Investigaciones Europeas de Dirección y Economía de la Empresa 19 (2013) 1–7
www.elsevier.es/iedee
Product efficiency in the Spanish automobile market
Eduardo González a,∗ , Juan Ventura a , Ana Cárcaba b
a
b
Department of Business Administration, University of Oviedo, Av. Cristo s/n, 33071 Oviedo, Spain
Department of Accounting, University of Oviedo, Av. Cristo s/n, 33071 Oviedo, Spain
a r t i c l e
i n f o
Article history:
Received 7 March 2012
Accepted 2 July 2012
Available online 7 September 2012
JEL classification:
C61
D13
D61
L62
L81
a b s t r a c t
This paper evaluates product efficiency in the Spanish automobile market. We use non parametric frontier techniques in order to estimate product efficiency scores for each model. These scores reflect the
minimum price for which each car could be sold, given the bundle of tangible features it offers in comparison to the best-buy models. Unlike previous research, we use discounted prices which have been
adjusted by car dealerships to meet sale targets. Therefore, we interpret the efficiency scores as indicators of the value of the intangible features of the brand. The results show that Audi, Volvo, Volkswagen and
Mercedes offer the greatest intangible value, since they are heavily overpriced in terms of price/product
ratios. Conversely, Seat, Kia, Renault and Dacia are the brands that can be taken as referent in terms of
price/product ratios.
© 2012 AEDEM. Published by Elsevier España, S.L. All rights reserved.
Keywords:
Product efficiency
DEA
Automobile
Spain
Eficiencia de producto en el mercado español del automóvil
r e s u m e n
Códigos JEL:
C61
D13
D61
L62
L81
Palabras clave:
Eficiencia de producto
DEA
Automóvil
España
Este artículo analiza la eficiencia de producto en el mercado español del automóvil, utilizando para ello
métodos de frontera no paramétricos. Los índices de eficiencia de producto indican el precio mínimo a que
cada automóvil podría venderse, dado el conjunto de atributos tangibles que ofrece en comparación con
las mejores compras. A diferencia de investigaciones previas, se utilizan los precios ajustados que incluyen
los descuentos hechos por los concesionarios para poder alcanzar los objetivos de ventas. Esto permite
interpretar los índices de eficiencia como un indicador del valor de las características intangibles de cada
modelo. Los resultados muestran que son Audi, Volvo, Volkswagen y Mercedes las marcas que ofrecen un
mayor valor intangible, dado que sus modelos tienen una relación precio/producto muy desfavorable. Por
el contrario, Seat, Kia, Renault y Dacia son las marcas que pueden tomarse como referencia en términos
de su relación precio/producto.
© 2012 AEDEM. Publicado por Elsevier España, S.L. Todos los derechos reservados.
1. Introduction
Business competitive analysis is concerned with the ability of
competitors to deliver products with similar or superior product/price ratios in the marketplace, which can be obtained at a
∗ Corresponding author.
E-mail addresses: efidalgo@uniovi.es (E. González), ventura@uniovi.es
(J. Ventura), acarcaba@uniovi.es (A. Cárcaba).
similar or lower cost. Competitive advantage exists when the firm
either offers a higher added value for a given price (through product differentiation) or when costs are lower for a similar quality.
Actually, market price is the variable that splits the value created
between the firm and the customer. While the difference between
price and cost provides a profit margin for the firm, the difference
between the value of the product and its price provides a rational reason for a customer to purchase. And, in fact, no competitive
advantage can emerge for a firm if customers do not purchase its
products. Price setting is a critical decision in this regard. If the price
1135-2523/$ – see front matter © 2012 AEDEM. Published by Elsevier España, S.L. All rights reserved.
http://dx.doi.org/10.1016/j.iedee.2012.07.003
2
E. González et al. / Investigaciones Europeas de Dirección y Economía de la Empresa 19 (2013) 1–7
is just too high for the merits of the product, sales (and profits) will
tend to be low. Alternatively, if the price is too low for the merits
of the product, sales will be high, but margins will be unnecessarily
low. The right price is the one that reflects the merits of the product
appropriately, while providing a reasonable profit margin for the
firm.
But how can we determine how valuable the merits of a product
are? There is a growing body of literature aimed at evaluating
the relative merits of competing products on the basis of product
attributes and prices. This line of research is rooted in the seminal
work of Lancaster (1966) who described a product as a combination of attributes or a vector in the quality-price space. Following
this representation, it is straightforward to construct a theoretical
frontier with the products showing the highest quality/price ratios.
The competitiveness or appeal of a product could then be inferred
by the distance of the product vector to the frontier containing the
best-buy products. Most customers are not attracted to buy either
the highest quality or the lowest priced product. Instead, products
with the best quality/price ratio will be favored by the bulk of
the market, since prospective customers will seek to maximize
that ratio (Rust & Oliver, 1994). Product efficiency, understood as
the comparison of the product to the best-buy frontier can then
be considered as an indicator of the relative customer’s perceived
value, i.e., the value received for the money paid (Smirlis, Despotis,
Jablonsky, & Fiala, 2004).
The estimation of customer’s perceived value is a central
research concern in business strategy and marketing (Zeithaml,
1988; Dodds, Monroe, & Grewal, 1991; Holbrook, 1994). The traditional approach relied on bi-dimensional maps of perceived value
(Gale, 1994; Brayman, 1996). The methodology requires listing the
relevant attributes of the product and asking well-informed consumers to evaluate those attributes for a given product and to
weight the relative importance of each attribute. This information
can then be combined into a composite indicator of relative quality
that can be compared with the relative price of the product. While
this approach is quite straightforward, it also introduces obvious
biases in product assessment, since subjective evaluation will vary
as a function of variables such as age or income of respondents
(Bolton & Drew, 1991).
More sophisticated non-parametric frontier techniques can be
applied to compare the measurable attributes of a set of competing
products. Data envelopment analysis (DEA) is a frontier-based tool
that has been extensively used during the last 30 years to measure efficiency in production by comparing input–output vectors
to an empirically constructed best-practice frontier (Emrouznejad,
Parker, & Tavares, 2008). The adaptation of the DEA methodology to the estimation of the relative perceived value of competing
products was first proposed by Doyle and Green (1991). They
applied DEA to compare 37 computer printers on the basis of
measurable and objective product attributes. After this pioneering application, many scholars have applied this technique to
different sectors such as notebooks (Fernández-Castro & Smith,
1996), numerical control machines (Sun, 2002), mobile phones
(Smirlis et al., 2004; Lee, Hwang, & Kim, 2005), computer printers (Seiford & Zhu, 2003) and, most notably, automobiles. To our
knowledge, the DEA approach has been applied to evaluate product efficiency in automobile markets by Papagapiou, Mingers, &
Thanassoulis (1997), Papahristodoulou (1997), Fernández-Castro
and Smith (2002), Fernández-Castro and Doldán (2002), Staat and
Hammerschmidt (2005) and, more recently, Oh, Lee, Hwang, &
Heshmati (2010). Within the automobile industry other papers
have focused exclusively on evaluating the environmental impact
of the products (Kortelainen & Kousmanen, 2007).
In this paper, we build on previous literature to evaluate product efficiency in the Spanish automobile market. In doing so, we
pay special attention to overcoming some of the most common
empirical limitations of this body of research from three basic
aspects. The first issue that has been largely overlooked in previous research is the fact that car dealers usually make big price
adjustments, cutting the model’s official price, in order to meet
sales targets. Using the official price list may be the right approach
for comparing computer printers, but will surely be misleading for
comparing automobiles. Some brands make huge discounts (even
official discounts) that are not registered in the official price list.
Real market prices can be completely different from official ones.
In this paper we will use real (discounted) market prices for the
models analyzed.
The second limiting aspect of previous literature is the focus on a
very narrow part of the market. The number of models and versions
included in the empirical applications is typically very small relative to the extent of the real market. In this paper, we use data on
more than 900 versions of 79 different models from 22 brands. Non
parametric frontier methods require extensive samples because the
frontier is not estimated as a function but as an envelope of the data
available. If few data are available, the DEA frontier will be a very
imperfect representation of the underlying market frontier.
The third limiting aspect of previous research that we want
to address in this paper is the number of attributes which are
accounted. In general terms, previous research has been limited to a
few visible and objective car attributes, such as horse power, speed,
fuel consumption or price. In this paper, we combine information
of more than 70 variables in order to obtain the final composite
indicator of product efficiency.
2. Data
In order to approach the efficiency value of a product, we have to
compile a complete data set to account for the product’s most relevant attributes. We limit our sample to passenger cars (excluding
superminis such as the SMART model) and multi-purpose vehicles (MPVs). Therefore, we explicitly exclude from the study the
segments of sports cars, superminis, off-road vehicles or pick-ups,
because we consider that the comparison of these vehicles with
the rest makes no sense. We only consider diesel versions, which
are (by and large) the most demanded in Spain. Therefore, gasoline, hybrid and electric vehicles are also excluded from the sample.
We used publicly available information about the most commonly
used models commercialized in Spain in November 2010. All the
data come from the printed car magazine AUTOFACIL and from
the online car magazine KM77.COM. These publications contain
updated information on all the relevant technical data of the different versions of each model and also on the standard and optional
equipment of each version. For each model’s version we compiled
the following data:
2.1. Technical data
Discounted price 1: best price offered by UNOAUTO, which is actually available by at least one car dealer
Discounted price 2: best price offered by AUTODESCUENTO, which
is actually available by at least one car dealer
Discounted price: average of discounted price 1 and discounted
price 2 (this will be the price used in the DEA analysis)
Size: length/width/height. We use the product of these three variables as a volume measure of the size of the vehicle
Boot space: in liters
Real horse power: engine maximum horse power divided by car
weight.
Fuel consumption: kilometers per liter
Speed: maximum car speed
E. González et al. / Investigaciones Europeas de Dirección y Economía de la Empresa 19 (2013) 1–7
Acceleration: average acceleration in meters per second squared
until the car reaches a speed of 100 km/h (this is obtained by dividing the constant 27.7 by the time in seconds that is required to
reach that speed). We make this transformation in order to use
the variable as an output in the DEA model, instead of an input.
Safety: passenger protection score in EuroNcap crash tests (the
models that did not perform the EuroNcap test were normalized
to a value 0.6, which is the minimum value obtained by the cars
that performed the test)
Ecology: the result of dividing 100 by CO2 emissions. Spanish car
taxes use 100 as a cutoff point to distinguish the most ecological
cars (i.e., those with emissions below 100). We make this transformation in order to use the variable as an output in the DEA
model.
2.2. Standard equipment
We registered with dummy variables whether the version does
or does not include the following items within the standard equipment (at the discounted price):
Active safety equipment: ABS, ESP, EBD, BAS, TCS
Passive safety equipment: front airbags, rear airbags, curtain
airbags, knee airbag, pre-safe, isofix
Comfort seats: adjustable, leather, heated, electric, sportive, etc.
Electronics: radio, DVD, bluetooth, GPS, parking sensors, parking
camera, special sound system, tire pressure monitoring system,
cruise control, USB and i-Pod connections, TV, on board computer,
etc.
Lights: fog lights, xenon lamps, bixenon lamps, adaptive lamps,
automatic lights
Aesthetics: alloy wheels, tinted windows, metallic or pearl paint,
metallic or wood details, spoilers, sport pedals, etc.
Comfort: panoramic roof, sunroof, electrically operated wing
mirrors, electrochromic rear-view mirror, keyless entry, central locking, power windows, steering wheel-mounted controls,
leather-wrapped steering wheel, air conditioning, automatic air
conditioning, power steering, number of doors, etc.
Mechanical aids: front/rear/4x4 drive, manual/automatic transmission.
The most complex part is the evaluation of the standard equipment. The number of potential elements is so great that introducing
them as dummies into the DEA program would generate absurd
results. Instead, it is preferable to aggregate all this information into
a synthetic indicator of the value of the standard equipment that is
included at the discounted price. This will allow comparing vehicles with different standard equipment levels. Our approach was
to check, for each of these elements, which was the average price
at which they were offered as an optional extra in other models.
For instance, a car model that includes ESP would receive a value
of about 600D (the average price of ESP when it has to be added as
an extra) for having that item included within the car’s equipment.
For computing the average prices, we considered all the models
in which each of the elements was listed as an optional extra (i.e.,
not only the models included in our sample). By applying these
prices to the list of standard equipment, we obtained a variable
(Equipment) that is entered as an additional output to the DEA
model.
3. Methods
In order to obtain the index of product efficiency for each
model’s version, we constructed a DEA best-buy frontier. This frontier is obtained from the comparison of the data on inputs and
3
outputs of all the versions in the sample. In the case of product
efficiency, the inputs would be the features that the customer
would like to minimize (e.g., price, fuel consumption, etc.). The outputs would be the features that the customer wants to maximize
(e.g., horse power, equipment, etc.). In this research we will only
consider one input: discounted price. The rest of the cars’ features
have been measured in a specific way in order to treat them as outputs (i.e., more is better). The outputs included in our DEA model
are: ecology, fuel consumption, real horse power, maximum speed,
acceleration, volume, boot space, safety and equipment. Therefore,
we propose a model with one input and nine outputs to construct
the best-buy frontier. Cars located on the frontier can be considered as best-buys since they offer a unique combination of input
and outputs, one that cannot be beaten by any other product that
is available in the market. It is just not possible to find another car
that costs the same and offers more of each of the nine outputs.
Therefore, it can be considered as a benchmark or a rational choice.
Even though there are numerous versions of the DEA programs,
in this paper we followed the original formulation of Charnes,
Cooper, & Rhodes (1978). The DEA program finds the maximum
radial contraction in the inputs (input orientation) or the maximum radial expansion in the outputs (output orientation). In our
case, we are interested in controlling all the inputs and outputs, but
we are only interested in knowing the right price for each bundle of
attributes (as represented by the outputs of each model’s version).
This can be easily done using Kopp’s (1981) single input efficiency
measure, or adapting the DEA setting to have just one input. We
follow this second approach, by converting all the features that
should be naturally considered as inputs (fuel consumption, CO2
emissions, etc.) into outputs. Therefore, our model will seek to minimize the price at which each outputs bundle can be purchased in
the market. The constant returns to scale DEA model with an input
orientation implies solving the following linear program:
min
s.t. :
J
j yjs ≥ yis , s = 1...S
j=1
J
j pj ≤ pi
j=1
j ≥ 0,
∀j
where yis is output s for model version i and pi is its discounted price.
The mathematical program searches the combination of other vehicles that would be equivalent to a virtual model with a similar or a
better outputs bundle (the same quantity of each output or more)
with the lowest discounted price. The weights j indicate, when
they are different from zero, the vehicles that compose the referent point on the best-buy frontier. When i = 1, there is no other
car (or linear combination of them) offering more outputs at the
same price or lower. Conversely, if the car is not on the frontier,
then there must exist, at least, another vehicle (h) in the sample
with h > 0, which offers a better deal (or a linear combination of
them). The objective function (i.e., the DEA product efficiency score)
reflects the overprice that customers are paying for that particular
model’s version, which cannot be rationalized on the basis of the
tangible features of the car.
Banker, Charnes, & Cooper (1984) added to this basic model
the constraint that the sum of the weights be equal to 1 ( = 1).
The result is that comparisons are restricted to the convex hull of
the data. In the context of productivity analysis, this is interpreted
as a variable returns to scale technology. The consideration of
variable or constant returns to scale is actually meaningless in our
setting, because we are not constructing a production frontier.
4
E. González et al. / Investigaciones Europeas de Dirección y Economía de la Empresa 19 (2013) 1–7
1
0,9
0,8
0,7
0,6
0,5
0,4
0,3
0,2
0,1
d
ag i
Sk en
M od
er a
ce
de
To s
yo
La ta
nc
M ia
az
da
Fo
P e rd
ug
eo
t
Bm
w
H
on
d
N a
M iss
a
its n
ub
is
C hi
itr
oë
n
O
pe
l
S
H ea
yu t
nd
R ai
en
au
lt
Fi
at
Ki
a
D
ac
ia
Vo
lk
sw
lv
Vo
Au
o
0
Fig. 1. Product efficiency score (brand averages).
Source: own elaboration.
However, Hollingsworth and Smith (2003) have shown that when
data on inputs and outputs represent ratios of other variables, the
Banker et al. (1984) model is more appropriate. Given that many
of our variables are constructed as ratios, we will use the variable
returns to scale version of the DEA program in the evaluation of
the automobile models in our empirical application.
Once the DEA scores are obtained, we will try to explore the relationship between those scores and the intangible features of the
model, which may be associated with the brand name. In the automobile market, consumers derive value from the status associated
with premium brands. Furthermore, the brand can be interpreted as
a signal of the mechanical reliability of a car and other aspects such
as design or technical service, which are learned from past history
and interaction. Therefore, image and reliability can be associated
with brand name. Other aspects that may condition the price policy
are some tangible features of the car. For instance, larger cars may
have a higher overprice, since buyers of large cars may not be as sensitive to price as buyers of small cars. Another policy that may be
related to overprice is the level of standard equipment. Models that
include more elements of equipment can be more overpriced, since
it would be more difficult for the buyer to establish comparisons
with competing products. We can operationalize these concerns in
a model with the following specification, in which the overprice of
a car is related to brand (B) and to a vector of car features (Y) that
include size and equipment:
1 − i = Bi + Yi + εi
Given that the DEA scores are bounded within the interval
(0, 1], we will use a truncated regression model in order to
estimate the coefficients efficiently.
4. Results
We ran the DEA program for each of the 919 car versions
included in the sample. Fig. 1 shows the brand averages of the
product efficiency scores. Not surprisingly, Dacia leads the product efficiency frontier with an average score of 0.976, with 60% of
its models on the frontier. In technical terms, Dacia is a best-buy
option, since it offers robust mechanical features at a comparatively
low price. Following Dacia, we observe a group of seven brands
with averages around 0.9. These brands are Kia, Fiat, Renault,
Hyundai, Seat, Opel and Citroën. Five of them are well established
European generalist brands. The other two (Kia and Hyundai) are
Korean brands, which belong to the same matrix company. These
firms price very aggressively in order to gain a place in this highly
competitive market. Volvo, Audi and Volkswagen are the most
overpriced brands according to the DEA frontier. The VAG group
has four brands present in the sample (Audi, Volkswagen, Skoda
and Seat). Many of their models are mechanically very similar, but
Seat offers the lowest prices. The comparison with Seat makes Audi,
Volkswagen and Skoda inefficient (overpriced) products, especially
those of Audi. Something similar seems to be happening with the
two brands of group PSA (Peugeot and Citroën) and the brands of
the Fiat group (Fiat and Lancia).
Even though Dacia is the brand offering the best average deal,
it is not a good referent for many models. Ford is the brand that
serves more often as a referent for other brands. It does not have
many models on the frontier, but one version of the Ford Focus
serves as referent in the DEA program of 224 other models and
one version of the Ford Mondeo is a referent for 235 models.1 On
average, each Ford car included in the sample serves as a referent for
13.2 overpriced models. Ford is followed by Peugeot (11.9), Renault
(11.8), Seat (11.0) and Opel (10.4). These figures, however, do not
account for the intensity with which a brand is considered as a
referent for others, which can be approached by the value of the
intensity vector () estimated through the DEA program. Table 1
shows for each brand the three referent brands that are used with
more intensity in the DEA program, indicating the intensity as a
percentage.
We find that Seat is the brand that serves as referent of inefficient models with more intensity. Globally, Seat is used as a
benchmark, with an average intensity of 37.7%, followed by Kia
with 16.4%, Renault with 12.8%, Dacia with 6.9% and, remarkably,
BMW with 4.7%. We must indicate, however, that BMW models
serve mainly as referents for other BMW models, which find no
comparison with best-buy (frontier) brands. At the other extreme,
Audi, Lancia, Volvo, Mazda, Honda, Fiat and Mercedes are the least
active benchmark brands. They do not reach a joint 1% and they
mostly serve as referents for other cars with the same brand.
Table 1 also shows the three main referents for the overpriced
models of each brand. For instance, we can observe that Audi’s models are considered overpriced when compared with those of Seat,
BMW and Kia. Seat accounts for 38.7% of the DEA comparisons of
Audi’s models, which is not surprising, since many Audi models are
1
In most cases, as part of a linear combination with other referent models.
E. González et al. / Investigaciones Europeas de Dirección y Economía de la Empresa 19 (2013) 1–7
5
Table 1
Brands as referents for other brands (intensity in brackets).
Brand as referent
Referent 1 of brand
Referent 2 of brand
Referent 3 of brand
Audi (0.1%)
BMW (4.7%)
Citroën (1.7%)
Dacia (6.9%)
Fiat (0.6%)
Ford (4.1%)
Honda (0.5%)
Hyundai (1.1%)
Kia (16.4%)
Lancia (0.2%)
Mazda (0.4%)
Mercedes (0.9%)
Mitsubishi (0%)
Nissan (0%)
Opel (1.1%)
Peugeot (4.6%)
Renault (12.8%)
Seat (37.7%)
Skoda (4.2%)
Toyota (1.7%)
Volkswagen (0%)
Volvo (0.3%)
Seat (38.7%)
BMW (48.7%)
Seat (28.3%)
Dacia (70.7%)
Seat (44.5%)
Dacia (30.4%)
Seat (34.9%)
Seat (39.2%)
Kia (40.0%)
Seat (38.5%)
Seat (45.0%)
Seat (24.0%)
Seat (79.2%)
Seat (38.9%)
Seat (48.4%)
Seat (31.4%)
Renault (40.2%)
Seat (55.7%)
Seat (33.7%)
Seat (43.4%)
Seat (46.9%)
Seat (48.8%)
BMW (12.1%)
Seat (15.2%)
Citroën (21.1%)
Kia (20.1%)
Kia (16.5%)
Seat (25.3%)
Honda (15.5%)
Hyundai (18.2%)
Seat (27.5%)
Renault (18.3%)
Dacia (15.6%)
Kia (17.3%)
Renault (7.6%)
Dacia (19.7%)
Renault (12.9%)
Kia (21.4%)
Seat (30.0%)
Kia (11.2%)
Kia (30.5%)
Kia (27.4%)
Kia (21.0%)
Renault (19.3%)
Kia (11.3%)
Renault (15.1%)
Kia (18.5%)
Renault (7.1%)
Toyota (10.5%)
Skoda (13.5%)
Kia (14.0%)
Kia (16.3%)
Renault (8.5%)
Lancia (15.0%)
Mazda (6.6%)
Mercedes (16.3%)
Fiat (6.8%)
Kia (16.5%)
Opel (11.5%)
Renault (12.8%)
Kia (14.0%)
Peugeot (9.0%)
Renault (9.2%)
Renault (7.2%)
Renault (10.1%)
Skoda (6.8%)
Source: own elaboration.
(Mercedes, Audi, BMW, Volvo) may be adding an important dose
of brand image to their products and, therefore, overpricing is the
expected outcome. We saw in Fig. 1 that Volvo, Audi and Mercedes
are indeed highly overpriced brands. However, this is not the case
with BMW, which is still an expensive brand, but offers a bundle
of output features that other brands can hardly match at the same
price. Fig. 2 reinforces this view, since Audi, Mercedes and Volvo
show considerable slack, while BMW is among the most adjusted
brands in terms of output slack. In the case of Audi, this is due
mainly to the sharing of most mechanical features with Seat, which
is clearly the frontier benchmark brand that makes Audi so overpriced. In the case of Mercedes and Volvo, they simply seem to be
heavily overpriced. But in the case of BMW, its customers’ purchasing decision can be rationalized not just in terms of brand image,
but also in terms of product/price ratio.
We have regressed the overprice scores (1 − ) on a list of brand
dummy variables and the two car features that we expected to
be related with overpricing (size and equipment). We excluded
Dacia from the list of dummy variables. Therefore, the results of
the brand dummy variables have to be interpreted in comparison with Dacia. The complete results are shown in Table 3. The
coefficients of the dummy variables confirm the results shown in
Fig. 1, with Audi and Volvo as the most heavily overpriced brands,
60
Safety
50
Boot space
Volume
40
Acceleration
30
Speed
Equipment
20
Horse power
Fuel cons
10
Ecology
Fig. 2. Output slacks in percentage.
Source: own elaboration.
Mitsubishi
Audi
Volvo
Ford
Toyota
Lancia
Mercedes
Skoda
Volkswagen
KIA
Nissan
Peugeot
FIAT
Hyundai
Seat
Opel
Mazda
Bmw
Dacia
Honda
Citroën
0
Renault
technically identical to the equivalent Seat models (with some differences in design and the quality of some materials). In the case
of BMW, we observe that almost half of the comparison comes
from other BMW models. Therefore, BMW seems to be less overpriced than other premium brands (Audi, Mercedes, Volvo) and,
additionally, half of the overpriced models are determined to be
so when compared with other BMW models. This indicates that
BMW would be a rational choice even if the customer bases the
purchasing decision exclusively on tangible attributes. The table
shows these figures for all the brands included in the sample.
It may also be interesting to know the models’ versions that are
considered as benchmarks for overpriced models most frequently
(Table 2). A very versatile version of the Renault Clio (G. Tour with
a very efficient and well-known engine of the brand) emerges as a
referent (in most cases as part of a composite linear combination)
for 482 overpriced models. This amounts to 62% of the overpriced
models in the sample. Another two versions of the Clio are among
the TOP20 benchmark models. Additionally, three versions of the
Seat Ibiza (with different engines and equipment) are included in
the TOP10 of Table 2.
The DEA program also shows the output features that could be
improved without raising the price of the car. As these results are
not the main objective of the DEA program, they are called ‘slacks’.
For instance, the slack for real horse power indicates the additional
increase in real horse power that a car should offer in order to be
totally comparable with the benchmark cars located on the bestbuy frontier. Fig. 2 shows the cumulative slacks in the nine output
features considered in the DEA program, as a percentage of the
actual value.
Mitsubishi and Volvo are not just overpriced (Fig. 1) but could
also increase the nine output features of their models by about a
cumulative 50%, which means about 5.5% per output feature (Fig. 2).
These brands behave especially poorly in terms of CO2 emissions,
as reflected by the Ecology slacks. Audi has the third largest cumulative slack in the outputs (being the second worst brand in terms of
product efficiency). Audi’s main slacks are observed in equipment
and fuel consumption.
It is not surprising to find generalist brands (Renault, Seat, Opel,
Citroën), low cost brands (Dacia) and Korean brands (Kia, Hyundai)
dominating the best-buy frontier. The market approach of these
brands is offering a good price/product relationship, and this is
exactly what the DEA frontier is composed of Premium brands
6
E. González et al. / Investigaciones Europeas de Dirección y Economía de la Empresa 19 (2013) 1–7
Table 2
Rank of benchmark models.
Rank
Model/version
Times as referent
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
RENAULT CLIO 1.5 dCi 105 Exception G. Tour 6V
SEAT IBIZA 2.0 TDI CR FR 6V 3P
SKODA SUPERB 2.0 TDI 140 Exclusive 6V Aut.Combi
SEAT EXEO 2.0 TDI Sport 6V
DACIA LOGAN 1.5 Dci Ambiance
KIA CEE‘D 1.6 CRI 90 Concept 3p
KIA CARNIVAL 2.9 CRDI Active
SEAT IBIZA 1.6 TDI CR 105 Sport 3P
SEAT IBIZA 1.2-60 CV-Emoción 5p
RENAULT CLIO 1.5 dCi 105 Exception 6V 5p
KIA CEE‘D 1.6 CRDi 90 Drive 3p
SEAT EXEO 2.0 TDI Sport 6V Aut.
FORD FOCUS 1.6 TDCI 109 Titanium 5p
PEUGEOT 308 1.6 HDI 112 Confort Aut. 6V 5p
PEUGEOT 207 1.6 HDI 90 99 Gramos 5p
SEAT ALTEA/ALTEA XL 2.0 TDI 140 Style 6V XL
PEUGEOT 207 1.6 HDI 92 Confort 3p
TOYOTA YARIS 1.4 D-4D Live 6V 3p
HYUNDAI i30 1.6 CRDi 128 Fde Style 6V
RENAULT CLIO 1.5 dCi 85 Expression G. Tour
482
317
315
288
268
235
224
216
203
200
143
117
112
94
83
80
70
70
59
51
Source: own elaboration.
Table 3
Determinants of car overprice.
Variable
Coefficient
t-test
Intercept
Audi
BMW
Citroën
Fiat
Ford
Honda
Hyundai
Kia
Lancia
Mazda
Mercedes
Mitsubishi
Nissan
Opel
Peugeot
Renault
Seat
Skoda
Toyota
Volkswagen
Volvo
Car’s size
Standard equipment
−0.738
0.592
0.444
0.298
0.246
0.430
0.382
0.296
0.260
0.448
0.465
0.542
0.331
0.417
0.353
0.399
0.287
0.352
0.458
0.474
0.539
0.582
0.41 × 10−7
0.23 × 10−5
−6.22***
5.38***
4.00***
2.61***
2.14**
3.92***
3.40***
2.38**
2.29**
3.99***
4.17***
4.89***
2.72***
3.46***
3.11***
3.62***
2.59***
3.16***
4.15***
4.27***
5.27***
5.27***
12.0***
1.14
Source: own elaboration.
***
Significance level 0.01.
**
Significance level 0.05.
followed by Mercedes and Volkswagen. Clearly this effect comes
from a perception of brand quality and image of these brands. In this
sense, the brand coefficients can be interpreted as the implicit value
that the brand adds to the model. Since we are using discounted
prices, we can assume that customers purchase those models at a
technical overprice, which means they obtain some compensating
utility from brand intangible attributes (image, post-sale service,
perceived reliability, perceived quality of components, etc.). Audi
(0.59), Volvo (0.58), Mercedes (0.54) and Volkswagen (0.54), would
be the brands for which customers are willing to pay more, for
a given bundle of tangible product attributes. Assuming that discounted prices have been adjusted to the levels that make the
models competitive in the marketplace, these are the brands that
offer more intangible value. The case of other premium brands, such
as BMW, is different. BMW’s customers pay a relatively small overprice for the intangible properties associated with the brand (0.44),
which suggests that the bundle of tangible product attributes is
more appealing and would be competitive even without such a
strong brand name.
The results also confirm the hypothesis that larger cars are more
overpriced than small ones. The coefficient of this variable is highly
significant. However, we must reject the idea that the level of standard equipment mediates overpricing. Even though the coefficient
is positive (as expected) it is insignificant at conventional levels.
5. Concluding remarks
The automobile industry offers a unique framework in which
to evaluate product efficiency for two reasons. The first is that,
while there is considerable product differentiation, there is plenty
of information about the tangible attributes of each car that is commercialized in the marketplace. The second is that real prices adjust
quickly to consumer preferences and market information. Every
year car manufacturers establish a price policy for each country that
includes an official price list. However, car dealers have considerable margin to adjust the prices down in order to meet sales targets.
At the end of the year, real (discounted) prices, can be dramatically
different from listed prices.
This paper is the first, to our knowledge, that uses discounted
prices to evaluate product efficiency. Using discounted prices
in the assessment of product efficiency is convenient, since it
allows the assumption that the products are comparably competitive. Therefore, if a product is found to be overpriced, given the
observable bundle of tangible attributes, we can infer that customers are willing to pay a premium for the intangible attributes
associated with the brand. This assumption allows us to identify
four brands (Audi, Volvo, Mercedes and Volkswagen) as those that
offer more intangible value to their customers. On the other hand,
other brands such as Dacia, Fiat, Citroën, Kia, Renault and Seat
base their competitiveness on offering a good bundle of tangible
attributes at a reasonably low price.
The research has some limitations that should be addressed
in future research. First, the sample has been limited to diesel
versions, which comprise the largest market share in Spain. However, including gasoline models could vary the assessment of some
brands (for instance, Japanese brands such as Honda and Toyota)
which may be more focused on gasoline models. Second, due to data
restrictions, we have used the same frontier to compare all the models in the sample. A larger sample would allow a finer comparison
E. González et al. / Investigaciones Europeas de Dirección y Economía de la Empresa 19 (2013) 1–7
by dividing the sample into segments by size (small, compact, large)
and by market orientation (generalist, premium). This segmentation could add valuable insights about the market and how brands
behave within each segment.
References
Banker, R. D., Charnes, A., & Cooper, W. W. (1984). Some models for estimating
technical and scale inefficiencies. Management Science, 39, 1261–1264.
Bolton, R. N., & Drew, J. H. (1991). A multistage model of customers’ assessment of
service quality and value. Journal of Consumer Research, 17, 375–384.
Brayman, J. M. D. (1996). Building a better business tool-customer value analysis.
In Paper presented at the AMA Conference, Conference on Customer Satisfaction
January 1996.
Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efficiency of decision
making units. European Journal of Operational Research, 2, 429–444.
Dodds, W. B., Monroe, K. B., & Grewal, D. (1991). Effects of price, brand, and store
information on buyers’ product evaluations. Journal of Marketing Research, 28,
307–391.
Doyle, J. R., & Green, R. H. (1991). Comparing products using data envelopment
analysis. OMEGA, 19, 631–638.
Emrouznejad, A., Parker, B., & Tavares, G. (2008). Evaluation of research in efficiency
and productivity: A survey and analysis of the first 30 years of scholarly literature
in DEA. Journal of Socio-Economic Planning Science, 42, 151–157.
Fernández-Castro, A. S., & Doldán, F. R. (2002). Aplicación del índice de productividad de Malmquist a la evaluación de productos. Revista Europea de Dirección y
Economía de la Empresa, 11, 173–180.
Fernández-Castro, A. S., & Smith, P. (1996). A mathematical programming approach
to product characteristics, Discussion Papers in Economics. The University of
York, 96/3.
Fernández-Castro, A. S., & Smith, P. (2002). Lancaster’s characteristics approach
revisited: Product selection using non-parametric methods. Managerial and
Decision Economics, 23, 83–91.
Gale, B. T. (1994). Managing customer value. New York: Free Press.
7
Holbrook, M. B. (1994). The nature of customer value: An axiology of services in the
consumption experience. In R. T. Rust, & R. L. Oliver (Eds.), Service quality: New
directions in theory and practice (pp. 21–71). Thousand Oaks: Sage Publications.
Hollingsworth, B., & Smith, P. (2003). Use of ratios in data envelopment analysis.
Applied Economics Letters, 10, 733–735.
Kopp, R. J. (1981). The measurement of productive efficiency: A reconsideration.
Quarterly Journal of Economics, 96, 477–503.
Kortelainen, M., & Kousmanen, T. (2007). Eco-efficiency analysis of consumer
durables using absolute shadow prices. Journal of Productivity Analysis, 28, 57–69.
Lancaster, K. J. (1966). A new approach to consumer theory. Journal of Political Economy, 74, 132–157.
Lee, J. D., Hwang, S., & Kim, T. Y. (2005). The measurement of consumption efficiency
considering the discrete choice of consumers. Journal of Productivity Analysis, 23,
65–83.
Oh, I., Lee, J. D., Hwang, S., & Heshmati, A. (2010). Analysis of product efficiency in the
Korean automobile market from a consumer’s perspective. Empirical Economics,
38, 119–137.
Papagapiou, A., Mingers, J., & Thanassoulis, E. (1997). Would you buy a used car with
DEA? OR Insight, 10, 13–19.
Papahristodoulou, C. (1997). A DEA model to evaluate car efficiency. Applied Economics, 29, 1493–1508.
Rust, R. T., & Oliver, R. L. (1994). Service quality: Insights and managerial implications
from the frontier. In R. T. Rust, & R. L. Oliver (Eds.), Service quality: New directions
in theory and practice (pp. 1–20). Thousand Oaks: Sage Publications.
Seiford, L. M., & Zhu, J. (2003). Context-dependent data envelopment analysis: Measuring attractiveness and progress. OMEGA, 31, 397–408.
Smirlis, Y., Despotis, D. K., Jablonsky, J., & Fiala, P. (2004). Identifying “best-buys”
in the market of prepaid mobile telephony: An application of imprecise DEA.
International Journal of Information Technology and Decision Making, 3, 167–177.
Staat, M., & Hammerschmidt, M. (2005). Product performance evaluation: A super
efficiency model. International Journal of Business Performance Management, 7,
304–319.
Sun, S. (2002). Assessing computer numerical control machines using data envelopment analysis. International Journal of Production Research, 40, 2011–2039.
Zeithaml, V. A. (1988). Consumer perceptions of price, quality, and value: A meansend model and synthesis of evidence. Journal of Marketing, 52, 2–22.