Documente Academic
Documente Profesional
Documente Cultură
HIROTADA HIROSE
Departmentof Psychology,Collegeof Arts and Sciences,Tokyo Woman'sChristian Unioersily,Suginami-ku,
Tokyo167
TOMOICHI ISHIZUKA
Japanese National Railways Labor ScienceResearch Institute, Kokubunji-shi, Tokyo 185
Key words: earthquake, panel survey, covariance structure analysis, preparing behavior,
earthquake anxiety, cognition of earthquake danger.
At a meeting of the Earthquake as the " Tokai earthquake " after the
Warning Council' in August 1976, a name of the region around Shizuoka
seismologist announced that there was Prefecture, it will cause damage over
imminent danger of a major earthquake an extremely wide area inland from
centered on Suruga Bay in Shizuoka Suruga Bay and is expected to occur in
Prefecture in Japan, which was widely the midterm or long-term future.
reported in the mass media and caused News about the medium- to long-term
extensive anxiety. In October of the same predictions of the Tokai earthquakes has
year, his announcement was generally con- been widely disseminated by television,
firmed in testimony given before the newspapers, and other mass media. The
Standing Committee on the Budget in threat of an earthquake by far the largest
the Diet by a professor (Rikitake, 1979). in personal history has stimulated popular
In the worst-case scenario, the focus of anxieties and urged the people toward
the earthquake, which is expected to have advance preparations.
a magnitude of 8.0 or more on the Richter
scale, could include such densely pop- Panel Survey
ulated areas as the cities of Numazu,
Shimizu, and Shizuoka, with the sprawling Given this background, the authors
Tokyo-Yokohama megalopolis less than conducted a three-wave panel survey of
150 kilometers away. Now referred to earthquake concern and preparing be-
havior among the people of two major
1 A liaison committee of the Geographical Survey tourist centers (Atami City and Ito City)
Institute authorized by the national government to
in Shizuoka Prefecture, where the scien-
conduct and publish observations of seismic ac- tific prediction of the Tokai earthquake
tivity. had seemed to damage on tourist enter-
104 H. Hirose and T. Ishizuka
prises. The survey was carried out on The person seeks some behavior that will
people who had been selected randomly modify this situation, and there ensues a
from town registers. The first and third decision-making process, which is followed
waves were carried out by having inter- by enactment of the selected behavior
viewers contact the sample, and the second (McGrath, 1982). Next, the enacted be-
wave was a mail survey. For the first havior is evaluated to see if it has suc-
wave, 1 300 men and women between the ceeded in mitigating the danger of the
ages of 20 and 69 were chosen. Interviews situation. If it has, vulnerability to the
were carried out in August 1980. Out of threat is reduced and the intensity of
the 1 300 subjects, interviews were com- anxiety and stress should subside (Krohne
pleted with 1 018 (78.3% completion & Rogner, 1982).
rate). The second wave questionnaires The mixture of different types of indi-
were sent to the 1 018 people who had viduals may complicate the phenomenon,
answered all of the first-wave survey ques- however we can make certain predictions
tions, and replies were received from 668 by restricting ourselves to the three factors
of them (65.6%) in February 1981. of earthquake anxiety, and preparing be-
Finally, the third wave was carried out in havior and restricting the period of study
October 1981, interviewing the same 1 018 to the relatively short period of fifteen
that had answered the first wave. This months. Our working hypotheses are given
time the completion rate was 81.2% (827 next,
respondents). The number who replied Hypothesis 1: Despite fluctuations
in all three waves totalled 576. among individuals, the two factors of
earthquake anxiety and danger cognition
Analysis of Earthquake Concern should be consistent and stable over time.
and Preparing Behavior Hypothesis 2: Attitude toward earth-
quake, which is composed by cognition
H)patheses of earthquake danger and earthquake
Developments in geophysics and seis- anxiety, should have a positive, causal
mology have made it possible to predict effect on earthquake preparing behavior.
the location, magnitude, and so forth of Hypothesis 3: Besides a contemporary
major earthquakes with fair accuracy and causal effect on preparing behavior
(Weisbecker, Stoneman, Ackerman, (in cross section), attitude toward earth-
Arnold, Halton, Ivy, Kauty, Kroll, quake should have a lagged causal in-
Levy, Mickley, Miller, Rainey, & Zandt, fluence on preparedness behavior at the
1977; Rikitake, 1976; Mogi, 1981). We next observed time.
have been presented with a new phenom-
enon: the social, ecomomic, psycho- Data
logical impact of information about a pre- The levels of earthquake preparing
dicted natural disaster on the area of behavior, earthquake anxiety, and cogni-
predicted damage. tion of earthquake danger are the variables
Information (scientific prediction) about in the three different waves of the survey.
an objective event like an earthquake con- The way in which questions were asked
fronts people with a kind of perceived and answered for each of the three vari-
situation. In a process that has been ables was as follows.
described as " cognitive appraisal" Preparing behavior. Subjects were pres-
(Lazarus, 1966), people attempt to assess ented with the question, " Please tell me
the degree of certainty of the danger. if you have done any of those things to
This cognitive process is reported to raise prepare for a future earthquake ? " with
anxiety and stress (McGrath, 1982). following alternatives to select from (mul-
Causal analysis of attitude toward an earthquake 105
Table I
Table 2
tiple answers allowed). tion, " Do you feel anxiety over the pros-
1. Fasten down breakables and furniture pect that the north Izu Peninsula may
2. Have an emergency kit ready 3. Keep have a major earthquake in the near
a bucket or tub filled with water 4. Keep future ? " were given 1 point for the reply
" Little anxiety
a fire extinguisher 5. Keep medicine and ," 2 for " Some anxiety,"
protective headgear 6. Keep food and 3 for " Considerable anxiety," and 4 for
" Extreme anxiety ."
drinking water 7. Have earthquake insur-
ance 8. Reinforce house or fence con- Cognitionof earthquakedanger. Replies to
structions 9. Decide on evacuation routes the question, " Tell me how much danger
10. Decide on emergency communication the earthquake is supposed to bring "
methods within the family 11. Participate were given 1 point for the reply " Little,"
in community earthquake drills 12. Stay 2 for " Some," and 3 for " Considerable."
away from dangerous places or dangerous Table I gives the variables and notation
vehicles 13. Not undertake new building used in the analysis. Since there were
14. Keep extra cash on hand 15. Others. three waves and three variables in each
One point was allotted for each answer, wave, a total of nine variables were an-
and one additional point was added to alyzed.
the individual's total for an affirmative
answer to the question, " Have you par- Results
ticipated in your community's disaster Table 2 gives the means, variances, and
prevention organization? " Accordingly, covariances of the nine variables. A com-
it was possible to rate each individual on parison of the means using a two-tailed
a scale running from a minimum of0 point t-test shows that Y1<Y2 (p<0.05)<Y3
to a maximum 16 points. (p<0.001). The average level of preparing
Earthquake anxiety. Replies to the ques- behavior thus increased over time, indi-
106 H. Hirose and T. Ishizuka
with lag. As can lie seen from both These relations can be described in
figures, to the model in Fig. la, the model terms of matrix equation as follows:
in Fig. lb adds paths by which attitude
X=Δxξ+δ (1)
toward earthquake at time T, influences
and )7i./.
where
shown below:
(8)
where
I= identity matrix
When the parameters in equation (8) fit, being not significant at even the 25%
are estimated by the maximum-likelihood level. The lag model is significant at the
method (McDonald, 1978, 1980) for the 25% level, but not at the 10% level. The
models in Figs. 1a and 1b, the results are lag model is constructed of the no-lag
shown in the first two columns of Table 5. model with the parameter r2 added, so
The parameters At, 22, and 23are scaled the contribution of this parameter can be
to unity to standardize the estimates. found by subtracting the chi-square value
The data are the 45 variance and covari- of the lag model from the chi-square of
ance values listed in Table 2. For Fig. la, the no-lag model and applying the chi-
the model without lag, there are 29 para- square test for one degree of freedom.
meters to be estimated. Since there are Since the difference between the two
45 equations, there arc 16 degrees of free- chi-square values is zero, the contributions
dom. For Fig. lb, the lagged model, of 7.2 is zero. The no-lag model is thus
there are 31 parameters to be estimated, better for the panel data at Tl and T2.
so there are 14 degrees of freedom. Both Figure 3 shows models without lag
models show a poor fit to the data, with (Fig. 3a) and with lag (Fig. 3b) for the
chi-square values signifiicant at the 1% Tr T, section, and the fifth and sixth
level. columns of Table 5 give the results of
Could the bad fit be due to bias in the maximum-likelihood estimation of their
parameter estimates caused by lumping parameters. The chi-square test shows
different structures together? It will be
appropriate to suppose that the T1-T2
and T2- T3 sections of the models might be
structurally dissimilar despite the stability
of each variable. Let us, therefore, divide
the three-wave model into two two-wave
models. Figures 2a and 2b show models
for the Tl-T2 section without and with
lag. If the maximum-likelihood method is
used to estimate the parameters in these
models, the results given in the third and
the fourth columns of Table 5 are ob- Fig. 2 a. A two-wave-three-variable covariance
tained. structure model with no lag effect (T2-73).
Both models now show good fits to the Fig. 2 b. A two-wave-three-variable covariance
data. The no-lag model shows the better structure model with lag effect (T,-T,).
Causal analysis of attitude toward an earthquake 109
Table 5
Maximum-likelihood estimates for the models in Figs. Ia, lb, 2a, 2b, 3a and 3b
of the no-lag model and apply the chi - lagged effect on earthquake-adaptive be-
square test for one degree of freedom . havior.
Once more, the difference between the
two chi-square values is zero , so the con- Discussion
tribution of the lag effect 14 is clearly not
significant. This result enables us to con- Never before in human history have
clude that the no-lag model is more ap- such broad regional societies been so seri-
propriate. ously affected by disaster prediction in-
We are thus left with two surviving formation. The first full-scale empirical
models, those of Figs. 2a and 3a , corre- social scientific investigation of the social
sponding to columns 3 and 5 in Table 5. impact of information stemming from a
These models will be used to examine major earthquake prediction was made by
hypotheses 2 and 3. Turner, Nigg, Paz, and Young (1980),
Concerning hypothesis 2, the fairly large whose survey dealt with the predicted
positive values of rr and r3 indicate that Southern California earthquake.
the earthquake attitude consisting of Meanwhile, similar events were taking
earthquake danger cognition and earth- place across the Pacific in Japan, where
quake anxiety has a positive effect on national and local governments began
preparing behavior. On the other hand, planning and implementing measures in
is has a small negative value, very close preparation for a major earthquake (the
to zero. This implies that, although Tokai earthquake) expected to occur in
attitude toward earthquake accelerate the near future.
the preparing behavior, the attitude loses The first result of the analysis was that
its function to motivate preparing behavior over the 15-month period of three-wave
after an individual's preparations have survey, levels of individual cognition of
advanced to a certain degree. Hypothesis earthquake danger and earthquake anx-
2, accordingly, is appropriate only before iety showed a sort of stability. The rela-
preparing behavior has been enacted, and tive stability of these factors provides some
becomes inappropriate, once it is ex- basis for concluding that there is very
ecuted. little chance that the sense of impending
Hypothesis 3 must be rejected, since the earthquake danger might suddenly flare
contributions of the lag-effect parameters up and ignite panic.
are all statistically insignificant. What has A second finding is that, while cognition
been confirmed is the opposite hypothesis of earthquake danger and earthquake
that attitude toward earthquake has no anxiety are capable of motivating earth-
quake adaptive behavior before earth-
quake preparations have been enacted,
they lose this function after the adaptive
behavior has been reached to a certain
degree. The mean values show that cog-
nition of danger and level of anxiety drop
as adaptive behavior progresses.
Finally, it has become clear that cog-
nition of earthquake danger and earth-
quake anxiety have a causal effect on adap-
Fig. 3 a. A two-wave-three-variable covariance tive behavior only in cross section, that
structure model with no lag effect (T2-T5) . is, only at the same time point. They have
Fig. 3 b. A two-wave-three-variable covariance no time-lagged influence. This suggests
structure model with lag effect (T2-T3). that cognition and anxiety cannot build
Causal analysis of attitude toward an earthquake 111