Abstract
The asymmetric and segmented causality between the monetary policy and real estate market in China is crucial but remains mystery. With the application of quantile causality test, this article investigates nonlinear dependence between property prices and money supply. Our results show that the tail causality exists in many cities in China. Moreover, we find that small-sized cities and inland cities are more sensitive to the broad money (M2) changes when the housing market return is in the tail quantile intervals. These findings can help the Chinese government formulate appropriate monetary policies regarding their implications in the real estate market.
Notes
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 This statement is officially written in ‘Guowuyuan guanyu Cujin Fangdichan Shichang Chixu Jiankang Fazhan de Tongzhi’ (‘Notice of the State Council on Promoting the Sustainable and Healthy Development of the Real Estate Market’), which is often named as Circular 18, by the State Council on 12 August 2003.
2 Data from National State Statistical Bureau of China at http://www.stats.gov.cn/tjsj/.
3 We use the price index of 68 large- and medium-sized cities in China as the proxy for China’s real estate market in the rest of the article, because they are the representative cities that provide sufficient information.
4 The literature taking the asymmetry into consideration includes, for example, Zhang et al. (Citation2016) who compare the impact of monetary policies on Chinese cities of different tiers, Chen et al. (Citation2012) who illustrate differential effects based on housing demands, and Tsai and Chen (Citation2009) and Tsai (Citation2013) who find that housing prices are asymmetrically adjusted to money supply in the time of downward and upward in the United Kingdom (UK) and the US respectively.
5 The results of ADF test are available upon request.
6 The urban area refers to downtown areas of a city. Counties under the city’s jurisdiction are excluded.
7 The maximum lag is set as 10 in practice. We reject the null hypothesis at a 10% significance level.