Closing the transport gender gap
Understanding differences in travel behaviour between genders
Introduction
In a previous blog post, Theodore introduced the value of using A2BM (activity and agent-based modelling) techniques to explore equity aspects of transport policies. An equity perspective helps us describe who benefits and disbenefits from potential policy interventions.
To further our modelling capability in equity, we’ve produced an analysis to identify differences in travel behaviour between women and men. This post discusses survey data where respondents have self-identified as female or male. However, we recognise how one identifies is personal, and this categorisation of gender may not capture the diversity of identities an individual may carry.
What’s gender got to do with it?
Gender-specific travel behaviour studies emerged in the 1970s, but transport planning continues to model the behaviour of the “average” user and relies on tools that do not disaggregate information to a finer granularity.
The OECD lists gender, household composition, income, and car ownership as the most important factors influencing mobility choice, although there are regional differences depending on the country of study. Research consistently indicates that gender differences should be considered in transport planning and our clients want to understand how gender influences mode choice and how genders respond differently to interventions.
A quantitative perspective
At the City Modelling Lab, our modelling utilises national travel surveys (NTS) and other publicly available data to describe travel behaviour. In England, the NTS is an annual anonymised household survey that monitors personal travel. It describes how, when, why, and where people travel and their individual characteristics (e.g., household structure, vehicle ownership, working status). An “Attitudes” survey accompanies this and interviews those over 16 years of age on their travel habits. This type of information may provide us with explanatory clues for how and why people travel.
The NTS datasets are massive, with nearly 5 million rows of data on trip behaviour and nearly 30 different attributes about each person. To narrow our line of inquiry, we developed a set of hypotheses through consultation of research on travel behaviour. Our hypotheses consider key factors of safety, access, reliability, comfort and convenience, the number and complexity of trips, children in the household, and propensity to walk/cycle or use public transport.
We then built a set of statistical models to test these hypotheses against variables of gender and additional controls. We analysed the Attitudes survey to determine whether women had a statistically significant difference in response from men. The combined analysis of travel diaries and survey results provides context to some of the travel behaviour observed. Below are two samples of this analysis.
How women travel and when
Have you ever changed your transport mode, or the route you typically take because of the time of day or how nice the path is? Do you notice if someone of another gender makes these considerations? Our analysis finds that both gender and daylight are statistically significant drivers of mode choice. Compared to men, women are more likely to walk or use a bus, but less likely to use rail or bike. After dark, the probability of choosing walk, bike, or bus drops for all users drops even further for women. Furthermore, women have a higher mode share of using the car after dark as compared to men.
To validate that this behaviour is statistically significant, we developed a multinomial logit model that describes the probability of choosing a specific mode. We specified the choice model to estimate this probability by controlling for gender, travel after dark (after sunset, before sunrise), and the interaction between these two variables.
When examining the interaction of trips where the surveyee is female and travelling after dark, we found that the probability of choosing to travel by bike, bus, and walking decreases at a statistically significant rate, validating that females are less likely to use these modes after dark.
Within the Attitudes survey, people were asked about the barriers to walking more. The top three responses that correlated most with female respondents included personal security concerns, ill health, and poor street lighting.
How children influence transport behaviour
Prams, schools, nap schedules! Travelling with a child can introduce a whole different set of needs that must be accommodated. Research in Finland has found that women make more “escort” trips than men — a trip where the primary purpose is to accompany someone else on their trip — and this does not depend on access to a car. Analysis from TfL also found that women tend to make more trips than men.
To determine whether the NTS data reflects this behaviour, we fit an ordinal logit model to estimate the number of trips and a binomial logit model to estimate the probability of at least one escort trip based on a person’s gender and parenthood.
In both models, we found that gender and parenthood have a statistically significant impact on the number of trips and the likelihood of escort trips, which may imply women make more trips due to escort trips. Within the Attitudes survey, there aren’t many questions around how children may impact a person’s travel behaviour, so it is challenging to extract further context.
Proceed with caution
The NTS trip diary information enables us to confirm that there are gendered differences in travel behaviour. The Attitudes survey sheds a bit of insight as to why there might be differences.
However, additional surveys and other methods (such as participatory workshop discussions and focus groups) are needed to allow for open conversations around sensitive topics to increase our understanding of gender differences and the latent potential in travel demand.
We understand how women are travelling, but it isn’t transparent what the desired trip experience looks like. The Attitudes survey doesn’t evaluate every specific trip a person makes, so we need to be careful in how we draw conclusions between one dataset that articulates daily behaviour and another that seeks more generalised feedback. The data doesn’t explicitly link how lifestyles shape transport behaviour, nor does it predict how a change in transport options may influence lifestyle changes. Will women be more likely to walk or cycle after dark if there is more well-lit and segregated infrastructure, or will the need to escort children and other home obligations ultimately determine the mode and time of travel?
What’s next?
This blog investigates the travel behaviour differences between women and men and women with children and women without. Other attributes of identity such as: employment, disability, and car ownership, were explored but are not included in this brief review. Given the range of attributes that can be modelled, it is important to conduct an equity exercise when scoping the model to ensure we are modelling responses appropriate for the policy intervention in question. The research we’ve put together has informed recommendations on incorporating gender-specific concerns in modelling and appraisal exercises.
A2BMs, with their ability to model behavioural response at an agent level, are a useful tool to measure the benefits and burdens of transport policies to specific groups of the population. In our modelling, we are especially keen to make the drivers of mode choice more transparent. This capability to model the levers of change will enable our clients to design policies that address these drivers and hopefully unlock pathways to more sustainable and equitable transport systems.
Divya is a senior data scientist at Arup’s City Modelling Lab, specialising in data science applied to transport decarbonisation policies.
Theodore is a senior data scientist and R&D Lead at Arup’s City Modelling Lab.
Paola Bueno is the Social and Equity Lead of Arup’s Transport Policy, Strategy and Economics Business area.
