Data & Methods

Every series the platform uses, grouped by originating source. The identifier under each name links to the source page on FRED; the muted line beneath each row is that series’ honest limitation, carried from the registry.

Raw series are transformed with a small, documented set of standard calculations, each implemented in the calculation library and covered by unit tests (docs/economic-methodology.md).

• Year-over-year percent change. yoyₜ = (Xₜ / Xₜ₋₁₂ₖ − 1) × 100 for monthly indices (CPI, average hourly earnings, retail sales); for quarterly series the lag is four quarters.
• Month-over-month difference. The change in level from the prior month — for example, the payroll change in thousands of jobs, the same measure reported in the monthly Employment Situation.
• Historical percentile. The percentile of the latest value within the series’ full published history (or a stated window). This answers “how unusual is this?” without imposing a good/bad judgment.
• Real growth. Nominal growth minus CPI inflation over the same window — the standard first-order approximation for small rates. Where compounding matters (the purchasing-power calculator), the exact ratio form ((1 + w) / (1 + π)) − 1 is used instead.
• Monthly averaging. Weekly and daily series (mortgage rates, Treasury yields, initial claims) are averaged to calendar months when compared against monthly series, so cross-frequency comparisons align on the same time step.

Honest caveats that apply across the platform, beyond the per-series notes in the catalog above.

• National averages. Nearly every series is a U.S. aggregate. National figures mask wide variation across regions, industries, and households — no single household experiences “the” inflation rate.
• Survey error. Many headline measures come from sample surveys with meaningful month-to-month noise: the household survey behind the unemployment rate, the wide confidence intervals on monthly housing starts, and sentiment surveys sensitive to salient prices.
• Revisions. As described above, current-vintage data means early readings can and do change; single-month moves should be read with corresponding caution.
• The late-2025 release gap. The disruption to federal statistical agencies in late 2025 delayed some scheduled releases and left others unpublished. Affected series carry a gap in their histories; the pipeline preserves those gaps rather than interpolating values that were never published.