EVERY so often comes a reminder that the science after which this newspaper is named is, shall we say, incomplete. A recent decision by America's Bureau of Economic Analysis, an obscure but excellent body, has exposed a ghastly hole at the centre of economics. By common consent, the big questions in economics have to do with productivity - that is, with the ability to grow richer by extracting more value from the same amount of labour. But economists are not sure what it is beyond the blindly obvious such as technology, education and investment that makes productivity grow. The profession's best recent advance is the 'endogenous' growth theory, an equally obvious notion that technology, education and investment interact interestingly. The latest argument shows that, on top of being perplexed by its causes, economists are lousy at measuring productivity. Until recently, officials could give a reasonably coherent answer to questions on how much the American economy had grown in the past five years, or how productivity in the past five years compared with the 1980s. No longer. Over the past five years American GDP grew by 3.1 per cent - or maybe by 2.6 per cent. Productivity over the past five years was either strikingly better than it was during the 1980s - or perhaps hardly better at all. The choice depends on whether you tote up economic statistics in the old way, or under new rules proposed by the Bureau of Economic Analysis. Why the change? Because statisticians are pursuers of truth. The old way ignored changes in relative prices, such as the halving of the price of a mainframe computer over the past decade. The new approach tries to correct this. Nonetheless a few economists, notably Stephen Roach of Morgan Stanley, an American investment bank, vehemently oppose the change. Mr Roach's main beef is that the change will rewrite American economic history at a stroke. In particular, it will undermine a theory - his theory, as it happens - that productivity in services has grown rapidly of late, that this reflects the long-awaited fruits of the trillion-plus dollars that America has invested in information technology over the past decade, and that these gains are already underestimated in the present statistics.