Faith fashions economy models

A mixture of faith, fashion and prejudice is what Kenneth Clarke, the Chancellor of the Exchequer, has ultimately to fall back on when taking advice from economic forecasters, according to a study published last week.

Based on interviews with leading forecasters, including the Treasury's panel of seven independent economic advisers, the study by Robert Evans of the University of Bath concludes that judgement is an important part of the forecasting and modelling process.

It is this component, he argues, which determines what the forecast will be. "This is demonstrated by the fact that even when using the same model, different forecasting teams will produce different forecasts."

Lecturing to the science policy support group, Mr Evans also argued that published economic forecasts are a joint product of the forecasting team and its model. By altering individual equations within the model, the forecasting team turns it into something more plausible. "It is this routine adjustment which is largely responsible for the convergence of economic forecasts to a consensus value. Unadjusted model forecasts are far more divergent," he says.

Mr Evans, who is based at Bath University's science studies centre, said that he was particularly surprised by how large the standard error of an econometric forecast is, and the many outcomes that can fit with a given forecast.

"For a one-year gross domestic product forecast anything within about 1.25 per cent of the published forecast would be regarded as consistent with the model. That is quite a lot of leeway," he says.

Another big problem is that econometric models are reliable only to the extent that history repeats itself -- models can only be estimated using historical data.

"This means that if the economy undergoes a period of rapid change, as it did during the 1980s, models are unlikely to be able to tell us much about what will happen next."

Mr Evans, whose research is backed by the Economic and Social Research Council, says that "although macro-econometric models support forecasts by imposing consistency, they do not actually produce forecasts and neither do they test economic theories."

Soothsaying or Science: Falsification and uncertainty in macroeconomic modelling by Robert Evans, Science Studies Centre, Bath University.