vintage_policy

Back to L1 | Browse all axes | Browse all options

Axis vintage_policy on sub-layer l1_a (layer l1).

Sub-layer

l1_a

Axis metadata

  • Default: 'current_vintage'

  • Sweepable: False

  • Status: operational

Operational status summary

  • Operational: 1 option(s)

  • Future: 1 option(s)

Options

current_vintage – operational

Use the latest available vintage of the dataset.

Loads the most recent FRED-MD/QD/SD snapshot bundled with the package. No real-time vintage tracking; revisions that happened after the snapshot date are not reflected.

This is the only operational option in v1.0. Real-time vintages (ALFRED-style) are tracked as a future axis – see GitHub issues #XXX.

When to use

Default for any pseudo-out-of-sample study using revised data.

When NOT to use

Real-time forecasting evaluations – those need ALFRED vintages.

References

  • macroforecast design Part 1, L1: ‘data definition is the recipe layer that pins source, target, geography, and horizon – everything downstream branches off these choices.’

Related options: information_set_type

Last reviewed 2026-05-04 by macroforecast author.

real_time_alfred – future

(no schema description for real_time_alfred)

TBD: option doc not yet authored for this value. The encyclopedia falls back to the bare schema description above. PRs adding a full OptionDoc entry under macroforecast/scaffold/option_docs/l1.py are welcome.