Research Methods in Political Science Exam 2

Classical Randomized Experimental Design

A study where subjects are randomly assigned to either an experimental group (receives the treatment) or a control group (does not). This allows researchers to isolate the experimental effect—the difference in outcomes due to the treatment.

Randomization

Prevents selection bias.

Experimental group

Receives treatment (e.g., gets campaign flyer).

Control group

No treatment (no flyer).

Experimental effect

Difference in turnout between the two groups.

Post-test only

Measure after treatment only. ✅ Easier; ❌ No baseline to compare.

Repeated-measurement (pre/post)

Measure before and after. ✅ Measures change; ❌ May influence behavior (Hawthorne effect).

Multiple-group design

More than one treatment (e.g., compare email, flyer, phone call). ✅ Allows comparisons; ❌ More complex, needs more subjects.

Quasi-Experimental / Observational Studies

Used when randomization isn't feasible—common in political science.

Cross-sectional design

Data at one point in time. 🗳 Ex: National Election Survey in 2020. ✅ Fast, large sample. ❌ No causal inference or time effects.

Longitudinal design

Follows same subjects over time. 🗳 Ex: Panel study following voters across multiple elections. ✅ Can assess change and causality. ❌ Costly, attrition.

Case Studies

In-depth study of one case (e.g., Watergate).

Hypothesis-generating Case Studies

Explore new theories (e.g., studying a new democracy to generate ideas about consolidation).

Hypothesis-testing Case Studies

Apply a theory to a case (e.g., test modernization theory in Tunisia).

Plausibility probe

Early test of a theory before large study.

Counterfactuals

Imagined alternative scenario. 🗳 "What if the U.S. hadn't invaded Iraq in 2003?" Helps think about causality and test claims in absence of experiments.

Internal validity

Confidence that X causes Y. ✅ Higher in experiments.

External validity

Results generalize to other settings. ✅ Higher in large-N studies.

Large-N (quantitative) research

High external, medium internal validity.

Comparative case studies

Balance both internal and external validity.

Case studies

High internal, low external validity.

Comparative Method

Used when randomization is impossible.

Method of Agreement

Different cases with same outcome → find common cause. E.g., Democracies in India and Costa Rica both had strong civil societies.

Method of Difference

Similar cases with different outcomes → find differing cause. E.g., Why Chile democratized but Argentina didn't? Compare institutions.

Content Analysis

A method to analyze written records systematically.

Running record

Systematic, ongoing written record (e.g., NYT articles, Congressional speeches).

Episodic record

Irregular, personal written record (e.g., memoirs, interviews).

Frequency

Number of times a value occurs.

Relative frequency

% of total, used to compare across datasets.

Cumulative frequency

Sum of frequencies up to a point.

Mean

Average; common measure of central tendency affected by outliers.

Median

Middle value; good with skewed data.

Mode

Most frequent value; useful for categorical data (e.g., favorite political party).

Variance

Average squared deviation from mean.

Standard deviation (SD)

√variance; easier to interpret.

Normal distribution

Bell curve where mean = median = mode; 68-95-99.7 rule applies.

Z-score

How far a value is from the mean in SDs. Z = 2 means the value is 2 SDs above the mean.

Hypothesis Testing

A method involving stating hypotheses, choosing alpha, calculating test statistic, and applying decision rule.

Type I error

Rejecting H₀ when it's true.

One-tailed test

Directional hypothesis (e.g., turnout increases).

Two-tailed test

Hypothesis that allows for any change.

Critical value

Threshold for decision rule in hypothesis testing.

Observed value

Your test result in hypothesis testing.

Cross-tab

Shows distribution across two variables (e.g., party ID by race).

Chi-square test

Tests if there's a relationship between variables.

Measures of Association

Tell how strong and in what direction variables are related.

Phi coefficient

Measure of association for 2x2 tables.

Cramer's V

Measure of association for larger tables.

Spearman's rho

Measure of association for ranked/ordinal data.