Research method
How we turn evidence on economies, institutions, societies, capital, production and resources into assessments of change, constraints and consequences.
Questions and scope
Each research item begins with a specific question, a time horizon and a defined set of actors or variables. This prevents the analysis from expanding until almost any evidence can be made to fit the argument.
We also identify what would count against the initial explanation. Evidence that weakens the working hypothesis is part of the analysis, not something to be removed from it.
Source selection
For quantitative claims and official actions, we prefer the original source when it is available: statistical agencies, central banks, governments, regulators, company filings, legislation, multilateral institutions and directly published datasets.
Academic work, specialist institutions and high-quality reporting are used for context, interpretation and developments not yet covered by primary data. Every source has limits, including revisions, definitions, incentives and gaps in coverage.
Verification
Several publications repeating the same statement do not constitute several independent confirmations. Where possible, we identify the original dataset, document, interview or official statement behind a claim.
When credible sources disagree, we describe the disagreement and examine whether it comes from different definitions, time periods, methods or genuinely conflicting evidence.
Evidence, analysis and scenarios
Observed evidence is information that can be checked directly against a source. An analysis is our interpretation of what that evidence implies. A scenario describes what could happen if stated conditions are met.
These categories can appear in the same report, but they are not interchangeable. A plausible interpretation is not a fact, and a scenario is not a forecast merely because it is internally coherent.
Mechanisms and causality
When two variables move together, the analysis asks what connects them and what other explanation could produce the same result. If the mechanism cannot be established, the conclusion should be stated more cautiously.
In political and geopolitical research, stated objectives are considered separately from the resources required to achieve them. Fiscal capacity, industrial output, logistics, institutions, military capability, technology, alliances and domestic politics can all constrain action.
Stocks, flows and valuation
A stock describes a position at a given time: reserves, debt outstanding, installed capacity or assets under management. A flow describes change during a period: investment, issuance, repayment, trade or purchases and sales.
Price and exchange-rate changes can increase the reported value of an asset without any new money changing hands. We therefore separate changes in quantity or ownership from valuation whenever the data allow.
Indirect effects
The initial effect of an event rarely ends the analysis. Sanctions can alter payment systems, sourcing and investment; war can affect energy prices, public spending and industrial capacity; demographic change can alter labor supply, savings and pension costs.
We follow these links when there is evidence for them and stop when the next step would depend mainly on speculation.
Scenario analysis
Scenarios are useful when important variables remain unresolved. Each scenario should state the assumptions or events required for it to become more plausible.
Where possible, reports identify observable indicators that would support or weaken each path. This makes later revision easier and reduces the temptation to treat one narrative as inevitable.
Uncertainty and confidence
Confidence refers to the quality, consistency and completeness of the evidence supporting an assessment. It is not a measure of how strongly the author feels about the argument.
High confidence is appropriate when several relevant sources point in the same direction and major alternatives are weak. Lower confidence is appropriate when data are incomplete, definitions conflict or the conclusion depends heavily on assumptions.
Revisions and corrections
Reports are tied to the information available at a stated date. Later releases, statistical revisions or new events can change the assessment.
When a material correction or revision changes the argument, the updated publication should make that change visible rather than silently preserving the earlier conclusion.