Μεταφέρω επιστολή αναγνώστη με τίτλο “Weather forecasting model won’t work for an economy”, η οποία δημοσιεύτηκε στην Φαϊνάνσιαλ Τάιμς της 6 Σεπτεμβρίου 2023. Το εμπειρικό ερώτημα ή, ακριβέστερα, η σύγκριση των προγνώσεων του καιρού με τις οικονομικές προγνώσεις δεν θίγεται στην επιστολή του αναγνώστη. Οι οικονομικές προγνώσεις είναι επιτυχέστερες διότι βασίζονται, μεταξύ άλλων, σε διαπιστωμένα εμπειρικά μοτίβα· η οικονομική συμπεριφορά είναι, σε ορισμένο βαθμό, προβλέψιμη. Τέτοια μοτίβα όμως είναι σπάνια στην συμπεριφορά του καιρού.
“With respect to Nicholas Gruen’s suggestions regarding ways to improve economic forecasts (Opinion, August 29), I agree there are several legitimate and reasonable criticisms that can be made of the performance of the economic models used in central banks and government for forecasting and policy analysis purposes.
However, to compare the atmospheric models used in weather forecasting with the models used for economic forecasting — as Gruen does — can result in misdiagnosis.
For weather is a classic example of a ‘chaotic’ system — it is dependent upon the non-linear simultaneous interactions between several well-defined and precisely measurable physical variables.
Weather forecasting models are based on the ‘hard science’ laws of physics and chemistry. The modelling and forecasting challenge lies in the frequency and the size of the three-dimensional area of the atmosphere in which they are measured and modelled, and, especially, the specification of the initial conditions.
It’s well to remember that even when using some of the world’s most powerful supercomputers, accurate weather forecasts only exist for about 10 days ahead.
Economic models (and especially financial ones) can also demonstrate some of the characteristics found in ‘chaotic’ systems. Economics is a ‘soft’ science: economic systems are complex, non-linear, dynamic and adaptive, concerned with the behaviour of a variety of economic agents and their complex relationships, its models subject to assumptions, simplifications and parameterisation. Its variables are typically concepts or abstractions, which are poorly defined, difficult to measure, only available with time lags, and subject to material revision. Yet economic forecasts are treated as accurate over months or years ahead!”
Santa Cruz, CA, US
Για την μεταφορά