Transportation Research Part F 78 (2021) 259–266

Contents lists available at ScienceDirect

Transportation Research Part F

journal homepage: www.elsevier.com/locate/trf

How to nudge drivers to reduce speed: The case of the left-digit

effect
Enrico Rubaltelli a,⇑, Dario Manicardi a, Federico Orsini b, Claudio Mulatti c, Riccardo Rossi b,
Lorella Lotto a
a
Department of Developmental Psychology and Socialization, University of Padova, Padova, Italy
b
Department of Civil, Architectural and Environmental Engineering, University of Padova, Padova, Italy
c
Department of Psychology and Cognitive Science, University of Trento, Trento, Italy

a r t i c l e i n f o a b s t r a c t

Article history: To decrease the negative consequences of a road crash, even a small reduction in driving
Received 18 September 2020 speeds can make a significant difference. We examined the possible application of the
Received in revised form 6 January 2021 left-digit effect as a nudge to reduce road speed. In the marketing field, this effect is based
Accepted 17 February 2021
on reporting price tags that are characterized by a low leftmost number and a high right-
most number (e.g., a price tag of €14.99 rather than €15.00). We applied the same strategy
to improve road safety. Participants were college students (43.75% female, mean age = 24.
Keywords:
06 years in Study 1; 50% female, mean age = 23.53 years in Study 2) who were asked to
Nudging
Traffic psychology
drive in a simulator on a route that had both usual unmodified road signs (e.g., 50 km/h)
Left-digit effect and the same road signs increased or decreased by one unit (e.g., 49, 51 km/h). We com-
Road safety pared the average median speeds in road segments with unmodified road signs with those
Driving behavior in road segments with the corresponding modified signs. The average median speeds in the
presence of a sign modified by the reduction of 1 km/h were significantly lower compared
with the median average speeds recorded with unmodified signs. We showed that the
application of psychological insights can reduce driving speeds and therefore increase road
safety.
Ó 2021 Elsevier Ltd. All rights reserved.

1. Introduction

A recent study showed that driver-related factors were present in almost 90% of crashes (Dingus et al., 2016). According to
the World Health Organization (WHO, 2018), in 2016 road crashes were the eighth leading cause of death in the world and
the leading cause of injuries. Specifically, speeding contributed to around 30% of these deaths in high-income countries and
up to 50% in some low-income and middle-income countries (WHO, 2018). For a comparison, 2017 data showed that, in the
United States, 29% of crashes involved alcohol-impaired drivers (National Highway and Transportation Safety
Administration, 2019).
Prior research has shown that increasing the speed limits can lead to higher travel speeds and an increased probability of
exceeding the new speed limit, resulting in a potential increase in crashes that are fatal or cause injury (Hu, 2017). Crucially,
it has been shown that a 5% increase in average speed leads to an increase of about 10% in crashes that cause injury to people
and a 20% increase in fatal crashes. Furthermore, pedestrians have a 90% chance of surviving if hit by a vehicle traveling at

⇑ Corresponding author at: University of Padova, Department of Developmental Psychology and Socialization, via Venezia 8, 35131 Padova, Italy.
E-mail address: enrico.rubaltelli@unipd.it (E. Rubaltelli).

https://doi.org/10.1016/j.trf.2021.02.018
1369-8478/Ó 2021 Elsevier Ltd. All rights reserved.
E. Rubaltelli, D. Manicardi, F. Orsini et al. Transportation Research Part F 78 (2021) 259–266

30 km/h, but less than 50% if the vehicle travels at more than 45 km/h (WHO, 2018). Based on a recent study, the reduction of
just 1 km/h would lead to an average reduction of 8.3% in fatal crashes (Elvik et al., 2019), which means saving up to 2100
lives each year only in the European Union.
Numerous studies have been conducted over the years to find the causes of different driving behaviors. For examples,
researchers have found that male drivers tend to engage in riskier behaviors more than women, such as driving at higher
speeds or performing dangerous maneuvers (Evans & Wasielewski, 1983; French et al., 1993). Similarly, younger and less
experienced drivers tend to risk more than those who are older or more experienced (Fildes et al., 1991; Galin, 1981). It
has also been shown that, although the motivations to put oneself at risk vary from individual to individual, most drivers
take risks unintentionally (Musselwhite, 2006) and, despite being aware of the dangers, many underestimate the risk asso-
ciated with driving (Lichtenstein et al., 1978). Finally, Elliott et al. (2003) showed that psychological factors can explain the
rate at which drivers comply with speed limits.
It follows that an intervention that aims to reduce the risk of crashes must take into account the psychological and behav-
ioral features of road users and, indeed, should focus on and take advantage of them to make roads safer. Specifically, we
aimed to test, using a driving simulator, an intervention based on behavioral science insights. Borrowing from marketing
(Lin & Wang, 2017; Manning & Sprott, 2009; Thomas & Morwitz, 2005), we modified the speed limit information reported
on usual road signs so that it could induce drivers to reduce their speed.
In this regard, behavioral science insights have become popular in recent years and have provided valuable solutions to
improve citizens’ behavior (Halpern, 2016; Thaler & Sunstein, 2008). One of the main approaches in this field is nudging, a
form of choice architecture that encourages individuals to adopt socially desirable behaviors, aimed at improving public
health and individuals’ well-being without limiting their freedom of choice (Thaler & Sunstein, 2008). In the field of road
safety, there are several nudge interventions that have been developed over the years. Some of these interventions are aimed
at increasing safety levels by assisting the driver (the Advanced Driver Assistance Systems; Brookhuis et al., 2019). Other
nudges take advantage of reward mechanisms, for example, a camera installed in Stockholm called ‘‘speed camera lottery”
rewards drivers who respect speed limits with a cash prize derived from the fines of those who do not respect them
(Zichermann & Cunningham, 2011). Finally, some interventions have been developed to increase road safety by applying per-
ceptual countermeasures such as optical speed bars, raised rumbles, chevrons, or optical circles (Allpress & Leland, 2010;
Hussain et al., 2018; Jamson et al., 2010; Martínez et al., 2013; Rossi et al., 2014). Nudges do not come without limitations
and/or potential backfiring effects (Sunstein, 2017; Willis, 2013). Nevertheless, this approach has shown very promising
results and can be very useful when the interventions are tested and assessed beforehand to ensure that they do not produce
any negative side effects (Thaler & Sunstein, 2008).
Our goal was to take advantage of the so-called left-digit effect, a strategy that is typically employed in marketing, and
use it to improve road safety. This effect is based on reporting price tags that are characterized by a low leftmost number and
a high rightmost one (e.g., a price tag of €19.99 rather than €20.00). Several studies have shown that a change in the leftmost
digit of a price significantly affects the perception of the magnitude (Lin & Wang, 2017; Manning & Sprott, 2009; Thomas &
Morwitz, 2005). The explanation for this effect is based on the analogue model of numerical cognition (Thomas & Morwitz,
2005). The model suggests that multidigit numbers are converted into an analogue representation (Dahaene, 1997); based
on this view, when people read a multidigit number, the leftmost digit exerts a disproportionate influence on encoding. As a
result, because an individual’s cognitive resources are focused on the leftmost digit, the higher rightmost one carries less
weight in encoding. In the domain of pricing, this means that a €2.99 price is encoded as much lower than a price of
€3.00 (Manning & Sprott, 2009).
We similarly modified the numbers on road signs that indicate the speed limits. The objective was to alter the perception
of the speed limit and make it feel less than what it actually was. We reduced the numbers on existing signs by a single digit,
so that the leftmost digit was reduced in turn (e.g., from 50 to 49 km/h). On a driving simulator, we then recorded the speed
held by drivers in the presence of both these new, unusual signs and the traditional ones. We hypothesized that, despite the
minimal difference (1 km/h), driving speeds should be lower with the modified signs compared with the traditional ones.

2. Experiment 1

2.1. Method

2.1.1. Participants
Thirty-two college students (43.75% female; M = 24.06, SD = 1.56 years) voluntarily participated in this experiment. They
had normal or corrected-to-normal vision and completed the study individually. All participants declared that they had held
a valid driving license for at least 1 year. The chance to participate in the study was advertised in several courses at both the
engineering and psychology departments.

2.1.2. Materials and procedure

The experimental apparatus consisted of a dynamic driving simulator that has been validated in several previous studies
(Rossi et al., 2018, 2020). The main components of the dynamic driving simulator used for the present study included: (i)
three PCs with an Intel i7 processor, 3.4 GHz frequency, two DDR3 2 Gb RAM modules, a Nvidia GeForce GTX 760 video card,

260
E. Rubaltelli, D. Manicardi, F. Orsini et al. Transportation Research Part F 78 (2021) 259–266

and Microsoft 7 Professional OS; (ii) five 60-inch plasma displays with 1920  1080 pixel resolution, arranged side by side to
cover a view greater than 300°; (iii) a control display with keyboard and mouse for the operator; (iv) a CKASÒ mobile plat-
form with the simulator cockpit on top, equipped with an adjustable seat, safety belt, dashboard with ignition lock, steering,
lever for turn signals and headlights, five-speed manual transmission and reverse gear, parking brake and pedal board with
clutch, brake, and throttle; (v) a 5.1 surround sound system consisting of three front speakers, two positioned at the rear of
the driver’s head and a subwoofer located on the side of the pedal board.
The test procedure was structured into three different phases. In the first phase (familiarization), each participant was
seated on the simulator and was invited to get into a normal driving position. Subsequently, a short driving session was
started on a different track than the test one; this track was specially designed to allow the participants to become familiar
with the simulator. The familiarization phase lasted about 5 min. In the second phase (driving test), the experimental track
was then uploaded to the simulator and instructions were communicated to participants. They were asked to drive as sim-
ilarly as possible to reality and not to talk to the experimenter except for retiring from the study. Each participant was also
advised of the possibility of interrupting the test at any time (e.g., in case of nausea). Finally, after completing the driving test,
participants completed the third part of the experiment (recall task) in which they were asked to perform both a sponta-
neous and a suggested recall of the experimental stimuli.
The driving route was characterized by very wide curves, alternating a curve to the left and a curve to the right. The road
had two lanes and was a two-way street, with a constant width and without any road markings other than the edge and the
center lines. The only road signs were those indicating the speed limits, positioned every 500 m. The software was pro-
grammed in a way that, as the driver proceeded along the route, he/she encountered some vehicles all moving in the oppo-
site direction except for one that was following him/her in the same direction. This vehicle respected speed limits, safety
distances, and could not pass the subject’s vehicle.
The driving route was repeated three consecutive times for each participant, without pauses, to create the illusion of trav-
eling on a single 30 km road. In each 10-km repetition of the route, three different segments were identified. These segments
had the following length and speed limit: 2 km with a limit of 70 ± 1 km/h, 2 km with a limit of 50 ± 1 km/h, and 6 km with a
limit of 90 ± 1 km/h. The 50 ± 1 km/h segment reproduced an urban road, whereas the other two segments reproduced sub-
urban roads. Since the 10 km road was repeated three times, we were able to present each version of a speed limit (e.g., 49,
50, and 51 km/h) within subjects. We counterbalanced the order in which each version of the speed limit was presented by
creating four alternative forms of the track (see Table 1 for details).
To have the same baseline and a common starting speed limit for all participants, we decided to have always the 70 km/h
speed limit in the first 2-km segment of the track. Choosing a sign with a rounded speed limit for the first segment, like those
commonly found in real-life roads, allowed us to make the start of the simulation look more realistic. The longest segment
(6 km) was purposely associated with the highest speed limits (90 ± 1 km/h).
To assess whether the participants remembered (and therefore attended to) the signs encountered along the track, they
were asked to complete two recall tasks: a spontaneous recall first and then a suggested recall. In the spontaneous recall
task, participants were asked to report the first three speed limits that they remembered from those seen during the test.
In the suggested recall task, 21 speed limits were shown: 9 of them were actually present in the track, while the other 12
were randomly chosen among signs that are present on real roads (e.g., 40, 60, and 80 km/h) or fictional (e.g., 39, 41, and
56 km/h). In this suggested recall task, participants were asked to mark all the limits that they remembered having encoun-
tered while driving on the simulator.

2.2. Results

We calculated the average speed recorded for each of the nine speed limits (i.e., from the first road sign indicating one of
the speed limits to the first one indicating a different speed limit) for each participant. We conducted a Shapiro–Wilk test to
assess the normality of the data. The data did not follow a normal distribution; therefore, for the analyses, we used non-
parametric tests assessing the equality of the medians for different groups.
We conducted a Kruskal–Wallis test to assess whether the order of presentation of the nine speed limits signs had an
impact on the median speeds at which participants drove along each of the four tracks. The effect of order was not significant
(ps  0.06), and thus we pooled together the data and did not include this variable in the following analyses. To compare the
speeds recorded for each of the signs, we performed a series of one-tailed Wilcoxon tests. Specifically, we looked at changes

Table 1
Pattern of road signs in Experiment 1.

Length of segments (km)

2 2 6 2 2 6 2 2 6
Track 1 signs (km/h) 70 51 89 71 49 90 69 50 91
Track 2 signs (km/h) 70 51 89 69 50 91 71 49 90
Track 3 signs (km/h) 70 49 91 71 50 89 69 51 90
Track 4 signs (km/h) 70 49 91 69 51 90 71 50 89

261
E. Rubaltelli, D. Manicardi, F. Orsini et al. Transportation Research Part F 78 (2021) 259–266

in speed between the unmodified version of the sign and each of its two modified versions (e.g., 49 vs. 50 km/h and 50 vs.
51 km/h). In this way, it was possible to assess, for each pair of signs, whether participants were slowing down significantly
in correspondence to the lower speed limit.
The results showed that the median speeds in the presence of the 49 km/h sign were significantly lower than the median
speeds in the presence of the 50 km/h sign, whereas there was no significant difference between the median speeds in the
presence of the 50 and 51 km/h signs. Similarly, median speeds were significantly lower when participants were presented
with the 89 km/h sign rather than the 90 km/h one, whereas no significant difference emerged between speeds recorded for
the 90 and 91 km/h signs. Finally, when looking at the 70 km/h sign and its variations, both differences were significant. In
particular, participants drove at higher speeds in the presence of a variation (69 or 71 km/h) than in the presence of the
unmodified sign (see Table 2).
When looking at the two manipulation checks, we found that in the spontaneous recall task, all participants correctly
reported the three speed limits. Answers to the suggested recall task showed that the new signs were reported by at least
62.07% of the participants (see Table 3), while no more than 20.69% of the participants reported a sign that was never
presented.

2.3. Discussion

Our hypothesis was that participants drive at lower speeds when presented with a road sign modified according to the
left-digit effect (and therefore with the reduction of the speed limit by 1 km/h) than when presented with the unmodified
speed limit sign. The results showed significant differences for the road signs reporting a speed limit of 49 (vs. 50) and 89 (vs.
90) km/h, but not for the sign reporting a limit of 69 (vs. 70) km/h.
Because all participants had begun the experiment seeing the 70 km/h sign in the first segment of the track and this is the
only case in which our manipulation was not effective, we believed that the participants maintained lower speeds at the
start because they had not acquired sufficient familiarity with the new track. To investigate this hypothesis, in Experiment
2 participants were presented with the 90 km/h sign (instead of 70 km/h) at the beginning of the track. As a result, in the new
study we were able to test whether there was a difference between 69 and 70 km/h when this sign was not presented at the
start of the track; similarly, we expected to find no effect for the 90 km/h sign.

3. Experiment 2

3.1. Method

3.1.1. Participants
Thirty-two college students (50% female; M = 23.53, SD = 0.59 years) voluntarily participated in this experiment. They had
normal or corrected-to-normal vision and completed the study individually. All participants declared that they had held a
valid driving license for at least 1 year. The chance to participate in the study was advertised in several courses at both
the engineering and psychology departments.

3.1.2. Materials and procedure

By and large, the design for Experiment 2 was the same as in Experiment 1. However, we made a few changes to the
placement of some of the speed limits. In Experiment 2, we switched the position of the 70 and 90 km/h signs. Similarly,
we switched the positions of the 71 and 91 km/h signs as well as the position of the 69 and 89 km/h signs. As a result, in
this experiment, the first segment was always associated with the 90 km/h speed limit (see Table 4 for details).
Since changing the position of the signs meant that speed limits in the final and longest segment were lower (with a con-
sequent increase in travel time) than the previous study, the route was shortened by 4 km to reduce the possible influence
that fatigue can have on participants’ attentional resources (Gastaldi et al., 2014).

Table 2
One-tailed Wilcoxon test results for Experiment 1.

Wilcoxon with Hyp: less Wilcoxon with Hyp: greater

Speed limit (km/h) Median speed (km/h) p z p z
49 vs. 50 53.67 vs. 56.74 0.021 2.314 0.980 0.025
50 vs. 51 56.74 vs. 56.16 0.839 0.203 0.166 1.387
69 vs. 70 73.44 vs. 67.03 1 0 0.002 4.752
70 vs. 71 67.03 vs. 71.21 < 0.001 5.868 1 0
89 vs. 90 84.25 vs. 85.54 0.034 2.121 0.968 0.041
90 vs. 91 85.54 vs. 83.79 0.844 0.197 0.160 1.402

262
E. Rubaltelli, D. Manicardi, F. Orsini et al. Transportation Research Part F 78 (2021) 259–266

Table 3
Percent correct responses of the suggested recall task for Experiment 1.

Signs 49 50 51 69 70 71 89 90 91
% correct 62.07 65.52 72.41 65.52 37.93 65.52 79.31 65.52 86.21

Table 4
Pattern of road signs in Experiment 2.

Length of segments (km)

2 2 2 2 2 2 2 2 2
Track 1 signs (km/h) 90 51 69 91 49 70 89 50 71
Track 2 signs (km/h) 90 51 69 89 50 71 91 49 70
Track 3 signs (km/h) 90 49 71 91 50 69 89 51 70
Track 4 signs (km/h) 90 49 71 89 51 70 91 50 69

3.2. Results

We conducted the same tests on the data as those described for the first experiment. The Kruskal–Wallis test showed
again that the order of the signs was not significant (ps  0.06), so we pooled together the data and did not include this vari-
able in the following analyses. The Wilcoxon test showed that the median speeds in the presence of the 49 km/h sign were
significantly lower than the median speeds in the presence of the 50 km/h sign, whereas there was no significant difference
between the median speeds in the presence of the 50 and 51 km/h signs (as in Experiment 1). Similarly, median speeds were
significantly lower when participants were presented with the 69 rather than the 70 km/h sign, whereas there was no a sig-
nificant difference between speeds recorded for the 70 and 71 km/h signs. Finally, when looking at the 90 km/h sign and its
variations, both differences were significant. In particular, participants drove at higher speeds in the presence of a variation
(89 or 91 km/h) than in the presence of the unmodified sign (see Table 5).
When looking at the two manipulation checks, we found that in the spontaneous recall task, all participants correctly
reported three speed limits. Answers to the suggested recall task showed that the new signs were reported by at least
56.25% of the participants (see Table 6), while no more than 12.50% of the participants reported a sign that was never
presented.

3.3. Discussion

Experiment 2 replicated the previous finding showing that participants significantly reduced their speed when presented
with a sign one unit lower than the usual one (e.g., 49 vs. 50 km/h). In addition, we showed that the lack of an effect for the
70 km/h sign (and its variation) in the first experiment was likely due to its positioning in the first section of the track. Con-
sistent with this explanation, in Experiment 2 we did not find a significant effect for the 90 km/h sign (and its variation),
which was now placed in the first section of the track, replacing the 70 km/h sign.

4. General discussion

The main purpose of this research was to assess the effectiveness of the left-digit effect as a nudge to improve road safety.
Our hypothesis was that the participants should drive at lower speeds in the presence of a road sign modified according to
the left-digit effect compared with another 1 km/h difference that does not involve the aforementioned effect. Both exper-
iments confirmed this hypothesis regarding the 50 km/h sign. Experiment 1 also confirmed this hypothesis regarding the
90 km/h sign and Experiment 2 confirmed it regarding the 70 km/h sign. With regard to the 70 km/h sign in Experiment
1 and the 90 km/h sign in Experiment 2, the effect was likely not significant because they were positioned at the beginning

Table 5
One-tailed Wilcoxon test results for Experiment 2.

Wilcoxon with Hyp: less Wilcoxon with Hyp: greater

Speed limit (km/h) Median speed (km/h) p z p z
49 vs. 50 60.21 vs. 62.48 0.008 2.660 0.993 0.009
50 vs. 51 62.48 vs. 62.76 0.825 0.221 0.180 1.341
69 vs. 70 70.34 vs. 73.55 0.026 2.226 0.975 0.031
70 vs. 71 73.55 vs. 73.94 0.562 0.579 0.445 0.764
89 vs. 90 85.50 vs. 76.40 1 0 <0.001 5.918
90 vs. 91 76.40 vs. 86.74 <0.001 5.918 1 0

263
E. Rubaltelli, D. Manicardi, F. Orsini et al. Transportation Research Part F 78 (2021) 259–266

Table 6
Percent correct responses of the suggested recall task in Experiment 2.

Signs 49 50 51 69 70 71 89 90 91
% correct 81.25 50.00 78.13 71.88 59.38 62.50 81.25 53.13 56.25

of the four tracks. We decided to keep the positioning of these initial (unmodified) speed limit signs constant for all partic-
ipants to make the simulation more realistic. It is likely that participants had driven more prudently when they started the
simulation and were not yet familiar with the route. Indeed, despite participants having a chance to become familiar with
the simulator, the results seemed to indicate that they started the experiment driving at a speed much lower than the limit
(unlike in the following segments of the route). An alternative explanation could be that participants were particularly atten-
tive to the speed limits when they had just started the simulation. At that point, they had not built enough confidence with
the simulator and were more focused on the cues on the road than on simply driving as they would do in real life. However,
this explanation is not supported by the results of the recall task, because in both experiments the signs presented in the first
segment (70 km/h in Experiment 1 and 90 km/h in Experiment 2) are those that were recalled the least often.
Crucially, we found that the median speed maintained by the subjects in the presence of signs modified with a 1 km/h
decrease was significantly lower than the median speed maintained in the presence of the corresponding unmodified signs.
In addition, the median speed maintained in the presence of signs modified with a 1 km/h increase was not significantly dif-
ferent than that maintained in the presence of corresponding unmodified signs. The different pattern of results that we found
by modifying speed limits by +1 versus 1 km/h is extremely important because it allows us to conclude that it is not the gap
in the speed limit that drives the effect. Indeed, it is the leftmost digit that makes people perceive the limit as lower and
therefore leads them to slow down. Otherwise, we should have found a difference in speeds even when the limit on the sign
was increased by 1 km/h.
In the recall task, participants were proficient at recalling both signs that were 1 km/h above (e.g., 51 km/h) or below (e.g.,
49 km/h) a usual one (e.g., 50 km/h). We expected this outcome because these signs are unusual and therefore should attract
people’s attention, something that has been proven to make them easier to recognize (Gregory et al., 2016). However, the
results of our driving simulation were not consistent with an explanation simply based on unusual information, because
people were more likely to change their speed when the sign was 1 km/h below the usual one rather than 1 km/h above it.
We also identified a reduction in driving speeds that is comparable to what has been found in the literature using per-
ceptual countermeasures (Jamson et al., 2010), especially when considering that participants drove on a road with long
sweeping curves and no traffic. Future work should investigate the combined effects of our modified signs with other per-
ceptual cues to assess whether these two measures can help treat sections of road with long straights or long radius corners
in which drivers find it more difficult to reduce speeds. Still, we believe that an important contribution of our work, and a
significant difference from the literature on perceptual countermeasures, is that signs can be repeated often along the road.
By contrast, many perceptual cues cannot be used on long stretches of road or may take up too much room (e.g., central and
peripheral hatching). As a result, these solutions can work well in specific sections of the road (e.g., near schools, in the pres-
ence of road work, or before a dangerous curve). Our work showed that modifying the speed signs could be a way to make
them more effective, and this endeavor could help extend the length of road sections in which an intervention to reduce
speed can be implemented.
Although the results of the present experiments are very promising in terms of finding a simple and effective way to make
drivers reduce their speed, some limitations must be acknowledged. The first limitation is that we ran the experiments on a
simulator. Despite being quite realistic, the simulation lacked some of the complexity that characterizes real-world roads. As
a result of this limitation, it is hard to understand whether this nudge intervention may have unwanted side effects or pro-
duce a backlash from road users once introduced in real life. For example, the ‘‘speed camera lottery” (Zichermann &
Cunningham, 2011) may inadvertently cause an increase in traffic that could lead to other road safety or environmental
issues. However, consistent with the general criteria behind good nudges (Thaler & Sunstein, 2008), we want to stress that
this study presents initial evidence of the effectiveness of the left-digit effect applied to speed limits. Real-life tests in specific
conditions (e.g., close to a school or in the presence of road works) or in short sections of roads are needed to make sure that
this nudge works and does not cause unexpected and potentially unsafe behaviors. This would also be a way to assess the
effectiveness of the intervention and whether it is applicable in light of the well-established procedures used to set speed
limits.
In accordance with the above reasoning, a second limitation is that it could be unrealistic to think that a town council or
country could modify all signs to implement this type of nudge. It would likely be very expensive to perform such an inter-
vention. However, it could be initially trialed in a small environment such as a single neighborhood or a small town and then
extended, if the results justify the investment. A way that could make this intervention more cost effective, flexible, and
implementable in a short- to mid-term period would be to place digital signs indicating the modified speed limit on roads.
Finally, our work does not allow us to draw any conclusions with regard to the effectiveness of this nudge over time. In
other words, it is not clear whether the intervention we tested will last even once the drivers are used to the new signs. An
initial test in the simulator could involve running several experimental sessions with the same participants. This would
make it possible to assess if after a while people get used to the signs and start not paying enough attention to them. Based

264
E. Rubaltelli, D. Manicardi, F. Orsini et al. Transportation Research Part F 78 (2021) 259–266

on the left-digit effect in marketing, we can establish a testable hypothesis and expected results. If, after quite some time,
price tags with a leftmost digit lower than the rightmost one (i.e., €19.99) are still effective, then we also expect to find a
similar result with the modified road signs.

5. Conclusions

The present work shows that the left-digit effect can potentially be used to modify driving behaviors, precisely to impact
the speed at which people drive. This relevant result is consistent with the nudge approach because it did not force drivers to
go slower or impose any cost to them. Therefore, from an applied point of view, the speed differences that emerged in our
experiments when the sign was reduced by 1 km/h are extremely relevant and could make a huge difference in terms of the
consequences of a crash. Indeed, by reducing participants’ speed, we achieved an outcome that previous research has shown
could significantly reduce the number of fatal crashes on roads and save thousands of lives each year (Elvik et al., 2019).
Many efforts and investments have gone into campaigns aimed to make drivers aware of the negative consequences of
driving too fast. However, these campaigns have often fallen short of reaching the goal of reducing the speed at which people
drive. In the present work, we showed an alternative way to reach the same objective by simply modifying a set of informa-
tion that is already present along all roads.

Data statement

Data will be made available upon reasonable request to the authors.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have
appeared to influence the work reported in this paper.

References

Allpress, J. A., & Leland, L. S. Jr., (2010). Reducing traffic speed within roadwork sites using obtrusive perceptual countermeasures. Accident Analysis and
Prevention, 42, 377–383. https://doi.org/10.1016/j.aap.2009.08.014.
Brookhuis, K. A., De Waard, D., & Janssen, W. H. (2019). Behavioural impacts of advanced driver assistance systems–an overview. European Journal of
Transport and Infrastructure, 1, 245–253. https://doi.org/10.18757/ejtir.2001.1.3.3667.
Dahaene, S. (1997). The number sense. Oxford University Press.
Dingus, T. A., Guo, F., Lee, S., Antin, J. F., Perez, M., Buchanan-King, M., & Hankey, J. (2016). Driver crash risk factors and prevalence evaluation using
naturalistic driving data. Proceedings of the National Academy of Sciences of the Unites States of America, 113, 2636–2641. https://doi.org/10.1073/
pnas.1513271113.
Elliott, M. A., Armitage, C. J., & Baughan, C. J. (2003). Drivers’ compliance with speed limits: An application of the theory of planned behavior. Journal of
Applied Psychology, 88, 964–972. https://doi.org/10.1037/0021-9010.88.5.964.
Elvik, R., Vadeby, A., Hels, T., & van Schagen, I. (2019). Updated estimates of the relationship between speed and road safety at the aggregate and individual
levels. Accident Analysis and Prevention, 123, 114–122. https://doi.org/10.1016/j.aap.2018.11.014.
Evans, L., & Wasielewski, P. (1983). Risky driving related to driver and vehicle characteristics. Accident Analysis and Prevention, 15, 121–136. https://doi.org/
10.1016/0001-4575(83)90068-4.
Fildes, B. N., Rumbold, G., & Leening, A. (1991). Speed behaviour and drivers’ attitude to speeding. Monash University Accident Research Centre (Report No.
16). https://www.monash.edu/__data/assets/pdf_file/0003/216426/muarc016.pdf.
French, D. J., West, R. J., Elander, J., & Wilding, J. M. (1993). Decision-making style, driving style, and self-reported involvement in road traffic accidents.
Ergonomics, 36, 627–644. https://doi.org/10.1080/00140139308967925.
Galin, D. (1981). Speeds on two-lane rural roads-a multiple regression analysis. Traffic Engineering & Control, 22, 453–460.
Gastaldi, M., Rossi, R., & Gecchele, G. (2014). Effects of driver task-related fatigue on driving performance. Procedia: Social and Behavioral Sciences, 111,
955–964. https://doi.org/10.1016/j.sbspro.2014.01.130.
Gregory, B., Irwin, J. D., Faulks, I. J., & Chekaluk, E. (2016). Differential effects of traffic signs stimuli upon speeding in school zones following a traffic light
interruption. Accident Analysis and Prevention, 86, 114–120. https://doi.org/10.1016/j.aap.2015.10.020.
Halpern, D. (2016). Inside the nudge unit. How small changes can make a big difference. Random House.
Hussain, Q., Pirdavani, A., Ariën, C., Brijs, T., & Alhajyaseen, W. (2018). The impact of perceptual countermeasures on driving behavior in rural-urban
transition road segments: A driving simulator study. Advances in Transportation Studies, 46, 83–96 http://www.atsinternationaljournal.com/index.php/
2018-issues/xlvi-november-2018/1006-the-impact-of-perceptual-countermeasures-on-driving-behavior-in-rural-urban-transition-road-segments-a-
driving-simulator-study.
Hu, W. (2017). Rising the speed limit from 75 to 80 mph on Utah rural interstates: Effects on vehicles speeds and speed variance. Journal of Safety Research,
61, 83–92. https://doi.org/10.1016/j.jsr.2017.02.006.
Jamson, S., Lai, F., & Jamson, H. (2010). Driving simulators for robust comparisons: A case study evaluating road safety engineering treatments. Accident
Analysis and Prevention, 42, 961–971. https://doi.org/10.1016/j.aap.2009.04.014.
Lichtenstein, S., Slovic, P., Fischhoff, B., Layman, M., & Combs, B. (1978). Judged frequency of lethal events. Journal of Experimental Psychology: Human
Learning and Memory, 4, 551–578. https://doi.org/10.1037/0278-7393.4.6.551.
Lin, C. H., & Wang, J. W. (2017). Distortion of price discount perceptions through the left-digit effect. Marketing Letters, 28, 99–112. https://doi.org/10.1007/
s11002-015-9387-5.
Manning, K. C., & Sprott, D. E. (2009). Price endings, left-digit effects, and choice. Journal of Consumer Research, 36, 328–335. https://doi.org/10.1086/597215.
Martínez, A., Mántaras, D. A., & Luque, P. (2013). Reducing posted speed and perceptual countermeasures to improve safety in road stretches with a high
concentration of accidents. Safety Science, 60, 160–168. https://doi.org/10.1016/j.ssci.2013.07.003.
Musselwhite, C. (2006). Attitudes towards vehicle driving behaviour: Categorising and contextualising risk. Accident Analysis and Prevention, 38, 324–334.
https://doi.org/10.1016/j.aap.2005.10.003.

265
E. Rubaltelli, D. Manicardi, F. Orsini et al. Transportation Research Part F 78 (2021) 259–266

National Highway and Transportation Safety Administration (2019). Traffic safety facts: 2017 data. https://crashstats.nhtsa.dot.gov/Api/Public/
ViewPublication/812780.
Rossi, R., Gastaldi, M., Gecchele, G., Biondi, F., & Mulatti, C. (2014). Traffic-calming measures affecting perceived speed in approaching bends. Transportation
Research Record, 2434, 35–43. https://doi.org/10.3141/2434-05.
Rossi, R., Gastaldi, M., & Meneguzzer, C. (2018). Headway distribution effect on gap-acceptance behavior at roundabouts: Driving simulator experiments in
a case study. Advances in Transportation Studies, 46, 97–110 http://www.atsinternationaljournal.com/index.php/2018-issues/xlvi-november-2018/
1007-headway-distribution-effect-on-gap-acceptance-behavior-at-roundabouts-driving-simulator-experiments-in-a-case-study.
Rossi, R., Meneguzzer, C., Orsini, F., & Gastaldi, M. (2020). Gap-acceptance behavior at roundabouts: Validation of a driving simulator environment using
field observations. Transportation Research Procedia, 47, 27–34. https://doi.org/10.1016/j.trpro.2020.03.069.
Sunstein, C. R. (2017). Nudges that fail. Behavioural Public Policy, 1, 4–25. https://doi.org/10.1017/bpp.2016.3.
Thaler, R. H., & Sunstein, C. R. (2008). Nudge: Improving decisions about health, wealth, and happiness. Penguin Books.
Thomas, M., & Morwitz, V. (2005). Penny wise and pound foolish: The left-digit effect in price cognition. Journal of Consumer Research, 32, 54–64. https://doi.
org/10.1086/429600.
Willis, L. E. (2013). When nudges fail: Slippery defaults. The University of Chicago Law Review, 80, 1155–1229. https://doi.org/10.2139/ssrn.2142989.
World Health Organization (2018). Global status report on road safety: Time for action. https://apps.who.int/iris/bitstream/handle/10665/44122/
9789241563840_eng.pdf?sequence=1.
Zichermann, G., & Cunningham, C. (2011). Gamification by design: Implementing game mechanics in web and mobile apps. O’Reilly.

266