Estimating the energy impact of electric, autonomous taxis: evidence from a select market

Expenditures considered when estimating consumer cost include vehicle financing, licensing, insurance, maintenance, cleaning, fuel and, for ATs specifically, safety oversight [16, 22]. Requisite safety oversight is assumed to decrease as AT technology advances. We also take account of operator-envisioned profit expectations and fluctuations in capacity utilization rates that reflect demand heterogeneity.

AT cost estimates also consider heterogeneity in vehicle operational lifespan and annual mileage. As the pro-rating of fixed costs over time impacts the financial proposition of ATs, both factors warrant attention. Mileage heterogeneity considers vehicle recharging requirements that may limit vehicle productivity and subsequently, profitability [23, 24]. Productivity may be further impeded when vehicle electrification is paired with vehicle automatization owing to increased vehicular weight, sensor load and aerodynamic drag, all of which limit vehicle range [25].

We also consider consumer travel time in terms of hourly wages and thus transform differences in travel time to money units [19, 26]. Literature suggests that productivity benefits are realized through the re-allocation of time to paid or leisure activities that replace the demands of driving on attention. Envisioned benefits include would-be drivers performing other valued activities [19].

The nascent nature of AT technology necessitates consideration of a hypothetical vehicle. Our AT is a mid-sized, electric auto with an internal volume between 110 and 120 ft³. The vehicle's energy profile is assumed to be consistent with an electric auto; one with a fuel economy and emissions footprint of 114 miles per gallon equivalent and 159.1 CO₂ grams per mile respectively [27]. Fuel economy estimates are within the range of estimates pairing vehicle electrification with vehicle automation [8, 28]. Emissions estimates account for vehicle production, extraction, processing, transportation, and fuel distribution [27].

The energy profile of CTs and PVs—both of which are also assumed to be mid-sized vehicles with an internal volume between 110 and 120 ft—are estimated using pre-existing data [27]. CTs are—consistent with the fleet profile in our target market—assumed to be hybrid-electric vehicles whereas PVs are powered by ICEs. Both vehicles have a fuel economy of 52 and 34 miles per gallon and emit 237.7 and 380.2 grams of CO₂ grams per mile, respectively. These figures reflect the lifetime mileage of CTs and PVs in our target market.

Our model also considers the emissions profile of public transit in our target market. Doing so accommodates instances where AT introduction may prompt mode switches from publicly funded transportation modes [29]. Our public transit energy profile is estimated—leveraging market specific demand characteristics—as a weighted average of heavy rail, commuter rail, light rail and transit buses [30, 31]. Consumption and emissions estimate per mile are based on existing sources and are occupancy adjusted [30]. These values are subsequently weighted based on their proportional contribution to overall miles traveled [31]. Proportions exclude passenger miles traveled via other motorized and nonmotorized transportation modes.

Uncertainty surrounding an AT's precise financial and energy profile warrant acknowledgement. The validity of our assumptions and the predictions estimated by our model are admittedly based on imperfect information, this given the nascent nature of the technology. For example, there is little publicly available data on what AT financing costs will be. We address concerns regarding the magnitude of input parameters (like financing), the relevance of their inclusion (particularly where licensing fees are concerned) and their subsequent impact on the cost proposition of ATs through elasticity testing and sensitivity analysis (detailed in section 3.1).

A similar approach is employed when estimating the energy impact of ATs. Impact estimates consider specific vehicle energy profiles and where electrified powertrains are concerned, grid carbon intensity and battery replacement frequency. This approach may well under or overestimate the true energy impact of ATs. Consequently, we also quantify the consumption/emissions footprint required of ATs to produce a net positive energy impact, acknowledging that variations in those assumptions would demand a less or more stringent AT energy profile (detailed in section 3.4).

AT fare-per-mile estimates drop as technology advances (figure 2). Cost parity between ATs and CTs is achieved at low levels of technological maturity regardless of travel time costs. However, achieving cost parity between ATs and PVs requires both high levels of technological maturity and consideration of travel time cost.

Zoom In Zoom Out Reset image size

Elasticity testing of AT-related input costs and parameters reveals several effects (table 1). Assuming high levels of technological maturity, capacity utilization approaches unit elasticity, dominating other model parameters. This is followed by annual mileage, envisioned profit-per-mile and licensing, whose changes exert disproportional influence on AT fares relative to other model parameters (i.e. financing, insurance, maintenance, cleaning, fuel, safety oversight and operational lifespan).

We use sensitivity analysis to identify the conditions under which AT fares become cost competitive with PVs. We focus on inputs/parameters for AT fleet operators—specifically, capacity utilization, annual mileage and envisioned profit-per-mile. This focus is intentional as reductions in other input costs/parameters, specifically financing, maintenance and fuel depend on a complex interaction between and across economic actors and conditions and hence are beyond the control of AT fleet operators.

Our model estimates that increasing AT capacity utilization rates (from 52% to 53%) produces—assuming high levels of technological maturity and consideration of the value of travel time—fares comparable with PVs. Absent consideration of travel time, cost parity between ATs and PVs requires—ceteris paribus—raising AT capacity utilization rates from 52% to 78%. Requisite utilization rates are lower as annual mileage increases and/or envisioned profit-per-mile decreases.

We draw four conclusions from the aforementioned results.

Firstly, ATs can offer fares comparable to or better than CTs today. At current $3.55 per mile fares, our model estimates that operator profits would rise from $0.27 to between $0.95 and $1.38 per mile respectively, subject to ATs' technological maturity. Alternatively, if fleet operators' profits did not increase, per mile fares would decline from $3.55 to between $1.42 and $2.24.

Second, while ATs should be competitive with CTs, they are unlikely to offer a financial proposition comparable to PVs. At high levels of technological maturity, AT fares remain—ceteris paribus—costlier than PV ownership ($1.42 versus $0.95 per mile). This effect persists even when we take account of generalized travel time costs. Consideration of these costs—which raises PV ownership costs to $1.40—improves the value proposition of ATs relative to PVs but fails to deliver cost parity.

Third, the lack of cost parity between ATs and PVs is influenced by previously overlooked factors—specifically capacity utilization and operator-envisioned profit expectations. Absent consideration of these factors, our model estimates that AT fares could be as low as $0.47 per mile, significantly lower than the cost of owning and operating PVs. However, our model estimates that the inclusion of the aforementioned factors, the most influential being capacity utilization, yields a minimum fare of $1.42.

Fourth, the impact that range reductions—potentially induced by pairing vehicle electrification and vehicle automation [23, 24]—have on AT fares, depends on whether these reductions prompt changes in occupied miles travelled. A 15% range reduction requires—ceteris paribus—that AT fare increase from the existing lower bound, $1.42, to at least $1.54 if the absolute number of miles covered by riders decreases proportionally. Conversely, no change in the absolute number of rider miles—which effectively increases the AT's utilization rate—produces fares as low as $1.31 ⁶ .

Our financial results admittedly differ from past studies demonstrating cost competitiveness of ATs with PVs [12–14, 19]. The primary reason for this is that our model accounts for capacity utilization considerations and operator-envisioned profit expectations. Although the inclusion of these factors 'worsens' an AT's financial proposition, their consideration is timely and consistent with commercial fleet operator business practices [20, 21].

Turning to the energy impact of ATs, we first consider instances in which AT fares are comparable to those offered by CTs today—a scenario our model predicts is plausible. Under this scenario, presuming demand for AT services parallels that associated with CTs, our model predicts net consumption and emissions reductions of up to 54.39% and 33.07% respectively, the precise figure depending on the magnitude of fleet overhaul (i.e. number of CTs that are replaced by ATs).

Predicting the environmental impact of ATs with fare reductions relative to the status quo is more challenging. On the one hand, our model predicts that fleet operators can—at high levels of technological maturity—offer lower fares and earn profits consistent with those realized today. However, these fares remain costlier than PV ownership.

At first glance, the inability of ATs to fiscally compete with PVs suggests that no change in energy demand/emissions is likely. Given the high(er) cost of using ATs versus PVs, drivers should choose the latter for mobility. Yet, recent research suggests that ATs offering fares lower than CTs but higher than PVs may cannibalize trips away from existing mobility options and induce greater demand [29]. Our model predicts AT fares within that range.

Cannibalization need not necessarily produce a negative environmental impact. Mobility mode switches from PVs to ATs may reduce consumption owing to the latter's improved energy efficiency [6–9]. However, consumption may increase if originally selected mobility modes are more fuel efficient [27, 29]. Further increases are likely owing to induced demand in the form of trips that would otherwise not exist absent the increased affordability of AT services.

We explore this scenario by leveraging pre-existing findings, applicable assumptions and data for our target market [5, 27, 29]. Our model yields four results.

First, demand stimulated by introducing low-cost AT services cannot be accommodated without increasing the size of the existing taxi fleet. Our model estimates that—assuming single ridership—up to 1.98% of demand (both existing and new) may be accommodated. However, maintaining the status quo fleet size of taxis decreases net energy consumption and emissions, owing to an AT's energy advantage over CTs.

Second, accommodating demand totality via AT fleet size increases may reduce the total number of vehicles needed to meet aggregate demand. This effect is attributed to higher daily mileage and by consequence, the number of trips ATs complete relative to PVs. Our model estimates that the magnitude of overall fleet size decreases realized via AT introduction depends on homogeneity in trip timing behavior (figures 3(a) and (b)).

Zoom In Zoom Out Reset image size

Third, accommodating aggregate demand using ATs produces a net increase in energy consumption and emissions. This increase is foremost driven by induced demand followed by mode switches from pooled PVs to ATs. The magnitude of the net increase in energy depends on trip timing behavior. Homogeneity in trip timing—such as morning and evening peaks during which a significant proportion of all daily journeys is completed—produces higher net increases in consumption (figures 4(a) and (b)) and emissions (figures 5(a) and (b)) respectively.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Fourth, increasing capacity utilization rates improves an AT's energy proposition. The magnitude of the increase required to achieve net zero consumption and emissions depends on trip timing behavior. Homogeneity in trip timing requires the achievement of higher utilization rates (66% and 91% for consumption and emissions respectively) compared to instances in which trip timing is uniformly spread through the day (57% and 79% for consumption and emissions respectively).

Collectively, our findings suggest that the confluence of vehicle electrification, vehicle automatization and vehicle sharing may—based on our financial assessment, profiled AT and specific market characteristics—be insufficient to achieve favorable energy outcomes, relative to the status quo.

Given the aforementioned results, how can the energy benefits of ATs be realized absent the energy costs associated with their use? In this section, we identify pathways for achieving this outcome and quantify the thresholds required for each of these pathways.

3.3.1. Pathway 1—technological improvements

Achieving net zero consumption/emissions via ATs may be achieved via technological advances: specifically, a superior vehicle energy profile. Our model estimates that a favorable energy outcome requires—presuming homogeneity in trip timing behavior—a fuel economy of 145.32 MPGe and an emissions profile of between 91.14 CO₂ grams per mile (figure 6). Realization of a higher fuel economy advantage, specifically 245 MPGe, precludes the need for improving an AT's emissions profile given interdependencies between fuel economy and emissions [27]. Existing literature suggests pairing vehicle electrification with automatization makes this outcome plausible [8, 28].

Zoom In Zoom Out Reset image size

Absent this advantage, a favorable energy outcome may be achieved via improvements in fuel economy, lifetime distance travelled and reductions in emissions from vehicle manufacturing and fuel production. Our model identifies improvements in fuel economy and emissions reductions from fuel production as most responsive to producing an energy profile that can, at achievable capacity utilization rates [32–34], produce a favorable energy outcome. Presuming the realization of a consumption advantage required to achieve net zero emissions, the requisite emissions profile may be achieved—ceteris paribus—by reducing emissions from fuel production from 121 CO₂ grams per megajoule to 73 CO₂ grams per megajoule ⁷ .

3.3.2. Pathway 2—behavioral improvements

3.3.3. Pathway 3—logistical improvements

The energy predictions surmised thus far and the efficacy of the pathways proposed may be challenged given uncertainties endemic to any nascent technology. For example, while our hypothetical AT assumes an energy profile similar to existing electric vehicles, the reality could be different. Higher AT computing demands could increase energy consumption while drive cycle optimization may reduce it [6–9, 25]. Speed selection and conformance may also influence an AT's energy profile [8, 28]. Our estimates may hence under or overrepresent the true energy impact of ATs. We have addressed this critique by estimating the energy profile, ride-pooling requirements and smart deployment thresholds required of ATs to achieve a favorable energy outcome.

However, the precision of these figures are themselves determined by market specific attributes. For example, the energy profile of existing transportation alternatives (specifically CTs, PVs and PT), the frequency of their use following AT introduction, the magnitude of induced demand and homogeneity in trip timing behavior, all influence the thresholds for the aforementioned pathways. But to what extent? Our final analysis addresses this question (table 2).

Our model estimates that the stringency of proposed net zero pathways is influenced principally by homogeneity in trip timing and to a lesser extent, the energy profile of PVs currently servicing the market. The former reflects—presuming fluctuations in rider demand—an increase in the number of ATs required to meet this demand, which by consequence, necessitates more stringent technological, behavioral and logistical thresholds to ensure a favorable energy outcome. PVs' role in influencing threshold requirements reflects this mobility mode's dominance in meeting existing mobility needs. Consequently, changing the energy profile of this mode significantly influences the aforementioned requirements.

Lessening light-duty vehicle emissions is timely given their contribution to climate change and societal morbidity [3, 4, 12]. However, practical and politically viable solutions entail recognition of the economic benefits linked to existing mobility on demand options [44]. Auto ownership in particular offers not only access to jobs, but also to educational opportunities, health care, food and recreation [45]. Consequently, addressing externalities linked to light-duty-vehicle use cannot—for reasons of economic mobility—impede upon their key offering: inexpensive, convenient point-to-point travel.

ATs have been seen as an optimal solution [30, 31]. The confluence of vehicle electrification, vehicle automation and vehicle sharing are viewed as a means of offering both cost-effective mobility on demand and reduced energy consumption and emissions [12, 24, 25]. Our results suggest otherwise. Firstly, we demonstrate that—contrary to previous work—widespread proliferation of ATs may be impeded by financial considerations (table 1). Secondly, we establish that even limited uptake of the technology may produce a net increase in energy consumption and emissions.

However, we document three pathways—conformance to an AT specific energy profile, ride-pooling and smart deployment—that would facilitate achieving a favorable energy outcome. We identify—for the first time to our knowledge—the numerical thresholds each of these pathways must meet; specifically showing what an AT's energy profile must be, how much ride-pooling is required and how many fewer miles ATs must traverse in order to lower energy consumption and emissions. We further demonstrate that these thresholds are principally influenced by homogeneity in trip timing and to a lesser extent, the energy profile of PVs currently servicing the market.

Changing PV energy profiles—with the aim of lowering AT specific technological, behavioral and logistical thresholds—is admittedly challenged by the relatively slow rate of PV fleet turnover [46]. Similar challenges do not, however, necessarily persist for influencing trip timing homogeneity. Remote work, staggered work schedules and leisure activity engagement during non-peak times could reduce AT threshold requirements via more heterogenous trip timing distribution. However, we acknowledge that a significant and sustained shift towards working from home in particular, could have impacts elsewhere in the energy system, such as those related to preferred modes of transport and demand for office space [47].