Is the past the best forecast of the future in a data rich environment?: the case of China
thesisposted on 2022-03-28, 13:49 authored by Qin Zhang
GDP growth rate and inflation are two of the most critical issues facing China's economy. To improve GDP growth rate and inflation forecasts in a data rich environment, this thesis studies forecasting of China's four leading macroeconomic variables using six models. These variables are the consumer price index (CPI). industrial production, electricity production, and producer price index:industrial goods. The three factor models used are: the diffusion index (DI), factor-augmented autoregressive integrated moving average (FARIMA) model and factor-augmented vector autoregressive (FAVAR) model. The three univariate time series models are: autoregressive model (AR), autoregressive integrated model and simple exponential smoothing model. The predictors are summarised using a small number of indexes constructed by principal component analysis and then are used to construct one-, three-, and six-month-ahead forecasts using 36 predictors from 1997 through 2014. Compared to benchmark AR forecasts, the forecasting results of the factor models showed that the DI and the FARIMA model generally do not improve forecasting performances for CPI, industrial production, and production of electricity in one-,three-, and six-month-ahead. Rather, the FAVAR model yields significant improvements over the benchmark AR model except for CPI in one-month-ahead forecast. Another notable result is that the two wining models in three-, and six-month-ahead forecasts: exponential smoothing and factor-augmented VAR modelessentially produce the naive forecasts except for PPI:industrial goods. This implies that the benefits of using complicated forecasting models such as diffusion index or factor-augmented VAR model are minor; naive forecasts are sufficient to explain the predictable dynamics of the CPI, industrial production and electricity production in three-, and six-month ahead. Overall, this study provides interesting results on forecasting China's macroeconomy in a data rich environment.