-------
129
consequences are reported in the literature: fatalities, injuries,
and dollar losses. Insurance costs as hedges against dollar losses
are also reported.1
For example, the CONSAD data reports for the 21-year period
in question totals as follows:2
Number of Incidents 810
Deaths 13,782
Injuries 12,472
Reported Property Damage $2.075 billion
One is tempted to determine the reported property damage per death or
per injury. While this can be done, it must be recognized that this
is a report of the ratio of deaths to property damage, not the amount
spent to avoid a fatality, nor the indirect sosts of the fatality,
nor compensation paid.
2. Common versus Catastrophic Risks
For the purpose of investigating events in the past that can
be considered catastrophes, the definition of catastrophe used is an
event that results in one or more of the following conditions:
(1) 10 or more fatalities, (2) 30 or more injuries, and (3) three
million dollars or more in damages.
Q
Data derived from Starr-5 give individual risk rates for
deaths from worldwide natural catastrophes at a level of 9.9 x 10~"
per year, and for man-made disasters, a level of 3.2 x 10~^ per CONSAD
indicates a U.S. man-made disaster rate of 2 x 10~6 per year, nearly
seven times greater than the Starr basis. Using the Starr data for
natural disasters, the CONSAD data for man-made disasters, and Accident
Facts (1973) for common accident and disease data, the total death rate
may be approximated and is shown in Table 8-2. It can be seen from
this array that: (1) catastrophes contribute about 0.12 ± .01% to the
total death rate, (2) catastrophic accidents are about 2 ± 1% of all
accidents, and (3) death from disease is about 12 times higher than
from accidents and is the largest factor.
Insurance Facts, Insurance Information Institute, 110 William Street,
New York, New York 10038 (1973).
2CONSAD Report, op. i
accidents reported.
2CONSAD Report, op. cit., p. 130. Includes both U.S. and foreign
Q
Starr, op. cit.
-------
130
Table 8-2
DEATH RATES BY SOURCE
Class
Natural Catastrophes
Man-Made Catastrophes
Common Accidents (1972)3
Motor Accidents
Work Related
Home
Public Non-motor Vehicle
TOTAL
Disease (1969)3
Heart Disease
Cancer
Stroke
Pneumonia
Diabetes Mellitus
Arteriosclerosis
TOTAL
o
Homicide, Suicide, OtherJ
TOTAL
Death Rate/100,000/Yr
0.991
0.0321 - 0.222
27.2
27.2
6.8
13.0
11.3
56.2
367
160
103
31
19
16
696
200
954
1
Starr, op. cit.
2CONSAD, op. cit.
3Accident Facts, National Safety Council, Chicago, Illinois, 1973.
-------
131
A risk adjustment factor (R.A) for catastrophic risks for
the overall death rate is:
(R.A.) = (1.2 ± 0.1) x 10-3 (8-])
while a risk adjustment factor for catastrophic risks over all acci-
dents is:
(R.A.) = (2.0 ± 0.1) x 10-2 (8-2)
3. Military versus Societal Risk Bases
One method of comparing military and societal risk bases is
to estimate the death rates from military and commercial airline
catastrophes. In doing so, one must use the proper populations at
risk (voluntary risks only). For the military, the population is
the number in the armed forces. One could argue that military trans-
port of dependents and civilian military employees should also be
included. However, such data is difficult to determine. For commer-
cial aircraft, the total U.S. population was used, including military
personnel flying commercially.
The results are shown in Table 8-3 for three years, 1970-1972.
The ratio of military death rate to that of civilians for air travel
ranges from 36 to 53. Thus, a risk adjustment factors between mili-
tary and civilian risks for aircraft would be about:
(R.A.) = (2.35 ± 0.45) x 10~2 (8-3)
A second method is to follow the lead of Starr,! who used the
Vietnamese War as a test case. During the war's heavy years, he esti-
mated that for a U.S. military population of 500,000 about 10,000
deaths per year occurred. This is an individual death rate of 2 x 10~2
per year, or 2,000 deaths per 100,000 per year. This is about twice
the total U.S. death rate of 954 per 100,000 per year, as shown in
Table 8-2, and is about 35 times the rate for accidents. However,
when one adjusts the figure for age distribution, assuming participants
1-Chauncey Starr, "Social Benefit versus Technological Risk," Science,
Vol. 169, 19 September 1969, pp. 1232-1238.
-------
132
CO
I
OO
rO
CC
H
M
Q
u o
Pi r-^
W ON
rt M
O fw
W O
H PH
W
H C/3
W P5
Q U
O M
O
CO
r-
n
^ -K
H Ol
0 C
rH -H
0) rH
F
B -H
0 <
c_
3
rH
cd
01 3
4-1 T)
cd -H
rt >
H
45 T3
4-
C
Cd M
a
C
cn
^£2
4J
cd
0)
Q
0)
C
H
1
1-1
H
"^
rH
H
s
o
>
cc
4-1
r
t
r
^
rH
cd
4J
O
H
5-j
0)
PH
rH
CO
4-1
0
H
i-i
O
PH
c/3
*
-i
K
rH
0)
(5
(5
O
cn
rJ
0)
PH
cd
CD
*
**o **o r**-
i i i
o o o
rH rH rH
XXX
CM O ON
rH i 1 ^O
in in in
i i i
000
rH rH rH
XXX
1^^ CO **O
^£> CO ON
m rH
4-j a)
to Q
00
CM
o
1
ON
rH
O
£
cd
4J
H
r-H
H
|
cd
H
O
o>
1
o
o
*£
,
&
00
o
CO
o
o
H
PH
O
PH
i-3
cn
4-1 a)
01 45
GO cn
T3 Cd
3 rH
PQ U
O
45
CO C/3
K
*
(5
01
01
0)
!H
45
H H
K
K *
* *
* *
-------
133:
ages range from 18 to 24, the picture of war versus peace time changes.
Accident Facts-*- for 1970 data shows that the total death rate from
all causes for males in this age group is 186.5 per 100,000, and for
accidents, 109.4 per 100,000. In this case, the ratio is about 11 to 1
for war deaths to all causes of death for this male age group, and
about 18 to 1 for war to accidental death. Risk adjustment factors
for conversions from war to peace are in the order of: 0.48 for total
population; 0.09 for 18-24 population total risk; and 0.55 for 18-24
population accident risk.
C. FACTORS INVOLVING TYPES OF RISKS
1. Voluntary versus Involuntary Risks
In Chapter IV, a risk was defined to be involuntary if the
risk taker did not receive the direct benefits of the activity
causing the risk, or if the risk information was purposely withheld
from the risk taker. This definition will continue to be used here,
but in order to compare results given here with other sources with
different definitions, it is necessary to review some of Starr's^
conclusions on this subject.
Starr's definition of voluntary and involuntary risks is
somewhat different:
Societal activities fall into two general cate-
gories - those in which the individual partici-
pates on a 'voluntary1 basis and those in which
the participation is 'involuntary' imposed by
the society in which the individual lives. The
process of empirical optimization of benefits
and costs is fundamentally similar in the two
cases - namely, a reversible exploration of
available options - but the time required for
empirical adjustments (the time constants of
the system) and the criteria for optimization
are quite different in the two situations.3
The case for the difference in time constants is explored
subsequently in terms of observing the effective discount rate for
voluntary and involuntary societal risks.
-^-Accident Facts, op. cit.
^Starr, op. cit.
3Starr, op. cit., p. 1233.
-------
134
Starr's definition of voluntary and involuntary activities is
also more general;1
In the case of 'voluntary* activities, the
individual uses his own value system to
evaluate his experiences. Although his even-
tual trade off may not be consciously or
analytically determined, or based upon objec-
tive knowledge, it nevertheless is likely to
represent for that individual, a crude opti-
mization appropriate to his value system.
'Involuntary1 activities differ in that the
criteria and options are determined not by
the individuals affected but by a controlling
body. Such control may be in the hands of a
government agency, a political entity, a lead-
ership group, an assembly of authorities or
'option makers,' or a combination of such
bodies. Because of the complexity of large
societies, only the control group is likely
to be fully aware of all the criteria and
options -involved in their decision process.
With these general criteria, Starr derives a risk multiplier
of four orders of magnitude difference between voluntary and involun-
tary exposure for equivalent benefits as shown in Figure 8-1.2
However, data derived on the basis of the more restrictive
definition used here show a much smaller risk multiplier. Table 7-2
and Table 7-4 from Chapter VII provide one basis for comparison
between total risks and involuntary risks, respectively. In making
such a comparison, however, the size of the populations at risk must
be taken into account.
a. Commercial Airlines
If one assumes that all citizens have access to commercial
airlines, and that risks to people on the ground are to the same
Istarr, op. cit.
2Starr, op. cit., p. 1234.
-------
135
Figure 8-1
SUMMARY OF RISK FACTORS
10*
in iv -
O
a.
: 10-6
o
(E
g 10-7
I
1 10-8
oc
UJ
0.
S 10 -9
lO-io
10-"
HUNTING SKIING
SMOKING vV-S-'-v
RAIL-
ROADS
COMMERCIAL
' 'IATION
AVERAGE Pf DUE TO
DISEASE FOR ENTIRE
US-
. MILITARY AGE GROUP
~' INVOLUNTARY
UOTOR
VEHICLES
0 40O 800 1200 1600 2000 2400
AVERAGE ANNUAL BENEFIT/PERSON INVOLVED (DOLLARS)
Risk (/?) plotted relative to benefit (/i) for various kinds of voluntary and
involuntary exposure.
From Starr, Social Benefit versus Technological Risk.
-------
136
population,1 then for an average value of 206.2 voluntary fatalities
per year (total risk minus the involuntary risk) and an average value
of 2.1 involuntary fatalities per year, a factor of nearly 100 results.
Since a smaller higher risk population around airports was
not used, one would expect this factor to be on the high side. How-
ever, Rasmussen, et al^ show a factor of between 30 and 70 for the
probability of total air crashes with fatalities between 1 and 100
over events with size consequences for persons on the ground. The
probable value of the factor is most likely between 10 and 100.
b. Military Air Crashes
For a military population of three million for voluntary
risks, a total of 88.6 fatalities per year on the average and a total
U.S. population subject to the mean involuntary risk of 5.0 fatalities
per year, the risk factor is
83.6 fat/yr 5.0 fat/yr
= 1,115 (8-4)
3 x 10& people 2 x 10B people
This is a value which is perhaps 20 times or more greater than for
commercial airlines.
c. Railroads
Assuming like populations for voluntary and involuntary
risks on railroads since all have access to trains and rights of way,
a factor of greater than 20 results, based upon averages per year.
However, since only two voluntary events are listed, the validity of
the result is questionable.
d. Marine and Mines
The data show no involuntary fatalities for catastrophic
marine and mining accidents. The number of involuntary deaths for
marine accidents involves a very small population who live near or
on water. No conclusions can be drawn. Although the 1972 Statistical
-'-There is little question that the risk near commercial airports is
higher than for the general population, but this population size was
not obtainable.
^Norman Rasmussen, et al, "An Assessment of Accident Risks in U.S.
Commercial Nuclear Power Plants." WASH-1400, U.S. Atomic Energy
Commission, August 1974, p. 189.
-------
137
Abstracts of the United States/provide data on the industrial fatalities
of all industries, including nrLneral recovery, the record of involuntary
deaths from such events as subsidence are not as easily available.
The difficulties in obtaining data on involuntary risks from mines is
discussed in Appendix B, since this area provides a good example of
data deficiencies.
e. Bus, Auto, Truck
Pedestrian and bystander fatalities for catastrophic events
are relatively high, with 2.2 involuntary fatalities per year for a
total of 16.4 voluntary fatalities per year on the average; a ratio
of about 7.5 to 1.
f. Total
Aggregating data in Tables 7-2 and 7-4 for total fatalities
over the 21-year period result in a 21 to 1 ratio. The average
fatalities per year result in the same ratio. The validity of such
aggregation is questionable.
g. Conclusion
The risk multiplier for voluntary versus involuntary
risks, at least for catastrophic events, ranges from about 10 to
1,000 for different accident sources, with the larger value repre-
senting military operations. Airline data provide the largest data
base, and result in a multiplier of about 100. This value is two
orders of magnitude lower than shown in Starr's data.
2. Avoidability of Risks and Alternatives to Risks
When it is possible to avoid risks by simply choosing not to
accept the risks, the existence of reasonable alternatives makes such
choices more attractive. Therefore, the avoidability of risk must
always be considered in conjunction with available alternatives. When
the only alternative is to avoid the risk, the threshold of action is
a function of the perceived degree of satisfaction by the risk taker
of his status quo. When the threshold is exceeded, the risk taker
will flee or avoid taking the risk. In the previous chapter, it was
hypothesized that the threshold of avoiding the risk was a function
of the degree of perceived satisfaction with the status quo, the type
of risk, and the magnitude of the risk. At this time, the hypothesis
-'-Starr, op. cit.
-------
138
remains untested, since there is sparse data available where people
have chosen to avoid risk by fleeing on a permanent basis. There is
even less data on the perceived degree of satisfaction with the status
quo. Unfortunately, the design of experiments to gather such data is
difficult because of the subjective nature of the concept. As a
result, this concept remains a speculative untested hypothesis.
The investigation of the effect of different alternatives
is another proposition. Hedging and insurance are means used to
alter the risk consequences to the risk taker or spread the risks
among larger numbers of risk takers respectively. When the risks
are convertible to financial terms, the following section uses a set
of non-financial data looking at the alternative risks of suscepti-
bility to contagious diseases and the risks of vaccination.
a. A Regulatory Decision ^_ Involuntary Risk
The highly contagious disease, smallpox (cariola), is no
longer endemic in the United States and no challenges to the immunity
of the U.S. population have occurred since 1949.1 Until 1971, approxi-
mately 15 million smallpox vaccinations were performed in the United
States each year. Of these, six million were primary vaccinations.2
The United States defense against smallpox until 1972 was based upon
four principles:-^ (1) routine vaccination of the population, (2) vac-
cination of travelers, (3) inspection of vaccination certificates of
travelers returning or entering the U.S., and (4) investigation of
suspect smallpox cases with rapid isolation and control of smallpox
importations.
The risk of infection and spread of smallpox in the U.S.
is such that one can expect one importation of smallpox every 12
years.^ Two of every three importations will cause spread, for which
an average of 23 cases will occur in each outbreak, and a death to
lj. Michael Lane, J. Donald Miller, and John M. Neff, "Smallpox and
Smallpox Vaccination Policy," Annual Review of Medicine, Vol. 22,
1971, pp. 251-272.
2Ibid.
3Ibid.
^"Vaccination Against Smallpox in the United States, A Reevaluation of
the Risks and Benefits," U.S. Department of Health, Education, and
Welfare, Public Health Service, Center for Disease Control, Atlanta,
Georgia 30333. Revised February 1972.
-------
139
case ratio of one-third will result in eight deaths per outbreak.!
Based upon a population of 2 x 10^ people, and a frequency of occur-
rence of 0.44 fatalities per year, the individual risk is 2.2 x 10~9
fat/yr/ind.
While vaccination is effective in preventing smallpox when
properly administered, there are risks associated with vaccination
(vaccinia). There are approximately seven deaths per year associated
with vaccination and in 1968 close to 500 cases involving morbidity.2
For adult primary vaccinations, this results in a death rate of three
per million, i.e., 3 x 10~6, with a significantly higher rate for
children under one-year of age.
The question addressed by the U.S. Department of Health,
Education, and Welfare, the regulating agency, consisted of two
alternatives: (1) continue the present policy of compulsory measures
as they relate to routine smallpox vaccination, and (2) immunize per-
sonnel involved in health services and all travelers only, and drop
compulsory measures. In the latter case, only about one to four pri-
mary adult vaccinations would be given each year, while six to fifteen
million, including children, would be in the first case.
If one assumes that the population has the opportunity to
travel and to be vaccinated (either as a traveler or routinely), the
individual risk rates are now lower than 9 x 10~8 fat/yr/ind for
routine vaccination and 2.1 x 10~8 for travelers only. A change in
policy from routine vaccination to vaccination of travelers and
health personnel (i.e., dropping the first of the four principles
of smallpox defense, but retaining the other three) results in a
reduction in individual risk of about 7 x 10~^ fat/yr/ind. This risk
avoidance alternative avoids present risks by over an order of magni-
tude from the risk of smallpox importation of 2.2 x 10~9 fat/yr/ind.
Thus, the selection of the second alternative, drop routine vaccination,
was made by the Public Health Service in 1972,3 a regulatory decision
involving involuntary risk to the population.
Section b, involving a medical decision on alternatives
of tetanus vaccination and antitoxin, and Section c, involving a per-
sonal decision for influenza vaccine, are not complete at this time.
^Lane, et. al, op. cit., p. 66.
^Lane, et. al, op. cit., p. 266.
-^Vaccination Against Smallpox in the U.S., op. cit., p. 17.
-------
140
Expected data have not yet been received. Later drafts will include
these sections.
In any case, risk alternatives, especially when the risks
are of the same type in each alternative, can be examined quantita-
tively for both voluntary and involuntary risks, resulting in accept-
able risk decisions.
The existence of alternatives does affect the valuation of
involuntary risks if they are not man-originated. For example, people
living in earthquake prone or tornado prone areas have the option to
move to other less exposed areas. Having such an option gives the
risk taker and risk evaluator an illusion of voluntary risk, and re-
sults in a lower negative valuation of the consequence. The conten-
tion here is that unless attractive risk alternatives are available,
a benefit-risk trade off cannot be made. Thus, the risk is still
involuntary be definition, but the valuation of the consequence is
sometimes similar to voluntary risk. For example, a fisherman who
knows no other trade could hardly move inland to avoid hurricanes,
unless there was a new attractive alternative. Knowing he could move
inland, even without an alternative, gives the risk taker some
apparent aspect of control over his destiny in the sense that he is
"voluntarily" taking the imposed risk. This apparent controllability
changes the risk evaluation, but by definition the risk is still
involuntary. It is a case of involuntary risk where the risk taker
changes his valuation of consequences through a perceived degree of
personal control.
If attractive alternatives do exist, and personal benefit-
risk trade offs can be made, then the risk is indeed voluntary.
3. Discounting in Time
There are many aspects to discounting in time and very little
data to support quantitative assessments of risk multipliers in this
area. Furthermore, there are some special problems that arise when
latent risks, risk to progeny, and irreversible world commitments are
involved. However, an attempt is made here to indicate the nature of
time discounting factors.
a. The Discount Function
The discount function is usually considered to be a
negative exponential function of the form:
e-at (8-5)
-------
141
where t is the number of time periods from the initiating event, and
a is a constant reflecting the discount rate. When one applies this
concept and attempts to identify the value of "a" for different socie-
tal discount functions, either probability of occurrence or the magni-
tude of the risk valuation can be discounted.
The underlying concept involved is based upon the implica-
tion that often a choice must be made in the present time frame among
two or more alternatives which have different effects on future risks.
The risk taker is assumed to be indifferent between two risks at the
present time if the discounted value of the future increased risk ratio
of the two risk alternatives is reduced to unity. The resultant dis-
count rate is called the effective discount rate.
From this model, one can infer minimum levels of effective
discount rate from observation of societal behavior although, gen-
erally, such experience results from ad hoc decisions as opposed to
knowledgeable, analytical choices. The future risk ratio can be based
upon the lowest future risk condition so that the risks of the proposed
alternatives are normalized. If the probability of occurrence of the
given consequence, t, periods in the future is denoted as po for the
lowest risk consequence, and pj is the probability of occurrence of
the higher risk consequence, t periods in the future, then the ratio
of pj/PO is the increased risk that must be discounted to unity in
terms of present value. On this basis, the present discounted value
of the increased risk is made equal to the present risk value of the
least risk alternative.
Pj = P0e-at = Pod + i)~t (8-6)
The right hand form of the equation is in discount rate form where:
i = ea - 1 (8-7)
a = In (1 + i) (8-8)
and i the discount rate in decimal form. The percent discount rate
is found by multiplying i by 100%. The form of some of these functions
is shown in Figure 8-2.
The original probability of occurrence, po, represents
the increased probability of occurrence of a delayed consequence re-
sulting from an event initiated at time zero (no delay). The proba-
bility of the same delayed consequence is unity. Thus, po is the
increased risk over similar risks due to the initiating event. When
the value of the discounted probability equals unity, then discounted
-------
142
oo
UJ
D;
LU
CM I
i :zr
oo :D
o
CO O
s- oo
CD
Q
OO
ZD
O
o
LL.
-------
143
risk, p0, is equivalent to risks should the event not be initiated.
Then, assuming that p-; equals unity, the ratio p^/po indicates how
many time periods must elapse at a given discount rate for the
increased risk to be discounted totally. This method derives an
"effective discount rate" for balancing a future risk. The risk taker
does not usually consider real discount in his deliberations on taking
risks.
This form is reasonable if the exact delay time of the
consequence is known. Most often the delay factor is itself a sta-
tistical function. For example, in the exposure of specific popula-
tions to increased levels of radiation, the delay of onset distribu-
tion tends to be Sigmoidl or approximately Gaussian. For example,
the Woodward and Fondiller study^ on lung cancers in uranium miners
predicts the onset of cancers with a mode of 12 to 14 years for a
Signioidal distribution. In another study-^ on the latency period for
185 cases of thyroid carcinoma from children with early childhood
exposure to x-ray, the Sigmoidal distribution shows a modal latency
period of nine years with a mean about 12 years. There are other
similar studies available.
The proper discounting function would have to be integrated
over the latency period distribution. However, a central value esti-
mate of the distribution, such as the mean or the mode, can be used
to provide a first order estimate. Since only gross conclusions will
be drawn here, such use of central values should suffice.
b. Voluntary Risk Effective Discount Rates
One particular voluntary risk with delayed consequences
is cigarette smoking. Here, for the individual benefit of smoking,
The Sigmoidal curve is skewed more toward higher time values from the
mode than a Gaussian curve.
Probable Numbers and Cost through 1985 of Lung Cancer Cases,"
Woodward and Fondiller, Inc., 1967, Appendix 7, Hearing before the
Subcommittee on Research, Development, and Radiation of the Joint
Committee on Atomic Energy, Ninetieth Congress, First Session, Part
2, pp. 975 and 1007.
3G. W. Dolphin and S. A. Beach, "The Relationship Between Dose
Delivered to the Thyroids of Children and the Subsequent Development
of Malignant Tumors," Health Physics, Pergamon Press, Vol. 9, No. 12,
December 1963, pp. 1385-1390.
-------
144
the individual increases his probability of certain diseases after
smoking for a period of time in years. The 1974 report on "The
Health Consequences of Smoking"! provides the following ranges of
increased risk of disease to smokers:
Range of Increased Risk of _ .
Disease Smokers over Non-Smokers ^° ^j
Lung Cancer^
All Smokers 7.61 to 14.20
Heavy Smokers 4.9 to 23.9
Chronic Bronchitis3 3.6 to 21.2
Emphysema 6.9 to 25.3
Coronary Heart Disease
(Male Smokers)4 2
The time of onset of the disease ranges from five years
upward to 20 years or more. The exact time distribution of disease
onset and the distribution of increased risk are not well established.
However, one can speculate that true distribution lies within certain
ranges for their values. With this in mind, a table of discount rates
may be constructed as shown in Table 8-4 for mean values of the ratio
of increased risk and the mean value of the delay of onset distribution.
For example, if the increased risk to smokers has a mean value of 10
and a mean value of onset of 20 years, then the smoker who accepts
this increased risk is using a 12.2% effective discount rate.
While one may question the validity of this example, it
does indicate that the discount rate for this voluntary risk most
Likely exceeds 6%.
c. Involuntary Discount Rates
One example of involuntary risk is that of individuals
exposed to radiation from nuclear industry facilities. These people
who have potential exposure are outside the fence of plant and receive
no direct benefit from it. If one assumes a linear dose effect
lMThe Health Consequences of Smoking," U.S. Department of Health,
Education, and Welfare, Public Health Service, January 1974.
2Ibid., p. 53.
3Ibid., p. 101.
4Ibid. , p. 15.
-------
145
CO
O)
-O
to
CO
cu
03
C£ CO
4-> CD
c c:
o cc:
o
co cu
O J3
CD CO
C CO
i- O
+-> Q-
o
CU CD
M- C
M- -r-
UJ 2
o
s- ^:
O CO
0) S-
cu
to o
I E
OO
c s-
3 O
O <4-
o
co
un
CM
OO
o
LU
o
LU
ex
LU
o
o
h-
00
o
(NJ
o
o;
LU
CO
co
oo
CO
LT)
CT)
CM
CM
O
LO
CM
LO
OO
CM
OO
CD
CM
-------
146
relationship as recommended by the BEIR Report,1 an estimated background
dose rate of 150 millirems per year and an exposure of 25 millirems per
year as recommended by the EPA, the following discount rate calculation
can be made:
P0 175 mrems
= _ _ = (1 + i)20 (8-9)
Pj 150 mrems
Where 20 years is an estimate of the latency period from the BEIR
Report, the effective discount rate is:
i = 0.77% (8-10)
This represents an estimate of the effective discount
rate specified by the EPA2 as an acceptable delayed involuntary,
statistical risk. It is a discount rate of a factor of one to two
orders of magnitude less than that for voluntary risks.
As a check on this, the present standard of five rems per
year for radiation workers'* who receive direct benefits for accepting
risk up to these levels involves a discount rate of 22% or comparing
this value with the 0.77% above shows a ratio of 28 to 1 between volun-
tary and involuntary exposure discount rates established by regulatory
agencies of the Government, Note that the purpose is to estimate
societal discount rates set by society or regulating bodies operating
for society, not to determine acceptability here.
Another regulatory example of prevention of exposure of
involuntary population to the delayed consequence of increases in
Effects on Populations of Exposure to Low Levels of Ionizing Radia-
tion, Advisory Committee on the Biological Effects of Ionizing Radia-
tion, Division of Medical Sciences, National Academy of Sciences,
National Research Council, 1972.
r\
EPA, in its proposed uranium fuel cycle standard, has proposed a level
of 25 millirems per year for exposure to individuals whereas in pro-
posed Appendix I to AEC Regulation 10CFR50, AEC recommends a design
level of 5 millirems via air pathways and 5 millirems via water path-
ways to a maximum exposed individual. The AEC level represents an
effective discount rate of 0.32%
3Federal Radiation Guide.
-------
147
cancer is the Delaney Clause of the Federal Food, Drug, and Cosmetic
Actl which states for food additives:
That no additive shall be deemed safe if it
is found to induce cancer when ingested by
man or animal, or if it is found, after
tests which are appropriate for the evalua-
tion of the subtlety of food additives, to
induce cancer in man or animal, ....
In this case, no discount rate is acceptable. Zero exposure (within
an ability to define zero in terms of measurement) is the only
recourse allowed. There is considerable controversy on this matter,
but even if relaxation should occur, this author expects that prudent
public health actions by regulatory agencies would keep the effective
discount rate very low.
d. Progeny
The exposure to risk by an individual can result in the
consequence occurring to his progeny as opposed to himself. Any risk
that involves mutagenetic consequences is a direct example where
future generations are affected. The impact on the raising of
children in a household without parents as a result of accident
involving both parents causes parents to have a different valuation
of consequences than they might otherwise consider.
Basically, although the consequence is delayed, the problem
is also a problem of spatial distribution. Identification of progenic
consequence recipients is different from identification of the
individual exposed to risk event. Furthermore, the effects may be
to the specific offspring of the exposed individuals or it may be a
statistical increase of effects to future generations, such as
changes in the genetic pool.
An example of increased risk to specific progeny is the
exposure of young people, and most likely women are more susceptible
than men, to radiation exposure from occupational, medical, and acci-
dental related activities. The first two activities bring specific
benefits to the individuals at risk and the latter may not. However,
there is little evidence available to draw conclusions. As larger
populations are exposed, increase in genetic pool changes becomes more
likely.
^Federal Food, Drug, and Cosmetic Act, Section 409c(3)A.
-------
148
e. Irreversible Commitments
When events taking place in the present affect risk
consequences in the future and cannot be altered once committed, an
irreversible condition occurs. Contamination of the world environ-
ment from long-lived radionuclides (such as plutonium from nuclear
weapons and nuclear power sources) and freon contamination of the
upper atmosphere ozone layer are two examples.
The problem is one of trading short-term benefits for
an identified group against long-term involuntary risks for a large
statistical population, in some cases yet unborn. Can such long-
term involuntary risks be discounted? Although subject to change as
society becomes more knowledgeable about such commitments, the first
conclusion would have to be answered negatively based upon the
following reasoning: (1) voluntary risks to individuals may be dis-
counted by some risk takers for discount rates that range from 6% to
50% per year; (2) involuntary risks to individuals are lower than
these levels; (3) involuntary risks to populations are regulated by
Government in the U.S. in the fractional percentage discount range,
e.g., 0.5%; (4) irreversible risks would be expected to be discounted
at some factor below involuntary risks to the population; the latter
are already at fractional levels; and lower levels as required for
irreversible commitments would tend to be meaningless in terms of
discount; (5) risks that are committed in the future with essentially
an infinite fine space will occur (theoretically) since the "laws of
proability" are perfect in eternity.
4. Spatial Distribution of Risks
There seems to be a great deal of qualitative information on
spatial distribution of risks, especially identifiable versus statis-
tical risk differences. For example, society will spend millions of
dollars to extricate children who have fallen into abandoned wells
while little or no money is spent in eliminating the hazards. An
ounce of prevention may be worth a pound of cure, but the investment
in prevention is not often made for low probability events. It is
not until the cure to an identifiable individual or group is considered
that sizeable investments are made.
Another axiom is that all medical practitioners should have
considerable clinical experience, lest the medical researcher treat
patients as statistics while the clinical practitioner treats them
as individuals (at least attempts to).
Qualitative data for voluntary risks are available, but this
is a measure of the propensity for risk taking, not for consideration
-------
149
of spatial distribution of risks. Unfortunately, in the several
cases involving involuntary risk and regulatory decisions that were
investigated, the basis for decisions did not involve consideration
of risk to risk recipients and no consideration of statistical dis-
tribution of risk. The risks examined in these cases were legal risks
in the sense of what was the chance of being brought to court in a
class action suit if no action was taken. Further investigation is
in progress, but no quantitative data can be reported at this time.
5. Controllability of Risks
a. Natural versus Man-Originated Risks
It is important to distinguish that there are avoidable
and unavoidable catastrophes that society faces. Those which are
avoidable generally stem from two conditions: (1) where society is
able to do something to prevent the catastrophe; or (2) where indi-
viduals, by changing their exposure to potential risks, can reduce
the chances of being involved.
In order to differentiate between both types of avoidable
catastrophes and unavoidable catastrophes, two classifications of
catastrophes can be considered: (1) natural disasters, and (2) man-
made disasters. In the former case, natural disasters are avoidable
only by choosing a place to live which has a lower probability of
such disasters, while man-made disasters are avoidable both directly
and by choice of potential exposure.
Two sources of data provide some estimate of the difference
in risk between natural and man-originated events. The first in the
Starr and CONSAD data shown in Tables 8-1 through 8-4. For catastro-
phic events, the Starr data result in average risk rates for natural
catastrophes for the world population in the order of 0.99 fat/yr/105,
and 0.032 fat/yr/105 for man-originated catastrophes. The multiplier
here is on the order of 30. However, if the CONSAD data for man-
originated risks of 0.22 fat/yr/105 are used, the multiplier is
closer to 5.
The second source of information is derived from two
separate presentations in the Rasmussen Study,! which are combined
in Figure 8-3 to show the frequency of events versus the number of
fatalities for both natural and man-caused events. From this data
are seen that man-caused events have risk rates for events that exceed
natural events where the events have smaller numbers of fatalities per
Rasmussen, op. cit. , pp. 227-228.
-------
150
Al
C£
=c.
LU
>-
c/o
o-
UJ
10
1 --
10'4-f
10
-7
I TOTA
MAN CAUSED
EVENTS
TOTAL \V.
NATURAL*,
EVENTS
T
10
100 1000 10,000 100,000 1,000,000
N (FATALITIES)
Figure 8-3
FREQUENCY OF TOTAL NATURAL AND MANMADE EVENTS
WITH FATALITIES GREATER THAN N
Data from: Reactor Safety Study (Draft) WASH-1400
U.S. Atomic Energy Commission, August 1974
pp. 227-228 combined
-------
151
event, but this condition is reversed at higher numbers of fatalities
per event. A high dependence on the size of the consequence is
indicated.
The two sets of data are not directly comparable since
the Rasmussen datal does take into account the populations at risk,
only the event frequency for varying consequence magnitudes. In any
case, a risk multiplier of between 5 and 30 would seem reasonable and
one order of magnitude difference, a reasonable approximation.
b. Protected and Exposed Populations
The extent to which protected or exposed populations are
willing to reject or accept higher levels of catastrophic risk depends
upon many factors, many subjective and situational. However, several
generalizations may be in order.
First, it would seem from the data above that society will
accept nature as an adversary at levels of risk much higher than for
man and man-made activities. Nature's catastrophes are sometimes
called "acts of God" which provide some insight to society's regard
for these events. They are beyond control by man, therefore, one
accepts them. One may argue that the populations at risk could move
to safer areas. However, one must consider the fact that all popula-
tions cannot live in the low seismic activity, low flood potential,
low weather event areas, since these are limited in number. Often,
one natural threat cancels another such that many East Coast areas of
low seismic activity are areas of high hurricane and flood potential.
More importantly, the way of life of many people has been developed
over many generations based upon the assumption of natural risk for
their livelihood. The fishermen living on the sea coast and the
operators of tourist industries are examples. They have learned to
live with these risks imposed by nature and their way of life may be
centered on these and associated risks.
Secondly, many risks are assumed voluntarily with reasonable
knowledge of the risk. Passengers on airplanes, autos, trains, and
ships are usually aware of the risks assumed on at least a qualitative
basis. However, the convenience, mobility, and timesaving benefits of
using these modes of travel offset the additional risks in the mind of
the risk agent. In these cases, that is, as a passenger, the risk
agent is getting the benefit as well as the risk. In many cases, such
as the air crash into an apartment house near an airport runway,
-------
152
the residents in the apartnent house are assuming risks, however small,
without directly receiving the benefits. Thus, these latter risks
are accepted on an involuntary basis. The conclusion that one finds
inescapable is that various groups in society will willingly and knowl-
edgeably accept relatively high risks to obtain particular benefits if
the risks are taken on a voluntary basis. Conversely, those upon whom
risks are imposed on an involuntary basis are willing to accept these
risks if they are man-made and avoidable, and the risk taker does not
directly receive the benefits.
c. Perceived Degre_e of Personal Control
When one gets behind the wheel of an automobile, a trade
off is made on a voluntary (perhaps not consciously, or the actual
decision was made when one decided to be a driver) basis where the
risk of an accident is balanced against the benefits of low cost,
rapid mobility. In 1973, there were 55,800 fatalities and 2,000,000
disabling injuries in the United States,^ a death rate of 27.9 per
100,000 population. About 12,000 of the fatalities involved pedes-
trians or pedal cycles where it may be presumed that the driver was
not personally at risk.
Does the average driver accept this risk on his own, or
are other processes involved for which the driver discounts this
level of risk? Any answer to this question must involve psychological
response of drivers; and in the absence of experiments to attempt to
identify and quantify such processes, the author can only make ob-
servations based upon his own experience. At least two processes
seem to exist.
First, spatial distribution is involved in the "it won't
happen to me" syndrome. Automobile risks are considered as statis-
tics until someone the driver knows personally is involved in an
accident or the driver witnesses a serious accident. In these cases,
the risks are brought home to the risk taker, and for some time there-
after tends to drive more cautiously than usual.
Secondly, the driver possesses certain skills and perceives
that he has some control over the risks involved through driving
skillfully and cautiously, and perceives that his reflexes will help
him avoid serious situations. The perception of such control may be
quite different from reality since, in 1973, 67.1% of all fatal
Accident Facts, op. cit.
-------
153
accidents and 79.9% of all accidents involving improper driving of
some type. Further, many readers may have experienced the situation
as a passenger with another driver where he unconsciously presses a
nonexistent brake pedal. A good driver seems to feel he is in control
of the situation.
Thus, the risk factor considered is the perceived degree
of control of the risk taker, not the actual degree of control which
may be quite different than subjective perception. For example, the
risk taker may be a poor, accident prone driver who will not admit
his shortcomings, even to himself.
Driving is but one case for which most readers have
personal experience. Motorcycling, swimming, skiing, and other sports
provide other examples. However, when traveling on a commercial air-
plane one looks to others for special expertise to minimize hazards,
namely the pilot and crew. In this case, positive systemic control
is sought, not personal control.
d. Degree of Systemic Control
^
Accident Facts provides some indication of the degree of
systemic control for accident related activities in terms of both
absolute numbers of deaths and the death rate. These are summarized
in Table 8-5 based upon death rates. The trends indicate that some
systems do show "learning curves" and are under positive systemic
control. Others that show lack of control involve personal voluntary
risks, such as ingestion of food or poisons, and are not easily sub-
ject to control without restriction of personal freedom.
Regulatory agencies, such as the Occupational Safety and
Health Administration, Food and Drug Administration, Department of
Health, Education, and Welfare, have had major programs aimed at pro-
viding systemic control in the work and home environments with obvious
success. These agencies have spent billions of dollars over the years
to achieve systemic control and, as a result, society as a whole has
accepted the idea that positive systemic control of risks for different
activities is, indeed, desirable and worth the money invested.
F. SUMMARY OF RISK FACTORS
It has not been possible to cover all risk factors in this chapter
at this time, primarily due to the paucity of available data and the
. , p. 48.
2Ibid. , pp. 10-13.
-------
154
H
O
LO
I
00
cfl
H
pq
CO
w
O
O
w
W
Q
u
O
PS
H
PH
O
W
W
w
Q
rH
O
SM
4-1
C
0
u
14-1
o
QJ
QJ
V-i
60
cu
Q
SM
O
60
CU
4J
cd
u
4-1
3
43
*1
O
^O
OX
rH
0
4-1
O
rH
OX
rH
E
O
M
UH
rH
o
!M
4-J
£2
O
o
CU
^
H
4-J
H
CO
O
PM
CO
4J
C
QJ
13
-H
U
O
QJ
rH
K^>
rH
rH
Cfl
(-1
Q)
£2
QJ
60
42
4-1
O
o
"
CO
QJ
4-1
03
3
4-1
a
3
rH
PM
QJ
rH
a
H
f"1
0)
[>
rl
O
4J
O
rH
0
rl
4-J
£2
0
o
QJ
H
4-1
CO
60
0)
C
0
C*")
OX
, |
QJ
O
£2
H
cn
rH
O
^-i
4-1
£2
O
U
MH
0
0)
0)
^_J
60
QJ
T3
CU
H
4-1
H
CO
O
PM
^
^_j
O
.
O
<}
ax
rH
Q)
O
a
H
CO
rH
O
V-l
4-J
£2
0
a
MH
o
Q)
01
JM
60
0)
*T3
cu
£>
H
4-J
H
Cfl
O
PM
0)
E
o
. I
rH rH
0 0
4J 4-J
£2 £2
O O
O 0
r i i L(_(
0 0
QJ 0)
QJ CU
!-l SM
60 60
QJ QJ
""O *"O
QJ QJ
> >
H -H
4-1 4-1
H -H
CO Cfl
0 0
PM PM
CO
cu
rH
a
rH
42
01
£>
^i
O
4 '
0
6 to
1 C2
C2 rM
O 3
& PQ
1
co cn
O rH QJ
H rH >M
rH CO -H
43 PM PM
3
PM
O
OX
i 1
QJ
a cn
C2 V-t
H Cfl
CO CU
J>^
MH
m cu
O 01
J-)
60 42
C 4-J
H
rH 4J
Q) CO
> Cfl
QJ rH
rH
rH
4-J -H
3 4-J
43 f2
3
0)
> rH
H 0
4-1 M
H 4J
CO f2
O O
ft O
>^ 0)
rH >
4-J -H
42 4-J
60 CO
H 60
rH QJ
CO ^
CO
4J
a
a)
r-)
o
"
-d
o
0
4H
0
CO f2
60 0
C2 -H
H 4-J
n cn
S CU
O 60
S-i £2
Q M
.
rH
cu
>
cu
rH
^
rH
rH
Cfl
SM
0)
C
0)
60
4-J
£3
o
X
CO
0)
4-J
rt
3
4-J
O
3
rH
PM
cn
B
M
cfl
QJ
SM
H
PM
rH
O
(M
4J
£2
O
O
QJ
>
H
4->
Cfl
60
0)
, ^
x)
H
rH
O
cn
*
TJ
*rH
3
cr
H
rH
/
cn
£2
o
cn
H
O
PM
.
rH
O
J_j
4-1
£2
O
U
rH
0)
>
CU
rJ
cn
cfl
O
K^
Jn
£2
O
CO
H
O
PM
O
p-
H
O
O
cfl
PM
4-J
S
QJ
T3
H
U
U
-------
155
preliminary nature of this study. Previously, a summary of risk
factors have been included in Table 8-1 as a means of indicating the
scope of risk factors and their complexity. The table is self-
explanatory.
1. Risk Factor Interrelationships
Up to this point, risk factors have been discussed for con-
venience, as if they were independent. However, the risk factors are
not necessarily independent of one another and possible interrelation-
ships are shown in Table 8-6. The basic separation of risk factors
is for voluntary and involuntary risks. Each remaining risk factor
is attacked differently by voluntary and involuntary risks.
The risk factors discussed can be related to the risk struc-
ture of Chapter III in the sense that risk evaluation is altered. A
summary of such interrelationships is shown in Table 8-7.
2. Numerical Summary of Societal Risk Experience
In order to provide a base line for risk comparison, a compila-
tion of societal risk experience levels for different kinds of risk is
a first step. A brief compilation for the United States follows, with
emphasis on how the information was derived for basic societal risk
experience from all causes of risk of specific type. Three types of
consequences are considered: (1) fatalities for which the data base
is most firm, (2) injuries, and (3) property damage for which the
data is less substantiated. The risk experience level is denoted as
AIJ for a risk of type i of consequence type j.
a. Man-Made Involuntary Catastrophic Risk ^
This is the base data taken from the CONSAD Report^ and
probably represents the least tolerable type of risk for which society
is concerned.
Fatalities AH = 10"? fat/yr/ind
Injuries A^2 = 5 x 10~7 inj /yr/ind
Property Damage A]^ = $.02/yr/ind
1CONSAD Report, op. cit.
-------
156
VD
00
43
CO
H
CO
Pd
O
H
%
fu
M
cn
H
Pi
O
a
O
C/3
P-i
CO
13
O
M
S
iJ
W
3
Pd
!>»
H
co
4-1
c
3
rH
O
>
C
M
>,
r-l
a
4->
c
a
rH
o
>
M
o
4J
O
CD
Pn
4»J
CO
H
Pi
4J
H
<4-l 13
0) r-l O >-. 4-1 |
e o) 3 4-1 o> o aicu
0) 43 13 O> -H 01 4-1 * rH T3
MrC'aj N 3 IH IH o o> A; a. to
u M -H tr mo o> en g g
4-1 -H 6 (1) 01 S-i -H -H 1
O & 4-i -HP! O 0> 0) 00 S-i C
01 3C'H/-N4-io)S-iO)o -co
S-i >, ' Xi -H 4-1 5-i 00133 T3 44 g
H rH
Q 0)
W
01 O
> PL.
H 5-1
a> 3
O PL,
0)
Pi w
H
4J
O 0)
S M
T3
co a)
01 rH
0 &
Q 0
C
(-1 «
O)
^ 4«i
03 co
H -rj
Pi
A!
CO M
H O
Pi
4-1
H
IH 13
0) iH
c T3
H 0)
0) iH
0 ?
Q) O
pei C
«
S-l
01 ^
^J CO
nj -H
H Crf
^ -o
CO Pi
H CO
Pi
a
o
H
4-1
H
pi
H
4-1
O)
p
g 4J4H oibo a) cr cort co
CO ^ -HOC > 0) T3C 010) -HS-i
Q) 4-1 -H -HT3 OCO >S-l HO
>O113 Q)rH 4-1 T) 3 -HP- 4-1
H T 1 O P14-I COT.1 0)134-1 4JO1 0)
4-1,060 0) CO) > 0) 03 ClJ UT-|
CO-H 42"-3 H> -HC04J Cl 3O
CCOU-I^^ 0)-H CUCfltn M 13H
WCOO ^ OCOOI-H4-1
X'HOJpiN -H'Pi O -HOX1CO
OJOJpiH -H Olpir-H t-Hctj IPCcfl^
r-l 30)g iH OOJO4-I Cfl i 1 CO
43C^-ICfl'd-H 4DrHMt3HCO Q) i-H t-l iH
CO'HCOOCtiC-1 03 03 4J CO 4J -c-l rHC03O4J^jJ
T3 MPL.S-H 13HP1ISC4-I ,C-HCOHCCO
HC3XIB -H3OIOCO C04JC4J01-H
OO4-10)C!^ O4JOCOCO TJCfl'HCBH
>-HCO nj >co nj -HCXO
«4J^;S coss ocou
C C >
C3 C3 <
CD J2 1
0) M 01 1C
ri> £1 >
CO rH r-l . CO C
0) CO Cl C S-i -H
C > 0 0 -H
O -H 0) -H CO C
H 4-1 r-l CO CO 0) CO
4-1 0) 43 -H -H T3
H C -H O CO)
C S-i CO >, Q) 3 ^
H 0) CO 4-1 13 -H
4-1 4J O -H CO rH
Q) r-l CX r-H S-I 0)
13 03 -H O > 0)
>-, 4-1 4-1 -H H
>i (-H (-1 3 4J O
rQ O 0 C Cfl B 4^
T34-lgPL,-H COOW
H CO S X W 0) 03
O -H 60 0) CJ T-I pq 4-1
> X r-l
O W , >,
4-1 4-1
H -H
rH rH
H -rl
43 ,0
03 a)
13 rH
H -H
0 CO
> >
,
0 C
4-1 0)
too
4-1 0
C S-i
3 a
o
0 0
CO 4J
H
T3 ^
CO
O -H
& H
1 !-J
O
5-1
CO CO
0) A!
>, CO
H
H H
0)
O, Q)
rH
s-? ja
rH -H
CO
C 5-1
CO 01
4-1 >
4J 0)
5-i
co 5-i
CO -H
0)
i-J
O
4-1
^sd 4-1
CO -H
H
Pi ^ W
rH 43
1 C 00
0 3
5-- 0
O) 13 CO
4= 01
60 4-1 4J
H C O
4= 3 C
O
HOC
O CO 0)
H H
5-1 13 13
CO rH
0) ^ -H
^ C 43
Q) O
5-1 60
0) O £
a s-i 01
& C
&*s
in
4-1
C
3
O
0
CO
H
P
0)
g
H
EH
-------
157
c
o
u
00
CD
tfl
H
f>^
J_l
CO
4-J
q
3
rH
O
q
M
r^
M
CO
4-1
q
3
rH
O
S-I
0
4-J
O
cd
[V ,
^
M
I-l
pi
y
> cn
Q) -r-l T3
0 SH CD
q - -H
CO O 4-1
S-l 4-J 0) -H
3 4J
cfl q -H q
q CD CD
H Pi *v *-rJ ^
CX ^i -H Cfl
Cfl CO -H
AJ 43 -H 1 l-i
Cfl S-l
H 4-1 Cfl rH
t-l O 4-J >-l Cfl
q co CD o
TS 42 -H
Cfl >, CD 4J 4J
CD co q o cn
SH ^O *H
p I>1 1 '
c/J | ^ 42 cfl
rH 4-1
1 r^ q TJ Cfl
cn o CD
A! -H in -
cn pi 4-J cd cn
H O 43 IH
pi CD z; cn CD
I 1 ^
CD 42 CO Cd
rH Cd -H 4J
42 TJ SH
Cd -H CD ^ ^
-a a o M cn cn
H 4-1 > Cd -H -H
O CD CO 4-J S-i IH
> c
, M-l
4-1 O
rl O
Cfl CD 3
q CD cr
CD S-i
ex eo co
0 CD 3
CX cfl
tj 4-1
rH CD cn
cfl >
3 -H 43
T3 CD 4-J
H CJ -H
r* SH £3
H a)
13 ex q
q o
H ^ "H
4-1
q -^ o
o cn co
H UH
cn S-i cn
TJ -H
q SH 4-1
CD O CO
PL, M-I cn
CD
P
q
rH O
cfl -H
U 4-1
H 3
4-J 42
cn -H
H S-l
4-1 4-J
CO Cfl
4-1 -H
cn Q
s s
CO
CD CD
CD a MH
rH . q -H
UH rH CD ^
O CD 3
4-1 SH SH CT 4J
CX 4-J 3 CD 43
CD q cn cn 60
a o o q -ri
x o ex o 4H
CD X CJ
rH CD <-&
Cfl cfl MH
CD q CD O rH
> o o o
H cn 3 CD S-J
4-J SH T3 T3 4J
CO 0) CD 3 C
q ex IH 4-1 o
IH CD -H CJ
CD -a MH q q
4-J cfl O cfl 60 O
rH B O Cfl -H
cfl 1 CD B B
q CD 4-J CD
O Cd IH 3 <& 4J
g; B 00 M cn
CD >*1 K^
I LM T-) | ij tn / s
H -H CD
CD tJ CD rH MH t3
rH » CD rH -H O CO
42 4J ^> 42 42 B
cfl 43 -H cfl cfl CD 1
13 60 CD T3 42 CD q
H -H cJ -H O SH CO
O MH SH O IH 60 B
> CD > CX CD
CO <-3 PH CO -~ ' Q
q q
to &
MH
O
*
CD rH
CD O
60 4J
CD q
T3 0
U
13 t~H
CD O cfl
> -H SH
H B 3
CD CD 4-i
CJ 4-J Cfl
IH cn q
CD >^
CX cfl cfl
£>
CO rH 14-|
> O O CD
IH T3
B J-> CD cfl
o q CD B
13 0 H 1
q o 60 q
CO CD cfl
p^ Q S
4J
H
rH
H
42
cfl
rH
rH
O
SH
4-J
q
o
CJ
<
rH
CO O
0) IH
> 4-1
H q
4-1 O
CO O
q
1-1 rH
CD Cfl
4-J q rH
rH O O
CO Cfl S-4
SH 4-1
LM CD q
0 CX 0
a
CD MH
U O CJ
C -H
CD CD B
4J CD CD
CO l-i 4-1
H 60 CO
X CD >,
W T3 Cfl
1 TJ MH
CD o
CD >
rH -H CD
42 CD CD
cd a JH
TJ SH 60
rl CD CD
O P-i Q
!>
<
-------
158
42
cfl
H
Pi
o
H
CJ
o
M
EH
Pi
O
W
W CO
Pi M
P Pi
H
CJ Pi
P w
Pi H
CO <
cn <;
M u
Pi
CO
JZ; pi
W O
W H
ts o
EH ^
M PH
PQ
pa Pi
CO
o o
M PC
H ^
W
Pi
0)
CJ
c
0)
3
cr
0)
cn
C
O
CJ
01
4J
m
o
0)
.H
cfl
4-J
O
£**!
4-1
H
iH
H
42
cfl
42
0
V-|
PM
M
0)
^
cfl
H
^
cn
H
Pi
0)
4C
4-1
O
4-1
4-1
0
0)
CJ
ca
J_(
{_!
3
O
CJ
O
OJ
o
ca
C 42 >^ H
C Cfl -H 4J
M 4-1 S-l -H O
>i W C 4-1 ^H 4J
1 4-1 O -H 42 [>-,
>> i-l -H O Q Cfl 4J
^ -H 4J W i-H >H
Cfl 42 cfl .H ^H i 1 CO
4-J Cfl C Q CO O C
C T3 ^ vH S-l 0)
3 -H 0) 01 4J 4J d,
iH 0 4J g cfl C 0
O > iH -H DJ O >-i
£>
-------
159
A risk adjustment factor to convert from fatalities to
injuries of five is evident from this data.
b. Man-Mad e Involuntary Ordinary Risk ^_ A2j
A risk adjustment factor for ordinary risk from catastrophic
risk factors was derived in equation 8-2. This represents a factor of
50 over the A'S.
Fatalities A21 = 5 x 10~6 fat/yr/ind
Injuries A22 = 2.5 x 10~5 inj/yr/ind
Property Damage A23 = $1.00/yr/ind
c. Natural Involuntary Catastrophic Risk j^ A3j
A risk adjustment factor of 10 is used to convert man-made
risk levels to natural risk experience levels. This represents a
central value of the range shown in Table 8-2 as a ratio of natural
to man-made catastrophes.
A31 = 10~6
A32 = 2.5 x 10-6 inj/yr/ind
A33 = $.20/yr/ind
Accident Facts-*- reports natural cataclysms for three
years (1968-1970) which resulted in an average of 239 fatalities per
year (standard deviation - 166). This represents a death rate of
1.2 x 10~6 fat/yr/ind and compares quite closely with A-31-
d. Natural Involuntary Ordinary Risk ^ A4j
A risk adjustment factor of 500 over AIJ is used. It
is made up of a factor of 10 for conversion from man-made to natural,
and a factor of 50 from catastrophic to ordinary risk.
A41 = 5 x 10-5 fat/yr/ind
A42 = 2.5 x 10-4 inj/yr/ind
A43 = $10.00/yr/ind
-^-Accident Facts, op. cit., p. 12.
-------
160
The A43 property damage number seems to be somewhat high.
Table 8-8 shows an average value of $1.02 for insured fixed property
losses. Assuming that only 70% of insured losses are fixed property,
and only one-half of all losses are insured, A43 would be closer to
$3.00 than $10.00 per year per individual. Therefore:
A43 = $3.00/yr/ind
A44 = $1.02/yr/ind
where A44 refers only to fixed property losses.
The A4i fatality number is also high. The Accident Factsl
data show an average value of 1,344 fatalities per year (standard
deviation - 40.5) for ordinary (cataclysms subtracted out) natural
fatalities. This is a rate of:
= 6.7 x 10~5 fat/yr/ind
using a fatality-injury factor of 5:
A42 = 3.6 x 10"5 inj/yr/ind
These revised numbers indicate a man-made to natural
adjustment factor of about 7 instead of 10.
e. Man-Made Voluntary Catastrophic Risk ^ Ajjj
Data here are obtained directly from the CONSAD data
shown in Chapter VII.
A51 = 2.1 x 10~6 fat/yr/ind
A52 = 2.3 x 10-6 inj/yr/ind
A53 = $.35/yr/ind
Note that the risk adjustment factor of 5 between fatali-
ties and injuries does not hold here.
f. Man-Made Voluntary Ordinary Risk j^ A5j
Overall accident death rates may be derived from Accident
Facts^ after natural causes are removed. This results in 113,713
1Ibid., p. 12.
2Ibid., p. 12.
-------
161
Table 8-8
PROPERTY DAMAGEl FROM NATURAL DISASTERS2
(From Insurance Facts)3
Total Dollars Dollars per Individual
Year (x millions) in U.S. 4
1963 $ 11 $ .06
1964 148 .74
1965 652 3.26
1966 57 .29
1967 160 .79
1968 90 .45
1969 185 .92
1970 360 1.80
1971 160 .79
1972 212 1.06
TOTAL $2,035 $10.16
MEAN 204 1.02
STANDARD DEVIATION 184 .92
For insured fixed property losses only.
r*
^Hurricanes, tornados, floods, earthquakes, windstorms, hail.
-'Insurance Facts, op. cit. , pp. 43-45.
^Based on population of 2 x
-------
162
fatalities (standard deviation - 770) per year over a three-year
period. Thus:
A61 = 5.7 x 10-4 fat/yr/ind
Accident Facts^ reports 14,028,000 bed disability injuries
in 1973, 20,703,000 injuries which were not disabling, but involved
restricted activity, and 26,189,000 injuries without restricted
activity, a total of 60,921,000 persons injured.
(bed disabling) = 7 x 10~2 inj/yr/ind
(activity restriction) = 1.0 x 10~1 inj/yr/ind
A&2 (no restriction) = 1.3 x 10~1 inj/yr/ind
A62 (total) = 3.0 x 10~1 inj/yr/ind
Insurance facts indicate that total insurance premiums
for property damage of about $38 billion were written in 1972. Pro-
perty damage losses of about $32 billion were experienced as follows:
Fire - $ 2.3 billion
Auto - 19.1 billion
Work - 10.4 billion
Other - 0.2 billion
This results in a property damage figure per individual
of about $160.00.
A£3 = S160/person/yr
These risk experience factors are summarized in Table 8-9.
It should be noted that all of the figures derived here are approxi-
mations (e.g., a low population figure for the U.S. of 2 x 10° people
has been used) and are yec to be validated. They are shown here to
provide some feeling for the different magnitudes of risk levels
experienced by society for different types of risks.
llb±d. , p. 2.
-------
163
i
00
,0
03
H
M
co
ii
Pi
CO
w
PH
>-<
H
H
W
Pi
w
Pi
O
CO
Pi
w
m
CO
I I
Pi
w
O
w
w
PH
0
O
CO
rH
>,
S-l
cd
4J
P
rH
O
>
>>
S-i
Cfl
J J
p
D
rH
O
>
rH
I
T)
P
i-t
rJ
!>
CO-
TS
P
H
^^
S-l
^^
1 )
P
H
T)
P
H
S-l
>,
4J
cd
<4-J
T3
P
H
1-1
>-,
T3
p
H
^-
1-1
>.
i ;
p
H
O
rH
X
LO
n
vO
I
o
rH
X
CO
.
CN
VO
|
O
1 1
X
rH
CM
CM
|
0
1 i
X
CM
r^
1
0
i 1
X
LO
TJ
P
H
^-
r^
>^
4-1
CO
m
r~-
1
O
i~H
X
rH
a
H
ji
a,
o
^
w
H
pi
14-1
o
-
H
4J
cd
M
4J
CO
Cfl
0
0)
TJ
cd
f
P
g
CM
O
rH
X
O
rH
rH
|
O
rH
X
o
ro
<»
1
O
r~H
X
r~-
LO
rH
LO
1
O
rH
X
LO
CM
O
1
O
1
X
LO
>.
rJ
cd
P
H
T3
!-j
O
(1)
TJ
cd
B
P
CO
g
K
rH
1
O
i~H
X
CM
^D
0
rH
X
LO
^0
1
o
rH
X
rH
O
H
J3
&
O
rJ
4-1
CO
cd
4J
cd
0
rH
CO
1-1
3
4J
CO
SZ
X
CO
LO
1
o
rH
X
^0
co
^o
1
o
rH
X
r^-
v£>
>.
S-l
CO
p
H
TJ
rJ
O
rH
CO
rJ
P
4-1
Cfl
£3
S-i
ca
w
c
3
ii
O
>
p
R
O
H
p
H
U-l
0)
o
0)
cn
o
P.
H
ra
M
4-J
cfl
-------
TRACK A & B
CHAPTER IX
DETERMINATION OF ACCEPTABLE LEVELS OF SOCIETAL RISK
A. INTRODUCTION
A major aim of investigating the nature of risk is to determine
what levels of risks of different types are acceptable to society,
and to use this information as a basis for determining the accepta-
bility of a new activity that involves new risks and benefits.
There has been considerable controversy as to whether it is possible
to rationally determine an acceptable level of risk for society,
especially in the case of large but low probability accidents. The
intent here is to demonstrate that it is possible to set acceptable
levels of risk for activities that affect society by developing a
rational, repeatable, visible methodology that accomplishes this end.
This does not imply that this methodology is the only one, or the best
one, or even a good one; but by its existence demonstrates that the
setting of acceptable risk levels is indeed possible.
B. REGULATORY IMPLICATIONS
The Governmental role in such determinations involves regulatory
activities through establishing standards, regulations, and guide-
lines to limit the magnitude and inequitable distribution of invol-
untary risks for which man has control. This regulatory function
generally only extends to limitations on involuntary risks when the
activities impact other aspects of society (as opposed to the volun-
tary risk taker) in an adverse manner. It can be extended to cover
voluntary risks in some cases. For example, the act of suicide has
a consequence not only to the individual involved but to his survivors,
his insurance company, his creditors, etc. Further, the benefit of
the act to the individual involved, if it may be thought of as a
benefit in the form of relieving oneself of problems of living, is
situational and, in some cases, irrational. Both the State and the
Church have laws, regulations, and moral codes which attempt to make
this act as unattractive as possible. Thus, Government is involved in
regulating voluntary as well as involuntary risks to some extent.
The object here is to develop a methodological approach to deter-
mination of an acceptable level of risk for new activities based upon
comparison with risk experience in society for similar types of risks.
C. METHODOLOGY FOR DETERMINING ACCEPTABLE LEVELS OF SOCIETAL RISKS
A methodology for establishing acceptable levels of societal risk
for a new activity can be stated in a general fashion along with some
164
-------
165
quantitative value judgments that are certainly arguable. The purpose
is to outline an approach to this problem, and then apply it to speci-
fic cases to see if it has validity and utility in setting such levels.
As such, the methodology involves several sequential steps.
1. Balancing Costs and Benefits
The direct and indirect societal benefits of a proposed
activity must be balanced against the total direct and indirect
societal costs of the activity. Risks are one aspect of the societal
cost. This balance is the type overall cost-benefit analysis sought
in environmental impact statements under the National Environmental
Policy Act of 1969 and a goal of technology assessment activities.
These balances must be made on at least three different levels of
impact: (1) local balance, (2) national balance, and (3) world balance,
and often result in qualitative value judgments as opposed to numerical
balances. This is primarily due to the difficulties of measuring
intangible values and of obtaining adequate data. However, the quali-
tative balancing often is precise enough to allow most neutral parties
to agree on a ranking of four different levels: (1) favorable - the
balance is overwhelming in favor of benefits over costs at the societal
level; (2) marginal - the balance is slightly positive or even in
considering benefits over costs; (3) unfavorable - costs generally
outweigh the benefits; and (4) unacceptable - costs far outweigh
societal benefits. The qualitative levels will provide a means of
determining the acceptable risk levels for a new activity in a sub-
sequent step.
2. Achieving "As Low As Practicable" Risk Levels
Once a balance is achieved, the question of further risk
reduction must be addressed in terms of its cost-effectiveness. In
other words, have the risks to achieve a given level of benefit been
made as low as possible by increasing efforts to reduce risk? Incre-
mental costs to achieve lower risk levels must be factored into the
balance of the previous step.
Although there have been many attempts to define the concept
of "as low as practicable," one definition considered here is: when
the incremental cost per risk averted is equivalent to similar costs
for similar risks in society, the system risk will be as low as prac-
ticable. This implies a relative level of "as low as practicable"
based upon societal risk as a whole. An alternative definition is:
when the incremental cost per risk averted is such that a very large
expenditure must be made for a relatively small decrease in risk as
compared to previous risk reduction steps, then the activity causing
the risk is as low as practicable. This implies a relative risk for
-------
166
the particular activity in question. Another alternative involves
defining the average practice of the best industry processes. In any
case, for whatever method selected, quantification of this level is
made in the same manner as acceptable risks in the final steps of this
methodology, but is also affected by the level of balance in the first
step. That is, when the benefits are overwhelming, one may tend to
spend more to reduce risk than otherwise. In a marginal situation,
costs to reduce risk may begin to tip the decision balance for going
ahead with the activity one way or the other, while in favorable cases
it may call for spending on risk reduction to be as safe as possible-'-
and to buy public acceptance may be warranted. Some aspects of control
of planned releases of iodine from nuclear power plants are considered
to be examples of the latter situation by some observers. The in-
creased cost of mine safety as a result of new laws has made some
mines marginal producers and is an example of the first case.
The amount society will pay to avoid a risk and the acceptable
level of societal risk are different concepts. The first is a rela-
tive concept, and is based upon what society does to reduce other
similar risks. The second concept is an absolute one and involves
direct valuation of the residual risks of an undertaking even after
as low as practicable levels have been achieved.
3. Reconciling Identified Risk Inequities
When the overall cost-benefit analysis is made and is
favorable, still various inequities may exist for specific value
groups. Those who assume the risks may not always receive the bene-
fits or the risk may not be evenly distributed among the benefit
receivers. If this condition occurs, the risk must be identified and
the nature and type of risk must be ascertained. These risks can then
be compared against the level of risk that society is experiencing for
similar types of risk.2 In the absence of actual data on similar types
of risk, the risk multipliers developed in the previous chapter may be
applied to determine the threshold of acceptability in terms of proba-
bility of occurrence for the particular type of consequence. This
level of societal risk experience for a specific type of risk will be
designated by AI, the societal risk experience factor for risk of
-'-"Safe as possible" implies implementation of a level of safety well
beyond the "as low as practicable" level.
^Note activities causing risk are not compared. The risks of acti-
vities are compared with similar risks in society independent of
source.
-------
167
type i; and can be measured either in terms of individual risk per
year,or number of risks per 100,000 people per year, or the risks to
a specific value group expressed individually or collectively.
These risk inequities are imposed since an undertaking with
some benefits to a segment of society are to be implemented. What pro-
portion of the involuntary risk inequity, as expressed by the societal
risk experience factor A-^, is acceptable in terms of increased risk
to obtain broad societal benefits? That is, society is already
experiencing a given level of involuntary risk of the type being
considered. What increase in the present risk level (as a fraction
of the existing level) will society accept to gain the broad benefits?
For example, should the introduction of a new pesticide to assist
agriculture be allowed to double the existing chance of exposing the
general population to low levels of carcinogens? This would be too
high a price to pay in the eyes of many, including the author.
The following value judgments are offered by the author to
provide a systematic approach to discounting the proportion of risk
experience for a new or existing activity. When steps one and two
are made in establishing a cost-benefit balance, a proportionality
factor, P, will be selected as follows:
Benefit-Cost Balance Range of P
Favorable 10~1 - 10-2
Marginal 10~2 - 10~3
Unfavorable 10~4 - IQ-5
Unacceptable 0
In other words, the product of A^ and P provides a ball park
estimate of that portion of acceptable risk levels for all similar
risks that a new activity would be allowed to impose on society on an
inequitable basis to obtain the overall benefit of the activity.
4. Determining Degree of Systemic Control
It is not enough to accept the level of risk that society is
experiencing at any one time as acceptable to society. Society may
be dissatisfied with the level of risks that they are experiencing,
and if so, will want to reduce the risks (never raise them unneces-
sarily) . While different expenditures can be more or less effective
for different activities in reducing risks, the concept of degree of
systemic controllability, as discussed in detail in Chapter V, must
be considered. Five levels of controllability are readily evident:
(1) demonstrated positive control - positive control exists in the
form of demonstrated learning curves based upon empirical data;
-------
168
(2) proposed positive control - positive control has been proposed and
designed, but not yet implemented or even proven feasible in an empiri-
cal manner; (3) demonstrated level control - control a steady level of
safety as demonstrated by empirical data; (4) proposed level control -
a system proposed to operate with no worse than a steady level of
control; and (5) uncontrolled - no control evident and risks may
increase with implementation and use.
The desirability of these systemic levels of controllability
is a value judgment for any given case, but it is possible to examine
some generic values for controllability. The controllability factor
is denoted by G.
Degree of Control Range of G
Demonstrated Positive Control 1.0
Proposed Positive Control 5 x ICT* - 1 x 10~1
Demonstrated Level Control 10~1 - 10~2
Proposed Level Control 10~2 - 10~3
Uncontrolled 10~4 - 10~5
5. Risk Acceptability
As a result, the acceptability is the product of three
factors, A-^, P, G, such that the acceptable level of inequitable
risk impact, R^, is:
R± = A± x P x G (9-1)
The actual level of risk by the project of type i must not exceed R^
or at least be in the same order of magnitude.
For example, a project with a favorable balance and a proposed
positive control might allow some exposure to man-made involuntary
catastrophic risks such that:
AI = 10~7 deaths/yr/ind
A2 = 5 x lO"7 injuries/yr/ind
A3 = $.02 property damage/yr/ind
P = 5 x 10~2 risk proportionality factor
G = 2 x 10"1 controllability factor
R! = 10-7 x 5 x 10-2 x 2 x 10-1 = io-9 deaths/yr/ind
R2 = 5 x 10~7 x .05 x .2 = 5 x 10~9 injuries/yr/ind
R3 = $.02 x .05 x .2 = 2 x 10~4 dollars/yr/ind
-------
169
These must then be compared to the actual risk level from the activity
to be undertaken. Thus, the actual level of risk cannot exceed 10"?
deaths/yr/ind or 5 x ICT? injuries/yr/ind or 2 x 10~4 dollars/yr/ind,
or at least be in the same order of magnitude. The latter rule allows
some of the uncertainty and subjective imprecision to be preserved in
the final judgment.
Sensitivity analysis can be used to test the value judgments
and allow examination of their criticality.
D. JUSTIFICATION OF THE VALUE JUDGMENTS
1. Risk Proportionality Factor
The risk proportionality factor is based upon the concept that
any activity undertaken by man produced some inequity in the balancing
of benefits and costs to different groups in society. For example, an
extremely beneficial program to society, such as elimination of cancer
as a cause of death, might very well decrease the life span of those
not susceptible to cancer, since the resultant lower death rate might
increase the age of the population and competition for scarce resources,
including food and other medicines. The question that must be
addressed, then, is: how much increase in risk will society accept
for a new beneficial activity?
First, the new risk must be compared to the totality of
similar risks to which the involuntary risk takers are subject, since
absolute risk by itself has little meaning until one gets down to
the threshold of absolutely uncontrollable risk, such as the proba-
bility of the sun exploding in one's lifetime or the risk of being
killed by a meteor. There seems little question that if a single
activity doubled man's total involuntary risk probability it would
most likely be unacceptable. However, strictly as a value judgment,
an extremely beneficial activity to society as a whole might be
acceptable if the increase of involuntary societal risks were less
than 10% of the total involuntary risk level. This, then, is the top
level for the risk proportionality factor. At a level of one part in
a hundred for increased risk, there would probably be little question,
and even risks with less total benefit would probably be acceptable.
However, as the benefit-cost balance becomes marginal, at
least another factor of 10 seems justified. If the activity is
unfavorable, but the benefits to some members of society still warrant
implementation, a factor of 100 below marginal would seem reasonable.
Unacceptable balances must not result in increased risk requiring a
zero multiplier.
-------
170
This reasoning is solely the value judgment of the author.
At least one can consider the values as "straw men" to be "torn down."
Further, the values expressed are quite gross and are further subject
to uncertainty in the rigor of definition of favorable, marginal, and
unfavorable balances.
2. Degree of Control Factor
Since the level of acceptance of the risk proportionality
factor is based upon a fraction of the type of risk that society is
experiencing at any given time, those systems which demonstrate de-
creasing levels of societal risk are usually desirable. This is evi-
denced by the investment in existing technological systems to con-
tinually reduce risk. In addition, if the risks in the system are
demonstratably under control, those exposed to risks are usually
fully aware of the risks and the degree of protection afforded. For
imposed involuntary risks, there is little incentive for those who
are satisfied with their "status quo" to flee although they may have
such an option. However, if such control has not been demonstrated,
but is only proposed or in the process of implementation, confidence
in such a system will be lacking. Depending upon the level of commit-
ment, a derating by a factor of two to five does not: seem unwarranted
for proposed positive control that has not yet been demonstrated.
If the rate of risk is not decreasing or increasing, one has
"level" control. This is less desirable than positive control since
it represents a different commitment by system benefactors to invol-
untary risk takers or an achieved physical limiting point. A de-
rating factor of about one to two orders of magnitude is proposed.
Further derating is based upon even less apparent commitment.
For example, although building codes attempt to minimize risks to
fire, etc., in new buildings, the variability of these codes around
the nation, and the general attitude of "let the buyer beware" indi-
cate that the degree of control in buildings and structures seems to
be relatively uncontrolled. To some extent the risks may seem to be
voluntary under a "buyer beware" but if information on risk is with-
held with the intent to hide it from the risk taker, the risks are
involuntary. Again, the concept and the value assignments may be
argued independently or jointly.
3. Societal Value Judgments
The values assigned to the risk proportionality and degree of
control factors are basically non-technical and represent the type of
value judgment that all members of society can participate in making,
individually and collectively. The judgments are stated in broad
-------
171
terms with consequences which everyone can relate to. This is a
major factor in the usefulness of such a methodology.
The societal risk experience factors are measured directly,
and the cost-benefit balances and degree of control determination
require technical analysis. However, when these are measured or
analyzed in a valid, credible manner, the basic value judgments are
entirely societal and non-technical.
E. SUMMARY OF THE METHODOLOGY
When a new activity imposing risks upon society is to be imple-
mented, the first step is to make a balance of the cost and the bene-
fits, at least on a qualitative basis. In making this examination,
one must identify all of the risks from the activity that are imposed
upon society and various groups within it to determine if the risks
identified have been reduced to as low as practicable levels. The
degree of favorability of the cost-benefit balance is used to deter-
mine what proportional increase in similar type risks, already being
experienced by society, will society accept to achieve that benefit.
The risk acceptance level that results is further altered by con-
sidering the degree of control that the new system being implemented
exhibits in attempting to reduce the risk to society over reasonable
time periods. The resultant risk acceptance level for each type risk
is then compared with the actual level of risk of the activity for
similar types of risk.
If the actual levels of risk exceed the acceptance level by more
than an order of magnitude, then the activity is deemed unacceptable.
In this case, two options remain. The first is to drop the activity
and not implement it. The second is to apply higher degrees of con-
trol on the activity to reduce the risks upon society. Of course,
in doing this, the cost-benefit balance will change and the system
may be forced to operate at levels that stretch technology to the
limits. Actions, such as demonstrating positive control as opposed
to just proposing it, can also be used to change the level of accept-
ability factor. However, a new proposed activity may be stuck with
a "proposed control." The point is that if one does not achieve a
level of acceptability, it does not necessarily mean that the activity
should be discarded forever, but that further steps must be taken to
reduce the risks or change the risk, cost-benefit balance. When this
is accomplished, the system may then achieve an acceptable level.
-------
172
F. ACCEPTABLE LEVELS OF RISK FOR NUCLEAR POWER PLANT CATASTROPHES
1. Background
The Rasmussen Report^ has provided an estimate of the invol-
untary, catastrophic risk of an accident at a nuclear power plant on
the general public. The analysis attempts to argue that the risk is
low comparable with other types of risks, but no attempt is made (and
properly so) to justify these risks as acceptable. However, this
does provide a vehicle to test the methodology.
2. Implementation
Step 1. Cost-benefit relationship. Most paper studies on
nuclear reactors have shown a favorable cost-benefit relationship;
however, actual practice leaves something wanted. Until the following
questions can be finally answered in a favorable manner, nuclear
energy has a marginal cost-benefit balance at best: (1) suitable
means for ultimate disposal of wastes; (2) adequate safeguards to
prevent diversion of nuclear materials; and (3) ability of plants to
operate at planned levels with achieved, planned efficiencies. The
first two considerations are primarily environmental and social, the
last one economic. On this basis, the risk proportionality factor,
P, as shown in Section C.3 of this chapter, ranges from 10" 2 to 10~3.
The experience factors, AI, as derived previously, are:
AI (fatalities from catastrophic incidents) = 10~7 fat/yr/ind
A2 (fatalities from ordinary incidents) = 5 x 10~6 fat/yr/ind2
A3 (injuries and chronic diseases) = 5 x 10~7 inj/yr/ind
A4 (damage to property) = $.02/yr/ind
Step 2. Achieving "as low as practicable" risk levels. Based
upon proposed 10 CFR 50 Appendix 13 and forthcoming Environmental
iRasmussen Report, op. cit.,
^Based upon a catastrophic/ordinary risk multiplier of 50.
-^Concluding Statement of the Position of the Regulatory Staff, Public
Rulemaking Hearing on Numerical Guides for Design Objectives and
Limiting Conditions for Operation to Meet the Criterion "As Low As
Practicable" for Radioactive Material in Light-Water-Cooled Nuclear
Power Reactors, Docket No. RM-50-2, U.S. Atomic Energy Commission,
February 1974.
-------
173
Protection Agency standards for the uranium fuel cycle,-'- the reactor
design technology will meet the "as low as practicable" test to the
extent that these standards and regulations are complied with.
Step 3. Degree of systemic control. There is little question
that the record of operation demonstrates level control and that posi-
tive control is proposed. Although no serious accidents have occurred
in the short period that reactors have been operating, it has not been
adequately demonstrated that learning curves for accidents of all
kinds actually exist. As a value judgment, a factor of 0.33 for G
seems a reasonable choice.
Step 4. Reconciliation of risk inequities. The inequitable
risk is an involuntary, catastrophic risk to the general population
living near reactors. Although they may receive power from these
reactors and tax relief as close neighbors, other forms of energy pro-
duction could provide these same benefits, but with differing levels
of involuntary, catastrophic risks. The acceptable risk levels are
calculated from equation 9-1 as follows, with the limits of uncertainty
shown:
1CT7 x ID"2 x .33 = 3.3 x 1Q-10
(cat fat) =
10~7 x ID"3 x .33 = 3.3 x lO"11
5 x 10-6 x 10-2 x .33 = 1.6 x 10-8
5 x 10-6 x 10-3 x .33 = 1.6 x 10~9
5 x 10-7 x io-2 x .33 = 1.6 x 10~9
5 x 10-7 x iQ-3 x .33 = 1.6 x IQ-i
$.02 x ID"2 x .33 = 6.7 x 10~5
$.02 x 10-3 x .33 = 6.7 x 10~5
3. Comparison with Calculated Risk
(ord fat) =
R3 (inj & pd)=
(prop dam)=
fat/yr/ind
fat/yr/ind
inj/yr/ind
dol/yr/ind
The Rasmussen Report2 assigns a probability to the individual
risk of an acute fatality for 100 reactors of 3 x 10"9 per individual.
U.S. Environmental Protection Agency, "Environmental Radiation Stand-
ards for Nuclear Power Operations," 40 CFR, Part 190, unpublished
draft, January 1975.
"Rasmussen Report, op. cit., p. 188.
-------
174
However, this is for all fatalities. Therefore, the probability of
fatalities from events with less than 10 fatalities must be calculated.
This may be accomplished from the probability distribution for
fatalities^ by integrating this distribution for the probability of
events with numbers of fatalities from one to nine as described in
the general case in Appendix A. This integration results in a proba-
bility of a fatality from an ordinary event of 3 x 10~4 fat/yr.
The population exposed is 15,000,0002 and the risk rate is 2 x 10"!-1
fat/yr/ind. When this is subtracted from the total fatality rate of
3 x 10~9 fat/yr/ind, the remaining catastrophic rate is within 1% of
being identical to the total. However, there is some argument that
the estimate of the Rasmussen Report for large accidents is under-
stated by as much as a factor of 10.3 This only holds for large
events and does not affect: the ordinary risk rate. Furthermore, the
initial factor of 3 x 10~9 fat/yr/ind is only for acute fatalities.
Delayed fatalities from latent cancers must also be included. These
are essentially equal in magnitude to the acute effects.^ These can
only be discounted at a rate less than 1% per year since these risks
are involuntary. (See the previous chapter on effective discount
rates.) So, for a 20-year latency period, these can only be weighed
by a factor of 0.8. Thus, the ordinary and catastrophic risk rate
for fatalities must each be doubled or multiplied at best by a
factor of 1.8.
M! (cat fat) = 5.4 - 54 x 10~9 fat/yr/ind
M2 (ord fat) = 3.6 x 10~H fat/yr/ind
M3 (inj & pd)5= 3.2 x 10~8 inj/yr/ind
$.10/yr for 15 x 106 people
MA (prop dam)6=
$8.0 x 10-3/yr for U.S. population
1Ibid., p. 153.
2Ibid., p. 156.
-^See Environmental Protection Agency comments on the Rasmussen Report.
^Rasmussen, op. cit. , p. 161.
5Ibid. , p. 174. Total of risks from acute illness, latent cancers,
thyroid injury, and genetic damage.
. , p. 174. Based upon a total of 1.6 x 10& dollars/yr.
-------
175
These risk rates are tabulated against the acceptable levels
of risk in Table 9-1. In making risk level comparisons, the lowest
end of the range of the risk rate exceeds the highest end of the
range of acceptable level of risk by over an order of magnitude for
catastrophic fatalities, injuries, and property damage. Conversely,
the fatality rate from events with less than 10 fatalities is well
below the acceptable level.
It should be recognized that this is only a demonstration of
the methodology, not a rating of nuclear power acceptability. How-
ever, assuming the results are reasonable, one can conclude that
nuclear reactor accidents contribute involuntary catastrophic risk
to society above acceptable levels. This risk contribution is from
the potential large consequence, low probability event since the
ordinary risks are acceptable. It Is this type of event which differ
entiates the nuclear reactor from other types of energy production.
The comparison over the ranges for uncertainty for the three
unfavorable balances result in risk rates exceeding the acceptable
levels by factors of:
Fatality - Catastrophic - 16 - 1640
Injury - 20 - 200
Property Damage - 1.2 x 102 - 1.6 x
The lower end of these ranges indicate that, with slight
improvement, fatalities and injuries can be brought within the order
of magnitude band of methodology imprecision. More is required for
property damage, but adequate insurance can offset this condition.
4. Sensitivity Analysis - An Optimistic Case
In order to test the assumptions made above, an optimistic
case for nuclear power may be postualted. Assume that all outstanding
problems on waste disposal, etc., have been solved and that positive
systemic control has been demonstrated. For this case, the risk pro-
portionality factor, P, and the degree of control factor are optimum.
P = ID"1
G = 1
On this basis, the risk acceptance levels become:
R! (cat fat) = 10~8 fat/yr/ind
R2 (ord fat) = 5 x 10~7 fat/yr/ind
R3 (injury) = 5 x 10~8 inj/yr/ind
R4 (prop dam) = $.002 yr/ind
-------
176
rH
1
a\
W
i_3
9
H
0
^
CO
M
P^
pL4
O
CO
rJ
W
W H
t 3 r^j
^
W J
rJ PM
-^
H
*^
CO
H
fvj
X3
C
H
M
^*>
4-1
td
<4H
rH
I 1
1
O
rH
*
CO
CO
1
CO
CO
XI
"^
^
^»
^^,
4-1
td
MH
CT*
1
O
M
-a-
to
I
-*
m
CO
C
01
w
4-1
cd
0
1
CO
OJ
H
4J
iH
rH
td
4-1
td
p*4
XI
a
^
S-l
\
4J
«2
cr.
I
o
rH
*
VQ
rH
1
\O
,_5
*O
c
H
i-t
^
~^^
4-J
n)
<4-J
i-H
rH
1
O
H
^
vO
CO
CO
JJ
C
0)
W
xl
^4
o
1
CO
O)
r)
4-1
H
i-H
td
4-1
td
f=M
X)
C
H
~Vj"
^
4-1
cd
14-1
o
rH
I
O
X
vO
rH
X
vo
i-H
XI
a
H
^.4
^>
**^s.
p- j
c
H
00
1
0
rH
X
CM
CO
CO
01
H
^4
3
p j
C
M
g
^.^
.^
(X,
CO
td
r-\
1-4
O
vO
0
rH
X
-^^
VO
Vf
1
r-.
"
>*j
. ,
n3 T3
a c
rH -H
M J-f
^*> P^
^^^ **^
CO CO
cd td
rH rH
rH rH
O O
X) XI
rH CO
| |
0 O
rH rH
X X
rH O
rH 00
*
(DO O
60 PL, fX<
rt
pi vO CO
td o o
Q rH rH
^s X X
ij
V4 IA CM
0) rH
o.
o
eu
-------
177
However, if the benefit-cost balance is as favorable as postulated,
one would expect more nearly 1,000 reactors to be built than the 100
used in the Rasmussen Report. As a result, the measured risk rates
might have to be increased by a factor of 10, but requires further
study. Table 9-2 summarizes these. In this case, the lower end of
the range for catastrophic fatalities and the measured risk rate for
injuries are within the one order of magnitude range of methodology
imprecision. Only property damage exceeds the acceptable level of
risk, but as indicated before, adequate insurance can cover this risk.
The cost of the insurance does affect the benefit-cost ratio.
In other words, assuming that the lower level of catastrophic
fatalities can be met, and that adequate insurance for property
damage is bought, the postulated no-problem nuclear industry would
be acceptable.
G. ACCEPTABLE LEVELS OF RISK FOR LIQUIFIED NATURAL GAS (LNG) AND
LIQUID PROPANE GAS TRANSPORT (LPG)
As a further demonstration of the utility of the methodology, the
acceptable level of risk for transportation of liquified natural gas
by special tanker and for transportation of liquid propane gas by
truck and truck-pipeline combinations is examined here. The basic
data on risk of fatalities are obtained from a study made by John A.
Simmons for the Environmental Protection Agency.1 This study provides
an event-tree analysis of the risks involved in these transportation
modes, and makes estimates of the fatalities that might be expected
in the industry at the present time, and provides some perspective on
future risks. In using this data, no attempt is made to ascertain
the validity of the Simmons estimates, since they will be reviewed
and argued on their own merits separately in a different forum. The
results are used here primarily as an exercise to demonstrate the
utility of the methodology for determining risk acceptance, not for
a final determination as to the acceptability of these technologies.
The uncertainties in the estimates, which range from 10 to 1/10,
preclude this at this time.
1. Background Information on LNG and LPG Transportation Hazards^
The reason for considering both LNG and LPG in this study is
the similarity of their hazards and the large volume transported (or
Ijohn A. Simmons, Risk Assessment and Transport of LNG and LPG. Draft
version - final report for contract 68.01.2695, Environmental Protec-
tion Agency, Washington, D.C. November 25, 1974.
^Excerpted from Simmons, op. cit.
-------
178
CN
1
ON
w
! 1
PQ
E 1
0
O
0
^
Pi
0
t^H
C/D
M pi
Pi M
fe O
O P*
to pi
H-3 *£
W U
hJ ! 1
£5
w
. i ~
pq S
Pi .
4-1
CO
MH
OO
1
0
rH
13
C
H
^V^
M
^*
--.^
4-1
CO
<4H
00
1
O
rH
X
^J-
m
i
*
m
CO
4J
a
^
w
J
CO
c_?
1
CO
0)
H
4-1
H
rH
CO
4_1
cO
fe
. .
rO "T3
pj C
H -H
>> ^
4J -r-)
CO C
MH -H
t^- oo
1 1
0 O
rH rH
tw* Kjt
rN HS
m in
, ^
*^ *T3
c c
H -H
*^^^ *NV.
M J-i
^*> ^>
"*^. ""^
4J -r-i
CO C
MH -H
O
i 1 l"^
1 1
0 O
rH rH
X M
vO CNJ
CO CO
CO
^J
£*
cu
^
w
T3
M
O
1
CO
cu
H CO
4J CU
H -H
rH H
CO 3
4J !-)
CO C
pen M
,
rrj
C!
Jj
^"J
**
CO
CO
rH
O
T3
fl
1
0
rH
X
T3 -^
a co
H M
\ cO
t-l rH
?% rH
^~. O
CO T3
^1
CO CNI
r-H I
rH 0
O rH
13
[xj
rH 0
i-H OO
cu
W) ft ft
cO O O
S ft ft
cO
f] \Q
O OO
>, rH O
4-1 t-H
!-i X
CU X
ft m
O r-l CN
^_l
PH
-------
179
planned to be transported). Both are highly volatile liquids such
that spills may create extensive flammable plumes. The plumes are
negatively buoyant and tend to lie flat on the ground or water. On
the other hand, there are important differences. LNG is a cryogenic
liquid consisting primarily of methane and is stored at atmospheric
pressure. LPG is a compressed gas consisting primarily of propane
and is stored at ambient temperature at a pressure of about 100 psig.
Because of this, a spill results in the immediate flashing of 30 to
35 percent into vapor.
About 90% of the LPG, 20 x 1Q9 gallons in 1973, is transported
by truck or a combination of pipeline and truck. The average truck
load is 4,370 gallons. By 1980 to 1985, it is expected that the U.S.
will be importing 24 x 109 gallons of LNG, which is equivalent to
2.06 x 1Q12 cubic feet of natural gas, in tank ships carrying approxi-
mately 32.5 x 1Q6 gallons each. Experiments with a few thousand
gallons of LNG suggest that the accidental rupture of a single cargo
tank could result in an LNG pool in water 1,500 feet in diameter and
a vapor air cloud which might remain flammable for a distance of
several miles downwind. Examination of accidents involving spills
of LPG and other volatile fuels indicates that a major fraction of
deaths, injuries, and property damage is caused by flash fire in a
large flammable vapor plume formed prior to ignition. In some cases,
explosions and even detonations may result when a portion of the
plume has infiltrated a building or other confined region.
Other types of fires, such as liquid pool fires, often cause
extensive property damage, but only infrequently cause fatalities
and injuries. The explosive rupture of LPG tanks because of over-
heating in a fire is another mechanism which may cause fatalities.
However, because it is stored at ambient pressure, this type of
accident is unlikely for LNG.
2. Implementing the Methodology for LNG and LPG Risks
Step 1. Cost-benefit relationship. Assuming the need for
energy, liquified and natural gases and liquid propane gas are highly
desirable alternates because of their low pollution burden when
burned. The cost of LNG may or may not be competitive with other
forms of energy, but sufficient potential for profit would seem to
exist based upon the rate of investment of the industry.
LPG has already demonstrated its cost-effectiveness for
areas where natural gas pipelines are unavailable or natural gas is
in short supply. On this basis, a P-factor of 10~1 seems to be rea-
sonable- to assign for both LNG and LPG, since benefits far outweigh
costs.
-------
180
Step 2. Achieving "as low as practicable" risk levels. New
LNG tankers use double-wall containers and new methods of storage
for LNG indicate that such storage facilities are relatively immune
to accidents caused by failure of the containers due to the low
temperatures involved. Their susceptibility of earthquake, airplane
crash, ship collision, and other types of major accidents, still
require further analysis. The double-walled tanker with separate
containers seems to be a reasonable means of minimizing risk, but
other cost-effective measures may still exist.
For the purpose of this exercise, it will be assumed that
tankers meet the criteria for "as low as practicable" risk reduction.
Thus, the P factor for LNG tankers remains 10~1.
LPG trucks and truck-pipeline combinations have been used for
many years, and reviewing the types of accidents that have occurred
over the years, as reported by Simmons,! it seems evident that more
safeguards are possible at reasonable incremental costs. This is
only a "snap value" judgment, but a reduction of the P factor by a
factor of two to five seems reasonable.
The adjusted P factors are:
PLNG = lo-1
PLPG = 5 x 10-2 - 2 x 10-2
Step 3. Degree of systemic control. The degree of systemic
control for LNG and LPG can be examined on either an absolute basis
through examination of the trend of the number of fatalities per year
or on a relative basis associating the number of fatalities with the
amount produced. A plot: of LPG production and fatalities per year
is shown in Figure 9-1. Both actual numbers of deaths per year and a
five-year running average are shown. On an absolute basis, the
number of fatalities is slowly rising. On the other hand, LPG pro-
duction is growing on an exponential rate (the amount produced is
nearly equal to the amount transported by pipeline, truck, or truck-
pipeline combination). The number of fatalities per million gallons
produced is decreasing.
The degree of sjrstemic control is, therefore, dependent upon
which criterion is chosen to represent the control function. On an
absolute basis, slightly negative, or at best level systemic control,
is observed. On a relative basis of amount transported, positive
Simmons, op. cit.
-------
181
o:
<
HI
Q
Q
Z
<
O
<
o
I
suo||BQgOl ui peonpoJd ^unoijuv
-------
182
systemic control is observed. A value judgment may be made in order
to assure that further possible controls are not ignored, and a
derating factor of two to five might be reasonable.
There is an inadequate data base for LNG shipments, but since
segregated, double-walled tankers are used, the same value as for LPG
will be assumed.
Step 4. Reconciliation of risk inequities. The risks imposed
by the transportation of LNG and LPG are involuntary risks resulting
from proximity to routes used in transportation. Both catastrophic
risks and ordinary accident risks are involved. The A factor for
involuntary risks is 10"7 fatalities per year per individual for
catastrophic risks, a factor of 50 higher for ordinary risks as dis-
cussed in the previous chapter.
AI (cat fat) = 10-7 fat/yr/ind
A2 (ord fat) = 5 x 10~6 fat/yr/ind
A3 (injuries) = 5 x 10"7 ini/yr/ind
A4 (prop dam) = $.02/yr/ind
Step 5. Risk acceptance levels. Sets of risk acceptance
levels can be determined for both LNG and LPG transport risks.
LNG - Catastrophic risk acceptance level:
10~7 x 10-1 x 2 x 10-1 = 2 x 10-9
R! (LNG) =
10"7 x io-l x 5 x 10-1 = 5 x 10-9
LNG - Ordinary risk acceptance level:
5 x 10-6 x 10-1 x 2 x 10-1 = 10-7
R2 (LNG) =
5 x 10-6 x ID-1 x 5 x 10'1 = 2.5 x 10~7
LNG - Injuries risk acceptance level:
5 x ID'7 x 10~! x 2 x 10~1 = 10~8
R3 (LNG) =
5 x 10-7 x 10-1 x 5 x 10-1 = 2.5 x 10~8
LNG - Property damage risk acceptance level:
$.02 x 10-1 x 2 x 10-1 = 4 x 10-4
R4 (LNG) =
$.02 x 10-1 x 5 x 10-1 = 1 x 10-3
-------
183
LPG - Catastrophic risk acceptance level:
ID"7 x 2 x ICr2 x 2 x 10-1 = 4 x lO"10
R! (LPG) =
10-7 x 5 x 10-2 x 5 x 10~1 = 2.5 x 10~9
LPG - Ordinary risk acceptance level:
5 x 10-6 x 2 x 10~2 x 2 x lO'1 = 2.0 x 10~8
R2 (LPG) =
5 x 10-6 x 5 x 10-2 x 5 x 10~1 = 1.25 x 10~7
LPG - Injury risk acceptance level:
5 x 10-7 x 2 x 10-2 x 2 x 10~1 = 2 x 10~9
R3 (LPG) =
5 x 10-7 x 5 x 10-2 x 5 x 10-2 = 1.25 x 10-9
LPG - Property damage risk acceptance level:
$.02 x 2 x 10-2 x 2 x 10~1 = 8 x 10~5
R4 (LPG) =
$.02 x 5 x 10-2 x 5 x 10~1 = 5 x 10~4
It is important to note that the ranges of uncertainty have
been preservedl by specifying the upper and lower limit.
3. Risk of Fatalities from LNG and LPG Accidents
The accidents considered in the Simmonsl study are limited to
accidents which lead to the formation of a flammable plume. Table 9-3
reproduces the results of the study for leaks and tank ruptures for
LPG tank trucks and LNG tank ships. An examination of this data shows
(based upon the definition for catastrophic accidents of 10 or more
fatalities without consideration of injury or property damage) that
the data may be divided up into two categories; namely, that for the
failure rate for catastrophic accidents and failure rate for other
types of accidents.
For LNG tank or transportation the catastrophic rate is 0.4
fatalities per year, and for the other lesser accidents, it is .015
fatalities per year. For LPG truck and truck pipeline transportation,
-------
184
TABLE 9-3
LNG and LPG Risk Estimates*
Accident Frequency
LPG Tank Trucks LNG
Fatalities
Per Accident
0.001-0.003
0.003-0.01
0.01-0.03
0.03-0.1
0.1-0.3
0.3-1.0
1.0-3.0
3.0-10
10-30
30-100
100-300
300-lxlO3
Ixl03-3xl03
3xl03-lxl04
A 4
1x10-3x10
*Reproduced from
Transport of LNG
Leak
3.5
4.6
2.9
1.4
0.50
0.15
0.046
0.011
1.2xlO~3
4.2xlO~3
_
_
_
_
Tank
RUpture
1.30
1.7
1.1
0.86
0.54
0.17
0.047
0.021
3.3xlO~3
1.2xlO~4
_
_
Leak
_
_
3.5xlO~3
4.9xlO~3
1.8xlO~3
1.2xlO"3
7.3xlO~4
4.3xlO~4
1.7xlO"4
3.3x!0"5
1.2xlO~6
_
_
_
(per year)
Tank Ships
Tank
Rupture
_
-
2.0xlO~3
1.6xlO~3
1.3xlO"3
9.0xlO"4
6.2xlO"4
3.8xlO~4
2.2xlO"4
l.lxlO"4
5.6xlO~5
1.8xlO~5
- 7.4xlO~7
John A. Simmons, Risk Assessment of Storage and
and LPG. November 25, 1974
(Draft) Final
Report
for Contract 68.01.2695, Sponsored by the Environmental Protection
Agency, p.5.
-------
185
the catastrophic failure rate is . 1 fatality per year, and that for
other types of accidents is 1.1 fatalities per year.
Of these fatalities, Simmons estimates that 76% are fatalities
that do not involve employees of the company and are in the area of
involuntary risk. The total U.S. population is subject to risk from
LPG tank trucks since the shipments of liquid propane by truck occurs
throughout the country on almost all of our roads. The population
exposed to risks from LNG tankers is harder to estimate since it is
primarily located at and near LNG equipped seaports. A population-at-
risk of 10 million people has been chosen to be used in this example
and has not been verified, but seems to be a reasonable number from
examination of the number of seaports and the people living within
reasonable distances of those seaports at any given time.
For LPG, Simmonsl has provided a "soft" estimate on injuries.
For the 36-year period from 1938-1973, there were 453 reported
injuries from LPG fires and explosions with a mean number of 12.6
injuries per year with a standard deviation of 46.5. No data were
found for LNG for injuries, and data on property damage for both are
even more suspect.2
A summary of these risk estimates and the involuntary risk
rates to individuals is contained in Table 9-4.
4. Comparison Risk Rates with Risk Acceptance Levels
Table 9-5 compares measured risk rates (M) against risk
acceptance levels (R) for LNG and LPG fatalities. It should also
be noted that this example only represents a partial examination of
acceptable risk levels since only dealing with consequences involving
fatalities and LPG injuries. As a result, the conclusions here are
incomplete since injury data for LNG and property for both must still
be obtained.
For LNG the measured risk rates are at least an order of
magnitude below the lowest end of the range for both ordinary and
catastrophic risk acceptance levels. At least in terms of risks
leading to fatalities, LNG must be judged as meeting acceptable levels
of risk.
llbid.
^Data for property damage for both LNG and LPG were never requested
prior to 1971. Data for 1972 and 1973 exist, but are too soft to
include as yet.
-------
186
CO
W
H
H
H
crt
W
H
O
W ,J
< §
H . cfl 0
--- o
4J rl Cfl
Cfl CU
M-I ft SH
OJ
* ^D Pi
0) rH
4-1 CD
CO Pi )H
pj CO O
rfi e
> 4J
rl CO O
rH CO
Cfl 0) O
4J h4 rH
CO
pD
CO
o
VD
r^ X
C-J
K
SO O
1^ rH
CO
rH
Cfl
3
a
60 -H
£ >
H -H
K* T3
rH C
0 M
C 0
rl 4-1
CO 01
0) 4J
H CO TJ
4-1 V{ M fl
H CO -rl
'H 'rl *^ **^
CO CO M CO rl
4-1 44 TH rS
CO CO 4J PH ^~
I-H *H Cfl 4->
rl >, CO
'4H C rl m
O >> O Cfl
Jj >H 4J
M Cfl 4-1 CJ
C 4J Cfl 3
(UP! rH rH
03 3 O
rl rH ft >
mo o pi
CL, > PH H
Pi
H
cr\ ON co
1 1 1
O 0 0
rH rH rH Cfl
4-1
XX X cfl
mo ro
rrj pj (f-J
H (U T3
CJ T3 CJ
a -H M
CO O
O O T3
rl Cfl 4J C
0) -H
0) 4-1 rl O>
O ft cfl £>> 60
rH ctf -~^ cfl
cu -n g
C rl ^ C CO
Cfl O M -H Q
rC 0 -H
4~* fvj ^,
M 1 1
CO O >-, rl
co s-i at
CU O 3 ft
rl rH !-, 0
Pi rl
M PM
0)
4-1
cfl
g
rl
4J
CO
W
*
-------
187
O)
,_j
w
t*
W C/>
t-I H
W W
CJ Q
a M
<3 u
H U
PH
cu x^
t-4
13
CU C
0 -H
13 -^
ct) Vj
4-> X
&i ^^^
CU 4-1
O CO
O 4-1
-<3j ^^
s^
CO
H
Pi
s~ *.
g
cu
^ '""" \
cu
(3
r-^M *lH
CO ^
"H >-4
Pi ^P-.
-O 4J
C1J OJ
1-1 4-1
^ ^^^
CO
cd
cu
^1
CU
0.
H
*^
co
H
Pi
o\
O^
I ON CT* ON 1
0 1 1
i-H 0 0
.-1 rH
X
»J» ^t
^< ^<
m
0
CM m co
l-H r-l
1
1 1
^3"
O CM
O CN|
1 0
O r-l
t-H
X
X
O
o
m
m CM
I I
CM O
o
i__i
O^N O"s CT> O^ ON
i i i
o o o
^H r~H rH
XXX
o m co
^" *>Q
rH
CO
43
CO 4->
43 cd
4-1 QJ
rt 13
0)
'O 4J O 4-*
(3 .H t3
cu cu cu
O -H CO -H W
00 43 0 (U
O 4-1 O -H
M n) nj ^
O CO D
)-4 CO >-l T-)
o cu cu cu c
^ O, i-4 O. M
O cj
s s
i i
o o
1 1
^1 r~l
X X
^- ^>
*
O *d"
CD
(~j
CO 4-1
43 CO
4J CU
0) 13
0)
O 4-> O 4-1
C! r-\ C
cu cu cu
}-j rO C3 T3
O -H CO -H
60 43 0
o 4J a
!-( CO CO
O CO
S-J CO t-l
o cu cu cu
rH a- h4 a,
-------
188
LPG is a different case. Ordinary event and injury risk rates
are over an order of magnitude below the low end of the range of the
risk acceptance level. However, for catastrophic events, the measured
risk rate exceeds the highest end of the range of the risk acceptance
level by a factor of 40. Even if complete positive systemic control
was demonstrated (a degree of control factor of unity, leading to a
risk acceptance level of 5 x 10~9 fat/yr/ind), the measured risk
would exceed the acceptable level of risk by almost an order of magni-
tude. One must conclude that catastrophic, involuntary risk from
LPG transport is too high and is not acceptable to society in its
present form, and increased effort and expenditure for risk reduction
seem warranted.
H. EXTENDED USE OF THE METHODOLOGY
The methodology can be used in a broader sense than illustrated
previously. Alternate technological systems can be evaluated against
one another by applying the methodology to each system and then com-
paring the results.
The comparison can take place on different levels. First, one
can make a risk acceptance level comparison by calculating acceptable
levels of risk for the different activities without regard to
measured levels of risk. This comparison provides some insight as
to relative risk among the different systems. For example, by com-
paring the risk acceptance levels for fatalities in Tables 9-1 and
9-5, both LNG and LPG have higher levels of acceptable risk than a
nuclear industry with 100 power plants. On the other hand, the
optimistic case for nuclear power, as shown in Table 9-2, shows a
balance in favor of nuclear energy over LNG and LPG.
Secondly, one can make a comparison of the ability of measured
risk levels to meet acceptable levels of risk for the alternate
systems. In this case, the ability of systems to meet acceptable
levels is compared.
In either case, this comparison is made only in terms of risk.
It provides one parameter for comparison in an overall risk-cost-
benefit analysis. As such, the methodology only addresses the diffi-
cult question of risk acceptance. It may be used as part of a
broader analysis, not as a substitute.
-------
TRACK A & B
CHAPTER X
CONCLUSIONS
Risk is indeed a complex matter with many factors and variables.
Too often risks are compared with one another erroneously since
different risk factors are involved. Essentially, "comparing apples
with oranges" misrepresents the underlying processes and can provide
a major disservice to society. The purpose of this study has been
to identify, to the extent possible, the existence of these factors,
to provide some estimate of their sensitivity and impact on risk
decisions, to attempt to quantify the effect of the different factors,
and as a result, provide a basis for comparing like types of risks
with each other.
The difficulties in obtaining data on risks and risk factors are
considerable. For example, little data has been taken on cancer
fatalities to determine which cases are involuntary or voluntary (such
as workers taking risks on a knowledgeable basis). Nevertheless,
pertinent data can be obtained or synthesized (for sensitivity
analysis) to provide useful insight and aid in decision making pro-
cesses. The risk factors associated with man-originated involuntary,
catastrophic accidents is a case in point where reasonable data are
available.
Such data can be used to assure that new risks of the same type
can be compared with existing levels of risk. These comparisons allow
methodologies, such as the one prescribed in Chapter IX, to be formu-
lated to determine acceptable levels of societal risk. Value judg-
ments made in a visible, repeatable manner are an inherent part of
such methodologies, imparting a subjective element into their applica-
tion. However, the impact of these value judgments must not be
masked by improper data comparison resulting from oversimplification
of the problem.
There are several major conclusions which can be identified
as a result of this effort. They are: (1) the factors affecting
risk valuation can be identified and studied in detail to provide
better understanding of this extremely complex individual and societal
problem; (2) oversimplification of the risk problem can lead to mis-
representation of risk conditions and levels of risk acceptance;
(3) the effect of risk factors and individual propensity for risk
taking can be measured, but detailed data in forms suitable for such
analysis is indeed difficult to obtain since the existing basis for
189
-------
190
obtaining data does not usually allow the identified factors to be
easily analyzed.
Acceptable levels of risk for society can be obtained through
examination of historic societal behavior to existing risks (when
risks are known) as an external referent, and comparisons of new
risks against existing societal behavior for similar risks by pre-
established methodologies. The methodologies involve value judgments
but, if made in a visible manner, can be argued and agreements and
disagreements made specific.
Two types of societal value judgments are involved: gross level,
non-technical judgments as to the burden society will accept for an
activity, and technical judgments as to the degree which given sys-
tems most different levele benefit-cost balances and degrees of
control. The separation of these value judgments makes it possible
for all risk takers to participate in the key judgments made at
gross levels. These gross value judgments do not imply simplifica-
tion; quite the contrary, they indicate the level of precision that
is meaningful to risk agents, and that further precision is fruitless.
There is no question that considerably more effort must be expended
in this area.
It is hoped that this categorization and identification of risk
factors and demonstration of at least one risk acceptance methodology
will stimulate further efforts in the field.
-------
APPENDIX A
DIFFERENCES IN PRESENTATION OF DATA
The validity of conclusions drawn from analysis of data depends
wholly on the quality of the data base and validity of statistical
inferences. Comparison of results among different data bases are
even more difficult to do in a valid manner. Problems that must be
considered involve the comprehensiveness of the data reporting system,
the definition of reportable events and resulting ambiguity in report
generation, along with valid use of statistical presentations.
An example of the need to resolve statistical presentation methods
is provided by the Rasmussen Report.1 Data are given for 51 events
associated with the major U.S. hurricanes between the years 1900-1972.
The consequence range from 6,000 fatalities to no fatalities. Of
these, 46 specific events having fatalities are individually listed,
along with a single grouping of five events with no fatalities, but
damages over $5 million.
A cumulative distribution of the frequency of hurricanes with
consequences greater than N is given, and reproduced as Figure A-l.
Figure A-2 is a histogram of the number of events occurring in decade
intervals for the 46 events involving fatalities. On the basis of
this histogram, the mode of the distribution is between 10 and 100.
The mean value of the 46 events is 273.11 with a standard deviation
of 915.95, and the total number of fatalities is 12,577. However,
when a log-normal distribution is used for the data, the mean number
of fatalities is 34.24 with a standard deviation of 7.76. The mean
frequency of occurrence is 0.64 events per year, and probability of
fatalities per year is:
TT (normal distribution) = 0.64 x 273.11 = 174 fat/yr (A-l)
N (log-normal distribution) = 0.64 x 34.24 = 22 fat/yr (A-2)
In order to gain the same information from the cumulative distribution
shown in Figure A-l, the curve in Figure A-l must be integrated over
the number of fatalities per event. The expected value (EV) can be
defined as follows:
EV = p(l) x 1 + p(2) = P(3) x 3 H +p(n)max x nmax (A-3)
where p(n) x n = probability of exactly n fatalities. However,
iRasmussen Report, op. cit. , pp. 202-204.
191
-------
192
Figure A-l
HISTOGRAM OF MAGNITUDE OF U.S. HURRICANES (1900-1972)
IN TERMS OF NUMBER OF FATALITIES (N)
ON A LOGARITHMIC SCALE
20
15
en
-P
C
Q)
H 10
4-1
O
0)
14
19
11
10 100 1,000 10,000
Number of Fatalities per Event (N)
-------
193
Figure A-21
FREQUENCIES OF METEORS WITH CONSEQUENCES GREATER THAN N
10-7
10?
100
N (DOLLAMS)
108
1,000
N ( Fatalities )
1010
10->
1C«
105
10,000
- 10'
100000
-"-Rasmussen Report, op. cit., p. 211.
-------
194
p(N > 1) = p(l) + p(2) + p(3) P(n)max (A-4)
and:
p(N > n) + p(n+l) + P(n+2) H +p(nmax) (A-5)
Therefore:
EV = p(N > 1) + p(N > 2) + p(N > 3) H p(N > nmax) (A-6)
n
max
where
EV = > p(N > n)
n=l
number of events meas N > n no. of events (N > n)
p(N > n) = -- = - (A-8)
number of years measured 72
= probability of an event with N > n occurring per year
EV = 255.72 fat/yr for the integral of the curve in Figure A-l, and is,
by definition, the best estimate of expected value. This value is 46%
greater than the mean value for a normal distribution.
The integration of the cumulative curve for acute fatalities from
100 nuclear power plants, as given in the Rasmussen Report,! results
in an expected value of:
EV = 0.04 fat/yr
for a population of 15 million people, resulting in an individual risk
rate of 2.67 x 10"^ fat/yr /ind which compares closely with the value
of 3 x 10~9 fat/yr/ind provided by the study.
Rasmussen Report, op. cit.
-------
APPENDIX B
AN EXAMPLE OF DIFFICULTY OF ANALYZING RISK DATA
There is a wealth of data available concerning voluntary risks in
the coal mining industry. Annual reports from the U.S. Department of
Interior, Bureau of Mines, provide in-mine and work related fatalities
in the form of Mineral Industry Surveys. The data extends back at
least as far as 1941.1 The overall industry trend in total fatalities
per year is downward and the frequency rates of fatalities per million
man-hours show a slight downward trend, indicating positive systemic
control. Analysis of different types of mining operations and accident
situations shows a variety of processes, some positively controlled,
some not.2 The decline of the number of fatalities is even better if
it is not measured in fatalities per year per individual but per tons
of coal mined. The efficiency or productivity of the industry/mine-
worker (due to mechanization) has increased manyfold in the last 20
to 30 years. Hence, for the coal mined, fewer workers are employed.
In 1973, the total individual voluntary risk rate was about 1.1 x 10~3
fat/yr/ind for 132 fatalities for about 120,000 mine workers.3
On the other hand, involuntary risks associated with coal mining
are derived from subsidence of structures, refuse pile movement,
refuse dam failure and ultra-active nuisance from abandoned mines
and refuse piles. The refuse dam failure at Saunders, West Virginia,
in 1972, is well documented,1^ and indicated that 125 involuntary
fatalities occurred. However, these failures occur infrequently, and
while workers in the field remember others, there is little documenta-
tion to provide frequency and average number of fatalities. More
importantly, there is no record of the failures or failure rates of
"Coal-Mine Fatalities in 1972," U.S. Department of Interior, Bureau
of Mines, January 14, 1972.
o
Coal-Mine Fatalities in 1970," U.S. Department of Interior, Bureau
of Mines, January 10, 1971.
3"Coal-Mine Fatalities in 1973," U.S. Department of Interior, Bureau
of Mines, January 1974.
^"Preliminary Analysis of the Coal Refuse Dam Failure at Saunders,
West Virginia," February 26, 1972, U.S. Department of Interior Task
Force to Study Coal Waste, HAZARDS, March 12, 1972.
195
-------
196
such dams1 or the populations at risk as a result of a dam's existence
that was found after exhaustive search and direct communication with
the Mining Enforcement and Safety Administration (MESA). In other
words, there is inadequate data to compute the involuntary risk rate.
This is also true for subsidence fatalities (if any exist) or involun-
tary fatalities from U.S. refuse pile movement. Likewise, the Bureau
of Mines does not keep records of involuntary fatalities for all mines.
They only record for U.S. controlled mines. For example, there are
some fatalities from old abandoned mines where workers or children
may fall into a shaft, etc. There are no records for such accidents.
The problem is that, even if the events and magnitude of conse-
quences were well documented, estimation of the actual population at
risk is usually lacking. The inescapable conclusion is that society
has evidently never had a serious concern in the past with involuntary
risks from industrial activity resulting in the lack of data. The
recently signed contract of the UMW has no provisions for involuntary
fatalities from coal mining activities, while it has a great deal for
the protection of the miners at work. It may well be that the possi-
bility of a large number of involuntary fatalities from the nuclear
power industry are, for the first time, making the problem of invol-
untary risk from man-made activities one of concern.
'-There is only a record in West Virginia of the most important of such
dams as a result of the 1972 disaster.
-------
APPENDIX C
CONSADl REPORT COMMENT ON CLASSIFICATION
OF INVOLUNTARY RISKS
Classifying each accident within the chronology - as involving
voluntary or involuntary risk - was problematic. Accidents whose
description specifically noted casualties to bystanders or other
victims who could not reasonably be expected to have anticipated the
possibility of such an event were classified as having an involuntary
risk factor; in the absence of such specific mention, accidents were
classified as involving only voluntary risk. Undoubtedly, this type
of arbitrary classification scheme, and its apparent lack of speci-
ficity, poses some very real problems which should be recognized
before attempting to use these figures in any specific context.
An example of the weaknesses in this scheme can be found in the
classification of accidents involving oil refineries and chemical
plants. In most cases, available information was very sketchy -
sketchy in the sense that no clear indication was given regarding
what segment, employees or residents of the surrounding area, suffered
the casualties. For example, an explosion at a chemical plant could
involve injuries to employees, residents of the surrounding area, or
both. If the explosion involved injuries to employees only, the risk
factor was considered voluntary on the grounds that those who work at
the plant consciously accept the possibility of being involved in an
accident when they make the decision to work there. If the explosion
involved injuries to residents of the surrounding area, the risk
factor was involuntary - those who reside in the vicinity of the plant
do not consciously accept the possibility of an accident at the chemi-
cal plant directly involving them.
Obviously, this interpretation is only one of many interpretations
possible - another being that those residing in the vicinity of the
plant consciously accept the risk of an accident at the plant in-
volving them when they decide to establish residence near the plant -
making the risk factor to them a voluntary one. A case can be made
for the viability of either instance. Keeping the feasibility of
those interpretations in mind, one could then deduce that there are
very few instances of accidents involving involuntary risk - again,
classification is very difficult and not entirely reliable.
We fully recommend this classification - voluntary or involuntary
risk factor - should be used always, keeping in mind the apparent
weaknesses of the scheme.
icONSAD, op. cit.
197
-------
GLOSSARY
ACCURACY - The quality of being free from error. The degree of accuracy
is a measure of the uncertainty in identifying the true measure of a
quantity at the level of precision of the scale used for the quantity.
ALGORITHM - A standard set of procedures for solving a mathematical
problem (as of determining the greatest common divisor) that frequently
involves repetition of an operation. An algorithm is an expression
utilizing factors (some of these factors may even be intangibles)
which can be assigned value; that is, which can be quantified.
BAYESIAN STATISTICS - "Bayes 'rule' (Thomas Bayes, a 19th Century
English mathematician and clergyman) states that the probability that
both of two events will occur is the probability of the first multi-
plied by the probability that if the first has occurred the second
will also occur. Bayesian statistics is a way of making quantity of
information substitute for quality of information. The problem lies
in the fact that there are two kinds of probabilities: the classical
type derived from empirical information, and the subjective proba-
bilities. Bayesian statistics is based upon these 'subject proba-
bilities'." It is also called the joint probability of A and B. The
probability of the second event occurring if the first has occurred
is called the conditional probability of the second, given the first.
Stated another way, the probability of any event P(A) is always posi-
tive but never greater than 1. Symbolically, 0 v< P(A) ^ 1. If P(A) =
0, then the occurrence of the event (A) is considered impossible. If
P(A) = 1, then the occurrence of the event (A) is considered to be
certain.
CARDINAL SCALE (Interval Scale) - A continuous scale between two and
points, neither of which is necessarily fixed.
COMPONENT UTILITY FUNCTION - The utility assigned to a subgoal.
COMPOUND UTILITY FUNCTION - The resultant utility function formed by
combining a multiplicity of component utility functions by some mathe-
matical or logical rule.
DECISION MAKING - A dynamic process of interaction, involving informa-
tion and judgment among participants who determine a particular policy
choice. Decision models denote either models of the decision making
process itself, or analytical models (for example, decision trees,
decision matrices) used as aids in arriving at the decisions. Deci-
sion theories usually are in relation to the process itself.
198
-------
199
DECISION MATRICES - Matrices whose elements exhibit quantitative
relationships (cardinal or ordinal) among sets of factors coming into
play in the decision making process, such as tht relevance of a set
of technologies to a set of missions (technology/mission matrix) or a
set of scientific disciplines to a set of technologies (science/tech-
nology matrix), or a set of resources (physical or otherwise) required
for a set of research tasks (task/resources matrix), etc.
DECISION TREES - A graphical display of logical relationships among
actions or events to be considered in the pursuit of a given objec-
tive, and exhibiting branch points where decisions must be made.
EXTRINSIC PARAMETER - A variable whose value may be determined empiri-
cally by direct nhysical measurement.
HEURISTIC - An operational maxim derived from experience and intuition.
IMPRECISION - The degree of inexactness for which a quantity can be
measured.
INACCURACY - The degree of error in identifying the true measure of
a quantity.
INTRINSIC PARAMETER - A variable whosa measurement is based upon the
value system of an individual and his prevention of these values.
NOMINAL SCALE (Taxonomy) - A classification of items which may be
distinguished from one another by one or more proper ties.
ORDINAL SCALE (Rank Scale) - An ordering (ranking) of items by the
degree they obtain some criterion.
PARADIGM - A structured set of concepts, definitions, classifications,
axioms, and assumptions used in providing a conceptual framework for
studying a given problem.
PRECISION - The exactness with which a quantity is stated, i.e., the
number of units into which a measurement scale of that quantity may
be meaningfully divided. The number of significant digits is a
measure of precision.
PREFERENCE - Assignment of rank to items by an agent when the criterion
used is the utility to the ranking agent.
RELEVANCE TREES - A synonym for decision trees.
-------
200
SATISFICING - The selection of a decision function which optimizes
an individual's freedom from anxiety as opposed to a function which
optimizes overall organization goals.
TRANSIVITY - A property of some ordinal scales where, if A is pre-
ferred to B, and B is preferred to C, then A is preferred to C.
UNCERTAIN UTILITY FUNCTION - A cardinal utility function with a finite
level of precision and/or accuracy.
UTILITY - A scale expressing the satisfaction of a rational, economic
man's wants and desires.
UTILITY FUNCTION - A scale of preference (ordinal) or value (cardinal)
to a decision maker or a multiplicity of decision makers.
VALUATION - The act of mapping an ordinal scale onto an interval
scale, i.e., assign a numerical measure to each ranked item based
upon its relative distance from the end points of the interval scale.
VALUE - A scale expressing the satisfaction of man's intrinsic wants
and desires.
VALUING - The act of assigning a value to a risk consequence.
-------
BIBLIOGRAPHY
Alfven, Hannes. "Fission Energy and Other Sources of Energy."
Bulletin of the Atomic Scientists, Vol. 30, No. 1, January 1974,
pp. 4-8.
Atomic Energy Commission. 1972 Atomic Energy Programs: Operating
and Developmental Functions. Washington, B.C., U.S. Government
Printing Office, 1973.
. "Theoretical Possibilities and Consequences of Major
Accidents in Large Nuclear Power Plants (WASH-740)." Report to
the Joint Committee on Atomic Energy, Congress of the United
States, March 1957.
. Proposed Appendix I to AEC Regulation 10 CFR 50.
Atzinger, Erwin, M., Brook, Wilbert J., Chernick, Michael R., Eisner,
Brian, and Foster. Ward V. "Compendium on Risk Analysis Tech-
niques." U.S. Army Materiel Systems Analysis Agency Special
Publication Number 4, Aberdeen Proving Grounds, Maryland,
July 1972.
Baram, Michael S. "Technology Assessment and Social Control."
Science, Vol. 180, No. 4085, May 4, 1973, pp. 465-473.
Basselman, Fred, Callies, David, and Banta, John. The Taking Issue.
President's Council on Environmental Quality, 1973. Superinten-
dent of Documents, U.S. Government Printing Office, Washington,
D.C. 20402 (Stock No. 4111-00017).
Bernoulli, Daniel. "Specimen Theoriae Novae de Mensura Sortis," 1738,
as referenced bv David W. Miller and Martin K. Starr, The Struc-
ture of Human Decision, Englewood Cliffs, New Jersey: Prentice-
Hall, 1967).
Boffey, Philip M. "Radiation Standards: Are the Right People
Making Decisions?", Science, Vol. 171, No. 3973, February 26,
1971, pp. 780-783.
Brown, Richard S. and Enke, Steven. "Economic Worth of Preventing
Death at Different Ages in Developing Countries," Journal of
Biosocial Science, Vol. 4, No. 3, July 1972, pp. 299-313.
Coates, J. Calculating the Social Costs of Automobile Pollution -
An Exercise, Symposium on Risk vs Benefit, Los Alamos, November
1971.
201
-------
202
Conner, James E. "Prospects for Nuclear Power," The National Energy
Problem, edited by Robert H. Connery and Robert S. Gilmour.
Montpelier, Vermont: Capital City Press, 1974.
CONSAD Research Corporation. Extract from the "Consequences and
Frequency of Selected Man-Oriented Accident Events." U.S.
Environmental Protection Agency Contract Report No. 68-01-0492.
Coombs, H. C. and Pruitt, D. C. "Components of Risk in Decision
Making; Probability and Variance Preferences," Journal of Experi-
mental Psychology, Vol. 60, 1960, pp. 265-277.
Corliss, William R. and Seaborg, Glenn T. Man and Atom, New York:
E. P. Button and Co., 1971. ~~
Dalkey, N. C. "Quality of Life," The "Quality of Life" Concept,
Washington, D.C., U.S. Government Printing Office, 1972, pp. V-19-
V-29.
David, H. A. and Moeschberger, M. L. "Life Tests Under Competing
Causes of Failure and the Theory of Competing Risks," Biometrics,
Vol. 27, No. 4, December 1971, pp. 909-933.
Davis, Joyce P. "The Regulation of the Environmental Effects of
Nuclear Power Plants," Nuclear Safety, Vol. 14, No. 1, January-
February 1973, pp. 6-13.
. "The Regulation of the Environmental Effects of Nuclear
Power Plants," Nuclear Safety, Vol. 14, No. 3, May-June 1973,
pp. 165-181.
Dolphin, G. W. and Beach, S. A. "The Relationship Between Dose
Delivered to the Thyroids of Children and the Subsequent Develop-
ment of Malignant Tumors," Health Physics, Vol. 9, No. 12,
December 1963, Pergamon Press, pp. 1335-1390.
Edwards, W. "Subjective Probability in Decision Theories," Psycho-
logical Review, Vol. 79, 1962, pp. 109-135.
Eisenbud, Merrill. Environmental Radioactivity, New York: McGraw-
Hill Book Company, Inc., 1963.
Farmer, F. R. "A British View - Safety and Nuclear Power Plants,"
Bulletin of the Atomic Scientists, Vol. 27, No. 9, November
1971, pp. 47-49.
-------
203
Federal Food, Drug, and Cosmetic Act, Section 409c(3)A.
Federal Radiation Guide, Reports No. 1-7, Federal Radiation Council.
Fine, William T. "Mathematical Evaluations for Controlling Hazards,"
Naval Ordinance Laboratory Technical Report, NOLTR 71-31, White
Oak, Maryland, March 8, 1971.
Fishburn, Peter C. "Utility Theory with Inexact Preferences and
Degrees of Preference," Synthese 21, 1970, pp. 204-221.
Foreman, Harry, ed. Nuclear Power and the Public, Minneapolis,
Minnesota: University of Minnesota Press, 1970.
Fraas, A. P. and Steiner, D. "Preliminary Observations on the
Radiological Implications of Fusion Power," Nuclear Safety,
Vol. 13, No. 5, September-October 1972, pp. 353-362.
Friedman, Milton, and Savage, L. J. Journal of Political Economy,
Vol. LVI, 1948, pp. 279-304.
Gardenier, John S. "Concepts for Analysis of Massive Spill Accident
Risk in Maritime Bulk Liquid Transport," U.S. Coast Guard Office
of Research and Development Report No. 723111, Washington, D.C.,
December 1971.
Giddings, J. Calvin. "World Population, Human Disaster and Nuclear
Holocaust," Bulletin of the Atomic Scientists, Vol. 29, No. 7,
September 1973, pp. 24-50.
Gillette, Robert. "Nuclear Reactor Safety: A Skeleton at the Feast?",
Science, Vol. 172, No. 3986, May 28, 1971, pp. 918-919.
. "Nuclear Safety: AEG Report Makes the Best of It,"
Science, Vol. 179, No. 4071, January 26, 1973, pp. 360-363.
. "Radiation Spill at Hanford: The Anatomy of an Accident,"
Science, Vol. 181, No. 4101, August 24, 1973, pp. 728-730.
"Radiation Standards: The Last Word or at Least a
Definitive One," Science, Vol. 178, No. 4064, December 1, 1972,
pp. 966-967 and p. 1012.
Gilson, Charlotte R.' "individual Differences in Risk Taking,"
Yale University, Department of Psychology Technical Report No. 13,
New Haven, Connecticut, June 1968.
-------
204
Golden, A. S., Moeller, D. W., and Yeats, D. B. "Natural Radiation
in the Urban Environment," Nuclear Safety, Vol. 13, No. 4,
July-August 1972, pp. 275-286.
Greenfield, M. A. and Starr, Chauncey. "Public Health Risks of Thermal
Power Plants," Nuclear Safety, Vol. 14, No. 4, July-August 1973,
pp. 267-274.
Gruenstein, Peter and Sandier, Richard H. "Power from Fission:
Potential for Catastrophe," The Progressive, Vol. 37, No. 11,
November 1973, pp. 36-41.
Gupta, R. B. and Rao, G. Rama. "Effect of Elimination of Different
Causes of Death on Expectation of Life - Bombay, 1960-61,"
Indian Journal of Medical Research, Vol. 61, No. 6, June 1973,
pp. 950-961.
Haaland, Gordon A., Pruitt, Dean G., St. Jean, Richard, and Teger,
Allan I. "A Re-examination of the Familiarization Hypothesis
in Group Risk Taking," State University of New York Technical
Report No. 9, New York., New York, February 17, 1969.
Hartman, Robert S. "The Structure of Value: Foundations of Scienti-
fic Axiology," Carbondale, Illinois: Southern Illinois University
Press, 1967.
Health, Education, and Welfare - U.S. Department of. "The Health
Consequences of Smoking," January 1974.
. "Vaccination Against Smallpox in the United States: A
Reevaluation of the Risks and Benefits," Public Health Service,
Center for Disease Control, Atlanta, Georgia 30333, Revised
February 1972.
Heinrich, H. W. Industrial Accident Prevention, Third Edition,
New York, McGraw-Hill B.ook Company, 1950, pp. 332-334.
Henderson, W. Paul. "Pyrotechnic Hazard Evaluation and Risk Concepts,"
Paper presented at the Conference on Hazard Evaluation and Risk
Analysis, Houston, Texas, August 18-19, 1971.
Higbee, Kenneth L. and Streufert, Siegfried. "Group Risk Taking
and Attribution of Causality," Purdue University Technical Report
No. 21, Lafayette, Indiana, April 1969.
-------
205
Holder, Lawrence C. "National Network of Physicians Investigates
Claims of Radiation Injury in the Non-Occupationally Exposed
Population," American Journal of Public Health, Vol. 62, No. 10,
October 1972, pp. 1387-1390.
Homans, George. The Human Group, New York, Harcourt, 1950.
Howard, R. A. "Decision Analysis: Applied Decision Theory," Pro-
ceedings of Fourth International Conference on Operational
Research, Vol. SSC-4, No. 3, September 1968, pp. 211-219.
Huebner, S. S. and Black, Kenneth. Life Insurance, Fifth Edition,
New York, Appleton-Century-Crofts, Inc., 1959.
. The Economics of Life Insurance, Third Edition, New
York: Appleton-Century-Crofts, Inc., 1959.
Insurance Information Institute. Insurance Facts, 1966 Edition,
New York, New York 10031.
Joint Committee on Atomic Energy. "Probable Numbers and Cost Through
1985 of Lung Cancer Cases," Woodward and Fondiller, Inc., 1967
Appendix No. 7, Hearings before the Subcommittee on Research
Development and Radiation, 90th Congress, First Session, Part 2
(Appendix 7), pp. 975, and 1007.
Interior, U.S. Department of. "Coal-Mine Fatalities in 1970,"
Bureau of Mines, January 10, 1971.
. "Coal-Mine Fatalities in 1972," Bureau of Mines,
January 14, 1972.
"Coal-Mine Fatalities in 1973,"Bureau of Mines,
January 1974.
Jessen, Peter J. "Defining 'Quality of Life' Measures - The State
of the Art," The "Quality of Life" Concept, Washington, D.C.,
U.S. Government Printing Office, 1972, pp. I-1-I-15.
Jewell, William S. "The Credible Distribution," paper presented at
the 10th ASTIN Colloquim, University of Essex, Colchester, U.K.,
September 4-8, 1973.
Keeler, Emmett and Zeckhauser, Richard. Another Type of Risk Aversion,
RM-5996-PR, Santa Monica, California: The RAND Corporation,
May 1969.
-------
206
Kogan, N. and Wallach, M. A. Risk Taking: A Study in Cognition and
Personality, New York: Rineburt and Winston, 1967.
Lane, J. Michael, Miller, J. Donald, and Neff, John M. "Smallpox
and Smallpox Vaccination Policy," Annual Review of Medicine,
Vol. 22, 1971, pp. 251-272.
Lapp, Ralph. "Nuclear Power Boom - Scourge or Blessing?", The
Washington Post, April 14, 1974, pp. Bl and B4.
Lederberg, Joshua. "Squaring an Infinite Circle - Radiobiology and
the Value of Life," Bulletin of the Atomic Scientists, Vol. 27,
No. 7, September 1971, pp.43-45.
Lenox, Hamilton D. "Risk Assessment," Unpublished Thesis, Air Force
Institute of Technology, June 1973.
Lewis, Richard S. The Nuclear Power Rebellion, New York: The
Viking Press, 1972.
Libby, L. M. Technological Risk Versus Natural Catastrophe, P-4602,
Santa Monica, California: The RAND Corporation, March 1971.
Lichtenstein, S. "Bases for Preferences Among Three Outcome Bets,"
Journal of Experimental Psychology, Vol. 69, 1965, pp. 162-169.
Lieblich, Amia. "The Effects of Stress on Risk Taking," Journal of
Psychonomic Science, Vol. 10, No. 8, 1968, pp. 303-304.
Linstone, Harold A. "Planning: Toy or Tool?", IEEE Spectrum,
April 1974, pp. 42-49.
Luce, R. Duncan and Raiffa, Howard. Games and Decisions: Introduc-
tion and Critical Survey, New York: Wiley and Sons, 1917.
Maslow, Abraham. Motivation and Personality, Harper and Row, 1954.
Metzger, H. Peter. The Atomic Establishment, New York: Simon and
Schuster, 1972. ~
Mihalasky, John. "Decision Risk Analysis: Problems in Practice,"
Paper presented at the Conference on Hazard Evaluation and
Risk Analysis, Houston, Texas, August 18-19, 1971.
National Academy of Sciences. "Air Quality and Automobile Emission
Control," Washington, D.C. 20006, September 1964.
-------
207
. "The Effects on Populations of Exposure to Low Levels of
Ionizing Radiation," Report of the Advisory Committee on the
Biological Effects of Ionizing Radiation (BEIR), Division of
Medical Sciences, National Academy of Sciences, National Research
Council, Washington, B.C. 20006, November 1972.
National Safety Council. Accident Facts, Chicago, Illinois, 1973.
Nogami, Glenda Y. and Streufert, Siegfried. "Time Effects on Per-
ceived Risk Taking," Purdue University Technical Report, No. 11,
Lafayette, Indiana, July 1973.
Novick, Sheldon. The Careless Atom, Boston, Massachusetts: Houghton
Mifflin Company, 1969.
Orsenberg, Herbert S., Eilers, Robert D., Hoffman, G. Wright, Kline,
Chester A., Melone, Joseph J., Snider, H. Wayne. Risk and Insurance,
Prentice-Hall, Inc., Englewood Cliffs, New Jersey, 1964, p. 62.
Otway, H. J. "The Quantification of Social Values," Risk vs. Benefit:
Solution or Dream, LA 4869-MS, February 1972.
Pouliquen, Louis Y. Risk Analysis in Project Appraisal, London:
The Johns Hopkins Press, 1970.
Pruitt, D. C. "Pattern and Level of Risk Taking in Gambling Decisions,"
Psychological Review, Vol. 69, 1962, pp. 187-201.
Rasmussen, Norman, et al. "An Assessment of Accident Risks in U.S.
Commercial Nuclear Power Plants," WASH-1400, U.S. Atomic Energy
Commission, August 1974, p. 189.
Reider, Roy. "Technology and Safety - A Qualitative View," Paper
presented at the Conference on Hazard Evaluation and Risk Analy-
sis, Houston, Texas, August 18-19, 1971.
Rowe, W. D. "Decision Making with Uncertain Utility Functions,"
Doctoral Thesis: American University, Washington, D.C., 1973,
pp. 27-33.
Rowe, W. D. "The Application of Structural Value Analysis to Models
Using Value Judgments as a Data Source," Technical Report M 70-14,
The MITRE Corporation, McLean, Virginia, 1970.
Sagan, L. A. "Human Costs of Nuclear Power," Science, Vol. 177,
No. 4048, August 11, 1972, pp. 487-493.
-------
208
Savage, Leonard. The Foundation of Statistics, New York: Wiley &
Sons, 1954.
Schlaiffer, Robert. Analysis of Decisions Under Uncertainty, New
York: McGraw-Hill, 1969.
Selman, Jerome. "Decision Risk Analysis: Risk Theory," Paper
presented at the Conference on Hazard Evaluation and Risk
Analysis, Houston, Texas, August 18-19, 1971.
Siccama, E. H. "The Environmental Risk Arising from the Bulk Storage
of Dangerous Chemicals," Paper presented at the Conference on
Hazard Evaluation and Risk Analysis, Houston, Texas,
August 18-19, 1971.
Simmons, John A. "Risk Assessment and Transport of LNG and LPG,"
Draft version - final report for contract 68.01.2695, Environ-
mental Protection Agency, Washington, D.C., November 25, 1974.
Simon, H. A. Models of Man, Bailey: New York, 1957.
Starr, Chauncey. "Social Benefit versus Technological Risk," Science,
Vol. 165, No. 3899, September 19, 1969, pp. 1232-1238.
Starr, Chauncey. "Benefit-Cost Relationships to Socio-Technical
Systems," Environmental Aspects of Nuclear Power Stations,
IAEA SM-146/47, IAEA, Vienna, 1971, p. 900.
Stevens, S. S. "Mathematics, Measurement, and Psychophysics," Hand-
book of Experimental Psychology, Edited by S. S. Stevens, New
York: Wiley & Sons, 1951.
Storer, John B. "Late Effects: Extrapolation to Low Dose Rate
Exposures," Health Physics, Vol. 17, No. 1, July 1969, pp. 3-9.
Streufert, Siegfried and Streufert, Susan C. "The Effect of Informa-
tion Quantity and Time Spent in Military Decision Making on the
Degree of Risk Taking: An Analysis Via an Experimental Simula-
tion Technique," Purdue University Technical Report No. 15,
Lafayette, Indiana, September 1968.
Streufert, Siegfried and Taylor, Eugene A. "Objective Risk Levels
and Subjective Risk Perception," Purdue Univeristy Technical Report
No. 40, Lafayette, Indiana, August 1971.
Swelm, Ralph 0. "Utility Theory Insights into Risk Taking, Harvard
Business Review, November-December 1966, pp. 123-136.
-------
209
Terrill, James G. "Cost-Benefit Estimates for the Major Sources of
Radiation Exposure," American Journal of Public Health, Vol. 62,
No. 7, July 1972, pp. 1008-1013.
Torrance, E. P. and Ziller, R. C. "Risk and Life Experience: Develop-
ment of a Scale for Measuring Risk-Taking Tendencies," Research
Report AFPTRC-TN-57-23, ASTIA Document No. 098926, Randolph Air
Force Base, Texas, Air Force Personnel and Training Center,
pp. 5-7 (as quoted in Denenberg, p. 61).
Van der Meer, H. D. "Decision Making: The Influence of Probability
Preference, Variance Preferences, and Expected Value on Strategy
in Gambling," Act a Psychologica, Vol. 21, 1963, pp. 231-259.
von Neumann, John and Morgenstern, Osker. Theory of Games and Economic
Behavior, March 1953.
Weinberg, Alvin M. "Social Institutions and Nuclear Energy," Science,
Vol. 177, No. 4043, July 7, 1972, pp. 27-34.
Wilson, R. "Tax the Integrated Pollution Exposure," Science, Vol. 178,
October 1972.
Zeckhauser, Richard. "Risk Spreading and Distribution," in Redis-
tribution through Public Choice, edited by Heckman and Peterson.
-------
-------
-------
Protection Agenc*
us Environm r
Street,^
60604-
-------
-------
-------