ML20086R869
| ML20086R869 | |
| Person / Time | |
|---|---|
| Site: | Shoreham File:Long Island Lighting Company icon.png |
| Issue date: | 02/27/1984 |
| From: | Mileti D, Sorenson J SUFFOLK COUNTY, NY |
| To: | |
| References | |
| ISSUANCES-OL-3, NUDOCS 8403010244 | |
| Download: ML20086R869 (38) | |
Text
_
geg@iO I LILCO, February 27, 1984 DCLKETED USNkr UNITED STATES OF AMES!C6EB 29 A10:44 NUCLEAR REGULATORY COMMISSION
- LFF'CE OF 'E!4 ??.
00ChEims & SEiWU.
MANCH Before the Atomic Safety and Licensing Board In the Matter of
)
)
LONG-ISLAND LIGHTING COMPANY
)
Docket No. 50-322-OL-3
)
(Emergency Planning
-(Shoreham' Nuclear Power Station, )
Proceeding)
Unit 1)
)
SURREBUTTAL TESTIMONY OF DENNIS S. MILETI AND JOHN H. SORENSEN ON CONTENTION 23 (SHADOW PHEMONENON)
PURFOSE This testimony replies to rebuttal testimony filed by Suffolk County' witnesses Dr. Cole and Dr. Tyree on February 1, 1984.- At issue is the interpretation of a path model' analysis of:Three Mile Island telephone survey data performed by LILCO's i
- witness Sorensen.
The basic point Drs. Cole and Tyree argue is that the variable THREAT (which measured the response to the question j.
.how dangerous the TMI~ situation seemed at the time) and EMITB (the response to question about how concerned the respondent was before the accident about radioactive emissions) both mea-
.sure the same thing, fear of radiation.
Likewise, Cole and Tyree argue that the variable. CONCERNS (a measure of concern
-over such things as the economic, social, and climatic disadvantages'of the power station) measures the same thing.
94030iO244 040227
_ {_ -. -(DRADOCK 05000322 P
gu J
4
, In fact, the correlation coefficient between EMITB and THREAT _is only
.28, between EMITB and CONCERNS only
.20.
-Moreover, the correlation between EMITB and evacuation is
.11, which illustrates that preemergency fears at TM1 accounted for at best only 1 per cent of the variation in evacuation behav-ior.
Also, the-beta weight of the relationship between EMITB (the one variable that does attempt to measure preemergency fear of radiation) and evacuation is given by Drs. Cole and Tyree as
.000.
Perhaps more important, the Cole-Tyree testimony makes inappropriate use of the path model coefficients.
Because the TMI data did not include measures of many of the determinants of behavior in emergencies, one simply cannot draw conclusions about the relative strengths of the relationships in the path model. -Yet this is precisely what Cole and Tyree try to do.
Finally, the Cole-Tyree testimony places far too much
' emphasis on the path model.
Ignoring all the other empirical I
evidence' about the determinants of behavior in emergencies, Cole and Tyree claim, incredibly, th'at the TMI path model is l
all the evidence there is.
In fact, the LILCO witnesses' opin-
-ions are based on a large body of empirical evidence.
The sig-nificance of the TMI path model is only that, despite poor or nonexistent data on some of the determinants of behavior at t
TMI,.the path model is consistent with the rest of the empiri-cal-literature on emergency response.
L i
r
~,.. - - -,. - - -.... _ -..
o.a-LILCO, February 27, 1984
.i I
UNITED STATES OF AMERICA NUCLEAR REGULATORY COMMISSION Before the Atomic Safety and Licensing Board In'the Matter of
)
)
LONG' ISLAND LIGHTING COMPANY
)
Docket No. 50-322-OL-3
). (Emergency Planning (Shoreham Nuclear Power Station, )
Proceeding)
-Unit-1)
)
SURREBUTTAL TESTIMONY OF DENNIS S. MILETI AND. JOHN H. SORENEN ON CONTENTION 23 (SHADOW PHENOMENON)
Qualifications u
1.
Q.
~Please' state your names.
A.
[Mileti)' My name is Dennis S. Mileti; my address l
is Department of Sociology, Colorado State Univer-sity, Fort Collins, Colorado 80523.
[Sorensen]
My name is John H. Sorensen; my address is Oak Ridge Laboratory,. Box X, Building 4500 North, Room H-11c, Oak Ridge, Tennessee 37830 2.
Q.
Please state your professional qualifications.
l L
=
w e-w,
.,m,-eyym,
..,..,-,,---.-yw--
.,,+v w...,,
,-wire-
__..---.,......-..-w_.z.
-~..m
t 4,
1 A.
[Mileti]
I am an Associate Professor of Sociology and Director of the Hazards Assessment Laboratory at Colorado State University.
A full statement of my professional qualifications was attached to my prior written testimony on contention 23.
I would like.to add that I teach a Ph.D seminar at Colorado State University on path analysis and multiple re-gression.
I have published some dozen or so works that have used this technique, and, to the best of my knowledge, I was the first sociologist to use this technique (in the early 1970's) to study how the public behaves in response to emergency infor-mation.
[Sorensen)
I am a staff member of the Energy Divi-sion at Oak Ridge National Laboratory.
For a com-plate summary of my qualifications, pleads see our testimony on Contention.23.
3.
Q.
What is the purpose of this surrebuttal testimony?
l:
I t
A.
We are replying to the Rebuttal-Testimony of Stephen Cole and Andrea Tyree on Behalf of Suffolk i
' County'Regarding. Contention 23 (Evacuation Shadow Phenomenon), which was served February 1, 1984.
We will refer to this throughout as the " Cole-Tyree testimony."
,m e
r
--g y: ve g -g cgv wirw++ww
<-w+--.-
y-yw--
9-t-w y
-w,s
-y.ya_--y-s-%y,wweq
,y vvg y---r--.s.wg-,-p py,-,-y n y,yo, 9y99 y,y sve + - - *
'4.
Q.
The Cole-Tyree testimony addresses a paper by Sorensen and Richardson, doesn't it?
A.
-Yes.
The paper, by John Sorensen and Brad Richardson, is entitled " Evacuation Behavior at TMI:
Review and Examination" (ff. Tr. 1839).
It has been entered into evidence as Suffolk County Exhibit EP-3 (Tr. 1864).
Drs. Cole and Tyree refer to this as "the TMI paper"; we will refer to it as
,the "Sorensen and Richardson paper."
5.
Q.
And this paper is based on data collected by Cynthia B. Flynn?
A.
Yes.
The report of Flynn's work is NUREG/CR-1093, Three Mile Island Telephone Survey: Preliminary Report on Procedures and Findings (October 1979).
Misrepresentation of LILCO's Testimony 6.
Q.
Do you feel the Cole-Tyree testimony accurately l
characterizes the LILCO testimony on Contention 23?
A.
No.
The LILCO testimony. reveals our position on i
the. determinants of public response in an l
emergency -- in this case, evacuation.
This posi-tion is that response is determined by two general i
factors.
The first factor can be labeled emergency information.
It can in turn be broken down into about 10 different sub-factors, which we called
~4-
" sender determinants" on page 26 of our original testimony on Contention 23.
The second of these two general factors is per-ceived risk at the time of the emergency (situa-tional, risk perception).
This factor is shaped or determined to a large extent by the various emer-gency information factors.
In addition, other fac-tors, which we have called " receiver determinants,"
shape situational risk perceptions.
Among these receiver determinants we include pre-emergency at-titudes and perceptions.
This position is based on a large number of em-pirical studies of how people behave in emergencies under various conditions.
These include studies of evacuation behavior at TMI and in other natural and technological emergencies.
The primary conclusion we have reached, based upon the overwhelming weight L
of the research evidence, is that emergency infor-mation is the key factor because it directly af-facts evacuation response as well as shaping situa-l tionally perceived risk.
Given this knowledge, the Sorensen and Richardson paper merely examined the only data available on evacuation behavior at TMI that are l
m_
...m.
t s.
-5 t,.
uitable for analysis to. determine if this pattern held at TMI.
It'did.
Furthermore, it did despite 0
extremely poor and inte lete measures of the fac-
.i.
tors' we-thi nk expla' er determine < response.
Therefore, our-t'estimony treats it accordingly.
It
.,)
is one more_ empirical study that confirms what we d J already knew about' behavior in emergencies.
Importance'of Situa[ional. Perception of Risk' 7.
Q.
Does.the Cole-Tyree testimony--appear to agree with you on any fundamental principles?
7 Th$, Cole-Tyree' testimony appears to accept that A.
s
'/
-situational perception of risk (" THREAT" in the-Sorensen-Richardson model),/ determines behavior in w
t an omargency (in th'is case' evacuation or " EVAC" in.
the model)..-
Apparently' accepting the effect of THREAT on EVAC,.the Cole-fyree testimony devotes l-its.attenEion to showing that pre-emergency-fear of p
L radiation (EMITB) determines situational risk,per-L ception (THREAT) or that they are both. measures of the sameIthing (fear of radiation).
L Miile _we take ; issue with ' the Cole-Tyree view of the effect of pre emergency fears on situational risk perception, we ' agree with their appare".t rodceptance of.the' significance ~of the effect of situational E
' ed -s W
4_
4
(
~
6
-L?
.3
+
g S
.7 risk perception on evacuation behavior.
We believe this acceptance is a significant concession by the County witnesses.
1 8.
Q. --
What indications are there that the Cole-Tyree tes-timony-accepts the-importance of situational risk perception?
.A.
Thoughout the testimony are statements that suggest the importance of situational " fear of radiation."
For example, on page 3 the testimony says this:
[T]he County witnesses have argued-
. that the primary reason why so-many-people left'their homes at TMI was their fear'of being exposed to nucle-ar radiation.
This statement appears to refer to situational per-ception of risk. - By the.same token, in a flash flood people fear.getting-caught by flood waters and drowned; thet.is why they evacuate.
g Likewise, elsewhere in the testimony (pages 5, 10, 19, :21, 22, and 23) Cole and Tyree talk about-
" fear,"'" fear'of radiation," or " fear of emissions"
- influencing. behavior.. Again,_the Cole-Tyree posi--
tion appears'to be that.(1) fear of radiation at the time of an accident (situational perception of risk)'is'important in determining behavior and (2)
- pre-emergency fear is overwhelmingly important in s
w c
wew.
,4
+-,,,4
,,,._e w
I
..,,.--..---.-,..%w----
,vw.-,-,,,
w-w.-w,.-
w.-e,-,..n.-..-.
w - -. -. ~ ~
determining fear at the time of the accident.
We agree with the first of these points and disagree with the second.
9.
Q.
Have LILCO and its consultants misinterpreted the Sorensen and Richardson paper, as the Cole-Tyree testimony claims?
A.
No.. One of the major "new analyses" conducted for the Cole-Tyree testimony concludes that " pre-emergency fear" in the public had no direct effect on evacuation.
This is exactly what is concluded in the Sorensen and Richardson paper, and what is claimed in the LILCO prefiled testimony on Conten-tion 23 based on generic knowledge and the Sorensen and Richardson paper.
The claim is made in the Cole-Tyree rebuttal testimony that the LILCO position has been that '
" pre-emergency fear" has no caucal effect upon evacuation.
This is a gross misinterpretation of the LILCO-position as represented on page 61 of
'LILCO's prefiled testimony:
The Sorensen and Richardson analysis demonstrates that only 3 of these factors had direct effects on evacua-tion:
perceived threat (4), atti-tudes towards information (3) and satisfaction'with information (2) were all significant (at the.01 level) causes-of evacuation.
Pre-accident fear of radiation was not a
l
~
us
= ',
significant cause of evacuation.
It was, however, a significant cause of the other three accident-specific factors..= Consequently, while pre-accident fear helps shape how people use information and perceive the
" threat of an accident, it does not directly cause evacuation.
Testimony of Matthew C. Cordaro et al.
on Contentions 23 (Shadow Phenomenon) and 65.C.2 and 65.F '(Panicked Drivers) 61 (Nov. 18, 1983).
This passage states that x affects y, that y affects z,.and that x does not directly affect z.
The Cole and Tyree "re-analysis" of the Sorensen and Richardson paper, the Sorensen and Richardson paper itself, and the LILCO'prefiled testimony are in agreement about the character of direct versus indirect effects.
There is no disagreement -- as the Cole 1and Tyree rebuttal testimony would lead a reader to believe -- about indirect versus direct effects, and indirect effects were neither "discov-ered" by the Cole-Tyree testimony analysis nor ig-nored by the original prefiled LILCO testimony.
Indirect effects, it is alleged in the Cole-
.Tyree testimony, may be more important in causing evacuation'than intervening direct effects.
Cole and Tyree also claim that this is the case in the I
p
Sorensen=and Richardson paper.
The implication is
/that the LILCO prefiled testimony misinterpreted
'some' statistical " fact" in the Sorensen and Richardson paper.
However, on close inspection of
-the' basis that' Cole and Tyree provide for this con-clusion, it=is readily apparent th t the Cols-Tyree a
testimony hka misinterpreted the Sorensen and 4
' Richardson paper, and also misinterpreted data used in their'" reanalysis" in such a way as to lead to
-their wrong conclusions.
10.
12 Mhat did Cole'and Tyree do-that leads to wrong con-clusions?
A.
What Cole and Tyree have done in their " reanalysis"
- knd rebuttaE testimony is to take the Sorensen and Richardson measure of " situational risk perception"
-and redefine it?to be " pre-emergency fear."
They are then'in a position to claim that the data sup-port the County _ position versus the LILCO position and,cthen, to accuse 'the LILCO consultants of mis-
~
representing.the'Sorensen and Richardson paper.
Ell..
Q.
- How precisely have' they tried to transmute situa-
}
tional risk perception into pre-emergency fear?
A.
They-claim that-EMITB-(pre-accident concerns about
-radiation in the TMI population) and THREAT L
I
% m
,w~-+.*-v w
~,-+-+-vw-v
-~~w--wwv~
w---wr-
- eev+-'s'w--m-----+-sv---e-~w m-----S-+-----~~-mm+--~-
--r
---e,-n----~v=m-w+=,n=~==v+='--- - ~ = -~
s
, i (situational risk perceptions) actually measure the same variable -- fear of radiation:
[I]t is probable that both the ques-tion used to measure pre-accident fear and the question used to measure fear at the time of the accident are two different indicators or measures of the same variable:
fear of radia-tion from nuclear power plants.
Cole-Tyree testimony 11-12.
A more logical interpretation of their (Sorensen's and Richardson's]
data would view all three questions 4
(pre-accident fear (EMITB), CONCERN, and THREAT) as different indicators of the same variable:
fear of nucle-ar radiation.
Id. 14-15 (emphasis in original).
However, the correlation coefficient between EMITB and THREAT in the Sorensen and Richardson paper was
.28; this means that the one variable explains only 8% of the variance in the other (leaving 92% unexplained).
This low correlation is not evidence of the two variables " measuring the same variable"; in fact, it is evidence that they do not measure the same variable.
Cole and Tyree then go even further in their effort to redefine situational risk perception into pre-emergency fear.
They claim that if x and y
4 i
11-
..c e
- (the independent variables) both measure the-same
, hing, then any correlation between x and z (the t
dependent variable) will disappear if y is con-
-trolled for in.a statistical analysis:
If-[EMITB and THREAT both measure the same variable), then it should not be
- surprising that when'we control for one measure of'the variable,.the other measure will not'be signifi-cantly : correlated -with the dependent variable.
Cole-Tyreeltestimonyfl2 (footnote omitted).
They
' fail-to point out, however,Ethat if x affects y and i.
L ytaffects z but x has-no direct effect on z, then b,
one would expect a correlation between.x and z to Ldisappear when y.is controlled for. - That is, their argument shows that-the theoretical logic of the LILCO position ~ holds in the Sorensen and Richardson H
data on TMI.
I' In addition, tdue correlation between pre-emergencyTfearsE(EMITB) and evacuation in the Sorensen and Richardson paper-is
.11.
This sug-
. gests that, in fact, the Sorensen and Richardson.
p<
' data on zero-order correlations 1/ show that aut
- 1/
A zero-order correlation measures how two variables corre-s late. statistically, notLtaking any;other variables into ac-
= count..
D""FP M
e t-tm-grse wt ty-v-m-1 sp M m ur e-t
-*qe-'
r p-d y g eMwN p.mw t s-
- rw9 w*ggy--7rFPWW*T"tW*'NW9'M9'WMM'PY'Wh9-'1'*"W W-9' frw *" ("' M' M F TTf 9NW rir
12
^
pre-emergency concern about radiation increased in
. the population at TMI, the odds of evacuating actually decreased.
This is not, however, evidence that fear lod to an " anti-shadow."
It is important to-remember that the path analysis revealed that there is no' direct.effect between these two vari-ables, contrary'to what the zero-order cerrelation would lead some to believe.
Nevertheless, the zero-order correlation does illustrate that pre-emergency fears accounted for at best only 1% of the variation in evacuation.
This is a very low effect.regardless of what intervening variables-are introduced into.the analysis.
12. -
Q.
Why is it wrong to. interpret THREAT as a measure of pre-emergency fear?~
.A.
. It is important to look at the actual question used 1
- by Cynthia Flynn to measure situational risk per-ceptions (which is what Cole and Tyree would have us take as a measure of pre-emergency fear):
I would like to ask you some ques-tions about the accident at the Three Mile Island Nuclear Station that deal specifically with-the two-week emergency period immediately after the accident on March 28.
- 21. How serious a threat did you feel the Three Mile Island Nuclear Station was for you and your fam-ily's safety at the time?
y
,;y4
.._.n
,_c~,
,,,4
,,,,,.-,..,,_,,,_,,,,,,,,,-n,..
,.,_,,-_.,-,,n_g_,_,,.
K M:
I C.B. Flynn, Three Mile Island Telephone Survey A8 (October 1979) (emphasis added).
To redefine this question-as_a measure of " pre-accident fear" and then to use it to illustrate that pre-emergency
= fears have a big effect is not a reanalysis but rather a misrepresentation -- of'the Sorensen and Richardson paper, of the Flynn data on which it is
-based, and of the-prefiled LILCO testimony.
13._
12 Is it true that indirect effects can be greater than direct-effects?
'A.
~It is theoretically possible in the world of sta-tistical models for indirect _ effects to be greater than direct-effects; however, because this is theo-retically possible in a statistical sense does not-make it the case in a study (like the Sorensen and Richardson. paper) where it is in fact'not the case.
It has always been the position of LILCO and its-consultants that public response in~an emergen-cy results from the public's situational percep-tions of risk,'which are largely determined by the emergency infort
- ion that the public receives dur-ing the emergency, and that emergency.information
- also has n' direct effect on response in addition to its effect through situational risk perceptiot..
v Other factors affect situational risk perceptions as well I.see the " receiver determinants" section of the LILCO prefiled testimony on Contention 23), but emergency information is the prime determinant of cublic response both directly and indirectly through situational risk perceptions.
Cole and Tyree, at page 8 in their rebuttal testimony, pro-vide a " reanalysis" -(see regression equations 2 through 6) of the Sorensen and Richardson analysis that has the purpoae of determining which factors (emergency information versus pre-accident fear)
L had the greatest effect cn1 situational risk percep-tions.
However, in conducting this analysis they
'misr'epresent the Sorensen and Richardson paper in
-several very significant ways.
l I
Misrepresentation of the Sorensen and Richardson Paper i
14.
Q.
How has the Cole-Tyree testimony micrepresented the Sorensen and Richardson paper?
A.
First, the Cole-Tyree testimony inappropriately draws conclusior.s about the strengths of the rela-tionships in the path model.
The Sorensen and Pichardson paper acknowledges the limitations of the data its authors use in the path model.
Ac-
.cordingly, they state:
^
g y
y
-ma
,y--q-m
-g
,-e-rw,--
7---,
,w,
o
'Here we are placing several restric-tions on the model and more closely following the philosophy of path analysis as a method for substantive
-theory-testing rather than examining data for predictive relationships.
-Sorensen and Richardson paper (Suffolk County Ex.
EP-3) 13..(citations omitted).
This statement means that in the context of the pape,r it is totally in-appropriate to draw any conclusions about the rela-tive strengths of-the relationships in the path model.
By doing so, the Cole-Tyree testimony mis-interprets.the purpose and conclusions of the Sorensen and Richardson paper.
The reason it is inappropriate to draw conclu-sions'about the relative strengths of the relation-ships lies at the heart of the method of path mod-eling.
TheEtechnique of path analysis requires that the model_being used be free of " specification
[
error."" This means that no relevant variables have been excluded from the analysis and no irrelevant ones included.
Yet, given existing knowledge and theory, it is clear that the path model used in the Sorensen and Richardson paper has specification error:
tha information-factors that are so impor-
' tant were excluded. : Conclusions about the relative strengths of beta coefficients are therefore likely I
1[?
,1 wrong, because these coefficients are certainly bi-ased because of specification error.
The Sorensen-and Richardson paper explicitly recognizes that emergency information-related fac-tors that we know explain emergency response (the ten sub-factors listed as " sender determinants" in the LILCO prefiled testimony) were not measured in the Flynn survey:
The addition of warning-specific con-structs to the analysis would likely explain additional variance not tapped by our theory.
Suffolk County Exhibit EP-3 at 24.
Thus, there is no misrepresentation in the LILCO testimony of the 1
conclusions of the Sorensen and Richardson paper-with respect to the importance it places on emer-gency information.
Second, the Cole-Tyree testimony commits a se-rious error in analyzing the relative importance of.
i the effects each indicator had on THREAT or situa-tionally perceived risk..In order to inductively rank relative importance, as is done on page 13.of the Cole-Tyree testimony, it is proper to use an inductive regression approach.
Instead, it appears as though Cole and Tyree used a pre-set structural
~_. - __.
~
equations model in which the sequence of entering the variables is fixed.
This assumes a theoreti-cally determined and grounded causal sequence in
. order for the-beta weights to be unbiased and trust-worthy.
The analysis in the rebuttal pres-ents no rationale for the sequence and is, indeed, a wrong one.
It is not the same as that used in the Sorensen-Richardson paper.
It is therefore
-invalid-to draw conclusions about the relative strengths'of'the relationships between the indica-stors and perceived risk'on the-basis of the analy-sis they performed.
Again, thic is a misrepresen-tation of the Sorensen and Richardson paper, in which such an error was not made.
Third, on page 14 of the Cole-Tyree rebuttal the variable CONCERNS is described as a measure of
" concern with radiation."
This is not supported by either a description of the variable in the Sorensen and Rich 3rdson paper or the question (s) on which the indicator is based.
As the Sorensen and Richardson paper says (on page 10), CONCERNS is "a measure of concern with other issues (such as eco-nomic, social, and climatic disadvantages of the area), relative-to the perceived disadvantages of the location of the power station."
Once again,
.the Cole-Tyree rebuttal testimony misleadingly de-
-picts'the indicators used in the Sorensen and Richardson paper.
15.
Q.
The Cole-Tyree testimony suggests that three indi-cators used in the path model, "EMITB," " CONCERNS,"
and " THREAT," all measure " fear of radiation" be-cause they were collected at the same time.
In fact, all of these questions were asked at the same time after the ac-cident at TMI occurred.
Sorensen and Richardson have no way of separating the attitudes of respondents towards radiation: from nuclear plants before the accident from the attitudes ex-isting at the time of the accident, or from those attitudes existing after the accident.
Cole-Tyree testimony 14.
Is this right?
A.
No.
To gain a better idea of what these indicators measure, we-can look at the questions used-to col-lect data on them.
A review of Table 1 (page 11)
- in-the Sorensen and Richardson paper shows that only one of these indicators (EMITB) is about pre-L emergency " fear of radiation."
Specifically the i
question used was:
p There has been some talk about possi-ble radioactive emissions or releases from nuclear plants.
Before the ac-l:
cident at Three Mile Island, how con-cerned were you with radioactive emissions from the plant.
l
~.
3 Flynn at A18 (question 71) (emphasis in original).
This is quite-specific both as to concern with ra-dioactivity and as to the time period in question.
The.other two indicators, CONCERNS and THREAT, do not concern emissions or radioactivity, nor do they specify a tima pariod before the accident.
16.
.Q.
But.they might measure the same thing, mightn't they, even though there is nothing to suggest that they do?
A.
Let us assume that all three variables measure the
- same thing, namely pre-emergency fear of radiation.
If this were the case, then we would expect to find
.a high. correlation among the three variables; if they measured the same thing, this would be re-
. flected in the answers people.gave to the ques-tions.
However, from observ$ng the correlation coeffi-cients in Table 2.(page 15) of the sorensen an'd Richardson paper, we find relatively low zero-order correlations.
The correlation coefficient between
' EMITB and THREAT is
.28, which means that only about 8% of the variance in one variable is associ-
- ated with the other.
The correlation between EMITE and CONCERNS is
.20 -- about 4% of the variance.
This provides solid evidence that CONCERNS and
...,m,-
..,,--,-,m-------c--.-.
-w
i h
R a -
THREAT do not measure fear of radiation, as is n
claimed in the Cole-Tyree testimony.
The Tabula Rasa 17.
Q.
The Cole-Tyree rebuttal testimony represents that
'4 LILCO and its consultants have attempted to con-g vince the Board that the public vill be a tabula rasa in an einergency and that the public would not evacuate until advised to do so.
Is this a correct characterization of-the LILCO prefiled testimony?
90iii A.
No.
The prefiled LILCO testimony at pages 36-41 catalogues a considerable number of " receiver de-
^
terminants" that, to use the language of Cole and Tyree, profile the "non-tabula rasa" character of f.
1 people involved in emergency response.
In addi-4, 3
tion, page 119 of the prefiled LILCO testimony states "It has been documented in 'many cases' that j
evacuation starts or occurs without official recom-i mendations."
It was also reported, at page 61 of
]L "si
' hat " pre-accident y
the prefiled LILCO testimony, t
4 fear helps shape how people ure information and perceive the threat of an accident" -- that is, that pre-emergency fear does affect situational risk perception.
In r'edirect examination on Con-Ij.
tention 23, Mileti said that he would be surprised V
u if, with good public emergency information, "over-y 1"
response" would be as high as 25%.
2
'I l
l
- J l 2
t:
e w
Thus, it is misleading and unfair for Cole and Tyree-to suggest that LILCO and its consultants have asked the Board to believe that a public is a tabula rasa_.-
What LILCO and its consultants have provided the Board is a summary of knowledge about the determinants of public. response in emergencies.
This detailed knowledge rests on dozens of empiri-cal sociological and social psychological studies of public response in actual emergencies.
The bot-tom line of these studies is knowledge that can be
-used to help a public respond-in An emergency re-
.gardless of and taking into account the non-tabula
>rasa character of-the human' animal.
That is, given the known effect'of' emergency information (sender determinants), the non-tabula-rasa character of
-people (receiver determinants), and the role of situational risk perceptions, it is (to quote the LILCO prefiled testimony) "possible.to use a warn-ing system to create accurate risk perceptions through good warning information principles, and, therefore, decrease or avoid over-or under-reaction" (p. 61).
LILCO and its consultants have always held that this is possible in spite of the non-tabula rasa character.of people, and not be-cause' people are a " tabula rasa" as Cole and Tyree represent in their testimony.
,,,,w.m-,w,--
_,--r---
m -
Deep-seated fear
[>
18.
Q.
Cole and Tyree state in their rebuttal testimony at page 16 the following:
[T]he primary cause of over-response at TMI, as demonstrated by Sorensen's and Richardson's data,.was the deep-seated fear that residents have of radioactive emissions from the plant during an accident.
In your opinion, is this interpretation of the data correct?
A.
This is possibly the most serious misrepresentation of the data in the Sorensen and Richardson paper
.that can be made.
First, this conclusion is based on the rebuttal testimony's redefinition of indica-tors used in the path analysis.
Rather than use a meaning attached to the indicators based on the original questions used in the Flynn survey, the County has given them new meanings, which fail to conform to either the questions used by Flynn to elicit the information or the descriptions used in the Sorensen and Richardson paper.
Second, in the Cole-Tyree testimony the rela-tionship between the one-variable that does in fact measure fear of radiation (EMITB) and evacuation (EVAC) has a beta weight of
.000.
Zero. simply
^
- s
-does.not' demonstrate a " primary cause" in any dis-cipline that uses statistical analysis.
The Sorensen and Richardson Paper in Perspective 19.=
. Q.'
Is the.Sorensen.and Richardson paper the principal basis.for your_ views about-the role of emergency information?
- A.
No.
The' Cole-Tyree testimony incorrectly claims that the Sorensen and Richardson paper is the only
" empirical" or " quantitative" evidence in support of LILCO's position on. Contention 23 (Cole-Tyree
. testimony 5, 21).
In fact, the LILCO position 3
rests on the knowledge gained from findings from dozens of-studies of public response in actual
' emergencies and.the knowledge that comes from tak-
'ing all of them together.
Most of these studies
- did contain quantitative data.
It is true that the prefiled LILCO testimony did not recite the individual data from these y
i<
sociological studies;-instead, the knowledge gained d'
.from these' studies was presented.
At the same
- time, tdue profiled LILCO testimony sought to put
.the Sorensen and Richardson paper, as well as the self-reports about reasons for evacuation at TMI in fthe Flynn data, in an appropriate perspective 2
g
(
- v
1 vis-a-vis the generic knowledge summarized in the
..c bulk of'the prefiled testimony resting on other em-pirical and scientific studies.
The Sorensen and Richardson paper and the Flynn
.self-reported reasons for evacuation support the generic knowledge gained from dozens of empirical stadies over the last several decades; they are not a tast of that kaowledge -- nor are they claimed to be in either the LILCO prefiled testimony or the Sorensen and Richardson paper.
It is~ misleading for-the Cole and Tyree rebuttal testimony to single out one empirical work used by LILCO and its con-
- sultants while ignoring the dozens of other empiri-cal works on which the prefiled LILCO' testimony rests.
The truth of the matter'is that the Sorensen and Richardson paper is the only study used-by LILCO that Drs. Cole and Tyree appear to have exam-q.
- ined in' sufficient' detail to be able to determine.
- whether it-is quantitative or not.
Actually, many of the~ studies referenced or summarized in the LILCO testimony.are quantitative.
Indeed, the LILCO testimony reported, principally, the knowl-edge derived from these studies.
The evidence L
lu i
I
^
.m
f f
brought to bear from the TMI paper was also the
-knowledge produced by that paper.
In no case did the testimony pretend that any data or numbers from that' paper were'important to the testimony.
What the LILCO testimony;found important was that the paper's findings were consistent with the other studies on which the testimony primarily rests.
To impute precise relative importance to the path co-efficients misrepresents the nature and purpose of
'the Sorensen and Richardson paper, which explicitly states that.the path analysis is used for theory building and not for predicting precise relation-ships.among variables.
Effects of the Governor's Advisory R20.
-Q.
The Cole-Tyree testimony discusses the effect of having_ preschool children.
Do you agree with this discussion?
A.
The Cole-Tyree rebuttal claims to have conducted an analysis of the effect of the content of the evacu-ation advisory at TMI on perce.ved risk and evacua-d
. tion (p. 17).
Drs. Cole and Tyree suggest that the indicator " KIDS," which actually measures the pres-ence of children' age five and under in a family, instead measures _the receipt of the Governor's ad-visory.
They also assume that this is the only
~ __
l-information that this group received that could in-fluence evacuation and risk perceptions.
While we disagree with both the claim that KIDS
-provides a measure of receiving the advisory and the assumption that it was the only information re-ceived, we will, for the sake of clarifying the re-sults of the path analysis, assume it measures receiving the advisory.
Given that assumption, our testimony on Contention 23 would lead us to hypoth-esize that KIDS could have a direct effect on evac-uation, an indirect effect through perceived risk (THREAT) on evacuation, or both.
Furthermore,'we would hypothesize that the nature of the effects would be primarily' determined by the content of the warning information.
The effects would be through perceived risk if the message contained information that-attempted to arouse fear.
The effects would be direct if the information was primarily instruc-
'tional.
The advisory was instructional in that it told people who met certain conditions what to do.
It did not use fear arousal.
This is verified by the path analysis.
KIDS had a direct effect on evacua-tion and did not alter risk perceptiens.
This
' ' demonstrates the advisory was effective in getting people who-met the conditions of the advisory to evacuate withoutLcreating undue alarm and fear.
This is consistent with the goals of a warning sys-p_
tem andl supports'the LILCO testimony on Contention
~
23.
e 21.
Q.
The Cole-Tyree rebuttal testimony (see page 19) claims.that the causal model presented in Figure 3 on page 14 of the Sorensen and Richardson paper as-sumes that~ emergency.information variables are a prior cause of situational risk perception (or per-a ceived threat), but that there is no evidence to support =this assumption of causal order, and that it'is equally likely that risk. perception was caus-ally prior to emergency information.
Do you agree?
A.
The Sorensen'and Richardson model correctly orders variables in terms of cause and effect.
This order
.is obvious and-familiar to those acquainted with
- the research area and studies of public-response in
- emergencies.
Preemergency fears come before emer-gency information (that'is, pre-emergency' factors precede emergency factors), and.these are followed by situational ~ perceptions of risk, assuming that situational-emergency risk perceptions do not emerge until after-someone learns'that an emergency is occurring.
This causal ordering not only rests on estab-lished and validated theory, it also conforms to a
. - =...-..
28-commonsense: time sequence that is essential to es-tablish causality in the social sciences; this lat-ter point is well addressed, for example, in Professor Stephen Cole's textbook on research meth-ods.
It is true that the operationalized measures for the: theoretical concepts / constructs in the
- Sorensen and Richardson paper were limited in quality -- Sorensen and Richardson had to work with i
data collected by Cynthia Flynn that was gathered for other research purposes and passed along to them second-hand.
The Sorensen and Richardson u
paper-does suffer from the use of poor measures for f
some of the theoretical'cancepts in the path model.
This-is a common flaw'in most sociological.research and ' explains why confidence in the firidings from.a e
single piece of research should not be high unless l
L
- those. findings are viewed in terms of all knowledge in a particular sociological subfield.
This', by the.way,-is exactly how the Sorensen and Richardson paper was used in the LILCO profiled testimony.
It
- is quite inappropriate to divorce this one paper from. generic knowledge and validated theory in the
- field,.as the Cole-Tyree testimony would have us S
do.
Y e
N w-* - - -Mv e
- $7 er_7,a
%gg,.
,, -.n
.,wp, g-w,,,-sw,g-,-_,w,
,_,,,,,,,,p_.w_p+.w.
r w-e-erw-w
- -*-m-
- * ---=-=r-.
J
- x. -
Different Accidents as the Units of' Analysis 22.
Q.
The Cole-Tyree testimony (page 20) sayc that "[i]n order to prove that the quality of information is a significant cause of over-response, it would be necessary to take different accidents as the unit of analysis, to measure independently the quality j
of information given out at the time of the acci-i dent, and then to measure the size of the over-J response."
What do you think of this?
A.
In a perfect world for sociological research, Cole and Tyree are correct; the research design that they propose would be an excellent one.
- However, the design would require that we wait for some 30 or so cases (enough.to be able to use path analy-sis) of emergencies at nuclear power plants to hap-pen, hope that some had good emergency information and others did not, and use standard measures in all cases of relevant variables.
This will never hsppen.
Nevertheless, the strategy suggested by Cola and Tyree (to look across emergencies for general conclusions) is precisely what was done for the
-.LILCO prefiled testimony.
The results of this work (which by necessity had to be qualitative, since different. researchers measured different variables in their studies,.and the variables measured were often measured differently) are summarized on pages im -
-sr y
w
.ar--
30-21.through 48, and at other places, in the LILCO prefiled testimony on Contention 23.
This research base provides reasonable assurance and confidence that'eme,rgency information can help most people
..come to do the correct thing in emergencies.
It
'does not provide perfect proof, but then proof'of this sort is well outside the abilities of the so-cial sciences.and.beyond the-state-of-the-art.
Exogenous Variables 23.
Q.
On page 21 of the Cole-Tyree testimony the signifi-cance of " exogenous" variables is discussed.
What do you think of this discussion?
P A.
Cole and Tyree provide an incomplete definition of
-exogenous variable.
In a structual equation model,
(
exogenous variables are given and cannot be changed,-as the Cole-Tyree testimony says (p. 21).
l-l But " change" here refers to a change caused by
~
other variables in the model.
Thus, a change in age,.an exogenous variable, is assumed not to ef-fact other exogenous variables defined in the
'model, such as income.
L While it is true that such variables are
.largely fixed and are non-manipulatable, this means at the time of the accident.
Exogenous variables g
y y =
e g-
,--m
,, - - - ~
--~,----y.
.re,--..-y,-..w,..
.,m.r
-e,
31 can change over time.
People age, get richer or poorer, form different attitudes, and so forth.
Thus, while pre-existing fears of. radiation will be fixed at the time of the accident, there is no evi-dence to suggest that they will not change with 1
time.
In the context of the path analysis, fears will not change as.a result of a change in another variable'in the model.
Controlling the Flow of Information
' 24.
Q.
At'pages 21-22 of the Cole-Tyree testimony the fol-lowing passage occurs:
[T]he County experts. bel'ieve that it is impossible for a utility to con-trol'the' flow of information during an accident.
Indeed, it is virtually inevitable that were an accident to
~ occur at~Shoreham, this would become
.the focus ~of widespread media atten-tion and conflicting reports and opinions'would inevitably be broad-cast.
If over-response is determined
_by fenr of radiation and if this fear canno.t be overcome by the type of EBS messages issued by a utility, then
, evacuation shadow is a' serious prob-lem(which must be dealt with in'de-
- veloping an adequ te evacuation plan.
1 Cole-Tyree testimony 21-22.
Do you agree with this?
-A.
Yes.
That is,.we agree that'it is impossible fort a utility-(or anyone:else for that matter) to control O
j n
me
4 all the information that would be available for a public to " hear" in an emergency.
It is inevitable that the emergency would become a media event, and likely that conflicting reports and opinions would be broadcast for public consumption.
However, this likely state of affairs does not
- mean.that the public must be left to the confusion that would result, as they were at TMI.
Emergency plans can be developed with the understanding that good emergency information must be able to be heard, understood, believed, personalized appropri-ately, decided upon, and responded to, despite the occurrence of misinformation, rumor, and conflict-ing information that would characterize the emer-gency.
This can be accomplished if the knowledge
~ that exists in the social sciences about how to ac-Ecomplish it is put to use in planning for public emergency information, as it has been in the LILCO Transition Plan.
We also agree that "if" over-response is deter-mined by fear, and "if" fear cannot be overcome by emergency information, that what has been called "the evacuation shadow" could occur in any emergen-cy.
However, we believe that the assumptions
, __ -. =
a
./
4
,,j s
contained in this set of qualified conclusions of-
.fere'd by Cole and Tyree are unfounded.
Emergency information can be used in such a way as to help most members of an affected population make good decisions and follow through with a good response
,l in an emergency.
This is possible to accomplish
-gi regardless of pre-emergency population characteris-tics like, for example, pre-emergency fear of a di-ysaster agent, apathy, and so on.
The'LILCO pre-
, filed testimony reveals how this goal can be 4
achieved, and the LILCO Transition Plan has incor-
, g pora'ted this knowledge.
F
.. Radiological Nature of Emergency i
25.
Q.
.,,On page 23 the Cole-Tyree rebuttal-testimony implies that Sorensen and Richardson say that if 6
l the emergency is radiological, more people will be likely to evacuate than if the emergency is of a different type.
Do you agree with this?
, ;r '
?
i-
,s A.
The~ precise passage from the Cole-Tyree testimony
~
reads:Tas-follows:
What Sorensen and Richardson are -
saying in.the TMI paper is that the j overall size of the group evacuating will be primarily influenced by the nature of the event.
We agree with this.. If-the emergency is ra-idiological in nature, more people
~ ill be.likely to evacuate than if w
the emergency is of a different type.
f~
'E
,s
.E a
y sr
+
-y
,u
..,--y
9.o
' a Cole-Tyree testimony 23.
This is based on a state-ment in the Sorensen and Richardson paper that the nature of the event influences overall response patterns.
The paper never states, however, as is implied in the Cole-Tyree passage just quoted, that
" radiological" is a relevant characteristic of an event that will influence response levels.
The Sorensen and Richardson paper is correct in not doing so, because no empirical evidence exists that U
will support the claim that the " radiological" na-ture of an emergency determines response.
The Cole-Tyree testiraony proceeds to correctly interpret the implications of the Sorensen and Richardson paper on the same page:
The-more fear with which the event is viewed, the higher the proportion of people who will eyacuate.
Cole-Tyrne testimony 23.
We ag,ree with this state-l ment.
It has~ consistently been our position that situational factors, such as the one illustrated in j
the above relationship, are causes of response.
This is empirically demonstrated, as are the rela-tionships between emergency information and risk perception and emergency information and responee r
Emergency.information can, if managed properly, be r+g-y,
,.e,,
,-.--,nm,m
,1-.r.m--.#
w--
-,w,w.,,
..*,..-..,-,e-,e-,.,--
e
." ~
used to guide human response in a nuclear power plant accident, as well as in other types of emergencies.
The public is nat predisposed to "overrespond" in such a way that it cannot be in-fluenced by emergency information.
Summary 26.
Q.
Will you please summarize your testimony?
A.
The rebuttal testimony of Drs. Cole and Tyree pro-videa further support for our main premise that evacuation reponse is primarily caused by situa-tional factors.
Numerous statements by Cole and Tyree suggest'that response is caused by situa-tional factors, primarily perceived fear of being exposed to harmful radiation during an accident.
j In other respects, however, the Cole-Tyree tes-
'timony misrepre'eents the LILCO position.
It has never been the LILCO position that usefulness of information is the sole cause of response, as im-I plied in the Cole-Tyree testimony.
Rather, it is the LILCO position that emergency information, as characterized by ten important sender determinants, and situationally perceived risk, which is shaped by emergency-information, as well as certain
- . t, receiver determinants, are the chief factors that enable us-to understand response.
In this mix emergency information stands out as the primary determinant because of the overriding strengths of its~ direct and indirect effects on re-sponse.
Behavior in an emergency is situationally determined, although factors such as pre-accident fears, as well as a multitude of other variables, will affect situational perceptions of risk and in-formation.
n v
m._
_-