時間序列模型對我國產業成長預測之優劣比較

This paper compares the forecast performances between the factor and conventional time series models for Taiwan’s industries. The factor model adopts the principle component analysis, generating factors from abundant information to represent the complicated economy. We find that the factor model has forecasting advantages over the autoregressive and vector autoregressive models. The factor model, in particular, greatly improves the forecasting performance on Taiwan’s manufacturing sector relative to conventional models. These results are robust to alternative model specifications.