Ecological and evolutionary dynamics of range expansions are shaped by both dispersal and population growth. Accordingly, density-dependence in either dispersal or growth can determine whether expansions are pulled or pushed, i.e. whether expansion velocities and genetic diversity are mainly driven by recent, low-density edge populations, or by older populations closer to the core. Despite this and despite abundant evidence of dispersal evolution during expansions, the impact of density-dependent dispersal and its evolution on expansion dynamics remains understudied. Here, we used simulation models to examine the influence of individual trait variation in both dispersal capacity and dispersal density-dependence on expansions, and how it impacts the position of expansions on the pulled-pushed continuum. First, we found that knowing about the evolution of density-dependent dispersal at the range edge can greatly improve our ability to predict whether an expansion is (more) pushed or (more) pulled. Second, we found that both dispersal costs and the sources of variation in dispersal (genetic or non-genetic, in dispersal capacity versus in density-dependence) greatly influence how expansion dynamics evolve. Among other scenarios, pushed expansions tended to become more pulled with time only when density-dependence was highly heritable, dispersal costs were low and dispersal capacity could not evolve. When, on the other hand, variation in density-dependence had no genetic basis, but dispersal capacity could evolve, then pushed expansions tended to become more pushed with time, and pulled expansions more pulled. More generally, our results show that trying to predict expansion velocities and dynamics using trait information from non-expanding regions only may be problematic, that both dispersal variation and its sources play a key role in determining whether an expansion is and stays pushed, and that environmental context (here dispersal costs) cannot be neglected. Those simulations suggest new avenues of research to explore, both in terms of theoretical studies and regarding ways to empirically study pushed vs. pulled range expansions.