Security Contact

Policy Center

Research Methodology

A repeatable process for framing market questions, ranking evidence, checking time alignment, analyzing uncertainty, documenting assumptions, and separating research-based market context from trade instructions.

A methodology improves consistency; it does not eliminate model risk, source error, interpretation differences, or unexpected market events.

Level 1

Official filings, regulator or government releases, exchange notices, and original provider documentation.

Level 2

Issuer materials, original datasets, audited information, and established academic work.

Level 3

Established reporting, research commentary, and professional analysis used for context.

Level 4

Unverified social posts, anonymous claims, promotional assertions, and unsupported summaries requiring caution.

1. Define the research question

Start with a specific question, timeframe, instrument or market, and intended use. “What happened?” requires a different method from “What might matter next?” or “How does a strategy behave under a defined rule?” The scope should state important exclusions.

2. Collect and rank evidence

Identify the best available evidence for each material claim. Prefer primary sources where practical. Record publication time, effective period, revision status, unit, currency, and whether a figure is actual, estimated, seasonally adjusted, or derived.

Source hierarchy is a guide, not an automatic truth score. An official source may be incomplete, revised, or limited to a specific definition. A secondary source may add valuable context if it accurately represents the underlying material.

3. Check timing and comparability

Market research often fails because observations from different periods or definitions are compared as though they were identical. Review the timestamp, timezone, session, corporate actions, benchmark composition, inflation basis, reporting period, and measurement methodology.

4. Prepare and document data

Where calculations are used, identify missing values, outliers, survivorship bias, look-ahead bias, symbol changes, delistings, adjusted prices, currency conversion, fees, slippage, and transaction assumptions. A transformed dataset should be described sufficiently for a knowledgeable reviewer to understand the main steps.

5. Analyze without overstating precision

Use methods appropriate to the question and data. Avoid false precision, selective start dates, unsupported causal language, and conclusions that rely on a single favorable measure. Where practical, compare alternative definitions, periods, and assumptions.

Qualitative analysis

Qualitative work may examine incentives, policy language, market structure, management statements, legal changes, competitive conditions, or source credibility. It should identify which conclusions are interpretive.

Quantitative analysis

Quantitative work may use descriptive statistics, historical comparisons, scenario analysis, factor relationships, or rule-based illustrations. Statistical association does not automatically establish causation or future profitability.

6. Evaluate uncertainty and risk

Identify evidence that would weaken the conclusion, plausible alternative explanations, missing variables, concentration, liquidity, leverage, regime change, model decay, and tail events. Risk discussion should be proportionate to the claim and the potential consequence of misuse.

7. Distinguish scenarios from forecasts

Scenarios, forecasts, targets, and model outputs are evaluated according to their stated assumptions, evidence, time horizon, and limitations. Research should distinguish observed facts from judgment and identify what could change a conclusion.

8. Review and challenge

Review should test factual support, calculations, source use, assumptions, internal consistency, disclosures, and presentation. Where practical, a reviewer should look for the strongest reasonable counterargument rather than only confirming the draft conclusion.

9. Publish with context

The final material should identify the relevant timeframe, main sources, assumptions, important limitations, and whether the content is fact, analysis, scenario, or opinion. A chart should retain necessary source and provider attribution.

10. Monitor, update, and correct

Time-sensitive material may become stale. Updates should be considered when a source is revised, an instrument changes, a provider corrects data, a material assumption fails, or a reader identifies a supportable error. Corrections are handled under the Corrections Policy.

Methodology limitations

No methodology guarantees accuracy, profitability, or completeness. Public information may be delayed or wrong; private facts may be unavailable; markets may change faster than a publication can be updated; and reasonable analysts may disagree. Research supports judgment but does not replace responsibility for the final decision.