A nonparametric approach to monetary aggregatio

There have been many attempts to reconcile the concept of aggregation in money stock. However, neither the conventional simple sum aggregates (which do not consider the user cost or the interest rate elasticity) nor the theoretically preferred approach, the Divisia (which considers the optimizing behavior of economic agents), performs as well as expected in econometric studies, estimation, and forecasting. Thus, I propose an alternative aggregate in an attempt to reconcile this issue by applying a nonparametric estimation.;First, a nonparametric estimation on the unknown transformation of monetary aggregate as dependent variable is explored. Based on this estimation, I compare the forecasting ability of three different monetary aggregates (simple sum, the Divisia, and the nonparametric estimate, HM) for real GDP, price, and the interest rate. The forecast encompassing test and the Diebold-Mariano equality test confirm the superiority of HM over the other two aggregates.;Next, I introduce more nonparametric models. With the benchmark VAR model of monetary components (model I), the semiparametric transformation of simple sum monetary aggregate (model II), and the semiparametric estimate of monetary components (model III) are compared for forecasting ability. The forecast encompassing test and the Diebold-Mariano equality test illustrate the relative superiority of each model in forecasting real GDP, price, and the interest rate.;Overall the nonparametric estimates reveal their relative competitiveness in forecasting relevant variables, which gives us an additional avenue worth exploring.