International Review of Environmental and Resource Economics, 2013, 6: 1–32
The Economics of Eco-Labeling:
Theory and Empirical Implications∗
Charles F. Mason
Department of Economics & Finance,
University of Wyoming, 1000 E. University Ave., Laramie, WY 82071
Over the past several years, environmental economists have been
increasingly attracted to the use of information as an alternative to
traditional methods for regulating externalities. An example of this
approach is ‘‘eco-labeling,’’ where a third party certifies firms’ prod-
ucts; this approach is particularly popular in practice, having been
adopted in a variety of countries. With this widespread adoption of
eco-labeling, a literature has developed in environmental economics. In
this paper, I survey the equilibria that may occur with eco-labeling,
and discuss the resultant welfare effects.
Keywords: Asymmetric information; eco-labeling; environmental
economics; signaling; testing.
JEL Codes: Q5, D8, L15
∗I thank, without implicating, two referees and the editor for constructive input that
greatly improved the manuscript. This paper was written in part during my visit at
the University of California-Santa Barbara, under the sponsorship of the University
of California Energy and Environmental Economics program, whom I thank for their
splendid hospitality. Work reported in this paper was funded in part by the National
Science Foundation, under grant # SES-0214160. Any opinions, ﬁndings, and conclu-
sions or recommendations expressed in this material are those of the author(s) and do
not necessarily reﬂect the views of the National Science Foundation.
ISSN 1932-1465; DOI 10.1561/101.00000054
2013 C. F. Mazon
Over the past several years, environmental economists have been
increasingly attracted to the use of information as an alternative to tradi-
tional methods for regulating externalities (Arora and Cason, 1996, 1999;
Konar and Cohen, 1997, 2001). The use of information as an alternative
regulatory strategy has been termed the ‘‘third wave’’ of pollution con-
trol (Tietenberg, 1998). A particularly popular example of this regulatory
approach is an environmental labeling program, which have emerged in a
wide range of countries (Karl and Orwatt, 2000; OECD, 1997; Vossenaar,
1997). Some of these certification programs have become quite popular, as
with the German ‘‘Blue Angel,’’ Japanese ‘‘Eco-Mark,’’ Swedish ‘‘Environ-
mental Choice,’’ and ‘‘Nordic Swan’’ programs (OECD, 1997) or the Ameri-
can ‘‘ENERGY STAR’’ label (Houde, 2012). These labels, typically referred
to as ‘‘eco-labels,’’ are often applied to products where consumers would
generally be individually unable to determine the environmental friendliness
of the product, for example the biodegradability of a paper product, or of
the production process itself. Many of the eco-labeling programs currently in
operation consider production-related criteria in their assessments of firms
that seek certification.
There is abundant evidence that consumers express a willingness to
pay a premium to ‘‘protect the environment’’ (Amacher et al., 2004;
Bjorner et al., 2004; Cairncross, 1992; Cason and Gangadharan, 2001;
Haji-Gazali and Simula, 1997; Levin, 1990; Wasik, 1996; Winterhalter and
Cassels, 1993).1While firms that use environmentally friendly production
techniques, which I refer to as ‘‘green firms,’’ would like to capitalize on
this demand, they face a problem of asymmetric information: Consumers
cannot typically tell the type of production process a particular firm has
used, so they cannot determine when it is green. Since the environmen-
tally friendly technique is generally more costly, firms would be disin-
clined to choose such a technique, with larger pollution flows resulting.
One possible remedy for this informational asymmetry is for firms to make
use of ‘‘eco-labeling.’’ With eco-labeling, a third party certifies a vendor
1Much of this evidence is based on surveys, i.e., it reﬂects stated preferences. As such, it is
conceivable that individuals claim to prefer green products, but in practice they seek out
less expensive products, irrespective of the associated environmental attributes. Empirical
evidence that labeling schemes can serve to protect the environment include Garc´ıa et al.
The Economics of Eco-Labeling: Theory and Empirical Implications 3
as being green, i.e., that its product results from an environmentally friendly
A parallel literature investigates the role of labeling in agricultural
markets, often with reference to food safety or genetically modified organ-
isms. Many of these papers consider the role of information provision
by a third-party; often this third party acts of his or her own volition,
i.e., without having been retained by the firm (Crespi and Marette, 2007;
McCluskey, 2000; Roe and Sheldon, 2007). In this setting, one can think of
the third party undertaking some sort of random monitoring, and then pro-
viding information to the general public. As with the eco-labeling problem,
these papers are based in a situation where the buyer is unlikely to have the
capacity to determine product characteristics even with repeated consump-
tion; such products are termed ‘‘credence goods’’ (Darby and Karni, 2001;
Feddersen and Gilligan, 2001). By its nature, the class of goods for which
eco-labeling is a potentially useful institution are credence goods. Unlike the
problem typically confronting buyers and sellers of agricultural goods, con-
sumers of eco-labeled products may well obtain no direct personal benefits
from the consumption of the product. Rather, the utility they get from con-
suming a green product —which is, in the end, the source of their enhanced
willingness to pay —is a form of altruism or a contribution to some public
good (Kirchhoff, 2000).
It is important to note here that two externalities are present: there is
the familiar environmentally-based externality, which will result from firms
using the environment as a costless input into their production process. But
there is also an externality related to imperfect information. The two are
linked: with imperfect information, dirtier firms produce more than perfectly
informed consumers would demand, while cleaner firms produce less than
perfectly informed consumers would demand. So while the ultimate goal
of eco-labeling is to address the former externality, its most direct impact
obtains from addressing the first externality.
Suppose there are two types of firms: those that are environmentally
friendly and those that are not. In this paper, I call the first type of firm
‘‘green’’ and the second type of firm ‘‘brown.’’ If certification is absolute,
i.e., errorless, then consumers know that any certified firms are green.
2This third party could be a governmental agency, an industry association or a non-governmental
organization. See Van’t Veld and Kotchen (2011) for a discussion of the similarities and diﬀer-
ences between these examples.
Accordingly, one might expect a first-best outcome to result. But the ability
of the eco-label to solve the asymmetric information problem depends
on the nature of the underlying supply and demand curves (Mattoo and
Singh, 1994; Sedjo and Swallow, 2002). Moreover, even if the label does
allow consumers to determine each firm’s environmentally friendliness, the
result need not be an improvement in environmental performance (Swallow
and Sedjo, 2000). This unpleasant result occurs because of the derived
demand for environmental services; raising demand for green firms’ final
products will increase those firms’ demand for inputs, which can have
unanticipated and undesirable general equilibrium effects. And there is a
subtle point here that was not clearly brought out in the early literature
on eco-labels: if a certification program can costlessly and perfectly inform
consumers as to a firm’s environmental friendliness, then in essence the
asymmetric information problem is assumed away.
Viewing certification as absolute only makes sense if the third party can
perfectly identify compliance with the eco-label’s avowed standards at a
reasonable cost, an assumption which is surely suspect. The alternative is to
think of certification as noisy. A simple way to model this effect by assuming
the certification process yields a positive report with some probability for
both green and brown firms, though one presumes that green firms are more
likely to pass the certification test than are brown firms. In addition, it seems
highly likely that pursuing certification is costly, if for no other reason than
the fact that the certifying agency must bear costs, which one expects are
passed along (at least in part) to those firms seeking certification. With these
points in mind, two questions seem relevant: First, what is the implication
of noisy certification for the resultant equilibrium? Second, how do various
parameters, such as the magnitude of certification costs or the accuracy of
the certification test, impact the equilibrium?
The paper proceeds as follows. In Section 2, I start discuss a deterministic
scenario; this variant corresponds to much of the the extant literature.
Depending on the various underlying market characteristics, introducing
a labeling program may or may not expand production of green units,
and may or may not deliver welfare improvements. In Section 3, I turn
to the more realistic framework, wherein certification is probabilistic. As
with the simple version, labeling need not expand the market, and even
if it does welfare gains are not guaranteed. Unlike the simple story, here
one finds a rich variety of potential equilibrium configurations that depend
on the key parameters associated with the certification test. In particular,
The Economics of Eco-Labeling: Theory and Empirical Implications 5
there are parameter configurations that support a separating equilibrium,
configurations that support a pooling equilibrium, and combinations that
support a partial pooling equilibrium. In Section 4, I discuss the extension
of a model to allow for endogenous choice of environmental friendliness; such
a scenario allows firms to respond to a labeling scheme by adopting more
environmentally benign technologies. In Section 5, I describe the situation
where firms have heterogeneous costs within each cohort. The potential for
endogenous labels, either because of multiple, competing labels or because
of endogenously selected standards, is discussed in Section 6. In Section 7,
I discuss briefly the existing empirical literature on eco-labeling. Section 8
offers some concluding thoughts, including the potential application of
certification to low-carbon technologies.
2 Deterministic Certiﬁcation
Most papers in the existing literature that investigates eco-labels assume
perfect (as opposed to probabilistic) certification, either explicitly or implic-
itly.3Matto and Singh (1994) offered the first serious attempt to model
eco-labeling in an economic framework. Their model implicitly assumes that
eco-labeling provides consumers fully accurate information regarding the
firm’s environmental friendliness, at little or no cost. This of course is an
approximation, and one that is arguably unrealistic. Nevertheless, Mattoo
and Singh’s work provides us with a useful starting point to discuss the
economics of eco-labeling.
Suppose there are two types of consumers: green consumers and brown
consumers.4Green consumers have a preference for products that are
environmentally-friendly, and so are willing to pay more for such products.
This so-called ‘‘green premium’’ forms the basis of the motivation for firms
to undertake eco-labeling in the first place. Brown consumers, on the other
hand, care only about price. Accordingly, there are two demand curves for
products that are in fact environmentally-friendly: one for green consumers
and one for brown consumers. In Mattoo and Singh’s model, there are two
potential demand curves for green consumers; provision of information about
3Representative examples include Amacher et al. (2004); Baksi and Bose (2006); Karl and
Orwatt (2000); Mattoo and Singh (1994); Robertson (2003, 2007); Sedjo and Swallow (2002);
Swallow and Sedjo (2000).
4Mattoo and Singh’s model has one of each type. One can think of these as prototypical con-
sumers, so that there individual demand curves are surrogates for market demand curves.
firms’ environmental attributes will shift out green consumers’ demand for
green products (while leaving their demand for brown products unchanged).
Mattoo and Singh interpret this effect as resulting from a triggering of the
green consumer’s environmental awareness following the provision of infor-
mation via the eco-label.5While individual green consumers are willing to
pay more for green products, the market demand curves represent horizontal
summations. Accordingly, one cannot make ex ante predictions about the
relative location of green and brown demand curves at the aggregate level.
The relative location of the two curves depends in part upon the number
of each type of consumer, so if there are many more brown consumers than
green consumers it is conceivable the brown demand curve lies above the
green demand curve. If, by contrast, the two demand curves are perfectly
elastic, then the green demand curve must lie above the brown demand
curve (as some consumers are willing to pay more for green products, and
no consumers are willing to pay more for brown products).
On the supply side of the market, the cost of doing business for brown
firms is presumably lower than that for green firms. This fact suggest that
the supply curve for brown firms lies below the supply curve for green firms,
but again the number of each type of firm matters. The market supply
curve for green firms is the horizontal summation of individual green firm
supply curves; if the number of green firms does not exceed the number of
brown firms are equal, then the green supply curve must lie above the brown
supply curve. But if there are sufficiently more green than brown firms, it
is conceivable the green supply curve lies below the brown supply curve.
In a market with no labels, the market equilibrium price balances total
quantity supplied (which combines output from green and brown firms) with
total quantity demanded (which combines purchases by green and brown
consumers). Call the resultant price P0, which for expositional convenience
I refer to as the ‘‘no-information price.’’ Now suppose consumers can be per-
fectly informed about each firm’s product type. In this way, green products
5This assumption implies that green consumers’ willingness to pay in a market where there are
no labels depends only on the total amount of the product consumed, and not on the distri-
bution of consumption between products produced by green and brown ﬁrms. An alternative
assumption, and one that is much closer in spirit to the vast literature on the economics of
information, would be that the amount green consumers are willing to pay reﬂects a weighted
average of the values placed on each of the two types of product. The weight placed on the
value for green products might be interpreted as the probability the product selected by a
consumer turns out to be green. With no information regarding the ﬁrm’s type of product, this
weight would presumably equal the fraction of green output in the market.
The Economics of Eco-Labeling: Theory and Empirical Implications 7
enter into a separate market, one that contains only green products and
form and hence involves the ‘‘green’’ demand curve discussed above. As only
green sellers are present in this market, the supply curve reflects the costs of
doing business for green firm. The resultant equilibrium price is paid only to
green firms; call that price P0
G. In the other market, one supposes that only
brown firms are present, and so the demand curve in this market reflects
consumer’s willingness to pay for brown products; the resultant price would
B. If the relative locations of green and brown supply curves on the one
hand, and green and brown demand curves on the other hand, is such that
there is excess demand in the green market at the no information price, then
the market equilibrium prices with labeling must satisfy P0
as illustrated in Figure 1. Notice that all agents strictly prefer the actions
they take in this equilibrium: environmentally conscious (green) consumers
prefer green products to brown at the observed prices; green sellers strictly
prefer to obtain the eco-label; and brown consumers, who prefer to purchase
the cheapest available products, all purchase brown products.
While this result might seem intuitive, Mattoo and Singh argue that
because the relative positions of green and brown supply and demand curves
cannot be predicted ex ante, it is conceivable that there will be excess sup-
ply, as opposed to excess demand, in the green market at P0, as in Figure 2.
If that is the case, then the green price cannot rise above P0, and a com-
pletely different scenario will emerge. Because brown consumers will happily
buy whichever product is cheaper, one cannot have P0
B. The result is
that arbitrage by brown consumers will force the prices together. Because
green consumers are willing to pay more for a labeled green product than for
Figure 1. The market with perfect certification.
Figure 2. Mattoo and Singh’s counterexample.
an unlabeled green product, Mattoo and Singh argue that P0
eco-labeling can push up the price of both products. Notice that in this equi-
librium, green sellers are indifferent between the market they enter (as they
would receive the same price in each segment), as are brown consumers;
green consumers strictly prefer to purchase certified products.
There are two problems with Mattoo and Singh’s line of reasoning. First,
as I pointed out above, assuming perfectly accurate information about each
firm’s environmental attributes can be costlessly delivered to consumers
essentially assumes away the asymmetric information problem. But even if
costs are incorporated into the analysis, there is a second problem: In the
situation where P0
Bgreen firms would be disinclined to pursue costly
certification (Swallow and Sedjo, 2000; Sedjo and Swallow, 2002). If there
are costs associated with certification, as seems most likely, then these costs
would have to be offset by gains that certification yields, so that the green
price would necessarily exceed the brown price. In that event, one would
observe both types of sellers, green and brown, participating in the erst-
while brown market. The result would be a price in the brown market that
is somewhat somewhat smaller than P0(because of the increase in supply);
at the same time, with some green sellers switching out of the green mar-
ket, products labeled as green will fetch a higher price than P0. Note that
these effects only depend on the existence of labeling costs, not whether
the costs are fixed or variable. As in the combination proposed by Mattoo
and Singh, green sellers and brown consumers are indifferent between their
potential actions, while green consumes strictly prefer to purchase labeled
The Economics of Eco-Labeling: Theory and Empirical Implications 9
3 Probabilistic Certiﬁcation
While modeling certification as perfect, i.e., errorless, has the advantage
of relative simplicity, this approach implicitly assumes the third party can
perfectly identify compliance with the eco-label’s standards at a reasonable
cost. In practice, this seems unrealistic. Consider for example, the Dutch
agency Stichting Milieukeur: when they cannot determine the environmental
effect of a particular product in a certain dimension, they ‘‘consider the mat-
ter in qualitative terms’’ (Giezman and Verhees, 1997). Similarly, it seems
unlikely that the certifying organization could be certain the firm always
uses an environmentally friendly technique, nor does it seem likely that the
monitoring scheme would be able to errorlessly detect any violations (Dosi
and Moretto, 2001; Hamilton and Zilberman, 2006). Indeed, third party
eco-labeling programs often gauge the firm’s environmentally friendliness
by random monitoring, in which case certification must is surely noisy: the
third party could mistakenly certify some brown firms as environmentally
friendly, or that some environmentally friendly firms could find certification
There is also considerable doubt that the certification standards
are perfectly correlated with ‘‘environmental friendliness’’ (Arda, 1997;
Morris, 1997). This issue is relevant whenever there are environmental
considerations at multiple stages in a product’s life cycle (extraction of
raw ingredients, production, packaging, consumption, and disposal) (Karl
and Orwatt, 2000; OECD, 1997). As an example, paper products might
be produced in a developing country from virgin timber with a relatively
clean production process, and in a developed country using recycled paper
but a less clean production process. One can easily imagine similar envi-
ronmental impacts arising from these two approaches, but if an eco-label
focuses on the amount of recycled paper used it would favor the devel-
oped country’s technique. Similar examples might include energy efficiency
(Houde, 2012) and the recent dispute between the United States and vari-
ous countries in south-eastern Asia regarding shrimp harvesting (Zhang and
Assuncao, 2004). In this context, one can imagine a locus of combinations of
product attributes which contribute to the firm’s environmental footprint,
6For example, a World Trade Organization case in the late 1990s found Brazilian textile pro-
ducers had an unduly diﬃcult time certifying that their products did not use pesticides
with consumer utility tied to this footprint. As eco-labeling schemes often
impose a specific weighting scheme for the various elements, which need
not match consumers’ preferences, inevitably there will be cases where con-
sumers place higher utility on an uncertified combination than they would
place on some certified combination. In such a scenario, certification surely
cannot be thought of as errorless.
One can imagine modeling probabilistic certification in a handful of ways.
A simple variation would entail randomization by green firms, along the
lines described above in the context of Sedjo and Swallow (2002). Ibanez
and Grolleau (2008) consider such a scenario; they analyze a three-stage
duopoly game: in the first stage each firm chooses a production technology
(e.g., brown or green), in stage 2 green firms decide whether or not to label
their product, and in stage 3 consumers make purchase decisions. In their
model brown firms are perfectly identified, but some green firms will remain
unlabeled. Alternatively, firms might use different forms of capital in their
production process, with environmental friendliness related to capital stock
composition, and where the probability of obtaining an eco-label is tied to
the firm’s environmental capital stock (Dosi and Moretto, 2001). In a third
variation, the certifying agency conducts ex post audits of firms that have
been awarded a label or that self-reports itself as green (where, presumably,
relatively greener firms are less likely to determined to be in violation);
Baksi and Bose (2006) and Hamilton and Zilberman (2006) consider this
In general, uncertainty regarding certification can be modeled by assum-
ing the certifying agency uses a certification test. One can think of a test
that involves the monitoring of some attribute of the production process,
such as emissions, that is correlated with the production technology. Since
it is prohibitively costly to monitor continuously, the third party monitoring
is conducted in a fashion analogous to random monitoring of emissions by
a government agency. Uncertainty from the certification test can be mani-
fested in two ways: One imagines that some green sellers will fail to obtain
certification (i.e., there can be false negatives) while at the same time one
imagines that some brown sellers will achieve certification (i.e., there can
7One might also imagine a model where ﬁrms choose to be green or brown, and then a third party
probabilistically identiﬁes the ﬁrm’s environmental friendliness — either because it randomly
monitors, or because its monitoring technology is imperfect (Feddersen and Gilligan, 2001;
Lizzeri, 1999; McCluskey, 2000). Such a model has some common features to the scenarios I
consider in this paper, one aspect it does not include is the decision to seek certiﬁcation.
The Economics of Eco-Labeling: Theory and Empirical Implications 11
be false positives). These probabilistic appraisal errors need not be large
in magnitude to have an effect on the market equilibrium prices. In any
event, it seems reasonable to expect that that environmentally friendly firms
would be more likely to obtain certification than environmentally unfriendly
firms, i.e., the probability with which a brown seller achieves certification
will surely be smaller than the probability with which a green seller achieves
certification. These attributes are present in Mason (2006, 2011). In his eco-
labeling model, the certification process passes green firms with a probability
φG, and brown firms with a probability φB, where φG>φ
There are three classes of equilibria that might obtain, depending on the
various parameters of the model.8In discussing these classes, I highlight the
parameter combinations that are required in order to support this outcome;
the reader can make his or her own judgment as to the empirical relevance
of such parameter combinations. The first class is a separating equilibrium:
all green sellers and no brown sellers pursue the eco-label. This class is qual-
itatively equivalent to a regime where the eco-label perfectly identifies green
products, as in the literature discussed in the previous section. The second
class is a pooling equilibrium, where all sellers pursue the eco-label. The
third class involves a hybrid or ‘‘partial pooling’’ equilibrium, wherein one
class of sellers plays a pure strategy and the other plays a mixed strategy.9
This third class is perhaps the most interesting and empirically relevant.
In fact, a number of critics of eco-labeling programs have complained that
the labels are not ‘‘pure,’’ in the sense that some ‘‘unworthy’’ products
been certified. While such erroneous certification might be the result of
fraud on the part of sellers who obtain certification, and then purposely
change their production scheme, it is also plausible that the certification
test is subject to false positives. On the other hand, there seems to be little
evidence to suggest that all brown firms are attempting to obtain certifi-
cation. Thus, a middle ground, in which the test cost is neither too large
8Loosely speaking, these three possible conﬁgurations can be linked to the cost of certiﬁcation.
If certifying costs are suﬃciently large, sellers of brown units do not pursue the eco-label, and a
separating equilibrium emerges. (One can think of the examples with deterministic certiﬁcation,
discussed in the previous section, as an example of this class of equilibrium.) If certifying
costs are suﬃciently small, all sellers seek certiﬁcation, and a pooling equilibrium results. For
intermediate values of certifying costs, or for particularly large values, the equilibrium is partial
9In fact, the variation of Sedjo and Swallow (2002) where some green sellers opt out of the
certifying market is an example of this variant, as is one of the versions considered by Ibanez
and Grolloeau (2008).
nor too small, seems likely to be empirically significant. The third class of
equilibrium has important features in common with two types of scenar-
ios that have generated interest in the environmental regulation literature
at least since Roberts and Spence (1976) and Kwerel (1977): the situa-
tion where polluters are tempted to under-report emissions, and situations
involving environmental fraud (Hamilton and Zilberman, 2006). An inter-
esting and counter-intuitive outcome of the partial pooling equilibrium is
that small increases in certifying costs can make green sellers better off, as
I discuss below.
The socially efficient level of green (respectively, brown) production
equates supply with full-information price PG(respectively, PB). Because
demand is perfectly elastic, the socially efficient combination also maximizes
industry profits. With an inefficiently large quantity of brown products and
an inefficiently small quantity of green products, as in the no-information
equilibrium, any change which reduces the volume of brown units while
raising that of green units would reduce deadweight loss and raise indus-
try profits. These gains must then be compared against aggregated testing
costs to determine the net gain from the information associated with cer-
tification. In addition to the market failure due to asymmetric information
there is a second market failure associated with environmental damages.
This adverse eternal effect will be larger for brown firms than for green
firms. Because the difference between prices for green and brown units are
the result of consumer preferences explicitly reflect divergent externalities
for green and brown firms (which would require consumers explicitly under-
stood the nature of those externalities), this second effect is not reflected in
To flesh out the nature of equilibrium under probabilistic certification,
consider the following scenario.10 Let Arepresent the cost of taking the
certifying test, common to all firms.11 Suppose the demand curve for green
products lies above the demand curve for brown products, with both curves
being perfectly elastic, with prices fixed at PGand PBfor green and brown
products, respectively. For now, assume there are exogenously fixed number
of firms, N, with NBbrown firms and NGgreen firms. Production costs
are convex in output; the cost curve for all green firms is cG(q) and the
10 This discussion is loosely based on Mason (2011).
11 This precludes, for example, schemes where eco-labeled ﬁrms that are found to be brown are
required to pay a penalty to the certifying company, as in the Canadian Environmental Choice
Program (Wasik, 1996). For an analysis of a model with such ﬁnes, see Kirchhoﬀ (2000).
The Economics of Eco-Labeling: Theory and Empirical Implications 13
cost curve for all brown firms is cB(q). Firm’s supply curves are therefore
upward sloping, for both green and brown firms.12 To fix ideas, suppose
cG(q)=αcB(q), with α>1 (reflecting the fact that green production is more
expensive), and αcommonly known by all market participants; cB
(reflecting positive, upward-sloping marginal costs); and αcB
There are three possible outcomes from the certifying process. First, the
firm could be certified, in which case it would receive the price Pc. Second, it
might fail the certification test, in which case it would receive the price Pf.
Third, the firm might not pursue certification, in which case it would receive
the price Pun. All these prices are formed endogenously, through rational
expectations. The three prices therefore depend on consumers’ predictions
of three conditional probabilities: that a randomly selected unit is green,
given that its characterization as c,f,orun. Plausibly, consumers might
expect that failed units and untested units will be lumped together as ‘‘unla-
beled,’’ in which case only two prices prevail: Pcand Pun.14 Of course,
whether consumers would form such pessimistic expectations is an empir-
ical matter; some evidence on this point is provided by Indonesia’s Public
Disclosure program. Under this scheme, firms are labeled with one of five
color-coded factors, ranging from black (the dirtiest factories) to gold (the
cleanest plants). As Tietenberg (1998, Table 1) reports, the lion’s share of
plants are in the 2nd or 3rd dirtiest category, so could interpret the 3rd
dirtiest category as containing firms that have passed the test and the 2nd
dirtiest category as containing firms that are unlabeled.15
12 For equilibrium outputs to be well-deﬁned, marginal revenue must be downward-sloping or
marginal costs must be upward-sloping. As the former option is inconsistent with price-taking
behavior, constant marginal costs require some form of oligopoly behavior.
13 This assumption ensures that some green ﬁrms will participate in any equilibrium, and so
rules out the famous “lemons equilibrium” (Akerlof, 1970), in which all higher-quality ﬁrms
are driven from the market.
14 As a general rule, information on certiﬁcation applications that are denied is generally unavail-
able (Vossenaar, 1997). And even if this information were available, should consumers believed
that any seller who marketed its units as “failed” was in fact brown, then all sellers who failed
would (weakly) prefer to market their units as “untested.” In that scenario, there would be
no units would be oﬀered for sale at the failed price, and so Bayes’ rule could not be applied
(Mason and Sterbenz, 1994). Occasionally this feature — which often crops up in signaling
games — can be resolved by applying a reﬁnement such as the Intuitive Criterion (Cho and
Kreps, 1987) or one of the Divinity Criteria (Banks and Sobel, 1987). However, in the case at
hand these reﬁnements have no bite.
15 Similarly, in India’s Green Rating Program, ﬁrms receive a ranking of 0 to 5 leafs, depending
on their level of environmental performance; most ﬁrms in the pulp and paper sector receive
either 1 or 2 leaves (Powers et al., 2011). For discussion of a model with multiple levels of
environmental quality, see Roe and Sheldon (2007).
In this framework, the equilibrium is intimately tied to two probabilities:
the probability that a randomly selected unit is green, conditional on it
being eco-labeled (µ), and the probability that a randomly selected unit is
green, conditional on it being unlabeled (ν). From Bayes’ law, these posterior
where θis the fraction of units on offer in the no-information equilibrium that
are green, pr(c|G) is the probability that a unit will be certified, conditional
on its seller being green, pr(c) is the marginal probability of observing a
certified unit, pr(un|G) is the probability that a unit will be unlabeled,
conditional on its seller being green, and pr(un) is the marginal probability
of observing an unlabeled unit. It turns out that φG>φ
Rational expectations prices in this framework are based on the conditional
probabilities µand ν16:
It also turns out that the expected price for brown products will be less
than the no-information price, while the expected price for green firms who
pursue the eco-label exceeds the no-information price (Mason, 2006). This
effect induces green firms to increase their production, so long as there are
some green firms who pursue certification —as is true in any equilibrium
that differs from the no-information equilibrium.
If the firm produces a positive output, then that output will equate
marginal cost to the price the firm anticipates receiving. If the firm has
its product tested, and passes, then the firm anticipates receiving Pc;ifit
enters the untested segment, or if it fails the test, then it anticipates receiv-
ing Pun. Let πkc represent the profit a type k=Gor Bfirm earns when it
receives the price Pcand produces qkc, and denote the profit a type kfirm
16 It can also be shown that µ>θ>ν (Mason, 2006). It then follows the test leads to a higher
price for eco-labeled units and a lower price for unlabeled units than the no-information price
un). Thus, the original bracketing result from Mattoo and Singh (1994) obtains,
although in a framework where agents employ rational expectations.
The Economics of Eco-Labeling: Theory and Empirical Implications 15
earns when it receives the price Pun and produces qku by πku. The difference
between the expected payoff from testing and the certain payoff from not
k−πku, measures the anticipated gain from pursuing the
eco-label. This expected gain depends on the probability of passing the test,
the nature of costs, the cost parameter, and the test cost. Under plausible
B, so that green firms are more inclined to seek
the eco-label than are brown firms.17 In this event, all green units pursue
certification whenever any brown firms do so.
Before turning to a discussion of comparative statics, I first discuss the
partial pooling equilibrium in more detail. In the version where brown sell-
ers are indifferent, one observes all green and some brown sellers pursuing
certification. But for this to be an equilibrium, WB= 0, which provides
important information linking prices in the two market segments:
B)] = A/φB,(1)
Bis the profit-maximizing output for a certified brown firm (i.e.,
B)=Pc) and q∗∗
Bis the profit-maximizing output for an uncertified
brown firm (i.e., cB
Partial pooling equilibria have an interesting feature that is absent in both
the separating or pooling equilibria: increases in test cost can raise social
welfare. In the partial pooling equilibria, an increase in Awould force the
certified price to increase by an amount that keeps brown firms indifferent
between pursuing certification on the one hand and entering the untested
segment of the market on the other. But the certified price can only increase
if the conditional probability that a certified product is green also increases,
which then requires a smaller fraction of brown firms attempt to masquerade
as environmentally friendly firms. It then follows that green firms produce
more, on average, which is socially attractive: Because consumers value green
products at a higher level than the certified price; an increase in green pro-
duction then reduces the deadweight loss associated with under-provision of
green output. At the same time, the expected output of brown sellers must
fall. This too is socially attractive: Because consumers value brown prod-
ucts at a lower level than the certified price; a decrease in brown production
17 Since it is less costly to produce brown units than green units, is clear that the proﬁts available
in the unlabeled segment are strictly greater for brown ﬁrms. Accordingly, if the proﬁts a
certiﬁed green ﬁrm earns are at least as large as those earned by a certiﬁed brown ﬁrm,
B. This is not a terribly restrictive assumption.
then reduces the deadweight loss associated with over-provision of brown
output. In addition, since green firms are associated with smaller environ-
mental damage than are brown firms, total externalities are likely to be
smaller with the larger test cost.
Regarding the comparative static effects of changes in the pass probabili-
ties, an increase in φGor a decrease in φBunambiguously lowers ν, which in
turn causes a reduction in the unlabeled price, Pun. While this is seemingly
welfare-enhancing, the indifference relation for brown firms requires a com-
pensating reduction in the certified price, Pc, in compensation for the lower
value of Pun. The result is that increases in φGlead to a reduction in Pc, the
equilibrium price paid to certified units. In contrast, a reduction in φBhas
two effects: it raises the right-hand side of Equation (1), which induces an
increase in Pc; it also pushes Pun down, which induces a reduction in Pc.Itis
straightforward but tedious to show that the former effect is more important
than the latter effect, so that the net effect is an increase in Pc. Accordingly,
the net effect on equilibrium prices from a marginal decrease in φBexceeds
the effect from a marginal increase in φG.
Figure 3 characterizes the welfare implications of the three types of equi-
libria.18 The graph plots the difference between net surplus in the testing
equilibrium and net surplus in the no-information equilibrium against test
cost. There are three ranges of note. For small test costs, the equilibrium
configuration entails pooling; in this regime, all sellers seek certification,
and so increases in the test cost do not change the configuration of firms in
the market, nor do they impact output levels. As a result, higher test costs
unambiguously lower net surplus. But as the test cost rises, at a certain
point brown sellers start to drop out of certified segment of the market. In
this second regime, the equilibrium configuration is partial pooling. Here,
increases in test cost raise profits for green sellers, while lowering profits for
brown sellers that seek certification. Initially, almost all brown sellers seek
certification, and so the (negative) impact on these sellers dominates the
(positive) impact on green sellers, so that net surplus continues to decline —
albeit at a somewhat slower rate than in the pooling equilibrium. Eventually,
the fraction of brown sellers that pursue the eco-label falls to the point where
18 The ﬁgure is based on an example with iso-elastic costs, of the form ck(qk)=αkq1.75
αB= 1 and αG=1.05. Because demand is perfectly elastic, net surplus corresponds to
The Economics of Eco-Labeling: Theory and Empirical Implications 17
Figure 3. Equilibrium welfare effects.
the impact on green sellers becomes dominant; further increases in the test
cost then raise net surplus. This effect continues until the point where all
brown seller eschew certification, i.e., the market configuration converts to
a separating equilibrium. At this point, only green sellers seek the eco-label;
further increases in the test cost do not change this fact, nor do they impact
outputs, and so —like in the pooling equilibrium configuration —further
increases in test cost serve to lower net surplus. A point of note here is that
there is a range of test costs where net gains are positive, and a range of test
costs where net gains are negative. That is, eco-labeling can either increase
or decrease net surplus.
4 Ecolabeling and Endogenous Types
In the short-run, firms cannot adjust their type: they are either brown or
green. An investigation of what happens in the long run, when firm type
is endogenous, is facilitated by allowing firms to choose their type prior to
deciding whether or not to pursue certification. To facilitate comparison
with the model in the preceding section, I suppose initially there are N
identical firms remains fixed at N, where the number of green firms, NG,is
chose endogenously. Each firm as makes decisions at three stages: first, they
choose their type; based on their type they choose whether to have their
products tested; finally, they choose an output.
As in the earlier model, untested products sell at the unlabeled price.
Because it is cheaper to produce using the brown technology, no firm that
chooses to be green will then choose to not have their product tested. On the
other hand, firms that choose to be brown might choose to be tested or not.
As a result, there are three combinations of potential interest: the firm might
choose to be green and have its product tested; it might choose to be brown
and have its product tested; or it might choose to be brown and not have
its product tested. If no firm chooses the second or third cohort, all tested
products must be green, and therefore both the certified and unlabeled prices
must equal PG. But then a firm that switched into the second cohort (brown,
tested) would receive the same price but bear lower costs, and so its profits
would rise. Therefore, it cannot be an equilibrium for all firms to choose to
be green. If no firm chooses to be green, all products sell at the price PB.
In this scenario, all firms choose the third cohort (brown, do not test). This
could be an equilibrium, but only if consumers hold implausibly pessimistic
expectations about tested products.19
Because it cannot be an equilibrium for all firms to elect to be green,
and an equilibrium where all firms are brown depends on implausible con-
sumer expectations, in equilibrium firms must be indifferent between the
first cohort (green, test) and either the second cohort (brown, test) or
the third cohort (brown, do not test). Suppose firms are indifferent between
the first and third cohorts. Then all tested products are green, so that the
certified price must be PG; the unlabeled price would be less than PG. While
this scenario has some similarities to the partial-pooling equilibrium dis-
cussed above, there is an important difference. Here, the indifference between
first and third cohorts requires expected profits from testing for green sellers
19 Because no products are tested, there is no data upon which to form expectations. But then
no expectation is inconsistent with the data — an awkwardness that can arise in signaling
games — so that the prediction that any certiﬁed unit must be brown cannot be ruled out by
Bayes’ law. If one were to insist that consumers believed that passed units were more likely
to be green than brown, the certiﬁed price would exceed the unlabeled price. Consider then
the thought experiment whereby the diﬀerence between the proposed probability a passed unit
was green and the proposed probability a passed unit was brown was increased to ever-larger
amounts, at a certain point it would pay some ﬁrms to deviate from the putative equilibrium,
by choosing to be green. This idea is an application of the divinity criterion (Banks and
Sobel, 1987); it applies so long as the test cost does not exceed the diﬀerence between PG
and Pun (in which case the test would be so costly that no ﬁrm would ever be willing to
The Economics of Eco-Labeling: Theory and Empirical Implications 19
to equal certain profits from not testing for brown sellers (with both larger
than the expected profits from testing for brown sellers). It follows that this
can only be an equilibrium for sufficiently large test costs.
The final possibility is for sellers to be indifferent between the first and
second cohorts (green, test; brown, test). Although green sellers pass the
test with greater probability, brown sellers have lower costs and hence would
stand to earn larger profits, conditional on passing the test. For sellers to be
indifferent between choosing to be green and test on the one hand, or brown
and test on the other, these two effects must exactly offset:
B)] −(1 −φG)[Pun q∗∗
Gis the profit-maximizing output for a certified green firm (i.e.,
G)=Pc) and q∗∗
Gis the profit-maximizing output for an uncertified
green firm (i.e., cG
G)=Pun). Because Pcdepends on µwhile Pun depends
on ν, Equation (2) induces a relation between µand ν.
The probabilities µand νcan also be calculated from Bayes Law; for a
given value of NGthis induces a unique relation µ(ν) between the values
of µand ν. The equilibrium configuration must satisfy both Equation (2)
and the relation µ(ν). There are three potential equilibria: one that contains
only brown units and two equilibria containing some brown units and some
green units. The equilibrium with the smaller positive number of green units
is unstable, as a small decrease (respectively, increase) in the number of
green units would create pressures leading to further decreases (respectively,
increases) in the number of green units. On the other hand, the equilibrium
with the larger positive number of green units is stable.
By manipulating Equation (2) and µ(ν), the fraction of sellers that choose
to be green can be expressed in terms of the equilibrium configuration of
µand ν. This relation depends on the various exogenous parameters (the
cost of the certification test, the two pass probabilities, and the difference
between the price consumers would be willing to pay for green and brown
products). The relation between the price premium and the fraction of sell-
ers that choose to be green in the stable equilibrium is depicted in Figure 4,
for three different values of φB: 0.1, 0.2, and 0.3. Notice that the relation
between the price premium and the fraction of sellers that choose to be
green is increasing and concave (For a given value of φB). This is intuitive:
increases in the extra amount consumers would pay for green units make
Figure 4. Impact of changes in Green price premium.
it more attractive to choose to be green, which must increase the fraction
of sellers that choose to be green in equilibrium. However, because all sell-
ers can not be green in equilibrium, increases in the price premium must
yield ever-smaller marginal incremental increases in the number of green
sellers. Note also that increases in φBmake it less attractive to be green,
and so decrease the fraction of sellers that choose to be green in equilibrium.
Increases in φGhave the opposite effect —they make it more attractive to
be green —and so increase the fraction of sellers that choose to be green in
5 Heterogeneous Costs
To this point, I have focused on a scenario where all green sellers are iden-
tical, i.e., they all have the same marginal cost curve; likewise, all brown
sellers have the same marginal cost curve. One can easily imagine, however,
a scenario where sellers within a given cohort vary in terms of their effi-
ciency. In such a setting, each firm iof type k=B, G will be characterized
by a cost parameter αki, with the cost parameters falling within a type-
dependent range [αk, αk]. Without loss of generality, let αB= 1, so that all
costs are viewed as relative to the most efficient brown seller. Letting this
The Economics of Eco-Labeling: Theory and Empirical Implications 21
seller’s cost function be cB(q), the cost function for type kseller iwould be
The probability distribution functions over these cost parameters are
G; the associated cumulative distribution functions are FBand FG.
Because it is cheaper to produce using the brown technology, a sensible
assumption is that fBﬁrst-order stochastic dominates fG, which is to say
that for any value α∈[αB, αG] it is the case that FB(α)≥FG(α), with
strict inequality for at least some values of α. In other words, there is a
tendency for brown sellers to have lower cost parameters.
In this setting, the labeling equilibrium will typically involve cutoff rules:
all sellers of type kwith cost parameters at or below some ˜
certification, while the remaining type ksellers will go directly into the
unlabeled market (Mason, 2006). To the extent the certification test provides
useful information to consumers, i.e., it is correlated with environmental
friendliness, one expects the cutoff values to satisfy ˜
αG. In this way,
eco-certification provides an avenue for a greater fraction of green sellers to
convey their status to consumers.
Though such a regime is more involved, and the analytics more demand-
ing, than the model described above it is easy to see that the nature of
equilibrium is qualitatively similar to the partial pooling equilibrium, in that
there will be some brown products that obtain the eco-label, and some green
products that are marketed as unlabeled. The key distinction is that each
seller will generally have a strict preference regarding labeling; in particular,
those brown sellers that pursue the eco-label will generally prefer to do so, as
compared to the previous model wherein they were indifferent. That point
noted, from a population perspective, it will appear as if brown sellers are
indifferent, in the sense that only a fraction of them pursue certification.
Moreover, the key comparative statics identified above carry over; in partic-
ular, more costly tests disproportionately screen out brown sellers, thereby
raising expected profits for green sellers.
This variation is readily extended to a long-run setting, where sellers can
choose their production type. Now, each firm iis ‘‘endowed’’ with a pair
of cost parameters αBi,α
Gi. The firm faces two decisions: it chooses a type
(brown or green) and a certification strategy (seek the eco-label or go directly
into the unlabeled market). It is easy to see that no seller would choose to
be green if it were planning to enter the unlabeled segment, so the decision
boils down to a comparison of the three combinations: (brown, unlabeled);
(brown, test); (green, test). Depending on the firm’s cost parameters, it may
choose any of these three combinations. Accordingly, there will be a cutoff
rule describing the set of cost parameters wherein the firm prefers to be
brown, but is indifferent between seeking certification and not; there will be
another frontier describing the set of parameters wherein the firm intends
to seek certification, but is indifferent between producing with the brown or
with the green technology.20
The main feature of such a model is that the presence of an eco-label
provides a benefit to switching from brown to green that does not appear in
a model without the label. Indeed, the generic equilibrium in the endogenous
type model with heterogeneous firms includes some green firms, whereas the
equilibrium with no label can only include brown firms. (Since it is cheaper
to be brown than green, without a label no firm would opt to be green). As
such, there is a clear long-run benefit associated with eco-labeling.
6 Endogenous Labels
The discussion above is couched in a framework with a single exogenous
label. In practice, however, there can be multiple competing labeling pro-
grams, which implies a degree of choice is available to the labeling agency
in terms of the attributes it screens for. There is the additional issue that
raising the stringency of certification standards may serve to screen out a
greater number of sellers, making the label more exclusive and potentially
more valuable. Finally, the pricing of its services is a margin the certification
agency may be able to exploit.
The notion that there may be multiple labels underscores the point I made
above regarding the potential inaccuracies in certification tests: if a certifi-
cation test were fully accurate there would be no scope for any subsequent
labels. It also provides a natural linkage to the idea that labeling organiza-
tions must summarize information from multiple sources of potential envi-
ronmental friendliness. With that in mind, one can think of a product as a
bundle of attributes, with values associated with each of the multiple dimen-
sions. One product may have a greater tendency to use recycled materials,
while another may use less energy in the manufacturing process (thereby
20 In principle, the ﬁrm might also be indiﬀerent between (brown, do not test) and (green, test),
though this seems less likely to obtain in practice. See Mason (2006) for discussion.
The Economics of Eco-Labeling: Theory and Empirical Implications 23
inducing a lower carbon footprint); a third might use packaging products
that are more readily degradable. If only one label is available, the certifica-
tion standards underlying that label will necessarily represent a set of value
judgments regarding the importance of each of these attributes. But con-
sumers may differ in the weight they would attach to these attributes, and
this heterogeneity provides some scope for the presence of multiple labels.
In a world with multiple competing labels, three questions merit scrutiny.
First, there is the issue as to whether multiple labels will provide con-
sumers with more accurate information. At one level the answer seems
transparent —more information is surely better. But curiously many pol-
icy pundits fear the presence of multiple labels, arguing that consumers will
suffer from ‘‘information overload.’’ In its most dangerous manifestation,
this overload is alleged to drive consumers to ignore labels altogether, so
that the original motivation (providing an avenue for green firms to convey
that information to potential buyers) is utterly undermined. This sort of
argument has its roots in behavioral economics; whether it has any merit is
an open question.
The second question has to do with the nature of competition between
the entities offering labels. One might imagine a game between the orga-
nizations; the nature of the equilibrium in this game will depend on the
players’ perceptions as to consumers’ tastes, the motivations of the play-
ers and the timing of moves. An obvious line of inquiry is one in which
one player is a government organization and the other player is a private
entity.21 In such a setting, it is most plausible that one agency moves first,
perhaps the private entity, with the other entity responding. For example,
one might imagine a political economy story in which interest groups pres-
sure a government regulator to offer some sort of environmental label to
offset perceived bias on the part of industry-sponsored labeling program.
This sort of interaction amongst labeling organizations does occur, for exam-
ple, in the forest product sector (Sasser et al., 2006). In general, the presence
of multiple labels is likely to raise industry profits because of the added
flexibility to signal firm type. On the other hand, if the industry-sponsored
label has a first-mover advantage, as seems likely, it is conceivable that this
21 Roe and Sheldon (2007) discuss the potential for a ﬁrm to seek private certiﬁcation in the event
it dislikes the results of a government test result; they do not discuss strategic interactions
between government and private certiﬁers. Fischer and Lyon (2012) investigate such strategic
interactions, as well as motives for new certifying ﬁrms to enter.
labeling organization will undertake actions that induce the second-mover,
be it a government regulator or a non-governmental organization (NGO)
to relax its standards slightly; in that event, it is entirely possible that the
presence of multiple labels will produce a deleterious environmental effect
(Fischer and Lyon, 2012).
The third question relates to the potential equilibrium configuration of test
cost and accuracy. To the extent that more accurate tests are more expen-
sive, labeling firms face a natural tradeoff associated with strengthening the
accuracy of their test. As I noted above, an increase in test accuracy that
lowers the potential for false positives would tend to raise expected producer
surplus, and presumably then would increase firms’ collective willingness to
pay to be tested. This effect would induce certifying firms to seek out more
accurate tests. But the extra accuracy increases the certifier’s costs test,
blunting the incentive to use more accurate tests. The ultimate outcome of
these competing pressures is not clear ex ante; while it is conceivable such
pressures might create a push toward very accurate tests, it is also possi-
ble the pressures would move the certifying industry towards noisier (and
In the context of a single label, the certification scheme may be within
the organization’s ability to control. One can think of conscious efforts to
make the test more stringent, for example. Such actions will likely impact
the demand for the label, and so can impact the labeling organization’s
profits. If the organization is a government regulator or an NGO, it is not
at all clear these profit considerations will matter. But if the organization
is a private entity, they may matter very much. Here the essential issue is
one of comparing the impact of a more stringent test upon revenues with
the impact upon costs. On the revenue side, tighter standards are likely
to reduce the number of firms that pursue certification, and so lower the
number of firms that pay the test fee. But there is every reason to believe
that the reduction in demand for the label will come disproportionately
from brown firms, thereby rendering a clearer signal to consumers, and so
making the label worth more to firms. As such, the certifying organization
can charge a larger fee for the service, i.e., it can increase the test cost. With
a smaller number of firms paying a larger fee, the impact on revenues is
ambiguous, but it is easy to imagine scenarios where tighter standards raise
revenues. The impact on costs is also ambiguous. While a tighter standard
will lower the number of firms that seek the eco-label, which will lower the
certifying organizations’ costs, the fact that the standard is more stringent
The Economics of Eco-Labeling: Theory and Empirical Implications 25
suggests a need for greater effort, e.g., a more detailed evaluation, which
will raise costs. Ultimately, whether an increase in the stringency of the
certification test raises revenues more than it raises costs, or alternatively,
lowers costs by more than it lowers revenues, is an empirical question. But
it does seem that there can be regimes in which it pays the certifying agency
to tighten its standards; such a regime seeks likely to lower the fraction of
output associated with brown sellers, and so may render an environmental
improvement. Finally, there is the matter of pricing the certification services.
The certifying firm might set a fixed price for its services, as in the model
I described above, it might charge a per-unit price based on the number of
units tested, or it might employ a two-part tariff, including a fixed charge
together with a per-unit licensing fee. Related to this question is the question
of the certifying firm’s costs: are these mainly made up of fixed costs, or
does the certifier firm bear costs per firm tested?22
7 Empirical Implications
The fundamental appeal of eco-labeling as a policy instrument is its poten-
tial to mitigate environmental damages, by shifting production away from
brown firms and towards green firms. For this mechanism to work, it must
be the case that some consumers are willing to pay more for green prod-
ucts, and it must be the case that information can be credibly conveyed
to consumers. As I noted in the introduction, there is evidence that some
consumers would be willing to pay a price premium for green products,
though this evidence arises mainly from surveys. In this section I discuss
the potential for empirical measurement of the green premium, as well as
the ultimate impact on measures of environmental impacts such as pollution
For expositional concreteness, I couch this discussion in terms of determin-
istic certification as that highlights the issues at play. Imagine the analyst
has data on prices paid for a product before and after the introduction of
an eco-label, and suppose the features corresponding to Figure 1 apply. As
green products sell at a higher price, it would be tempting to regard the
Bas the price premium. But the implicit premium corre-
sponds to the added value environmentally conscious (or green) consumers
22 For a detailed discussion of these issues, see Crespi and Marrett (2007).
place on green products, that is, the difference between the levels such a
consumer would pay for green and brown products. In Section 2, I argued
that green consumers strictly preferred to purchase labeled products in each
of the possible equilibria.23 Accordingly, one cannot infer the magnitude of
the premium green consumers would be willing to pay.
To be able to identify the green premium, the analyst would need to be
able to identify the green consumers’ demand curves for both green and
brown products; equivalently, information before and after the introduction
of the label would be needed. Data on prices after the introduction of the
eco-label will not suffice in that regard.
Suppose data were available on individual consumer demands before and
after the introduction of the eco-label. It would be straightforward to identify
green consumers (they will be the ones paying the higher price to obtain
the labeled product), so comparing their demands before and after might
be a way forward. Prior to the imposition of the label, the initial price can
be thought of as a weighted average of the values placed on the two types
of product; if the analyst had information regarding the volumes of the two
types of products prior to the introduction of the eco-label, it would be
possible to back out an implied value for green products. If we were willing to
assume the green consumers’ price premium was a constant dollar amount,
as suggested in Figure 1, then it would be possible to estimate the price
premium. But this empirical strategy will not work if the premium green
consumers are willing to pay is a percentage of the brown price, as Sedjo
and Swallow (2002) suggest.24 Figure 5 illustrates this case.
Of course, if we knew that the price premium were a percentage, as
opposed to an absolute amount, then observing one value for the price wedge
(as described above) would suffice to infer the percentage premium. The
problem is that there is no reason ex ante to think one variant has greater
empirical validity; in point of fact, there is no reason the extra value green
consumers might place upon green products is not composed of a fixed part
(as in much of the discussion up to this point and a variable part as in
23 There is a slight caveat here: if the price wedge P0
Bthat would allow green sellers to
cover their certiﬁcation costs were larger than the premium green consumers were willing to
pay, the market would revert to the no-information conﬁguration. In that case, there would be
no observable impact from the introduction of the eco-label.
24 The idea is that the extra value placed upon a green unit will likely be relatively smaller for
very cheap products than for very expensive products.
The Economics of Eco-Labeling: Theory and Empirical Implications 27
Figure 5. Certification with a percentage price premium.
Sedjo and Swallow’s model). In the end, this is an identification problem,
the successful resolution of which will require careful thought.25
Evaluating the impact of an eco-label upon market prices is somewhat
more straightforward. The idea here is to see if certified prices exceed uncer-
tified prices, and tangentially, to see how the introduction of the label altered
both labeled and unlabeled prices. Examples of such a line of inquiry include
Bjorner et al. (2004), for the Nordic Swan program, and Teisl et al. (2002)
for the Dolphin-safe Tuna labeling program. In both cases there is evidence
that eco-labeled products do earn a modest price premium.
Presumably, the ultimate goal associated with the introduction of an eco-
label is a reduction in environmental damage. It is interesting in this regard
to consider programs introduced in Indonesia and India. Evaluating the
PROPER program in Indonesia, Garc´ıa et al. (2007, 2009) argue that pollu-
tion levels were lowered in industries associated with the labeling program,
and so judge the scheme to be a success. Similarly, the Green Rating pro-
gram in India appears to have successfully lowered pollution levels, at least
in the paper and pulp sector (Powers et al., 2011). As this sector is asso-
ciated with significant levels of water pollution, one imagines the benefits
25 In some ways, this identiﬁcation problem is akin to that faced by empirical analysts in the ﬁeld
of industrial organization who are interested in identifying the correct solution concept in an
oligopoly. To determine the appropriate model there, one has to be able to identify the ﬁrm’s
perceived marginal revenue curve; as discussed by Bresnahan (1982), this resolution requires
that certain delicate assumptions regarding interaction terms in the ﬁrm’s demand curve be
present. Perhaps similar stylistic assumptions might be invoked in the eco-labeling context.
associated with lower effluents to be significant. These two cases hint at the
potential for substantial benefits related to the introduction of a certification
scheme, particularly in less-developed countries. As such, they represent an
important policy avenue, as well as an intriguing area for research.26
When firms are privately informed about production and abatement costs, as
in the context of my model, environmental regulation is notoriously difficult.
Whether society opts for a command-and-control approach, using standards,
or a market-based approach, using effluent taxes or tradable permits, there
is generally a welfare loss associated with the informational asymmetries.
Appealing to outside interests, as with third party certification, to reduce
the informational asymmetries therefore provides an intriguing alternative.
Indeed, Tietenberg (1998) refers to this as the ‘‘third wave’’ of pollution con-
trol. One interpretation of the results I obtain above is that any reduction
in net surplus (related to the information effects) resulting from the intro-
duction of third party certification should be compared against the costs
attendant to other forms of environmental regulation, such as monitoring
and enforcement costs, or expected welfare losses attributable to asymmetric
information. To the extent that net surplus rises as a result of the improved
information, eco-labeling would be an attractive alternative to other forms of
regulatory control. That being said, it is conceivable that the inclusion of an
eco-labeling option with a more traditional form of environmental regulation
would yield an outcome that is socially preferable to the second-best out-
come typically found in models of environmental regulation. Identifying con-
ditions where such an improvement could be expected to occur would have
important implications for public policy toward environmental regulation.
A few possible complications related to eco-labeling bear some thought.
First, one might wonder if failed green firms would be tempted to dispute
the test result, or take some other action, so as to convey their environ-
mental friendliness to potentially interested buyers. While at first blush this
seems a natural avenue for the firm to pursue there is a subtle, but cru-
cial, difficulty: Because any firm that fails the certification test would be
26 These examples can also be viewed as components of a broader line of research, related to
information provision and corporate environmental strategies, which have particular currency
in developing and transitional economies (Earnhart et al., 2012).
The Economics of Eco-Labeling: Theory and Empirical Implications 29
interested in convincing potential buyers that the firm was in fact green,
the action in question has to have some potential for distinguishing between
green and brown firms. Seeking legal redress for a test failure can not sat-
isfy this requirement: if the perception is the mere act of seeking a legal
injunction will induce consumers to treat the firm as green, then any failed
firm will do so. Only if the legal challenge winds up in court, and only if
the legal authority is somehow able to tell failed green from failed brown
sellers, would the act of bringing suit serve to distinguish green from brown
sellers. And even if such an outcome were plausible, it only makes sense if
the reward (in terms of higher price) is sufficient to cover the cost associ-
ated with the legal challenge. On a related note, one might wonder if firms
that failed the certification test might be found out, and then suffer further
losses from associated stigma. The potential for failed firms to be identified
seems greater with publicly provided eco-labels, since a private eco-labeler
might be concerned that revealing the identity of failed sellers would deter
some potential customers of the certification service. That noted, if failed
firms were identified, the expected payoff from pursuing certification would
presumably be reduced (as the payoff associated with failing the test would
be reduced). As this bad outcome is more likely to impact brown firms than
by green firms (because brown firms are more likely to fail than are green
firms), it could increase green sellers’ expected payoffs in a partial-pooling
equilibrium. This would likely generate increase the net change in welfare
associated with eco-labeling, enhancing the case for its use as a policy tool.
A potentially important application of eco-labeling, which is starting
to gain some traction, is the certification of low-carbon technologies.27
This seems like a natural application of certification schemes: determin-
ing whether a particular firm is or is not low-carbon seems hopelessly out
of reach for a typical individual. Accordingly, carbon certification schemes
could yield important benefits. Even so, I fear a less optimistic view is
in order. Because measuring carbon emissions has proven notoriously dif-
ficult, it seems likely that identifying green firms will also prove challeng-
ing, so that carbon certification schemes are likely to be noisy. As a result,
it is not clear that the introduction of a carbon certification program will
27 See Hamilton et al. (2009) and Wallis and Chalmers (2007) for discussion. Examples of car-
bon certiﬁcation schemes include the Carbon Neutral Protocol (www.CarbonNeutral.com),
the Gold Standard program (www.cdmgoldstandard.org), and the Green-E Climate Standard
deliver meaningful reductions in carbon emissions. That point noted, future
improvements in verification methods, which can be interpreted as increases
in the accuracy of the certification test, do seem likely to generate welfare
gains. Accordingly, devoting energy and resources toward identifying such
improvements in testing technologies could be an important direction for
research and development efforts in the near future.
Akerlof, G. A. 1970. “The Market for Lemons: Qualitative Uncertainty and the Market
Mechanism.” Quarterly Journal of Economics 84: 488–500.
Amacher, G. S., E. Koskela, and M. Ollikainen. 2004. “Environmental Quality Compe-
tition and Eco-Labeling.” Journal of Environmental Economics and Management 47:
Arda, M. 1997. “Environmentally Preferable Commodities.” In Eco-Labelling and Inter-
national Trade, eds. S. Zarrilli, V. Jha, and R. Vossenaar, New York, N.Y.: Macmillan
Arora, S. and T. N. Cason. 1996. “Why Do Firms Volunteer to Exceed Environmental
Regulations? Understanding Participation in EPA’s 33/50 Program.” Land Economics
Arora, S. and T. N. Cason. 1999. “Do Community Characteristics Inﬂuence Environmental
Outcomes? Evidence from the Toxics Release Inventory.” Southern Economic Journal
Baksi, S. and P. Bose. 2006. “Credence Goods, Eﬃcient Labelling Policies, and Regulatory
Enforcement.” Environmental and Resource Economics 37: 411–430.
Banks, J. S. and J. Sobel. 1987. “Equilibrium Selection in Signaling Games.” Econometrica
Bjorner, T. B., L. G. Hansen, and C. S. Russell. 2004. “Environmental Labeling and
Consumers’ Choice — an Empirical Analysis of the Eﬀect of the Nordic Swan.” Journal
of Environmental Economics and Management 47: 411–434.
Bresnahan, T. F. 1982. “The Oligopoly Solution Concept is Identiﬁed.” Economics Letters
Cairncross, F. 1992. Costing the Earth. Cambridge, MA: Harvard University Press.
Cason, T. N. and L. Gangadharan. 2001. “Environmental Labeling and Incomplete Con-
sumer Information in Laboratory Markets.” Journal of Environmental Economics and
Management 43: 113–134.
Cho, I.-K. and D. M. Kreps. 1987. “Signalling Games and Stable Equilibria.” Quarterly
Journal of Economics 102: 179–222.
Crespi, J. M. and S. Marette. 2007. “How Should Food Safety Certiﬁcation be Financed?”
American Journal of Agricultural Economics 83: 852–861.
Darby, M. R. and E. Karni. 2001. “Free Competition and the Optimal Amount of Fraud.”
Journal of Law and Economics 16: 67–88.
Dosi, C. and M. Moretto. 2001. “Is Ecolabelling a Reliable Environmental Policy Mea-
sure?” Environmental and Resource Economics 18: 113–127.
Earnhart, D., M. Khanna, and T. P. Lyon. 2012. “Corporate Environmental Strategies in
Developing and Transition Economies.” University of Kansas Working Paper.
Feddersen, T. J. and T. W. Gilligan. 2001. “Saints and Markets: Activists and the Supply
of Credence Goods.” Journal of Economics and Management Strategy 10: 149–171.
Fischer, C. and T. P. Lyon. 2012. “Competing Environmental Labels.” Stephen M. Ross
School of Business, University of Michigan Working Paper.
The Economics of Eco-Labeling: Theory and Empirical Implications 31
ia, J. H., S. Afsah, and T. N. S. Sterner. 2009. “Which Firms are More Sensitive to
Public Disclosure Schemes for Pollution Control? Evidence from Indonesia’s PROPER
Program.” Environmental and Resource Economics 42: 151–168.
ia, J. H., T. N. S. Sterner, and S. Afsah. 2007. “Public Disclosure of Industrial Pol-
lution: The PROPER Approach for Indonesia?” Environment and Development Eco-
nomics 12: 739–756.
Giezman, I. and F. Verhees. 1997. “Eco-Labelling: Practical Use of the Cradle-to-Grave
Approach.” in Eco-Labelling and International Trade, eds. S. Zarrilli, V. Jha, and R.
Vossenaar, New York, N.Y.: Macmillan Press.
Haji-Gazali, B. and M. Simula. 1997. “Certiﬁcation Schemes for all Timber and Timber
Products.” in Eco-Labelling and International Trade, eds. S. Zarrilli, V. Jha, and R.
Vossenaar, New York, N.Y.: Macmillan Press.
Hamilton, K., M. Sjardin, A. Shapiro, and T. Marcello. 2009. “Fortifying the Foundation:
State of the Voluntary Carbon Markets 2009.” Technical Report, Ecosystem Market-
place and New Carbon Finance, London, England.
Hamilton, S. F. and D. Zilberman. 2006. “Green Markets, Eco-certiﬁcation and Equilib-
rium Fraud.” Journal of Environmental Economics and Management 52: 627–644.
Houde, S. 2012. “How Consumers Respond to Product Certiﬁcation: A Welfare Analysis
of the ENERGY STAR Program.” Stanford University Working Paper.
Ibanez, L. and G. Grolleau. 2008. “Can Ecolabeling Schemes Preserve the Environment?”
Environmental and Resource Economics 40: 233–249.
Karl, H. and C. Orwatt. 2000. “Economic Aspects of Environmental Labelling.” In Inter-
national Yearbook of Environmental and Resource Economics, ed. T. Tietenberg, Chel-
tenham, UK: Edward Elgar.
Kirchhoﬀ, S. 2000. “Green Business and Blue Angels,” Environmental and Resource Eco-
nomics 15, 403–420.
Konar, S. and M. A. Cohen. 1997. “Information As Regulation: The Eﬀect of Community
Right to Know Laws on Toxic Emissions.” Journal of Environmental Economics and
Management 32: 109–124.
Konar, S. and M. A. Cohen. 2001. “Does the Market Value Environmental Performance?”
Review of Economics and Statistics 83: 281–289.
Kwerel, E. 1997. “To Tell the Truth: Imperfect Information and Optimal Pollution Con-
trol.” Review of Economic Studies 44: 595–601.
Levin, G. 1990. “Consumers Turning Green: JWT Survey.” Advertising Age 61: 74.
Lizzeri, A. 1999. “Information Revelation and Certiﬁcation Intermediaries.” The RAND
Journal of Economics 30: 214–231.
Mason, C. F. 2006. “An Economic Model of Eco-labeling.” Environmental Modeling and
Assessment, vol. 11, no. , pp. 131–143, 2006.
Mason, C. F. 2011. “Eco-Labeling and Market Equilibria with Noisy Certiﬁcation Tests.”
Environmental and Resource Economics 48: 537–560.
Mason, C. F. and F. Sterbenz. 1994. “Imperfect Product Testing and Market Size.” Inter-
national Economic Review 35: 61–85.
Mattoo, A. and H. V. Singh. 1994. “Eco-Labelling: Policy Considerations.” Kyklos 47:
McCluskey, J. J. 2000. “A Game Theoretic Approach to Organic Foods: An Analysis of
Asymmetric Information and Policy.” Agricultural and Resource Economics Review 29:
Morris, J. 1997. Green Goods? Consumers, Product Labels and the Environment. London:
OECD. 1997. “Eco-Labelling: Actual Eﬀects of Selected Programs.” OECD Working Paper
vol. 5, no. 44, Paris.
Powers, N., A. Blackman, T. P. Lyon, and U. Narain. 2011. “Does Disclosure Reduce
Pollution? Evidence from India’s Green Rating Project.” Environmental and Resource
Economics 50: 131–155.
Roberts, M. J. and M. Spence. 1976. “Eﬄuent Charges and Licenses under Uncertainty.”
Journal of Public Economics 5: 193–208.
Robertson, P. 2003. “Eco-labels and Free Riding.” University of New South Wales Dis-
cussion Paper 2003/09.
Robertson, P. 2007. “Global Resources and Eco-labels: A Neutrality Result.” Review of
International Economics 15: 735–743.
Roe, B. and I. Sheldon. 2007. “Credence Good Labeling: The Eﬃciency and Distribu-
tional Implications of Several Policy Approaches.” American Journal of Agricultural
Economics 89: 1020–1033.
Sasser, E. N., P. Aseem, B. Cashore, and G. Auld. 2006. “Direct Targeting as an NGO
Political Strategy: Examining Private Authority Regimes in the Forestry Sector.” Busi-
ness and Politics 8: 1–32.
Sedjo, R. and S. Swallow. 2002. “Voluntary Eco-labeling and the Price Premium.” Land
Economics 78: 272–284.
Swallow, S. and R. Sedjo. 2000. “Eco-labelling Consequences in General Equilibrium: A
Graphical Assessment.” Land Economics 76: 28–36.
Teisl, M. F., B. Roe, and R. L. Hicks. 2002. “Can Eco-labels Tune a Market? Evidence
From Dolphin-safe Labeling.” Journal of Environmental Economics and Management
Tietenberg, T. 1998. “Disclosure Strategies for Pollution Control.” Environmental and
Resource Economics 11: 587–602.
Van’t Veld, K. and M. J. Kotchen. 2011. “Green Clubs.” Journal of Environmental Eco-
nomics and Management 62: 309–322.
Vossenaar, R. 1997. “Eco-labelling and International Trade: The Main Issues.” In Eco-
Labelling and International Trade, eds. S. Zarrilli, V. Jha, and R. Vossenaar, New York,
N.Y.: Macmillan Press.
Wallis, N. and J. Chalmers. 2007. Environmental Accreditation and Carbon Certiﬁcation
of Biofuels for Road Transport the UK Experience, European Council for an Energy
Eﬃcient Economy, chapter pp. 1767–1776.
Wasik, J. F. 1996. Green Marketing and Management: A Global Perspective. Cambridge,
MA: Blackwell Publishers, Inc.
Winterhalter, D. and D. C. Cassels. 1993. United States Hardwood Forests: Consumer
Perceptions and Willingness to Pay. West Lafayette, IN: Purdue University Press.
Zhang, Z. and L. Assuncao. 2004. “Domestic Climate Policies and the WTO.” The World
Economy 27: 359–386.