E-Bike Charging Infrastructure in the Workplace—Should Employers Provide It?

Abstract

Constructing charging infrastructure for e-bikes at home or in other locations is necessary to enable motor support while riding. This paper focuses on charging facilities at work and study locations. It analyzes the charging frequency preference of 281 e-bike commuters who work or study at RWTH Aachen University, using survey data with 1091 choices for hypothetical free charging, as well as the same conditions for hypothetical paid charging. We use a mixed logit model to estimate the factors influencing the charging frequency, focusing on the commuting distance, an e-bike’s resale value, the age of the owner, student status, and employment group. One charging event per day can be expected for four e-bike commuters when free charging is available. In the case of paid charging, there is one charging event per 12 e-bike users. The magnitude of the reduction caused by a charging fee depends on group membership and, probably, on income. Commuting distance only has a statistically significant influence on the charging frequency when charging is free, raising the question of whether charging at work is necessary to cover trip chains that include stops at work. Owners of more expensive e-bikes charge less often, likely due to higher battery capacities, while the influence of age is inconclusive. However, providing charging infrastructure for employees and guests could be used as a low-cost measure to promote cycling among commuters.

1. Introduction

1.1. Background

The average length of trips traveled by e-bikes is longer than that of conventional bicycle trips. Therefore, they are considered to be range extenders that should increase the share of cycling, especially for commuting, and, thus, the sustainability of transportation [1,2]. The ‘range’ of a conventional bicycle is usually based on the rider’s fitness and the difference in travel time compared to other modes, as they do not rely on energy storage. In contrast, e-bikes have exhaustible batteries. But, what does their limited range say about the need to construct charging infrastructure? For example, should companies provide charging infrastructure to encourage commuters to cycle to work?

When an e-bike’s battery is flat, riders can still use human power to cover the remaining distance to their destination or the nearest charging station. However, e-bikes are heavier than conventional bicycles. In addition to containing an electric motor, the battery accounts for 30% of the weight of an e-bike [3]. As a result, using an e-bike without motor assistance is very strenuous. Consequently, the utility of e-bikes depends on the availability of battery power and, thus, the availability of charging infrastructure.

The energy consumed by e-bikes is highly dependent on riding style, as most of them allow different levels of pedal assistance. Some also enable negative assistance, where pedaling recharges the battery, increasing the theoretical range up to the physical limits of the rider. Therefore, it is not possible to determine a precise range for e-bikes [4].

Current batteries have capacities between 250 and 500 Wh, and some batteries have even greater capacities. The range is often quoted as being between 25 and 100 km (15/60 miles) [3,4,5]. Although battery capacity and range have increased in recent years, in part because lithium batteries have replaced heavy lead–acid batteries, the issue of charging location and infrastructure remains.

In general, most e-bike owners charge their bikes regularly at home. Charging infrastructure is usually only required elsewhere for longer trips or trip chains beyond the range of the e-bike. Such trip chains can include stays at work. However, it is questionable whether employees with long commutes need charging infrastructure at work in order to commute by e-bike. Nevertheless, regular charging at work would be a solution for people with no or inconvenient charging options at home.

1.2. Approach

This paper analyzes how often students and employees who commute by e-bike want to charge e-bikes at their place of work or study. We focus on the charging demand of commuters and do not consider the charging of company fleets, sharing systems, etc.

We analyze the influence of the following factors on charging frequency:

• Commuting distance;
• Resale value of the e-bike;
• Age of the owner;
• Free versus paid charging;
• Student status or employment group.

For the analysis, we surveyed students and staff at RWTH Aachen University, one of the largest universities in Germany (45,000 students and 8000 employees). We asked e-bike commuters about their potential charging frequency during a regular work or study week at the university of using both free and paid charging. We use descriptive analysis and mixed logit models to estimate the factors that influence their preferred charging frequencies. This analysis aims to answer the question regarding the extent of the charging infrastructure that universities or employers should provide. We also aim to formulate a recommendation regarding whether free or paid charging of e-bikes should be provided.

1.3. Literature Review

In Germany, annual sales of e-bikes multiplied from 0.3 to 2 million between 2011 and 2021 [6], increasing the relevance of this field. Previous studies focused mainly on the following areas:

• The use of e-bikes, their (potential) influence on mobility, and their environmental and health impacts [1,2,7,8,9];
• The road traffic safety of e-bikes [10,11,12];
• E-bike sharing systems and their (charging) stations [13,14];
• Technical charging methods [15,16].

Studies have shown that e-bikes contribute to more sustainable mobility by increasing cycling’s share of transportation methods [1,2]. E-bike owners are older than conventional bicycle owners, and medical conditions are an important reason for buying an e-bike [1,9]. The physical activity of employees who use e-bikes to commute to work is similar to that of those who use conventional bicycles, resulting in fewer sick days [8,17]. Parked bicycles occupy far less space than parked cars [18]. In addition to the image factor, these are the reasons that entice companies to promote commuting by conventional or e-bike.

Although the provision of e-bike charging infrastructure could encourage e-bike use [7], research has not yet focused on the issue of charging points and the necessary charging infrastructure outside of sharing schemes. As a result, the amount of available scientific literature regarding e-bike charging, especially at places of work or study, is limited [19].

In [4], e-bikers said that they had ‘range anxiety’, which is the fear of running out of battery power during a ride, as well as that they would feel more comfortable taking longer trips if more charging stations were available. The study also found that participants charged their e-bikes at work when commuting up to 20 miles each way.

E-bikes can be charged at dedicated charging stations or via regular power outlets. Apart from the battery capacity, the battery mount varies between e-bike types. There are fixed and removable e-bike batteries. While the charging infrastructure for e-bikes with fixed batteries needs to be built at the parking location, removable batteries can be charged elsewhere [19]. Although regular power outlets are usually available, charging equipment is not standardized. As a result, e-bike users must carry the appropriate equipment if charging is required during a trip chain [20].

Studies have shown that offering a daily financial incentive to cycle to work (up to £5/5.7 EUR) is effective in increasing cycling rates [21]. In comparison, the energy cost of a charging process appears low. As the price of a charging event is usually substantially less than 1 EUR, it is trivial compared to charging a battery electric car [4]. Free charging could, therefore, be a cost-effective measure for employers to encourage employees to commute by e-bike.

In summary, there is a lack of literature on the charging behavior of e-bike commuters and the frequency with which they charge their e-bikes. There is also a lack of scientific sources regarding the necessary charging infrastructure that employers should provide for their employees and guests. For this reason, we analyzed the charging frequency in relation to free and paid charging.

2. Materials and Methods

2.1. Data Collection

We conducted a survey of RWTH employees and students on the topic of bicycle parking. The RWTH is spread over the western part of Aachen. The city lies in a valley basin near to the Belgian and Dutch borders. The mountains of the High Fens, the Eifel, and the Ardennes are located near to the city. Many commutes to the RWTH are relatively hilly, making e-bikes an attractive mode of transportation.

We invited students and employees via email to participate in our survey in July 2022, and we received data from 1583 students and 1377 employees. The latter group included professors, scientific employees (mostly PhD students), and non-academic administrative and technical staff. At the time of data collection, while most COVID-19 restrictions had been relaxed, the university still had a COVID-influenced homeworking-friendly policy.

For this analysis, we only considered people who commuted by e-bike at least once a week, and we excluded those with missing values. As a result, we had 281 respondents in our sample. Most use standard e-bikes (pedal-assisted up to 25 km/h, i.e., legally considered a bicycle), as shown in

Table 1. Sample overview.

	Students	Professors	Scientific Employees	Administrative and Technical Staff
Standard e-bikes	19%	7%	27%	43%
S-pedelecs	0%	0%	0%	1%
Cargo pedelecs	0%	0%	1%	2%

. A minor share of this group used s-pedelecs (pedal-assisted up to 45 km/h, i.e., legally classified as mopeds) or cargo pedelecs.

Although students are the largest group of university members, they were a minority in our sample of e-bike commuters. This result occurred because the proportion of e-bike users compared to conventional bicycle users differs substantially between the groups in our survey: 4% of students use e-bikes, whereas 12% of scientific employees, 31% of professors, and 48% of administrative and technical staff did so. As a result, the largest age group in our sample was between 50 and 59 years old, as shown in

Table 2. Age distribution in the sample.

<20	20–29	30–39	40–49	50–59	≥60
1%	24%	20%	20%	25%	11%

.

2.2. Commuting Behavior in the Sample

Before focusing on charging behavior, we looked at the commuting frequency in general, as well as by e-bike specifically, in our sample. As shown in Figure 1, 40% of the respondents commuted daily or almost daily by e-bike. However, 31% of our sample had a higher commuting frequency in general than by e-bike alone, indicating multimodality.

Figure 2 shows the self-reported distance between the home or usual place of residence and the university. The median distance was 6 km (mean: 8 km), and about 10% of our sample have commutes of more than 20 km. We assumed that workplace charging infrastructure was more relevant to those commuters, as they were more likely to run out of battery power during their commute. Overall, this distribution suggests that most commuters with home charging options did not rely on workplace charging infrastructure for regular commuting due to their e-bikes having sufficient battery capacity.

2.3. Charging Frequency in the Sample and Model Estimation

We asked participants to state how many days in a typical work or study week they would charge their e-bike at RWTH, depending on whether or not there was a charging fee. Each respondent received both questions in a randomized order. In a few cases, participants reported charging their e-bike slightly more often than using it to commute to work. In these cases, we limited the charging frequency to the maximum number of e-bike commutes.

We analyzed the charging frequency values using descriptive and multivariate statistics. To create a binomial mixed logit model, we converted the number of days per week with or without charging into choices, with each event corresponding to one day. Therefore, the number of choices per participant were equal to the number of days per week that each participant cycled to work. To calculate the number of charging events per day of the week, we also converted the participants’ responses into a daily probability of charging using the factors shown in

Table 3. Factors for translating participants’ responses into daily charging frequency values.

Answer of Participant	Charging Frequency
	Per Week	Per Weekday
(Almost) daily	5	1
Three to four times a week	4	0.8
Two to three times a week	3	0.6
Once a week	1	0.2

.

As a result, we created a dataset with 1091 observations each for free and paid charging. Due to the multiple observations made per participant, we chose a mixed logit model that included a random coefficient reflecting individual preferences, which used the following utility function:

$$U_{itq}={\beta }_q{X}_{itq}+{\eta }_{iq}+{\epsilon }_{iq}$$

(1)

The utility $\left(U_{itq}\right)$ of each alternative resulted from the multiplication of coefficients (${\beta }_q)$ and parameters ($X_{itq})$, depending on the alternative $\left(i\right)$, the situation $\left(t\right)$, and the individual $\left(q\right)$. There was also an error term and a random coefficient (${\eta }_{iq})$, which allowed a correlation between each individual’s choices, as they were not independent:

$${\eta }_{iq}={\mu }_{\left(\beta \right)}+{\sigma }_{\left(\beta \right)}\cdot {\mathsf{\xi }}_{iq}$$

(2)

The random coefficient consisted of a ${\mu }_{\left(\beta \right)}$ for the mean and a ${\sigma }_{\left(\beta \right)}$ for the standard deviation. The ${\xi }_{iq}$ was normally distributed symmetrically at around zero. We chose draws from the Modified Latin Hypercube Sampling type and used an n of 500.

Finally, the probability of each alternative $p\left(i,t\right)$ is given as follows:

$$p\left(i,t\right)=\frac{e^{U_{itq}}}{\sum _i{U}_{itq}}$$

(3)

For more information on mixed logit modeling, see [22,23].

We used the Apollo package in R to perform the calculations [24]. We performed bootstrapping to estimate standard errors and p-values and drew 100 samples from our dataset. For interpretation, we chose a p-value level of 0.1 as a cutoff because our dataset is comparatively small.

3. Results

3.1. Descriptive Analysis of the Charging Frequency

While there are no estimates of employee charging frequency in the literature, we found that nearly 18% of respondents would never charge their e-bike at the place of work or study, as shown in Figure 3. A likely reason for this decision is the need to carry charging equipment to work or leave a charger permanently at the destination if an interoperable charging station is unavailable. When paid charging is concerned, the percentage of respondents who would never charge at their place of work or study is more than three times higher than the percentage of those who would do so if charging was free. While half of the respondents would charge at least once a week if charging was free, only 16% would do so if they had to pay to charge. The proportion of respondents who would charge at least twice a week drops from 26 to 7%, if they had to pay for charging.

By converting the weekly charging frequency into a daily frequency, we can estimate the mean and median charging frequency per e-bike commuter per weekday, as shown in

Table 4. Daily charging frequency for free and paid charging.

	Free Charging	Paid Charging
Mean	0.27	0.08
Median	0.2	0
SD σ	0.32	0.2

. One charging event per day can be expected for four e-bike commuters when free charging is available. In the case of paid charging, there is one charging event per 12 e-bike users. However, the standard deviation for free charging is 0.32, while for paid charging, the value is not much lower, being 0.2. Therefore, the number of charging events could more notably fluctuate in the case of paid charging.

3.2. Multivariate Analysis of the Charging Frequency

To analyze the frequency of free and paid charging, we estimated a combined mixed logit model using the coefficients shown in

Table 5. Coefficients of the mixed logit model.

Reference: Scientific Employees		Free Charging					Paid Charging
		Est.	Std. err.	t-Ratio	p-Value		Est.	Std. err.	t-Ratio	p-Value
No charging	${\mu }_{\left(\beta \right)}$	Fixed to 0
	${\sigma }_{\left(\beta \right)}$	2.607	0.298	8.756	<2 × 10−10	***	(see left)
Charging	${\mu }_{\left(\beta \right)}$	−1.545	0.562	−2.748	0.006	**	−4.130	0.701	−5.895	<2 × 10−10	***
	${\sigma }_{\left(\beta \right)}$	−1.057	0.753	−1.405	0.160		−1.073	0.803	−1.337	0.181
Commuting distance2 [km2]	$\beta$	0.004	0.002	1.959	0.050	.	0.001	0.003	0.224	0.823
E bike resale value > 2000 €	$\beta$	−1.214	0.452	−2.684	0.007	**	−1.132	0.666	−1.701	0.089	.
Age ≥ 50 years	$\beta$	−1.025	0.482	−2.129	0.033	*	0.548	0.596	0.919	0.358
Daily commuting by e bike	$\beta$	0.340	0.459	0.741	0.459		0.980	0.587	1.668	0.095	.
Men	$\beta$	0.711	0.468	1.519	0.129		0.158	0.560	0.283	0.777
Students	$\beta$	0.197	0.670	0.294	0.769		−1.467	0.816	−1.797	0.072	.
Professors	$\beta$	0.424	0.591	0.718	0.473		−1.009	0.790	−1.277	0.202
Administrative and technical staff	$\beta$	0.916	0.949	0.965	0.334		1.224	1.195	1.024	0.306

Significance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1.

. The reference was a scientific employee with a female or non-binary gender. We fixed the ${\mu }_{\left(\beta \right)}$ for ‘no charging’ at zero and estimated a ${\sigma }_{\left(\beta \right)}$ that applies to both free and paid charging. All other estimates refer to free or paid charging. The table shows a substantially negative coefficient for free charging compared to the reference category of no charging. This result is consistent with the finding that even with free charging, e-bike commuters would generally charge infrequently. For paid charging, the coefficient is even more negative, implying that e-bike commuters would largely avoid charging at work in this scenario.

We expected the need for charging to increase disproportionately relative to commuting distance. Therefore, we used the square of the commuting distance as a variable. As the distance increases, the probability of using free charging grows (p-value of 0.05). For paid charging, the increase is much smaller and statistically insignificant. We also estimated a model where the commuting distance was not squared and obtained a similar result. Using stepwise linearization for the influence of the commuting distance only resulted in insignificant coefficients.

Inconsistently, an age above 50 years significantly reduces the frequency of free charging and insignificantly increases the frequency of paid charging. As expected, a resale value of the e-bike above 2000 EUR reduces the charging frequency for both free and paid charging, probably due to less charging being required due to higher battery capacities. The respective influence of the variables ‘daily commuting by e-bike’, ‘men’, ‘students’, and the employment group on the charging frequency is statistically insignificant or very weak. However, daily commuting by e-bike increases, and student status decreases, the frequency of paid charging. The second result is probably related to the incomes of the respondents.

4. Discussion

This article analyzes the potential charging frequency of e-bike commuters who work or study at RWTH Aachen University. Calculations based on a mobility survey conducted at the university estimate that about 1000 people primarily commute to RWTH by e-bike. Since the members of the university received a single email invitation to participate in this survey, we consider the response rate to be acceptable.

The university, as an example, is comparable to other employers because our model considers ‘students’ as a separate and statistically insignificant factor for free charging. Although we only analyzed members of one university, the age distribution and the proportion of students in our sample are not markedly different from those of other studies that analyzed e-bike use in general, except for the lack of retirees [1,8]. Therefore, we believe that the results are generalizable to companies and other locations where e-bike charging is relevant, as long as local conditions (e-bike types and battery capacities, availability of home charging, e-bike mode share, distribution of trip distances, and parking duration) are similar.

In general, the results show that most commuting e-bike users do not need daily charging opportunities at the university. While statistically, one charging event per day occurs for every fourth e-bike commuter in the free charging scenario, the frequency of use is only one charging event for every twelfth e-bike commuter in the paid charging scenario. However, the standard deviations for both scenarios are high. Furthermore, it is unlikely that cyclists would choose the days on which to charge in a way that maximizes the use of the charging infrastructure. Instead, they would want to charge whenever they needed to do so. Therefore, when providing charging infrastructure for employees, it is better to install a few additional power outlets than those calculated as necessary. In addition, it is necessary to know the (future) proportion of employees who commute by e-bike and their destination at the work site in order to locate charging facilities where they are useful. It is conceivable that the proportion of people working from home will continue to decrease as commuting patterns become more normalized once the COVID-19 pandemic is a thing of the past. This trend will increase the frequency of commuting and, therefore, the frequency of charging.

Our model also shows that charging frequency substantially decreases if users have to pay for charging. Since the cost to employers of providing charging facilities and electricity is relatively low, free charging could be a useful measure to encourage cycling [7]. For example, we might suppose that a bottleneck in the provision of charging infrastructure is unavoidable. In this situation, paid charging could be a solution to reduce the frequency of charging, as e-bike users without a strong need to charge would avoid charging. This approach would increase the practical availability of the charging infrastructure. It should be noted that the potential revenue from a charging fee may, in most cases, not economically justify the necessary billing process.

While [4] found that range anxiety exists among e-bikers and commuters charge at work when the commute distance is up to 20 miles each way, the influence of commuting distance on the charging frequency was only statistically significant for free charging in our model. This significant influence may be due to a generally greater need to charge as the number of kilometers cycled increases, rather than to avoid running out of battery power during a particular trip chain. Otherwise, we would have expected the influence of the distance on the frequency of use of paid charging to also be significant. The reason for this result could be the small proportion of e-bike users who travel commuting distances that result in trip chain lengths being close to the range of their e-bikes’ batteries. In other geographical areas, where the proportion of e-bike users with long commutes is higher, the influence of the commuting distance on the charging frequency may be stronger.

For the e-bikes’ resale values, the obtained results are as we expected. For both free and paid charging, a resale value above 2000 EUR reduces charging frequency. We assume that this result is due to the higher battery capacity of higher-value e-bikes. With future technological innovations and, thus, greater battery capacities, charging frequency will decrease even further.

As previous research has shown that medical conditions are an important reason for purchasing an e-bike [9], we would have expected older respondents to use the motor support more intensively and, therefore, need to charge more often. In our results, the frequency of free charging decreased for those aged over 50, while the influence on paid charging was statistically insignificant and in the opposite direction; thus, the effect of age is unclear. One possible explanation is that our sample mainly consists of working people, who are likely to be fit enough not to need excessive motor support and, therefore, do not need to charge more often than others. If the sample had included retirees, the results might have been different. However, this group is mainly out of scope when considering charging performed at work or study locations.

An analysis of current charging behavior would provide additional insight into the habits and preferences of e-bike users. Unfortunately, we did not have any information about the current charging behavior of the participants because the RWTH does not officially provide charging infrastructure for its employees and students. This problem leads to ambiguity about the current availability of charging infrastructure. Providing charging infrastructure would raise a number of administrative and tax issues, as the electricity used to charge employees’ e-bikes is considered by law to be a taxable benefit in kind. We assume that riders who currently need to charge their e-bike will bring it, or just the battery if it is removable, to the office and charge it there without official permission and in violation of fire safety regulations. We also expect this behavior to continue in the future if the charging infrastructure is inadequate or too expensive. This gray area makes it difficult to analyze current charging habits. Nevertheless, research should validate our results by evaluating current or future charging behavior.

Our results are based on surveys completed by RWTH employees and students who own an e-bike and use it to commute. Potential new e-bike commuters with trip chains that are too long to only charge at home are unlikely to be represented in our sample. However, for such long distances, the travel times for cars or public transportation are much shorter than for e-bikes. Consequently, e-bikes are unlikely to be competitive for making such trip chains and will only be used in exceptional cases. As a result, the range of e-bikes should be sufficient for most commuting distances. Although s-pedelecs made up a negligible proportion of our sample, their future proliferation could increase the distances that employees commute by e-bike and, thus, the need for charging. In addition, the provision of charging facilities at work could also support commuters with no or inconvenient charging facilities at home, e.g., if there is no power supply available at the e-bike parking location. Overall, their provision is a kind gesture by employers to show their support for employees who commute by cycling.

5. Conclusions

As e-bike sales and usage increase, the issue of e-bike charging becomes more relevant. From an employer’s perspective, supporting e-bike use by providing charging infrastructure could encourage e-bike commuting, thereby reducing the need for car parking space and improving the health of employees. Although it is not necessary to provide a charging point for every e-bike commuter, the number of charging points should exceed the number of e-bike users willing to charge, as it should not be necessary for someone to disconnect their e-bike during the working day in order for another e-bike to be charged.

Overall, the following conclusions can be drawn from our results:

• The frequency of charging at a place of work or study depends on the commuting distance for free charging. However, this influence was not statistically significant for paid charging. Therefore, the provision of charging infrastructure at the place of work or study is, probably, not necessary to avoid running out of battery on the way home for people who regularly charge their e-bike at home.
• More expensive e-bikes are less likely to be charged at work, probably due to their higher battery capacities.
• The influence of age is unclear, probably because in our work-related sample, the oldest participants are still too fit to use motor support to a significantly greater extent than younger participants.
• The charging frequency is much lower for paid charging than for free charging. Paid charging can be implemented if it is not possible to provide sufficient charging infrastructure.
• The willingness of e-bike users to pay for charging depends on their status as students or employees, and probably on their income. Therefore, it is useful to consider potential users before deciding to implement a charging fee.

Under German law, as well as legislation in other countries, even free charging of e-bikes is theoretically taxable. However, this law is likely to be ignored in practice by both employers and employees; thus, legislative reform is needed to avoid excessive documentation efforts and litigation over small amounts of money, as well as to facilitate the provision of charging infrastructure by employers.

Beyond this study, it would be interesting to analyze real-world charging data from employees, as well as different pricing schemes, price levels, and payment methods, to improve our understanding of e-bike charging. In addition, a future study should include the ability of employees to charge at home and in other locations.