Speculating as we do in precarious futures of the weather kind, it is reassuring for meteorologists to hear that economists, financiers and other pundits of the more material world may sometimes get it wrong.
It was therefore with some sympathy, and yet with not a little shamefaced pleasure, that I read Conor O'Clery last weekend on the egg-faced economic experts who failed to predict the current south-east Asian economic turmoil. "In the misfortunes of our friends," de la Rochefoucauld remarked astutely, "we tend always to find something not displeasing."
Like weather forecasts made by a computer, economic predictions are very often right, but sometimes wrong. Rather than the behaviour of the atmosphere, however, the computer models used in economic forecasting describe, essentially, the behaviour of us human beings.
Economists have distilled their beliefs and theories about people into equations that encapsulate a myriad of variables. But be it the economy or meteorology, the forecast will be wrong if the basic model is inadequate, or if the input data are untimely or inaccurate - like, for example, one theory which related economic well-being, not just to the weather but even to the moon.
It was in the early 1930s that an American economist called Simon Kuznets suggested that the US economy was subject to more or less regular cycles of about 20 years' duration. Throughout each cycle, he maintained, agricultural and industrial production, financial investment and various demographic indicators could be seen to rise and fall in a pattern repeated every 20 years.
Decades later someone remembered that the moon exerts a tidal influence on our atmosphere, which as the Earth revolves, produces a very small twice-daily rise and fall in barometric pressure. Moreover, because of peculiarities in the lunar orbit, this tidal effect varies over time, one component of the variation having a cycle of 18.613 years.
This, it was reckoned, was near enough to 20. The next step was to connect this 18year cycle to the weather, which was achieved by relating it to US rainfall; it was claimed that there was a corresponding cycle in the rainfall, with wet years in 1917, 1936, 1954 and 1973, all coinciding with times of maximum lunar tidal force.
It was then easy enough to relate this information to figures for the grain production of the US which, very credibly, showed excellent yields in years when rain was plentiful.
The final step was to postulate that variations in the crop production must ripple through the largely agrarian economy of the United States, and thus produce the lunar "swings" that were recognised so many years ago by Kuznets.