X Close

STS Observatory



1/2 idea No. 18: 1968 IMF/government computing breakdown

By Jon Agar, on 30 July 2021

(I am sharing my possible research ideas, see my tweet here. Most of them remain only 1/2 or 1/4 ideas, so if any of them seem particularly promising or interesting let me know @jon_agar or jonathan.agar@ucl.ac.uk!)

The late 1960s in the UK witnessed intensely choppy economic waters. In November 1967, following crises in sterling, the Labour Prime Minister Harold Wilson devalued the pound. Four visits from by the International Monetary Fund were scheduled in 1968. In all of this, knowing what was happening in the economic was difficult but crucially important.

In National Archives file T 433/23 there is an annex, ‘Forecasting in 1968’, part of a draft history of forecasting, I think by the economist and government adviser Andrew Roy, that was under Treasury review. The story it contains is fast-moving and detailed. In brief, the Wilson government needed accurate economic forecasts on which to base policy.

These forecasts were already partly computerised. What is clear from Roy’s accounts is that further computerisation was urgently being considered. Professor Jim Ball, the computer model expert from the London Business School, was consulted.

It is also clear that the forecasts were also deeply sensitive. The Economist magazine ran an editorial, ‘Full, publicity please’ in May 1968 demanding that the forecasts be published, and the Permanent Secretary of the Treasury, William Armstrong. summarised the argument in favour of doing so in terms of accountability in a democracy: “a democratic government ought to provide for public scrutiny the basis of the economic strategy that it intends to pursue”. But “publication could be an embarrassment to the Government … The forecast will sometimes be wrong”. He also stated “we ought, as a matter of principle, never to publish forecasts that reveal a situation compelling government action”. At one stage the forecasts were moved from SECRET to TOP SECRET, reflecting the extreme sensitivity.

Furthermore there was clearly sharp differences between advisers about what the forecast models should contain. Thomas Balogh, for example, argued the existing method was “based on a mechanical conception of the economic system which unwarranted and leaves out of the account not merely the basic psychological questions but also the close inter-relationship between policy and reactions to that policy”.  The Chief Economic Advisor. Alex Cairncross, replied that Balogh misconceived the purpose and character of forecasting. Cairncross also expressed doubts about a computer’s capacity to reduce the burden of essential forecasting’.

There’s quite a lot here that interests me – computerisation of government work is my thing, and the word ‘mechanical’ is catnip.

Here are some notes about possible research leads:

  • Was further computerisation a response to the fast-moving, urgent economic crises that the Wilson government faced?
  • How does this story look when placed into a longer history of public knowledge of the economy? The dilemma expressed by Armstrong is a real one. The solution adopted, discretion and “minimum disclosure”, and even then only to trusted parties, was perhaps a typical position for the British state, and was a long way from open government.
  • Look at the differences between the so-called ‘mechanical’ model, and the more dynamic one proposed by Balogh. The dynamic model computerises psychological responses – just how was that human agency turned into code? – and more feedback loops that have a cybernetic feel, even if that word was not used.
  • More generally there is the material for a line-by-line analysis of the evolving Treasury computer model. This would be a history of software, and not just any software but the program that generated the economic forecasts on which the UK’s economic policy was set. If different interests, or values, shaped that code – and the disagreement between Cairncross and Balogh reflected deeper commitments – then the history of this software (and the policies based on them) is a history shaped by them.





Comments are closed.