Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Ibrahem Shatnawi, Juan Nicolas Gonzalez, and Jeyhun Mikayilov

October 2025 | Doi: 10.30573/KS--2025-DP48

Driving the Future

How Connected Autonomous Vehicles Reshape Travel and Energy Demand

in Urban and Rural Areas – An Empirical Modeling Approach

Discussion Paper

About KAPSARC

KAPSARC is an advisory think tank within global energy

economics and sustainability providing advisory services to

entities and authorities in the Saudi energy sector to advance

Saudi Arabia’s energy sector and inform global policies through

evidence-based advice and applied research.

This publication is also available in Arabic.

or materials contained therein) (the “Document”) shall not be used without the proper attribution to KAPSARC. The Document shall

not be reproduced, in whole or in part, without the written permission of KAPSARC. KAPSARC makes no warranty, representation or

undertaking whether expressed or implied, nor does it assume any legal liability, whether direct or indirect, or responsibility for the

accuracy, completeness, or usefulness of any information that is contained in the Document. Nothing in the Document constitutes

or shall be implied to constitute advice, recommendation or opinion. The views and opinions expressed in this publication are those

of the authors and do not necessarily reﬂect the ofﬁcial views or position of KAPSARC.

Legal Notice

Abstract

The rapid advancement of connected autonomous vehicles (CAVs) is reshaping

transportation by signiﬁcantly impacting travel behavior, fuel efﬁciency, and mobility

patterns in both urban and rural settings. This study develops a nonlinear regression

framework to analyze how improvements in fuel efﬁciency and reductions in time costs –

two key features of CAVs – affect vehicle miles traveled (VMT), thereby contributing to the

energy rebound effect. While prior research has provided a national-level analysis of these

factors, this study extends the scope by differentiating between urban and rural areas and

incorporating nonlinear relationships to capture behavioral heterogeneity. Using the 2022

National Household Travel Survey in the United States, this study estimates the elasticity

of VMT concerning changes in fuel and time costs.

The ﬁndings highlight substantial quantitative impacts within

the broader context of CAV adoption. Speciﬁcally, the elasticity

of travel time cost is estimated at -0.565, while fuel cost

elasticity is -0.337. Across all CAV scenarios, the study identiﬁes

a consistent and positive rebound effect: improvements in fuel

efﬁciency (ranging from 0% to 40%) and reductions in time

costs (up to 45%) result in increased VMT, potentially offsetting

anticipated energy savings. This effect is notably more

pronounced in rural areas, where longer travel distances and

greater reliance on private vehicles amplify the response.

The rebound effect also varies signiﬁcantly by income level,

with time cost elasticity values ranging from -0.51 to -1.84.

Higher-income households exhibit the strongest rebound

effects, reﬂecting their greater sensitivity to time cost

reductions. These ﬁndings underscore a critical challenge for

Keywords: Connected Autonomous Vehicles (CAVs), Fuel Cost Elasticity VMT, Energy Rebound, Income Level Travel Behavior, Time

Cost Elasticity VMT, Rural Travel Demand, Urban Travel Demand

energy efﬁciency and emissions reduction policies: increased

vehicle miles traveled (VMT) resulting from the rebound

effect varies signiﬁcantly by income level, with time cost

elasticity estimates ranging from -0.51 to -1.84. Higher-income

households exhibit the strongest rebound effects, reﬂecting

their greater sensitivity to time cost re ductions. This pattern

may substantially diminish the anticipated beneﬁts of

such policies, particularly in rural contexts. The study also

highlights differences in travel behavior across income groups,

showing that higher-income households are more responsive

to time cost savings.

This research reﬁnes the understanding of CAV-induced

energy rebound effects and provides actionable insights for

policymakers and urban planners seeking to balance CAV

adoption with sustainable mobility goals.

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Introduction

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Central to understanding these impacts is the interplay between

VMT, fuel efﬁciency, and time costs. These factors are key

drivers of the energy rebound effect, where improvements in

fuel efﬁciency and reductions in travel time costs may induce

additional travel, potentially offsetting anticipated energy

savings. While the pioneering work of Taiebat, Stolper, and

Xu (2018) provided a robust framework for estimating travel

demand elasticity and forecasting induced travel behavior, their

analysis focused primarily on aggregate national trends. It did

not sufﬁciently differentiate between urban and rural contexts

or explore the potential nonlinear relationships in travel

demand elasticities.

Building on this foundational research, our study extends the

analysis in two critical directions. First, we explicitly disag-

gregate the energy rebound effects between urban and rural

settings, recognizing the distinct differences in travel behavior,

infrastructure availability, and accessibility across these regions.

By addressing these spatial distinctions, we aim to offer a more

nuanced understanding of how CAV adoption impacts energy

consumption, providing targeted insights for policymakers and

urban planners. Second, we address the limitations of linear

assumptions in elasticity modeling by employing a variety

of nonlinear modeling techniques to verify the relationships

between travel demand and its determinants. This approach

allows us to assess the heterogeneity in behavioral responses

to changes in travel costs, particularly across diverse socio-

economic and demographic groups.

To achieve these objectives, our study leverages the most recent

and comprehensive data available – the National Household

Travel Survey (NHTS) 2022 (Federal Highway Administration

2022). This updated dataset reﬂects the latest trends in travel

behavior, technological adoption, and energy use, allowing us

The rapid evolution of connected autonomous vehicles (CAVs) is ushering in a transformative

era in transportation, with profound implications for travel behavior, fuel efﬁciency, and

mobility dynamics in both urban and rural areas. These advancements promise unparalleled

convenience, safety, and accessibility. However, they also present complex challenges that

warrant careful examination, particularly regarding their long-term impact on energy

consumption, environmental sustainability, and societal mobility patterns.

to produce forecasts that are more representative of the current

transportation landscape. Through these contributions, our

work provides a more reﬁned and spatially sensitive simulation

of CAV-induced energy rebound effects, enhancing our

understanding of sustainable mobility pathways and offering

actionable insights to mitigate the challenges posed by this

transformative technology.

While this study relies on U.S.-based data, its relevance

to the Saudi Arabian context is supported by empirical

evidence demonstrating structural and behavioral similarities

in mobility patterns and vehicle usage characteristics

between the two countries.

First, both the U.S. and Saudi Arabia exhibit high car

dependency, low public transit ridership, and dispersed urban

development, particularly outside major city centers. These

commonalities underpin comparable household travel behaviors

and vehicle usage patterns (Aljouﬁe et al. 2013; Litman 2025b).

Second, the U.S. vehicle market has historically served as a

benchmark for Saudi Arabia’s ﬂeet policy development. Notably,

Saudi Arabia adopted the U.S. 2012 Corporate Average Fuel

Economy (CAFE) standards in 2016, reﬂecting a four-year lag

in regulatory alignment (Sheldon and Dua 2018). This decision

was inﬂuenced by the fact that, by 2016, the composition and

technical proﬁle of Saudi Arabia’s vehicle ﬂeet closely mirrored

that of the U.S. market as it stood in 2012.

Such alignment underscores the structural similarities

betweenthe two markets and strengthens the rationale for

leveraging U.S. data and behavioral trends as a forward-

looking proxy to guide transport and energy policy

development in Saudi Arabia.

Introduction

October 2025

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

The paper is structured as follows: Section 2 presents a compre-

hensive review of the existing literature on VMT dynamics, CAV

implications, and associated energy and travel rebounds. It

identiﬁes key drivers of VMT, including fuel and time costs, and

examines how CAVs may alter these relationships. Section 3

outlines the methodology employed in this study, detailing the

data sources, econometric models, and analytical techniques

used to estimate the impact of CAVs on VMT across different

income groups and geographical contexts. Section 4 provides

a detailed discussion of the empirical results, comparing the

rebound effects of fuel economy and travel time cost reductions

in urban and rural areas. Finally, Section 5 concludes with policy

implications, highlighting the challenges and opportunities

of CAV adoption in managing travel demand and promoting

sustainable mobility.

By examining these dynamics, this paper contributes to a

deeper understanding of the opportunities and challenges

posed by CAVs. While the study draws on ﬁndings from the

United States, its insights are broadly applicable to other

regions, including Saudi Arabia, with a focus on tailoring energy

and transportation policies to the unique characteristics of

urban and rural areas.

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Literature Review

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Literature Review

The literature review comprises three subsections: (i) determinants of vehicle miles

traveled, which examines key drivers of VMT, such as travel time, fuel costs, and

socioeconomic factors; (ii) inﬂuence of connected autonomous vehicles on VMT

determinants, focusing on how CAVs reshape VMT patterns and mobility; and (iii) energy

and environmental rebound effects of CAV-induced travel demand, which explores their

energy efﬁciency, eco-routing, and the potential rebounds from increased travel demand.

2.1 Determinants of Vehicle

Miles Traveled

VMT is a fundamental metric in transportation studies,

reﬂecting the total distance covered by vehicles within a given

timeframe. Understanding the factors that inﬂuence VMT offers

valuable insights into travel behavior, energy consumption,

and broader economic and environmental impacts. This review

examines key drivers of VMT, including travel time costs, fuel

prices, urban-rural differences, and income levels. It also

explores the transformative potential of CAVs in redeﬁning VMT

patterns, enhancing fuel efﬁciency, reducing time costs, and

reshaping mobility.

The complex interplay between fuel costs, travel time, and

VMT is a central focus for researchers, policymakers, and

urban planners. Globally, city planners strive to reduce VMT

without disrupting business activities or household mobility.

Achieving this balance requires a comprehensive understanding

of the factors inﬂuencing VMT, along with the development of

effective transportation infrastructure and policy strategies.

VMT is inﬂuenced by a range of interrelated factors, with travel

time emerging as a key determinant.

Research indicates that reductions in travel time, enabled by

infrastructure improvements or technological advancements

often lead to increased VMT. This phenomenon, known as

induced demand, occurs as enhanced accessibility encourages

more travel (Cervero 2002; Duranton and Turner 2011).

Additionally, changes in travel time affect the elasticity of travel

demand, with short-term impacts differing from long-term

trends. Reduced time costs can also prompt shifts from public

transportation to personal vehicle use, further increasing

VMT (Litman 2025a).

These dynamics have important implications: while CAVs may

offer individual beneﬁts such as improved convenience and

productivity, they risk triggering systemic increases in energy

use, emissions, and congestion – potentially offsetting many of

the environmental gains associated with their adoption.

Fuel prices play an equally signiﬁcant role in shaping VMT by

directly inﬂuencing travel costs. When fuel prices rise, VMT

generally decreases, with greater elasticity observed in the

short term as individuals quickly adjust their travel behavior

(Hughes, Knittel, and Sperling 2008). Over time, policy inter-

ventions such as fuel taxes and subsidies for alternative

energy vehicles can further inﬂuence VMT by encouraging

shifts toward sustainable travel practices (Sterner 2007).

Prolonged periods of high fuel prices also drive behavioral

adaptations, including increased carpooling, fewer discretionary

trips, and greater adoption of fuel-efﬁcient vehicles (Cao,

Wongmonta, and Choo 2013).

The dynamics of VMT differ signiﬁcantly between urban and

rural areas due to variations in population density, infra-

structure, and accessibility. Metropolitan areas, characterized

by shorter trip lengths and higher densities, typically show

lower VMT per capita compared to rural regions (Ngo et al.

2024). In contrast, rural areas often lack robust public

transportation systems, resulting in greater reliance on

personal vehicles. Additionally, socioeconomic factors such

as income levels and employment opportunities inﬂuence

travel behavior, with rural populations generally displaying

distinct VMT patterns compared to their urban counterparts

(Aderibigbe and Gumbo 2022).

Discussion Paper

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Income levels also play a role in travel behavior and VMT.

Higher incomes facilitate greater vehicle ownership and support

discretionary travel, resulting in increased VMT (Federal

Highway Administration 2018). In contrast, economic downturns

or recessions reduce disposable incomes, often resulting in

decreased travel activity. For instance, during the 2007-2009

recession in the United States, there was a notable decline in

auto output’s share of gross domestic product (GDP), reﬂecting

reduced travel activity amid economic hardships (Samaddar and

Bachman 2022). These shifts underscore the sensitivity of VMT

to economic conditions and the affordability of travel.

As these traditional factors continue to inﬂuence VMT dynamics,

emerging technologies are poised to redeﬁne the transportation

landscape. Among these, CAV technology stands out as a trans-

formative force, ushering in a new era of mobility. By integrating

advanced communication systems and autonomous capabilities,

CAVs promise to enhance mobility, safety, and convenience.

However, they also introduce uncertainties, particularly

regarding their potential impacts on energy demand, travel

patterns, and environmental outcomes.

2.2 Inﬂuence of Connected

Autonomous Vehicles on VMT

Determinants

The implications of CAV adoption on VMT, travel behavior, and

urban planning warrant closer examination as we consider the

future of transportation systems. CAVs are expected to signif-

icantly inﬂuence VMT trends due to their convenience, acces-

sibility, and operational ﬂexibility. One of the primary drivers of

increased VMT is the reduction in travel effort and time costs.

With CAVs taking over the driving task, passengers can travel

more comfortably, which may lead to an increase in discre-

tionary travel (Milakis et al. 2017). Additionally, CAVs provide

mobility opportunities for previously underserved populations,

including the elderly and disabled, further contributing to

higher VMT (Martínez-Buelvas et al. 2022).

Another critical factor is the emergence of empty vehicle

travel. Unlike traditional vehicles, CAVs may undertake trips

without passengers, such as returning home after dropping

someone off or repositioning themselves to a different

location. This phenomenon poses unique challenges to VMT

management, as these empty trips add to overall road usage

and congestion (Zhang and Guhathakurta 2017). To mitigate

these effects, policymakers should consider implementing

strategies such as dynamic pricing or operational restrictions

on empty (unoccupied) vehicle movements (Fagnant

and Kockelman 2015).

CAVs hold the promise of improving fuel efﬁciency through

optimized driving patterns. By leveraging advanced algorithms

and communication technologies, these vehicles can achieve

smoother acceleration and deceleration, reduce idling times,

and select energy-efﬁcient routes (Taiebat, Stolper, and Xu

2018). Such improvements could contribute to substantial

energy savings and reduced emissions. However, these

beneﬁts must be weighed against potential energy trade-offs.

The anticipated rise in VMT due to CAV adoption could offset

gains in fuel efﬁciency, leading to increased overall energy

consumption. This necessitates the implementation of robust

energy policies, such as integrating renewable energy sources

for electric CAVs or encouraging energy-efﬁcient vehicle

designs (Taiebat et al. 2019). Balancing these trade-offs will

be essential to ensuring that the environmental beneﬁts of

CAVs are fully realized.

Beyond energy considerations, the shift in driving responsi-

bilities from individuals to CAV systems is likely to transform

travel behavior and perceptions of time costs. Freed from the

need to actively drive, passengers can repurpose travel time

for work, leisure, or other productive activities. This reduction

in the perceived cost of travel could make longer trips more

acceptable and desirable to travelers (Cohen and Hopkins 2018;

Yueshuai He, Jiang, and Ma 2022).

This behavioral shift, along with the expansion of Trans-

portation Network Companies (TNCs), such as Uber and Lyft,

is expected to impact VMT with the anticipated adoption of

autonomous vehicles (Martin, Shaheen, and Wolfe 2024;

Choi, Guhathakurta, and Pande 2022). By offering convenient,

affordable, and accessible travel options, TNCs reduce the

effort and cost associated with transportation, making

travel more appealing and less burdensome. This increased

convenience could lower the perceived value of travel time,

potentially leading to a rise in VMT (Clewlow and Mishra

2017). As CAVs become more integrated into transportation

systems in the coming decades, they are expected to play a

pivotal yet disruptive role in reshaping transportation markets

and inﬂuencing traveler behavior (Cohen and Hopkins 2018;

Garikapati and Shetiya 2024).

October 2025

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

2.3 Energy and Environmental

Rebound Effects of CAV-

Induced Travel Demand

One approach to evaluating the travel and energy implications

of CAVs is to analyze the demand response to changes in energy

efﬁciency and travel costs introduced by CAV technologies.

The energy and travel implications of CAVs are primarily

inﬂuenced by two key factors, as highlighted in the

literature: travel demand elasticities (such as travel costs,

income levels, and travel mode) and variations in cost

structures (including travel time and fuel costs) (Fagnant

and Kockelman 2015; Taiebat, Stolper, and Xu 2018; Wadud,

MacKenzie, and Leiby 2016).

Current studies report a range of estimates for the energy

efﬁciency improvements that CAVs could achieve compared to

traditional vehicles, driven by technological enhancements such

as optimized driving, eco-routing, congestion reduction, and

electriﬁcation. While these estimates tend to be optimistic, they

vary depending on the assumptions and scope of each study.

The magnitude of potential energy savings, however, remains

uncertain and is heavily dependent on the pace of technological

adoption and behavioral adjustments. Table 1 presents key

ﬁndings from the literature.

Table 1. Energy-efﬁciency enhancements enabled by CAVs.

Aspect Description Efﬁciency improvement Source

Eco-routing CAVs equipped with real-time trafﬁc data

and machine learning algorithms can

select routes that minimize fuel consump-

tion and emissions. Eco-routing alone

could achieve signiﬁcant fuel savings

depending on trafﬁc conditions and route

alternatives.

4% to 20% Guo et al. (2022)

Impact of platooning Automated platooning reduces aerody-

namic drag by allowing vehicles to travel in

tightly spaced groups.

4% to 5% (lead vehicle); 10% to 14%

(following vehicles)

Pi et al. (2023)

Fuel savings through

congestion reduction

Improved trafﬁc ﬂow via vehicle-to-

vehicle (V2V) and vehicle-to-infrastructure

(V2I) communication reduces idling and

stop-and-go driving in congested areas,

contributing to additional energy savings.

V2I connectivity in urban driving:

• Improved fuel economy by 6% to 13%

forbattery electric vehicles (BEVs).

• Improved fuel economy by 9% to 15%

forfuel-cell electric vehicles (FCEVs).

V2V connectivity in highway driving:

• Enhanced fuel economy by 5% to 32%

for BEVs.

• Enhanced fuel economy by 5% to 26%

for FCEVs.

Fong, Lane, and Samuelsen

(2024)

Improvement in energy

efﬁciency

A commonly cited range for energy

efﬁciency improvements over traditional

vehicles. Savings are attributed to smooth-

er driving patterns (e.g., reduced hard

braking and acceleration), route optimiza-

tion, and maintaining constant speeds in

free-ﬂowing trafﬁc.

5% to 15% as per the documented road

experiments.

Fagnant and Kockelman

(2015), Vahidi and Sciarretta

(2018), and Faghihian and

Sargolzaei (2023)

Discussion Paper

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

While energy efﬁciency improvements remain a core focus of

CAVs, their potential beneﬁts extend beyond just fuel savings.

One signiﬁcant advantage is the reduction in travel time costs

(TTC), which varies between 25% and 45%, as reported in

the literature (International Transport Forum 2018; Victoria

Transport Policy Institute 2023). This variation depends on

factors such as the purpose of the trip, with work-related or

leisure trips beneﬁting more signiﬁcantly due to the ability

to utilize travel time for productivity or relaxation. However,

while these reductions are notable, researchers have reported

60% as the theoretical upper limit for achievable TTC savings

(Fagnant and Kockelman 2015; Taiebat, Stolper, and Xu 2018).

This ceiling is primarily attributed to the residual in-vehicle

attention demands that persist, even with highly advanced

levels of automation.

While CAVs have the potential to signiﬁcantly reduce the

cognitive demands of travel – by allowing occupants to engage

in work or leisure activities – this beneﬁt is not absolute.

Occasional driver intervention remains necessary, particularly

under Level 3 and Level 4 automation1 (Shatnawi, Gonzalez,

and Masoud 2025), limiting the full elimination of mental

workload. Additionally, drivers may still be required to monitor

vehicle operations and respond to unexpected system failures,

maintaining a baseline of cognitive engagement. Key ﬁndings

from the literature highlight the following constraints:

• Partial automation still requires human oversight, especially

in transitional scenarios.

• Residual cognitive demand persists due to the need to

monitor driving tasks and intervene during failures.

• Maximum travel time cost (TTC) reductions are bounded,

with 60% often cited as a theoretical upper limit (Fagnant

and Kockelman 2015; Taiebat et al. 2019).

• Perceived beneﬁts vary by use case, with higher impacts on

work and leisure-related trips.

These constraints underscore the need for realistic

expectations about the beneﬁts of CAVs. Our study builds on

this understanding by empirically quantifying how reductions

in time and fuel costs inﬂuence travel behavior across urban

and rural areas and across income levels. This contributes to a

deeper understanding of the behavioral rebound effects

associated with CAV adoption and informs future transportation

and energy policy.

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Methodology

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Methodology

3.1 Data Processing and Descriptive Statistics

To ensure the robustness of our analysis, we processed detailed microdata from the latest

National Household Travel Survey (NHTS) dataset, published in 2022 (Federal Highway

Administration 2022), which includes over 100,000 trip records from U.S. households.

We extracted information on trip purpose, travel time, distance,

household composition, number of vehicles, driver counts,

income categories, and urban versus rural classiﬁcation. This

core dataset was augmented with vehicle-speciﬁc fuel economy

data (miles per gallon, MPG) from the EPA’s Fuel Economy

database (DOE 2024), matching it with NHTS records based on

vehicle make and model (Federal Highway Administration 2009).

This integration allowed for the calculation of the average MPG

for each household’s vehicles, including those with multiple

cars. Using the 2022 NHTS trip data, we also derived the

average VMT per household.

To enrich the analysis, we incorporated gasoline prices and

travel time data from the 2022 NHTS survey. We also included

contextual household information, such as household size,

the number of trips per household, counts of adults, drivers,

and vehicles, as well as area characteristics (e.g., urban vs.

non-urban classiﬁcation). For a more nuanced understanding,

we categorized households into six income groups based

on reported income ranges in the 2022 NHTS survey, which

spanned from less than $25,000 to more than $200,000

annually. These income groups provided a critical dimension for

both descriptive and inferential analyses, enabling meaningful

comparisons across socioeconomic strata.

The ﬁnal cleaned dataset comprises 5,115 households (Table

2), including 4,093 urban (Table 5) and 1,022 rural (Table 6)

households. Sample sizes for each income group are reported

in Tables 7-9. Data cleaning followed three steps: (i) restricting

the sample to motorized trips (excluding non-motorized trips

such as walking and cycling); (ii) removing households with

zero reported VMT (VMT ≤ 0); and (iii) excluding records with

undeﬁned fuel cost or efﬁciency values (e.g., missing values).

Income groups were derived from the 11 NHTS household

income levels and consolidated into six categories: 1st income

group (levels 1-3: <$24,999); 2nd income group (levels 4-5:

$25,000-$49,999); 3rd income group (level 6: $50,000-

$74,999); 4th income group (levels 7-8: $75,000-$124,999);

5th income group (level 9: $125,000-$149,999); and 6th income

group (levels 10-11: ≥$150,000). For continuous approxi-

mations, midpoints were assigned to each income level (e.g.,

$12,500 for level 2, $20,000 for level 3), and an hourly wage

proxy was derived by dividing annual midpoint income by

2,080 working hours.

Following Taiebat et al. (2019) the dependent variable in all

our models is the VMT per household. Our analysis focuses

on three primary predictors: fuel costs, time costs, and

their combined value.

To calculate fuel costs, we used Equation (1):

Pf 5

i 5 1

∑n VMTj

∑

j = 1

nVMTj

MPGj

(1)

Where, Ø represents the price of gasoline (dollars per gallon),

VMT

is the vehicle miles traveled for the j-th vehicle in a

household, and

MPG

is the fuel economy (miles per gallon)

of the j-th vehicle.

To compute time costs, we used Equation (2):

Pt 5γWw 1

∑n VMTj

i 5 1

2γNWw

3 ∑Tvmt (2)

Where, γ

and γ

represent wage rates for work-related and

non-work-related travel, respectively; w is the average hourly

wage;

vmt is the total travel time associated with VMT, and

i 5 1

∑n

VMTj is the total VMT across all vehicles in a household.

This equation calculates the per-mile time cost by incorporating

both work-related and non-work-related travel time, weighted

by their respective wage rates.

October 2025

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

In our study, w represents the imputed hourly wage for

each household in the dataset, while γw and γ

denote the

proportions of work-related and non-work-related trips within

a household, respectively. Additionally, we deﬁne

vmt as the

aggregate cost of fuel and time, calculated as the sum of fuel

costs Pf and time costs Pt:

vmt 5 Pf + Pt.

The analysis includes both descriptive and inferential

components, applied to the entire dataset and stratiﬁed across

the six identiﬁed income groups. This stratiﬁcation enables

a more nuanced examination of the relationships between

variables within different income brackets.

Table 2 provides a comprehensive overview of transportation

behaviors, time valuation, and household characteristics across

six income groups in the United States, along with the national

average. Several key patterns emerge:

• Time allocation differs by income: The lowest income

group (1st) has the highest share of leisure trips (89.3%)

and the lowest share of work-related trips (10.7%).

In comparison, the 3rd income group shows the highest

work-related trip share (17.2%), indicating differing patterns

of labor force engagement and time use.

• Fuel cost and efﬁciency improve with income: Fuel cost

per city mile decreases from $0.162 in the 1st group to $0.140

in the 6th group, while city fuel economy improves from

25.4 MPG to 30.0 MPG. This suggests that higher-income

households have access to more efﬁcient vehicles.

• Time costs rise steeply with income: The average time cost

per mile increases more than tenfold, from $0.31 in the lowest

income group to $3.42 in the highest group, emphasizing the

higher opportunity cost of time for afﬂuent households.

• Higher income equals greater mobility: Annual VMT

increases from 7,782 miles in the 1st group to nearly 20,000

miles in the 6th group. Similarly, the number of trips rises

from 4.2 to 6.5 per day, and vehicle ownership climbs from

1.5 to 2.5 vehicles per household.

• Larger and better-resourced households: Household size

increases from 2.35 to about 2.94 people, and the number

of drivers and adults per household grows with income,

highlighting broader resource availability.

The same analysis was conducted to compare urban and

rural households across income groups (see Table3),

revealing clear differences in travel behavior and

household characteristics.

• Fuel efﬁciency is higher in urban settings: In most income

groups, urban households exhibit slightly better fuel

economy. For instance, in the 6th income group, the average

city MPG is 30.2 in urban areas vs. 29.1 in rural areas, and

highway MPG is 42.2 vs. 40.8, respectively.

• Time costs per mile are consistently higher in urban

areas: For the highest income group, urban households face

an average time cost of $3.64 per mile, compared to $2.26

in rural areas, reﬂecting higher opportunity costs of time in

cities.

• Rural households drive more: Annual VMT in the 6th

income group is 24,375 miles in rural areas vs. 19,134 miles

in urban ones. This trend is consistent across all income

brackets, underscoring rural reliance on personal vehicles.

• Larger and more vehicle-rich rural households: Rural

households tend to be larger and have higher vehicle

ownership, particularly in second to ﬁfth income groups. For

example, in the 5th income group, rural households average

3.13 people compared to 2.99 in urban areas and own more

vehicles (2.96 versus 2.44).

• Trip-making varies by setting and income: Urban

households generally take more trips than rural households,

especially at the lowest (4.41 versus 3.60) and highest (6.63

versus 5.73) income groups. However, rural households

surpass urban ones in the middle-income categories, the 3rd

(5.07 versus 4.60) and 4th (6.0 versus 5.7) groups.

• Fuel costs show moderate differences: Fuel cost per city

mile in the 1st income group is slightly higher in urban

households ($0.164) than in rural ones ($0.157), with the gap

narrowing in higher-income groups.

These quantiﬁed comparisons conﬁrm that urban households

prioritize efﬁciency and time. In contrast, rural households

prioritize access and coverage, leading to distinct energy and

travel demand proﬁles across geographies and income groups.

Discussion Paper

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Table 2. Descriptive statistics – general (mean and standard deviation).

Attribute U.S.

average

1st income

group

2nd income

group

3rd income

group

4th income

group

5th income

group

6th income

group

Work share

0.154 0.107 0.151 0.172 0.162 0.158 0.147

(0.004) (0.016) (0.011) (0.011) (0.008) (0.009) (0.012)

Leisure share

0.846 0.893 0.849 0.828 0.838 0.842 0.853

(0.004) (0.016) (0.011) (0.011) (0.008) (0.009) (0.012)

Fuel cost ($/mile) (city)

0.144 0.162 0.146 0.144 0.141 0.143 0.140

(0.001) (0.003) (0.002) (0.002) (0.002) (0.002) (0.003)

Fuel cost ($/mile) 0.100 0.110 0.101 0.100 0.097 0.100 0.098

(highway) (0.001) (0.002) (0.001) (0.001) (0.001) (0.001) (0.002)

Average time cost ($/mile)

1.522 0.314 0.678 1.072 1.455 2.194 3.415

(0.033) (0.023) (0.025) (0.112) (0.041) (0.056) (0.142)

Annual VMT (miles)

15280 7782 10843 14870 16025 19177 19968

(483) (682) (916) (1232) (770) (1596) (1525)

Annual driving time 494.258 313.936 387.005 453.233 515.560 602.178 621.777

(hours) (10.237) (19.813) (20.897) (23.497) (17.398) (29.937) (36.765)

Fuel economy (MPG) 28.678 25.397 27.944 28.969 28.929 29.403 30.043

(city) (0.121) (0.409) (0.308) (0.321) (0.203) (0.268) (0.380)

Fuel economy (MPG) 40.411 36.760 39.596 40.611 40.759 41.204 41.948

(highway) (0.121) (0.444) (0.310) (0.315) (0.200) (0.271) (0.351)

Gas price 4.076 4.015 3.970 4.066 4.055 4.146 4.219

($/gallon) (0.012) (0.045) (0.032) (0.033) (0.024) (0.033) (0.041)

Household size 2.642 2.354 2.432 2.443 2.629 3.008 2.937

(people) (0.028) (0.121) (0.073) (0.073) (0.047) (0.063) (0.064)

Number of trips

per household

5.397 4.225 4.421 4.699 5.742 6.362 6.485

(0.079) (0.336) (0.143) (0.164) (0.152) (0.196) (0.251)

Count of adults

2.007 1.675 1.827 1.891 2.016 2.299 2.198

(0.016) (0.059) (0.040) (0.037) (0.024) (0.044) (0.039)

Count of drivers

1.932 1.508 1.658 1.824 1.965 2.265 2.186

(0.015) (0.052) (0.035) (0.039) (0.024) (0.039) (0.037)

Count of vehicles

2.113 1.512 1.713 2.019 2.175 2.513 2.476

(0.021) (0.056) (0.040) (0.058) (0.042) (0.051) (0.053)

Urban areas 0.805 0.777 0.801 0.787 0.785 0.857 0.841

(1=urban; 0=rural) (0.006) (0.026) (0.018) (0.017) (0.013) (0.014) (0.019)

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban and Rural Areas – An Empirical Modeling Approach

October 2025

Table 3. Descriptive statistics – urban and rural (mean and standard deviation).

Attribute

1st income group 2nd income group 3rd income group 4th income group 5th income group 6th income group

Urban Rural Urban Rural Urban Rural Urban Rural Urban Rural Urban Rural

Work share

0.111 0.093 0.153 0.140 0.176 0.157 0.153 0.194 0.151 0.196 0.147 0.147

(0.019) (0.026) (0.012) (0.022) (0.012) (0.019) (0.009) (0.020) (0.010) (0.025) (0.013) (0.031)

Leisure share

0.889 0.907 0.847 0.860 0.824 0.843 0.847 0.806 0.849 0.804 0.853 0.853

(0.019) (0.026) (0.012) (0.022) (0.012) (0.019) (0.009) (0.020) (0.010) (0.025) (0.013) (0.031)

Fuel cost ($/mile)

(city)

0.164 0.157 0.146 0.148 0.145 0.142 0.140 0.142 0.142 0.147 0.140 0.141

(0.004) (0.005) (0.003) (0.004) (0.002) (0.005) (0.002) (0.003) (0.002) (0.005) (0.003) (0.007)

Fuel cost ($/mile)

(highway)

0.111 0.105 0.100 0.102 0.100 0.099 0.098 0.097 0.100 0.101 0.098 0.097

(0.002) (0.003) (0.002) (0.002) (0.001) (0.003) (0.001) (0.002) (0.002) (0.003) (0.002) (0.004)

Average time cost

($/mile)

0.344 0.211 0.704 0.573 1.185 0.653 1.574 1.018 2.304 1.535 3.635 2.255

(0.028) (0.027) (0.025) (0.077) (0.141) (0.029) (0.049) (0.051) (0.062) (0.094) (0.163) (0.132)

Annual VMT

(miles)

7275 9543 9070 17987 12887 22191 14435 21828 18534 23016 19134 24375

(780) (1393) (753) (3366) (1362) (2746) (853) (1708) (1801) (2832) (1642) (3982)

Annual driving

time

(hours)

311.09 323.79 359.75 496.84 419.60 577.38 489.01 612.41 597.59 629.59 622.06 620.27

(23.11) (37.62) (19.58) (68.47) (24.51) (60.87) (19.47) (37.78) (33.21) (64.93) (39.59) (98.01)

Fuel economy

(MPG)

(city)

25.308 25.709 28.082 27.388 28.986 28.907 29.278 27.658 29.745 27.357 30.231 29.051

(0.463) (0.865) (0.334) (0.758) (0.359) (0.714) (0.233) (0.414) (0.294) (0.598) (0.392) (1.164)

Fuel economy

(MPG)

(highway)

36.568 37.428 39.793 38.801 40.693 40.309 41.104 39.501 41.551 39.131 42.173 40.761

(0.494) (0.981) (0.335) (0.778) (0.357) (0.667) (0.229) (0.407) (0.294) (0.676) (0.369) (1.007)

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban and Rural Areas – An Empirical Modeling Approach

Discussion Paper

Attribute

1st income group 2nd income group 3rd income group 4th income group 5th income group 6th income group

Urban Rural Urban Rural Urban Rural Urban Rural Urban Rural Urban Rural

Gas price

($/gallon)

4.058 3.866 3.975 3.948 4.075 4.031 4.074 3.984 4.154 4.095 4.234 4.141

(0.054) (0.069) (0.037) (0.061) (0.037) (0.073) (0.028) (0.049) (0.036) (0.072) (0.045) (0.098)

Household size

(people)

2.387 2.237 2.405 2.541 2.348 2.793 2.575 2.829 2.987 3.134 2.967 2.781

(0.142) (0.221) (0.079) (0.181) (0.081) (0.160) (0.053) (0.098) (0.070) (0.127) (0.071) (0.146)

Number of trips

per household

4.406 3.598 4.443 4.329 4.598 5.072 5.669 6.008 6.444 5.874 6.628 5.728

(0.421) (0.313) (0.153) (0.374) (0.167) (0.453) (0.168) (0.344) (0.215) (0.466) (0.273) (0.624)

Count of adults

1.675 1.674 1.806 1.908 1.840 2.080 1.986 2.123 2.296 2.314 2.218 2.093

(0.070) (0.103) (0.044) (0.092) (0.040) (0.089) (0.029) (0.044) (0.050) (0.061) (0.044) (0.071)

Count of drivers

1.487 1.580 1.624 1.795 1.741 2.130 1.928 2.101 2.249 2.365 2.192 2.155

(0.060) (0.102) (0.039) (0.077) (0.041) (0.090) (0.028) (0.044) (0.044) (0.067) (0.041) (0.086)

Count of vehicles

1.454 1.712 1.664 1.911 1.902 2.452 2.020 2.740 2.439 2.961 2.412 2.812

(0.059) (0.134) (0.044) (0.090) (0.067) (0.103) (0.034) (0.137) (0.056) (0.127) (0.056) (0.155)

Table 3. Continued

October 2025

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

3.2 Methodology: Modeling

VMT Response to Travel Costs

This section presents the empirical modeling framework used to

estimate how household VMT responds to changes in travel-

related costs – speciﬁcally, per-mile fuel costs and time costs.

We build upon the econometric structure proposed by Taiebat

et al. (2019), which models VMT as a function of the time-in-

clusive marginal cost of travel. Our study extends this

framework by applying it across different household income

groups and geographic settings (urban vs. rural), enabling a

disaggregated assessment of energy rebound effects under

varying socioeconomic conditions. All models discussed in this

subsection are estimated using generalized least squares (GLS)

regression to account for heteroskedasticity and to incorporate

the sampling weights provided by the NHTS dataset, ensuring

consistent and unbiased coefﬁcient estimates.

3.2.1 Theoretical Framework and Functional

Form

Following Taiebat et al. (2019) to estimate the impact of the

time-inclusive marginal cost of travel (

vmt) on VMT, we use the

following speciﬁcations:

Ln(VMT ) 5 a0 1 a1 Ln(

vmt ) (3)

Where,

is the per-mile fuel cost of VMT and

is the cost of

time spent on VMT and

vmt 5 Pf 1 Pt (4)

To estimate the individual impacts of fuel and time costs on VMT,

we link VMT directly to these two variables via Equation (3).

Using Equation (4), the following relationship can be written:

Ln(VMT) 5 Ln(Pf 1 Pt ) (5)

Applying Taylor expansion to the right-hand side of Equation

(5) up to second-order derivatives, about the point (a, b),

gives us the following:

Ln(Pf 1 Pt ) ≈ ln(a 1 b) 1 1

a 1 b [(Pf 2 a) 1 (Pf 2 b)]

2 1

2(a 1 b)2 [(Pf 2 a)2 1 2 (Pf 2 a) (Pt 2 b) 1 (Pt 2 b)2] (6)

Then, dropping higher-order and cross-product terms

and making some simpliﬁcations, Equation (6) can be

expressed as follows:

Ln(VMT ) 5 a0 1 a1 Pf 1 a2 Pt (7)

Then, considering the relationship between the functions

f(x ) 5 x and g(x ) 5 Ln(x ) through the Taylor expansion leaves us

with the following formula:

Ln(VMT ) 5 b0 1 b1 Ln(Pf) 1 b2 Ln(Pt) (8)

Equation (8) corresponds to Model 3 in Taiebat et al. (2019).

The existence of quadratic terms of fuel and time cost in

equation (6) raises a potential non-linear relationship between

these two costs and VMT. The existence of a non-linear

(quadratic) relationship can be tested using the following

simpliﬁed speciﬁcation:

Ln(VMT ) 5 c0 1 c1 Ln(Pf ) 1 c2 (Ln(Pf ))2

1 c3 Ln(Pt ) 1 c4 (Ln(Pt ))2 (9)

If c2 5 c4 5 0, then equation (9) reduces to Equation (8).

In cases where only c2 5 0 (c4 5 0) model (9) becomes

non-linear only with respect to Pt (Pf). The same potential

non-linearity applies to the speciﬁcation where we have

vmt as a

left-hand side variable:

Ln(VMT) 5 d0 1 d1 Ln(

vmt ) 1 d2 (Ln(

vmt ))2 (10)

Analogically, if d2 5 0, then model (10) reduces to a

linear speciﬁcation (3).

Considering the discussions above, we utilize a general-

to-speciﬁc modeling approach, starting with a non-linear

case to select the most appropriate model. In other words,

we begin with speciﬁcations (9) and (10) and test them to

select the relevant model.

3.2.2 Empirical Model Speciﬁcations

For econometric modeling purposes, the following model

speciﬁcations will be used.

Model 1: Ln(VMTi ) 5

0 1

1 Ln(Pf, i ) 1

i 1

1, i

Ln(VMTi ) 5

0 1

1 Ln(Pf, i ) 1

2 (Ln(Pf, i ))2

d9X

i 1

2, i (11)

Discussion Paper

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Model 2:

Ln(VMTi ) 5

0 1

1 Ln(Pt, i ) 1

«X

i 1

3, i

Ln(VMTi ) 5

«9

0 1

«9

1 Ln(Pt, i ) 1

«9

2 (Ln(Pt, i ))2

«9X

i 1

4, i (12)

Model 3:

Ln(VMTi ) 5

0 1

1 Ln(Pf, i ) 1

2 Ln(Pt, i ) 1

i 1

5, i

Ln(VMTi ) 5

0 1

1 Ln(Pf, i ) 1

2 (Ln(Pf, i ))2 1

Ln(Pt, i ) 1

4 (Ln(Pt, i ))2 1

u9X

i 1

6, i (13)

Model 4:

Ln(VMTi ) 5

0 1

1 Ln(

vmt, i ) 1

i 1

7, i

Ln(VMTi ) 5

0 1

1 Ln(

vmt, i ) 1

2 (Ln(

vmt, i ))2

q9X

i 1

8, i (14)

Model (13) nests Models (11) and (12). All variables are deﬁned

above;

incorporates all control variables, characterizing

households.

j, i is an error term and numbered to specify

different error series across models. The parameters in front of

the variables are coefﬁcients to be estimated econometrically,

and these are differentiated by different letters and prime

signs(

) across models. In addition to the general case, we have

estimated models for the urban and rural areas separately to

determine if there are any differences in consumers’ behaviors

in these two areas. Each model is estimated via generalized

least squares (GLS) regression, utilizing the sampling weights

provided by the NHTS.

3.2.3 Relative Energy Demand

The formula used in our analysis estimates the relative energy

demand (RED) as a function of changes in fuel economy and

time costs, incorporating the elasticity of VMT with respect to

travel costs. The following steps provide a detailed step-by-step

explanation of the formula and its derivation, highlighting how

changes in these parameters inﬂuence travel behavior.

Building on this foundation, the forecasting analysis examines

the energy and travel demand rebounds associated with

CAV adoption. Speciﬁcally, it evaluates how variations in fuel

economy (denoted as MPG) and travel time cost (denoted as Pt)

affect VMT. To capture these dynamics, we simulate changes

in fuel economy and time costs using the variables r1 and r2,

respectively.

Here:

[0. 0, 0.4] indicates fuel economy improvements

ranging from 0% to 40%.

[0. 0, 0.45] indicates travel time cost reductions

ranging from 0% to 45%.

The key metric of interest is the change in VMT, expressed as

a percentage difference compared to the pre-CAV business-

as-usual (BAU) scenario. The following provides a detailed,

step-by-step explanation of how the rebound effect from CAV

adoption is estimated in terms of VMT and energy demand.

Step 1: Initial Cost Per Mile (Baseline Travel

Cost, BTC)

The baseline travel cost (BTC) per mile (see Equation (15))

represents the initial cost incurred by drivers and is the

sum of Fuel cost per mile 5Fuel cost

Fuel Economy and the time

cost per mile (

TCT

BTC 5Fuel cost

Fuel Economy 1 TCT (15)

Where: fuel cost is expressed as the cost of fuel per gallon

($/gal), fuel economy is measured in miles per gallon (MPG),

and TCT represents the time cost per mile ($/mile).

Step 2: Adjusted Travel Cost (ATC)

When fuel economy increases by r1 % and time cost decreases

by r2 % the adjusted travel cost (ATC) is expressed as (16):

ATC (X, Y) 5Fuel cost

(1 1 r1

) * Fuel Economy 1 (1 2 r2) * TCT (16)

Where: Fuel cost

(1 1 r1

) * Fuel Economy and (1 2 r2) * TCT are the

fuel cost per mile and the time cost per mile after CAV

adoption, respectively.

1 1 r1 reﬂects an r1

% increase in fuel economy

(reducing cost per mile)

1 2 r2 reﬂects a r2

% decrease in time cost

(improving travel efﬁciency).

October 2025

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Step 3: Relative Energy Demand

The

RED

measures the percentage change in energy

consumption, assuming it is proportional to VMT

changes, see Equation (17).

RED 5

VMTCAV 2 VMTbaseline

VMTbaseline

VMTCAV

VMTbaseline

52 1

(17)

Considering that VMT 5 e Pf Pt

ˆ1

ˆ2 from the estimated

exponential version of linear Model 3, and assuming

there is no change in the fuel cost, we can derive the

following Equation (18):

ˆ0

ˆ1

ˆ2

RED 5VMTCAV

VMTbaseline

2 1 5 e Pf, CAV Pt, CAV

e Pf, baseline Pt, baseline

Pf,CAV

Pf,baseline

Pt,CAV

Pt,baseline

ˆ1

ˆ2

52 1

(18)

Fuel cost

(1 1 r1

) Fuel Economy

Fuel cost

Fuel Economy

ˆ1

2 1

(1 2 r2) TCT

TCT

ˆ2

Simplifying Equation (18), we end up with

(1 1 r1)

ˆ1

RED 5

ˆ2

(

(1 2 r2)

)

2 1 (19)

Where:

ˆ1 = Elasticity due to fuel cost changes via fuel economy

improvement,

ˆ2 = Elasticity due to time cost changes,

1 1 r1 reﬂects an r1

% increase in fuel economy (reducing

cost per mile) and 1- r2 reﬂects a r2

% decrease in time cost

(improving travel efﬁciency).

2 1

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Results and Discussion

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Results and Discussion

4.1 Empirical Results

Table 4 presents the estimation results for the models in which the coefﬁcients for fuel

and time costs were found to be statistically signiﬁcant. In this context, speciﬁcation refers

to the functional form used in the estimation – either linear or log-linear (log-log) – which

deﬁnes how VMT responds to changes in fuel and time costs.

Table 4. Elasticity estimation across various models – general context.

Variable Model 1 Model 2 Model 3 Model 4

log (fuel cost) -0.366 ***  -0.337 ***  

0.038  0.034  

(log (fuel cost))2    

    

log (time cost)  -0.569 *** -0.565 ***  

 0.016 0.016  

(log (time cost))2 -0.139 *** -0.137 ***  

 0.011 0.010  

log(time-inclusive cost)    -0.593 ***

   0.020

(log(time-inclusive cost))2   -0.145 ***

   0.016 

Intercept 7.352 *** 7.794 *** 7.087 *** 7.863 ***

0.098 0.055 0.089 0.054 

Household size -0.082 *** -0.115 *** -0.111 *** -0.114 ***

0.015 0.014 0.014 0.014 

Household trips 0.120 *** 0.125 *** 0.122 *** 0.125 ***

0.004 0.004 0.004 0.004 

Drivers in the household 0.099 ** 0.173 *** 0.156 *** 0.169 ***

0.031 0.028 0.028 0.028 

Vehicles in household 0.117 *** 0.134 *** 0.140 *** 0.135 ***

0.018 0.016 0.016 0.016 

Household in urban/rural area -0.473 *** -0.200 *** -0.198 *** -0.197 ***

0.041 0.038 0.037 0.037 

Workers in household 0.144 *** 0.262 *** 0.255 *** 0.263 ***

0.021 0.019 0.019 0.019 

Observations 5115 5115 5115 5115

Adjusted R20.255 0.391 0.403 0.395

AIC 17258.0 16227.1 16130.6 16200.8

BIC 17316.9 16292.5 16202.5 16266.2

* p < 0.05, ** p < 0.01, *** p < 0.001

Each of these model forms was estimated, and only those

yielding statistically signiﬁcant coefﬁcients are reported in

Table 4. Estimation results for the pooled (general) sample

appear in Table 4, while Table 5 and Table 6 present results

disaggregated by urban and rural areas, respectively.

Discussion Paper

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Based on the estimation results given in Table 4, the model selec-

tion criteria – speciﬁcally the Akaike Information Criterion (AIC)

and Bayesian Information Criterion (BIC) – indicate a preference

for Model 3. This is because Model 3 exhibits lower AIC and BIC

values compared to Models 1, 2, and 4, signaling a better over-

all ﬁt to the data while penalizing model complexity. In terms of

coefﬁcient estimates, Model 3 captures the effects of both fuel

cost and time cost jointly, whereas Models 1 and 2 estimate these

effects separately, and Model 4 assesses the combined impact.

This integration allows for a more complete representation of the

travel cost burden.

Moreover, the quadratic term for time cost is statistically

signiﬁcant in Model 3 (coefﬁcient ≈ 0.137, p < 0.001), suggesting

a non-linear relationship between time cost and VMT, which

Models 1 and 2 partially overlook. The signs of the estimated

coefﬁcients are consistent with economic expectations: both

fuel and time costs have negative elasticities, indicating that

VMT decreases as travel costs increase. The estimated fuel

cost elasticity is approximately -0.337 in the preferred model,

while the non-linear form of time cost elasticity provides a more

ﬂexible and realistic depiction of behavioral response.

Geometrically, this non-linearity is illustrated in Figure 1, which

plots the relationship between VMT and time cost using the

results from Model 3.

Figure 1. Non-linear relationship between time cost and VMT.

–12.5

–6

–4

–2

In(VMT)

–10.0 –7.5 –5.0

In(P_t) = –9.54

In(VMT) = –0.137(In( ))2 – 0.565In( ) + 7.087

Plot of In(VMT) = –0.137(In( ))2

– 0.565In( ) + 7.087

In(P_t) = 5.42

In(VMT) = 7.087 at In(P_t)=0

Vertex (In(P_t)=–2.06, In(VMT)=7.67)

–2.5 0.0 2.5 5.0 7.5

In( )

Source: Authors.

The validity of the quadratic relationship is subject to the

condition of the turning point being within the boundaries of

the sample values (Hasanov, Hunt, and Mikayilov 2021) of the

related explanatory variable, time cost in our case. Based on

information provided in the 3.1 Data Processing and Descriptive

Statistics section and ﬁndings in Figure 1, we see that this

condition is met. Therefore, we conclude that the relationship

between time cost and VMT demonstrates a U-shaped form

as shown in Figure 1. Using the elasticity formula, one

can obtain the corresponding elasticity (from the second

equation of Model 3) to be:

EPt 5

3 1 2

4 Ln(Pt, i ) (20)

October 2025

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Table 5. Elasticity estimation across various models – urban context.

Variable Model 1 Model 2 Model 3 Model 4

log (fuel cost) -0.086  -0.323 *** 

0.158  0.037 

log (fuel cost)20.049   

0.029   

log (time cost)  -0.560 *** -0.558 *** 

 0.018 0.018 

(log (time cost))2 -0.132 *** -0.130 *** 

 0.012 0.012 

log(time-inclusive cost)    -0.597 ***

   0.025

(log(time-inclusive cost))2   -0.133 ***

   0.018

Intercept 7.305 *** 7.588 *** 6.969 *** 7.672 ***

0.206 0.046 0.085 0.045

Household size -0.068 *** -0.104 *** -0.098 *** -0.103 ***

0.018 0.016 0.016 0.016

Household trips 0.119 *** 0.126 *** 0.122 *** 0.125 ***

0.005 0.004 0.004 0.004

Drivers in the household 0.078 * 0.159 *** 0.140 *** 0.154 ***

0.036 0.032 0.032 0.032

Vehicles in household 0.116 *** 0.128 *** 0.137 0.131 ***

0.022 0.020 0.020 0.020

Workers in household 0.144 *** 0.266 *** 0.261 *** 0.268 ***

0.023 0.021 0.021 0.021

Observations 4093 4093 4093 4093

Adjusted R20.23 0.377 0.389 0.382

AIC 13886.6 13016.2 12943.7 12989.4

BIC 13943.5 13073.1 13006.9 13046.2

* p < 0.05, ** p < 0.01, *** p < 0.001

Applying the estimation results in Table 4

(Model 3), we end up with:

EPt 5 2 0.565 2 2 * (2 0.137) * Ln(Pt, i )

5 2 0.565 2 0.274 * Ln(Pt, i ) (21)

The time cost (price) elasticity of VMT, as in Equation (21),

indicates that households respond to changes in time cost

non-linearly. Speciﬁcally, the elasticity in Equation (21) implies

that

) time cost negatively impacts VMT and

) an additional

increase in time cost makes households more responsive,

further reducing driving distance.

Discussion Paper

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Table 6. Elasticity estimation across various models – rural context.

Variable Model 1 Model 2 Model 3 Model 4

log (fuel cost) -0.428 ***  -0.386 *** 

0.079  0.072 

(log (fuel cost))2   

   

log (time cost)  -0.683 *** -0.672 *** 

 0.048 0.047 

(log (time cost))2 -0.213 *** -0.207 *** 

 0.029 0.028 

log(time-inclusive cost)    -0.685 ***

   0.047

(log(time-inclusive cost))2   -0.269 ***

   0.044

Intercept 7.086 *** 7.825 *** 6.948 *** 7.912 ***

0.207 0.104 0.193 0.105

Household size -0.145 *** -0.159 *** -0.159 *** -0.157 ***

0.033 0.030 0.030 0.030

Household trips 0.123 *** 0.122 *** 0.121 *** 0.122 ***

0.008 0.008 0.008 0.008

Drivers in the household 0.200 ** 0.228 *** 0.218 *** 0.226 ***

0.065 0.060 0.060 0.060

Vehicles in household 0.115 *** 0.148 *** 0.150 *** 0.147 ***

0.029 0.027 0.027 0.027

Workers in household 0.152 *** 0.248 *** 0.235 *** 0.247 ***

0.044 0.041 0.041 0.041

Observations 1022 1022 1022 1022

Adjusted R20.272 0.377 0.394 0.381

AIC 3369.5 3211.4 3184.9 3205.5

BIC 3408.9 3255.8 3234.2 3249.8

* p < 0.05, ** p < 0.01, *** p < 0.001

Model 3 was also selected for both rural and urban areas based

on the selection criteria (see Tables 5 and 6). The impact of fuel

cost is found to be higher for rural households. The time cost

elasticities for the general, urban, and rural household cases

are demonstrated in Figure 2. As can be seen from Figure2,

the elasticity for households living in rural areas is larger (i.e.,

more negative) than that of their urban counterparts. This

ﬁnding aligns with economic reasoning, as households in rural

areas tend to drive more, making an increase in time costs more

signiﬁcant for them.

October 2025

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Figure 2. Time cost elasticities of VMT.

–1.0 –0.5 0.0 0.5 1.0 1.5 2.0–1.5–2.0

–1.50

–1.25

–1.00

–0.75

Elasticity Values

–0.50

–0.25

0.00

In( )

General, –0.565 – 0.274ln( )

Urban, –0.558– 0.260ln( )

Rural, –0.672– 0.414ln( )

Plots of Elasticities in Terms of In (

)

Source: Authors.

Table 7 presents the estimation results across different income

groups. We have estimated all models across income groups,

and the results are presented in Appendix Table A1, Table A2,

and Table A3. Table 7 only presents the results of the selected

models. Table 8 and Table 9 present the results of estimations

across income groups for urban and rural areas, respectively.

As can be seen from Table 7, the time cost elasticity of VMT

is higher for households with higher income levels. The same

result holds for households living in rural and urban areas.

The fuel cost elasticity also changes across groups; however,

it is difﬁcult to identify a consistent pattern in any case. The

time cost elasticity follows a non-linear trajectory across

income groups in both urban and rural areas. However, it is

challenging to make a general statement regarding which

households experience a higher magnitude of elasticity

across income groups.

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban and Rural Areas – An Empirical Modeling Approach

Discussion Paper

Table 7. Elasticity estimation by income group – general context.

Variable

1st income group 2nd income group 3rd income group 4th income group 5th income group 6th income group

Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4

log (fuel cost) -0.191  -0.430 ***  -0.289 ***  -0.166 ***  -0.233 ***  0.094 

0.129  0.067  0.066  0.047  0.061  0.067 

log (time cost) -0.835 ***  -1.280 ***  -1.376 ***  -1.423 ***  -1.504 ***  -1.504 *** 

0.060  0.042  0.039  0.035  0.046  0.058 

log(time-inclusive cost)  -1.286 ***  -1.574 ***  -1.579 ***  -1.568 ***  -1.606 ***  -1.557 ***

 0.087  0.052  0.044  0.038  0.049  0.061

Intercept 6.247 *** 6.592 *** 5.829 *** 6.939 *** 7.036 *** 7.872 *** 7.819 *** 8.376 *** 8.701 *** 9.397 *** 9.960 *** 9.892 ***

0.300 0.154 0.159 0.078 0.153 0.062 0.117 0.063 0.159 0.083 0.175 0.101

Household size -0.048  -0.086 *** -0.095 *** -0.079 *** -0.078 *** -0.051 * -0.046 * -0.054 * -0.054 *  

0.034  0.025 0.025 0.022 0.022 0.021 0.021 0.022 0.023  

Household trips 0.126 *** 0.112 *** 0.130 *** 0.129 *** 0.137 *** 0.135 *** 0.097 *** 0.097 *** 0.087 *** 0.087 *** 0.111 *** 0.109 ***

0.013 0.012 0.009 0.009 0.008 0.008 0.005 0.005 0.006 0.006 0.007 0.007

Drivers in the household   0.117 * 0.131 *   0.143 *** 0.137 ***    

  0.052 0.053   0.042 0.042    

Vehicles in household   0.094 ** 0.096 ** 0.065 ** 0.060 * 0.052 * 0.051 * 0.076 ** 0.073 **  

  0.036 0.037 0.024 0.024 0.023 0.023 0.024 0.024  

Households in urban/rural areas -0.240 * -0.207          

0.116 0.115          

Workers in household 0.141 * 0.104 0.184 *** 0.189 *** 0.187 *** 0.198 *** 0.138 *** 0.139 *** 0.158 *** 0.157 *** 0.154 *** 0.153 ***

0.066 0.062 0.038 0.038 0.033 0.033 0.026 0.026 0.031 0.031 0.039 0.039

Observations 363 363 830 830 903 903 1489 1489 943 943 587 587

Adjusted R20.486 0.5 0.63 0.615 0.665 0.66 0.634 0.634 0.621 0.617 0.639 0.635

AIC 1075.2 1063.5 2213.2 2243.9 2345.7 2359.8 3924.0 3924.9 2397.7 2406.9 1525.2 1529.9

BIC 1106.4 1086.8 2255.7 2281.6 2384.1 2393.4 3971.7 3967.4 2436.5 2440.8 1551.4 1551.8

* p < 0.05, ** p < 0.01, *** p < 0.001

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban and Rural Areas – An Empirical Modeling Approach

October 2025

Table 8. Elasticity estimation by income group – urban context.

Variable

1st income group 2nd income group 3rd income group 4th income group 5th income group 6th income group

Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4

log (fuel cost) -0.163 -0.416 ***  -0.293 ***  -0.153 **  -0.161 *  0.110 

0.136 0.066  0.072  0.053  0.065  0.077 

log (time cost) -0.893 *** -1.240 ***  -1.371 ***  -1.430 ***  -1.546 ***  -1.488 *** 

0.064 0.048  0.044  0.038  0.050  0.064 

log(time-inclusive cost)  -1.367 ***  -1.561 ***  -1.569 ***  -1.588 ***  -1.665 ***  -1.557 ***

 0.097  0.060  0.051  0.042  0.053  0.067

Intercept 5.952 *** 6.438 *** 5.977 *** 7.013 *** 7.023 *** 7.852 *** 7.888 *** 8.427 *** 8.989 *** 9.524 *** 10.035 *** 9.984 ***

0.273 0.137 0.147 0.076 0.156 0.070 0.121 0.067 0.162 0.090 0.184 0.116

Household size -0.087 * -0.075   -0.088 *** -0.082 *** -0.054 * -0.048 * -0.050 * -0.063 *  

0.039 0.042   0.025 0.026 0.024 0.024 0.024 0.027  

Household trips 0.126 *** 0.113 *** 0.122 *** 0.120 *** 0.144 *** 0.143 *** 0.104 *** 0.104 *** 0.082 *** 0.082 *** 0.108 *** 0.105 ***

0.014 0.014 0.010 0.010 0.010 0.010 0.006 0.006 0.006 0.006 0.008 0.007

Drivers in the household  0.146     0.182 *** 0.173 ***  0.082  

 0.083     0.042 0.042  0.046  

Vehicles in household 0.112 0.093 ** 0.099 ** 0.048 0.042   0.049   

0.066 0.035 0.036 0.027 0.028   0.027   

Households in urban areas 0.126 0.169 *** 0.176 *** 0.256 *** 0.264 *** 0.125 *** 0.124 *** 0.173 *** 0.155 *** 0.128 ** 0.130 **

0.077 0.039 0.039 0.039 0.039 0.029 0.029 0.033 0.037 0.042 0.043

Workers in household -0.163 -0.416 ***  -0.293 ***  -0.153 **  -0.161 *  0.110 

0.136 0.066  0.072  0.053  0.065  0.077 

Observations 271 271 667 667 699 699 1166 1166 796 796 494 494

Adjusted R20.524 0.527 0.616 0.599 0.663 0.652 0.644 0.644 0.628 0.625 0.634 0.63

AIC 797.0 793.3 1730.6 1757.9 1813.7 1834.6 3023.5 3024.3 2024.3 2029.0 1287.5 1292.3

BIC 825.8 814.9 1762.1 1784.9 1850.1 1866.4 3064.0 3059.7 2061.7 2061.8 1312.7 1313.3

* p < 0.05, ** p < 0.01, *** p < 0.001

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban and Rural Areas – An Empirical Modeling Approach

Discussion Paper

Table 9. Elasticity estimation by income group – rural context.

Variable

1st income group 2nd income group 3rd income group 4th income group 5th income group 6th income group

Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4

log (fuel cost) 0.130  -0.501 *  -0.217  -0.135  -0.535 ***  0.101 

0.361  0.245  0.144  0.094  0.135  0.134 

log (time cost) -0.510 ***  -1.351 ***  -1.456 ***  -1.394 ***  -1.296 ***  -1.839 *** 

0.147  0.094  0.101  0.101  0.149  0.191 

log(time-inclusive cost)  -0.810 ***  -1.620 ***  -1.835 ***  -1.545 ***  -1.412 ***  -1.842 ***

 0.218  0.111  0.114  0.110  0.164  0.187

Intercept 7.271 *** 6.917 *** 5.611 *** 6.945 *** 7.153 *** 7.839 *** 7.979 *** 8.455 *** 7.416 *** 8.803 *** 9.850 *** 9.727 ***

0.833 0.303 0.575 0.185 0.343 0.144 0.251 0.120 0.360 0.191 0.413 0.232

Household size   -0.309 *** -0.319 *** -0.106 * -0.113 *      

  0.055 0.056 0.053 0.051      

Household trips 0.196 *** 0.182 *** 0.146 *** 0.147 *** 0.120 *** 0.120 *** 0.075 *** 0.074 *** 0.118 *** 0.115 *** 0.127 *** 0.127 ***

0.037 0.035 0.020 0.020 0.013 0.013 0.010 0.010 0.015 0.015 0.017 0.017

Drivers in the household   0.487 *** 0.514 *** 0.241 ** 0.219 *      

  0.112 0.113 0.090 0.087      

Vehicles in household       0.082 * 0.082 * 0.187 *** 0.200 ***  

      0.033 0.033 0.047 0.049  

Households in rural areas   0.161 0.130   0.208 *** 0.214 ***   0.297 ** 0.276 **

  0.092 0.093   0.058 0.057   0.099 0.098

Workers in household 0.130  -0.501 *  -0.217  -0.135  -0.535 ***  0.101 

0.31  0.245  0.144  0.094  0.135  0.134 

Observations 92 92 163 163 204 204 323 323 147 147 93 93

Adjusted R20.317 0.335 0.671 0.665 0.614 0.638 0.514 0.516 0.557 0.522 0.662 0.66

AIC 275.1 271.6 465.8 467.9 526.3 512.1 905.4 903.3 365.1 375.2 239.9 239.5

BIC 287.7 281.7 490.5 489.6 549.5 532.0 931.8 925.9 383.0 390.2 255.1 252.2

* p < 0.05, ** p < 0.01, *** p < 0.001

October 2025

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

4.2 Comparative Insights into

Travel Time Cost and Fuel Cost

Impacts on VMT

This section examines the elasticity of VMT in response to

changes in travel time and fuel costs, drawing comparisons

between our ﬁndings based on the new 2022 survey data and

those reported by Taiebat et al. (2019).

In the Taiebat et al. (2019) study, the estimated elasticity

of VMT with respect to travel time cost is -0.4, indicating a

relatively strong behavioral response to changes in travel time

cost. The study also ﬁnds that high-income households exhibit

greater elasticity with respect to travel time costs compared to

lower-income groups. This suggests that wealthier households

are more responsive to reductions in time cost, likely due to the

higher value they place on their time.

Our ﬁndings similarly underscore that high-income households

exhibit greater elasticity with respect to travel time costs

compared to lower-income groups. The study further reveals

that rural households exhibit higher elasticity to travel time cost

compared to urban households. This reﬂects the heightened

value of time savings in urban settings, where congestion and

higher opportunity costs make time reductions more impactful.

Additionally, rural travelers rely more heavily on personal

vehicles, making them more responsive to time cost changes.

Our analysis estimates higher elasticity values than those

reported by Taiebat et al. (2019), with travel time cost elasticity

around -0.565 in general contexts. The greater elasticity in our

ﬁndings (-0.565) compared to Taiebat et al. (2019) (-0.4486)

is likely due to the use of updated data (2022 vs. 2016). This

difference is supported by changes in average income levels:

Taiebat et al. (2019) reported an average income of $70,237,

whereas our analysis, based on the most recent dataset,

indicates an average income of $103,303.

Regarding fuel cost, the elasticity of VMT with respect to fuel

cost is -0.0989 in the Taiebat et al. (2019) study, indicating a

lower sensitivity compared to time cost. Our ﬁndings align,

showing that fuel cost reductions contribute to increased

VMT, but the effect is less pronounced compared to time

cost reductions. Our analysis estimates elasticity values that

are higher than those of Taiebat et al. (2019), with fuel cost

elasticity around -0.337 in general contexts. The higher fuel

cost elasticity (-0.337) in our ﬁndings compared to Taiebat et al.

(2019) (-0.0989) likely results from updated data (2022 versus

2017). Fuel cost and economic conditions changed between

these periods, likely affecting how travelers responded to fuel

cost variations. For example, households facing higher fuel

prices in 2022 might have adjusted their travel more signif-

icantly compared to 2017.

4.3 Simulation: Travel and

Energy Rebounds of CAVs

Figure 3 and Figure 4 illustrate the rebound effect of CAVs

on VMT in urban and rural settings across different income

groups. The X-axis represents the percentage increase in fuel

economy, while the Y-axis shows the percentage reduction in

travel time cost. The heatmaps depict the rebound effect in VMT

as a percentage, with darker colors (red to black) indicating

higher rebound effects.

Discussion Paper

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Figure 3. Rebound effect of CAVs on VMT – urban.

Source: Authors.

10%

15%

20%

25%

30%

35%

40%

45%

Rebound Effect

VMT (%)

0% 10% 20%

% Increase in fuel economy

30% 40%

10%

15%

20%

25%

30%

35%

40%

45%

% Reduction in travel time cost

10%

15%

20%

25%

30%

35%

40%

45%

Rebound Effect

VMT (%)

0% 10% 20%

% Increase in fuel economy

30% 40%

10%

15%

20%

25%

30%

35%

40%

45%

% Reduction in travel time cost

10%

15%

20%

25%

30%

35%

40%

45%

Rebound Effect

VMT (%)

0% 10% 20%

% Increase in fuel economy

30% 40%

10%

15%

20%

25%

30%

35%

40%

45%

% Reduction in travel time cost

10%

15%

20%

25%

30%

35%

40%

45%

Rebound Effect

VMT (%)

0% 10% 20%

% Increase in fuel economy

30% 40%

10%

15%

20%

25%

30%

35%

40%

45%

% Reduction in travel time cost

10%

15%

20%

25%

30%

35%

40%

45%

Rebound Effect

VMT (%)

0% 10% 20%

% Increase in fuel economy

30% 40%

10%

15%

20%

25%

30%

35%

40%

45%

% Reduction in travel time cost

10%

15%

20%

25%

30%

35%

40%

45%

Rebound Effect

VMT (%)

0% 10% 20%

% Increase in fuel economy

30% 40%

10%

15%

20%

25%

30%

35%

40%

45%

% Reduction in travel time cost

10%

15%

20%

25%

30%

35%

40%

45%

Rebound Effect

VMT (%)

0% 10% 20%

% Increase in fuel economy

30% 40%

10%

15%

20%

25%

30%

35%

40%

45%

% Reduction in travel time cost

1st income group

4th income group

Average all income groups

5th income group 6th income group

2nd income group 3rd income group

October 2025

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Figure 4. Rebound effect of CAVs on VMT – rural.

Source: Authors.

1st income group

Average all income

groups

4th income group 5th income group 6th income group

2nd income group 3rd income group

% Reduction in travel time cost

10%

15%

20%

25%

30%

35%

40%

45%

10%

15%

20%

25%

30%

35%

40%

45%

10%

15%

20%

25%

30%

35%

40%

45%

10%

15%

20%

25%

30%

35%

40%

45%

10%

15%

20%

25%

30%

35%

40%

45%

10%

15%

20%

25%

30%

35%

40%

45%

10%

15%

20%

25%

30%

35%

40%

45%

10%

% Reduction in travel time cost

15%

20%

25%

30%

35%

40%

45%

10%

% Reduction in travel time cost

15%

20%

25%

30%

35%

40%

45%

0% 10% 20%

% Increase in fuel economy

30% 40% 0% 10% 20%

% Increase in fuel economy

30% 40% 0% 10% 20%

% Increase in fuel economy

30% 40%

10%

15%

20%

25%

30%

35%

40%

45%

10%

15%

20%

25%

30%

35%

40%

45%

10%

15%

20%

25%

30%

35%

40%

45%

Rebound Effect

VMT (%)

Rebound Effect

VMT (%)

Rebound Effect

VMT (%)

Rebound Effect

VMT (%)

Rebound Effect

VMT (%)

Rebound Effect

VMT (%)

Rebound Effect

VMT (%)

0% 10% 20%

% Increase in fuel economy

30% 40% 0% 10% 20%

% Increase in fuel economy

30% 40%

0% 10% 20%

% Increase in fuel economy

30% 40%

0% 10% 20%

% Increase in fuel economy

30% 40%

10%

15%

20%

25%

30%

35%

40%

45%

10%

15%

20%

25%

30%

35%

40%

45%

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Discussion Paper

A key observation across all scenarios is the positive rebound

effect, where improvements in fuel economy and reductions in

travel time costs lead to an increase in VMT. As fuel economy

improves (moving rightward on the X-axis) and travel time costs

decrease (upward on the Y-axis), the rebound effect becomes

more pronounced, as indicated by the darker regions on the

heatmaps (shades of red to black).

The comparison between rural and urban areas highlights

distinct patterns in the rebound effect of VMT based on fuel

economy improvements and reductions in travel time costs. In

rural areas, the rebound effect (average of all income groups)

is signiﬁcantly more pronounced, as indicated by the darker

regions on the heatmap. This can be attributed to the longer

travel distances typically required in rural settings, where fuel

economy improvements lead to substantial cost savings per

mile. Additionally, rural residents often rely heavily on personal

vehicles due to limited alternative transportation options, which

ampliﬁes the impact of reduced travel costs, even with small

increases in fuel economy. For instance, a study by Ahmadnia

and Rowangould (2024) analyzed data from 132,141 households

in Vermont and found that a 10% increase in fuel efﬁciency

led to a 1.4% increase in miles driven, a 14% rebound effect.

This effect was more pronounced in rural areas, where people

are more automobile-dependent and drive longer distances.

The rebound effect in rural areas is noticeable and grows more

signiﬁcant as travel time cost reductions increase. This suggests

that rural drivers are particularly sensitive to both factors,

likely because of the essential nature of their trips and the

relative lack of congestion.

In contrast, the rebound effect in urban areas is less

pronounced, as evidenced by the lighter regions in the corre-

sponding heatmap. Urban drivers tend to have shorter trip

distances and greater access to alternative transportation

options such as public transit, walking, or biking. These factors

reduce their dependency on personal vehicles and limit the

extent to which cost reductions translate into increased

VMT. Even at similar levels of fuel economy improvements or

travel time cost reductions, the urban rebound effect remains

moderate. Congestion in urban areas further dampens the

potential for induced travel, as time savings from reduced travel

costs are often offset by trafﬁc conditions (Litman 2023).

Income levels play a signiﬁcant role in inﬂuencing the rebound

effect. Higher-income groups demonstrate a more pronounced

rebound effect, as indicated by the darker regions on the

heatmaps, even at moderate improvements in fuel economy

and reductions in travel time costs. For these groups, the value

of time savings is particularly signiﬁcant, making them more

responsive to changes in travel costs. While higher-income

individuals may not be as constrained by fuel expenses,

reductions in travel time costs can incentivize additional travel,

leading to a notable increase in VMT. This suggests that, for

higher-income groups, the rebound effect is driven primarily by

time-related efﬁciencies rather than direct fuel cost savings, as

shown in Table 8 and Table 9.

The implications of this rebound effect are signiﬁcant,

particularly for energy efﬁciency and emissions reduction

policies. The increase in VMT due to the rebound effect offsets

the expected energy savings from fuel economy improvements,

especially in rural areas. This presents a critical challenge for

policymakers aiming to reduce transportation emissions. In

urban areas, strategies such as congestion pricing or road

tolls can be effective in counteracting the rebound effect by

reducing the ﬁnancial incentive to increase travel. In rural

areas, policies could focus on reducing dependency on private

vehicles by promoting public transportation or improving

rural transit options.

The role of connected vehicles in improving travel efﬁciency

and reducing costs underscores the need for complementary

measures. Promoting alternative transportation modes, imple-

menting better land-use planning, and managing induced travel

demand are crucial to ensuring that the beneﬁts of connected

vehicles do not come at the expense of increased emissions

and energy consumption.

Table 10 summarizes the rebound effects of energy efﬁciency

improvements or travel cost reductions in both urban and

rural areas, detailing three key aspects: Initial Rebound Effect,

Sensitivity to Parameters, and Policy Focus.

October 2025

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Table 10. Rebound effects of energy savings in urban and rural areas.

Aspect Urban Rural

Initial Rebound Effect Lower, due to congestion and shorter trip distances. Higher, as rural trips are longer, and cost reductions have a

greater impact.

Sensitivity to

Parameters

Signiﬁcant rebound only at high fuel economy improvements

and travel cost reductions. Rebound effect noticeable even at moderate changes.

Policy Focus Congestion pricing, road tolls, and alternative modes of

transport.

Road tolls, public transport incentives to reduce private

vehicle dependency.

Conclusions and

Policy Insights

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

By applying econometric models to household-level data from

the 2022 National Household Travel Survey (NHTS), the analysis

uncovers essential insights into the behavioral shifts induced

by CAV technology, shedding light on the complexities of the

energy rebound effect.

Key ﬁndings reveal that time cost elasticity is signiﬁcantly

higher than fuel cost elasticity, indicating that reductions in

perceived travel time – enabled by CAVs – will be the primary

driver of induced travel. In the preferred linear model, the

time cost elasticity of VMT is -0.565 for the general population,

-0.558 in urban areas, and -0.672 in rural areas, suggesting

that rural households are more responsive to changes in

time-related travel costs due to their longer travel distances and

lack of modal alternatives.

Fuel cost elasticity, while lower, remains substantial. The

general sample shows a fuel cost elasticity of -0.337, with rural

areas displaying the highest sensitivity (-0.386) compared

to urban areas (-0.323). These ﬁndings highlight that both

components – fuel and time – must be considered in energy

rebound assessments.

Moreover, the income-stratiﬁed analysis reveals substantial

heterogeneity in behavioral response. For the lowest-income

group, time cost elasticity is -0.835, indicating a moderate but

meaningful sensitivity to travel time reductions. In contrast, the

highest-income group exhibits a markedly higher elasticity of

-1.504, suggesting that wealthier households are signiﬁcantly

more responsive to time cost savings and thus more susceptible

to the energy rebound effect.

These elasticity estimates are further reinforced by the

simulation analysis, which models improvements in fuel

efﬁciency (0%-40%) and reductions in time costs (0%-45%).

Results show consistently positive rebound effects across all

scenarios, with rural areas experiencing the most pronounced

VMT increases. The heatmap analysis conﬁrms that rural

households are more elastic with respect to time cost

reductions than their urban counterparts – reﬂecting their

greater dependence on personal vehicles and limited access to

alternative modes of transport.

From a policy perspective, these ﬁndings emphasize the

importance of proactive strategies to mitigate potential

unintended consequences of widespread CAV deployment.

Policymakers and urban planners should consider mechanisms

such as congestion pricing, demand management strategies,

and incentives for shared mobility solutions to ensure that CAVs

contribute to energy savings and sustainable mobility rather

than exacerbating VMT and emissions.

Additionally, the implications of this study extend beyond the

U.S., offering valuable insights for regions like Saudi Arabia,

where energy and transportation policies are rapidly evolving

in response to emerging mobility technologies. Future research

can explore how policy instruments – such as congestion

pricing, VMT-based taxation, and shared mobility incentives –

can be optimized to mitigate rebound effects while promoting

sustainable, inclusive, and energy-efﬁcient transport systems.

Conclusions and Policy Insights

This study provides a critical examination of the impact of CAVs on travel demand, fuel

efﬁciency, and time costs, with a particular focus on urban versus rural settings.

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Endnote

1 SAE International deﬁnes vehicle automation across six levels (0-5). Level 3 (“Conditional Automation”) allows the vehicle to

perform all driving tasks in speciﬁc scenarios without driver input; however the driver must remain available to take over. Level 4

(“High Automation”) enables full driving without human intervention, but only within limited operational domains (e.g., geofenced

urban areas).

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Aderibigbe, Oluwayemi-Oniya, and Trynos Gumbo. 2022. “Inﬂuence

of Socio-Economic Attributes on Travel Behaviour in the Rural

Areasof Nigeria: Towards a Sustainable Rural Planning and

Development.”

Urban, Planning and Transport Research

10 (1):

181–99. https://doi.org/10.1080/21650020.2022.2072946.

Ahmadnia, Narges, and Gregory Rowangould. 2024. “A Spatial

Analysis of the Fuel Economy Rebound Effect in Vermont.”

VTrans

Research and Innovation Symposium

. Vermont. https://vtrans.

vermont.gov/sites/aot/ﬁles/Research/2024 Symposium/Session

Presentations/D4 - VTrans_Highlights_ReboundEffect.pdf.

Aljouﬁe, Mohammed, Mark Zuidgeest, Mark Brussel, and Martin van

Maarseveen. 2013. “Spatial–Temporal Analysis of Urban Growth and

Transportation in Jeddah City, Saudi Arabia.”

Cities

31 (April): 57–68.

https://doi.org/10.1016/j.cities.2012.04.008.

Cao, Xinyu (Jason), Sasiwooth Wongmonta, and Sangho Choo. 2013.

“Examining the Adaptation Process of People’s Behavioral Response

to High Gasoline Costs.”

KSCE Journal of Civil Engineering

17 (4):

815–23. https://doi.org/10.1007/s12205-013-0208-1.

Cervero, Robert. 2002. “Induced Travel Demand: Empirical Evidence,

and Normative Policies.”

Journal of Planning Literature

17 (1): 3–20.

https://doi.org/10.1177/088122017001001.

Choi, Yunkyung, Subhrajit Guhathakurta, and Anurag Pande.

2022. “An Empirical Bayes Approach to Quantifying the Impact of

Transportation Network Companies (TNCs) Operations on Travel

Demand.”

Transportation Research Part A: Policy and Practice

161

(July): 269–83. https://doi.org/10.1016/j.tra.2022.04.008.

Clewlow, Regina R., and Gouri Shankar Mishra. 2017.

Disruptive

Transportation: The Adoption, Utilization, and Impacts of Ride-

Hailing in the United States

. Research Report – UCD-ITS-RR-17-07.

Davis, CA: Institute of Transportation Studies, University of

California, Davis. https://escholarship.org/uc/item/82w2z91j.

Cohen, Scott A., and Debbie Hopkins. 2018. “Autonomous Vehicles

and the Future of Urban Tourism.”

Annals of Tourism Research

(January): 33–42. https://doi.org/10.1016/j.annals.2018.10.009.

Duranton, Gilles, and Matthew A. Turner. 2011. “The Fundamental

Law of Road Congestion: Evidence from US Cities.”

American

Economic Review

101 (6): 2616–52. https://doi.org/10.1257/

aer.101.6.2616.

Faghihian, Hamed, and Arman Sargolzaei. 2023. “Energy Efﬁciency

of Connected Autonomous Vehicles: A Review.”

Electronics

12 (19):

4086. https://doi.org/10.3390/electronics12194086.

Fagnant, Daniel J., and Kara Kockelman. 2015. “Preparing a Nation

for Autonomous Vehicles: Opportunities, Barriers and Policy

Recommendations.”

Transportation Research Part A: Policy and

Practice

77 (July): 167–81. https://doi.org/10.1016/j.tra.2015.04.003.

Federal Highway Administration. 2009. “Make and Model,

National Household Travel Survey.” https://nhts.ornl.gov/2009/

pub/2009farsmakemodel.pdf.

Federal Highway Administration. 2018. “Chapter 3 Travel.” In

Status

of the Nation’s Highways, Bridges, and Transit

, 23rd ed. https://

www.fhwa.dot.gov/policy/23cpr/smallbook.cfm.

Federal Highway Administration. 2022. “2022 National Household

Travel Survey.” Washington, DC. https://nhts.ornl.gov.

Fong, Eric, Blake Lane, and Scott Samuelsen. 2024. “Impact

of Vehicle-to-Everything Connectivity on Fuel Economy.”

Transportation Research Part D: Transport and Environment

136

(November): 104423. https://doi.org/10.1016/j.trd.2024.104423.

Garikapati, Divya, and Sneha Sudhir Shetiya. 2024. “Autonomous

Vehicles: Evolution of Artiﬁcial Intelligence and the Current

IndustryLandscape.”

Big Data and Cognitive Computing

8 (4): 42.

https://doi.org/10.3390/bdcc8040042.

Guo, Huizhong, Bo Yu, Shan Bao, Zhenhai Li, Eric M. Rask, Henry Liu,

and James Sayer. 2022.

How Vehicle Connectivity Based Eco-Routing

Choices Will Impact on Driver Decision Making

. Final Report.

UMTRI-2023-17. Ann Arbor, MI: Center for Connected and Automated

Transportation. https://doi.org/10.7302/22364.

Hasanov, Fakhri J., Lester C. Hunt, and Jeyhun I. Mikayilov. 2021.

“Estimating Different Order Polynomial Logarithmic Environmental

Kuznets Curves.”

Environmental Science and Pollution Research

(31): 41965–87. https://doi.org/10.1007/s11356-021-13463-y.

Hughes, Jonathan E., Christopher R. Knittel, and Daniel

Sperling.2008. “Evidence of a Shift in the Short-Run Price

Elasticityof Gasoline Demand.”

Energy Journal

29 (1): 113–34.

https://doi.org/10.5547/ISSN0195-6574-EJ-Vol29-No1-9.

References

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

International Transport Forum. 2018.

What Is the Value of Saving

Travel Time? Summary and Conclusions

. ITF Roundtables.

Litman, Todd. 2023.

Congestion Costing Critique. Critical Evaluation

of the ‘Urban Mobility Report.’

Victoria Transport Policy Institute.

www.vtpi.orgInfo@vtpi.org.

Litman, Todd. 2025a.

Understanding Transport Demands and

Elasticities: How Prices and Other Factors Affect Travel Behavior

Victoria Transport Policy Institute.

http://www.vtpi.org/elasticities.pdf.

Litman, Todd. 2025b.

Evaluating Transportation Equity: Guidance for

Incorporating Distributional Impacts in Transport Planning.

Victoria

Transport Policy Institute. https://vtpi.org/equity.pdf.

Martin, Elliot, Susan Shaheen, and Brooke Wolfe. 2024.

“Environmental Impacts of Transportation Network Company

(TNC)/Ride-Hailing Services: Evaluating Net Vehicle Miles Traveled

and Greenhouse Gas Emission Impacts within San Francisco, Los

Angeles, and Washington, D.C. Using Survey and Activity Data.”

Sustainability

16 (17): 7454. https://doi.org/10.3390/su16177454.

Martínez-Buelvas, Laura, Andry Rakotonirainy, Deanna Grant-Smith,

and Oscar Oviedo-Trespalacios. 2022. “A Transport Justice Approach

to Integrating Vulnerable Road Users with Automated Vehicles.”

Transportation Research Part D: Transport and Environment

113

(December): 103499. https://doi.org/10.1016/j.trd.2022.103499.

Milakis, Dimitris, Maaike Snelder, Bart Van Arem, Bert Van Wee,

and Gonçalo Homem De Almeida Correia. 2017. “Development and

Transport Implications of Automated Vehicles in the Netherlands:

Scenarios for 2030 and 2050.”

European Journal of Transport and

Infrastructure Research

17 (1): 63–85. https://doi.org/10.18757/

ejtir.2017.17.1.3180.

Ngo, Nicole S., Zhenpeng Zou, Yizhao Yang, and Edward Wei.

2024. “The Impact of Urban Form on the Relationship Between

Vehicle Miles Traveled and Air Pollution.”

Transportation

ResearchInterdisciplinary Perspectives

28 (November): 101288.

https://doi.org/10.1016/j.trip.2024.101288.

Pi, Dawei, Pengyu Xue, Weihua Wang, Boyuan Xie, Hongliang Wang,

Xianhui Wang, and Guodong Yin. 2023. “Automotive Platoon Energy-

Saving: A Review.”

Renewable and Sustainable Energy Reviews

179

(June): 113268. https://doi.org/10.1016/j.rser.2023.113268.

Samaddar, Monali, and Daniel Bachman. 2022. “No Longer

a SmoothDrive: How Automobiles’ Role in the US Economy

HasEvolved.”

Deloitte Insights

, February 28. https://www2.deloitte.

com/us/en/insights/economy/spotlight/automobile-impact-us-

economy.html.

Shatnawi, Ibrahem, Juan Nicolas Gonzalez, and Neda Masoud. 2025.

“Toward AV-CAV Deployment in the Kingdom of Saudi Arabia: A

Readiness Assessment Based on Expert Feedback.”

Research in

Transportation Business and Management

60 (June): 101378.

https://doi.org/10.1016/j.rtbm.2025.101378.

Sheldon, Tamara, and Rubal Dua. 2018. “Drivers of New Light-

Duty Vehicle Fleet Fuel Economy in Saudi Arabia.” Riyadh: King

Abdullah Petroleum Studies and Research Center (KAPSARC).

https://doi.org/10.30573/KS–2018-DP55.

Sterner, Thomas. 2007. “Fuel Taxes: An Important Instrument

for Climate Policy.”

Energy Policy

35 (6): 3194–3202.

https://doi.org/10.1016/j.enpol.2006.10.025.

Taiebat, Morteza, Austin L. Brown, Hannah R. Safford, Shen Qu,

and Ming Xu. 2018. “A Review on Energy, Environmental, and

Sustainability Implications of Connected and Automated

Vehicles.”

Environmental Science & Technology

52 (20): 11449–65.

https://doi.org/10.1021/acs.est.8b00127.

Taiebat, Morteza, Samuel Stolper, and Ming Xu. 2019. “Forecasting

the Impact of Connected and Automated Vehicles on Energy Use:

A Microeconomic Study of Induced Travel and Energy Rebound.”

Applied Energy

247 (March): 297–308. https://doi.org/10.1016/j.

apenergy.2018.03.174.

U.S. Department of Energy (DOE). 2024. “Datasets for All Model

Years (1984–2025).” Fuel Economy Data. https://www.fueleconomy.

gov/feg/download.shtml.

Vahidi, Ardalan, and Antonio Sciarretta. 2018. “Energy Saving

Potentials of Connected and Automated Vehicles.”

Transportation

Research Part C: Emerging Technologies

95 (October): 822–43.

https://doi.org/10.1016/j.trc.2018.09.001.

Victoria Transport Policy Institute. 2023.

Transportation Cost and

Beneﬁt Analysis II – Travel Time Costs

https://www.vtpi.org/tca/tca0502.pdf.

References

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

References

Wadud, Zia, Don MacKenzie, and Paul Leiby. 2016. “Help or

Hindrance? The Travel, Energy and Carbon Impacts of Highly

Automated Vehicles.”

Transportation Research Part A: Policy and

Practice

86 (April): 1–18. https://doi.org/10.1016/j.tra.2015.12.001.

Yueshuai He, Brian, Qinhua Jiang, and Jiaqi Ma. 2022. “Connected

Automated Vehicle Impacts in Southern California Part-I: Travel

Behavior and Demand Analysis.”

Transportation Research Part

D: Transport and Environment

109 (August): 103329.

https://doi.org/10.1016/j.trd.2022.103329.

Zhang, Wenwen, and Subhrajit Guhathakurta. 2017. “Parking

Spaces in the Age of Shared Autonomous Vehicles.”

Transportation

Research Record: Journal of the Transportation Research Board

2651

(1): 80–91. https://doi.org/10.3141/2651-09.

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban and Rural Areas – An Empirical Modeling Approach

Table A1. Elasticity estimation by income group – general context.

Variable

1st income group 2nd income group 3rd income group 4th income group 5th income group 6th income group

Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4

log (fuel cost) -0.177  -0.430 ***  -0.285 ***  0.416 *  -0.233 ***  0.110 

0.127  0.067  0.066  0.205  0.061  0.067 

(log (fuel cost))2      0.099 **     

      0.034     

log (time cost) -1.312 ***  -1.280 ***  -1.376 ***  -1.374 ***  -1.504 ***  -1.848 *** 

0.141  0.042  0.039  0.041  0.046  0.153 

(log (time cost))2-0.171 ***    0.044  -0.077 *    0.139 * 

0.046    0.025  0.034    0.058 

log(time-

inclusive cost)

 -1.495 ***  -1.510 ***  -1.657 ***  -1.552 ***  -1.755 ***  -1.968 ***

 0.154  0.055  0.047  0.053  0.113  0.169

(log(time-

inclusive cost))2

 -0.145  0.195 ***  0.150 ***  -0.018  0.091  0.159 **

 0.088  0.057  0.032  0.041  0.063  0.061

Intercept 6.015 *** 6.584 *** 5.829 *** 6.865 *** 7.022 *** 7.817 *** 8.631 *** 8.380 *** 8.701 *** 9.425 *** 10.143 *** 10.091 ***

0.294 0.153 0.159 0.080 0.153 0.063 0.294 0.064 0.159 0.085 0.190 0.127

Household size   -0.086 *** -0.095 *** -0.079 *** -0.080 *** -0.054 ** -0.046 * -0.054 * -0.055 *  

  0.025 0.025 0.022 0.022 0.021 0.021 0.022 0.023  

Household trips 0.113 *** 0.111 *** 0.130 *** 0.131 *** 0.138 *** 0.138 *** 0.096 *** 0.096 *** 0.087 *** 0.088 *** 0.113 *** 0.111 ***

0.012 0.012 0.009 0.009 0.008 0.008 0.005 0.005 0.006 0.006 0.007 0.007

Appendix

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban and Rural Areas – An Empirical Modeling Approach

Appendix

Variable

1st income group 2nd income group 3rd income group 4th income group 5th income group 6th income group

Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4

Drivers in the house-

hold

  0.117 * 0.131 *   0.154 *** 0.137 ***    

  0.052 0.053   0.042 0.042    

Vehicles in household   0.094 ** 0.092 * 0.065 ** 0.063 ** 0.045 * 0.051 * 0.076 ** 0.072 **  

  0.036 0.037 0.023 0.023 0.023 0.023 0.024 0.024  

Households in urban/

rural areas

-0.218 -0.210          

0.114 0.114          

Workers in household 0.102 0.097 0.184 *** 0.195 *** 0.183 *** 0.185 *** 0.136 *** 0.139 *** 0.158 *** 0.160 *** 0.154 *** 0.152 ***

0.062 0.062 0.038 0.038 0.033 0.033 0.026 0.026 0.031 0.031 0.039 0.039

Observations 363 363 830 830 903 903 1489 1489 943 943 587 587

Adjusted R20.503 0.502 0.63 0.62 0.666 0.668 0.637 0.634 0.621 0.617 0.642 0.639

AIC 1063.1 1062.7 2213.2 2234.3 2344.6 2339.5 3914.3 3926.7 2397.7 2406.7 1521.3 1525.1

BIC 1094.3 1090.0 830 830 2387.9 2377.9 3972.7 3974.5 2436.5 2445.5 1551.9 1551.3

* p < 0.05, ** p < 0.01, *** p < 0.001

Table A1. Continued

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban and Rural Areas – An Empirical Modeling Approach

Appendix

Table A2. Elasticity estimation by income group urban context.

Variable

1st income group 2nd income group 3rd income group 4th income group 5th income group 6th income group

Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4

log (fuel cost) -0.145  -0.416 ***  -0.291 ***  0.324  -0.161 *  0.113 

0.135  0.066  0.071  0.216  0.065  0.077 

(log (fuel cost))2      0.090 *     

      0.039     

log (time cost) -1.284 ***  -1.240 ***  -1.407 ***  -1.382 ***  -1.546 ***  -1.808 *** 

0.151  0.048  0.045  0.047  0.050  0.171 

(log (time cost))2-0.152 **    0.105 ***  -0.066    0.124 * 

0.050    0.030  0.036    0.062 

log (time-

inclusive cost)

 -1.367 ***  -1.514 ***  -1.717 ***  -1.588 ***  -1.904 ***  -2.010 ***

 0.097  0.061  0.058  0.042  0.133  0.195

(log (time-

inclusive cost))2

   0.208 **  0.182 ***    0.142 *  0.166 *

   0.065  0.036    0.072  0.067

Intercept 5.888 *** 6.438 *** 5.977 *** 6.951 *** 6.969 *** 7.789 *** 8.494 *** 8.427 *** 8.989 *** 9.574 *** 10.192 *** 10.223 ***

0.269 0.137 0.147 0.078 0.156 0.070 0.284 0.067 0.162 0.093 0.200 0.151

Household size -0.097 * -0.075   -0.091 *** -0.087 *** -0.055 * -0.048 * -0.050 * -0.065 *  

0.042 0.042   0.025 0.025 0.024 0.024 0.024 0.027  

Household trips 0.117 *** 0.113 *** 0.122 *** 0.123 *** 0.148 *** 0.146 *** 0.102 *** 0.104 *** 0.082 *** 0.084 *** 0.110 *** 0.107 ***

0.014 0.014 0.010 0.010 0.010 0.010 0.006 0.006 0.006 0.006 0.008 0.008

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban and Rural Areas – An Empirical Modeling Approach

Variable

1st income group 2nd income group 3rd income group 4th income group 5th income group 6th income group

Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4

Drivers in the

household

0.181 * 0.146     0.186 *** 0.173 ***  0.083  

0.085 0.083     0.042 0.042  0.046  

Vehicles in household   0.093 ** 0.095 ** 0.053 0.050   0.049   

  0.035 0.036 0.027 0.027   0.027   

Households in

urban areas

  0.169 *** 0.177 *** 0.248 *** 0.253 *** 0.121 *** 0.124 *** 0.173 *** 0.159 *** 0.127 ** 0.130 **

  0.039 0.039 0.039 0.039 0.029 0.029 0.033 0.036 0.042 0.043

Workers in household -0.145  -0.416 ***  -0.291 ***  0.324  -0.161 *  0.113 

0.135  0.066  0.071  0.216  0.065  0.077 

Observations 271 271 667 667 699 699 1166 1166 796 796 494 494

Adjusted R20.535 0.527 0.616 0.605 0.668 0.664 0.646 0.644 0.628 0.626 0.636 0.634

AIC 790.8 793.3 1730.6 1749.7 1803.5 1811.6 3018.8 3024.3 2024.3 2027.1 1285.4 1288.2

BIC 819.6 814.9 1762.1 1781.2 1844.4 1848.0 3069.4 3059.7 2061.7 2064.6 1314.8 1313.4

* p < 0.05, ** p < 0.01, *** p < 0.001

Table A2. Continued

Appendix

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban and Rural Areas – An Empirical Modeling Approach

Table A3. Elasticity estimation by income group – rural context.

Variable

1st Income Group 2nd Income Group 3rd Income Group 4th Income Group 5th Income Group 6th Income Group

Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4

log (fuel cost) -3.990  7.065 **  -0.242  0.588  -0.535 ***  -1.529 * 

2.953  2.216  0.138  0.454  0.135  0.762 

(log (fuel cost))2-0.863  1.545 ***    0.112    -0.239 * 

0.607  0.450    0.066    0.110 

log (time cost) -1.352 ***  -1.342 ***  -1.897 ***  -1.368 ***  -1.296 ***  -1.904 *** 

0.398  0.091  0.145  0.102  0.149  0.189 

(log (time cost))2 -0.263 *    -0.275 ***  -0.208     

0.115    0.067  0.118     

log(time-inclusive cost)  -1.385 **  -1.620 ***  -2.029 ***  -1.545 ***  -1.412 ***  -1.842 ***

 0.442  0.111  0.147  0.110  0.164  0.187

(log(time-inclusive

cost))2

 -0.307    -0.232 *      

 0.205    0.113      

Intercept 1.850 6.746 *** 14.819 *** 6.945 *** 7.135 *** 7.895 *** 8.956 *** 8.455 *** 7.416 *** 8.803 *** 7.471 *** 9.727 ***

3.615 0.322 2.738 0.185 0.330 0.146 0.725 0.120 0.360 0.191 1.168 0.232

Household size   -0.292 *** -0.319 *** -0.114 * -0.115 * -0.067     

  0.054 0.056 0.051 0.051 0.045     

Household trips 0.205 *** 0.181 *** 0.140 *** 0.147 *** 0.120 *** 0.120 *** 0.078 *** 0.074 *** 0.118 *** 0.115 *** 0.121 *** 0.127 ***

0.038 0.035 0.019 0.020 0.013 0.013 0.010 0.010 0.015 0.015 0.017 0.017

Drivers in the household   0.404 *** 0.514 *** 0.195 * 0.198 * 0.164     

  0.111 0.113 0.088 0.087 0.100     

Appendix

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban and Rural Areas – An Empirical Modeling Approach

Variable

1st Income Group 2nd Income Group 3rd Income Group 4th Income Group 5th Income Group 6th Income Group

Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4 Model 3 Model 4

Vehicles in household       0.066 0.082 * 0.187 *** 0.200 ***  

      0.034 0.033 0.047 0.049  

Households in

rural areas

  0.220 * 0.130   0.216 *** 0.214 ***   0.295 ** 0.276 **

  0.091 0.093   0.063 0.057   0.097 0.098

Workers in household -3.990  7.065 **  -0.242  0.588  -0.535 ***  -1.529 * 

2.953  2.216  0.138  0.454  0.135  0.762 

Observations 92 92 163 163 204 204 323 323 147 147 93 93

Adjusted R20.356 0.344 0.693 0.665 0.642 0.644 0.521 0.516 0.557 0.522 0.676 0.66

AIC 271.6 271.3 455.8 467.9 511.7 509.7 904.5 903.3 365.1 375.2 237.0 239.5

BIC 289.3 283.9 483.7 489.6 538.3 533.0 946.0 925.9 383.0 390.2 254.7 252.2

* p < 0.05, ** p < 0.01, *** p < 0.001

Table A3. Continued

Appendix

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

Ibrahem Shatnawi

Ibrahem Shatnawi is a Senior Fellow in Transportation and Infrastructure at KAPSARC

with more than 19 years of experience in transportation engineering. He currently

leads the development of the project “An Integrated Transport Energy Demand

Model for KSA.” His research focuses on transport policies and technologies for

decarbonizing road transport, as well as integrated transport demand modeling to

estimate energy demand and emissions. Ibrahem holds a B.Sc. in Civil Engineering,

specializing in transportation, from Jordan University of Science and Technology, and

a Ph.D. in Civil Engineering, specializing in trafﬁc engineering, from the University of

Akron, Ohio, U.S.

Juan Nicolas Gonzalez

Juan Nicolas Gonzalez is a Postdoctoral Researcher at KAPSARC. He hold an M.Sc.

in Civil Engineering with a specialization in trafﬁc and transportation, as well as a

Ph.D. in Civil Engineering Systems with a focus on transport and territory. He has

contributed to projects on transport policy, parking management, low-emission zones,

new shared mobility services, and urban logistics. His research interests include

transport policy and modeling, data science in transport, energy and transport, and

sustainable urban mobility.

About the Authors

Jeyhun I. Mikayilov

Jeyhun Mikayilov is a Research Fellow at KAPSARC and leads the AI for Energy

Demand Modeling and Forecasting project. His primary research interests include

applied time series econometrics, the economics of energy and environment, and

sustainable development. He holds a Ph.D. in Applied Mathematics.

Driving the Future: How Connected Autonomous Vehicles Reshape Travel and Energy Demand in Urban

and Rural Areas – An Empirical Modeling Approach

The project “An Integrated Transport Energy Demand Model for KSA” aims to develop

acomprehensive framework covering all transport sectors at both national and urban

levels, while also incorporating cross-border interactions with neighboring countries.

As a decision-support tool, it will enable policymakers and stakeholders to evaluate the

impacts of policy measures, infrastructure developments, and decarbonization strategies

on the transport sector.

About the Project

/kapsarcinfo@kapsarc.orgkapsarc.org