The implications of aggregation theory and non-linear dynamics for understanding the role of monetary assets and monetary service flows.

摘要

Aggregation theory can be used to produce monetary quantity and price aggregates under assumptions that are no more restrictive than the assumptions necessary to produce quantity and price aggregates for other decision variables such as consumer goods and services. The standard aggregation framework can also be used to aggregate interest rates. In this dissertation, I investigate the empirical relationship between the aggregation-theoretic monetary aggregates, short-term interest rates, and the other key macroeconomic business cycle variables. In Chapter 1, I show that both monetary quantity and interest rate aggregates improve predictions of the change in real output over the post-war period, but that this result is sensitive to the time period. Over some sub-periods, neither monetary aggregates nor interest rates are informative. I use partial coherence analysis to show that the correlation between the growth rates of money and real output is significant only at low frequencies, and argue that this may explain the instability of statistical results in the time domain. In Chapter 2, I use an alternative econometric technique to examine the long-term properties of the monetary aggregates. The levels of monetary and price aggregates can be modeled as integrated stochastic processes. If the non-stationary stochastic trends in these time series are identical, then a linear combination of them will have a reduced order of integration. This property is called co-integration. I show that nominal money and prices are integrated of order two, but that real money is integrated of order one. I further demonstrate that the integration and co-integration properties of the aggregation-theoretic monetary variables are more stable than the properties of atheoretic monetary variables. In Chapter 3, I use a representation theorem to show that co-integrated economic variables will generally have a non-linear data generating process. I use non-parametric methods to test a co-integrated set of short-term interest rates for non-linear dynamics, and find robust evidence of non-linearity. These results reinforce earlier findings that the aggregation-theoretic monetary variables are non-linear. The implications of non-linearity for econometric identification are explored.