A DSGE model of China

Abstract

We use available methods for testing macro models to evaluate a model of China over the period from Deng Xiaoping’s reforms up until the crisis period. Bayesian ranking methods are heavily influenced by controversial priors on the degree of price/wage rigidity. When the overall models are tested by Likelihood or Indirect Inference methods, the New Keynesian model is rejected in favour of one with a fair-sized competitive product market sector. This model behaves quite a lot more ‘flexibly’ than the New Keynesian.

Keywords:

• China
• DSGE
• Bayesian inference
• indirect inference

JEL Classification:

• C11
• C15
• C18
• E27

Notes

¹ We also considered the now standard developed country forward-looking rule and found that both models were rejected strongly under this form – see Appendix 2.

² If we alter our error processes as indicated, the Wald statistic of the NK model is 18.87 (p-value = 0.09) and that of the HY model is 12.33 (p-value = 0.42). The results are essentially no different from those obtained under our data – which are respectively 19.61 (p-value = 0.07) and 13.85 (p-value = 0.31) as in Table 1.

³ $k$ is set at eight quarters, in line with the usual central bank practice of looking 2 years ahead.