This study used earthquake records from China to investigate comprehensively the correlation coefficients between various intensity measures (IMs), including peak ground acceleration, peak ground velocity, spectral acceleration, spectrum intensity, acceleration spectrum intensity, Arias intensity, cumulative absolute velocity, and significant duration. After collection of metadata information, 681 three‐component ground‐motion recordings with magnitudes of 4.9–6.9 were carefully processed and extracted from the China National Strong‐Motion Observation Network System dataset (2007–2015). The applicability of both the Next Generation Attenuation (NGA)‐West2 ground‐motion model (GMM) and of other GMMs was verified for different IMs, regarding the China dataset. Then, empirical correlation coefficients between different IMs were computed, considering the uncertainty due to the different sample sizes of the observational data using the bootstrap sampling method and Fisher transformation. Finally, the median values of the correlation coefficients were fitted as a continuous function of the vibration period in the range of 0.01–10.0 s and compared with the results of similar studies developed for shallow crustal regions worldwide. The developed region‐specific correlation coefficient prediction model yielded tendencies approximately like those reported in other studies. However, obvious differences were found in long‐period ranges of amplitude‐based IMs, cumulative effect IMs, and significant duration. These results suggest the necessity of using region‐specific correlation coefficients for generalized IMs in China. The presented results and parametric models could be easily implemented in a generalized IM ground‐motion selection method or a vector‐based probability seismic hazard analysis procedure for China.