The effects of fuel prices and vehicle sales on fuel‐saving technology adoption in passenger vehicles

• Date: 01 July 2020
• Author: Thomas Klier,Joshua Linn,Yichen C. Zhou
• Published date: 01 July 2020
• DOI: http://doi.org/10.1111/jems.12384

J Econ Manage Strat. 2020;29:543–578. wileyonlinelibrary.com/journal/jems © 2020 Wiley Periodicals LLC

|

543

Received: 29 March 2019

|

Revised: 23 January 2020

|

Accepted: 20 May 2020

DOI: 10.1111/jems.12384

ORIGINAL ARTICLE

The effects of fuel prices and vehicle sales on fuel‐saving

technology adoption in passenger vehicles

Thomas Klier

1

|Joshua Linn

2

|Yichen C. Zhou

3

1

Federal Reserve Bank of Chicago,

Chicago, Illinois

2

Department of Agricultural and Resource

Economics, University of Maryland,

College Park, Maryland

3

John E. Walker Department of

Economics, Clemson University,

Clemson, South Carolina

Correspondence

Joshua Linn, Department of Agricultural

and Resource Economics, University of

Maryland, College Park, MD.

Email: linn@umd.edu

Abstract

Although economic theory suggests that both sales and fuel costs affect tech-

nology adoption by vehicle manufacturers, there is very little empirical evi-

dence on either effect. We document a strong connection between a vehicle's

sales and its energy efficiency. Using a demographics‐driven demand shifter to

isolate demand‐side changes in sales, we find that a one standard deviation

increase in sales raises efficiency by 0.2%, compared with a mean improvement

rate of 1.4% per year between 1997 and 2013. Higher fuel prices also increase

technology adoption directly by increasing willingness to pay for fuel cost

savings. The results have two implications: manufacturers will continue to

focus technological improvements on top selling vehicles; and fuel taxes will

have larger effects on technology adoption than fuel economy standards and

feebates.

1|INTRODUCTION

Improving vehicle fuel economy is a central part of worldwide efforts to reduce the risks of climate change. In the

United States, passenger vehicles account for about 15% of greenhouse gas emissions and half of transportation sector

emissions (IPCC, 2014).

Meeting near‐and long‐term emissions targets requires substantial technology adoption (Knittel, 2012). An ex-

tensive literature (e.g., Goldberg, 1995; Jacobsen, 2013; Klier and Linn, 2012; Reynaert, 2015; Roth, 2015) has examined

the welfare consequences of fuel economy and greenhouse gas standards for passenger vehicles. Klier and Linn (2016)

find that these standards have increased the rate of technology adoption. However, the vast majority of this literature

has assumed that technology adoption is exogenous and thus has missed a potentially important aspect of the

standards.

The literature suggests that several factors may affect technology adoption. According to NRC (2015), variable costs

of technology and resulting fuel savings affect adoption. Newell, Jaffe, and Stavins (1999) document the effects of

consumer demand and energy prices on innovation and technology adoption in air conditioners and other industries.

1

More broadly, the literature also suggests that a vehicle model's sales can affect adoption of technology through directed

technical change (Acemoglu, 2002), fixed costs of technology adoption with variable markups (Berry, Levinsohn, &

Pakes, 1995), or learning by doing in vehicle production.

In this paper, we focus on a manufacturer's decision to adopt fuel‐saving technology across its individual vehicle

models. We show that fuel costs and vehicle demand substantially affect this decision, and we discuss implications for

greenhouse gas emissions policy.

Motivated by the technology literature, we distinguish three demand‐side drivers of fuel‐saving technology adop-

tion. The first channel is the direct effect of fuel costs. A vehicle's fuel costs increase with the price of fuel and decrease

with fuel economy. An increase in fuel prices raises a consumer's willingness to pay for a given fuel economy

improvement (Klier & Linn, 2012), and thus the profitability of adopting fuel‐saving technology.

The second and third demand drivers operate via sales. As noted above, an increase in a vehicle's sales can cause

technology adoption because of fixed costs of adoption or other reasons. Therefore, a demand shock that increases a

vehicle's sales could also cause the adoption of fuel‐saving technology. Busse, Knittel, and Zettelmeyer (2013) and

Allcott and Wozny (2014) demonstrate that high gasoline prices raise the market shares of vehicles with high fuel

economy. Thus, an increase in fuel prices can raise sales of vehicles with high fuel economy, leading to more technology

adoption for those vehicles. We label this demand driver the indirect effect of fuel costs because it operates via sales.

The distinction between the direct and indirect effects is that the direct effect holds sales fixed. For the consumers

who would purchase a particular vehicle with low fuel prices, an increase in fuel prices raises their willingness to pay

for fuel‐saving technology, stimulating technology adoption. In addition, if that vehicle has high fuel economy, the

increase in fuel prices results in additional consumers wanting to purchase the vehicle. The increase in sales creates the

indirect effect. For example, an increase in gasoline prices raises fuel costs for all vehicles, such as the Toyota Prius.

Higher fuel costs raise the incentive for Toyota to make the Prius more efficient, representing the direct effect. The

gasoline price increase also raises the market share of the Prius, since it has relatively high fuel economy, increasing

sales and further incentivizing technology adoption. Thus, for the Prius the direct and indirect effects both encourage

more technology adoption. As we explain below, the two effects can work in the opposite direction for vehicles with low

fuel economy, such as the Toyota Tundra (a pickup truck).

The third driver includes any other demand shock that increases sales, besides fuel prices. We denote this effect as

the direct effect of sales. For example, the increase in demand for crossover vehicles during the 2000s and 2010s raised

sales of those vehicles. Adding fuel‐saving technology raises consumer willingness to pay for the vehicle, allowing the

firm to raise the vehicle's price. The greater the vehicle's sales before adoption, the greater the revenue increase, making

it more likely that manufacturers adopt technology when crossover demand is high than when it is low.

We use a novel identification strategy to estimate empirically the magnitudes of the three effects. We focus on fuel‐

saving technology for the internal combustion engine, including gasoline‐and diesel‐powered engines, and associated

transmissions. We use unique data on consumer demographics (consumer preferences for new vehicles), and vehicle‐

level characteristics. Notwithstanding the media attention around electric vehicles and other alternative technologies,

the internal combustion engine still accounted for about 99% of new vehicles sold in the United States in 2015 and

about 98% in 2018 (authors' calculations).

We begin the analysis by defining a vehicle's powertrain efficiency, which is our measure of fuel‐saving technology

and is distinct from the vehicle's fuel economy (miles per gallon [mpg]). For a given level of powertrain efficiency, a

manufacturer can trade off fuel economy, horsepower, and weight, analogously to movement along a production

possibilities frontier. By definition, when a manufacturer adds fuel‐saving technology, it shifts the frontier and can

increase fuel economy without affecting other characteristics. We define the increase in efficiency that results from

technology adoption as the increase in fuel economy that is feasible holding other vehicle characteristics constant. This

definition accounts for the possibility that manufacturers adopt fuel‐saving technology and use additional efficiency to

boost horsepower or increase weight. We estimate the powertrain efficiency of each vehicle model by model year from

1997 to 2013 similarly to Klier and Linn (2016).

Firms generally choose efficiency by deciding which of a set of existing technologies to install. Consequently, the

current or expected sales may affect the choice of efficiency. The main empirical challenge is the endogeneity of a

vehicle's sales. The endogeneity problem, which is common to nearly all empirical analysis of market‐driven techno-

logical change, arises from both potential reverse causality and omitted variable bias. Improving a vehicle's efficiency

may increase its demand, causing sales to increase and resulting in reverse causality. Furthermore, omitted supply

variables, such as a vehicle's production costs, can be correlated with both its sales and efficiency.

To address this challenge, we construct an instrumental variable (IV) that takes advantage of variation in consumer

demographics over time, combined with variation in purchasing behavior across consumer groups. The IV is a demand

shifter that captures changes in demand for a particular vehicle, relative to demand for other vehicles, which arise from

changes in consumer demographics over time. For example, larger households tend to purchase more minivans than

smaller households. The decrease in the share of large households in the United States over the sample period has

reduced demand for minivans relative to other market segments. To construct our instrument, we use consumer

preferences by demographic group that are measured at a specific point in time, combined with temporal variation in

demographics. Acemoglu and Linn (2004) and DellaVigna and Pollet (2007) have similarly used demographic trends as

exogenous determinants of sales in the pharmaceuticals and toys markets (other papers, such as Blundell, Griffith, &

544

|

KLIER ET AL.

Van Reenen, 1999), have used presample information to address endogeneity). The validity of the instrument rests on

(a) a positive correlation between the IV and actual sales; and (b) the exclusion restriction, that the IV affects tech-

nology only via sales. We document a strong positive correlation between the IV and actual sales, and provide strong

evidence supporting the exclusion restriction. Specifically, because consumer purchasing patterns are held fixed in

constructing the instrument, changes in supply‐side factors that affect purchasing patterns, such as changes of a

vehicle's position in product space, do not affect the instrument. Also, we show that the instrument isolates variation in

the potential demand for a vehicle, that is, uncorrelated with supply‐side factors that affect sales, such as imperfect

competition in nonprice vehicle attributes. In keeping with the recent literature (e.g., Acemoglu, Akcigit, Hanley, &

Kerr, 2016), we identify the effects of fuel costs on technology adoption assuming that fuel prices are exogenous to the

market.

Sales positively affect a vehicle's efficiency. A one standard deviation increase in sales, which corresponds to about a

10% increase, raises a vehicle's efficiency by 0.2%. This estimate constitutes a substantial and statistically significant

increase relative to the observed average annual efficiency increase of about 1.4% between 1997 and 2013. As an

alternative interpretation of the estimated magnitude, the sales effect implies annual consumer benefits of $100–400

million (2010 USD) for consumers who purchase high‐rather than low‐selling vehicles. In addition to sales, we test

whether a vehicle's efficiency responds to the efficiency of competing vehicles or to the manufacturer's stock of

efficiency‐related patents. We find some effects of competing vehicles, but these effects are less precisely estimated than

the primary effects of sales and fuel costs on efficiency. We also find that the main results are robust to alternative

functional forms and constructions of the instrument.

Fuel prices affect technology adoption both directly and indirectly via sales. After controlling for sales, fuel prices

have a positive and statistically significant effect on technology adoption. A fuel price increase causes average fuel

economy to increase for two reasons: first, by increasing markets shares of vehicles with high fuel economy (Busse

et al., 2013), and second, by inducing fuel‐saving technology adoption. It turns out that the estimated effects are similar

in magnitude, demonstrating the economic significance of the direct fuel price effect.

We illustrate the magnitudes of the sales and fuel cost effects using three sets of simulations. First, we compare the

indirect and direct channels through which fuel costs affect efficiency, by focusing on the gasoline price increase that

occurred between 2003 and 2007. The indirect effect of sales works as follows. The 80% increase in real gasoline prices

in that period raised the sales of vehicles with high fuel economy relative to vehicles with low fuel economy. In turn, the

changes in sales caused efficiency of the lowest‐fuel‐economy vehicles to be lower than they would have been if fuel

prices had remained at the low 2003 levels. Likewise, efficiency of the highest‐fuel‐economy vehicles was higher in 2007

than if fuel prices had remained at 2003 levels. In contrast, the direct fuel cost effect works in the opposite direction,

causing more technology adoption for low fuel economy vehicles. The increase in gasoline prices raises fuel costs

disproportionately more for low fuel economy vehicles than high fuel economy vehicles. This change raises the

willingness to pay for fuel cost savings more for low fuel economy vehicles than high fuel economy vehicles, causing

more technology adoption for low fuel economy vehicles. In the simulations, the direct fuel cost effect was about twice

as large as the indirect sales effect.

Second, we show that demographics affected the efficiency distribution across models in different market segments.

The overall shifts in demographics between 1980 and 2013 caused a shift in cumulative efficiency improvements away

from light‐duty trucks and toward cars. This effect occurred simultaneously with other demand‐or supply‐side effects

on relative efficiencies of cars and light trucks, such as changes in gasoline prices and fuel economy standards, which

affected consumer demand and manufacturer technology adoption.

Third, we find that changes in sales for crossovers and sport utility vehicles (SUVs) have affected technology

adoption. Between 2001 and 2004, per‐model sales of crossovers increased sharply and per‐model sales of SUVs

decreased sharply. The increase in crossover sales raised crossover efficiency and the decrease in SUV sales reduced

SUV efficiency, relative to a counterfactual in which sales remained unchanged.

The empirical results have two main implications for policies aiming to improve passenger vehicle fuel economy

and reduce greenhouse gas emissions. First, the strong sales effect implies that manufacturers will continue to improve

the efficiency of vehicles with internal combustion engines, which currently dominate the market. Most analysts argue

that vehicle electrification, combined with decarbonization of electricity generation, is necessary for substantially

reducing transportation greenhouse gas emissions. The sales effect will increase consumer demand for internal com-

bustion engines, and increase the challenges faced by alternative‐fuel vehicles to gain market share, compared to a

hypothetical scenario in which there is no sales effect for internal combustion engines. Existing welfare analyses of fuel

KLIER ET AL.

|

545