methods (for example in hermeneutics) Presented techniques rely on a dichotomization of the observed variables (yes/no, high/low, 0/1 etc.) More varied characteristics must be dichotomized with the use of certain thresholds dummy variables be formed

18 Basic introduction Macro-qualitative methods: Dichotomization (or other transformations 'simplifying' distributions) entails loss of information (also characteristic of numerous statistical methods) However, it allows the implementation of new and more complex methods on the basis of Boolean algebra, of set theory and elaborated 'similarity' and 'dissimilarity' levels, which represent a certain 'compensation' for the occurring information loss

19 Basic introduction Macro-qualitative methods: In contrast to statistical methods, macroqualitative analyzes are more strongly caseorientated Each case has in principle the same value for the analysis Selection of cases must be as hypothesis- and theory-guided as the selection of variables A minimum amount of homogeneity amongst the cases to be chosen (e.g. historical-regional similarities) must be ensured in order to analyze them meaningfully

20 Basic introduction Macro-qualitative methods: In contrast to statistical methods, macroqualitative analyzes are more strongly caseorientated Among the more limited number of cases selected in this way, a high level of heterogeneity not only with regard to the dependent variable but also for the possible independent variables is desirable The smaller and less studied countries or strongly 'deviating' cases can often supply interesting information relating to the validity and range of certain hypotheses

21 Basic introduction Macro-qualitative methods: In contrast to statistical methods, macroqualitative analyzes are more strongly caseorientated Most importantly, high level of familiarity with the cases is a prerequisite for every 'macroqualitatively' inclined social scientist in order to obtain the necessary sensibility for the often complex and historically determined facts

22 Basic introduction Macro-qualitative methods: In contrast to more universal-statistical explanations and exclusively historicalidiographical (individualising) case studies, macro-qualitative analyses can also reveal 'conjunctural' causal relationships (different patterns of factor combinations) The range of these patterns can be ascertained and in certain cases modified by a step by step expansion of the field of examination

23 Basic introduction Macro-qualitative methods: In contrast to more universal-statistical explanations and exclusively historicalidiographical (individualising) case studies, macro-qualitative analyses can also reveal 'conjunctural' causal relationships (different patterns of factor combinations) Qualitative Comparative Analysis (QCA) allows including hypothetically possible cases ('logical remainders') in the analysis and developing at least a hypothetical generalisation over and above the cases taken into account

24 Let's see how it works :-) Now classical example: Testing an empirical theory of democracy conditions favoring the emergence of democratic political systems Only socio-economic and structural factors Dirk Berg-Schlosser & Lasse Cronqvist (2005) The data set used is derived from an international research project dealing with the conditions of authoritarianism, fascism and democracy in Europe in the inter-war period (Berg-Schlosser & Mitchell 2000, 2002)

25 Let's see how it works :-) Now classical example: Theory Lipset s Political Man (1963): The more well-to-do a nation, the greater the chances that it will sustain democracy (p. 31) Some 'special' cases in the original analysis were identified, however, not explained: Germany's breakdown in interwar period due to 'a series of adverse historical events'? Czechoslovakia, Finland and France 'unstable' in interwar period, however, survived as democracies Later studies 'confirmed' socio-economic conditioning of democracy

26 Let's see how it works :-) Now classical example: Socio-economic factors: Gross national product per capita (ca. 1930) Urbanization (percentage of population in towns with and more inhabitants ca. 1930) Literacy (percentage of literate population ca. 1930) Industrialization (percentage of industrial labor force including mining ca. 1930) Dependent/outcome variable: survival of democracy See raw data file (Democracy.raw.4in.csv)

27 Let's see how it works :-) Now classical example: Quantitative analysis (p. 161):

28 Let's see how it works :-) Now classical example: Quantitative analysis (Norkus & Morkevicius):

29 Let's see how it works :-) Now classical example: Quantitative analysis (p. 162):

30 Let's see how it works :-) Now classical example: Quantitative analysis (p. 162): Over and above problematic with quantitative procedures is that we do not know which variables were responsible for the respective groupings and classifications of cases All in all with these procedures we identified one major factor (GNP p.c.), followed by literacy, and some problematic cases which cannot be explained so easily

31 Let's see how it works :-) Now classical example: Quantitative analysis (p. 162): For this reason, more 'diversity-oriented' macroqualitative procedures can, at least, meaningfully supplement such an analysis and point to specific constellations of cases and factors (specific 'conjunctural' causations), but may also help to develop more refined theories

32 Let's see how it works :-) Now classical example: Qualitative comparative analysis (QCA): QCA is a technique based on Boolean algebra, which can reduce variables related to a specific outcome to its minimal 'prime implicants' also allowing for possible 'conjunctural causations' Each case is described by including all independent variables in the data set and the respective outcome In a Boolean sense each variable has to be dichotomized, i.e. each variable is either T (True) or F (False)

33 Let's see how it works :-) Now classical example: Qualitative comparative analysis (QCA): Dichotomization employing 'thresholds' should not be purely mechanical consider the actual distribution of cases avoiding to set the threshold between some closely neighbouring cases In the original version of QCA True conditions are written with uppercase letters, while False conditions are written in lowercase In the next step, all cases are placed in a 'truth table'

34 Let's see how it works :-) Now classical example: Qualitative comparative analysis (QCA): In the truth table, cases with the same configuration of variable values are grouped together, and each configuration is assigned its QCA characteristic Cases with the same configuration and the same outcome value (0 or 1) have this outcome value as the QCA characteristic But if they have identical configurations of independent variables and different outcomes, then the QCA characteristic is set to C (contradiction).

35 Let's see how it works :-) Now classical example: Qualitative comparative analysis (QCA): The main step performed in QCA is logical minimization rule: If two Boolean expressions differ in only one causal condition yet produce the same outcome, then the causal condition that distinguishes the two expressions can be considered irrelevant and can be removed to create a simpler, combined expression (Ragin 1987: 93)

36 Let's see how it works :-) Now classical example: Qualitative comparative analysis (QCA): So, if there are two logical configurations in the truth table with the same outcome and they only differ in one variable, then this variable can be considered irrelevant for the outcome and does not have to be retained For example, ABC + ABc can be reduced to AB

37 Let's see how it works :-) Now classical example: Qualitative comparative analysis (QCA): Two steps in calculations of QCA: First, all possible minimization steps are carried out. In this part the prime implicants are calculated by executing all possible minimizations by pairwise comparisons of all possible configurations of factors when no more minimizations are possible, these logical expressions are considered to be 'prime implicants'

38 Let's see how it works :-) Now classical example: Qualitative comparative analysis (QCA): Two steps in calculations of QCA: In the second step, these prime implicants are combined into the shortest logical expression possible covering all the cases with the same outcome, where the length of the logical expression is given by the number of variables included in each implicant The solutions found by QCA are annotated by the use of Boolean AND (*) and Boolean OR (+).

39 Let's see how it works :-) Now classical example: Results of QCA: Dichotomization thresholds: Gross National Product per capita: 0 if below 600$; 1 if above Urbanization: 0 if below 50%; 1 if above Literacy: 0 if below 75%; 1 if above Industrialization: 0 if below 30% of active population; 1 if above See dichotomized file (Democracy.csQCA.4in.csv)

40 Let's see how it works :-) Now classical example: Results of QCA: Dichotomization thresholds: Gross National Product per capita: 0 if below 600$; 1 if above Urbanization: 0 if below 50%; 1 if above Literacy: 0 if below 75%; 1 if above Industrialization: 0 if below 30% of active population; 1 if above See dichotomized file (Democracy.csQCA.4in.csv)

41 Let's see how it works :-) Now classical example: Results of QCA: Resulting 'truth table' (p. 165):

42 Let's see how it works :-) Now classical example: Results of QCA: Lipset s positive conditions are fulfilled in a 'pure' sense in only three of our eighteen cases (Belgium, Great Britain, and the Netherlands). However, this is contradictory to the German case G * U * L * I Contradictory In a negative sense, Lipset s hypothesis is confirmed in the cases of Greece, Hungary, Italy, Poland, Portugal, Romania, and Spain g * u * l * i Breakdown of democracy

43 Let's see how it works :-) Now classical example: Results of QCA: Czechoslovakia is a (democratic) case which ranks high on all indicators, except for wealth which is somewhat below the threshold. Ireland, another 'survival' case, has relatively low values of urbanization and industrialization G*u*L*i (IRE) + g*u*l*i (CZE) Survival of democracy Other cases contradictory (but we know their configurations) We can 'solve' them and employ other 'nice' techniques for data analysis within the QCA

44 Thank you! :-)