GOVT2991 Comparative Method

Crucial case study

"least likely case," find a case where a thoery is least likely to be confirmed and observe that it is confirmed

Case selection strategies

1. Diagnostic/Interpretive 2. Hypothesis-generating 3. Deviant/Outlier 4. Critical/Crucial

Van Evera's tests for the value of evidence

1. Straw-in-the-wind 2. Hoop 3. Smoking Gun 4. Doubly Decisive

characteristics and comparative politics

1. unites case orientation with variable orientation 2. detailed knowledge of cases important for analysis and research goal 3. studies cases as configurations 4. analytic frame not as flexible as in case study but more flexible that quantitative

Hypothesis-generating case study

Associated with grounded theory and ethnographic methods, and aim to establish new general theories to be tested in other cases

why is comparison important for case studies?

Case studies and concepts exist within wider universe of cases and concepts and case studies are often interested in within-case analyses aimed at indetifying intevrening vairables and establishing causal mechanisms

example of MoA

Compare three countries that have not suffered from acts of terrorism and ask whether they also had low net immigration, were religiously fragmented, and whether security services share intelligence. The latter condition is common however prevalence of this condition makes it a potentially trivial cause.

example of 1960s comparative work

Lipset's Political Man employed aspects of cross-case comparison to understand the relationships between economic development and democratic stability

Example of application of MoA and MoD

Skocpol's States and Social Revolutions, which explains that in the administrative collapse in the face of war losses and peasant revolts successful revolutions occurred in France, China and Russia

methodology (general)

WAY (how, not what) of thinking about methods, principles and criteria of scientific research

Ragin reasoning for comparison

a deliberate choice in order to study patterns of diversity and unravel different causal conditions connected to different outcomes

variable-oriented research is characterised by

a focus on identifying correlations between independent and dependent variables. Large-N statistical analysis of this sort applies tests across cases in order to assess the probability with which state reflects the hypothesised relationship while controlling for other variables.

benefits of MSSD

ability to use purposive sampling to mimic experimental method

fuzzy sets

allow for partial membership between 1 and 0

advantage of truth tables

allows for systematic comparison of empirical cases and transparent evaluation of counterfactuals

potnetial choice in political research

ask questions which can be meaningfully answered using large-N comparative research methods or interesting small-N questions but answer in inferior way

process-tracing

attempt to provide a framework for identifying causal mechanisms within cases

limitations of MoA

biased sampling due to selecting on the dependent variable.

Diagnostic case study

chosen with an interest in explaining a particular case in light of existing theories

Doubly Decisive

combines hoop and smoking gun to provide evidence necessary and sufficient for the confirmation of a hypothesis

Joint method of agreement and differences

combines logic of comparing across cases that agree in order to then indirectly apply the method of comparing cases that disagree on the outcome.

example of MoD

compare two cases to answer what causes democratisation along variables of economic development, British colonial heritage and abundance of natural resources exports. In country D absence of natural resources is accompanied by democracy while the opposite is true from country E

Method of Difference

comparing instances where a phenomenon does occur with one where it does not

Hoop

concern evidence is necessary for hypothesis to be true

method of process-tracing

consider a set of alternative explanations of an event or sequence fo events and ask each explanation: if this particular explanation were true, what evidence would we expect to see?

types of truth table

crisp and fuzzy

Most different systems design

deliberately selecting cases that are most different in as many important respects as possible for explaining an outcome yet agree on one factor as well as the outcome

problems with experiment method

difficult to apply to politics

MoA, MoD and joint method are methods of

elimination

two types of methods

experiment and non experiment

in Comparative Politics and the Comparative Method, Lijphart distinguishes that comparative method from

experiment and statsical methods

benefit of MDSD

extreme difference provides greater grounds for claiming that the common factor is causal than if applying same procedure to similar cases

Mills joint method

first identifies common factors across cases where the outcome was present than check whether those factors were not present in negative cases.

methodology (fundamentally)

general exploration of how scientific knowledge can be developed and obtained

crisp sets

have dichotomised presence or absence

Smoking gun

high uniqueness and provides strong evidence for hypothesis. Evidence is unlikely to be attributable to any other hypothesis

Method of agreement

identifying only cases where an outcome is observed (or not) and identifying the one conditions (or set) that agrees in the presence (or absence) among all cases of a phenomena

Ragin argues that social scientists ask

if economic development is a necessary or sufficient condition for democratisation instead of whether economic development increases the likelihood of democratisation

more or less of x leads to more or less of y

increase/decrease question that assumes a symmetrical relationship of covariation across all cases and can be tested by regressing proxy variables and basing inference on linear algebra

1950 and 1960s saw

increasign numdner of states which state out the compare two or more states, regions or individuals

John Stuart Mills developed

inductive methods of experimental inquiry which have become synonymous with small-N comparative research in social science

problems with case study

lack of contextualisation

problems with statistical method

lack of thick description, sequences and interaction

types of non-experiment methods

large-N statistical, single N case study, small-N comparative method

Straw-in-the-wind

low in uniqueness and uncertainty, evidence that passes the test for a hypothesis does not really distinguish its credentials from other hypotheses, nor would it be completely necessary to confirm a hypothesis

what technique had profound impact on discipline, allowing for more controlled comparsions and inferences from relatively large samples of cases to the population level

mass survey

according to Lijphart, the comparative method is

no equivalent to the experiment method but only a very imperfect substitute

comparative method

one of the four fundamental methods which can be used to test the validity of general empirical propositions which balances focus on cases and variables

benefits of fuzzy sets

partial membership can reduce problem of contradictory truth table rows and allow for richer analysis of variation in presence/absence conditions across cases

comparativists working with small-N need to

provide clear justifications that they are not merely selecting cases to suit a priori argument

Qualitative Comparative Analysis

provides an integrated method for causal interpretation of set theoretical relationships

Deviant case study

researcher deliberately selects an outlier case that is known to contradict the typical cases which otherwise provide evidence for the veracity of a theory

most similar systems design

select cases that are as similar as possible in important factors to predict the outcome but that differ on the outcome, in an attempt to mimic experimental control and reduce extraneous variance

small-N comparisons are limited by

sensitivity to the way an investigation is specified and cases selected.

criticism of MoD

small sample size

size of comparative reseach

small-N

advantages of small-N research

smaller number of cases allows researcher to take time to engage with a number of ancillary investigations to try and answer questions of potential error

does a combination of x and y always produce z/ is x required for y

sufficiency/necessity question allowing for asymmetry in relationships among condtions and proses that cases will and will not be found in particular subset/superset relations

MoD seeks to identify

the variable that is present in the positive case and absent from the negative case and vice versa

Ragin's The Comparative Method argues that

there are at least 2 broad apporaches to social science: variable oriented and case-based

why compare?

to answer aproblem when we don't have enough cases and several variables

Case-based research is said to

treat cases more holistically as conjunctions of conditions. researchers tack back and forth between theory and evidence to try to understand and describe important invariant relationships in a smaller set of carefully selected cases

QCA limitations

unreliable and sensitive to case selection and specification, also static nature of analysis

truth table

visualises combinations of properties that observed cases take, mapped on those combinations that they could logically take.

we use the comparative method when

we cannot perform an experiment because we cannot control an intervention and we cannot make use of inferential statistics as we do not have sufficient sample of cases

example of joint method

when asking what explains high or low voter turnout across countries, there are no two cases that differ on outcome and only one explanatory conditions. So first apply MoA to pairs that agree on outcome in order to eliminate irrelevant conditions, leaving compulsory voting as remaining potential variable

example of truth table

when investigating conditions required for higher voter turnout, we are interested in the conjunction of high campaign spending, proportional rep, and compulsory voting. Table will have 8 rows, one for each possible combination of the 3 variables.

dimensions of theory confirmation

whether evidence can be uniquely predicted by a particular hypothesis and the certitude with which observed evidence is necessary fo hypothesis to be true