Full text of DOI:10.1093/geronb/57.2.S96

Journal of Gerontology: SOCIAL SCIENCES
Copyright 2002 by The Gerontological Society of America
2002, Vol. 57B, No. 2, S96–S107
Religion, Death of a Loved One, and Hypertension
Among Older Adults in Japan
Neal Krause,1 Jersey Liang,1 Benjamin A. Shaw,1 Hidehiro Sugisawa,2 Hye-Kyung Kim,2
and Yoko Sugihara2
1School of Public Health and Institute of Gerontology, University of Michigan, Ann Arbor.
2Tokyo Metropolitan Institute of Gerontology, Japan.
Objective. To see whether three dimensions of religion (private religious practices, religious coping, and belief in the
afterlife) buffer the effect of the death of a significant other on change in self-reported hypertension over time.
Methods. Interviews were conducted with a nationally representative sample of people aged 60 and older in Japan at
two points in time, 1996 and 1999. Complete data were available on 1,723 older Japanese. Respondents were asked a se-
ries of questions about their religious beliefs and practices, whether a family member or close friend had died in the past
year, and whether they had hypertension.
Results. The data suggest that older adults in Japan who experienced the death of a loved one but who believed in a
good afterlife were less likely to report they had hypertension at the follow-up interview than elderly people in Japan
who lost a close other but did not believe in a good afterlife.
Discussion. The results suggest how one overlooked dimension of religion (i.e., religious beliefs) may bolster the
health of older people in the face of adversity.
convincing body of research has suggested that older
people who are religious tend to enjoy better health
figure out whether the potentially salubrious effects of reli-
gion are restricted to the Judeo-Christian religions of the
West, or whether there is something common to faiths in all
cultural settings that may benefit health.
Second, Ellison and Levin (1998) argued that if further
headway is to be made in the study of religion and health, in-
vestigators must use data that have been gathered prospec-
tively. We aimed to contribute to the knowledge base by ana-
lyzing data that came from a nationally representative sample
of older people in Japan who were interviewed at more than
one point in time. Doing so allowed us to observe the effects
of religion on change in health over time. Throughout, we fo-
cus on one particular chronic illness, hypertension.
Third, far too many epidemiologic studies have examined
a limited range of religion measures. In fact, many have fo-
cused solely on the frequency of church attendance (e.g.,
Strawbridge, Cohen, Shema, & Kaplan, 1997). Although
this work has produced a number of valuable insights, many
investigators do not take advantage of the fact that religion
is a complex multidimensional phenomenon. If further ad-
vances are to be made in the field, researchers must examine
the effects of multiple facets of religion on health within a
multivariate framework. Consistent with this view, three
dimensions of religion are examined in the analyses that
follow: private religious practices, religious coping, and
religious beliefs. This measurement strategy follows the
guidelines specified by Ellison and Levin (1998), who
maintained that researchers should investigate the func-
tional aspects of religious involvement (e.g., coping) in con-
junction with behavioral markers of religion (e.g., private
religious practices).
A
than elderly people who are less religious (Koenig, 1999;
Koenig, McCullough, & Larson, 2001). The relationship
between these constructs appears to hold for an impressive
array of outcomes, including global measures of self-rated
health, functional disability, hypertension, heart disease,
cancer, and mortality (see Koenig et al., 2001, for a compre-
hensive review of this literature).
A key task facing researchers is to explain why religion
may exert a salubrious effect on the health of older people.
Although a number of causal factors are likely to be in-
volved, one line of reasoning that appears to be especially
promising is pursued in this study. In particular, as Ellison
and Levin (1998) pointed out, religion may exert a benefi-
cial effect on health because it helps elderly people cope
with the deleterious effects of stressful life events. Even
though several investigators have already explored the in-
terface between religion, stress, and health in late life (e.g.,
Krause, 1998), we hope to contribute to the literature in
three potentially important ways.
First, we explore the relationships among religion, stress,
and health in a diverse cultural setting—Japan. As Musick,
Traphagan, Koenig, and Larson (2000) argued, the wide
majority of studies on religion and health have focused on
older Americans with Judeo-Christian religious orienta-
tions. Consequently, these investigators have argued that a
top priority be placed on conducting research in culturally
diverse settings so researchers can learn more about the ro-
bustness of the relationship between religion and health.
More specifically, this type of work will help researchers
S96

RELIGION AND HYPERTENSION
S97
In the discussion that follows, we begin by developing
the theoretical underpinnings of this study. The sample and
survey measures are introduced after this. Finally, the re-
sults are presented and discussed.
religious beliefs and rituals are among the most important
factors eliciting the relaxation response. Second, Benson re-
viewed research showing that hypertensive patients who
were able to cultivate the relaxation response subsequently
experienced a decrease in blood pressure. Given these pro-
vocative possibilities, we need to learn more about how reli-
gion may foster the relaxation response in the face of adver-
sity.
THEORETICAL OVERVIEW
To the best of our knowledge, there is no published em-
pirical research on the relationships among religion, stress,
and physical health among older adults in Japan. Conse-
quently, we initially turn to studies that were conducted in
the West to frame the study hypotheses. This is accom-
plished by first examining research on the relationship be-
tween religion and hypertension. Following this, we expand
the scope of inquiry by reviewing research on religion and
stress. Finally, select studies on religion in Japan are re-
viewed in an effort to embed the emerging hypotheses in a
proper cultural context.
Religion and Stress
There are several ways in which religion may help pro-
mote the relaxation response when older people are con-
fronted by stressful life events. However, to gain a better
understanding of how this may happen, it is important to
first examine a general issue that must be addressed when
studying stressful life events. There are two broad ap-
proaches to the study of stress. First, some investigators rely
on long checklists that assess exposure to a diverse array of
life events. Here, the goal is to carefully enumerate all life
events a person has experienced during a given time period.
This approach is important because it rests on the plausible
assumption that one cannot understand the impact of any
one event without also examining other stressors that were
present at the same time. In essence, this perspective sug-
gests that it is the cumulative effects of stress that matter
most. The second approach to studying life stressors in-
volves focusing on single life events. Studying events indi-
vidually is important because this strategy allows investiga-
tors to probe more deeply into the rich psychosocial context
surrounding a particular stressor, thereby arriving at a more
complete understanding of how specific coping resources,
like religion, operate.
In the analyses that follow, we focus on a single life
event, the death of a loved one. There are two reasons why
it is important to examine this particular type of stress. The
first has to do with Japanese culture. As Doi (1973) and
many others have pointed out, Japan is a collectivist society.
In this tightly integrated culture, people are highly depen-
dent on each other, and individual identities are overshad-
owed by the group. Individual success arises from group
success, just as individual failure is viewed as a failure of
the group. In fact, Doi coined the term amae to underscore
the high level of interdependence among people in Japan. If
individual lives are fused so tightly, it follows that the loss
of a close other may be especially challenging for older peo-
ple in Japan. Evidence of this may be found in a recent
study by Brannigan (1998). His research revealed that the
death of a close other in Japan “constitutes the death of a
part of one’s self” (p. 165).
Religion and Hypertension
In their comprehensive review of the literature, Levin and
Vanderpool (1989) made a compelling case for why reli-
gion may be associated with hypertension. In fact, they pro-
vided 12 hypotheses that help explain why these constructs
may be related. This important conceptual work, coupled
with a surprisingly large number of empirical studies that
have already appeared in this area, led Koenig and his col-
leagues (2001, p. 259) to argue that if “the salutary effects
of religion are going to be found anywhere, they will be
found in hypertension studies.”
Even though Levin and Vanderpool (1989) developed an
impressive array of hypotheses to explain the relationship
between religion and hypertension, they did not devote suf-
ficient attention to the potentially significant role of stress.
This is important because researchers working outside the
context of religion have been studying the relationship be-
tween stress and hypertension since 1932 (see Gerin et al.,
2000, for a review of this research). Although far from con-
clusive, this secular literature has indicated that people who
are exposed to stressful experiences may be at risk for de-
veloping hypertension. Unfortunately, as Gerin and his as-
sociates pointed out, there is little consensus on the precise
physiological mechanisms linking stress and hypertension,
but most explanations focus on cardiovascular reactivity
(e.g., increased heart rate and arterial wall pressure follow-
ing exposure to stress; Gerin et al., 2000). Nevertheless,
Gerin and his colleagues went on to identify an intriguing
possibility that has not been examined frequently in the lit-
erature. In particular, they argued that instead of being due
solely to cardiovascular reactivity per se, the key risk factor
may involve the person’s ability to return to prestress rest-
ing levels following a period of physiological hyperarousal.
Extensive research by Benson (1996) on the relaxation
response is especially helpful for shedding light on how
some people are able to return to prestress levels of cardio-
vascular functioning. Benson defined the relaxation re-
sponse as a physiological state in which “blood pressure is
lowered and heart rate, breathing, and metabolic rate are de-
creased” (p. 17). Two key findings from Benson’s research
are especially helpful for studies attempting to examine
stress, religion, and hypertension. First, Benson argued that
The second reason for focusing on the death of a loved
one may be found in a series of empirical and theoretical pa-
pers suggesting that religion may be especially useful for
coping with the effects of bereavement. Evidence for this
may be found in the work of Hood, Spilka, Hunsberger, and
Gorsuch (1996, p. 153), who observed that “institutional-
ized faith . . . plays many roles in life, but the issue of death
lies at its core.” Although Hood and colleagues explored a
number of ways that death figures prominently in religion,
they noted that it may be especially helpful for coping with
the loss of a loved one. Evidence of this may be found in re-

S98
KRAUSE ET AL.
search by Mattlin, Wethington, and Kessler (1990). These
investigators examined whether a wide range of coping re-
sponses reduces the deleterious effects of six different types
of stressful events on psychological well-being. Their find-
ings revealed that religion is especially useful for helping
people cope with the death of a loved one, but it is less ef-
fective for dealing with other types of stressors (e.g., inter-
personal difficulties).
If religion is a complex multidimensional phenomenon,
then it may help people deal with grief and bereavement in
more than one way. As noted earlier, three dimensions of
religion are examined in this study: private religious prac-
tices, religious coping, and religious beliefs. In the discus-
sion that follows, we speculate on how each dimension may
offset the noxious effects of losing a significant other. In the
process, we show how each facet of religion may elicit the
relaxation response (Benson, 1996).
less likely to develop hypertension than those who do
not rely on religious coping responses.
Religious beliefs are obviously an integral part of any
faith, yet the study of this conceptual domain remains elu-
sive because the number of specific religious beliefs seems
almost limitless. However, this daunting task is simplified
in studies that focus on a particular type of life event be-
cause researchers can examine only those beliefs that per-
tain to the particular stressor at hand. We capitalize on this
strategy below by focusing on one specific type of religious
belief that is especially relevant for studies of bereavement,
belief in the afterlife. More specifically, we evaluate the po-
tentially salubrious effects of believing that those who die
go on to live in another place that is better than the present
life.
To see why belief in the afterlife may play a key role in
reducing the noxious effects of bereavement, it is important
to examine the psychosocial challenges that arise when a
loved one dies. Three issues are especially important in this
respect. First, existing ties with the deceased are clearly sev-
ered. As a result, those who were once an important source
of social support are no longer directly accessible. Believing
there is life after death holds out the prospects of seeing the
deceased again and speaks to the likelihood of resuming
close ties with them following one’s own death. Second, for
many death is an uncertain outcome. No one has direct
knowledge about whether there is life after death. More-
over, even if there is an afterlife, no one knows for certain
what it is like. Consequently, when a significant other dies,
those who are left behind may worry about what has hap-
pened to someone they love. More specifically, deep con-
cerns may arise about whether the deceased continue to ex-
ist at all and, if they do, whether they are safe, cared for, and
happy. Finally, the death of a significant other clearly brings
thoughts of one’s own mortality to the foreground (Wuth-
now, Christiano, & Kuzlowski, 1980). Believing there is no
life after death, or believing that the deceased may face a
precarious existence in the next life, may be a source of sig-
nificant distress for those who remain behind.
Given these challenges, Marrone (1999) argued that one
of the most important tasks in dealing with the loss of a
loved one involves a cognitive restructuring of the bereaved
person’s worldview. In particular, the bereaved individual
must make sense of what happened and assimilate the death
of a close other into a cognitive schemata that helps them
get past the uncertainties of death and function effectively
in the present life. Marrone maintained that some people are
able to meet this challenge by relying on specific religious
beliefs. In particular, he suggested that “the belief in an af-
terlife may be the most common religion-based assimilation
strategy observed in individuals dealing with loss” (p. 500).
In fact, several empirical studies have shown that those who
believe in the afterlife recover more quickly (i.e., derive a
better sense of well-being) from the death of a loved one
than those who do not believe in life after death (Smith,
Range, & Ulmer, 1991–1992).
A number of specific behaviors are typically included in
measures of private religious practices, including prayers
offered at home when one is alone and the reading of sacred
scriptures. Also, in Japan, people typically make offerings
of water and food to deceased ancestors (Klass, 1996). As
Benson (1996) argued, these regularly occurring ritual be-
haviors help elicit the relaxation response by providing a
temporary respite from troubling thoughts and concerns.
Moreover, as Levin and Vanderpool (1989) pointed out, the
regular practice of religious rituals may also lower the risk
of hypertension by easing a sense of dread, reducing anxi-
ety, allaying fears, and moderating loneliness. The latter
factor (i.e., moderating loneliness) may be especially rele-
vant when a loved one has died.
The rationale developed up to this point leads to the first
hypothesis to be evaluated in this study:
H1: Compared with older adults who do not engage in
private religious practices, older adults who regularly
perform private religious rituals will be less likely to
develop hypertension when a loved one dies.
A vast literature has also suggested that one of the pri-
mary functions of religion is to provide people with an ex-
plicit set of coping responses that helps reduce or offset the
potentially deleterious effects of stress on health and well-
being (Pargament, 1997). Although religious coping is a
complex multidimensional domain in its own right, mea-
sures of this facet of religion typically focus on things like
offering prayers for guidance, direction, and understanding
(i.e., meaning), as well as pleas for the resolution of a diffi-
cult situation. As Pargament pointed out, religious coping is
important for a number of reasons. However, one is espe-
cially important for the purposes of this study. In particular,
he noted that those who engage in religious coping re-
sponses often find a sense of comfort, solace, and relief.
These potential important benefits are closely akin to the
key characteristics of Benson’s (1996) relaxation response.
Consistent with this literature, the analyses presented below
also empirically evaluate the following hypothesis:
Benson (1996) argued that worry is one of the primary
factors that promotes physiological hyperarousal. More-
over, he suggested that factors eliciting the relaxation re-
H2: Older adults in Japan who turn to religious cop-
ing responses when a significant other dies will be

RELIGION AND HYPERTENSION
S99
sponse counteract these effects directly by helping people
dispel their concerns. Cast within the context of bereave-
ment, these observations suggest that belief in the afterlife
may allay worries and concerns about the departed, and the
resulting comfort and solace afforded by these beliefs may
ultimately lower the risk of developing hypertension. Stated
more formally, the third, and final, hypothesis to be exam-
ined in this study states the following:
tal religious orientation is geared specifically toward find-
ing relief from stress. This is captured succinctly in a
Japanese phrase that is identified by Musick and his col-
leagues: Kurushii toki no kamidanomi, which means “turn
to the gods in time of distress.”
Although older people in Japan may be inclined to turn to
religion during stressful times, there is evidence that they
may be especially likely to do so when a significant other
dies. In fact, elaborate rituals that traditionally last 49 days
are often performed when a loved one passes on (see Klass,
1996, for a detailed discussion of these ancestor worship rit-
uals). The purpose of these rituals is to convince the de-
parted that they are dead and help ensure their entrance into
the Buddhist Pure Land (i.e., heaven). However, if the pre-
scribed rituals are not performed, Klass reported the dead
will be turned into harmful spirits, who cause bad things to
happen in the world of the living.
If religion is used to deal with adversity in Japan, and a
good deal of religious ritual focuses on care for the dead,
then it follows that religion may help older people in Japan
avoid the deleterious health effects associated with bereave-
ment. However, as relatively little is known about religion,
stress, and health in Japan, it is difficult to determine
whether religion is an effective stress buffer and, if it is,
how these health-protective effects may arise. The three hy-
potheses derived above were designed to take a modest first
step toward resolving this important issue.
H3: Older adults who believe in the afterlife will be
less likely to develop hypertension following the
death of a loved one than those who do not believe in
life after death.
Religion in Japan
At first glance, religion in Japan may seem puzzling to
those with a Judeo-Christian religious orientation. More
specifically, it is difficult to identify the particular faith of
Japanese people because most belong to more than one reli-
gion. As Miller (1992) pointed out, the majority of people in
Japan practice a mixture of Shinto and Buddhism. This is
reflected in data provided by the Japan Agency for Cultural
Affairs, who reported that 93% of Japanese say they are
Shinto, and 74% also indicate they practice Buddhism (Ka-
sulis, 1990; see also Musick et al., 2000). Although it may
appear that straddling two faiths would create difficulty in
daily life, it is not a major problem because most people use
the two religions for different purposes that reflect differ-
ences in the underlying worldviews of these faiths. More
specifically, themes of birth and life are more evident in
Shinto, whereas Buddhism is more concerned with suffer-
ing and issues related to death. Consequently, it is not un-
usual to have a child “confirmed” by a Shinto priest, yet
have a relative buried by a Buddhist priest (Miller, 1995).
Religion in Japan also differs significantly from Judeo-
Christian orientations because regular institutionalized par-
ticipation in formal worship services is not as deeply in-
grained as regular Sunday worship is in the United States
(Musick et al., 2000). Instead, a good deal of worship in Ja-
pan takes place in the home, where people often maintain
either a Buddhist altar (butsudan) or a Shinto god shelf (ka-
midana). It is especially important for the purposes of this
study to note that the primary function of the butsudan is
veneration of the dead.
Although there are differences between Japanese reli-
gions and those of the West (e.g., Christianity and Judaism),
there are also similarities. For example, research has indi-
cated that, as in the West, older people in Japan tend to be
more religious than their younger counterparts. Evidence of
this may be found in research reviewed by Swyngedouw
(1993). He reported that only 28% of people aged 20 be-
lieve in Shinto deities and only 31% in this age group be-
lieve in Buddhist deities. But for age 70, these figures rise to
65% and 48%, respectively. This suggests that if religion
improves health in Japan, then these effects should be espe-
cially evident among older people in this culture.
METHODS
Sample
The data for this study come from the 1996 (Wave 4) and
1999 (Wave 5) rounds of interviews in the National Survey
of the Japanese Elderly. This ongoing project is conducted
jointly by the University of Michigan and the Tokyo Metro-
politan Institute of Gerontology. Prior waves of interviews
were administered in 1987, 1990, and 1993. All waves of
interviews were conducted face to face in the homes of the
study participants. The survey was designed to create a
panel data set for use in cross-cultural analyses of aging in
Japan and the United States (see Liang, Borawski-Clark,
Liu, & Sugisawa, 1996). In 1996, a supplemental sample (n ꢀ
898) was added to the respondents who participated in ear-
lier waves of interviews (n ꢀ 1,549). This supplemental
survey involved a self-weighted, two-stage probability sam-
ple of Japanese elders aged 60 and older. The sampling de-
sign included 192 primary sampling units stratified by geo-
graphical region and city size. Altogether, 2,447 Wave 4
interviews were completed successfully. Of this group, a to-
tal of 1,929 older people (79%) were reinterviewed success-
fully at Wave 5.
The analyses presented below focus solely on the Wave 4
and Wave 5 interviews because questions on religion were
administered for the first time in the Wave 4 survey. As
noted earlier, including the Wave 5 data allows us to exam-
ine the effects of religion at Wave 4 on change in the proba-
bility of having hypertension at Wave 5.
After taking item nonresponse into account with listwise
deletion of missing data, we based the analyses presented
below on the responses of 1,723 Japanese elders. The aver-
As Musick and colleagues (2000) pointed out, Japanese
religious behavior tends to be focused on specific goals,
such as improvement in health or success in academic en-
trance examinations. Viewed more broadly, this instrumen-

S100
KRAUSE ET AL.
age age of the people in this group at Wave 4 was 69.17
years (SD ꢀ 6.76 years). Approximately 43% of the study
participants were men. Moreover, 71% were married at the
time of the Wave 4 survey. Finally, the respondents in this
group reported that they had completed an average of 9.62
years of schooling (SD ꢀ 2.59 years).
did not have hypertension at Wave 4 indicated they had hy-
pertension when they were reinterviewed at Wave 5.
The fact that we relied on self-reports of hypertension
raises some concern about the accuracy of the data. There
are two ways to address this issue. First, the Japanese gov-
ernment provides health insurance coverage to its entire
population. Consequently, it is not surprising to find that in
1995, people in Japan averaged 16 physician visits per year
compared with only 6 visits per year by people in the United
States (Graig, 1999). This is important because having easy
access to health care, and making a large number of yearly
visits to a health care provider, increases the odds that hy-
pertension will be detected and treated. If this is true, then
there is a good chance older people in Japan would know
whether they have hypertension. The second reason why we
have faith in these self-reports of hypertension may be
found in a recent study by the Japanese Health Services Bu-
reau, Ministry of Health and Welfare (1997). Part of the
protocol for this study involved directly measuring the
blood pressure of a representative nationwide survey of
older people in Japan. Subsequent analysis of these data re-
vealed that 32% of men 70 years of age and older had hy-
pertension, whereas 32% of the women in this age group
had hypertension (Health Services Bureau, Ministry of
Health & Welfare, 1997). These figures are close to those
obtained in our Wave 4 (28%) and (especially) our Wave 5
interviews (32%).
Measures
Table 1 contains the measures of the main study con-
structs examined in this study. Information on the scoring
procedures for these indicators is provided in the table’s
footnotes.
Before reviewing the study measures, it is important to
briefly discuss procedures that were used to address a com-
mon problem in cross-cultural research. Because the key
study measures were initially drafted in the United States,
steps were be taken to ensure they were translated properly
into Japanese and that they would be understood in the man-
ner intended by the research team. As described elsewhere,
we used standard back-translation methods (Marsella, 1987),
a rigorous pilot-testing strategy, and sophisticated confirma-
tory factor analyses to address this issue (see Krause, Liang,
Jain, & Sugisawa, 1998).
Hypertension.—Hypertension was assessed through self-
report in the Waves 4 and 5 surveys. The same question was
used at both interviews. Preliminary analyses revealed that
28% of respondents indicated they had hypertension at
Wave 4, and 32% said they had this chronic illness at Wave
5. Perhaps more important, 191 respondents who said they
Death of a loved one.—The respondents in our study
were also asked if a family member or (separately) a close
Table 1. Study Measures
Measure
Item
1. Hypertension (Waves 4 and 5)a
2. Death of a loved one (Wave 4)b
Do you have high blood pressure?
During the past year did a member of your family die?
During the past year did a close friend die?
3. Private religious practices (Wave 4)c
At my home, I pray and offer prayers for my ancestors (includes offering water, incense, and rice).
How often do you read Buddhist scriptures or the Bible in your home?
How often do you watch or listen to religious programs on TV or the radio?
When facing difficulties, I calm myself when I pray to God or Buddha.
When making an important decision, it is important for me to consult with God or Buddha.
Praying to God or Buddha helps me overcome my stress and worries.
Compared to life on earth, life after death will be a much better place.
Death is just one point on the way to eternal happiness.
How tall are you without shoes on?
4. Religious coping (Wave 4)d
5. Belief in the afterlife (Wave 4)e
6. Body mass index (Wave 4)
About how much do you weigh?
Do you smoke at present?
How often do you do yard work?
7. Smoking (Wave 4)f
8. Physical exercise (Wave 4)g
How often do you exercise or engage in sports?
How often do you take a stroll or walk a considerable distance (over 1 km or ten blocks)?
Do you drink beer, sake, or other alcoholic beverages?
9. Alcohol intake (Wave 4)h
In the past month, on approximately how many days did you drink alcohol?
On days that you drank, how much did you drink?
aThis variable was scored in the following manner: yes, has hypertension (1); no, does not have hypertension (0).
bThis variable was scored in the following manner: had both a family member and a close friend die (2), had either a family member or a close friend die (1), neither
a family member nor a close friend died (0).
cThese variables were scored in the following manner: always (4), sometimes (3), seldom (2), never (1).
dThese variables were scored in the following manner: always (4), sometimes (3), not usually (2), never (1).
eThese variables were scored in the following manner: strongly agree (5), agree (4), uncertain (3), disagree (2), strongly disagree (1).
fThis variable is scored in the following manner: respondent currently smokes (1), respondent does not smoke at this time (0).
gThese variables were scored in the following manner: often (4), sometimes (3), seldom (2), not at all (1).
hThese measures were used to compute the total number of Japanese gos of alcohol consumed in the past month.

RELIGION AND HYPERTENSION
S101
friend had died in the year before the Wave 4 interview.
Preliminary analyses indicated that approximately 17% of
the study participants (n ꢀ 304) had a family member or
friend die during this time. A measure consisting of a simple
count of deaths among family and close friends was used in
the analyses presented.
pants did yard work, engaged in regular physical exercise or
sports, and walked at least 10 city blocks. A high score on
this measure meant a respondent was more physically ac-
tive.
Alcohol intake.—Finally, there is some evidence that
high consumption of alcoholic beverages is associated with
an elevated risk of developing hypertension (Rabbia et al.,
1995). Information on alcohol consumption was assessed at
the Wave 4 survey by asking respondents how often they
had drunk alcoholic beverages in the past month and how
much they typically drank on days they consumed alcohol.
These data were used to compute an overall estimate of al-
cohol consumption that is expressed in terms of a traditional
Japanese go. A go is a container for holding sake. To get a
better sense of the size of a go, it is helpful to point out that
a 350-ml can of beer equals 0.6 of a go.
Private religious practices.—As shown in Table 1, pri-
vate religious practices were assessed at Wave 4 with three
items that asked whether respondents offered prayers at
home, read sacred scriptures, and watched or listened to re-
ligious programs on TV or the radio. A high score on these
items denotes greater involvement in religion. The internal
consistency reliability estimate for this brief index was .572.
Religious coping.—Religious coping was also evaluated
at Wave 4 with three indicators that assessed whether re-
spondents turned to a deity during difficult times, whether
they consulted one of the gods when adversity arose, and
whether they asked a deity to help them overcome the stres-
sors they faced. A high score on this brief composite mea-
sure represents greater use of these religious coping re-
sponses. The reliability estimate for this index is .857.
Demographic control measures.—The relationship be-
tween religion, death of a loved one, and changes in self-
reported hypertension was evaluated after the effects of the
following demographic variables were controlled for statis-
tically: age (scored continuously in years), sex (1 ꢀ male,
0 ꢀ female), marital status at Wave 4 (1 ꢀ presently mar-
ried, 0 ꢀ otherwise), and education (the total number of
years of completed schooling).
Belief in the afterlife.—As shown in Table 1, two ques-
tions were administered in the Wave 4 interviews to see if
study participants believed in life after death. However, these
indicators captured more than just beliefs in the afterlife, per
se. In particular, they also gauged whether a respondent be-
lieved the deceased is going to a better place. Consistent
with the theoretical rationale provided earlier, believing that
a deceased significant other is going on to a better life may
help allay concerns about their safety and well-being,
thereby providing an important prerequisite for eliciting the
relaxation response. Preliminary analyses revealed that the
correlation between the two afterlife items was .433 (p ꢁ
.001).
Data Analysis Strategy
The hypotheses developed for this study specify that reli-
gion will reduce the deleterious effects of bereavement on
change in self-reported hypertension over time. Stated in
more technical terms, this perspective calls for a statistical
interaction between each dimension of religion (i.e., private
religious practices, religious coping, and belief in the after-
life) and death of a loved one on hypertension. The follow-
ing logistic regression equation was used to test for these ef-
fects:
HTN5 ꢀ a ꢂ b1 HTN4 ꢂ b2 DEATH4 ꢂ b3 PRAC4 ꢂ
b4 COPE4 ꢂ b5 AFTER4 ꢂ b6 (PRAC4 ꢃ DEATH4) ꢂ
b7 (COPE4 ꢃ DEATH4) ꢂ b8 (AFTER4 ꢃ DEATH4) ꢂ
Body mass index.—To place greater faith in the main
study findings, controls were included in the analyses for a
number of well-known risk factors for hypertension. The first
was body mass index (BMI). As a number of investigators
have pointed out, high BMI values are associated with a
greater risk of developing hypertension (Lu, Tang, Wu,
Yang, & Chang, 2000). BMI ratios were calculated from
study participants’ self-reported height and weight at Wave 4.
ꢄciZi.
(1)
In this equation, HTN4 and HTN5 denote hypertension at
Waves 4 and 5, respectively; DEATH4 is death of a loved
one at Wave 4; PRAC4 represents private religious practices
at Wave 4; COPE4 stands for religious coping at Wave 4;
and AFTER4 is belief in the afterlife at the Wave 4 inter-
views. Zi denotes the control variables (i.e., BMI, smoking,
alcohol consumption, exercise, age, sex, marital status, and
education). Finally, a is the intercept, whereas the bi and ci
are logistic regression coefficients. Following the recom-
mendations of Selvin (1991), all independent variables were
deviation scored (i.e., centered on their means) before Equa-
tion 1 was solved.
Smoking.—Research has also indicated that people who
smoke are more likely to develop hypertension than individ-
uals who do not smoke (Jackson, Jatulis, & Fortman, 1992).
To control for this risk factor, a simple binary variable was
created from self-reported smoking at Wave 4. Approxi-
mately 22% of the study participants indicated they smoked
at that time.
Equation 1 is estimated in a hierarchical fashion. The ad-
ditive effects of the independent variables are entered in the
first step (i.e., HTN4, DEATH4, PRAC4, COPE4, AFTER4,
and Zi). Following this, tests for the statistical interaction
between the religion measures and death of a loved one are
performed in Step 2 by entering the multiplicative or cross-
Physical exercise.—A number of studies have suggested
that engaging in regular light physical exercise helps to re-
duce the risk of developing hypertension (Nothwehr, Elmer,
& Hannan, 1994). Three questions were administered at the
Wave 4 interviews to determine how often study partici-

S102
KRAUSE ET AL.
product terms into the model (i.e., DEATH4 ꢃ PRAC4,
DEATH4 ꢃ COPE4, and DEATH4 ꢃ AFTER4). If the coeffi-
cients associated with the multiplicative terms are statisti-
cally significant, it is important to determine whether the in-
teraction effects are in the hypothesized direction. We
accomplished this by using the formula provided by Hos-
mer and Lemeshow (1989; see p. 69). Focusing on the inter-
action between death of a loved one and private religious
practices illustrates how this formula was used. If Hypothe-
sis 1 is correct, the effects of the death of a loved one on
changes in hypertension should become progressively
weaker at higher levels of private religious practice. This ef-
fect can be shown clearly by performing some hand calcula-
tions that use the estimates derived from Equation 1. The
formula used for these hand calculations is
they are well cared for. If this is true, then beliefs in the af-
terlife may be correlated with the death of a loved one.
However, if this correlation is sufficiently large, we may
have difficulty disentangling the impact of death of a loved
one on belief in the afterlife from the interaction between
death of a loved one and belief in the afterlife on hyperten-
sion. To evaluate the extent of this problem, the second set
of supplementary analyses presented in the Results focuses
on the relationship between death of a loved one and belief
in the afterlife at Wave 4. Ideally, it would have been nice to
examine the relationships among these constructs with lon-
gitudinal data as well. Unfortunately, data on belief in the
afterlife were gathered only at the Wave 4 survey.
RESULTS
The findings from this study are presented below in three
sections. The potential bias created by the loss of study par-
ticipants over time is evaluated first. Following this, tests of
the three main study hypotheses are presented in the second
section. Finally, results from the supplementary analyses
are reviewed in the third section.
bDEATH ꢀ b2 ꢂ b6 (PRAC4).
(2)
4
In this equation, bDEATH stands for the effects of death of a
loved one on change in4self-reported hypertension at select
levels of private religious practice. The estimates of b2 and
b6 come directly from Equation 1. Equation 2 is solved by
substituting select values of private religious practices
(PRAC4) into the formula. After this, additional formulas
provided by Hosmer and Lemeshow are used to compute
standard errors for the estimates (this permits the calcula-
tion significance tests) as well as odds ratios for each of the
coefficients derived with Equation 2.
Effects of Sample Attrition
Owing to the supplemental sampling strategy that was
implemented at Wave 4, the data from this wave of inter-
views constitute a nationally representative sample of peo-
ple who are at least 60 years of age and living independently
in Japan (i.e., those who do not live in nursing homes).
However, some participants who were interviewed at Wave
4 did not participate in the Wave 5 survey. The loss of par-
ticipants over time in panel studies may bias empirical find-
ings if sample attrition occurs nonrandomly. Although it is
difficult to determine the extent of the problem precisely,
some preliminary insight can be obtained by using select
data from the Wave 4 interviews to see if those who were
lost to follow-up differed significantly from those who re-
mained in the study (see Norris, 1987, for a detailed discus-
sion of this approach). These analyses were performed in
the following manner. First, a binary outcome measure was
computed by assigning a score of 1 to all participants who
were lost to follow-up at Wave 5 and a score of 0 to all who
participated in the Wave 5 interviews. Then, using logistic
regression, this binary outcome was regressed on the fol-
lowing Wave 4 measures: hypertension, death of a loved
one, private religious practices, religious coping, belief in
the afterlife, BMI, smoking, physical exercise, alcohol con-
sumption, age, sex, marital status, and education. If any of
the independent variables are related significantly to the bi-
nary outcome, then it would be reasonable to assume that
sample attrition did not occur in a random manner.
Supplementary analyses.—We performed two additional
sets of analyses that are designed to provide greater confi-
dence in the findings. The first has to do with the causal lag.
As discussed in the previous section, the main analyses
tested for the impact of religion and death of a loved one at
Wave 4 on change in self-reported hypertension at Wave 5.
Stated in more technical terms, this specification calls for a
lagged effect of religion and bereavement on health. In our
case, the lag was 3 years (1996 to 1999). But it is also possi-
ble that the effects of death of a loved one and religion on
self-reported hypertension may appear in a much shorter pe-
riod of time. Again, casting this issue in more technical
terms, one might argue that there are contemporaneous ef-
fects of religion and bereavement on hypertension. One way
to address this issue is to look at the relationships among re-
ligion, death of a loved one, and hypertension using mea-
sures that have all been gathered at the same point in time
(i.e., at Wave 4). Consistent with this rationale, the first set
of supplementary analyses involved estimating the relation-
ships among the Wave 4 waves of religion, bereavement,
and health only.
The second set of supplementary analyses involves a po-
tential problem with confounding. As discussed earlier, the
work of Marrone (1999) suggested that when a significant
other dies, bereaved study participants face critical cogni-
tive restructuring tasks that are geared toward altering their
belief systems so that the loss of a loved one can be inte-
grated into worldviews that guide daily behavior. Cast
within the context of the present study, Marrone’s (1999)
observations suggest that older people in Japan may alter
their beliefs about the afterlife in an effort to convince
themselves that their departed loved ones still exist and that
The sample attrition analyses revealed that the loss of
participants over time did not take place in a random man-
ner. More specifically, the data suggest that, compared with
older adults who were reinterviewed successfully, elderly
people who did not participate in the Wave 5 survey were
more likely to be older (ꢅ ꢀ .047, p ꢁ .001), male (ꢅ ꢀ
.297, p ꢁ .05), and more highly educated (ꢅ ꢀ .083, p ꢁ
.001) and to have a lower BMI (ꢅ ꢀ ꢆ.050, p ꢁ .01). More-
over, older adults who did not participate at Wave 5 were
also less likely to exercise (ꢅ ꢀ ꢆ.118, p ꢁ .001) and less

RELIGION AND HYPERTENSION
S103
likely to engage in private religious practices (ꢅ ꢀ ꢆ.077,
p ꢁ .01). However, significant differences failed to emerge
with respect to Wave 4 hypertension, marital status, smok-
ing, alcohol consumption, religious coping, belief in the af-
terlife, and death of a loved one (a table containing the re-
sults of these analyses is available from Neal Krause). Even
so, the potential bias associated with nonrandom attrition
should be kept in mind as the findings from this study are
reviewed.
cedure involves multiplying an unstandardized logistic re-
gression coefficient by the standard deviation of the inde-
pendent variable. Cast in the context of the present study,
these standardized estimates indicate the change in log-odds
of developing hypertension for a one standard deviation in-
crease in a given independent variable.
Three main findings emerge from the model assessing the
additive effects of the study measures on change in self-
reported hypertension over time (see Model 1). To begin
with, the three dimensions of religion do not appear to exert
a substantial additive effect on hypertension. More specifi-
cally, religious coping responses (ꢇ ꢀ ꢆ.076, ns), as well
as belief in the afterlife (ꢇ ꢀ .035, ns), fail to exert a statis-
tically significant effect. In contrast, private religious prac-
tices have a slight impact (ꢇ ꢀ .161, p ꢁ .05), but the rela-
tionship is not in the anticipated direction. In particular, the
data suggest that a one-unit increase in private religious
practices is associated with about a 7% increase in the odds
of developing hypertension over the course of the follow-up
period (odds ratio [OR] ꢀ 1.072). Although this finding
was not anticipated, similar results have been obtained by
other investigators. More specifically, Levin and Markides
(1985) reported that respondents in their study with high
levels of self-rated religiosity were more likely to have hy-
pertension than respondents who did not feel that religion
was very important to them.
Religion, Death of a Loved One, and Hypertension
Table 2 contains the estimates derived from evaluating
the three main study hypotheses. The additive effects of the
study measures on change in self-reported hypertension
over time are provided in the left-hand portion of this table
(see Model 1), and tests for the interaction between death of
significant others and the religion measures are presented in
the right-hand side of this table (see Model 2). Three coeffi-
cients are provided in each part of the table so that the ef-
fects of the independent variables may be described more
fully. In addition to the unstandardized logistic regression
coefficients (ꢅs) and odds ratios, Table 2 also contains stan-
dardized logistic regression coefficients (ꢇs). Before turn-
ing to the substantive results, it is important to briefly dis-
cuss why standardized estimates are important and how
they are computed.
One problem with logistic regression analysis is that it is
difficult to determine whether the impact of one indepen-
dent variable is greater than that of another. Comparing
odds ratios does not help because, as Selvin (1991) pointed
out, these coefficients are influenced by the metric of the in-
dependent variable. Fortunately, Selvin provided a simple
way of rank ordering (i.e., standardizing) the relative effects
of the independent variables in logistic regression. This pro-
The second finding to emerge from Model 1 has to do with
the effects of bereavement on hypertension. More specifi-
cally, the data in the left-hand portion of Table 2 reveal that
the death of a loved one is not related significantly to change
in self-reported hypertension over time (ꢇ ꢀ ꢆ.066, ns).
Finally, the data in Model 1 indicate that the impact of
some of the control measures on hypertension is similar to
what other investigators have observed. More specifically,
Table 2. The Effects of Religion and Recent Deaths on Changes in Hypertension (N ꢀ 1,723)
Model 1a
Model 2b
ꢅc
ꢇd
Odds Ratio
ꢅ
ꢇ
Odds Ratio
Independent Variable
Hypertension (Wave 4)
Age
Sex
Marital status
Education
2.813**
0.001
ꢆ0.080
0.025
0.005
ꢆ0.016
0.001
1.257
0.003
ꢆ0.040
0.011
0.013
ꢆ0.007
0.011
0.295
ꢆ0.143
ꢆ0.076
0.161
0.035
ꢆ0.066
—
16.666
1.001
0.923
1.025
1.005
0.984
1.001
1.100
0.939
0.973
1.072
1.018
0.849
—
2.849**
0.002
ꢆ0.080
0.040
0.003
ꢆ0.018
0.001
1.273
0.013
ꢆ0.040
0.018
0.009
ꢆ0.007
0.019
0.291
ꢆ0.146
ꢆ0.086
0.167
0.030
ꢆ0.064
—
17.273
1.002
0.923
1.040
1.003
0.982
1.001
1.098
0.938
0.969
1.074
1.016
0.855
0.914
1.044
0.766
Smoking
Alcohol
Body mass index
Exercise
Religious coping
Private religious practices
Belief in the afterlife
Recent deaths
Religious Coping ꢃ Recent Deaths
Private Religious Practices ꢃ Recent Deaths
Belief in Afterlife ꢃ Recent Deaths
0.095**
0.094**
ꢆ0.063*
ꢆ0.028
0.069*
0.018
ꢆ0.164
—
ꢆ0.064*
ꢆ0.031
0.072*
0.015
ꢆ0.157
ꢆ0.090
0.043
—
—
—
—
—
—
—
—
ꢆ0.267**
ꢆ2 log likelihood
Hosmer-Lemeshow goodness of fit
1,559.552
9.633e
1,546.043
3.539e
aModel 1 tests additive effects only. bModel 2 tests for the interaction between recent deaths and the religion measures. cUnstandardized logistic regression coeffi-
cient. dStandardized logistic regression coefficient computed by multiplying the unstandardized logistic regression coefficient by the standard deviation of the indepen-
dent variable. eChi-square value is not significant at the .05 level.
*p ꢁ .05; **p ꢁ .005.

S104
KRAUSE ET AL.
Table 3. The Impact of Recent Deaths on Change in Hypertension
the findings suggest that a higher BMI is associated with an
increase in self-reported hypertension over the course of the
follow-up period (ꢇ ꢀ .295, p ꢁ .001). Similarly, the re-
sults further reveal that older adults in Japan who exercise
more frequently tend to have a slightly lower risk of report-
ing they have developed high blood pressure at Wave 5 (ꢇ ꢀ
ꢆ.143, p ꢁ .05).
Viewed broadly, the data that have been examined so far
suggest that neither religion nor the death of a significant
other appears to have a substantial effect on change in self-
reported hypertension over the course of the study. How-
ever, as the findings in the right-hand side of the table re-
veal, this conclusion may be premature.
at Select Values of Belief in the Afterlife (N ꢀ 1,723)
Value of Belief in the Afterlife
Statistic
2
5
8
10
ꢅa
ꢇb
0.540*
.219
1.716
2.913
1.011
ꢆ0.260
ꢆ.105
0.771
1.086
0.548
ꢆ1.059**
ꢆ.429
0.347
0.700
0.172
ꢆ1.592**
ꢆ.644
0.204
0.565
0.073
Odds ratio
Upper confidence limit
Lower confidence limit
aUnstandardized logistic regression coefficient. bStandardized logistic re-
gression coefficient computed by multiplying the unstandardized logistic regres-
sion coefficient by the standard deviation of the independent variable.
*p ꢁ .05; **p ꢁ .005.
The data provided by Model 2 indicate that neither pri-
vate religious practices (ꢅ ꢀ .043, ns) nor religious coping
(ꢅ ꢀ ꢆ.090, ns) offset the noxious effects of bereavement
on changes in self-reported hypertension (unstandardized
coefficients are reported when focusing on the multiplica-
tive terms because standardized estimates are meaningless
in this context). In contrast, the multiplicative term repre-
senting the interaction between belief in the afterlife and
death of a loved one is significant at the .005 level (ꢅ ꢀ
ꢆ.267).
Although the interaction effect between belief in the af-
terlife and death of a loved one is statistically significant, it
may be somewhat difficult to determine whether this rela-
tionship is in the hypothesized direction. Fortunately, as dis-
cussed earlier, the data in Table 2 can be used in conjunc-
tion with Equation 2 to provide estimates of the impact of
death of a loved one on change in self-reported hyperten-
sion at select values of belief in the afterlife. Scores on the
afterlife measure ranged from 2 to 10. Even though any of
these values could be used to illustrate the interaction effect,
we selected four that are almost equally spaced: 2, 5, 8, and
10. The results of the hand calculations that were performed
by substituting these scores into Equation 2 are presented in
Table 3.
The data in Table 3 reveal that at the lowest possible
score on belief in the afterlife, death of a loved one is asso-
ciated with a sharp increase in the odds of developing hy-
pertension over time (OR ꢀ 1.716). This means that when
older Japanese do not believe in a good afterlife, the odds of
reporting they have high blood pressure following the death
of a significant other increase by about 72%. This finding is
important because approximately 22% of the study partici-
pants indicated they did not believe at all in a good afterlife
(i.e., 381 respondents had a score of 2 on the afterlife mea-
sure).
A value of 5 on the afterlife measure represents the me-
dian score for the older participants in this study. The data
in Table 3 reveal that, at this level, the noxious effects of be-
reavement are offset completely, and there is a slight de-
cline in the odds of developing hypertension following the
death of a loved one (OR ꢀ .771). However, as the upper
confidence limit for this estimate exceeds 1, this finding
must be viewed with caution.
6.55). At an afterlife score of 8, the odds of reporting hyper-
tension at the follow-up interview dropped by about 63%
for older people who experienced the loss of a significant
other but who maintained a strong belief in a good afterlife
(OR ꢀ .347).
Finally, the potentially protective effects of belief in a
positive afterlife become especially apparent among older
people in Japan with the highest score on measure of belief
in the afterlife (i.e., 10). At this level, the death of a loved
one is associated with a fairly substantial drop in the odds of
reporting that one has developed hypertension over time
(OR ꢀ .204). However, this coefficient must be viewed
with caution, because only 35 elderly people had a score of
10 on the afterlife measure.
Supplementary Analyses
The findings presented up to this point suggest there may
be a lagged effect of death of a loved one and belief in the
afterlife on change in hypertension over time. However, as
discussed above, greater confidence may be placed in these
results if we explore other temporal specifications. Follow-
ing the supplemental data analytic strategy discussed ear-
lier, we tested for contemporaneous effects by examining
the relationships among religion, death of a loved one, and
hypertension with data that have been gathered at the same
point in time (i.e., Wave 4). These additional analyses failed
to reveal significant main effects of any religion measure or
death of a loved one on self-reported hypertension. More-
over, these analyses failed to uncover any significant inter-
action effects between the religion measures and death of a
loved one on hypertension (a table containing the results of
these analyses is available from Neal Krause). Coupled with
the findings presented in Table 2, these analyses reveal that
the effects of religion and bereavement on hypertension are
lagged and that it may take some time for the effects of
these independent variables to become manifest.
As discussed earlier, the second set of supplemental anal-
yses involved estimating the correlation between death of a
loved one and belief in the afterlife. The findings reveal that
the two constructs are not related significantly (r ꢀ .006).
Three useful insights emerge from these results. First, belief
in the afterlife does not appear to be stress responsive. In-
stead of altering beliefs to ease the pain of a loss, older peo-
ple in Japan appear to be relying on beliefs that have arisen
The potentially protective effects of belief in a good af-
terlife become evident for those Japanese elders with a
score of 8. This score is slightly higher than one standard
deviation above the mean score for belief in afterlife (SD ꢀ

RELIGION AND HYPERTENSION
S105
from, and are maintained by, other factors. This finding is
consistent with studies in the West, which have shown that
other dimensions of religion are not stress responsive either
(Krause & Tran, 1989). Pargament (1997) provided some
insight into the factors that may be at work here. In particu-
lar, he noted that when people grapple with problematic
events, they turn to cognitive frameworks that are firmly in
place because these belief systems are familiar and plausi-
ble and have frequently worked in the past. Second, these
results do not necessarily negate the points made by Mar-
rone (1999). Rather, they suggest that cognitive restructur-
ing following the loss of a loved one may involve factors
other than belief in the afterlife. Third, and most important,
the additional analysis reveals that data estimation problems
are not likely to have arisen in our study and that the inter-
action between death of a loved one and belief in the after-
life is not confounded by a strong relationship between be-
reavement and belief in the afterlife.
how the deleterious effects of other specific types of stress
are offset by different religious beliefs that are matched
with, and take care of, the needs created by these life events.
Second, beliefs about sacred matters lie at the core of all
religions, yet as Levin (2001) pointed out, religious beliefs
have rarely been used to study physical health outcomes. In-
stead, the little work that has been done with religious be-
liefs focuses primarily on measures of psychological well-
being, such as life satisfaction (Krause, 1993). The findings
in the present study suggest that in addition to studying
well-being, researchers in the West, as well as in Japan,
should pay far more attention to the potentially important
role that religious beliefs may play in the restoration and
maintenance of physical health.
Third, the findings from this study are important because
we examined the interface between religion, stress, and
health in a culture that differs substantially from that of the
United States. This is noteworthy because research in the
West has already shown that belief in the afterlife helps
people cope effectively with the death of a loved one (Smith
et al., 1991–1992). Consequently, our work provides some
evidence that at least this facet of the religion–health con-
nection is fairly robust across cultures.
Fourth, the findings from this study are noteworthy be-
cause they are based on longitudinal data provided by a na-
tionally representative sample of study participants. This
has enabled us to examine the effects of religion and stress
on change in self-reported hypertension over time. This rep-
resents a significant contribution to the literature because, as
Koenig and his colleagues (2001) observed, the majority of
studies on religion and health rely primarily on cross-sec-
tional research designs and samples that are not representa-
tive.
Although the findings from our study may be thought
provoking, a considerable amount of work remains to be
done. For example, research in the West has revealed that
religion is associated with a wide range of health-related
outcomes, including mortality, a number of specific acute
and chronic conditions, and several dimensions of psycho-
logical well-being (e.g., depression; Koenig et al., 2001).
We need to know whether religion also exerts this type of
broad-based effect on health in Japan.
DISCUSSION
Some scholars have concluded that men and women have
believed in life after death since the year 50,000 B.C. (e.g.,
Eliade, 1978). Since that time, many world religions have
embraced the notion that people live on in another world
following death. This is true in Japan, where elaborate ritu-
als have arisen around communication with, and care of,
those who have passed on to the next life (Klass, 1996). If
belief in life after death has flourished for approximately 50
millennia, then it must serve some purpose. The central aim
of the present study was to explore one possibility. In partic-
ular, the main thrust of our analyses was to see if older peo-
ple in Japan who believed in a good afterlife were less likely
to report they developed hypertension following the death
of a loved one than those who did not believe in a good af-
terlife. Data from a nationwide longitudinal survey support
this hypothesis. Viewed broadly, these findings provide evi-
dence of yet another way in which religion may influence
the health of elderly people. However, these results also re-
veal that the beneficial effects of religion appear to be fairly
circumscribed when a significant other dies. Rather than in-
volving private religious practices or more general religious
coping responses, the data from this study suggest that the
potentially salubrious effects of religion arise from specific
beliefs about life after death.
Only three dimensions of religion were examined in this
study. However, investigators have argued for some time
that religion is a complex multidimensional phenomenon
that may be assessed in a number of ways (Ellison & Levin,
1998). Further research is needed to see whether other di-
mensions of religion in Japan are related to health as well.
For example, several studies in the West have examined the
relationship between religiously motivated forgiveness and
health (see McCullough, Pargament, & Thoresen, 2000, for
a review of this research). Research is needed in Japan to
see whether religion is related to forgiveness and, if it is,
whether forgiveness is associated with health. Similarly, re-
search in the West has suggested that religious affiliation or
preference may also have an effect on health (Levin, 2001).
We need to know whether the same is true with respect to
Buddhist and Shinto affiliation in Japan.
There are four reasons why these findings are notewor-
thy. First, studies on religion tend to focus primarily on reli-
gious practices and behaviors while paying less attention to
religious beliefs. By focusing on a single stressful experi-
ence (i.e., death of a loved one), we provide evidence show-
ing how a particular religious belief (i.e., belief in the after-
life) may perform an important stress-buffering function.
This more focused approach to studying religion, stress, and
health is consistent with research in the wider stress litera-
ture on the matching hypothesis (Wills & Shinar, 2000).
According to this perspective, greater insight into the stress
process will be obtained if investigators endeavor to show
how the noxious effects of particular types of stress are off-
set by specific coping resources that directly address the
needs created by the stressor at hand. Researchers wishing
to continue this type of research may benefit by examining
But instead of merely assessing dimensions of religion
that have been identified in the West, research is also

S106
KRAUSE ET AL.
1420 Washington Heights, Ann Arbor, MI 48109-2029. E-mail: nkrause@
umich.edu
needed on facets of religion that are relatively unique to
Asian cultures. For example, research on ancestor worship
was reviewed earlier (Klass, 1996), but we did not measure
this unique dimension of religion directly. Because ancestor
worship is a core component of religion in Japan, it is impor-
tant to devise specific survey measures of it and see whether
ancestor worship is related to health and well-being.
In the process of devising plans for further research, it is
important to keep the limitations of our study in mind.
Three shortcomings are discussed briefly below.
The first limitation in our work has to do with the mea-
surement of religious coping. Three indicators were used in
the present study to assess this construct, but as the work of
Pargament (1997) revealed, religious coping is a vast con-
ceptual domain that contains a large number of dimensions
and a wide range of specific attitudes and behaviors. Clearly,
more extensive measures of religious coping are needed for
further research in Japan.
Second, our measure of hypertension is based on respon-
dent self-reports. This raises obvious questions regarding
self-report bias. Although we presented data earlier that
suggest that self-reports may be reliable, researchers would
benefit from using other ways to assess hypertension, in-
cluding physician diagnoses, medical record checks, or di-
rect measurement blood pressure during the interview.
Third, as reported earlier, the reliability of the private re-
ligious practice measure used in this study is somewhat low
(.572). Although there is not complete consensus in the lit-
erature, some investigators have recommended that internal
consistency estimates should be at least .700 (Nunnally,
1978). However, the issue may not be as straightforward as
it initially appears. Our measure of private religious prac-
tices consists of only three indicators. This is important be-
cause it is well-known that coefficient alpha is sensitive to
the number of items in a scale. More specifically, it is
harder to attain estimates of the size recommended by Nun-
nally (1978) when there are only a few measures in an index
(Cortina, 1993). Even so, those wishing to conduct further
research on private religious practices would benefit from
using more reliable measures.
References
Benson, H. (1996). Timeless healing: The power and biology of belief.
New York: Scribner.
Brannigan, M. (1998). On asking the right questions: Personal death vs.
brain death in Japan. Death Studies, 22, 157–169.
Cortina, J. M. (1993). What is coefficient alpha? An examination of theory
and applications. Journal of Applied Psychology, 78, 98–104.
Doi, T. (1973). The anatomy of dependence. Tokyo: Kodansha International.
Eliade, M. (1978). A history of religious ideas, Volume 1: From the Stone
Age to the Eleusinian mysteries (W. Trask, Trans.). Chicago: Univer-
sity of Chicago Press.
Ellison, C. G., & Levin, J. S. (1998). The religion-health connection: Evi-
dence, theory and future directions. Health Education & Behavior, 25,
700–720.
Gerin, W., Pickering, T. G., Glynn, L., Christenfeld, N., Schwarz, A., Carroll,
D., & Davidson, K. (2000). An historical context of behavioral models of
hypertension. Journal of Psychosomatic Research, 48, 369–377.
Graig, L. A. (1999). Made in Japan: Universal health insurance. In L. A.
Graig (Ed.), Health of nations (pp. 91–120). Washington, DC: CQ
Press.
Health Services Bureau, Ministry of Health and Welfare. (1997). The na-
tional nutrition survey. Tokyo: Author.
Hood, R. W., Spilka, B., Hunsberger, B., & Gorsuch, R. (1996). The psy-
chology of religion. New York: Guilford.
Hosmer, D. W., & Lemeshow, S. (1989). Applied logistic regression. New
York: Wiley.
Jackson, C., Jatulis, D. E., & Fortman, S. P. (1992). The Behavioral Risk
Factor Survey and the Stanford Five-City Project Survey: A compari-
son of cardiovascular risk behavior estimates. American Journal of
Public Health, 82, 412–416.
Kasulis, T. P. (1990). Intimacy: A general orientation in Japanese religious
values. Philosophy East and West, 40, 434–449.
Klass, D. (1996). Ancestor worship in Japan: Dependence and the resolu-
tion of grief. Omega, 33, 279–302.
Koenig, H. G. (1999). The healing power of faith: Science explores medi-
cine’s last great frontier. New York: Simon & Schuster.
Koenig, H. G., McCullough, M. E., & Larson, D. B. (2001). The handbook
of religion and health. New York: Oxford University Press.
Krause, N. (1993). Measuring religiosity in late life. Research on Aging,
15, 170–197.
Krause, N. (1998). Neighborhood deterioration, religious coping, and
changes in health during late life. The Gerontologist, 38, 653–664.
Krause, N., Liang, J., Jain, A., & Sugisawa, H. (1998). Gender differences
in health among older adults. Archives of Gerontology and Geriatrics,
26, 141–159.
The death of a loved one is a truly universal event that is
experienced by virtually everyone around the world. More-
over, the odds of losing a significant other through death in-
crease as people grow older. Religion is also a universal
phenomenon in the sense that it is found in virtually every
major culture in the world. Consequently, it makes sense to
explore the interface between religion and the loss of a
loved one in cross-cultural settings. Doing so provides a ex-
citing forum for exploring how older people around the
world differ and are alike. We hope the findings reported in
this study encourage others to pursue this integral facet of
aging around the globe.
Krause, N., & Tran, T. V. (1989). Stress and religious involvement among
older blacks. Journal of Gerontology: Social Sciences, 44, S4–S13.
Levin, J. S. (2001). God, faith, and health: Exploring the spirituality-heal-
ing connection. New York: Wiley.
Levin, J. S., & Markides, K. S. (1985). Religion and health in Mexican
Americans. Journal of Religion and Health, 24, 60–69.
Levin, J. S., & Vanderpool, H. Y. (1989). Is religion therapeutically signif-
icant for hypertension? Social Science and Medicine, 29, 69–78.
Liang, J., Borawski-Clark, E., Liu, X., & Sugisawa, H. (1996). Transitions
in cognitive status among the aged in Japan. Social Science and Medi-
cine, 43, 325–337.
Lu, F. H., Tang, S. J., Wu, J. S., Yang, Y. C., & Chang, C. J. (2000). Hy-
pertension in elderly persons: Its prevalence and associated cardiovascular
risk factors in Taiwan City, Southern Taiwan. Journal of Gerontology:
Medical Sciences, 55A, M463–M468.
Marrone, R. (1999). Dying, mourning, and spirituality: A psychological
perspective. Death Studies, 23, 495–519.
Marsella, A. J. (1987). The measurement of depressive experience and dis-
order in different cultures. In A. J. Marsella, R. M. Hirschfeld, & M. M.
Katz (Eds.), The measurement of depression (pp. 376–397). New York:
Guilford.
Mattlin, J. A., Wethington, E., & Kessler, R. C. (1990). Situational deter-
minants of coping and coping effectiveness. Journal of Health and
Social Behavior, 31, 103–122.
Acknowledgments
This research was supported by National Institute on Aging Grants R37
AG15124 (Jersey Liang, Principal Investigator), RO1 AG14749 (Neal
Krause, Principal Investigator), and T32 AG00134 (Public Health and Ag-
ing Program).
Address correspondence to Neal Krause, Department of Health Behavior
and Health Education, School of Public Health, University of Michigan,

RELIGION AND HYPERTENSION
S107
McCullough, M. E., Pargament, K. I., & Thoresen, C. E. (2000). Forgive-
ness: Theory, research, and practice. New York: Guilford Press.
Miller, A. S. (1992). Conventional religious behavior in modern Japan: A
service industry perspective. Journal for the Scientific Study of Reli-
gion, 31, 207–214.
Smith, P. C., Range, L. M., & Ulmer, A. (1991–1992). Belief in afterlife as
a buffer in suicidal and other bereavement. Omega, 24, 217–225.
Strawbridge, W. J., Cohen, R. D., Shema, S. J., & Kaplan, G. A. (1997).
Frequent attendance at religious services and mortality over 28 years.
American Journal of Public Health, 87, 957–961.
Miller, A. S. (1995). A rational choice model of religious behavior in Ja-
pan. Journal for the Scientific Study of Religion, 34, 234–244.
Musick, M. A., Traphagan, J. W., Koenig, H. G., & Larson, D. B. (2000).
Spirituality in physical health and aging. Journal of Adult Develop-
ment, 7, 73–86.
Norris, F. H. (1987). Characteristics of older nonrespondents over five
waves of a panel study. Journal of Gerontology, 40, 627–636.
Nothwehr, F., Elmer, P., & Hannan, P. (1994). Prevalence of health behav-
iors related to hypertension in three blood pressure treatment groups: The
Minnesota Heart Health Program. Preventive Medicine, 23, 362–368.
Nunnally, J. C. (1978). Psychometric theory. New York: McGraw-Hill.
Pargament, K. I. (1997). The psychology of religion and coping: Theory,
research, and practice. New York: Guilford.
Swyngedouw, J. (1993). Religion in contemporary Japanese society. In
M. R. Mullins, S. Susumu, & P. L. Swanson (Eds.), Religion and so-
ciety in modern Japan (pp. 49–72). Berkeley, CA: Asian Humanities
Press.
Wills, T. A., & Shinar, O. (2000). Measuring perceived and received social
support. In S. Cohen, L. G. Underwood, & B. H. Gotlieb (Eds.), Social
support measurement and intervention: A guide for health and social
scientists (pp. 86–135). New York: Oxford University Press.
Wuthnow, R., Christiano, K., & Kuzlowski, J. (1980). Religion and be-
reavement: A conceptual framework. Journal for the Scientific Study of
Religion, 19, 408–422.
Rabbia, F., Veglio, F., Russo, R., Schiavone, D., Oliva, S., & Chiandussi,
L. (1995). Role of alcoholic beverages in essential hypertensive pa-
tients. Alcohol and Alcoholism, 30, 433–439.
Received April 26, 2001
Selvin, S. (1991). Statistical analysis of epidemiologic data. New York:
Oxford University Press.
Accepted June 20, 2001
Decision Editor: Fredric D. Wolinsky, PhD