ABSTRACT Air voids (density) is a key volumetric property of asphalt mixes and play an unignorable role in anti-cracking. While lower voids were expected to result in more durable asphalt mixtures, contradictory trends were found in lab testing. In this study, several plant-produced mixtures were compacted to three levels of air voids (7%, 9%, and 11%) and tested to evaluate the sensitivity to the crack resistance. The results indicate that the higher the air voids, the better the cracking resistance as measured by the CTindex. To understand the cause, a stepwise regression analysis was conducted to model the IDEAL-CT parameters with volumetric and gradation parameters. The air void, percent passing the 0.6 mm (P0.6) and binder content (Pb) are found to have significant effects on CTindex and post-peak slope. The analysis of the CT change rate shows that the Pb, dust to binder ratio (D/B), and P0.6 can be used to predict the effect of the voids with a correlation factor (R2) of 0.96. For further understanding, the asphalt film thickness was calculated and found that compared with using film thickness as the single fitting parameter, combining the film thickness with D/B in modelling can dramatically increase the R2 of correlation from 0.67 to 0.92.