Are Jewish Deathdates Affected by the Timing of Important Religious Events?
Peter Lee
Postbaccalaureate Premedical Program
Columbia University
New York, New York 10027
Gary Smith
Fletcher Jones Professor of Economics
Pomona College
Claremont, California 91711
Abstract
Earlier studies reported a decline in September mortality in New York City and Budapest during years when Yom Kippur was in the interval September 28 through October 3, and fewer deaths among Californians with Jewish surnames during the week preceding Passover than during the week after Passover. These studies suggest that some Jews are able to postpone their deaths until after the celebration of an important religious event. We reexamine these findings using new data and find no statistically persuasive evidence that Jewish deaths decline before religious holidays. We do find an increase in deaths in the weeks shortly before and after birthdays.
Introduction
Phillips and Feldman (1973) reported evidence of a decline in deaths before
Yom Kippur, which they called the holiest day of the Jewish year,
suggesting that some Jews are able to postpone death until after the celebration
of this religious event. In a later study, Phillips and King (1988) reported
fewer deaths during the week preceding Passover than during the week after Passover.
The data in both studies are questionable because the decedents were not necessarily
Jewish, and it is puzzling that the authors did not report results for both
holidays using each data set. In this paper, we use data on 5,111 verified Jewish
decedents to reconsider the question of whether Jewish deaths are affected by
the timing of religious holidays and birthdays.
Review
The Phillips-Feldman study examined mortality data for New York City during
a time (19211969) when its population was 28 percent Jewish and Budapest
(18751915) when its population was 22 percent Jewish. These authors report
that September mortality rates during ten years in New York City and eleven
years in Budapest when Yom Kippur was in the interval September 28 through October
3 were generally lower than in adjacent years. Despite the large sample sizes,
the conclusion that some Jews postpone their deaths until after Yom Kippur is
weakened by the fact that most of the residents of these cities were not Jewish.
Phillips-Feldman also report that religious affiliation is identified in five years of the Budapest data, and that there is a September deathdip among Jews in four of these five years and among non-Jews in only two of these years. In the two years with a deathdip among non-Jews, the deathdip among Jews was somewhat larger. These results are suggestive, but hardly definitive. In addition, it is puzzling that Phillips-Feldman did not report a comparison of the number of September deaths to the number of deaths in other months (such as October), particularly since this is the procedure they use when they investigate the possibility of a deathdip before birthdays.
The Phillips-King study used data from California computerized death certificates that do not identify the religion of the decedents. Instead, Phillips-King compiled a list of surnames that are probably Jewish because they are mentioned prominently in a Jewish reference book and also listed at least ten times in the central Los Angeles telephone directory. They report fewer deaths among those with Jewish surnames (particularly males) during the week preceding Passover than during the week after Passover. They also find statistically significant patterns in the 24-week period surrounding Passover, though the patterns are not unambiguous. There were more deaths than expected three to five weeks before Passover, fewer than expected during the two weeks immediately before Passover, more than expected in the week after, and fewer than expected in the next two weeks.
While it is true that Los Angeles has a significant Jewish population, the number of surname listings in this telephone book is surely a weak predictor of whether or not a person is indeed Jewish. In addition, there were inevitable ambiguities; for example, Phillips-King excluded people named Ash or Bach, but included people named Asher or Brody.
It is not certain which religious holiday is the most meaningful for individual Jews, and unsettling that these two studies report results for different holidayssuggesting a willingness to look for patterns near several holidays and report those that are the most statistically persuasive. There are four important Jewish holidays that might plausibly be studied. Yom Kippur is the Day of Atonement and is devoted to fasting, prayer, and repentance. At Passover, relatives come together for a Sedar service recounting the story of the Exodus of the Jewish people from Egypt. Channukah is the Festival of Lights, which commemorates the victory of the Maccabees and rededication of the Temple. Rosh Hashanah marks the Jewish New Year and the start of the Ten Days of Penitence. The significance of these holidays varies from person to person.
A number of other studies have reported conflicting or ambiguous findings. For example, Idler and Kasl (1985) report marginally significant results for a sample of New Haven Jews; however, their sample was small and they miscalculated their p values. For example, their lowest reported p value is 0.019 for 23 more-observant Jews who died within 30 days of any of three different holidays (7 before and 16 after). They used a normal approximation with a continuity correction for the probability of fewer than 7 deaths. Instead, they should have calculated the probability (using either the exact binomial or a normal approximation) of 7 or fewer deaths; the exact binomial probability is 0.047. Of their 20 statistical tests for Jews, only this one has a one-sided p value less than 0.05.
Nonsuicidal people seldom have precise information about when they will die and, other than their choice of medical treatment, it is not clear that they have any effective means of changing the timing. If we knew our death dates and had the power to change them, wouldnt everyone who was not in extreme pain choose to live longer? Our research hypothesis is that the timing of Jewish deaths is not affected by Jewish holidays; earlier results to the contrary may have reflected data limitations or the selective reporting of statistical tests.
Data
To investigate the timing of Jewish deaths near these four holidays, we examined
information provided by Sinai Memorial Chapel in San Francisco about all 5111
services performed for Jews who died between January 1, 1987 and December 31,
1995. These persons were not only definitely Jewish, but felt strongly enough
about their faith to be given services in a Jewish mortuary. We compared these
death dates to the dates of the four major Jewish holidays: Channukah, Passover,
Rosh Hashanah, and Yom Kippur. A complicating factor is that Rosh Hashanah and
Yom Kippur are always nine days apart. If there turn out to be few or many deaths
in this interval, we will not know if this is because the interval is after
Rosh Hashanah or before Yom Kippur.
The Jewish day begins and ends at sundown and we used Phillips-Kings procedure of classifying those persons who died on the morning after the evening services as having died after the celebration of the holiday. Because Phillips-Feldman and Phillips (1972) report that some persons are able to postpone their deaths until after the celebration of birthdays, we also compared the deathdates of those in our sample to their birthdates. (Royer and Smith, 1999, reexamine Phillips original claim.)
We looked at three different time periods centered on each holiday (2 weeks and 24 weeks, as did Phillips-King, and 8 weeks, following Phillips-Feldmans consideration of the month preceding Yom Kippur), and also divided our data by sex (as did Phillips-King)resulting in 45 separate statistical tests: three time periods for females, males, and both sexes for five holidays. This proliferation of tests increases the chances of incorrectly rejecting at least one null hypothesis. Our tests are, of course, not independent since the time periods overlap, the data for both sexes are an aggregation of the female and male data, and the proximity of a death date to one holiday is not unrelated to its proximity to other holidays. Nonetheless, we should assess the results of 45 tests conservatively. One way to do so is to require lower p values for statistical significance.
If the timing of deaths is independent across individuals, then the binomial distribution is appropriate for testing the null hypothesis that a death that occurs during a specified time interval has a 0.5 probability of occurring in the first half of the interval. Although some earlier studies suggest fewer deaths before an event, we use two-sided p values as we cannot a priori rule out the possibility that the anxious anticipation of an event increases the probability of dying.
For each sex category and holiday, we used the binomial distribution to compare (as did Phillips-King) the number of deaths during the 7-day period preceding the holiday and during the 7-day period following the holiday. Because Phillips-Feldman looked at the month preceding Yom Kippur, we also used the binomial distribution to compare the number of deaths during the 4-week periods preceding and following each holiday. Phillips-King also report the number of deaths during each of the 12 weeks preceding Passover and each of the 12 weeks following Passover. We consequently categorized our data in this fashion too, and used a chi-square statistic to test the null hypothesis that deaths during this 24-week period are equally likely to be in each of the 24 weekly categories.
Results
Table 1 summarizes the results. Only two of these 45 tests have p values less
than 0.05; one relates to Passover and one to birthdays. Although these results
should be interpreted cautiously because of the large number of tests conducted,
it is noteworthy that the lowest p values both contradict the conclusions reported
in Phillips-King and Phillips-Feldman.
Table 2 shows the female, male, and combined deaths during the 12 weeks preceding and following Passover. The column labels show the expected number of deaths each week if the deaths are equally likely to occur in each week. Females had somewhat fewer deaths than expected during the three weeks preceding Passover and somewhat more the week after. For males, the pattern is reversed with more deaths than expected during 5 of the 6 weeks preceding Passover and fewer deaths than expected in 4 of the 5 weeks following Passover. Combining the male and female data, there are more deaths than expected in 6 of the 7 weeks preceding Passover and fewer than expected in 6 of the 7 weeks following Passover. Figure 1 displays these combined female and male data.
The two-sided p values for a binomial comparison of the number of deaths during the four weeks preceding and following Passover are 0.038 for males and 0.098 for both sexesin each case, these reflect an increase in deaths before Passover and a dip afterward, the reverse of the pattern reported by Phillips-King.
Figure 2 shows the power functions for the 2-week and 8-week Passover data. The 8-week test is more powerful because of the larger number of observations in this interval. If the actual probability of dying in the first half of a 2-week interval is 0.43, there is a 50 percent chance of observing a death dip that is statistically significant at the 5 percent level. If the actual probability of dying in the first half of an 8-week interval is 0.46, there is a 58 percent chance of a statistically significant death dip. If the mortality effects are even weaker than this, a larger sample is needed to detect them reliably.
Table 3 shows female, male, and combined deaths during the twelve weeks preceding and following birthdays; Figure 3 displays the combined data. The largest number of deaths is during the week preceding the birthdaycontrary to Phillips theory that some people can postpone their deaths until after the celebration of a birthday. If anything, the relatively large number of deaths shortly before and after birthdays suggest that the stress associated with this milestone may sometimes be fatal. Our results can perhaps be reconciled with Phillips-Feldman by the fact that they compared the deathmonth with the birthmonth, while we compared the deathday with the birthday. Similarly, Grigsby (1985) found an increase in deaths during the birthmonth in the U.S. National Mortality Survey.
Discussion
We find no persuasive evidence that Jews can postpone their deaths until after
the celebration of religious holidays or birthdays. For Jewish male deaths that
occur near Passover, the case emphasized by Phillips-King, the patterns in our
data are the opposite of those reported by them, with somewhat more deaths during
the weeks before Passover than during the weeks afterward. Our data do suggest
that there may be an increase in deaths in the weeks shortly before and after
birthdays
REFERENCES
Grigsby, J. 1985. Special Occasions,
Stress, and Mortality: Do People Tend to Die During Their Birth Month? Social
Biology 32: 102114.
Idler, E. L. and S.V. Kasl. 1985. Religion. Disability, Depression, and the
Timing of Death. American Journal of Sociology 97: 10521079.
Phillips, D. 1972. Deathday and birthday: an unexpected connection. In J. Tanur
(ed.) Statistics: a Guide to the Unknown. Holden-Day, San Francisco, 5265.
Phillips, D. and K. Feldman. 1973. A Dip in Deaths Before Ceremonial Occasions:
Some New Relationships Between Social Integration and Mortality. American Sociological
Review 38: 678696.
Phillips, D. and E. King. 1988. Death Takes a Holiday: Mortality Surrounding
Major Social Occasions. The Lancet ii: 728732.
Royer, H. and G. Smith. 1999. Can the Famous Really Postpone Death? Social Biology
45, 302-305.
Table 1 The Statistical Results
Binomial
Tests
|
Chi-Square
Test for
|
|||||||
2 Weeks
|
8 Weeks
|
24 Separate
Weeks
|
||||||
Before
|
After
|
p
|
Before
|
After
|
p
|
p
|
||
Channukah | ||||||||
Females |
61
|
59
|
0.927
|
213
|
232
|
0.394
|
0.838
|
|
Males |
47
|
48
|
1.000
|
173
|
197
|
0.232
|
0.876
|
|
Both |
108
|
107
|
1.000
|
386
|
429
|
0.141
|
0.573
|
|
Passover | ||||||||
Females |
51
|
59
|
0.505
|
213
|
206
|
0.769
|
0.319
|
|
Males |
48
|
41
|
0.525
|
197
|
157
|
0.038
|
0.786
|
|
Both |
99
|
100
|
1.000
|
410
|
363
|
0.098
|
0.150
|
|
Rosh Hashanah | ||||||||
Females |
58
|
57
|
1.000
|
217
|
198
|
0.377
|
0.862
|
|
Males |
38
|
45
|
0.510
|
158
|
162
|
0.867
|
0.939
|
|
Both |
96
|
102
|
0.722
|
375
|
360
|
0.606
|
0.732
|
|
Yom Kippur | ||||||||
Females |
53
|
50
|
0.844
|
215
|
203
|
0.591
|
0.863
|
|
Males |
47
|
41
|
0.594
|
167
|
171
|
0.870
|
0.668
|
|
Both |
100
|
91
|
0.563
|
382
|
374
|
0.799
|
0.596
|
|
Birthday | ||||||||
Females |
66
|
57
|
0.471
|
234
|
239
|
0.854
|
0.056
|
|
Males |
59
|
57
|
0.926
|
186
|
194
|
0.720
|
0.546
|
|
Both |
125
|
114
|
0.518
|
420
|
433
|
0.681
|
0.025
|
Table 2 Twelve Weeks Before and After Passover
Females
|
Males
|
Combined
|
|
Week
|
(expected
= 53.6)
|
(expected
= 45.0)
|
(expected
= 98.6)
|
-12
|
67
|
40
|
107
|
-11
|
69
|
48
|
117
|
-10
|
69
|
51
|
116
|
-9
|
62
|
54
|
116
|
-8
|
58
|
38
|
96
|
-7
|
54
|
45
|
99
|
-6
|
54
|
51
|
105
|
-5
|
53
|
47
|
100
|
-4
|
64
|
55
|
119
|
-3
|
53
|
49
|
102
|
-2
|
45
|
45
|
90
|
-1
|
51
|
48
|
99
|
1
|
59
|
41
|
100
|
2
|
52
|
45
|
97
|
3
|
46
|
32
|
78
|
4
|
49
|
39
|
88
|
5
|
44
|
34
|
78
|
6
|
49
|
49
|
98
|
7
|
49
|
48
|
97
|
8
|
57
|
47
|
104
|
9
|
50
|
41
|
91
|
10
|
49
|
40
|
89
|
11
|
50
|
48
|
98
|
12
|
38
|
45
|
83
|
Total
|
1287
|
1080
|
2367
|
Table 3 Twelve Weeks Before and After Birthdays
Females
|
Males
|
Combined
|
|
Week
|
(expected
= 53.9)
|
(expected
= 44.4)
|
(expected
= 98.3)
|
-12
|
36
|
46
|
82
|
-11
|
52
|
43
|
95
|
-10
|
40
|
43
|
83
|
-9
|
45
|
42
|
87
|
-8
|
55
|
37
|
92
|
-7
|
53
|
52
|
105
|
-6
|
49
|
50
|
99
|
-5
|
49
|
42
|
91
|
-4
|
63
|
41
|
90
|
-3
|
49
|
41
|
90
|
-2
|
56
|
45
|
101
|
-1
|
66
|
59
|
125
|
1
|
57
|
57
|
114
|
2
|
62
|
44
|
106
|
3
|
70
|
46
|
116
|
4
|
50
|
47
|
97
|
5
|
61
|
52
|
113
|
6
|
69
|
49
|
118
|
7
|
53
|
41
|
94
|
8
|
55
|
32
|
87
|
9
|
57
|
41
|
98
|
10
|
62
|
42
|
104
|
11
|
42
|
38
|
80
|
12
|
43
|
35
|
78
|
Total
|
1294
|
1065
|
2359
|
Figure 1 Total Deaths During the Twelve Weeks Before and After Passover
Figure 2. Power Functions for Combined Deaths Near Passover
Figure 3. Total Deaths During the Twelve Weeks Before and After Birthdays