A century of suicide in Japan
By William Wetherall
First drafted August 1992
Last updated October 2002
This is a draft of an article in progress.
Please to not quote or cite without permission.
Tables, Figures, Graphs, and Maps
Patterns and Trends
1. Sex and Age
2. The Unfortunate Generation
3. The Yukko Effect
4. Delayed Marriage
5. Less Risk, More Concern
Notes and References
Tables, Figures, Graphs, and Maps
|Table 1||Male suicide rates, by year and age group|
|Table 2||Female suicide rates, by year and age group|
|Table 3||Elderly male suicide rates, crude and adjusted by year and age group|
|Table 4||Elderly female suicide rates, crude and adjusted by year and age group|
|Figure 1||Artificial data Illustrating Contour Interpolations|
|Graph 1||Middle-age male suicide rates, rate-by-age-group curves by year|
|Graph 2||Middle-age male suicide rates, rate-by-year curves by age group|
|Map 1||Topography of male suicide rates, by year and age|
|Map 2||Topography of female suicide rates, by year and age|
|Map 3||Topography of male suicide/death ratios (logarithmic)|
Japan's suicide rates are at record, or near-record, lows for most age groups of both sexes. Rates for the second half of the 20th century average lower than those for the first half, for all age groups, but are especially lower for younger people, particularly women. Suicide rates have been notably higher for the generation born between 1930 and 1940, especially men. While suicide rates have been falling, other mortality rates have been falling much faster, and so suicide as a cause of death has been rising, particularly among younger people. And as more young people have been marrying later in life, their suicide rates have been peaking at higher ages. Such trends are most dramatically shown with topographic maps representing nearly a century of suicide data.
However, Japan's population has been rapidly aging. And changes in age structure are greatly biasing crude suicide rates, not only for the entire population, but also for large age groups, such as those of the elderly, or that of all people aged 65 and older taken as one group. To make such rates comparable over a period of years, they must first be adjusted to a common--usually earlier--population.
I wish to thank Drs. Masaaki Kato, Tooru Takahashi, Kazuko Kamiya, and George De Vos, and Rev. Yukio Saito, for their encouragement and support. Parts of this paper first appeared in Japanese.
I also wish to thank the library staff at the National Institute of Mental Health, at the National center of Neurology and Psychiatry, in Japan, where I carried out most of my research on suicide, first as a Research Fellow, then as an Associate Researcher in the Department of Adult Mental Health.
This article examines nearly a century of Japan's official suicide statistics by sex and age. It also features age-adjusted rates and topographic maps, to illustrate major changes and trends, and to dispel a number of misconceptions that have been nurtured by innumeracy.
Suicide statistics are numbers, and bureaucrats, journalists, and scholars are vulnerable to innumeracy: difficulty understanding numbers. In Japan, too, innumeracy fosters misreporting, misuse, and other abuses of suicide statistics, and allows doomsayers to press the slippery numbers into their alarmist social commentaries. The most primitive form of innumeracy is the use of body counts instead of rates to measure suicide frequency. But innumeracy is also evident in the use of crude rates when comparing suicide frequencies in different or changing populations.
This article argues that, to improve statistical comparability, it is necessary to compensate for differences or changes in demographic composition by adjusting suicide rates to a common population. It also introduces, for the first time in English, the author's application of contour mapping techniques to the plotting of three-dimensional suicide statistics. The topographies created by contour plotting both clarify familiar patterns observable from tables and graphs, and reveal new patterns not readily discernible with conventional methods.
Suicide statistics in Japan are independently compiled by two government groups: the Ministry of Health and Welfare (MHW) and the National Police Agency (NPA). MHW suicide statistics are part of the mortality data generated from cause-of-death information recorded on death certificates filed at local municipal offices. NPA suicide statistics are based on inquest data compiled by local police.
MHW was established in 1938, but its statistics are available for all years from 1899, except 1944 and 1945 and the first nine months of 1946. [n 1] NPA was created in 1954, but nationwide police data on suicide are available for all years from 1878 except 1944 and 1964-1977. Police data for Tokyo prefecture go back to 1874, while nationwide police data combining suicide with other "unnatural deaths" go back to 1876. [n 2]
MHW publishes suicide data by sex, age, method, month, prefecture, place, occupation, marital status, and nationality. In addition to sex, age, method, month, and place, NPA classifies suicide by motive, type of household, and cases involving more than one person.
Of these two sets of suicide statistics, MHW's are more "official" in the sense that they are reported to WHO. But NPA's figures get more media attention because they are made public two months earlier, are more sensationally presented, and are limited to suicide (whereas MHW's report covers all causes of death and other vital statistics).
A few archival studies have noted some superficial differences between NPA and MHW statistics. [n 3] One field study has investigated structural disparities in local data, but raises more questions than it answers. [n 4] In recent years, the totals and sex/age breakdowns of NPA and MHW suicide tallies have been converging, apparently because of changes in NPA's compilation criteria. [n 5]
There are no simple answers to questions about the reliability and validity of suicide statistics in any country. Japan has its share of public scandals that expose attempts to cover up suicide for the usual human motives. And undoubtedly many other cases of unintended if not intentional misclassification never come to light.
The only practical problem is how misclassification affects gross counts and their sex, age, and other distributions. Short of knowing more than may be possible about data errors, however, one assumes that Japan's suicide statistics are good approximations of what they are meant to be: objective tallies of suicide.
Considering the pervasiveness of Japan's bureaucracies, and their dedication to social control and the counting of all things that foster control, including vital events, Japan's suicide figures are probably not less reliable than those of any other reporting country. Suicide tallies are probably within ten percent of their supposedly true values, while age, sex, and other patterns are most likely not greatly distorted by unequal distributions of errors. Population counts are probably no less accurate.
A far more serious problem is how some bureaucrats, journalists, and scholars have been mishandling Japan's suicide statistics. One result is a misinformed public. Another is misguided social policy.
The impression is often given that Japan's suicide rates have been increasing with economic success, especially among the elderly, who are often portrayed as victims of family nuclearization. In fact, suicide rates for all age groups in Japan have generally, if not steadily, been decreasing (Table 1 and Table 2). The topographies of decline are shown on contour maps, plotted here for the first time, spanning nearly a century of data (Maps 1 and 2).
Specific examples of how bureaucrats, journalists, and academics in Japan unwittingly help spread misinformation about suicide in Japan abound. [n 6] The most telling case outside Japan is found in Mamoru Iga's thesis that suicide in Japan has been increasing with economic growth. [n 7]
Preconvictions led Iga to take deceptive newspaper accounts of misleading NPA statistics as evidence of increase. Iga observed that the suicide rate had peaked in 1955, declined until 1967, and then risen again. He cited successively higher rates for 1970 and 1974, and then stated that "the rate seems to be still rising"--all because a major newspaper had reported that "'suicides both in Tokyo and the nation hit a record high'" in 1983. [n 8] In fact, the "record high" referred to an NPA body count, not a suicide rate. Iga got some of his rates from a 1977 special MHW report on suicide, [n 9] but he cited the higher crude rates and ignored the much lower age-adjusted rates in the report's main table. [n 10]
NPA's press summary on suicide, and media reports and editorials based on it, focus on body counts: raw tallies of how many people have committed suicide. Emphasis is placed on numbers of cases, and changes in case load, rather than rates of occurrence in a standard population. So even when the suicide rate has decreased, if the police have handled more cases, then NPA will report that suicide is increasing. While such reports may be correct as statements about cases, they are too commonly misconstrued to mean that suicide is being committed more frequently.
Unfortunately, NPA's attitude toward the reporting of statistics is shared by many bureaucrats, journalists, and academics who facilely associate increasing body counts with social degeneration. In the transition from number to image, "record high" becomes "worst in history" and otherwise encourages all manner of doomsday analyses. Some educators and policy makers have blamed the "increase" in suicide on the "Americanization" of education and the "nuclearization" of the family. Even a few suicide preventionists have confused the public picture of trends by failing to distinguish body counts from epidemiologic rates, or by overlooking the effects of demographic changes on rates.
Problems with statistics are compounded by a fashionable urge on the part of many analysts, Japanese and non-Japanese alike, to impute cultural meanings to ciphers that contain little cultural information. Though cultural determinists and relativists have created the impression that Japan is the Galapagos Islands of human behavior, clinical evidence suggests that suicide in Japan is essentially no different than suicide in other countries.
In Japan, too, suicidal acts are committed by people in reaction to complex mixes of personal, social, and other factors that may have been a lifetime in the making. The factors that induce people in Japan to kill themselves run the human gamut. Culture, as a concept invoked in exceptionalist explanations that people in Japan kill themselves for peculiar reasons, has more to do with how the living are compelled by ideologies and stereotypes to regard acts like suicide. But how and why many social scientists are prone to confuse cultural (religious, social, political) explanations of suicide (which the living contrive for their benefit) for the behavior of suicide itself (which seems to have changed much less over time than the explanations of it) is another story. [n 11]
One statistical problem with suicide rates is that their stability declines with population size. Japan's 125 million total population, though one of the world's largest, is too small to yield stable annual suicide rates in the most elderly age groups when it is broken down into 5-year age groups by sex. This problem is partly solved by averaging over 5-year periods (as was done one way in Table 1 and Table 2, and another way on Maps 1 and 2). Apart from problems with the size of the age-group population, 5-year age groups represent a sufficiently small increment (differential) of the age curve to be comparable over a period of years during which the entire age distribution is rapidly changing.
For populations with different or changing compositions, rates become less comparable as the increment of the age-group increases. Thus crude rates for an entire population (no matter how large), or for a significant part of the population (such as all elderly people aged 65 and older), will be biased by the composition of the population in terms of sex, age, occupation, marital status, and other traits that are used to compile suicide statistics.
To compare the suicide rates for different or changing populations that vary in composition, it is necessary to compute them relative to a common population (whether natural or contrived). Though compensations for differences and changes in population composition are most commonly made with respect to age (though even age-adjusted mortality rates are rarely found in suicide literature), the basic adjustment procedure can be applied to any trait; and if cross-trait data is available, it can be compounded for more than one trait.
A crude rate for an entire population is computed by dividing the sum of suicides in each age group of the population, by the sum of the age-group populations. This method minimizes the effects of instability because it weighs the suicides in each age group in proportion to the population of the age group that produced the suicides.
An ideal total rate would equally weigh each age-group, so the rate would be independent of age composition. In principle, such a rate can be obtained by averaging the age-group rates. Unfortunately, this method is vulnerable to instability in small-population age-group rates.
A compromise method is to adjust the rates of, say, populations X and Y to a common population, which could be X, Y, or Z. The rate for X can be adjusted to the age composition of Y, for example, by computing the rate for X as though it had been produced by Y. This is done by taking the age-group rates of population X, multiplying them by their respective Y age-group populations, and then dividing the sum of these products (which correspond to the numbers of suicides that X's suicide rates would have produced with Y's age composition) by the sum of all the age group populations of Y (the total population of Y).
Such adjustments are rarely made, probably because they are not understood or require more effort to compute, or for technical reasons like lack of necessary data. When made, they usually take into account only age, though sometimes sex, since cross-trait data is almost always available for these two factors. In principle, though, multiple adjustments can be made for all mutually crossed traits.
No population is unbiased. But with respect to age composition, this drawback does not necessarily limit our ability to compare rates that have been adjusted to a common population. While adjustments for all factors known to affect suicide would be ideal, taking into account at least age, relative to practically any population, will usually improve the comparability of data from different periods or places.
While age-adjusted mortality rates have been officially computed for a number of major causes of death in Japan, for several decades, no one has indicated the need for adjusted rates for large age-groups like all elderly people aged 65 and older. Table 3 and Table 4 show crude and age-adjusted rates for males and females. Note that since the 1970s for men and the 1960s for women, the rates adjusted to the 1935 census (column i) become increasing lower than the crude rates (column g), and that the percentage difference (column j) accelerates about the same as the acceleration in the percent of the elderly population (column e).
The percentage of the disparity between the crude and adjusted rate of an aged population signifies the extent to which the crude rate is biased too high in comparison with that of an unaged population. This interpretation also applies to age-adjusted morbidity or mortality rates for any disease or cause of death which is characterized (as is suicide in Japan) with an incidence that increases with age.
Though not shown here, the disparities between the crude and age-adjusted rates for Japan's total population have become even greater. These disparities reflect mainly the population aging that began to in the 1950s. The transition from a population pyramid with a wide base and a rapidly tapering top like Eiffel Tower, to a more less tapered distribution that reflects lower fertility and mortality rates, has been smooth except for the trough created by World War II, and the peaks created by the postwar baby boom and its second generation (which is just now coming of age). This aging, which is gathering considerable speed, is expected to peak by 2020 or 2025, when roughly one-fourth of Japan's total population will be elderly-a very top heavy pyramid.
Since older people have higher suicide rates than younger people, population aging means that the crude suicide rate would increase even though the incremental age-group rates were to remain constant. If the rate of aging is high enough, the crude rate can increase even as the age-group rates slightly decrease. This means that, when comparing past and present suicide rates in Japan, or rates in Japan with rates in other countries, one must consider the effects of demographic change, such as population aging (though other kinds of change may also affect the total suicide rate).
Japan's MHW has been computing age-adjusted mortality rates for a number of major causes of death, by prefecture and for all Japan, since 1960. From 1992, some MHW statisticians began to adjust mortality rates relative to the age composition of a model population based on the 1985 national census, and they intend to make these adjustments retroactive to 1947. This will not facilitate prewar-postwar comparisons; and other technical problems may limit the usefulness of the corrected rates. But as an experiment with a promising methodology, it is a step in the right direction. The main difficulty will be to educate others-not only the general public, but also other bureaucrats, journalists, and academics, including suicidologists-in the rationale, use, and interpretation of adjusted rates.
Suicide rates are usually plotted two-dimensionally, either as year curves or as age-group curves. A family of such curves can be superimposed on the same plot (Graphs 1 and 2); but when doing so, the three-dimensional relationship (surface topography) between the curves is difficult to read. The curves on such plots can be untangled, and the three-dimensional relationship can be clearly revealed, by viewing the table of suicide rates used to generate the curves as the matrix of a topography.
A table of suicide rates by year and age-group is equivalent to the elevation data a surveyor would obtain by setting out a grid of stakes at regular north-south (year) and east-west (age) intervals, and then using a level and rod to shoot the elevation at each stake. To show the shape of the surface of the land, the surveyor has to estimate where between two adjacent grid points the elevation of the land is the same as between other pairs of grid points. Joining these interpolated points together results in a contour.
Interpolations (usually linear) are made along the sides of each cell in a grid, or each set of four data points that form the corners of a square or rectangle. The process of developing a contour map from a table of data is shown in Figure 1. In the grid, observed values are shown in bold face, interpolated points are shown in italics, and the contour lines are drawn by connecting the interpolated points. The coordinates of the observed rates are taken as the year and the central value of the age group. The coordinates of the interpolated points are computed in relation to the coordinates of the observed rates. [n 12]
For example, if X is the year, Y the age, and Z the suicide rate, then the coordinates (X, Y, Z) of three points in the upper-left corner of the grid are (1970, 22, 45), (1970, 24.5, 50), and (1970.5, 22, 50). The first of these three points is an observed rate. The second and third points, both interpolated rates, can be connected because they represent two adjacent points on the same contour.
The contour lines in Figure 1 show part of the slope of a mountain that rises toward the upper-right corner of the grid. The 70 contour in the upper-right corner is adjacent to a much longer 60 contour that meanders down and to the left, and then switches back to the right. The slope is evenly steepest on its south-southwest face in the lower-right corner, and of less average steepness on its west-northwest face toward the upper-left corner; both of these faces fall below the 50 contour, but at different rates. The fall-off is the least steep along the ridge that runs west toward the left side, and turns west-southwest toward the lower-left corner; and the terrain remains above the 50 contour.
Any table of data such as that shown in Figure 1 can be plotted three-dimensionally as a relief map. But the features of the terrain represented by the matrix of values in the table can also be projected two-dimensionally as a set of contour lines that pass through points in the table that are interpolated to be of equal value.
Topographic maps of suicide rates have an advantage over conventional graphs in that their isorate lines more clearly show how rates have simultaneously been changing in terms of year and age. An isorate line is the contour passing through a set of year-age points at which the interpolated suicide rate is the same, just as on a weather map an isobar connects points that have equal barometric pressure at a given altitude.
Map A, shown as a box on Map 1, is the result of interpolating contours from the table of rates that were used to plot Graphs 1 and 2. Graph 1 shows a set of vertical cross-sections for selected years (as though looking at the topography of Map A from the bottom to the top), while Graph 2 shows cross-sections by age group (as though looking at Map A from the left to the right). Note that on both graphs, the dotted horizontal lines for rates 20, 30, 40, and 50 intersect the curves at points corresponding to the contour lines for these rates on Map A.
Maps 1 and 2 show the topographies of male and female suicide rates for all available year/age data, at a contour interval of 10. [n 13] These maps are read like any topographical map using elevation contours. That is, the closer the lines, the steeper the slope, and the more rapid the vertical change in the terrain at right angles to the lines.
Map 3 uses an anti-log function interval to show the patterns of change in the percentage of suicide among all causes of death. Since this is the equivalent of using semi-log graph paper (on which quadratic functions like a parabola appear as a straight line), the fairly even distances between the logarithmically spaced contour lines, going left to right from the youthful highs to the elderly lows, mean that the suicide/death ratio tends to drop off exponentially with age.
Patterns and Trends
Japan's suicide rates do not support the thesis that modernization drives more people to kill themselves, for they have generally declined to record lows in most age groups of both sexes. This decline does not, of course, mean that modern life has made people happier; suicide rates may measure a population's collective will to live, but not necessarily the quality of its existence, since the religio-political purpose of life may be to suffer in peace or to die in war.
The inability of suicide rates to answer ontological questions has not, however, stopped observers of the human condition from speculating that civilization is in decline, as is evident in some Japanese reports which are laden with a nostalgic variety of social commentary that falls beyond the pale of science. [n 14] Here I wish to consider the data in the tables, graphs, and maps as pure numbers, rather than as indicators of change in human vitality or social well-being.
1. Sex and Age
Table 1 and Table 2 show average suicide rates for five-year periods, and average rates for early, prewar, postwar, and recent periods, by sex, based on MHW suicide counts from 1899 through 1991. Maps 1 and 2 represent the topographies of the full set of annual suicide rates that were used to compute the averages in the tables.
The tables show that suicide rates in the first half of the 20th century averaged higher than they have so far for the second half. Moreover, excepting men and women in their 20s, average rates for the quarter-century period before World War II were higher than rates for the quarter-century period after the war. Thus suicide rates do not support the contention of romantic nationalists that life in Japan (as measured by suicide) was better in "the good old days" before the war.
The tables also show that, for all except 5-9 and 35-59 year-old males and 10-14 year-old females, the average suicide rates of the most recent four-year (Present) period, are lower than those of any of the earlier (Early, Prewar, Postwar, Recent) periods. For all age groups of both sexes, except middle-aged males in their late 40s and early 50s, the 1991 suicide rates are lower than the 1900 rates.
Suicide rates have fallen considerably more for younger people than for older people, but more for younger women than for younger men, and (exceptionally) more for middle-aged and older men than for women of these age groups. Older male rates generally peaked shortly before the war, while older female rates generally peaked shortly after the war.
While the rates for older people have not fallen as rapidly as those for younger people, they have nonetheless been steadily falling. Alarmist reports of "increases" in elderly suicides have been based either on body counts that naturally increase because the population is rapidly aging, or on crude rates that fail to account for changes in the age composition of the population.
2. The Unfortunate Generation
Table 1 and Map 1 generally, and Map A on Map 1 (Graphs 1 and 2) more specifically, show how middle-aged male rates rose during the 1970s, peaked in the mid 1980s, and are now retreating. Table 2 shows (as would Map 2 with a smaller contour interval) that middle-aged female rates have similarly been influenced by what seems to be a generational effect.
It is far from clear why the 1925-1945 generation, especially those born in the 1930s, should have produced such high suicide rates. In its formative years-before, during, and after World War II-this generation experienced considerable death, disruption, and deprivation. It set record suicide rates both in the 1950s when it was in its teens and twenties, and again in the 1980s when its survivors were in their forties and fifties.
Be as they may, these patterns could be entirely coincidental, in that they may have arisen from non-generational factors that just happen to have affected mainly the members of one generation. Unfortunately, speculation about generational causes does not lend itself to the kind of scrutiny that might explain why even under the most traumatic social conditions, few individuals kill themselves, while the vast majority of people are willing to struggle through life as it comes.
Over the past few decades, a number of attempts have been made, some more successful than others, to determine the generational effects of suicide. In all these cohort studies, though, the data would have been graphically clearer had topographic maps been used.
3. The Yukko Effect
Suicide contagion occurs when suicide spreads through a vulnerable population like an infectious disease. This phenomenon, known from antiquity, and sometimes called the "Werther effect" after the hero of Goethe's 18th century novel, is now added and abetted by mass media.
The most recent statistically observable case of suicide contagion in Japan occurred in the spring of 1986, when idol teenage singer Okada Yukiko leapt to her death under very sensational and even scandalous conditions. [n 15] In the weeks following Okada's death, cause-of-death statistics reflected two to three times the expected number of youth suicides, many of which were infectiously linked with Okada's death by suicide notes and other evidence. The contagion was statistically large enough to be clearly evident in the topographies of Maps 1 and 2.
4. Delayed Marriage
Marriage is generally associated with lower suicide rates. In Japan, too, married people of all generations have much lower suicide rates than never married, divorced, or widowed people. And so the ages by which most young men and women marry will affect how their suicide rates vary with age.
Table 1 and Map 1 show that sometime in the 1960s, the young male peak shifted from the 20-24 to the 25-29 age group. This phenomenon is even more evident on Map 3, which shows the percentage of suicide among all deaths by year and age. This shift is mainly due to the fact that the age at which most men marry has been drifting from the early and late 20s to the late 20s and early 30s.
More women, who tend to marry a few years younger than men, are also either remaining single longer or never marrying. Table 2 shows that the age peak in young female suicides did not shift from the lower to the upper twenties until around 1980. Map 2, with a smaller contour interval (say 5), would also show this shift.
5. Less Risk, More Concern
Maps 1 and 3 suggest why there has been so much concern about suicide among younger people compared to older people. Though these two maps show only male rates and ratios, the females maps are similar. For men in their 20s, the suicide/death ratio broke the 5 percent level in the 1910s, 1920s, and 1930s. After this, the ratio has twice peaked at over 30 percent, first in the 1950s, and again in the 1980s.
The peak in the 1950s corresponds to record high suicide rates, but it also reflects early postwar declines in general mortality. This decline, especially among younger people, accelerated so much that by the 1980s, the suicide-death ratio climbed to a comparable peak, even though the suicide rates barely reached even the lowest prewar levels, which were only one-forth to one-half the postwar records of the 1950s.
While many factors fuel the flares of attention that mass media pays to youth suicides, two seem unrelated to the actual suicide rate (which rises and falls like commodity prices): first, baby boomers, which will soon swell the older age groups, are generating more cases of suicide, even though their suicide rates have been steady or falling; and second (related to Map 3), diseases that once overshadowed suicide as a cause of death among teenagers and young adults have been controlled or eliminated, so that most deaths in these age groups are accidents or suicides-both of which, though at lower rates, have a greater impact as a death experience in the family and community.
The contour lines on Map 3 very slightly drift to the right for middle-aged and older people, as one moves down from the middle of the century to the present. This means that among these age groups, also, suicide is becoming more conspicuous as a cause of death. But as age increases, the percent of suicide as a cause of death rapidly falls, so that even the slight drift to the right of the contour lines for the elderly represents very little increase in the impact of suicide as a cause of death among older people. While suicide rates continue to be relatively higher for older people, suicide remains a minor cause of death among the elderly. As the population rapidly ages, however, the number of cases of elderly suicide will probably increase for purely demographic reasons; and without regard to the suicide rate, which will probably continue to decline, the higher body counts will undoubtedly (and perhaps deservedly) get a certain amount of attention as a form of death experience for those who live by the numbers.
Studies of suicide in Japan should not be limited to statistics. At the present time, however, only statistics afford a panoramic and relatively objective overview of suicide behavior such as that presented in this article. The general patterns revealed here could be correlated with cultural, economic, and political events and developments, but with great risk of nurturing ecological fallacies. In any event, sweeping socio-cultural theories are not very helpful if ultimately one wants to tackle the more vital problems of motivation, and possibly prevention.
While there are many excellent case and community studies of suicide in Japan, the incompatibility of their diverse criteria, if not their small scale, leaves much to be desired when it comes to advancing general theories of behavior. Clinical and other studies are extremely sporadic in terms of both the kinds and numbers of cases reported, and the quantity and quality of information disclosed. Without a concerted effort on the part of counselors, clinicians, police, coroners, medical examiners, and epidemiologists to collaborate in long-term nationwide research projects that collect and collate data according to common criteria, suicidology in Japan will remain a fairly primitive science, as it generally does elsewhere. Comparative studies inevitably rest on even shakier scientific footings, and the achievement of higher quality global research will require even greater planning and coordination.
Notes and References
- I have found MHW data for all years except those noted. Okazaki gives a total for 1944, but I cannot confirm it. See Okazaki Ayanori, Jisatsu no shakai tokei teki kenkyu [A social statistical study of suicide], Tokyo: Nihon Hyoron Shinsha, 1960, page 35.
- I have located police statistics for all other years since 1876 except 1887, 1888, 1937, and 1938. For an account of Japan's first suicide statistics, see Takahashi Jiro, "Honpo jisatsu tokei no raireki" [The beginning of suicide statistics in this country], Tokei shushi [Statistics abstracts], 351, May 1910, pages 397-399.
- Okazaki Ayanori, Statistical study, op. cit. n. 1, pages 34-36; Koseisho Daijin Kanbo Tokei Joho Bu [Ministry of Health and Welfare, Minister's Secretariat, Statistics and Information Department], ed, Jisatsu shibo tokei: Jinko dotai tokei tokushu hokoku [Suicide death statistics: Vital statistics special report], Tokyo: Kosei Tokei Kyokai [Health and Welfare Statistics Association], 1990, page 148; Uezarooru Uiriamu [William Wetherall], "Jisatsu tokei no goho to goyo: Kancho, masukomi, oyobi gakkai ni okeru 'sujimosho' no shindan to chiryo" [The misreporting and misuse of suicide statistics in Japan: A diagnosis and treatment of innumeracy in the bureaucracy, mass media, and academia], Seishin hoken kenkyu [Journal of Mental Health], 4(37), 1991, (forthcoming).
- Yoshida Koji, "Hokkaido ni okeru jisatsu (Jinko dotai tokei to keisatsu tokei to no hikaku)" [Suicide in Hokkaido (A comparison of vital statistics and police statistics)], Jisatsu yobo to kiki kainyu [Suicide prevention and crisis intervention], 13, May 1989, pages 20-29.
- Wetherall, "Misreporting", op. cit. n. 3, Table 4.
- For a detailed account of the widespread misuse of statistics in the bureaucracy, mass media, and academia in Japan, see Wetherall, "Misreporting", op. cit. n. 3.
- Mamoru Iga, The Thorn in the Chrysanthemum (Suicide and Economic Success in Modern Japan), Berkeley: University of California Press, 1986.
- Ibid., page 13.
- Ibid., page 14ff., notes to tables.
- Koseisho Daijin Kanbo Tokei Joho Bu [Ministry of Health and Welfare, Minister's Secretariat, Statistics and Information Department], ed, Jisatsu shibo tokei: Jinko dotai tokei tokushu hokoku; Showa 49 nendo jinko dotai (shibo) shakaikeizaimen chosa hokoku [Suicide mortality statistics: Vital statistics special report; 1974 vital (mortality) socioeconomic survey report], Tokyo: Kosei Tokei Kyokai [Health and Welfare Statistics Association], 1977, page 27.
- For part of this story, see William Wetherall, "Japan's Anti-Suicide Traditions", in Essays on Japanology, 1978-1982, compiled and translated by the International Cultural Association of Kyoto, Kyoto, 1983, pages 105-123 (Japanese translation pages 125-142). An abridged version appeared as "Anti-Suicide Traditions in Japan: Past and Present", The Japan Times, 14 February 1981, pages 10-11.
- Mathematical details on computational methods are presented in Uezarooru Uiriamu [William Wetherall], "Jisatsu no topogurafiizu: Toritsusen ni yoru gurafikku hoho" [The topography of suicide in Japan: A graphic method using isorate lines], Kosei no shihyo [Journal of Health and Welfare Statistics], 38(15), December 1991, pages 27-35.
- I first drew contour maps of suicide rates in Japan by hand, using coordinates individually computed with a programmable "electronic slide rule" (HP-55), in 1975. Since 1990, I have used a slightly modified version of a commercial Japanese contour map program written in BASIC. I run the program through MS-DOS on an NEC 9801, using data converted to BASIC files from spreadsheets created with a Japanese version of Lotus 1-2-3. The program will also color the contour intervals on a color monitor or printer. The maps in this article were constructed with high-tech scissors and paste from the computer printouts of the contour plots, plus headings, rulings, and notes produced with a word processor. The maps were labeled and marked with dry decals and a ballpoint pen.
- See, for example, Kichinosuke Tatai, "Suicide in the Elderly: A Report from Japan", Crisis, 12(2), September 1991, pages 40-43.
- For an analysis of Okada's death, its media coverage, and the contagious aftermath, see William Wetherall, "Japanese youth and the Yukko syndrome", Far Eastern Economic Review, 17 July 1986, pages 45-47.