The matter of scale is often ignored when analysing accessibility and its effects on population change. While accessibility has a concentrating effect on the activities of society on a regional scale, it also has a decentralising effect within urban areas in developed countries, including Finland. Population change is the outcome of numerous individual location choices. However, the scale on which these choices form a pattern related to transport accessibility is unclear, because the increasing stochasticity in an accurate resolution forces consideration of the effect of accessibility on population dynamics in the context of local characteristics. The matter of scale on modelling the relationship between transport accessibility and population change is considered in this article by comparing results using six different resolutions: namely side lengths of grid cells involving 2, 4, 8, 12, 16 and 24 kilometres. Road, air and rail accessibility were related to population change with non-linear regression, generalised additive models (GAM). Road accessibility was assessed by potentials originating from a gravity model. For revealing the effect on population change, air and rail accessibility were calculated as travel times to airports and stations. Analyses were based on exact population grid-cell data, an accurate model of road and rail networks and using geographical information systems (GIS). The study shows that population change was strongly related to potential accessibility. Airport accessibility also had high importance, whereas railway accessibility did not have any significant effect. The relevance of modelled relationships was noted to be clearly dependent on geographical scale. The models have good predictive ability at a 24 × 24-km resolution, but in resolutions more accurate than 12 × 12 km, much of the predictive power decreases. Our results strongly indicate that scale matters in accessibility analysis, and it should be taken into account in forthcoming population change studies.