Research Methodology, Strategic Analysis and Contemporary Issues
This domain tests design literacy: whether you can move from a question to measurement and recognise what a finding can support.
Design and inference
A hypothesis states an expected relationship between an independent and a dependent variable; intervening variables transmit the effect. Deductive designs test theory; inductive designs build it. Comparative case studies use Mill's methods of agreement and difference; process tracing opens the causal mechanism; statistical designs test covariation across cases. Internal validity concerns causation, external validity generalisability and construct validity whether the measure captures the concept; reliability is consistency of measurement.
Sampling and data
Probability sampling (simple random, stratified, cluster) permits inference; non-probability sampling (purposive, snowball, convenience) is common in elite interviewing but limits generalisation. Standard datasets include SIPRI for military expenditure and arms transfers, the Correlates of War project, the Uppsala Conflict Data Program for organised violence, and UN Comtrade; government budget documents are primary sources. Triangulate, and distinguish primary from secondary material. Ethics require informed consent, anonymity for officials and no fabrication.
Worked example
Testing whether defence spending reduces conflict: operationalise spending as share of GDP and conflict as battle deaths per year from UCDP, control for regime type and income, and address reverse causality, since war raises spending. That is why correlation is not a finding.
Traps
Confusing reliability with validity; using significant without a statistical test; treating think-tank estimates as primary data; sampling on the dependent variable; ignoring endogeneity.