Research campaign · Iteration 1 · unassessed

RecessionAlert RFD — inferred public diffusion proxy campaign

What can be learned from the SHA-verified RecessionAlert RFD warning rule using transparent public macro proxies when its observation series, publication lag, and portfolio semantics are incomplete?

Expressions: 6
Logged trials: Not recorded
Independent events: Not assessed
Evidence: unassessed

What the research found

RFD is documented as a fifteen-model 0–15 recession-warning count with a strict trigger above five, but the checked-in snapshot has no observation-level native series.
A transparent FRED proxy diffusion was executable over a 1982-09 to 2026-06 feature window, with six lag/lookback/threshold variants and causal SPY/IEF overlays.
All primary proxy overlays trailed aligned SPY; timing changed results materially, so this is an interesting negative and fidelity-mismatch result rather than a promotion candidate.

Mechanism and falsifiers

Not recorded in this iteration.

Confidence and limitations

No source-linked native RFD observations, component formulas, proprietary workbook, synchronized release clock, vintage policy, or exact publication lag are captured.
The four-state FRED diffusion is an inferred analogue, not the disclosed fifteen proprietary model count; its 0–15 scaling and >5 threshold neighborhood do not establish source fidelity.
FRED histories are revised snapshots rather than point-in-time vintages, and daily Treasury inputs are reduced to last available monthly observations.
SPY/IEF/BIL are inferred overlay sleeves; RecessionAlert discloses no target, sizing, rebalance, hold, exit, re-entry, cost, or slippage rule.
Lag 0 is a same-available-period research-only sensitivity; published-facing interpretation uses one available-month lag plus one engine execution bar.
All six overlays trail aligned SPY in this exploratory window; no official indicator registry entry or investment recommendation is made.

Not recorded in this iteration.

Compare expressions

Download evidence

Exploratory results. Check each period, proxy and cost assumption before comparing. — means not recorded.

Expression / family	CAGR	Sharpe	Max drawdown	Test period	Assessment
recessionalert.1346.public_diffusion.lb1.avail1.t5p0public macro diffusion	—	—	—	Not recorded	exploratory negative or sensitivity
recessionalert.1346.public_diffusion.lb3.avail1.t5p0public macro diffusion	—	—	—	Not recorded	exploratory negative or sensitivity
recessionalert.1346.public_diffusion.lb6.avail1.t6p0public macro diffusion	—	—	—	Not recorded	exploratory negative or sensitivity
recessionalert.1346.public_diffusion.lb12.avail2.t4p0public macro diffusion	—	—	—	Not recorded	exploratory negative or sensitivity
recessionalert.1346.public_diffusion.lb6.avail0.t5p0public macro diffusion	—	—	—	Not recorded	exploratory negative or sensitivity
recessionalert.1346.public_diffusion.lb21.avail1.t5p0public macro diffusion	—	—	—	Not recorded	exploratory negative or sensitivity

Open an expression to inspect its rules and request confirmation. The request must be submitted by a trusted repository collaborator.

recessionalert.1346.public_diffusion.lb1.avail1.t5p0

claim kind

inferred

equity signal path

research/campaigns/recessionalert-rfd/lab-series.json variant_signals[]

exploratory

True

family

public_macro_diffusion

recessionalert.1346.public_diffusion.lb1.avail1.t5p0

parameters

availability lag months: 1
execution lag bars: 1
lookback: 1
threshold: 5.0

status

exploratory_negative_or_sensitivity

Confirm as strategy Confirm as indicator

recessionalert.1346.public_diffusion.lb3.avail1.t5p0

claim kind

inferred

equity signal path

research/campaigns/recessionalert-rfd/lab-series.json variant_signals[]

exploratory

True

family

public_macro_diffusion

recessionalert.1346.public_diffusion.lb3.avail1.t5p0

parameters

availability lag months: 1
execution lag bars: 1
lookback: 3
threshold: 5.0

status

exploratory_negative_or_sensitivity

Confirm as strategy Confirm as indicator

recessionalert.1346.public_diffusion.lb6.avail1.t6p0

claim kind

inferred

equity signal path

research/campaigns/recessionalert-rfd/lab-series.json variant_signals[]

exploratory

True

family

public_macro_diffusion

recessionalert.1346.public_diffusion.lb6.avail1.t6p0

parameters

availability lag months: 1
execution lag bars: 1
lookback: 6
threshold: 6.0

status

exploratory_negative_or_sensitivity

Confirm as strategy Confirm as indicator

recessionalert.1346.public_diffusion.lb12.avail2.t4p0

claim kind

inferred

equity signal path

research/campaigns/recessionalert-rfd/lab-series.json variant_signals[]

exploratory

True

family

public_macro_diffusion

recessionalert.1346.public_diffusion.lb12.avail2.t4p0

parameters

availability lag months: 2
execution lag bars: 1
lookback: 12
threshold: 4.0

status

exploratory_negative_or_sensitivity

Confirm as strategy Confirm as indicator

recessionalert.1346.public_diffusion.lb6.avail0.t5p0

claim kind

inferred

equity signal path

research/campaigns/recessionalert-rfd/lab-series.json variant_signals[]

exploratory

True

family

public_macro_diffusion

recessionalert.1346.public_diffusion.lb6.avail0.t5p0

parameters

availability lag months: 0
execution lag bars: 1
lookback: 6
threshold: 5.0

status

exploratory_negative_or_sensitivity

Confirm as strategy Confirm as indicator

recessionalert.1346.public_diffusion.lb21.avail1.t5p0

claim kind

inferred

equity signal path

research/campaigns/recessionalert-rfd/lab-series.json variant_signals[]

exploratory

True

family

public_macro_diffusion

recessionalert.1346.public_diffusion.lb21.avail1.t5p0

parameters

availability lag months: 1
execution lag bars: 1
lookback: 21
threshold: 5.0

status

exploratory_negative_or_sensitivity

Confirm as strategy Confirm as indicator

Interactive lab

Interactive history unavailable

Loading available evidence…

Research record

Source claims, inferred rules, experiments and the evidence behind the assessment.

Source

article timestamp: 2022-03-18T11:27:57-0400
fetched at: 2026-08-22T23:48:08Z
kind: RecessionAlert inventory row
native span: Narrative page snapshot only; no source-linked RFD observation span
publication lag: Unknown; monthly PRO context without synchronized release clock
redacted url: https://recessionalert.com/a-recession-forecasting-diffusion-rfd/
sha256: sha256:e705e5996621202e0e72744fdf229c4a15a8d85d1de3eff8427c7ac76bd99066
source fidelity: documented fifteen-model count and >5 warning only; no executable native series or portfolio rule
title: MODELS: A Recession Forecasting Diffusion (RFD) | RecessionAlert

What the source claims

The March 18, 2022 RecessionAlert page describes a diffusion count of fifteen diversified recession models and says RFD ranges from 0 to 15.
The page states an optimum actionable warning trigger of RFD > 5, meaning at least six models are in recession territory.
The page claims approximately six months of lead time, about two months of variability, and a 4.5–7.5-month Golden Lead Zone, but the source-linked probability table and validation history are not captured.
The chart is updated and subscribers alerted whenever RFD rises, and historical values are said to be in a Monthly PRO Excel Data File; no release clock or workbook is captured.
The fifteen named model families are disclosed, but their formulas, thresholds, directions, state encodings, missing-input rules, and revisions are not.
No source-linked RFD observations, component histories, native date span, exact publication lag, buy/sell/hold mapping, target, sizing, rebalance, costs, or execution state machine is present in the checked-in snapshot.
INDPRO, UNRATE, DGS10, DGS3MO, and HOUST form a transparent public macro diffusion analogue; SPY/IEF/BIL overlays are inferred research harnesses, not RecessionAlert rules.

Rules actually disclosed

The March 18, 2022 RecessionAlert page describes a diffusion count of fifteen diversified recession models and says RFD ranges from 0 to 15.
The page states an optimum actionable warning trigger of RFD > 5, meaning at least six models are in recession territory.
The page claims approximately six months of lead time, about two months of variability, and a 4.5–7.5-month Golden Lead Zone, but the source-linked probability table and validation history are not captured.
The chart is updated and subscribers alerted whenever RFD rises, and historical values are said to be in a Monthly PRO Excel Data File; no release clock or workbook is captured.
The fifteen named model families are disclosed, but their formulas, thresholds, directions, state encodings, missing-input rules, and revisions are not.

What had to be inferred

INDPRO, UNRATE, DGS10, DGS3MO, and HOUST form a transparent public macro diffusion analogue; SPY/IEF/BIL overlays are inferred research harnesses, not RecessionAlert rules.

Research questions

What can be learned from the SHA-verified RecessionAlert RFD warning rule using transparent public macro proxies when its observation series, publication lag, and portfolio semantics are incomplete?

Data

canonical source

sha256:e705e5996621202e0e72744fdf229c4a15a8d85d1de3eff8427c7ac76bd99066

commands

python3 -m research.research_campaign.recessionalert_1346 --run --output-dir research/campaigns/recessionalert-rfd
python3 -m research.reports notebooks --id 1346
python3 -m research.reports research-library
python3 research/scripts/check_static_reports.py docs

input hashes

recessionalert page: sha256:e705e5996621202e0e72744fdf229c4a15a8d85d1de3eff8427c7ac76bd99066

proxy datastore

data/market.sqlite opened through SQLite mode=ro

proxy sources

FRED INDPRO
FRED UNRATE
FRED DGS10
FRED DGS3MO
FRED HOUST
cached SPY
cached IEF
cached BIL

public boundary

Derived IDs, scores, signals, normalized equity, aggregate metrics, spans, and hashes only; no raw cached source payloads, credentials, cookies, or private account data.

source fetched at

2026-08-22T23:48:08Z

source snapshot

data/recessionalert.sqlite read-only; research/findings/1346.md; research/artifacts/recessionalert/manifest.json

source url

https://recessionalert.com/a-recession-forecasting-diffusion-rfd/

Baseline implementation

claim kind

inferred

equity signal path

research/campaigns/recessionalert-rfd/lab-series.json variant_signals[]

exploratory

True

family

public_macro_diffusion

recessionalert.1346.public_diffusion.lb1.avail1.t5p0

metrics

benchmark spy total return: 11.39742578500792
primary spy ief total return: 4.174145746380111
risk off events: 25
risk off fraction: 0.3231357552581262
secondary spy bil total return: 2.3835958571066675

parameters

availability lag months: 1
execution lag bars: 1
lookback: 1
threshold: 5.0

status

exploratory_negative_or_sensitivity

What to try interactively

default
5.0
label
Proxy warning threshold (scaled 0–15)
name
threshold
source fidelity
inferred neighbor around documented native RFD >5
values
- 4.0
- 5.0
- 6.0
default
1
label
Available-month lag
name
availability_lag_months
source fidelity
native publication lag unknown; lag 0 is research-only
values
- 0
- 1
- 2
default
1
label
Trailing monthly smoothing
name
lookback
source fidelity
inferred
values
- 1
- 3
- 6
- 12
- 21

Suggested next research

Can an authorized, redistributable Monthly PRO workbook or native RFD observation export be obtained?
What are the exact component thresholds, directions, state encodings, exclusions, and revision/vintage rules?
What recurring publication timestamps make the native RFD signal point-in-time executable?
Can the documented six-month lead and 6/12/18/24-month probability mappings be tested on native vintages?
If a market overlay is desired, what target, sizing, persistence, costs, and exit semantics should a researcher explicitly approve?

Trial ledger

iteration
1
status
published
summary
Offline public-proxy exploration with six executed variants; no native admission.

Agent assessment

Research-only exploratory campaign; no official RecessionAlert indicator registration or promotion is proposed.

Combinations

Fifteen pairwise state correlations are retained for transparency; no combination is promoted as native RFD.

Historical events

Event rows in diagnostics.json list every inferred risk-off transition and six-month forward SPY/IEF comparison; these are proxy events, not native RFD alerts.

Parameter sensitivity

Lookbacks 1/3/6/12/21, scaled thresholds 4/5/6, and available lags 0/1/2 were tested. Lag 0 is research-only; lag 1 is causal default.

Regime behavior

Diagnostics split outcomes into 2002–2009, 2010–2019, and 2020–2026 regimes without claiming source-level stability.

State-space exploration

Six inferred variants were executed across the longest common local macro window, with primary SPY/IEF and secondary SPY/BIL overlays.

Timing and cost sensitivity

Timing materially changes results. Native release lag and source costs are unknown, so no source transaction-cost claim is fabricated.

What did not work

No tested public proxy overlay beat aligned SPY, and the native RFD series cannot be validated from the snapshot.

What the signal looks like

The interactive lab plots the inferred monthly public diffusion score and lets readers vary threshold, lag, and smoothing; all six variant overlays remain in the sidecar.

What worked

The runner is reproducible offline, preserves missing/warm-up states, exposes source hashes and spans, and makes the proxy assumptions inspectable.

Why the failures appear to happen

The true fifteen-model basket and publication timing are unavailable; revised proxy data and defensive states can lag or remain active through recoveries.

Next questions

Can an authorized, redistributable Monthly PRO workbook or native RFD observation export be obtained?
What are the exact component thresholds, directions, state encodings, exclusions, and revision/vintage rules?
What recurring publication timestamps make the native RFD signal point-in-time executable?
Can the documented six-month lead and 6/12/18/24-month probability mappings be tested on native vintages?
If a market overlay is desired, what target, sizing, persistence, costs, and exit semantics should a researcher explicitly approve?

Review status

published. Research publication does not imply official admission.

Return to pending research