Research campaign · Iteration 1 · unassessed

A new monthly leading US index | RecessionAlert — exploratory proxy campaign

What can be learned from the SHA-verified RecessionAlert monthly leading-index narrative using transparent public proxies when its observations, transformations, lag, thresholds, and market rule are incomplete?

Expressions: 9
Logged trials: Not recorded
Independent events: Not assessed
Evidence: unassessed

What the research found

The SHA-verified article discloses monthly component context, four qualitative probability interpretations, a highest-two preference, and Long-/Short-leading views, but not an executable signal.
INDPRO, UNRATE, and T10Y3MM are transparent public aggregate proxies selected for local history; they do not recover the source's component identity.
Nine labeled variants ran over the BIL-limited common window; 9 variants underperformed the SPY benchmark and all outcomes remain in the ledger.
The causal baseline uses one available monthly bar; lag zero is timing-only and warm-up/missing inputs remain unavailable.
No inferred variant is promoted or registered as an official RecessionAlert indicator; this is Other / Research evidence.

Mechanism and falsifiers

Not recorded in this iteration.

Confidence and limitations

Source-linked observations, identifiers, units, transformations, and native history for all 23 components and the later 21-component views.
Component membership, weights, normalization, score range, missing-input behavior, and reconciliation of the 23-versus-21 discrepancy.
Four recession-probability model equations, calibration, outputs, and exact highest-two averaging behavior.
Numeric action threshold, persistence, smoothing, buy/sell/hold side, instrument, sizing, rebalance, holding period, costs, exits, and re-entry.
Recurring monthly release calendar, publication lag, timestamp, timezone, vintage, revision, and point-in-time availability policy.
Source-linked market/recession outcome history and an independent validation window for the article's warning claims.
All evaluated families below are public inferred proxies and must not be presented as native RecessionAlert fidelity or registered as an official indicator.

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.1344.indpro_health.lb6.lag0.t0p0indpro health	—	—	—	Not recorded	exploratory
recessionalert.1344.indpro_health.lb12.lag1.t0p0indpro health	—	—	—	Not recorded	exploratory
recessionalert.1344.indpro_health.lb21.lag2.tm0p5indpro health	—	—	—	Not recorded	exploratory
recessionalert.1344.unrate_regime.lb6.lag0.t0p0unrate regime	—	—	—	Not recorded	exploratory
recessionalert.1344.unrate_regime.lb12.lag1.t0p5unrate regime	—	—	—	Not recorded	exploratory
recessionalert.1344.unrate_regime.lb21.lag2.t0p0unrate regime	—	—	—	Not recorded	exploratory
recessionalert.1344.curve_stress.lb6.lag0.t0p0curve stress	—	—	—	Not recorded	exploratory
recessionalert.1344.curve_stress.lb12.lag1.t0p0curve stress	—	—	—	Not recorded	exploratory
recessionalert.1344.curve_stress.lb21.lag2.tm0p5curve stress	—	—	—	Not recorded	exploratory

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

recessionalert.1344.indpro_health.lb6.lag0.t0p0

available signal observations: 1285
claim kind: inferred
execution: one available monthly bar is the causal baseline; two-bar lag is conservative sensitivity; lag 0 is research-only.
execution status: same_close_research_only
exploratory: True
family: indpro_health
family description: inferred INDPRO activity proxy below a trailing mean
hold months: 1
id: recessionalert.1344.indpro_health.lb6.lag0.t0p0
input names: INDPRO
lag bars: 0
lookback: 6
market inputs: SPY
IEF
BIL
missingness: warm-up and unavailable states remain unavailable, never converted to risk-on
publication lag: source unknown; campaign uses monthly cutoff and explicit execution lag
risk off events: 91
risk off fraction: 0.3027237354085603
rule label: binary inferred risk-off overlay; no native RecessionAlert buy/sell rule is claimed
same close status: research_only_sensitivity
signal span: 1919-06-30 to 2026-06-30
source identity: public aggregate proxy; not the undisclosed 23-component source index
spec id: recessionalert.1344.indpro_health.lb6.lag0.t0p0
threshold: 0.0
unavailable signal observations: 5

Confirm as strategy Confirm as indicator

recessionalert.1344.indpro_health.lb12.lag1.t0p0

available signal observations: 1279
claim kind: inferred
execution: one available monthly bar is the causal baseline; two-bar lag is conservative sensitivity; lag 0 is research-only.
execution status: causal_one_available_bar
exploratory: True
family: indpro_health
family description: inferred INDPRO activity proxy below a trailing mean
hold months: 1
id: recessionalert.1344.indpro_health.lb12.lag1.t0p0
input names: INDPRO
lag bars: 1
lookback: 12
market inputs: SPY
IEF
BIL
missingness: warm-up and unavailable states remain unavailable, never converted to risk-on
publication lag: source unknown; campaign uses monthly cutoff and explicit execution lag
risk off events: 48
risk off fraction: 0.2720875684128225
rule label: binary inferred risk-off overlay; no native RecessionAlert buy/sell rule is claimed
same close status: not_used
signal span: 1919-12-31 to 2026-06-30
source identity: public aggregate proxy; not the undisclosed 23-component source index
spec id: recessionalert.1344.indpro_health.lb12.lag1.t0p0
threshold: 0.0
unavailable signal observations: 11

Confirm as strategy Confirm as indicator

recessionalert.1344.indpro_health.lb21.lag2.tm0p5

available signal observations: 1270
claim kind: inferred
execution: one available monthly bar is the causal baseline; two-bar lag is conservative sensitivity; lag 0 is research-only.
execution status: causal_two_available_bars
exploratory: True
family: indpro_health
family description: inferred INDPRO activity proxy below a trailing mean
hold months: 3
id: recessionalert.1344.indpro_health.lb21.lag2.tm0p5
input names: INDPRO
lag bars: 2
lookback: 21
market inputs: SPY
IEF
BIL
missingness: warm-up and unavailable states remain unavailable, never converted to risk-on
publication lag: source unknown; campaign uses monthly cutoff and explicit execution lag
risk off events: 28
risk off fraction: 0.18661417322834645
rule label: binary inferred risk-off overlay; no native RecessionAlert buy/sell rule is claimed
same close status: not_used
signal span: 1920-09-30 to 2026-06-30
source identity: public aggregate proxy; not the undisclosed 23-component source index
spec id: recessionalert.1344.indpro_health.lb21.lag2.tm0p5
threshold: -0.5
unavailable signal observations: 20

Confirm as strategy Confirm as indicator

recessionalert.1344.unrate_regime.lb6.lag0.t0p0

available signal observations: 935
claim kind: inferred
execution: one available monthly bar is the causal baseline; two-bar lag is conservative sensitivity; lag 0 is research-only.
execution status: same_close_research_only
exploratory: True
family: unrate_regime
family description: inferred unemployment rate-of-change risk proxy
hold months: 1
id: recessionalert.1344.unrate_regime.lb6.lag0.t0p0
input names: UNRATE
lag bars: 0
lookback: 6
market inputs: SPY
IEF
BIL
missingness: warm-up and unavailable states remain unavailable, never converted to risk-on
publication lag: source unknown; campaign uses monthly cutoff and explicit execution lag
risk off events: 57
risk off fraction: 0.3497326203208556
rule label: binary inferred risk-off overlay; no native RecessionAlert buy/sell rule is claimed
same close status: research_only_sensitivity
signal span: 1948-07-31 to 2026-06-30
source identity: public aggregate proxy; not the undisclosed 23-component source index
spec id: recessionalert.1344.unrate_regime.lb6.lag0.t0p0
threshold: 0.0
unavailable signal observations: 6

Confirm as strategy Confirm as indicator

recessionalert.1344.unrate_regime.lb12.lag1.t0p5

available signal observations: 929
claim kind: inferred
execution: one available monthly bar is the causal baseline; two-bar lag is conservative sensitivity; lag 0 is research-only.
execution status: causal_one_available_bar
exploratory: True
family: unrate_regime
family description: inferred unemployment rate-of-change risk proxy
hold months: 1
id: recessionalert.1344.unrate_regime.lb12.lag1.t0p5
input names: UNRATE
lag bars: 1
lookback: 12
market inputs: SPY
IEF
BIL
missingness: warm-up and unavailable states remain unavailable, never converted to risk-on
publication lag: source unknown; campaign uses monthly cutoff and explicit execution lag
risk off events: 30
risk off fraction: 0.348762109795479
rule label: binary inferred risk-off overlay; no native RecessionAlert buy/sell rule is claimed
same close status: not_used
signal span: 1949-01-31 to 2026-06-30
source identity: public aggregate proxy; not the undisclosed 23-component source index
spec id: recessionalert.1344.unrate_regime.lb12.lag1.t0p5
threshold: 0.5
unavailable signal observations: 12

Confirm as strategy Confirm as indicator

recessionalert.1344.unrate_regime.lb21.lag2.t0p0

available signal observations: 920
claim kind: inferred
execution: one available monthly bar is the causal baseline; two-bar lag is conservative sensitivity; lag 0 is research-only.
execution status: causal_two_available_bars
exploratory: True
family: unrate_regime
family description: inferred unemployment rate-of-change risk proxy
hold months: 3
id: recessionalert.1344.unrate_regime.lb21.lag2.t0p0
input names: UNRATE
lag bars: 2
lookback: 21
market inputs: SPY
IEF
BIL
missingness: warm-up and unavailable states remain unavailable, never converted to risk-on
publication lag: source unknown; campaign uses monthly cutoff and explicit execution lag
risk off events: 16
risk off fraction: 0.3673913043478261
rule label: binary inferred risk-off overlay; no native RecessionAlert buy/sell rule is claimed
same close status: not_used
signal span: 1949-10-31 to 2026-06-30
source identity: public aggregate proxy; not the undisclosed 23-component source index
spec id: recessionalert.1344.unrate_regime.lb21.lag2.t0p0
threshold: 0.0
unavailable signal observations: 21

Confirm as strategy Confirm as indicator

recessionalert.1344.curve_stress.lb6.lag0.t0p0

available signal observations: 530
claim kind: inferred
execution: one available monthly bar is the causal baseline; two-bar lag is conservative sensitivity; lag 0 is research-only.
execution status: same_close_research_only
exploratory: True
family: curve_stress
family description: inferred T10Y3MM inversion proxy below a smoothed level
hold months: 1
id: recessionalert.1344.curve_stress.lb6.lag0.t0p0
input names: T10Y3MM
lag bars: 0
lookback: 6
market inputs: SPY
IEF
BIL
missingness: warm-up and unavailable states remain unavailable, never converted to risk-on
publication lag: source unknown; campaign uses monthly cutoff and explicit execution lag
risk off events: 6
risk off fraction: 0.1018867924528302
rule label: binary inferred risk-off overlay; no native RecessionAlert buy/sell rule is claimed
same close status: research_only_sensitivity
signal span: 1982-06-30 to 2026-07-31
source identity: public aggregate proxy; not the undisclosed 23-component source index
spec id: recessionalert.1344.curve_stress.lb6.lag0.t0p0
threshold: 0.0
unavailable signal observations: 5

Confirm as strategy Confirm as indicator

recessionalert.1344.curve_stress.lb12.lag1.t0p0

available signal observations: 524
claim kind: inferred
execution: one available monthly bar is the causal baseline; two-bar lag is conservative sensitivity; lag 0 is research-only.
execution status: causal_one_available_bar
exploratory: True
family: curve_stress
family description: inferred T10Y3MM inversion proxy below a smoothed level
hold months: 1
id: recessionalert.1344.curve_stress.lb12.lag1.t0p0
input names: T10Y3MM
lag bars: 1
lookback: 12
market inputs: SPY
IEF
BIL
missingness: warm-up and unavailable states remain unavailable, never converted to risk-on
publication lag: source unknown; campaign uses monthly cutoff and explicit execution lag
risk off events: 3
risk off fraction: 0.08587786259541985
rule label: binary inferred risk-off overlay; no native RecessionAlert buy/sell rule is claimed
same close status: not_used
signal span: 1982-12-31 to 2026-07-31
source identity: public aggregate proxy; not the undisclosed 23-component source index
spec id: recessionalert.1344.curve_stress.lb12.lag1.t0p0
threshold: 0.0
unavailable signal observations: 11

Confirm as strategy Confirm as indicator

recessionalert.1344.curve_stress.lb21.lag2.tm0p5

available signal observations: 515
claim kind: inferred
execution: one available monthly bar is the causal baseline; two-bar lag is conservative sensitivity; lag 0 is research-only.
execution status: causal_two_available_bars
exploratory: True
family: curve_stress
family description: inferred T10Y3MM inversion proxy below a smoothed level
hold months: 3
id: recessionalert.1344.curve_stress.lb21.lag2.tm0p5
input names: T10Y3MM
lag bars: 2
lookback: 21
market inputs: SPY
IEF
BIL
missingness: warm-up and unavailable states remain unavailable, never converted to risk-on
publication lag: source unknown; campaign uses monthly cutoff and explicit execution lag
risk off events: 1
risk off fraction: 0.040776699029126215
rule label: binary inferred risk-off overlay; no native RecessionAlert buy/sell rule is claimed
same close status: not_used
signal span: 1983-09-30 to 2026-07-31
source identity: public aggregate proxy; not the undisclosed 23-component source index
spec id: recessionalert.1344.curve_stress.lb21.lag2.tm0p5
threshold: -0.5
unavailable signal observations: 20

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: 2016-03-03
database: data/recessionalert.sqlite opened read-only
fetched at: 2026-08-22T23:45:29Z
inventory: research/findings/recessionalert_inventory.md
manifest: research/artifacts/recessionalert/manifest.json
native span: single March 3, 2016 HTML article snapshot; narrative shown through January 2016; no component observations
publication lag: unknown; no recurring release clock, timestamp, timezone, vintage, or point-in-time policy disclosed
redacted url: https://recessionalert.com/a-new-long-leading-us-index/
sha256: sha256:deedcf22d1d280248542375b8093cf84c98462b27f8eb4f92d09c0d5e1c8b246

What the source claims

The article names 23 monthly components and says the index is shown through January 2016, but supplies no source-linked observations or transformations.
The article describes four recession-probability interpretations and a preferred average of the highest two, but gives no equations, calibration, or output series.
The article describes Long-leading and Short-leading views from 21 components, without reconciling the 23-component list.
This campaign tests INDPRO, UNRATE, and T10Y3MM public proxies with explicit lookback, threshold, and lag neighbors; they are not native source fidelity.
SPY risk-on versus IEF/BIL defensive sleeves are descriptive inferred overlays because the source provides no target, sizing, or exit rule.

Rules actually disclosed

Monthly component context, 23 named components, four probability interpretations, highest-two preference, and 21-component Long-/Short-leading context; no executable state machine.

What had to be inferred

INDPRO/UNRATE/T10Y3MM proxies, lookbacks, thresholds, one/two-bar execution, and SPY/IEF/BIL overlays.

Research questions

What can be learned from the SHA-verified RecessionAlert monthly leading-index narrative using transparent public proxies when its observations, transformations, lag, thresholds, and market rule are incomplete?

Data

canonical native span: single March 3, 2016 HTML article snapshot; narrative shown through January 2016; no component observations
canonical publication lag: unknown; no recurring release clock, timestamp, timezone, vintage, or point-in-time policy disclosed
canonical sha256: sha256:deedcf22d1d280248542375b8093cf84c98462b27f8eb4f92d09c0d5e1c8b246
canonical source: research/artifacts/recessionalert/manifest.json + research/findings/recessionalert_inventory.md + read-only data/recessionalert.sqlite
database policy: All reads use SQLite mode=ro; no canonical crawl or market cache mutation.
primary common window: 2007-05 through 2026-06 after monthly alignment and BIL intersection
proxy inputs: FRED INDPRO/UNRATE/T10Y3MM and Yahoo SPY/IEF/BIL snapshots represented by checked-in data/market.sqlite
redistribution: Publish derived metrics, proxy labels, source hashes, gap notes, charts, and notebook only; no raw page/media, credentials, cookies, or private account data.

Baseline implementation

available signal observations

1285

claim kind

inferred

execution

one available monthly bar is the causal baseline; two-bar lag is conservative sensitivity; lag 0 is research-only.

execution status

same_close_research_only

exploratory

True

family

indpro_health

family description

inferred INDPRO activity proxy below a trailing mean

hold months

recessionalert.1344.indpro_health.lb6.lag0.t0p0

input names

INDPRO

lag bars

lookback

market inputs

metrics

SPY BIL

annualized return: 0.09297474022891805
benchmark total return: 5.9368716911788315
end: 2026-06-30
max drawdown: -0.16221310705313907
observations: 229
risk off fraction: 0.37554585152838427
sharpe: 0.8544601344677404
start: 2007-06-30
total return: 4.455190068956827
volatility: 0.11163990440234395

SPY IEF

annualized return: 0.09588816367698749
benchmark total return: 11.69295042059387
end: 2026-06-30
max drawdown: -0.17612333323309304
observations: 287
risk off fraction: 0.3310104529616725
sharpe: 0.8545124270567294
start: 2002-08-31
total return: 7.934614480068149
volatility: 0.11528209476306092

missingness

warm-up and unavailable states remain unavailable, never converted to risk-on

publication lag

source unknown; campaign uses monthly cutoff and explicit execution lag

risk off events

risk off fraction

0.3027237354085603

rule label

binary inferred risk-off overlay; no native RecessionAlert buy/sell rule is claimed

same close status

research_only_sensitivity

signal span

1919-06-30 to 2026-06-30

source identity

public aggregate proxy; not the undisclosed 23-component source index

spec id

recessionalert.1344.indpro_health.lb6.lag0.t0p0

threshold

0.0

unavailable signal observations

What to try interactively

default
recessionalert.1344.indpro_health.lb12.lag1.t0p0
description
Select an inferred public proxy variant; no setting is disclosed by RecessionAlert.
name
variant
values
- recessionalert.1344.indpro_health.lb6.lag0.t0p0
- recessionalert.1344.indpro_health.lb12.lag1.t0p0
- recessionalert.1344.indpro_health.lb21.lag2.tm0p5
- recessionalert.1344.unrate_regime.lb6.lag0.t0p0
- recessionalert.1344.unrate_regime.lb12.lag1.t0p5
- recessionalert.1344.unrate_regime.lb21.lag2.t0p0
- recessionalert.1344.curve_stress.lb6.lag0.t0p0
- recessionalert.1344.curve_stress.lb12.lag1.t0p0
- recessionalert.1344.curve_stress.lb21.lag2.tm0p5
default
1
description
Lag 1 is the causal baseline; lag 0 is research-only same-close sensitivity.
name
execution_lag
values
- 0
- 1
- 2
default
IEF
description
Descriptive inferred sleeve; the source discloses no target instrument.
name
defensive_sleeve
values
- IEF
- BIL

Suggested next research

Can an authorized redistributable source-linked monthly panel or chart be obtained?
What exact transformations, weights, component membership, probability calibration, and highest-two missing behavior did the source use?
What recurring publication timestamps, vintages, revisions, and point-in-time cutoffs apply?
Can the source's warning claims be recomputed against a defined market outcome?
Does any complete buy/sell/hold/exit rule exist beyond the qualitative narrative?

Trial ledger

iteration
1
status
published
summary
Offline public-proxy campaign published while native source-fidelity gaps remain.

State-space exploration

Nine labeled variants tested; negative outcomes preserved in state-space.json.

What did not work

Aggregate proxies cannot validate the native RecessionAlert identity or source performance claims.

Next questions

Can an authorized redistributable source-linked monthly panel or chart be obtained?
What exact transformations, weights, component membership, probability calibration, and highest-two missing behavior did the source use?
What recurring publication timestamps, vintages, revisions, and point-in-time cutoffs apply?
Can the source's warning claims be recomputed against a defined market outcome?
Does any complete buy/sell/hold/exit rule exist beyond the qualitative narrative?

Review status

published. Research publication does not imply official admission.

Return to pending research