15 Discussion

It is well known that model parameters are not invariant to shifts in policy regimes. Since Lucas' excellent paper on the subject, economists have paid lip service to the problems raised by regime shifts. Most policy analyses, however, ranging from columns in national magazines and newspapers to more formal work, continue to use models whose parameters are implicitly assumed to be immune to regime shifts. Carl Walsh departs from standard practice by analyzing how the parameters of a money market model might be affected by the October 1979 change in the Federal Reserve's operating procedures. When the Fed switched from a policy of stabilizing within-month fluctuations in the federal funds rate to a policy of aiming at a path for nonborrowed reserves-while allowing large fluctuations in the funds rate-the probability distributions of asset returns were affected and the pararne-ters of the system changed. Walsh is correct to be distrustful of models whose parameters were estimated from data generated under the old regime to analyze the behavior of the system under the new regime. There appears to be good reason to be concerned about possible structural changes following the shift in policy regimes. The behavior of financial markets has been difficult to predict and some troublesome puzzles have emerged. While the variance of short-term interest rates rose markedly, as predicted by existing models, the variance of short-term money growth also increased. This was not predicted by existing models and created considerable embarrassment for the Fed since it claimed that the change in operating procedures was needed to achieve closer control over money growth. Finally, long-term interest rates have behaved in a mysterious fashion. Not only has their average level remained higher than seems consistent with reasonable expectations of