Dark energy can modify the dynamics of dark matter if there exists a direct interaction between them . Thus a measurement of the structure growth , e.g. , redshift-space distortions ( RSD ) , can provide a powerful tool to constrain the interacting dark energy ( IDE ) models . For the widely studied Q = 3 \beta H \rho _ { de } model , previous works showed that only a very small coupling ( \beta \sim \mathcal { O } ( 10 ^ { -3 } ) ) can survive in current RSD data . However , all of these analyses had to assume w > -1 and \beta > 0 due to the existence of the large-scale instability in the IDE scenario . In our recent work [ Phys . Rev . D 90 , 063005 ( 2014 ) ] , we successfully solved this large-scale instability problem by establishing a parametrized post-Friedmann ( PPF ) framework for the IDE scenario . So we , for the first time , have the ability to explore the full parameter space of the IDE models . In this work , we reexamine the observational constraints on the Q = 3 \beta H \rho _ { de } model within the PPF framework . By using the Planck data , the baryon acoustic oscillation data , the JLA sample of supernovae , and the Hubble constant measurement , we get \beta = -0.010 ^ { +0.037 } _ { -0.033 } ( 1 \sigma ) . The fit result becomes \beta = -0.0148 ^ { +0.0100 } _ { -0.0089 } ( 1 \sigma ) once we further incorporate the RSD data in the analysis . The error of \beta is substantially reduced with the help of the RSD data . Compared with the previous results , our results show that a negative \beta is favored by current observations , and a relatively larger interaction rate is permitted by current RSD data .