New paper – Car segments and fuel types

In a comparatively smoother process than the previous one, the second paper from my PhD thesis has just been published by Applied Energy. It was, again, co-written with my PhD supervisor, Professor Elisabetta Cherchi.

I started working on this paper around 2021, when I was looking at the information available to make sense of the demand for electric vehicles in the UK. The National Travel Survey is the main transport-related household survey in the country, and it has a lengthy history in collecting information about personal travel, including – among many other dimensions – car ownership and use. I set up to analyse fuel types and vehicle types among households in England – as the survey does not cover the other home nations anymore – and found that there were significantly higher proportions of electric and hybrid cars for certain vehicle types – mostly SUVs and MPVs. This has a fairly obvious interpretation, albeit one that does not always appear in choice studies of alternative fuel cars. From the buyer’s perspective, fuel type is just one part of the choice. Vehicle type plays a significant role and is also highly correlated.

Discrete choice models offer has several interesting modelling alternatives for this case. We used a cross-nested logit model which, in short, allows the modeller to address correlation in more than one dimension simultaneously. We used the NTS data in combination with a privately sourced dataset to characterise all the available vehicle alternatives, and the results prove that these choice dimensions are strongly correlated. Then we carried out several simulation and policy analysis.

I think the main contribution of this manuscript is using a national household survey to harness valuable data at the individual level, that can then be used to estimate models that can potentially be useful in policy and practice. Governments and public agencies spend a vast amount of resources in surveys and data collection, and this information is available for everyone, which provides attractive incentives for stronger and practice-focused research, a strong necessity in these troubled times. I feel very proud to have been able to synthesise all this information and arrive at reasonable results (at least that’s what I would like to think).

I must say that one of the reviewers pointed out a significant issue that the first version of the paper had, and this led to much stronger models estimated for the final version. It is always gratifying when reviewers contribute towards better research and science, instead of being pedantic and self-serving, as is also the case sometimes.

The paper is open access and can be downloaded here. Curiously, it has a publishing date of 2024 which means it will still be “new” by next year. I think that’s rather nice.