Abstract
Public stereotypes of young people throughout the industrialized world are often grim, with researchers and popular writers warning of a generation in crisis, discontented with the present and pessimistic about the future [1–4]. In support of this view, commentators point to social indicators such as the epidemiological evidence that depressive disorders and suicidality, particularly among young men, have increased sharply over recent decades. We could refer to this line of reasoning as ‘the absolute misery hypothesis’: the intuitively straightforward notion that the suicide rate of a generation is a surrogate measure of that generation's happiness. The more suicides, the greater the level of unhappiness. As Eckersley [5], p.S16] once put it, ‘I believe that behind suicide and other youth problems also lies a profound and growing failure of the culture of western industrial societies – a failure to provide a sense of meaning, belonging and purpose in our lives, and a framework of values’. Except for declines during and immediately after both World Wars, suicide mortality has been increasing for most of this century [6]. Presumably, then, the tide of social malaise has been rising for many years now.
The proposal that suicide rates reflect social maladjustment in such a straightforward fashion must be reconciled with the rises in national suicide rates that tend to accompany good fortune and the falls that accompany national adversity. Durkheim's [7] observation that suicide rates decline during war, for example, has been reliably replicated many times in recent history [8]. Somewhat less well publicized, however, have been the sharp rises in suicide rates that frequently follow national jubilation such as has occurred in Eastern European countries since the fall of Communism and the so-called ‘peace dividend’ in Northern Ireland [9]. Research has also shown a positive linear correlation between various economic indicators of quality of life and suicide rates, both within [10] and between countries [11, 12]. Following Durkheim [7], the absolute misery hypothesis predicts that during times of adversity, the country's level of social integration will rise as its citizens pull together to defeat a common enemy, but in times of peace and prosperity, patriotism and superordinate values will be eroded, causing a decline in national morale and a consequent rise in suicidality.
The first study reported in this paper used secondary data to compare the psychological wellbeing of adolescents from a diverse group of countries. In so doing, it sought to establish whether the youth suicide rate in those countries was a valid indicator of social maladjustment, as the absolute misery hypothesis predicts.
Study I
Method
Overview
The project on which this study was based was designed to examine aspects of adjustment to school of high school students in eight communities: Hong Kong, Taipei, Osaka, Berlin, Winnipeg, Phoenix, Canberra and Brisbane [13]. English-language questionnaires were translated into Cantonese, Mandarin, Japanese and German. The fundamental objective was to identify associations between adaptive problems and personality characteristics, on the one hand, and specific family patterns on the other. Items measuring social adjustment were extracted from this data set and compared with suicide rates for these same countries [14].
Respondents
Selection of the eight communities and the schools within them was purposive rather than random, depending on the presence of local collaborators and on the willingness of schools to participate. A preliminary version of the questionnaire was sent out by mail to district superintendents in these and other communities. Many of the superintendents declined to approve the study and once a school district agreed to participate, it was still necessary to find willing principals and teachers. Participation rates within schools also varied. In China, Hong Kong and Japan the student participation rate was close to 100% because no prior parental permission was required. In both Australian cities, the participation rate was at least 80%, but in the USA, Canada and Berlin, participation rates were unknown because only the names of those students whose parents signed a consent form were released to the researchers. The final sample (Table 1) involved 2619 children from 31 schools in grades seven to 12 (ages 11–20 years, mean = 15.5, SD = 1.7). Forty-four per cent of the sample (1139) were male and 56% (1479) were female. One student failed to record his or her gender.
Profile of respondents
Measures
The questionnaire was administered in class during 1989 and contained 122 questions. Sixty-five of the questions covered student satisfaction with school, self and friends, with the remaining questions examining students' perceptions of their family. Questionnaires were translated from English into Cantonese, German, Japanese and Mandarin, and the adequacy of the translations was checked by the original study collaborators, including through the use of forward-and back-translation [13].
As standardized measures of adjustment were not included in the original data set, items from the student questionnaire were examined for their relevance to self-esteem, adjustment at school and social adjustment. In all, 16 items were extracted: five for self-esteem, nine for social adjustment and two for school adjustment. The 16 items are presented in Table 2.
Items used to construct composite measures of psychological adjustment
Although responses to all adjustment questions were made on Likert-type scales, the number of points varied from item to item. Thus, to ensure that each item received equal weighting when combined into a single measure, all items were converted to 5-point rating scales. Internal consistency for all re-weighted adjustment scales was satisfactory: α= 0.79 for self-esteem, and also social adjustment; and the correlation between the two school adjustment questions was quite high at r(2616) = 0.64. When all items were combined to produce a single adjustment measure, internal consistency for the sample as a whole reached α= 0.87, and within-country alphas ranged from a low of α= 0.75 in Japan to α= 0.83 in Canada.
Results
Summary scores for self-esteem, social adjustment and school adjustment scales were obtained by adding the relevant items. Thus, scores on the self-esteem scale could range from five to 25; on the social adjustment scale from nine to 45; and on the school adjustment scale, from two to 10. In every case, higher scores indicate higher levels of adjustment. Scores for males and females on the three adjustment subscales and for total adjustment scores are presented in separate tables broken down by country. Tables 3 and 4 also record suicide rates (1991–1993) among the 15–24 years age group for these same seven countries. The youth suicide rate for each country was obtained from 1994 World Health Organization statistics and was the latest available data at that time. In Germany, Hong Kong, China (urban) and Canada, the figures in Tables 3 and 4 relate to the 1992 calendar year, Japanese and Australian figures are for 1993, and the USA figures relate to 1991. In recording these figures, participating countries referred to the selected causes of the Basic Tabulation List of the International Classification of Diseases, Ninth Revision (ICD-9) which appeals to the concept of the underlying cause of death, defined as ‘the disease or injury which initiated the train of morbid events leading directly to death, or the circumstances of the accident or violence which produced the fatal injury’ [15].
Mean (SD) adjustment scores and youth suicide rates across the seven countries (males)
Mean (SD) adjustment scores and youth suicide rates across the seven countries (females)
Tables 2 and 3 reveal considerable variation in male and female suicide rates across the seven countries. The male suicide rate ranged from a low of 5.6 per 100 000 in China to a high of 24.7 in Canada, while the female suicide rate ranged from a low of 3.3 per 100 000 in Germany to a high of 10.6 in China. In the case of males, suicide rates tended to cluster into four groups: the lowest rate was recorded in China; followed by Hong Kong and Japan; followed by Germany; followed by the USA, Australia and Canada. In the case of females, lowest suicide rates were reported in Germany; followed by the USA, Australia and Japan, which reported similar rates; followed by Hong Kong and Canada, and finally by China which reported much higher female suicide rates than any other country.
The four adjustment measures were subjected to one-way analysis of variance followed by Fisher (least significant difference) post-hoc contrasts for males and females separately. For males and females, all analyses proved significant. In the case of males: total adjustment F = 69.62, df = 6,1132, p < 0.0001; self-esteem F = 120.41, df = 6,1132, p < 0.0001; social adjustment F = 19.04, df = 6,1132, p < 0.0001); and school adjustment, F = 72.83, df = 6.1132, p < 0.0001. In the case of females: total adjustment F = 127.26, df = 6,1472, p < 0.0001; self-esteem F = 143.82, df = 6,1472, p < 0.001; social adjustment F = 54.79, df = 6,1472, p < 0.001; and school adjustment F = 88.51, df = 6,1472, p < 0.001. Results of post-hoc comparisons are also presented in Tables 2 and 3. The post-hoc comparisons identified a fairly consistent rank ordering: for males, at the highest level of adjustment were the North Americans followed by Australia, then Germany, and, finally by the Asia-Pacific countries. The female rankings were similar, except that Australian females belonged with the North Americans in the group of best-adjusted nations.
Of more importance in the present study than international comparisons of adjustment is the relationship between mean adjustment scores and suicide across the seven countries as a whole. Despite the small sample size and truncated distribution of adjustment scores involved in these analyses, three of the four correlations between adjustment and suicide proved statistically significant for males. Specifically, the product moment correlation for total adjustment and suicide rate was r (7) = 0.74, p < 0.05; for self-esteem r(7) = 0.87, p = 0.01; and for school adjustment r(7) = 0.81, p < 0.05. The correlation between mean social adjustment score and suicide was also high, although not statistically significant, at r (7) = 0.65. For females, all correlations were moderate to weak and in the opposite direction: total adjustment r(7) = −0.37; self-esteem r(7) = −0.46; social adjustment r(7) = −0.33; and school adjustment r(7) = −0.23.
Discussion
At least in the case of males (who account for the majority of suicides in the industrialized world), results of this study do not support the ‘absolute misery hypothesis’ outlined earlier. On the contrary, higher rates of male suicide were associated with higher levels of psychological adjustment among the general adolescent population. The strength of these associations between adjustment and suicide was remarkable. In the case of self-esteem, for instance, more than 75% of the variance in suicide rates was accounted for by country-level scores on the self-esteem subscale. Another striking result was the apparent effect of gender on the relationship between suicide and adjustment. The moderate negative correlations between country-level female adjustment measures and suicide rates suggests that young women may be more likely to act in accord with the absolute misery hypothesis, although the absence of statistical significance in this small sample means that this interpretation must be cautiously proposed.
One interpretation consistent with the findings is a concept which can be referred to as ‘relative misery’. According to the ‘relative misery hypothesis’, young men's predisposition to suicide is influenced by their social comparisons. When those around them are perceived to be better off than they are, the distress of vulnerable youth is magnified and their susceptibility to suicidality increased. The relative misery hypothesis would therefore predict a rise in suicide rates whenever there is a rise in the overall mood or happiness of a population, whether that rise is due to the end of war or any other event of general and positive social significance. At such times, the misery of the community's unhappiest young men will be compounded by the isolation they experience at witnessing the happiness of those around them. By contrast, suicide in young women appears to be more influenced by internal cues (absolute misery) and less by their appraisal of others.
It is important to recognize that while the analysis performed here confirms the existence of what Cryer [16] has called an ‘ecological correlation’ between adjustment and suicide, there is no certainty that associations at the national level will also hold at the level of individuals within those countries. Another important limitation of the present study is that the data set contained no measure of the central construct – ‘relative misery’ – which is hypothesized to underlie young male suicide; nor was any standardized measure of psychological adjustment available. Finally, study I contained something of an age difference between the groupings used for suicide and adjustment comparisons. The mean age of respondents to the adjustment questionnaire was 15.5 years and a sizeable proportion fell below the 15–24 years age grouping for whom suicide statistics were available. On the other hand, the fact that the suicide statistics used in this study were collected between 2 and 3 years after the survey compensates for the age discrepancy to some degree, because more of the survey respondents would have come within the 15–24 years age bracket by that time.
In the next study, an unrelated secondary data set was used to examine the association between suicidality and relative misery at the level of individual respondents. Along with a measure of suicidality, this study incorporated a standardized measure of psychological adjustment (self-esteem) and a measure of social comparison. Furthermore, all these measures were performed on the same individuals. Under the relative misery hypothesis, it was expected that psychological adjustment alone would not be significantly associated with suicidality in adolescent males, but that a negative social comparison would also be necessary. In the case of adolescent females, it was expected that suicidality would be significantly associated with absolute levels of psychological maladjustment.
Study II
Method
Overview
The project on which the second study is based was designed primarily to assess the prevalence and correlates of drug-taking behaviour among Canadian adolescents living in the Province of Alberta. However, the questionnaire also contained measures of depressed affect, self-esteem, social comparison and suicidality which were extracted from the data set for use in this study. Students from years 7 to 12 across nine school districts were administered the anonymous in-class questionnaire.
Respondents
The sample consisted of 2111 junior and senior high school students drawn from 95 schools in the Canadian Province of Alberta. One school was randomly selected from each of the nine school districts in the province. Within these schools individual students were selected by stratified random sampling according to age group from 12 through 17 years. In this way, the proportion of students in each age group from the total population within each school district was reflected in the final sample. Individual school districts were responsible for drawing the sample of students based on a required sample size. In a few cases where individual sampling was not possible, a classroom sampling method was adopted. In these cases, one or more classes that were mandatory for all students within the schools were randomly selected for participation in the study, thereby enhancing the representativeness of the sample. The pupils who agreed to participate represent 72.5% of the children selected by the sampling strategy. The final sample was 49% female with an average age of 15.03 years (SD = 1.70). Twelve per cent of the sample was drawn from the seventh grade, with 19%, 24%, 15%, 19% and 12% from grades eight to 12, respectively.
Measures
The questionnaire consisted of four sections: (i) questions relating to family background and demographic characteristics; (ii) questions concerning how respondents felt about themselves and their achievements; (iii) questions about the school environment, family functioning and the respondents' social lives; and (iv) questions about respondents' experiences with cigarettes, alcohol and other drugs. We selected from this pool of questions those items that came closest to operationalizing the key constructs involved in the relative misery hypothesis: psychological adjustment, social comparison and suicidality. Psychological adjustment was measured using Rosenberg's 10-item scale for adolescents [17] and a single item possessing high face validity was used as an indicator of affective state. The single item was, ‘I am unhappy, sad or depressed’ and it was scored on a three-point scale: ‘1: Never or not true; 2: Sometimes or somewhat true; 3: Often or very true’. As shown in Table 5, this item correlated with self-esteem significantly and in the expected direction for a measure of depressed affect. With regard to the self-esteem measure, Rosenberg cites extensive research in support of the reliability and validity of this instrument, including an internal consistency coefficient of α= 0.85 and significant associations between self-esteem scores and nurses' assessments of respondents' levels of depressed affect, physiological indicators of neurosis and frequency of psychosomatic symptoms. Two items from the questionnaire assessed suicidal ideation and suicidal behaviour respectively: ‘I think about killing myself’, and ‘I deliberately try to hurt or kill myself’. Both items were scored on three-point scales, and the correlation between them was r = 0.61. The items were added together to provide an overall measure of suicidality. Finally, social comparison was assessed by the single item, ‘I am not as happy as other children’. This item was scored on the same three-point scale as the depressed affect and suicide items.
Means (SD) and first-order correlations between suicidality, depressed affect and social comparison for males and females
Results
Bi-variate correlation analysis
Table 5 presents mean scores and first-order correlations between suicidality, depressed affect, self-esteem and social comparison for adolescent males and females, and shows that all four measures were significantly correlated.
Multiple regression analysis using depressed affect
In order to determine the relative importance of depressed affect and social comparison in the prediction of suicidality, a multiple regression analysis was conducted. This analysis also examined the interaction between social comparison and depressed affect to determine whether a combination of depressed affect and negative social comparison had an effect additional to those observed for depression and social comparison alone. Following Cohen and Cohen [18], a hierarchical entry was used in which the main effects (social comparison and depressed affect) were entered into the equation first, followed by the interaction term (social comparison × depressed affect). The results are shown in Table 6.
Hierarchical multiple regression analysis using depressed affect
As indicated by Table 6, suicidality in males was not directly related to depressed affect or social comparison on their own, but was significantly associated with the interaction between them. Given that negative social comparison and depressed affect were scored in a positive direction, the positive ?-value suggests that greater suicidality was associated with the combination of depressed affect and negative social comparison. The same interaction effect was also observed for girls, although the effect was smaller. In addition, there was a direct association between depression and suicidality after the interaction was taken out. Gender differences in the significance of depressed affect and the product term were assessed using Williams' [19] procedure for comparing slopes. This procedure revealed significant differences between males and females both for depressed affect (t = 2.20, p < 0.05) and for the interaction term (t = 2.60, p < 0.01), indicating that depressed affect was more important for females and the interaction term was more important for males. The same procedure also revealed that the effect of the interaction term was significantly greater than the effect of depressed affect for males (t = 7.76, p < 0.0001), but not for females (t = 1.86, NS).
Multiple regression analysis using self-esteem
The same analyses described above were performed with self-esteem as the measure of psychological adjustment. Because self-esteem was negatively correlated with suicidality (see Table 5), the scoring of social comparison was reversed for this analysis in order to simplify the interpretation of the product term. Thus, unlike in Tables 5 and 6, a negative correlation between social comparison and suicidality in Table 7 means that suicidality is associated with negative social comparison. Table 7 shows that a similar pattern of effects applied to self-esteem as applied to the indicator of depressed affect. For males, the model identified no direct association between self-esteem and suicidality, but a highly significant effect of the interaction between self-esteem and social comparison. There was also a main effect of (negative) social comparison. For females, both main effects were significant: greater suicidality was associated with lower self-esteem and negative social comparison. However, these results must be qualified by the highly significant interaction term, indicating that the combination of the two variables was also of importance in explaining female suicidality. Finally, slope comparisons revealed a significant difference between males and females in the importance of self-esteem (t = −19.80, p < 0.0001), but not of the product term (t = 1.71, NS) (n = 2079).
Hierarchical multiple regression analysis using self-esteem
General discussion
This paper began by asking whether suicide rates were a reasonable proxy for the level of psychological maladjustment within the youth population. Results of study I suggested just the opposite: that the male suicide rate is a proxy for the level of psychological wellbeing within the general population. Study II then provided evidence regarding the psychological mechanism responsible for this counter-intuitive finding. In short, study II found that suicidality in the young male population is related to upward social comparison. More specifically, suicidality in males requires both a level of psychological maladjustment and the perception that one is worse off than one's peers. I have labelled this phenomenon ‘the relative misery hypothesis’ because it proposes that the rate of suicide among young men is a function of the level of relative rather than absolute unhappiness within the community. Study I produced equivocal results in relation to the association between female suicide rates and aggregate levels of psychological adjustment, and study II again helped to clarify why this was the case. In essence, study II found that suicidality in females is influenced both by absolute and relative misery. While relative misery is undoubtedly an important determinant of young women's suicidality, it is less important than for males. Moreover, unlike males, the effect of absolute misery on suicidality in young women remains even after the contribution of relative misery has been removed.
The relative misery hypothesis provides an elegant explanation for the paradox described in the introduction to this paper, that suicide rates rise during national jubilation and fall during national adversity. In short, when national morale rises and falls, the relative misery of the community's unhappiest citizens moves in concert. Unlike Durkheim's [7] celebrated theory of suicide, then, the relative misery hypothesis does not appeal to fluctuations in the overall level of social integration, but to fluctuations in the relative affective state of a small minority. But while the relative misery hypothesis is inconsistent with Durkheim's sociological account of suicide, it is consistent with recent epidemiological research into the health effects of income inequality [20–23]. While not denying that absolute deprivation is bad for one's health, researchers in the field of ‘relative poverty’ argue that income inequality matters in its own right for population health. For example, drawing on USA data, Kennedy et al. [23] found that income inequality was strongly correlated with mortality rates even after controlling for median income, poverty rates, smoking prevalence and race. According to this line of reasoning, deprivation and its opposite are relative constructs and, within limits, where we stand in relation to others matters a great deal to our health and wellbeing. Just as the ‘relative poverty’ of a society can be defined as the difference between income levels, so relative misery can be defined as the difference (or perceived difference) between affective states in that society.
Further research into the relative misery hypothesis is now needed to replicate and elaborate the phenomena identified in these pages. In the first place, results of study I should be replicated on a larger and more diverse sample of nations. As Lester and Stack [11] have shown, cross-cultural correlations between suicide rates and quality of life can critically depend on the sample of nations chosen. In their study, for example, the authors found a strong association between suicide rates and quality of life in the sample as a whole, but the correlation did not hold when the analysis was conducted only on European nations. Similarly, Zhang [12] showed that indicators of modernization were better predictors of the suicide rate in developing than developed countries. And since all of the countries in this study except China were industrialized nations, results may not apply to developing countries. Moreover, the samples used in studies I and II were drawn from non-random samples of the general adolescent population, and the extent to which suicidality in the clinical population is governed by the mechanisms identified in the general population is unclear. Indeed, even the extent to which the samples are representative of the nations as a whole is unclear. Finally, since standardized measures of key constructs were not always available in the secondary data sets used here, replication studies are needed which incorporate only instruments of known psychometric properties.
In summary, then, these studies contain numerous methodological flaws, particularly the use of nonrandom samples and unstandardized measures of key constructs. Nevertheless, the results have potentially important implications for suicide prevention. Until now, the psychosocial factors that have received most attention, at least in psychological autoposy research, are psychiatric illness and stressful life events such as interpersonal loss and problems with authority [24–26]. Not surprisingly, such stressors have been related to psychiatric illness in the suicidal, but life stress is known to be associated with suicide even after adjusting for mental illness [27]. Results of the present studies suggest the existence of another important, but previously unidentified environmental stressor: happy, well-adjusted peers. As with many new discoveries in science, this concept of relative misery has arisen from some anomalous findings in studies that were designed for other purposes. What is clearly needed now are studies designed explicitly (and therefore more adequately) to test this new ‘relative misery hypothesis’.