The Missing Tank - Explaining Boom Bust Economics & the Coming Crash of 2026

What the Phillips Machine left out of the boom and bust, and why the 18-year land cycle keeps happening anyway

A subscriber left a comment recently, prompted by a piece on this Substack, saying he liked the old Phillips machines that the LSE built in the 1960s, the ones that used tanks of coloured water to model debt and interest rates. He wanted to know something more specific: How do economic rent, land prices and productivity actually interrelate to produce Fred Harrison’s 18-year property cycle, the one that gives us our regular booms and busts, and that appears, on the data available in mid-2026, to be delivering another bust roughly on schedule.

The Missing Tank

So I built a small hydraulic-style simulation of my own, land price in one column, productive capacity in the other, with a lever marked land value tax. What follows is the reasoning behind it: a history of the original Phillips machine, an honest account of what it did well and what it left out, and an explanation of the cycle it was never built to show.

Play with the model here:

https://phillipsmachienerentcycle.netlify.app/

A machine built in a garage in Croydon

Bill Phillips was not a conventional economist. He trained as an electrical engineer in New Zealand, served with the RAF, spent three years in a Japanese prisoner-of-war camp, and arrived at the London School of Economics after the war to study sociology, reportedly graduating with only a third-class degree. None of that suggested he was about to build one of the more celebrated teaching devices in the history of economics.

In the summer of 1949, working in his landlady’s garage in Croydon with a junior Leeds academic called Walter Newlyn, Phillips built a machine out of Perspex tanks, tubing, and salvaged parts from decommissioned Lancaster bombers. It cost around £400, a little over £12,000 in today’s money. He called it the Monetary National Income Analogue Computer, MONIAC for short, a name that nodded to ENIAC, the early digital computer built a few years earlier in the United States.

He demonstrated it to LSE economists in November 1949. It went down well enough that he was offered a teaching post on the strength of it. Between twelve and fourteen machines were eventually built and sold on to Leeds, Cambridge, Harvard, Melbourne, the Ford Motor Company, and the Central Bank of Guatemala among others. One sits in the Science Museum in London, conserved in 1995 by the curator Doron Swade, who once described being called away mid-lecture to fix an “incontinent machine” that had disgraced itself in front of a class. A working model is still demonstrated annually at Cambridge by the engineer Allan McRobie.

What the water actually showed

The MONIAC modelled the circular flow of income familiar to anyone who has done first-year economics: Y = C + I + G + (X − M). National income flows from a top tank, representing the Treasury, down through households, taxation, government spending, saving, investment and trade, and back round again. Turn a valve to raise the tax rate, or the interest rate, and you could watch the consequences ripple through the whole system in real time, in a way that no chalkboard diagram could manage. For a piece of 1949 engineering, it was genuinely effective. It was not built purely as a demonstration toy either. It gave broadly sound answers, and several of the machines went to institutions that used it for actual policy training rather than decoration.

What it did not do, and could not do, was show land. There is no tank for it. Land does not appear as a distinct category anywhere in the Y = C + I + G + (X − M) framework, because mainstream post-war economics did not treat land as a distinct category. Capital, in the neoclassical tradition Phillips inherited, absorbed land into itself: a factory, a field and a city-centre plot were all just “capital”, differing only in degree. Mason Gaffney, the American land economist, made the case at length that this collapsing of land into capital was not an innocent simplification but a specific and traceable move within the economics profession, one he set out in the essay collection he co-wrote with Fred Harrison, The Corruption of Economics (Shepheard-Walwyn, 1994). Whether or not one accepts Gaffney’s account of motive, the descriptive point stands on its own: a hydraulic model with no rent tank cannot show what happens when rent rises faster than output, because the machine has no plumbing through which that story could flow.

This is not really a criticism of Phillips personally. He was building an illustration of Keynesian national income accounting, which itself had no land sector, so the machine reproduced the theory faithfully. The limitation lives further upstream, in the accounting framework, not in the tubing.

Homer Hoyt and the 18-year pattern

The empirical pattern behind Fred Harrison’s cycle is older than Harrison. In 1933, in the depths of the Depression, the American land economist Homer Hoyt published his doctoral study One Hundred Years of Land Values in Chicago, tracing Chicago land values from 1830 onward. Hoyt found total land value rising from around $160,000 in 1833 to roughly $1.5 billion by 1893 and $5 billion by 1926, before collapsing to about $2.5 billion by 1932, and he identified recurring peaks roughly every eighteen years: 1836, 1854, 1872, 1890, 1907, 1925.

Fred Harrison picked up this empirical thread and gave it a Georgist explanation. In The Power in the Land (1983) he argued the pattern had survived the interruption of two world wars, and used it to call the early 1990s property downturn in advance. In Boom Bust: House Prices, Banking and the Depression of 2010 (2005), he set out roughly two hundred years of UK and US land price data showing the same fourteen-years-up, four-years-down rhythm, and used it to warn Gordon Brown, as early as 1997 by his own account, that the UK would peak around 2007 and enter depression by 2010. The 2007 to 2009 crash duly arrived. The economist Dirk Bezemer, at the University of Groningen, later listed Harrison among a small number of economists who had flagged the crisis years ahead of time. Working independently within the same Georgist tradition, Fred Foldvary wrote in the American Journal of Economics and Sociology in 1997 that the next bust after 1990 would land around 2008, eleven years before the fact, and he subsequently pointed to 2026, eighteen years on, as the next scheduled peak.

The mechanism, not just the timetable

The interesting part is not the arithmetic of adding eighteen to a previous trough. It is the mechanism, and this is exactly what the reader’s comment was asking about.

After a crash, credit is cheap and land is cheap, so early in the cycle it genuinely is possible to build, manufacture and employ people more cheaply. Real productive capacity expands. But newly created bank credit does not flow evenly into productive investment. A large share of it flows into the purchase of an asset that cannot be manufactured in response to demand: land. Because supply is fixed, rising demand for land shows up almost entirely as a rising price rather than a rising quantity. Economic rent, the return to that fixed, unimproved factor, climbs, and it climbs on top of whatever genuine productivity gains have been achieved.

For a while these two lines run together, rising land values partly reflecting genuinely improving productive capacity in the surrounding economy. But because land price is capitalised future rent, and expectations of future rent become increasingly speculative as the cycle matures, the land price line eventually detaches from the productivity line and keeps climbing on credit and expectation alone. At some point, wages and business turnover cannot service both the rents landowners are charging and the debt buyers took on to acquire land at inflated prices. Productive capacity stalls, then falls, while the land price line is still rising. That is the crisis: an economy where the cost of access to land and buildings has outrun what production can actually support, forcing default, foreclosure, and a reset back down to where rent and output can align again.

This is the story the original MONIAC had no way of telling, because there was no rent tank sitting between “investment” and “output” to show the two flows separating.

Building the tank back in

None of that undermines the basic point the reader was after. Whatever one makes of the precision of the eighteen-year clock, the underlying mechanism, credit flowing preferentially into a fixed-supply asset until rent extraction outruns productive capacity, is not controversial in Georgist economics and is not something the standard circular-flow model, or the machine built to teach it, was ever equipped to show.

That is what the simulation does differently. It runs the same land price and productivity lines through several cycles under the current tax mix, wages, trade and some investment taxed, land largely untaxed, and land price visibly pulls away from productive capacity each time, with output stalling as rents rise. Switch on a land value tax, so that the publicly created rise in land value is collected as public revenue rather than capitalised into private purchase prices, and the two lines behave differently: land prices fall back, and output, no longer being squeezed by rising rent and debt service, starts to rise instead. It is not a claim that a single lever fixes every part of a modern economy. It is a demonstration, in the same hydraulic spirit as Phillips’s original machine, of the one relationship that machine never had the plumbing to show.

More from Mason Gaffney: