Professional Documents
Culture Documents
net/publication/226602407
CITATIONS READS
45 2,460
5 authors, including:
Richard J Siegert
Auckland University of Technology
250 PUBLICATIONS 6,946 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by John Mcclure on 17 May 2014.
w
Author for correspondence: E-mail: matt.spittal@cancervic.org.au
16
measures with four or five items (e.g., Dooley et al., 1992; Farley et al.,
1993; Showalter, 1993; McClure et al., 1999). Eight of the 18 studies in Ta-
ble I have used longer scales (between 12 and 27 items) to assess earth-
quake preparation (e.g., Hirose and Ishizuka, 1983; Hirose, 1986; Turner
et al., 1986; Mileti and Fitzpatrick, 1992, 1993).
These studies have yielded important insights into people’s preparation
for and response to natural disasters such as earthquakes. Unfortunately,
few of these studies have reported estimates of reliability or other relevant
psychometric data for their scales (Lindell and Perry, 2000). Any interpre-
tation of the results is limited therefore by the absence of a numerical esti-
mate of the amount of unsystematic measurement error in the scales.
These data, however, are available for some studies.
Russell et al. (1995) used exploratory factor analysis to identify three
factors: survival, planning, and hazard mitigation. Reliability estimates for
these factors ranged from 0.42 to 0.73. More substantial reliability coeffi-
cients were obtained by Mulilis et al. (1990), who developed a 27-item scale
for measuring the earthquake preparedness of individuals and small busi-
nesses. The Mulilis–Lippa Earthquake Preparedness Scale (ML-EPS) also
measures the perceived difficulty of preparing for an earthquake. For each
item, participants state whether they are prepared by answering Yes, No, or
Unsure, and also indicate the difficulty of performing each item. Using four
samples of respondents, Mulilis et al. (1990) reported estimates of internal
consistency (Cronbach’s alphas) ranging from 0.68 to 0.97 for the prepared-
ness items and estimates of 0.84 to 0.94 for the difficulty items. Test–retest
reliabilities ranged from 0.78 for a 4–6 week interval, down to 0.64 for a
3 month period for both preparedness and difficulty items.
Despite the use of psychometric criteria to assess the usefulness of the
ML-EPS, this scale has two possible limitations. First, in most earthquakes
leading to fatalities, the major cause of death is from buildings collapsing
on people (Russell et al., 1995; McGeary, 1999). Although the ML-EPS
has several questions assessing whether the contents of the home have been
secured (e.g., fastening tall furniture), this scale has no items that assess
whether people have ensured that their homes are sufficiently sound struc-
turally to prevent them collapsing. Because one major goal of earthquake
preparation is to prevent the loss of life, and since a major cause of loss of
life in earthquakes is buildings collapsing, the absence of a measure of this
component of preparation represents a possible limitation of the scale. A
second possible limitation of the ML-EPS is the inclusion of items that
may be unrelated to earthquake survival. In particular, the items ‘‘Do you
attentively listen to or watch radio or television messages about earth-
quake preparedness’’ and ‘‘Do you vote on bills dealing with earthquake
resistant buildings’’ may not necessarily identify those who have prepared
for an earthquake.
MEASUREMENT OF EARTHQUAKE PREPARATION 19
Previous research on this issue has not focused on the issue of interpret-
ing scale scores – that is, what it means to be at different points on the
scale. With most scales, it is unclear how a person with a score of say,
two, differs from a person with a score of 10 (although decile-based norms
do provide some useful information on this issue). The basis for interpret-
ing scores in this way is referred to as scale calibration. Calibration has
been performed successfully on other measures, for instance the Economic
Living Standards Index (Jensen et al., 2002); and is useful when a scale is
used for descriptive purposes. Thus information on the calibration of a
preparedness measure would usefully augment other psychometric informa-
tion.
Finally, for many uses of a measure it is convenient to group responses
into score intervals (for example, where there is a need to present results
indicating the distribution of a set of scores or to compare the distribu-
tions of two sets of scores). This is usually done by means of a table or
histogram showing the numbers of scores falling within specified score
intervals. In general, however, no attempts are made to develop formal
score intervals. Instead, different researchers have interpreted scores along
the continuum of a scale in a sample-specific manner. Although this meth-
od is not flawed and is applied to scales in many domains, it does limit
comparisons between studies, particularly where presenting descriptive
information is of interest. Thus, we regard the specification of a standard
set of score intervals as useful for other researchers – particularly when
combined with the calibration information.
In sum, a review of measures of earthquake preparation shows that 18
scales have been published in the last 25 years. Few studies have reported
data that indicate the level of measurement error within the scales. Of the
studies that report psychometric data, Russell et al. (1995) found poor esti-
mates of reliability for two of their three subscales, while Mulilis et al.
(1990) found higher estimates of reliability for a 27 item scale. However,
not all items in this scale necessarily measure earthquake preparedness.
Other problems identified in measures of preparation (and in measures
generally) include a lack of an interpretative framework for guiding analy-
sis and the lack of standard score intervals for guiding description.
To address the limitations identified in previously developed earthquake
preparation measures, a new scale was developed to assess household pre-
paredness for a major earthquake: the Earthquake Readiness Scale (ERS).
Several goals were applied to the development of this scale: (a) A unidi-
mensional questionnaire that assessed people’s preparation for a major
earthquake; (b) A scale with a high level of internal consistency; (c) A
scale with a normal spread of responses (for parametric tests); (d) To cali-
brate the scale for interpretive purposes; (e) To be able to specify scale
scores as a set of intervals that would suit most relevant applications.
20 MATTHEW J. SPITTAL ET AL.
Schmitt (1996) and others have criticised the general use of coefficient
alpha as a measure of unidimensionality. Coefficient alpha is an estimate
of internal consistency that refers to the interrelatedness of a set of items;
whereas unidimensionality refers to the existence of one latent trait under-
lying the data (Hattie, 1985; Cortina, 1993). Thus items can be interrelated
to one another yet form different latent variables. The development of the
ERS scale provided an opportunity to develop an earthquake preparation
measure using goodness of fit procedures to assess unidimensionality. We
hypothesised that a unidimensional model would fit the 23 earthquake pre-
paredness items better than competing models (e.g., a two factor model).
The validity of the ERS was explored using three different approaches.
Mulilis et al. (2000) showed that homeowners are better prepared for disas-
ters such as tornadoes than non-homeowners. One way of testing the con-
struct validity of the ERS is to see whether this finding holds for the ERS.
We therefore predicted that homeowners would be better prepared for an
earthquake than non-homeowners. Second, previous research has shown
that age is related to preparation, with older people on average being better
prepared (Dooley et al., 1992). If the ERS is a valid measure of earthquake
preparation, then this finding should be replicated. Thus we hypothesised
that age would be positively related to preparation. Our final indicator of
validity was a comparison between participants’ self-ratings of their level of
preparation and the actual preparation that they had undertaken, as shown
by their responses on the ERS. The purpose of this indicator is to examine
the face validity of the ERS, because a global judgement of participants’
preparation should be positively related to their actual level of preparation.
To test this, we hypothesised that a global self-rating of participants’ prepa-
ration would be significantly correlated with their ERS scores.
Finally, a calibration of the ERS was obtained by showing how the
occurrence of different types of preparedness activities changes across the
continuum of responses on the scale. The two types of activities that were
examined were low-level preparation activities, such as storing food and
water, and high-level activities, such as structural modifications to dwell-
ings. The low level actions may be regarded as activities that facilitate sur-
vival, and those at a higher level, as activities that mitigate damage.
1. Method
1.1. PARTICIPANTS
The exploratory sample included 295 respondents, of whom 49% were fe-
male and 51% were male. Twenty-five percent were aged 24 or under, 41%
were between 25 and 40, 30% were between 41 and 60, and 4% were 61 or
older. Forty-six percent owned their own home and 54% did not.
The second group of respondents (referred to as the Residential Survey
Sample) were 358 residents of Wellington City, of whom 59% were female
and 41% were male. Twelve percent of this group were aged 24 or under,
32% between 25 and 40, 37% between 41 and 60, and 19% were 61 and
over.
earthquake preparation. Based on the review of the sources, the scale in-
cluded two sets of items. The first set included ‘‘mitigation’’ actions that
are specifically intended to limit earthquake damage, and the second set
consisted of ‘‘planning’’ actions intended to facilitate survival following a
major earthquake. This second set of items included some actions, such as
access to an alternative cooking source, that would not be performed so-
lely for the purpose of earthquake preparation. Other researchers have no-
ted these multi-purpose items in their scales (e.g., Dooley et al., 1992;
Weinstein et al., 2000).
To check that the respondent had actually performed the preparatory
action, each item was phrased as ‘‘I have’’ followed by a verb (e.g., consid-
ered, fastened, ensured), followed by the action. Each question was designed
to be answered with a ‘‘yes’’ or ‘‘no’’ response. For two items that referred
to major structural features (4 and 5), the item allows for respondents to
have checked and satisfied themselves that the adjustment was not neces-
sary, having been performed on the structure by the builder or a previous
owner of the home. This design was used in order to reduce the likelihood
of participants replying in the negative if this feature had been dealt with by
a previous owner. The items included in the scale are presented in Table II.
Following the ERS scale were several demographic items: Age (24 or
under, 25–40, 41–60, 61+); gender; and in the first sample: Do you own
your own home? (Yes/No).
1.3. PROCEDURE
For the first sample, members of the public sitting in parks in the central
business district of Wellington city were invited to complete a question-
naire. Those who agreed were given a copy of the questionnaire. Before
answering the ERS questions, participants in this sample were asked ‘‘Do
you think that you are prepared for a major earthquake?’’ Participants re-
sponded to this question on a 7-point scale ranging from 1 (not at all pre-
pared) to 7 (very prepared), with a midpoint of 4 (somewhat prepared).
For the second sample (the Residential Survey Sample), 1,000 question-
naire booklets containing the ERS were delivered to households across 10
area sampling units that represented all five levels of socioeconomic strata
across Wellington City. A stamped envelope was provided for participants
to return responses. The response rate for this survey was 38.8%, which
yielded a total of 358 complete questionnaires.
2. Results
An exploratory factor analysis was undertaken on the first sample to
establish the factor structure of the scale. The unrotated principal compo-
MEASUREMENT OF EARTHQUAKE PREPARATION 23
Table II. Percentage of participants undertaking each of the earthquake preparation activi-
ties listed in the Earthquake Readiness Scale.
nents suggested that the data did form a clear general factor with a high
mean loading of 0.51 on the first component (SD=0.08) and very low
mean loadings on the other components (M=0.01, SD=0.27 for Factor 2
and M=0.01 SD=0.26, for Factor 3). Analysis of the rotated solution
suggested that there might be a coherent two-factor structure to the items:
one factor seemed to comprise activities undertaken solely for disaster
preparation (e.g., fastening hot water cylinder, obtained working fire extin-
guisher, etc.). The other factor contained activities that could potentially
be undertaken for a number of reasons (e.g., obtained working torch, first
aid kit, etc.). It is interesting that if activities are ordered from lowest to
highest according to the percentage of people who have performed them,
items related to Factor 2 appear predominately at the top of this list (most
people have undertaken these activities) and items related to Factor 1 ap-
pear at the bottom (few have performed them) (see Table II). The three-
factor solution appeared less clear, and could not be interpreted.
To test the competing one and two factor structures, confirmatory fac-
tor analysis was undertaken using the responses from the Residential Sur-
vey Sample. To examine the unidimensionality of the scale, we employed
the method recommended by Hattie (1985). In this analysis, items were
parcelled into three groups of 6 items and one group of 5 items (Kishton
and Widaman, 1994). Eight indices of fit were examined – the v2 test; the
Goodness of Fit Index (GFI); the Adjusted Goodness of Fit Index (AGFI)
the Comparative Fit Index (CFI); the Normed Fit Index (NFI); the Root
Mean Square Error of Approximation (RMSEA); the Root Mean Square
Residual (RMR); and the ratio of the v2 to the degrees of freedom. This
model was contrasted against the two-factor model identified by the
exploratory analysis.
The data did not differ significantly from the unifactorial model
(v2 (2)=5.77, NS; GFI=0.99, AGFI=0.96, CFI=0.98, NFI=0.98;
RMSEA=0.07, RMR=0.02; ratio of v2 to degrees of freedom=2.88). The
observed data, however, did differ from the two-factor model derived from
the exploratory analysis. With items grouped in seven parcels, v2
(13)=60.18, p>0.0001; GFI=0.95, AGFI=0.89, CFI=0.93, NFI=0.92;
RMSEA=0.10, RMR=0.05; and the ratio of v2 to degrees of free-
dom=4.63, indicating that this is a less adequate model. Examination of
the difference between the v2 values found for the two models, (v2
(11)=54.41, p>0.001) shows that the unifactorial model is significantly
better than the two factor model.
In addition to this empirically-derived two factor model, a competing the-
oretically-derived two factor model was also evaluated. This two factor mod-
el contrasted items relating to survival facilitation (items 7–12, 14–18, 20–23)
and items relating to damage limitation (items 1–6, 19). For this analysis,
items were grouped in five parcels. The data did not differ significantly from
MEASUREMENT OF EARTHQUAKE PREPARATION 25
Table III. Standard score ranges and distribution of the Earthquake Readiness Scale.
1 0–4 14.9
2 5–8 20.9
3 9–12 26.8
4 13–16 25.9
5 17–23 11.5
3. Discussion
This study developed and evaluated a unidimensional questionnaire, the
Earthquake Readiness Scale (ERS), which was designed to assess people’s
preparedness for a major earthquake. Further goals of the study were to
demonstrate that the developed scale has a high level of reliability, an
acceptable level of validity, and a calibrated interpretative framework. To
achieve these goals, the generation of items for the scale was based on rec-
ommendations made by several relevant agencies, and took account of
steps that other researchers have regarded as necessary for preparation
(e.g., Hirose, 1986; Turner et al., 1986; McClure et al., 1999).
The results indicate that the ERS is a unidimensional measure that has
an acceptably high level of internal consistency. Confirmatory factor analy-
sis showed a close fit between the data and a one-factor model. This model
was better on goodness of fit indices and error estimates than the competing
two-factor model derived from the exploratory factor analysis of the other
data set. A model that identified survival facilitation and damage limitation
as separate factors was also examined. This model was not rejected on the
basis of confirmatory factor analysis, but it was considered that the imbal-
ance in the scale lengths in this model gave the unidimensional model an
edge. In addition, the correlation between the two latent variables was so
high that little utility is gained by separating them into separate factors.
Thus, the parsimonious one-factor model is more appropriate as a measure
of earthquake preparation. A subsequent independent analysis of data from
358 respondents confirmed the appropriateness and the reliability of the
unidimensional model of the Earthquake Readiness Scale.
MEASUREMENT OF EARTHQUAKE PREPARATION 27
Tests of the validity of total ERS scores showed the predicted differ-
ences for substantial groups who would be expected to differ in their
scores: home-owners and non-homeowners, and different age groups. A
third indicator of the validity of the scale is that participants who thought
they were better prepared for an earthquake had higher scores on the
ERS.
The basis for interpreting ERS scores is the calibration information
which contrasts two different types of preparedness – low-level survival
facilitation preparation (e.g., having a store of water, food, spare batteries,
etc.) and high level damage limitation preparation (e.g., structural modifi-
cations to dwellings, fastening heavy objects, etc.). The calibration infor-
mation showed that those with low ERS scores tended to have undertaken
only a small percentage of the survival facilitation activities and effectively
none of the damage limitation activities. As ERS scores increased, the level
of survival facilitation increased rapidly. Damage limitation increased also
but to a lesser extent.
The distribution of ERS scores in Table III coupled with the calibration
information shows the percentage of the participants that have undertaken
some survival facilitation activities and no damage limitation activities
(36% of the sample – 14% from level 1 and 22% from level 2). About
26% of the participants had undertaken moderate levels of both survival
facilitation activities and damage limitation activities (level 3). High levels
of survival facilitation were observed for about 37% of the participants
(levels 4 and 5), but only 12% of the participants had undertaken a signifi-
cant amount of damage limitation preparation (in level 5). Based on this
information, people in level 1 could be regarded as ‘‘very poorly pre-
pared;’’ people in level 2 as ‘‘poorly prepared;’’ people in level 3 as ‘‘mod-
erately well prepared;’’ people in level 4 as ‘‘well prepared;’’ and people in
level 5 as ‘‘very well prepared.’’ The mean score from these data suggests
that, on average, Wellington residents are only moderately well prepared
for a major earthquake.
The ERS has been designed to include a wide range of earthquake
preparation indicators. As with any measure, however, the question arises
as to whether other items should be included (e.g., having a wrench to turn
of water, etc.). Items such as this are likely to be highly correlated with the
scale. Thus, individuals who fail to store water, have a first-aid kit, or
make structural modifications to their home are also unlikely to have a
wrench to turn off water. A primary criterion for the inclusion of addi-
tional items should be the discriminative power that an additional item can
offer in measuring earthquake preparation. Using the example above, such
an item is likely to offer greater discrimination at the bottom end of the
scale’s distribution (i.e., it may discriminate between those at level 1 and
level 2, but not between people at levels 4 and 5). Given that there is
28 MATTHEW J. SPITTAL ET AL.
already a large number of items that discriminate at the bottom end of the
scale, any revisions of the ERS could focus on incorporating items that
discriminate at the top end of the continuum. However, the responses from
the present samples show that the ERS has an approximately normal dis-
tribution, suggesting that the scale does not need additional top-end items
to successfully measure earthquake preparation.
In summary, the data presented here support the usefulness of the ERS
as a measure of earthquake preparation. It includes items on structural
preparation, a component that is not included in other measures of earth-
quake preparation. It has a stable, unifactorial structure that is suitable for
use as a dependent variable, and has high reliability and good validity. The
scale can be expressed as a continuous variable ranging from 0 to 23, or it
can be classified into five different score intervals. The latter will be partic-
ularly useful when descriptive information, particularly between different
sub-groups, is of interest. Finally, the scale has been calibrated to enable
interpretation of scale scores.
Acknowledgements
We thank members of the Friday Research Group for their assistance in
the Scale development, and Douglas Paton and Kathy Parkes for valuable
comments on the study.
References
Cortina, J. M.: 1993, What is coefficient alpha? An examination of theory and application, J.
Appl. Psychol. 78, 98–104.
De Man, A. and Simpson-Housley, P.: 1987, Factors in the perception of earthquake hazard,
Percept. Motor Skills 64, 815–820.
Dooley, D., Catalano, R., Mishra, S., and Serxner, S.: 1992, Earthquake preparedness:
predictors in a community survey, J. Appl. Social Psychol. 22, 451–470.
Edwards, M. E.: 1993, Social location and self-protective behavior: implications for
earthquake preparedness, Int. J. Mass Emergencies Disasters 11, 293–303.
Farley, J. E., Barlow, H. D., Finklestein, M. S., and Riley, L.: 1993, Earthquake hysteria,
before and after: a survey and follow-up on public response to the Browning forecast, Int.
J. Mass Emergencies Disasters 11, 305–322.
Hattie, J.: 1985, Methodology review: assessing unidimensionality of tests and items, Appl.
Psychol. Meas. 9, 139–164.
Hirose, H.: 1986, The psychological impact of the Tokai earthquake prediction: individual’s
responses and the mass media coverage, Jpn. Psychol. Res. 26, 64–76.
Hirose, H. and Ishizuka, T.: 1983, Causal analysis of earthquake concern and preparing
behavior in the North Izu Peninsula, Jpn. Psychol. Res. 25, 103–111.
Hurnen, F. R.: 1997, Perceived Damage Preventability, Knowledge, and Preparation for
Earthquakes, Victoria University of Wellington, Wellington. Unpublished MA thesis.
MEASUREMENT OF EARTHQUAKE PREPARATION 29
Hurnen, F. R. and McClure, J.: 1997, The effect of increased earthquake knowledge on
perceived preventability of earthquake damage. Aust. J. Disaster Trauma Stud. 1, Massey
University, New Zealand, http://www.massey.ac.nz/-trauma.
Jackson, E. L.: 1981, Response to earthquake hazard: the West Coast of North America,
Environ. Behav. 14, 387–416.
Jackson, E. L. and Mukerjee, T.: 1974, Human adjustment to the earthquake hazard in San
Francisco, California, In: G. F. White (ed), Natural Hazards: Local, National and Global,
New York: Oxford University Press, pp. 160–166.
Jensen, J., Spittal, M., Crichton, S., Sathyandra, S., and Krishnan, V.: 2002, The Direct
Measurement of Living Standards: The New Zealand ELSI Scale, Ministry of Social
Development, Wellington.
Kiecolt, K. J. and Nigg, J. M.: 1982, Mobility and perceptions of a hazardous environment,
Environ. Behav. 14, 131–154.
Kishton, J. M. and Widaman, K. F.: 1994, Unidimensional versus domain representative
parceling of questionnaire items: an empirical example, Educ. Psychol. Meas. 54, 757–765.
Lindell, M. K. and Perry, R. W.: 2000, Household adjustment to earthquake hazard: a review
of research, Environ. Behav. 32, 461–501.
McClure, J., Walkey, F., and Allen, M.: 1999, When earthquake damage is seen as
preventable: attributions, locus of control and attitudes to risk, Appl. Psychol. Int. Rev. 48,
239–256.
McGeary, J.: August 30, 1999, Buried alive, Time 16–21.
Mileti, D. and Darlington, J. D.: 1997, The role of searching in shaping reactions to
earthquake risk information, Soc. Probl. 44, 89–103.
Mileti, D. and Fitzpatrick, C.: 1992, The causal sequence of risk communication in the
Parkfield earthquake prediction experiment, Risk Anal. 12, 393–400.
Mileti, D., Fitzpatrick, C.: 1993, The great earthquake experiment: Risk communication and
public action. Boulder, CO: Westview.
Mileti, D. and O’Brien, P. W.: 1992, Warnings during disasters: normalizing communicating
risk, Soc. Probl. 38, 40–57.
Mulilis, J. P. and Lippa, R.: 1990, Behavioral change in earthquake preparedness due to
negative threat appeals: a test of protection motivation theory, J. Appl. Soc. Psychol. 20,
619–638.
Mulilis, J. P., Duval, T. S., and Bovalino, K.: 2000, Tornado preparedness of students, non-
student renters, and non-student owners: issues of PrE theory, J. Appl. Soc. Psychol. 30,
1310–1329.
Mulilis, J. P., Duval, T. S., and Lippa, R.: 1990, The effects of a large destructive local
earthquake on earthquake preparedness as assessed by the Earthquake Preparedness Scale,
Nat. Hazards 3, 357–371.
Russell, L. A., Goltz, J. D., and Bourque, L. B.: 1995, Preparedness and hazard mitigation
actions before and after two earthquakes, Environ. Behav. 27, 744–770.
Schmitt, N.: 1996, Uses and abuses of coefficient alpha, Psychol. Assess. 8, 350–353.
Showalter, P. S.: 1993, Prognostication of doom: an earthquake prediction’s effect on four
small communities, Int. J. Mass Emergencies Disasters 11, 279–292.
Sullivan, R., Mustart, D. A., and Galehouse, J. S.: 1977, Living in earthquake country, Calif.
Geol. 3, 3–8.
Turner, R. H., Nigg, J. M., and Paz, D. H.: 1986, Waiting for Disaster: Earthquake Watch in
California, University of California Press, Los Angeles.
Weinstein, N., Lyon, J. E., Rothman, A. J., and Cuite, C. L.: 2000, Preoccupation and affect
as predictors of protective action following natural disaster, Brit. J. Health Psychol. 5,
351–363.