Explanation and forecasting

In macroeconomics, models are built for explanatory, predictive or policy purposes. The success of these models with respect to these purposes depends on whether they capture sufficient invariance. To know whether the invariance is sufficient one needs to ascertain the domain and scope of the involved invariant generalisations. The invariance domain is determined by the potential influences that are non-negligible. But for an explanation or prediction we need a complete list of these non-negligible causal factors. To solve this problem the Cowles Commission suggested to give theory the responsibility of completing this list. Friedman suggested an alternative methodology: to investigate the scope of the model to determine for which economic systems, time period or phenomena the model applies by testing its predictive performance. But to know whether the future will fall within this scope, one needs to investigate whether the magnitudes of the potential influences are stable across time and cannot be influenced by policy interventions. To deal with this problem Prescott suggested to use for the determination of these magnitudes the stable facts of an economy, that is to say, stable across a very long period.