This dissertation presents an empirical analysis of United States monetary policy over the period ranging from January 1973 to December 1985. Switching regressions are introduced as a method of allowing for discrete shifts in the coefficients of reduced form money supply equations. Two switching models are presented. The first is a monetary policy reaction function using a switching mechanism developed by Goldfeld and Quardt. In this first model, the probability that a particular set of coefficients, or monetary rule, was in place in each sample month is determined by the level of a set of important economic aggregates. The second specification is a time series model in which M-1 money may follow one of two possible autoregressive processes in any given period. The particular process followed in each period is modeled as the outcome of a first order Markov process. Maximum likelihood estimates of each model are presented and interpreted. The results indicate that there were two periods of substantial monetary instability during this sample period. The first period approximately begins in April 1973 and ends in February 1975. The second period corresponds closely to the set of months between the Fed's October 1979 and October 1982 changes in operating procedure. Results from the Goldfeld-Quandt model also show that the levels of four macroeconomic series, the interest rate, the inflation rate, the unemployment rate, and the trade weighted value of the dollar, may be used as indicators of the monetary rule being employed by the Federal Reserve.Ph. D.