Skip to main content

Energy efficiency trade-offs in small to large electric vehicles

A Correction to this article was published on 29 May 2020

This article has been updated



As electric kick scooters, three-wheelers, and passenger cars enter the streets, efficiency trade-offs across vehicle types gain practical relevance for consumers and policy makers. Here, we compile a comprehensive dataset of 428 electric vehicles, including seven vehicle types and information on certified and real-world energy consumption. Regression analysis is applied to quantify trade-offs between energy consumption and other vehicle attributes.


Certified and real-world energy consumption of electric vehicles increase by 60% and 40%, respectively, with each doubling of vehicle mass, but only by 5% with each doubling of rated motor power. These findings hold roughly also for passenger cars whose energy consumption tends to increase 0.6 ± 0.1 kWh/100 km with each 100 kg of vehicle mass. Battery capacity and vehicle mass are closely related. A 10 kWh increase in battery capacity increases the mass of electric cars by 15 kg, their drive range by 40–50 km, and their energy consumption by 0.7–1.0 kWh/100 km. Mass-produced state-of-the-art electric passenger cars are 2.1 ± 0.8 kWh/100 km more efficient than first-generation vehicles, produced at small scale.


Efficiency trade-offs in electric vehicles differ from those in conventional cars—the latter showing a strong dependency of fuel consumption on rated engine power. Mass-related efficiency trade-offs in electric vehicles are large and could be tapped by stimulating mode shift from passenger cars to light electric road vehicles. Electric passenger cars still offer potentials for further efficiency improvements. These could be exploited through a dedicated energy label with battery capacity as utility parameter.


Policy makers support the electrification of road transport for several reasons—to decrease urban air and noise pollution, to mitigate transport-related CO2 emissions, and to secure energy supply for the mobility of citizens [1, 2]). In Europe, most attention is paid to electric passenger cars, for which subsidies and other incentives opened a growing market [3, 4]. The policy focus is justified because passenger cars account for the bulk of road transport [5, 6] and its health and environmental impacts [7, 8]. But there is also a risk of overlooking the larger potentials of electromobility for transforming road transport in general.

Not only do electric powertrains allow operating vehicles without direct CO2 and air pollutant emissions, but also they relax important design constraints of conventional vehicles that need to accommodate a voluminous cylinder block, a crankshaft, and a transmission. Electric powertrains can generate high torque and power by multiple small motors placed in versatile configurations on one or multiple axles or directly in the wheel hub [9,10,11,12,13,14]. Traction batteries still offer a 50–100 times lower energy density than gasoline [15] and require more space than comparable fuel tanks. However, they allow for flexible integration into the rolling chassis and their size may decrease once occasional charging—at home, work, or in the public space—becomes feasible.

Without the design constrains of internal combustion engines, more diverse road vehicles can emerge. Electric kick scooters, tandem scooters [16], and e-bikes [17] represent just examples of new mobility solutions (see Table 3 in the Appendix) for urban areas where trip distances are short and required drive ranges low. This way, electromobility facilitates niche applications, situational mode choice, and frequent shifts between transport modes.

As diverse electric vehicles enter the market, trade-offs between energy consumption and other vehicle attributes become relevant. Efficiency trade-offs have been studied for conventional passenger cars [18,19,20,21,22,23] but not yet for a wider range of electric vehicles.

Here, we analyze such trade-offs for a comprehensive set of vehicles, ranging from light electric skate- and hover-boards, e-bikes and kick scooters to passenger cars and heavy-duty trucks. We hypothesize that:

  • the energy efficiency of electric vehicles is strongly related to vehicle mass and power;

  • state-of-the-art and mass-produced electric passenger cars are more efficient than first-generation cars produced at small scale;

  • despite high overall efficiencies, there is scope for further efficiency improvements, specifically for electric cars, which are covered by the CO2 labeling scheme for passenger cars in the European Union [24] but not yet by a dedicated energy label.

The results can provide policy makers with rationale for shaping the electrification of road transport, thereby contributing to a climate neutral European economy [25].


General aspects

This article covers seven major types of electric vehicles: (i) hover-boards and skateboards (vehicle Type 0 in this analysis), (ii) stand-up and kick scooters (Type 1), (iii) e-bikes, including bicycle-like three-wheelers (Type 2), (iv) larger two- and three-wheelers such as mopeds and step-through scooters classified as L1e vehicles, motorcycles classified as L3e and L4e vehicles, and three-wheelers classified as L2e and L5e vehicles (Type 3; [5]), (v) light four-wheelers classified as L6c and L7e vehicles (Type 4; [26]), (vi) passenger cars classified as M1 vehicles (Type 5; [27]), and (vii) light commercial and heavy-duty vehicles classified as N1-3 vehicles (Type 6; [11]). An indicative taxonomy of electric vehicles is provided in Table 3 in Appendix.

We consider certified and real-world energy consumption as both parameters can deviate from each other depending on the operating conditions of vehicles. Certified energy consumption is understood here as the consumption value declared by a manufacturer or certification body. For larger vehicles, such as passenger cars, the energy consumption is certified in a standardized regulatory test procedure [28]. For light vehicles such as e-bikes or kick scooters, the energy consumption is not yet certified with a standardized test but declared by manufacturers according to their own test protocol. Real-world energy consumption refers to the energy consumption observed by vehicle users on the road. For conventional passenger cars in the European Union, real-world CO2 emissions (g/km)—being proportional to energy consumption—systematically exceeded certified values (on average 40% in 2015; [22]). Similar deviations were already observed for electric cars, whose energy consumption (kWh/100 km) tends to exceed certified values, e.g., at low ambient temperature [29, 30].

Data collection and processing

Electric passenger cars

Based on Wikipedia [31], we identify which models of electric passenger cars are or have been available on the market. We identify for each model in an extended web search the following attributes: brand, model version, year of market introduction, country of production and certification, vehicle mass (kg), power (kW), battery capacity (kWh), motor and battery type, and certified energy consumption (kWh/100 km). We complement the data with information on the real-world energy consumption (kWh/100 km) and the country of vehicle operation. The bulk of data are obtained from Spritmonitor [32] reflecting vehicle use in Germany. However, from miscellaneous web sources, we also collect information about the real-world energy consumption of electric cars in China, Norway and the USA. The resulting dataset comprises 218 electric car models (Additional file 1: Table S1).

Other electric vehicles

To identify vehicles and their technical attributes, we scanned market portals, technical magazines, product tests, and the web pages of vehicle manufacturers with a focus on the German market (e.g., [33,34,35]). This search was subsequently extended to cover also vehicles sold and operated elsewhere by combining search terms such as “hover-board”, “skateboard”, “kick scooter”, “e-bike”, “three-wheeler” “electric motorcycle”, “moped”, “step-through scooter”, “motorcycle”, “light-weight vehicle”, “transporters”, “heavy-duty vehicle” with “electric”, “energy consumption” and “real-world energy consumption”. This approach adds 210 vehicle models to our dataset but it does not allow us to cover all models offered globally. Yet, we think our overall dataset of 428 models, which includes 218 passenger cars and 210 models of other vehicle types, captures the broad range of electric vehicles and can be considered a representative sample of electric vehicle types available to date (see also discussion in “Strengths and limitations” section).

Data processing

Before proceeding with the data analysis, we calculate for all vehicles certified and real-world drive range (km) by dividing battery capacity (kWh) by the respective energy consumption (kWh/100 km). We consider vehicles in running order by including the mass of a 70-kg driver. This correction acknowledges that the driver can exceed the mass of light electric vehicles and, thus, have an impact on their energy consumption. The vehicle mass presented in the “Results” section, thus, includes the vehicle and the driver, unless otherwise stated.

Data analysis

We characterize the vehicle data by calculating mean, standard deviation, minimum and maximum value of attributes separately for each vehicle type (Table 1). We then apply regression analyses to quantify efficiency trade-offs across all electric vehicles. We begin by modeling energy consumption Ei of vehicle i as a function of its mass Mi and rated motor power Pi in separate bivariate linear relationships as (Models 1, 2):

$$E_{i} = \alpha_{1} + \beta_{1} M_{i} + \varepsilon_{i} ,$$
$$E_{i} = \alpha_{2} + \beta_{2} P_{i} + \varepsilon_{i} ,$$

where β1,2 represents the regression coefficients, α1,2 stands for the constants, and ɛi denotes the unexplained residuals. The two regression models are applied separately to certified and real-world energy consumption.

Table 1 Descriptive statistics of vehicle attributes; mass includes the vehicle and a generic driver of 70 kg; SD standard deviation; indirect CO2 emissions from electricity consumption during vehicle use is calculated based on a carbon intensity of 300 ± 230 g CO2/kWh (mean of the European Union ± standard deviation of country data; [36])

Models 1 and 2 neglect relevant attributes such as powertrain efficiency or vehicle type. To account for these, we follow Mellios et al. [21] and Tietge et al. [22] and apply a multiple linear regression model. In this model, we approximate powertrain efficiency by the variable year Yi in which a vehicle was introduced to the market and we include vehicle type Ti as categorical variable. The energy consumption of electric vehicles is then modeled as (Model 3):

$$E_{i} = \alpha_{3} + \beta_{3} M_{i} + \beta_{4} P_{i} + \beta_{5} Y_{i} + \beta_{6} T_{i} + \varepsilon_{i} .$$

This model is also applied separately to certified and real-world energy consumption. Models 13 assume a linear relationship between dependent and independent variables, which may not hold across all vehicle categories. To address this shortcoming, we follow Knittel [19] and also model energy consumption as a power-law function of vehicle attributes, which equates to a linear relationship between the logarithms of dependent and explanatory variables. The model specifications are as follows (Models 4, 5 and 6):

$$\log E_{i} = \alpha_{4} + \beta_{7} \log M_{i} + \varepsilon_{i} ,$$
$$\log E_{i} = \alpha_{5} + \beta_{8} \log P_{i} + \varepsilon_{i} ,$$
$$\log E_{i} = \alpha_{6} + \beta_{9} \log M_{i} + \beta_{10} \log P_{i} + \beta_{11} Y_{i} + \beta_{12} T_{i} + \varepsilon_{i} ,$$

where log depicts the common logarithm to the base 10. Sample sizes for vehicle types differ from each other; thus, sampling biases are likely to affect our results. To control for sample size, we apply in a sensitivity analysis Models 16 to the median attribute values for the seven vehicle types.

Accounting for the needs of policy makers, we then zoom in on passenger cars and apply Models 3 and 6 to this vehicle type only. As battery size contributes to the mass of electric cars, trade-offs between battery capacity, drive range, and energy consumption can become relevant. To analyze effects, we expand the regression models for passenger cars by: (1) considering energy consumption as a function battery capacity and drive range, respectively, and (2) exploring the inter-relation between vehicle mass, battery capacity, and drive range.

A preliminary screening of residual plots reveals heteroscedasticity, which can bias the regression errors. To minimize this effect, we follow Tietge et al. [22] and estimate heteroscedasticity-robust standard errors for all regression coefficients with the R package ‘estimatr’ [37]. All regression analyses are conducted with R [38]).


Descriptive statistics of vehicle attributes

Vehicle attributes span over a wide value range (Table 1; Fig. 1). Vehicle mass ranges from 6.3 kg (76.3 kg including the driver) for skateboards to 27070 kg for trucks. Rated motor power ranges from 0.25 kW for e-bikes to 575 kW for passenger cars; battery capacity ranges from 0.10 kWh for skateboards to 300 kWh for trucks.

Fig. 1
figure 1

Types of electric vehicles (y-axis) and their attributes mass (a), power (b), battery capacity (c), certified energy consumption (d), and real-world energy consumption (e); dots represent individual data points, box and whiskers depict the median, the upper and lower quartiles, and 1.5 times the interquartile range of the data; dots are semitransparent to visualize overlaying data points; grouping of vehicle types as in “General aspects” section and Table 1

Certified and real-world energy consumption vary between 0.17 and 2.25 kWh/100 km for e-bikes and 13 kWh/100 km to 153 kWh/100 km for light commercial vehicles and trucks (Table 1, Fig. 1). Light-commercial and heavy-duty vehicles appear to consume less energy on the road than during certification. But this observation is likely the result of the small data sample on real-world energy consumption, which only includes two observations.

Passenger cars account for 218 out of 428 data points; their power, mass, and battery capacity span over a factor five to one hundred between the smallest and largest value (630–3370 kg; 5–575 kW; 8–100 kWh). Certified and real-world energy consumption only span over a factor of six and three (5–29 kWh/100 km and 10–33 kWh/100 km, respectively; Table 1). This observation suggests that vehicle mass, which spans a similar range, could indeed be the most relevant factor for the energy consumption of electric cars.

E-bikes and electric kick scooters save more than 90% (around 17 kWh/100 km per vehicle) of the energy consumed by electric passenger cars; light four-wheelers still save around half (9 kWh/100 km per vehicle) of the energy consumed by cars. At a carbon intensity of 300 ± 230 g CO2/kWh (mean of the European Union ± standard deviation of country data; [36]), these savings decrease the indirect carbon emissions from electricity by 50 ± 39 g CO2/km and 39 ± 21 g CO2/km, respectively, per vehicle.

Regression analysis—all vehicle types

The bivariate regression Models 1 and 2 suggest mass and power are significant, each explaining 74% and 15% of the certified energy consumption and 81% and 40% of the real-world energy consumption of electric vehicles, respectively (5% significance level; Table 2; Fig. 2). Applying a log-transformation to variables increases the coefficients of determination, suggesting changes in mass and power can explain individually 89% and 84% of changes in certified energy consumption and 92% and 90%, respectively, of changes in real-world energy consumption (Models 4 and 5 in Table 2). Consistently, the bivariate models reveal a stronger relation of mass and power with real-world energy consumption than with certified energy consumption. Overall, power appears to be less relevant than mass for the energy consumption of electric vehicles.

Table 2 Summary statistics—regression analysis of certified and real-word energy consumption as function of vehicle attributes; significance at 1% level (***), 5% level (**), and 10% level (*); light commercial and heavy-duty vehicles are excluded from the regression analysis of real-world fuel consumption in Models 3 and 6 as sample size is insufficient; vehicle type 0: hover- and skateboards, type 1: stand-up and kick scooters, type 2: e-bikes, type 3: larger two- and three-wheelers, type 4: light four-wheelers, type 5: passenger cars, type 6: light commercial and heavy-duty vehicles
Fig. 2
figure 2

Certified and real-world energy consumption of electric vehicles as a function of mass and power; shaded areas represent the 95% confidence interval of the fitted regression lines; dots are semitransparent to visualize overlaying data points

The explanatory power of the regression models increases, if we include year of market introduction and vehicle type as explanatory variables. Together, the four independent variables explain 93% and 84% of the certified and real-world energy consumption of electric vehicles, respectively (Model 3). Log-transforming variables increase the coefficients of determination to roughly 97% (Model 6; Table 2).

Considering all vehicle types, the multiple regression analysis indicates that:

  • each 100 kg increase in vehicle mass increases certified and real-world energy consumption by 0.4 kWh/100 km and 0.6 kWh/100, respectively (linear Model 3); each doubling of mass increases certified and real-world energy consumption by 58% and 42%, respectively (power-law Model 6);

  • rated vehicle power is weakly correlated with energy consumption; coefficients are not significant at a 5% level in Models 3 and 6 for certified energy consumption; Model 6 suggests each doubling of rated vehicle power increases the real-world energy consumption by 6%;

  • despite overall high efficiencies, electric vehicles still become 0.8% and 1.0% more efficient each year during certification and real-world use, respectively (Model 6); under real-world conditions, their energy consumption has been decreasing by 0.1 kWh/100 km each year (linear Model 3; Table 2);

  • vehicle types differ significantly in their energy consumption even after correcting for mass and power (Models 3 and 6 in Table 2); larger vehicles consume more energy than smaller ones; certified energy consumption of light vehicles (Types 0–3) increases more steeply with vehicle mass than the energy consumption of larger three–four-wheelers and passenger cars (Fig. 3).

    Fig. 3
    figure 3

    Certified energy consumption as a function of vehicle mass; shaded areas represent the 95% confidence interval of the fitted regression lines; dots are semitransparent to visualize overlaying data points

The effect of vehicle attributes on energy consumption depends on whether certification or real-world operating conditions are considered and whether a linear or power-law relationship between variables is assumed. Vehicle mass has a larger effect on certified than on real-world energy consumption in the linear Model 3, while it is the other way around in the power-law Model 6. This observation suggests our findings are subject to uncertainty. The sensitivity analysis based on the median attribute values for each vehicle type confirms this expectation (Additional file 2: Table S3). Models 1 and 2 now yield high coefficients of determination. Given the small sample size of only 7 data points, the coefficients of the multiple regression Models 3 and 6 are not significant at a 5% level. Moreover, the sensitivity analysis (Additional file 2: Table S3) suggests:

  • each 100 kg increase in vehicle mass increases certified and real-world energy consumption by 1.0 kWh/100 km and 1.3 kWh/100 km, respectively (Model 1; Additional file 2: Table S3), which is more than the 0.7 kWh/100 km and 0.9 kWh/100 km determined by Model 1 in Table 2; a doubling of vehicle mass roughly doubles energy consumption (108% and 96% increase in certified and real-world energy consumption, respectively (Model 4), which is consistent with the 102% and 99% previously observed by Model 4;

  • power is un-correlated with energy consumption at the 5% significance level in Model 2 but there is a significant relationship in the power-law Model 5, suggesting each doubling of vehicle power leads to a 61% and 54% increase in certified and real-world energy consumption of electric vehicles;

  • electric vehicles have not become significantly more efficient in recent years of their manufacturing.

Taken together, our regression analyses reveal that vehicle mass is the most important parameter for the energy consumption of electric vehicles. Each 100 kg vehicle mass increases the energy consumption by some 0.4–1.3 kWh/100 km, which is equivalent to 1.2 ± 0.9–3.8 ± 3.0 g CO2/km of indirect emissions from electricity at a carbon intensity of 300 ± 230 g CO2/kWh (mean of the European Union ± standard deviation of country data; [36]).

Regression analysis—passenger cars

Passenger cars follow the same pattern as electric vehicles in general (Fig. 4). Table 4 in Appendix shows that:

Fig. 4
figure 4

Certified and real-world energy consumption of electric passenger cars as a function of mass, rated motor power, and year of market introduction; shaded areas represent the 95% confidence interval of the fitted regression lines; dots are semitransparent to visualize overlaying data points

  • each 100 kg increase in vehicle mass increases both certified and real-world energy consumption by 0.6 kWh/100 km (linear Model 3); a doubling of vehicle mass leads to a 60% and 46% increase in the certified and real-world energy consumption of electric cars, respectively (power-law Model 6);

  • power does generally not affect energy consumption at a 5% level of significance; an exception is real-world energy consumption that tends to increase 6% with each doubling of vehicle power according to Model 6;

  • over the years, electric passenger cars have not become more efficient during certification; however, their real-world energy consumption appears to decrease on average by 0.1 kWh/100 km (linear Model 3) or 1% (power-law Model 6) each manufacturing year.

Modern mass-produced and purpose-designed electric cars consume under real-world conditions on the road 2.1 ± 0.8 kWh/100 km less energy than the first-generation electric cars produced at small scale based on chassis component of conventional cars (Fig. 5). This observation and the large spread of consumption values for cars of similar mass suggests scope for future efficiency improvements. As batteries are relatively heavy, such improvements could be attained by simply decreasing battery size and the drive range of vehicles (Fig. 6, Additional file 2: Table S2). We find that each 10 kWh of battery capacity increases vehicle mass by 15 kg, drive range by 40–50 km, and energy consumption by 0.7–1.0 kWh/100 km (Fig. 6, Additional file 2: Table S2).

Fig. 5
figure 5

Certified and real-world energy consumption of electric passenger cars as a function of vehicle mass; differences between the three groups of electric cars are significant at 5% level for real-world energy consumption only; shaded areas represent the 95% confidence interval of the fitted regression lines; dots are semitransparent to visualize overlaying data points

Fig. 6
figure 6

Regression analysis for passenger cars exploring the relationship between battery capacity and vehicle mass (a), drive range and battery capacity (b), and certified and real-world energy consumption and battery capacity (c, d); shaded areas represent the 95% confidence interval of the fitted regression lines; dots are semitransparent to visualize overlaying data points

A visual breakdown of certified and real-world energy consumption of passenger cars for several countries reveals that in China, real-world energy consumption scatters around certified values; whereas in Germany, Norway, and the USA, electric cars tend to consume more energy on the road than during certification (Fig. 7).

Fig. 7
figure 7

Real-world and certified energy consumption of electric passenger cars, disaggregated per country; shaded areas represent the 95% confidence interval of the fitted regression lines; dots are semitransparent to visualize overlaying data points


Strengths and limitations

We compile a comprehensive dataset of vehicle attributes for the major types of electric road vehicles (Additional file 2: Table S1). Our analysis reveals: (i) large mass-related efficiency trade-offs that could be tapped by mode shift from passenger cars to light electric vehicles and (ii) scope for improving the already high energy efficiency of electric passenger cars, which could be exploited through dedicated energy labeling. The results provide climate, energy, and transport policy with rationale for shaping the transition towards a sustainable transport system. Our findings are robust, albeit subject to several limitations.

First, vehicle types are not equally represented in our dataset. Passenger cars alone account for half of all data points (Table 1). This sampling bias introduces an error, which is not negligible as the sensitivity analysis suggests. We, therefore, regard the actual relationships of parameters to be in range of—but not necessarily identical with—the coefficients of our regression models (Table 2 and Additional file 2: Table S3).

Second, real-world energy consumption reflects the actual operating conditions of vehicles but may neither captures average vehicle use nor the specific operating conditions of any individual vehicle user. Variability in real-world and declared energy consumption is particularly large for e-bikes, likely owing to modulating torque (typically ranging between 40 and 100 Nm [39]) and thus power requirements.

Third, assuming a generic driver of 70 kg introduces a random error into our results, which is large for light-weight vehicles such as e-bikes and kick scooters for which the driver’s mass exceeds the mass of the vehicle.

Fourth, regression models are only robust if residuals are randomly distributed along the value range of independent variables and if explanatory variables are unrelated with each other. Both requirements are only partially met in our analysis. The diagnostic plots in Additional file 2: Figures S1–S16 suggest residuals are heteroscedastic. They appear to be clustered and not normally distributed, which is likely the result of large variability in the energy consumption of e-bikes. For passenger cars, residuals appear to scatter randomly along the range of fitted values. We address heteroscedasticity, as far as feasible, by estimating heteroscedasticity-robust standard errors for all regression coefficients [37]. Collinearity is tested for by estimating variance inflation factors (VIF) for Models 3 and 6. For Model 3, the VIF is smaller than five for certified energy consumption, which can be interpreted as unproblematic collinearity [40]. The same applies to Models 3 and 6 for passenger cars. But for real-world energy consumption (linear Model 3) and both certified and real-world energy consumption (power-law Model 6), VIFs reach up to 15, indicating strong collinearity. Therefore, we test the robustness of Models 3 and 6 in a stepwise regression. In this analysis, power and vehicle type are excluded from the regression models and regressed separately against the residuals of the adapted Models 3 and 6. For Model 3, the residuals are uncorrelated with vehicle power at a 5% significance level. For Model 6, however, residuals are correlated with power and vehicle type. Excluding both variables from the adapted Model 6 suggests a doubling of vehicle mass leads to a 108% and 104% increase in certified and real-world energy consumption, respectively. These values are higher than our original findings in Model 6 (Table 2) but they are consistent with the univariate power-law Model 4 and the results of the sensitivity analysis (see Model 4 in Additional file 2: Table S3). We, therefore, regard the value range of 60–110% and 40–100% as indicated by Models 4 and 6 to be indicative of the increase in certified and real-world fuel consumption across all types of electric vehicles with each doubling of vehicle mass.

Finally, the coefficients of determination for Models 4–6 (Table 2) suggest energy consumption is best modeled as a power-law rather than a linear function of mass, specifically when value ranges are large and several vehicle types are covered in the analysis. If vehicle attributes vary little, then the linear model provides a reasonable approximation of parameter relationships as a comparison of Models 3 and 6 for passenger cars suggests (Table 4).

Comparison of findings

Our findings are broadly in line with previous studies and add insight into the trade-offs between the various attributes of electric vehicles. The observation that the energy consumption of electric passenger cars increases by 0.6 kWh/100 km with each 100 kg vehicle mass is broadly consistent with the 0.4 kWh/100 km modeled by Redelbach et al. [41]. Likewise, the observed 46–60% increase in energy consumption with each doubling of vehicle mass is consistent the 34–42% increase with each doubling of vehicle mass observed by Carlson et al. [42].

The observation that the energy consumption of electric passenger cars is closely related to vehicle mass but less so to rated motor power contrasts findings for conventional passenger cars, for which power is an important, if not the most important, driver of fuel consumption: Each doubling of rated engine power tends to increase fuel consumption by around 30–50% [19, 21]. A 10 kW increase in rated engine power raises the fuel consumption by 0.3 l/100 km in compact gasoline cars and 0.2 l/100 km in compact diesel cars [23].

Electric cars only become marginally more efficient with time, mainly under real-world driving conditions on the road. By contrast, conventional passenger cars have been showing considerable efficiency improvements in the past, both during certification and real-world driving [19, 23]. The differences between electric and conventional cars could be attributed to at least three factors:

  • Electric motors have a higher tank-to-wheel efficiency (73–90%) than internal combustion engines (16–37%) over all relevant engine loads and speeds [12, 13].

  • Electric vehicles recuperate their kinetic energy through regenerative breaking, whereas conventional vehicles do not. With energy recuperation, power and related accelerations become less relevant for energy consumption than rolling resistance and aerodynamic drag—the former being directly proportional to vehicle mass, the latter being proportional to front area and increasing with vehicle speed squared.

  • Electric motors do not generate idling losses, which are proportional to displacement and rated power in naturally aspired internal combustion engines. The energy consumption of an electric motor is, therefore, largely dependent on its instantaneous power output rather than its maximum rated power.

Together, these factors influence the energy consumption of passenger cars during certification and real-world driving. Under real-world conditions, the energy consumption varies, moreover, depending on the actual operating conditions. Part of the low dependency of energy consumption on rated motor power in electric cars could then result from drivers who do not exploit the full acceleration and speed potential of their vehicles [43] to preserve still limited drive range. Nevertheless, occasional high-load motor operations during high speeds and uphill driving specifically for larger and comparatively powerful cars may explain why real-world energy consumption in passenger cars is more closely related to rated motor power than is certified energy consumption.

Implications for policy makers and vehicle manufacturers

Our findings have four major implications. First, the strong relationship between energy consumption and vehicle mass suggests mode shift from electric passenger cars to light electric vehicles can decrease the energy consumption of road transport. The technical characteristics of electric powertrains facilitate such shifts as the market success of electric kick scooters, e-bikes, and light three-wheelers (e.g., [44]) suggests. In urban areas, light electric vehicles can increase travel speed, mitigate ambient air and noise pollution, and decrease the direct CO2 emissions of individual vehicles. Mode shift towards light electric vehicles would also decrease demand for road infrastructure, which in turn opens new opportunities to revitalize cities and make them resilient to climate change by mitigating heat-island effects through enlarged vegetation areas [45, 46]). Other merits of mode shift include lower investment needs for recharging infrastructure—light vehicles can be charged with standard wall outlets in buildings—and decreased material and energy requirements (per vehicle) for manufacturing.

But challenges remain. Mode shift away from passenger cars is not straightforward but requires a broader strategy to provide adequate infrastructure, strengthen inter-operability with other transport modes including public transport, and facilitate innovative mobility solutions around vehicle sharing and renting services. Users of light vehicles are vulnerable in case of accidents [47], which may require updated vehicle certification schemes, traffic rules, and safety regulation before light electric vehicles may realize their full market potential [48].

Second, large variability in the energy consumption of passenger cars with a similar mass (Figs. 3 and 4) suggests there is scope for further efficiency improvements, e.g., through purpose design, wheel-hub motors, improved energy recuperation, decreased coasting resistance, and the application of light-weight chassis components [49]. These potentials could be tapped by introducing minimum efficiency requirements or a dedicated energy label that classifies vehicles according to their energy efficiency. If done, our findings caution against the use of vehicle mass as a utility factor to normalize energy consumption—as it is implemented for the fleet average CO2 emissions target of conventional passenger cars [50]. Ideally, a label on efficiency would directly classify efficiency, that is, distance-specific energy consumption of electric vehicles (kWh/100 km; km/kWh). However, this way efficiency improvements could be achieved by curbing vehicle mass through a small battery, which in turn may impair vehicle utility. If policy makers prefer to use a utility factor to classify energy efficiency, battery capacity could be a suitable choice. Normalizing energy consumption by battery capacity would force both efficiency improvements of vehicles and increasing energy densities of batteries. Our study suggests the real-world energy consumption of electric passenger cars E (kWh/100 km) is related to battery capacity c (kWh) (coefficient of determination 0.34) as follows:

$$E = \left( {0.09 \pm 0.01} \right) c + \left( {14 \pm 1} \right).$$

Future research could expand the analysis (Additional file 1: Table S1) towards a concrete proposal for an efficiency label. Research could also investigate whether there is scope for implementing energy labels to other types of electric vehicles.

Third, our findings suggest the current certification procedure of electric passenger cars in Europe may underestimate the real-world energy consumption (Fig. 7). If future research confirms systematic artifacts in the certification test, policy makers should consider adapting test procedures to ensure consumers are accurately informed about the energy consumption of vehicles when deciding about a purchase.

Fourth, electromobility is still in its infancy and rapid technological learning will likely improve cost competitiveness [51,52,53] and technical features, including the efficiency of battery charging [54]. Already to date, however, expensive high-power vehicles offer marketing opportunities that manufacturers began to exploit. If rated motor power is less important for the energy consumption in electric cars than in conventional ones, electric cars with high motor power, torque, and unmatched acceleration capabilities have a competitive edge over their conventional counterparts. If so, a superior driving performance may turn the high price of electric cars [51, 53] into a status symbol (as it is the case of conventional sports cars), transforming a market barrier into a purchase argument for status seeking consumers.


We draw the following conclusions:

  • The energy consumption of electric vehicles is strongly related to vehicle mass but less so to rated power and the year a vehicle entered the market. Given the high overall efficiency of electric motors, electric vehicles become only marginally more efficient with time.

  • Assuming a linear relationship between vehicle attributes suggests each 100 kg vehicle mass increases energy consumption by some 0.4–1.3 kWh/100 km, which is equivalent to 1.2 ± 0.9–3.8 ± 3.0 g CO2/km of indirect carbon emissions at the current European electricity mix. Mode shift to e-bikes and kick scooters can save more than 90% (around 17 kWh/100 km per vehicle) of the ‘tank-to-wheel’ energy consumption of passenger cars; mode shift to light three- and four-wheelers can still save half of the energy consumed by electric passenger cars (9 kWh/100 km per vehicle). Climate and transport policy could support mode shift through dedicated road and recharging infrastructure, giving priority to small and light-weight electric two–four-wheelers and bicycles in urban environments.

  • The energy consumption of electric passenger cars (kWh/100 km) scatters over a wide range even for vehicles of similar mass (e.g., 10–15 kWh/100 km for cars of 2000 kg; Fig. 4). This finding suggests there is scope for efficiency improvements that could be exploited by a dedicated energy label. To this end, the prospects of using battery capacity as a utility factor should be explored.

  • The weak statistical relationship between energy consumption and rated motor power implies electric cars can be positioned on the market as high-power, status revealing vehicles. Such marketing strategy would turn the high price of electric passenger cars into a sales argument for status seeking consumers. On the other hand, large high-power electric vehicles usually come with a higher vehicle mass and thus resources consumption.

  • This article covers energy consumption in the use phase of electric vehicles only. The environmental costs and benefits of the electrification of road transport, however, require a holistic perspective and should be investigated by comprehensive life cycle assessments.

Availability of data and materials

The datasets supporting the research are included in additional files.

Change history

  • 29 May 2020

    An amendment to this paper has been published and can be accessed via the original article.















Standard deviation


  1. Briec E, Mazal C, Meyer G, Müller B (2012) European roadmap—electrification of road transport. 2nd edn, European Green Vehicles Initiative. Accessed 29 May 2019

  2. EC (2017) Electrification of the transport system—studies and reports. EC—European Commission. Accessed 29 May 2019

  3. IEA (2018) Global EV outlook 2018. IEA—International Energy Agency. Accessed 29 May 2019

  4. McKinsey (2014) Electric vehicles in Europe: gearing up for a new phase. Accessed 18 July 2019

  5. EC (2017) 2017—EU transport in figures. Statistical pocketbook. EC—European Commission, Eurostat. Accessed 26 June 2019

  6. Löhr E, Kirsch F, Lones L (2016) Exploration of EU road vehicle fuel consumption and disaggregation. Ricardo Energy & Environment. Accessed 27 June 2019

  7. EEA (2018) Emissions of air pollutants from transport. European Environmental Agency. Accessed 26 June 2019

  8. Thomas S (2018) Drivers of recent energy consumption trends across sectors in EU28. Samuel Thomas Consulting. Accessed 29 May 2019

  9. Chau KT (2014) 21—Pure electric vehicles. In: Alternative fuels and advanced vehicle technologies for improved environmental performance. Towards Zero Carbon Transportation, pp 655–684

  10. Ehsani M, Gao Y, Gay SE, Emadi A (2005) Modern electric, hybrid electric, and fuel cell vehicles—fundamentals, theory, and design. CRC Press LLC.,%20Hybrid%20Electric%20&%20Fuel%20Cell%20Vehicles%20-%20Mehrdad%20Ehsani.pdf. Accessed 2 July 2019

  11. EM (2019) Electric power train. EM—Ecomove. Accessed 18 July 2019

  12. Gustafsson T, Johansson A (2015) Comparison between battery electric vehicles and internal combustion engine vehicles fueled by electrofuels. Master’s Thesis, Chalmers University of Technology, Gothenburg, Sweden

  13. Helmers E, Marx P (2012) Electric cars: technical characteristics and environmental impacts. Environ Sci Eur 24(1):14

    Article  Google Scholar 

  14. R3E (2019) Put it all in the wheel. R3E. Accessed 19 July 2019

  15. Ding Y, Cano ZP, Yu A, Yu A, Lu J, Chen Z (2019) Automotive Li-ion batteries: current status and future perspectives. Electrochem Energy Rev 2(1):1–28

    Article  CAS  Google Scholar 

  16. Toll M (2019) This dangerous looking two-person electric scooter wants to challenge Bird. Electrek. Accessed 3 July 2019

  17. Weißenborn S (2019) Fahrrad ersetzt Kleinwagen. Spiegel Online. Accessed 5 Sept 2019

  18. Bandivadekar AP (2008) Evaluating the impact of advanced vehicle and fuel technologies in U.S. light-duty vehicle fleet. PhD thesis, Massachusetts Institute of Technology, Cambridge, USA

  19. Knittel C (2012) Automobiles on steroids. Product attribute trade-offs and technological progress in the automobile sector. Am Econ Rev 101:3368–3399

    Article  Google Scholar 

  20. Ligterink NE, Smokers R, Spreen J, Mock P, Tietge U (2016) Supporting analysis on real-world light-duty vehicle CO2 emissions. Report TNO 2016 R10419v3. TNO—Netherlands Organisation for Applied Scientific Research, Delft

    Google Scholar 

  21. Mellios G, Hausberger S, Keller M, Samaras C, Ntziachristos L (2011) Parameterisation of fuel consumption and CO2 emissions of passenger cars and light commercial vehicles for modeling purposes. JRC Report 66088, EUR 24927 EN. European Commission—Joint Research Centre. Ispra, Italy

  22. Tietge U, Mock P, Franco V, Zacharof N (2017) From laboratory to road: modeling the divergence between official and real-world fuel consumption and CO2 emission values in the German passenger car market for the years 2001–2014. Energy Policy 103:212–222

    Article  Google Scholar 

  23. Weiss M, Irrgang L, Kiefer AT, Roth JR, Helmers E (2020) Mass- and power-related efficiency trade-offs and the CO2 emissions of compact passenger cars. J Clean Prod 243:118326

    Article  CAS  Google Scholar 

  24. EC (1999) Directive 1999/94/EC. EC—European Commission. Off J L 12:1–10

    Google Scholar 

  25. EC (2019) The European green deal. COM(2019) 640 final. EC—European Commission. Accessed 12 Jan 2020

  26. EC (2013) Regulation 168/2013 on the approval and market surveillance of two- or three-wheel vehicles and quadricycles. EC—European Commission. Off J Eur Union L60:1–117

    Google Scholar 

  27. EC (2007) Directive 2007/46/EC establishing a framework for the approval of motor vehicles and their trailers, and of systems, components and separate technical units intended for such vehicles. EC—European Commission. Off J Eur Union L263:1–437

    Google Scholar 

  28. EC (2017) Commission regulation (EU) 2017/1151. EC—European Commission. Off J Eur Union L175:1–839

    Google Scholar 

  29. Haakana A, Laurikko J, Granström R, Hagman R (2013) Assessing range and performance of electric vehicles in Nordic driving conditions—End of Project Report. Nordisk Energieforskning, December 2013, pp 74. Accessed 19 Mar 2019

  30. Schaufenster (2015) Questions about electric vehicles: how does data about energy consumption and range of electric vehicles come about? Parallel Impact Research of the Electric Mobility Showcase. Schaufenster Elektromobilität, results document No 6, May 2015, pp 8 (removed from the internet)

  31. Wikipedia 2019. Liste von Elektroautos in Serienproduktion bis 2019. Wikimedia Foundation. Accessed 2 July 2019

  32. Spritmonitor (2019) Verbrauchswerte real erfahren. Fisch und Fischl GMbH. Accessed 2 July 2019

  33. Autobild (2019) Elektro-Skateboards-Vergleich 2019. Axel Springer SE. Accessed 5 July 2019

  34. E-bike (2019) Die besten E-bike Marken in Deutschland. e-motions experts GmbH. Accessed 5 July 2019

  35. MM (2019) Der große E Scooter Test 2019. Medienhaus Müller GmbH &Co. KG. Accessed 5 July 2019

  36. EEA (2019) CO2 emission intensity. EEA—European Environmental Agency. Accessed 26 Dec 2019

  37. Blair G, Cooper J, Coppock A, Humphreys M, Sonnet L, Fultz N (2018) Package ‘estimatr’. R Foundation for Statistical Computing, Vienna, Austria. Accessed 23 Apr 2018

  38. R Core Team (2017) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Accessed 9 Apr 2017

  39. FR (2019) Die E-Bike Motoren in der Übersicht. FR— GmbH & Co. KG. Accessed 20 Nov 2019

  40. Frost J (2019) Multicollinearity in regression analysis: problems, detection, and solutions. Accessed 5 Nov 2019

  41. Redelbach M, Klötzke M, Friedrich HE (2012) Impact of lightweight design on energy consumption and cost effectiveness of alternative powertrain concepts. In: European electric vehicle congress, 19–22 November 2012, Brussels, Belgium

  42. Carlson R, Lohse-Busch H, Diez J, Gibbs J (2012) The measured impact of vehicle mass on road load forces and energy consumption for a BEV, HEV, and ICE vehicle. SAE International 2013-0-1457

  43. Helmbrecht M, Olaverri-Monreal C, Bengler K, Vilimek R, Keinath A (2014) How electric vehicles affect driving behavioral patterns. IEEE Intell Transp Syst Mag 6(3):22–32

    Article  Google Scholar 

  44. MCD (2019) Europe—Electric scooter and motorcycles market grew up 64% in 11 months 2019. MCD—Motor Cycles Data. Accessed 27 Jan 2020

  45. EPA (2019) Climate change and heat islands. EPA—United States Environmental Protection Agency. Accessed 10 Nov 2019

  46. Lapola DM, Braga DR, di Giulio GM, Torres RR, Vasconcellos MP (2019) Heat stress vulnerability and risk at the (super) local scale in six Brazilian capitals. Clim Change 154:477–492

    Article  Google Scholar 

  47. Hertach P, Uhr A, Niemann S, Cavegn M (2018) Characteristics of single-vehicle crashes with e-bikes in Switzerland. Accid Anal Prev 117:232–238

    Article  Google Scholar 

  48. E-mobil (2019) Elektrische Klein- und Leichtfahrzeuge. E-mobil BW GmbH - Landesagentur für neue Mobilitätslösungen und Automotive Baden-Württemberg, Stuttgart

  49. Zhao C, Gong G, Yu C, Liu Y, Zhong S, Song Y, Deng C, Zhou A, Ye H (2019) Research on key factors for range and energy consumption of electric vehicles. SAE Technical Paper 2019-01-0723,

  50. EC (2019) Regulation (EU) 2019/631. EC—European Commission. Off J Eur Union L111:13–53

    Google Scholar 

  51. Safari M (2018) Battery electric vehicles: looking behind to move forward. Energy Policy 115:54–65

    Article  Google Scholar 

  52. Weiss M, Dekker P, Moro A, Scholz H, Patel MK (2015) On the electrification of road transportation—a review of the environmental, economic, and social performance of electric two-wheelers. Transp Res Part D Transp Environ 41:348–366

    Article  Google Scholar 

  53. Weiss M, Zerfass A, Helmers E (2019) Fully electric and plug-in hybrid cars—an analysis of learning rates, user costs, and costs for mitigating CO2 and air pollutant emissions. J Clean Prod 212:1478–1489

    Article  CAS  Google Scholar 

  54. Channegowda J, Pathipati VK, Williamson SS (2015) Comprehensive review and comparison of DF fast charging converter topologies: improving electric vehicle plug-to-wheels efficiency. IEEE, pp 263–268. Accessed 31 Jan 2020

Download references


We thank Uwe Tietge for supporting the collection of vehicle data and Juliana Stropp and three anonymous reviewers for their constructive comments on previous versions of the manuscript.


Not applicable.

Author information

Authors and Affiliations



EH and MW conceived the original idea. KCC collected the data. KCC and MW analyzed the data. EH and MW wrote the manuscript. All authors read and approved the final manuscript.

Corresponding authors

Correspondence to Martin Weiss or Eckard Helmers.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Additional file 1: Table S1.

Data of vehicle attributes.

Additional file 2:

Additional Figures S1–S36, Tables S2, S3.



See Tables 3 and 4.

Table 3 Electric vehicle types covered in our analysis
Table 4 Summary statistics—multiple regression analysis of certified and real-word energy consumption of passenger cars as a function of vehicle attributes; significance at 1% level (***), 5% level (**), and 10% level (*)

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Weiss, M., Cloos, K.C. & Helmers, E. Energy efficiency trade-offs in small to large electric vehicles. Environ Sci Eur 32, 46 (2020).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: