Block 1: Introduction to X-centered research designs (Lesson 1) Flashcards
Method
Specific procedure for gathering and analyzing data
Methodology
Tasks, strategies and criteria governing scientific inquiry
Research design
Plan for carrying out a given study, commonly involving a sequence of research steps
X-centred methodology
Population / variable oriented research Main book: Gerring Main article: Ross (2006): “Is democracy good for the poor?”
Y-centred methodology
Case oriented research Main book: Blatter & Haverland Main article: Beland & Hacker (2004): “Ideas, private institutions and American welfare state ‘exceptionalism’: the case of health and old-age insurance, 1915–1965
Between X and Y-centered methodologies
Interpretive (critical) research Main book: Schwartz-Shea & Yanow Main article: Best (2010): “The IMF’s Constructivist Strategy in Critical Perspective”
What is the difference between the social and natural sciences?
Focus of social sciences is mankind, not as part of nature, but as creator and product of history
What is the difference between social science and the humanities?
Social science studies human actions in a systematic and falsifiable way
Experiments
Focus on the treatment effect (x-centered/ variable) Extrapolate to a bigger polulation than the analyzed sampe (population-oriented reserach, not case, not interpretive)
Issues with experiments
How representative is the sample of the overall population? Was the correct methodology used?
Gerring’s conceptualization of a case study
Study of a single unit for the purpose of understanding a class of similar units (Gerring’s conceptualization).
This is actually a sample!
Definition of a unit (case study)
A spatially limited phenomenon (e.g. nation- state) observed at a single point in time or over some limited time period.
Population in a case study, definition and what does it consist of?
The empirical boundaries of the set of cases relevant to an investigation A population is comprised of a sample as well as unstudied cases
Sample in a case study
The sample is all the studied cases
What does a case consist of?
Variables (relevant dimensions), built on observation(s)
Definition of a variable
Unidimensional factor that can have various values
Data cell (table)
[Image]
Variable (table)
[Image]
Observation (table)
[Image]
Case (table)
[Image]
Case study/ sample (table)
[Image]
Comparative study sample (table)
[Image]
Population (table)
[Image]
Single-outcome study
Case which does not belong to a larger category of cases Note: Case selection is NOT a problem Example: The French Revolution
Aim of a single-outcome (case) study
Investigate a bounded unit in an attempt to explain a single outcome within that unit
Critique of single-outcome (case) studies
Population-oriented researchers say you cannot learn much from such studies
Why is the choice of population important in population-oriented research?
Due to missing negative cases (cases which don’t show the outcome of interest)
What does the research question determine (in population-oriented research)
Size of population and which variables/ observations are included
Scope conditions
Criteria specifying the appropriate range of cases based on theory OR scope conditions = define the population Based on a condition/ variable, BUT can also be both a causal effect (causal condition)
Causal condition
Causal effect put on the population
Should a defined population be changed?
King, Keohane & Verba: No, too easy to omit unfavorable cases Ragin, Yes, but only with very good reasons
Ragin’s reasons for changing a defined population
Unrecognized heterogeneity = There are sub-groups in the population Inflexibility = Never changing the population means that you assume “causal homogeneity” (same causal relationship for all cases) without checking for appropriateness
Why is case selection important in x-centered research?
Affects the answers you get
Main problems of case selection (x-centered)
1) Defining the population (stretching vs relevance) 2) How to select the case which is worth analyzing
What is the goal of sampling? (x-centered)
Make a representative sample for the population so that you can make an (unbiased) inference from the sample to the population
Definition of sampling bias
Distortion in the representativeness of the sample, where some members of the population have a lower chance of being selected for inclusion in the sample
Main distinction in sampling strategies
Random and non-random selection
Random selection sampling strategies
SSSM
- Simple random sample: Complete randomization
- Systematic sample: First pick randomly and then follow a selection rule
- Stratified random sampling: Independent sampling in each level
- Multi-stage cluster sampling: Hierarchical structure of data, with sampling at each level
Non-random selection sampling strategies
Convenient Snow Quotas
- Convenience sampling
- Snowball sampling (ask interviewees to create a contact or to forward survey)
- Quota sampling (interviewer needs to satisfy quotas, determined by research design)
Why do we use random sampling?
To ensure that the sample average represents the population
Strategies to use if there are too few cases?
- Select all cases (sample = population) - Convenience sampling (e.g. data availability) - Deliberate (theory-driven) selection
Deliberate (theory-driven) selection methods
“PC Teddy” PC TEDDI
pathway,
crucial (least likely and most likely),
typical,
extreme,
deviant,
diverse,
influential
Typical selection method
- Goal:* How want to study the mechanism linking X to Y / hypothesis testing
- Case selection*: A low-residual case (on-lier)
- Desired outcome:* Your cross-case argument can be observed also within-case
Diverse cases selection method
- Goal:* Analysis of variation of X and Y (hypothesis generating)
- Case selection:* Combine values of X and Y
- Desired outcome*: Improve understanding of how X and Y are connected
Extreme case selection method
- Goal:* (Explorative) study by looking at extreme examples
- Case selection:* (Very) high scores on X and Y
- Desired outcome:* Hypothesis generation
Deviant case selection method
- Goal:* Identify the missing element of a valid X-Y relationship
- Case selection:* Find an outlier
- Desired outcome:* Complements the X-Y relationship and does not disprove it
Influential case selection method
- Goal:* Check robustness of X-Y relationship by studying cases that heavily influence the relationship (shifts it)
- Case selection*: Statistical
- Desired outcome*: X-Y relationship holds
Least likely case selection method
- Goal:* You want to show that a theory is correct
- Case selection:* You pick the case where your theory is least likely to hold
- Desired outcome*: Your theory holds even in the least likely case
Most likely case selection method
- Goal:* You want to show that an (established) theory is wrong
- Case selection:* You pick the case for which the theory is most likely to hold
- Desired outcome*: The theory does not even work in the most likely case
Pathway case selection method
- Goal:* You know that X causes Y, but you are not sure about the causal mechanism linking them together
- Case selection:* You pick a case in which X and Y are present, but all other potential causes of Y are absent
- Desired outcome:* You can find the mechanism linking X to Y
On what do you base selection on for necessary conditions?
The dependent variable (y)
When using deliberate (theory-driven) selection, what do you base the selection on?
The independent variable (x), unless you have used random sampling
Describe Gerring’s criterial approach
Based on hypothetical ideal types and not real world features. Stress certain elements to further understanding. Ideational constructs that help us better understand the world
Implications of Gerring’s criterial approach
Criteria are true, under ceteris paribus conditions Tradoffs are necessary Experimental template is the ideal
Goals of causal inference in social science
- CAD*
- Causality: To say that a factor, X, is a cause of an outcome, Y
- Appraisal: Is it falsifiable?
- Discovery: Is it new?
Gerring’s criteria for a good argument (general, NOT causal)
BCCG PPRT
Boundedness: Scope conditions?
Coherence: How consistence is it?
Commensurability: How well does it cumulate with other theories?
Generality: How broad is its scope?
Parsimony: How concise is it? Number of assumptions?
Precise: Is it specific?
Relevance: Everyday importance?
Truth: Is it true?
Gerring’s criteria for a good analysis
CAST Cumulation Accuracy Sampling Theoretical fit
Criteria for a good analysis: Accuracy
Are the results (a) valid, (b) Reliable and ( c) accompanied by an estimate of uncertainty with respect to (d) internal validity and (e) external validity
Criteria for a good analysis: Sampling
Are the chosen observations (a) representative of the population, (b) sufficiently large in number and (c) at the principal level of analysis?
Criteria for a good analysis: Cumulation
(a) Is the research design standardized with other research? (b) Does it replicate current findings? (c) Is it transparent?
Criteria for a good analysis: Theoretical fit
(a) Does the research design provide an appropriate test for the inference (construct validity)? (b) Is the test hard (severity)? (c) Is the test segregated from the argument under investigation (partition)?
