Robustness of statistical models for the relationship between accessibility and regional development – A case study of Swedish municipalities

Abstract

This paper examines the relationship between car accessibility and regional growth across all 290 Swedish municipalities, using accessibility data from the PIPOS platform for 2014 and 2020 and population and gross regional product data for 2020–2023. We estimate a range of regression models, varying the inclusion of control variables, the treatment of outliers, and the separate estimation for municipalities in high-growth and low-growth counties. Simple linear models show a strong positive correlation between accessibility and regional growth. However, this relationship is not robust: adding control variables reverses the effect of accessibility, removing outliers severely weakens the explanatory power, and the relationship disappears entirely in municipalities in low-growth regions. Our main conclusion is that researchers using statistical approaches to analyse the relationship between accessibility and regional development should demonstrate that their results are robust across model specifications, as conclusions can depend heavily on methodological choices.