Lecture 1 Flashcards
a tale of two cultures
1994
= heavy read (book: designing social inquiry)
focus on examples + clarification
no focus on the maths etc.
applies quantitative methods (linear regression analysis) guidelines to qualitative research -> controversial
became canonical (western mainstream) political science for a good 2 decades
traditions
a tale of two cultures
authors compare scholars through a ‘naturalistic’ approach, a sort of ‘anthropology of political science’
instead of trying to define hard liens of what’s scientific, they give a useful characterization of two ideal types of a researcher. these types are identified through two sets of values, beliefs and practical norms (cultures or traditions) of political research
!ideal types are analytical distinctions, they are hardly observed in pure forms (mixed methods are the rule, bc they have the same objectives)
quali and quanti are not differentiated by:
*these criteria are ambiguous, all these an be combined into the same project
- types of data (quali also uses statistics)
- sample size (small N vs large N)
quali and quanti are differentiated by:
- immediate goals
!ultimate goal is the same: to produce valid descriptive and causal inferences (there is a diff between positivists and interpretivists -> 3 cultures) - norms of research practices/procedures
a tale of two cultures
ideal types
10 criteria to identify the quali vs quanti ideal-type researchers realted to their goals and practices (see table in slides)
!focus on two ideal types of positivist research: scholars that care about causality
a tale of two cultures
approaches to explanation
quali = causes of effects
- explain specific set of outcomes in particular set of cases
- move backward to find the causes of these particular cases
e.g. why did India become a democracy and Pakistan an autocracy after independence despite similar dev. levels?
-> criticism: why explain just two cases?
bc it shows that not all regime transitions are the same, it shows that context matters, same outcome may happen through different causal paths
quanti = effects of causes
- estimate average effects (holding other things constant): direction, sign, magnitude
- important that there are many individual cases for the purpose of estimation
- can’t explain and does not aim to explain specific set of outcomes. only cares about whether a factor has causal meaning
e.g. does development cause democratization
-> critcism: can you explain why India has been a democracy since independence, having been a low-income country for decades and having so much population in poverty?
answer: GDP per capita levels are important factor to make democracy possible, however it is not the only factor
a tale of two cultures
conceptions of causation
quali =
- necessary / sufficient causes (sometimes only implicitly depending on the author)
*you don’t see this in quanti bc many diff causes - the “why” question (causal mechanisms, why does this cause this)
- research project acknowledges and aims to recognize multiple causal paths to the same outcome (equifinality) depends on context
equifinality = there are multiple causal paths to the same outcome
e.g. explain why India became demcoratic but not Pakistan given similar levels of dev. it was diff than the western path, where dev. was important (aristocracy lost power, labor unions etc)
e.g. David Waldner’s comparison between Syria/Turkey vs Korea/Taiwan = path dependency argument
situation of elite cohesion and mass-excluding political system (Syria/Turkey had conflict) is necessary to explain why Korea/Taiwan became “developmental states” (state also nurtured other industries) but not Syria/Turkey (they did industrialize, became manufacturers)
-> it doesn’t make sense to talk in probablistic terms: even if onebelieve that the world is probablistic and not deterministic. we are talking about one single event
-> does it matter to compare only a few cases?
- qualis: yes, bc equifinality (context matters)
quanti =
- correlational causation
- average effects
- counterfactual model of causation (if you could go back in time and change the cause, the outcome would have looked diff)
- research project does not concern whether this or that is the cause of something, only whether a given factor matters
e.g. as GPD per capita increases by this amount, the likelihood that a country experiences democratization raises by 10%. counterfactual formulation: the same country is more likely to be a democracy with greater eco growth
a tale of two cultures
multivariate explanations
when scholars in each tradition say something like “there are multiple variables involved in explaining this phenomenon”, they mean very different things
- the authors demonstrate how these two meanings are hardly compatible since they have different underlying mathematical logics
quali = variables combine to form explanations, these explanations may be multiple paths to the same outcome
quanti = variables combine to modify the direction and magnitude of effects, effects accumulate
- no diff explanation going on, its jus that diff effect accumulate
- always includes an error term (unobserved variables)
e.g. Political consequences of assassination
- leader assassination increases likelihood of turmoil , this doubles if succession is not institutionalized (interaction effect)
is not the same as 4 neighboring cases that share common histories, assassination as sufficient but not necessary term of turmoil
- turmoil = assassination * non-inst. succession
- Turmoil = assassination * oil
- no turmoil = no assassination * non-inst. succession
- turmoil = non-inst. succession * oil
a tale of two cultures
equifinality
= there are multiple causal paths to the same outcome
multiple explanations for the same outcome
discussing it is not the objective of quanti-oriented projects (the IV matters for DV, i don’t know what other variables matter for the DV)
this concept is summarized by saying “context matters”: A and B can lead to outcome 1 in multiple contexts, however in some contexts it won’t lead to outcome 2 bc there are other factors as well
with a small N the nr of causal paths is finite, you can point at them
-> doesn’t make sense to talk about probability, you want to find out THE factor for a specific (set of) case(s)
-> quali delivers an explanation for a small phenomenon, uses comparable language for knowledge accumulation with the hope that maybe one day it will add up to a law (but it’s impossible to know now)
a tale of two cultures
scope and generalization
quali = narrow scope
-> favor contextualized explanations that specify when and where a variable is connected to an outcome
- explain the outcomes in a small set of cases (still, generalizable to arguably similar cases)
- assumption: expanding the sample requires modifying explanation
- takes seriously the problem of causal heterogeneity (same IV can lead to diff DV in different contexts)
quanti = broader scope, thinner explanation
-> often focus on just one variable
- explain changes in the outcomes of large sets of cases with the ultimate purpose of understanding populations
- assumption: expanding the case sample will not require to change the explanation much
a tale of two cultures
case selection practices
quali = purposeful selection on the DV
no variation(e.g. one or two cases)
- makes sense with immediate goals: explaining rare events, analsis of causal mechanisms within cases, explaining outcomes in a contextually bounded set of events/cases
- no variation, but comparative perspective: is it a deviant, crucial or representative case?
- it does not the immediate goal to generalize findings?
quanti = random selection on variables
- experimental golden standard: cases are comparable (he’s gonna rewrite the slide bc makes no sense)
- with observational data (e.g. cross-ational dataset) not possible
a tale of two cultures
weighting observations
quali = not all pieces of data count equally, theories are “one critical observation away from being falsified”
- like detective work: given what i know how much this pieces of evidence affect my beliefs?
- uses “causal process” observations: often come from diff sources, can’t be studied within a single sample and method (e.g. interviews, photos, speeches, statistics all in one research project)
- with historical research, sources are often incomplete and spotty -> needs to be supplemented
quanti = all observations weighted equally, one observation can’t lend decisive support or critically undermine a theory
- uses “dataset” observations: entries in a spreadsheet, each entry is a case, use estimate average effects
a tale of two cultures
substantively important cases
quali = prior theoretical knowledge often makes certain cases especially interesting and theoretically important
- theories can get big cases right and worry when they don’t
- e.g. worry when we can’t explain India by looking at correlation between Development and Democracy
e.g.
- realism in IR should be able to explain the end of the Cold War, given the Cold War was a special source of inspiration
- ??Marx’s historical materialism should be able to explain the French revo, but it did not??
a tale of two cultures
lack of fit
quali = case that does not fit is not ignored as an outlier in a distribution
- non-fitting caes help to specify scope conditions and to refine theories
quanti =
a tale of two cultures
concept and measurement
quali = eliminating measurement error by devising clear and context-sensitive concepts
quanti = focus on indicators and measurement validity
- central concern = modelling measurement error and avoiding systematic error (biased indicators)
quali: want to apply concepts to the right kind of cases, concepts should not be stretched
Florida’s Panhandle case example (reading)
2000 George W. Bush won Florida (over Al Gore) by 537 votes
Florida gave Bush the presidential win in the Electoral College
thus: Bush’s …..
Brady’s analysis:
said Lott was wrong:
chapter shows: methodologies can complement each other (Lott’s data was incomplete + process tracing can undermine it)
quali and quanti diff yet compatible
quali = it’s own thing
not a measurement of last resort when there is no large-N data
it is not a starting point for quant research, although it may be (and vice versa)
requires diff skills and training
STILL: compatible in many ways: contribute to same debates with different aspects + can co-author (often now)
now: respect for other traditions
qualitative research can help understand broader populations by clarifying explanations
quali methods pair with a variety of theory styles e.g. formal theory (analytical narratives)
