Langer, Ashley; Miller, Nathan H.

Working Paper
Automobile Prices, Gasoline Prices, and Consumer
Demand for Fuel Economy

EAG Discussion Paper, No. EAG 08-11

Provided in Cooperation with:

Economic Analysis Group (EAG), Antitrust Division, United States Department of Justice

Suggested Citation: Langer, Ashley; Miller, Nathan H. (2008) : Automobile Prices, Gasoline Prices, and
Consumer Demand for Fuel Economy, EAG Discussion Paper, No. EAG 08-11, U.S. Department of
Justice, Antitrust Division, Economic Analysis Group (EAG), Washington, DC

This Version is available at:

https://hdl.handle.net/10419/202377

Standard-Nutzungsbedingungen: Terms of use:

Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Documents in EconStor may be saved and copied for your personal
Zwecken und zum Privatgebrauch gespeichert und kopiert werden. and scholarly purposes.

Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle You are not to copy documents for public or commercial purposes, to
Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich exhibit the documents publicly, to make them publicly available on the
machen, vertreiben oder anderweitig nutzen. internet, or to distribute or otherwise use the documents in public.

Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen If the documents have been made available under an Open Content
(insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, Licence (especially Creative Commons Licences), you may exercise
gelten abweichend von diesen Nutzungsbedingungen die in der dort further usage rights as specified in the indicated licence.
genannten Lizenz gewährten Nutzungsrechte.
 ECONOMIC ANALYSIS GROUP
DISCUSSION PAPER

Automobile Prices, Gasoline Prices, and

Consumer Demand for Fuel Economy

By

Ashley Langer* and Nathan Miller**

EAG 08-11 December 2008

EAG Discussion Papers are the primary vehicle used to disseminate research from economists in the
Economic Analysis Group (EAG) of the Antitrust Division. These papers are intended to inform interested
individuals and institutions of EAG’s research program and to stimulate comment and criticism on
economic issues related to antitrust policy and regulation. The analysis and conclusions expressed herein
are solely those of the authors and do not represent the views of the United States Department of Justice.
Information on the EAG research program and discussion paper series may be obtained from Russell
Pittman, Director of Economic Research, Economic Analysis Group, Antitrust Division, U.S. Department
of Justice, BICN 10-000, Washington, DC 20530, or by e-mail at [email protected]. Comments
on specific papers may be addressed directly to the authors at the same mailing address or at their e-mail
address.
Recent EAG Discussion Paper titles are listed at the end of this paper. To obtain a complete list of titles or
to request single copies of individual papers, please write to Janet Ficco at the above mailing address or at
[email protected]. In addition, recent papers are now available on the Department of Justice website
at http://www.usdoj.gov/atr/public/eag/discussion_papers.htm. Beginning with papers issued in 1999,
copies of individual papers are also available from the Social Science Research Network at www.ssrn.com.
_________________________
* Department of Economics, University of California, Berkeley. Email: [email protected].

** Economist, Economic Analysis Group, Antitrust Division, U.S. Department of Justice. Email:
[email protected]. We thank Severin Borenstein, Joseph Farrell, Richard Gilbert, Joshua Linn,
Kenneth Train, Clifford Winston, Catherine Wolfram, and seminar participants at the George Washington
University and the University of California, Berkeley for valuable comments. Daniel Seigle and Berk
Ustun provided research assistance. The views expressed are not purported to reflect those of the United
States Department of Justice.
 Abstract
The relationship between gasoline prices and the demand for vehicle fuel efficiency is im-
portant for environmental policy but poorly understood in the academic literature. We provide
empirical evidence that automobile manufacturers price as if consumers respond to gasoline
prices. We derive a reduced-form regression equation from theoretical micro-foundations and
estimate the equation with nearly 300,000 vehicle-week-region observations over the period 2003-
2006. We find that vehicle prices generally decline in the gasoline price. The decline is larger
for inefficient vehicles, and the prices of particularly efficient vehicles actually rise. Structural
estimation that ignores these effects underestimates consumer preferences for fuel efficiency.
1 Introduction
The combustion of gasoline in automobiles poses some of the most pressing policy concerns of
the early twenty-first century. This combustion produces carbon dioxide, a greenhouse gas that
contributes to global warming. It also limits the flexibility of foreign policy – more than sixty
percent of U.S. oil is imported, often from politically unstable regimes. These effects are classic
externalities. It is not clear whether, in the absence of intervention, the market is likely to produce
efficient outcomes.
One topic of particular importance for policy in this arena is the extent to which retail gasoline
prices influence the demand for vehicle fuel efficiency. If, for example, higher gasoline prices induce
consumers to shift toward more fuel efficient vehicles, then 1) the recent run-up in gasoline prices
should partially mitigate the policy concerns outlined above and 2) gasoline and/or carbon taxes
may be reasonably effective policy instruments. However, a small empirical literature estimates an
inelastic consumer response to gasoline prices (e.g., Goldberg 1998; Bento et al 2005; Li, Timmins
and von Haefen 2007; Jacobsen 2008). For example, Shanjun Li, Christopher Timmins, and Roger
H. von Haefen conclude that:

[H]igher gasoline prices do not deter American’s love affair with large, relatively fuel-
inefficient vehicles. Moreover, a politically feasible gasoline tax increase will likely not
generate significant improvements in fleet fuel economy.

Interestingly, the findings of the academic literature are seemingly contradicted by a bevy of recent
articles in the popular press. Consider Bill Vlasic’s article in the New York Times titled “As Gas
Costs Soar, Buyers Flock to Small Cars”:

Soaring gas prices have turned the steady migration by Americans to smaller cars into
a stampede. In what industry analysts are calling a first, about one in five vehicles sold
in the United States was a compact or subcompact car...1

One might be tempted to point out that four in five vehicles were neither compact nor subcompact
cars. But suppose that the spirit of the article is correct. Can its perspective be reconciled with
the academic literature?
We approach the topic from a new perspective. We ask the question: “Do automobile
manufacturers behave as if consumers respond to gasoline price?” Our approach starts with the
observation that consumer choices have implications for equilibrium automobile prices: if gasoline
price shocks affect consumer choices then one should see corresponding adjustments in automobile
prices. We derive the specific form of these adjustments from theoretical micro foundations. In
particular, we show that a change in the gasoline price affects an automobile’s equilibrium price
1
The article appeared on May 2, 2008. Other recent press articles include CNN.com’s May 23, 2008 article titled
“SUVs plunge toward ‘endangered’ list,” the LA Times’ April 24, 2008 article titled “Fueling debate: Is $4.00 gas
the death of the SUV?” and the Chicago Tribune’s May 12, 2008 article titled “SUVs no longer king of the road.”
through two main channels: its effect on the vehicle’s fuel cost and its effect on the fuel cost of the
vehicle’s competitors.2
To build intuition, consider the effect of an adverse gasoline price shock on the price of an
arbitrary automobile. If consumers respond to the vehicle’s fuel cost, then the gasoline price shock
should reduce demand for the automobile. However, the gasoline price shock also increases the
fuel cost of the automobile’s competitors and should therefore increase demand through consumer
substitution. The net effect on the automobile’s equilibrium price is ambiguous. Overall, the
theory suggests that the net price effect should be negative for most automobiles, but positive for
automobiles that are sufficiently more fuel efficient then their competitors. We believe that the
framework is quite intuitive. For example, the theory formalizes the idea that an adverse gasoline
price shock should reduce demand for fuel inefficient automobiles (e.g., the GM Suburban) more
than demand for fuel efficient automobiles (e.g., the Ford Taurus), and that demand for highly fuel
efficient automobiles (e.g., the Toyota Prius) may actually increase.
In the empirical implementation, we test the extent to which automobile prices respond to
changes in fuel costs. We use a comprehensive set of manufacturer incentives to construct region-
time-specific “manufacturer prices” for each of nearly 700 vehicles produced by GM, Ford, Chrysler,
and Toyota over the period 2003-2006, and combine information on these vehicles’ attributes with
data on retail gasoline prices to measure fuel costs. We then regress manufacturer prices on fuel
costs and competitor fuel costs and argue that manufacturers set prices as if consumers respond
to the gasoline price if the first coefficient is negative while the second is positive. Overall, the
estimation procedure uses information from nearly 300,000 vehicle-week-region observations; as we
discuss below, identification is feasible even in the presence of vehicle, time, and region fixed effects.
By way of preview, the results are consistent with a strong and statistically significant con-
sumer response to the retail price of gasoline. Manufacturer prices decrease in fuel costs but increase
in the fuel costs of competitors. The median net manufacturer price change in response to a hypo-
thetical one dollar increase in gasoline prices is a reduction of $792 for cars and a reduction of $981
for SUVs; the median price change for trucks and vans are modest and less statistically significant.
Although the fuel cost effect almost always dominates the competitor fuel cost effect, the manu-
facturer prices of some particularly fuel efficient vehicles do increase (e.g., the 2006 Prius or the
2006 Escape Hybrid). The manufacturer responses that we estimate are large in magnitude. Rough
back-of-the-envelope calculations suggest that, for most vehicles, manufacturers substantially offset
the discounted future gasoline expenditures incurred by consumers.
The results have important policy implications. The manufacturer price responses that we
2
By “fuel cost” we mean the fuel expense associated with driving the vehicle. Notably, changes in the gasoline
price affect the fuel costs of automobiles differentially – the fuel costs of inefficient automobiles are more responsive
to the gasoline prices than the fuel costs of efficient automobiles. One can imagine that the gasoline price may affect
equilibrium automobile prices through other channels, perhaps due to an income effect and/or changes in production
costs. Our empirical framework allows us to control directly for these alternative channels; we find that their net
effect is small.

2
document should dampen short-run changes in consumer purchase behavior by subsidizing relatively
fuel inefficient vehicles when gasoline prices rise. Structural estimation that fails to control for
these manufacturer responses may therefore underestimate the short-run elasticity of demand with
respect to gasoline prices. Further, counter-factual policy simulations based on such estimation are
likely to understate the effects of gasoline prices or gasoline/carbon taxes, even if the simulations
allow for appropriate manufacturer responses.3
Thus, the evidence presented here may help reconcile the academic literature with the per-
spective of the popular press. That is, consumers may consider gasoline prices when choosing which
automobile to purchase but, due to the manufacturer price response, changes in the gasoline price
are not fully reflected in observed vehicle purchases. We speculate that a major effect of gasoline
price changes (or gasoline/carbon taxes) may occur in the long-run. The manufacturer responses
that we estimate reduce the profit margins of fuel inefficient vehicles relative to those of fuel efficient
vehicles. It is possible, therefore, that increases in the gasoline price (or in the gasoline/carbon tax)
provide a substantial profit incentive for manufacturers to invest in the development and marketing
of fuel efficient vehicles.
The paper proceeds as follows. We lay out the empirical model in Section 2, including the
underlying theoretical framework and the empirical implementation. We describe the data and
regression variables in Section 3. Then, in Section 4, we present the main regression results and
discuss a number of extensions related to historical and futures gasoline prices, pricing dynamics,
selected demand and cost factors, and manufacturer inventory levels. We conclude in Section 5.

2 The Empirical Model

2.1 Theoretical framework
We derive our estimation equation from a model of Bertrand-Nash competition between multi-
vehicle manufacturers. Specifically, we model manufacturers, = = 1, 2, . . . F that produce vehicles
j = 1, 2, . . . Jt in period t. Each manufacturer chooses prices that maximize their short-run profit
over all of their vehicles:
X
π=t = [(pjt − cjt ) ∗ qjt − fjt ] (1)
j∈=

where for each vehicle j and period t, the terms pjt , cjt , and qjt are the manufacturer price, the
marginal cost, and the quantity sold respectively; the term fjt is the fixed cost of production. As
we detail in the empirical implementation, we assume that marginal costs are constant in quantity
but responsive to certain exogenous cost shifters.4
We pair this profit function with a consumer demand function that depends on manufacturer
3
We formalize this argument in Appendix A.
4
We abstract from the manufacturers’ selections of vehicle attributes and fleet composition, as well as any entry
and/or exit, which we deem to be more important in longer-run analysis.

3
prices, expected lifetime fuel costs, and exogenous demand shifters that capture vehicle attributes
and other factors. We specify a simple linear form:

Jt
X
q(pjt ) = αjk (pkt + xkt ) + µjt , (2)
k=1

where the term αjk is a demand parameter and the terms xkt and µjt capture the fuel costs
and the exogenous demand shifters, respectively. One can conceptualize the demand shifters as
including the vehicle’s fixed attributes and quality, as well as maintenance costs and any other
expenses that are unrelated to the gasoline price. We consider the case in which demand is well
defined (∂qjt /∂pjt = αjj < 0) and vehicles are substitutes (∂qjt /∂pkt = αjk ≥ 0 for k 6= j). The
equilibrium manufacturer prices in each period are then characterized by Jt first-order conditions:

∂π=t X X
= αjk (pkt + xkt ) + µjt + αkj (pkt − ckt ) = 0. (3)
∂pjt
k k∈=

We solve these first-order equations for the equilibrium manufacturer prices as a function of the
exogenous factors.5 The resulting manufacturer “price rule” is a linear function of the fuel costs,
marginal costs, and demand shifters:

X X
p∗jt = φ1jt xjt + φ2jkt xkt + φ3jlt xlt
k∈=
/ l∈=, l6=j
X¡ ¢ X ¡ ¢
+ φ4jt cjt + φ5jt µjt + φ6jkt ckt + φ7jkt µkt + φ8jlt clt + φ9jlt µlt . (4)
k∈=
/ l∈=, l6=j

The coefficients φ1 , φ2 , . . . , φ9 are nonlinear functions of all the demand parameters. The price rule
makes it clear that the equilibrium price of a vehicle depends on its characteristics (i.e, its fuel cost,
marginal cost, and demand shifters), the characteristics of vehicles produced by competitors, and
the characteristics of other vehicles produced by the same manufacturer.6 For the time being, we
collapse the second line of the price rule into a vehicle-time-specific constant, which we denote γjt .
The sheer number of terms in Equation 4 makes direct estimation infeasible. With only Jt
observations per period, one cannot hope to identify the Jt2 fuel cost coefficients, let alone the
5
The solution technique is simple. Turning to vector notation, one can rearrange the first-order conditions such
that Ap = b, where A is a Jt × Jt matrix of demand parameters, p is a Jt × 1 vector of manufacturer prices, and b is
a Jt × 1 vector of “solutions” that incorporate the fuel costs, marginal costs, and demand shifters. Provided that the
matrix A is nonsingular, Cramer’s Rule applies and there exists a unique Nash equilibrium in which the equilibrium
manufacturer prices are linear functions of all the fuel costs, marginal costs, and demand shifters.
6
We divide the terms into these three groups because Equation (3) can be rewritten:
X X
αjj (2pjt + xjt − cjt ) + αjk (pkt + xkt ) + (αjl (plt + xlt ) + αlj (plt − clt )) = 0,
k∈=
/ l∈=, l6=j

in which each group has a distinctly different functional form.

4
vehicle-time-specific constant. We move toward the empirical implementation by re-expressing the
price rule in terms of weighted averages:
X X
p∗jt = φ1jt xjt + φ2jt 2
ωjkt xkt + φ3jt 3
ωjlt xlt + γjt , (5)
k∈=
/ l∈=, l6=j

2
where the weights ωjkt 3 both sum to one in each period.7 Thus, the equilibrium price
and ωjlt
depends on its fuel cost, the weighted average fuel cost of vehicles produced by competitors, and the
weighted average fuel cost of vehicles produced by the same manufacturer. Under a mild regularity
condition that we develop in Appendix B, the equilibrium manufacturer price of a vehicle decreases
in its fuel cost (i.e, φ1jt ∈ [−1, 0]) and increases in the weighted average fuel cost of vehicles produced
by competitors (i.e., φ2jt ∈ [0, 1]). Further, the equilibrium price of a vehicle is more responsive to
changes in its fuel cost than identical changes to the weighted average fuel cost of its competitors
(i.e., |φ1jt | > |φ2jt |). The relationship between the equilibrium price of a vehicle and the weighted
average fuel cost of vehicles produced by the same manufacturer is ambiguous (i.e., φ3jt ∈ [−1, 1]).8
The intuition that manufacturer prices can increase or decrease in response to adverse gasoline
price shocks can now be formalized. Assume for the moment that the gasoline price does not
affect marginal costs or the demand shifters, and therefore does not affect the vehicle-time-specific
constant (we relax this assumption in an extension). Denoting the gasoline price at time t as gpt ,
the effect of the gasoline price shock on the manufacturer price is:

∂p∗jt ∂xjt X X
2 ∂xkt ∂xlt
= φ1j + φ2jt ωjkt + φ3jt 2
ωjlt , (6)
∂gpt ∂gpt ∂gpt ∂gpt
k∈=
/ l∈=, l6=j

where fuel costs increase unequivocally in the gasoline price (i.e., ∂xjt /∂gpt > 0 ∀ j). The first
term captures the intuition that manufacturers partially offset an increase in the fuel cost with a
reduction in the vehicle’s price. This reduction is greater for vehicles whose fuel costs are sensitive
to the gasoline price (e.g., for fuel-inefficient vehicles). The second and third terms capture the
intuition that an increases in the fuel costs of other vehicles can increase demand (e.g., through
consumer substitution) and thereby raise the equilibrium price. Although the first effect tends to
dominate, prices can increase provided that the vehicle is sufficiently more fuel efficient than other
vehicles.9
7 i
The weights have specific analytical solutions given by ωjkt = φijkt /φijt for i = 2, 3, so that closer competitors
2 3
receive greater weight.PThe coefficients φjt and φjt are the sums of the φ2jkt and φ3jkt coefficients, respectively.
Mathematically, φijt = φijkt for i = 2, 3.
8
As we show in Appendix B, if demand is symmetric (i.e., αjk = αkj ∀ j, k), then changes in the fuel costs of
other vehicles produced by the same manufacturer have no effect on equilibrium prices, and φ3jt = 0.
9
The exact condition for a price increase is:
0 1
∂xjt 1 X 2 ∂xkt X ∂xlt A
< − 1 @φ2jt ωjkt + φ3jt 2
ωjlt .
∂gpt φjt ∂gpt ∂gpt
k∈=
/ l∈=, l6=j

5
2.2 Empirical implementation
Our starting point for estimation is the reduced-form outlined in Equation 5. The empirical im-
i
plementation requires that we specify the fuel costs (xjt ), the weights (ωjkt for i = 2, 3), and the
vehicle-time-specific constants (γjt ). We discuss each in turn.
We proxy the expected lifetime fuel cost of vehicle j at time t as a function of the vehicle’s
fuel efficiency and the gasoline price at time t, following Goldberg (1998), Bento et al (2005) and
Jacobsen (2007). The specific form is:

gpt
xjt = τ ∗ ,
mpgj

where mpgj is the fuel efficiency of vehicle j in miles-per-gallon and τ is a discount factor that
nests any form of multiplicative discounting; one specific possibility is τ = 1/(1 − δ), where δ is the
“per-mile discount rate.”10 The fuel cost proxy is precise if consumers perceive the gasoline price
to follow a random walk because, in that case, the current gasoline price is a sufficient statistic for
expectations over future gasoline prices. As we discuss below, we fail to reject the null hypothesis
that gasoline prices actually follow a random walk, but also provide some evidence that consumers
consider both historical gasoline prices and futures prices when forming expectations.
To construct the weighted average variables, we assume that the severity of competition
between two vehicles decreases in the Euclidean distance between their attributes. To that end,
we take a set of M vehicle attributes, denoted zjm ; m = 1, . . . , M , and standardize each to have
a variance of one. Then, for each pair of vehicles, we sum the squared differences between each
attribute to calculate the effective “distance” in attribute space. We form initial weights as follows:

∗ 1
ωjk = PM .
2
m=1 (zjm − zkm )

To form the final weights that we use in estimation, we first set the initial weights to zero for vehicles
of different types and then normalize the weights to sum to one for each vehicle-period. We perform
this weighting procedure separately for vehicles produced by the same manufacturer and vehicles
produced by competitors; the result is a set of empirical weights that we denote ω 2 and ω
ejkt 3 .11
ejkt
The use of weights based on the Euclidean distance between vehicle attributes is analogous to the
We say that manufacturer prices tend to fall in the gasoline price because |φ1jt | > |φ2jt | and φ3jt ≈ 0 provided that
demand that is approximately symmetric.
10
It may help intuition to note that the ratio of the gasoline price to vehicle miles-per-gallon is simply the gasoline
expense associated with a single mile of travel.
11
Thus, the weighting scheme is based on the inverse Euclidean distance between vehicle attributes among vehicles
of the same type. There are four vehicle types in the data: cars, SUVs, trucks and vans. We use the following set of
vehicle attributes in the initial weights: manufacturer suggested retail price (MSRP), miles-per-gallon, wheel base,
horsepower, passenger capacity, and dummies for the vehicle type and segment. Although the initial weights are
constant across time for any vehicle pair, the final weights may vary due to changes in the set of vehicles available on
the market. An alternative weighting scheme based on the inverse Euclidean distance of all vehicles (not just those
of the same type) produces similar results.

6
instrumenting procedures of Berry, Levinsohn, and Pakes (1995) and Train and Winston (2007).
Turning to the vehicle-time-specific constants, recall from that Equation (4) that the con-
stants represent the net price effects of marginal costs and demand-shifters:
X¡ ¢ X ¡ ¢
γjt = φ4j cjt + φ5j µjt + φ6jk ckt + φ7jk µkt + φ8jl clt + φ9jl µlt .
k∈=
/ l∈=, l6=j

In the empirical implementation, we decompose this function using vehicle fixed effects, time fixed
effects, and controls for the number of weeks that each vehicle has been on the market. Let λjt
denote the number of weeks that vehicle j has been on the market as of period t, and λ̄A,t denote
the weighted average number of weeks since the vehicles in the set A were first produced. The
decomposition takes the form:

γjt = δt + κj + f (λjt ) + g(λ̄k∈=,t

/ ) + h(λ̄k∈=, k6=j,t ) + ²jt

where δt and κj are time and vehicle fixed effects, respectively, and functions f , g, and h flexibly
capture the net price effects of learning-by doing and predictable demand changes over the model-
year.12 In the main results, we specify the functions f , g, and h as third-order polynomials; the
results are robust to the use of higher-order or lower-order polynomials. The error term ²jt captures
vehicle-time-specific cost and demand shocks.
Two final adjustments produce the main regression equation that we take to the data. First,
we incorporate regional variation in manufacturer prices and gasoline prices and add a correspond-
ing set of region fixed effects.13 Second, we impose a homogeneity constraint that reduces the
total number of parameters to be estimated; the constraint eliminates vehicle-time variation in the
coefficients, so that φijt = φi ∀ j, t (in supplementary regressions we permit the coefficients to vary
across manufacturers and vehicle types). The regression equation is:

gptr X X
3 gptr gptr
pjtr = β 1 + β2 ω
ejkt + β3 ω 2
ejlt
mpgj mpgk mpgl
k∈=
/ l∈=, l6=j

+ f (λjt ) + g(λ̄k∈=,t
/ ) + h(λ̄k∈=, k6=j,t ) + δt + κj + ηr + ²jt , (7)

where the fuel cost coefficients incorporate the discount factor, i.e., β i = τ φi for i = 1, 2, 3; for
reasonable discount factors, these coefficients should be much larger than one in magnitude. Thus,
we estimate the average response of a vehicle’s price to changes in its fuel costs, changes in the
weighted average fuel cost among vehicles produced by competitors, and changes in the weighted
average fuel cost among other vehicles produced by the same manufacturer.
12
Copeland, Dunn and Hall (2005) document that vehicles prices fall approximately nine percent over the course
of the model-year.
13
Adding regional variation in prices does not complicate the weight calculations because there is no regional
variation in the vehicles available to consumers.

7
 We estimate Equation 7 using ordinary least squares. We are able to identify the fuel cost
coefficients in the presence of time, vehicle, and region fixed effects precisely because changes in
the gasoline price across time and regions affects manufacturer prices differentially across vehicles.
We argue that manufacturers price as if consumers respond to gasoline prices if the fuel cost
coefficient is negative (i.e., β 1 < 0) and the competitor fuel cost coefficient is positive (i.e., β 2 > 0).
The theoretical results suggest that the fuel cost coefficient should be larger in magnitude than
¯ ¯ ¯ ¯
the competitor fuel cost coefficient (i.e., ¯β 1 ¯ > ¯β 2 ¯); more generally, the relative magnitude of
these coefficients determines the extent to which average manufacturer prices fall in response to
an adverse gasoline shock. We cluster the standard errors at the vehicle level, which accounts for
arbitrary correlation patterns in the error terms.14

3 Data Sources and Regression Variables

3.1 Data sources
Our primary source of data is Autodata Solutions, a marketing research company that maintains
a comprehensive database of manufacturer incentive programs. We have access to the programs
offered by Toyota and the “Big Three” U.S. manufacturers – GM, Ford, and Chrysler – over the
period 2003-2006.15 There are just over 190,000 cash incentive-vehicle pairs in the data. Each lasts
a fixed period of time, and provides cash to consumers (“consumer-cash”) or dealerships (“dealer-
cash”) at the time of purchase.16 The incentive programs may be national, regional, or local in
their geographic scope; we restrict our attention to the national and regional programs.17 Thus, we
are able to track how manufacturer incentives change over time and across regions for each vehicle
in the data.
By “vehicle,” we mean a particular model in a particular model-year. For example, the 2003
Ford Taurus is one vehicle in the data, and we consider it as distinct from the 2004 Ford Taurus.
Overall, there are 681 vehicles in the data – 293 cars, 202 SUVs, 105 trucks, and 81 vans. The
data have information on the attributes of each, including MSRP, miles-per-gallon, horsepower,
wheel base, and passenger capacity.18 We impute the period over which each vehicle is available to
14
The results are robust to the use of brand-level or segment-level clusters. Brands and vehicle segments are finer
gradations of the manufacturers and vehicle types, respectively. There are 21 brands and 15 vehicle segments in the
data. Examples of brands (and their manufacturer) include Chevrolet (GM), Dodge (Chrysler), Mercury (Ford), and
Lexus (Toyota). Examples of vehicle segments include compact cars, luxury SUVs, and large pick-ups. The results
are also robust to the use of manufacturer and vehicle type clusters, though the small number of manufacturers and
vehicle types makes the asymptotic consistency of the standard errors suspect.
15
The German manufacturer Daimler owned Chrysler over this period. We exclude Mercedes-Benz from this
analysis since it is traditionally associated with Daimler rather than Chrysler.
16
Consumer cash includes both “Stand-Alone Retail Cash” and “Bonus Cash.”
17
We consider an incentive to be regional if it is available across an entire Energy Information Agency region. We
exclude incentives that are available in only a single city or state.
18
Attributes sometimes differ for a given vehicle due to the existence of different option packages, also known as
“trim.” When more than one set of attributes exists for a vehicle, we use the attributes corresponding to the trim

8
consumers as beginning with the start date of production, as given in Ward’s Automotive Yearbook,
and ending after the last incentive program for that vehicle expires.19 For each vehicle, we construct
observations over the relevant period at the week-region level.
We combine the Autodata Solutions data with information from the Energy Information
Agency (EIA) on weekly retail gasoline prices in each of five distinct geographic regions. The
EIA surveys retail gasoline outlets every Monday for the per gallon pump price paid by consumers
(inclusive of all taxes).20 In addition to the regional measures, the EIA calculates an average
national price. Figure 1 plots these retail gasoline prices over 2003-2006 (in real 2006 dollars).
A run-up in gasoline prices over the sample period is apparent. For example, the mean national
gasoline price is 1.75 dollars-per-gallon in 2003 and 2.57 dollars-per-gallon in 2006. The sharp
upward spike around September 2005 is due to Hurricane Katrina, which temporarily eliminated
more than 25 percent of US crude oil production and 10-15 percent of the US refinery capacity
(EIA 2006). Although gasoline prices tend to move together across regions, we are able to exploit
limited geographic variation to strengthen identification.
We purge the gasoline prices of seasonality prior to their use in the analysis. Since automobile
manufacturers adjust their prices cyclically over vehicle model-years (e.g., Copeland, Hall, and
Dunn 2005), the presence of seasonality in gasoline prices is potentially confounding. Further, the
use of time fixed effects alone may be insufficient in dealing with seasonality because gasoline prices
affect the fuel costs of each vehicle differentially (e.g., Equation 7). We employ the X-12-ARIMA
program, which is state-of-the-art and commonly employed elsewhere, for example by the Bureau of
Labor Statistics to deseasonalize inputs to the consumer price index.21 Figure 2 plots the resulting
deseasonalized national gasoline prices together with the seasonal adjustments. As shown, the
program adjusts the gasoline price downward during the summer months and upwards during the
winter months. The magnitude of the adjustments increases with gasoline prices.
In an extension (presented in Section 4.2), we explore whether consumers consider historical
and futures prices when forming expectations about future gasoline prices. Interestingly, statistical
tests based on Dicky and Fuller (1979) fail to reject the null that gasoline prices follow a random
walk – the p-statistic for the deseasonalized national time-series is 0.7035 and the p-statistics for the
deseasonalized regional time-series are similar. These tests suggest that knowledge of the current
with the lowest MSRP.
19
The start date of production is unavailable for some vehicles. For those cases, we set the start date at August 1
of the previous year. For example, we set the start date of the 2006 Civic Hybrid to be August 1, 2005. We impose
a maximum period length of 24 months. In robustness checks, we used an 18 month maximum; the different period
lengths did not affect the results.
20
The survey methodology is detailed online at the EIA webpage. The regions include the East Coast, the Gulf
Coast, the Midwest, the Rocky Mountains, and the West Coast.
21
We use data on gasoline prices over 1993-2008 to improve the estimation of seasonal factors, and adjust each
national and regional time-series independently. We specify multiplicative decomposition, which allows the effect of
seasonality to increase with the magnitude of the trend-cycle. The results are robust to log-additive and additive
decompositions. For more details on the X-12-ARIMA, see Makridakis, Wheelwright and Hyndman (1998) and Miller
and Williams (2004).

9
gasoline price is sufficient to inform predictions over future gasoline prices. The result is consistent
with the academic literature and statements of industry experts. For example, Alquist and Kilian
(2008) find that the current spot price of crude oil outperforms sophisticated forecasting models
as a predictor of future spot prices, and Peter Davies, the chief economist of British Petroleum,
has stated that “we cannot forecast oil prices with any degree of accuracy over any period whether
short or long...” (Davies 2007). If consumers form expectations efficiently, therefore, one would
not expect historical and/or futures prices of gasoline to influence vehicle purchase decisions.

3.2 Regression variables

The two critical variables that enable regression analysis are manufacturer price and fuel cost.
We discuss each in turn. To start, we measure the manufacturer price of each vehicle as MSRP
minus the mean incentive available for the given week and region. We also show results in which
the variable includes only regional incentives and only national incentives, respectively. From an
econometric standpoint, the MSRP portion of the variable is irrelevant for estimation because the
vehicle fixed effects are collinear (MSRP is constant for all observations on a given vehicle). It
is the variation in manufacturer incentives across vehicles, weeks, and regions that identifies the
regression coefficients.
At least two important caveats apply to our manufacturer price variable. First, the variable
does not capture any information about final transaction prices, which are negotiated between the
consumers and the dealerships. Changes in negotiating behavior could dampen or accentuate the
effect we estimate between gasoline prices and manufacturer prices. Second, although we observe
the incentive programs, we do not observe the actual incentives selected. In some circumstances,
it is possible that consumers may stack multiple incentives or choose between different incentives.
To the extent that manufacturers are more lenient in allowing consumers to stack incentives when
gasoline prices are high, our regression estimates are conservative relative to the true manufacturer
response.22
We measure the fuel costs of each vehicle as the gasoline price divided by the miles-per-
gallon of the vehicle. As discussed above, this has the interpretation of being the gasoline expense
associated with a single mile of travel. Since the gasoline price varies at the week and region levels
and miles-per-gallon varies at the vehicle level, fuel costs vary at the vehicle-week-region level. In
an extension, we construct alternative fuel costs based on 1) the mean of the gasoline price over the
previous four weeks and 2) the price of one-month futures contract for retail gasoline. The futures
data are derived from the New York Mercantile Exchange (NYMEX) and are publicly available
from the EIA.23 The alternative variables permit tests for whether consumers are backward-looking
22
To check the sensitivity of the results, we construct a number of alternative variables that measure manufacturer
prices: 1) MSRP minus the maximum incentive, 2) MSRP minus the mean consumer-cash incentive, 3) MSRP minus
the mean dealer-cash incentive, and 4) MSRP minus the mean publicly available incentive. None of these alternative
dependent variables substantially change the results.
23
We use one-month futures contracts for reformulated regular gasoline at the New York harbor. In order to ensure

10
and forward-looking, respectively.
Table 1 provides means and standard deviations for the manufacturer price and the gasoline
price variables, as well as for five vehicle attributes used in the weighting scheme – MSRP, miles-
per-gallon, horsepower, wheel base, and passenger capacity. The statistics are calculated from the
299,855 vehicle-region-week observations formed from the 681 vehicles, 208 weeks, and five regions
in the data. As shown, the mean manufacturer price is 30.344 (in thousands). The mean fuel cost is
0.108, so that gasoline expenses average roughly eleven cents per mile. The means of MSRP, miles-
per-gallon, horsepower, wheel base, and passenger capacity are 30.782, 21.555, 224.123, 115.193,
and 4.911, respectively.
Table 2 shows the means of these variables, calculated separately for each vehicle type. On
average, cars are less expensive than SUVs but more expensive than trucks and vans. The mean
manufacturer price for the four vehicle types are 30.301, 35.301, 24.482, and 24.658, respectively.
Cars also require far less gasoline expense per mile. The mean fuel cost of 0.087 is nearly thirty
percent smaller than the means of 0.121, 0.133, and 0.120 for SUVs, trucks, and vans, respectively.
The means of the attributes used in the weights also differ across type, and reflect the generalization
that cars are smaller, more fuel efficient, and less powerful than SUVs, trucks, and vans. Of course,
the vehicles also differ along unobserved dimensions. We use vehicle fixed effects to control for all
these differences – observed and unobserved – in our regression analysis.

4 Empirical Results
4.1 Main regression results
We regress manufacturer prices on fuel costs, as specified in Equation 7. To start, we impose the
full homogeneity constraint that all vehicles share the same fuel cost coefficients. The estimated
coefficients are the average response of manufacturer prices to fuel costs. Table 3 presents the re-
sults. In Column 1, we use the baseline manufacturer price – MSRP minus the mean of the regional
and national incentives. In Columns 2 and 3, we use MSRP minus the mean regional incentive and
MSRP minus the mean national incentive, respectively. Although the first column may provide
more meaningful coefficients, we believe that the second and third columns are interesting insofar
as they examine whether manufacturers respond at the regional and national levels, respectively.
As shown, the fuel cost coefficients of -55.40, -56.96, and -63.75 are precisely estimated and
capture the intuition that manufacturers adjust their prices to offset changes in fuel costs. The
competitor fuel cost coefficients of 50.76, 50.16, and 50.09 are also precisely estimated and support
the idea that increases in competitors’ fuel costs raise demand due to consumer substitution. In
each regression, the magnitude of the fuel cost coefficient exceeds that of the competitor fuel cost
that the regression coefficients are easily comparable, we normalize the futures price to have the same global mean
over the period as the national retail gasoline price.

11
coefficient, which is suggestive that the first effect dominates for most vehicles.24 We make this
more explicit shortly. The same-firm fuel cost coefficients are nearly zero and not statistically
significant.25 Finally, a comparison of coefficients across columns suggests that manufacturers
adjust their prices similarly at the regional and national levels in response to changes in fuel
costs.26
We explore the effect of retail gasoline prices on manufacturer prices in Figure 3. The gasoline
price enters through the fuel costs, average competitor fuel costs, and average same-firm fuel costs.
We calculate the effect of a one dollar increase in the gasoline price for each vehicle-week-region
observation:
X ω 2 X ω 2
∂pjrt βb1 ejkt ejkt
= + βb2 + βb3 .
∂gprt mpgj mpgk mpgk
k6=j k6=j

We plot these derivatives (in thousands) on the vertical axis against vehicle miles-per-gallon on
the horizontal axis. We focus on the first dependent variable, i.e., MSRP minus the mean regional
and national incentive.27 The median effect of a one dollar increase in the gasoline price per gallon
is a reduction in the manufacturer price of $171. The calculation varies greatly across vehicles –
for example, the effects range from a reduction of $1,506 for the 2005 GM Montana SV6 to a rise
of $998 for the 2006 Toyota Prius. Although the manufacturer price drops for 83 percent of the
vehicles, the price response for fuel efficient vehicles tends to be less negative, and the prices of
extremely fuel efficient vehicles such as hybrids actually increase. Overall, the own fuel cost effect
dominates the competitor fuel cost effect for most vehicles; the converse is true only for vehicles
that are substantially more fuel efficient than their competitors.
We use sub-sample regressions to relax the homogeneity constraint that all vehicles share the
same fuel cost coefficients. In particular, we regress manufacturer prices on the fuel cost variables
for each combination of vehicle type (cars, SUVs, trucks, and vans) and manufacturer (GM, Ford,
Chrysler, and Toyota). The sub-sample regressions may be informative, for example, if the market
for cars is more (or less) competitive than the market for SUVs, if region- and time-specific cost and
demand shocks affect cars and SUVs differentially, or if consumers who purchase different vehicle
types are heterogeneous (for instance if they drive different mileage or have different discount
factors).28 For expositional brevity we focus solely on the baseline manufacturer price and present
24
The fuel cost coefficients contribute substantially to the regression fits. For example, the R2 of Column 1 is
reduced from 0.5260 to 0.4133 when the fuel cost variables are removed from the specification, so that changes in
vehicle fuel costs explain more than ten percent of the variance in manufacturer prices.
25
As we develop in Appendix B, this is consistent with demand being roughly symmetric.
26
The results to not seem to be driven by outliers; the coefficients are similar when we exclude the extremely fuel
efficient or fuel inefficient vehicles from the sample.
27
We plot each vehicle only once because the derivatives do not vary substantially over time or regions. Indeed,
the only variation within vehicles is due to changes in the set of other vehicles available.
28
One might additionally suspect that the response of manufacturer prices to fuel costs changes over time. To test
for such heterogeneity, we split the observations to form one sub-sample over the period 2003-2004 and another over
the period 2005-2006; the results from each sub-sample are quite close. Similarly, we divide the sample between the
2003-2004 model-years and the 2005-2006 model-years without substantially changing the results. We conclude that

12
the results using figures. The regression coefficients appear in Appendix Table A-1.
Figure 4 plots the estimated effects of a one dollar increase in the gasoline price on manu-
facturer prices against vehicle miles-per-gallon, separately for each vehicle type.29 Converted into
dollars, the median estimated effect is a reduction in the manufacturer price of $779, $981, and
$174 for cars, SUVs, and trucks, respectively, and an increase of $91 for vans. Among cars and
SUVs, the fuel cost effect almost always dominates the competitor fuel cost effect: 91 percent of
the cars and 95 percent of the SUVs feature negative net effects. Still, the estimated manufacturer
price response is less negative for more fuel efficient vehicles, so that the univariate correlation co-
efficient between the price response and miles-per-gallon is 0.6610 for cars and 0.7521 for SUVs.30
By contrast, the magnitude of the estimated effects are much smaller for trucks and vans, as is the
strength of the relationship between the effects and vehicle fuel efficiency.
In order to provide some sense of the economic magnitude of these results, we use back-of-
the-envelope calculations to (roughly) estimate the extent to which manufacturers offset changes
in consumers’ cumulative gasoline expenses. We assume an annual discount rate of five percent,
a vehicle holding period of thirteen years, and a utilization rate of 11,154 miles per year (the
Department of Transportation estimates an average vehicle lifespan of thirteen years and 145,000
miles). Under these parameters, the cumulative gasoline expense associated with a one dollar
increase in the gasoline prices ranges between $1,972 and $7,953 among the sample vehicles; the
expense for the median vehicle (miles-per-gallon of 21.40) is $5,073. We divide the estimated
manufacturer responses, based on the regression coefficients shown in Appendix Table A-1, by the
computed cumulative gasoline expense. The resulting ratio is the percent of cumulative gasoline
expenses, due to a change in the retail gasoline price, that is offset by changes in the manufacturer
price.
Figure 5 plots this “offset percentage” against vehicle miles-per-gallon, separately for each
vehicle type. The median offset percentage is 18.17 and 15.27 for cars and SUVs, respectively, but
climbs as high as 52.17 for cars (the 2006 Ford GT) and as high as 33.92 for SUVs (the 2004 GM
Envoy XUV). These percentages fall in vehicle fuel efficiency, so that the univariate correlation
coefficients between the offset percentage and miles-per-gallon for cars and SUVs are -0.6292 and
-0.6681, respectively. By contrast, the offset percentage is smaller for trucks and vans. We wish
to emphasize that these numbers should be interpreted with considerable caution. Alternative
assumptions regarding the discount rate, the vehicle holding period, and the utilization rate could
push the offset percentages higher or lower. Further, as previously discussed, the manufacturer price
we use to estimate the regressions – MSRP minus the mean available incentive – could understate
the manufacturer responses and the offset percentages if some consumers stack multiple incentives.
Returning the regression results of Appendix Table A-1, in Figure 6 we plot the estimated
the effects of any time-related heterogeneity are relatively small.
29
Each plot combines the results of four regressions, one for each manufacturer.
30
Appendix Table A-2 lists the largest positive and negative price effects for both cars and SUVs.

13
manufacturer price effects against vehicle miles-per-gallon for cars, separately for each manufac-
turer. The estimated effects are negative for all GM and Ford cars, and negative for 92 percent of
the Toyota cars (all but the 2003 Echo and the four Prius vehicles). Converted into dollars, the
median estimated effect for these manufacturers is a reduction in price of $610, $1180, and $758,
respectively. By contrast, only 38 percent of the Chrysler estimated effects are negative and the
median effect is an increase of $107. This difference between Chrysler and the other manufactur-
ers remains even for a given level of fuel efficiency. For example, the mean effects for cars with
between 25 and 35 miles-per-gallon are reductions of $529, $843, and $719, respectively, for GM,
Ford and Toyota, but an increase of $239 for Chrysler. One might conclude that Chrysler pursues
a different pricing strategy than GM, Ford, and Toyota. However, an alternative explanation is
that Chrysler vehicles are simply more fuel efficient than their competitors (e.g., Chrysler vehicles
could be closer to inefficient vehicles in attribute space). We compare the manufacturers’ pricing
rules more explicitly in Section 4.2.
We plot the estimated manufacturer price effects among SUVs separately for each manufac-
turer in Figure 7. Among the GM, Ford, and Toyota SUVs, the estimated price effects are positive
for only four vehicles: the 2006 (Ford) Mercury Mariner Hybrid, the 2006 Ford Escape Hybrid,
the 2006 Toyota Highlander Hybrid and the 2006 Lexus RX 400 Hybrid. The median estimated
effects for GM, Ford, and Toyota are reductions in price of $1315, $663, and $754, respectively. The
price effects are more negative for fuel inefficient SUVs. By contrast, the estimated price effects
are positive for nearly 30 percent of the Chrysler SUVs and the price effects are actually more
negative for fuel efficient SUVs.31 The unexpected pattern among Chrysler SUVs exists because
the estimated fuel cost coefficient is positive and the competitor fuel cost coefficient is negative (see
Appendix Table A-1), inconsistent with the profit maximizing pricing rule derived in the theoretical
framework.

4.2 Extensions
4.2.1 Lagged retail gasoline prices and gasoline futures

The main results are based on the premise that consumers form expectations about future retail
gasoline prices based on current retail gasoline prices. We explore that premise here. In particular,
we examine whether manufacturers set vehicle prices in response to information on historical gaso-
line prices and gasoline futures prices. We construct two new sets of fuel cost variables. The first
uses the mean retail gasoline price over the previous four weeks, and the second uses the one-month
futures price for retail gasoline. To the extent that consumers are backward-looking and forward-
looking, respectively, manufacturers should adjust vehicle prices to these new fuel cost variables.
The units of observation are at the vehicle-week level; we discard regional variation because futures
31
The univariate correlation coefficients between the price effects and miles-per-gallon are 0.9062, 0.8584, and
0.9447 for GM, Ford, and Toyota, respectively, and -0.1765 for Chrysler.

14
prices are available only at the national level. The results are therefore comparable to Column 3
of Table 3.
Table 4 presents the regression results. Columns 1 and 2 include variables based on mean
lagged gasoline prices and gasoline futures prices, respectively. The fuel cost coefficients are -64.55
and -47.66; the competitor fuel cost coefficients are 50.01 and 63.32. The coefficients are statistically
significant and consistent with the theoretical model. Still, the more interesting question is whether
these variables matter after controlling for the current price of retail gasoline. Columns 3 and 4
include variables based on mean lagged gasoline prices and gasoline futures prices, respectively,
together with variables based on the current gasoline price. Each of the coefficients takes the
expected sign and statistical significance is maintained for all but two coefficients. Finally, Column
5 includes variables based on mean lagged gasoline prices and variables based on gasoline futures
prices. The coefficients are precisely estimated and again take the correct sign.
The finding that consumers may use historical gasoline prices and gasoline futures prices to
form expectations for gasoline prices is interesting, in part because both the empirical evidence
and the conventional wisdom of industry experts suggest that gasoline prices follow a random walk
(as we outline Section 3). One could argue that some consumers form inefficient expectations
for future gasoline prices. Alternatively, some consumers may be imperfectly informed about the
current gasoline price; these consumers could rationally turn to alternative sources of information,
such as historical prices and/or futures prices. We are skeptical that our data can untangle these
informal hypotheses and hope that future research better addresses the topic.

4.2.2 Impulse Response Functions

In this section, we examine manufacturer price responses for hypothetical, “perfectly average”
vehicles. We define a perfectly average vehicle as one whose miles-per-gallon, weighted-average
competitor miles-per-gallon, and weighted-average same-firm miles-per-gallon are all at the mean
(for cars the mean is 25.99; for SUVs it is 18.80). Hypothetical vehicles are advantageous for
comparisons of manufacturers because they strip away the vehicle heterogeneity that may not be
apparent in the main results (e.g., Figures 6 and 7); one can essentially compare the performance
of manufacturer price rules under identical circumstances.32
We use impulse response functions to track the effects of a gasoline price shock, during the
week of the shock and each of the following ten weeks. The approach may be of additional interest
to the extent that it captures dynamics. To compute the impulse response function, we add ten
lags of each fuel cost variable to the baseline specification, and estimate the specification separately
for the cars and SUVs of each manufacturer. We then calculate the predicted effects of a one dollar
increase in the gasoline price for the perfectly average car and SUV (in principle, one could examine
32
For example, based on Figure 6 alone, it is not clear whether Chrysler employs a fundamentally different pricing
rule than GM, Ford, and Toyota, or whether its vehicles are simply more fuel efficient than their competitors (e.g.,
they could be closer to inefficient vehicles in attribute space).

15
any hypothetical vehicle).
Figure 9 shows the results.33 Starting with the cars, GM, Ford, and Toyota reduce prices by
$516, $495, and $691, respectively, immediately following the gasoline price shock, while Chrysler
increases prices by $106. The discrepancies between the manufacturer grow steadily over the
following ten weeks; by the final week, the net price changes are reductions of $1,495, $2,767,
$1,673, and $21 for GM, Ford, Toyota, and Chrysler, respectively. Turning to the SUVs, GM,
Ford, and Toyota reduce their prices by $121, $105, and $569, respectively, immediately following
the gasoline shock, while Chrysler increases prices by $63. Again, the discrepancies between the
manufacturer grow steadily over the following weeks; by the final week, the net price changes are
reductions of $831, $612, $1,422, and $72 for GM, Ford, Toyota, and Chrysler, respectively. Overall,
Ford reacts most aggressively relative to the other manufacturers in adjusting its car prices; Toyota
reacts most aggressively for SUVs. Chrysler’s reactions are negligible for both vehicle types.
Two of the results merit further discussion. First, we find Chrysler’s price responses puzzling
because the theoretical framework indicates that demand for the perfectly average vehicle must
fall in response to an adverse gasoline shock.34 We are reticent to conclude that Chrysler’s pricing
rule is suboptimal, however, in the absence of more sure evidence. It is possible that Chrysler’s
consumers are distinctly unresponsive to fuel costs, or that Chrysler adjusts its prices without using
incentives.35 Second, the result that manufacturer prices continue to fall after the initial gasoline
price shock is consistent with the hypothesis that consumers internalize gasoline price shocks slowly
over time. The result could also be consistent with some forms of dynamic competition or certain
supply-side frictions; we leave the exploration of these possibilities to future research.

4.2.3 Demand and cost factors

In the main regressions we estimate a separate time fixed effect for each of the 208 weeks in the data.
These fixed effects capture the combined influence of demand and cost factors that change over time
through the sample period. In this section, we use a second-stage regression to decompose the fixed
effects into contributions from specific time-varying demand and cost factors. We are particularly
interested in whether the retail gasoline price affects manufacturer prices after having controlled
for its impact on vehicle fuel costs. Such an effect could be present if higher gasoline prices increase
manufacturer production costs or reduce consumer demand through an income effect.36 One might
33
Appendix Tables A-3 and A-4 provide the regression coefficients. The individual coefficients are difficult to
interpret due to the high degree of co-linearity among the 33 fuel cost regressors, but the net manufacturer price
effects are reasonable, easily interpretable, and consistent with the main results.
34
A corollary is that the fuel cost coefficient should be larger in magnitude than the competitor fuel cost coefficient,
i.e., |φ1jt | > |φ2jt |. In the main regression results, shown in Table A-1, this holds for GM, Ford, and Toyota, but not
for Chrysler.
35
Chrysler dealerships may adjust prices. We note, however, that our data include cash incentives paid to both
consumers (“consumer-cash”) and dealerships (“dealer-cash”).
36
For example, Gicheva, Hastings, and Villas-Boas (2007) identify an income effect of gasoline prices using scanner
data on grocery purchases.

16
expect these two channels to partially offset; we can identify only the net effect.
Figure 8 plots the time fixed effects estimated in Column 3 of Table 3, together with the prime
interest rate and the unemployment rate (which may shift demand), price indices for electricity
and steel (which may shift manufacturer costs), and the retail gasoline price (which may shift
demand and costs). The fixed effects units are in thousands, so that a fixed effect of 0.25 represents
manufacturer prices that are $250 on average higher than manufacturer prices during the first
week of 2003 (the base date). The fixed effects are higher in the winter months than in the summer
months, consistent with the notion that manufacturer prices fall as consumers anticipate the arrival
of new vehicles to the market in the summer months (e.g., Copeland, Dunn, and Hall 2005). The
prime interest rate increases over the sample while unemployment decreases; the means of these
variables are 5.64 and 5.30, respectively. The electricity and steel indices are defined relative to
January 1, 2003; the prices of these cost factors increase over the sample by 10 and 61 percent,
respectively. The mean gasoline price is $2.16 per gallon, and gasoline prices increase over the
sample.37
We regress the estimated time fixed effects on different combinations of the demand and
cost factors.38 Table 5 presents the results. Column 1 features only the gasoline price, Column 2
features the gasoline price and the other demand factors, Column 3 features gasoline price and the
other cost factors, and Column 4 features all five demand and cost factors. The coefficients are
remarkably stable across specifications. In each column, the gasoline price coefficient is small and
statistically indistinguishable from zero; gasoline prices appear to have little effect on manufacturer
prices after controlling for vehicle fuel costs. The remaining coefficients take the expected signs.
Based on the Column 4 regression, a one percentage point increase in prime interest rate reduces
manufacturer prices by $164 and a one percentage point increase in the unemployment rate reduces
manufacturer prices by $104 (though the latter effect is not statistically significant). Similarly, ten
percent increases in the prices of electricity and steel raise manufacturer prices by $283 and $55,
respectively.

4.2.4 Vehicle inventories

We use the assumption that manufacturers have full information about consumer demand condi-
tions to generate a simple linear pricing rule. It is not clear whether the assumption is appropriate.
For example, manufacturers may receive only noisy signals about demand, and accurate informa-
37
The electricity index is publicly available from the EIA, and the steel index is publicly available from Producer
Price Index maintained by the Bureau of Labor Statistics. We deseasonalize both indices using the X12-ARIMA
prior to their use in analysis.
38
Each regression includes week fixed effects to help control for seasonality. To be clear, we estimate 52 week fixed
effects using 208 weekly observations; equivalent weeks in each year are constrained to have the same fixed effect. We
use the Newey and West (1987) variance matrix to account for first-order autocorrelation. The standard errors do
not change substantially when we account for higher-order autocorrelation. We are unable to use the more general
clustering correction because the data lack cross-sectional variation. Of course, the standard errors may be too small
because the dependent variable is estimated in a prior stage.

17
tion may be costly to obtain. In such an environment, one might expect manufacturers to set their
prices primarily based on their observed inventories; demand conditions would affect prices only
indirectly. As a specification test, we re-estimate the empirical model controlling for inventories.
The main theoretical framework – and its simple pricing rule – should gain credibility if the fuel
cost coefficients remain important.
To implement the test, we collect data on the “days supply” of inventory from Automotive
News, a major trade publication. Days supply is the current inventory divided by sales during the
previous month (the units are easily converted from months to days). The measure is frequently
used in industry analysis (e.g., Windecker 2003). Intuitively, the days supply should be high when
demand is sluggish and low when demand is great. The units of observation are at the month-model
level. To be clear, the inventories data do not vary across weeks within a month, and the data
lump all vehicles within a given model (e.g., the 2003 Dodge Neon and 2004 Dodge Neon). We map
the data into the main regression sample by using cubic splines to interpolate weekly observations.
We then apply the days supply to every vehicle in the model category. The procedure generates a
regression sample of 500 vehicles and 41,822 vehicle-week observations.39
Table 6 presents the regression results. In Column 1, we re-estimate the same specification as
in Table 3, Column 3 using only those observations for which we have information on inventories.
The fuel cost and competitor fuel cost coefficients are -69.23 and 53.16, respectively.40 We add
the days supply measure to the specification in Column 2. The fuel cost and competitor fuel cost
coefficients of -69.11 and 53.00 are virtually unchanged.41 The result suggests that manufacturers
respond to changes in demand conditions before these changes affect inventories; one might infer
that manufacturers are well informed about consumer preferences. The result also strengthens our
interpretation of the main empirical results: manufacturers intentionally set prices as if consumers
respond to gasoline prices.

5 Conclusion
We provide empirical evidence that automobile manufacturers adjust vehicle prices in response to
changes in the price of retail gasoline. In particular, we show that the vehicle prices tend to decrease
in their own fuel costs and increase in the fuel costs of their competitors. The net effect is such that
adverse gasoline price shocks reduce the price of most vehicles but raise the price of particularly fuel
efficient vehicles. We argue, based on theoretical micro foundations, that these empirical results are
39
We have inventory data for 500 of the 589 domestic vehicles in the data; the Toyota data are insufficiently
disaggregated to support analysis. The mean days supply among the 41,822 vehicle-week observations is 92.18. The
25th , 50th , and 75th percentiles are 62.26, 84.63, and 109.42, respectively.
40
The fact that these coefficients are close to those produced by the full sample provides some comfort that the
smaller inventory sample does not introduce sample selection problems or other complexities.
41
The days supply coefficient is small and statistically indistinguishable from zero. We are wary of interpreting this
coefficient too strongly because inventories may be correlated with the vehicle-time specific cost and demand shocks
that compose the error term in the regression equation.

18
consistent with the notion that automobile manufacturers set prices as if consumers value (low) fuel
costs. In terms of policy implications, the results suggest that gasoline and/or carbon taxes may be
effective instruments in mitigating the negative externalities associated with gasoline combustion in
automobiles. The results do not speak, however, to the optimal magnitude of any policy responses;
we leave that important matter to future research.

19
References

[1] Alquist, Ron and Lutz Kilian. 2008. What do we learn from the price of crude oil futures?
Mimeo.

[2] Bento, Antonio, Lawrence Goulder, Emeric Henry, Mark Jacobsen, and Roger von Haefen.
2005. Distributional and efficiency impacts of gasoline taxes: an econometrically based multi-
market study. American Economic Review – Papers and Proceedings, 95.

[3] Berry, Steve, Jim Levinsohn, and Ariel Pakes. 2004. Estimating differentiated product demand
systems from a combination of micro and macro data: the market for new vehicles. Journal of
Political Economy, 112: 68-105.

[4] Copeland, Adam, Wendy Dunn, and George Hall. 2005. Prices, production and inventories
over the automotive model year. NBER Working Paper 11257.

[5] Corrado, Carol, Wendy Dunn, and Maria Otoo. 2006. Incentives and prices for motor vehicles:
what has been happening in recent years. FEDS Working Paper.

[6] Davies, Peter. 2007. What’s the value of an energy economist? Speech presented at the Inter-
national Association of Energy Economics, Wellington, New Zealand.

[7] Dickey, D. A., and W. A. Fuller. 1979. Distribution of the estimators for autoregressive time
series with a unit root. Journal of the American Statistical Association, 74: 427431.

[8] Department of Energy, Energy Information Agency. 2006. A primer on gasoline prices.

[9] Gicheva, Dora, Justine Hastings, and Sofia Villas-Boas. 2007. Revisiting the income effect:
gasoline prices and grocery purchases. CUDARE Working Paper.

[10] Goldberg, Pinelopi. 1998. The effects of the corporate average fuel economy standards in the
automobile industry. Journal of Industrial Economics, 46: 1-33.

[11] Jacobsen, Mark. 2008. Evaluating U.S. fuel economy standards in a model with producer and
household heterogeneity. Mimeo.

[12] Li, Shanjun, Christopher Timmins, and Roger H. von Haefen. 2007. Do gasoline prices affect
fleet fuel economy? Mimeo.

[13] Makridakis, Spyros, Steven C. Wheelwright, and Rob J. Hyndman. 1998. Forecasting Methods
and Applications. (3rd ed.). New York: Wiley.

[14] Miller, Don M. and Dan Williams. 2004. Damping seasonal factors: Shrinkage estimators for
the X-12-ARIMA program. International Journal of Forecasting, 20: 529-549.

[15] Newey, Whitney K., and Kenneth D. West. 1987. A simple positive semi-definite, heteroskedas-
ticity and autocorrelation consistent covariance matrix. Econometrica, 55: 703-708.

[16] Windecker, Ray. 2003. The battle of the bulge: the intricacies and lessons of days supply.
Automotive Industries, November.

20
[17] Train, Kenneth and Cliff Winston. 2007. Vehicle choice behavior and the declining market
share of U.S. automakers. International Economic Review, 48: 1469-1496.

21
A Elasticity Bias
In our introductory remarks, we argued informally that structural estimation can understate con-
sumer responsiveness to fuel costs if it fails to account for manufacturer price responses. We
formalize our argument here in the context of logit demand. In particular, we demonstrate that
1) estimation yields a fuel cost coefficient that is biased downwards and 2) one can estimate the
magnitude of bias with data on gasoline and manufacturer prices.
Under a set of standard (and restrictive) assumptions, the logit demand system generates the
well-known regression equation:

log(sjt ) − log(s0t ) = ψ(pjt + xjt ) + κj + νjt , (A-1)

where sjt and s0t are the market shares of vehicle j and the outside good, respectively, pjt is the
vehicle price, xjt captures the expected lifetime fuel costs, κj is vehicle “quality,” and νjt is an error
term that captures demand shocks.
Assuming away the obvious endogeneity issues, one can use OLS with vehicle fixed effects to
obtain consistent estimates of ψ, the parameter of interest. However, suppose that one observes
the mean price of each vehicle rather than the true price. The regression equation becomes:

log(sjt ) − log(s0t ) = ψxjt + κ∗j + νjt

∗
, (A-2)

where κ∗j = κj +ψpj and νjt ∗ = ν +ψ(p −p ). The problem is now apparent. Gasoline price shocks
jt jt j
affect not only xjt but also the composite error term νjt∗ through the manufacturer response. Since,

as we document above, adverse gasoline shocks typically induce manufacturers to lower prices, the
OLS estimate of ψ is biased downwards. Going further, the regression coefficient has the expression:
P P
(xjt − xj )νjt (xjt − xj )ψ(pjt − pj )
b
ψ = ψ+ P + P
(xjt − xj )2 (xjt − xj )2
µ P ¶
(xjt − xj )(pjt − pj )
→p ψ 1 + P . (A-3)
(xjt − xj )2

Thus, it is possible to estimate the magnitude of bias simply by regressing vehicle prices on expected
lifetime fuel costs and a set of fixed effects; one need not have market share data or any other inputs
to the structural model.
Such a procedure has its difficulties. Perhaps the most central is constructing an appro-
priate proxy for expected lifetime fuel costs.42 We use the discounted price-per-mile, i.e., xjt =
(gpt /mpgj )/(1 − δ), and impose a per-mile discount rate of δ = 0.999995401; this corresponds to an
annual discount rate of 0.95, assuming 11,154 miles per year.43 We measure manufacturer prices in
dollars, rather than thousands of dollars, to sidestep any problems associated with unit conversion.
We then regress manufacturer prices on lifetime fuel costs, vehicle fixed effects, and time fixed
effects. The resulting coefficient of -0.141 (standard error = 0.019) corresponds to a downward bias
of 14 percent.44
42
Of course, structural estimation also requires one to proxy fuel costs. Goldberg (1998), Bento et al (2005) and
Jacobsen (2007) all use measures based on price-per-mile.
43
The Department of Transportation estimates the average vehicle lifespan to be thirteen years and 145,000 miles;
based on these data, the average number of miles per year is 11,154.
44
The calculation is sensitive to the discount rate. An annual discount rate of 0.99 produces a bias of 2.7 percent;

22
 Although we hope our empirical estimate of bias provides a useful benchmark, we caution
against taking the calculation too literally. Data imperfections and/or specification errors could
result in an estimate that is too high or too low. For example, our measure of manufacturer prices
is based on incentives offered to consumers and does not fully capture transaction prices or even the
actual incentives selected. Our proxy for expected lifetime fuel costs imposes both a specific form
of multiplicative discounting and an arbitrary discount rate. Aside from these estimation issues,
the bias formula itself is based on logit assumptions that are generally considered too restrictive.
More flexible structural models still understate consumer responsiveness to fuel costs – the negative
correlation between fuel costs and unobserved price responses remains – but the bias is nonlinear
and could be substantially larger or smaller than what we estimate here.

B Analytical solutions to the theoretical model

B.1 Three single-vehicle manufacturers
We derive analytical solutions to the theoretical model for the specific case of three single-product
manufacturers that compete in prices. The profit equation specified in Equation 1 takes the form:

πj = (pj − cj ) ∗ qj (e
p· ) − fj , (B1-1)

where pj is the price of vehicle j, the scalar cj captures the marginal cost of production, the quantity
demanded qj is a function of the “full” vehicle price, inclusive of fuel costs, and fj is a fixed cost.
We specify the linear demand system:
X
qj = αjj (pj + xj ) + αjk (pk + xk ) + µj (B1-2)
k6=j

in which the scalar xj is the fuel cost of vehicle j, and the scalar µj is an exogenous demand shifter.
We are concerned with the case in which demand is well-defined (so that αjj < 0 ∀j) and vehicles
are substitutes (so that αjk > 0 ∀j 6= k). The first-order condition for the equilibrium price of
vehicle j can be expressed as follows:
µ ¶
∗ 1 1 1 1 X αjk
pj = cj − µj − xj − (pk + xk ) (B1-3)
2 αjj 2 2 αjj
k6=j

an annual discount rate of 0.90 produces a bias of 28.9 percent.

23
We solve the system of equations for the equilibrium vehicle prices as functions of the non-price
variables. The equilibrium price for vehicle 1 has the expression:
· ¸
∗ 1 α23 α32 1 α12 α21 1 α13 α31 1 α12 α23 α31 1 α13 α32 α21
p1 ∗ 1− − − + +
4 α22 α33 4 α11 α22 4 α11 α33 8 α11 α22 α33 8 α11 α33 α22
· ¸
1 1 α23 α32 1 α12 α21 1 α13 α31 1 α12 α23 α31 1 α13 α32 α21
= − 1− − − + + ∗ x1
2 4 α22 α33 2 α11 α22 2 α11 α33 4 α11 α22 α33 4 α11 α33 α22
· ¸ · ¸
1 α12 1 α13 α32 1 α13 1 α12 α23
− − ∗ x2 − − ∗ x3 (B1-4)
4 α11 2 α11 α33 4 α11 2 α11 α22
· ¸ µ ¶
1 1 α23 α32 1
+ 1− ∗ c1 − µ1
2 4 α22 α33 α11
· ¸ µ ¶ · ¸ µ ¶
1 α12 1 α13 α32 1 1 α13 1 α12 α23 1
− − ∗ c2 − µ2 − − ∗ c3 − µ3
4 α11 8 α11 α33 α22 4 α11 8 α11 α22 α33

The equilibrium prices for vehicles 2 and 3 are analogous. One can combine the two competitor
fuel cost terms into a single term that captures the influence of the weighted average competitor
fuel cost. This single term has the expression:
· ¸
1 α12 α13 1 α12 α23 1 α13 α32
− + − − ∗ (ω12 x2 + ω13 x3 ) , (B1-5)
4 α11 α11 2 α11 α22 2 α11 α33

where the weights w12 and w13 sum to one. The weights are functions of the demand parameters:
α12 1 α13 α32
α11 − 2 α11 α33
ω12 = α12 α13 1 α12 α23 1 α13 α32
α11 + α11 − 2 α11 α22 − 2 α11 α33
(B1-6)
α13
α11 − 12 αα12 α23
11 α22
ω13 = α12 α13 1 α12 α23 1 α13 α32 .
α11 + α11 − 2 α11 α22 − 2 α11 α33

A single regularity condition generates the following results regarding the relationship between
equilibrium prices and fuel costs:
∂p∗1 ∂p∗1
Result A1-1: ∈ [−1, 0] and ∈ [0, 1]
∂x1 ∂(ω12 x2 + ω13 x3 )
¯ ∗¯ ¯ ¯
¯ ∂p1 ¯ ¯ ∂p∗1 ¯
Result A1-2: ¯ ¯>¯ ¯
¯ ∂x1 ¯ ¯ ∂(ω12 x2 + ω13 x3 ) ¯

Thus, in any empirical implementation, one should expect that the regression coefficient on fuel
costs should be negative, that the coefficient on the weighted average competitor fuel costs should
be positive, and that the first coefficient should be larger in magnitude than the second. If one
proxies cumulative fuel costs using a measure of current fuel costs – for example, the “price per-
mile” variable that we employ – then the coefficients may be much larger than one in magnitude.

24
The same regularity condition generates the following results regarding the weights:

Result A1-3: ω12 ∈ (0, 1) and ω13 ∈ (0, 1)

∂ω12 ∂ω13
Result A1-4: > 0 and >0
∂α12 ∂α13
Since the parameters α12 and α13 govern the severity of competition between vehicles, it is appro-
priate to weight “closer” competitors more heavily when constructing the empirical proxies for the
weights. The regularity condition that generates these results is:
1 α12 1 α13 1 α12 α21 1 α13 α31 1 α23 α32
1> − − + + +
2 α11 2 α11 2 α11 α22 2 α11 α33 4 α22 α33
(B1-7)
1 α12 α23 1 α13 α32 1 α12 α23 α31 1 α13 α32 α21
+ + − − .
4 α11 α22 4 α11 α33 4 α11 α22 α33 4 α11 α33 α22
The condition holds provided that the own-price parameters are sufficiently large relatively to the
cross-price parameters. For intuition, it may be useful to note that each right-hand-side term
enters as a positive because the own-price parameters are negative and the cross-price parameters
are positive. Although these results extend naturally to cases with J > 3 manufacturers, the
algebraic burden associated with obtaining analytical solutions increases exponentially with J.

B.2 One manufacturer with three vehicles

We derive analytical solutions to the theoretical model for the specific case of a single manufac-
turer that produces three distinct products. The first-order conditions for profit maximization are
identical to those presented in Section 2, i.e.,
∂π=t X X
= αjk (pkt + xkt ) + µjt + αkj (pkt − ckt ) = 0. (B2-1)
∂pjt
k k

25
We solve the system of equations for the equilibrium vehicle prices as functions of the non-price
variables. The equilibrium price for vehicle 1 has the expression:
· ¸
1 (α23 + α32 )2 1 (α12 + α21 )2 1 (α13 + α31 )2 1 α12 + α21 α23 + α32 α13 + α31
p∗1 ∗ 1− − − +
4 α22 α33 4 α11 α22 4 α11 α33 4 α11 α22 α33
·
1 1 (α23 + α32 )2 1 α12 + α21 α21 1 α13 + α31 α31
= − 1− − −
2 4 α22 α33 2 α11 α22 2 α11 α33
¸
1 α12 + α21 α23 + α32 α31 1 α13 + α31 α23 + α32 α21
+ + ∗ x1
4 α11 α22 α33 4 α11 α33 α22
·
1 α12 1 α12 + α21 1 α13 + α31 α32 1 α13 + α31 α32 + α23
− − − +
2 α11 2 α11 2 α11 α33 4 α11 α33
(B2-2)
¸
1 α12 + α21 α23 + α32 α32 1 α23 + α32 α23 + α32 α12
+ − ∗ x2
4 α11 α22 α33 4 α22 α33 α11
·
1 α13 1 α13 + α31 1 α12 + α21 α23 1 α12 + α21 α23 + α32
− − − +
2 α11 2 α11 2 α11 α22 4 α11 α22
¸
1 α13 + α31 α23 + α32 α23 1 α23 + α32 α23 + α32 α13
+ − ∗ x3
4 α11 α33 α22 4 α33 α22 α11

+f (c1 , c2 , c3 , µ1 , µ2 , µ3 ),

where, for brevity, we focus on the fuel cost terms. Again, one can combine the fuel cost terms of
vehicles 2 and 3 into a single term that captures the weighted average influence of these vehicles.
A single regularity condition, slightly stronger than that presented in Equation B2-4, generates the
following results:
∂p∗1 ∂p∗1 ∂p∗1
Result A2-1: ∈ [−1, 0] ; ∈ [−1, 1] and ∈ [−1, 1]
∂x1 ∂x2 ∂x3
¯ ∗¯ ¯ ∗¯ ¯ ∗¯
¯ ∂p1 ¯ ¯ ∂p1 ¯ ¯ ∂p1 ¯
Result A2-2: ¯ ¯ ¯ ¯ ¯ ¯
¯ ∂x1 ¯ ≶ ¯ ∂x2 ¯ ≶ ¯ ∂x3 ¯

Thus, the manufacturer partially offsets changes in fuel costs of a specific vehicle with changes in
price of that vehicle. Changes in the fuel costs of other vehicles produced by the same manufacturer,
however, have ambiguous implications for the vehicle price. Interestingly, in the specific case of
symmetric demand (i.e., αjk = αkj ∀ j, k), changes in the fuel costs of other vehicles produced
by the same manufacturer have no effect on the equilibrium price. The regularity condition that

26
generates these results is:

1 (α12 + α21 )2 1 (α13 + α31 )2 1 (α23 − α32 )2

1 > + +
4 α11 α22 4 α11 α33 4 α22 α33
1 α12 + α21 α21 1 α13 + α31 α31
+ + (B2-3)
2 α11 α22 2 α11 α33
1 α12 + α21 α23 + α32 α31 1 α13 + α31 α23 + α32 α21
− −
4 α11 α22 α33 4 α11 α33 α22
The condition holds provided that the own-price parameters are sufficiently large relatively to the
cross-price parameters. Again, the intuition underlying these results extends naturally to cases
with J > 3 vehicles.

27
 Retail Gasoline Prices by Region
2003−2006

3
Price Per Gallon

2.5

National East Coast

1.5 Gulf Coast Midwest
Rocky Mts West Coast

2003 2004 2005 2006 2007

Figure 1: The weekly retail price of gasoline by region over 2003-2006, in real 2006 dollars.

Seasonally Adjusted Retail Gasoline Prices

January 1993−February 2008

2.5 Seasonal Adjustment

Price Per Gallon

1.5 .2
.1
1 0
−.1
Price Adjustment −.2

1994 1996 1998 2000 2002 2004 2006 2008

Figure 2: Seasonally adjusted retail gasoline prices at the national level over 1993-2008, in real
2006 dollars. Seasonal adjustments are calculated with the X-12-ARIMA program.

28
 The Effect of a $1 Increase in the Gasoline Price

Manufacturer Price Change (000s) 1

.5

−.5

−1

−1.5

−2
10 20 30 40 50 60
Miles Per Gallon

Figure 3: The estimated effects of a one dollar increase in the retail gasoline price on the manu-
facturer price, based on the regression results in Column 1 of Table 3. Each point represents the
price effect for a single vehicle. See text for details.

The Effect of a $1 Increase in the Gasoline Price

Cars SUVs

1
Manufacturer Price Change (000s)

−1

−2

−3

Trucks Vans

−1

−2

−3

0 20 40 60 0 20 40 60
Miles Per Gallon

Figure 4: The estimated effects of a one dollar increase in the retail gasoline price on the manu-
facturer price, based on the regression results of Appendix Table A-1. Each point represents the
price effect for a single vehicle. See text for details.

29
 Manufacturer Offsets of Gasoline Price Changes
Cars SUVs
.75

Percent of Consumer Expenses Offset .5

.25
0
−.25
−.5

Trucks Vans
.75
.5
.25
0
−.25
−.5
0 20 40 60 0 20 40 60
Miles Per Gallon

Figure 5: The percentages of consumer cumulative gasoline expenses, due to changes in the retail
gasoline price, that are offset by changes in the manufacturer price. Each point represents the
percentage for a single vehicle. Based on back-of-the-envelope calculations and the regression
results of Appendix Table A-1.

The Effect of a $1 Increase in the Gasoline Price

Vehicle Type = Car
GM Ford

1
Manufacturer Price Change (000s)

0
−1
−2
−3

Chrysler Toyota

1
0
−1
−2
−3

0 20 40 60 0 20 40 60
Miles Per Gallon

Figure 6: The estimated effects of a one dollar increase in the retail gasoline price on the manu-
facturer price, based on the regression results of Appendix Table A-1. Each point represents the
price effect for a single vehicle. See text for details.

30
 The Effect of a $1 Increase in the Gasoline Price
Vehicle Type = SUV
GM Ford

1
Manufacturer Price Change (000s)

0
−1
−2
−3

Chrysler Toyota

1
0
−1
−2
−3
0 20 40 60 0 20 40 60
Miles Per Gallon

Figure 7: The estimated effects of a one dollar increase in the retail gasoline price on the manu-
facturer price, based on the regression results of Appendix Table A-1. Each point represents the
price effect for a single vehicle. See text for details.

Demand and Cost Factors

8

Time Fixed Effect Gasoline Price

Prime Interest Rate Unemployment Rate
Electricity Price Index Steel Price Index

2003 2004 2005 2006 2007

Figure 8: Time-series plots of time fixed effects together with selected demand and cost factors.
There are 208 weekly values over 2003-2008. The time fixed effects are estimated in the regression
presented in Table 3, Column 3.

31
 The Effect of a $1 Increase in the Gasoline Price
Cars SUVs

Manufacturer Price Change (000s)

1

−1

−2

−3
−1 0 1 2 3 4 5 6 7 8 9 10 11 −1 0 1 2 3 4 5 6 7 8 9 10 11
Week
GM Ford
Chrysler Toyota

Figure 9: The estimated effects of a one dollar increase in the retail gasoline price for a hypothetical,
“perfectly average” vehicle, in the ten weeks following a gasoline price shock. A perfectly average
vehicle is one whose miles-per-gallon, weighted-average competitor miles-per-gallon, and weighted-
average same-firm miles-per-gallon are all at the mean (for cars the mean is 25.99; for SUVs it is
18.80). The impulse response function is calculated based on the regression coefficients shown in
Appendix Tables A-3 and A-4.

Table 1: Summary Statistics

Variables Definition Mean St. Dev.
Manufacturer price MSRPj − INCjrt 30.344 16.262
Fuel cost gprt /mpgj 0.108 0.034
MSRP MSRPj 30.782 16.299
Miles-per-gallon mpgj 21.555 5.964
Horsepower 224.123 71.451
Wheel base 115.193 12.168
Passenger capacity 4.911 1.633
Means and standard deviations based on 299,855 vehicle-region-week ob-
servations over the period 2003-2006. The manufacturer price is defined
as MSRP minus the mean regional and national incentives (in thousands).
The fuel cost is the gasoline price divided by miles-per-gallon, and cap-
tures the gasoline expense per mile. The manufacturer price, the fuel cost,
and MSRP (in thousands) are in real 2006 dollars; wheel base is measured
in inches.

32
 Table 2: Means by Vehicle Type
Variables Cars SUVs Trucks Vans
Manufacturer price 30.301 35.782 24.482 24.658
Fuel cost 0.087 0.121 0.133 0.120
MSRP 30.835 36.124 24.881 25.048
Miles-per-gallon 25.991 18.803 17.121 18.815
Horsepower 209.947 241.858 254.383 191.152
Wheel base 107.723 114.880 129.527 123.196
Passenger capacity 4.799 5.849 3.763 4.451

# of observations 125,660 90,270 46,615 37,310

Means based on vehicle-region-week observations over the period 2003-
2006. The manufacturer price is defined as MSRP minus the mean regional
and national incentives (in thousands). The fuel cost is the gasoline price
divided by miles-per-gallon, and captures the gasoline expense per mile.
The manufacturer price, the fuel cost, and MSRP (in thousands) are in
real 2006 dollars; wheel base is measured in inches.

33
 Table 3: Manufacturer Prices and Fuel Costs
Incentive level:
Regional+ Regional National
National Only Only
Variables (1) (2) (3)
Fuel cost -55.40*** -56.96*** -63.75***
(7.73) (7.86) (8.77)

Average competitor 50.76*** 50.16*** 50.09***

fuel cost (7.15) (7.39) (8.12)

Average same-firm 1.15 2.62 1.31

fuel cost (2.29) (1.78) (2.30)

R2 0.5260 0.6763 0.5289

# of observations 299,855 299,855 59,971
# of vehicles 681 681 681
Results from OLS regressions. The dependent variable is the manufac-
turer price, i.e., MSRP minus the mean regional and/or national incen-
tives (in thousands). The units of observation in Columns 1 and 2 are at
the vehicle-week-region level. The units of observation in Column 3 are
at the vehicle-week level. All regressions include vehicle and time fixed
effects, and Columns 1 and 2 include region fixed effects. The regressions
also include third-order polynomials in the vehicle age (i.e., weeks since
the date of initial production), the average age of vehicles produced by
different manufacturers, and the average age of other vehicles produced
by the same manufacturer. Standard errors are clustered at the vehicle
level and shown in parenthesis. Statistical significance at the 10%, 5%,
and 1% levels is denoted by *, **, and ***, respectively.

34
 Table 4: Gasoline Price Lags and Futures Prices
Variables Metric (1) (2) (3) (4) (5)
Fuel cost Lagged -64.55*** -36.51*** -30.08***
Retail (8.77) (10.65) (8.42)

Average competitor Lagged 50.01*** 23.19** 30.24***

fuel cost Retail (8.16) (10.09) (9.93)

Fuel cost Futures -47.66*** -35.52** -31.69***

(7.11) (16.42) (9.39)

Average competitor Futures 63.32*** 19.87 27.73**

fuel cost (10.44) (24.95) (13.21)

Fuel cost Retail -29.70*** -22.58

(10.83) (16.46)

Average competitor Retail 27.70*** 33.38*

fuel cost (8.14) (18.87)

R2 0.5291 0.5286 0.5295 0.5295 0.5305

Results from OLS regressions. The dependent variable is the manufacturer price, i.e., MSRP minus the mean
national incentive (in thousands). The sample includes 59,971 observations on 681 vehicles at the vehicle-week
level. Fuel cost variables labeled “lagged retail” are constructed using the mean retail gasoline price over the
previous four weeks. Fuel cost variables labeled “futures” are constructed using the one-month futures price
of retail gasoline. Fuel cost variables labeled “retail” are constructed using the current retail gasoline price.
All regressions include the appropriate average same-firm fuel cost variable(s). The regressions also include
vehicle and time fixed effects, as well as third-order polynomials in the vehicle age (i.e., weeks since the date
of initial production), the average age of vehicles produced by different manufacturers, and the average age of
other vehicles produced by the same manufacturer. Standard errors are clustered at the vehicle level and shown
in parenthesis. Statistical significance at the 10%, 5%, and 1% levels is denoted by *, **, and ***, respectively.

35
 Table 5: Demand and Cost Factors
Variables (1) (2) (3) (4)
Gasoline Price -0.015 0.011 -0.102 -0.096
(0.036) (0.059) (0.088) (0.067)

Interest Rate -0.128*** -0.164***

(0.027) (0.034)

Unemployment Rate -0.315*** -0.104

(0.073) (0.091)

Electricity Price Index 0.950* 2.832***

(0.540) (0.726)

Steel Price Index 0.405*** 0.549***

(0.113) (0.152)

R2 0.5160 0.6117 0.5829 0.6454

Results from OLS regressions. The data include 208 weekly observations over the
period 2003-2006. The dependent variable is the time fixed effect estimated in
Column 3 of Table 3. The regressions also include 52 week fixed effects; equivalent
weeks in each year are constrained to have the same fixed effect. Standard errors
are robust to the presence of heteroskedasticity and first-order autocorrelation.
Statistical significance at the 10%, 5%, and 1% levels is denoted by *, **, and ***,
respectively.

36
Table 6: Manufacturer Prices, Fuel Costs, and Inventories
Variables (1) (2)
Fuel cost -69.23*** -69.11***
(11.57) (11.54)

Average competitor 53.16*** 53.00***

fuel cost (9.79) (9.76)

Average same-firm 1.95 1.94

fuel cost (3.36) (3.36)

Vehicle inventory 0.0001

(0.0001)

R2 0.6202 0.6203
Results from OLS regressions. The dependent variable is the
manufacturer price, i.e., MSRP minus the mean national incen-
tive (in thousands). The sample includes 41,822 observations on
500 vehicles over the period 2003-2006, at the vehicle-week level.
The regressions include vehicle and time fixed effects, as well as
third-order polynomials in the vehicle age (i.e., weeks since the
date of initial production), the average age of vehicles produced
by different manufacturers, and the average age of other vehicles
produced by the same manufacturer. Standard errors are clus-
tered at the vehicle level and shown in parenthesis. Statistical
significance at the 10%, 5%, and 1% levels is denoted by *, **,
and ***, respectively.

37
 Table A-1: Manufacturer Prices by Vehicle Type and Manufacturer
Vehicle Type: Cars SUVs
Manufacturer: GM Ford Chrysler Toyota GM Ford Chrysler Toyota
Fuel cost -97.52*** -146.80** -152.09*** -77.13*** -75.98*** -72.10*** 45.87* -62.11***
(17.85) (71.68) (30.81) (21.55) (23.26) (23.50) (24.92) (17.82)

Average competitor 85.78*** 80.61 159.99*** 46.38*** 64.08*** 66.75*** -29.56 44.25***
fuel cost (18.25) (57.25) (41.91) (18.52) (21.67) (23.79) (19.01) (16.41)

Average same-firm -6.47 41.10 -2.56 6.12 8.19 -1.51 -17.88* 1.73
fuel cost (8.79) (29.64) (11.33) (4.48) (9.02) (6.28) (10.42) (4.27)

R2 0.6173 0.5254 0.5294 0.7282 0.7861 0.6758 0.7126 0.8352

# of vehicles 101 92 34 66 94 50 24 34
Vehicle Type: Trucks Vans

38
Manufacturer: GM Ford Chrysler Toyota GM Ford Chrysler Toyota
Fuel cost -43.10*** -61.49*** 26.07 2.70 2.26 4.02 8.60 30.47
(5.46) (17.91) (20.50) (15.35) (1.75) (6.93) (8.61) (14.26)

Average competitor 37.77*** 57.54*** -30.63 -0.68 -1.12 -1.51 -4.40 -28.52*
fuel cost (5.74) (17.39) (19.31) (14.07) (3.51) (7.48) (10.78) (11.91)

Average same-firm 2.70* 5.13 -1.69 -0.36 0.88 -0.39 -15.33*** -5.33
fuel cost (1.44) (3.49) (2.27) (1.18) (2.28) (1.50) (4.84) (3.65)

R2 0.8946 0.7959 0.8248 0.5659 0.9051 0.8610 0.7074 0.8769

# of vehicles 59 22 16 8 30 19 28 4
Results from OLS regressions. The dependent variable is the manufacturer price, i.e., MSRP minus the mean regional and national incentives (in
thousands). The units of observation are at the vehicle-week-region level. All regressions include vehicle, time, and region fixed effects, as well
as third-order polynomials in the vehicle age (i.e., weeks since the date of initial production), the average age of vehicles produced by different
manufacturers, and the average age of other vehicles produced by the same manufacturer. Standard errors are clustered at the vehicle level and
shown in parenthesis. Statistical significance at the 10%, 5%, and 1% levels is denoted by *, **, and ***, respectively.
 Table A-2: Fuel Efficient and Inefficient Vehicles
The Most Positive Manufacturer Price Responses
d
∂p d
∂p
Cars Brand mpg ∂gp SUVs Brand mpg ∂gp
2003 SRT4 Dodge 32.85 0.9148 2006 Escape Hybrid Ford 30.25 0.6485
2004 Prius Toyota 55.05 0.5268 2006 RX 400h Lexus 30.25 0.4304
2006 Prius Toyota 55.05 0.5227 2006 Mariner Hybrid Mercury 30.80 0.3944
2005 Prius Toyota 55.05 0.4971 2006 Highlander Hybrid Toyota 30.25 0.3111
2005 SRT4 Dodge 26.40 0.4661 2003 Wrangler Jeep 19.10 0.1551
2004 SRT4 Dodge 26.40 0.3740 2005 Wrangler Jeep 19.65 0.1442
2003 Prius Toyota 48.15 0.3414 2006 Liberty Jeep 20.20 0.1284
2004 Neon Dodge 32.85 0.3305 2003 Liberty Jeep 21.75 0.1246
2003 Neon Dodge 32.85 0.3244 2003 Durango Dodge 16.75 0.1155
2005 Neon Dodge 32.85 0.2981 2006 Wrangler Jeep 19.65 0.0691

39
The Most Negative Manufacturer Price Responses
d∂p d
∂p
Cars Brand mpg ∂gp SUVs Brand mpg ∂gp
2003 XKR Jaguar 19.85 -2.0168 2003 H2 Hummer 13.65 -2.3293
2004 GTO Pontiac 18.75 -2.0239 2006 H2 SUV Hummer 13.65 -2.3618
2004 Marauder Mercury 20.30 -2.0617 2004 H1 Hummer 13.65 -2.3711
2005 Viper Dodge 16.95 -2.1401 2003 9-7X Saab 13.65 -2.4298
2003 Viper Dodge 16.95 -2.1462 2003 H1 Hummer 13.65 -2.4511
2004 Viper Dodge 16.95 -2.1880 2003 Escalade Cadillac 13.65 -2.5031
2003 Marauder Mercury 20.30 -2.2581 2006 H2 SUT Hummer 13.65 -2.5640
2006 Viper Dodge 16.40 -2.4917 2005 H2 SUT Hummer 13.65 -2.578
2005 GT Ford 17.40 -3.2390 2006 H1 Hummer 13.65 -2.6173
2006 GT Ford 17.40 -3.2552 2005 Envoy XUV GMC 13.65 -2.6979
Based on Appendix Table A-1 and Figures 6 and 7.
 Table A-3: Multiple Fuel Cost Lags – Cars
Variables Weeks Lagged GM Ford Chrysler Toyota
Fuel cost 0 -37.39 -50.31 -46.24 -81.33***
Fuel cost 1 -3.56 -13.28 5.21 22.78*
Fuel cost 2 -19.01 -36.08** -78.41* -2.99
Fuel cost 3 17.41 -46.24** -18.12 17.15
Fuel cost 4 18.63 -5.47 -17.10 -18.97
Fuel cost 5 8.79 7.48 18.23 11.85
Fuel cost 6 -29.07*** 21.31** 66.43* -26.99***
Fuel cost 7 13.72 8.66 29.70 -0.59
Fuel cost 8 9.20 -55.16*** -29.28 -2.15
Fuel cost 9 38.37** 61.03*** 87.15 15.40
Fuel cost 10 -128.67*** -76.56* -186.37* -19.28
Competitor fuel cost 0 62.56** 96.30 179.61** 39.14**
Competitor fuel cost 1 -6.61 -50.99 -33.56** -8.47
Competitor fuel cost 2 14.27 13.95 -0.29 4.88
Competitor fuel cost 3 8.94 -0.07 -29.92 -4.58
Competitor fuel cost 4 -7.34 -12.45 17.16** -5.11*
Competitor fuel cost 5 4.24 -4.73 0.70 6.23
Competitor fuel cost 6 -14.43 -11.32 3.00 3.67
Competitor fuel cost 7 11.55 18.85 -8.89 1.24
Competitor fuel cost 8 0.78 -30.56 39.08 1.67
Competitor fuel cost 9 -19.88 38.13 -48.17 -5.92*
Competitor fuel cost 10 24.95 16.55 53.38 2.42
Same-firm fuel cost 0 -38.58 -58.84 -130.65 24.22
Same-firm fuel cost 1 10.12 58.21*** 27.63 -12.61
Same-firm fuel cost 2 1.84 18.16 77.65 -6.11
Same-firm fuel cost 3 -27.85 42.70* 45.63 -12.74
Same-firm fuel cost 4 -13.41 14.15 1.77 20.82
Same-firm fuel cost 5 -15.89 -9.79 -20.53 -20.91
Same-firm fuel cost 6 41.49*** -16.04 -71.12* 19.82**
Same-firm fuel cost 7 -26.16 -27.69 -23.03 -1.58
Same-firm fuel cost 8 -13.21 77.31*** -6.22 -2.72
Same-firm fuel cost 9 -17.86 -95.73*** -44.45 -8.20
Same-firm fuel cost 10 93.18** 36.61 139.47 5.50
Results from four OLS regressions. The dependent variable is the manufacturer price, i.e., MSRP minus the
mean regional and national incentives (in thousands). The units of observation are at the vehicle-week-region
level. All regressions include vehicle, time, and region fixed effects, as well as third-order polynomials in the
vehicle age (i.e., weeks since the date of initial production), the average age of vehicles produced by different
manufacturers, and the average age of other vehicles produced by the same manufacturer. Standard errors are
clustered at the vehicle level but omitted for brevity. Statistical significance at the 10%, 5%, and 1% levels is
denoted by *, **, and ***, respectively.

40
 Table A-4: Multiple Fuel Cost Lags – SUVs
Variables Weeks Lagged GM Ford Chrysler Toyota
Fuel cost 0 -71.20** -43.14* -83.16** -53.31***
Fuel cost 1 19.08* 14.33 57.27** 8.93
Fuel cost 2 -7.88 -26.46 24.12 -3.87
Fuel cost 3 -17.00 1.86 7.55 -0.311
Fuel cost 4 15.59 -10.92 22.44** 5.93
Fuel cost 5 -11.67 9.79 3.11 -8.23**
Fuel cost 6 14.21 -5.77 3.41 -26.99***
Fuel cost 7 12.78 4.16 0.93 6.12
Fuel cost 8 -26.82** -2.80 -3.14 -2.73
Fuel cost 9 16.32 9.02 11.18 7.51
Fuel cost 10 -26.35 -31.32 13.24 -31.48***
Competitor fuel cost 0 54.88*** 51.93** -47.89 19.78
Competitor fuel cost 1 -17.24*** -2.81 26.82 7.39
Competitor fuel cost 2 13.70** -1.54 5.40 1.49
Competitor fuel cost 3 0.64 1.41 -4.15 -8.32
Competitor fuel cost 4 -10.06 -12.30 7.75 6.37
Competitor fuel cost 5 1.93 3.82 -14.28 -4.54
Competitor fuel cost 6 -1.86 4.04 16.51 10.18**
Competitor fuel cost 7 -3.10 13.56* -9.35 4.41
Competitor fuel cost 8 5.59 3.78 -7.94 12.13**
Competitor fuel cost 9 7.89 6.02 22.83 -10.01
Competitor fuel cost 10 23.06* 0.60 -36.55 6.40
Same-firm fuel cost 0 14.05 -10.78 132.23* 22.83
Same-firm fuel cost 1 -3.03 -11.78 -86.08* -15.56
Same-firm fuel cost 2 -6.04 27.00 -29.04 0.12
Same-firm fuel cost 3 15.90 -3.69 -4.66 8.66
Same-firm fuel cost 4 -7.62 21.91 -29.28* -15.09*
Same-firm fuel cost 5 8.73 -14.59 10.42 10.50
Same-firm fuel cost 6 -13.94 1.12 -21.26 -9.30
Same-firm fuel cost 7 -10.22 -17.76 8.46 -10.77
Same-firm fuel cost 8 20.09 2.19 11.08 -11.27
Same-firm fuel cost 9 -8.55 -14.54 -34.20 4.19
Same-firm fuel cost 10 -1.68 26.61 24.88 17.74
Results from four OLS regressions. The dependent variable is the manufacturer price, i.e., MSRP minus the
mean regional and national incentives (in thousands). The units of observation are at the vehicle-week-region
level. All regressions include vehicle, time, and region fixed effects, as well as third-order polynomials in the
vehicle age (i.e., weeks since the date of initial production), the average age of vehicles produced by different
manufacturers, and the average age of other vehicles produced by the same manufacturer. Standard errors are
clustered at the vehicle level but omitted for brevity. Statistical significance at the 10%, 5%, and 1% levels is
denoted by *, **, and ***, respectively.

41