Data for: Evaluating heterogeneity in household travel response to carbon pricing: a study focusing on small and rural communities

The spike in gasoline and diesel fuel prices during the spring of 2022 provided a unique opportunity to evaluate how changes in transportation costs affect travel behavior, including changes in small and rural communities. We created an internet-based survey that asked people living in Vermont how they responded to the recent increase in gasoline and diesel fuel prices and what they plan to do if prices remain high. We included questions about changes in the vehicles they use and plans to purchase more fuel-efficient or electric vehicles. We also asked about changes made to travel for essential trips, including trips for work, education, medical appointments, and food, and less essential trips including visiting friends and family, recreational activities, and going to social events. For each trip type, essential and less-essential, respondents could make one or more selections from a list of possible actions that could have been taken in response to higher fuel costs. Actions included using alternative modes of transportation, completing more activities at or from home, carpooling and ridesharing, seeking closer destinations for essential and non-essential activities, trip chaining, moving to a more transportation-efficient location, and adjusting household budgets. Similarly, respondents were asked to indicate the likelihood that they would make any of the previously listed actions if prices were expected to remain high for more than one year on a 5-point Likert scale. We also include a set of questions about potential barriers to taking action to avoid higher fuel costs. This included questions about concerns surrounding EV range, charging access, and performance. We also asked if respondents would change (decrease or increase) the amount of driving they do if various improvements to public transportation and active travel infrastructure were made or if the number of local employment, shopping, and education opportunities were increased. Prior research finds that attitudes and beliefs can be significant factors in explaining travel behavior and choices; therefore, we also consider attitudes and beliefs in our study. We asked about attitudes and beliefs related to sustainability, car culture, and dependence, the role of government, and technological advancement. We hypothesized that these attitudes and beliefs are not only important in explaining how people respond to increasing fuel prices but that they may also vary across community types (urban to rural). Lastly, we asked respondents to describe the type of community they live in (rural, suburban, or urban) and provide the name of the town where they reside along with a standard set of socioeconomic questions.

Methods
We began by recruiting participants using a geolocated database of about 40,000 Vermont e-mail addresses (geolocated to the town level) obtained from a marketing company. The sample collected from this recruitment method skewed much older than the Vermont population but was otherwise broadly representative. We therefore recruited additional participants through Facebook and Instagram advertisements. The sample collected from the social media advertisements was, on average, 29 years younger. All participants were given a chance to enter a drawing for one of ten cash cards each worth $50, as an incentive. The survey was distributed in March 2022 and received 911 responses. After filtering out surveys that were less than 50% complete (these were mostly surveys that were started but never completed), the final size of the sample used in our analysis was 749. Missing values in the final sample were imputed in R using the MICE package. Numerical variables we imputed using Predictive Mean Matching (PMM) and categorical variables were imputed using Polytomous Regression (Polyreg) which is used for categorical variables with two or more than two levels. We used principal axis factor analysis with an orthogonal (varimax) rotation to reduce the number of attitudinal and behavioral variables and to identify latent attitudinal factors using the R psych package. Based on the evaluation of a scree plot of eigenvalues, we determined that four factors were optimal and labeled them as technology concern, car travel enjoyment, environmental concern, and political activity. Factors scores were estimated for each respondent using the Thurston method (a regression approach)