4 Empirical results
Recall that, the null hypothesis of the KPSS test is stationarity, complementing the ADF test. Remember that the ADF test has low power against stationary near unit root processes. Table II shows that the variables are integrated of first order.
Now, we estimate a VAR with 1 lag (according to the minimum AIC) and then, we proceed with a VEC model testing the cointegration relationship for the pre‐ and post‐reform periods and for the whole period 1952‐2007. Table III summarizes the results.
Note that there is a cointegration relationship for the whole period, it is positive and the real per capita GDP is weak exogenous at 1 percent (χ2‐statistic is 6.05 producing a p‐value of 0.014). Moreover, Table III shows that testing Toda and Yamamoto (1995) Granger causality for the whole period indicates that causality goes from inequality to economic growth.
The cointegration test for the first period suggests an inequality‐growth elasticity of 0.29 where GDP per capita is weak exogenous (χ2‐statistic is 1.36 producing a p‐value of 0.243). Once again, Toda and Yamamoto Granger causality test indicates that causality goes from inequality to per capita GDP in the pre‐reforms period. This unidirectional causality can be explained by the composition of the Chinese economy. The pre‐reform period, nearly 80 percent of China's population lived in rural areas and was primarily involved in agriculture but there was not perfect equality in rural China (Benjamin et al., 2005). There were differences across localities and regions reflecting differences in endowments and natural conditions which can predetermine economic growth[6].
The cointegration test indicates one relationship for the second period with elasticity of 0.37 and GDP per capita is weakly exogenous (χ2‐statistic is 1.147 producing a p‐value of 0.284). However, there is not directional causality in this period.
In addition, we present the impulse response functions explaining how a positive shock in the per capita GDP produces an effect on inequality (Gini coefficient). Figure 1 shows, the pre‐reform period, a positive response of inequality by a positive shock of income during the first three terms but then there is a decline on inequality in the subsequent three terms and then, once again a positive response.
Figure 2 shows that during the post‐reform period, a positive shock of income produces a rapidly increment of inequality after the second term.
Instead, Figure 3 shows for the whole period (1952‐2007), the response of an increase in inequality has a negative impact for nine periods and then a high positive impact.