By the time the Bureau of Economic Analysis tells you the economy grew 2.3% last quarter, you are looking at data that is 30 days stale, will be revised twice, and averages together components moving in opposite directions. GDP is a lagging indicator that arrives too late to be useful and too aggregated to be true.

The real question is not “what is GDP?” It is “what is GDP about to become?” And that question cannot be answered by headline data. It requires understanding the underlying mechanics of how an economy produces output.

This is why growth sits at the nexus of our framework, not because GDP itself matters, but because growth dynamics determine everything downstream: corporate profits, Fed policy, credit conditions, and asset prices. Get the growth call right, and you have triangulated the policy response. Miss it, and you are trading yesterday’s narrative while tomorrow unfolds.

Note: Due to the Q4 2025 government shutdown, several agency releases (Census, BEA) are operating on a significant lag. All data below reflects the most current official releases as of February 5, 2026.

The Core Insight: The Second Derivative

Here is the conceptual unlock that separates useful analysis from headline-watching.

Growth is not a level. It is a rate of change. And what matters most is the change in that rate of change.

When industrial production decelerates from +3.5% to +1.2%, that is not “still growing.” That is a momentum break. The economy is still in positive territory, but the direction has shifted. Deceleration tells you more than the level.

Why does this work? Because economic decisions are made at the margin. Companies do not wait for GDP to go negative before cutting investment. They sense weakening demand, see order books thinning, feel customers hesitating. Then they freeze hiring, defer capital expenditure, and let inventory accumulate. The hard data catches up later.

The second derivative captures this. When growth is still positive, but the rate of improvement is slowing, something has changed. The momentum that carried the expansion is fading. What follows depends on whether the deceleration stabilizes or accelerates into contraction.

This is not theory. We validated it across every post-war recession. Industrial production typically peaks well before recession (median ~20 months, though highly variable). The deceleration phase (second derivative turning negative) is the early warning that the level decline is coming.

Figure 1: Industrial Production YoY with second derivative shading. Orange bands show periods when momentum is breaking (growth still positive but decelerating).

The chart makes it visible. The shaded regions show recessions. The second derivative (change in growth rate) turns negative before the level does. Every time. The pattern does not fail because the mechanics behind it do not change.

What to Watch and Why

We approach growth analysis through three lenses. Not a checklist, but a framework for organizing the signal from the noise.

Hard data captures what actually happened. Industrial production. Retail sales. Capital goods shipments. Aggregate hours worked. These are not surveys or sentiment. They measure physical output, real transactions, actual labor input. Hard data is slower to release but harder to manipulate. When industrial production contracts, something real is happening.

Soft data captures what decision-makers expect to happen. Purchasing managers’ indices. CEO confidence surveys. Consumer sentiment. Soft data is faster, often released before hard data for the same period. But it can lie. People say one thing and do another. Sentiment can stay elevated while actual production slumps, or collapse while output holds steady. Use soft data to generate hypotheses, hard data to confirm them.

Real-time trackers capture what is happening in real time. Weekly economic indices. Credit card spending. Freight volumes. Electricity consumption. These run circles around the official data in timeliness. Some update daily. The trade-off is noise: week-to-week volatility can obscure the signal. But for nowcasting between official releases, real-time trackers are indispensable.

The discipline is not picking one lens. It is triangulating across all three. When hard data, soft data, and real-time trackers all point the same direction, the signal is high-confidence. When they diverge, that divergence itself is information worth understanding.

The Indicators That Matter

Industrial Production: The Monthly GDP Proxy

Industrial production measures the real output of the manufacturing, mining, and utilities sectors. It is monthly (unlike quarterly GDP), hard data (unlike survey PMIs), and highly correlated with GDP (r = 0.86).

Why it works: IP captures the physical production decisions that flow through to GDP with a lag. When factories cut output, they are responding to weakening orders, rising inventories, or tightening credit. Those same forces will eventually show up in the expenditure-side GDP calculation. But IP shows you earlier.

Figure 2: Industrial Production YoY vs Real GDP YoY. Highly correlated, with IP typically peaking slightly before GDP at cycle turns.

The chart shows the relationship. IP and GDP move together with a correlation of 0.86. At cycle peaks, IP typically turns down slightly before GDP, though the lead time varies. IP is not a perfect proxy for the whole economy, since manufacturing is only 11% of GDP. But manufacturing is cyclical in ways that services are not. When manufacturing contracts, it is usually telling you something about the broader cycle.

Historical validation: Industrial production peaked before all post-war recessions. The median lead time at peaks is around 20-24 months before the recession start, though this varies widely. IP troughs tend to coincide almost exactly with the end of the recession. This is not a pattern that has worked “sometimes.” It has worked every time.

Current reading: IP is running at +2.0% year-over-year as of December 2025, with manufacturing specifically at +2.1%. After contracting for much of 2023-2024, the goods-producing economy has turned positive. But watch the second derivative: the rate of improvement matters as much as the level.

ISM Manufacturing: The Forward-Looking Survey

The Institute for Supply Management surveys purchasing managers at manufacturing firms about new orders, production, employment, supplier deliveries, and inventories. The headline index is a diffusion index: readings above 50 mean more firms reporting expansion than contraction.

Why it works: Purchasing managers are the front line. They see order books before revenue shows up. They sense customer hesitation before it becomes cancellation. The ISM captures this real-time intelligence, releasing on the first business day of each month for the prior month’s data.

Figure 3: Regional Fed Manufacturing Surveys (Empire State, Philly Fed) serve as a proxy for ISM. Above 0 = expansion, similar to ISM above 50. ISM itself (box) is 52.6 as of January 2026.

The key level is 50. Above 50 means expansion. Below 50 means contraction. But context matters. Sustained readings below 48 have historically preceded recession. A single month below 50 can be noise. Six consecutive months below 50 is a different signal entirely.

The sub-indices matter as much as the headline. New Orders leads the composite by 1-2 months. When New Orders minus Inventories goes negative, demand is weaker than supply. That is a forward-looking signal that the headline has yet to reflect.

Current reading: ISM Manufacturing surged to 52.6 in January 2026, the first expansion in 12 months after 10 consecutive months of contraction. Before that, manufacturing managed just two months of expansion following a punishing 26-month contraction streak, the longest since 2000-2002. In total, manufacturing has spent 36 of the last 40 months in contraction territory. New Orders jumped to 57.1. The question now is whether this represents genuine demand recovery or tariff-related front-running. Supporting the front-running thesis: Prices Paid rose to 59.0 while Customers’ Inventories dropped to 38.7 (a “too low” reading). Low customer inventories plus rising prices suggests panic buying ahead of expected tariffs. ISM Chair Susan Spence flagged that some buying may be positioning ahead of expected price increases. Services (ISM Services at 53.8) remain solid. If manufacturing improvement sustains, the two-speed economy narrative inverts.

Core Capital Goods Orders: CEO Confidence in Real Dollars

Durable goods orders are volatile. Aircraft orders spike and crash. Defense contracts are lumpy. The noise obscures the signal.

Core capital goods orders (nondefense, ex-aircraft) strip away that volatility. What remains is business equipment: machinery, computers, transportation equipment. When CEOs commit capital to these items, they are betting on future demand. It is not cheap to build a new production line. They only do it when they expect returns.

Why it works: Capital expenditure decisions are forward commitments. Unlike hiring (which can be reversed with layoffs) or inventory (which can be liquidated), capex represents multi-year bets. When core capital goods orders decline, CEOs are saying they see weakness ahead. They are pulling back on the very investments that would drive future growth. And when orders accelerate, it signals confidence in the outlook.

Figure 4: Core Capital Goods Orders YoY vs Nonresidential Fixed Investment. Orders lead investment by 3-6 months.

Historical validation: Core capital goods orders have led business equipment investment by 3-6 months at every turning point since 2000. The lead relationship held in 2000, 2007, 2014-15, 2018-19, and 2022. Orders peaked in early 2022 at roughly +12% year-over-year, then decelerated through 2023 before stabilizing.

Current reading: Core capital goods orders are running at +5.3% year-over-year as of November 2025, hitting a record $78.2 billion. This is a genuine bright spot. CEOs are committing real capital to equipment investment. The question is whether this confidence survives tariff uncertainty and tightening financial conditions.

Aggregate Hours Worked: The Labor Input to Output

From our labor pillar: before companies lay off workers, they cut hours. Before they cut hours formally, they reduce overtime. Hours are the most reversible lever employers have.

Aggregate hours worked captures the total labor input to the economy: employment times average weekly hours. It is the volume of labor feeding into the production function. GDP cannot grow sustainably if labor input is shrinking, unless productivity surges to offset it. (It usually does not.)

Figure 5: Aggregate Hours Worked YoY. Hours contract before headcount cuts, making this an early warning signal.

Current configuration: Aggregate hours growth has been running below +1% year-over-year through the second half of 2025. Private employment growth has decelerated to roughly +0.5% YoY while average weekly hours remain flat at 34.2. That combination is positive but tepid, well below the 1.5%+ growth rate typical of healthy expansions. Labor input is growing, but barely.

Housing Starts: The Long Lead

Housing operates on a different timescale than most economic indicators. Building permits precede starts by 1-2 months. Starts precede residential investment by 6-9 months. Residential investment feeds into GDP with further lags.

The result: housing starts are one of the longest-leading indicators in the toolkit. They peaked before every recession in the post-war era, often by 18-24 months or more. The 2006 housing peak preceded the December 2007 recession start by 23 months. The median lead time across all post-war recessions is around 32 months. By the time that recession was officially dated, housing had already collapsed 40%.

Figure 6: Housing Starts YoY vs Real GDP YoY. Housing typically peaks 18-24+ months before recession.

Why it works: Housing is rate-sensitive and credit-sensitive. When the Fed tightens, mortgage rates rise. Affordability declines. Buyers pull back. Builders respond by cutting permits and starts. The transmission is mechanical and predictable. Housing is always an early casualty of Fed tightening cycles.

Current reading: Housing starts peaked at 1.8 million (annualized) in April 2022. They are now at 1.25 million (October 2025, the final release before the shutdown-induced reporting freeze). That is down -7.8% year-over-year on a single-month basis, or -6.3% on a smoothed 3-month average as shown in the chart. Housing has been weak for over 30 months.

The Consensus Trap

Here is the pattern that repeats every cycle.

Surface narrative: “GDP is growing 2.3%. Manufacturing is recovering. Capex is at record levels. Soft landing achieved.”

What is actually happening: The headline improvement is real but uneven. The January ISM surge may reflect tariff-related front-running rather than organic demand. Capex is strong, but labor input growth is tepid and housing is in recession. Services and government are doing the heavy lifting. Inventory dynamics can flip from tailwind to headwind quickly.

Consensus gets trapped by three biases.

Aggregation bias. GDP combines components moving in opposite directions into a single number. A reading of +2.3% can mask housing at -5% and services at +3.5%. The average tells you nothing about the underlying dynamics. The economy can be “fine” in aggregate while critical sectors are already in recession.

Lag bias. GDP is released 30 days after the quarter ends. By the time you see Q4 GDP in late January, you are six weeks into Q1. Meanwhile, monthly indicators (IP, retail sales, PMIs) have already shown you what January looks like. Trading GDP data means trading stale information.

Level bias. Positive GDP growth sounds reassuring. But the level matters less than the trajectory. A single month of ISM expansion after nearly three years of mostly contraction is not “recovery.” It is a data point that needs confirmation. One month does not make a trend. Three months starts to.

The “headline looks fine, composition is uneven” trap catches consensus every cycle because humans anchor to single numbers. The nuance requires more work.

Where We Are Now

Applying the framework to current conditions.

The headline metrics look solid. Real GDP is running at +2.3% year-over-year. The Atlanta Fed GDPNow is tracking Q4 2025 at 4.2% SAAR. At the aggregate level, the expansion is accelerating.

The recent data shows improvement, but with important context.

Manufacturing just snapped back. ISM Manufacturing surged to 52.6 in January 2026, the first expansion reading in 12 months. Before that, manufacturing managed only a brief two-month expansion interrupting what has been a brutal stretch: 36 of the last 40 months spent in contraction. New Orders jumped to 57.1. Industrial production is running at +2.0% year-over-year, with manufacturing specifically at +2.1%. The manufacturing recession appears to be ending.

But context matters. ISM Chair Susan Spence noted that January is a seasonal reorder month after holidays, and some buying appears to be front-running expected price increases from tariff uncertainty. Whether this reflects genuine demand recovery or inventory positioning ahead of trade policy changes remains to be seen. One month does not make a trend. We need 3+ months of expansion to confirm the turn.

Business investment is a bright spot. Core capital goods orders are running at +5.3% year-over-year, hitting a record $78.2 billion in November 2025. CEOs are committing real capital. Watch whether this confidence holds through tariff uncertainty or whether January’s ISM improvement feeds through to even stronger equipment orders in the coming months.

Labor input is growing slowly. Private employment growth has decelerated to roughly +0.5% YoY with flat average weekly hours, putting aggregate hours growth below +1%. That is well below the 1.5%+ pace typical of healthy expansions. Growth is positive but tepid.

Housing remains weak. Starts are down -7.8% year-over-year (October data, latest available due to Census release delays), down from the April 2022 peak. Residential investment has been contracting for over two years.

What is supporting growth?

Services (ISM Services at 53.8) remain in expansion. Services are 80% of the economy. As long as services hold, the headline stays positive.

Figure 7: Goods vs Services Consumption YoY. Goods turn first at cycle turns; services follow.

Consumers (real retail sales +0.4% year-over-year) are hanging on. Barely. That +0.4% is the difference between nominal sales growth and inflation. Consumers are spending more dollars to buy roughly the same amount of stuff.

Figure 8: Real Retail Sales YoY. Adjusting for inflation, consumer spending is barely positive.

Government spending is additive. Fiscal policy remains expansionary, supporting headline GDP.

Figure 9: GDP Component Contributions (YoY). PCE (+1.8 pp) drives growth. Net exports (+0.5 pp) added modestly.

The chart shows YoY contributions to GDP growth (total: +2.4 pp). PCE added +1.8 percentage points. Net exports contributed +0.5 pp. Government added +0.2 pp. GPDI was flat (-0.0 pp). Note: This reflects Q3 2025 data. The Q4 2025 Advance Estimate has been delayed until February 20 due to the government shutdown, making monthly trackers like the GCI even more critical for nowcasting.

Where does our composite stand? The Growth Composite Index (GCI) is hovering near neutral territory. The January ISM surge, if sustained, would push GCI into modest expansion. But one month of manufacturing improvement after nearly three years of mostly contraction does not erase the structural challenges in housing and labor input. Confirmation is required.

Figure 10: The Growth Composite Index (GCI) with regime bands. Synthesizes IP, capex orders, hours, housing, and retail into one indicator.

The Cross-Pillar Connection

Growth does not exist in isolation. It connects to every other pillar in our framework.

Growth to Labor: When output contracts, labor demand follows. IP contracting and hours shrinking precede payroll declines by 2-4 months. Current configuration: IP is positive at +2.0%, and aggregate hours growth is positive but tepid, running below +1% YoY.

Growth to Prices: When growth slows, the output gap widens. Slack builds. Pricing power fades. Capacity utilization at 76.3% (below the ~80% long-run average) signals slack in the system. Goods prices remain subdued relative to services. If the manufacturing recovery fades, pricing pressure on the goods side weakens further.

Growth to Credit: When growth weakens, revenues decline, earnings miss, and credit quality deteriorates. Spreads should widen to reflect higher default risk. Current paradox: growth is mixed but high-yield spreads remain tight. Credit is pricing a continuation of expansion. If the manufacturing improvement is real and capex holds, credit may be right. If either fades, spreads are too tight.

Growth to Fed Policy: When growth slows, the Fed should cut. But the Fed is constrained by elevated inflation. Rate cuts are delayed even as parts of the economy soften. The transmission mechanism is impaired.

The cross-pillar signals paint a nuanced picture. The bright spots are real: manufacturing inflecting, capex at record levels, IP positive. But labor fragility (LFI +0.93) and thin liquidity (LCI -0.8) remain elevated. The question is whether the positive momentum in production and investment is enough to stabilize the labor and housing weakness, or whether those drags eventually pull the expansion apart. History says that tension resolves one way or the other within 6-12 months.

How to Track This Pillar

Industrial Production YoY. The monthly GDP proxy. Released mid-month (~15th) for the prior month. Contraction below -1% signals manufacturing recession.

ISM Manufacturing PMI. The forward-looking survey. Released 1st business day of each month. Watch the 50 level and the New Orders minus Inventories spread. (Source: ISM, first business day)

Core Capital Goods Orders YoY. CEO confidence in real dollars. Released ~26th of month. Contraction below -5% signals capex collapse.

Aggregate Hours Worked YoY. Labor input to output. Released with employment situation (~first Friday). Contraction below 0% signals labor input shrinking.

Housing Starts YoY. The long lead. Released ~17th of month. Contraction below -10% confirms housing recession.

GDPNow. Real-time GDP tracking from the Atlanta Fed. Updates with each major data release. Compare to consensus to gauge whether data is surprising up or down. (atlantafed.org/cqer/research/gdpnow)

Release schedule strategy: Use early-month data (ISM on the 1st) to set expectations. Use mid-month data (IP, housing on the 15th-17th) to refine. Use late-month data (capex, GDP on the 26th-30th) to confirm.

Invalidation Criteria

Every thesis needs an exit door.

Bull Case (Growth Reaccelerating) Invalidation:

If the following occur simultaneously for 3+ months, the contraction risk thesis is wrong:

• Industrial Production YoY sustains above +2%

• ISM Manufacturing exceeds 52 (expansion confirmed)

• Core Capital Goods Orders YoY sustains above +5%

• Aggregate Hours YoY exceeds +1.5% (labor input expanding)

• Housing Starts YoY turns positive (+5%+)

• GCI exceeds +0.3 (neutral regime)

Action if invalidated: Rotate from defensive to cyclical. Increase equity allocation to 60-65%. Add exposure to industrials, materials, discretionary.

Note: As of today, IP, ISM, and capex are already meeting or approaching these thresholds. The key missing pieces are housing and labor input. If those follow, the expansion is broadening and the bull case is intact. We are watching.

Bear Case (Full Recession) Confirmation:

If the following occur, growth weakness is accelerating beyond contraction risk into recession:

• Real GDP turns negative for 2 consecutive quarters

• Industrial Production YoY exceeds -2% (severe contraction)

• ISM Manufacturing drops below 45 (deep contraction)

• Payrolls 3-month average turns negative (job losses)

• GCI drops below -1.0 (recession regime confirmed)

Action if confirmed: Maximum defensive posture. Reduce equity to 30-40%. Overweight bonds, gold, cash. Avoid all cyclical exposure.

Framework drives positioning, but the framework can be wrong. Data determines outcome.

The Bottom Line

Growth is not GDP. It is the velocity of economic metabolism. The speed at which the system converts inputs (labor, capital, innovation) into outputs (goods, services, income).

After spending 36 of the last 40 months in contraction, ISM Manufacturing just surged to 52.6 with New Orders at 57.1. Industrial production is positive at +2.0% YoY. Core capital goods orders are at a record +5.3% YoY. The goods-producing economy may be inflecting.

But skepticism is still warranted on the durability:

• The ISM surge may reflect tariff front-running rather than genuine demand recovery

• Labor input growth is tepid (aggregate hours below +1% YoY, well under the 1.5%+ expansion norm)

• Housing is still in recession (starts down 31% from peak)

• The manufacturing improvement is one month old after nearly three years of mostly contraction

The key insight remains: the second derivative matters more than the level. One month of improvement does not make a trend. We need 3+ consecutive months of expansion to confirm manufacturing has turned. Watch whether January’s ISM surge sustains into February and March, or fades as tariff-related distortions work through.

The data is improving. Some of it meaningfully so. The sustainability is unproven. Framework drives positioning, not headlines.

This is how we analyze growth.

Join The Watch.

Bob Sheehan, CFA, CMT

Founder & CIO, Lighthouse Macro

This is the third in a 12-part series on the Lighthouse Macro framework. Next up: Pillar 4 (Housing) and the Wealth Effect Transmission.

MACRO, ILLUMINATED.

GDP can disappoint and the Fed shrugs. Employment can soften and they wait for confirmation. But if inflation stays elevated, the Fed stays restrictive. There is no discretion here. Inflation is the only macro variable that directly controls the policy rate, and the policy rate controls everything else.

The transmission chain runs in one direction:

Prices → Fed Policy → Real Rates → Financial Conditions → Asset Prices → Growth → Employment

Get the inflation call right, and you’ve triangulated the Fed’s reaction function. Miss it, and you’re trading narratives about “transitory” nonsense while the Fed tells you exactly what it’s going to do.

This is why prices sit at the center of the monetary transmission mechanism. Not because inflation is one indicator among many, but because it is the binding constraint. Understanding inflation composition means understanding what the Fed can and cannot do.

The Core Insight: Composition Over Level

Here is the conceptual unlock that separates useful analysis from headline-watching.

Inflation data splits into three fundamentally different economies, each with different drivers, different transmission speeds, and different implications for policy.

Goods are globally traded, supply-chain responsive, and volatile. When goods inflate, it’s fast. When they deflate, it’s fast. Goods inflation is the shock absorber of the CPI basket.

Services are locally produced, wage-driven, and structurally sticky. You can’t offshore a haircut. Labor costs are 60-70% of service sector expenses, and wages don’t deflate without unemployment rising substantially. Services inflation is the structural floor.

Shelter is the lagging anchor. Measured via lease renewals and owners’ equivalent rent, it represents 34% of the CPI basket but operates on a 12-18 month lag relative to market conditions. It distorts the headline in both directions.

Why does this matter? Because a headline reading of 2.7% tells you nothing about the underlying regime. A uniform 2.7% is structurally different from +1.4% goods + 3.0% services averaging to 2.7%. The first is symmetric disinflation. The second is bifurcation with a structural wage floor preventing the last mile. Fed policy responds to composition, not averages.

The pattern repeats across every cycle since 1980: goods are the shock absorber, services are the structural floor, shelter is the lagging confirmation. In 2008, goods deflated violently while services stayed positive until unemployment hit 10%. In 2021-2023, goods peaked 13 months before services. Composition tells you where you’re going. The headline tells you where you’ve been.

What to Watch and Why

We approach inflation analysis through three lenses.

Leading indicators move first. Producer prices signal what manufacturers are paying before those costs reach consumers. Market rents signal where CPI shelter will be 12 months from today. The dollar signals where goods prices are heading.

Persistence indicators reveal whether inflation pressure is broad-based or concentrated. The Atlanta Fed’s sticky CPI separates components that change infrequently from those that change rapidly. The Dallas Fed’s trimmed mean strips the most extreme price changes to reveal the underlying trend.

Confirming indicators validate what the leading data suggested. Core PCE is the Fed’s actual target. Services ex-shelter (supercore) is the wage-driven core the Fed watches most closely.

The transmission speed hierarchy matters: goods respond in months, services adjust over quarters, shelter takes over a year. That lag between “goods are fixed” and “services are still stuck” is where consensus gets trapped every cycle.

Direction changes often matter more than levels. The 3-month annualized rate compared to the 12-month rate reveals momentum. When 3M runs below 12M, the trend is improving. When it runs above, pressure is building. Watch the slope, not just the spot.

The Indicators That Matter

The Gap That Matters: CPI vs Core PCE

The Fed doesn’t target CPI. It targets Core PCE. This distinction is not academic. It’s mechanical.

CPI uses fixed weights based on what people bought last year. PCE uses chain-weighted spending based on what people are buying now. PCE also captures a broader set of expenditures, including employer-provided healthcare. The practical result: PCE tends to run 0.3-0.5 percentage points below CPI because it accounts for substitution effects. When beef gets expensive, people buy chicken. CPI misses that behavioral shift. PCE captures it.

Shelter is 34% of CPI but only 18% of PCE. This weight difference means CPI overstates the shelter problem relative to what the Fed actually watches. When shelter normalizes, CPI will drop faster than PCE, and the gap between “headline victory” and “Fed reality” will temporarily widen. Track Core PCE if you want to know what the Fed is thinking.

Figure 2: Headline CPI vs. Core PCE. The Fed watches Core PCE, and it’s still meaningfully above target.

The Great Divergence: Goods vs Services

Goods inflation is volatile and self-correcting. Services inflation is sticky and persistent. This divergence defines the current phase of the cycle.

Why goods self-correct: globally integrated markets with elastic supply. When demand fades, inventory clears, supply chains heal, the dollar compresses import prices. Why services persist: local markets with inelastic supply. The primary input is labor, and labor costs don’t adjust downward without unemployment rising.

The 2021-2023 episode confirmed this. The Global Supply Chain Pressure Index normalized by mid-2022, yet core inflation persisted through 2023. If supply disruptions were the root cause, inflation would have collapsed when supply healed. It didn’t. Demand was doing the heavy lifting.

The goods-services spread tracks the cycle. When goods deflate while services stay elevated, headline improvement masks structural stickiness. Does convergence happen through goods re-inflating (tariffs, dollar weakness) or services decelerating (labor cracking)? That distinction is the call.

Figure 1: Core goods CPI vs. core services CPI. The divergence is the defining feature of the current inflation regime.

The Shelter Lag Trap

Shelter is 34% of the CPI basket. And it operates on a mechanical lag that distorts the headline in both directions.

The BLS measures shelter through surveys of existing leases, not new leases. When market rents spike, it takes 12-18 months for those increases to flow through the lease renewal cycle into CPI. We tested the lag against every turning point since 2015. Average: 12.7 months. The pattern is mechanical and predictable.

The trap: even after shelter normalizes, services ex-shelter remains sticky. Shelter was masking the problem, not causing it. When shelter cooperates, supercore becomes the binding constraint. And supercore is where the wage floor lives.

Figure 3: Shelter CPI, Rent of Primary Residence, and Owners’ Equivalent Rent. All three declining but remain above target.

Sticky vs Flexible: The Persistence Signal

The Atlanta Fed decomposes CPI into two categories. Flexible CPI includes items that change frequently: gasoline, food, airfares. Sticky CPI includes items that change infrequently: rent, insurance, medical care, education. These embed expectations and wage costs.

Sticky CPI is the best available predictor of future core inflation, with more than twice the predictive power of flexible CPI over a six-month horizon. What happens in flexible prices today flows into sticky prices roughly 12 months later.

Historically, sticky CPI has never broken below 3.0% without a recession. That threshold is the structural marker. Until it moves decisively lower, the last mile remains incomplete.

Figure 4: Sticky CPI vs. Flexible CPI (shifted 12 months). Flexible inflation leads sticky by roughly 12 months.

The Pipeline: PPI Leads CPI

Producer prices are the upstream signal. What manufacturers and service providers pay eventually passes through to consumers. PPI leads CPI by 3-6 months. When producers face cost increases, they absorb what they can, then pass the rest through. The lag reflects the adjustment process: inventory drawdowns, contract renegotiations, menu cost frictions.

The PPI-CPI spread is the directional signal. When PPI runs below CPI, producers are absorbing cost declines that haven’t yet reached consumers. That’s disinflationary pressure building in the pipeline. When PPI runs above CPI, inflationary pressure is coming. The spread tells you what’s in the pipe before it hits the consumer.

Figure 5: PPI Final Demand vs. CPI. Producer prices lead consumer prices by 3-6 months.

Inflation Expectations: The Anchoring Test

If businesses and consumers expect inflation to remain elevated, they price accordingly. The Fed’s credibility depends on keeping expectations anchored at 2%.

5Y5Y Forward captures the bond market’s view of where inflation will be 5-10 years from now. When it’s near 2.0%, the market trusts the Fed. Above 2.5%, trust is eroding. Above 3.0%, the Fed has lost the anchor.

UMich 1Y Expectations captures what households feel. Consumers don’t watch Core PCE. They watch grocery bills and insurance premiums. Consumer expectations can stay elevated even when professional expectations are calm. That gap reveals the bifurcation in lived experience.

Figure 6: The 5-Year, 5-Year Forward Inflation Rate vs. University of Michigan 1-Year Consumer Expectations.

The Signal Beneath the Noise: Trimmed Mean

Standard inflation measures get distorted by outliers. Motor vehicle insurance surging 14%+ tells you nothing about the underlying trend, but because it’s roughly 3% of CPI, it contributes about 0.4 percentage points to headline.

The Dallas Fed Trimmed Mean PCE strips the most extreme price changes each month to reveal what’s actually happening underneath. It produces meaningfully lower forecast errors than headline CPI at predicting core PCE twelve months forward.

When the trimmed mean and sticky CPI agree, the signal is high-confidence. Track both. Trust the convergence.

Figure 7: Dallas Fed Trimmed Mean PCE vs. Core PCE. The trimmed mean strips outlier noise and confirms the underlying trend.

The Spiral Check: Wages and Unit Labor Costs

The wage-price spiral is the inflation scenario that keeps central bankers awake. Workers demand higher wages because prices are rising. Businesses raise prices to cover higher labor costs. Repeat.

The Employment Cost Index is the gold standard for measuring labor cost pressure because it controls for compositional shifts. Average Hourly Earnings can fall just because the economy is adding more low-wage jobs. That looks like wage disinflation but it’s actually mix shift. ECI holds the job mix constant. If ECI is elevated, wage pressure is real.

Unit labor costs equal compensation growth minus productivity growth. At 3.5% compensation and 1.5% productivity, ULC runs at roughly 2.0%. That’s consistent with services inflation in the 2.5-3.0% range. Not crisis. Not target. Stuck.

The Phillips curve, dormant through the 2010s, has reasserted. At current unemployment near 4.4%, it suggests services inflation consistent with 3-4%, not 2%. For services inflation to reach 2%, unemployment likely needs to rise toward 4.5-5.0%. This is the narrow path the Fed is attempting to navigate.

From our labor framework (Pillar 1): the Labor Fragility Index is elevated and quits have fallen to the 2.0% threshold. If labor continues to soften, wage growth will decelerate, and that flows into services inflation with a 6-9 month lag. The pillars are connected through the wage channel.

Figure 8: ECI Total Compensation vs. Core PCE. Wages above inflation means positive real wages but sticky services.

The Dollar Channel

The trade-weighted dollar is the mechanism behind goods price suppression. Dollar strength flows directly into goods prices with a 9-18 month lag. A 10% appreciation typically results in 1.0-1.5 percentage points of reduction in goods CPI.

The implication cuts both ways. If the dollar weakens in a rate-cutting cycle, goods inflation reaccelerates. The Fed’s own easing could reignite goods inflation, constraining how aggressively they can cut. There’s also a tariff channel: a broad 10% tariff could add 0.3-0.5 percentage points to headline CPI over 6-12 months.

Figure 9: Trade-weighted dollar (inverted) vs. Core Goods CPI, shifted 18 months to show the lead relationship.

The Prices Composite Index

We synthesize the key inflation signals into a composite we call the Prices Composite Index (PCI). The purpose is regime classification: is the inflation environment one where the Fed has flexibility, or one where the Fed is constrained?

The PCI combines six components, each measured against a Fed-consistent target level:

• Core PCE momentum (3-month annualized): The Fed’s primary target, highest weight

• Services inflation trend: Wage-driven persistence

• Shelter trajectory: Largest single CPI component, lagging but mechanical

• Sticky price persistence: Atlanta Fed decomposition, predicts future core

• Expectations anchoring (5Y5Y forward): The credibility gauge

• Goods price level: Disinflationary force when negative

PCI above +0.5 means the Fed is constrained. PCI above +1.0 means no cuts possible. When multiple inflation indicators remain elevated simultaneously, the composite stays in restrictive territory, and the Fed stays boxed in.

Figure 10: The Prices Composite Index (PCI). Regime bands show the inflation environment.

The Consensus Trap

Here is the pattern that repeats every cycle.

Surface narrative: “Inflation is beaten. Headline CPI has fallen from 9.1% to 2.7%. Soft landing achieved.”

What’s actually happening: Goods have normalized and are no longer pulling headline lower. Services remain elevated with a wage floor intact. Sticky CPI sits at 1.5x the target. The easy disinflation is done.

Consensus gets trapped by three biases.

Recency bias. Inflation fell from 9.1% to 2.7% in roughly two years. Extrapolating that pace suggests sub-2% soon. But the disinflation was driven almost entirely by goods reversal and energy normalization. Those forces are spent. The remaining improvement has to come from services, and services don’t deflate without labor market deterioration.

Headline focus. Media covers headline CPI. Markets trade headline CPI. The Fed watches core PCE. When headline and core diverge, consensus misses the Fed’s constraint.

Lag misunderstanding. As shelter continues to normalize, headline will improve. But the market will shift to watching services ex-shelter, which remains elevated. The goal post moves. When shelter finally cooperates, supercore becomes the binding constraint.

Where We Are Now

The surface metrics look constructive. Headline CPI at 2.7%, down from the 9.1% peak. Core goods at +1.4%, near flat. Core PCE 3-month annualized at 2.3%, trending in the right direction.

The persistence indicators tell a different story. Core PCE at 2.8%, still 40% above target. Sticky CPI at 3.0%, right at the threshold that historically requires recession to break. Trimmed Mean PCE at 2.5%, converging but still above target. ECI at 3.6%.

The pipeline has flipped. PPI at 3.0% now runs above CPI at 2.7%, suggesting inflationary pressure still in the pipe. Expectations remain bifurcated: 5Y5Y forward at 2.19% (well-anchored) while consumer expectations stay elevated.

The PCI currently sits in the Elevated regime. Shelter is unwinding. Goods have normalized. But services ex-shelter, sticky CPI, and wages remain elevated enough to keep the composite above the +0.5 threshold.

The headline may approach 2% eventually. The question is whether it gets there through genuine services moderation or through demand destruction.

How to Track This Pillar

Core PCE YoY. The Fed’s actual target. Sustained move below 2.5% signals the last mile is breaking. (Monthly, BEA, FRED: PCEPILFE)

Trimmed Mean PCE 12M. The cleanest trend signal. Convergence toward 2% confirms broad-based disinflation. (Monthly, Dallas Fed, FRED: PCETRIM12M159SFRBDAL)

Sticky CPI. The persistence gauge. Break below 3.0% historically requires recession. (Monthly, Atlanta Fed, FRED: CORESTICKM159SFRBATL)

Shelter CPI. The lagging anchor, 34% of headline. Compare to Zillow/Apartment List for the 12-month forward signal. (Monthly, BLS, FRED: CUSR0000SAH1)

PPI Final Demand. The upstream pipeline signal. PPI above CPI means pressure is coming. (Monthly, BLS, FRED: PPIFIS)

5Y5Y Forward Inflation. The expectations anchor. Above 2.5% is caution. Above 3.0% is alarm. (Daily, FRED: T5YIFR)

ECI Total Compensation. The gold standard wage measure. Deceleration below 3.0% is consistent with 2% inflation. (Quarterly, BLS, FRED: ECIALLCIV)

Release schedule: CPI mid-month (~13th), PCE end-of-month (~30th), PPI mid-month (typically one day before CPI). Use CPI for real-time signals, PCE for the Fed’s reaction function.

Invalidation Criteria

Every thesis needs an exit door.

Bull Case: If any three of the following trigger for three consecutive months, the elevated inflation regime ends:

• Core PCE 3M annualized drops below 2.5%

• Services ex-shelter drops below 3.0%

• Sticky CPI drops below 3.0%

• 5Y5Y forward drops below 2.3%

• PCI drops below +0.5

Bear Case: If any two of the following trigger, inflation is reaccelerating:

• Core PCE 3M annualized exceeds 4.0%

• Goods CPI accelerates above 3.0%

• 5Y5Y forward exceeds 2.75%

• Services ex-shelter exceeds 4.5%

Framework drives positioning, but the framework can be wrong. Data determines outcome.

The Bottom Line

Inflation is not a number. It is a regime. And the regime has changed.

Pre-pandemic, 2.0% PCE was the equilibrium. That world required globalization tailwinds, subdued wages, and fiscal austerity. All three have reversed. Five structural forces, deglobalization, energy transition, demographics, housing underbuilding, and fiscal deficits, suggest a floor of 2.5-3.0%.

Watch composition, not the headline. Track the PCI regime bands. When PCI is above +0.5, the Fed is constrained. When sticky CPI is at or above 3.0%, the persistence problem hasn’t been solved. When services ex-shelter stays above 3.0%, the last mile is incomplete.

This is how we analyze inflation.

Bob Sheehan, CFA, CMT

Founder & CIO, Lighthouse Macro

This is the second in a 12-part series on the Lighthouse Macro framework. Next up: Pillar 3 (Growth) and the Second Derivative.

People make things, build things, produce things. It is the literal foundation on which everything else rests.

When we talk about economic cycles, we are really talking about labor cycles dressed up in different clothes. The transmission chain runs in one direction: employment leads to income, income enables spending, spending generates revenue, revenue justifies credit, credit supports housing, and housing anchors household wealth. Change the labor picture, and everything downstream changes with it.

If the engine sputters, the car eventually stops. It does not matter how much gas is in the tank or how smooth the road is. Mechanics dictate outcome.

This is why labor sits at the center of our framework. Not because it is one indicator among many, but because it is the source code. Understanding labor means understanding the economy.

The Core Insight: Flows vs. Stocks

Here is the conceptual unlock that separates useful analysis from headline-watching.

Labor data splits into two categories: flows and stocks. The distinction matters more than almost anything else we will discuss.

Stocks measure the cumulative result of past decisions. The unemployment rate. Total payrolls. These are the numbers you see in headlines, the figures politicians cite, the data points that anchor consensus narratives. They tell you where the labor market has been.

Flows measure what is happening at the margin right now. How many people are quitting their jobs. How many are being hired. How temp agencies are adjusting their staffing. Flows tell you where the labor market is going.

Why do flows lead stocks? Because decisions happen before outcomes.

When employers sense trouble on the horizon, they do not fire workers immediately. They freeze hiring. They let headcount decline through attrition. They stop bringing on temps. The actual layoffs come later, sometimes much later, because firing people is expensive, disruptive, and hard to reverse. Companies remember 2021 and 2022 when they could not find workers at any price. That scar tissue makes them hoard labor longer than they probably should.

The same dynamic operates on the worker side. When employees sense the labor market tightening, they do not wait to get laid off. They stop quitting. They hold onto the job they have rather than chasing a better one. The quits rate is the economy’s truth serum. It strips away narrative and exposes what workers actually believe about their options. And workers sense the shift before management admits it.

The early signals show up in flows. The headlines focus on stocks. That gap is where consensus gets caught.

This chart makes the relationship visible. The quits rate turns down six to nine months before unemployment turns up. The pattern has held across every recession since JOLTS data began in 2000. When workers stop quitting, unemployment eventually rises. The lag is predictable. The sequence does not fail.

What to Watch and Why

We approach labor analysis through four lenses. Not a checklist, but a mental model.

Leading indicators move first. They are the canaries. These capture decisions being made today that will show up in headline data months from now. When companies stop hiring temps, when workers stop quitting, when the hiring rate begins to slide even as unemployment stays low, these shifts register in real time. Their implications will not become obvious until the lagging data catches up.

Breadth indicators reveal whether weakness is localized or spreading. A single sector struggling is not necessarily alarming. But when stress starts showing up across multiple states, across different industries, across both manufacturing and services, that is a different signal entirely. Breadth is how you distinguish between noise and regime change.

Confirming indicators validate what the leading data suggested. When unemployment starts rising, it is not telling you something new. It is confirming what the flows already showed months earlier. But that confirmation matters because it eliminates the possibility that the early warnings were false positives.

Direction changes often matter more than levels. This is counterintuitive but critical. An indicator’s turn can be leading even when its level is lagging. The unemployment rate is the classic example. It is called a “lagging indicator” because its level peaks after recession ends. But its trough, the point where it stops falling and starts rising, actually precedes recession by several months. Watch the turn, not just the level.

We have validated these relationships against every post-war recession. They hold. The lag times vary, but the sequence does not.

The Indicators That Matter

The key signals, what they measure, and why they work.

The Quits Rate measures voluntary separations as a percentage of total employment. When a worker quits, they are making a statement: “I believe I can find something better.” This is a real-time confidence barometer. Unlike surveys that ask people how they feel, the quits rate captures what they actually do.

The quits rate is arguably the single most important flow indicator in labor market analysis. Workers have asymmetric information about their own situations. They sense weakness in their company, their industry, their local market before it shows up in aggregate data. When they stop quitting, they are telling us something management has not admitted yet.

Historical validation: The quits rate has declined from cycle peak before all three post-2000 recessions with lead times of 6-9 months. Current reading sits at 2.0%, down from a peak of 3.0% in April 2022. That is a one-third decline from cycle high. Workers have already adjusted their behavior.

Temporary Help Services employment measures workers employed through temporary staffing agencies. This is the canary in the coal mine with an exceptional track record.

Why it works: Temp workers are the adjustment valve for corporate labor demand. When companies sense softening demand, they do not immediately fire permanent employees. That is expensive, disruptive, hard to reverse. They stop renewing temp contracts. Temp help is the first to go and the first to come back.

Historical validation: Temp help employment YoY has gone negative before all three post-2000 recessions with lead times ranging from 1 to 12 months. Current reading shows YoY contraction of approximately 3-4%, persisting for over two years.

A caveat worth addressing directly: recent academic research (Tito and Bowdle, 2024) found that pandemic-era labor shortages boosted temp demand by 25-85%, and the subsequent easing reduced temp employment by 5-20%. That is a meaningful distortion. But even after accounting for pandemic-era overshoot, the contraction exceeds what normalization alone would explain. Since March 2022, temp employment has fallen by 577,000 jobs (18.1%), the largest decline outside a recession in the indicator’s modern history. The signal may be noisier than pre-2020, but it has not lost its directional value. We weight it accordingly: lower than before, but still part of the mosaic.

The Hiring Rate shows new hiring as a percentage of employment. When paired with the layoffs rate, it reveals the “hiring freeze without firing wave” phenomenon that characterizes late-cycle labor markets.

Current configuration: The hires rate sits at 3.2-3.3%, down from 4.5% at the 2022 peak. The layoffs rate remains at 1.0-1.1%, historically low. Companies are not yet cutting workers because they remember how hard it was to hire in 2021-2022. But they have stopped adding headcount. The labor market is frozen, waiting for the break.

This can persist for extended periods, but history says when it breaks, it tends to break toward layoffs rather than renewed hiring.

Aggregate Hours Worked captures something payrolls miss entirely. Before companies lay off workers, they reduce hours. Before they reduce hours on paper, they cut overtime. This is the most reversible lever employers have, and reversible levers get pulled first.

The chart shows aggregate hours (YoY) plotted against employment (YoY). Hours consistently turn down before employment at cycle inflection points. The logic is straightforward: cutting hours costs nothing. Cutting workers costs severance, institutional knowledge, and the risk that you cannot rehire when conditions improve. So hours absorb the initial shock. When hours decline while employment holds steady, that divergence is a warning. The economy is weakening at the intensive margin (fewer hours per worker) before it weakens at the extensive margin (fewer workers).

Current configuration: aggregate hours growth has decelerated sharply from post-pandemic highs and is now oscillating near zero. Employment growth has followed the same trajectory but with a lag. The pattern is consistent with prior late-cycle behavior.

State Labor Market Diffusion measures the percentage of state-level labor markets experiencing rising unemployment over a three-month window. It captures geographic breadth of weakness.

Why it works: A single state struggling is not necessarily alarming. That could be idiosyncratic. But when weakness spreads across geographies, it reveals systemic stress. In recessions, this measure approaches 100%. Everyone hurts simultaneously.

Historical validation: State diffusion sustained above 50% has preceded every recession in the dataset with lead times of 2-14 months. Current reading hovers around 47-52%, oscillating near the majority threshold. Weakness is not universal, but it is not localized either.

The Job-Hopper Wage Premium is a microstructure signal most people miss. Throughout history, workers who switch jobs earn higher wage gains than those who stay put. That makes intuitive sense. You quit your current job when you have a better offer somewhere else. The premium varies with the cycle, widening when labor markets are tight and workers have options, narrowing when conditions deteriorate and staying put becomes the safer play.

What most people miss is that this premium can actually reverse. When job-switchers start earning less than job-stayers, something unusual is happening. Workers are taking new positions out of necessity rather than opportunity. They are moving laterally or even down, accepting whatever is available because their current situation became untenable.

Historical pattern: The premium collapsed to near zero around every recession since the Atlanta Fed Wage Growth Tracker began. Current reading sits at 0.1-0.2 percentage points, down from over 2 percentage points in 2021-2022. The grass is no longer greener. Workers know something.

Long-Term Unemployed Share measures the percentage of unemployed workers who have been jobless for 27 weeks or more. This is a lagging indicator by level but a confirming indicator by direction.

Why it matters: Long-duration unemployment has permanent effects. Skills atrophy. Networks fade. Employer bias against long gaps is real. Once someone crosses the six-month threshold, their probability of finding work drops significantly. Rising long-term share is early structural damage, the kind that leaves scars.

The turn in this indicator actually leads recessions by 3-13 months even though long-term unemployment peaks after recession ends. Current reading sits at 24-26%, up from cycle lows around 17-18%. More than one in four unemployed workers has been looking for over six months. That is meaningful deterioration even as the headline unemployment rate stays manageable.

The Consensus Trap

Here is the pattern that repeats every cycle.

The economy is late in an expansion. Headlines say “strong labor market” because unemployment is low and payrolls are still positive. Analysts point to the unemployment rate and conclude there is nothing to worry about. Markets price accordingly.

Meanwhile, under the surface, the flows are deteriorating. Temp help is contracting. The quits rate has fallen materially from its cycle highs. Hiring has slowed even as layoffs remain subdued. Duration of unemployment is extending for those who do lose jobs. Weakness is starting to spread geographically.

But because the stocks look fine, consensus dismisses the warning signs. “The labor market remains resilient” becomes the mantra. And it stays the mantra right up until the lagging indicators finally catch up and everyone collectively discovers what the leading indicators showed months earlier.

We call this the “stocks look fine, flows are deteriorating” trap. It catches consensus every cycle because humans anchor to what they can see in headlines. The flows require more work to track, more nuance to interpret. Most people do not do that work.

By the time unemployment is rising and payrolls are contracting, the information is already fully priced. The edge exists earlier, in the gap between leading and lagging.

Where We Are Now

Applying the framework to current conditions.

The surface metrics look unremarkable. Unemployment sits at 4.4%. Payrolls remain positive, adding 50,000 jobs in December. Initial claims have not spiked, hovering around 215-220k on a four-week average. If you only read headlines, you would conclude the labor market is holding up reasonably well.

The flow indicators tell a different story.

Temporary help employment has been contracting for over two years. The canary stopped singing long ago.

The quits rate sits at 2.0%, down one-third from its cycle peak. Workers are not quitting like they used to because they are not seeing the same opportunities they saw a year or two ago. When the quits rate peaked in 2022, workers had unprecedented bargaining power. That era has clearly passed.

Hiring has slowed considerably even as layoffs remain historically low. Companies are in a holding pattern: not adding headcount, not yet cutting it. This “hiring freeze without firing wave” configuration can persist for extended periods, but history suggests when it breaks, it tends to break toward layoffs rather than renewed hiring.

Duration of unemployment is extending. The share of unemployed workers who have been looking for six months or more has risen from 17-18% to 24-26%. This is early structural damage.

Breadth is broadening. State diffusion is oscillating around 50%, meaning roughly half of all state labor markets are experiencing rising unemployment. The weakness is not confined to one region or one industry. It is spreading.

The job-hopper premium has collapsed to near zero. Workers who switch jobs are earning the same as those who stay put. The Atlanta Fed Wage Growth Tracker shows the premium sits at 0.1-0.2 percentage points, down from over 2 percentage points during the Great Resignation. That means the grass is no longer greener. Workers sense the shift before it shows up in aggregate data.

Where do our composites stand? The Labor Pulse Index (LPI) currently reads approximately -0.4, placing us in the “softening” regime where fragility is developing and cyclical exposure should be reduced. The Labor Fragility Index (LFI) sits near +0.9, elevated and approaching the pre-recessionary threshold of +1.0. These are not recession readings. But they are not expansion readings either. The labor market is in transition, and the direction is not encouraging.

The question is not whether the flows are deteriorating. They are. The question is whether stocks follow. History says they do.

The Framework in Practice

We are not in the business of calling recessions. We are in the business of understanding probabilities and positioning accordingly.

What the labor data tells us right now is that the economy’s engine is running cooler than headlines suggest. The transmission chain we described at the outset, labor into income into spending into credit into housing, is experiencing friction at the source.

Does that mean recession is imminent? Not necessarily. These configurations can persist. Companies can hoard labor for quarters while waiting for visibility to improve. Workers can stay put while hoping conditions stabilize.

But it does mean the margin for error has narrowed. When flows deteriorate and stocks have not yet caught up, you are in a fragile state. Any incremental shock, policy mistake, financial stress, or exogenous event will transmit faster through an economy that is already showing early strain.

That is the kind of information you can act on, whether you are running a portfolio, making business decisions, or simply trying to understand what is actually happening beneath the noise.

How to Track This Pillar

The core indicators we track monthly:

JOLTS Quits Rate. The truth serum. Watch for sustained readings below 2.0%. This is the single most important flow indicator.

JOLTS Hires and Layoffs. Track both rates. The gap between them reveals whether we are in hiring freeze or firing wave territory.

Temporary Help Services Employment. The canary. Year-over-year contraction is the signal. Anything below negative 3% sustained starts flashing red.

Long-Term Unemployed Share. Calculate this as workers unemployed 27+ weeks divided by total unemployed. Rising share confirms flow deterioration is translating into structural damage.

State Labor Market Diffusion. Percentage of states with rising unemployment over a three-month window. Sustained readings above 50% signal systemic stress, not idiosyncratic weakness.

Industry Employment Diffusion. Same concept, different cut. What percentage of industries are adding jobs versus shedding them. We track both one-month and three-month diffusion. When the majority of industries contract simultaneously, breadth has broken.

Sector Quits Diffusion. Not just the aggregate quits rate, but how many sectors are seeing quits decline. Weakness concentrated in one or two sectors is rotation. Weakness across eight or ten sectors is regime change.

Cohort Analysis by Firm Size. Small businesses (under 50 employees) hire and fire first. Large firms (500+ employees) move later but with more momentum. When small business employment peaks and rolls over while large firms still expand, that divergence is an early warning. The small guys see it first.

Cohort Analysis by Demographics. Prime-age workers (25-54) are the core. But younger workers (16-24) and older workers (55+) behave differently across the cycle. Youth employment is more volatile, turning sooner. We watch for peak-to-trough timing across cohorts to understand cycle phase.

Atlanta Fed Wage Growth Tracker. Median wage growth for job switchers versus job stayers. The premium collapse reveals what workers actually believe about their options. When the job-hopper premium drops below 0.5 percentage points, that is a late-cycle signal. When it goes negative, as it has in recent readings, that has historically only occurred around recessions.

Initial Jobless Claims. Weekly data, four-week moving average. This is our high-frequency pulse check. Sustained break above 250k confirms what monthly data suggested.

Continued Claims. How many workers remain on unemployment benefits week after week. Rising continued claims while initial claims stay flat means duration is extending. People are filing but not finding work quickly.

The release schedule: JOLTS arrives monthly with a 40-day lag, so November data drops in early January. Employment Situation (payrolls, unemployment, demographics) releases first Friday of each month. Jobless claims report weekly every Thursday morning. Wage tracker updates monthly.

We do not wait for the data to tell us what happened. We use the data to understand what is happening now and what is likely to happen next. Flows before stocks. Breadth alongside direction. Turns over levels.

Invalidation Criteria

Every thesis needs an exit door. Here is what would make us wrong about labor market fragility.

The bearish labor thesis is invalidated if we see three consecutive months of:

Quits Rate rising above 2.1%. Workers regaining confidence, chasing better opportunities again.

Hires Rate rising above 3.8%. Companies competing for talent, not just holding pattern.

Long-Term Unemployed Share dropping below 20%. Quick reabsorption of job seekers, no structural scarring.

Temp Help Employment YoY turning positive. Companies adding flexible labor, signaling confidence in demand.

Initial Claims four-week average dropping below 200k sustained. No layoff wave materializing.

If we see sustained improvement across these measures, the late-cycle stress signal is wrong. We rotate from defensive to cyclical and add risk. Framework drives positioning, but the framework can be wrong. Data determines outcome.

The Bottom Line

The labor market does not move in a straight line from expansion to contraction. It gives you signals. The question is whether you are watching the right ones.

We watch flows before stocks. We measure breadth alongside direction. We track the turns, not just the levels. And we remain humble about timing while staying confident about framework.

Labor is not a sector. It is the economy itself, measured in real time. There is no production without workers. No consumption without income. No credit cycle without employment.

The key insight remains: flows vs. stocks. Flows move first because workers and employers at the margin sense changes before they show up in aggregate data. Stocks lag because they are the cumulative result of many individual decisions.

When quits decline, workers are seeing something management has not admitted. When temp help contracts, employers are cutting the most flexible labor first. These are not forecasts of recession. They are the recession beginning, six to twelve months before NBER makes it official.

Current state: quits at 2.0% and oscillating near threshold. Temp help contracting for two years. Hiring frozen while layoffs stay low. Duration extending. Breadth widening. Premium collapsed. LPI at -0.4 (softening). LFI at +0.9 (elevated fragility).

The flows are flashing. The stocks are sleeping. One of them is wrong.

This is how we analyze the labor market.

Bob Sheehan, CFA, CMT

Founder & CIO, Lighthouse Macro

This is the first in a 12-part series on the Lighthouse Macro framework. Next up: Pillar 2 (Prices).