How the research is produced

These standards apply to all contributors.

Sources and evidence

Research starts with primary material: official statistics, budget documents, national treasuries, debt management offices and central banks. IMF, World Bank and other multilateral datasets support broader analysis and comparisons, with differences in definitions and coverage identified where relevant.

Commercial and market data are used within their licensing terms. Market prices provide evidence about expectations, liquidity and risk appetite; they do not independently establish a country’s underlying financial position. Press reporting provides context and helps establish event timelines. Quantitative claims are checked against their underlying sources.

Dates, releases and revisions

Economic data changes as new information becomes available. Figures therefore identify the period they describe and the date of the information used. Where revisions could affect the argument, the relevant release or data vintage is recorded.

Historical analysis uses the information available at the time. Charts reconstructed with subsequently revised data are labelled accordingly.

Signals and forecasts

A signal describes an observed reading relative to a stated baseline. It can identify a development that warrants closer examination.

A forecast assigns a probability to a defined future event. It specifies a deadline, the evidence that will determine the outcome and a rule for resolving the forecast.

The Signal Monitor does not generate forecasts automatically. Published probabilities are assigned by the named author, with a written rationale; they are not model-generated. Revisions are dated and retained alongside the original forecast.

Reading z-scores and percentiles

A z-score expresses the distance between a reading and its historical average in standard deviations, a measure of how widely the observations vary. A score of +2 means two standard deviations above the average for the stated baseline period. A score of −2 means two below.

A historical percentile shows where a reading sits within that period’s distribution. A reading at the 95th percentile is higher than approximately 95% of observations in the comparison window.

Both measures depend on the baseline selected. Neither, on its own, gives the probability of a future event.

Interpreting unusual readings

An unusual reading needs an explanation. It may reflect deteriorating conditions, a deliberate policy change or a break in how the data is measured.

Each indicator therefore identifies the direction associated with greater risk and includes a written interpretation. The current level is shown alongside its historical deviation: countries with similar z-scores can face very different financial pressures.

Baseline periods, transformations and material changes in definitions are disclosed. A colour or statistical threshold does not replace the accompanying analysis.

Missing data and comparisons

Publication schedules vary across countries, and some releases lag by several months. Series may also be incomplete, irregular or revised without clear notice.

Missing, stale or otherwise unusable data are marked “Insufficient or stale data” in gray. That status provides no basis for a directional conclusion and should not be read as evidence of low risk.

Peer comparisons are included only where the countries and measures are sufficiently comparable. The selection and its limitations are explained.

Assessing forecasts

Resolved binary forecasts are evaluated using the Brier score:

Brier score = (p − y)²

Here, p is the published probability between 0 and 1. The outcome, y, is 1 if the event occurred and 0 if it did not. Lower scores indicate smaller errors. Aggregate results report the average score across the included forecasts.

The initial probability is the primary basis for assessing performance. The latest eligible revision is reported separately and must have been published before both the deadline and the outcome becoming known.

Results identify the scoring basis, evaluation period and number of resolved forecasts. Comparisons use a base-rate benchmark declared in advance for the relevant forecast category. Unresolved or voided forecasts remain visible, with any exclusion explained. Performance claims require enough comparable outcomes to support them.

Corrections and changes to the method

Substantive corrections are recorded on the affected article or chart. Changes to methods are dated and explained. Original forecasts and their revisions remain on the record so readers can distinguish the original assessment from later information.

Sources under evaluation

The following sources are being assessed for possible future use. They are not live integrations, and their data is not currently included in the site:

• IMF PortWatch
• GDELT Project
• ALFRED, Federal Reserve Bank of St. Louis

Scope

EQLID publishes research and commentary for informational purposes. Its content does not constitute investment advice or a credit rating and does not take account of any reader’s individual circumstances.