The Dawson LedgerMichael W.J. Dawson
Search

← Investing & judgement

Finance · A reading of the cycles

The expansion is in its middle years. The debt underneath is not.

Three clocks run at once: a loop of five to ten years, an arc of about eighty, and a far longer one for empires. This is where I think each stands today, and how sure I am.

Reading as of 9 October 2026Static snapshot, refreshed by hand; each input carries its own date

Educational. Not investment advice. Framework after Ray Dalio

A waterwheel beside a working harbour, with changing weather and autumn trees.
The water runs in a loop, and the weather changes without asking. AI-created editorial illustration for The Dawson Ledger, 2026.

What a cycle is

A loop with a memory

Most of what we are told about the economy arrives as a line: growth was 2.5 percent, so expect 2.3. I find it more useful to ask where we are in a loop, because that shapes what is likely next.

The loop runs on credit. Borrowing lets people spend more than they earn. That spending is someone else’s income, which lifts asset prices, which supports more borrowing, until the cost of carrying the debt becomes a burden and the loop turns the other way. Ray Dalio describes this plainly in How the Economic Machine Works, and it is the picture I keep returning to.

Three clocks run at once, each slower than the last.

  • Short5 to 10 yearsThe business cycle: recovery, expansion, tightening, recession. Credit conditions set the pace.
  • Long75 to 100 yearsThe debt cycle: leverage builds over decades, then something forces it back. The two sectors, private and public, need not move together.
  • Big200 to 300 yearsThe rise and fall of empires: education, invention and trade build the lead, debt and division wear it down.
“Mechanistic thinking asks where you are in a cycle, because that shapes what is likely next.”
Mechanistic Over Narrative, essay 32
“In How the Economic Machine Works, the economy is a system of transactions in which credit does most of the work.”
Right About Mexico, Wrong About the Market, essay 11

The three-cycle framework is after Ray Dalio. The stage rules, the thresholds and the scoring here are my own adaptation, not his.

The short cycle · 5 to 10 years

Mid expansion, but only just

The business-cycle loop, five regimes A loop that runs clockwise: Early Expansion, Mid Expansion, Late Expansion, Early Contraction, Deep Contraction, then back to Early Expansion. The reading today is Mid Expansion, with Early Expansion the next most likely. the short cycle Early Expansion Mid Expansion LateExpansion EarlyContraction DeepContraction we are here
The loop runs clockwise. The dashed ring marks the runner-up.
  1. Mid ExpansionThe long, steady middle. Growth is solid and nothing is yet strained.
  2. Late ExpansionSlack runs out. Prices, wages and borrowing costs climb.
  3. Early ContractionBorrowers start to struggle and credit tightens.
  4. Deep ContractionRecession. Policy turns to rescue and the loop resets.

The reading sits at the second of five regimes, with the first close behind.

Five regimes in a rowA five-segment bar. A marker sits on Mid Expansion; a dashed tick marks Early Expansion as the runner-up. The segments are drawn equal and carry no probability scale.Early ExpansionMid ExpansionLate ExpansionEarly ContractionDeep Contractionwe are hererunner-up

Computed from four inputs by rules I set myself. Segments are equal in width and are not a probability scale.

My best reading is Mid Expansion. It is a moderate-confidence one. Early Expansion is next. On the model’s scores for the five named regimes it is 42 against 31 out of 100; those scores show how far apart the regimes sit, not calibrated probabilities, and when the gap is that small I hold the label loosely. Read it as “somewhere in the healthy middle, drifting neither way”. Over the last three months it has not moved.

The reason is a calm set of signals: the yield curve is positive, joblessness is modest, credit spreads are tight, and a recession alarm built from unemployment is quiet. The one that disagrees is inflation, which is still well above where it would be in a clean recovery.

In this model, with the other inputs held fixed, Mid Expansion becomes Early Expansion if inflation falls to about 1.6%, or if the yield curve steepens to about one point. It heads for Early Contraction if the curve inverts to about minus 0.4.

The signals behind it

Four readings, and what each one means

I use four plain measures. Each is in the same place as the economy it describes: the price of time, the state of work, the price of things, and the price of risk.

The price of time

The yield curve

+0.44percentage points, 9 Oct

Ten-year Treasuries pay a little more than two-year ones.

Monthly average, last ten years. Dashed line is zero.

What this means

Lenders are being paid to wait, which is how a healthy economy normally looks. The curve inverted for much of 2022 to 2024, the usual early warning, and has since returned to normal. A positive curve says the alarm is off. It does not say the danger is gone.

Source: US Treasury via Federal Reserve (FRED, T10Y2Y).

The state of work

Jobs

4.2%unemployed, September

Modest, and a touch above early 2024 (3.9%).

US unemployment rate, last ten years. The 2020 spike is the pandemic.

What this means

Work is plentiful but not tight. A separate recession alarm built on how fast unemployment is rising (the Sahm rule) reads 0.0, where it would need 0.5 to sound. Recessions show up here late, so this is the signal I trust least for turning points.

Source: US Bureau of Labor Statistics via FRED (UNRATE, SAHMREALTIME).

The price of things

Inflation

3.4%a year, August

Well above the 2% most central banks aim for.

Consumer prices, year on year, last ten years. Dashed line is zero.

What this means

This is the reading holding the cycle back from Early Expansion. Inflation this high limits how far the central bank can cut, so the cheap-credit recovery stays out of reach. If it fell towards 2%, my reading would shift a step earlier in the loop.

Source: US Bureau of Labor Statistics via FRED (CPIAUCSL). Year-on-year change computed here.

The price of risk

Credit spreads

0.93percentage points, September

How much extra a riskier company pays over the government to borrow. Today that gap is among the narrowest in ten years.

Monthly, last ten years. The 2020 spike is the pandemic.

What this means

When lenders get nervous, this gap widens before almost anything else breaks, which is why it matters to the loop. Today it sits in the lowest 8% of the last ten years: lenders are relaxed about risk. That favours the earlier, healthier regimes, and it is the reading that pulls my answer towards Mid Expansion. Tight spreads can persist for years, so this is a reason for calm, not a reason to relax.

Source: Federal Reserve Board, Gilchrist–Zakrajšek credit spread (public domain). Published monthly, about four business days into the next month, so I treat it as one month late. public domain Cross-check, not scored: the US Treasury’s high-quality corporate bond yield minus the 10-year Treasury, 0.78 points in September.

Why it matters

Right about the mechanism, wrong about the outcome

In 1982 Ray Dalio saw that Latin America had borrowed more than it could repay, and he was right. Then he expected a depression, the Federal Reserve eased instead, and he lost nearly everything. I find it the most instructive story in modern investing, precisely because the analysis was good.

That is why a reading like this one states its doubts as plainly as its numbers. A cycle tells you where to look first: at credit, at what debt service is doing to spending, at what the central bank is permitted to do about it. It does not tell you the outcome. The question I would carry from Dalio is the one he asked himself afterwards.

How do I know I’m right?

Locating yourself in a cycle is a way to make judgement answerable. For how this reading is made, see Method and limits. For background on combining incomplete signals, see The State Estimator; for why a model comes second to a mental model, Mental Models Before Financial Models.

“Being right about a mechanism is not the same as being right about the outcome.”
Right About Mexico, Wrong About the Market, essay 11
“I would rather watch the machine than admire the narrative.”
Mechanistic Over Narrative, essay 32
“A credit state is not a temperature an instrument reads. It is a model’s interpretation, and a number does not remove the uncertainty.”
The State Estimator, essay 34

The long cycle · about 75 to 100 years

Private debt has fallen relative to GDP. Federal debt has not.

Debt builds for decades, then something forces it back. Households and firms sit on one arc, the government on another, and they need not be in the same place.

Private debt has fallen relative to GDP; federal debt reads Top, in a three-way tie with Reflation and Normalisation.

The long debt arc, with today’s positionA pale archetype curve of debt over time: it builds through early leveraging and a bubble, peaks at the top, then falls through deleveraging, reflation and normalisation. Federal debt is marked at the top. Household and company debt is marked at reflation, with a dashed ring around the neighbouring normalisation stage where the rules cannot separate them.Early leveragingBubbleTopDeleveragingReflationNormalisationFederal government: you are hereHouseholds and firms

Households and firms: Reflation, tied with NormalisationFederal government: Top, tied with Reflation and NormalisationThe pale curve is an archetype, not a forecast. Dashed rings are stages the rules cannot tell apart.

Federal debt, share of GDP

123%

Since 1966, quarterly. Latest: Q1 2026.

Higher than in 98% of the record I hold. Debt on this scale is not a crisis by itself. It is what leaves a government with fewer choices in the next one.

Interest, share of tax receipts

21%

About one dollar in five of federal receipts now goes to interest, which is high against the last sixty years, around the 76th percentile. The bill rises when rates stay high and old cheap debt rolls over.

Household and company debt, share of GDP

115%

Down about 13 points over five years. Private debt has fallen relative to GDP while the government borrowed: it grew more slowly than the economy rather than being repaid. That is the shape of a deleveraging that has mostly run its course. Stand-in series

I read the long cycle by sector, because a single national “debt cycle” hides the interesting thing. Households and companies have spent the last few years borrowing less against their incomes, while the weight of debt has moved to the public balance sheet.

The rules I use cannot always tell neighbouring stages apart. Private debt is Reflation tied with Normalisation. Federal debt is Top, tied with Reflation and Normalisation. I show those ties as ties. The honest summary is “high, stretched, and not yet forced to adjust”.

Where patterns repeat

The structure rhymes. The timing does not.

Since 1945, American expansions have lasted about 64 months on average and recessions about ten. That tells you the structure repeats without telling you when the next turn comes. The same is true on the longer clocks. What recurs is the sequence: credit eases, confidence builds, debt service bites, policy responds.

The interesting part, to me, is what survives each turn. When an empire ends, some of its institutions disappear and others are quietly adopted by whoever comes next. The same goes for a family, a currency or a lodge. The question is not whether something fell, but which part, and who kept the rest running.

“Turnings recur. Elites overextend. Institutions lose legitimacy. Social capital depletes.”
On the secular-cycle theorists, in The Institutions That Held the Line
“We should want to know which institution persisted, who maintained it and what changed in its use.”
What outlasts an empire
1982

Mexico could not service about $80 billion. Dalio expected a depression. The Federal Reserve eased and a long expansion followed.

Essay 11

2007

The story said subprime was contained. The credit signal said otherwise, and the recession would be dated from December 2007.

Essay 32

2020

Headlines were catastrophic and the Federal Reserve Board’s credit spread jumped from 2.0 points in February to 3.8 in March, its highest monthly reading since 2009. By August it was back near 1.9. Credit had told the story earlier.

Essay 32

The big cycle · 200 to 300 years · A dossier

The lead is intact. The margins are thinning.

Empires rise on education, invention, productivity and trade, and decline when the lead in those narrows while debts and internal quarrels grow. I follow eight of these determinants for the United States.

The United States still leads, and by a distance, in military spending and in the reach of its financial markets. But on six of the seven determinants with a usable ten-year trend, its position is lower than ten years earlier (latest year 2023 to 2025 against a decade before): measured shares of world military spending, world output in purchasing-power terms and world goods exports, and proxies for research, education and productivity. Each uses its own world total. The strain measures sit near the top of their histories.

Author’s judgement, not a computed stage

Excess / Top, with decline pressure building

I keep the label at Top because the lead positions are intact. The numbers lean a little further, towards fiscal and financial decline, and I would not argue hard with someone who read them that way. No rule produced this stage. I chose it.

A fuller dossier, with each determinant’s sources and the challengers, is planned. For the idea behind it, see What outlasts an empire.

The United States’ share of the world, and ten years of change

  • Military spendingSpend, not capability33.4%−4.9 pts
  • Economic outputShare of world GDP, purchasing power14.6%−1.5 pts
  • Goods exportsChina 14.3%8.3%−0.8 pts
  • Stock-market valueValuation-sensitive proxy48.8%+8.5 pts
  • Research spendingSpend, not output proxy19.6%−2.0 pts
  • Foreign official custody at the FedCustody holdings against federal debt; a funding measure, not reserve share proxy7.5%−9.1 pts

Education and productivity are tracked too and also lower than ten years ago; both are proxies and live in the full dossier. The only measure up is stock-market value, which is partly a function of how high prices are.

  • Measured
  • Proxy: an indirect stand-in
  • Author’s judgement or illustration
Illustration, author’s judgement

Empires rise and fall in turn, and the leader rarely sees it from inside.

Four empires, 1500 to the present: an illustrative arcFour hand-weighted relative-power curves on one axis. The Dutch peak c.1650, the British c.1900, the United States is high and shown dashed beyond today, and China peaks c.1700, falls to a trough around 1900 and rises again. Illustrative, not measured.164818151945Dutch, peak c.1650British, peak c.1900Qing China c.17002026150016001700180019002000BritishUnited StatesChina

Hand-weighted shapes after Ray Dalio’s framework, drawn for this page. They show the pattern, not measured power. There is no forecast on it. Dotted hairlines mark 1648, 1815 and 1945.

Measured, dates approximate

The money lags the power by decades.

The baton: three reserve currenciesA timeline. The Dutch guilder leads from the Bank of Amsterdam’s founding in 1609 to 1781, the pound from about 1815 to about 1925, and the dollar from about 1925, with the formal dollar system dated 1944. Shaded overlap: US output passed Britain in 1862, about sixty years before the dollar overtook sterling in central-bank reserves. A dotted empty band after today is marked next, unknown.GuilderPoundDollarnext?US output ahead, pound still leads1862c.19251944 Bretton Woods2014: China’s output passes the US16001700180019002000

US output passed Britain’s in 1862 and China’s passed the US in 2014, both on Maddison Project GDP (CC BY 4.0). Reserve-currency dates are approximate: the guilder from the Bank of Amsterdam’s founding (1609) to its loss of status in 1781–92 (Quinn and Roberds, “Death of a Reserve Currency”); the pound from about 1815; the dollar overtaking sterling in central-bank reserves in the mid-1920s (Eichengreen and Flandreau, 2009), though sterling regained the lead in the 1930s and the formal dollar system dates from 1944. The “next” band is empty on purpose: this is not a prediction.

Archetype, illustrative

Education peaks first. Reserve status leaves last.

The lag: eight determinants peak in orderEight curves for an archetypal rising and falling empire. Education peaks first, then innovation, competitiveness, output, trade, military and financial centre; reserve-currency status peaks last. The gap between the first and last peaks is shaded and labelled the lag. Archetype, not measured.the lagEducationReserve currencyearlierlater

An idealised empire with eight determinants drawn in the order they tend to peak. It explains why a reserve currency can look strong long after the underlying lead has narrowed.

Measured Proxy

The United States still leads on four of the seven measures that have challengers. Three have slipped to others.

CChinaEEuro areaIIndiaEach row is on its own scale. The United States is the vertical line; a dot to the right is ahead of it.

Eight determinants: the United States against challengersFor each of eight determinants, dots show China, the euro area, Japan and India relative to the United States, on a scale where the United States is the vertical line. A dot to the right is ahead of the United States. China is ahead on output and trade; the euro area is ahead on trade and education. Reserve currency has no comparable challenger series.United Statesa quarterhalfdoubleEducationproxy · US down 9.4 pts in 10 yrsChina: 76.9 against US 79.36CEuro area: 81.9 against US 79.36EIndia: 34.4 against US 79.36IInnovationproxy · US down 2.0 pts in 10 yrsChina: 18.7 against US 19.58CEuro area: 10.6 against US 19.58ECompetitivenessproxy · US down 8.8 pts in 10 yrsChina: 95 against US 309.8CEuro area: 243.6 against US 309.8EIndia: 49 against US 309.8IOutputmeasured · US down 1.5 pts in 10 yrsChina: 19.5 against US 14.57CEuro area: 11.5 against US 14.57EIndia: 8.1 against US 14.57ITrademeasured · US down 0.8 pts in 10 yrsChina: 14.3 against US 8.28CEuro area: 24 against US 8.28EIndia: 1.7 against US 8.28IMilitarymeasured · US down 4.9 pts in 10 yrsChina: 11.7 against US 33.42CEuro area: 12.6 against US 33.42EIndia: 3.2 against US 33.42IFinancial centreproxy · US up 8.5 pts in 10 yrsChina: 11 against US 48.79CIndia: 7.5 against US 48.79IReserve currencyproxy · US down 9.1 pts in 10 yrsno like-for-like challenger series

Latest year per measure, from the World Bank (CC BY 4.0) and the Federal Reserve. Filled US dot: measured. Hollow: proxy. The reserve-currency row has no like-for-like challenger series.

Measured

China’s share of world output has risen from about 4% to nearly 20% in thirty-five years.

Shares of world output, 1990 to 2025Stacked shares of world GDP in buying-power terms. The United States falls from about 20 to 14.6 percent; China rises from about 4 to 19.5 percent.United States 15%Euro area 11%Japan 3%India 8%China 20%Rest of world 43%1990200020102025

Shares of world GDP in purchasing-power terms, 1990 to 2025. World Bank WDI (CC BY 4.0).

Measured Proxy (dashed)

Output, trade and military shares have slipped. Stock-market value is the exception.

OutputShare of world GDP, buying-power terms
19902025

Measured · now 15%

Goods exportsShare of world goods exports
19702025

Measured · now 8%

Military spendShare of world military spending
19902025

Measured · now 33%

Stock-market valueShare of world listed value; rises with prices
19902025

Proxy · now 49%

US share of the world total. The stock-market line rises with prices as well as with weight, so it is a proxy for financial standing.

Measured, three measures

Financial strain moves in waves, and today’s sits near the top of its range.

Financial strain since 1970A wave of financial strain, the average of three debt measures ranked against their own history. It rises through the 1980s, falls in the 1990s, climbs to the 2008 crisis, and is now near 83 out of 100.typical197019801990200020102020now 83

The average of three debt measures (total credit-market debt to GDP, federal debt to GDP and federal interest to receipts), each ranked against its own history. This is the financial strand only; the full strain composite adds internal and external order and scores about 90 out of 100 on a rank-based scale today. Power and strain are different waves, and they do not always move together.

Direction of travel

Steady on the loop. Heavier on the arc. Thinner at the edges.

Putting the three clocks side by side, no single one is raising an alarm. What stands out is the combination: a short cycle in an unremarkable middle, a long cycle where the burden has moved from households to the state, and a big cycle where the lead is intact but narrower each decade.

Short cycle: steady

The reading has not moved in three months. Inflation is what keeps it from sliding earlier in the loop.

Long cycle: the public burden is rising

Federal debt has risen about seven points of GDP in three years, while household and company debt has fallen relative to GDP. Interest now takes about a fifth of receipts.

Big cycle: slipping, from a high position

Six of seven determinants with a usable trend are lower than ten years ago (three measured, three proxies). Strain is high. The lead is still large.

What would change the direction
  • Inflation under about 1.6%, or a steeper curve: the short cycle moves earlier in the loop.
  • An inverted curve near −0.4: it heads towards contraction.
  • Interest taking a still larger share of receipts: the long and big cycles darken.
  • Polarisation and external conflict, the two strain measures I watch most closely.
“Its weights are choices rather than discoveries.”
The Ledger on its own method, in An Institution That Cannot Confess Cannot Learn. The same is true of this page.

Related finance essays

The thinking, in the Ledger’s own words

The method comes from Ray Dalio’s picture of the economy as a machine of credit, which I have tried to keep honest rather than clever.

More in Investing & judgement · What outlasts an empire · What Makes a Civilization?

About this reading

Method and limits

A reading is only as useful as your sense of how it could be wrong.

How it is made

The three-cycle framework is after Ray Dalio; the rules, thresholds and drawings are mine. The short cycle uses four inputs (the yield curve, jobs, inflation and credit spreads) and scores each of the five regimes. The long cycle reads debt by sector. The empire cycle follows eight determinants, then I make a judgement. All the rules are my own priors, not fitted to history. I checked them against the past afterwards. Some were revised after those checks, so the results below are in-sample, not a clean out-of-sample test.

How well it has worked

  • Short cycle. Tested on data as it was published at the time, the contraction regimes covered 47 of 58 recession months (81%), within three months either side. My bar was 80%, so this sits on the bar rather than above it: the commercial spread I first used scored 46 of 58 (79%), and the whole gap is one month in 2001. Small changes, such as how long a look-back the scores use or how late the spread arrives, move the figure between about 79% and 83%. It missed the 1990–91 recession entirely. That downturn was mild, credit spreads barely widened before or during it, and unemployment still looked low against a ten-year window that held the early 1980s, so a credit-led read had little to react to. It also raised false alarms in about 100 months. The simple Sahm rule on its own did better (96.6% and fewer false alarms), so I show it beside the stage and do not claim this is a better recession detector.
  • Long cycle. Of nine historical sector-eras, six matched what I expected, two were a stage off and one missed. Nine is too few to call it proof.
  • Empire cycle. Selected directional checks against past handovers came out as expected. They are illustrations, not a test, and there is no fit statistic, because the stage is a judgement.

What is estimated or stood in

  • Stand-in Household and company debt uses the Federal Reserve’s own flow-of-funds data, which tracks the BIS series closely (correlation 0.99).
  • Proxy In the empire cycle: stock-market value, research spending, education, productivity and the reserve-currency role are indirect measures, and each is labelled where it appears. The reserve-currency stand-in (foreign official custody at the Fed against federal debt) is a funding and custody measure. It tracks the level of the IMF reserve-share series but not its changes, so I do not read its direction as a change in reserve share.
  • Stand-in Credit spreads use the Federal Reserve Board’s own GZ spread, which is public domain. It tracks the commercial corporate-bond spread I first used closely (correlation 0.90 over 1986–2026), and swapping it in changed the back-test only slightly. It arrives about a month late, and it is not available in its original, unrevised form, so the back-test is a little kinder than real life would have been.

What is missing

True ISM survey data, US-specific conflict involvement and mass polarisation are not measured. Where data are late, such as jobs and inflation, the reading is as late as they are.

What is illustrated, not measured

Illustration The empire arc, the lag curves and the pale debt archetype are drawings of a pattern. They are not data, and none of them is a forecast.

What it is not

It is not a forecast, a trading signal or advice. It describes where I think we are, so that the next piece of news can be read in context.

Show the working: each input, its date and source

The model reading is as of 9 October 2026. “Period” is the period each observation describes, which is earlier for monthly and quarterly data. On narrow screens the Source column is to the right; scroll the table sideways.

ReadingLatestPeriodSource
10-year minus 2-year Treasury spread+0.44 pts9 Oct 2026US Treasury via FRED (T10Y2Y)
Unemployment rate4.2%Sep 2026BLS via FRED (UNRATE)
Consumer price inflation, year on year3.4%Aug 2026BLS via FRED (CPIAUCSL)
Sahm recession indicator0.0 (alarm at 0.5)Sep 2026St. Louis Fed via FRED (SAHMREALTIME)
Corporate bond spread (GZ), scored+0.93 ptsSep 2026Federal Reserve Board (Gilchrist–Zakrajšek), public domain. Monthly; counted as stale 45 days after its month ends
High-quality corporate bond spread, cross-check only+0.78 ptsSep 2026US Treasury (HQM) less 10-year Treasury, via FRED. Public; not scored
Federal debt / GDP123%Q1 2026Treasury and BEA via FRED (GFDEGDQ188S)
Federal interest / receipts21%4 quarters to Q2 2026BEA via FRED (A091RC1Q027SBEA, FGRECPT)
Household and company debt / GDP115%2026Federal Reserve Z.1 and BEA via FRED
US share of world military spend33.4%2025SIPRI via World Bank WDI, CC BY 4.0
US share of world GDP (PPP)14.6%2025World Bank WDI, CC BY 4.0
US share of world goods exports8.3%2025World Bank WDI, CC BY 4.0
US share of world stock-market value48.8%2025World Bank WDI (World Federation of Exchanges), CC BY 4.0
US share of world research spending19.6%2023World Bank WDI (UNESCO UIS), CC BY 4.0
Foreign official custody at the Fed / federal debt (funding proxy, not reserve share)7.5%2026Federal Reserve H.4.1 and Treasury via FRED

Credits. Data from FRED (Federal Reserve Bank of St. Louis), US Bureau of Labor Statistics, US Bureau of Economic Analysis, US Treasury, the Federal Reserve Board, the World Bank World Development Indicators (CC BY 4.0, incl. SIPRI, UNESCO UIS and the World Federation of Exchanges), recession dates from the National Bureau of Economic Research, used here to test the method, and the Maddison Project Database 2023 (CC BY 4.0) for historical output comparisons. Framework after Ray Dalio; figures drawn originally from our own data.

Educational and editorial. Not investment advice. No trade signals. Figures are as published at the dates shown and may be revised.

Beta. This reading is a static snapshot as of 9 October 2026, refreshed by hand, and may lag. Source: FRED, Federal Reserve Bank of St. Louis.

This product uses the FRED® API but is not endorsed or certified by the Federal Reserve Bank of St. Louis.