Asian-American Deaths Near the Harvest Moon Festival

Gary Smith, Ph.D.
Fletcher Jones Professor
Department of Economics
Pomona College
Claremont, California 91711
gsmith@pomona.edu
909.607.3135
Fax 909.621.8576

* I am very grateful for the careful reading and detailed suggestions of the editor and three expert referees.

Abstract

Key words: death postponement, mortality dip, Harvest Moon Festival

Asian-American Deaths Near the Harvest Moon Festival

Numerous studies have investigated the theory that stressful life events can be unhealthy, even fatal.^1-4 A very different theory is that the anticipation of a meaningful event, such as a birthday or religious holiday, can prolong life. If this theory is valid, health care providers should be sensitive to the effects of symbolically meaningful occasions on at-risk patients, and medical researchers should interpret their data with care when deaths occur near symbolically meaningful events.

To date, the data have not supported the death-postponement theory conclusively. For example, Phillips and King report a dip in Jewish death rates preceding Passover; however, the decedents in this study were not necessarily Jewish, but rather Californians whose names seemed Jewish.⁵ A study using data from a large Jewish mortuary in California found no evidence that Jews were able to postpone death until after the celebration of religious holidays.⁶ For deaths that occurred near Passover, there were somewhat more deaths (particularly male deaths) during the weeks before Passover than during the weeks afterward.

Another example is a study by Idler and Kasl of elderly New Haven residents which indicated that some Christians and Jews postponed their deaths until after the celebration of religious holidays.⁷ However, the p values were calculated incorrectly and the correct p values are larger than reported and make their conclusions less convincing, especially for Jews.⁸

One of the most important Chinese holidays is the Harvest Moon Festival (HMF), which occurs on the 15th day of the eighth moon of the lunar calendar. The family customarily gathers for bonding and celebration and eats a festival meal including traditional moon cakes outdoors in the moonlight at midnight. Phillips and Smith analyzed California mortality data for the years 1960-1984 and concluded that some elderly Chinese-American women evidently postpone their deaths until after the celebration of the Harvest Moon Festival.⁹ Their strongest evidence is a death dip in the week preceding the Harvest Moon Festival and a rise in the week afterward for Chinese women at least 75 years old. However, they also observe that, “It is a scientific truism that statistical significance alone is not sufficient to justify substantial confidence in the reliability of a finding. It is crucial that the finding be replicated also.” (p. 1947)

We now have 16 years of independent data on mortality near the Harvest Moon Festival that can be used to test the theory that elderly Chinese women are able to postpone their deaths. In addition, Koreans and Vietnamese also use the Chinese lunar calendar and celebrate the Harvest Moon Festival on the same day as the Chinese. According to the PBS show Hidden Korea: “Perhaps the greatest Korean holiday is Ch’usok, or Harvest Moon Festival.”¹⁰ The AdoptVietnam web site says that, “The Mid-Autumn Moon Festival is rated second only to Tet as the most favorite Vietnamese holiday.”¹¹ Independent mortality data for Korean-, and Vietnamese-Americans can also be used to determine whether the earlier finding is replicated.

METHODS

Phillips and Smith analyzed deaths recorded by the California Department of Health Services for the years 1960 through 1984. Records are now available through the year 2000.¹² The entries for racial and ethnic status identify Chinese for the entire period, Vietnamese since 1978, and Koreans since 1985. All deaths from natural causes that occurred within seven days of the Harvest Moon Festival are analyzed. Any day-of-the-week mortality patterns are avoided by examining two 7-day periods, each of which contains every day of the week.

The most straightforward test of the research hypothesis that there is a difference in the number of deaths during the seven days before and after the festival is a test of the null hypothesis that for each death within seven days of the festival, there is a 0.5 probability that it occurs in the week before the festival. The binomial distribution gives the appropriate p-value.

There may be confounding influences due to the timing of a deadly event (such as an earthquake) during the weeks studied. It is also true that, particularly with elderly people, the probability of death increases with each passing day though the number of people still living diminishes. It is unlikely that death rates will change much over such a short interval of time. Nonetheless, a way to control for confounding influences is to compare the death patterns to a control group, such as white Californians, for whom the Harvest Moon Festival has no special significance. To compare Asian-American deaths to the white control group, Fisher’s exact test for a 2-by-2 contingency table was used with the data separated by race and by whether the death occurred during seven days before or after the festival.

Phillips and Smith argue that one-sided p values are appropriate; however, Phillips, Van Voorhees, and Ruth argue that holidays can be either a lifeline (prolonging death) or a deadline (provoking death).¹³ If true, two-sided p values should be used, and these are reported here.

Phillips and Smith justify their focus on women at least 75 years old by arguing that, “The senior woman of the household takes charge of preparing an elaborate meal.” However, age 65 was used by one of the coauthors to identify elderly Jewish males in the study mentioned earlier of Jewish deaths near Passover. The present paper examines ages 65-74 and at least 75 for the original 1960-1984 data and at least 75 for 1985-2000 replication tests.

Another issue is how to categorize deaths that occur on the Harvest Moon Festival day. The Passover study notes that, “The main ceremonial ritual of Passover (the retelling of the Exodus story at a large family meal) takes place on the evening before [emphasis added] the first day of Passover.” (p. 729); the authors consequently count deaths that occur the following day as having been postponed until after the celebration of Passover. The main ceremonial ritual of the Harvest Moon Festival occurs at midnight at the end of the HMF day. The logic used in the Passover study implies that deaths occurring before the HMF’s main ceremonial ritual should be categorized as before the ceremony; deaths that occur after midnight should be counted as after the celebration. Instead, Phillips and Smith count deaths that occur on the HMF day as having occurred after the celebration. The present results are tabulated both ways.

RESULTS

For the 1960-1984 period studied by Phillips and Smith, Table 1 shows that the data are not statistically persuasive for women in the age group 65-74, no matter how deaths on the festival day are recorded. Contrary to the postponement hypothesis, the p values are actually above 0.5 for binomial tests comparing Chinese deaths before and after the festival and for Fisher tests comparing Chinese deaths to the white control group. For women 75 and older, the data are not statistically persuasive unless a death on the Harvest Moon Festival day is considered to have occurred after the festival.

We now have data for the years 1985-2000 and we have data for Koreans and Vietnamese as well as Chinese women. Table 2 shows the results using the Phillips/Smith cutoff of 75-years old. No matter how deaths on the festival day are recorded, elderly Chinese women had more deaths during the week before the festival than during the week after. The same is true of elderly Vietnamese women. Elderly Korean women had either more deaths before or the same number before and after. None of these differences are statistically persuasive. In fact, for these three ethnic groups, the only p values less than 0.05 are for young Chinese women using the Phillips/Smith rule for counting deaths on the festival day: they died in unusually high numbers during the week before the festival. This shouldn’t be taken too seriously since for the three ethnic groups, four age/gender groups, and two rules for counting deaths on the festival day, only 1 of 24 categories has a two-sided p value less than 0.05.

A possible explanation for this failure to replicate the results is that the festival may be less important now than in the past. However, the annual difference between the number of deaths during the week before and after the Harvest Moon Festival does not support this explanation. The first graph in Figure 1 shows these data using the Phillips/Smith categorization of deaths that occurred on the festival day as having occurred after the festival. If the festival has lost importance over time, one would expect to see a trend, with before-minus-after gradually going from negative to neutral. In fact, 1984 shows the largest before-minus-after disparity in the 1960-1984 data, and 1998 shows the largest before-minus-after disparity in the 1985-2000 data. For the five years 1980-1984, there were 15 deaths before and 32 after; for the next five years (1985-1989) there were 39 deaths before and 40 after. There was not a gradual change in the before-minus-after disparity over time; the disparity was largest at the end of the Phillips/Smith data period and abruptly stopped when their data stopped.

This interpretation is confirmed by an examination of the annual before-minus-after data using the categorization of deaths that occurred on the festival day as having occurred before the festival. The second graph in Figure 1 shows again that the largest disparity is at the end of the 1960-1984 data set and that there is no trend in these data.

DISCUSSION

It is possible that the day of death may sometimes be misreported. However, it is unclear which way this bias would work and how it might be quantified. Phillips/Smith do not consider this possibility and the objective of the present paper is simply to see whether the application of their methodology to new data supports their conclusion.

An odd feature of the 1960-1984 data used by Phillips and Smith is that 15 elderly Chinese women died on the Harvest Moon Festival day, which was the most deaths on any day within a week of the festival. One explanation might be that the stress associated with the festival is sometimes fatal; another possibility is that this was a fluke—which is supported by the fact that there is no similar spike in the 1985-2000 data.

The statistical significance of the results reported for 1960-1984 hinges on whether the 15 women who died on the festival day are counted as having successfully prolonged their lives. Since the main festival activity is the midnight meal in the moonlight, it seems reasonable to count these people as not having been able to postpone their deaths for the festival celebration. If a person is really able to postpone death until after the celebration of an important ceremonial occasion, shouldn’t she be able to postpone death until after the main ceremonial activity?

In situations where there is no compelling theoretical rationale for categorizing the data, it is especially important to determine whether the statistical results are replicated with independent data. Here, the results reported for elderly Chinese women for the years 1960-1984 are not replicated for elderly Chinese, Korean, or Vietnamese women for the years 1985-2000.

References

1. Herbert TB, and Cohen S Health Psychology: Psychological Factors and Physical Disease from the Perspective of Human Psychoneuroimmunology, Annual Review of Psychology 1996; 47:11342.

2. Byrne DG. Personality, life events and cardiovascular disease, Journal of Psychomatic Research 1987; 31:661-671.

3. Tennant CC, Palmer KJ, Langeluddecke PM, et al. Life event stress and myocardial reinfarction: a prospective study. European Heart Journal 1994; 15: 472-478.

4. Tennant CC. Experimental stress and cardiac function. Journal of Psychomatic Research 1996; 40: 569-583 .

5. Phillips DP, King EW. Death takes a holiday: Mortality surrounding major social occasions. The Lancet 1988; 2: 728–732.

6. Lee P, Smith G. Are Jewish Deathdates Affected by the Timing of Holidays?, Social Biology 2000; 47: 127–134.

7. Idler EL, Kasl SV. Religion, disability, depression, and the timing of death. American Journal of Sociology 1992; 97: 1052–1079.

8. Smith G. Idler and Kasl’s p Values: A Cautionary Lesson, Psychosomatic Medicine forthcoming.

9. Phillips DP, Smith DG. Postponement of death until symbolically meaningful occasions. JAMA 1990; 263: 1947–51.

10. Hidden Korea. Retrieved September 24, 2002, from http://www.pbs.org/hiddenkorea/chusok.htm

11. Adopt Vietnam. Retrieved September 24, 2002, from http://www.adoptvietnam.org/parenting/autumn.htm

12. California Department of Health Services, Death Statistical Master Files, 1960-2000. Sacramento, CA: California Department of Health Services (CD-ROM data files).

13. Phillips DP, Van Voorhees CA, Ruth TE. The birthday: Lifeline or deadline? Psychosomatic Medicine 1992; 54: 532–542.

Table 1 Deaths of Elderly Chinese Women Within a week of the Festival, 1960-1984

Note: The binomial test uses the number of deaths during the 7 days before and after the festival to test the null hypothesis that a death during this 14-day period is equally likely to be before or after the festival. The Fisher test compares Asian deaths during the 7 days before and after the festival to the white control group.

Table 2 Deaths Within a Week of the Harvest Moon Festival, 1985-2000 (old is >= 75)

Counting deaths on the HMF day as having occurred after the festival

Counting deaths on the HMF day as having occurred before the festival