Plenty of scholars have studied social health insurance and private health insurance in different economies, and many concentrate on crowd-out effect. Different results have been presented due to various data choices and empirical methodologies, but few of them have revealed the transaction mechanism. We study this issue based on updated Chinese provincial panel data from 2002 to 2014 and improved methods. We focus on how the crowd-out effect happens through 3 channels-saving, demographic factor, and medical expenditure using multiple mediator models. Besides, we apply Panel VAR model to study magnitude of the impact these 3 channels contribute, corporately and respectively. Results clearly show that social health insurance has crowd-out effect in penetration but crowd-in effect in density caused by different mediation variables. In addition, the 3 channels have shown lasting and dynamic influences on the crowd-out (in) effect. Finally, our paper provides the anti-crowd-out solutions from both perspectives according to our empirical analyses. For social health insurance system, it is necessary to improve the efficiency and fairness of the funds. For the private health insurance companies, insurance products innovation and privilege policy should be made to decrease the negative impact of saving and medical service overuse.