Library

Phillip Anderson’s The Secret Life of Real Estate and Banking assembled, between roughly 1998 and 2008, the most complete public history of the US land cycle in the modern era. The book’s authority is not in its theorising — Anderson is forthright that the theoretical scaffolding was already in place, principally from Homer Hoyt’s 1933 study of Chicago, from Fred Harrison’s The Power in the Land, and from Roy Wenzlick’s 1936 work assembling data across more than 120 US cities. Anderson’s contribution was to walk every single cycle in the historical record, chapter by chapter, and to show that the 18-year rhythm has held since 1818 with a single war-disrupted exception.

That record is the foundation under everything the modern cycle forecaster does. This essay walks it. It begins with the pattern Anderson found, lays out the five-phase structure that every cycle has gone through, explains why 14 + 4 = 18 in the underlying arithmetic of credit, and closes on how a serious forecaster uses the record without overfitting it.

Read alongside our essay on Harrison’s three-force mechanism, which explains why the cycle must repeat, this essay covers what it has done.

The pattern Anderson documented.

The opening pages of The Secret Life of Real Estate and Banking state the case directly. The United States began selling off its real estate, officially, under a defined legal structure on 10 May 1800. From that starting point, Anderson identifies the following realised peaks in speculation and land value: 1818, 1836, 1854, 1869, 1888, 1908, 1926. Each was followed by a recession. The downturn that should have followed the 1908 peak was cut short by the build-up to the First World War; the rhythm reasserted itself at 1926; the Second World War disrupted the next normal peak, but a government-financed construction boom peaked in 1944.

For the first 144 years of US real estate enclosure, in Anderson’s words, “land sales and/or real estate construction peaked consistently every 18 years.”

After the Second World War, the rhythm reasserted with vigour. Land rent bottomed in 1955. The first major bust in listed real estate stocks came in May 1962 — exactly seven years after the low, a textbook mid-cycle slowdown. Construction peaked in the early 1970s, followed by the 1973–74 property crash and the recession that defined the decade. The next cycle ran 1974/75 to 1989/90, the next 1992 to 2006/07, and the present cycle from a trough that Anderson, writing in 2008, forecast for around 2010.

That is the spine of the record. Eleven peaks, one war-distorted reset, an 18-year average that has held across two centuries of wildly different economic, technological and political conditions.

The five-phase structure.

Anderson’s chapter-by-chapter walk is held together by a structural reading he takes from Harrison’s Boom Bust Graph 17.1 (p.87). Every cycle, regardless of its surface story, has moved through the same five phases. Naming them clearly is the first piece of equipment a cycle forecaster needs.

Phase I — Recovery (years 0 to 7).

The cycle begins quietly. Land prices are low, the crash is fresh, banks are cautious, and construction is minimal. Smart money enters slowly. The crowd is still licking its wounds from the previous contraction. The most important behavioural fact about this phase is that nobody who experienced the previous bust trusts that another cycle is starting. That distrust is what keeps the early prices low. Half of the seven years passes before sentiment shifts, and by then real-estate stocks have been rising off the bottom for some time.

Concretely, in the US record: 1837–44 after the Panic of 1837; 1893–1900 after the 1890s depression; 1933–40 after the Great Depression; 1955–62 after the post-war reset; 1975–82 after the 1973–74 crash; 1993–2000 after the S&L crisis; 2010–17 after the global financial crisis. Each is a seven-year quiet rebuild after a major contraction.

Phase II — Mid-cycle Slowdown (year 7).

The single most-misread moment in the entire cycle. Seven years after the trough, the construction wave reaches a natural pause. Real-estate stocks correct, sometimes sharply. Almost every observer at the time reads this as the top. It is not. Selling here is the most expensive mistake a forecaster can make, because the longest and most profitable phase of the cycle comes immediately after.

The textbook example is May 1962. Anderson is precise: “After 1955, the first bust in listed real estate stocks came on May 28, 1962.” Exactly seven years from the 1955 land low. The pause was real — the Kennedy-era stock correction was sharp. But the construction cycle was not done. The next ten years were among the best in real-estate history. The same shape repeats at 1982–83 (the Volcker pause from the 1975 cycle low), 1999–2000 (the dot-com pause from the 1992 cycle low), and 2017–18 (the rate-hike pause from the 2010 cycle low).

Phase III — Second Expansion (years 8 to 14).

The long upward leg. Six years on average, sometimes longer. Credit expands rapidly, lending standards stay sensible for a time, construction returns at scale, and land prices compound upward. The rent of land — created by the surrounding society — is being capitalised into rising land prices, which become the collateral for further credit creation, which finances further construction. The loop is benign through this stretch because the productive economy is still strong enough to service the debt being created.

In the US record: 1845–54 (the 1850s land boom), 1901–08 (the Edwardian expansion), 1963–72 (the long post-war run), 1983–89 (the Reagan expansion), 2000–06 (the sub-prime build-up), 2018–26 (the current cycle’s long leg). The forecaster’s discipline in this phase is patience — the cycle wants to be held, not sold.

Phase IV — Winner’s Curse (years 14 to 16).

The frenzy. Two years. Land prices go parabolic, real-estate stocks reach all-time highs, construction overshoots into oversupply, and bank lending becomes reckless. This is where the marginal buyer enters — the buyer who only bought because the price had already moved, who had to stretch to get the loan, who is convinced that prices cannot fall.

Anderson is unambiguous about the outcome: “It is usual for a majority of the mortgaged real-estate buyers in this phase to end up with negative equity once the cycle turns down.” The very worst time to be leveraged is the very time it looks safest to be so. The realised examples are the visible peak years — 1836, 1888–90, 1925–26, 1972–73, 1988–89, 2005–06. The marker on the current cycle sits 2024–26.

Phase V — Contraction (years 16 to 18+).

The crash. Three to four years on average. Marginal buyers exit first. Foreclosures cascade. Banks freeze new lending. A land-price-led recession follows, usually with a 12–24 month lag. The cycle does not end with a bang at the peak; the bang comes a year or two later when the contraction has worked through the credit system.

In the US record: 1837–43 (Panic of 1837), 1873–79 (Long Depression), 1893–97 (1890s depression), 1929–33 (Great Depression), 1973–75 (oil-shock recession), 1990–93 (S&L crisis), 2008–12 (GFC, foreclosure cascade). The next one, forecast on Anderson’s record, runs 2026–30.

Why 14 + 4 = 18.

Anderson devotes Chapter 17 to the arithmetic question that any honest reader of the record arrives at eventually: why 18 years? Not 10, not 30, not random. The answer Anderson gives, drawing on Harrison’s Boom Bust research, is the most satisfying piece of detective work in the literature.

The long-run rate of interest in industrial economies has historically hovered near five per cent. At five per cent compound interest, a sum of money doubles in approximately 14 years. The first mass mortgage-style lending in the English-speaking world emerged in late-18th-century Birmingham, where ale drinkers in a public house pooled savings, drew lots, and financed one another’s housing. The first published advertisements for these Terminating Societies appeared in 1778. The model required exactly 14 years for every member to be housed and the society to wind itself up.

The Halifax Building Society launched in 1853 with shares at £120, payable at ten shillings a lunar month at 5 % interest — a share would be fully paid up in 13 years and 7 months. The Temperance Permanent Building Society followed in 1854 on the same plan, with £80 set aside for a house. Anderson summarises: “The interest rate of 5 % determined the timetable — it took 14 years to raise enough cash for everyone to secure a home.”

The construction cycle inherits its length from this arithmetic. The +4 of speculation and crash sits on the end of every cycle as the final two years of frenzy plus the first two years of contraction. 14 + 4 = 18. The forecaster does not have to memorise this. It is a piece of structural arithmetic that emerges naturally from the underlying credit instruments.

The seven-year half-cycle that appears in Phase I and again at the boundary of Phase III is not an accident either. UK and Australian home-tenure data show that owners live in their homes for around seven years before moving — half of fourteen. UK commercial property data show the average time from conception to delivery of a major building runs about seven years. US urban-fringe data show that farmland passes through the hands of speculators and developers over an average 15-year period before reaching its final use. The 14-year construction cycle is woven into the operating timescales of the institutions that build the cycle.

The mid-cycle slowdown — why it fools people.

If there is a single piece of the record worth re-reading every time the forecaster gets nervous, it is Phase II. The mid-cycle slowdown looks exactly like a top. The price action is similar — a sharp correction in real-estate stocks, a stall in construction starts, tightening from the central bank, headlines about an end to the boom. Almost every cycle has produced a sincere “the top is in” reading at this moment. Almost every one of those readings has been wrong.

The diagnostic that separates a mid-cycle slowdown from a true peak is structural. At Phase II, mortgage debt is still moderate relative to GDP, lending standards have not collapsed, construction has not overshot into oversupply, and the productive economy can still service its debt. None of those things are true at Phase IV. The forecaster’s job is to count the years from the last cycle low, look at where mortgage debt sits as a share of GDP, look at lending standards, and refuse to confuse year 7 with year 16.

Looking at the present cycle: from a 2010 trough, year 7 was 2017, year 16 is 2026. The rate-hike pause of 2017–18 had the surface features of a top — the Fed was lifting, the curve was flattening, the late expansion was widely declared finished. It was Phase II. The expansion that followed ran another seven to eight years and produced the largest household-real-estate equity in US history. The same trap will appear at the next cycle’s year 7, around 2035. The forecaster should mark it now.

The Winner’s Curse and what it means for buyers.

Phase IV is the most psychologically interesting part of the cycle because the human behaviour it produces is so consistent. Late-cycle buyers fall into a pattern that has been documented across every realised peak in the record. They have watched prices rise for years. They have heard from acquaintances who bought five or ten years earlier and made significant equity. They have been told, in the language of their cycle, that this time is different — that the city is special, that the technology has changed the rules, that the central bank will not let it fall.

The instrument they use to enter is almost always credit at the limit of what they can service. The loan is justified by the assumed continued rise in price. The buyer is, in effect, betting their household balance sheet on the cycle continuing beyond its statistical median life. Anderson’s number — a majority of these buyers end with negative equity within roughly 18 months of the peak — is not a rhetorical flourish. It is what the record shows.

The forecaster’s role in this phase is not to predict the exact peak month. It is to refuse to participate in the Winner’s Curse. The cycle has a peak; the peak has a year. The forecaster’s job is to know which year the cycle is in, regardless of what the surface stories about that cycle’s specifics happen to be.

Two cycle peaks measured: 1989 vs 2007.

The two most data-rich peaks in the record are 1989 and 2007. They are the two cycles for which we have full modern data series on land value, real-estate stocks, construction starts, mortgage debt, land speculation indices, and bank credit growth. They are the two cycles a present-day forecaster should benchmark against, with 1926 added in where data permits.

Our companion Stress Dashboard diagram lays the six instruments out as gauges with rings marking the 1989 and 2007 readings. The pattern that emerges across both peaks is uniform. By the time of the cyclical top, five or six of the six instruments are sitting above their normal range, with several at all-time highs. No single instrument is the signal. The signal is the cluster.

The 2024–26 readings, as snapped in the published dashboard, show five of six instruments already exceeding 2007 levels — mortgage debt is the only laggard, reflecting the post-2008 regulatory clamp on US household borrowing. The 2026 forecast peak ring sits above all six current values across the dashboard. That is what the architecture predicts at the close of Phase IV.

The 1944 caveat.

The only cycle in 200 years that did not follow the rhythm is 1944. Anderson treats this carefully. The Second World War distorted the US economy in a way that previous wars had not — a government-financed construction boom (largely for war materiel and the housing of war workers) peaked in 1944, but there was no corresponding peak in private land speculation, because the war environment suppressed the kind of psychological climate in which speculation runs. Anderson’s word for that is precise: “There was little real estate speculation during the Second World War.”

The cycle did not vanish; it was held in suspension. Land rent reached its post-war low in 1955, and from that low the 18-year rhythm reasserted itself: 1955–62 (Phase I), 1962 mid-cycle bust, 1963–72 expansion, 1972–73 peak, 1973–75 crash. From there onward the record is regular.

Anderson’s reading of the 1944 anomaly is that only world-scale war has the cultural and financial weight to break the rhythm. The First World War did the same thing in 1908–18 — the 1908 peak was followed by a normal contraction that was cut short by war mobilisation; the cycle re-emerged at 1926. The forecaster’s takeaway is sobering: barring a global conflict at the scale of the two world wars, the cycle continues. Modern regional conflicts, recessions, financial crises, regulatory regimes and policy shifts have all been absorbed within the rhythm without breaking it.

How a forecaster uses the record.

The record is not a stopwatch. Reading it as one is the most common misuse. The 18-year average has held across two centuries, but the realised cycles have ranged from 15 to 28 years in length. The 1869 peak came 15 years after 1854; the 1972 peak came 28 years after 1944. The forecaster who tries to set their watch to “exactly 18 years from the last peak” will be wrong on most cycles.

The correct use of the record is structural. It tells the forecaster how to recognise which phase a cycle is in — by counting years from the last realised low, by reading the cluster of gauges in the Stress Dashboard, by watching for the year-7 mid-cycle pause that fools everyone, by recognising the late-stage psychology of the Winner’s Curse. The exact peak year is harder to call than the phase the cycle is in. The phase is enough for almost every forecasting decision.

The record is also a humility instrument. A forecaster who claims to call the top of the cycle to the month is overfitting. A forecaster who knows the cycle is in Phase IV is doing useful work. The difference between those two postures is the difference between forecasting and showmanship.

Where the record points next.

If the rhythm holds, and if there is no world-scale war between now and the late 2020s, the record forecasts a peak in the current cycle around 2026 (Phase IV) followed by a contraction running into the early 2030s. The next cycle’s Phase I should run from roughly 2028 to 2035, with a Phase II mid-cycle pause around 2035–36, a Phase III expansion through the late 2030s and early 2040s, and a Phase IV peak around 2044. The cycle after that points at a peak around 2062. These are forecasts — they should be marked as such, and they will be tested by reality between now and the time they happen.

The forecaster’s discipline is to keep the record open and not to refuse the next reading just because the present cycle is hard. The 1955 trough was hard to spot at the time. The 1962 mid-cycle was hard to read. The 1992 trough was hard to call. The 2007 peak was forecast by Jones & Harrison in 1997 — a decade in advance — and was widely dismissed. The record is uncomfortable because it asks the forecaster to commit to a view that everyone else’s commentary will be against, and to hold that view through periods when the cycle looks like it is breaking.

The cycle has not broken. The record is the most reliable instrument the long-cycle forecaster has. It should be read against Harrison’s mechanism — described in our companion essay on Ricardian rent, credit, and speculation — and placed inside the larger cycle hierarchy, where the 18-year rhythm sits beside the 18.6-year McWhirter economic cycle, the 36-year Saturn cycle, and the 84-year American conflict rhythm. The 18-year cycle is one voice. The forecaster’s job is to hear the chord.

A reading list.

Anderson’s The Secret Life of Real Estate and Banking (2008, with later reprints under that revised title) is the primary text and remains the most complete public record of the US cycle from 1800 to its publication date. Read it twice: once for the chapter-by-chapter history, once with the Part III machinery in mind to see how the architecture appears in every chapter.

Hoyt’s One Hundred Years of Land Values in Chicago (1933) is the foundational empirical work that Anderson and Harrison both build on. Wenzlick’s The Coming Boom in Real Estate and What to Do About It (1936) is the proof-of-concept forecast made from the same record, and worth reading for its discipline alone. Harrison’s The Power in the Land (1983) and Boom Bust (2005) are the theoretical companion volumes that explain why the record looks the way it does. Jones & Harrison’s The Chaos Makers (1997) made the 2007 forecast a decade in advance and stands as one of the cleanest pieces of long-cycle forecasting in the modern literature.