_r»tec States C*fice o« =cnc> ==a 23C/C8-39-064
L-.vor.rvia -»ctecnon 3amfxj ara Evahjanoo August iSa9
Agency wasmgton IX 20460
Communicating Effectively
About Risk Magnitudes
Risk Communication Series
-------
COMMUNICATING EFFECTIVELY ABOUT RISK MAGNITUDES
Neil D. Weinstein, Peter M. Sandman
and Nancy E. Roberts
Rutgers, The State University of New Jersey
September 12, 1989
The research described in this document has been funded by the United
States Environmental Protection Agency under Cooperative Agreement
CR-814506. It has been subject to the Agency’s peer and administra-
tive review, and it has been approved for publication as an EPA
document. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use. Additional support
from the New Jersey Agncultural Experiment Station as NJAES Project
26502 is gratefully acknowledged.
This document can be ordered from the Environmental Communication
Research Program, Cook College, Rutgers University, P.O. Box 231, New
Brunswick, New Jersey 08903. It is also published by the U.S.
Environmental Protection Agency as document EPA-230-08-89-064.
-------
CONTENTS
EXECUTIVE SUMMARY i
I NTRODUCTION • 1 1
Previous Research . 11
Criteria for Evaluating Success in Risk Communication 12
Selection of Risk Communication Formats 15
Hazard Selection. . . . . . 19
METHOD...... ... ... . ,• •••••• 20
Research Plan . . . . . . . . . . . . . . . . . 20
Sample. . . . . . . . . . . . 21
Materials...... . .... .... .. 21
Procedure •. . . . . . . 24
Measures of Communication Success in the 26
Present Investigation
RESULTS 27
Comprehension 28
Acceptance 31
Consistency . . . 35
Uniformity Across Participants 44
Audience Evaluation. . . . . . . . . ‘ 44
Direction of Communication Errors 46
Other Variables 46
INTERPRETATIONAND CONCLUSIONS . 49
The Value of an Action Standard 49
The Value of Advice If There Is a Standard 50
The Value of Probabilities and Probabilities 50
Plus Comparisons
The Value of Risk Comparisons and Graphical. 51
Presentations
Information Overload. . . . . . . . . . . . . . . . 51
Matching Formats to the Communication Situation 51
Responsiveness to Risk Magnitude Information 51
REFERENCES. ............. ........... ...... .. . 55
APPENDIXA.RISKCOMXUNICATIONFORXATS...... Al
Asbestos Format 1 (with complete brochure) A2
Asbestos Format 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A6
Asbestos Format 3 . . . . . . . . . . . . . . . . . A7
besets Format 4 . . . . . . . . . . . A8
Asbestos Format 5 . . . . . . . . . . . A9
Asbestos Format 6 . . . . . . . . . . . . AlO
Asbesots Format 7 . . . . . . . . . . . . All
-------
TABLES
Table 1 Ingredients in different formats tested . 16
Table 2 Summary of results by format 29
Table Cl Means for key variables by format C2
Table C2 Designated errors for different risk probability.... C4
choices
Table C3 Criteria for acceptance of mitigation advice C6
Table C4 Measures of acceptance for different criterion C8
patterns
Table C5 Differences in acceptance of mitigation advice C9
Table C6 Analysis of consistency in perceived threat Cig
and action plans
FIGURES
Figure 1 Format F7 (example) 18
Figure 2 Error rate in judgments of illness probabilities.... 30
Figure 3 Absolute value of errors in judgments of 30
illness probabilities
Figure 4 Judged probability of eventual illness if no 32
action is taken
Figure 5 Errors in judgments of illness probabilities 32
Figure 6 Measures of acceptance of action advice 33
Figure 7 Disagreements with advice to act (Underreactions)... 34
Figure 8 Disagreements with advice to act (Overreactions).... 34
Figure 9 Highest level at which subject would feel satisfied. 36
Figure 10 Lowest level at which friends still need to act 36
Figure 11 Consistency of perceived threat with actual risk.... 37
Figure l2Meanperceivedthreat................. 39
Figure 13 Effects of formats and assigned levels on 40
perceived threat (log scale)
Figure 14 Consistency of mitigation plans with actual risk.... 41
Figure l5Meanmitigatienplans. 42
Figure 16 Effects of formats and assigned levels on action.... 43
plans (log scale)
Figure 17 Helpfulness for understanding test result 45
Figurs l8Understandingofrisklevels............... 47
Figurm 19 Effects of education on perceived threat and 48
mitigation plans
Figure Cl Effect of format and assigned level on perceived... C13
threat (by test result)
Figure C2 Effect of format and assigned level on mitigation.. C14
plans (by test result)
Figure C3 Effect of format and assigned level on perceived... C15
threat (by rung on ladder)
Figure C4 Effect of format and assigned level on mitigation.. C16
plans (by rung on ladder)
-------
2
Successful communication about this c 1 ass of hazards is
particularly impor .ant because it can increase the likeli. 1 ood
th it penpie take actions to reiuue health r ks when such
actions are appropriate and decrease the likelihood of excessive
worry and unneeded action when risk levels are low. Different
ccmniunicationstratcgies rn.iy be appropriate for h ard thdt do
not permit ii dividuals to assess their own risk and make their
own decisions about mitigation.
Seven formats were evaluated, as follows:
Fl. Ris)c Pr abi1iti . Information about expected
lifetime mortaiity (deaths per thousand people) at
various 1 vels of exposure.
P2. Risk_f oba it es d Q japj . Fl with
comparisons to ,noking risks added.
F3 . raphic_oabi J i . Fl displayed in histogram form.
F4. Standard . Information about the recommended action
level only.
F5. Standard + Ri5k ProbabjUties and Comparisons . F4
F2.
F6. ansjard±_Advice. F4 with dotailed a’ t!on advice cirid
veLba iabels for four ranges of exposure levels.
F7. Standard + Advice 4- Risk P obabilities and C ris
-------
3
discussing the hazard in questior. The first three pages were
constant across conditions; the fourth page consisted of the
experimental manipulation. Each subject ecei ed one
hypothetical reading for either radon or asbestos and one
brochure explaining the hazard and its risk.
Subjects then responded to an evaluation questionnaire.
Issues covered included: subjects’ evaluation of the brochures,
perceived risk seriousness, likelihood of illness, concern,
fear, remediation difficulty, intentions to take action,
acceptable exposure levels, and numericai i]lness probability.
Responses to the risk seriousness, illness likelihood, concern,
and fear questions were combined to form a single measure of
“perceived threat” because they were so high’y correlated with
one another.
The experimental design thus comprised seven presentation
formats, four exposure levels, and two hazards—-a total of 56
cells. The final sample consisted of 1948 subjects, an average
of 35 subjects per cell.
The discussion that follo :s sur marizes the major findings
and their implications.
1. The “glue of an action standard . Formats F4 through F7
included an action standard—-a level (midway between the second
and the thirc exposure levels) below which mitigation was not
recommended, and above which it was. Formats Fl through F3 did
not. An important goal for communication (an aspect of the
“consistency” criterion) is that people with low readings see
their threat as smaller than people with high readings and less
often undertake remedial action. The formats with an action
ta’idard were superior to the formats without an action standard
according to this goal.
The presence of t ndard not surprisingly increased the
likelihood that subjects’ action intentions matched the action
recommendations; in other words, subjects were more likely to
p!an tc remediate at levels above the action standard and to
plan not to remec iate at 1evel below the standard if they were
told what the standard was. Though Obvious, this is not a
trivial finding; it shows that explicit action standards
significantly ffoct individual action plans, and thus can
contribute meaningfully to public health.
However, the presence of a stanlard also created an
artificial dis ’ ontirwity in hazard responses as one goes from
just below th.- standard to just above the standard. This
discontinuity or “step” exaggerated the minimal increase in
actual risk between the second specified i xposure level and the
third. This effect, too, is not surprising; many commentators
have complained about the oublic’s tendency to dichotomize
around a standard, perceiving leves just below the standard as
safe and those just above as risky. The discontinuity found
was, if anything, ma11er than miqht have been expcwted.
-------
4
ihFldicla. Fl
disti. I W afld
Lt ] evO’
sky than 1, st i dä -rd:..
differ enti-ati-on betwe n h igh nd very high
th’ ’ act i’Oñ dá-rd d-i d’ n’ot a féct. the
to hi’ h and a tión o ery hi
2. stañdard- onIv. cènditjQfl. . Ai.though
four formats (F4 th ougñ F7)’ in lude$ añ-âdtiofl z dard, only F-4
rovi éd the standard wtthout any add tfonal infoima;txon-—t1 at
is, w4hout a risk ladder, without riCk robabilit].es or risk
:rnpári ons, and ,i th tit á i àe arid ve b l ‘láb -1s keyed to
a ioüs Thvè] s.
This cox i tion stood out from the others in ma y ways. Most
importà tly, it l d to the highest perçèptions of th-reat, to the
g te t ii t-entions to te édi te (e pecia1iy at levels below th’e
standà•rd), and tO the io é t levels judqe p’ersonally
acbè t ble. it also prdüced the moSt risk averse responses on
several other outcome me suz es. These find4-ngs strongly suggest
that a standard without addThi-ona l risk inf ábi on is usefu l
when an apathetic resp nCe is ‘feared t i - äl is to provoke
rnö e i’sk-a ver-sidn. Where panic and o ér eääti án are likely
on the other hand, a standar -oiil ’ Oommunication
Chbu-ld be avOldéd.
The differences Obtained between format P4 and formats F
t ro h F? are greátér th ñ can be explained by the risk
probá ili’ty infotmatiøn (F5 afid p7), risk ôm arisons (P5 and
F 4, ‘ r d ice (F5 and Fl) th t these others cbnta-in. It seeais
m st Hkely that the uniquely ri k—ave se responses to P4 r uit
from ti fact that P4 lacks the risk l a d der pr,eCeht in 5
tWrough P7. in the absence of a ladder, we believe, subjects
had no -way of telling what test read r1gs were unusual, and they
the’ref e reacted as thou h the f -ind-ii o -f aT y :am ount of radon
‘r asbestos we -re a serious problem. In c ntrást, subjects ‘th -o
ha a ladder and a low test •resu] -t -san that their exposure level
• - 16W g the page and f- lt reassure 1 -. E,Vè if their level was
b e the standard, the presence of still higher ruht s on the
L 1der was reassuring. Thbse with results at the Very top rung
-O thä tadaer fñight ha e become niore 1ttei ed - than subjects
W t -h3 i o la’ddê-r at all ( th:i was not tested). Før ôveryohe else,
‘heye , an exposure la ddCr appears to be reassuring, and the
abséñce of a ladder in fbrmat F4 appe rs - “to. C e increased
ttsk—àve siveness.
3. The:. a ue of.-a’d -iee - . Two fo -rm -ts 6 and P7) went
beyc nd the dic)ioto iy create’ by the sta idard to provide ve ba-l
l Ls and ctthn - advice at diffe ent Levels. Aithough people
t t e: -
-------
5
ifl th is study t -n4ed .to be more ri a rs e than the acti-on
i àéñ A € ióhS (i )iey f en sa id they jou1d i1-tigate levis
bel ów thé t àndara), those receiving action ad icie. shc wed his
tendency l ást. That is, they were the most ii -kely to c pt
the reconnrenc atioi not to take action at low levels, the least
likely t-o “ôvérr act” viC-à-vis the standard. Subj ects also
re thte d less LAn ttainty and better understanding of their risk
when advice was provided-.
It is not su ptising that the presence of mqre graduat’ed ärT d
detail -’ed advxc than a simple standaid increased t ie L.kelihoo
that subja €s would act in ways cori iste,nt with what was
advised. -However, thé effe t was not true for al-i ii s)c levels
STubjects re eivi ng or dats FE and F7 were less risk-averse at
low levels thai s bjects receiving format F 4 (standard-only),
but they were not ñbre risk-averse at iiriqh l eivel-s. rovidihg
explic-i t kdvice istfiu especial3y üsefu-l for panic preventioh,
to deter *er -reaction at low risk levels. Its value for
incr asing r n,é’di -l act-jon at high level (beyond what wou ld be
expe ted with a tandard alone) has not been demonstrated.
4. i?he value p tisk .probabIl-itv da ta r roba j Jties.
plus comDaz is 15 to si cin q . i: k praba 1iity in o matfon as
pr -o rid éd in .ói’ ’ats fl., F2, , F5, and P7; it was acconipar ied
by cotn ari of s to smoking in F2, F5, and P7. Subjec-ts recei iing
these fbrm ts did a better job of est-ifflating illness
probabili ties at their levels than sub-j ts receiving no risk
probability information. This shows ti at the information wa
not totally ignored or totally ±ncompreh flsibie. But the
i-mprovemen .. in illness probability estimation was not matched by
any change in perceived threat or mitigation plans. Analysis ef
F4 versus PS and P6 ersu P7, for exainpl-e, shows that the
addition of risk probability data (plus comparisons) did not
lead to a further differentiation bet zeen h gh and low risks
beyo;.d that produced by the standard alone (P4) or the standard
plus advice (P6).
On some evaluation criteria, risk probability information
did help slightly in enabling subjects to distinç ish the risks
of radon from the risks of asbestos. Subjects receiving th-is
information gave illness probability estimates that were higher
for radon than for asbestos (they still g reitly underestimated
the 25-fold difference between radon risks and asbestos risks at
the same exposure level). Comparing P5 to P4 and P7 to PG
reveals that this contribution o! risk probabi] .ty information
to illness probability estimates continued even in the presence
of standards that did not reflect the 25—fold between-hazard
risk differential. Probability data also produced a slight
difference between radon and asbestos in the highest level
people said they would find personally acceptable, but only wher
no standard was present (P1 through F3). In the presence of a
standard tha ignored the difference between radon and asbestos
risics, info r-tion aba-ut risk probabilities did not improve
-------
6
sub.j:ects’ abilit to see th-is d&fference when ChO iflg.
a c eptab1e exposure levels In spite of these effects on
ilTh,ess piobabthty ti uates and dcceptable levei.s, i-t hotflri
be i t d’tWa t tne p;r-óbá ±li t-y data (-or probabiliti e ;p ps
sin king cöiñ a-r-i h ) d-Ld hot 1-cad radon subjects- to farm
perceptioi s of thr éa-t or action .plãns that diffe’red .-ftoi those
formed by asbestos subjects, eve i thot gh the two q rçups received
probao-ility i-n-formatthri wtth a 2 5—fOld differenc in risk.
-I -n short, people seem somewhat able to under ’and r-j;s. -1c oba-
bility i-hfor ation, but we have not fotthd a for t á pabre -61
hel-pi-ng them integrate this information into t -heir vi(ews 0_f the
seriousness of the threat or their need for a ti óh -. Wieri- .a
standard is provided, furthermore,. the e-ff’ect of the tah dard
may vitiate the much weaker effect o’f the risk probäb’ility
information.
5. The value of risk comparis and a g-raphical
presentation . Format F2 added compari sOns to siiIoking to the
risk probability information in format Fl; format F3 displayed
the information from Fl in the form of a histogram. Both
comparisons to smoking and graphical display improved sü .ects’
ratings of the helpfulness of the brochure and their ce taThty
about their risk. On two of the fou-r measures o-f
risk-averslveness, risk comparisons and graphLca-1 d-i -1a.y also
had the effect of making subjects less risk-ave 5e; for é amp le,
when they were asked what levels they would find persona 1l-y
acceptable, higher responses were given by subjects receiving
formats F2 and F3 than by subjects receiving format Fl.
However, there were no differences between F2 and Fl or
between F3 and Fl in the extent to which subjects di stinguished
high levels from low levels or radon from asbestos. That 4s,
comparisons and graphical display had no effect or. the a c . acy
of illness probability estimates, or on the variation in threa-t
perceptions or action plans with level or with hazard.
Comparisons and graphical. display, in short, helped subjects
feel that they understood their risk better, and made them less
risk-averse, but did not in fact strengthen the relationship
between the actua] risk and subjects’ responses to that risk.
6. The value of providing maximum informp.tiojn . Pormat F7
presented more information than any other (i.e., risk
probabIlity data, risk comparisons, an action standard, advice
and verbal labels). Except for the histogram in format P3, F7
had everything that any other format had. Yet it scored as well
as or becter than the other formats on almost all measures of
communication success, including subjectsv eva]uations of the
helpfulness of the brochure and their certaint s about their
risk. We found no evidence that subjects were confused by
“information overload” in format F7.
-------
7
Of course, format F7 is not an option if the. ôq tpr i;s
hot prepared to provide an explicit standard dnd act1on,.ad iióe
And It is probably not advisab’e in cases where à,th fi sä-
major problem 4nd the c mmunicatir wiShes to en o e k Iium
risk-aversiven ss in the aud ence. For most oth ’r. pir s ’,
however, P7 i th’ for.wat f choi’ e a-mong the refr’ ‘ed.
7. . : .,pr ,sjvenes... to risk inaanitucle jn ’or i1a’t.i ’0n-.and-ehe
jQcatjona y hesis.. As mentioned earlier, zn e-. of ‘the
fox-mats test succeedea in giving su:bjects an- a’ ura ,te ’Cser e of
the mag nitude of their risk. All of them did ‘a f’ã r ’ bb ’ôf
producing thre t perccptions nd action plans t’hãt varied w i-th’
the level of rauon or asbestos (P4 through did l est), but
none of them was really able tc’ produce threat percept-ions -ahd
action plans that were sizably different for radon than for
asbestos.
It is not surprising that formats P4 and F6 failed in this
latter task; they provided an action standard that ign re’d the
differer.ce between radon and asbestos and no ris k probability
information to point out that difference. Thus, P4 and P6
suggested implicitly that radon and asbestos riskS are about
equal. It is somewhat more surprising that the risk robabi1ity
information n P5 and F7 did little to overcome the misleading
impr ssion created by the standard. But it is believab e that
an action standard is a stronger cue than risk data.
What is most surprising is the fact that, for most outcome
vr r E hlcs, formats ?1 through F3-—formats with risk probability
intorma ion and no action standard--produced no greater
‘ ifferencc between the two hazards than the other formats. Of
thr’ seven formats tested, we would expect Fl through F-3 to yield
the greatest realization that radon is more hazardous than
asbestos at the same exposure level. (Formats explicitly
pcinting out tne difference between radon nd asbestos or
offering a different action guideline for t e two hazards were
not tested.) In fact F] through 1 did produce slightly more
awareness of the radon-asbestos distinction in estimates of
illness likelihood, but rot in threat perceptions or action
plans.
Why were all forlnatE able to produce a level effect while no
format was able to produce an appropriately large
effect? A format that suc.cessfu]ly communicates risk
probability information should produce both. We be 4eve. tha
the tJ t an&nUtigat ipn intentic ns wj ii .
rtsA .ng ex i ure level with,in a _ did not reflect the.
the ne ir iLi _inj rmation. Instead, we hypothesize that
people were responding to the position of their test results on
the exposure level ladder in the brochure. In other words, we
believe that responses varied with level in formats Fl, F2, P5,
P6, and P7 chiefly because higher levels were “farther up” the
ladder. Simiiar]y, we believe that subjects who received format
-------
8
F3 responded to thE relative heights of the hist-og.ram bars, not
to the numerital risks associated with these bar heigh ts Even
those receiving the standard only condition ( ‘4), we believe,
used this guidepost to create their own scale. a-nd responded
chiefly to ho . far above or below the .. tandard their Level was.
This “locational hypoThesis” asserts that t:he within—hazard
variations with test result were chiefly a product of the
placement of the level on the page, a purely arbitrary
“locational” factor, arid not an appreciation of the magnitude o.
the risk. The locational hypothesis neatly accounts for the
failure to achieve an appropriate respc nse to the &tfference
between the two hazards. Phet e were no locatio ,al differences
between tadon and asbestos; a parti ’:ular level of radon (in
pic ocuries per liter) was located at about the same poi.iit on the
risk ladder as the comparable level of asbestos (in fibers per
liter), even though the radon exposure represented a 25-fo]d
higher risk. Thus the sludy’s succeSs in achieving a i y 1
effect and its failure to achieve a ard effect can both be
explained by the locational hypothesis.
The naturai tcndcbncy when preparing informational materia1
on risk is to offer exposure 1 ciders that start at the lowest
levels that are normally encountered for that particular hazard,
and that end at the highest ]evels normally encountered.
According to the preceding argument, this practice may se2.iously
interfere With the communication of risk probability
information, since readers may tend to “few the bottom of the
page as very low risk and the top as very high risk, regardIes ;
of what risk information accompanies Lhe ]adder.
These conclusions can be tested easily in future research
A risk ladder that covered the same range of risks for asbestos
as f or radon (rather th in the ame range of exposures, as in th
present formats) would locate the asbestos readings low on the
ladder and the rad’n readings high on the ladder. This should
produce qreatcr recognition of the diftercnce in risk betwe-?
the two hazards. Thus, the locational effect, if verified,
could b. used to communicate risk probabilities more
effectively, rather than obscuring them as it does in the
form;its tested so far.
R.
(a) If the natu e of the hazard i,s such that apathy is a
problem, US’? of a rtandard—on]y approach (P4) is likely to
produce the most risk—averse rL ponse and the fewest
underr’ aci:ior.s.
(b) If overreaction or panic is a prob]em, then a standard
plus advice, with or withot.t risk probabilities and comparisons
(i.e., P ’6 or p7), perforn s best, discouraging action by people
w ith low leveis and leadJnq to less risk aversion than the •ther
formats.
-------
9
(c) If a cotnrnunicator is not prepared to offer an action
standard or advice, a d wishes to help citizens thiderstand the
magnitude of their risk so they can arrive at their cwn
decisiois about mitigation, none of the formo.ts tes-t d is
successful. People can not 3asily use risk magn•ltude
information to appreciate the seriousness of a threat or to make
decisions about action plans.
(d) For most purposes, format Fl, out- most complete
condition, is an appropriate choice. It helps pe ple
distinguish between different risk levels for a single hazarr3,
3eads them to mitigation decisions in accot’d with
recommendations, and gives them a feeling of conftdence that
they understand their risk.
(e) No format yet tested seems able to help people assess a
particular hazard in a way that i sensitive to the differences
between hazards. That is, no format was successful in conveyin
the fact that radon is nuch more serious than asbestos at the
exposure levels in the study. Developing communication
approaches that help people appreciate differences among the
hazards they face shoeld be first on the agenda for future
research—-and testing the locati nai hypothesis is, we think,
tt?e place to start.
(f) These findings may help information and education
programs communicate about many other individually remediable
hazards, inc]uding lead in drinking water, elevated blood
cholesterol levels and blood pressure levels, earthquake and
flood risks, and decisions about foods containing pesticide
residues. Risk decisions, however, are sometimes sensitive to
the precise manner in which the decision problem is posed. We
have tried to create a decision problem close to that
encountered by people who test their homes for dangerous
substances, hut we canr.Dt tell whether there are important
differences that would Jimit the applicability of the
conclusions reached here. Before putting the preceding
recommendations to use, they shou’d be tested with hazards r ther
than radon and asbestos, and with actual test results rather
than with the hypothetical test results used here.
-------
ii
IN?RODUCT ION 1
The difficu’ty of informing the public about the magnitude
of risks and about the changes in risk that can be achieved by
remediation is widely recognized. Citizens often seem to ignore
nformation designed to alert them to significant ar.d reducible
azards. Yet, these same citizens may insist that the
c vernment undertake remedial action for community risks that
a too small or too irremediable to merit attention. If these
individuals test their own homes for health hazards, they may
find it quite djffj u1t to understand information intended to
help them decide what action to take (Weinstein, KJot , &
Sandman, in press)
Although researchers have identified many factors that
influence nonexperts’ risk judgments (e.g., Slnvi , Fischhoff,
& Lichtenstein, 1985; Vaughan, 3986), very few studies have
examined methods br explaining risk magnitudes. There are
hardly any data to support claims that one approach works better
than another. The goal of the present investigation was to
begin to fill this void by testing a variety of promising risk
‘resentation formats. We focused on two remediable home
izards, asbestos and radon. Successful messages about such
nazards should encourage action when risks are high and
discourage unnecessary action when risks are low.
We have searched the existing scient]fic literature for
information about communicating risk magnitudes and risk
changes. Journals trom various disciplines, including
psychology, communication, and public health, wer’ ‘-he focus of
attention. Specifically, we hoped that such journals as RJ J
1 Major funding for this research was provided by the risk
communication program, Office of Policy, Planning, and
Evaluation, U.S. Environmental Protection Agency. We wish to
thank Ann Fisher, director of that program, for assistance
during the planning of this project and during the preparation
of this report. The suggestions of those who reviewed the draft
report for the Agency are very gratefully acknowledged. The
opinions expressed in this report, however, are solely those of
the authors.
-------
12
AnaLyjjjg, the J oi xna’ of Comrnun icatiQfl , y i zon flt, the
Aii rJ çãn Jp l and t e
ycf oiioay wouLd ‘be ouices o-f information about èf fecttve
techniques f r onin uiiicat-ing risk through pri ,ted ‘mater.ials.
It became ã parent that “format effectivené’ss” has b an
overlooked by resea-rchers. In fact, we are awa re ot only One
study that has examined this issue, and its f-Inä:l p r is ntt
yet available (Sth ’i h, Desvousg s, Fis-her, & 3olThsor , 19 87).
Queries to vatiOus experts in the field of risk c6ithnuiii t1on
were equally unsucceasful in uncovering research on fôri a t
effectivenes .
CJ U ERiA EW5LU I’ING .SUCCE$S IN RISK cOM3N IC X N
Thers are many possible criterta for judging the
effectiveness of a risk communication. As a consequence,
evaluating success is a surprisingly complex undertaking. To
date, the criteria that are appropriate have not been examined
systematically. The following paragraphs set out five major
issues: comprehension, acceptance, consistency, uniformity, arid
audience evaluation. Al] five will be used in evaluating the
formats included in the present investigation. A sixth
criterion reflects the notion than some types of comii unieation
errors may be more acceptable than others. It can be applied,
however, only if judgments have been made about the relative
advantages of ‘different errors (especially the advantages of
o erreaction vs. underreaction) in a given situation.
Com rehens on
Sometimes we attempt to communicate factual information or
information that is widely accepted as fact. In such cases
there is one “correct” interpretation of the message, the one
that accurately reflects the information transmitted. A
successful communication is one that results in comprehension,
that is, agreement between the audience’s interpretation and the
correct interpretation. An example of a fact that might be
included in a risk communi ation is the probability of illness
at a given exposure level. If this risk probability
2 There is considerable disagreement among experts about
the probability of illness at different radon and asbestos
expos res, despite genera] agreement that both are serious
hazards. To adopt comprehension as a risk communication
criterion, one must accept that the factual claim being advanced
is indeed “correct.” If it is seen as a debatable opinion,
acceptance is a more appropriate criterioa.
-------
13
iñ orrna i-on i s 1: luded in a s ag; , it is hop e that the
audieice will un’derstand it and that audience perceptions w ill
t herëaiftet match tI dse conta ijied iii the message-.
ceptance
When a -r-iAk message conta-i i s a recommenda-tt-On- or áh ë*p.licit
interp etat iofl, o e can ask to What ektent the t d ienô äcdepts
the öómün-icatht’f judgment. or example, a b ochuré ma
tecominend that hoineb yier vedU be- rád i, levels -dyer 4 picócuries
per liter (pC i/i) or it Thay provide ‘he interpretation that 100
fibers of áêoe tos per liter Cf/i-) is -a “se rióus risk.”
The individuals, agencies, or corporati ns that issue such
messages may f e1 that th ir recommendations or ihte- p e €ati’ons
a-re “correct,” ev.en though they are ai.tLa-11- tha ter of
opinion. However, it is !-mpoitañt to disti ngu-i h between a
situation whe- e an audiénc misundërstands a inessagk, i- d’icating
poor co municat on, and a iituation where an audi-èncê disagrees
witn a message, indicating that it understa.nds the mes-sa e out
has re-ached a different conclusion. It would certai-n-iy be
inappropriate to conclude that people who disagree w-ith a
recomm’endation are necessarily wrong. 3
Cons istencv
In many instances there is no objectively correct
interpretation, no recommendation, and no expli-cit
interpretation. No particular response is anticipated. For
example, government agencies do not have specific expectations
concerning how worried people should be at a given level of
exposu-re. In this case, success can still be judged, but by
relative rather than absolute criteria: A message is successful
if responses vary systematically with variations in risk.
The criterion of systematic variation refers both to
variations in risk that occur with changes in level within a
single hazard and to variations in risk between hazards. People
exposed to high radon levels, for example, should feel more
worry than people exposed to lower radon levels (a with-in-hazard
comparison), and people wi’.h radon should be more worried than
peopJe with asbestos if their radon exposure present-s a greater
risk (a between-hazard comparison).
The between-hazard and within-hazard aspects of consistency
3 The-re is consider’ble debate over whether or r.ot it is
appropriate for government agencies to offer recommendations on
ma-tt-ers of opinion, rather than simply explaining the facts and
leaving audiences to decide for themselves.
-------
14
-db not -ñecê sa r1Iy h eèanIe-
- / ; . — $ ‘ .
for!n ts; A, format that -is. succepsfu1 in’exp lainin g that -20 - t 4i
of asbe tos is a rea ei ri 3rthaci 5 f/ li y t béS ti âás
in -e1aining tha1 2O’f/l of asbestos i-s -i ess ri-sJ y thân20
At times, the di 11 i bhà .b t ééñ
co p ensión-, rid a é t h ä 1 é- -at -bluf
increase with ekposure, a d we war t ze 1 his -
fact, small changes in risk may be wot1-wan t
people exposed to 1 pCi/ i df radon to rel ently thaf
people exposed to 20 -pCi/i. But if one- p-e s o rn - a- -I?ëk e1 Of
1.0 pCi/i and another a level of 1.1, we would ot want them to
behave differently Thus, analyzing hazard response bo see
they vary consistently with risk must be- guided by infor na ion
about what differences in risk are sig iificant- ‘petermining
what di-f-fer n-c are s-i hiridant i an t é iiái - thiè , not just
a matter of consistency.
Un-iformitv
Other factors be Lng egual, a mes a e . at. ié. inter .preted
similarly by various audience members wouX dbe preft rred to one
that is int rpreted differèhtly i y eacrk- jn :jv-jè üa -1. Thus, low
variance in response i-s one measure of süccèssful
communication.
Audkence Eva-that-ion
Some messages may be perceived by audier es as more -or 1ess
he- pfu’1, clear, or aôcurate thar otheT S S. A id es u1
contxñünlcatibn is one that i-s rated pOs-i t-i e-2 y - the âüd-1ëi ä on
the e and similar d4 ens ions. Th h Sa4.é that PØ e I -i-kb
best, however, may not be the one that iè ôSt ef ’fe tive WWe
judged by other criteria.
Di-rect-i o,n of Communication ‘E rQrs
i-na-lly, it shou1 d also b-e kept ih -ñth -that the var-iols-
possible discrepancies bi twèen riessai ès ändau’dience
i-ñtërptetätions are ñ0’t riecessa’Iily eqt1iya ent?. SOme -
discrepancies n ay be more ccep able ‘th-ah óAers For example,
the Onse- uáflcGs of t8k’i -I g a t i when adtió is n t rê en d d
a.y be quite differeht t Ôm the c éà uei ices of not ta!k.in
a override
-------
15
action when action is recommended. Testing for radpn in a
low-risk area may be a waste of $20, but not testing in a
high-risk area may be a matter of life and death..
Furthermore, if it has proved difficult to encourage p e:op le
to take a needed precaution, formats that lead peôp1e. to
overestimate the threat are preferable to formats thá;t 1-cad
people to underestimate the threat. If panic is. a i ikeiy
outcome, on the other hand, a format that leads people to
underestimate the risk may be preferable. I any .particdlar
situation the communicator needs to aSk whether- ,érrêäct dn or
underreaction is (or is more like-ly to be) a p : än’d take
this into account w1 en evaluating a risk corn nuñ.thabion.
The tendency of a communication format to provoke
underreaction or overreaction constitutes an ádäi•ttoñal measure
of success. However, one cannot use this criterion f b se1 ct
communication formats w 4 thout making a judgment about whLbh type
of error is worse. Furthermore, to speak of overreact-ion or
underreaction implies that we know what hazard resp onse i-s
“right.” Thus, this c”iterion ca’n be applied whe; cøinprehension
or acceptance is involved (in which cases a spwif c reaction is
sought), but it is not relevant when uniformity r audience
evaluation is the measure of success. Analyses of consistency
can also reveal a tendency of certain formats to produce
risk-tolerance or risk-aversiveness, but because the consistency
critericn does not assume that a particular answer is correct,
it is not proper to speak of underreactions or overreactions.
SELECTION OF RISK COMMUNICATION FORMATS
Many different risk communication formats were considered in
the course of planning this investigation. In making our
choices we attempted to balance Fhree factors: a) the desire to
study approaches that are in frequent use (e.g., a standard
above which action is required or recommended); b) the desire to
study variables that have been of interest to risk communicators
(e.g., risk comparisons); and C) the need to arrive at Sn
interpretable design so that it would be possible to determine
what format elements were responsible for the effects observed.
After much discussion w decided to study seven printed
formats. Table 1 displays the variables that were included in
each of the seven.
Fl. PROBABILITY condition.
*information about the numerical probability of harm
(deaths per 1000 people receiving lifetime exposures).
-------
16
Table 1.
Iujr diehts in Different Fon ts t d
Format
Prthability
(deaths per
1000 pe ple
Hiito ran i
i p iscns
to snoking
tarx3añ1
$d IriC +.
e bá1
1 bë
T ádd r
‘v 1
of
‘
.
-
F2
,
X OC
-
—
,00C
—
- _
—
‘
F3
OOC
)OOC
-
-
-
-
F4
F5
F6
-
XXX
-
-
-
—
-
)0OC
—
‘ O cX
X XX
)OCX
-
-
‘ O c X
-
)O
F7
)OcX
-
‘ OcX
)OcX
ôcX-
&
-------
17
£2. PROBABILITY + RISK COMPARISONS condition.
*probabjljty information as in Fl
*colnparisons to smoking risks.
Since both radon and asbestos increase the risk of ‘ ng
cancer, the comparison to smoking was appropriate.
F3. GRAPHIC PROBABILITY condition.
*probability lr.formation as in Fl, displayed in histogram
form.
These first three formats convey information about the
magnitude of the risk without iu-posing evaluative judgments
about which risk levels are “serious” or which ones warrant
action. All three contain risk probabilltie . Readers are
required to digest the information presented and make their own
evaluations about the risk. Formats Fl to P3 allow us to
determine whether the addition of risk comparison’s or the u e of
graphical devices enhances the communication of risk magnitude
information.
The last four formats incorporate evaluative information;
they all contain a standard or action guideline. Standards are
so widely used that we felt it important to include them in
several variations. On the other hand, we did not include the
standard in all of our conditions because it is often necessary
to communicate risk information when no standard exists.
F4. STANDARD condition.
F5. STANDARD + PROBABILITY + RISK COMPARISONS condition.
(combination of P2 + F4)
F6. STANDARD + ADVICE condition.
*standard
*detailed action advice
verbai labe]s
Verba] Labels are so often part of advice that we
decided not to try to separate these two factors.
F7. STANDARD + PROBABILITY + RTSK COMPARISONS + ADVICE condition.
(combination of F2 + F6)
Except for F4 (the standard-only condition) and P3 (the
histogram), all used a vertical scale (an “exposure ladder”) to
show possible exposure levels. An example is shown in
Figure 1. (The ladder was suggested by its use in the U.S.
Environmental Protection Agency booklet, A Citizen’s Guide to
Radon. ) The ladder was omitted from 14 because informational
rn teria1s that contain a standard but no dose—response
infori ’ation rarely if ever contain a scale of exposures.
-------
18
Figure 1
Format F7
Raden
L.vól
(pC Ill)
EzSrl
Caac.r
(OuI ó#
1000-- o.oá,d
100 500 In I00 —
40 ‘00 In 1000
20 100 In —
10 50 In 1000
4 20 In 1000
2 70 In 1000
SIn 1000 —
0.S 25 In 1000
- f -
Risk From Lifetime Radoii Fxp6su 1
1 -
: !r T F1 -
tO PIckildiy
Ti
2 Icks day
c: :
a Clq.r.It..Id.y
Advlc.
VERY LOW TO LOW RAlX N LEVELS
Measurcweflu-in th.i,ange 4re noht hec than he outdoor
b ckgroundl vc in ma,w 8reCS. Ekposure to cftaac lcvcls
d a not call fOu act.oif. Even wIhCsC .lbw-kvela. there a a
small risk.a cd ied’wlUel,me ek IcsUretoraddn
Ho ver, au;hotiIiàl: t4at’cffOrta;o-riduce radon lc els
still fur’ her arc likely iobe e cn á and i,iéffaciivê.
HIGH TO VERY HIGH-RM)ON !..EVELS
much
higher than ih;EPA;clion;guidelmC. E d urcIo sUch
lesciS La ycrv4a lgcroU5. For rcssdçnis li tngin bomcsiai
the higher and oFihi range sci loat ahbi ld bC taken w th,n
_ the extcoUplc;orwccki)0 s’ubsia t allytc4ucc Ihclr
e post.re.- ti -p rnptaco is i o poi ib e orls not
errecitv.:- hçy,s ould conSider mo, flgun ll -ihe radon levels
arc, reduced. Exp brC t&lcyels ariheio ç1 Cnd of thc
lange is also ünsa C. ,ReaiqenL lMQg1fl. homei ajthçie
revçIa snould act to rcducc the rea hngs-witbin the ncit
couple nt monilu.
M ThTOHLGHRA)ON L
1ienI in Ibis ii -
oiutc r these
2 Ctg.r.lt..day —
0 1 0.5 In bOO
-------
19
Except for the fact tbat F4 lacks a ladder of exposures,
formats F4 to F? constitute a 2 x 2 factorial design, with one
factor being the presence or absence of “enhanced” probability
information (probabilities in cases per thousand plus risk
comparisons) and the second factor being the presence or absence
of advice. This set of formats can suggest the additional value
of advice and risk magnitude information with comparisons when a
standard is already present.
Although conditions Fl to F3 differ from conditions F4 to F7
in several respects, a comparison of these two g-roups gives an
indication of the impact of providing a standard or action
guideline.
HAZARD SELECTION
The class of hazards we elected to study has several
distinguishing properties. It contains h izards that confront
individual homeowners rather than being community—wide problems;
hazards for which tests can be carried out to indicate the
seriousness of the risk (and therefore present the problem of
explaining the test finding); andhazards for which
individual-level remediation is possible. Among the hazards
that fall into this important category are: lead contamination
from water pipes within the home; contamination of home wells by
toxic chemicals in qroundwater; asbestos in the home; naturally
occurr. ng radon; vapors from urea—formaldehyde foam used as
insulation; and pesticide contamination.
Successful communication about such hazards is particularly
important. Effective messages should increase the likelihood
that people take action to reduce health risks when such action
is appropriate and should decrease the likelihood of excessive
worry and unneeded action when risk levels are low. When
hazards, such as toxic :aste sites, do not permit individua]s to
assess their own risk and to make their own decisions about
mitigation, other com nunication strategies may prove ‘iore
successful than the ones that proved most successful in this
investigation.
We chose to include two different hazards in this
investigation in order to deternine whether the success of
different format4 is consistent across hazards. Including more
than one hazard also allowed us to test the ability of formats
to convey between-hazard risk magnitude differences. A format
may be able to show that risk increases with exposure within a
single hazard but be unable to communicate the fact that one
hazard is more serious than another. Although an even larger
sample f hazards would be desirable, it was necessary to
restrict our focus to two in order to test a variety of formats
with sufficient statistical power.
-------
20
-seiected were asbestos and- na-tura--ily
c these- -hazatds - f-Iécts several
flit was the availabfI ity a-f
øt only were døse—respon e data not
ias many diffefrent effé c , has
t -popu’iat’ion góup , ind -has 1 ea-lth
cues-ad- in terms ‘of- —b] od-
i - onmenta-i coiicenttaietoris. To
- ñ, We- wou-ld nee& to ask
t iac —they get their own water from
- - flf d ±t d ff:i ui t to re la-te to th&s
si-tuati-on. Also, well contam;i h ti-on is usua-1iy due to
i-ñdu ri-al ch inical , so it raises the compli ating issue of
blame. Formaldehyde foam and pesticide contam-ination a-iso raise
the -i-sst.e Of bla.mé,, and the risk appeaTs to be lpwe than for
the hazards -sCiected -a-nd is not as well estab ishèd-.
In addition t the similarities already mentioned, radon and
asbestos are similar i-n that both cause lung cancer anà i-n that
health e-ffec-ts take years tø app-ear. These two haza ds also
di-f ,èr in soiwe respects. ‘or example, geological radbn is
-hatü-ra-l-l-y ocôurrin hereas the presence of asbestos in homes is
due -tO humCn aOtion; the dangerousness of asbëst s has been
kñq ri1 for many years, but radon ha-d received mofe recent media
aé it- on at the t1i e of this study. It was imposs ibie to
detè mine the impact of such attributes on s ibj:ects-’ responses,
but this was not a s-ignificant probi.em because this research was
ocused on differences among forniat rather than on differences
among- hazards.
ME tm IIOD
RES-EkRCH PLAN
r ,r r asons of feasibility and budget, the people who
participa-tea in this study had not actually tested their homes
for -asbestos or radon. They were res-idents of New Jersey
sé éctêd at ra idorn from telephone listings. Th ese i-ñd-iv-iduals
were contacted b telephone and asked to he p test materials
d-e rgned t explain environmenta-l risks. Volunteers were
ra-ñdthnly assigned a hazard, a format, and a hypothetical home
test result.
Thur different test results were used for each. h-az-a-rd ,
representing levels that were sub tantialiy be-low the action
gu-id 1 he., lc1 tly below the action gu-idel-ine, slightly above
t-e ac.t-iOh u dèI-ihe, a-nd substantially above the action
ü f’de ]Jiifé’. :TIie kôvels -for .as ,eetos were 0.8, 2.5, 3.5:, a-nd 2-4
/1. Th é. els for ± ádon -w re 0.8, 3.5, 4.5, arid 24 pc-i/i. In
eabh- -cá è, t-kie- h i hest risk is 3O timCs the lowest risk, so that
we would e cpeot significantly different reactions from people
ái iedthe highes’t and lowest levels if the difference in risk
communicated. The middle two levels
-------
21
are s1igh -iy di Uerent for radon than for ss h cause the
action gu±delirces are 3.0 f/i for asbestos’ and 40 pC l
for radon. The four levels, seven formats, and two hazards
constitute a research desi4n ,.-ith 56 cellS. Data collection
was planhëd- to y. e)d a minimum of 30 subje ts i-ne c’è]i-l -,
gi-i’iI g a total o at ieast 1680 subject .
The use of hypothetica-l test results has several major
advantages. ‘First, it avoids the ethical prbblem .of €äst ’i ’ng
communication materials that may be ineffective on people who
are -actively attempting to evaluate a r ál Séc d, i t
greatly simplifies subject recruitment by Lnc’reaSiñ t he pool
of potential partici an€s, allow-ing us to test a i à ger nuniber
of communication formats. The d isadvantage is that we dO not
know wl ethe.r resp,o ses to hypothetical tes.’ s di -f-fé r f-tom
tes onses to real te tS. Cbnclus tons from the pfesent research
paradigm eventually need to be evaluated On c it izèns who have
cairied out real test-s on their homes, but such evaluations
will be much more difficult to carry out.
SAMPLE
The potential population was made up of all households with
listed telephone numbers in the New Brunswick, New 3 ersey
telephone directory. This directory includes many different
communities with significant variations in soci-oecohomic
stat-us. The area includes communities in whic) i high school
graduates predominate and ones in which college graduates
predominate. Study iarticipant-s had to be at leaèt 18 years o1d
and own their curre.. t residence.
MATE l*LS
Risk Communication Brochures
Four—page brochures were developed from in-formation
contained in various government publications. The first three
pages, with basic information about either radon or asbestos,
were constant across conditions. The last pa ge of the
brochures was the format being tested. (The 14 brochures are
presented in Appendix A.) All brochures except format 4, the
standard—only condition, and format 3, the h rstogram, used a
vertical ladder to show possible exposure levels. Successive
rungs on the ladder differed by about a factor of two, so the
ladder constituted a logarithmic scale of asbestos or radbn
levels.
4 No standard or guideline for home asbestos currently
exists. We derived the 3 f/i level used here from a standard
that does exist for schools. The weekl acceptab-le school
exposure is the product of the 0.1 f/cm maxi mum acceptable
exposure level and the exposure duration in the school of 40
-------
22
With radon measured in picocuries/liter a nd asbestos in
fibers/liter, the same range of levels could be inc1ud d (.1 to
100). Controlling this range of levels as irnpgrtant. It
ensured that any differences that might be fouhdb twèen
responses to radon and responses to asbestos wóu] d: not be due
to differences in the numbers used to describe th e test re ul’ts
(for xantp1e, if asbestos were measured in fibers/cc, the
typical unit, rather than in fibers/liter, the asb estos LeveLs
ei’nployed here would fall into the range .0001 to .1).
The radon action guideline was the 4 pCi/i action level
adopted by the U.S. Fnvironmgntal Protection Agency and the
Centers for Disease Control . There is no standard or
guideline for home asbestos levels, but there is a standard for
;choo1s, and ths level, approximately 3 fibers/liter, was used
as an action guideline in our research. It is important to
note that th action guideline for radon represents a much
greater risk than the action guideline for asbestos. In fact,
at any radon level, measured in pCi/i, the risk is roughly 25
times g eater than the risk at the same asbestos level measured
in f/i.
For the advice condition:;, the regions above and below the
guidelines were each divided into two parts, yielding four
regions. A paragraph of advice was written for each region.
The advice tr. ::hed the action guideline rather than the ctn?
hours per week. Home effects are typically based on an assumed
exposure of 18 hours a day, o: 126 hours a week. Converting the
maximum acceptab]e school exposure to an equal home exposure
yields an “acceptable” home level of about 3 f/i.
?AdcJ I ng r t n 1a rd to - î brochij re would have had 1 i tt I e
effect if the subjects we recruited were already aware of the 4
pCi/l action guideline. A small telephone survey was conducted
in February 1988 (N = 29) to determine how many peopiC could
correctly idertify the radon guideline. People were asked “Do
you recognize any of the foliewing numbers as the radon level at
which people are supposed to take action: 0.1, 0.4, 2, 4, 10,
20, or 100?” Ot the 2 responderts who had not tested their
homes, only one picked the correct response. This procedure was
repeated in October 1988 (N 28). At that time, none of the 26
respondents who had not already tested picked the correct
response.
°Theie is uncertainty about the precise magnitude ot
asbestos and radon risks. In order to test different formats we
chose not to att’ mnt to explain the uncertainties in the
estimatts. We b l eve that the risk estimates cited in Lhe
materials develop d are close to current scientific estimates,
but we do not clairn that they are definitive values.
-------
i- -isk. That i ; thé ad i - 1ór tiTh 4 o s€• crIt±ln: n-Th th’ •-t1 e
asbestos and the rador booi€let was that no act .on was
nèces ary. 6r the t
h a zards was th action sho’uld be
actua r-i sk in c the
the. .risk in the
in :ö d 3 1 - is
guideline; -we b I e è’. th,&s i -s p ifiüa.ril y be
levels re cmménded -for b StàS :a n a h -i
of ho n-e*pbsu-rè; -
The hi stogr iiis preseñtrn asbe to and âion p rob bi1 f.½.es
were clearly labeled to indipate the actdarrisJ .evei .
liowever, the histoqrams for both hazards were st ared to ut ilIz -e
the full page Thus, the heights ot - the asbe to rnd -radon.
bars were the same even thou h they rëpfesented 1f sk “ma4n 4 udés
that differed by a factor ot 2-5.
Ev-aluati,.on Ou s-tion j
The feedback questionnaire ex3mined a variety of róspon e
dimensions. A copy is iri-cluded in AppendixB..
Audienç eva -l j ion. Fou:r questions t ’er e .q êd..
as assessed with a 4-point scale: 1-very di ff’j u’ t to
understand to 4-very easy to understand. ‘rfie helpful-ness
question also offered four cho-icës: 1-did -understand
my test result to 4-very helpful. IThspor de itS-’eval’Uàted the
amount of information as follows: l-iii ich too little; 2-too
littJe; 3—about right; 4-too much: 5—muc i €ob i itich;. A final
question asked whether the brochure provided a ood
understanding of the risk: 1-very good inderstat d±ng
4-very uncertain
Perceived risk maqnjtude . A 7-point sc iie asked about the
likelihood of harmful effects (1-no chance; 7—cCr - air to
happen). A 6-point scale required subje t to áte thè-ir
asbestos or radon level from 1—no risk to 6 very s-erious risk.
CoLncern and fear . 5-point scales ranq -ing -f in l-n’ t at all
to 5-extremely were used to ask about the co ice ii- and ea r
respondents would feel from their test result.
bi’ fic i-1ty
through
Reuiediat .ion difficulty .
difficult to 4-very easy.
Choices ranged f-r- m 1—very
emediation decision . The alternatives offered we-re: I -dc
not feel action is needed at my level; 2-undecided and leaning
toward not acting; 3-und cided and leaning towa rd acting;
4—shou)d take action only if it is relatively easy to reduce
the level; 5-should take action even if it Ls hai d to reduce
the level. Very few people chose the response “undecided nd
leaning toward not acting.” Consequently, the responses on the
mitigation scale were not normally distributed. The two
undecided choices were co nbined to yield a 4-point scale with a
isk
-------
4_
-dis’tribution t at cämé dioser to rneeti the nor i-t ’
á -ésu pti’ohs deri -n the sta -ti’st -id i i ode s used
Acce tab .l e. e osu -re . eve s. P.ëspondeflt-s w th
level Ln thèii uia i -n living area they would feel sati’sfled - so
tfiat they bu1d hat s ênd rnbr n ohè tr i -r g to ge t- é ‘ 1ê -1
even l ’o ér. C!id cës an ed - fr’dTh -O -t•o 20 /i or
Respo de1ts also were ask .sd to thagine that a s t o. f xends
had tested tl eIr homes and had found different asDestos or-
radon levels ranging from 0.8 to 50r-pCi/l or f/i. The t do -part
question asked wh.ch of these friends should try to roduce
their radon leveis if the cost wëré ‘S 00 and- i
$2000.
lUness probabilities . As a comprehension test, a question
was included which asked about the probability of i:.1irie s’ at
the respondent’s exposure level. This informatian was
contained in formats Fl to F , F5, and F7, but not th fbi iñats
F4 or F6. A 9-point scale, with every second point lab ] d
was used. The following !.abeis were used: 1—no cthancë 3--i
chance in 1000 (0.1%), 5-1 chance in 10-0 (1%), 7—1 dhance in 10
(10%), and 9—certain (100%).
Because the two risk magnit .ide questions (likelihood and
seriousness) and the conc ern and fear questions were a-li very
highly correlated with one another (correlations of .56 to
.72), they were summed into a single measure of perceived
threat. The strong inter-item correlations led to a composite
measure with high reliability (reliability coefficient alpha =
.88) . Similarly, the two questions concerning mini’ iUm levels
at which friends should act (r = 0.73) were combined to form a
single measure with reliability 0.84.
1 QCEP! 3
Potential respondents were contacted by telephone and asked
to participdte in this tudy. To ensure a balance, males and
fema]cs wer recruitc d from households on an alternating basis.
Volunteers we-re screened to en3ure that they were 18 years of
age or older arid owned their home. They were also asked
whether they had carried out a radon or asbestos test in their
home.
7 Reliabilities would be inflated if the correlations among
items were based on the combined sample. Instead, correlations
were calculated within each cell of the research design. These
;orre1ations were subjected to Fisher’s 2 transformation,
averaged across cells, and then reconverted from Z scoies to
correlations before the reliability was determined. It is worth
noting that perceived threat was normally distributed. There
was no evidence of a bimodal distribution.
-------
25
Those recruited ihto the tudy were sent ohe q- the ri-sk
comznun át-ion -btochures for ei ther asbestos or rãdôn- and an
evaluati on -uest:ibnnaire. They were also sent a ‘ tio ne test
result-” that gave the result of a recent, hypothet-icá l test i-fl
theIr home. The cover letter accompanying these ni têrials rt ade
c-i-ear that th-i-s as -an imaginary value, used t m kë--t &e task
of evaluat4h out material-s more concrete. Ins€riicti-bns st-át’ëd
that responder ts we-re to “use thi-s level when atis r-inq
questions about høw serious a probi-em you have:,” and that tthOy
should “f el free to look back at the brochure when f i1-i1-ng
out” the questionna-ire. Study parti-cLpants were è ed cä ies
of the other brochures being tested after they t:ü± zied the
feedback questionnaire.
Participants were randomly assigned to one of the 56 cells
in this study. However, forma-t effects would be unlikely to
a ppea-r if individuals were already knowledgable about a
hazard. Therefore, volunteers who had tested t-heir homes for
asbestos were assigned only to radon conditions; those who had
tested for radon were assigned only to asbestos conditions.
The few who had tested for both were exc1uded from the study.
Because no subject received information about both radon
and asbestos, conclu ions about differences in reactions to
these two hazards are between—subject comparisons. Most
communications about individually remediab]e hazards focus on a
single hazard (perhaps with comparisons to another hazard, such
as smoking, that is thought to bc familiar enough to aid the
communication process) . Sometimes, however, messages do try to
explain about two hazards. For example, physicians may have
to explain the risks from both radiation and surgery in order
to help a patient choose a cancer treatment. The ability of
people to see differences in risk may be greater when two
hazards are discussed simultaneously than when each is
discussed separately.
The task f3ced h respondents was rather demanding. They
had to read three pages of information about a hazard,
interpret their imaginary test result, and complete a 4-page
questionnaire giving their evaluation of the brochure and their
perceptions of their imaginary risk. Telephone reminders were
used to maximize response rates.
In September 1988, the U.S. Environmental Protection Agency
issued a recommendation urging anyone living on the second
flooi of a building or lower to test his or her residence for
radon. The extra interest in radon that this announcement
generated might have affected public perceptions of the
seriousness of high radon levels or affected the perceived need
for mitigation. The attention might also have increased the
number of people who were aware of the 4 pCi/i action
guideline. because it seemed possible that homeowners
recruited after this date would react differently to our
brochures than people alreidy in our sample, we decided not to
test any more radon brochures after that date.
-------
2 6
Al]. together., ap roximaté1y 8500 households were contacted,
it -h 2905 34;-2 %) a éeinq to take part in this project-. Of
these, 1980 ret frned 6npleted feedback aesti-ói ai- , 6 .l%.
k few of these were fl-I-led out incorrectly, yielding a final
usable sample of 1948, 67.0% of those mailed out. These
figures were practiäall.y constant across format ., Shb :iñg that
di -fferenc s that mi ht ppear in eva1ua-tion of the sêv h
formats we-re nbt caused by different reSponse rät S.
The final sample was 46.8% male, ar 1 d the med-ian age was
41. A total o’f 25.6% had a high school education or lesà,
45.9% had attended college, and an additional 28.6% had pursued
graduate study. People living in single family houses made up
89.2% of the sample, and 95.2% owned their current re idënce.
Although only a limited proportion of the population
provided feedback on the brochures, it seems likely that the
people who volunteered to take part in this study are a-iso the
type of people most likely to request and read inforrnation
about environmental hazards As might be expected, the sample
prov-ed to be quite well-educated. The relationship between
education and risk communication success is one of the issues
thdt will be examined in the results section.
j ASURES OF COMMUNICATION SUCCESS
IN THE PRESENT INVE$I ATION
All six of the criteria for evaluating risk communications
that were previously described can he used to evaluate the
success of the formats developed in this research. Each
criterion is listed below, along with the specific issue that
was examined and the question numbers in the feedback survey
(Appendix B) that were involved.
Comprehension
Communicated and perceived risk probabilities should agree
(Q13). (Better results were expected for those formats that
provided such probabiliti€s: Fl, F2, F3, F5, and F7).
Acceptance
Mitigation plans (Ql0) and message action recommendations
should agree. (Better results were expected for those formats
that provided action recommendations: F4 through F7.)
The hazard exposure levels judged to be acceptable (Ql1,
Q12a, Q12b) should agree with the recommended action levels.
-------
27
-Pérc p ioflsc ot. threat (Q5., Q6., Q-7, arid Q ) - and is ic ns
abó it i n 1 (Qi? shoürct vary with é sure. Th 1è1 1 thi i
bà a)da d g êii x ure level -o f /I),
-for ä sb 5
rj i e ti ai arc- oi u . -j ê vei Ø judg d €0 :á á) ] :( Ql -’i;
Q12a, Q12b ) s aou] d highér fór a bestos than fór -radbff. Tile
mortality risk at the radon level judged acceptable should be
about the lame as- the moi tà-l.ity ri- k at the asbestos l evêl
j:u ged á cè tãbl ë (e.g., 1. £ i 3 OO).
Unifoz nity
The variance in -perceived threat (Q5, 6, Q -7, ánd- 8) ánd
in piähs to mitigate (Q’lO) should -bC low.
Audience E taa-uation
Ratings of clarity (01), helpfulness in understanding the
test result çQ2), amount of infotinat-ion (Q3), and perceived
certainty about the risk (Q4) should be high.
Comparing Comnlun -jcption Errors
The tendency of any format to produce over- or
underestimates of risk probäbilitiès (Q13) waè exam-med.
The tendency of any format to produce parti ü arly high or
low feelings of threat (Q5, Q6, Q7 and Q8) w s exetñined.
The tendency of any fot mat to produce mitigati on intentions
(QlO) when mitigation action was not recommended or intentions
not to act when mitigation action was recommended was examined.
The tendency of any format to produce judgments about
acceptable levels (Qil, Q12a, Q12b) th&t were consistently
higher or lower than recommendations was examined.
REBTJLTB
The results will be presented in the same ord er as the
previous section, “Measures of Coml%iunication Success,” except
that the direction of errors will be described when the
variable to which they apply is first discussed. For example,
information about the discrepancies between action plans and
action recommendations will be presented under the “Acceptance”
heading where action plans are first examined.
After all six success criteria have been eonsidered,
findings for other variables (such as education) are reported.
-------
28
The reader WISh -tà -refer back to TabTh 1 on -page 16-
where Obi ten€ o f aqh - t he siven -formats is described-;
Ifl the f 11o ;ing., e er fl eae ct of : a d ;
level ig mentibned, herëid ’r can aa’sum’e that the ékect i
statisticil-ly significant beyond the .01 level. o tuf erit s’ b 9 ut
différence 6ng specifi fo ’rmatS are s-ignif’iicafft-”beybna the
.05 1 évèi -;°
Details of the analyses are provided in Appendi xq,’. r pr
each of the result stb ec ii ns hera , there is a corresponding
section in Append-tx C vidi’ng a ‘cbinplltè -1ájiãt-i’oh &f- the
statistical próceduré thà t was used.
Table 2 summarizes the resu-rts by format fpr ,a;X - á-i-x
criteria of co unicatión suâqeSs. Some read rs ,may wa,nt to
skip the rest ô f the “R eü1 ’ èe t Oñ -and é y- zi’ -15 2 -ár d
the “interpretation and Cónó iA*i’ôñ1” séOt i-on iñ eä,d’
COMPREH HS ON .
Illness probabilities
The nuntber of scale divisions separating the correct answer
from the answer given was the primary rieasUre of
comprehension. (The answers that were considered correct are
detailed in Appendix C.) In the abserce o relevant
information, totally randOm answers wc,u-ld yield er±ors
averaging 2.07 scale units. S-imply pióking the middle choice
on the scales would give average e-r-rt rs o-f 1.25.
The percentage of jud men-ts in errc..r by more than one scale
unit is shown in Ff ure 2 and the mean size of the e- rors is
shown in Figure 3. As suggested by th se figures, formats
8 For statistical tests involving the entire data set (for
example, a test of a difference among formats.) the .01
significance level was chosen as a criterion because -effects
less significant than this, g-iven the large saáple sIze, are
very small and have little practical importance. PoCt-hoc tests
comparing specific formats (for example, a test of the
difference between formats Fl and F2) use only part of the
sample and are less powerful, so t-he .05 criterion l-evc’l was
used in such cases.
9 1•n all figures where results are presented by format, the
formats will be presented in consistent order, with Fl farthest
left and F7 farthest right.
3 °Errors are measured in scale units. Correct answers are
included in the determination of average error tze (with error
size zero).
-------
-
JCrjterIan of cSi$s ticaiicn
;SUcCfls!
- —-- - - - -
- p
-
Präb
--
4 r2c
;- ) ‘ - • - -
Prob S iiiijo áii , -
••-
• •:
- -I - -‘ ‘ -‘- - - --
‘
bdtter ‘bitter jo6rest’ bette A
C Pehens ion
Low rrär rites in tiide?naid
ing-1tl ,%as-probabLt-ities
better
-AdCSta i c e
AcEepts récomiendat ions for
mttigaticws action
‘boean’t act at tow levels
*Acts at htgh levels
poore st
poor
good
-: c - • -; -
poorest pooreSt better bettef but’.
I -
poor poor poor p &r Pb € r -
good g o Od -b uitê i befi i -. :A6eifef -:
b st
?—
Sttjr
; b t ij ;
$ i s tency
Personally acceptable levels
varç with harerd
poor
poor poor worst wors t •ttãist.
cwolit
LO ut friend should act at
vary with hazard
5oor
poor pOor p&or -
-
- ji r
Alress probability varies
witn haza;d
poor
poor poor woi st ‘poo
poor
Pwrceived threat varies with
hazard
very poor
very poOr vSry-podr very poor very- j Or l vert poor
Very poor
Action plans vary with hazard
very poor
very poor very poor very poor ver9 poor te ry poor
very poor
Perceiied threat vai lea with
level
fair
fair fair better benâ? -b OttEr
-bej iOr-
Action plans vary with level
fair
fair fair better bitt(r
bettii,
Uniformity
low var i ante in response
average
average average avera9e á óu gé .ii 4Va9e
v6ge
Evaluation
Helpful in urderstanding result
slightly
worse
slightly average slightly ave a9e li titly
better woise Sitar
sl-i Iitly
b tter-
Low cncertainty about say risk
RLgk Senqitivi tv/Aversion
Own iltnose probability
slightLy
worse
tower
average averaga average avefIji sL-ightty
- h it ler
lower lower highest lower lower
-stlihtl?
b utter
lowor
wets that would be personally
acceptable
Perceived threat
average
tower
highest htghest Lowest averae average
lower tower highest lower lo i ter
avOra e-
lower
Need to act
avenge
lowest average highest avero9e bisect
lowest
-------
30
Figure 2
Error Rote In Judgments
of Illness Probabilities
Fl F2 F3 F4 F5 F6 F7
. ——-
Absolute
Figure 3
Value of Errors
In Judgments of Illness
smokin g+
advice
, El F2 F3 F F5 F6 F7
P&,mbers N mbers+ Histogram
smoking
0 Mbestos U Radon
smoking
c
a)
c i
L
‘I-
00 )
OU)
E>
J
z
100
80
60
40
20
0
3
2.5
2
1.5
1
0.5
0
Numbers Numbers+ Histogram Standard Std+nos . Std+ Std+nos.
+ advice +
smoking
(scale divisions)
Probabilities
1
Standard Std+nos+ Std+ Std+nos.+
smoking advice smoking+
advice
-------
31
P4 and F6, the two formats that did not provide risk
probabilities, yielded more errors and larger errors than the
ormats that did contain this information. There were no
significant differences among the formats containing the
probabilities. To some extent, people were able to use the risk
probabilities presented to answer this question.
Figure 4 shoWs that there were differences among formats in
the amount of perceived ris k. In addition to producing more and
larger errors in risk estimation, format F4 also produced higher
risk estimates than the other formaes, with this difference
being greater for asbestos than for radon. None of the other
formats differed from one another. Figure 5 shows that radon
risks were consistently (moderately) underestimated and asbestos
risks were consistently (substantially) overestimated.
CEPTANCE
Mitigation de Lsjons
Several different criteria were used to examine agreement
between subjects’ action plans and the recommendations in the
brochures. For example, the appropriate response for a person
with a level just below the action guideline was ambiguous.
“Undecided,” “do not act” and “act If easy” could each be
considered as correct. The three sets of criteria developed
were labeled “Dichotomy,” “Diagonal,” and “Comp1e ” and are
explained in Appendix C. As seen in Figure 6, the results were
relatively consistent across criteria. (Recall that formats F
through F3 contained no recommendation.) Not surprisinq].y,
conformity with advice was greater when a standard was present
(F4 through F7). Provision of detailed advice (F6 and F7)
increased acceptance of recommendations still further. Risk
probabilities and smoking comparisons did not increase
acceptance.
Next, the disagreements occurring with test t ults above
and below the action guideline were examined separately (see
Figures 7 and 8). Overall, there were many more disagreements
with advice at low levels than at high levels; compared to the
suggested response, people tended o overreact. At high 1evel ,
providing a standard (F4 through F?) increased acceptance, but
detailed guidance (F6 and F7) had no further effect. At low
levels, a standard did not reduce the “overreaction.” Only
detailed guidance led more people to listen to the
recommendation that action was not neces$arily appropriate. The
addition of risk probabilities and smoking comparisons reduced
neither underreaction nor ovetreaction.
Accept ble exposure levels
The highest level at which people said they would be
satisfied and the lowest level at which they said friends
-------
certain
1 In 10
Odds/
Percent
I In 1.000
no chance
Figure 4
Judged Probability of Even 4 ual W.ne ss
if NO Action Is Taken
Errors (scale
Figure 5
d lvTs lons)
of Illness
In Judgments
Proba bilitles
Mimbers M.arnbers+ Histogram Standard Std+nos.+ Std+ Std+nos+
smoking smoking odv!ce
fl Asbestos I Radon
smoking-f
advice
Mirnbers Mimber3+ Histogram Standard Std4-nøe.+ Std+ Std+nos+
smoking smoking advice smoking-f
advice
2 Fl F2 F3
15
z
F5 F6 F7
O 5
0
—0.5
—1•
-------
33
Figure 6
Measures of Acceptance
(Deif a indlca*es
advice plans
of Action Advice
degree of
agreem ent)
11111 Dichotomy I Diagonol I Complex
.Ei
Asbestos
0.5
0.4
0.3
02
01
0
F7
I
1
I.
0
0
0
a)
a
Numbers Numbers+ Histogram Standard Std+nos.+ Std+ Std+nos.+
smoking smoking advice smoking+
advice
Fl F2
Radon
0.4 -
0.3
FLi
F7
0.2
I
0.1
Numbers Numbers+ Histogram Standard Std+nos.+ Std+ Std+nos+
smoking smoking advice smoking+
advice
-------
- ‘4
FIgure 7
Disagreements with Advice to Act at
Levels Above Action Guidefln.e
(Underreactions)
Disagreements
Figure 8
with Advice
Not to Act at
Levels Below Action Guideline
(Overreacfio ns)
Fl F2 F3 F5 F6
1110 Asbestos I Radon
C
a)
C )
L
a,
a
100
80
20
0
40
‘1
Numbers Nurnbeis+ Histogram Standard Std+nos.+ Std+ Std+nos.+
smoking smoking advice smqkjnq+
advic
C
ii )
L
a.)
a
100
80
60
40
20
0
-I
F7
-l
Numbers Numbers+ Histogram Standard Std+nos.+ Std+ Std+nos +
smoking smoking advice smoking+
advice
-------
should take further action were both considerably lower than the
action guideline. As seen in Figures 9 and 10, this aversion to
risk was greatest for F4.
It is also interesting to note that the higher the test
result people received, the higher the level at which they said
they would be satisfied and would advise friends to be
eatisfied, even for respondents receiving a standard. The more
serious their problem, in other words, the less people insisted
on a “complete” solution.
CONSISTENCY
erceived threat: -
For all formats, perceived thLeat was slightly greater for
radon than for asbestos at the same exposure level. In other
words, the perceived radcn threat at 3.5 pCi/i was slightly
greater- than the perceived asbestos threat at 3.5 f/i. The
difference was considerably less than would be expected from the
difference in actual risk (i.e., the radon risk at 3.5 pCi/i has
a ]ung cancer probability 25 times that of asbestos at 3.5
f/i). Furthermore, the difference between the percei”ed threat
cf radon and the perceived threat of asbestos was not
significantly larger with one format than anotner.
This failure to reflect between—hazard d±fferences in risk
is not terribly surprising for formats F4 through F7. These
recommend action at nearly the same exposure level for asbestos
as for radon. However, even when only risk data were present,
in formats Fl through F3, the inform ’tion was apparently not
well—absorbed. It seemed that people responded more to their
position on the risk ladder (or the height of the bar in format
F3) than to the risk numbers that were attached to the ladder.
The greater perceived threat for radon can be seen in the
fact that the lines for radon are slightly higher on the y-axis
in Figure 11 than the lines for asbestos. However, if the
thrc at perceptions properly reflected the magnitude of the risk
differences, the asbestos lines would appear much lower (or the
radon lines much higher), so that the asbestos lines would merge
smoothly with the radon lines. This would eliminate the
discontinuity that occurs when one switches front asbestos to
radon while moving along the x—axis.
Perceived threat: Variations with level
The variations in perceived threat as level increases within
a single hazard are indicated by the slopes of the lines in
Figure 11. Because Fl, F2 and F3 performed similarly, as did F6
and F7, these have been combined to simplify the figure. The
slope of F4 was not significantly different from the slopes of
F6 and F7. Still, it is kept separate in Figure 11 because
perceived threat was significantly higher overall for F4 than
-------
36
Figure 9
Highest Level at Which Subject
Would Feel Satisfied
El F2 [ 3 F5 F6 F7
Lowest
smoking
I
Figure 10
L2vel at Which Friends
Need to Act
ll Asbestos
smoking
U odon
Numbers Numbers-i- Histogram Standard Std+nos.+ Std± Std+nos.+
smokinç smoking
advice smoking+
adv 1 ce
Stfll
_i ___ j _
r7
Numbers Numbe’s+ Histogram Standard Std+nos.+ Std+ Std+nos.+
advice
smoking+
advice
-------
37
Cor.sistency
Fl
44
I’
gure
of Perceived
with Actual
Risk
1’O
Actual Mortality Risk
(deaths/i 000)
ido boo
Threat
*
4-
1
• -. Asbestos -‘- Asbestos - Asbestos Asbestos
F1—F3 F4 F5 F6—F7
- Radon - Radon -. Radon Radon
F1—F3 F4 F5 F6—F7
-------
38
for afl. of the other formats (see Figure 12). Forrna-t F also
appeared to he somewhat different, so it is k ept distinct.
As seer clearly in Figures 1]. and 13, perceived threat
increased with level, but there was a very significant
discontinuity or step at the action guideline. In other words,
people with test results slightly be1o the action guideline
were significantly less concerned than people wi-th test results
slightly above the guideline. This is not a desirable outcome,
since the degree of risk in these two cases is essentially the
same. The size of the discontinuity varied with format: It was
much greater with formats F4 through F7 (all conta-ini-ng action
guidelines) than with F ” through F3 (none of which contained a
guideline). (Some analyses suggested that the discontinuity was
smaller for format F5.) Below the guideline, formats with a
standard had a greater slope; in other words, the sta-ndard
helped people distinguish low from very low levels. Above the
guideline, all the slopes were the same.
Mtig tion ec i
These iiridings paralleled the threat perception results that
have just been described (see Figures 14-16). People were
slightly more likely to say that they would act with radon than
with asbestos at the same exposure levels, but the differences
between hazards were small compared to the difference in actual
risk. The between-hazard differences did not vary significantly
with presentation format.
Mitigation plans to act increased with exposure, but again
there wa . a substantial discontinuity at the recommended action
level. This step appeared to be somewhat greater with formats
F4 through F7 than with Fl through F3. However, the step
appeared to be negligible for F5 (just as it was for threat),
and there appeared to be a step for F3 when none would be
expected. Below the guideline, formats with a standard had a
greater slope; that is, the standard helped people distinguish
low from very low levels. Above the standard, all the slopes
were the same.
Intentions to remediate were greatest with F4, significantly
greater than with F2, F6, or F7.
Maximum_Acceptable_Exposure
The highest exposure levels deemed personally acceptable for
radon co ’-responded to risks that were much greater than the
risks at the levels deemed acceptable for asbestos. For the
tormats containing an action guideline (F4 through F7), in fact,
there werE no significant dfferences between the radon levels
(in pCi/l) found personally acceptable and the asbestoc levels
(in f/i) found personally acceptable despite the 25-fold
difference in risk. For the formats without an action guideline
-------
39
Figure 12
Mean Perceived
Threat
0
c i)
L
H-
D
ci)
>
ci)
0
L
ci)
3-
19
17
15
13
11
9
7
Numbers umbers+ Histogram Standard Std+nos.+ Std+ Std+rtos.+
smoking advice smoking+
advice
smoking
111111 Asbestos • Radon
-------
40
Figure 1 3
Effects of Formats arid Assigned
Levels on Perceived Threat
15- -
13- -
11-
9- -
7- — ________________ - 1
0.1 1 10 100
Test Resuft
-i -- Numbers -i’ - Numbers-f- - Histogram -- Standard
(Fl) smoking (F3) (FLI)
.° Std.+nos.+ Standord+ Std.+nos.+
smoking advice smoking+
(F5) (F6) adVICe
4J
0
a,
L
D
a,
>
a.)
0
L
a,
0
-------
41
act even
if hard
act if easy
Mitigation
Plans
undeci ded
do not act
0.1
Co r si.Stency
with
Figure 14
of Mitigation
Actual Risk
Plans
1
1 10
Actual
100 icioo
Mortality Risk
(deaths/i 000)
- Asbestos - ‘ Asbestos -° Asbestos - Asbestos
F1—F3 F4 F5 F6—F7
Radon - Radon - Radon - Ration
Fl —F3 F4 F5 F’â—F7
-------
42
ac t even
it hard
act if
easy
MitIgation
Plans
rdecided
1
A
Figure. 15
Mean MfflgratIon Pt
an
If] Asbestos Radon
Fl F2 F3 F4
3.5.
25
F5
not
needed
Std+nos
+
smoking
Std+
advice
+
smOlthg+
advice
-------
43
Figure 16
Effects af Formats
Le eIs on Action
and Assigned
?Ian
3ur -
act if
easy
Action 25 -
plans
undecided 2- -
1.5--
not
needed
0.1
Test Result
Numbers - - Numbers ± - - Hi stogam -- Standard
(Fl) smoking (F3) (FLI).
(F2)
-o Std. +nos. - Standard+ -- Std. +nos. +
smoking advice smoking+
(F5) (F6) advice
(F7)
--- I
1 10 10
3- -
act even
If hard
4- -
-------
44
people accepted higher asbestos tAst results. Note from Figure
9, however, tL.at the difference in acceptable levels for those
not receiving a standat L came nowhere near the 25—fôl higher
exposure level for asbestos that would properly reflect the
25-fold higher risk for radon. Adding smoking’ comparisons to
the risk probabilities did not significantly improve the
recognition that a personally acceptable exposure should be much
lower for radon (in pCi/i) than for asbestos (in f/i).
As far as friends’ action is concerned, none of the formats
was any better than others in helping people understand that
friends could (at constant risk) tolerate much h igher levels of
asbestos (in f/i) than of radon (in pCi/i).
Illness probabilities
Difference in perce:ived illness probabilities were greater
(though still small) for the formats that contained risk
probability information than for F4 and F6 that did not.
UNIFORMITY ACROSS PARTICIPANTS
Perceived Threat and Action Plans
Although it is desirable that different people come to
similar conclusions about a hazard, the consistency in the
responses given by those assigned to the same hazard, format,
and test result was no greater for one format than for another.
AUDI ENCE EVALUATION
Difficulty and Amount of Information
These two items did not refer specifically to the last page
of the brochures, the .± hat contained the risk presentation
formats and the only one t t differed among brochures. One
woulQ expect people to consider the whole brochure when
answering these questions, so it is not surprising that no
significant differences among formats were found. The mean
response for difficulty was 3.50, halfway between “fairly easy
to understand” and “very easy to understand.” The mean response
for information was 2.86, where 2 = “too little information” and
3 = “about right.”
Helpfulness
Since this question referred to helpfulness for
understanding the test result, format differences might be
expected. As Figure 17 suggests, differences were found, but
they were not particularly large. The rank order was F6 (best),
£7, £2, F5, F3, F4, and Fl (worst), but only some of these
contrasts were large enough to be significant. Format F6 was
ratec significantly higher than Fl, F3, F4, and F5. Formats F2
-------
45.
very
helpful
Helpful
ddnt
help
4
3.5
3.
2.5.
2
1.5
1
Figure 1.7
Helpfulness for Understanding Test Result
111111 Asbestos S Radon
odvce
• •. - •‘,
rs
Mimbers t’&imbers+ Histogram Standord Std+nos.+ Std+ Std+nos.+
smoking srnokin.g advice smoklng+
-------
46
and F7, rated nearly as high as F6, were significantly better
than Fl and F4. None of the formats was rated lower than
“moderately helpful”——a “3” on the l—to—4 sca:le——ar d none was
judged above 3.6.
Risk Certainty
For all brochures, subjects felt that their unders-ta’nding of
the risk fell between “fairly good” and “very good,” with means
closer to the former response. Formats F6 and F7 were judged
slightly better than the others and format Fl was judged
slightly worse. Only the contrasts between F6 and Fl and
between F7 and Fl were significant. As is evident in Figure 18,
all between—format differences were small.
DIRECTION OF COMMUNICA’rION ERRORS
Findings that bear upon the issues of underreaction vs.
overreaction and risk aversion vs. risk to-lerance have appeared
in the preceding subsections. These include:
*Format F4 produced the highest risk probability estimates.
*Formats with advice (F6 and F7) decreased overreactions at
low exposures.
*Formats with a standard (F4 through F7) decreased
underreactions at high exposures.
*Format F4 led people to want particularly low levels
before they would be satisfied.
*Format F4 led to the highest perceived threat.
*Format F4 led to the greatest intended action.
OTHER VARIABLES
There were no differences among the formats in perceived
difficulty of mitigation or in subjects’ interest in actually
testing their homes for radon or asbestos. (There was more
interest in testing for radon than for asbestos.) Subjects
rated asbestos as more difficult to reduce than radon, 2.03 vs.
2.65 respectively, where 2 = fairly difficult and 3 = fairly
easy.
Additional analyses were conducted to determine whether the
success of different formats might vary with the educational
level of the reader. Although the sample was well—educated,
about a quarter had no more than a high school education.
Examination of the maximum level that would be personally
acceptable, perceived threat, and intentions to take action
found no significant interactions between education and format.
It did appear, however, that the more educated subjects were
somewhat more sensitive to the magnitude of their test results.
This can be seen in Figure 1.9, where the variation in response
as one goes from the lowest to the highest test result is
smallest for the high school group.
-------
47
very uncertain
about my risk
3.5
-a 3
C
a
2.:
very good
under ’tanding
Figure 18
Gave a Good Understanding of My Risk
, Fl F2 F3 FL4 F5 F6 F7
lUll Asbestos • Radon
smoking
M.jrnbers M.jmbers+ Histogram Standard Std+no .+ Std+ Std+nos .+
smoking advice smoking+
advice
-------
48
17
15 -
Perceived 13-
Threat
11•
9.—
0.1
act even
if hard
3-
Mitigation 2
Plans
not
needed i•
0.1
Effects
Threat
on Perceived
Plans
Test Result
Figure 19
of Education
and Mitigation
1•0
iào
—
100
1
10
-. Post—
college
High School - College
-------
49
INTERPRETATION AND CONCLUSIONS
Although many differences were found among the seven formats
included in this study. it is not easy to go from the preceding
results to general conclusions about formats for communicating
risk magnitudes. The following paragraphs attempt to summarize
the main lessons demonstrated or suggested by the data gathered
in this investigation. It should be kept in mind, however, that
decisicns about risks are sometimes sensitive to the precise
manner in which the decision problem is presented. Although we
iave tried to create a dccision problem close to that
encourtered by people who test their homes for dangerous
substances, we cannot tell whether there are important
differences that would limit the applicability of the
conclusions reach d in this research. Before putting the
following lessons to use, they should be tested with hazards
other than radon and asbestos, and with actual test results
rather thin with the iypothetical test results used here.
THE VALUE OF AN ACTION STANDARD
Comparing formats Fl through F3 with formats F4 through F7
i.cates that a standard increases the likelihood that pcople
will follow recommendations. A standard also helps people to
distinguish between very low and low risks. (Even when nothing
other than a standard is provided, as in format F4, people buiir
a scale around this reference point.) However, a standard also
creates an artificial discontinuity in hazard response as one
goes from just below the standard to just above the standard. A
standard does no better than other formats in helping people
distinguish high from very high levels.
The standard-only condition, F4, star.ds out from the others
in many ways. It leads to the highest perceptions of threat, tc
the greatest intentions to remediate (especially at levels below
the action guideline;, and to the lowest acceptable levels. On
several other variables it produces the most risk—averse
responses.
The differences betwcen format F4 and formats F5 through F7
seem gLeater than can be explained by the risk magnitude
information or advice that these others contain. In fact, F4
differs from F5 through F7 in one additional respect: It is the
only condition that does not contain a ladder of exposure
levels.
We believe that in the absence of a ladder, subjects have no
way of telling what types of re ;ults are ui .aual, and they rea- t
therefore as thoi 1 gh the finding of any amount of hazardous
substance were a serious problem. (Note that the lowest level
-------
.n this study was 0.8; we cannot be ce rtain how people
ct with still lower levels.) In contrast, people who
• dder and a low test result see that their exposure
iS low on the page and feel reassured. If their level is
the standard, the presence of still higher rungs on the
: is also reassuring. Those with results at the very top
f the ladder might become n’ore frightened, but an exp ü e .
appears to be rcassurir g for everyone else.
THE VALUE OF ADVICE IF THERE IS A STANDARD
n this study people were more risk—averse than the act ior
endations. They often said they would -emediate at ëvél s
the guideline. Those receiving action advice (F6 and F7)
the most likely to accept the reccmmendation not to take
n at ]ow levels. Tt was not that advice led to a greater
rentiation among hazard levels, hut that advice reduced
perceptions across all l ve13 when conpared to the
ard-only condition, F4.
ith advice, subjects also judged their formats more helpful
derstanding their te. ,t results and they reported that they
better understanding of their risk.
nalyses of perceived threat and mitigat ion plans for ‘he
formats containing an action guideline suggested that the
ion of paragraphs of advice did not help people distinguish
low and high risks.
j ALUE O EROBABII ITIE$ AND
PROBABILITIES PLUS COMPARISONS
eople gr ’ atly ‘ nderastimated the difference between the
risks and a oestos risks in this study, regardless of what
t they recejved. However, when they received information
sk probabilities or probabilities pitis comparisons to
ng, they were able to recognize some difference between the
iazarcjs in the likelihood of illness at their test results.
!erl j?i the comparison of F5 and F7 to F4 and F6, probability
1sK comparison information modestly increased the
zen—hazard differences in estimated illness likelihoods even
e presence of a standard. yet, probabilities and
irisons had little effect on threat perceptions or action
Probabilities arid comp3risons produced a sliqht
rence between radon an asbestos ir the highest level
e said they would find acceptable, but only when no
iard w present. In the presence of a standard,
bilities and comparisons did not improve subjects’ ability
e this di.fferen e between radon and asbestos.
nalyses of perceived threat and mitigation plans for the
formats containing the action guideline suggested that the
zion of risk prob. bilities and smoking comparisons did not
-------
51
lead to a further differentiation between high and low risks
beyond that provided by the standard alone.
Th ‘JAL,UE OF RISK COMPARISONS AND GRAPHICA.L PRESENTATIONS
Contrasting Fl with F2 or F3 shows that eomparisons to
smokthg and graphical display of information improved ra-t ings of
heipfulness arid certainty and lowered two out of four
risk—aversiver.ess criteria. However, a format that succeeds in
communicating risk magnitude information should also help people
form threat perceptions and action plans that vary appropriately
across hazards and across levels within hazards. By these
consistency criteria, risk comparisons and histograms did not
significantly improve communication.
INFORMATION OVERLOAD
Condition F7 presented more information than any other. Vet
it also scored as well as or better than other formats on almost
all measures of communication success. Je found no evidence
that respondents were confused by the amount of information in
this condition.
MATChING FORMATS TO THF COMMUNICATION SITUATION
If the nature nf a hazard is such that apathy is a problem,
format F4, standard—only, is likely to produce the best
results-—that is, the most risk-averse response and the fewest
underreac h.ns. If overreaction is a problem, then formats F6
and F7 perfnrin best; they discourage action by people with low
levels and lead to less risk aversion than the standard a]one.
If a communicator is not prepared to offer a guideline or
advice, and wishes to help citizens understand the magnitude of
the risk so that they can arrive at their own decisions about
mitigation, none of these formats is successful. This last
finding suggests that people do not easily comprehend risk
magnitude information; neither numbers alone nor numbers plus
comparisons to smoking were very helpful.
RESPONSIVENESS TO RIS ’ MAGNITUDE TNFORMATIO
As mentioned earlier, and as indicated by the data in
Figures 11 and 14, none of the formats tested really succeeded
in giving respondents an accurate sense of the magnitude of
their risk. All of the formats did a fair job of prodncing rsk
perceptions and action plans th t varied with of radon or
asbestos (F4 through F7 did best), but if th formats had
successfully communicated r sk magnitude itself, risk
perceptior s and action plans would have been much different for
asbestos than for radon Instead, risk perceptions dnd action
plans were nearly the same for the two hazards when the
exposures (in pci/i cr radon and f/i for asbestos) were the
same. The small differences between radon and asbestos that did
-------
52
appear or some response vari les varied little or not at all
from one format to the next.
Formats F4 and F6 contained no risk magnitude information,
only the action guideline or the action guideline plus advice
Keyed to the guideline. Because the guidelines for asbestos and
radon weie at nearly the same exposure level, F’4 and F6
suggested implicitly that radon and asbestos risks are about
equal. Thus, it is not surprising that differences in the
perceptions of radon and asbestos were negligible for these two
for uats. It may also seem reasonable that the histograms in
format F3 did not yield differences in risk perceptions.
Although the bars were labeled with appropriate risk numbers,
they were scaled to fill the page, so the heights of the bars
were the same for asbestos as for radon.
More surprising is the fact that the addition of risk
probabilities and risk comparisons to the guidelines in F5 and
F7 did little to overcome the misleading impression generated by
the guidelines. Most surprising is the fact that Fl and F2,
formats with no action guideline, yielded no greater difference
between the two hazards ir most outcome variables.
In .. iort, neither illness probabilities nor illness
probabilities plus smoking comparisons enabled people to
recognize the 25—fold difference in risk. This information was
used by respondents to some extent when answering questions
focused directl’ on illness probabilities and levels that would
be p’ r .onally aLceptable, so it was not completely
A1though this report occasionally refers to subjects’
ability or inability to recognize the relative risks of rac on
and asbestos, it is important to bear in mind that each subject
was assigned Just one hazard, ej j er radon or asbestos. The
comparisons referred to in the report are thus between—subjects
comparisons, not within—subjects comparisons. (It may be
misleading to draw within-subjects inferences from
between—sujects data; Nickerson and McClelland, in press.)
Presumably, if subjects had received a single presentation that
compared radon and asbestos risks directly, the difference
between the wo hazards would have been more obvious. However,
even this pre .umption may be misleading. People often mistrust
or relect risk ccmparisons, especially when the hazards being
compared differ along other dimensions, such as voluntariness or
fairness (Covello, Sandman, and Slovic, 1988). For example,
geological radon is natural, while asbestos is technological;
there is someone to blame for asbestos but not radon. Such risk
attributes might have a different effect on direct risk
comparisons than on separate messages that mention only one of
the hazards.
-------
53
incomprehensible, but it improved illness probability estimates
only slightly and had no impact on overall perceptions of threat
or miticiat-ion decisions.
While no format succeeded in producing respons iveness to
risk magnitude differences between hazards, all formats
succeeded in producing responsiveness to risk magnitudes within
hazards——that is, to exposure level (see Figures 11 and: 1-4). We
believe that the increases in perceived threat and mitigation
decisions with rising exposure 1-eve-i within hazards do not
reflect comprehension of the risk magnitude information.
Instead, we believe that people were responding to the position
of their test results on the exposure level ladder in the
brochure. In other words, we believe that responses varied with
level in formats F l and F2 because higher levels were “farther
up” the ladder (formats F5 through F7 also contain an exposure
ladder). The within—hazard variations with test result seem to
be a matter of the placement of the level on the page, a purely
arbitrary factor, and not an appreciation of the magnitude of
the risk.
Similarly, we believe that people who received format F3
responded to the relative height of the bars, not to the
r imerical risks associated with these bar heights. We believe
t: at people receiving the standard only, F4, used this guidepost
in creating their own scale.
The natural tendency when preparing informational materials
on risk is to have exposure ladders that start at the lowest
levels that are normally encountered for that particular hazard
and end at the highest levels normally encountered. According
to the preceding argument, this practice may seriously interfere
with the communication of risk magnitude information, since the
bottom of the page tends to be viewed as very low risk and the
top as very high risk.
These conclusions can be tested easily in future research.
For example, if ladder were created that covered the same
range of risk for asbestos as for radon (rather than covering
he same range of exposures , as in the present versions), the
asbestos readings would appear low on the ladder and the radon
readings high on the ladder. This should produce grEater
recognition of the differences in risk between the two hazards.
Thus, the ladder “geography” effect might be used constructively
to communicate risk magnitudes more effectively.
We have discovered much about the impact of standards,
advice, probabilities, and risk comparisons on risk
communication. Our most complete condition, F7, performed
welJ: helping people distinguish between different risk levels
for a sing)e hazard; leading people to mitigation decisiont3 in
accord with recommendations; and giving people a feeling of
confidence that they understood their risk. Still, neither this
-------
5:4
-hor other conv yed the fact that rd t. i ’rñuc more
serious than asbestos, at the exposure levels in this study.
Developing conimunication approaches that help pe6ple apprecriate
diTfe éncés- ainbng the hazards that they face should be fi: st on
the agenda for future research.
-------
55
Covello, V. T., Sandman, P. N., & S1ov ic, P. 9 Bi. Risk
Communication. Risk Statistics. and Risk ComDarlsons: A
Manual for. P1an T M&ña ers 1 . Washiffgthn, D C: chë i I
Minufacturers Assbáiát ion.
Fli”ss., 3. L. (-1981). . Sta is ca 1 rnethdds for- 5 rate and
Dr000rtions (2nd edition). New York: Wilè i.
Hildebrand, D. K., Laing, J. D., & Rosenthal, H. (].977a).
Analysis of ordinal data . Beverly H-ills, CA: Sage.
Hildebrand, D. K., La-ing, J. D., & Rosenthal, H. (1977b).
Prediction aja.ly.sis f cross classificati . New York:
Wiley.
Nickerson, C. A. and McClelland, G. H. (in press). Across—
Persons vs. Within—Persons Tests of Expectancy—Value Models:
A Methodological Note. JournaT . of Behavioral
Decision-Making .
Sandman, P. M., Klotz, M. L., & Weinstein, N. D. (1987)
Public response to the risk from geologica-1 radon. Journal
of Communication , 1:1(3), 93—ba.
SAS Institute (1987). SAS/S PAT Guide for oersonal computers .
(version 6 edition). Cary, N.C.: Author.
Slovic, P., Fischhoff. B., & Lichtenstein, S. (1985).
Characterizing perceived risk. In R. Kates, C. Hohenemser,
& R. Kasperson (Eds.), Perilous roaress: managina t-he
hazards of technology (pp. 91-125). Boulder, CO: Westview.
Smith, V. K., Desvousges, W. H., Fisher, A., & Johnson, R.
(1987, July). Communicating Radon Risk Effectively: A
Mid—Course Evaluation. U.S. Environmental Protection Agency
Cooperative Agreement No. CR-811075. Report
EPA—230—07—C7—029.
Vaughan, E. (1986). Some i ctors influencing the nortexpert’s
perception and evaluation of environmental risk.
Unpublished doctoral dissertation, Stanford University,
Stanford, CA.
Weinstein, N. D., Kbotz, M. L., & Sandman, P. N. (in press).
Promoting remedial response to the risk from radon: Are
information campaigns enough? Science. Technoboav and Human
Values .
Winer, B. 3. (1971). Statistical DrinciDles in exDerimental
design. New York: McGraw Hill.
-------
. ttw;4*. !E 1 1 aiip *4 1i%stos:
-------
A2
P1 -Asbest os
Test Brochs*re
2-/88
BASIC ASBESTOS INFORMATION
WHATIS ASBESTOS?
Asbestos i ,a mineral fiber found in rocks. i’here-are-sev al
kinds of asbestos fibers, all of which are fire resistantand nbt
easily destroyed by natural processes. Because of its-desir bJe
qualities, asbestos has been used in a wide variety of pro hict
including appliances, ceilings, wall and pipe covetings, flOor
tiles, and some roofing materials.
IS ASBESTOS DANGEROUS?
Although asbestos has many benefits for humans, it is also a
very dangerous mineral. B eathin airborne asbestos fibers has
been shown to cause: (1) Asbestosis - a serious lung disease which
can lead to disability and death; (2) Lung cancer - a disease that
is incurable and almost always fatal; and (3) Mesothelioma - cancer
of the lining of the lung or abdominal cavities. The greater the
exposure to a hestos, the more likely it is that one of these
serious diseases will develop. Workers who handle or come into
contact with asbestos on a daily basis are open to the greatest
health risks.
There is no level of exposure to asbestos fibers that is
completely safe The greater the concentration of asbestos, the
greater the risk.
HOW DOES ASBESTOS AFFECT US?
The danger arises when asbestos fibers are released from the
product or material. These fibers are so small that they cannot be
seen. They can float in the air for a long time and can pass
ihrough the filters of normal vacuum cleaners and g t back into the
air. Once inhaled, asbestos fibers can become lodged in tissue for
a long time. After many years cancer or asbestosis can develop.
Cigarette smoking and asbestos together are especially
hazardous. Exposure to asbestos piLls smoking gives an even greater
risk of lung cancer than adding the risk from smoking alone to the
risk from exposure to asbestos alone.
Asbestos found in “friable” materials is most dangerous.
Friable materials are materials that can be crumbled, pulverized,
13ccau c ihic is an experimenial brochure, please check with
other auihoritic bcforc hiking any aiclions in your home.
-------
‘k3
or reduced to-pQwder by-hand:.pressure. Ashe tos:i I ifspra cJ
on a ceiling is an example of a friable material lncoñfrast;
vinyl asbestbs f1oor tile is not usually friable. Th L s os,-
fibers are firmly .bound . or sealed into the tile and:cç be’r leased
into the-air only iftbe tile is. cut ground, or san4ed;
WHERE IS ASBESTOS I Ii(ELY TO- .BE FO NN1E IIOM
There are several areas-in-the home where sb sto,s r le.msaire
most likely to arise. Tbese- indude:
Wall construction materials and- pipe i nsul tiQn
especially those dating between 1920:and49 .7-2. (This
includes materials found in and behind. plaster.;or
wallboard and in paper tape.)
Friable ceilings in buildings built or remodel d-between
1945 and 1978.
* Material found on stoves and furnaces such-asinsutation
and miliboard and door gaskets.
Other asbestos-containing products that you may find in the home
include:
Patching compounds and textured paints. ($ince the use of
asbestos in these products was banned in 1975, you are
most likely to find it when sanding or scraping old or
damaged material in older houses.)
Vinyl floor tiles and flooring.
Roofing, shingles, and siding.
Appliances with asbestos-containing parts or components,
such as toasters, broilers, slow cookers, dishwashers,
refrigerators, ovens, ranges, clothes dryers, electric
blankets, and popcorn poppers. (Unless broken or misused,
most appliances with asbestos are safe. There has been a
general decline in the use of asbestos in these appliances
during recent years. If asbesto ; is still used, it is in
parts which will probably not release fibers during use.)
Having significant amounts of asbestos in the home is not rare.
Many old homes in New Jersey could create health problems for
residents because of materials that may release asbestos fibers
ir.to the air.
HOW CAN I TELL IF I HAVE ASBESTOS IN MY HOME?
The manufacturer of a product may be able to tell you, based on
the model number and age of the product, whether or not it contains
asbestos. People who have frequently worked with asbestos (such as
plumbers, or building or heating contractors) can often tell you
whether or not material contains asbestos by looking at it.
Problems may occur in the home where asbestos-containing
materials are worn, damaged, or exposed to the air. If you have
ceiling or wall material that is crumbling, or you are preparing a
major renovation which will expose material contained behind a
-------
A4
wall, you should have the material analyzed-to det ri iine-’if-ii
contains-asbestos. I .aboratory analyses range -from bout $20 -to $40
per sample. Several samples may be required- to g inaT!; -accurate
determination of asbestos content.
If you suspect that you have a problem, you may also want to
have an air sample taken to measure the amount of asb stos.fibets
circulating inside your home. To collect the sample, a l b ratory
will send a technician to your home. A pump:is-used tp draw air
from the room into a filter that will trap the asbestos Alt-
electron microscope is used to count the number offibeps trapped
in the filter. It takes about six hours to collect the s inpl : and
costs between $ 100 and $400, depending on the laboratory and
techniquc used. The results of the test can he reported in units
1)1 “fibers per liter of air,” abbreviated as f/I. This unit tells
how much asbestos there is in one liter of air.
WHAT SHOULD I DO IF I HAVE AN ASBESTOS PR BLEM?
If you discover that you have an ashes’os problem, the best
thing to do is to contact a contractor who h. s experience in the
proper procedures for repairing and removing asbestos. The-re are
special guidelines for handling asbestos-containing materials. You
should avoid drilling, scraping, sandin , brushing, sweeping or
vacuuming asbestos materials. This will disturb tiny asbestos
fibers, make them airborne, and increase the risk of breathing
them. It is highly recommended that you hire an experienced
contractor or get professional advice if you are thinking of doing
the work yourself. A contractor will seal off the contaminated
area from the rest of the house and workers will use protective
clothing and a special respirator while they are handling the
asbestos. Using improper techniques can make an existing problem
much worse by contaminating the entire house. For more information
about identifying, testing, handling, and fixing asbestos problems
call the U S. Environmental Protection Agency toll-free at
1.800-638-2772.
INTERPRETING YOUR TEST RESULT:
Using the imaginary test result we have given you, look at the
left column on the next page, under the heading, “Asbestos
Level,” and find the level nearest to your test result. Next
look across to the right-hand column, under the heading “Extra
Cancer Deaths,” and find the number that is associated with
your asbestos test result. This is the extra risk of cancer
from living in a home with that level of asbestos.
-------
- .5 es::s
Cancer Risk From Lifetime Asbestos Exposure
Es Cr c
Asbestos
Canc•r
L.ve
O.aths
( I)
(O .t OF
1000 people)
100 20 In 1000
50 10 In 1000
20 4 In 1000
10 2 In 1000
3 p.60 In 1000
2 7.40 In 1000
1 7.20 In 1000
0.5 7.10 In 1000
0.1 p.02 In 1000
-------
CANCER RISK FROM LIFETIME
ASBESTOS EXPOSURE
20
>1
I
101
EXTRA
CANCER
DEATHS
(Out of 1000
people)
1 iILIL1L
0.2
I I I I —I— i
0.1 0.5 1 2 3 10 20 50 100
Asbestos Level (f/I)
-------
4 -Asz est s
. .tl. “ou should :‘a’ the T L1i .:ed o ‘e rn’n fit
contains a nestos L coruor, nai es ran e from inouL S2() :o S O
per sam le. Several samples ma -e equ.red to gain an accurate
determination of asbestos content.
If you suspect that you have a problem, you may also want to
have an air sample taken to measure the amount of asbestos fibers
circulating inside your home. To collect the sample, a laboratory
will send a technician to your home. A pump is used to draw air
from the room into a filter that will trap the asbestos. An
electron microscope is used to count the number of fibers trapped
in the filter. It takes about six hours to collect the sample and
costs between $100 and $400, depending on the laboratory and
technique used. The results of the test can he reported in units
of “fibers per liter of air,” abbreviated as f/I. This unit tells
how much asbestos there is in one liter of air.
WHAT SHOULD I DO IF I HAVE AN ASBESTOS PROBLEM?
If you discover that you have an asbestos problem, the best
thing to do is to contact a contractor who has experience in the
proper procedures for repairing and removing asbestos. There are
special guidelines for handling asbestos-containing materials. You
should avoid drilling, scraping, sanding, brushing. sweeping or
vacuuming asbestos materials. This will disturb tiny asbestos
fibers, make them airborne, and increase the risk of breathing
them. It is highly recommended that you hire an experienced
contractor or get professional advice if you are thinking of doing
the work yourself. A contractor will seal off the contaminated
area from the rest of the house and workers will use protective
clothing and a special respirator while they are handling the
asbestos. Using improper techniques can make an existing problem
much worse by contaminating the entire house. For more information
about identifying, testing, handling, and fixing asbestos problems
call the U. S. Environmental Protection Agen toll-free at
1-800-638-2772.
INTERPRETING YOUR TEST RESULT:
The U. S. Environmental Protection Agency has evaluated the
risk from asbestos. The following information about the EPA
regulation for schools and public buildings can help you interpret
your own (imaginary) asbestos level:
A home level of 3 f/I or above is the risk for
which EPA requires schools and public buildings
to take action to reduce the asbestos.
-------
—j
-As es s
Risk From Lifetime Asbestos Exposure
Extra
Asbestos Cancer
Equivai.ru
Level
(f/I)
Deaths
(Out of
1000 people)
Smoking
Risk
Advice
I
.i
8 CIg.r.tt..Idsy
— 1 Q 2
Cigarettes/day
1000 —
1/10
C lgar.tt.Id.y
Cancer
100 20 In 1000
50 10 In 1000
20 4 In 1000
10 2 In 1000
3 p.60 In 1000 —
2 1.40 In 1000
1 1.20 In
0.5 1.10 In 1000
1/4
Cigar.tt.Iday
—
—
a
a
a
a
iorr e Ieve Q f’I pr abqve: ss the
risk for wtncli PA requires schoolz
and public buiIdinqs to take action. ‘2
0.1 ).02 In 1000
-------
F -Asbestos
Advice for Asbestos
Exposure
Asb. to 5
Level
Adv Ice
(f/I)
100
50
20 —
—
10
3 —
2
1
0.5
HIGH TO VERY HIGH ASBESTOS LEVELS
Measurements near the uppcr limit of this rangc are far above thc EPA action
level fur schools and public buildings and indicate a strong sourc of friable
.i lic ios. Ncat the upper cnd of thus range, immcdiaic plans should he made to
— redtiu . tlic .isbestos lc el. Ncar the Iowcr cnd of the range. homeowners houkl
,i i iihun a fcw months In the meantimc, asbestos rnatcrials should not he
di itirbcd sincc this would increase the lcvcl still further. Rcsidents should
niinirni/c thc amount of time spent in the contaminated area.
\IODERATE TO HIGH ASBESTOS LEVELS
Mcasurcmcnts in this range are above the EPA action level for schools and
puhli buildings and indicate a signiflcant sourcc of friable asbestos Most
would advise that VOU attempt to idcntify and remove thc source of
— .iiui.imin4ti()n R ideiit should minuinue the time spent in arL.is ul the hot s
th.it in.uy ont.iin friable .i%he Ioc Materials suspected of containing asbestos ‘
h iultl not be tli turhcd sin e this would incrc.I%e the levcl still lurther
LOW TO MODERATE ASBESTOS LEVELS
Me.i urcmcnts in this range fall below thc EPA action lcvcl for schools and
— public building While the hcalth risk at this level is rclativcly low,
muliods are jvailjbl which can clfcctivcly rcducc th amount of asbestos
present in the home yct further
VERY LOW TO LOW ASBESTOS LEVELS
Me.isurcmcnts in this range arc not much diffcrcnt from outdoor urban
oncvnt rations of a hcsios You may have no sourcc of asbcstos in your home
— and even if you do havc a source, you may not bc able to reduce that level
further. The health risk at these lcvcls is very small. Most experts do not
re.ommcnd taking action it thcsc levels.
0.1
-------
F . ; -As e5 S
—! I !!
8 Clgsr.tt.sId.y
—
CIg.r.tt.ilday
1110
Cig.r.tt.Iday
Cancer
Risk From
Lifetime Asbestos
Extra
Asbestos
Cancer
Equivalent
Level
Deaths
Smoking
Risk
Advice
(ff1)
(Out of
1000 people)
Exposure
HIGH TO VERY HIGH ASBESTOS LEVELS
Mcasurcrnents near the upper limit of this range are far
above the EPA action level for schools and public buildings
and indicate a strong source of friable asbestos. ilear the
upper end of this range, immediate plans should be made to
— reduce the asbestos level ?Jear the lower end of the range.
homeowners should act within a few months In the meantime,
isbestos materials should not be disturbcd since this would
increase thc level still further. Residents should minimize
the amount of time spent in the contaminated jrea.
MODERATE TO HIGH ASBESTOS LEVELS
Measurements in this range are above the EPA action level
for schools and public buildings and indicate a significant
source of friable asbestos. Most experts would advise chat
you attempt to identify and remove the source of contamma-
— tion Residents should minimize thc time spent in areas of
the house that may contain friable asbestos. M.itcrials
suspected of containing asbestos should not be dbiutbcd
since this would increase the lcvel still further.
100
50
20
10
3
2
I
0.5
20 In 1000
10 In 1000
4 In 1000
2 In 1000
760 In 1000 —
7.40 In 1000
7.20 In 1000 —
7.10 In 1000
0
1/4
Clgar.tt.Idsy
a
—
—
—
Ilorne Ievej. o f/J or: above ‘L t e
risk or which Ef A ecp. res sc$iools
and public buildings to take action.
LOW TO MODERATE ASBESTOS LEVELS
— Measurements in this range fall below the EPA action Level
for schools and public building. While the health risk at
this level is relatively low, methods are available which can
effectively reduce the amount of asbestos present in the home
yet further.
VERY LOW TO LOW ASBESTOS LEVELS
Measuremcnu in this range are not much different from
outdoor urban concentrations of ubeszos. You may have no
— source of asbestos in your home and even if you do have a
source, you may not be able to reduce that level further.
The health risk at thcse levels is very small. Most experts
do not recommend taking action at these l vcls.
0.1 1.02 In 1000
-------
Fl -R3dc
Test Brochure
3/88
BASIC RADON INFORMATION
WHAT IS RADON?
Radon is a radioactive gas that occurs in nature. It has rio color,
odor, or taste.
WHERE DOES RADON COME FROM?
Radon comes from the natural breakdown (radioactive decay) of the
uranium present in rocks and soil. This is not an unusual situation;
rocks and soil often contain small amounts of uranium. Radon can move
through the soil into the open air.
WHY THE CONCERN ABOUT RADON?
Exposure to elevated levels of radon increases the risk of lung
cancer. Radon is not known to cause any other health problem, only lung
cancer. Not everyone çxposed to elevated radon levels will develop lung
cancer. Still, in some houses the level of risk is very high. Studies
suggest that between 5,000 and 30,000 lung cancer deaths a year in the
U.S. are caused by radon.
The time between exposure and the onset of the disease may be many
years. There are no symptoms or early-warning signs to tell you that
you have a high radon level in your home.
The risk increases as the level of radon and the 1en th of exposure
increase. Therefore, exposure to a low level for a long time may
present a greater chance of cancer than exposure to a higher level for
only a short time.
HOW DOES RADON AFFECF US?
Radon naturally breaks down and forms decay products that cling to
dust and other particles in the air. As we breathe, these particles can
become trapped in our lungs with the radon decay products still attached.
As these trapped radon products decay further, they release small bursts
of radiation that can damage lung tissue and lead to lung cancer.
ARE WE SURE ThERE IS A RISK?
Our understanding of the risk posed by radon is based on a number
of scientific studies of mine workers in several different countries.
The results are generally very consistent. Because the studies have
involved people, rather than laboratory animals, and because the
amount of radon in mines is not greatly different from the amount of
radon in houses with elevated levels, our information about radon is
more dependable than that for most other environmental hazards. We
still have more to learn about radon, however, and the degree of risk
at different levels is not known precisely.
•Because this is an experimental brochure, please check with
other authorities before taking any actions in your home.
-------
DO \R\Y HO\1E5 l\ \E JIRSE\ H \E RADO\ PROBLE\IS ’
Y c R cc ‘ .:‘. ‘ flL ’ 1t :o be i ?robIer orv ..i ’
,f r othe rt \c ’.’. . Re. c g Pror’ L : s s ho’.¼c ..
.: th ur s •T O !e rejd E e’ a ed r:: , . Jr ;e’.e
rure occurrence: they ure ruirlv common. if ‘.ou Ii e north of
Trenton here is a 8 oud chance that some part or ‘our home has
excessive radon. It your neighbor has rested and round little radon.
you could still have a problem, since radon levels can vary greatly
from one house to the next.
HOW DOES RADON ENTER A HOME?
Outdoors, radon from the soil mixes with the rest of the air. This
mixing dilutes the radon, producing concentrations that are usually
quite low. Inside an enclosed space there is usually less fresh air (E
dii Lite the radi 1i I lie rc lu . radi in i iisade a liunic can uecti mu Li ic. u
some times it reaches d.in e rou le els I fumes I hut are .i irt igh t hold
rudon longer and pre e nil re h on c side air (rum entering t lie hoii e.
keeping windows open bind ‘ entiLi ring the home, on the inher hutid. v ill
usually lower the concentration at radon in the air inside the home.
Radon can seep mw a home through any opening in the walls or
ii our ot t he I ULI nd.u ion, ope iii ng u n )U iid pipes, su mp hump holes, u nd
unp:i c cl cra I spaces Radon c in also enter the water of private eIls
unU be released into u home hen the water is used. This is not
u Li 1lI a problem with public water supplies, where the radon ould
likely be released into the outside air during treatment before the
water ever reached the home.
The soil also affects the level of radon that enters the home.
High—uraniLim soil will produce more radon. Porous soil will allow
radon gases to seep into the foundation of the house. Household
appliances that eth 1 iList air to the outside, such as clothes driers,
furnaces, and tirepLices, will .ilso contribute IC) U radon problem.
These aI)pliances creute a small vacuum if they do not have fresh air
in takes going directly to them. The vacuum draws radon from the su iii
into the house.
HOW IS RAI)ON DETECTED?
Since you cannot see or smell radon, special equipment is needed to
detect it. The devices that may he used to determine if your home has
high radon levels include the alpha-track detector, the charcoal
canister, a 5-minute air sample, and others. The first two devices are
very simple to use. They involve removing the seal or top of a small
canister and placing the canister on a shelf or table in the area you
want to test. By removing the seal a special material is exposed which
will register the amount of radon in the air. After the prescribed
amount of time (several days to several months) you simply replace the
top of the canister and mail it to a laboratory for analysis. Typical
costs range from S12 to S50 per test. Because radon levels in a home
often vary from day to ciay. measurements over a brief time period give
only a very rough indication of the situation.
-------
lf .’our name was testec a -.jrcoai or abna-rrack e:ec: r
.-e resuits . ill prooabiv oe reported n mrs Or picocuries Der ter
r air.” abbreviated as pCi 1 ’l This unit tells now rnucn rauon
:r one liter of air.
IS IT DANGEROUS IF! FIND I HAVE HIGH LEVELS?
Radon is a serious risk. There is no level of exposure to radon in
the home that experts can agree is completely safe. The chances of
suffering harm from elevated radon levels are much greater than the
chances of suffering harm from most other pollution problems. As
Consumer Reports has stated, “While there are uncertainties in
pinpointing low-level radon risks, there is no doubt that the risks of
radon vastly exceed the risks from aflatoxin, PCBs, nuclear wastes, and
virtually all other environmental hazards” (July, 1987, p. 442). The
disease caused by radon, lung cancer, is incurable and almost always
fatal. For these reasons, it is important to identify radon problems
and take sc ps to reduce your exposure before any illness occurs.
CAN I DO ANYI’HING IF I HAVE HIGH RADON LEVELS AT HOME?
If you find an elevated level in your home, there are many steps
>ou can take. Even very high radon concentrations can be lowered. In
fact, the extremely high levels that were found in several Clinton, NJ
homes have all been brought down to acceptable levels. On a short-term
basis, just keeping windows open will usually substantially lower the
radon level. But this is too expensive to continue during the heating
season. Other, more permanent methods are described in a booklet
entitled “Radon Reduction Methods: A I lomeowner’s Guide’ that is avail-
able from the New Jersey Department of Environmental Protection (DEl ’)
radon information line, 1-800-648-0394. The DEP also provides training
for contractors to help them gain expertise in radon reduction tech-
niques. You can call the information line to ask questions and to get
the names of companies offering radon measurement and reduction
services.
INTERPRETING YOUR TEST RESULT:
Using the imaginary test result we have given you, look at the left
column on the next page, under the heading ‘Radon Level, and find
the level nearest to your test result. Next look across to the
right-hand column under the heading, “Extra Cancer Deaths” and find
the number that is associated with your radon test result. This is
the extra risk of cancer from living in a home with that level ot
radon.
-------
Cancer Risk From Lifetime Radon Exposure
Extra
Radon
Canc.r
L.v.I
D..tha
(pCI/I)
(Out of
1000 p•epl.
100 500 In 1000
4 0 200 In 1000
2 0 100 In 1000
10 50 In 1000
4 20 1n1000
2 l OIn l 0 0 0
I SIn 1000
0.5 2.5 In 1000
0.1 0.5 In 1000
-------
Cancer Risk From Lifetime
Radon Exposure
Extra
Radon
Canc.r
Equivalent
Lev•i
O.atha
Smoking
Risk
(pCi/i)
(Out ol
1000 p.opl.
I ’
--
—
II
100
40
20
10
4
2
1
0.5
500 In lOOC
200 In 1000
100 In 100( —
50 In 1000
20 In 1000
10 In 1000
5 In 1000 —
2.5 In 1000
10 Psck.Idsy
.rr
2 Pecks/day
8 Cigarstt.s/day
2 Ctger.tt•.ld.y
0.1
0.5 In 1000
-------
CANCER
RISK FROM
LIFETIME
RADON EXPOSURE
500
0.1 0.5 1 2 4 10 20 40 100
EXTRA
CANCER
DEATHS
(Out of 1000
people)
200
100
50
20
5
10
Radon Livel (pCI/I)
-------
If your home was :ested ‘ . ith a charcoal or aipha-track detecwr.
:he results will probably be reported in units Otpicocuries per i:e
of air,’ anhreviated as pC i 1 I. This unit tells how much radon there is
in one liter of air.
IS IT DANGEROUS IF I FIND I HAVE HIGH LEVELS?
Radon is a serious risk. There is no level of exposure to radon in
the home that experts can agree is completely safe. The chances of
suffering harm from elevated radon levels are much greater than the
chances of suffering harm from most other pollution problems. As
Consumer Reports has stated, “While there are uncertainties in
pinpointing low-level radon risks, there is no doubt that the risks of
radon vastly exceed the risks from aflatoxin, PCBs, nuclear wastes, and
virn ally all other environmental hazards” (July, 1987, p. 442). The
disease caused by radon, lung cancer. is incurable and almost always
fatal. For these reasons, it is important to identify radon problems
and take steps to reduce your exposure before any illness occurs.
CAN! DO ANYThING IF I HAVE hiGh RADON LEVELS AT HOME?
If you find art elevated level in your home, there are many steps
you can take. Even very high radon concentrations can be lowered. In
fact, the extremely high levels that were found in several Clinton, NJ
homes have all been brought down to accepcabfe ‘evels. On a short-term
basis, just keeping windows open will usually substantially lower the
radon level. But this is too expensive to continue during the heating
season. Other, more permanent methods are described in a booklet
entitled “Radon Reduction Methods: A Homeowner’s Guide” that is avail-
able from the New Jersey Department of Environmental Protection (DEP)
radon information line, 1-800-648-0394. The DEE’ also provides training
for contractors to help them gain expertise in radon reduction tech-
niques. You can call the information line to ask questions and to get
the names of companies offering radon measurement and reduction
services.
INTERPRETING YOUR TEST RESULT:
The U. S. Environmental Protection Agency has evaluated the risk from
radon and has issued the following guideline which can help you interpret
your (imaginary) radon level:
At 4 pCI/I or above, homeowners should
take action to reduce their radon level.
-------
Cancer Risk From Lifetime Radon Exposure
Extra
Cancer
Equivalent
Level
O.aths
Smoking
Risk
Advice
(pCI /i)
(Out at
1000 p.o p1.
hrrri
iftE
100
40
20
10
4
2
I
0.5
500 In 100( —
200 In 1000
100 In lOOC —
50 In 1000
20 In 1000
10 In 1000
5 In 1000 —
2.5 In 1000
10 Psckaid.y
2 Packs/day
—05050005
• Clgar.tt..Iday
55
2 CIgar.ttss/day
4 pCi /I or above PA recQmmend
Ihat you reduce your radop ;evei. :
0.1 0.5 In 1000
-------
Advice for Radon Exposure
Radon
Advice
Level
(PCI/I)
HIGH TO VERY HIGH RADON LEVELS
Measurements near the upper end of this range are much hither than the EPA
action guideline. Exposure to such levels is very dangerous. ‘ or residents
living in homes at the higher end of this range, action should be taken within
— the ncxt couple of weeks to substantially reduce their exposure. If prompt
1 0 0 action is not possible or is not effective, they should consider moving until
the radon levels are reduced. Exposure to levels at the lower end olthe range
is also unsafe. Rcsidcnis living in homes at these levels should act to reduce
the readings within the next couple of months.
40
—
MODERATE TO HIGH RADON LEVELS
2 0 Measurements in this range are above the EPA action g iideline. Exposure tO these
Ievcls is a si nificant risk ii it extends over many years. Rccidcnts should
carc(ully evaluate the causes of thcir elevated lcvcls and make plans to reduce the
— levels permanently. To minimize the cumulative risk, this permanent action should
be completed in the next year or two. In ihc meantime, residents may want to avoid
1 o prolonged exposure to areas of the home where the levels arc highest. “
4 —
2
—
I —
0.5
0.1
Al 4 pC, or above. EPA re corn n e rids
you reduce your. adon eve .
LOW TO MODERATE RADON LEVELS
Measurements in this range fall below the EPA action guideline.
— Radon levels at the lower end of this range present a low health nsk. Radon
levels at the higher end of this range, extended over a lifetime, present j
moderate health risk. Any plan to lower the lcvels should he reiulIy
evaluated to be sure that ii is likcly to be effective, since it is often
difficult to reduce levels below this range. Many authorities do not recommend
trying to reduce levels in ibis range, especially for homes near the lower
end. Rcsidcnts who decide to try to reduce their levels below this range can
lake several years to act without .idding significantly to their risk.
VERY LOW TO LOW RADON LEVELS
Measurements in this range are no higher than the outdoor ‘background
— lcvel in many areas. Exposure to these lcvcls does not call tor action. Even
at thesc low cvcIs, there is a small risk associated with lifetime exposure to
radon. However, authorities . grcc that culorts to reduce radon lcvels still
further arc likely to he expensive and incifcctivc.
-------
-Ra cr
Cancer Risk From L;fet me Radon
Exposure
Extra
-
Radon
Cancer
Equivat
ant
Level
Deaths
Smo Irig
Risk
Advice
(pCi.i)
(Out of
1000 people
b1 J
HIGH TO VERY HIGH RADON LEVELS
Measurements near the upper end of this range arc much
higher than thc EPA action guidclinc Exposure to such
Icscls is vcry dangerous. For residents living in homc at
thc higher cnd olihis rangc. action should be takcn within
— thc ncxi couple of wccks to substantially reduce their
cspnsurc Ii prompt auiun is not possihic or is not
elleuivc, they should consider moving until the radon levels
.ire reduced Exposure to levels , the lower end of the
ranec is also unsafc. Rcsidcnts living in homes at these
levt 1c should act to rcducc the rcadings within the ncxs
couple of months.
MODERATE TO HIGH RADON LEVELS
Measurements in this range arc abo c the EPA action guide-
line Exposure to these levels is a signilicant risk if it
e tcnds over many vcars Residents should carefully cvaluate
he CaUSeS ol their elevated levcls and m.uke plans to reduce
thc levels pcrmanenily To mirumiLe the cumulative risk,
this permanent action should he completed in ihF nc year or
t so In the meantime. residcnts may want to avoid pro1on ed
e posurc In areas of the home where tli levels af highest.
cslq bov PA
that you reduce
LOW TO MODERATE RADON LEVELS
Measurements in this range fall below thc EPA action
guideline. Radon levels at the lower end of this range present
— a low health risk. Radon levels at thc higher end orihis
range, extended over a liktime, present a moderate health
risk. Any pl.in to lower the levels should bc carefully evalua-
ted I C ) be sure that is likely in he effective, since it is
uii n diflicult to reduce l vel below this range Many author-
ties do not recommend tryingto reduce levels in this range,
especially for homes near the lower end Residents who decide
to try to reducc their levels below this range can take several
‘.cars to act without adding significantly to their risk.
VERY LOW TO LOW RADON LEVELS
Measurcmcnts in this range arc no higher than the outdoor,
background level in ma y areas. Exposure to these levels
— does not call for action. Even at these low lcvcls, thcrc u.s a
small risk associated with lifetime exposure to radon.
However, authorities agree that efforts to reduce radon levels
still further arc likely to be expensive and inelIcctivc
iIi1
—
100
40
20
10
3
2
I
0.5
500 In 10CC —
200 In 1000
100 In 10CC —
50 In 1000
20 In 1000
10 In 1000
5 In 1000 —
2.5 In 1000
10 Packs/day
N i
2 Pecksiday
8 Clgar.tt.slday
2 Clgar.tt.slday —
0.1 0.5 In 1000
-------
Appendix B
Instruction Letters and Evaluation Questlonnaires
-------
Asbestos est_ rmaire ___________________
C.: ee - ::
C 3cx 23 New ‘.e .,e’ e,
Dear New Jersey Resident:
Thank you for talkir with us on the phone aid for agree ir to help
our pro)ect. At Rutgers we are tasti.ng different brochures to f .nd
effective ways of helping people lea about asbestos. The feedback
questionnaire will show us what u ression you get about asbestos fran the
brochure we sent you aid whether you find the brochure helpful.
There are no right or wrong answers. We ne i to get your reactions
aid your opinions to evaluate the Lnfor tion we sent. All your answers
will be kept can letely confidential. The ocde rn.miber on the fe hack
questionnai.re is only us to show us which questionnaires have been
return (so we don’t call aid reztu.rd people who have already mail back
their answers)
DIRECTIONS
Pretend that you have just had your house tested for asbestos.
The testing company tells you that you have a reading of
______ fibers per liter in your first floor and you are
trying to decide whether you should do anything about it.
Read the “Test Brochure” and then fill out the blue
questionnaire just as you would if you were a homeowner who
had received these results. FEEL FREE TO REFER TO THE
BROCHURE WHEN ANSWERING THE FEEDBACK QUESTIONS. When you
have finished, mail the questionnaire back to us in the
envelope that we have provided. The Test Brochure is yours
to keep.
ThANK YW V Y P R YC(JR P!
Sincerely,
(IT
Neil D. Weinstein, Associate Professor
Peter M. San nart, Professor
Nancy E. Roberts, Project Director
-------
‘E2ACY sT:: :R
Please feel free to Lock back at e broc rure wt en fil1jj out t s
qiest.oTir.a.ire.
* *
* Your in%agiziazy asbestcs *
* test result is: *
a a
* fibers/liter (f/i) *
* *
a Use this level when answerir questions *
* l i serious a probles you have. *
* *
aaa**aaaaaeaa*aa*e.*a**a.****a***********.**a**aea.aaa*
Overall, h i would you rate the br hure we sent you? (Please circle one
ans ier in each r i.)
1. very difficult fairly difficult fairly easy very easy
to understand to understand to understand to understand
2. didn’t help a little moderately very helpful
lYE understand helpful for helpful for for understanding
my test result unde.rstanding understandir my test result
my test result my test result
3. much too too little about right too much much too
little infor t ion information much
irtfor tation thf o r t ion
4. Did the brochure give you a good understandthg of the risk fran your
asbestos level?
I have a very good understanding of the risk
( J I have a fairly good understanding of the risk
( ) I feel fairly uncertain about the risk
( 3 I feel very uncertain about the risk
5. H likely do you think it is that oontinus exposure to your asbestos
level .ild eventually have harmful effects? (Even though you may feel
uncertain, please circle an answer to tell us what iii ress ion you got from
the brochure.)
no very %.t hkeIy nvderste likely very certain
chance .t (ikeIy chance likeLy to happen
-------
—,
T rase ;sz.on 5 d-er:y.
:- o r .s,
very sl . : r:sk
slight risk
( ) rrcderate risk
( ) serious risk
very serious risk
do you think you would feel about
(Please circle one answer in each row.)
7. not at all slightly n derately very extreii ly
ncerned ncerned concerned concerned concerned
8. not at all slightly rrcderately very extremely
frightened frightened frightened frightened frightened
9. Prom what you have read in the brochure, how difficult or easy does it seem to
reduce the asbestos i.n houses that have a problem?
very difficu.Lt
( ) fairly difficult
fairly easy
] very easy
10. Given your test result and the information in the brochure, do you think
you need to r ce the asbestos level;
( ] should take action even if it is hard to reduce the
asbestos level
( ] should take action onlY if it is relatively easy to reduce
the asbestos level
do not feel action is needed at my level
undecided
\ > Which way you be leartirq?
reduce the asbestos level
do not reduce the asbestos level
11. At what asbestos level (in your main living area) do you think you would
feel, satisfied, so that you ld not spend more money trying to get the level
even lower?
(]NOASB It (1
)O.Sfiber/l
( ]lfibers/l ()
t)2fibers/] . ()
c’-. .C4. Id :‘C r as e. ::s .Ci’
the asbestos level found by the test?
3 fibers/i
10 fibers/i
20 fibers/i
other —>What level’ -
-------
L-ag e at seve.i- i cf f .e.’xs -ad - eLr cres as:a2 r
as es: s, w .th t.-.e f l ç resu.l s f r .a._- area:
H e asbesots level
Pr e.rd Ln fthers/lite.r
A 0.8
B 1.5
C 2.6
0 4.0
E 9.0
F 24.0
G 50.0
a. In your opi.ruon, if it costs $500, which of these friends sho 1d try to
reduce their asbestos level? Below, put a check i.n the box for each
friend who ought to try to reduce his or her level. C eck as many as
you wish.
() () 1) (1 1] (3
Fri erd Friend Pr iend Fri end Frier Frier Frierd
A B C 0 E F C
b. What if it turned out that the cost was $2000?. Which of your friends
ought to spend this ney to reduce the asbestos level? C eck as many as
you wish.
(3 (] (3 (1 C) C) (3
Friend F’rierd Friend Friend Friend Friend Friend
A B C 0 E F G
13. What do you think are the chances that you uld eventually have sane
illness due to asbestos if you continued to live i.n the home with your
test result and didn’t do anything about tZie asbestos? (Please put a
check in the box that canes closest to your opinion.)
(3 (1 (3 (1 (3 (3 C l (1 (3 (1
no cPi cs I ChWIC• 1 ch I charc. certain
in 9 000 in 100 10 (100%) coinion
(0.1%) (1%) (10%)
-------
Fcr classificat n p pcses. please .e .1 • .s;
a. Your sex: ( ) ma 1 e
b. Your age: ______
fez le
c. H i iTu.ich school have you cc leted?
] sane elenientary school
ele nentary school
) sane high school
( ] finished high school
sane oollege
finished 2-year college
( 3 finished 4—year coll. e
sane graduate study
graduate degree
d. Prior
(
C
to receiving our brochure, h
3 very little
a en ll amount
rm.ach had you read about asbestos:
( 3 a ucderate an unt (at least
one information booklet or a
magazine article)
a lot
e. Do you have any i.nte.rest in having your real home checked for asbestos:
no
probably will not have it tested
ur ecided one way or the other
3 probably will have it tested
( 3 definitely will have it tested
already had an asbestos test in my home
f. At present I live in a:
3 single family house
3 rT .ltifa1TLily building
g. Do you rent or am your current residence?
(]rent (3am
ThANK YW FOR YWR Tfl€ AND HELP! If you want a copy of sane of the other
brochures we’re testing, please fill out the mailing label ar include it in the
envelope when you mail back the questionnaire.
-------
RjT 1 ER 4 S
: :.e • —e’ :
- . . .e .- - ‘ e 2::
New Jersey Resident:
Thank you for talkirg with us on the phone and for agreeing to help
our project. At Rutgers we are testing different brochures to find
effective ways of helping people learn about radon. The feer1hack
questionnaire will show us what in ression you get about radon from the
brochure we sent you and whether you find the brochure helpful.
There are no right or wrong answers. We need to get vou.r reactions
and your opinions to evaluate the information we sent. All your answers
will be kept ccii letely ccnfidential. The code n m ber on the fe rlhack
questi.onnai.re is only used to show us which questionnaires have been
returned (so we don’t call and remind people who have already mailed back
their answers).
DIRECTIONS
Pretend that you have Just had your house tested for radon.
The testing company tells you that you have a reading of
picocuries per liter in your first floor and you are
trying to decide whether you should do anything about it.
Read the “Test Brochure” and then fill out the blue
questionnaire just as you would if you were a homeowner who had
received these results. FEEL FREE TO REFER TO THE BROCHURE
WHEN ANSWERING THE FEEDBACK QUESTIONS. When you have finished,
mail the questionnaire back to us in the envelope that we have
provided. The Test Brochure is yours to keep.
ThANK Y J V Y MUC! FOR YCUR 1 P!
Sthcerely,
/ ;.? U
Neil D. Wethstein, Associate Professor
Peter M. Sar ran, Professor
Nancy E. Roberts, Project Director
-------
Pl a.se feel free to lock back at the br chu.re wt en filling out the quest .or..r.a..:e
a Your imaginary radon a
a test result is: *
a a
a pioocuries/liter (pci/I.) a
a *
* Use this level when an.swerthg questions a
a about ) w serious a probl€!n you have. *
a a
Overall, how would you rate the brochure we sent you? (Please circle one
answer in each row.)
1. very difficult fairly difficult fairly easy . very easy
to understand to urx erstand to understand to understand
2. didn’t help a little i derately very helpful
me understand helpful for helpful for for undørstanding
my test result understand ing understanding my test result
my test result my test result
3. much too too little about right too much much too
little information information much
i nfor t ion information
4. Did the brochure give you a good understanding of the risk from your radon
level?
( ) I have a very good understanding of the risk
( ] I have a fairly good understanding of the risk
( ) I feel fairly uncertain about the risk
( ] I feel very uncertain about the risk
5. How likely do you think it is that oontinued exposure to your radon level
would eventually have harmful effects? (Even though you may feel
uncertain, please circle an answer to tell, us what i ression you got from
the brochure.)
very tiilikely amdera e likely very certa in
chance ullikely chance likely o happen
-------
5. T e - se ct.e.s:. ..:r. 5 f erer: , ;c ‘/ j .C raccr vej’
no ‘. .s,
very s1. r t :. .sk
sl .ght risk
e.rate r:sk
serious risk
very serious risk
How do you think ya.i would feel about the radon level four by the test?
(Please circle one answer in each row.)
7. not at all slightly moderately very extr e1y
concerned concerned concerned concernet concerned
8. not at all slightly, moderately very extrenely
frightened frightened frightened frightened frightened
9. From what you have read in the brochure, how difficult/easy do you think it is
to reduce the radon level in houses that have a problem?
very difficult
fairly difficult
( ] fairly easy
very easy
10. Given your test results ar the thforTI tion in the brochure, do you think you
need to reduce the radon level?
( ] should take action even if it is hard to reduce the radon level
( ) should take action only if it is relatively easy to reduce the
radon level
( ] do not feel action is needed at my level
urxiecided
\ > Which way ld you be leaning?
( ] reduce the radon level.
( 3 do not reduce the radon level
11. At what radon level (th your main living area) do you think you ild feel
satisfied, so that you ld not sper more money trying to get the level
even ] mr?
3 NO RAC J ( 3 4 pico iries/l
0.5 picocurie/l ( ] 10 pico .iries/l
3 2. pico iries/l ( 3 20 pioocuries/l
( 3 2 picocuries/l ( 3 other—> What level’
-------
:2. — . g.-e that several cf c fr . .eri .s r c -e ..r cnes :e.s-z .
t. e fo 1:. c.i res-u. t.s f r .- e . r a .rt I ....-ç a.r a:
H e radon level
Friend L’3 o cocur.es/11ter
A 0.3
B 2.0
C 3.6
D 5.0
E 9.0
F 24.0
C 50.0
a. In your opinion, if it costs $500, which of these frierds should try to
reduce their radon level? Below, p t a check in the box for each friend
who ought to try to reduce his or her level. O eck as n ny as you wish.
C) C) (3 (3 (1 (3 C]
Friend Friend F’riend Friend F’riend Friend Friend
A B C D E F C
b. What if it turned out that the cost was $2000? Wh.ich of your friends
ought to spend this n ney to reduce the radon level? Q eck as n ny as
you wish.
C] C) C] C) C] C] C]
Friend Friend Friend Friend Friend Friend Friend
A B C D E F C
13. What do you think are the chances that you uld eventually have sane
illness due to radon if you continued to live in the hane with your test
result and didn’t do anythi.rq about the radon? (Please pit a check in
the box that cl est to your opinion.)
(3 (3 (3 (1 (3 (3 (3 (3 (3 (3
rio charics I charce I chance I ctiwice cer ein rio
in 1,000 in 300 in 10 (100%) op,n,an
(0.1%) (1%) (1O )
-------
F r class f1.at .cn ses, p: se :: ‘ :
a. tour sex:
le
fer.ale
b. Your age:
c. How much school have you T 1eted?
sare elementary school
finished elementary school
sare high school
finished high school
sate oo1l e
finished 2-year colleqe
finished 4-year col l e
sare graduate study
graduate deree
d. Prior
to receiving our brochure, how much
very little ( I
ll amount
had you read about radon:
u derate ancunt (at least ore
i.nforTnation boolUet or a
magazi.ne article)
a lot
e. Do you have any interest in testing your real home for radon:
no
probably will not have it tested
ur ecided one way or the other
probably will have it tested
definitely will have it tested
already had a radon test in rirj home
f. At present I live in a:
single family house
multifazn.ily building
g. Do you rent or own your current residence?
(]rent ()own
ThANK YCU F R YCUR TI7€ AND HELP! If you want a copy of sate of the other
brochures we’re testing, please fill out the mailirq label ar include it in
the envelope when you mail back the questionnaire.
-------
APPENDIX C
STATISTICAL ANALYSES
STANDARD STATISTICAL MODEL
The statistical procedure used most often in this research
employed the method of least squares to fit response data to a
linear model (Proc GLM, SAS Institute, 1987). The independent
variables were HAZARD (asbestos or radon; 1 df 1 ), FORMAT (Fl
to F7; 6 df), LEVEL (1st, 2nd, 3rd, or 4th; 3 df), and all
interactions. This model was used when variations with level
were not a major focus, so the spacing among the different
levels was not a consideration and level could be treated as a
class or categorical variable. If initial calculations revealed
FORMAT effects or interactions with FORMAT, the differences
among the formats were examined further with Tukey’s studentized
range test. Tukey’s test controls the experimentwise error
rate.
In the rest of Appendix C this analysis will be called the
“standard model.” The standard model permits tests of the
following effects:
Hazard main effect (indicates overall differences between
asbestos and radon means)
Format main effect (indicates differences among mean
responses to the seven different formats)
Hazard x format interaction (indicates whether differences
in response to asbestos and radon vary with format)
Level (indicates whether responses vary with exposure level)
Hazard x level interaction (indicates whether changes with
level are different for asbestos than for radon)
Format x level interaction (indicates whether changes with
level are different for some formats than for others)
Hazard x format x loalevel interaction (indicates whether
the effects of format on changes with exposure level are
different for asbestos than for radon)
MEAN RESPONSES FOR DIFFERENT FORMATS
Many differences among formats are apparent in the figures
that have already been presented, but the figures do not reveal
1 df = degrees of freedom.
-------
f- - I
— C.
Table Ci
Means for Key Variables by Format
(Averages Over Hazard and Test Result)
Forost
Resp
1
2
3 4
5
6
7
Likelihood of illness Cl tO 9 sCale)
4.09k
3.95
6.12w 5 9 b
6.27
6 ’ 1 a
27
Perceived saf. level ( I/l or f/I)
1.46
91 b
2 • 09 b 1.16•
172 ab
139 ab
1 59 ab
Neon of levels •t diicli fri.,
sh .,ld sct (pCI/I or f/I) (level
for frte at $500 cost • at
$2000 cost)
9.O6
1O.S5
9 . 34 k
7.a5
7 • 30 b
Threat (4 to 23 scale)
12.7
12.3
12.5 15 °b
l2.8
• 13 °a
12.5
Mitigation plane Ct to 4 scale)
2 • 76 th
2.59k
2.7S 3 01 b
2 • 74 aI
2 58 a
2 53 a
Difficulty (I to 4 scale)
3.48
3.53
3.5l 3.51
3 49 a
3.50k
Helpfulness (1 to 4 scale)
3.I9
346 b
3.3l 3.26
3.33
3.4%
MioIglt of inforiMt ion (1 to 5 scale)
2.8l
2.8S
2.36 2.8l
2 87 a
2 92 a
Uncertainty Cl to 4 scale)
1.81•
1 2 ab 73 .b
‘ 0 ab
1.621 ,
L 61 b
Note: Neane sharing a ca n s cript are mt si ,ficontIy different.
-------
C3
wnich differences are significant. Table Cl c3nta1 s, fo key
response variables, meafl responses averaged across hazards ard
levels. The table shows which formats tended to produce higher
or lower responses on these variables regardless of the hazard
or test result. For example, the risk aversion produced by
format F4 is evident in many rows of the table.
MEASURES OF CO!*IUNICATION SUCCESS
The following subsections present examinations of
comprehension, acceptance, consistency, uniformity, audience
evaluation, and other variables. Information bearing upon the
tendency of certain formats to produce particularly risk-averse
or risk—tolerant reactic ns arises in the analyses of
comprehension, acceptance, and consistency and can be found in
those subsections.
Com rehens ion
The criteria in Table C2 were used to decide what risk
probability estimates by subjects constituted errors and how
large those errors were. Analyses were based on scale values (1
to 9 scale) rather than on the probabilities themselves. If
probabilities had been chosen, the results would have been
dominated by errors at the higher risk levels where the scale
intervals were much larger. Errors made at lower risk levei,s
would have made little difference.
Several types of analyses were carried out. Figure 2
indicates the fraction of responses in that were in error (i.e.,
entries different from 0 in Table C2). Chi-square analyses
confirmed the impression given by this figure that error rates
were higher for F4 and P6, chi-square (6) — 76.5, p < .0001, and
chi-square (6) 49.5, p c .0001, for asbestos and radon,
respectively. Figure 3 depicts the mean error size (treating
correct answers as having errors of size zero), based on the
absolute value of the entries in Table C2. Statistical analysis
of this variable was carried out with the standard model. There
was a significant FORMAT effect, F(6, 1763) = 14.02, p < .0001,
and a significant HAZARD x FORMAT interaction, F(6, 1763) =
2.87, p < .01. The interactions are apparent in Figure 3, but
they do not affect the main finding that formats P4 and F6 have
the largest errors. This is not surprising since F4 and F6 do
not contain information about risk probabilities. The remaining
formats do not differ significantly from one another.
The results in Table Cl (“Likelihood of illness”) were based
on the untransformed (original 1 to 9) scale choices on the risk
probability question. These were analyzed using the standard
model. There were significant effects of HAZARD, F(l, 1763) =
70.67, < .0001, and FORMAT, F(6, 1763) — 8.08, < .0001, and
a significant HAZARD x FORMAT interaction, F(6, 1763) — 4.16,
< .001. Radon was given higher risk probability estimates than
-------
C4
Table C2
Designated Errors for Different
Risk Probability Choicesa
Test result
(no
] .
chance)
Response
2 3
Choice (scale
4 5
units)
6
7
8
9
(certain)
0.8
0
0
1
Asbestos
2 3
4
5
6
7
2.5
0
0
0
1. 2
3
4
5
6
3.5
—1
0
0
0 1
2
3
4
5
24.0
—2
—1
0
0 0
1
2
3
4
Radon
0.8
3.5
—2
—4
—1
—3
0
—2
0 1
—1 0
2
0
3
1
4
2
, 5
p
3
4.5
—4
—3
—2
—1 0
0
1
2
3
24.0
—6
—5
—4
—3 —2
—1
0
0
0
Entries
md
icate the
size
of
scale
the error (in units).
Responses were regarded as correct if they fell within one full
scale unit of the information given in the brochures that
contained risk probabilities.
-------
as estos (scale means of 4.78 and 3.99, respectively). seer
F gure 4 and Table Cl, format F4 produced sign ficar.tiy
higher risk probability estimates than all other formats.
Furthermore, the difference between radon and asbestos was
significantly larger for all the formats that contained risk
probability data than for formats F4 arid F6 that did not.
Figure 5 is based directly on the entries in Table C2,
taking signs into account.
Acceptance of Recommendations
Mitigation Dlans . The relationship between recommendations
and stated intentions was analyzed according to a procedure
developed by Hildebrand a 2 d colleagues (Hildebrand, Laing, &
Rosenthal, 1977a, 1977b). The size of the correlation
between plans and levels (or the size of ordinal measures of
association such as Sonier’s D, Gamma, and Kendall’s tau) is not
suitable to test acceptance. The issue is not Just whether
intentions to act increase as levels increase, but whether plans
to act occur at the particular level where action is
recommended. A strong correlation can exist even if people plan
much more action (or less action) than recommended.
To carry out these calculations, one has to decide, for each
level, what response or responses represent an acceptance of the
action recommendations. Given the test results employed in ,this
investigation, several different options were possible. We
decided to look at the performance of the formats for three
different sets of criteria.
These criteria are set out in Table C3. The “dichotomous”
pattern represents a rather simpleminded following of the action
advice. It assumes that people are accepting advice if they say
that they will take action (if easy or if hard) for any test
result above the action guideline, and that people are following
advice when they say they will not take action, even if action
is easy, for any test below the guideline. The “diagonal”
pattern implies that action is not worthwhile at the lowest test
result and that action should be taken even if difficult at the
highest level. At levels just above and just below the
guideline, however, responses of “undecided” and “act if easy”
are both regarded as consistent with recommendations. The
final, “complex” pattern suggests that all responses of
“undecided” represent communication failures. Furthermore,
sinc, the brochures with advice or risk probabilities indicated
that some risk remains at even the lowest levels, both “no
action” and “act if easy” are regarded as appropriate responses
at the 0.8 test results. People assigned to test results just
2 We are grateful to David Hildebrand for providing us with
a computer program for carrying out these calculations.
-------
Table C3
Criteria for Acceptance
of Mitigation Advice
Act even if hard
Act if easy
Undecided
Don’t act
Act even if hard
Act if easy
Undecided
Don’t act
Act even if hard
Act if easy
Undecided
Don’t act
Test Results
0.8 2.5 3.5
0.8 3.5 4.5
Dichotomous
0 0 +
0 0 +
0 0 0
+ + 0
Diagonal
0 0 0
0 + +
0 + +
+ 0 0
Complex
0 0 +
+ + +
0 0 0
+ + 0
+
+
0
0
+
0
0
0
+
0
0
0
Response
Asbestos
Radon
24
24
Note: A plus indicates acceptance of mitigation
advice, a zero indicates failure to follow advice.
-------
below tne action guideline are considered correct .f they c.iose
either “no action” or “act if easy.” Just above the act n
guideline, “act if easy” and “act even if hard” are both
accepted as consistent with the advice, but at the highest
level, “act even if hard” is the only response considered
appropriate.
The method of Hildebrand et al . also allows one to consider
a response to be partially correct, but we did not make use of
this flexibility. The statistical procedure yields a measure of
agreement, Delta. The distribution of Delta is asymptotic
normal, and a computer program produces the significance level
of Delta and the asymptotic standard error, in addition to Delta
itself. Values of Delta are portrayed in Figure 6 for each of
the criterion patterns. Delta and its standard error are
reported in Table C4.
The values of Delta for different formats were compared by
extending the method described by Fliess (1981, Chapter 10).
Delta divided by the square of its standard error is distributed
as a chi—square variable with one degree of freedom. Adding
these 14 terms together (7 formats for each of the two hazards)
yields a chi—square variable with 3.4 degrees of freedom. This
can be partitioned into two new chi—square variables. One
variable, analogous to the mean of the 14 terms, can be used to
determine whether the Deltas are different from zero. This is
obviously correct and the test is highly significant. The,,
second new variable is of greater interest. It tests whether
the 14 Deltas differ significantly front one another. Regardless
of the criterion used, the Deltas do turn out to be different,
p’s < .0001. -
Differences among the Deltas can then be compared with
planned contrasts that have one degree of freedom. The results
are shown in Table C5. According to the first and third rows of
the table, a standard improved acceptance by all three
criteria. In the presence of a standard, advice further
improved acceptance according to two of the three criteria
used. Finally, in the presence of a standard, probabilities and
risk comparisons did not improve acceptance. (It is also clear
from formats Fl and F2 in Figure 6 that the addition of risk
comparisons to risk probabilities did not significantly improve
acceptance of mitigation advice).
Disagreements with action advice were also examined
separately for test results above and below the asbestos and
radon guidelines. At levels below the guidelines, the response
of “no action needed” was considered correct. At levels above
the guidelines the responses “act if easy” and “act even if
hard” were considered correct. (These choices are equivalent to
the “dichotomous” criterion.) Here the outcome variable has
only two values (either “accepted” or “disagreed”). The
independent variables were hazard and format. These data were
analyzed by a linear model suitable for categorical dependent
variables (Proc CATMOD, SAS Institute, 1987).
-------
Table C4
Measures of Acceptance for
Different Criterion Patterns
Format
Dichotomy
Delta ASE
Criterion Pat
Diagonal
Delta
tern
ASE
Complex
Delta
ASE
Asbestos
Fl
.069 .050
.076
.060
.118
.060
F2
.096 .053
.100
.062
.180
.054
F3
.098 .046
.144
.058
.163
.053
F4
.212 .050
.212
.057
.427
.055
F5
.176 .052
.224
.060
.260
.059
F6
.287 .054
.359
.055
.366
.059
F7
.289 .053
.292
.060
.320
.059
Radon
Fl
.157 .057
.062
.070
.200
.065
F2
.138 .057
.178
.069
.192
.068
F3
.195 .062
.192
.070
.276
.078
F4
.224 .050
.222
.064
.311
.063
F5
.282 .054
.275
.065
.356
.065
F6
.358 .061
.373
.066
.359
.066
F7
.432 .060
.346
.066
.457
.066
Note: ASE = asymptotic standard error.
-------
Table C5
Differences in Acceptance of Mitigation
Advice for Different Sets of Formats
Criterion of Acceptance
Contrast Dichotomy Diagonal Complex
Standard vs. no standard < .0001 < .0001 < .0001
(F1—F3 vs. F4—F7)
Advice vs. no advice 2 < .001 g < .01
(F4—F5 vs. F6—F7)
No standard vs. standard •g < .0001 < .005 p < .0001
but no advice (F1-F
vs. F4—F5)
Probabilities and risk p > .2 p > .2 p > .2
comparisons vs. none
(F4 & F6 vs. F5 & F7)
-------
The analysis of the disagreement rate at low levels
(. verreaction) fcund a significant effect for FORMAT (ch -s a
= 45.as, df = 6, < .0001) and no HAZARD effect or HAZARD x
FCRMAT interaction. Post-hoc contrasts showed that the FORMAT
effect was due to the fact that the overreaction rate was
smaller for the two formats with advice, F6 and F7, than for the
other formats. The contrasts of F6-F7 with Fl-F3 and with F4-F5
were highly significant (chi-square = 17.36, 2 < .001, and
chi—square = 22.13, 2 < .0001, respectively). The Contrast of
Fl—F3 with F4-F5 was not significant, p > .2.
The analysis of the disagreement rate at high levels
(underreaction) presented a somewhat different picture. There
was a significant HAZARD effect (chi—square 10.66, df = 1, p =
.001), indicating that asbestos underreactions were more
frequent than radon underreactions. There was a significant
FORMAT effect (chi-square 18.49, 2 .005). Again the FORMAT
x HAZARD interaction was not significant. Post—hoc contrasts
revealed that formats F4-F5 led to fewer underreactions than
Fl—F] (chi—square = 10.59, 2 — .001) and also fewer than F6—F7
(chi-square = 3.83, 2 = .05).
Acce tab1e exposure levels . Two variables were examined:
the level at which respondents said that they would feel safe
and the composite measure of risk aversion formed from the two
questions that referred to levels at which friends should spend
$500 or $2000 to remediate (Figures 9 and 10). The standa d
model was used for both.
For the first measure, there was no main effect for HAZARD,
> .15, a barely significant FORMAT effect, F(6, 1806) 2.57,
p < .02, and a significant HAZARD x FORMAT interaction, F(6,
1806) 3.05, c .01. Although format F4 provoked the greatest
risk aversion overall, the level deemed acceptable for F4 was
significantly lower than for just two other formats, F2 and F].
Across hazards, the level deemed acceptable was considerably
below the action guideline. The interpretation of the
interaction term is clear from Figure 9. When no standard was
provided (formats F1—F3), people were correct in showing more
risk aversion with the higher radon risks than with the lower
asbestos risks, PCi, 772) = 8.82, 2 < .003. But with a standard
that gave the same advice for a radon test result (in pci/i) as
for the same asbestos test result (in f/i), in spite of the
difference in risk, the radon-asbestos difference was
eliminated, F(i, 1034) 3.05, p .08. Calculations showed
that this between—hazard contrast of F1-F3 with F4-F7 was
significant, F(l, 1845) = 14.16, p = .0002. However, risk
probabilities and smoking comparisons in the presence of a
standard (P5 and P7) did not help people see the difference in
risk between the two hazards, > .15. Also, adding smoking
comparisons did not increase the radon—asbestos difference
beyond that resulting from the probabilities alone, p > .2.
-------
There was also a very strong effect of LEVEL, witn
acceptable levels cli rwing with the test result the subject
received, F(3, 1806) = 5.86, 2 < .001. The mean values of tne
highest acceptable level (averaged across hazard and format)
were 2.51, 2.77, 2.76, and 3.03 for people with test results far
below the guideline, just below, just above, and far above,
respectively.
For the second measure, the two levels at which friends
should act (under different cost assumptions) were averaged
together. Neither the HAZARD effect nor the HAZARD x FORMAT
interaction was significant, F < 1, and F(6, 1844) = 1.85,
.08, respectively. The FORMAT effect was significant, F(6,
1844) 6.67, 2 < .0001. Post hoc comparisons showed that
people who received format F4 were most risk-averse, with the
mean acceptable levels significantly, lower for F4 than for Fl,
F2, F3, or F7. There was also a very large LEVEL effect, F(3,
1844) = 31.77, 2 < .0001. The lowest recommended action levels
(averaged over the the two questions and averaged over hazard
and format) were 6.7, 6.8, 7.8, and 12.1 for people with test
results far below the guidelines, just below, just above, and
far above, respectively.
Consistency Between Hazards and Between Levels within Hazards
Acceotable levels of exoosure . The sensitivity of these
variables to hazard, an issue of consistency, has already been
described under the AcceDtance of Recommendations heading.
Perceived threat and mitiaation Dlans . In these
calculations we are interested in determining whether perceived
threat and action plans vary more with exposure level for some
formats than for others and whether the difference between radon
and asbestos risks is better communicated by some formats than
by others.
To fully consider the relationship between responses and
exposure levels, it is not enough to consider the four test
results as simply four different categories; the spacing of the
test results is also relevant. Yet what measure of spacing to
use is not clear—cut. If the measure selected reflects the
format feature to which people are responding, we would expect
that for those formats without a standard or advice (F1-F3),
r.sponses would vary smoothly with level. (The addition of a
standard might be expected to produce a discontinuity or step at
the standard.) For formats Fl, P2, and F3, in other words,
there should be no sudden increases or decreases in the slopes
of the three line segments connecting the four test results.
The magnitude of the slope itself would then constitute a
measure of the success of the format in helping people
distinguish among the test results.
Among the variables that might be used to reflect the
spacing, the following are most plausible as capturing the
dimension to which people are reacting.
-------
C12
The actual test result in f/l or pCi/i. This variacie ...s
proportional tQ t e actual risk. Figures Cl and C2,
however, suggest that people are not responding in a lir ear
or simple nonlinear way t3 the concentration reported in the
test result, even for Fl through F).
The logarithm of the test result . Since the ladder of
exposures is essentially logarithmic, it seemed likely that
people would respond approximately linearly to this
variable. Examining the results for formats Fl through F3
in Figure 13 suggests that this is correct for perceived
threat (all four points for each format fall along a
straight line). However, it is hard to understand why
Figure 16 shows a sudden rise in action plans for Fl through
F3 when passing the guidance level, since none of these
formats mention the guidance level. (According to Figure
14, the jump is attributable to the radon subjects.
However, evidence described in footnote 5 suggests that
subjects in this study were not aware of the radon action
guideline.)
Physical location (inches alona the risk ladder . Since the
exposure ladders used are logarithmic, an actual distance
measure would not be distinguishable from the logarithm of
the test result.
Runas alona the exposure ladder . Using this variable, the
four test results would correspond to rungs 3, 5, 6, and 8,
as shown in Figures C3 and C4. Although this also seems
like a possible determinant of responses, the decline in th&
slope of perceived threat between the middle two test
results (Figure C3) for formats F2 and F3 (and also F5) is
difficult to explain. Using ladder rung as the independent
variable seems to separate the middle two test results more
than study participants actually did when forming
impressions of threat. On the other hand, with formats Fl
through F3, a logarithm created a step for radon action
plans around the action guideline (see Figure 14), and this
is is essentially eliminated when the ladder rungs are the
ordinate of the graph.
Of course, there really is no ladder for formats F3 and F4,
so the rung approach is applicable only to certain types of
ceamunication approaches. Consequently, we will use the
logarithm of the test result, realizing that the step in
mitigation plans at the radon action guideline for formats Fl
through F3 has not been adequately explained.
The three degrees of freedom corresponding to the four test
results will be partitioned in the following manner. One degree
of freedom will represent the logarithm of the test result
(LOGLEVEL). This is a numerical, rather than categorical,
variable, and the effects due to this variable in the final
statistical model reflect the slopes of the regression lines
linking level to response. In other words, this variable
-------
Figure Cl
Effect of Format and Assigned
Level on Perceived Threat
Test Result (f/I
or pCi/I)
(BY TEST RESULT)
1
1
1
1
11
a
w
-c
w
1 )
0
9.
0 5 10 15 20
2
Number3 - Numbers+ F-ilstogrcm - Standard
smoking
(Fl) (F2) (F3) (FLI)
Std.+nos.+ - Standard+ Std.+nos.+
smoking advice smoking-I-
advice
(Fc) (F6) (F7)
-------
Figure C2
Effect of Format and Assigned
Level on Mitigation Plans
act even (BY TEST RESULT)
3.
act if
easy
C
0
a
undecided 2
1.
not
needed 0 10 15 20
Test Result (f/I or pCi/I)
Numbers ‘ - Numbers+ Histogram Standard
smoking
(Fl) (F2) (F3) (F )
Std.+nos.+ - Standard+ a- Std.+nos.+
smoking advice smoking+
advice
(Fc (FFI’ (F7)
-------
Effect of Format and Assigned
Level on Perceived Threat
Figure 03
1 9
(sv RUNG ON
LA DD ER)
I
1
1
11
a
I -
-c
I-
-a
2
U
0
7
4
Runq on Risk Ladder
7
8
Numbers Numbers+ -“ Histogram -.‘ Standard
smoking
(Fl) (F2) (F3) (F4)
Std.+nos+ - Stondard+ Std.+nos+
smoking advice smoking+
advice
(F5) (F6) FF7)
-------
015
Figure
Effect of Format
Level
04
and Assigned
Plans
on Mitigation
oct even
if hard
(BY RUNG ON LADDER)
3.
3.
2.5
act if
easy
C
r S
=0
undecided
not
needed
1
1
4
- Numbers
Rung
5
on Risk Ladder
Histogram
(Fl)
Std.+nos.
+
smoking
Numbers- i -
smoking
(F2)
- Standard+
advice
- Standard
(F3)
Std.+nos.+
smoking+a
dvice
(F’4)
(F5)
(F6) (F7)
-------
1
L. I
shows how well people are able to distinguish the different
levels of r .sk. A variable with the second degree of freedc i
distinguishes between the first two levels and the last two
levels. This dichotomous variable, STEP, indicates whether
there is a sudden jump in threat perception or action plans as
one goes from below the action guideline to above the
guideline. Finally, the third degree of freedom is assigned to
the interaction of the preceding two variables. A significant
interaction indicates that the slope is different above the
action guideline than below the guideline. Because LOGLEVEL and
STEP are very strongly correlated, .76, it is difficult to
separate slopes from steps and the order in which the two
variables enter any statistical model has to be carefully
considered. The effect of STEP should be assessed both before
and after LOGLEVEL is entered to help clarify what effects are
present.
The full model contains HAZARD, FORMAT, LIOGLEVEL, STEP, and
all interactions. The results of the analyses (with LOGLEVEL
preceding STEP) are shown in Table C6. These are the Type I (or
sequential) sums of squares, used because LOGLEVEL is a
numerical variable and coding effects make it important to enter
it into the model after HAZARD, FORMAT, and their interaction.
Also, Type I sums of squares are used because of the correlation
between LOGLEVEL and STEP. Following the criterion established
earlier, F values are included only for effects significant at
or beyond the .01 level.
Perceived threat . The mean perceived threat is 12.5 for
asbestos and 13.6 for radon (see Figure 12). Although in the
proper direction and consistent across subjects (as indicated by
the statistical significance of the first term of Table C6), the
between-hazard difference is very small compared to the 25-fold
difference in actual risk. Overall, the brochures tended to be
inadequate in communicating the difference in risk between the
two hazards. Furthermore, the third line of the table (and
Figure 12) indicates that all formats were equally poor in
conveying this information.
The significant FORMAT effect is visible in Figures 1]. and
13. According to Table Cl, format F4 yielded significantly
higher threat perceptions than any of the others.
The significance of the STEP effect, when entered after
LOGLEVEL, shows that there was a significant jump in perceived
threat at the action guideline. Furthermore, the significant
STEP x FORMAT interaction indicates that some steps were higher
than others. Post—hoc comparisons revealed that the step was
larger for formats with a standard than for those without a
standard, F(l, 1805) — 11.15, < .001. Adding advice to the
standard did not increase the size of the step, > .17.
However, format F5 had a significantly lower step than the other
formats with a standard, F(1, 1805) = 3.95, < .05.
-------
Table C6
Analysis of Consistency in Perceived
Threat and Action Plans
Source Perceived
Threat
Action
Plans
F
2
F
2
HAZARD 43.82 .0001 8.53 .0045
FORMAT 17.98 .0001 6.38 .0001
HAZARD x FORMAT ns ns
LOGLEVEL 640. .0001 316.5 .0001
LOGLEVEL x HAZARD ns 6.83 .009
LOGLEVEL x FORMAT 9.55 .0001 6.64 .0001
LOGLEVEL x HAZARD x FORMAT ns ns
STEP 13.91 .0002 24.02 .0001
STEP x HAZARD nS ns
STEP x FORMAT 2.92 .008 ns
STEP x HAZARD x FORMAT ns ns
STEP x LOGLEVEL flS 11.85 .0006
STEP x LOGLEVEL x HAZARD ns 5.81 .002
STEP x LOGLEVEL x FORMAT ns
STEP x LOGLEVEL x HAZARD x FORMAT ns ns
-------
c:9
It is important to note that if the calculations are
repeated with STEP appearing before LOGLEVEL, we find that
LOGLEVEL still is a powerful determinant of perceived threat,
< .0001. However, the LOCLEVEL x FORMAT interaction that
appears strong in Table C6 is no longer significant. This
suggests that once different steps are introduced for different
formats, the slopes of the different formats (away from the
standard) are really the same. When the variables are analyzed
in this sequence (STEP then LOGLEVEL), it appears that no format
is superior in helping people distinguish the different threat
at different test results. Those containing a standard have a
greater variation in response as one goes from the lowest to the
highest test results, but only because of the artificial
increase in perceived threat as one passes the action guideline.
The simplest model that accounts for the threat data
explains 32.7% of the variance and contains the following
terms: HAZARD, FORMAT, LOGLEVEL, STEP, and STEP x FORMAT. All
terms are highly significant, ‘s < .0001.
The absence of significant LOGLEVEL x HAZARD X FORMAT and
STEP x HAZARD x FORMAT interactions is the reason why Figures
13, Cl and C3 do not distinguish between asbestos and radon.
Except for a difference in overall means, the relationships with
different formats and levels are the same for the two hazards.
Mitigation Dlans . The means for action plans are 2.65 ‘and
2.79 for asbestos and radon, respectively. This is a very small
difference considering the 25-fold difference in risk at the
same test results. Furthermore, the absence of a significant
HAZARD x FORMAT interaction (in Table C6 and in Figure 14) shows
that all formats are equally ineffective in conveying the real
risk differences. Figure 15 and Table 2 show that format F4
leads to the greatest risk aversion (for both hazards), though
it is significantly different only from F2, F6, and F7.
Both LOGLEVEL and STEP are significant, regardless of the
order in which they enter the prediction model. There is also a
significant LOGLEVEL x STEP interaction. These results show
that action plans increase with test results, that there is a
jump in action plans at the action guideline, and that
(independent of the jump) the slope is smaller above the action
guideline than it is below (as if a ceiling effect were
operating). This pattern is visible in Figure 16.
The significance of the interactions between test result and
format depends totally on the order in which the terms enter the
prediction equation. The analysis cannot determine whether the
different formats have different slopes or have different size
steps or both. If LOGLEVEL x FORMAT enters first, it is highly
significant, F(6, 1921) 6.70, < .0001, and the STEP x FORMAT
term is not significant. But if STEP x FORMAT appears first, it
is highly significant, F(6, 1921) — 11.20, < .0001, and the
LOGLEVEL x FORMAT interaction is not significant.
-------
---‘
Post-hoc cor parisonS are difficult to 2.nterpret. if a cde.
contains both the LOGLEVEL x FORMAT and STEP x FORMAT terms,
neither is sigrti.ficant. if LOGLEVEL x FORMAT .s left out, sloo
variations appear to be differences in step (i.e., if there sa
positive slope, the two points below the action guideline are
lower than the two points above the action guideline, suggesting
a positive “step” as one moves past the guideline). If STEP x
FORMAT is left out, step variations become differences in
slope. However, post—hoc comparisons reveal the same
differences among formats, regardless of whether interactions
with steps or slopes are omitted. Formats with a standard
(Fl—F4) have larger steps than formats without a standard, F(l,
1921) — 34.22, < .0001. The step is also slightly larger for
formats with advice and a standard, F6 and F l, than for formats
with a standard but no advice, F4 and F5, F(l, 1921) = 3.86, =
.05. Similarly, if a model considers slopes ahead of steps, the
slopes are greater for F4-F7 than for F1-F3 and are slightly
greater for F6 and F7 than for F4 and P5.
The simplest model that accounts for the action plans data
explains 18.7% of the variance and contains the following
terms: HAZARD, FORMAT, LOGLEVEL, STEP, LOGLEVEL x STEP, and
either LOGLEVEL x FORMAT or STEP x FORMAT. All terms are highly
significant, ‘s < .001.
The absence of significant LOGLEVEL x HAZARD x FORMAT 4nd
STEP x HAZARD x FORMAT interactions (i.e., not significant at
the .0]. level) is the reason why Figures 14, C2 and C4 do not
distinguish between asbestos and radon.
Alternative analysis . Because the preceding analyses were
equivocal with respect to format effects on slopes and steps,
additional calculations were carried out. These analyzed the
effects of level by looking at the test results in pairs. The
LEVEL x FORMAT interaction based on the first (lowest) and
second test results indicates the effect of format on the
exposure—response relationship (slope) below the standard. The
LEVEL x FORMAT interaction train the second and third test
results indicates the effect of format on the step that may
occur around the action guideline. The LEVEL x FORMAT
interaction based on the third and fourth (highest) test results
indicates the effect of format on the slope above the standard.
This method of analysis is not ideal, since slope effects
and step effects cannot be completely separated in this way.
For example, assume that one format produces a particularly
steep straight—line response as one goes from the lowest to the
highest test result, with no additional step as one passes the
action guideline. The steep slope in this case applies to an
two different test results, including the point just above and
the point just below the standard. It would be easy to
misinterpret the large difference in response to these two
points as a substantial step.
-------
C2
When approached in this way, we find the following fornat
effects:
Below the standard, slopes vary with format for both
perceived threat, F(6,905) 2.43, 2 < .025, and action
plans, F(6, 965)=2.85, p < .01. Planned comparisons reveal
that the slopes are greater if formats have a standard
(F4—F7) than if they do not (Fl—F)), F(1, 905) = 10.83, p
.001, and F(1, 965) 14.80, p < .0001, for perceived threat
and action plans, respectively. Neither the addition of
advice to standards nor the addition of probabilities and
smoking comparisons had any further effect on the slope.
Around the standard, the step (slope) varies with format for
perceived threat, F(6, 873) 3.12, p < .005, but not for
action plans, F(6, 926) = 1.16, n.s. If attention is
focused on the effect gf the standard, one finds that a
standard increases the step for perceived threat, F(1, 873)
= 11.78, p < .001, and also increases th! step somewhat for
action plans, F(1, 926) = 3.90, p < .05.
Above the standard, there are no effects of format on slope,
F’s < 1, for either perceived threat or action plans.
Adding these results to the preceding analyses leads to the
conclusion that a standard helps people distinguish among risks
at levels below the standard. A standard also creates an’,
artificial step as test results pass the standard level. This
step effect is clear for perceived threat and small for action
plans (probably because of the unanticipated step that appears
to be present for radon with formats F1-F3). Finally, above the
standard, all formats are equally good (or poor) in helping
people recognize the difference between high levels and very
high levels.
Uniformity Across Particioants
Possible differences in the variability of response among
the formats ware examined using Scheffe’s test for homogeneity
of variance (Winer, 1971, pp. 219—220). This test is relatively
insensitive to departures from normality and can be used for
unequal—sized groups. The logarithms of the variances for each
of th. eight cells of the design within each format (four levels
and two hazards for each format) were the independent variables
3 Norma l ly, it is not appropriate to conduct additional
tests on specific conditions after an initial overall test is
not significant. The question of whether a standard would
produce a step function for action plans, however, was an issue
from the beginning of this study. The analysis of this effect
should be considered a planned comparison rather than a post—hoc
comparison.
-------
:22
in the calculations. Unifor nity of variance was exarn ned o iy
for perceived threat and action plans. The FORMAT effects er
not significant for either response variable, Q’s > .2.
Audience Evaluation
These calculations made use of the standard model. The
FORMAT effect was not significant for difficulty ratings, 2 >
.2. The FORMAT effect for amount of information was marginally
significant, F(6, 1844) = 2.50, 2 < .05. This is a very small
effect, significant only because of the la’ ge sample size. It
does not reach the .01 criterion set in this study for main
effects and was not examined further.
FORMAT effects for helpfulness and certainty were more
significant, F(6, 1845) 8.69, < .0001, and F(6, 1886) =
2.94, < .005, respectively. There were no significant HAZARD
x FORMAT interactions. Post—hoc comparisons were carried out
and significant differences among the formats are shown in Table
Cl.
Other Variables
Responses to questions about subjects’ actual plans to test
and their beliefs about mitigation difficulty were analyzed with
the standard model. People were less likely to say that t ey
planned to test for asbestos than for radon, with means of 2.36
and 3.23, respectively, F(1, 1870) — 228, < .0001, but there
were no format differences. There were also no format
differences in perceived mitigation difficulty, though asbestos
was seen as harder to reduce (means of 2.03 and 2.65 for
asbestos and radon, respectively, F(1, 1862) = 370, < .0001).
The possibility that communication success would vary with
education was examined by adding education (treated as an
8—point numerical scale rather than as a categorical variable)
to the prediction models used previously. The response
variables examined included the level that would be personally
acceptable, perceived threat, mitigation plans, and the four
evaluation variables (difficulty, amount of information,
helpfulness, and certainty). No significant FORMAT x EDUCATION
interactions were observed. The calculations did, however,
reveal a significant EDUCATION effect for threat perceptions and
a CATION x LOGLEVEL interaction for perceptions of threat and
mitigation plans (see Figure 19). For threat, both effects were
highly significant, ‘s < .0001. Less education was associated
with greater perceived risk, but as seen in Figure 19, this
applied only at lower levels. People with the least education
also showed the least variation in response as they went from
the lowest to the highest test result. For action plans, only
the interaction was significant, < .0003. Again, people with
less education differentiated less among the different test
results.
-------
C2 3
People with high school educations rated the brochures as
more difficult to understand, < .0001, and as less helpful for
understanding their levels, < .01, but these differences were
quite small in absolute terms. There was & o effect of education
on judgments of the amount of information contained in the
brochures or in ratings of how uncertain people felt about the
risk from their test result.
------- |