Attitudes have been used as explanatory variables of travel behaviour for decades, typically under the assumption that there is a causal effect of attitudes on behaviour. However, recent research has shown that the relationship between attitudes and travel behaviour is bi-directional. In this study we use a longitudinal modelling technique on panel data to 1) separate within-person effects from between-person associations and 2) test whether the within-person effects changed during the COVID-19 pandemic. We find that the within-person effects were weaker during the pandemic than they were before the pandemic. In addition, the within-person effects were much smaller than would be expected based on methods that do not separate within-person effects from between-person associations. This means that researchers should be careful when basing policy recommendations on cross-sectional correlations between attitudes and behaviour for two reasons: first, the problem of endogeneity, and second, the highly relevant separation of within-person effects from between-person relations.

@en

Societal and political attention to the effects of climate change and possible mitigation and adaptationpolicies has increased sharply in the last decades, resulting partly from increasing awareness about therole of humanity and partly from the ever more noticeable changes in our world caused by climatechange. This societal interest has highlighted a lack of knowledge about the effects that a changingclimate will have on many aspects of our lifes, one of which is the transport section. To understand theimpact climate change will have on our transport system we need to know how travel behaviours areaffected by weather circumstances, which is the main topic of our research.We focus on four aspects of the relationship between weather and travel behaviour: (1) how weatheris taken into account in the decision-making process (2) if the influence of singular weather variables(such as temperature) depend on the value of other parameters (3) if the influence of weather is differentfor urban and rural areas and (4) whether there are groups of people whose response to travelbehaviour are distinctly different from one-another. This knowledge can be used for climate changeadaptation measures, such as ensuring that our supply of travel infrastructure will be able to copewith changes in travel demand resulting from a changed climate, and mitigation measures, such asincreasing the number of people that use more sustainable travel options like the bicycle.For our analyses we use travel data provided by the KiM Netherlands Institute for Transport PolicyAnalysis, which is the result from a travel diary survey held in autumn. We use weather data asmeasured by weather stations, provided by the Royal Netherlands Meteorological Institute (KNMI).This data is used to estimate the influence of weather on travel demand and mode choice in the Netherlands,using regression and choice models respectively. Within our analyses we try to find factorsthat moderate the relationship between weather and travel behaviour, such as urban density and sociodemographics.With respect to the four aspects identified above, we report the following findings:(1) that people use a general perception of the weather during the whole day for their mode choicetravel decisions, which contrasts with the most common practice of using the weather at the trips’departure time.(2) by accounting for the fact that meteorological variables always co-occur in our models we areable to more accurately capture its effect on travel behaviour. The difference with the current practiceof estimating separate effects for each weather variable is particularly stark for days at the extreme endof the observed range of weather variables.(3) The influence of weather on travel behaviour differs more qualitatively between rural and urbanareas: the total effect size of the weather similar, but they are brought upon by different weathervariables. The difference is also very specific to travel modes. For bicyclists the effects of wind speedseem to be more sizeable in urban environments, whilst temperature, rain, and sunshine have smallereffects in urban environments.(4)We find multiple groups of travellers whose responses to weather variations are different from oneanother. These differences seem to be caused by the set of travel modes that are used during averageweather conditions. People that only use the car during average conditions are not very affected, withonly enjoyable weather conditions prompting increased bicycle use. If the car and the bicycle are usedoften people swap between the modes, although use of the bicycle during inclement conditions isrelatively much higher than for the other two groups. The last group has a more multi-modal travelpattern, which results in the largest variations caused by weather. Inclement conditions favour bothpublic transport and the car, with car use increasing quite sharply during wet weather with high windspeeds.
Additionally we find that weather variations account for differences in travel behaviour across boththe spatial and temporal dimensions. A particularly surprising finding is that the smaller number ofbike trips in the western provinces of the Netherlands can be fully explained by the fact that there arehigher average wind speeds and lower temperatures in this part of the country.
Our results have several implications for the research community and policy makers. We advice researchersto account for the fact that the weather is perceived as a whole and thus that the effect ofone single variable (such as temperature) will depend on the values of other variables. We also foundinteresting subgroups with different reactions with regards to weather. We advice researchers to moreclosely investigate the effects of weather for the separate subgroups. Finally we find sizeable differencesin the effect of weather between different regions, even within the relatively small country of theNetherlands. Researchers studying a relatively large study area would do well to estimate separateeffects for regions within their study area, for example based on population density and geographicallocation.
For policy makers our findings imply that there is a sizeable effect of weather that could be usedto improve the forecasts of future travel demand, both in the short- and long terms. Whilst policymakers obviously can’t control the weather, we have found that changing travel patterns or attitudesto travel modes will have repercussions for the effect weather has on travel behaviour. We think thatpolicies aimed at allowing commuters to gain experience with using the bicycle for their daily commuteduring summer, coupled with temporary financial incentives when weather conditions becomeless favourable, could be one way of achieving more cyclists during inclement conditions. Policy makerscould even target younger professionals specifically, as they are much more likely to have alreadydeveloped such habits during their education.