Multivariate filter estimation and ARDL model analysis of China’s potential output

ABSTRACT

In this article, we apply a new multivariate filter approach to estimate China’s potential output. Furthermore, we build an ARDL model to analyse the influence on potential growth caused by important factors that contribute to estimation and China’s development. Our results show that the current economic slowdown is not a cyclical phenomenon and China’s potential growth has declined since 2010. We also show that fixed asset investment and trade, which have a long-run relationship with potential output, exert negative long-run effect on potential growth justifying the implementation of China’s recent supply-side reforms.

KEYWORDS:

• Potential output
• economic growth
• multivariate filter
• ARDL

JEL CLASSIFICATION:

• C32
• C51
• E32

Acknowledgement

This work was supported by the Fundamental Research Funds for the Central Universities, and the Research Funds of Renmin University of China under Grant [16XNH036].

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ The reason for using Equation 4 is that we add dynamic terms in observation equation and may cause the scaling factor setting too little weight on reducing potential output variability compared to standard HP filter. Detailed explanations can be found in Borio, Disyatat, and Juselius (2014).

² The properties that make economic variables significant in observation equation include two aspects. One is that variables should be correlated with output at specific frequencies setting by scaling factor and the other one is that variables should have stable means. Thus, we follow Borio, Disyatat, and Juselius (2014)’s method using Cesàro mean to de-mean all economic variables so that they can meet the requirement of the approach.

³ The growth rates are computed by using the data from ‘China Statistical Yearbook’.

⁴ Assumptions for prior mean and SD of auto-regressive parameter and scaling factors are following Borio, Disyatat, and Juselius (2014) and Chang et al. (2016)’s work. As prior distribution, we assume the gamma distribution with SD of 0.3 for all parameters and the prior mean for auto-regressive parameter is 0.65 of which the interval is between 0 and 0.95. Coefficients for economic information variables have prior means equal to 0.4.

Additional information

Funding

This work was supported by the Renmin University of China [16XNH036].