The geometric features of rural roads are the basis of most studies and safety models, but one of the other factors that can affect the safety of these roads is the residential areas located on their abutting lands. In this study, fifteen rural residential areas located in Kurdistan province were studied using the Data Envelopment Analysis (DEA), and the effect of these areas on the safety of rural roads was investigated. In each of these areas, four criteria of driveway density, the presence of pedestrian barriers, pedestrian overpasses, and the amount of crossing of vulnerable users have been investigated and determined as inputs to the DEA method. The inverse of the expected crash frequency is the output term of this method, which is known as the safety score of the units. The expected crash frequency is obtained from the Empirical Bayes (EB) method by combining the frequency of observed and predicted crashes. In this study, by using the DEA and obtaining the contribution of the input's effect on efficiency, we came to the conclusion that these values are equal to 0.5 for the driveway density criterion (R_DD), for the pedestrian barrier criterion (R_PB) equal to 0.02, for the crossing pedestrian criterion (R_PC) is equal to 0 and for the measure of traffic of vulnerable users (VRU_C) is equal to 0.48. It can be concluded that the criteria of driveway density and the amount of crossing of vulnerable users in these units are more effective than other criteria.