Type Ia supernovae (SNe Ia) are explosions of white dwarf stars that facilitate exquisite measurements of cosmological expansion history, but improvements in accuracy and precision are hindered by observational biases. Of particular concern is the apparent difference in the corrected brightnesses of SNe Ia in different host galaxy environments. SNe Ia in more massive, passive, older environments appear brighter after having been standardized by their light-curve properties. The luminosity difference commonly takes the form of a step function. Recent works imply that environmental characteristics that trace the age of the stellar population in the vicinity of SNe show the largest steps. Here we use simulations of SN Ia populations to test the impact of using different tracers and investigate promising new models of the step. We test models with a total-to-selective dust extinction ratio $R_V$ that changes between young and old SN Ia host galaxies, as well as an intrinsic luminosity difference between SNe from young and old progenitors. The data are well replicated by a model driven by a galaxy-age varying $R_V$ and no intrinsic SN luminosity difference, and we find that specific star-formation rate measured locally to the SN is a relatively pure tracer of this galaxy age difference. We cannot rule out an intrinsic difference causing part of the observed step and show that if luminosity differences are caused by multiple drivers then no single environmental measurement is able to accurately trace them. We encourage the use of multiple tracers in luminosity corrections to negate this issue.