Raw Year-over-Year Recession Signal Stack
Combine unadjusted labor and trade trends to detect weakening cycles early
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 93%
The signal stack begins with non-seasonally adjusted continuing unemployment claims because they measure people who remain on benefits rather than those who merely apply. Analysts compare the raw series with the same period a year earlier, reducing dependence on opaque seasonal adjustments. A persistent sequence of year-over-year increases is treated as evidence of labor-market softening, not as a conclusive recession call. The signal is then cross-checked against world-trade volumes and other independent measures. Contracting trade adds severity because historical US recessions accompanied by declining global trade have tended to be more damaging and internationally synchronized. Eligibility changes, shortened benefit periods, and participation rates must be reviewed before interpreting the results. The output is a graduated recession-risk assessment rather than a binary prediction.
Origin
Extracted from Masters in Business, where Danielle DiMartino Booth described Quill Intelligence's use of raw continuing-claims data alongside global trade trends.
Core principles
- 01Prefer raw year-over-year comparisons when seasonal adjustments obscure trends
- 02Use continuing claims rather than applications to measure persistent labor weakness
- 03Require confirmation across independent economic channels
- 04Treat contracting world trade as an amplifier of recession risk
How to run it
- 1
Select the durable labor series
Use continuing unemployment claims rather than initial applications to focus on people who remain unemployed and receive benefits. Obtain the non-seasonally adjusted series.
Pro tip Document revisions and state-level reporting differences.
Watch out Claims exclude unemployed people who are ineligible or no longer receiving benefits.
- 2
Calculate year-over-year movement
Compare each observation with the equivalent period one year earlier. Look for direction and persistence rather than reacting to one weekly print.
Pro tip Track a rolling sequence of at least several observations.
Watch out A low comparison base can exaggerate percentage changes.
- 3
Audit structural distortions
Review whether states changed eligibility rules, benefit duration, or employer premiums. Adjust confidence in the signal when coverage has narrowed.
Pro tip Supplement the national series with state-level policy notes.
Watch out Policy-driven undercounting can make current claims appear healthier than comparable historical periods.
- 4
Confirm with global trade
Check whether world-trade volumes are also contracting year over year. Treat simultaneous labor deterioration and trade contraction as a stronger warning than either measure alone.
Pro tip Use data vintages that align in time, since global trade releases can lag.
Watch out Do not attribute old trade data to a new shock that had not yet occurred.
- 5
Grade recession risk
Classify the evidence as isolated noise, domestic softening, or synchronized contraction. Update the grade as new observations confirm or invalidate the trend.
Pro tip State the confidence level and known data limitations with every call.
Watch out The stack measures economic risk, not the exact timing or direction of asset prices.
In the wild
Quill Intelligence observed four consecutive months in which non-seasonally adjusted continuing claims were higher than a year earlier. Because benefit rules had changed and participation was incomplete, the signal was not accepted blindly. It was paired with declining world-trade volumes to evaluate whether the weakness could develop into a synchronized recession.
→ The combined evidence produced a more cautious macro outlook than the headline unemployment rate alone suggested.
An investor sees a sharp weekly increase in initial claims and considers selling risk assets. Applying the framework, the investor checks continuing claims year over year, finds no persistent increase, and sees world trade still expanding.
→ The isolated release is classified as noise rather than a confirmed recession signal.
Common mistakes
Relying on a single release
One weekly observation can reflect holidays, reporting delays, weather, or a weak comparison base. Persistence matters more than a lone spike.
Ignoring eligibility changes
Shorter benefit periods and stricter qualification rules can suppress claims even when labor conditions deteriorate.
Mixing incompatible data vintages
Lagged trade data should not be explained by a shock that occurred after the measurement period.
Is it for you?
Best for
It is best for macro researchers, investors, and risk managers monitoring turning points in the business cycle.
Not ideal for
It is not ideal as a standalone market-timing signal or when structural policy changes make historical comparisons invalid.
From the transcript
“So we follow a not seasonally adjusted number because it's just raw data.”
“If a US recession is accompanied by contracting world trade, it's a nasty recession. It's a global recession.”
From the episode
Danielle DiMartino Booth on What’s Wrong With the Fed