The energy partition in the products of ethylene photodissociation (including C2H4, C2D4, D2CCH2, cis- and trans-HDCCDH) at 193 and 157 nm and the rate constants of H loss channels were computed based on ab initio ethylene ground-state surfaces of which most were reported earlier. In the calculations of the energy partitions, a simple model was used in which the excess energy above the transition state is distributed statistically and the energy released by the exit barrier is described by the modified impulsive model. The rate constants of the ethylene H(D) elimination were calculated according to the variational RRKM (Rice-Ramsperger-Kassel-Marcus) theory, and the RRKM rate constants with tunneling corrections were obtained for vinyl decomposition at 193 nm. In contrast with previous conclusions drawn by LIF (laser induced fluorescence) studies, the rate constant calculations suggest that the H loss may be a nonstatistical process. However, the computed variational transition states for H loss appear reasonable as indicated by the translational energy. That with present investigation indicates that the atomic elimination proceeds via the predicted transition states though the process is nonstatistical. Analysis of the H-2 translational energy measured at 193 and 157 nm by molecular beam experiments gives evidence that the overall mechanisms of the molecular elimination are different at the two wavelengths, which is also in disagreement with previous belief. At 193 nm, both H-2 elimination channels may occur through the predicted transition states. On the other hand, further comparison of the theoretical and experimental translational energy of hydrogen molecule at 157 nm suggests that the observed (1,1E) reaction path is most likely of much higher "exit barrier'' than the one computed. For the (1,2E) channel, the calculations are still in support of the computed transition state being the one along the experimentally observed pathway at 157 nm.