Abstract Manager

Transportation researchers have been investigating equity in transportation and specifically in the provision of transportation investment and services for decades.  In Canada, nearly all wide-scale investigations of transportation equity have been based around household income (or more rarely work status) due to the lack of information on race and/or ethnic background on regional travel surveys, such as the Toronto Tomorrow Survey (TTS) or the Metro Vancouver Regional Trip Diary.  Travel studies sponsored by municipalities are increasingly likely to ask about race or ethnic background, but the regional surveys that support regional transportation decision-making have not asked these demographic changes, though the 2022-23 TTS has added these questions.

This study borrows a data fusion technique used to import propensity to consider purchasing an autonomous vehicle (AV) drawn from a City of Toronto survey focused on AV attitudes into the 2016 TTS.  In this case, the researchers have three separate data sources: the 2018 and 2021 waves of the AV attitude survey and a survey Metrolinx conducted to measure COVID impacts on transit usage.  Notably, race and ethnic background were asked in these surveys, as well as household variables that can be used to align these surveys with the TTS and carry out the data fusion.

The first step is to develop a model for predicting the ethnic background of a respondent based upon household characteristics.  The models were developed from the two surveys independently as well as a pooled model, with the pooled model proving to be more satisfactory.  Note that 15% of the records were held-out for testing of the final model.  An additional validation exercise was executed with the Metrolinx data to calculate the synthesized ethnic background variables for every record.  Then the impact of actual versus synthesized data was measured on status as an essential worker, which leads to some interesting conclusions on the data fusion technique.  Finally, ethnic background variables were generated for the 2016 TTS, and the impact on auto ownership and work and non-work travel mode was explored.  As expected, the ethnic background was statistically significant even after controlling for income and work status.  This data fusion approach, while very promising in situations where critical data is missing, should not be considered a replacement for asking income and race/ethnic background on regional travel surveys.