Abstract: Based on the standard structural time series (STM) model, the macro-economic series can be decomposed into trend, cycle, season and irregular components. By adding explanatory and intervention components, we extend it into a complicated structural time series model. Cycles of different frequencies can be tested based on goodness of fit. We compare different mixes of components in forecasting China’s quarterly GDP and compare the STM model results with those of other models. The results indicate the STM model has superior forecasting performance.

Full report ：Structural Time Series Models for Economic Forecasting.pdf