An Electric Vehicle Adoption Model for Nigeria—A Fuzzy MCDA Policy Analysis Tool with Implications for Developing Nations

Abstract

The dominant source of the vehicle fleet in developing nations is the used vehicle market in developed nations. As the automotive fleet in developed nations electrifies, so will the used vehicle market. In many cases, developing nations’ electric infrastructure is inadequate to support significant vehicle electrification. Therefore, there is an immediate need for developing nations to plan and prepare for vehicle electrification before scarcity of fossil fueled vehicles constitutes a national emergency. This research presents the Developing Nations Electric Vehicle Adoption Model (DN-EVAM), a decision support tool designed to help decision makers in developing nations address this challenge. We then use DN-EVAM to analyze the vehicle adoption landscape of Nigeria. First, we survey subject matter experts in Nigeria to identify antecedent and moderating variables relevant to Nigeria. Then we implement DN-EVAM to model the Nigerian vehicle electrification landscape. Finally, we conduct a scenario-based analysis to identify those antecedent and moderating variables most important to vehicle electrification in Nigeria. We find that for Nigerian policy makers, adoption incentives and infrastructure investments are the most critical areas of focus in the near term while investments in technology development are generally not the most attractive option.

1. Introduction

1.1. The Need for Developing-Nations-Specific EV Policy Research

Understanding the cause-and-effect relationships between the elements of techno-economic systems is a key element of effective policy design and assessment. The need for effective policy is perhaps nowhere greater than in the realm of climate change. Climate change’s roots in the most fundamental building blocks of techno-economic systems mean that effective climate change mitigation and adaptation will require coordinated global effort to achieve a fundamental shift of technology away from fossil fuels [1]. One key aspect of this shift is the decarbonization of transport systems through vehicle electrification. Electric vehicle (EV) market share has grown from 0.01% in 2010 to about 14% in 2022. Moreover, the rate of transport decarbonization is accelerating rapidly: EV market share grew by approximately 60% between 2021 and 2022 [2].

While significant attention and resources have been devoted to supporting passenger vehicle electrification in the developed world, far less attention has been paid to this question as it applies to developing nations. This is a critical gap, as developing nations represent both a large decarbonization opportunity and a vital pillar of the global used vehicle market. Between 2015 and 2020, approximately 23 million vehicles were exported to low- and middle-income countries [3]. These vehicles are often in poor mechanical condition and lack modern emission control equipment, making them some of the most polluting vehicles on the road [4].

As the developed world electrifies their transport networks, the used vehicle market will shift toward EVs. However, many developing nations currently lack the capacity to use significant numbers of electric vehicles due to low incomes, insufficient infrastructure, lack of effective EV policies, and other factors. Moreover, unlike developed nations, where EV adoption is being driven from within, developing nations are facing external pressure to adopt EVs due to their dependence on the international used vehicle market.

The need for developing nations to adapt to transport electrification is urgent and growing. This work addresses this need with the development of the Developing Nations Electric Vehicle Adoption Model (DN_EVAM). DN-EVAM is a multi-criteria decision analysis (MCDA) tool that builds on Kumar and Alok’s [5] EV-specific technology adoption model. It provides an assessment framework to assist planners in developing nations to assess their situation, identify threats and opportunities, and prioritize policy actions.

There is no universally accepted definition of developing nations [6,7,8]. However, developing economies are often characterized as having lower levels of income and infrastructure development compared to developed economies, as well as a more rapid pace of change in key social and economic indicators such as urbanization and infrastructure development [9]. While no one country can be characterized as a “typical” developing nation, these broad similarities in economic and infrastructure development often lead to similar challenges to EV adoption across different nations.

Although this work applies DN-EVAM to the specific case of Nigeria, the methods described here and the framework of DN_EVAM can be applied more broadly to transport electrification in other developing nations. Specifically, users in other countries can customize the antecedent and moderating variables used in their analysis, as well as the levels of those variables, using the methods described in Section 2 below.

Nigeria was selected as the subject for our case study because it represents a large, complex, and interesting application for DN-EVAM. It is Africa’s most populus country [10] and its largest importer of used vehicles [11], with an annual vehicle demand of approximately 500,000 units, of which about 400,000 are imported [12].

At the same time, Nigeria’s infrastructure will require significant investment to support EVs. Crucially for EV adoption, Nigeria lags in the critical area of electric generation and distribution infrastructure. One 2006 study estimated that less than 45% of the Nigerian population had access to electricity and that an investment of US$262 billion would be required to meet Nigerian electricity demand in 2030 [13]. According to the World Bank, as of 2020 Nigeria, ranks 171 out of 190 in electricity access [14] and as of 2021, only about 60% of Nigeria’s population had access to electricity [15]. Even where it is available, electricity access is often unreliable; Nigeria’s World Bank electricity reliability score is 1.4/7 [16].

Finally, Nigeria’s economic and policy landscape create additional challenges for EV adoption. Gasoline pump prices are low compared to other sub-Saharan African nations and there are no stated EV adoption goals [16]. Moreover, the import tax on battery electric vehicles is nearly double the import tax on internal combustion vehicles [17].

1.2. Technology Acceptance Models

Technology acceptance models (TAMs) explore the relationships between factors influencing acceptance (use) of a new technology and users’ intention to use or the actual use of such technologies [18]. TAMs were first introduced by Fred Davis in 1989 [19] in the context of computer technologies. In Davis’s model, acceptance (use) of a new technology results from a process whereby a user forms a behavioral intention to use a technology based on three factors: the user’s attitude toward using the technology, the perceived ease of use of the technology, and the technology’s perceived usefulness. These factors are in turn influenced by external (to the user) variables that govern the user’s perceived usefulness and/or ease of use of the technology (Figure 1).

Technology acceptance models are evolving as scholars add new variables and apply the technology adoption model to different fields [18]. Of particular relevance to this work is the model of EV adoption by Kumar and Alok [5], which we discuss in detail below.

1.3. Kumar and Alok’s Electric Vehicle Adoption Model

Kumar and Alok’s work analyzed 239 scholarly papers about EV adoption and generated a relationship map of EV adoption (Figure 2). As with Davis’s original TAM, Kumar and Alok’s relationship map relates a user’s behavioral intention to adopt EVs and actual adoption to external variables. Kumar and Alok further describe the nature of the external variables as antecedent variables that exist prior to adoption and act to promote or impede adoption, moderating variables that influence the strength of the relationship between behavioral intention to adopt and actual adoption, sociodemographic variables that influence adoption based on users’ demographics, and consequence variables that measure the effect of adoption. A key distinction between antecedent and moderating variables is that antecedent variables are generally EV-centric and have a direct impact on users’ intention to adopt, while moderating variables are exogenous to the EV adoption landscape and characterize broader societal conditions such as demographic and political trends.

1.4. The Knowledge Gap: EV Adoption in the Developing World

Kumar and Alok’s relationship map provides a clear picture of the relationships between external variables, users’ behavioral intention to adopt, and actual adoption in an EV-specific context. This relationship map, therefore, provides a useful framework for structuring policy discussion and decision making with respect to EV adoption. However, because the variables were explored within the context of developed countries, they may not necessarily apply to developing countries. Kumar and Alok’s work focused primarily on developed nations: of the 239 papers considered in their review, only 38 (15.8%) were focused on developing nations as defined by the United Nations [20] and none considered African nations. Therefore, while Kumar and Alok’s work provides insight into EV adoption in developed nations, it may not adequately address EV adoption in developing nations. This work fills this gap by proposing a methodology for adapting Kumar and Alok’s relationship map to developing nations contexts and then using this adapted model to develop the DN-EVAM analysis tool. We then demonstrate this methodology by adapting DN-EVAM to the Nigerian context.

A key challenge faced when developing this model was how to identify relevant variables for EV adoption in the developing nations context. There is a lack of research on this topic for developing nations; therefore, the variables identified by Kumar and Alok were used as a starting point. A survey of Nigerian transportation subject matter experts (SME)s was used to validate or exclude these variables from DN-EVAM and to identify possible new variables not already identified.

2. Materials and Methods

We will address the knowledge gap discussed above in two parts. First, we characterize the Nigerian EV adoption context. We conduct a survey of subject matter experts in Nigeria’s auto industry. Our survey has two goals: first, to validate or exclude Kumar and Alok’s variables as either applicable or inapplicable to the developing nations context, and second to identify variables relevant to the developing nations context that were not identified by Kumar and Alok.

Second, we develop a fuzzy multi-criteria decision analysis model (DN-EVAM) to assist decision makers with developing policy interventions to promote EV adoption. This model will be based on the relationship map developed by Kumar and Alok, but the variables and weights are Nigeria-specific, as informed by the results of the survey.

2.1. Charactarizing the Nigerian EV Adoption Context

Technology adoption models must characterize potential users’ perceptions of a technology and their attitudes toward its adoption. Several methods can be used to accomplish this, including longitudinal studies, interviews, or surveys. Davis’s original work used a longitudinal study to develop the original TAM [19].

It is also necessary to characterize the external variables affecting adoption. In Kumar and Alok’s work, the external variables were identified through a literature review; however, as was noted above, that literature was heavily focused on developed nations. Therefore, before we adopt Kumar and Alok’s framework, these variables must be validated in the Nigerian context.

2.1.1. Determining Antecedent Variables: Survey of Nigerian Transportation Experts

To validate Kumar and Alok’s antecedent variables in the Nigerian context, this work uses a survey of subject matter experts (SME)s. An online survey was chosen because it offered the most cost-effective way to reach Nigerian SMEs. There is evidence that the length of surveys is inversely related to response and completion rates [21,22] and quality of answers [21,23]. Based on this information, our goal was to create a survey that could be completed by our respondents in 15 min or less, based on the mean ideal web survey length determined by [24].

This constraint limited the number of variables that could be considered. Therefore, we used expert judgement to select 11 of Kumar and Alok’s variables for inclusion in the survey as well as two additional variables not included in Kumar and Alok’s work. The antecedent variables considered in this work are listed in Figure 3.

Prospective subject matter experts were identified based on the approved inclusion criteria: age 18 or older, member of one of the participant classifications defined below, and located in Nigeria.

Automotive SME: persons with expertise in automotive manufacture and repairs.

Passenger Vehicle SME: persons with expertise using passenger vehicles from an operational and business perspective.

Current EV Users: persons who currently own an electric vehicle.

Prospective EV User: persons who are interested in transitioning to EVs in the future.

Government Official: any government official whose duties directly relate to the regulation or implementation of policy relating to access, use, and availability of passenger vehicles.

Invitations to participate in the survey were then distributed to SMEs via in-person recruitment, social media (Facebook, LinkedIn, and WhatsApp), and email. Each participant that completed the survey was compensated with a $5 phone card to each phone number submitted in the survey.

There was a total number of 80 respondents. A total of 7 responses were invalidated due to duplicate entries and invalid responses (

Table 1. Summary of survey responses.

Respondent Category	Number of Valid Responses
Automotive Subject Matter Experts	32
Passenger Subject Matter Experts	25
Current and Prospective EV Owner	12
Government Official	4
Invalid/Duplicate	7
Total	80

), leaving 73 valid responses.

Respondents were asked to rate the importance of each of the antecedent variables using a five-point Likert scale: “insignificant”, “not important”, “neutral”, “important”, or “most important”. Figure 3 summarizes these responses. Infrastructure and range-related antecedents are most heavily weighted, with approximately 75% or more of the respondents rating these antecedents as “most important”; this is consistent with the lack of infrastructure development in Nigeria. On the other hand, customer-focused antecedents such as perceived risks and symbolic attributes are less heavily weighted.

2.1.2. Quantifying Antecedent Performance Levels

One of the challenges in this study is capturing the uncertainty and variability associated with using linguistic terms to rate the condition of our variables. For example, different experts might assign different numerical values to the term “least desirable”; one expert might say that this corresponded to a rating of ten percent of ideal while another might assign the same term a rating of five percent. Similarly, variability in expert judgement will lead to different experts rating the same variable differently—where one expert sees the least desirable condition another may see the medium–low desirable condition. In this study, we address this ambiguity by converting linguistic descriptions about the current state of the antecedents into quantitative terms using fuzzy numbers [25,26].

When setting up the DN-EVAM model, the user rates the performance of each antecedent variable according to one of six linguistic terms, ranging from least desirable to very highly desirable conditions (

Table 2. Linguistic terms for antecedent variable weighting.

Linguistic Term	Linguistic Acronym	Triangular Fuzzy Number $({\mathit{m}}_{\mathit{L}},{\mathit{m}}_{\mathit{M}},{\mathit{m}}_{\mathit{H}})$
Least Desirable	LD	(0, 0.1, 0.2)
Medium Low Desirable	MLD	(0.2, 0.3, 0.4)
Moderately Desirable	MD	(0.4, 0.45, 0.50)
Medium High Desirable	MHD	(0.5, 0.6, 0.75)
Highly Desirabile	HD	(0.75, 0.8, 0.9)
Very Highly Desirable	VHD	(0.90, 0.95, 1)

). Each linguistic term is associated with a fuzzy number whose membership function is a triangular fuzzy distribution ($m_L,m_M,m_H$), where $m_L,m_M,$ and $m_H$ are the lowest possible, most probable, and highest possible values of the condition of the variable (Figure 4). Table 2 lists the distributions for each of the linguistic terms. These membership functions for each linguistic term were chosen such that the membership functions collectively span the range from zero to one (worst possible to best possible performance). The fuzzy numbers are then defuzzified using the Middle of Maxima method [27]. For our triangular fuzzy number distributions, the resulting crisp number will be equal to $m_H$. This crisp number is then assigned as the performance level of each antecedent variable $a_i$ according to Equation (1) below:

$$a_i=m_{Mi}$$

(1)

where $i$ indexes the antecedent variables, and $m_{Mi}$ is the $m_M$ (most probable) value for the fuzzy number mapped to antecedent $i$.

We apply the same technique to moderating variables (

Table 3. Showing the moderator fuzzy preference scale of impact linguistic terms and their corresponding triangular fuzzy number.

Linguistic Term	Abbreviation	Triangular Fuzzy Number for Impact (Mi)
Very Low Impact	VL	(0, 0.1, 0.2)
Low Impact	L	(0.2, 0.3, 0.4)
Medium–Low Impact	ML	(0.4, 0.45, 0.50)
Moderate Impact	M	(0.5, 0.6, 0.75)
High Impact	H	(0.75, 0.8, 0.9)
Very High Impact	VH	(0.90, 0.95, 1)

), except that for moderators there is also a directional component to the estimation. Therefore, the moderating impact $M$ is given by the piecewise defined function in Equation (2) below:

$$M_{jk}=\left\{\begin{array}{cc}1+m_{Mjk}& for promoting moderators\\ 1-m_{Mjk}& for inhibiting moderators\end{array}\right.$$

(2)

where $j$ indexes the antecedent categories, $k$ indexes the moderating variables, and $m_{Mjk}$ is the $m_M$ (most probable) value of the triangular fuzzy number for antecedent category $j$ and moderator $k$.

2.2. DN-EVAM

DN-EVAM implements a fuzzy multi-criteria decision analysis based on Kumar and Alok’s relationship map of EV adoption (Figure 2). This information is used to generate the adoption coefficient, a score that characterizes favorability of a given scenario for EV adoption. The purpose of DN-EVAM is to identify strategies that automotive policy makers can use to maximize the potential for EV adoption.

DN-EVAM’s composition differs slightly from Kumar and Alok’s relationship map in two ways. First, DN-EVAM does not consider consequence variables. Consequence variables are downstream impacts of EV adoption. Although these variables are an important and interesting aspect of EV adoption, they are out of scope for this work, which is focused on developing a tool to facilitate EV adoption.

Second, DN-EVAM considers sociodemographic variables collectively with moderating variables. Specifically, sociodemographic variables are represented in DN-EVAM as a single moderating variable. This structural layout simplifies implementation while remaining consistent with Kumar and Alok’s broad distinction between antecedent variables, which are generally EV-specific, and moderating and sociodemographic variables, which reflect broader societal conditions.

2.2.1. Antecedent Variables

Antecedent variables are weighted based on their relative importance as determined by the survey and assigned a current performance level based on the user’s expert judgement using the process outlined above. Antecedents are then classified into one of six antecedent variable categories: Technology, Infrastructure, Cost, Incentives, Business Model, and Human Factors (

Table 4. Showing the antecedent variable categories and the sub factors/antecedent variables considered in each category.

Antecedent Category	Component Antecedent Variables
Cost	Total cost of ownership
Incentives	Purchase-based incentives
	Use-based incentives
Technology	Performance measures
	Vehicle design and features
Infrastructure	Electric load distribution and management
	Charging infrastructure development
Business Model	Dealership network availability
Human Factors and Perceptions	Potential environmental benefit
	Symbolic attributes
	Perceived risks
	Awareness
	Range anxiety

).

Technology: antecedent variables directly connected to the performance of EVs themselves, without consideration of external factors like infrastructure, cost, and government policy.

Infrastructure: antecedent variables concerned with the electric grid as well as EV-specific supporting infrastructure, such as Electric Vehicle Supporting Equipment (EVSE) Development, Charging Network Development, and power generation and distribution capacity.

Cost: any costs contributing to the total cost of ownership.

Incentives: this category includes either use-based incentives such as free electric vehicle parking policy, road tax exemption, or highway lane access and purchase-based incentives such as cash-back incentives or EV purchase tax credits.

Business Model: The electric vehicle business model category includes factors such as EV sales and distribution channels; battery exchange, recycling and disposal systems; and EV maintenance and repair services. All of these are consolidated under the heading of dealership network because dealership networks currently serve internal combustion engine vehicle manufacturers in these areas.

Human Factors: All antecedent variables relating to the user. This includes user perceptions of environmental benefits, symbolic attributes (status symbols), risks, and awareness of EV options and characteristics.

2.2.2. Moderating Variables

Moderating variables are variables exogenous to the EV marketplace that either promote or diminish the impact of the antecedent variables. As such, moderating variables are not likely to be directly influenced by EV-specific policies; however, their impact on EV adoption must be considered when analyzing EV adoption. Moderating variables were identified through literature and from the survey respondents and aggregated into four moderator categories: macroeconomics, politics, demographics, and global trends. The factors affecting the impact levels of these moderating variables are discussed below and the moderator impact levels are summarized in

Table 5. Summary of estimated moderator impacts used in this study.

	Moderating Variable
Antecedent Variable Category	Global Trends	Macro-Economics	Demographics	Politics
Costs	1.3	0.7	0.4	0.2
Incentives	1.45	0.2	0.4	0.4
Technology	1.8	1	1	1
Infrastructure	1.45	0.2	0.4	0.4
Business model	1.6	0.55	1	0.6
Human factors	1.6	0.2	0.4	0.4

.

Global trends are those global processes that can impact the local EV adoption landscape. They include technological innovation [28], changing automotive market conditions [3,29], and global climate policies and development goals [30,31].

Macroeconomics include the overall health and characteristics of the economy such as GDP and currency exchange rates. Since the cost of electric vehicles is a significant antecedent variable for developing nations’ adoption of electric vehicles, an increase in the ability to afford more expensive products will likely positively influence electric vehicle adoption [32]. Because the supply of passenger vehicles in developing nations relies heavily on imports, currency and exchange rates strongly influence international EV markets and in turn importation of used electric vehicles at affordable rates. The current economic outlook in Nigeria indicates a declining GDP and soaring exchange rates [15]. These conditions result in an increase in the cost of goods including passenger vehicles. EV incentives have proven to be successful in developed countries; however, poverty remains a challenge in Nigeria and effective purchase-based incentives could be very expensive to implement.

Given the factors discussed above, we hypothesize that the macroeconomic factors may have a high diminishing impact on infrastructure, low diminishing impact on cost, no impact on technology, moderate diminishing impact on incentives, moderate–low diminishing impact on business model, and high promoting impact on human factors.

Demographic variables are those variables concerned with the population. An increase in population also increases the need for passenger transportation and the impact of the scarcity thereof [33]. Moreover, some younger age groups have affinity for trying new technologies, while older age groups seem to be the late adopters [34]. Forty percent of Nigeria’s 11.5 million vehicle fleet are privately owned vehicles belonging to the working age between 21 and 65 years old [35]. Education levels can influence exposure to new technologies [36] as well as higher disposable income [37].

Given all of the aforementioned considerations, we conclude that the current demographics will have a moderate diminishing impact on infrastructure, low impact on cost, no impact on technology, moderate impact on incentives, moderate impact on business model, and high impact on human factors.

The political climate can either enhance development of factors relevant for passenger vehicles electrification or resist the development of such factors, thus limiting electric vehicle adoption. Politics is expected to have a high impact on infrastructure, high impact on cost, moderate impact on technology, high impact on incentives, moderate impact on business model, and medium–low impact on human factors. This is because a government’s disposition towards electric vehicles could either create a suitable environment for the technology adoption to thrive or be stifled. Policies could highly influence the cost of a product through taxes or other regulations. Policies could highly affect the investment in infrastructure either through government or private investment. Policies could also highly impact what incentives are offered to encourage electric vehicle adoption.

On the other hand, although the Nigerian government could fund research and development of electric vehicles, years of catchup would be required to contribute to the existing technology. Therefore, the influence of Nigerian policies on technology innovation are expected to be minimal.

Given all of the aforementioned considerations, current politics will have a moderate diminishing impact on infrastructure, high diminishing impact on cost, no impact on technology, moderate diminishing impact on incentives, moderate diminishing impact on business model, and moderate–low diminishing impact on human factors. However, it is important to note that the impact of political factors can change with each new government, so the magnitude and direction of these impacts can (and probably will) change as successive administrations hold power.

2.2.3. The Antecedent Impact Score and Adoption Coefficient

For each scenario of antecedent and moderator weights and performance, DN-EVAM calculates two types of metrics: an antecedent impact score $ϵ$ for each antecedent category and an aggregate adoption coefficient $\mathsf{\Delta }$ for the entire scenario. The antecedent impact score is a dimensionless score, normalized on a scale from [0, 1], that measures the performance of each antecedent category within a given scenario relative to its best possible performance. Similarly, the adoption coefficient is a dimensionless score, normalized on a [0, 1] scale, that measures the relative favorability of a given scenario for EV adoption, compared to the most favorable EV adoption conditions possible (the best-case scenario). The best-case scenario is discussed and defined in the next section; Equations (4) through (9) and Figure 5 formally define and explain the antecedent impact score and adoption coefficient.

First, weights $w_i$ are assigned to each antecedent variable $a_{\mathit{i}}$ based on their relative importance as determined by the survey. Antecedents are assigned a point score based on the ratings given by survey respondents, with each rating of “insignificant” receiving 1 point, “not important” 2 points, and so on up to 5 points for a rating of “most important”. Each antecedent’s weight is the ratio of each antecedent’s score to the highest possible score according to Equation (3) below:

$$w_i={\displaystyle \frac{{\sum }_l{r}_l{n}_l}{5{\sum }_l{n}_l}}$$

(3)

where $l$ indexes the possible ratings on the Likert scale, $r_l$ is the point value of a given rating, and $n_l$ is the number of survey respondents that assigned rating $l$ to antecedent $i$.

Next, the weighted antecedent values ${\alpha }_i$ are calculated according to the following formula:

$${\alpha }_i=a_i{w}_i$$

(4)

These weighted antecedent values are then averaged across each antecedent category to determine the weighted categorical antecedent value ${\overline{\alpha }}_j$:

$${\overline{\alpha }}_j={\displaystyle \frac{{\sum }_{i\in j}{\alpha }_i}{p_j}} \forall j$$

(5)

where $p_j$ is the number of antecedent variables in category $j$.

The weighted categorical values are then adjusted for the influence of the moderating variables to generate the moderated categorical antecedent impact ${\mathrm{A}}_j$:

$$A_j={\overline{\alpha }}_j{\sum }_k{M}_{jk} \forall j$$

(6)

where $k$ indexes the moderating variables and $M_{jk}$ is the influence of moderating variable $k$ on antecedent category $j$. These moderated categorical impacts are summed across all categories to calculate the raw adoption coefficient $\delta :$

$$\delta ={\sum }_j{\mathsf{A}}_j$$

(7)

Finally, the raw adoption coefficient is normalized to a scale of 0 to 1:

$$\mathsf{\Delta }=\frac{\delta }{{\delta }_{max}}$$

(8)

where ${\delta }_{max}$ is the raw adoption coefficient that would result from the best-case scenario where all antecedent and moderating variables are at their highest possible performance level.

The antecedent impact score ${ϵ}_j$ is calculated by normalizing each moderated categorical antecedent $A_j$’s performance on a [0, 1] scale relative to its best possible performance:

$${\mathsf{ϵ}}_{\mathrm{j}}=\frac{A_j}{{\mathrm{A}}_{{\mathrm{j}}_{max}}}$$

(9)

2.3. Scenario-Based Analysis of Nigerian EV Adoption

Given the lack of prior research into the EV adoption landscape of Nigeria, our goal in this work is to broadly contextualize the status quo conditions in Nigeria within the space of possible improvements. This first step will provide a broad characterization of the Nigerian EV adoption landscape that will begin to identify and prioritize areas of high potential impact for policy action and/or further study.

We analyze four scenarios: the first is the Status Quo scenario, with antecedents and moderators set to current conditions in Nigeria. We compare the status quo with three hypothetical scenarios that define the extreme limits of possible improvement: the Best-Case scenario, where all antecedents and moderators are at their highest possible performance level; the Perfect Antecedent scenario, where the antecedents are at their highest performance levels while moderators remain at the status quo; and the Perfect Moderator scenario, where moderating variables are at their highest performance and antecedents are set to the status quo (Figure 6). Together, these scenarios define the range of possible improvements that could occur, and, thus, the space of possible outcomes that policy makers should consider.

2.3.1. The Status Quo Scenario

The Status Quo scenario reflects the current situation in Nigeria, based on expert judgement of the current conditions.

Table 6. Antecedent variable performance levels in the Status Quo scenario.

Antecedent Variable	Description	Linguistic Term	Performance Level
Total cost of electric vehicle ownership	Very expensive for masses	LD	0.1
Purchase-based incentives	No incentives	LD	0.1
Use-based incentives	No incentives	LD	0.1
Performance measures	Moderate performance	MD	0.45
Vehicle design and features	Some superior design features	MHD	0.6
Electric load distribution and management	Insufficient	LD	0.1
Charging infrastructure development	Underdeveloped	LD	0.1
Charging technology development	Some fast charging	MLD	0.3
Charging infrastructure resilience	Poor resilience	LD	0.1
Dealership network availability	Limited EV dealership network	LD	0.1
Concern for environmental benefit	Moderate concern	MD	0.45
Symbolic attributes	Reputable	MHD	0.8
Perceived risks	High risks	LD	0.1
Awareness	Very low awareness	LD	0.1
Range anxiety	Slight range anxiety	MLD	0.3

and

Table 7. Moderating variable impacts in the Status Quo scenario.

	Moderating Variable
Antecedent Variable Category	Global Trends	Macro-Economics	Demographics	Politics
Costs	1.3	0.7	0.4	0.2
Incentives	1.45	0.2	0.4	0.4
Technology	1.8	1	1	1
Infrastructure	1.45	0.2	0.4	0.4
Business model	1.6	0.55	1	0.6
Human factors	1.6	0.2	0.4	0.4
Moderator impact on category
If Moderator increases impact on adoption, then use (1 + Mj)
If Moderator diminishes the categorical impact, then use (1 − Mj)
If Moderator has no effect, use 1

below summarize the status quo levels of the antecedents and moderators, respectively. The only EVs currently available are imported, so EV technology in Nigeria is the same as in developed nations. EVs are moderately performing in comparison to internal combustion engine vehicles. Although most internal combustion engine vehicles currently have a superior range to electric vehicles, EVs are at least as reliable and performance measures like top speed, acceleration, and torque are comparable or superior. Currently, Nigeria is not offering any incentives for electric vehicle adoption.

The current infrastructure in Nigeria is insufficient for existing demand to the power grid. All of the sub-factors relating to electric vehicle supporting infrastructure are either non-existent or underdeveloped. Significant investment is needed to install charging infrastructure, establish charging networks, and build sufficient power generation and distribution capacity.

New electric vehicles are unaffordable for most Nigerians. Purchase costs are high compared to Nigeria’s per capita income of $2080 [15]. EV adoption is further hampered by high maintenance costs, limited availability of skilled technicians, charging costs, and non-existent infrastructure for battery disposal or recycling.

Human factors are set to moderate. Moderate concern for environmental impacts and EVs’ perception as status symbols promote positive perceptions of EV users. On the other hand, EVs are perceived as high risk; range anxiety, lack of an established dealership and repair network, and the risk of owning an expensive vehicle in the current security situation all detract from EVs’ attractiveness.

2.3.2. The Perfect Antecedents Scenario

The Perfect Antecedents scenario represents situation where all antecedent variables are at their highest possible performance (

Table 8. The Perfect Antecedents scenario. All antecedent variables are at their highest possible performance levels, and the associated fuzzy numbers are at their highest possible level.

Antecedent Variables	Description	Linguistic Term	Fuzzy Number
Total cost of electric vehicle ownership	Cheapest alternative	VHD	0.99
Purchase-based incentives	Abundant incentives	VHD	0.99
Use-based incentives	Abundant incentives	VHD	0.99
Performance measures	All-around superior performance	VHD	0.99
Vehicle design and features	Superior aesthetic designs and features	VHD	0.99
Electric load distribution and management	Abundant capacity	VHD	0.99
Charging infrastructure development	Highly developed	VHD	0.99
Charging technology development	Fastest charging	VHD	0.99
Charging infrastructure resilience	Excellent resilience	VHD	0.99
Dealership network availability	Effective dealership network	VHD	0.99
Concern for environmental impact (EI)	Assured of net positive EI	VHD	0.99
Symbolic attributes	Highly reputable	VHD	0.99
Perceived risks	Lowest possible risks	VHD	0.99
Awareness	Very high awareness	VHD	0.99
Range anxiety	High range confidence	VHD	0.99

) while factor weights and moderators remain at the status quo scenario levels. Comparing this scenario with the Status Quo scenario provides insights into how the relative importance of antecedent variables may change as their performance improves.

In this scenario, electric vehicles in Nigeria would match or outperform internal combustion engine vehicles on every comparable metric. There are abundant incentives for electric vehicle adoption. EV-supporting infrastructure is fully developed and optimized and the overall cost of EV ownership is comparable to other alternatives. This scenario gives users the highest quality and most affordable EV experience possible.

In addition to the technical factors, all of the human factor conditions are also at the most desirable conditions. This includes a high perception of EV’s net positive environmental impact, very positive symbolic attributes associated with EV ownership, very low perceived risks relating to use of electric vehicles, high awareness about the electric vehicle technology and very high range of confidence.

Finally, there would also be a well-developed and effective business strategy. Strong dealership networks facilitate widespread access to EVs and support for maintenance, financing, and supply chain stability.

2.3.3. Perfect Moderators Scenario

In the perfect moderating variables scenario, all of the moderators are acting to the greatest extent possible to promote electric vehicle adoption, while all antecedent variables remain at status quo levels (

Table 9. Showing the Perfect Moderators Scenario.

	Moderating Variable
Antecedent Variable Category	Global Trends	Macro-Economics	Demographics	Politics
Costs	2	2	2	2
Incentives	2	2	2	2
Technology	2	2	2	2
Infrastructure	2	2	2	2
Business model	2	2	2	2
Human factors	2	2	2	2
Moderator impact on category
If Moderator increases impact on adoption, then use $(1+M_j)$
If Moderator diminishes the categorical impact, then use $(1-M_j)$
If Moderator has no effect, use 1

). This scenario provides a point of comparison that allows policy makers to analyze how factors external to the EV market, such as macroeconomic, demographic, and political drivers, will impact EV adoption potential. Therefore, this scenario provides an important opportunity to assess the relationship between EV-specific policy (e.g., policies directed toward developing antecedent variables) and larger societal policies and trends.

2.3.4. The Best-Case Scenario

The Best-Case scenario combines the Perfect Antecedents and Perfect Moderator scenarios: All antecedent and moderating variables are at their highest performance levels. This policy provides a point of comparison that provides insight into the potential room for improvement between any given scenario and the maximum possible performance.

3. Results

The scenarios described above were evaluated using the DN-EVAM model.

Table 10. Results of the scenario-based analysis.

	Antecedent Category	Best-Case	Status Quo	Perfect Antecedents	Perfect Moderators
Normalized Categorical Impact	Costs	1.00	0.03	0.38	0.10
	Incentives	1.00	0.03	0.30	0.10
	Technology	1.00	0.21	0.59	0.52
	Infrastructure	1.00	0.04	0.25	0.15
	Business Model	1.00	0.05	0.46	0.10
	Behavioral Factors	1.00	0.11	0.32	0.35
	Adoption Coefficient	1.00	0.08	0.39	0.22

and Figure 7 summarize the results. Table 10 shows the antecedent impact scores ${ϵ}_j$ and adoption coefficient $\mathsf{\Delta }$ for each scenario. As discussed above, the Best-Case scenario establishes the upper limit of adoption potential and serves as a benchmark for evaluating other scenarios. Therefore, the antecedent impact scores and adoption coefficient of the Best-Case scenario are all at their best possible values of 1.0. In contrast, the Status Quo scenario shows poor performance across all antecedent categories. The best-performing antecedent category is technology, with an antecedent impact score of 0.21, with most other antecedent categories scoring at or below 0.05. Under the Perfect Moderators scenario, the antecedent impact scores improve only marginally compared to the status quo scenario. In the Perfect Antecedents scenario, technology is again the highest-performing antecedent category, with an antecedent impact score of 0.59. However, the other antecedents perform at much higher levels than in the Status Quo scenario, so the other antecedent impact scores do not lag as far behind.

The adoption coefficients of the scenarios are listed along the bottom row of Table 10. The status quo scenario’s adoption coefficient of 0.08 contrasts sharply with the Best-Case scenario’s adoption coefficient of 1.0. Perfectly performing moderators will improve the adoption coefficient considerably to 0.22, and perfectly performing antecedents will improve the adoption coefficient to 0.39, but it is clear to see that the bulk of possible improvements will require improvements to antecedent and moderating variables simultaneously.

These results are displayed graphically in Figure 7, which shows a radar plot of the antecedent impact scores. Here, the striking contrasts between scenarios are visually apparent. The Best-Case scenario defines the outer edge of the plot; the innermost plot is the Status Quo. As discussed above, the Perfect Antecedents and Perfect Moderators scenarios show improvement over the Status Quo; however, the significant area between the Perfect Antecedents and Best-Case scenarios illustrate how much of the possible performance improvements require coordinated improvements to both antecedent and moderating variable performance.

4. Discussion

The results discussed above illustrate the synergistic relationship between antecedents and moderators: nearly 60% of the possible improvement from the Status Quo to Best-Case scenarios results from synergies between antecedents and moderators. This suggests that the most effective approach to promoting EV adoption will be coordinated policy interventions designed to improve both antecedent and moderator performance. These results also illustrate how antecedent categories rank across different scenarios, which allows policy makers to optimize policy interventions as well as to evaluate the robustness—and therefore riskiness—of policies.

For this discussion we assume that the normalized moderated categorical impact $ϵ$ of each antecedent is a proxy for effort already expended to develop an antecedent variable, with higher values reflecting a more developed antecedent. Under the economic concept of diminishing marginal returns, which suggests that the first units of effort produce the highest impacts, this implies that the lowest-performing antecedents are the most desirable to improve.

In

Table 11. Comparison of the relative priority of antecedent categories across scenarios. The Best-Case scenario is not included because in the Best-Case scenario all antecedent categories are of equal importance.

Priority	Status Quo	Perfect Antecedents	Perfect Moderators
1	Incentives	Infrastructure	Costs
2	Costs	Incentives	Business Model
3	Infrastructure	Behavioral Factors	Incentives
4	Business Model	Costs	Infrastructure
5	Behavioral Factors	Business Model	Behavioral Factors
6	Technology	Technology	Technology

below sorts antecedents from lowest to highest moderated categorical impact for all scenarios except Best-Case (under Best-Case conditions, all normalized impacts are equal to 1.0). Under the assumptions discussed above, this sort order arranges the antecedents in order from highest to lowest priority to address. Several important patterns can be seen. Costs and incentives are frequently high priority, ranking as the first or second antecedent category across multiple scenarios. On the other hand, Technology is the lowest priority category across all scenarios. These patterns suggest that promoting incentives and lower costs are robust policy choices that are likely to have a strong impact across a wide range of possible scenarios while investing in technology development is the lowest priority choice because better options will always be available.

Table 11 offers insights into how policy priorities will be impacted as antecedent performance improves. If antecedent performance is perfect while moderator performance was held constant (the Perfect Antecedents scenario), the infrastructure and behavioral factors categories both gain importance relative to their priority in the Status Quo scenario and incentives and costs decrease in importance. Similarly, comparing the Status Quo scenario to the Perfect Moderators scenario shows that the incentives and infrastructure categories decrease in priority as moderator performance improves, while the business model and cost categories increase in priority.

Table 12. Change in the priority rank of antecedent categories under the Perfect Antecedents and Perfect Moderator scenarios.

Antecedent Category	Perfect Antecedents	Perfect Moderators
Incentives	Decreasing	Decreasing
Costs	Decreasing	Increasing
Infrastructure	Increasing	Decreasing
Business Model	Decreasing	Increasing
Behavioral Factors	Increasing	Neutral
Technology	Neutral	Neutral

summarizes the impact of improving antecedent and moderator performance. For each scenario, each antecedent category is listed as “increasing” if its priority rank increases when compared to its rank in the Status Quo scenario, “decreasing” if its priority decreases, and “neutral” if its priority remains the same. Incentives have a lower priority under both scenarios, suggesting that phasing out incentives is likely to be a robust policy choice. Similarly, Technology is always at the lowest priority, suggesting that other areas should be prioritized. Behavioral Factors increase in priority under Perfect Antecedents but are neutral under Perfect Moderators; this suggests that improving public perceptions of EVs and EV users is a robust policy choice.

The other antecedent categories have mixed results, increasing in priority under Perfect Antecedents and decreasing under Perfect Moderators, or vice versa. The policy implications for these antecedent categories are less clear, which implies that more work is needed to better understand how these antecedents behave under changing antecedent and moderating variable conditions.

5. Conclusions and Future Work

This work developed the DN-EVAM framework for analyzing EV adoption in developing nations. We applied this model to the Nigerian EV adoption environment through a scenario-based analysis and identified several important insights that can guide policy makers and researchers. Firstly, neither the antecedent variables that directly influence users’ perceptions of EVs nor the moderating variables that affect Nigerian society more broadly can have large impacts on EV adoption potential by themselves. Broad-based EV adoption will require joint effort to improve performance across both variable classifications.

Secondly, there are clear signals about the importance of several antecedent variable categories. Incentives are likely to decrease in importance as either antecedent or moderating variables improve in performance. Policies targeting behavioral factors such as public perceptions of EVs are likely to become a higher priority as antecedent performance improves and are not affected by improving moderators. Technology has the lowest priority under all scenarios examined. Taken together, these insights suggest that in Nigeria, promoting a positive public image of EVs is a robust policy choice, that incentives should be temporary, and that investment in technology development is not likely to be an effective means of promoting EV adoption.

The message for the other antecedent categories is less clear. The Cost, Infrastructure, and Business Model categories show mixed performance, depending on whether antecedent or moderating variables are improving. Therefore, more work is needed to understand the behavior of these antecedent categories under changing conditions.

Finally, nations must work together. While the Nigerian government can structure policy to promote improvement in antecedent variables, many of the moderating variables such as macroeconomics and global trends have significant international components, meaning that the international community must coordinate efforts to improve the moderating variables.

DN-EVAM is an important first step toward effective EV policy analysis in developing nations contexts. However, this work is just the beginning of an iterative process and has limitations that should be addressed through future work. First, although the relevant antecedents were identified by a survey of SMEs, estimating the performance level of these antecedents is left to the user. Although this represents a typical use case of DN-EVAM, there is a risk of inaccurate antecedent performance levels being used in the model. Second, the constraints on the length of the survey limited the number of prospective antecedents presented to respondents. While the survey asked open-ended questions about what factors survey respondents thought were important, which allowed respondents to note any significant omissions from our list, there is a risk that relevant antecedent variables were omitted from this study. Finally, the DN-EVAM model evaluates conditions at a static point in time. While useful, this static information may have a short shelf life, and policy decisions based on such static results may be suboptimal under dynamic real-world conditions.

Future work could address these limitations by developing a more robust methodology for setting antecedent performance levels, perhaps by using a formal expert elicitation or developing data-based metrics. A similar approach could be useful for identifying new antecedent and moderating variables. Finally, future development of the DN-EVAM model could add the capability to project results into the future, which would allow users to evaluate how policies may perform over time.

Another important area of future work is to identify possible future areas of study. This work has shown that coordinated development of both antecedents and moderators is essential for promoting EV adoption. The next step is to identify specific pathways that most effectively achieve this goal. This could be accomplished by a more broad-based scenario analysis, such as, perhaps, a Monte-Carlo analysis that samples many possible levels for antecedents and moderators. This analysis would add more detailed contours to the EV adoption landscape we have begun to characterize here. It would identify specific combinations of antecedents and moderators that are of particular importance, as well as identifying possible interactions between moderators that have synergistic or antagonistic effects. The results of this analysis would provide policy makers with important information about how to craft effective and efficient EV policy.