Article Text

Download PDFPDF

School and seasonality in youth suicide: evidence from Japan
  1. Tetsuya Matsubayashi1,
  2. Michiko Ueda2,
  3. Kanako Yoshikawa1
  1. 1Osaka School of International Public Policy, Osaka University, Toyonaka, Osaka, Japan
  2. 2Department of Political Science, Syracuse University, Syracuse, New York, USA
  1. Correspondence to Dr Tetsuya Matsubayashi, Osaka School of International Public Policy, Osaka University, 1-31 Machikaneyama, Toyonaka, Osaka 560-0043, Japan; tetsuya.matsubayashi{at}


Background Seasonality in youth suicide has been speculated to be associated with the school calendar, as it tends to increase at the beginning of the academic year or after a long break, but robust empirical evidence remains scarce.

Methods We examined the nationwide death records in the Vital Statistics of Japan to investigate the seasonal patterns of suicide among youth. Our data set included 108 968 suicides by individuals who died at 6–26 years of age between 1974 and 2014 in Japan. The daily frequencies of death were plotted against the Japanese school calendar, which has little regional and temporal variations. We also estimated a Poisson regression model to uncover the cyclical patterns of suicide deaths.

Results We found that the frequencies of suicide by middle school students (ages 12–15 years) and high school students (ages 15–18 years) sharply increased around the dates when a school session began in April and September. These tended to be low during school breaks. The results of regression analysis suggested middle school students were more than twice as likely to die by suicide when the summer break ended and the second semester began, compared with the baseline week in July. Similarly, the frequency of suicide for high school students also increased by ∼40% at the end of the summer break. Importantly, no such pattern was found for those aged 18–26 years.

Conclusions Our findings strongly indicate that the cyclical pattern of youth suicide is closely related to the school calendar.


Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Numerous studies have examined seasonality in suicides—a cyclical change in the frequency of suicide across seasons. Empirical support for the seasonal pattern has been found across various countries, including Belgium, Finland, France, Germany, Greece, Greenland, Hong Kong, Ireland, Italy, Japan, Norway, Slovenia, South Africa, Spain, Switzerland, Taiwan, the UK and the USA.1–3 Although there is some evidence to suggest that the seasonality has declined in recent years, the frequency typically increases in the spring and early summer months; however, the pattern varies across regions, times, and methods.1

When focusing on seasonality in youth suicide, however, the literature remains inconclusive. For example, using data from the USA, McCleary et al4 reported unimodal seasonality in the fall and winter among male youth aged under 16 years, while Frank and Lester5 and Hoberman and Garfinkel6 found no such pattern among those aged between 15 and 24 years, and under 20 years, respectively. Similar heterogeneous results were also found in other European nations, such as England and Wales,7 France,8 ,9 Germany,10 ,11 and Italy,2 ,12 and Asian nations such as Singapore13 and China.14

Further, the mechanism underlying the cyclical pattern in youth suicide, if such a pattern exists, remains unclear. In general, seasonal variations are associated with psychiatric and physical disorders, the availability of suicide methods, and the intensity of social activities.1 However, it remains uncertain whether these explanations can be applied to the cyclical change in youth suicide within a year.

To account for an underlying mechanism of seasonality in youth suicide, several previous studies offered arguments and evidence that highlight the importance of the school calendar. Lahti et al15 found a significant peak in the number of suicides from August to October among youth below 18 years in northern Finland. Since the school year begins in the middle of August in Finland, they argued that a stressful environment at school increases suicidal risk among vulnerable students. Similarly, Näyhä attributed a significant rise in suicide in the fall among the youth aged between 15 and 24 years in Finland to intensive stress in school life.16 McCleary et al,4 Chew and McCleary,17 and Hansen and Lang18 all studied suicide among students in the USA and found that the number of suicides increases when the academic year starts, but tends to be low during school breaks. In addition, a recent report by the Japan's Cabinet Office showed a sharp spike in suicide by young individuals aged 16 years or younger on the dates when schools start a new semester in the spring and fall.19 All these findings suggest that the number of youth suicides increases at the beginning of the academic year or after a long break because vulnerable students may be exposed to school-related problems, including violence, bullying, interpersonal issues, dissatisfaction with school, and academic performance which can become a cause of suicide.20–22 They also suggest that students are less likely to die by suicide during school breaks because they are detached from the potential stress associated with school life or from bullies at school.

This study seeks to offer additional evidence for the role of the school calendar in youth suicide by using national mortality data from Japan. It improves on previous studies in two major ways. First, we used daily records of suicide frequencies over a period of about 40 years, which allowed us to identify the exact timing of suicide in relation to the school calendar. Most previous studies used monthly data to examine the association between the school calendar and the timing of suicide deaths. Second, regional and temporal variations in the school calendar have been small in Japan, and this enabled us to design a nation-wide analysis for over a long period of time, including about 15 000 cases of suicides by young individuals below 18 years of age.

Using the mortality data from the Vital Statistics of Japan between 1974 and 2014, we compiled the daily frequency of suicides across the calendar among young individuals at ages 6–26 years. We then examined whether the frequency of suicides by students increases around the dates when the school session begins in January, April, and September. We also investigated whether suicides among students tend to decrease when school is not in session. We used the data of individuals aged 19 years and older as a comparison group, under the assumption that they are less likely to be affected by the typical Japanese school calendar.


The mortality data in this study came from the death records in the Vital Statistics of Japan compiled by the Ministry of Health, Labour and Welfare. The Vital Statistics data were collected for administrative purposes and anonymised by the Ministry prior to analysis. The data in the Vital Statistics are based on death certificates issued by physicians that are subsequently reported to the local government by a family member of the deceased. The local government then transfers the information to the Ministry of Health, Labour, and Welfare. The Ministry assigns the International Classification of Diseases (ICD) code to each case using the information reported in the death certificate. If it is not clear whether a particular death is intentional (suicide), unintentional (accident), or a homicide, it will not be treated as suicide. Thus, it is possible that the number of suicides reported in the Vital Statistics is lower than the actual number of suicides. The quality of the Vital Statistics is likely to be high throughout our study period because only physicians are legally allowed to issue a death certificate and a false report is subject to criminal penalty. The records cover all reported deaths in Japan and include the date of birth, the date of death, the place of death, marital status, and the underlying cause of death based on the ICD, ICD-8/ICD-9 standard (until 1994) and the ICD-10 standard (1995 till present).

Our analysis focused on deaths by suicide (ICD-8/ICD-9: E950-E959, ICD-10: X60-X84) among young individuals who died at ages 6–26 years. We categorised the individuals into five groups: elementary school students (ages 6–12 years), middle school students (ages 12–15 years), high school students (ages 15–18 years), college students and working individuals (18–22 years), and graduate students and working individuals (ages 23–26 years). In Japan, education is compulsory until the end of middle school, and almost all students (97%) attend high school. The first three groups follow a similar school calendar, as discussed below. The last two groups were included for the purpose of comparison because their behaviour is likely to be affected by different calendars such as the Japanese fiscal year which starts on 1 April.

We excluded death records on the basis of several criteria. First, we focused on Japanese citizens who were in Japan at the time of their death because those outside Japan were likely to follow a different school calendar. Second, we excluded records without any information on the dates of birth or death because this information is vital for calculating the person's exact age and the timing of the suicide. Third, we excluded records with no information on the place of death because an unknown place of death was likely to indicate that the authorities had no personal information about the deceased or could not specify the exact date of death.

Using individual suicide records, we counted the frequencies of suicides on each calendar date from 1 January to 31 December for each of the five age groups defined above. To simplify our analysis, we excluded the data of those who died on 29 February. Accordingly, the number of observations was 365.

In Japan, a new academic year begins in April and ends in March. Our survey of public elementary and middle schools in the 150 most populated municipalities, and public high schools in 46 of the 47 prefectures revealed that the following school calendar is typical across regions. The first (spring) semester begins around 8 April and ends around 20 July. It is immediately followed by the summer break, which ends around 31 August. The second (fall) semester begins typically on 1 September and continues until 25 December. The winter break follows and ends around 7 January. The third (winter) semester is in session between 8 January and 25 March, followed by the spring break. This typical academic calendar, as summarised in table 1, applies to students in the majority of elementary, middle and high schools.

Table 1

A typical school calendar in Japan

However, there is a slight variation in the school calendar across regions. According to our survey, about 30% of the surveyed municipalities start the fall semester nearly 10 days earlier than 1 September. These early starts are concentrated in the northern part of Japan that tends to have harsh winter conditions.

We examined the relationship between the school calendar and the frequency of suicides in two different ways. First, we plotted the frequencies of suicides against the school calendar dates by the five age groups. If the school calendar affects suicidal risks among young individuals who are vulnerable to a stressful situation around the start date of the school, we should be able to find a spike in the frequencies of suicides around 8 April, 1 September and 8 January, when the semester begins. In contrast, the frequencies of suicide were expected to decrease during the spring, summer, and winter breaks. Those aged 18–26 years were expected to show no spike on these dates because they are affected by different calendars.

Second, we estimated a Poisson regression model where the frequency of suicides on each calendar date was regressed on 52 dummy variables. These dummy variables were coded 1 if a particular calendar date was included in one of the 52 weeks of the year and zero otherwise. For example, the dates between 1 January and 7 January were set to one for the week 1 dummy, while all other dates were set to be zero. The dates between 8 January and 14 January were set to one for the week 2 dummy, etc. The baseline week excluded from the regression model was set to be week 26, from 25 June to 1 July, that is, the middle of the year. Using these dummy variables, we examined whether a particular week of the year was associated with higher frequencies of suicides among young individuals in school. More specifically, we expected that the weeks in which most schools begin a session (ie, week 2, 8–14 January; week 14, 2–8 April; and week 35, 27 August to 2 September) were associated with higher frequencies, while those weeks in which school is in break were associated with lower frequencies. To facilitate the interpretation of the Poisson regression results, we visually depicted the incidence rate ratios (IRR) for each dummy variable in the regression model.


Our analysis included 108 968 suicides by young individuals who died at 6–26 years of age between 1974 and 2014 in Japan. Among these, the number of suicides by elementary school students (ages 6–12 years) was 459. Further, 3274 suicides were by middle school students (ages 12–15 years), and 10 388 suicides were by high school students (ages 15–18 years). These groups are likely to be affected by the Japanese school calendar. In our control group, which is not likely to be affected by the school calendar, 45 312 individuals died by suicide at ages 18–22 years, and 49 535 died at ages 23–26 years. Table 2 contains the summary statistics for the frequency of suicides per calendar day by age groups.

Table 2

Summary statistics of daily suicide frequencies by age categories

The frequencies of suicides are plotted against the calendar dates by five age groups in figure 1. The shaded regions indicate the typical period of school break in spring (26 March to 7 April), summer (21 July to 31 August), and winter (26 December to 7 January). Accordingly, the right end of the shaded regions indicates the beginning of the school session. The figure shows the frequency of suicides by elementary school students in red, middle school students in blue, high school students in black, college students and working individuals aged 18–22 years in green, and working individuals aged 23–26 years in purple.

Figure 1

Daily frequencies of suicide by five age groups. Note: The number of suicide deaths by grade groups and age groups is plotted against the calendar dates. Twenty-nine February is excluded. The sample includes individuals who died by suicide at 6-26 years of age between 1974 and 2014 in Japan. The total number of suicide deaths is 108 968. Source: Death records, the Vital Statistics of Japan.

We found several notable patterns in figure 1. First, the frequencies of suicides by middle school students and high school students show large spikes when the school term starts, around 8 January, 8 April and 1 September. No such pattern is found for elementary school students, but this could be due to the small number of suicides in this age group. Second, the frequencies of suicides among school-age children and adolescents tend to decrease during the breaks. Most evidently, the beginning of the summer break is associated with a large reduction in suicides. The frequencies, however, tend to increase as the summer break approaches the end, resulting in the highest frequency on 1 September.

Third figure 1 shows no similar pattern in suicides by those aged 18–22 years and 22–26 years around 7 April and 1 September as these cohorts are unaffected by the typical Japanese school calendar. The number of suicides among such individuals tends to increase towards early April, but the timing seems to coincide more with the beginning of the Japanese fiscal year. We speculate that this is because colleges and companies typically start the new school term or budget year around 1 April.

The Poisson regression results are presented in figure 2. We used the frequencies of suicide by middle and high school students, and those at ages 18–22 years as the outcome variables. We excluded elementary school children from this part of the analysis because of the insufficient number of observations. The x-axis in figure 2 indicates the 52 weeks in the school calendar: week 1 (1 January to 7 January) to 52 (24 December to 31 December). The baseline category was set to be week 26 (25 June to 1 July). The y-axis shows the estimated influences of the corresponding weeks on suicide counts. The height indicates the per cent increase in the number of suicides from the baseline period.

Figure 2

Poisson regression estimation results by age groups. Note: The estimated incidence rate ratios (IRRs) are shown with 95% CIs (vertical lines). The baseline week is 25 June to 1 July. All estimates are based on Poisson regressions using the frequency of suicides per calendar day. The data include individuals who were middle school students (A), high school students (B) at the time of death as well as individuals who died by suicide at 18–22 years of age (C). Source: Death records (1974–2014), the Vital Statistics of Japan.

Figure 2A shows that the frequency of suicides among middle school students increases dramatically in the week of 27 August to 2 September (week 35). The IRR is 2.383 (95% CI 1.735 to 3.272), implying that the incidence rate of suicide doubled around the dates when the summer break ends and the second semester begins. Similarly, the incidence rate also increases by ∼40–50% in the weeks of 2–8 April and 8–14 January when the first and third semester begins. In contrast, the incidence rate decreases by half during the period of week 30 to 32 (23 July to 12 August), which corresponds with the summer break.

Figure 2B shows a similar pattern for high school students. The frequency of suicides among high school students significantly increases in the weeks of 8–14 January with the IRR 1.117 (CI 1.102 to 1.346), 9–15 April with the IRR 1.610 (CI 1.453 to 1.785), and 27 August to 2 September with the IRR 1.451 (CI 1.125 to 1.871). The incidence rate is smaller just before these weeks when the school is in the summer or winter break, as compared with the baseline category.

Finally, figure 2C shows quite a different pattern for young individuals between ages 18 and 22 years. Most of them are either in college or in the labour market, and thus follow a different calendar from those below 18 years of age. While we found a statistically significant increase in suicides in the spring, yet no spike was found in the weeks of 2–8 April and 9–15 April. Those at ages 18–22 years were more likely to die by suicide around the dates when the new budget year begins. Further, the incidence rate of suicide shows no increase in the week of 27 August to 2 September.


Using the daily frequencies of suicides by the youth in Japan between 1974 and 2014, we found that middle school students at ages 12–15 years, and high school students at ages 15–18 years had the highest frequency of suicide around 8 April and 1 September. These dates correspond to the beginning of the school session. Conversely, the lowest frequency of suicide was in late July and early August, when they were not in school. Importantly, no such pattern existed for those who died at age 18–26 years, who were unaffected by the school calendar. This implies that the timing of suicide is clearly related to the school calendar for middle and high school students; therefore, their decision to take their own lives is likely to be associated with school-related issues. Our results are also consistent with a so-called ‘broken promise’ effect;23 as students are free from school-related issues when school is in break, vulnerable students may increase their hope that their troubled situation may improve once a new academic year or semester starts. If these expectations are not met, however, it can enhance their hopelessness and may contribute to their decision to die by suicide.

The present study makes two important contributions to the literature. First, it used an extensive data set of suicides by young individuals, which offers strong evidence regarding the seasonality of suicide among the youth. The bimodal peak, as shown in figures 1 and 2, is consistent with the findings by Näyhä16 in Finland and Souêtre et al9 in France, while the peak in the fall is consistent with studies conducted in Finland,15 in the USA,4 and in Canada.24

Second, in contrast to the monthly data typically used in previous studies, the daily records of suicide in our analysis allowed us to compare the school calendar and the exact timing of the suicides. Our findings uncover a part of the underlying mechanism of seasonality in youth suicide because the cyclical pattern corresponds to the school calendar; thus, it is plausible that stressful environments and the ‘broken promise’ at the beginning of school sessions affect suicidal risk among vulnerable students.

This study has several limitations. First, the present study could not provide direct evidence that these school-aged individuals died by suicide for school-related reasons. Future studies should investigate whether students tend to experience a higher level of stress when school starts or resumes, and whether those who died by suicide in April and September had a school-related problem. Second, our findings are based on the data from a single country. A cross-national analysis would improve our understanding of the extent to which school calendars account for the seasonality in youth suicide.

This study has crucial policy implications for designing a prevention strategy for youth suicide. Previous studies on youth suicidal behaviour suggest that school-related problems, such as violence, bullying, interpersonal issues, dissatisfaction with school, and academic performance, are a significant risk factor.21–23 These risk factors may have a larger effect on suicidal behaviour among the youth particularly when a semester begins. Thus, our analysis suggests that any intervention effort to reduce these risk factors should be implemented when the school is in session, as well as before the school resumes from breaks. Potential school-based prevention strategies include suicide awareness programmes and skill trainings that facilitate help-seeking behaviour, self-report or individual interviews to identify vulnerable students, and gatekeeper training to improve the skills of school personnel to screen at-risk students.20 These strategies should be implemented during the semesters so that vulnerable students can seek out help or be contacted and treated by school personnel before a new semester begins.

In conclusion, our findings offer additional evidence that the school calendar plays an important role in accounting for the seasonality in youth suicide. The suicide prevention effort in the school community setting should take this seasonality into account to prevent an increase in suicides when the semester begins.

What is already known on this subject

  • Previous studies have provided suggestive evidence to show that temporal variation in youth suicide is associated with the school calendar, as it tends to increase at the beginning of the academic year or after a long break, but robust empirical evidence remains scarce.

What this study adds

  • Using nationwide death records in Japan between 1974 and 2014, we found that the frequency of suicide among middle and high school students was highest at the beginning of the new academic year in April and at the beginning of the fall semester. Our findings suggest that the timing of student suicide was related to the school calendar. Therefore, the suicide prevention effort in the school community setting should take this pattern into account.



  • Contributors TM and MU collected and analysed the data. TM, MU and KY wrote the manuscript.

  • Funding JSPS Grants-in-Aid for Scientific Research (26870326).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.