In a recent paper (summarized here) Alex Nikolsko-Rzhevskyy, David Papell, and Ruxandra Prodan applied statistical tests to identify periods in recent U.S. history when Fed policy was “rules-based”—which they define in practice as Taylor rule periods—and periods when it was discretionary—which they define as non-Taylor rule periods. (They used both a structural change test and a Markov switching test).
Their statistical identification procedures were not based on which periods had better economic performance, only on measures of monetary policy actions.
But now they have now gone back to the data to determine whether the rules-based or discretionary periods delivered better economic performance. This morning David Papell sent me the following table summarizing their latest results. For three different performance measures, panels A, B, and C in the table show that Taylor Rule periods had far better economic performance than non-Taylor Rule periods.