Cointegration and Causality between Housing Price Index and CPI: Empirical Evidence from Nanjing of China

The Park test, Durbin-Watson test, ADF test, error correction model (ECM) and Granger-causality (GC) techniques were applied on the data of Nanjing from July 2005 to June 2008 to examine cointegration and causality between housing price index (HPI) and the Consumer Price Index (CPI) in natural logarithm, called LNHPI and LNCPI. Firstly, the Park test approves that heteroskedasticity dose not exist in the linear regression model between them with the method of ordinary least-squares (OLS), and thus the method is suitable for the regression model. Secondly, Durbin-Watson statistic of the regression model above indicates that autocorrelation exists, but it disappears when lagged LNCPI is added in the regression model, which is applied in the following tests. Thirdly, ADF stationary tests and ADF cointegration tests show that LNHPI and LNCPI are both integrated of order 1 (1) and cointegration does exist between them. Fourthly, the ECM results demonstrate that LNHPI has a positive effect on LNCPI immediately and negative effect on LNCPI chronically. At last, the GC results detect one-way causality running from LNHPI to LNCPI in short range while from LNCPI to LNHPI in long range. In a word, both cointegration and causality exist between HPI and CPI, which are validated from the data of Nanjing, China.