ANALYSIS

Environment and happiness: Valuation of air pollution using life

satisfaction data
Heinz Welsch
Department of Economics, University of Oldenburg, 26111 Oldenburg, Germany
Received 23 March 2005; received in revised form 31 August 2005; accepted 12 September 2005
Available online 13 October 2005
Abstract
This paper explores the relationship between pollution and reported subjective well-being (happiness) in ten European
countries. Using a set of panel data from happiness surveys, jointly with data on income and air pollution, it examines how self-
reported well-being varies with prosperity and environmental conditions and calculates the implied valuation of changes in air
pollution. The paper finds that air pollution plays a statistically significant role as a predictor of inter-country and inter-temporal
differences in subjective well-being. The effect of air pollution on well-being translates into a considerable monetary value of
improved air quality. The improvements achieved in Western Europe in the 1990s are valued at about $750 per capita per year
in the case of nitrogen dioxide and about $1400 per capita per year in the case of lead. Due to synergies among the pollutants,
the value of simultaneously reducing nitrogen dioxide and lead is slightly higher than the sum of these values.
D 2005 Elsevier B.V. All rights reserved.
Keywords: Air pollution; Environmental valuation; Subjective well-being; Pollution control; Costbenefit analysis
JEL classification: Q510; Q530; D610; I310
1. Introduction
A fundamental question in the study of environ-
ment and pollution is how pollution affects human
well-being. The question is important because adverse
effects of environmental degradation on well-being
provide a major rationale for pollution control. His-
torically, concern over such effects was at the origin of
environmental regulation in many countries, even
before issues of environmental integrity and sustain-
ability gained prominence.
At a more specific level, a question that arises with
respect to the effects of environmental conditions on
well-being is how individuals value these effects.
Environmental valuationin monetary termsis a
basic ingredient to the costbenefit analysis of envi-
ronmental policy.
The present paper is concerned with both, the
effect of pollution on human well-being and its mone-
tary valuation. By correlating peoples reported sub-
jective well-being with both, pollution and income, it
0921-8009/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.ecolecon.2005.09.006
E-mail address: welsch@uni-oldenburg.de.
Ecological Economics 58 (2006) 801813
www.elsevier.com/locate/ecolecon
evaluates pollution directly in terms of life satisfaction
as well as indirectly, relative to income. This consti-
tutes a novel and potentially effective approach to
environmental valuation. It will be referred to as the
life satisfaction approach.
1
To put the approach in perspective, it is useful to
refer to more familiar techniques of environmental
valuation. These are usually classified into revealed
preference and stated preference methods. The former
use complementarity or substitutability relationships
between environmental goods and market goods to
infer the value attributed to environmental conditions
from observed behavior with respect to market goods.
Examples include hedonic price analysis, the travel
cost approach, and the averting or mitigating behavior
method. In the stated preference approach, individuals
are directly asked to value the environmental good in
question. The most prominent stated preference
method is contingent valuation, which attempts to
elicit the valuation of hypothetical changes in envir-
onmental conditions.
The revealed preference and stated preference
approaches both have their weaknesses. Revealed
preference approaches are based on stringent
assumptions concerning the rationality of agents
and the functioning of markets. Contingent valuation
surveys, on the other hand, use hypothetical scenar-
ios, which may entail unreliable results and strategic
behavior. Both types of approach can capture only
those aspects and effects of environmental conditions
of which the individual is consciously aware. Effects
which the individual does not attribute to these con-
ditions, though they exist, are not included in the
valuation.
2
The life satisfaction approach avoids some of these
difficulties. This technique does not rely on asking
people how they value environmental conditions.
Instead, individuals are asked in surveys how satisfied
they are with life, and econometric analysis is used to
identify if and how their en masse answers move with
environmental conditions. Thus, the approach does
not require awareness of causeeffect relationships
on the part of the individual. It does not even require
that people know the degree of pollution they are
exposed to. For these reasons the life satisfaction
approach is cognitively less demanding than contin-
gent valuation and does not evoke answers considered
desired by the respondents. Moreover, there is no
reason to expect strategic behavior, since the survey
questions do not relate to environmental conditions in
any way. In contrast to revealed preference methods,
the approach does not presume rational agents and
perfect markets.
Data on self-reported well-being are used in a
growing literature in economics.
3
In this literature,
reported subjective well-being has been used as an
empirically adequate and valid approximation for
individual utility. It is thus an obvious strategy to
use subjective well-being for a direct evaluation of
environmental conditions (or other public goods or
bads). By measuring the marginal disutility of pol-
lution, as well as the marginal utility of income, the
trade-off between income and pollution can be
determined.
The present paper applies this approach to air
pollution in Europe. Using a set of panel data on
self-reported well-being in ten European countries
jointly with data on air pollution and income, the
paper examines how well-being varies with air quality
and prosperity, and calculates the rate at which they
trade off against each other. The paper finds that air
pollution plays a statistically significant role as a
predictor of inter-country and inter-temporal differ-
ences in subjective well-being (even accounting for
country and period dummies). The linkage of well-
being to income and air pollution implies a substantial
value placed on air quality. More specifically, the
average (across countries) valuation of the air quality
1
See Frey et al. (2004) for a general appraisal of this approach.
2
See Freeman (2003) for a detailed account of valuation
techniques.
3
An early study of the economics of happiness is Easterlin
(1974). Later contributions examine the relationship between
income distribution and self-rated happiness (Morawetz et al.,
1977) and between unemployment and happiness (Clark and
Oswald, 1994; Winkelmann and Winkelmann, 1998). Di Tella et
al. (2001) use data from happiness surveys to estimate the trade-off
between inflation and unemployment in the framework of a macro-
economic social welfare function. Frey and Stutzer (2002) discuss
the insights from happiness research which may be important for
integrating into economics. In this literature, as in the present paper,
the terms happiness, life satisfaction, and subjective well-being are
used interchangeably. To economists the terminology experienced
utility may sound more familiar, see Kahnemann et al. (1997).
H. Welsch / Ecological Economics 58 (2006) 801813 802
improvements achieved in 19901997 amounts to
about $750 per capita per year in the case of nitrogen
dioxide and about $1400 per capita per year in the
case of lead.
Though there is by now a considerable literature on
the economics of happiness (see footnote 3), applica-
tions to environmental economics are rare. Van Praag
and Baarsma (2005) have examined the effect of air-
port noise on subjective well-being, and Rehdanz and
Maddison (2005) have studied the relationship
between climate and happiness. The linkage between
air pollution and happiness has been examined by
Welsch (2002) using cross-section data for 54 coun-
tries. However, the cross-sectional approach turned
out to have limitations because it crucially relies on
finding appropriate controls to deal with unobserved
heterogeneity. The present paper pursues a panel data
approach to study the relationship between air pollu-
tion and life satisfaction, thus avoiding unobserved
heterogeneity and the danger of spurious correlation.
In addition, the present paper augments that previous
study by computing not only the marginal value of air
pollution changes (marginal rate of substitution
between income and pollution), but the value of
observed infra-marginal changes, in terms of the
equivalent surplus.
The paper is organized as follows. Section 2
describes the methodological framework. Section 3
presents the results of estimating several versions of
a happiness function with air pollution included. Sec-
tion 4 calculates the effects of air pollution on sub-
jective well-being and their monetary value. Section 5
concludes.
2. Methodological framework
2.1. The basic approach
As mentioned in the Introduction, data from hap-
piness surveys are successfully used in a growing
literature in economics. In these surveys, people are
asked to state their degree of happiness or satisfaction
with life, for instance bvery satisfiedQ, bfairly satis-
fiedQ, etc.
In order to be useful for statistical work, happi-
ness data need to satisfy certain quality requirements.
Especially, a basic condition for using happiness data
as empirical approximations to individual utility is
that they be at least ordinal in character and satisfy
conventional quality standards. Whether happiness
measures meet these conditions has been widely
assessed in decades of validation research (see,
e.g., Frey and Stutzer, 2002). In these studies mea-
sures of happiness are generally found to have a high
scientific standard in terms of internal consistency,
reliability and validity, and a high degree of stability
over time (Diener et al., 1999). Different happiness
measures correlate well with each other and, accord-
ing to factor analyses, represent a single unitary
construct. Happiness responses are correlated with
physical reactions that can be thought of as describ-
ing true, internal happiness: people reporting to be
happy tend to smile more and show lower levels of
stress responses (heart rate, blood pressure). They are
more frequently described by others as being happy
and they are less likely to commit suicide. As con-
cerns the comparability across nations, no indication
has been found that cultural or linguistic bias may
prevent a comparison of happiness across nations
(Veenhoven, 1993).
In addition to these properties, using happiness
data for welfare analysis requires to presume some
sort of cardinality. While psychologists and sociolo-
gists usually interpret happiness scores as cardinal,
economists are inclined to be skeptical about this
claim. However, this may be less problematic at a
practical level than at a theoretical level (Kahneman,
1999), and recent analysis has produced evidence that
assuming ordinality or cardinality of happiness scores
has little effect on empirical results (Ferrer-i-Carbonell
and Frijters, 2004). In what follows, we therefore
assume that cardinal numbers can be assigned to the
verbal happiness categories used in survey questions.
As a consequence, we can compute average happi-
ness, and this has methodological advantages, as will
be seen immediately.
In specifying a happiness function, there are basi-
cally two varieties: a function with individual happi-
ness (micro approach) or a function with aggregate
(average) happiness as the dependent variable (macro
approach). The micro approach does not require to
cardinalize happiness scores (since estimators for
ordered variables can be used), but it requires to
control for heterogeneity in the individuals character-
istics. As far as socio-demographic and socio-eco-
H. Welsch / Ecological Economics 58 (2006) 801813 803
nomic characteristics are concerned, these are usually
available in the survey data. In addition, however,
there is unobserved heterogeneity across individuals
(especially with respect to personality traits). This
may affect results substantially (Bertrand and Mullai-
nathan, 2002; Ravallion and Lokshin, 2001; Ferrer-i-
Carbonell and Frijters, 2004) and should be controlled
for, if possible. In this regard, the macro approachif
based on representative surveyshas the advantage
that unobserved heterogeneity at the micro level can
be expected to even out.
4
Below, we will therefore follow the macro
approach, using average life satisfaction (by year
and country) from representative surveys as the
dependent variable. Explanatory variables are per
capita income and levels of air pollution. The partial
derivatives with respect to the explanatory variables
can be interpreted as the marginal utility of income
and the marginal disutility of pollution, respectively.
Their ratio yields the marginal rate of substitution
between income and pollution, that is, the marginal
monetary value of a change in pollution levels.
5
Moreover, because the indifference curve over income
and pollution is estimated directly, the relevant wel-
fare measures (equivalent surplus and compensating
surplus) can be computed even for infra-marginal
changes in pollution levels.
2.2. Data
The data set used in this paper comprises annual
data, 19901997, for the following countries: Bel-
gium, Denmark, France, Germany, Greece, Luxem-
bourg, Netherlands, Portugal, Spain, United
Kingdom. The variables and their summary statistics
are listed in Appendix A, along with correlation
coefficients.
The definitions of variables and the sources of data
are as follows.
! NITROGEN (nitrogen dioxide concentration,
NO
2
), PARTICLES (total suspended particulate
concentration, TSP) and LEAD (lead concentra-
tion) are based on time series of ambient air pollu-
tion from OECD (1999). The data represent
dnational trendsT based on country-wide networks
of measurement stations. For details see OECD
(1999).
6
The pollutant concentrations are represen-
tative figures by country and year. They are
expressed in microgram per cubic meter (Ag/m
3
).
! INCOME (GNP per capita at purchasing power
parity) is computed using data on GNP, population,
and purchasing power parities (PPP) from OECD
(1998). The resulting data are in thousand US$ (at
PPP and 1990 prices).
! The variable LIFESAT (average self-reported well-
being) comes from the World Database of Happi-
ness (Veenhoven, 2004). The data are based on the
Eurobarometer Survey Series, a representative sur-
vey
7
in which the following type of question
is asked: bOn the whole, are you very satisfied,
fairly satisfied, not very satisfied or not at all
satisfied with the life you lead.Q The qualitative
responses are rated as follows: bvery satisfiedQ=4,
bfairly satisfiedQ=3, bnot very satisfiedQ=2, bnot at
all satisfiedQ=1.
8
LIFESAT measures the average
level of life satisfaction in a given country.
Our choice of countries and years is largely based
on data availability. In addition, the choice of pollu-
tants to be considered is guided by their likely impor-
tance for subjective well-being. Nitrogen dioxide is a
precursor of photochemical smog and tropospheric
(low-level) ozone and as such highly relevant to
4
An alternative way of dealing with unobserved micro-hetero-
geneity is based on surveys in which a constant set of individuals is
surveyed over time. This allows to use dummy variables for each
individual as controls. The problem with this approach is that a
fixed set of survey respondents will not remain representative across
time.
5
Note that the marginal rate of substitution is an ordinal concept,
i.e., it is invariant with respect to the cardinalization of utility
(happiness) as long as alternative cardinalizations are monotonically
increasing transformations of each other.
6
The data in OECD (1999) take the form of indices (1985=100).
These indices have been converted back to their natural units using
information from WEF/YCELP/CIESIN (2002).
7
The Eurobarometer public opinion surveys are conducted on
behalf of the European Commission, DG Press and Communication.
Each consists of approximately 1000 face-to-face interviews per
Member State of persons aged 15 and over.
8
Although observations are available with three or five response
categories, the four-item data from the Eurobaometer is the only
complete data set available for European countries in the 1990s.
H. Welsch / Ecological Economics 58 (2006) 801813 804
human health (Cifuentes et al., 2001). Particulates
may affect well-being via respiratory problems.
Lead is implicated for increased incidence of hyper-
tension, hart attack and general mortality. Whereas
only a small proportion of lead intake might actu-
ally be due to inhalation, a larger part may be
related to dchance ingestionT of lead-rich dust on
the hands or in food (Dubourg, 1996). Thus, the effects
of lead and particles may be complementary to some
extent.
With respect to the various pollution variables it
may be noted that they occupy quite different ranges
and display considerable difference in variability.
Whereas NITROGEN and PARTICLES have mean
values of about 45 Ag/m
3
each, the mean value for
LEAD is less than 0.2 Ag/m
3
. LEAD also differs from
the other two pollutants in that its variation is much
larger. The largest value observed for LEAD is 80
times higher than the smallest one.
2.3. Empirical strategy
The happiness equation to be estimated reads as
follows:
log LIFESAT
it
a
i
b
t
clog INCOME
it

d
1
log NITROGEN
it

d
2
log PARTICLES
it

d
3
log LEAD
it
u
it
1
where i and t refer to countries and years, respectively.
The coefficient relating to INCOME is expected to
be positive while the pollution coefficients should be
negative. The specification in logarithms accounts for
interaction among the pollution variables: since the
partial derivative of LIFESAT with respect to one
pollutant depends on the level of the other pollutants,
the various pollutants mutually enforce each other.
9
The parameters a
i
and b
t
are country and period
dummies. The country dummies are included to
control for time-invariant omitted-variable bias. Espe-
cially, the country dummies accommodate country-
specific differences in the individuals personal char-
acteristics.
10
The period dummies are included to
control for global shocks, which might affect well-
being in any period but are not otherwise captured by
the explanatory variables.
Standard methods of estimating Eq. (1) are fixed
effects or random effects. The random effect estimates
are more efficient than the fixed effect estimates, but
may be inconsistent.
11
The random effects method is
not pursued here for dtechnicalT reasons: in contrast to
the fixed effects technique, the random effects estima-
tion requires that the number of cross sections is larger
than the number of coefficients to be estimated. This
condition is not met by the present model and data due
to the presence of 8 period coefficients in addition to
the coefficients on income and the various pollutants.
Therefore, the fixed effects method is used through-
out. To account for cross-section heteroskedasticity,
generalized least squares (GLS) is used.
It may be added that issues of simultaneity are
ignored in this paper because it is unclear what kind
of variable could serve as a persuasive instrument for
environmental conditions in a happiness regression
equation.
3. Empirical results
3.1. Structure of the data
We first consider simple regressions of log(LIFE-
SAT) on the various explanatory variables. These
regressions provide some information on the structure
11
The major difference between these two techniques is the infor-
mation utilized to calculate the coefficients. The fixed effects esti-
mates are calculated from differences within each country across
time; the random effects estimates are more efficient, since they
incorporate information across individual countries as well as across
periods. The random effects method is consistent only if the coun-
try-specific effects are uncorrelated with the other explanatory
variables. A Hausman specification test can evaluate whether this
independence assumption is satisfied.
9
As an alternative to the specification in logarithms we also
examined the corresponding specification in level variables. We
found all of our qualitative results (signs and significance of coeffi-
cients) confirmed. In what follows, we confine ourselves to the
logarithmic version.
10
In individual-level happiness functions, observable personal
characteristics show up explicitly as explanatory variables, but
problems may result from the neglection of unobserved character-
istics (see Section 2.1).
H. Welsch / Ecological Economics 58 (2006) 801813 805
of the data and will be used below as a background to
our discussion of the multiple happiness regressions
we are actually interested in. As Table B1 in Appen-
dix B shows, well-being is positively and significantly
related to prosperity and negatively and significantly
related to the various pollutants.
12
In order to understand several of the results pre-
sented below it is also useful to examine how the
various air pollutants are related to the level of pros-
perity. Table B2 in Appendix B reports the results of
regressing the pollutant concentrations on per capita
income. The result is that the concentration of all three
air pollutants is negatively linked to the level of
prosperity. In other words, the countries considered
in this paper are on the downward-sloping part of a
hypothetical environmental Kuznets curve. The nega-
tive association is particularly strong and highly sig-
nificant in the case of PARTICLES: an increase of per
capita income by 1% is associated with a reduction in
urban particulate concentration by more than 2.5%. In
the case of LEAD the relationship is of a similar
strength, but insignificant. For NITROGEN, the link-
age is significant, but much weaker. It may also be
noted that the coefficient of determination is particu-
larly high in the case of PARTICLES.
3.2. The happinessprosperitypollution relationship
The results of several versions of the happiness
regression (1) are shown in Table 1. It can be seen
that the coefficient of determination (R2) is quite high.
Since a substantial portion of the R2 (over 90%) is
explained by the country and year dummies, it can be
said that these dummies act as powerful controls.
Therefore, we can be confident that significance of
the variables of interest, if obtained, indicates a sub-
stantive role of these factors and is not the result of
spurious regression.
Regressions (A)(C) combine income with each of
the pollutants separately. It can be seen that the coef-
ficient of INCOME is positive and significant in each
of these cases. The coefficients of the pollutants are
negative and significant for NITROGEN and LEAD
while being negative and insignificant in the case of
PARTICLES. The result that PARTICLES becomes
insignificant when INCOME is included while being
significant on its own (see Table B1) is obviously
related to the circumstance that the two variables
exhibit a strong and highly significant (negative)
association, as shown in Table B2.
Regression (D) includes all pollutants jointly with
INCOME. The result is that all explanatory variables
except LEAD become insignificant. The lowest t-
statistic is obtained for PARTICLES and the coeffi-
cient turns positive. Eliminating this insignificant
regressor yields regression (E). The result is that all
remaining coefficients have the expected sign and the
two pollutants are significant at confidence levels of
about 3%; INCOME is significant at a level of 8%.
The circumstance that it is difficult to identify a
linkage between PARTICLES and well-being when
INCOME is included is mainly due to the strong
negative association between PARTICLES and
INCOME. The beneficial effect of reduced levels of
the former is therefore attributed to higher levels of
the latter. With respect to the other two pollutants
there is also a negative association with INCOME,
but it is weaker or more fragile than in the case of
PARTICLES and therefore does not prevent their
effect on well-being from being identified.
Table 1
Multiple regressions
(A) (B) (C) (D) (E)
log(INCOME) 8 (3.86) 0.259 (2.09) 0.262 (2.34) 0.191 (1.15) 0.208 (1.80)
log(NITROGEN) 0.044 (4.47) 0.085 (1.61) 0.057 (2.27)
log(PARTICLES) 0.022 (1.22) 0.028 (0.71)
log(LEAD) 0.011 (2.26) 0.020 (2.52) 0.012 (2.23)
R2 0.970 0.967 0.964 0.967 0.966
Adj. R2 0.959 0.953 0.946 0.946 0.949
Dependent variable: log(LIFESAT).
Note: Regressions include country and year dummies; method: GLS; n =80. Figures in parenthesis are t-statistics.
12
The significant negative relation between life satisfaction and
PARTICLES stands in contrast to the small positive simple correla-
tion reported in Table A2.
H. Welsch / Ecological Economics 58 (2006) 801813 806
With respect to the effect of PARTICLES it should
also be noted that there is a strong positive linkage of
this pollutant to NITROGEN and LEAD: in a log-
linear regression of PARTICLES on NITROGEN and
LEAD, the coefficients are 0.733 and 0.107, respec-
tively (with associated t-statistics of 7.09 and 4.58).
Thus, a 1% change in NITROGEN goes along with a
change in PARTICLES by more than 0.7%.
13
This
(multi)collinearity adds to the difficulty of establish-
ing an explicit relationship between LIFESAT and
PARTICLES in the multivariate framework, though
it is clearly visible in the univariate setup of subsec-
tion 3.1. In addition to these statistical considerations
it should be recalled that the effects of LEAD and
PARTICLES may be complementary to some extent
(see subsection 2.2).
Overall, regression (E) can be considered a satis-
factory representation of the linkage of subjective
well-being to prosperity and air pollution. Even
though it fails to capture the role of PARTICLES
explicitly, their effect may be thought of as being
represented by the coefficients of NITROGEN and
LEAD in an implicit fashion, due to the strong asso-
ciation between PARTICLES and these other two
pollutants.
4. Valuing air pollution
4.1. Effects on well-being
Our preferred specification, regression (E) from
Table 1, can be used as the basis for computing the
marginal disutility of air pollution. The result provides
an answer to the question: bHow many happiness
categories on a scale from 1 to 4 is a represen-
tative individual shifted downwards (upwards) when
the ambient concentration of pollutant X increases
(decreases) by 1 Ag/m
3
?Q
The problem with expressing the effects on well-
being in this way is obviously that the measure-
ment units of the pollutants (Ag/m
3
) are not very vivid.
Moreover, as discussed in Section 2.2, typical con-
centrations of NITROGEN and LEAD differ by a
factor of more than 100. Hence, a comparison of the
effects of the two pollutantsif expressed in this
wayis not straightforward.
For these reasons we choose to describe the effects
of air pollution on well-being (and likewise its mone-
tary valuation) differently: we ask how well-being is
affected by the observed changes over the sample
period 19901997?
More specifically, we consider our preferred
regression, written in the following general way:
LIFESAT
it
V INCOME
it
; NITROGEN
it
; LEAD
it
: 2
We compute the following changes in LIFESAT:
DV
N
i
: V INCOME
i1
; NITROGEN
i1
; LEAD
i1

V INCOME
i1
; NITROGEN
i0
; LEAD
i1

DV
L
i
: V INCOME
i1
; NITROGEN
i1
; LEAD
i1

V INCOME
i1
; NITROGEN
i1
; LEAD
i0

DV
NL
i
: V INCOME
i1
; NITROGEN
i1
; LEAD
i1

V INCOME
i1
; NITROGEN
i0
; LEAD
i0

where the subscript 0 refers to the initial year (1990)
and the subscript 1 to the final year (1997). DV
i
N
then
tells us how subjective well-being in the final year
differs from its hypothetical level that would have
obtained, had NITROGEN remained at the initial
level.
14
Of course, DV
i
N
N(b)0 as NITROGEN
i1
b(N)
NITROGEN
i0
. Similarly, DV
i
L
gives the effect on
well-being of the change in LEAD, and DV
i
NL
gives
the combined effect of both changes.
15
Before turning to the results, some information on
the development of air pollution in the countries
considered may be helpful. The percentage changes
14
For ease of notation, the dummy variables do not show up in Eq.
(2) and in the definitions of the changes in LIFESAT. In computing
these changes the dummy variables are fixed at their final-year
level.
13
Note that NITROGEN (that is, NO
2
) and one component of
PARTICLES, namely particulate nitrate, both originate from NO
X
.
15
For infra-marginal changes, the combined effect may differ from
the sum of the partial effects because the non-linear specification of
the life satisfaction regression allows for interactions among the
pollutants. In our results presented below, the difference sometimes
does not show up due to rounding.
H. Welsch / Ecological Economics 58 (2006) 801813 807
in NITROGEN and LEAD 19901997 are shown in
the second and fourth column of Table 2. The NO
2
concentration decreased in most countries. The rates
of decrease were between 8% in the UK and 29%
in Spain, whereas Portugal experienced an increase
by 58%. Among the reasons for the variance in the
rates of change are differences in the growth of
electricity generation and of car traffic. With respect
to power generation it is also important to note that
by 1990 the countries considered (which are all
members of the European Union) had achieved
varying degrees of completing the denitrification
programs for power stations prescribed by the Eur-
opean Unions legislation. As a result, some of the
countries, especially the poorer ones, had to under-
take more abatement than others in the 1990s. As
concerns car traffic, a factor that influenced NO
2
emissions was the market penetration of catalytic
converters, which was different in the various coun-
tries by 1990 and thereafter.
The drop in LEAD is even more pronounced than
that in NITROGEN. Due to the introduction of
unleaded gas, lead concentrations dropped by up to
90% in 19901997. However, in one of the countries
considered (Luxembourg) the lead concentration in
1997 was practically the same as in 1990, probably
because the diffusion of unleaded gas had already
been largely completed by the beginning of the
1990s.
From the third column of Table 2 it can be seen
that, in the countries where NITROGEN decreased,
average well-being by 1997 was between 0.014
(UK) and 0.056 (Spain) categories higher than it
would have been without the decrease in NITRO-
GEN. For Portugal, the increase in NITROGEN
implied that well-being was 0.069 categories below
the level that would have obtained without the
increase. With respect to LEAD, the effects are
more pronounced, ranging from 0.019 (Belgium) to
0.101 (Denmark) well-being categories. The com-
bined effects are between 0.034 (Luxembourg) and
0.121 (Denmark).
In interpreting these results, it should be clear that
the figures obtained are to be seen relative to the
chosen numerical scale for life satisfaction (14).
The range actually occupied on that scale in 1997 is
2.61 (Portugal) to 3.59 (Denmark). Against this back-
ground we see that observed changes in air pollution
lead to considerable shifts in subjective well-being.
4.2. Monetary valuation of air pollution
Another way of assessing the subjective effect of
air pollution changes is in terms of the marginal rate
of substitution (MRS) between income and air pollu-
tion. The MRS of pollution for income is the amount
of income that the representative individual must be
given to compensate her for a one-unit increase in
pollution, keeping her well-being unchanged. It thus
yields the monetary value of a marginal change in
pollution.
The MRS can be computed as follows:
MRS
BLIFESAT=BPOPULATION
BLIFESAT=BINCOME
; 3
The MRS is expressed in $ per microgram per
cubic meter, a variable that is not easy to grasp. An
Table 2
Effect of air pollution changes on average life satisfaction
% Change in NITROGEN Change in LIFESAT % Change in LEAD Change in LIFESAT Total change in LIFESAT
Belgium 14.0 0.025 40.1 0.019 0.043
Denmark 10.5 0.022 90.0 0.101 0.121
France 8.5 0.015 74.8 0.047 0.061
Germany 0.0 0.000 70.2 0.040 0.040
Greece 11.6 0.019 60.4 0.030 0.049
Luxembourg 16.6 0.034 0.0 0.000 0.034
Netherlands 9.2 0.018 79.7 0.067 0.085
Portugal 58.3 0.069 n.a. n.a. n.a
Spain 29.0 0.056 n.a. n.a. n.a
UK 7.9 0.014 n.a. n.a. n.a
H. Welsch / Ecological Economics 58 (2006) 801813 808
alternative measure would be an elasticity, that is, the
utility-constant drop in income if pollution goes up by
1%. In the logarithmic specification of the life satis-
faction regression, the elasticity is simply the ratio of
the coefficients of pollution and income. With respect
to NITROGEN the elasticity is about 0.27. With
respect to LEAD it is about 0.06. In other words, a
1% increase in pollution is equivalentin happiness
termsto a decrease in income by 0.27% and 0.06%,
respectively. In comparing these two figures, how-
ever, it should be recalled that the variability of
LEAD is much larger than that of NITROGEN (see
Table A1). Therefore, a 1% change in LEAD is a
relatively trivial event in comparison with a 1%
change in NITROGEN.
To account for these differences in the actual varia-
bility of the pollutants we choose to present our
estimates of the monetary value of air pollution
changes in a different way. We again refer to the
observed changes over the sample period, as dis-
cussed in the preceding subsection. Given that we
have estimated a life satisfaction function in income
and air pollution, we are able to determine the mone-
tary value of infra-marginal air pollution changes,
using the equivalent surplus. In our context, the
equivalent surplus is the (hypothetical) income an
individual would need to attain her current level of
utility at initial pollution levels (everything else fixed
at the current level), minus the current income (see
Johansson, 1987). In other words, the equivalent sur-
plus fixes utility at its current (final) level and mea-
sures the value of an observed decrease in pollution
from the initial to the final period in terms of the
additional income required, had the decrease not
taken place.
16
Referring to the notation of Eq. (2), the equivalent
surplus of a change in NITROGEN, ES
i
N
, is defined
implicitly as follows:
V INCOME
i1
; NITROGEN
i1
; LEAD
i1

V INCOME
i1
ES
N
i
; NITROGEN
i0
; LEAD
i1

Clearly, ES
i
N
N(b)0 as NITROGEN
i1
b(N) NITRO-
GEN
i0
. Likewise, the equivalent surplus of a change
in LEAD, ES
i
L
, and the equivalent surplus of a change
in both pollutants, ES
i
NL
, are defined in the following
way:
V INCOME
i1
; NITROGEN
i1
; LEAD
i1

V INCOME
i1
ES
N
i
; NITROGEN
i1
; LEAD
i0

V INCOME
i1
; NITROGEN
i1
; LEAD
i1

V INCOME
i1
ES
N
i
; NITROGEN
i0
; LEAD
i0

The equivalent surpluses derived from our estimate
of the life satisfaction function are presented in Table
3. In the countries where NITROGEN has decreased
19901997, this decrease is valued (in 1997) at $398
Table 3
Equivalent surplus of air pollution changes
ES of change in NITROGEN ES of change in LEAD Total ES
$ % $ % $ %
Belgium 894 4.1 667 3.1 1589 7.3
Denmark 649 3.1 3113 14.7 3859 18.2
France 500 2.5 1750 8.6 2292 11.2
Germany 0 0.0 1489 7.4 1489 7.4
Greece 398 3.4 657 5.6 1076 9.2
Luxembourg 1858 5.1 0 0.0 1858 5.1
Netherlands 541 2.7 2041 10.0 2637 13.0
Portugal 1576 11.8 n.a. n.a. n.a. n.a.
Spain 1554 9.8 n.a. n.a. n.a. n.a.
UK 442 2.3 n.a. n.a. n.a. n.a.
16
Computing the equivalent surplus or similar measures on the
basis of revealed preference methods requires to make strong
assumptions concerning the incomeutility relationship and the
consistency of choice, since these are essential for the expenditure
function. The present approach makes no explicit use of the expen-
diture function but determines the equivalent surplus implicitly from
the life satisfaction function.
H. Welsch / Ecological Economics 58 (2006) 801813 809
per person per year (Greece) up to $1858 (Luxem-
bourg). The difference between these two values
reflects not only the fact that the drop in NITROGEN
has been more pronounced in Luxembourg (16.6
percent) than in Greece (11.6 percent), but especially
the circumstance that more value is placed on envir-
onmental improvement in richer than in poorer coun-
tries (viz. the per capita incomes of roughly $12000
(Greece) vs. $37000 (Luxembourg)). In relative
terms, the value of reduced NITROGEN ranges
from 2.3% of per capita income (UK) up to 9.8%
(Spain). In Portugal, the drastic increase in NITRO-
GEN is valued at $1576 or 11.8% of per capita
income.
With respect to LEAD, the valuation is even some-
what higher, ranging from $657 (Greece) up to $3113
(Denmark). The latter represents the largest percen-
tage of income (14.7%) whereas the smallest percen-
tage is 3.1% (Belgium). The high value for Denmark
reflects the fact that ambient lead concentrations have
been reduced very drastically (90%).
The total value of air quality improvement (NITRO-
GEN and LEAD) ranges from $1076 (Greece) up to
$3859 (Denmark). The total value as a percentage of
income ranges from 5.1% (Luxembourg) up to 18.2%
(Denmark).
The combined value of reduced NITROGEN and
LEAD is slightly super-additive with respect to the
two pollutants. That is, reducing the two pollutants
simultaneously is worth more than the added values of
reducing them individually. This reflects the fact that
reducing any one of these pollutants raises the mar-
ginal utility of reducing the other,
17
and this synergy
is captured by the total equivalent surplus. Super-
additivity amounts to about 2%.
4.3. Discussion
Given that we have found a considerable monetary
value of reduced air pollution, a comparison with
estimates from the literature appears warranted. It
should be noted, however, that valuation of infra-
marginal changes in air pollution is difficult to per-
form using the familiar valuation methods. These
usually yield marginal valuations, comparable to the
marginal rate of substitution that can be derived from
the present study. Below, we will therefore focus on a
comparison of marginal values.
Previous studies of the valuation of air quality have
largely been concerned with particulate matter.
18
A
reason for this may be that particulate matter affects
visibility, which is the most obvious and most easy to
comprehend aspect of air pollution. Thus, the value of
changes in particulate matter canat least partlybe
captured and expressed without any recognition of
more subtle effects of air pollution which may be a
characteristic of other air pollutants. With respect to
such effects, Cropper (2000) asserts that it is unlikely
that individuals are very well-educated about the health
risks of dirty air. She notes that it is more likely that
property values capture all of the aesthetic benefits (e.g.
visibility), but only capture a portion of the health
effects. Zabel and Kiel (2000) note that it is still unclear
how individuals process air quality information when
determining their willingness-to-pay for housing. Simi-
lar reasoning may apply to willingness-to-pay for air
quality elicited in contingent valuation studies.
The present study found a sizeable and statistically
significant relationship of subjective well-being to
ambient concentrations of nitrogen dioxide and of
lead, whereas the relationship to particulate matter
was statistically insignificant in the multiple regres-
sion (though significant in the simple regression). As
explained above, the difficulty with respect to parti-
culate matter may reflect multicollinearity (negative
correlation with per capita income, positive correla-
tion with NO
2
).
Notwithstanding its low level of significance, we
did obtain an estimate of the linkage between particu-
late matter and well-being, and it may be useful to
compare the implied valuation of TSP with that
obtained in other studies. The regression reported in
column (B) of Table 1 implies that the reduction of total
suspended particulates by 1 Ag/m
3
is worth between
$13 and $211 (in 1990 dollars), where the former figure
refers to Portugal (dpoorT) in 1990 and the latter to
Luxembourg (drichT) in 1997. The mean, median and
standard deviation are $52, $35, and $47, respectively.
This compares with the mean and median marginal
willingness to pay of $110 and $22 (in 198284 dol-
lars) reported in the meta-analysis of property value
17
The marginal (subjective) damage of one pollutant is an
increasing function of the level of the other.
18
See, e.g., the meta-analysis by Smith and Huang (1995).
H. Welsch / Ecological Economics 58 (2006) 801813 810
studies by Smith and Huang (1995).
19
Basically, our
results for Europe based on happiness regressions
thus cover the same order of magnitude as those for
the US obtained from property value models.
20
This
gives us confidence that the happiness approach to
environmental valuation is capable of producing
reasonable results.
5. Conclusions
This paper has used a set of panel data on self-
reported well-being (happiness) in 10 European coun-
tries, jointly with data on air pollution and per capita
income, to examine how subjective well-being varies
with air quality and prosperity. The estimates obtained
yield information not only on the slope of indifference
curves, i.e. the rate at which a representative citizen is
willing to trade off income against air pollution at the
margin. In addition, the estimated preference function
can be used to evaluate infra-marginal changes in
pollution, by computing the equivalent surplus or
related measures. These measures are able to capture
synergies in the change of several pollutants.
Direct estimation of the preference function over
prosperity and air pollution using life satisfaction data
provides an obvious, but up to now rarely used,
strategy to evaluate the utility consequences of pollu-
tion and their monetary value. In contrast to both
contingent valuation and the hedonic property value
method, it does not rely on individuals being con-
sciously aware of causeeffect relationships between
pollution and their well-being. In contrast to contin-
gent valuation, the approach is not liable to strategic
misrepresentation of values. Finally, in contrast to the
property value and related methods, it does not rely on
rational agents and perfect markets.
A caveat to be observed in applying the life satis-
faction approach concerns the possibility of spurious
correlations that may arise from unobserved hetero-
geneity (omitted variables). The current paper has
used a cross-national panel data framework to control
for unobserved heterogeneity across countries. Unob-
served heterogeneity across individuals is controlled
by using average subjective well-being from represen-
tative country samples.
The paper has found that air pollution plays a
statistically significant role as a predictor of inter-
country and inter-temporal differences in subjective
well-being and that the effect of air pollution on well-
being shows up as a considerable monetary valuation
of improved air quality. The average (across coun-
tries) valuation of air quality improvements achieved
in 19901997 amounts to about $760 per capita per
year in the case of nitrogen dioxide and about $1390
per capita per year in the case of lead. Differences
across countries in the valuation of observed air qual-
ity changes reflect differences not only in abatement
rates but also in per capita income, due to the lower
marginal utility of higher income.
Due to a lack of data, the present study has been
concerned with countries as the cross-sectional units.
Future research may address the question how regio-
nal or local happiness profiles are affected by the
corresponding environmental conditions. It is concei-
vable that at a more disaggregate level the linkage
between environment and happiness is even more
articulate than it is with respect to national data.
Acknowledgements
I am grateful for useful comments by Udo Ebert,
Carsten Ochsen, participants of the 2005 Annual Con-
ference of the European Association of Environmental
and Resource Economists, and two anonymous referees.
Appendix A. Summary statistics and correlations
Table A1
Summary statistics of variables used
LIFESAT INCOME NITROGEN PARTICLES LEAD
Unit 14 1000 $
(PPP)
Ag/m
3
Ag/m
3
Ag/m
3
Mean 3.10 19.243 45.078 45.342 0.178
Median 3.11 19.475 42.650 44.250 0.100
Maximum 3.64 36.629 64.700 81.400 0.800
Minimum 2.44 10.619 31.900 13.200 0.010
S.D. 0.34 5.010 8.238 19.928 0.168
20
Results for the US may not be directly comparable to the present
ones for Europe, due to differences in preferences. Nevertheless, it
would be disturbing if they differed by orders of magnitude.
19
The meta-study comprises 37 studies, including 86 estimates of
the marginal willingness to pay for reducing total suspended parti-
culates in 18 US cities in the 1960s and 1970s.
H. Welsch / Ecological Economics 58 (2006) 801813 811
Appendix B. Simple regressions
References
Bertrand, M., Mullainathan, S., 2002. Do people mean what they
say? Implications for subjective survey data. American Eco-
nomic Review 91, 6772.
Cifuentes, L., Borja-Aburto, V.H., Gouveia, N., Thurston, G.,
Davis, D.L., 2001. Hidden health benefits of greenhouse gas
mitigation. Science, 12571259 (August 17).
Clark, A.E., Oswald, A.J., 1994. Unhappiness and unemployment.
Economic Journal 104, 648659.
Cropper, M.L., 2000. Has economic research answered the needs of
environmental policy? Journal of Environmental Economics and
Management 39, 328350.
Diener, E., Suh, E.M., Lucas, R.E., Smith, H.L., 1999. Subjective
well-being: three decades of progress. Psychological Bulletin
125, 276302.
Di Tella, R., MacCulloch, R., Oswald, A., 2001. Preferences over
inflation and unemployment: evidence from surveys of happi-
ness. American Economic Review 91, 335341.
Dubourg, W.R., 1996. Estimating the mortality costs of lead emis-
sions in England and Wales. Energy Policy 24, 621625.
Easterlin, R., 1974. Does economic growth improve the human
lot? Some empirical evidence. In: David, P.A., Reder, M.W.
(Eds.), Nations and Households in Economic Growth: Essays
in Honour of Moses Abramovitz. Academic Press, New York,
pp. 89125.
Ferrer-i-Carbonell, A., Frijters, P., 2004. How important is metho-
dology for the estimates of the determinants of happiness?
Economic Journal 114, 641659. T
a
b
l
e
A
2
C
o
r
r
e
l
a
t
i
o
n
c
o
e
f
f
i
c
i
e
n
t
s
L
I
F
E
S
A
T
I
N
C
O
M
E
N
I
T
R
O
G
E
N
P
A
R
T
I
C
L
E
S
L
E
A
D
L
I
F
E
S
A
T
1
.
0
0
0
0
.
4
9
7

0
.
5
5
3
0
.
1
8
9

0
.
6
5
3
I
N
C
O
M
E
0
.
4
9
7
1
.
0
0
0

0
.
5
3
9

0
.
2
4
9

0
.
3
8
4
N
I
T
R
O
G
E
N

0
.
5
5
3

0
.
5
3
9
1
.
0
0
0

0
.
0
6
0
0
.
4
4
2
P
A
R
T
I
C
L
E
S
0
.
1
8
9

0
.
2
4
9

0
.
0
6
0
1
.
0
0
0
0
.
0
0
7
L
E
A
D

0
.
6
5
3

0
.
3
8
4
0
.
4
4
2
0
.
0
0
7
1
.
0
0
0
Table B1
Simple happiness regressions
Coefficient t-statistic Adj. R2
log(INCOME) 0.398 4.90 0.967
log(NITROGEN) 0.035 3.41 0.958
log(PARTICLES) 0.032 2.81 0.954
log(LEAD) 0.012 4.18 0.951
Dependent variable: log(LIFESAT).
Note: Regressions include country and year dummies; method:
GLS; n =80.
Table B2
Simple regression of various pollutants on log(INCOME)
Dependent variable Coefficient t-statistic Adj. R2
log(NITROGEN) 0.248 2.01 0.829
log(PARTICLES) 2.584 4.65 0.954
log(LEAD) 2.543 0.99 0.896
Note: Regressions include country and year dummies; method:
GLS; n =80.
H. Welsch / Ecological Economics 58 (2006) 801813 812
Freeman III, A.M., 2003. The Measurement of Environmental and
Resource Values, Second Edition. RFF Press, Washington, DC.
Frey, B.S., Stutzer, A., 2002. What can economists learn from
happiness research? Journal of Economic Literature, 402435.
Frey, B.S., Luechinger, S., Stutzer, A., 2004. Valuing public goods:
the life satisfaction approach. CESifo Working Paper 1158.
Johansson, P.-O., 1987. The Economic Theory and Measure-
ment of Environmental Benefits. Cambridge University Press,
Cambridge.
Kahneman, D., 1999. Objective happiness. In: Kahneman, D.,
Diener, E., Schwarz, N. (Eds.), Well Being: the Foundations
of Hedonic Psychology. Russell Sage Foundation, New York.
Kahnemann, D., Wakker, P., Sarin, R., 1997. Back to Bentham?
Explorations of experienced utility. Quarterly Journal of Eco-
nomics 112, 375406.
Morawetz, D., et al., 1977. Income distribution and self-rated happi-
ness: some empirical evidence. Economic Journal 87, 511522.
OECD, 1998. National Accounts, Volume 1: Main Aggregates,
1998 edition. Organization for Economic Cooperation and
Development, Paris.
OECD, 1999. Environmental Data Compendium 1999. Organiza-
tion for Economic Cooperation and Development, Paris.
Ravallion, M., Lokshin, M., 2001. Identifying welfare effects from
subjective questions. Economica 68, 335357.
Rehdanz, K., Maddison, D., 2005. Climate and happiness. Ecolo-
gical Economics 52, 111125.
Smith, V.K., Huang, J.-C., 1995. Can markets value air quality? A
meta-analysis of hedonic property value models. Journal of
Political Economy 103, 209227.
Van Praag, B., Baarsma, B., 2005. Using happiness surveys to value
intangibles: the case of airport noise. Economic Journal 52,
111125.
Veenhoven, R., 1993. Happiness in Nations: Subjective Apprecia-
tion of Life in 56 Nations 19461992. Erasmus University
Press, Rotterdam.
Veenhoven, R., 2004. World Database of Happiness. www.eur.nl/
fws/research/happiness, September 16, 2004.
WEF/YCELP/CIESIN, 2002. Pilot Environmental Performance
Index, World Economic Forum (WEF), Yale Center for Envir-
onmental Law and Policy (YCELP), Center for International
Earth Science Information Network (CIESIN), www.ciesin.
columbia.edu/indicators/ESI.
Welsch, H., 2002. Preferences over prosperity and pollution: envir-
onmental valuation based on happiness surveys. Kyklos 55 (4),
473494.
Winkelmann, L., Winkelmann, R., 1998. Why are the unemployed
so unhappy? Economica 65, 115.
Zabel, J.E., Kiel, K.A., 2000. Estimating the demand for air quality
in four US cities. Land Economics 76, 174194.
H. Welsch / Ecological Economics 58 (2006) 801813 813