Demise of the Golden Guru: Recent Treasury predictions have been far from accurate and plans may be afoot to contract out the research. But that is not as easy as it sounds, writes Robert Chote

THE ENTRANCE hall of the Treasury's forbidding Whitehall home contains a glass-topped display cabinet. Inside, next to a battered budget briefcase and a presentation set of Japanese coins, sits a small painted Buddha - the 1986 Golden Guru award for being Britain's most accurate economic forecaster.

A lot has changed since then. The Treasury has failed to predict the biggest surge in economic growth in 15 years and the longest recession since the Second World War, and has spent two years predicting an imminent recovery which only now appears to be under way.

Although the Treasury's predictions have been as accurate on average as those of academia and the City, its crystal ball gazing has been subject to much greater public ridicule. Morale among its forecasters has slumped, not helped by unprecedented ministerial massaging of their predictions and the virtual absence of lucrative offers to defect to the City.

But even greater indignities may lie in wait. The Treasury is informally sounding out private sector institutions with a view to contracting out the background work for its forecasting.

The decision is keenly awaited by the 30 or so economists in the Treasury's Economic Analysis (EA) and Monetary Policy (MP) divisions, whose jobs may depend on the result. But they can take comfort from the fact that contracting out is not as easy as it sounds.

The Treasury - like the London Business School and the National Institute for Economic and Social Research - uses a large computer model of the economy to forecast.

Economists cannot predict the behaviour of millions of consumers and businesses to the last detail, so they distil their findings about how different parts of the economy interact into a set of mathematical equations, linking a number of variables. For example, an equation in the model may state that a 1 per cent rise in wages leads - after a time lag - to a 1 per cent rise in what people spend in the shops.

The 1992 version of the Treasury model contains 326 such equations, plus 60 which simply define one variable as a combination of others. There are 127 variables which the model does not seek to predict - such as the oil price or changes in policy (interest rates or tax rates) - which have to be plugged in directly by the forecaster.

Model-based forecasting assumes that past relationships between parts of the economy are resilient over time and can thus be projected, which is how many of the economists in EA and MP spend their time. It would be difficult for the private sector to do this as the process depends on market-sensitive assumptions about planned policy changes, which are never made public.

The Treasury forecasts published in budgets and autumn statements are also influenced by ministers, who are ultimately responsible for the predictions. Civil servants accept this as a fact of life, although the massaging of figures before the last election made an important contribution to the recent slump in morale.

Long-term projections of government borrowing were cut dramatically ahead of publication in the 1992 Budget to show the books returning to balance - the changes were so big that they were even cleared with the Prime Minister's office.

The Treasury argues that even if ministers and officials need to be involved with the final stages of the forecast, in theory it would still be possible to contract out the long-term development and maintenance of the computer model. But in practice it may be difficult.

The model evolves over time to capture changes in the structure of the economy. But it is only by consistently using the model for forecasting that many of these improvements are stumbled upon.

The changes in the 1992 vintage of the model are a case in point. The equations which link changes in national insurance contributions and VAT rates to the level of earnings and inflation have been changed, no doubt reflecting research undertaken when the Chancellor decided he might include them in last month's Budget.

Conveniently, the new version of the model assumes that only half any increase in VAT or NICs is passed through to earnings and inflation, rather than the full amount. This helped to keep the budget forecast for underlying inflation below the Chancellor's 4 per cent target ceiling.

There is little reason to believe that contracting out the Treasury's forecasting would do much to improve its accuracy. The chart shows the Treasury has been over-optimistic about the strength of the economy throughout the recession, but most other forecasters have done little better. They missed the boom and the bust.

Predicting consumer spending has been a big problem, although failures to predict investment and exports have also been important. Traditional Keynesian economic theory argues that consumer spending is directly related to income. But in the 1970s, economic models were changed so that people's wealth, such as shareholdings, became as important in explaining spending.

However the Treasury still dramatically underpredicted consumer spending in the late 1980s because people borrowed money secured on the rapid rise in the value of their houses.

The use of house prices has helped the Treasury predict most consumer spending in the downturn, but not that on big items such as fridges and carpets. The Treasury has commissioned Warwick University's macro-economic modelling bureau to investigate the problem, and its report has just been submitted. The inclusion of housing market turnover is the leading contender for the missing link, as 40 per cent of these expensive items are bought when people move house.

Financial deregulation - allowing people and companies to borrow more with less security - hammered the predictive abilities of economic models in the 1980s. The washing-out of this effect, combined with the aftermath of the resulting explosion in debt, must create doubts about the ability of all forecasters to predict events in the 1990s. Handing the Treasury model to outsiders is unlikely to make much difference.

(Photograph omitted)