Environment and Planning A, 1976, volume 8, pages 149-162
The measurement of mortality-air-pollution relationships
V K Smith Department of Economics, State University of New York at Binghamton, Binghamton, NY 13901, USA Received 7 April 1975, in revised form 28 August 1975
Abstract. This paper reviews past evidence on the properties of alternative specifications of the relationship between air pollution and mortality, and discusses the extent to which estimates derived from these relations are credible. Revised mortality-air-pollution equations are estimated with 1968 and 1969 data. These equations are tested for specification errors. While the results of the analysis indicate that the measured effect of air pollution on total mortality is quite stable, they do suggest that the findings of any of these studies must be viewed with caution. Air-pollution-mortality relations can only be regarded as weak approximations of a more complex underlying association. 1 Introduction Air pollutants have generally been assumed to affect people and property. As a consequence there has been and continues to be considerable public interest in the quality of our air and in public policy toward improving it. In the United States under the Clean Air Act (as amended in 1970) primary air-quality standards were established to protect public health. This legislation requires that the standards be met nationally by 1975 unless a two-year extension has been granted. Since there is increasing concern over the present and future availability of low-cost energy, it is not unreasonable to suggest that the mandated need for 'clean air' may have to be reevaluated, with due consideration being given to the costs associated with meeting the standards of the Clean Air Act. If these decisions are to be made efficiently, it is necessary to assemble information on the costs of air pollution as well as on those costs associated with higher prices and/or shortfalls in energy. For energy the problems are partially mitigated by the data available from the established markets for those natural resource commodities providing primary energy sources (see Hudson and Jorgenson, 1974, for discussion). Air pollution by contrast is more difficult to evaluate. It is an externality that brings costs to individuals (and firms) without their consent and in the absence of market-determined compensation. Accordingly the determination of the opportunity costs associated with the relaxation in the standards of the Clean Air Act is particularly difficult. The purpose of this paper is to review the results of one of the approaches frequently selected for measuring the impact of air pollution and to discuss the extent to which estimates derived from these relations are credible. Specifically the expanding literature on mortality and air pollution will be reviewed and the results of further analysis of these relationships discussed. Before proceeding, however, it should be noted that this approach (that is mortality-air-pollution relations) is not the only method which has been used in evaluating the costs of air pollution. Another method is based on a generalization of the Tiebout (1956) model and assumes that individuals reveal their dissatisfaction with air pollution by moving. Thus owner and rental housing markets ought to provide ceteris paribus an indirect means of measuring the revealed preferences of
V K Smith
150
individuals for 'clean air' (1) . Since Freeman (1974) has commented on the limitations of these models in some detail, the focus of the present paper will be on models seeking to determine the effects of pollution on health directly. Section 2 of the paper briefly reviews the trends in suspended particulate matter and sulfur dioxide over approximately the past decade. In section 3 a selective review of recent attempts to estimate mortality-air-pollution relations is presented. Section 4 discusses the findings from a reevaluation of mortality-air-pollution relationships for 1968 and 1969 and considers the problems associated with discriminating between the alternative specifications. Section 5 develops a generalized least squares estimator for one of the specifications considered 'best' in the set. The last summarizes the implications of the analysis for research and policy. 2 Trends in air quality Table 1 summarizes the reports of the urban National Air Surveillance Network (NASN) stations for 1971 relative to 1968 on the annual geometric mean and maximum daily levels of suspended particulate matter. These data indicate a relatively stable pattern of average suspended particulate matter with about the same number of stations recording increases as recording decreases. The pattern of increases and decreases seems to indicate that 'clean' air is getting more 'dirty', and the very 'dirty' air is becoming less so in relation to average levels. A comparison of the 1968 to 1971 average levels with previous periods, either 1960 to 1963 or 1964 to 1967, indicates a fairly consistent pattern of decline in the average levels of suspended particulates. This contrasts, as has been noted, with the recent data, which do not indicate any consistent trend in the average level over the 1968 to 1971 period. The 1971 maximum daily concentrations indicate a modest downward trend relative to 1968, with 98 stations recording lower daily maxima and 78 recording increased levels. It should be noted that similar comparisons of 1968 to 1971 average daily maxima with either 1960 to 1963 or 1964 to 1967 indicate (as with the average annual levels) a clear trend with a great many more stations recording decreases than increases. A review of the maximum daily concentrations for sulfur dioxide reveals a roughly comparable pattern. Comparisons of 1968 and 1971 levels indicate a less dramatic differential in the number of stations recording declines relative to those with increases than is suggested by comparisons of the full 1968 to 1971 period with Table 1. Trends in suspended particulates, 1968-1971 a. TSPb
Annual geometric mean TSP
Maximum daily levels
number of stations
number of stations
(jig m ~ 3 )
up 150-250 90-150 60-90
}-4
CO
^
o O
ON^ o r-
T f ON ^vf ^
1
O
CN,
l>
^f
^j
0)
5
' J^
CO
CO
cp i—i
o
00 ^ t > VO r f CO
/ ^ N
CO 1—1
1
CN
O
vo
^
^ S
^
^ CN
O 1
^H I
ON
1
CO O / ^ O in Cp vo
1
O
CN
o
^t-
r,1,
O VO in
o
ral logarithm,
/—\
iates in 1969; capita;
00 CO
is the ith
?L)
j
1
156
.
^j
"£
«>
co
-d o
^
< D O
JT»
CJ • -
II S~£
^
.. 3
^
2^
a>
in
O
CN
rLi
O
CO O /^^ O OO
CD t^
l>
O
m
O
CO
CO
oo
O
O
O
CO,
| |
r>
VO i—i
i—i
r-
vo
o
6
i—i
O
m
O
CO
o o
a
cd
cu
.3 ,
t^
C7>
i>
vo
\~s
CO
t >^ ^i-
CD
^H
O
C,
1
E^
3
CD
00
CD £ £
r?
CD
'r£
„
t^ XI
CO
>
^
CD Q
> vo
"to CD
CO J}
«
.
Ico .3 o
c §
^^h —^
CN
in
9
CN
O vo^
1
oo /^N VO /—N r- / - « s
CO
i
"cd
"^
CN,
vo
o
£ ^ oo co
O
VO VO
CO O /-—v O l> O -H
vo oo
0-0
£» ^
00
1
^ CN
1 J^
s—'
oo
^
V£,
00 00 Tf
0-6
1
0-0
O
VO CO
m co OO
1—(
1
1 J^
m
0-4
O
1
in
O
0-6
CN,
c^
O
0-4
-st"
VO
ON
CO
o
VO CN
5s oo
CN,
ON
i n oo
0-0
ON
o o
in
0-5
CN
^ CN
0-6
1
CN CN
q\ ^ ^?
0-0 0-7
o
o o
0-6
m
0-7
3
CO
ca
C3
O^X)
> ^ *«S r^
