The molecular depolarization ratio (MDR) is of great significance for polarization lidar techniques in terms of validating the measurement accuracy, etc. However, previous studies mainly focused on cases with narrowband laser linewidths, and the transmittance of the Cabannes line in the receiver has been assumed constant. In this work, the narrowband theoretical model of MDR has been re-examined by taking the transmittance of the Cabannes line into account. A large relative deviation of beyond 200% has been found if the wavelength-shift reaches up to 0.5 nm for a receiving bandwidth of 0.5 nm at 532 nm, which is much larger than the case without considering the transmittance of the Cabannes line, i.e., only 15%, reported in previous studies. Besides, a broadband theoretical model has been proposed to evaluate the MDR for polarization lidar using high-power multimode laser diodes as light sources. Simulation studies have revealed that the MDR is highly related to the laser linewidth, the receiving bandwidth, as well as the wavelength-shift between the laser wavelength and the center wavelength of the receiver. The MDR at 520 nm calculated by the broadband theoretical model is about 21% larger than the value evaluated without considering the laser linewidth, when the receiving bandwidth is equivalent to the laser linewidth (e.g., 2 nm). Validation measurements, employing a 520-nm imaging-based polarization lidar with a 3.4-nm laser linewidth and a 10-nm receiving bandwidth, illustrated that the volume depolarization ratio in a clean atmospheric region (0.129±0.0025) was highly consistent with the theoretical MDR (0.132). The good agreement between theoretical and experimental results demonstrated a high measurement accuracy of the imaging-based polarization lidar and excellent feasibility of the broadband theoretical model.