Economics
The Phillips Machine

The Phillips Machine is an analog computer from 1949 that describes the income flows in a modern economy.

Burgess COMMENTARY
I studied economics at Cambridge in the late 1950s. One of these machines was accessible in the basement of the Marshal Library in the Economics Department. It has helped me to understand the circularity of a modern economy and the fundamentals of the Keynsian economic ideas. Fast forward to now, and these ideas inform my thinking about how a system of metrics for the modern economy must be designed.
Peter Burgess

The Phillips Economic Computer

Doron Swade

Wednesday 22 March was United Nations World Day for Water. This may be an odd fact to tout in a bulletin such as this, but on Water Day a highly unusual piece of computing history went on public display at the Science Museum.

The machine was conceived and designed by Bill Phillips (1914-1975), a New Zealand-born engineer turned economist. The prototype caused something of a stir when first demonstrated at the London School of Economics (LSE) in November 1949. The contraption stood seven feet high and five feet wide and Heath Robinson would have been very pleased. The pumps were switched on and coloured water sloshed around through tanks, pipes, sluices and valves. The levels settled, pulleys turned and a pen plotter traced results. The machine was an hydraulic model of income flow in the national economy. Professors of economics were impressed.

Phillips outlined hydro-mechanical methods of modelling aspects of economic behaviour in a paper written in 1949 while a final year mature student at the LSE. He showed diagrams of economic models to Walter Newlyn, an undergraduate contemporary who had been a year ahead of Phillips at the LSE and had since taken up a lectureship at Leeds.

Newlyn encouraged Phillips to build a machine and was instrumental in securing £100 from Leeds University towards materials for the first prototype. At the LSE, Professor James Meade championed Phillips and his bizarre machine and, as an incentive to Phillips to complete it, promised an opportunity to demonstrate the beast at Professor Robbins' seminar. The demonstration took place on 29 November 1949. The presentation by Phillips and the performance of the machine (built in a garage in Croydon) confounded the sceptics.

Phillips formally described the use of mechanical models to show the behaviour of interacting economic variables in a paper 'Mechanical Models in Economic Dynamics' which appeared in Economica in August 1950. The publication of this paper and the impression made by the machine led to his appointment, without interview, as Assistant Lecturer in Economics at the LSE a few months later.

He was awarded his Ph.D ('Dynamic Models in Economics') in 1954, and became LSE's Tooke Professor of Economic Science and Statistics in 1958.

While Phillips is better known to economists for the 'Phillips Curve' which relates wage inflation to unemployment, it is his ingenious and original machine that brings him to the attention of the computing community.

The teaching potential of the Phillips' machine was obvious from the start and about 14 machines are believed to have been built. The prototype went to Leeds University where it still resides as a prized relic of the 1950s. Improved versions went to the universities of Cambridge, Liverpool, Birmingham, Manchester, Melbourne and Harvard, as well as to the New Zealand Institute of Economics. Two machines were built for the LSE.

Not all went to university departments. Ford Motor Company and the Central Bank of Guatemala were customers too. In the USA the machine was called Moniac to suggest money, mania and computation - this last by evoking echoes of Eniac.

The Phillips machine is an hydro-mechanical analogue computer designed to model the effects on total national income of a variety of factors including taxes and government spending, saving and investment, and imports and exports. The movement of money is represented by the flow of water and the accumulation of money (stocks) is represented by water collecting in tanks.

If incoming savings flow exceeds outgoing investment flow, for example, then a balance will accumulate in the Savings and Investment tank and the level will rise. If investment flow exceeds savings flow for any length of time, then the tank runs dry.

The model is a circular one (see diagram). Water is pumped to the top of the machine and cascades down a central column. Taxes, imports and savings are siphoned off into separate loops. Proportions of these rejoin the main flow as government expenditure, exports and investment. The net flow at the bottom of the central column accumulates in a tank. This level represents the working balance required for a given level of economic activity and is duly pumped back to the top of the machine to cascade back down through the system.

The strength of model is in the interactivity of economic factors. Rates of taxation, investment and levels of foreign trade can be altered by setting valves and sluices. More subtly, stocks (represented by the level of water in a tank) control flows elsewhere in the system by automatically altering settings of the valves.

Diagram of the Phillips machine. Source: LSE Quarterly, Winter 1988, Nick Barr.

A further level of sophistication is added by curves cut into Perspex templates to represent the particular form of a relationship between a level in a tank, for example, and a flow. The templates can be altered when the machine is set up and in effect control the transfer functions relating the independent and dependent variables. If the settings represent a viable economy the model stabilises and the quantitative results can be read from scales calibrated to 4%. The dynamic changes between one equilibrium state and another are also correctly modelled and drawn automatically by cord-driven pen plotters.

The Phillips machines were novel teaching aids. James Meade, then Professor of Economics at the LSE, used a Phillips machine to illustrate the destabilising effects of uncoordinated government intervention. One student would act as the Chancellor of the Exchequer with control over taxation and public spending, and a second student as the Governor of the Bank of England with control over monetary policy. Both were given the task of achieving a target level of national income. Crucially, each was instructed to ignore the actions of the other. The results were invariably disastrous.

Phillips machines, for all their ingenuity and appeal, fell into disuse with the advent of electronic computers in the 1950s. Few survive intact, and those that do have deteriorated.

The machine to be displayed at the Science Museum, thought to date from 1952, languished in the basement of the London School of Economics after being retired from active service in the classroom. It was rescued and restored to working order at the LSE in the 1980s and last ran (for a radio interview) in May 1992. The LSE was reluctant to part with the Phillips machine which had become the focus for the affection and respect felt for Phillips by colleagues, students and friends.

Professor Nick Stern, then Chairman of STICERD (Suntory-Toyota International Centre for Economics and Related Disciplines) at the LSE is acutely aware of the significance of the machine as a serious-minded memento of a distinguished episode in LSE's history, and he became properly concerned for its future.

In recent years several factors began to weigh. The small community of those familiar with Phillips and the history of his eponymous machine was dispersing and there was concern that the significance of the machine would fade with successive generations of staff. The two dedicated restorers of the machine in the 1980s, Reza Moghadam and Colin Carter, were pursuing careers elsewhere and could not provide the continuity of technical support necessary to maintain the increasingly frail device in working order. There was also concern for the gradual deterioration of the machine and the physical consequences of intermittent demonstration. Finally, there was the genuine wish to make the machine more accessible to a wider public.

These considerations weighed in favour of transferring the machine to the Science Museum despite the LSE's evident attachment to it. The basis of the transfer was outright donation but with an undertaking, unusual for the Museum, to display the machine, as a condition of acquisition. Expert conservation effort is reserved for fully inventoried objects only, and once acquired the machine was meticulously cleaned and stabilised.

Some 200 hours of conservation work was expended on it. The coloured water was drained and treatments applied. Conservation concerns militate against the machine being run again, and unless a working replica is built, the pleasure, not to mention the educational value of seeing the machine in operation, is a thing of the past.

However, alongside the display case is a video screen showing short clips of the working machine. The programme, which is stored on a one-off glass video disc, uses footage shot for Newsnight purchased from the BBC, newly story-boarded and with new voice-over. The Phillips hydro-mechanical model is an application crying out for software simulation which would have a ready market in university Economics departments.

With electronic computers a daily commonplace it is tempting to see the Phillips machine as an act of pre-electronic desperation - a string and sealing wax solution of heroic ingenuity forced on Phillips by the lack of a digital electronic solution. History confounds this view.

Eniac, the large-scale vacuum tube electronic calculator developed at the Moore School, Pennsylvania, was operational by the end of 1945, well before Phillips' first efforts. Nearer to home was Edsac, the first full-scale electronic stored program computer. Edsac came into service at Cambridge University in June 1949, the selfsame summer that Phillips began work on his prototype. Phillips was part of an academic community and it is inconceivable, given his interests and talents, that he was unaware of these developments.

Phillips' machine provided inspiration for this Emett cartoon, which appeared in Punch 15 April 1953.

In the early 1950s Phillips used an electronic analogue computer for macroeconomic modelling at the National Physical Laboratory where Richard Tizard was head of the Automatic Control Group. Later in the 1950s Tizard and Phillips used an English Electric Deuce, a digital vacuum tube machine, located at Aldwych, for joint economics research.

The new electronic computers were quite up to the computational task of economic modelling. The problem for Phillips was not computer processing power but visual display. Punched paper tape or numerical results tabulated offline by teleprinters are hardly distinguished for their visual immediacy. Without electronic VDUs the behaviour of arcane mathematical models was invisible and the process quite hidden.

Phillips wrote in 1950 that available electronic systems may well have offered greater accuracy and flexibility. But since his apparatus was intended for 'exposition rather than accurate calculation', it was desirable that the interactions were 'immediately comprehensible to the viewer'. So Phillips' choice of plumbing and Perspex was conscious and deliberate, and the visual immediacy of his hydro-mechanical creation vindicates his stand.

Economic theory has moved on and Phillips' theoretical model, though still relevant, is no longer complete. The machine now stands as a monument to an inspired piece of pedagogy. But the machine is symbolic in another respect. It suggests that economic behaviour can be described by mechanistic models and that economic modelling might submit to control theory.

We have been spoiled by science and engineering. We expect from economists the same degree of certainty offered by physicists and engineers. Despite the sophistication of economic theory it is not nearly as effective in predicting economic events as is control theory, for example, in predicting the behaviour of physical systems. Uncontrollable decline, an obstinate recession, and currency crises are not the best advertisements.

Economists now embrace chaos theory and increasingly use statistical models, and this tends to obscure the failure of mechanistic determinism. But perhaps we are wrong to see the Phillips' machine as a symbol of predictive certainty. When operated, the machines were sometimes temperamental and prone to spring leaks. Phillips was repeatedly called away to a repair an incontinent machine which had disgraced itself during a class. It may be that the machine, like the economy, is defying our efforts at control, by flaunting the irregularity of its ways, and giving us a cold shower in the process.

Doron Swade is Senior Curator of Computing and Information Technology at the Science Museum in London.

Editor's note: an edited version of this article has appeared in The Guardian, 16 March 1995, under the title 'Liquid assets'.