This article examines the changes of adverse selection over time during a 3-year subsidized, voluntary-based Community Health Insurance (CHI) scheme in rural China. The data came from a 4-year longitudinal social experimental study (2002-2006) on the CHI in Fengsan Township, Guizhou Province of China. A panel of 8198 observations (average of 2730 individuals) was analyzed using random effect logit model. We found that the effect of health status on the enrollment choice of the CHI scheme was significant. People with chronic condition history, with fair health, and with poor health were more likely to enroll in the scheme than those without chronic condition and with good health status. In addition, we found that almost all of the interaction terms of the health status variables and CHI wave variable were not significant, which indicates that the effects of adverse selection have not significantly changed over time. Furthermore, people with medium income and high income were more likely to enroll in the scheme compared to those with low income. This shows that adverse selection persisted in the subsequent enrollments of the CHI scheme, even with the government subsidy to the premium. However, adverse selection did not become more or less severe over time and worked through to a steady state. In addition, inequity of enrollment still exists under the current premium subsidy policy. Based on the findings, relevant policy implications are put forward to further improve the CHI scheme.