These definitions explain the language used across the workspace. For a particular chart or model, its own source definition, units and assumptions take precedence.
- Real GDP growth
- The percentage change in the volume of output, after allowing for price changes. It answers “did production grow?” rather than “did prices rise?”. It is a growth rate, not the size of the economy.
- Nominal GDP
- The value of output at current prices. It can rise because prices increased, real output increased, or both. GDP measured in US dollars also changes when the local currency’s dollar exchange rate changes.
- CPI inflation
- The rate of change in a consumer price index, which tracks a specified household basket. Annual-average inflation compares annual averages; a monthly year-on-year release compares one month with the same month a year earlier. The two readings can differ.
- Percent and percentage points (pp)
- If an inflation rate rises from 10% to 12%, it increases by 2 percentage points. Relative to the initial 10%, that is a 20% increase. Use percentage points when subtracting two percentage-valued rates.
- Policy rate
- A central-bank interest rate used to influence financial conditions. Countries can have several benchmark rates, such as deposit and lending rates. Check which rate is quoted; it is not automatically the rate a customer receives from a bank.
- Exchange-rate quote
- The units matter. EGP per USD means the pounds needed to buy one dollar: a higher value means a weaker pound against the dollar. USD per EGP reverses that interpretation. The website’s policy and currency scenarios retain their stated quote convention.
- Fiscal and calendar years
- A calendar year runs from January through December. A fiscal year is a different annual reporting period. Egypt’s GDP reporting basis can differ from Gulf calendar-year reporting, so a matching year label does not guarantee matching months.
- Source period
- The month, quarter or year that a figure describes. A release published today can describe an earlier period. Keep the period separate from both its publication date and the date Nile Atlas retrieved it.
- Data vintage
- The version of a dataset available at a particular point in time. The same historical year can have a different value in a later vintage. Save the downloaded file, source and retrieval date when reproducing a result.
- Revision
- A later change to a previously reported value for the same period. A revision is different from a new observation for a later period. The release desk preserves values the workspace actually observed; it is not a complete reconstruction of earlier vintages.
- Observations, estimates and projections
- An observation is a value reported for an economic period; the provider may still estimate or later revise it. An estimate uses available evidence to approximate an outcome. A projection extends into a not-yet-fully-reported period under stated assumptions. Always read the source’s own status label.
- Forecast and scenario
- A forecast is a model-based statement about future outcomes. A scenario asks what would follow if specified assumptions held. The editable policy, currency and business paths are conditional scenarios; their output is not a calibrated probability of a central-bank decision.
- Levels and changes
- A level is the value itself. A first difference subtracts the preceding period’s value and keeps the original unit. A percentage change divides that difference by the preceding value. A rate’s first difference is often measured in percentage points.
- Correlation and regression
- Correlation describes how two series move together. Regression estimates an association under a chosen specification and sample. Neither alone establishes that one variable causes the other; missing drivers, policy shifts and timing can change the result.
- AR(1) benchmark
- A simple model that relates a variable to its own previous value and an intercept. It provides a baseline to compare with other forecasts. A fitted point path alone does not show the full uncertainty around future outcomes.
- Bayesian VAR and shrinkage
- A vector autoregression models several variables together using their past values. A Bayesian prior expresses assumptions about its coefficients before observing the sample. Shrinkage pulls uncertain estimates toward the prior, which can help with small samples but does not guarantee a better forecast.
- Predictive and credible intervals
- A predictive interval describes uncertainty about a future observation under the model, including future disturbances. A credible interval describes uncertainty about a model parameter or other inferred quantity. An 80% model band is conditional on the model and assumptions; it is not a guarantee about real-world outcomes.
- Out-of-sample benchmark
- A forecast evaluation fits a model using earlier observations and checks predictions against later targets. The website’s latest-vintage benchmarks use data as available now, so they do not recreate exactly what a forecaster knew at the historical forecast date.