AbstractThe search for generalizations in the face of complex species–environment interactions is particularly important for minimizing the cost of managing populations of species. We tested whether we could generalize, at various nested scales, the species‐level demography of the widely invasive plant species Parkinsonia aculeata (Fabaceae) and whether these generalizations were representative of the demography observed locally. Full demographic surveys at all life stages of the species were conducted in 23 Australian sites during seven years (from 2000 to 2007), across a 1000‐km climatic gradient. Sites were nested across four climate regions (arid, semi‐arid, semi‐wet/dry tropics, wet/dry tropics) and three habitat types (upland, wetland, and riparian). We estimated the vital rates (growth/retrogression, survival, fecundity) at all life stages and size classes and combined them to create 99 site–year demographic matrix population models. With these models, we then estimated site–year‐specific asymptotic population growth rates and their corresponding prospective elasticity values to perturbation of the vital rates. We then developed a nested retrospective elasticity analysis (nested LTRE) to test whether and how upscaling the results (i.e., from site to habitat, to climate region, and to the invaded range) produced averaging bias, which could lead to spurious interpretations of the relationships between the retrospective elasticity values. The prospective analysis highlighted that site–year variation in the matrix population models, population growth rates, and corresponding elasticities could not be well summarized by a single species‐level analysis and that the spread was as diverse as found in previously reported multispecies‐level demographic analyses. The nested LTRE analysis showed that upscaling demographic models introduced for most sites new sources of errors in the estimation (in terms of magnitude and sign of retrospective elasticities), which increased drastically as we progressively aggregate the demographic information between nested scales of observation. Our findings suggest that regardless of the scale, demographic generalizations at the species scale are not always useful for managing P. aculeata across sites in its invaded range, given its plasticity in demography.