{"paper_id":"268def19-5b6c-4f61-8083-d732fdfd880f","body_text":"PREPRINT\nAuthor-formatted, not peer-reviewed document posted on 28/03/2025\nDOI: https://doi.org/10.3897/arphapreprints.e153999\nThe relationship between daylight hours and suicides\nseasonality in Russia\nJulia Korshunova, Ivan Kucherov\n\nThe relationship between daylight hours and suicides seasonality in Russia \nAbstract \n \nResearchers from various countries across the globe have found suicidal behaviour  to exhibit \nseasonality. In Russia and other countries located in the northern hemisphere it is observed that the \nnumber of suicides spikes during the spring to summer period and drops in winter. Researching the \nseasonal fluctuations of suicide mortality will allow us to better u nderstand this phenomenon and , \nconsequently, to develop effective measures of suicide prophylaxis, which help prevent future suicide \ncases. \n \nIn this article we research whether the seasonality of suicide levels in Russian countries is related to \nthe daylight hours in different months. To achieve this, seasonality indices for suicides and daylight \nhours have been constructed for 8 Russian cities located across different latitudes. For these indices \nPearson’s correlation coefficient has been computed. Granger causality test has been performed for \nthe Russian suicide mortality data obtained for the years from 2000 to 2021.  The authors have also \nattempted to estimate the real number of suicides by including  some other causes of death, which \nwere classified as events of undetermined intent. \n \nThe results of the study show significant high positive correlation between the seasonality indices of \nsuicides and daylight hours (ranging between 0.74  and 0.9  depending on t he city) as well as the \npresence of Granger causality for all researched cities when using 2 and 3 lags, which might imply a \npotential influence of the daylight hours on the suicidal behaviour in Russia. \n \nThis research contributes to academic literature on the seasonal patterns in suicides and their potential \ncauses. \n \nKeywords \nsuicidal behaviour, daylight hours, events of undetermined intent , correlation, Granger causality, \nsuicides estimate. \n \nIntroduction \n \nSuicides are a disturbing problem which is highly relevant in today’s society. Every 45 seconds a \nperson takes his own life, while the annual number of suicides amounts to a devastating 700 thousand \npeople (Suicide 2023) . Every case of suicide is a tragedy and a strong b low to the families, \ncommunities and even entire countries , causing a lasting negative impact . The World Health \nOrganization considers suicide prevention as one of their top priorities in the domain of public health. \nEspecially alarming is the data for suicides among young people aged 15 to 29, where suicide ended \nup being the fourth most common cause of death (Suicide 2023). \n \nThe suicide levels in Russia exceed those of many other countries, which shows why this problem is \nespecially relevant for it. In 2019 the Russian Federation became the top  4 country by the suicide  \ncases per 100 thousand persons (Ranking of countries by suicide rate 2019). On top of that, according \nto the Investigative Committee of Russia, in 2021 the number of child suicides increased to 37.4% \ncompared to the previous year (Maria Lvova-Belova presented the 2021 work results 2022). \n \nSuicides in Russia have a strongly apparent seasonality with the highest number of cases during spring \nand summer month s and  the lowest in winter  (Vishnevsky 2017). In order to develop effective \nmeasures of prophylaxis and international collaboration in suicide prevention it is vital to research \nthe seasonal behaviour of suicide mortality coefficients and to explain the nature of those fluctuations. \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\n \nThe aim of this research article is to determine the relationship between daylight hours and suicides \nseasonality in Russia. \n \nThe object of this research is the suicides seasonality in Russia. \n \nThe subject is the relationship between daylight hours and suicides seasonality in Russia. \n \nMethodology \n \nAnonymized data on individual death reports was obtained for the analysis from Rosstat for 8 Russian \ncountries for the years from 2000 to 2021 with the following ICD-10 causes: \n \n1) Intentional self-harm (X60-84) \n2) Poisoning, undetermined intent, except for narcotics and psychedelics (Y10-11, Y13-19) \n3) Hanging, strangulation and suffocation, undetermined intent (Y20) \n4) Firearm discharge, undetermined intent (Y22-24) \n5) Falling, jumping or pushed from a high place, lying or running before or into moving object, \nundetermined intent (Y30-31) \n \nUsing this data, monthly  mortality coefficients have been computed for 22 years as the number of \ndeaths divided by the person-years in the current month. \n \nFor this study only cities in Russia with a population of over 500 000 people and located at different \nlatitudes with a difference no less than 2° ± 0,08° have been considered. This is because the latitude \naffects the daylight hours of a given city. The sample contains 8 Russian cities (Table 1). \n \nTable 1. The list of cities chosen in the study and their characteristics for the year 2021. \n# City name Population size, people Latitude (degrees) \n1 Vladivostok 603 519 43°11,55’N \n2 Krasnodar 1 099 344 45°03,54’N \n3 Khabarovsk 617 441 48°48,02’N \n4 Saratov 901 361 51°53,35’N \n5 Tolyatti 684 709 53°50,78’N \n6 Moscow 12552559 55°75,58’N \n7 Perm 1 034 002 58°01,04’N \n8 Saint Petersburg 5283371 59°93,90’N \n \nThe data on population size in Table 1 for 2021 has been collected on the 2020 All-Russian population \ncensus. This study analyzes the city of federal importance Moscow, which also includes the following \ncities: Zelenograd, Moscow, Moskovsky, Troitsk, Shcherbinka.  Also, the city of federal importance \nSaint Petersburg includes the following cities: Zelenogorsk, Kolpino, Krasnoe Selo , Kronstadt, \nLomonosov, Pavlovsk, Peterhof, Pushkin, St. Petersburg, Sestroretsk. \n \nThe data on daylight hours for the years from 2000 until 2021 has been gathered from the webpage  \nhttp://voshod-solnca.ru. \n \nIn order to consider the seasonality of suicides in the chosen cities, the median deaths per month have \nbeen computed as well as the suicide seasonality indices according to the formula below: \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\n \n𝐼𝑠 = 𝑦̅𝑖\n𝑦̅  \nHere 𝑦̅𝑖 is the weighted average number of deaths in month 𝑖 ∈ {1,2, … ,12}, weighted by the number \nof days in it, and 𝑦̅ is the weighted average number of deaths in each year, weighted by the number \nof days in it. \n \nThe indices of daylight hours were computed similarly for each city as the day weighted average for \na given month. These seasonality indices of suicides and daylight hours were compared in order to \ndetermine their connection via the Pearson correlation coefficient. \n \nIn order to study the impact of daylight hours on the seasonality of suicides in Russian cities Granger \ncausality test was performed. Prior  to that, the linear trend had been removed from the number of \ndeaths variable by applying the first difference operator. \n \nA lot of Russian and foreign researchers study the phenomenon of seasonality in peoples’ suicidal \nbehaviour. Some works consider the psychopathological reasons for the seasonal fluctuations of \nsuicides (Kim et al. 2004), others consider sociological, biological and ecological factors instead  \n(Souêtre et al. 1987). \n \nMost academic research on the suicidal behaviour is focused on studying various suicide risk factors: \nauthors describe the influence of family on the suicidal behavio ur (Frey & Cerel 2015); the \nrelationship between suicides and religious beliefs (Hajiyousouf & Bulut 2022); the differences in \nethnicity and race and how these differences influence suicides (Lee & Wong 2020). Many \nresearchers are also considering the role of economic and social environment and the conditions \nwhich cause the number of suicides to increase, such as: the poverty and inequality levels (Piatkowska \n2020), the unemployment rate (Amiri 2022) and the economic crises (Marazziti et al. 2021). \n \nExperts in the field of medicine determined the link between seasonal levels of suicides and climate \nfactors such as air temperature  and daylight hours (Rozanov et al. 2018а). The authors of this research \nhave analyzed 17 years of  monthly data on suicides in the city of Odessa from 2001 util 2016 and \nconcluded that daylight hours and air temperature are highly and positively correlated with the \nfrequency of suicides (0.97 and 0.96 respectively). This way it was determined that the number of \nsuicides in May exceeds that of in December by 58% . To justify the relationship between suicidal \nbehaviour and climate factors the authors bring up the fact that  the number of suicides is close to the \naverage in September and March exactly, which correspond to the periods of autumn and spring \nsolstice. For other months the following pattern can be seen: the shorter the daylight hours, the smaller \nthe frequency of suicides in the city and vice versa.  The authors speculate that this link between \nsuicides and climate is due to the dynamics of serotonin and melatonin in the human body and due to \nneurohumoral mechanisms of thermal adaptation. \n \nThe authors of another study looked at the relation between suicides in south and north countries and \ndaylight hours  (Petridou et al. 2002). The researchers revealed a consistent seasonality in the \nprevalence of suicides, which peaks approximately in June in the northern hemisphere and in \nDecember in the southern hemisphere. They also found a positive relationship between the seasonal \namplitude of suicides and the number of sunny days  in the respective countries. These results point \nto a possible provoking effect of sunlight on suicides. \n \nResearchers from Finland and Switzerland also propose that seasonal fluctuations of suicides \nfrequency are connected with climate factors, in particular, with changes in air temperature  \n(Holopainen et al. 2013). By having the vastest continuous demographical statistics available in the \nworld, the authors were able to analyze the data on mortality from suicides in Switzerland and Finland \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\nsince the 1750s. Thanks to  these early records on demographic changes in these  countries, the \nresearchers had access to 260 years’ worth of time series data. According to the authors, the increases \nand decreases in the mortality levels from suicides can be caused by sharp changes in temperature \ntwice a year in May and October (it is exactly in these months that the number of suicides peaks in \nSwitzerland and Finland) . The authors explain this  phenomenon with the relationship between \ntemperature fluctuations and the activity of brown adipose tissue in the human body, which aggravates \ndepression, in turn, leading to suicidal behaviour. The authors think that there is evidence to suggest \nthat the farther is the region from the equator, the later the level of suicides peaks in spring. However, \nthey acknowledge that currently there is no  data that confirms the correlation of underlying brown \nadipose tissue activity with the indicators of anxiety or depressive episodes in people. \n \nThe seasonality of suicides is outlined by researchers of many countries. For instance, the researchers \nfrom Poland also noted an increased rate of suicides during spring (Młodozeniec et al. 2010). Having \nanalyzed 29,232 cases of suicide deaths registered in Poland from 1999 until 2003, they came to the \nconclusion that there was a clear seasonal component in suicidal behaviour among Polish men, but \nnot women. \nNevertheless, in the scientific community there are also studies that did not find the seasonality of \nsuicidal behaviour in some particular regions. For example, no evidence for the seasonality of suicides \nwas found in Tasmania (Lee & Pridmore 2014 ), Los Angeles and Sacramento  (Tietjen & Kripke \n1994), which may be due to the specific socio-economic factors of these regions or the lacking sample \nsize. \n \nIt has been  discovered that longer daylight hours associated with seasonal changes can shift the \nbalance of melatonin and its production levels in the body (Danilenko et al. 2021). This, in turn, \naffects the regulation of sleep and mood, which increases the risk of depression in individuals. During \nlong sunny days, the secretion of melatonin is reduced, which provokes the disruption of the body's \ncircadian rhythms. This can lead to decreased sleep quality and changes in the psycho-emotional state, \nwhich increases the likelihood of developing depressive disorders. \n \nResearch suggests that melatonin levels may be associated with the risk of developing depression \n(Wang et al. 2021). Low levels of melatonin have been found in people suffering from depression, \nwhich may indicate problems with the regulation of circadian rhythms and sleep. Melatonin also has \nantioxidant effects and protects the brain from stress, which may alleviate symptoms of depression \n(Wu et al. 2013). \n \nAdditionally, melatonin can influence the production of neurotransmitters such as serotonin, which \nplays a key role in mood regulation (Jenkins et al. 2016). There is evidence that melatonin and \nserotonin levels are interrelated, and that changes in melatonin levels can affect a person's \npsychological state (Dollins et al. 1993). \n \nSince melatonin is produced when there is  less light radiation, higher levels of this hormone are \nobserved in autumn and winter, when the nights are longer, and , conversely, lower levels are found \nin spring and summer (Danilenko et al. 2021). \n \nAs depression plays an important role in suicides (Angst et al. 1999) and daylight hours directly affect \nthe biorhythms of the body, melatonin regulation and the risks of depression, understanding this \nrelationship can be important for developing preventive measures for suicidal behaviour. \nSince the 1990s the level of mortality from events of undetermined intent has drastically increased. \nIn 2018 the losses from events of undetermined intent were 2. 1 times higher than from suicide for \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\nmale population and 3 times higher for female population ( Semyonova et al. 2020). According to \nYumaguzin and Vinnik, events of undetermined intent are a reservoir for latent assaults and suicides. \nBesides, since 2014 th e volume of events of undetermined intent in Russia exceeded the mortality \nlevels from assaults and suicides combined ( Yumaguzin & Vinnik 2019). In their paper the authors \nattempt to estimate the real level of suicides in Russia by redistributing the event s of undetermined \nintent among assaults, suicides and accidents. According to their estimates the suicide levels in Russia \nare underestimated by 30%. \n \nIn the article by Andreev, Shkolnikov, Pridemore and Nikitina an estimate of suicide levels and other \ncauses is also constructed by using the method of reclassifying the events of undetermined intent. The \nresults of this research indicated an underestimation of 24% in the official data sources for the natural \nmovement of the population in 2011 (Andreev et al. 2015). \n \nOther estimates for 2011 –2018 in Russia suggest that the real mortality from suicides exceeds that \nstated in the official data sources by 58.3% for males and by 85.7% for females (Semyonova et al. \n2020). \n \nKrenev and Vasin have outlined multiple causes among the events of undetermined events which are \nmechanically similar to suicides (Krenev & Vasin 2012). The following potential suicides have been \nselected from the ICD-10: \n \n1) Poisoning by and exposure to nonopioid analgesics, antipyretics and antirheumatics, \nundetermined intent (Y10) \n2) Hanging, strangulation and suffocation, undetermined intent (Y20) \n3) Drowning and submersion, undetermined intent (Y21) \n4) Injuries from firearms and explosives, undetermined intent (Y22 - Y25) \n5) Exposure to smoke, fire and flames, undetermined intent (Y26) \n6) Contact with steam, hot vapours and hot objects, undetermined intent (Y27) \n7) Contact with sharp object, undetermined intent (Y28) \n8) Contact with blunt object, undetermined intent (Y29) \n9) Falling, jumping or pushed from a high place, lying or running before or into moving object, \nundetermined intent (Y30 – Y31) \n10) Crashing of motor vehicle, undetermined intent (Y32) \n11) Other specified events, undetermined intent (Y33) \n12) Unspecified event, undetermined intent (Y34) \n \nThe practice of underreporting suicides due to the overuse of the events of undetermined intent is \npresent not only in Russia, but also in other countries. For instance, in Estonia the fraction of the \nevents of undetermined intent is higher than in Russia (Vasin 2015). The researchers attribute this to \nnegligent investigations in the country. \n \nIn most developed countries the data on mortality causes is more detailed and is available for scientific \nresearch (Vasin 2015). However, concealment of violent causes of death occurs in developed \ncountries as well. In the USA there have been systematic classifications of suicides as p oisoning, \nundetermined intent (Fingerhut & Cox 1998). Similar underreporting was also present in Australia \n(Snowdon & Choi 2020). Also, in research conducted in Sweden, it was found that about two thirds \nof all causes of death, which were classified as events of undetermined intent, were in fact cases of \nsuicide. It was discovered by interviewing relatives, friends and medics as well as by analyzing the \ndeath certificates. \nResults \n \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\nLet us consider the distribution of suicide cases in Russian cities by months. For this purpose, suicide \nseasonality indices were constructed. The distribution of these indices is displayed in Table 2. \n \nTable 2. Suicide seasonality indices in Russian cities. \nCity Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec \nVladivostok 1.16 0.89 1.00 0.91 1.03 1.11 1.38 1.04 0.92 0.82 0.91 0.84 \nKrasnodar 0.90 1.07 1.10 1.03 1.16 1.15 1.23 0.91 0.98 0.74 0.84 0.88 \nKhabarovsk 0.95 0.96 1.03 1.13 1.17 1.00 1.04 1.12 1.03 0.98 0.80 0.80 \nSaratov 1.01 0.93 1.00 1.08 1.13 1.02 1.27 1.09 0.98 0.89 0.84 0.77 \nTolyatti 1.20 0.94 1.06 1.09 1.12 0.97 1.02 0.87 1.07 0.87 0.83 0.96 \nMoscow 0.98 0.94 1.00 1.09 1.11 1.12 1.06 1.06 0.92 0.91 0.93 0.87 \nPerm 1.02 0.95 0.99 1.06 1.11 1.12 1.02 1.06 0.97 0.94 0.89 0.87 \nSaint Petersburg 1.06 0.97 1.00 1.12 1.08 1.03 1.10 1.03 0.93 0.94 0.86 0.88 \n \nThe months in which min and max suicides occur as well as some additional relevant characteristics \nof each Russian city in this study are presented in Table 3. \n \nTable 3. Suicide seasonality profiles for Russian cities. \nCity Latitude \n(degrees) \nAverage annual \nnumber of \nsuicides \nMonth of max \nsuicides \nMonth of min \nsuicides \nVladivostok 43°11’N 49 July October \nKrasnodar 45°03’N 112 July October \nKhabarovsk 48°48’N 113 May December \nSaratov 51°53’N 134 July December \nTolyatti 53°50’N 50 January November \nMoscow 55°75’N 705 June December \nPerm 58°01’N 230 June December \nSaint Petersburg 59°93’N 561 April November \n \nThe overall trend is seen for all cities aside from Tolyatti: an increase in suicides in spring and summer \nand a decrease in autumn and winter periods. Thus, for most cities the peak of suicides is seen in June \nor July and the minimum occurs  in the range from October to December. The one exception is the \ncity of Tolyatti, in which the peak occurs in January. However, as with the other cities, Tolyatti also \nhas a minimum occurring in the abovementioned period in November . This deviation for Tolyatti \ncould be explained by the relativ ely low number of the average annual number of suicides, only 50 \ncases per annum, which leaves the index sensitive to outliers for each given month. \n \nSince the average value, which was used to compute the seasonality indices, is sensitive to outliers, \nmedian monthly number of suicides has been computed for 22 years (Figure 1). \n \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\n \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\nFigure 1. Median monthly number of suicides by city. \n \nThe red color in Figure 1 indicates the month with the highest number of median monthly suicides, \nwhereas the green color indicates the month in which the minimum occurs. \n \nWhen using the median instead of the average the trend has slightly changed. In most cities the peaks \nnow occur in May, while the min months stay within the October -December range. That being said, \nonce again it is Tolyatti that is different within the group with the maximum occurring in January and \nminimum in May. Another deviation is the city of Vladivostok with a peak occurring in January. This \nsituation could again be connected with the low average annual number of suicides with less than 50 \ncases per annum. Another reason could lie in the underreporting of suicides (Figure 2). \n \nAccording to mortality researchers, excessive classification of external causes as undetermined intent \ncould indicate questionable data quality  (Andreev et al. 2015).  Some experts state that the frequent \nuse of this cause suggests the presence of data manipulation (Yumaguzin 2017). \n \n \nFigure 2. Monthly weighted average number of suicides and other causes of death in Tolyatti for 22 \nyears, weighted by the number of days in each month. \n \nFigure 2 displays that the suicide levels in Tolyatti are lower than those of  even such  causes as \npoisoning, undetermined intent, except for narcotics and psychedelics; hanging, strangulation and \nsuffocation, undetermined intent; f alling, jumping or pushed from a high place, lying or running \nbefore or into moving object, undetermined intent . Thus, we can assume the presence of latent \nsuicides among these causes of death. \n \nIt is also important to note that hanging, strangulation and suffocation, undetermined intent exhibit \nthe same seasonal pattern as that of Russia as a whole: the number of deaths increases in spring and \ndecreases in autumn-spring. \n \nThe monthly weighted average number of deaths from suicides and some events of undetermined \nintent, weighted by the number of days in each month is shown in Figure 3. \n \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\n \nFigure 3. The monthly weighted average number of deaths from suicides and some events of \nundetermined intent, weighted by the number of days in each month, for all studied Russian cities. \n \nSeasonality indices have also been computed for each city for all 5 causes. The average distributions \nof these indices for all 8 cities as a whole are presented in Figures 4, 5, 6 and 7. \n \n \nFigure 4. Seasonality indices for suicides and poisoning, undetermined intent, except for narcotics \nand psychedelics. \n \n0.0\n0.2\n0.4\n0.6\n0.8\n1.0\n1.2\n1.4\nJan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec\nIntentional self-harm\nPoisoning, undetermined intent, except for narcotics and psychedelics\nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\n \nFigure 5. Seasonality indices for suicides and hanging, strangulation and suffocation, undetermined \nintent. \n \n \nFigure 6. Seasonality indices for suicides and firearm discharge, undetermined intent. \n \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\n \nFigure 7. Seasonality indices for suicides and falling, jumping or pushed from a high place, lying or \nrunning before or into moving object, undetermined intent. \n \nWe can see that the seasonal fluctuations of suicides, hanging, strangulation and suffocation, \nundetermined intent as well as falling, jumping or pushed from a high place, lying or running before \nor into moving object, undetermined intent  have a similar profile with an increase in the number of \ndeaths during the spring-summer period and a decrease during autumn and winter. In order to check \nthe hypothesis of whether there is a relationship between the distributions of those causes and the \ndistribution of suicides, Pearson correlation coefficient has been computed.  The results of the \ncalculations are presented in Table 4. \n \nTable 4. Pearson correlation coefficients between the death causes. \nCauses Pearson correlation \ncoefficient \nP-value Significance at α = \n0.05 \n1 and 2 0.329 0.296 not significant \n1 and 3 0.597 0.04 significant \n1 and 4 0.298 0.347 not significant \n1 and 5 0.592 0.042 significant \n \nIn Table 3 the causes are numbered according to the scheme below: \n1) Intentional self-harm \n2) Poisoning, undetermined intent, except for narcotics and psychedelics \n3) Hanging, strangulation and suffocation, undetermined intent \n4) Firearm discharge, undetermined intent \n5) Falling, jumping or pushed from a high place, lying or running before or into moving object, \nundetermined intent \n \nThe correlation between suicides and hanging, strangulation and suffocation, undetermined intent as \nwell as falling, jumping or pushed from a high place, lying or running before or into moving object, \nundetermined intent  ended up being significant at the confidence level of α = 0 .05. The linear \nrelationship between the variables is positive. This can point to the presence of latent suicides among \nthese two causes. \n \nBased on this assumption, the seasonality indices have been recomputed again by including two \ncauses of death : hanging, strangulation and suffocation, undetermined intent ; falling, jumping or \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\npushed from a high place, lying or running before or into moving object, undetermined intent . Here \nand thereafter the three causes of death combined will be called the suicides estimate. \n \nThe distribution of seasonality indices for the suicides estimate by month for each city are displayed \nin Table 5. \n \nTable 5. Seasonality indices for the suicides estimate in Russian cities. \nCity Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec \nVladivostok 0.93 0.89 1.06 1.09 1.06 1.12 1.21 1.03 0.98 0.92 0.91 0.80 \nKrasnodar 0.90 1.05 1.05 1.00 1.13 1.15 1.16 1.00 1.05 0.79 0.85 0.87 \nKhabarovsk 0.99 0.90 0.94 1.07 1.13 0.98 1.09 1.13 1.05 0.98 0.92 0.82 \nSaratov 1.01 0.94 0.88 1.03 1.09 1.05 1.21 1.15 1.03 0.97 0.86 0.79 \nTolyatti 0.96 0.88 0.96 1.06 1.08 1.04 1.16 1.05 1.11 0.96 0.90 0.84 \nMoscow 0.90 0.88 0.95 1.05 1.09 1.13 1.11 1.08 0.99 0.97 0.95 0.92 \nPerm 0.99 0.94 0.97 1.05 1.13 1.14 1.03 1.08 0.99 0.95 0.88 0.86 \nSaint Petersburg 1.00 0.93 0.94 1.07 1.08 1.07 1.13 1.07 0.99 0.95 0.89 0.88 \n \nThe months in which min and max suicides  estimate occurs as well as some additional relevant \ncharacteristics of each Russian city in this study are presented in Table 6. \n \nTable 6. Suicides estimate seasonality profiles for Russian cities. \nCity Latitude \n(degrees) \nAverage annual \nnumber of  the \nsuicides estimate \nMonth of max \nsuicides \nMonth of min \nsuicides \nVladivostok 43°11’N 135 July December \nKrasnodar 45°03’N 146 July October \nKhabarovsk 48°48’N 162 May, August December \nSaratov 51°53’N 198 July December \nTolyatti 53°50’N 183 July December \nMoscow 55°75’N 1470 July February \nPerm 58°01’N 245 July December \nSaint Petersburg 59°93’N 909 July December \n \nAfter estimating the number of suicides by including additional causes of death, the number of \nsuicides in the cities has drastically increased. For example, in Tolyatti the average annual number of \ndeaths has increased by 266%, in Vladivostok by 176% and in Moscow by 109% . At the same time \nin Perm this increase constituted only 7%, which could indicate that the suicides in this city are \nrecorded fairly. \n \nSeasonality indices for daylight hours have been computed analogously to the seasonality indices for \nsuicides. The average distribution of daylight hours by month in all the Russian cities is presented in \nFigure 8. \n \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\n \nFigure 8. Daylight hours distribution by month for all Russian cities. \n \nThe longest daylight hours occur in summer months, in particular, June is the month with the longest \ndaylight hours. In September and March  the daylight hours are equal approximately to the annual \naverage. The month with the smallest value of the daylight hours is December. \n \nIn order to check the relationship between the seasonality indices for daylight hours and suicides for \nall 8 Russian cities, Pearson correlation coefficient has been computed. The results of the calculations \nare presented in Table 7. \n \nTable 7. Pearson correlation coefficients between the seasonality indices for daylight hours and \nsuicides in Russian cities. \nCity Pearson correlation \ncoefficient \nP-value Significance at  α = \n0.05 \nBefore accounting for hanging, strangulation and suffocation, undetermined intent ; falling, \njumping or pushed from a high place, lying or running before or into moving object, \nundetermined intent \nVladivostok 0.59 0.044 significant \nKrasnodar 0.7 0.012 significant \nKhabarovsk 0.78 0.003 significant \nSaratov 0.8 0.002 significant \nTolyatti 0.02 0.95 not significant \nMoscow 0.86 0 significant \nPerm 0.83 0.001 significant \nSaint Petersburg 0.67 0.018 significant \nAfter accounting for hanging, strangulation and suffocation, undetermined intent ; falling, \njumping or pushed from a high place, lying or running before or into moving object, \nundetermined intent \nVladivostok 0.898 0 significant \nKrasnodar 0.831 0.001 significant \nKhabarovsk 0.735 0.006 significant \nSaratov 0.791 0.002 significant \nTolyatti 0.816 0.001 significant \nMoscow 0.934 0 significant \nPerm 0.894 0 significant \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\nSaint Petersburg 0.901 0 significant \n \nIn all cities a strong positive and significant linear relationship was found between suicides and \ndaylight hours. The coefficients demonstrate a high, positive and significant correlation between the \nseasonality indices for suicides and daylight hours . This may indicate that an increase in daylight \nhours is related to an increase in suicides in these cities. \n \nAfter accounting for two additional causes of death, the correlation coefficients have increased for all \nanalyzed cities. On top of that, the correlation for the city of Tolyatti became significant at 5%. \nGranger causality \n \nIn order to establish the relationship (or lack thereof) between the daylight hours and the suicides in \nRussia, Granger causality test has been performed. \n \nGranger causality test is used to determine a weak form of causality between two time series variables \nand consists of a statistical test, which establishes the predictive power of some variable . This test \ncompares the forecast accuracy in the case where one time series is explained only by its lagged \nvalues and the case where additional lagged terms from the second time series are added . If the \nadditional terms improve the forecast accuracy, it is said that the second time series Granger causes \nthe first time series. Despite the fact that this test does not give a definitive answer as to whether the \nactual causality is present, it serves as a good benchmark for further investigations. \n \nSince this test is only applicable to stationary time series and the time series of suicide mortality \ncoefficients are not stationary, difference operators were applied on the time series to make them \nstationary. In order to check if the differ enced series are indeed stationary, the augmented Dickey -\nFuller (ADF) test has been performed. Indeed, it showed that all time series were stationary at all sane \nsignificance levels.  The tests were performed for both the actual suicide coefficients and the ir \nestimates by the authors.  The results of the Granger causality test for differenced actual time series \ncan be found in Table 8.  Note that the series of  differenced suicide coefficients was used as the \ndependent variable. \n \nTable 8. The results of the Granger causality test for suicide coefficients. \nCity P-value Significance at α = 0.05 Is Granger causality present? \nLags included = 1 \nVladivostok 0.44 not significant No \nKrasnodar 0.07 not significant No \nKhabarovsk 0.75 not significant No \nSaratov 0.90 not significant No \nTolyatti 0.61 not significant No \nMoscow 0.33 not significant No \nPerm 0.54 not significant No \nSaint Petersburg 0.22 not significant No \nLag included = 2 \nVladivostok 0.212 not significant No \nKrasnodar 0 significant Yes \nKhabarovsk 0.094 not significant No \nSaratov 0 significant Yes \nTolyatti 0.891 not significant No \nMoscow 0 significant Yes \nPerm 0.001 significant Yes \nSaint Petersburg 0 significant Yes \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\nLag included = 3 \nVladivostok 0.179 not significant No \nKrasnodar 0 significant Yes \nKhabarovsk 0.021 significant Yes \nSaratov 0 significant Yes \nTolyatti 0.493 not significant No \nMoscow 0 significant Yes \nPerm 0 significant Yes \nSaint Petersburg 0 significant Yes \n \nWe can observe that when using actual values for suicide coefficients, no Granger causality was found \nwhen using only one lag. As the number of lags started increasing, more and more cities started to \nexhibit Granger causality. This is a natural behaviour of the test as it is sensitive to the number of lags \nchosen and the results of the test tend to reject the null hypothesis more as the number of lags grows. \nHowever, since we see a rather non -uniform and unstable results for all cities, we cannot conclude \nanything from using the actual values for suicide coefficients. \n \nThe same procedure was performed to get Table 9, which shows the results of the Granger causality \ntest for the authors’ estimates of the true suicide coefficients. The hopes are that trying to estimate the \ntrue coefficients would yield more accurate results for the Granger causality tests. \n \nTable 9. The results of the Granger causality test for estimated true suicide coefficients. \nCity P-value Significance at α = 0.05 Is Granger causality present? \nLags included = 1 \nVladivostok 0.0073 significant Yes \nKrasnodar 0.2621 not significant No \nKhabarovsk 0.5613 not significant No \nSaratov 0.2714 not significant No \nTolyatti 0.3972 not significant No \nMoscow 0.5352 not significant No \nPerm 0.6723 not significant No \nSaint Petersburg 0.8939 not significant No \nLag included = 2 \nVladivostok 0.0002 significant Yes \nKrasnodar 0 significant Yes \nKhabarovsk 0.0203 significant Yes \nSaratov 0 significant Yes \nTolyatti 0.0002 significant Yes \nMoscow 0 significant Yes \nPerm 0.0001 significant Yes \nSaint Petersburg 0 significant Yes \nLag included = 3 \nVladivostok 0.0001 significant Yes \nKrasnodar 0 significant Yes \nKhabarovsk 0.0015 significant Yes \nSaratov 0 significant Yes \nTolyatti 0 significant Yes \nMoscow 0 significant Yes \nPerm 0 significant Yes \nSaint Petersburg 0 significant Yes \n \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\nWhen using only one lag no Granger causality was found, aside from the case of Vladivostok. This \nmeans that the first lag of the dependent variable is sufficient in forecasting its future values and there \nis no improvement when adding the lag of the second time series. However, when using 2 and 3 lags \nthe past values of daylight hours improve the predictive power of the model compared to not adding \nthem. \n \nThus, the test results allow us to conclude that forecasting the number of suicides by using its lags as \nwell as the lags of daylight hours improves the accuracy compared to only using the lags of the \nnumber of suicides when using 2 and 3 lags for each variable. This may indicate that there is a weak \ncausal relationship between suicides and daylight hours.  Indeed, we can see that when using the \nestimates, the results seem to be a l ot more uniform in the sense that all the cities exhibit Granger \ncausality starting from lag 2 at the same time.  This might be a hint at the fact that the actual values \npresented are not entirely representative of the truth given the hypothesis that daylight hours Granger \ncause suicide coefficients. \n \nDiscussion \n \nThis research is, to the best of our knowledge, the first to attempt to estimate the relationship between \ndaylight hours and suicidal behaviour in Russia. The link between suicides and climate factors is an \nactive field of research in other countries, while lacking its deserved attention in the Russian scientific \ncommunity. We have only found one Russian paper in which climate and ecological factors are \nmentioned with regards to being connected with the fluctuations of suicidal behaviour seasonality in \nRussia, although not statistically analyzed (Rozanov & Grigoriev 2018b). \n \nDespite analyzing 8 cities in Russia, which are located on different latitudes with the maximal \ndifference between any given two cities being 16 degrees, the suicides seasonality profile after \naccounting for possible latent cases turned out to be quite similar amo ng the m: the maximum \noccurring most commonly in June and minimum in December. This phenomenon could be connected \nwith the predominantly northern locations of Russian cities, thus, more data on some of the more \nsouth countries of the world is needed to more accurately analyze the difference between the seasonal \nfluctuations of suicides. \n \nStudying causality in time series data is an extremely difficult task. The variables in question may be \ninfluenced by hidden variables, the correlation could end up being spurious and mislead the \nresearchers into false discoveries. Currently there is no method to guarantee the presence or lack of \ncausal relationship between two time series. There is also a problem of the third variable, which can \ninfluence the two series, making it seem like the relationship is there, while in reality it is not. The \nGranger causality test does not control for that, making it an unreliable, yet, perhaps, the only viable \noption to try and at least attempt to find a causal relationship. \n \nConclusion \n \nThis scientific research provides an analysis of suicides seasonality in various Russian countries. The \noverall trend is an increase of the number of suicides in the spring -summer period and a decrease in \nautumn-winter, with Tolyatti not following the pattern and having its suicide peaks in January . This \ncould be connected with the underreporting of suicides and classifying them as other causes with \nevents of undetermined intent , in particular, hanging, strangulation and suffocation, undetermined \nintent; falling, jumping or pushed from a high place, lying or running before or into moving object, \nundetermined intent. Havning taken these causes into account changed the distribution of suicides by \nmonth in the researched cities . This seasonal profile in Russian cities may arise due to the varying \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\ndaylight hours in each month.  For instance, high, positive and significant correlation was found \nbetween suicides and daylight hours.  \n \nThe results of the Granger causality test showed th at forecasting the number of suicides by using its \nlagged values as well as the lags of daylight hours improves the accuracy when compared with just \nusing the lags of the number of suicides for the case of 2 and 3 lags being used in the model . This \ncould indicate that daylight hours Granger cause the number of suicides. However, when using ony \n1 lag no Granger causality has been found between the variables. \n \nResearch limitations \n \nIn order to conduct the Granger causality test, each time series for all countries had to be differenced \nto make them stationary. This could have tampered with the behaviour of the time series, skewing the \ntesting results.  \n \nThis research analyzes the suicides which were officially documented, thus, some suicides may have \nbeen missing entirely or classified as events of undetermined intent. The authors have attempted to \nincorporate latent suicides into the analysis by including two other causes of death: h anging, \nstrangulation and suffocation, undetermined intent ; falling, jumping or pushed from a high place, \nlying or running before or into moving object, undetermined intent . However, not all such cases are \nnecessarily latent suicides, it might only be some fraction. Also, not all latent suicides are located in \nthese two causes in particular, other causes may contain some latent suicides as well. For instance, \nsome researchers find latent suicides in relatively high quantities among medicinal poisonings \n(Semyonova et al. 2020). \nSince this work only analyzes actual suicides and does not consider attempted ones, it is not possible \nto see whether the same seasonal patterns are present in attempted suicides. \n \nDespite all the limitations, this paper contributes to existing knowledge o n suicides seasonality and \nits possible causes. \n \nReferences \n \n1. Amiri, S. (2022). Unemployment and suicide mortality, suicide attempts, and suicide ideation: \nA meta -analysis. International Journal of Mental Health, 51(4), 294 -318. \nhttps://doi.org/10.1080/00207411.2020.1859347  \n2. Andreev, E, Shkolnikov VM, Pridemore WA, Nikitina SY (2015) A method for reclassifying \ncause of death in cases categorized as “event of undetermined intent” . Population Health Metrics  \n13(23): 1–25. https://doi.org/10.1186/s12963-015-0048-y  \n3. Angst J, Angst F, Stassen HH (1999) Suicide risk in patients with major depressive disorder. \nJ Clin Psychiatry 60(2): 57–62. https://pubmed.ncbi.nlm.nih.gov/10073389/  \n4. Danilenko KV , Kobelev E, Zhanaeva SY , Aftanas LI (2021) Winter-summer difference in \npost-awakening salivary α-amylase and sleepiness depending on sleep and melatonin. Physiol Behav. \nhttps://doi.org/10.1016/j.physbeh.2021.113549  \n5. Dollins AB, Lynch HJ, Wurtman RJ, Deng MH, Kischka KU, Gleason RE, Lieberman HR  \n(1993) Effect of pharmacological daytime doses of melatonin on human mood and performance . \nPsychopharmacology (Berl). 112(4): 490–496. https://doi.org/10.1007/BF02244899  \n6. Fingerhut LA, Cox CS  (1998) Poisoning mortality, 1985 –1995. Public Health Rep  113(3): \n218–233. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1308672/  \n7. Frey, L. M., Cerel, J. (2015). Risk for suicide and the role of family: A narrative review. \nJournal of Family Issues, 36(6), 716-736. https://doi.org/10.1177/0192513X13515885  \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\n8. Hajiyousouf, I. I., Bul ut, S. (2021). Mental health, religion and suicide. Open Journal of \nMedical Psychology, 11(1), 12-27. https://doi.org/10.4236/ojmp.2022.111002  \n9. Holopainen J, Helama S, Björkenstam C, Partonen T  (2013) Variation and seasonal patterns \nof suicide mortality in Finland and Sweden since the 1750s // Environ Health Prev Med  18(6): 494–\n501. https://doi.org/10.1007/s12199-013-0348-4  \n10. Jenkins TA, Nguyen JC, Polglaze KE, Bertrand PP  (2016) Influence of Tryptophan and \nSerotonin on Mood and Cognition with a Possible Role of the Gut -Brain Axis. Nutrients 8(1): 56. \nhttps://doi.org/10.3390/nu8010056  \n11. Kim CD, Lesage AD, Seguin M, Chawky N, Vanier C, Lipp O, Turecki G  (2004) Seasonal \ndifferences in psychopathology of male suicide completers . Compr Psychiatry  45(5): 333–339. \nhttps://doi.org/10.1016/j.comppsych.2004.06.007  \n12. Krenev А. А., Vasin S. A. (2012)  «Rod smerti ne ustanovlen» - vedushchaya prichina \nnasilstvennoj smertnosti v Rossii [‘The manner of death has not been established’ is the leading cause \nof violent death s in Russia ]. Demoskop [ \nDemoscope] Weekly 535. https://publications.hse.ru/articles/77703821 (in Russian) \n13. Lee AY , Pridmore S  (2014) Absence of seasonality of suicide in Tasmania (Australia) . \nAustralas Psychiatry 22(2): 204–206. https://doi.org/10.1177/1039856214526287  \n14. Lee, C. S., Wong, Y . J. (2020). Racial/ethnic and gender differences in the antecedents of youth \nsuicide. Cultural Diversity & Ethnic Minority Psychology, 26(4), 532.  \nhttps://doi.org/10.1037/cdp0000326  \n15. Marazziti, D., Avella, M. T., Mucci, N., Della Vecchia, A., Ivaldi, T., Palermo, S., & Mucci, \nF. (2021). Impact of economic crisis on mental health: a 10-year challenge. CNS spectrums, 26(1), 7-\n13. https://doi.org/10.1017/S1092852920000140  \n16. Młodozeniec A, Brodniak WA, Polewka A, Bembenek A. (2010) Sezonowość samobójstw w \nPolsce. Analiza danych Głównego Urzedu Statystycznego z lat 1999-2003 [Seasonality of suicide in \nPoland. Analysis of the Main Statistical Of fice data for the years 1999 -2003]. Psychiatr Pol  44(1): \n61–69. https://pubmed.ncbi.nlm.nih.gov/20449981/  \n17. Petridou E, Papadopoulos FC, Frangakis CE, Skalkidou A, Trichopoulos D  (2002) A role of \nsunshine in the triggering of suicide . Epidemiology 13(1): 106–109. \nhttps://doi.org/10.1097/00001648-200201000-00017  \n18. Piatkowska, S. J. (2020). Poverty, inequality, and suicide rates: a Cross-National Assessment \nof the Durkheim Theory and the Stream Analogy of Lethal Violence. The Sociological Quarterly, \n61(4), 787-812. https://doi.org/10.1080/00380253.2020.1715308  \n19. Rozanov V .A., Grigoriev P.E. (2018) Environmental factors and suicide behavior in human \nbeing. Suicidology 9 (2): 30 -49. https://cyberleninka.ru/article/n/ekologicheskie-faktory-i-\nsuitsidalnoe-povedenie-cheloveka (in Russian) \n20. Rozanov V .A., Grigoriev P.E., Zakharov S.E., Kryvda G.F. (2018) Analysis of completed \nsuicides seasonality in relation to such external factors as length of the day and ambient temperature. \nSuicidology. 9 (3): 71-79. https://doi.org/10.32878/suiciderus.18-09-03(32)-71-79 (in Russian) \n21. Semyonova V .G., Ivanova A.E., Sabgaida T.P., Zubko A.V ., Mikhailov A.Yu., Evdokushkina \nG.N., Zaporozhchenko V .G. (2020) Social and economic determinants of the age structure of \nmortality from suicide in Russia. Health care of the Russian Federation. 64(5):  243-252. \nhttps://doi.org/10.46563/0044-197X-2020-64-5-243-252 (in Russian) \n22. Snowdon, J, Choi NG  (2020) Undercounting of suicides: Where suicide data lie hidden. \nGlobal Public Health 15(12): 1894–1901. https://doi.org/10.1080/17441692.2020.1801789  \n23. Souêtre E, Salvati E, Belugou JL, Douillet P, Braccini T, Darcourt G (1987) Seasonality of \nsuicides: environmental, sociological and biological c ovariations. J Affect Disord 13(3): 215–225. \nhttps://doi.org/10.1016/0165-0327(87)90040-1  \n24. Tietjen GH, Kripke DF (1994) Suicides in California (1968–1977): absence of seasonality in \nLos Angeles and Sacramento counties. Psychiatry Res 53(2): 161–172. https://doi.org/10.1016/0165-\n1781(94)90107-4  \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999\n\n25. Vasin S. А. (2015). Mortality from undetermined causes of death in Russia and in a se lected \nset of countries. Demographic Review 2(1): 89-124. https://doi.org/10.17323/demreview.v2i1.1790  \n26. Vishnevsky A. G. (2017) Smertnost ot vneshnih prichin v Rossii s serediny XX veka  \n[Mortality from external causes in Russia since the mid-20th century]. Izdatelskij dom Vysshej shkoly \nekonomiki [HSE Publishing House], Moscow 448 p. https://doi.org/10.17323/978-5-7598-1397-2 (in \nRussian) \n27. Wang Y , Jiang Y, Zou M, Liu J (2021) Antidepressant actions of melatonin and melatonin \nreceptor agonist: Focus on pathophysiology and treatment . Behav Brain Res 420 (9): 113724. \nhttps://doi.org/10.1016/j.bbr.2021.113724  \n28. Wu YH, Ursinus J, Zhou JN, Scheer FA, Ai-Min B, Jockers R, van Heerikhuize J, Swaab DF \n(2013) Alterations of melatonin receptors MT1 and MT2 in the hypothalamic suprachiasmatic \nnucleus during depression . J Affect Disord 148 (2–3): 357–367. \nhttps://doi.org/10.1016/j.jad.2012.12.025 \n29. Yumaguzin V. V. (2017) Smertnost ot povrezhdenij s neopredelennymi namereniyami kak \npokazatel kachestva statistiki smertnosti ot vneshnih prichin [Mortality from events of undetermined \nintent as an indicator of the data quality on mortality from external causes].  Aist na kryshe. \nDemograficheskij zhurnal [Stork on the roof. Demographic magazine ] 5: 40 -47. \nhttps://publications.hse.ru/articles/211645194 (in Russian) \n30. Yumaguzin V ., Vinnik M. (2019). Assessment of the Real Rates of Homicides and Suicides in \nthe Regions of Russia. Sotsiologicheskie issledovaniya 1: 116–126. \nhttps://doi.org/10.31857/S013216250003753-1 (in Russian) \n \nOther information sources  \n \nUPOLNOMOCHENNYJ PRI PREZIDENTE ROSSIJSKOJ FEDERACII PO PRA V AM \nREBENKA: Mariya L'vova -Belova predstavila itogi raboty za 2021 god . [COMMISSIONER \nUNDER THE PRESIDENT OF THE RUSSIAN FEDERATION FOR CHILD RIGHTS : Maria \nLvova-Belova presented the 2021 work results] https://deti.gov.ru/Press-Centr/news/958 (in Russian) \n \nNoNews: Ranking of countries by suicide rate. https://nonews.co/directory/lists/countries/suicide-\nrate \n \nWHO: Suicide. https://www.who.int/news-room/fact-sheets/detail/suicide  \n \n \nAuthor-formatted, not peer-reviewed document posted on 28/03/2025. DOI:  https://doi.org/10.3897/arphapreprints.e153999","source_license":"CC-BY-4.0","license_restricted":false}