Chapter 2. Literature review
2.2 The influence of ambient light on the decision to cycle
To rise the population of cyclists in the UK and other locations, it needs to be a viable transport mode at all times of the day, including when it is dark. However, there are a number of reasons why darkness may deter people from cycling. For example, it may be harder to see potential hazards. Research involving pedestrians has demonstrated that obstacle detection decreases as illuminance reduces (e.g.Fotios and Cheal, 2009; Uttley et al., 2017) and this is also likely to be true for cyclists. Illuminance is also associated with reassurance (e.g.Boyce et al., 2000; Fotios et al., 2018; Fotios et al., 2015a). For example, when it is dark people may feel less safe and therefore be discouraged from cycling. For prospective cyclists, the fear they will not be seen by vehicle drivers may also demotivate them. This is understandable, as rates of accidents and fatalities among pedestrians and cyclists due to car accidents are higher in poorly lit areas (Eluru et al., 2008). Darkness may therefore increase a prospective cyclist’s perceived risk of colliding with a car, dissuading them from using their bicycle when it turns dark.
The effect of light on the decision to cycle was investigated in several studies. For example, in a longitudinal study Heinen et al. (2011) found women to be less inclined to commute by bicycle during the after dark time. Whereas Spencer et al. (2013) who conducted their analysis on interview and focus group transcripts of 24 cyclists found light level factor to be determinate to the decision to cycle or not.
It therefore seems likely that light conditions may influence whether or not someone chooses to cycle. However, confirming this with robust evidence, and quantifying the size of any effect of darkness on cycling rates, is not straightforward. One approach would be to obtain subjective assessments of the impact of darkness on whether someone is likely to choose to cycle, and how safe they might feel when cycling after dark. Such subjective judgements can, however, be prone to bias and produce misleading conclusions (Poulton, 1977, 1982). This is illustrated by research into the effect of light levels on the pedestrians’ reassurance (Fotios et al.,2018). A number of past studies have assessed whether illuminance levels influence perceived safety after dark by asking participants to provide a subjective assessment of how safe they feel on roads whose average illuminances vary (e.g.Boomsma and Steg, 2014; Loewen et al., 1993; Rea et al., 2015). For example, Peña-García et al. (2015) asked participants to complete a series of rating scales in five different streets, with each street varying
27
in light intensity and lamp colour. Although Pena-Garcia et al. concluded that higher illuminance of road lighting correlates with people reassurance, this is a trivial finding as it fails to address the impact of stimulus range bias (Fotios, 2016). This is because Pena-Garcia and colleagues collected participants’ responses generated by different illuminance levels under varied contexts i.e. different street environments. Had the same street was evaluated and the illuminance levels were the only variable a different set of conclusions would probably have been drawn (Fotios and Castleton, 2017a). A further problem with using rating scales is that individuals may be forced to make a judgment about a phenomenon to which they may otherwise pay little or no attention (Fotios et al., 2015a). Flawed responses are also anticipated as a result of the way the assessment questions are structured (Toomingas et al., 1997). Consequently, concerns have been raised that it may not be possible to generalise the findings of subjective assessments about the impact of light and lighting on a behaviour such as the decision to cycle. An alternative, more objective approach is to examine actual behaviour rather than subjective judgements. This involve counting and comparing the number of cyclists during daylight with the number of cyclists when it is dark. However, this observational approach also has its potential drawbacks. For instance, several aspects increase the likelihood of obtaining cycling as a travel mode such as time of day, weather, and purpose of journey (commuting or pleasure), and these may confound any analysis of the effect of light.
What is required is a method that compares cycling rates at the same time of day, whether this is in darkness or daylight. Such a method of analysis has been conducted before, but in a different context – namely the influence of ambient light on road traffic collisions (Sullivan and Flannagan, 2002). For this analysis, the researchers compared the number of accidents involving pedestrians and vehicles at a given time of day in the weeks immediately prior and afterward of daylight saving time. The daylight-saving time (DST) clock change occurs twice a year, usually around the end of March and October. In March, the national clock in the UK is advanced by one hour and then reverts in October. This change is also implemented in other countries, including in Europe and North America. The aim is to ensure that a greater number of daylight hours can be used during the months of March to October. This also affects the times when dawn and dusk occur, meaning that an hour in darkness before daylight saving will be an hour in daylight once the time has changed. This is especially true in the
28
weeks immediately before and after the time change. With regard to the influence on the daily routine, this means that commuting in daylight before the Autumn clock change will become commuting in darkness after the clocks change. The reverse is then true for the Spring clock change. Sullivan and Flannagan (2002) found a significant difference in the number of road traffic accidents prior and afterward the DST clock change, see Figure 2.1.
Figure 2. 1. Rate of crashes resulting in pedestrian fatalities in the United States from 1987-1997,
prior and afterward the clocks were changed to Daylight Saving Time (DST) (Redrawn from Sullivan and Flannagan, 2002).
The clock-change method has been extended by other researchers to evaluate ambient light effect on the numbers of people walking and cycling, for example, Uttley and Fotios (2017) analysed an established database comprising counts of pedestrians and cyclists over a five-year period from automated counters installed at 31 locations across Arlington, Virginia, USA. The automated counters were located on different types of cycle routes, such as road cycle lanes and cycle tracks. A case hour of 17:00 to 17:59 was chosen for the Autumn clock change, as this time interval fell in daylight prior the clock change and darkness afterward the change. Another case hour of 18:00 to18:59 was selected for the clock change in Spring, where the ambient light was first in darkness and then in daylight after the clock change.
29
Any observed changes may, however, be attributable to other changes prior and afterward clock change, such as changes in temperature and rainfall. To counter this, four control periods were selected where there was no change in light conditions before and after each clock change: day, early day, dark, late dark. Using the odds ratio (OR) as shown in equation 2.1, they divided the frequencies of the case hour taken two weeks prior and afterward the clock change in both Spring and Autumn DSTs and compared this with the control hours.
Where:
A = frequency of pedestrians or cyclists during the case hour in daylight; B = frequency of pedestrians or cyclists during the case hour in darkness; C = frequency of pedestrians or cyclists during control hours when the case
hour is in daylight;
D = frequency of pedestrians or cyclist frequency during control hours when the case hour is in darkness.
The A/B ratio denotes daylight/darkness, where a higher ratio indicates a higher tendency to cycle during daylight than after dark. However, this does not account for other influential factors such as the weather. Assuming that the effect of such factors is consistent throughout the day, the C/D ratio thus serves to weight these changes (Szumilas, 2010).
The overall odds ratio of (all control hours were summed) was 1.38 (1.37 – 1.39 95% CI, p<0.001) indicating a significant effect of daylight on increasing the desire to cycle (see Equation 3.1 in Chapter 3 for calculations of confidence intervals (CI)).
It was concluded that there is a significant increase in cycling during daylight, which indicated that ambient light played a role in motivating active traveling (Uttley and Fotios, 2017).
However, one limitation of this approach is that only a small portion of time was analysed. This raises a concern about the extent to which the findings could be
𝑂𝑑𝑑𝑠 𝑟𝑎𝑡𝑖𝑜 =
𝐴/𝐵
30
generalised. Another limitation relates to the possibility that other peripheral events may influence the decision to cycle, such as public events and public holidays. These may generate outliers during the weeks before and after the time change.
To address these limitations, Fotios et al. (2017b) utilised the same odds ratio method used to assess the impact of darkness on cyclist numbers, but this time over the whole year rather than the short periods prior and afterward annual clock changes. An hour was selected that was in daylight during a segment of the year and dark for the remain of the year. Changes in cyclist frequencies during this hour were again evaluated with changes in control hours, where the light condition was constant over the same period. The overall odds ratio was 1.67 (1.66 – 1.68 95% CI, p<0.001), which again confirmed the negative impact darkness has on the public tendency to cycle.
Both studies (Fotios et al., 2017b; Uttley and Fotios, 2017) used cyclist count data from a single city in the USA. Different countries, and different cities within any given country, may have different tendencies with regard to cycling due to differences in terrain, cycling infrastructure, public activeness, and the correlation between residential, leisure, and industrial areas. For example, Pucher and Buehler (2008) found that the share of cycling trips relative to other means of transportation in the UK is much lower than the mean percentage of Netherlands, Denmark, Germany, and Finland, 1% and 17% respectively, see Figure 1.3. It is therefore not known whether the findings established from one city in the USA are generalisable to other cities within the USA or to locations in other countries. It is, however, reasonable to predict differences in cycling trends between distant countries such as the UK and US given variations in traffic, land use, culture, mean income, and transportation networks (Hallal et al., 2012; Pucher and Buehler, 2008). Further work is therefore needed to determine whether ambient light level influence the decision to cycle in other locations.