← The Net Labor Capacity family

Supply side: $BIRTH → S-series

The supply leg answers one question from observed data: given the people born and counted here, how many hours of work will exist five, ten and twenty years out? It starts from a count — realized births — and applies nothing but frozen schedules to it. No fertility forecast enters anywhere.

Status: The supply leg is the first part of the family to launch. $BIRTH has not yet published its first official print; the S-series follows it. This page describes the construction as specified in the methodology paper, not a series with a public track record.

What the series are

$BIRTH is a monthly index of realized US births, published as two series that are never merged. $BIRTH-P (panel observed) is the sum of registered births in the jurisdictions DMX reads directly, counted by the mother’s jurisdiction of residence, adjusted for registration lag, with no scaling and no imputation of jurisdictions outside the panel. $BIRTH-N (national estimate) is $BIRTH-P mapped to national coverage, published with a credible interval. They are kept apart because they fail differently: -P can be wrong only if a registry is wrong; -N can additionally be wrong because a model is wrong.

S(5), S(10), S(20) — the S-series — is forward labor supply at each horizon, in full-time-equivalent hours: everyone alive or born by today, carried forward by survival, participation and hours schedules that are frozen for a year at a time and published in the methodology paper’s Appendix A. A person aged twenty in twenty years is being born now, which is why a live count of births is the only input that updates the twenty-year supply leg in real time.

How it is built

S(t, h) = Σ over ages a of P(a, t) × surv(a, h) × LFPR(a + h) × H(a + h) summed over ages such that a + h falls in the working-age window 16–74; for long horizons the youngest cohorts enter as B(t) × surv(0, h) × LFPR(h) × H(h)

P is population by single year of age; surv is survival from a period life table; LFPR is participation by age; H is annual hours by age. All four are frozen schedules — nothing in the supply construction is estimated live. That is what makes S deterministic and independently replicable: anyone with the published schedules and the birth count can recompute it.

The birth count itself is built from provisional state files that fill in over months as registrations arrive. Every version of every file is archived with a cryptographic fingerprint, so DMX measures each jurisdiction’s filing curve from its own history and treats it as part of the methodology. Some states publish timely data on an occurrence basis (where the birth happened) rather than residence; those are converted for the research print by an observed ratio from that state’s own finals, frozen and published.

What is published

Free ($BIRTH Public)
National -P and -N side by side, monthly, for the reference month six months back — after the federal estimate for it has appeared — with credible interval, coverage shares, methodology paper and revision log.
Tier 01
The national print at full lead time — at least six weeks, typically eleven, ahead of the first federal estimate — with daily indicative updates between prints, full history and the revision log, by API.
Tier 02
State-level $BIRTH series and S(5/10/20), at full lead time, for jurisdictions whose published terms grant redistribution in writing (14 jurisdictions, 41.3% of US births at launch, growing by confirmation).
Cadence
Official print on the tenth US business day of the second month following the reference month, 09:00 New York time.

What settles, and how

$BIRTH-P and the S-series are the family’s first settleable objects, because their settlement construction contains no model layer: realized births, grossed by a coverage factor that is a published constant frozen at rebalancing. Settlement reads each jurisdiction’s count as it stood at a fixed settlement timestamp — 180 days after the reference month at launch, longer for a small set of jurisdictions listed in the paper — and never restates it. Later revisions are disclosed in a public revision triangle, not applied.

State-level series delivered under Tier 02 carry settlement values under the same rules, jurisdiction by jurisdiction, reproducible from a per-jurisdiction extract published with every settlement print. Settlement values are offered only once at least twelve months of DMX’s own archived vintages exist to set the lag; the free tier never carries them.

What it does not claim

It does not claim to know participation rates twenty years from now; it applies today’s frozen schedule and publishes the sensitivity of the index to schedule shocks. It does not claim migration is zero; a separate S-M variant carries the Census Bureau’s own low, middle and high migration series without a DMX point assumption. It does not claim complete coverage; panel coverage, live-signal coverage and the basis-converted share are published with every print, and the error budget is quantified in the paper.

Questions about this series?

The methodology paper is available on request; the FAQ answers the common ones on coverage, revisions and what each tier permits.