The role of dealer demonstration in the adoption of electric vehicles

3.1. Customer and vehicle characteristics

We consider a unit mass of customers with heterogeneous total mobility needs. We consider that each customer drives a total number of f trips throughout a vehicle’s useful life. Customers are heterogeneous in their trip frequency and we assume that f ∼ Uniform [ 0 , F ] . Moreover, each trip has a priori an uncertain length ℓ ∼ Uniform [ 0 , 2 L ] ; thus, trips can be long or short, with an average length of L . We normalize L = F = 1 without loss of generality. We denote a customer’s total mobility needs from all trips taken over a vehicle’s useful life by θ , which is then given by θ ≡ f E [ ℓ ] = f . Accordingly, the total mobility need of customers is heterogeneous, and thus, given by θ ∼ Uniform [ 0 , 1 ] . Note that while we capture heterogeneity in trip frequency f in our main model, we assume that customers are homogeneous in the distribution of their trip lengths, which allows us to obtain analytically tractable and clear insights. In Section 6, we generalize our model to consider heterogeneity in trip lengths and verify the robustness of our key results.

Customers maximize their net utilities by choosing among three options: (i) buy a conventional vehicle at price p g , (ii) buy an electric vehicle at price p e , or (iii) refrain from buying a vehicle; in this case, customers meet their mobility needs through their outside option (e.g., public transportation) and derive a nominal utility, which we normalize to zero. Customers who buy a vehicle derive a gross utility v per unit of mobility need (e.g., mile) met by the vehicle. ⁵ They also incur a per-mile operating (e.g., fuel and maintenance) cost denoted by k g and k e for the conventional and electric vehicle, respectively. We assume that the operating cost is lower for an electric vehicle than a gasoline vehicle (AAA, 2022a, 2022b), that is, k e < k g , and that the gross utility exceeds the operating costs, that is, v > k g , to avoid trivial cases in which no customers buy a vehicle; both assumptions are verified to hold with calibrated data in online E-Companion C. We also assume that the conventional vehicle can cover all trips due to the extensive refueling infrastructure (i.e., the dense network of gas stations). Thus, a customer with total mobility needs θ derives a net utility U g ( θ ) = ( v − k g ) θ − p g from buying a conventional vehicle.

In contrast, an electric vehicle may not cover trips longer than a certain distance due to shorter range, less extensive refueling infrastructure, and longer refueling duration than a comparable gasoline vehicle. Let x < 2 L = 2 denote the maximum distance an electric vehicle can cover under standard conditions, which we refer to as the designed range to emphasize its dependence on the vehicle’s design characteristics, such as battery capacity, weight, aerodynamics, and rolling resistance (see National Research Council, 2015: 19). In practice, customers can observe the designed range from the rated range published on the vehicle’s EPA Fuel Economy and Environment Label.

We draw an important distinction between an electric vehicle’s designed versus achievable range. We define achievable range as the extent to which an electric vehicle can meet a customer’s total mobility needs under idiosyncratic conditions. Electric vehicles typically achieve lower-than-designed ranges under, for example, high-speed driving, heavy acceleration, and extensive HVAC use. Since these factors are specific to each customer at the time of driving, their combined effects on range can be ascertained only through hands-on experience with an electric vehicle. ⁶

In light of these practical observations, we consider that the electric vehicle’s achievable range, as denoted by x ~ , is ex-ante unknown and can take either a high value equal to the designed range x , or a low value equal to ς x where ς ∈ ( 0 , 1 ) . Customers are uncertain about the achievable range and expect it to be high with probability α ∈ ( 0 , 1 ) and low with probability 1 − α . ⁷ This prior captures information consumers have from external sources (e.g., online reviews) before demonstration. We consider rational expectations, that is, after purchase, customers experience a high achievable range x ~ = x with probability α and low achievable range x ~ = ς x with probability 1 − α . Customers who experience an achievable range x ~ use the electric vehicle for all trips no longer than x ~ miles; for longer trips, they resort to the outside option and obtain zero net utility (see Lim et al., 2015 for a similar formulation). Therefore, in the absence of demonstration, a customer expects to obtain a net utility U e ( θ ) = ( v − k e ) f ( α ∫ 0 x ( ℓ / 2 ) d ℓ + ( 1 − α ) ∫ 0 ς x ( ℓ / 2 ) d ℓ ) − p e = ( α + ( 1 − α ) ς 2 ) ( x / 2 ) 2 ( v − k e ) θ − p e from buying an electric vehicle, where k e ∈ ( 0 , k g ) denotes the per-mile operating cost and p e the selling price of the electric vehicle. ⁸ To further simplify this and subsequent utility functions, we define r ≡ ( x / 2 ) 2 , τ ≡ ς 2 , and obtain U e ( θ ) = ( v − k e ) ( α + ( 1 − α ) τ ) r θ − p e . This suggests that the range limitation can be equivalently formulated in terms of drivers’ total mobility needs θ , with the electric vehicle meeting a fraction r versus τ r of these needs under a high versus low achievable range. For ease of exposition, we henceforth use r and τ r and refer to r as range.

3.2. Demonstration services

The uncertainty over the achievable range may prevent customers from purchasing an electric vehicle. The dealer can influence this decision by providing demonstration services (e.g., extended test drives) that enable customers to better understand the range an electric vehicle can achieve under their idiosyncratic driving conditions. ⁹ To capture the impact of demonstration services on customers’ purchase decisions, we adopt the established framework of demonstration for innovative products with “deal-breaking” attributes (Boleslavsky et al., 2017). Specifically, we assume that a customer, who after purchase experiences a low achievable range, draws from the demonstration an unfavorable impression with probability δ ∈ ( 0 , 1 ] , and a favorable one with probability 1 − δ ; hence, P ( unfav | r ~ = τ r ) = δ and P ( fav | r ~ = τ r ) = 1 − δ . In other words, demonstration accurately reveals a low achievable range with probability δ , which, therefore, represents the informativeness of demonstration. We also consider that a customer, who after purchase experiences a high achievable range, always draws a favorable impression from demonstration as he never encounters a deal-breaking attribute; hence, P ( fav | r ~ = r ) = 1 and P ( unfav | r ~ = r ) = 0 . For example, customers with high-speed driving and/or extensive HVAC use experience a low achievable range from daily usage (i.e., r ~ = τ r ), and may be able to infer this from an extended test drive upon observing fast battery discharge. Those without such driving habits experience a high achievable range (i.e., r ~ = r ) from daily usage and therefore, do not observe fast battery discharge during the test drive.

Based on the impression from the demonstration, customers update their beliefs over r ~ following Bayes’ rule; hence, P ( r ~ = τ r | unfav ) = 1 and P ( r ~ = r | fav ) = α / ( 1 − ( 1 − α ) δ ) ≡ α ′ > α . In words, customers drawing an unfavorable impression expect the electric vehicle to achieve a low range with certainty, whereas customers drawing a favorable impression expect a high achievable range with probability α ′ , which is increasing in the level of demonstration informativeness δ ∈ ( 0 , 1 ] . Specifically, when δ = 1 , we have α ′ = 1 and demonstration is fully informative—customers drawing a favorable versus unfavorable impression expect high versus low achievable range with certainty. We analyze the economic and environmental implications of offering demonstration at any informativeness level (full or partial) in the paper. In online E-Companion H.1, we numerically examine how improvements in demonstration informativeness affect these implications.

We use the subscript ψ ∈ { f , u } to distinguish between the cases where a customer draws a favorable ( ψ = f ) versus an unfavorable ( ψ = u ) impression. The corresponding net utilities that customers expect to derive after the purchase of an electric vehicle are given by U f e ( θ ) = ( v − k e ) ( α ′ + ( 1 − α ′ ) τ ) r θ − p e and U u e ( θ ) = ( v − k e ) τ r θ − p e . To simplify notation, let q g ≡ v − k g , q e ≡ ( α + τ ( 1 − α ) ) ( v − k e ) r , q f ≡ ( α ′ + τ ( 1 − α ′ ) ) ( v − k e ) r , and q u ≡ τ ( v − k e ) r . Customers’ net utilities can thus be simplified as U g ( θ ) = q g θ − p g , U e ( θ ) = q e θ − p e , and U ψ e ( θ ) = q ψ θ − p e for ψ ∈ { f , u } , where q g , q e , and q ψ can be seen as the effective, as perceived by the customer, qualities of the corresponding mobility options.

3.3. Problem formulation

The sequence of events is as follows. First, the dealer decides whether to invest C d and offer demonstration services. Second, the manufacturer sets wholesale prices of the electric and conventional vehicles, as denoted by w e and w g , respectively. Third, the dealer procures vehicles from the manufacturer and determines the selling prices p e and p g . Fourth, customers arrive and choose to buy a conventional versus electric vehicle or entirely refrain from buying. ¹⁰ Notably, the dealer decides the demonstration strategy before procuring and selling vehicles, because offering demonstration requires substantial resources and lead time for staff training, developing extended test drive programs, and upgrading service and information infrastructure. In online E-Companion H.6, we consider an alternative setting, where the dealer decides the demonstration strategy after the manufacturer sets wholesale prices. We numerically find that all our structural results continue to hold, with demonstration being relatively more attractive as the manufacturer can better incentivize it by setting a high electric vehicle wholesale price.

For a given demonstration and pricing strategy, customers maximize their net utilities by deciding whether to buy a conventional versus electric vehicle or entirely refrain from buying. The dealer maximizes her profit by setting the vehicle prices and, when offering electric vehicles, whether to provide demonstration. Let π d ( p g , p e ) and π n ( p g , p e ) denote dealer profits under a given price pair ( p g , p e ) with and without demonstration, respectively; her pricing problems are hence given by:

The manufacturer’s problem, at a given dealer demonstration strategy, is to determine the wholesale prices to maximize its profits from producing and selling the vehicles to the dealer as follows:

We assume c g < q g = v − k g to rule out trivial cases in which the conventional vehicle is unprofitable to sell, and similarly, c e < q e for the electric vehicle. In addition, we assume that the electric vehicle’s operating cost k e is sufficiently low, while the achievable range for customers receiving an unfavorable impression (parameterized by τ ) is sufficiently limited; see online E-Companion A for details. These assumptions allow us to focus on the interesting cases where the electric vehicle’s operating cost advantage and limited (achievable) range disadvantage are such that demonstration is salient in influencing customers’ purchase decisions.

Based on the optimal purchase decisions presented above, we derive the demand functions for the electric and conventional vehicles, substitute them into the dealer and manufacturer profits, and solve for their optimal (selling and wholesale) prices, with and without demonstration separately. Below, we present how offering demonstration affects customers’ purchase decisions in equilibrium.

When the production cost of the electric vehicle is prohibitively high (i.e., c e ⩾ q f − q g + c g ), the dealer never finds it profitable to sell electric vehicles at the equilibrium wholesale prices set by the manufacturer, and therefore, the demonstration has no impact on customers’ purchase decisions. In the remainder of the paper, we focus on the nontrivial case of c e < q f − q g + c g . To disentangle the different effects of offering demonstration, we present the equilibrium purchase decisions of customers who in the absence of demonstration would buy versus not buy an electric vehicle, separately.

Since the CAFE level is effectively the ratio of electric to conventional vehicle sales, the impact of offering demonstration on the CAFE level depends on its effect on both electric vehicle sales (as derived in Proposition 2) and conventional vehicle sales, which we analyze next.

Proposition 4

Offering demonstration leads to greater conventional vehicle sales if and only if c e ⩽ G ( r , c g , δ ) , which is increasing in r and c g , and satisfies G ( r , c g , δ ) ⩾ A ( r , c g , δ ) .

As expected, when offering demonstration leads to lower electric vehicle sales (i.e., c e ⩽ A ( r , c g , δ ) ), it also leads to greater conventional vehicle sales. More interestingly, even when offering demonstration leads to greater electric vehicle sales, conventional vehicle sales still increase when c e ∈ ( A ( r , c g , δ ) , G ( r , c g , δ ) ) . In this case, conventional vehicle sales increase because customers who receive an unfavorable impression from the demonstration switch from electric to conventional vehicles. Although this lowers electric vehicle sales, offering demonstration leads to greater electric vehicle sales overall through market expansion. Specifically, in the absence of demonstration customers with less total mobility needs do not purchase any vehicle, but in the presence of demonstration, they may purchase an electric one after receiving a favorable impression.

Building on the analysis above, we now examine the impact of offering demonstration on the manufacturer’s CAFE level. When c e ⩾ G ( r , c g , δ ) , offering demonstration leads to lower conventional and greater electric vehicle sales, thus a higher CAFE level. When c e ⩽ A ( r , c g , δ ) , offering demonstration leads to greater conventional and lower electric vehicle sales, thus a lower CAFE level. When c e ∈ ( A ( r , c g , δ ) , G ( r , c g , δ ) ) , offering demonstration leads to greater electric and conventional vehicle sales. Hence, the overall impact on the CAFE level is more nuanced and is further explored below.

Proposition 5

When c e ∈ ( A ( r , c g , δ ) , G ( r , c g , δ ) ) , offering demonstration leads to a lower CAFE level if and only if c e ⩽ F ( r , c g , δ ) , which is increasing in c g and satisfies F ( r , c g , δ ) < G ( r , c g , δ ) .

The above result shows that offering demonstration leads to a lower CAFE level when the electric vehicle technology is more developed and/or the conventional vehicle production cost is high (i.e., c e ⩽ F ( r , c g , δ ) which is increasing in c g ; ¹⁷ see Figure 6). Under these conditions, the electric vehicle is relatively more attractive even in the absence of demonstration. This further implies that: (i) electric vehicle sales are already high in the absence of demonstration, so more customers switch to conventional vehicles after receiving an unfavorable impression; (ii) only customers with very low total mobility needs purchase the conventional vehicle, so they are less attracted to the electric vehicle’s lower operating cost after receiving a favorable impression. In other words, offering demonstration leads to a larger increase in conventional than electric vehicle sales, hence the CAFE level increases.

5.2. Optimal wholesale pricing under CAFE compliance

We now examine the manufacturer’s wholesale pricing decisions to ensure CAFE compliance and the resulting effect on electric vehicle adoption in equilibrium. To this end, we formulate the manufacturer’s problem as

maximize w g , w e ⩾ 0 π n m ( w g , w e ) = ( w e − c e ) D e n ( p g n ( w g , w e ) , p e n ( w g , w e ) ) + ( w g − c g ) D g n ( p g n ( w g , w e ) , p e n ( w g , w e ) ) , s.t. D e n ( p g n ( w g , w e ) , p e n ( w g , w e ) ) ⩾ R D g n ( p g n ( w g , w e ) , p e n ( w g , w e ) )

in the absence of demonstration, and

maximize w g , w e ⩾ 0 π d m ( w g , w e ) = ( w e − c e ) D e d ( p g d ( w g , w e ) , p e d ( w g , w e ) ) + ( w g − c g ) D g d ( p g d ( w g , w e ) , p e d ( w g , w e ) ) , s.t. D e d ( p g d ( w g , w e ) , p e d ( w g , w e ) ) ⩾ R D g d ( p g d ( w g , w e ) , p e d ( w g , w e ) )

in the presence of demonstration. The objectives π n m and π d m are the same as before, but now the manufacturer is subject to the CAFE constraint D e / D g ⩾ R .

First, for a given demonstration strategy, we examine how the manufacturer should adjust the wholesale prices under a more stringent CAFE standard. Since this requires a larger ratio of electric relative to conventional vehicle sales, one might expect the manufacturer to lower the wholesale price of the electric vehicle and raise the wholesale price of the conventional vehicle. However, our analysis shows that this is not always the case.

Proposition 6

For a given demonstration strategy, there exist R w and R a ⩾ R w , such that: (i)

the optimal conventional vehicle wholesale price is increasing in R ;

(ii)

the optimal electric vehicle wholesale price is decreasing in R if and only if R < R w ;

(iii)

the equilibrium electric vehicle adoption is increasing in R if and only if R < R a .

Under a more stringent CAFE standard, the manufacturer should raise the wholesale price of the conventional vehicle w g to reduce its sales, and lower the wholesale price of the electric vehicle w e only when the CAFE standard is not very stringent (i.e., R < R w ). As the CAFE standard tightens further, the manufacturer should instead raise w e and meet the CAFE standard solely by decreasing conventional vehicle sales through a higher w g . This is more profitable than lowering the already low w e , which would further erode the profit margin. Moreover, raising w g not only incentivizes electric vehicle purchases but also leads customers with lower total mobility needs to refrain from purchasing altogether, further decreasing conventional vehicle sales. This allows the manufacturer to meet the tighter CAFE standard while profitably raising w e , which leads to decreased electric vehicle adoption when the CAFE standard is sufficiently stringent (i.e., R > R a ).

Next, we compare results with and without demonstration to gain further insights into its moderating role in the effects of a tighter CAFE standard R . As illustrated by Figure D.1(a) and (c) in online E-Companion D, threshold R w is higher in the presence of demonstration—it expands the range of CAFE standard wherein further tightening leads the manufacturer to lower w e (i.e., R < R w ). To see the intuition, recall from Lemma 3 that offering demonstration enables the manufacturer to raise w e and obtain a higher profit margin ( w e − c e ). As a result, it can meet a tighter CAFE standard more efficiently by lowering w e as it does not significantly erode electric vehicle profitability.

We also find that a tighter CAFE standard leads to lower electric vehicle adoption only in the presence of demonstration. This occurs when the tighter CAFE standard leads the manufacturer to raise w e , for example, R > R a = R w = 11.6 in Figure D.1(c) and (d). In the absence of demonstration, however, a tighter CAFE standard always leads to greater electric vehicle adoption, even though the manufacturer raises w e for a wider range, for example, R > R w = 4.2 in Figure D.1(a). This is because, for CAFE compliance, the manufacturer raises w g , which makes the electric vehicle relatively more attractive. Although it also raises w g in the presence of demonstration, the electric vehicle adoption may decrease as the CAFE standard tightens. This is because offering demonstration enhances the quality differentiation between electric and conventional vehicles—customers receiving an unfavorable impression from the demonstration will not buy an electric vehicle even when w g increases.

6. Heterogeneous trip length distributions

In the main model, we assumed that customers are heterogeneous in their trip frequency, but homogeneous in the distribution of their trip lengths. In particular, customers may have trips of different lengths that follow the same distribution ℓ ∼ Uniform [ 0 , 2 L ] , where the average trip length is given by L , normalized to L = 1 . We now consider an extension where customers have heterogeneous trip length distributions to verify the robustness of our key results and offer new insights.

We consider two customer segments that have the same total mobility needs but differ in trip frequency and length. A fraction λ of customers (referred to as high-type customers hereafter) are similar to customers in our main model; each customer takes f ∼ Uniform [ 0 , 1 ] trips, each of which has an a priori uncertain length ℓ ∼ Uniform [ 0 , 2 ] . The remaining 1 − λ fraction of customers (referred to as low-type customers hereafter) have more frequent trips that are shorter on average than the trips of high-type customers. In particular, we assume a trip frequency f l ∼ Uniform [ 0 , F l ] among these customers, with each customer’s trip length ℓ l ∼ Uniform [ 0 , 2 L l ] . Moreover, to ease comparisons, we assume that the distribution of the total mobility needs is the same across both types of customers. That is, we have that f l L l = θ ∼ Uniform [ 0 , 1 ] , or equivalently L l F l / 2 = E [ θ ] = 1 / 2 . This implies that F l = 1 / L l and that L l < 1 = L and F = 1 < F l , which captures that low-type customers have shorter but more frequent trips than the high type.

We now derive customers’ utilities from purchasing an electric vehicle. We assume 2 L l ⩽ ς x < x ⩽ 2 L = 2 to focus on the interesting case wherein high- and low-type customers are sufficiently different in trip lengths and therefore, affected by the demonstration differently. In the absence of demonstration, customers expect a high achievable range (i.e., x ~ = x ) with probability α , and a low achievable range (i.e., x ~ = ς x ) with probability 1 − α . Thus, the net utilities, as denoted by U h e ( θ ) for high- and U l e ( θ ) for low-type customers are

U h e ( θ ) = ( v − k e ) f h ( α ∫ 0 x ℓ 2 d ℓ + ( 1 − α ) ∫ 0 ς x ℓ 2 d ℓ ) − p e = ( v − k e ) ( α + ( 1 − α ) τ ) r θ − p e , U l e ( θ ) = ( v − k e ) f l ( α ∫ 0 2 L l ℓ 2 L l d ℓ + ( 1 − α ) ∫ 0 2 L l ℓ 2 L l d ℓ ) − p e = ( v − k e ) f l L l − p e = ( v − k e ) θ − p e .

In the presence of demonstration, customers update their beliefs based on the impression they receive. As in the main model, they expect a high achievable range with probability α ′ > α after receiving a favorable impression and zero probability after receiving an unfavorable impression; seeSection 3.2. Accordingly, we derive the utility expressions as U h ⋅ f e ( θ ) = ( v − k e ) ( α ′ + ( 1 − α ′ ) τ ) r θ − p e , U h ⋅ u e ( θ ) = ( v − k e ) τ r θ − p e , and U l ⋅ f e ( θ ) = U l ⋅ u e ( θ ) = ( v − k e ) θ − p e .

There are two important observations from the above: First, low-type customers are less affected by the range limitation due to their shorter trips. Although high-type customers can use the electric vehicle only for trips within range (i.e., ℓ ⩽ 2 L ), low-type customers can use the electric vehicle for all trips as they are within range, that is, 2 L l ⩽ ς x . Aggregating over trips, the electric vehicle only meets a fraction ∫ 0 x ~ ℓ 2 d ℓ = ( x ~ / 2 ) 2 < 1 of the high-type customers’ total mobility needs, but meets all the mobility needs of the low-type customers, under the achievable range x ~ ∈ { x , ς x } . That is, the low-type customers do not experience any utility loss, but the high-type customers face utility loss due to their longer trips being limited by the range of the electric vehicle. This implies that the effect of offering demonstration is different for low-type versus high-type customers. Second, in the presence of demonstration, the order of customers’ purchase decisions upon receiving an unfavorable impression does not switch for all customers (see online E-Companion E for details). As in our main model, high-type customers with high total mobility needs switch to preferring a conventional vehicle as they now perceive a lower quality due to inferring a low achievable range. However, low-type customers, whose trips are shorter and within range, still expect a higher quality from the electric vehicle, and continue to prefer an electric vehicle even if their total mobility needs are high.

While we can analytically characterize the customer purchase decisions under this generalization, it is analytically intractable to characterize the equilibrium wholesale prices and other outcomes. See online E-Companion E for all the technical details. Accordingly, we resort to numerical analysis to compare the equilibrium outcomes in the absence and presence of demonstration to offer insights for when the dealer should offer demonstration, and how this influences adoption and usage emissions. We find that all our structural results in Section 4 continue to hold as can be seen by comparing Figure 7 with Figures 3(a) and 5(a).

OPEN IN VIEWER

Figure 7.

When to offer demonstration services and equilibrium effect on adoption D e and usage emissions E M U .\emph{Note.} The figure is based on calibrated data, where v = ⧸ c 35 /mile , k e = ⧸ c 11.98 /mile , k g = ⧸ c 21.38 /mile , τ = 0.35 , α = 0.8 , δ = 0.8 , c g = $ 15 , 900 , λ = 0.9 , η U = 0.319 lb/mile, γ U = 3.03 , and the dashed line represents the current value of vehicle production cost.

We next focus on how heterogeneity in trip length distributions affects the adoption of the demonstration, and thus, the equilibrium adoption and usage emissions (see Figure 8, where the effect of greater heterogeneity can be inferred by considering a lower λ ). We find that greater heterogeneity exacerbates our result that offering demonstration can lead to lower electric vehicle adoption (see Figure 8(a)). A greater heterogeneity implies more low-type customers who are less affected by range limitation. Moreover, with the demonstration, high-type customers who receive an unfavorable impression prefer the conventional vehicle to the electric vehicle, so the dealer sells the conventional vehicle to them, which leads to lower electric vehicle adoption. Similarly, a greater heterogeneity also strengthens our result that the demonstration can lead to lower emissions (see Figure 8(b)). With more low-type customers, the dealer can profitably raise the price of the electric vehicle and induce more high-type customers to purchase a conventional vehicle. Offering demonstration mitigates the dealer’s incentive to do so as high-type customers who receive a favorable impression perceive a higher effective quality of the electric vehicle and prefer it to the conventional vehicle.

OPEN IN VIEWER

Figure 8.

Equilibrium effect of the fraction of high-type customers λ on electric vehicle adoption D e and usage emissions E M U with and without demonstration.Note. The figure is based on calibrated data, where v = ⧸ c 35 /mile , k e = ⧸ c 11.98 /mile , k g = ⧸ c 21.38 /mile , τ = 0.35 , α = 0.8 , δ = 0.8 , c e = $ 23 , 000 , c g = $ 15 , 900 , η U = 0.319 lb/mile, γ U = 3.03 , and the dashed line represents the current value of vehicle production cost. We use different r values in the left and right panels to more clearly illustrate the impact of offering demonstration on D e and E M U .

In this paper, we formally examined the impact of offering demonstration services on dealer profits, the manufacturer’s regulatory compliance, the adoption of electric vehicles, and overall environmental impact. Our results highlight that the economic, environmental, and policy implications of offering demonstration are nuanced, and depend critically on operational characteristics, including production costs and the electric vehicle’s designed range. These findings offer a set of key managerial and policy insights.

First, offering demonstration services can simultaneously improve dealer profits, promote electric vehicle adoption, and reduce emissions when the electric vehicle technology is relatively less developed; see region I in Figure 6. This highlights the potential of community-based organizations’ ongoing initiatives in facilitating dealer demonstrations (Drive Electric Northern Colorado, 2021; Evolve Houston, 2023; Peninsula Clean Energy, 2021) to generate both economic and environmental benefits. The effectiveness of such initiatives depends on the state of electric vehicle technology. As the production cost c e decreases and/or the designed range r increases due to larger battery capacity and improved charging infrastructure, offering demonstration becomes more likely to decrease electric vehicle adoption and increase emissions (see region IV in Figure 6).

Second, community-based organizations should exercise caution when using electric vehicle adoption as the primary metric for evaluating the environmental benefit of their initiatives that facilitate dealer demonstrations. When offering demonstration leads to greater electric vehicle adoption, it also leads to lower emissions when the electric vehicle technology is less developed, but can result in higher emissions once the technology becomes more advanced; see regions II and III in Figure 6.

Third, manufacturers should carefully consider the impact of dealer demonstrations on their CAFE level when deciding whether to facilitate demonstration by providing staff training and/or promoting extended test drive programs on their websites (Volvo, 2023). Even when dealer demonstration promotes electric vehicle adoption, it may lower the manufacturer’s CAFE level by leading to a proportionally greater increase in conventional vehicle sales (see region III in Figure 6).

Fourth, policies that disincentivize conventional vehicle production (e.g., through taxation or emission regulations) or tighten the CAFE standard have nuanced environmental implications through their interactions with dealer demonstrations. Although conventional vehicle production disincentives on their own may promote electric vehicle adoption, they can induce the dealer to refrain from offering demonstration services, which in turn can decrease adoption and increase emissions when the electric vehicle technology is less developed. Moreover, in the presence of demonstration, a more stringent CAFE standard may not lead to greater electric vehicle adoption.

To derive first-order insights into the role dealers can play in facilitating electric vehicle adoption, our model relied on simplifying assumptions, which naturally open avenues for future research. For instance, building on our analytical framework, one can further explore how dealer demonstrations could help mitigate competitive pressures from other dealers or auto manufacturers. Another promising direction is to introduce information asymmetry, for example, dealers may be better informed than customers on vehicle performance, durability, and safety features, and to examine the signaling role of demonstration services. Future research can examine other aspects that customers may learn from dealer demonstrations, such as the charging experience, driving experience, or overall lifestyle fit. Other promising directions include accounting for additional sources of customer uncertainty (e.g., travel needs or charging access) and information-provision mechanisms (e.g., product reviews and owner communities). Future research may also incorporate additional considerations, such as salesforce effort, planned obsolescence, secondary markets, consumer subsidies, or battery recycling policies that promote social welfare, as well as network effects in technology diffusion, to examine their interactions with dealer demonstrations. The impact of dealer demonstrations on the manufacturer’s ability to meet environmental regulations also highlights opportunities for negotiation and coordination between dealers and automakers.

Supplemental Material