The method behind macro
ยท Methodology note
Macro is an economic dashboard: inflation, unemployment and related series, pulled from FRED. This is the note that is supposed to sit beside it.
Where the numbers come from
FRED, the Federal Reserve Economic Data service run by the Federal Reserve Bank of St. Louis. The app reads published series rather than computing its own, which is the first thing worth knowing about it: nothing on that dashboard is my calculation, and anything wrong with a series is wrong at the source and not in my arithmetic.
That is a deliberate constraint, and it is the same one behind the rest of the portfolio. The apps are keyless or use the visitor's own key, they cache the last response and show its age, and they do not invent a number.
The three ways this gets misread
These are the ones I have watched happen, and they are general rather than specific to any one series.
Revisions. Most macro series are revised after first publication, sometimes substantially and sometimes months later. A screenshot of a dashboard is a screenshot of a vintage, not of a fact. Any chart of recent data is partly a chart of estimates that will move.
Units and transformations. The same series can be published as a level, a change, a percent change, or a percent change from a year ago, and they tell four different stories. An inflation chart that looks like a cliff and an inflation chart that looks like a plateau are frequently the same data under two transformations.
Seasonal adjustment. Adjusted and unadjusted series diverge in exactly the months people tend to ask about. Comparing an adjusted figure to an unadjusted one, or comparing December to November on unadjusted data, produces a confident conclusion about nothing.
None of those three is subtle, and all three are easy to walk into at speed, which is precisely the mode a dashboard encourages.
What this note cannot yet say
Not written yet, waiting on N-1
Which series macro actually pulls, with the FRED series ids, the transformation applied to each, and the refresh cadence. No figures for the product were supplied and none reproduce from a repo or report in scope as of 2026-09-21. The three cautions above are general and true; the series list has to be read off the running app rather than guessed.
Checked 2026-09-21.
Listing a plausible set of FRED series ids here would look authoritative and be unverifiable, which is the combination this desk exists to avoid.
What to do before trusting a chart on it
The same three checks that apply to any number, written up separately: how I read a number.