摘要：This paper compares the direct and indirect methods of predicting the velocity of
circulation in the Japanese economy. Forecasts are generated using the autoregressive (AR)
model and Harvey’s structural time series model. In addition to point forecasts, prediction
intervals (calculated by using the recently proposed bootstrap-after-bootstrap) are used as a
criterion for evaluating forecasting accuracy. The results indicate the superiority of the direct
method. While this result is not consistent with the theoretical appeal of the indirect method,
it can be explained on the grounds that the pooling of time series reduces the noise associated
with individual time series.
关键词：Monetary Velocity, Forecasting, Bootstrapping, Harvey’s Structural Time Series
Modelling
