Register data versus survey data for estimation of national forecasting models of car type choice

Abstract The car traffic sector faces a major challenge in the coming decades as CO2 emissions must be drastically reduced. A significant part of this transition is likely to occur through the electrification of the car fleet, which implies that new car types will be introduced on the market. Various transport policies can be used to boost this change towards electric vehicles. A car type choice model is essential for understanding which transport policies will achieve the intended effects and which are cost-effective. Such models have been used in Sweden for many years. Until now, car type choice models have either relied on register data without socioeconomic information or on survey data including socioeconomic information. Relying on survey data has unfortunately become increasingly difficult because of low response rates. However, opportunities to estimate car type choice models using national register data that include socioeconomic information have recently become available in Sweden. In this paper, we estimate a car type choice model based on register data including socioeconomic variables and compare it with an existing survey-based model to examine the advantages and disadvantages of the two data sources. Our findings show that a car type choice model estimated on register data can match the results of a survey-based model, suggesting that register data including socioeconomic variables may well be the preferred data source for this kind of models in the future.