Forecasts of Macroeconomic Variables in China: Combination Forecasts of Surveys and Models
LIANG Fang, SHEN Shihan, HUANG Zhuo
International School of Business & Finance, Sun Yat-sen University; University of California, Los Angeles; National School of Development, Peking University
Summary:
This paper uses the forecast combination method to predict the GDP growth and CPI growth in China. It also compares the out-of-sample performance of predictive model forecasts, expert forecasts, and combined forecasts in predicting GDP and CPI growth, and analyzes forecast errors to explore whether the forecast combination method can improve forecasting accuracy in different economic periods. We choose several predictive models from multiple perspectives. First, we use the regime-switching model (RSM) to reflect the dynamic patterns of GDP and CPI growth rates during stable and volatile periods. Second, we use the mixed data sampling model (MIDAS) to incorporate the information content of monthly data to improve the accuracy in predicting quarterly macro variables. Third, we use the mixed-frequency error correction model (MF-ECM) to consider the cointegration relationships between variables. To use a multi-dimensional and high-frequency macroeconomic dataset in prediction, we also resort to the mixed-frequency vector autoregression (MF-VAR). In addition, since the growth rates of GDP and CPI are both first-order single integral time series, the autoregressive integrated moving average model (ARIMA) is included as a benchmark. We use the forecasts of macroeconomic variables in the “Langrun Forecast” program to construct our expert forecast data. We choose the “Langrun Forecast” mainly for two reasons. First, it contains forecasts from various institutions and covers a long time period. Second, the forecasts included are all provided by well-known academic institutions or leading commercial organizations, which ensures the reliability and continuity of the data. Based on forecast series provided by models and experts, we use a variety of methods to carry out combination forecasts and explore whether forecast combination helps improve forecast accuracy. Specifically, the combination methods include simple averaging, weighting by forecasting errors, and the Bayesian model averaging method based on the Bayesian information criterion. The predictive information set includes fixed asset investment, total retail sales of consumer goods, total export value, total import value, industrial added value, M2 supply, Shanghai Composite Index volatility, national interbank market interest rate, financial institution RMB loan balances, newly started area of commercial housing, generated electrical energy, consumer expectation index, and national housing prosperity index. We use a multi-dimensional and high-frequency macroeconomic information set to make predicts. In the out-of-sample comparison, expert forecasts generally outperform model forecasts. The empirical results show that the expert forecasts contain almost all the information predicted by the models, indicating that the experts have considered the predictive content of the models when making forecasts. Furthermore, we find that the accuracy of expert predictions is significantly higher than model predictions during periods of economic instability (2008-2010), as experts can adjust expectations timely by grasping the actual economic environment and the direction of economic policies, and thus obtain more accurate forecasts. In addition, combined forecasting improves forecast accuracy. The robustness tests show that the improvement of forecast accuracy by combination forecasting does not depend on a specific benchmark, and changes in the length of the estimation window do not affect the main findings. This paper contributes to the existing literature mainly in two ways. First, previous studies that use forecast combination methods to predict China's macroeconomic variables focus on a specific combination forecasting model and discuss its predictive performance. Few studies have considered survey forecast information in the combination forecasting. Compared with model forecasts, expert forecasts are more sensitive to macroeconomic conditions and policy releases, and therefore can continuously update predictive information during forecasting, which improves forecast accuracy. This paper combines expert forecasts and model forecasts by using combination forecasting methods, and examines whether expert predictive information and model forecasting results can help to improve forecasting accuracy simultaneously. Second, this paper compares the forecast performance of expert forecasts and econometric models in different economic periods. We find that expert forecasts significantly outperform model forecasts during periods of economic volatility, and explains the reasons for the difference across economic states.
Ang A., G. Bekaert, and M. Wei. 2007. “Do Macro Variables, Asset Markets, or Surveys Forecast Inflation Better?”, Journal of Monetary Economics, 54(4):1163~1212.
[15]
Bates, J. M., and Granger, C. W. J. 1969. “The Combination of Forecasts”, Operations Research Quarterly, 20(4):451~468.
[16]
Curran, Declan, and M. Funke .2006. “Taking the Temperature-Forecasting GDP Growth for Mainland China”, Quantitative Macroeconomics Working Papers, 2006.
[17]
Diebold, F.X., and Mariano, R.S. 1995. “Comparing Predictive Accuracy”, Journal of Business and Economic Statistics, 253~263.
[18]
Elliott, G., and A. Timmermann. 2005. “Optimal Forecast Combination under Regime Switching”, International Economic Review, 46(4):1081~1102.
[19]
Ghysels, E., Santa~Clara, P., and Valkanov, R. 2004. “The MIDAS Touch: Mixed Data Sampling Regression Models”, UCLA: Finance.
[20]
Götz, T.B., Hecq, A., and Urbain, J. P. 2014. “Forecasting Mixed Frequency Time Series with ECM-MIDAS Models”, Journal of Forecasting, 33(3)198~213.
[21]
Hendry, F.D., and R.N. Ericsson. 2001. “Understanding Economic Forecast”, Cambridge, MA: The MIT Press.
[22]
Higgins, P. C., Zha, T., and Zhong, K. 2016. “Forecasting China's Economic Growth and Inflation”, FRB Atlanta Working Paper.
[23]
Nilsson, Ronny, and O. Brunet. 2006. “Composite Leading Indicators for Major OECD Non-member Economies: Brazil, China, India, Indonesia, Russian Federation, South Africa”, OECD Publishing.
[24]
Rapach, D.E., Strauss, J.K., and Zhou, G. 2010. “Out-of-sample Equity Premium Prediction: Combination Forecasts and Links to the Real Economy”, Review of Financial Studies, 23: 821~862.
[25]
Schorfheide, F., and Song, D. 2013. “Real-Time Forecasting with a Mixed-Frequency VAR”, Journal of Business & Economic Statistics, 33(3):366~380.
[26]
Stock J. H., and M.W. Watson. 2003. “Forecasting Output and Inflation: The Role of Asset”, Journal of Economic Literature, 41(3):788~829.
[27]
Stock J. H., and M.W. Watson. 2004. “Combination Forecasts of Output Growth in a Seven-country Data Set”, Journal of Forecasting, 23(6):405~430.
[28]
West, K. D. 1996. “Asymptotic Inference about Predictive Ability”, Econometrica, 1067~1084.
[29]
Wright J. H. 2009. “Forecasting US Inflation by Bayesian Model Averaging”, Journal of Forecasting, 28(2):131~144.
[30]
Yu. 2011. “Uncertainty and Risk Analysis of the Langrun Chinese GDP Forecast: Fan Charts revisited”, China Economic Journal, 2011(4):81~104.