Individuals undertake both solo and joint activities as part of their overall activity-travel patterns. Compared to work and maintenance activities, social and leisure activities differ in that they exhibit high levels of temporal and spatial flexibility. In this study we used data from an ego-centric social networks survey in the Greater Tokyo Area and follow-up group activity survey to estimate a joint eating-out destination choice model explicitly incorporating group-level impedance. Consistent with the literature, travel time has a large impact on destination choice as measured by its elasticity; however, the elasticities of group-level maximum, average and median travel times are larger than individual-level travel times. Furthermore, we show that incorporating group-level impedance increases model performance up to 49% against the ego-level impedance model, a substantial increase that underscores the need to incorporate group-level characteristics in travel behavior models.