Gender Differences in the Association between Psychosocial Factors and Depressive Symptoms in Japanese Junior High School Students

To determine psychosocial factors associated both with depressive symptoms and with gender differences in depressive symptoms among junior high school students, we conducted selfadministered questionnaires using a sample of 2,660 students of 13 public junior high schools in Okinawa, Japan. Depressive symptomatology was measured by the Center for Epidemiologic Studies Depression Scale (CES-D). The psychosocial variables examined were life stresses, social support, health practices, self-esteem, and locus of control. Depressive symptoms were positively associated with life stresses and negatively associated with health practices, social support, self-esteem, and internal locus of control. In addition, female students were likely to report more depressive symptoms, life stresses, and low selfesteem and poor health practices. After controlling for the psychosocial variables differed by gender, gender differences in depressive symptoms were eliminated. In conclusion, for depressive symptoms of junior high school students, life stresses might be risk factors, but positive health practices, social support, high self-esteem, and internal locus of control might be protective factors. Gender differences in depressive symptoms could be explained by the females' elevation on these psychosocial variables. J Epidemiol, 2000 ; 10 : 383-391

As for gender difference in depressive symptoms of adolescents, consistent with findings from previous studies of adult depression, first, female adolescents reported more depressive symptoms than males did 6-8).Second, gender difference in depressive symptoms appears to take place somewhere early and middle adolescence and is stable throughout the adolescent period to adulthood 9).Third, the female preponderance in depressive symptoms is demonstrated in Japanese high school students as well 10).Forth, several researchers have pointed out that the emergence of gender difference in adolescent depressive symptoms could be explained by the female's higher levels of psychosocial variables because gender difference was eliminated after controlling for effects of the psychosocial variables 4,11).We also found in our earlier studies of Japanese adolescents that females scored in a more pathological direction than did males on many psychosocial variables 12.13).Therefore, it seems reasonable to suppose that the association between gender and depressive symptoms in Japanese adolescents will be influenced by the psychosocial variables.However, this assumption has been not yet substantiated by research findings.
In this study, we explored the psychosocial variables associated with depressive symptoms among junior high school students in Okinawa, Japan and determined the influence of the psychosocial variables on the relation of gender and depressive symptoms.

Procedure and Subjects
Using a self-administered anonymous questionnaire, the survey was carried out from September through November in 1998, in classroom settings with the help of classroom teachers.Students were informed of the nature and intent of the study before responding to the questionnaire, and those who do not want to participate in the study could decline to respond and designate that on the front sheet of the questionnaire.We did not follow up students who were absent from school when the survey was conducted.
The population in this study was adolescents in grades 7 through 9 (12-15 years old) at thirteen public junior high schools in Okinawa, Japan.Okinawa Prefecture is located in the south-westernmost of Japan with a population of 1.3 million.It is divided into six school districts (Kunigami, Nakagami, Naha, Shimajiri, Miyako, and Yaeyama).Among them, Miyako and Yaeyama districts are located in remote areas, while the others are situated on Okinawa Island.All of the 13 schools were chosen one to four schools from each district depend on the size of student population, based upon willingness of school administrators to participate in the study.
The study sample was 2,660 students enrolled from one to two classes in each grade of each school.The questionnaires were collected from 2,482 students.Five students declined to participate.A total of 2,027 students completely responded in depressive scale (1,030 males, 997 females), and their data were used for analysis.The proportions of male and female subjects and school districts were nearly in accord with the results of the basic school survey conducted by the Prefectural Board of Education in 1998 14).Therefore, the sample in this study represented junior high school students in Okinawa as a whole.

Soeiodemographic variables
Gender, grade level, place of residence, way of commuting, participation in club activities, going to cram schools, and parental education level were assessed.

Depressive symptoms
Depressive symptomatology was measured using the Japanese version 15) of the Center for Epidemiologic Studies Depression Scale (CES-D) 16).The reliability and validity of the scale were verified both in the U.S. and Japanese adults 15,16) Several researchers also reported acceptable levels of internal consistency and test-retest coefficients when the scale was administered to adolescents 11.17, 18).Then, the use of this scale has been expanded to adolescents both in the U.S. and Japan 11.17 -19) The subjects were asked to rate frequency or duration of each symptom in the past week.Each item was scored on a 4-point scale (0 = rarely or none of the time [less than 1 day]; 1 = some or a little of the time [ 1-2 days]; 2 = occasionally or a moderate amount of time [3-4 days]; 3 = most or all of the time [5-7 days]).

Life stress
We measured life stress by the Adolescents Daily Events Scale (ADES) 2D).The ADES was developed to assess the frequency and negative impact to which subjects reported having experienced daily events during the past six months.It contained 25 items judged suitable for adolescent daily events to cover each of the following domains: club activity, academic, teachers, family, and friends.The validity and reliability of the ADES were verified in Japanese adolescents 20).Subjects first rated a 4-point scale of frequency of occurrence of each event (0 = not at all; 1 = rarely; 2 = sometimes; 3 = very often).Then, subjects rated those events that had experienced during the past 6 months on 4-point scales of negative impact level (0 = not impact at all; 1 = a little impact; 2 = considerably impact; 3 = extremely impact).The sum of the products of the frequency and negative impact scores for each event was counted as the scale.We obtained domain-specific scores by summing the products over each event in a specific domain.

Social support
We measured the perceived social support that indicated intensity of expectancy for receiving support in case of any trouble, using the revised version for junior high school students of the Scale of Expectancy for Social Support (SESS) 2u.Sixteen items were measured in accordance with sources of support such as father, mother, siblings, teachers, and friends respectively.This revised SESS seems to be reliable and valid for Japanese junior high school students with internal consistency and reducing effect of stress 21).Although the SESS used a 4-point scale, students were asked in our study to encircle all the persons from whom they can expect support to make the questionnaire simpler.Each encircled source was rated as 1 point and other sources were rated as 0 point.The total points of all resources of supports were used as scale score (0 to 80).

Health practices
We used the six common health practices including hours of sleep, eating breakfast, eating between meals, smoking, drinking, and physical activities, which were based on the Alameda County Study 22).Each of the following response was considered a positive health practice: sleeping 7-8 hours per night; eating breakfast every day; snacking between meal sometimes or never; not smoking within a month; not drinking within a month; doing physical activities more than once per week except attending physical education classes.Because previous studies have used a simple sum of the total number of positive health practices and have shown that it is a significant predictor of depression 23), the number of positive practices was summed to yield the health practices index ranging from 0 to 6.

Self-esteem
We evaluated self-esteem using the Japanese version A) of the Rosenberg Self-esteem Scale 4, a 10-item self-rating one with a 4-point scale (1 = strongly disagree; 2 = disagree; 3 = agree; 4 = strongly agree).Construct validity and high internal consistency of the scale have been demonstrated in middleschool age youth 1).In addition, this scale has been widely used to assess self-esteem of adolescents both in the U.S. and Japan 11.27)

Locus of control
Locus of control is an index of whether an individual perceives that reinforcement is contingent upon one S own behavior or attributes 28).Internal locus of control indicates that the individual believes in this contingency, whereas an external locus of control shows a lack of such believes.We measured the locus of control orientation using a 18-item scale developed by Kambara et al. 29), which originated from RotterS I-E scale 28) .This scale seems to be reliable and valid to apply for Japanese adolescents with internal consistency and factorial validity 30) .Each item was scored in the direction of internal locus of control on a 4-point scale (1 = strongly disagree; 2 = disagree; 3 = agree; 4 = strongly agree).

DATA ANALYSIS
Preliminary analyses examined descriptive statistics and zero-order correlations for depressive symptoms and psychosocial variables.We partly used non-parametric analyses to accommodate skewed distributions of data from several instru-ments.To determine the influence of the psychosocial variables on depressive symptoms, a hierarchical multiple regression analysis was conducted.A major advantage of the hierarchical multiple regression analysis is that once the order of sets of independent variables has been specified, a unique partitioning of the total variance in dependent variable accounted for by the sets of independent variables may be made.R2 change gives the increase in variance of dependent variable accounted for by a set of independent variables at that point beyond what has already been accounted for by prior sets 31).In this study, sociodemographic variables were entered before psychosocial variables to control for possible effects of these variables on depressive symptoms and to ensure that any effects found for psychosocial variables would not be attributable to shared variance with sociodemographic variables.Finally, gender was entered after the sociodemographic and psychosocial variables to examine the unique incremental contribution of gender to depressive symptoms.In each regression, any interaction effects were not computed due to many combinations of independent variables.All computations were performed using the SPSS release 9.0 statistical package 32).The hierarchical multiple regression analysis was carried out using the REGRES-SION command.We applied a criterion of p < 0.05 for significance.

RESULTS
Cronbach alpha coefficients of the main scales used in this study were moderate, ranging from .51 to .93.Therefore, it seems that these scales have been shown to have adequate internal reliability.

Effects of sociodemographic variables on depressive symptoms
The mean scores on CES-D by sociodemographic variables are provided in Table 1.The mean scores of males and females were 15.8 and 18.4, respectively.As expected, females had significantly more depressive symptoms than males.There were also significant differences in depressive symptoms across grade levels, 8th graders were more likely to report depressive symptoms than other graders.As for participation in club activities, students not participating in club activities had more depressive symptoms than those participating in such activities.Additionally, students not going to cram schools had more depressive symptoms than those going to those schools.There were no differences in depressive symptoms by place of residence, way of commuting, and parental education level.

Gender difference in psychosocial variables
Table 2 shows the mean scores of the psychosocial variables by gender.There were significant gender differences in all of the psychosocial variables except club activity stress and locus of control.Females reported higher scores on academic, family, teachers', and friends' stress and social support than males, whereas males reported higher scores on health practices and self-esteem.

Correlates of depressive symptoms
We examined interrelationships among the psychosocial variables.Spearman correlations among these variables for the combined sample of male and female students were low to moderate, ranging from -.003 to .39, and a consistent pattern was not detected.Thus, we retained all these variables for further analyses.
Because some sociodemographic variables such as gender, grade level, participation in club activities, and going to cram schools were found to affect depressive symptoms in the preceding section, these variables were considered to be confounding factors.Therefore, a hierarchical multiple regression symptoms.For females, sociodemographic variables accounted for 0.6% of the variance in depressive symptoms.These.control variables did not show significant relations.On step 2, the psychosocial variables accounted for an additional 46% of the variance in depressive symptoms.In the final equation, life stresses in academic, teachers', family, and friends' domains were significant correlates of increased depressive symptoms and positive health practices, higher self-esteem, higher internal locus of control, and more perceived social support were again related to lower level of depressive symptoms.
Influence of psychosocial variables on gender differences in depressive symptoms Given our earlier findings that males and females differ in levels of most psychosocial variables, it appears that gender variations in levels of these variables may explain the higher levels of depressive symptoms in female students.We expect to find that controlling for these variables result in a reduction or elimination of any association between gender and depressive symptoms.To determine whether the psychosocial variables influenced the relation between gender and depressive symptoms, we performed a hierarchical multiple regression analysis in which sociodemographic variables were entered on step 1, all of the psychosocial variables differed by gender were entered on step 2, and gender was entered last.As shown in Table 4, prior extraction of the psychosocial variables eliminated the association between gender and depressive symptoms.Although data is not shown in table, gender effect was significant when entered first (* = .163,p<.0001; model F= 44.817, p<.0001, R2= .03).After controlling for sociodemographic variables on step 2, the gender effect still remained (/3=.162,p<.0001; model F= 14.752, p<.0001, R2=.04).However, when psychosocial variables were added to the model, the gender effect was disappeared (*= .019,P=.347; model F= 108.322, p<.0001, R2= .42).Therefore, gender differences in depressive symptoms could be accounted for by the psychosocial variables.

DISCUSSION
The psychosocial variables examined were found to have independent effects on depressive symptoms in junior high school students.Judging from the direction of correlations, life stresses may be risk factors and other psychosocial variables such as positive health practices, more social support, high self-esteem, and internal locus of control may be protective factors.These findings were consistent with the findings from previous studies in the West that many of the psychosocial variables associated with adult depression were also related to adolescent depression 4.11).Additionally, the results of this study revealed similar findings from other studies in Japanese high school students as well as Hong Kong Chinese adolescents 5,12).Therefore, these findings suggest that many similari-ties in the psychosocial correlates of depressive symptoms for adolescents and adults may be common both in Western and Asian countries.
Life stresses in the domains of friends, family, and academic were significantly associated with depressive symptoms both in males and females, as reported in previous studies 5, 33•.34).Although it was noted in the West that the influence of friends' stress increased from early adolescence into late adolescence S, our finding revealed that friends' stress was most strongly associated with depressive symptoms in junior high students as early adolescence.This may evidence that Japanese students are strikingly influenced by peer stress since early adolescence, and reflect a considerable important role of peer relationships.It was pointed out that although Japanese adolescents today had thin and superficial friendships and deep concerned over communications through new media such as a cellular phone, most of them presented friendships as the first reason of happy school lives 36).Therefore, adolescents who have difficulties to keep company with classmates may not feel happy at school, then report high levels of depressive symptoms.
Stress in teachers' domain was weakly but significantly associated with depressive symptoms only in female students.There was inconsistency in the relation between teachers' stress and depressive symptoms of adolescents among the findings from Western studies.For example, Daniels and Moos n 34) reported that school stress composed of items assessing interpersonal stresses with teachers and students did not predict higher depressive symptoms.Whereas, Siddique and D'Arcy 34 showed that school stress was associated with depressive symptoms and no gender differences in the magnitude of association were found.Previous studies have shown that female adolescents were likely to report more negative events and to perceive those events as more stressful than male adolescents33.38) .Further, female adolescents were particularly more vulnerable to stresses in the interpersonal domains 33).Because the identity of female adolescents is more closely linked with success in interpersonal relationships, their failure to make interpersonal relationships may cause an extreme negative impact on psychological functioning 39).In this study, Japanese female students experienced higher levels of interpersonal stresses with not only friends and family but also teachers at school, and reported more depressive symptoms in relation to these interpersonal stresses.
Furthermore, female students in this study reported more academic stress and showed a greater association of this stress with depressive symptoms than male students.These findings were not consistent with the findings of Wagner and Compas 33) that there was no gender difference in reporting academic stress or associated distress.Recently, Japanese female junior high school students fairly intend to be a high educational background 36).In fact, the percentage of female graduates who go on to university is higher than that of their male counterparts.In this circumstance, female students may invest them-selves in academic achievement and be more aware to teachers' expectations and of competition among students.Therefore, our finding may be a reflection of the academic plight of Japanese female students.
Stress in club activity domain was not associated with depressive symptoms.This finding was consistent with the finding of Okayasu et al. 21) that this stress was not related to various stress responses in Japanese junior high students.In Japan, many students participate in club activities voluntarily as extracurricular activities.Because students who participate in club activities have a strong motive to be active, even if they were confronted with stress in club activity, they can weaken the impact of this stress 21).They are also likely to have many opportunities to receive teammates' support that may reduce depressive symptoms.Actually, our findings showed that participants in club activities reported less depressive symptoms than non-participants.Additionally, Newcomb et al. 40) noticed that each stress in different domains was related with different various symptoms.Thus, it is possible that stress in club activity domain may be associated with various symptoms except for depressive symptoms.
Among other psychosocial variables, self-esteem was the most strongly associated with depressive symptoms both in males and females.Students who had high self-esteem were more likely to score low on depressive symptoms.These findings support other in the depression literature 6. 11.41) Since a negative self-schema has been recognized to relate to depression 41), it is not surprising that low self-esteem is a strong correlate of depression in adolescents.The findings from Western studies also revealed that self-esteem in this age group functioned as an antecedent factor as well as a strong correlate of depressive symptoms in adolescents 4.11).It will be important to investigate whether low self-esteem may be a prospective predictor of future depression in Japanese adolescents.In either case, these findings suggest that promoting high self-esteem may be one of the effective reduction and prevention programs for adolescent depression.
Locus of control was also associated with depressive symptoms in the two genders.Students who were oriented toward external locus of control were more likely to report depressive symptoms.This finding was consistent with results of previous studies 0. It seems that external locus of control may produce feelings of helplessness characterizing depressive phenomena because they tend to perceive outcomes as beyond their control.This finding supports the learned-helplessness model in which individuals become helpless and depressed when they perceive events as uncontrollable 43).
Health practices and social support were also negatively related to depressive symptoms in the two genders.Students who had more positive health practices and perceived more social support were less likely to report depressive symptoms.These findings were consistent with the findings from previous studies 12. 23.44.0.Brown and Lawton 46) found that physical activity had a buffering effect on depressive symptoms.It is also considered that positive health practices such as physical activity may have the similar effects.
Our finding of the association of gender with depressive symptoms indicated that, at the univariate level, female students had more depressive symptoms than male students did.However, after controlling for the psychosocial variables differed by gender, gender could not account for a significant proportion of the variance in depressive symptoms.This finding was consistent with the findings of Allgood-Merten et al. 11) that the association between gender and the CES-D score was reduced to a non-significant level after controlling for differences between girls and boys in the psychosocial variables.They suggested that if female adolescents had equal levels of the psychosocial variables as did their male counterparts, they would not report so many depressive symptoms.Petersen et al. m also found that adolescent girls experienced more challenging and stressful circumstances than adolescent boys, and these differences accounted for gender differences in depressive symptoms.In this study, females scored in a more pathological direction than males on the psychosocial variables such as life stresses, self-esteem, and health practices.Particularly, females perceived interpersonal stresses with friends, family, and teachers as more challenging and stressful.Therefore, it is considered that the, relationship between gender and depressive symptoms in Japanese adolescents may be explained by the females' elevation on these psychosocial variables.
One limitation of this study was that we assessed depressive phenomena by a self-report symptom scale rather than diagnostic criteria based on a structured interview.It has been pointed out that such a scale is not preferable as a screening instrument because of many false positives on the self-reported measure 17).Accordingly, we focused on the self-reported depressive symptoms without classification and diagnosis as a case.Another limitation of this study was on our adolescent sample.This study was carried out exclusively on public junior high school students in Okinawa Prefecture.Therefore, the generalizability of the present findings for adolescents in Japan as a whole may be limited.A third limitation involved the cross-sectional study design.Clearly, our data cannot provide any information about causal linkages between psychosocial variables and depressive symptoms.As a reciprocal causation among these concepts is also considered, the results of this study must be interpreted cautiously.To address this issue, it will be necessary to conduct prospective study design.In summary, depressive symptoms in Japanese junior high students in Okinawa were positively associated with life stresses and negatively associated with positive health practices, more social support, high self-esteem, and internal locus of control.Among life stresses, friends' stress was the most strongly associated with depressive symptoms.On the other hand, of all other psychosocial variables, self-esteem was the most strongly related to depressive symptoms.In addition, female students were likely to report more depressive symptoms, life stresses, and low self-esteem and poor health practices.After controlling for the psychosocial variables differed by gender, gender differences in depressive symptoms were eliminated.
The results of this study indicate that many psychosocial variables shown in Western studies were also associated with depressive symptoms among junior high school students in Okinawa, Japan and gender differences in depressive symptoms could be explained by the females' elevation on those psychosocial variables.

INTRODUCTION
Many epidemiological studies of depression on adults have been conducted in the West.These previous studies have consistently indicated the female preponderance in depression u, and have also identified the many psychosocial characteristics as potential risk factors for depression a.In recent years, some studies focused on adolescents have shown that most psychosocial variables associated with adult depression are also related to adolescent depression.Hops et al. 3), for example, showed that self-reported depressive symptoms in high school students were associated with a variety of psychosocial variables, including other psychopathology, stressful life events, cognitive style, low self-esteem, low social support, and impaired coping skills.Lewinsohn et al. ') also indicated that similar psychosocial variables were related to depression met criteria for a diagnosis based on clinical interview with adolescents.In Asian countries, Stewart et al. 5) found that the psychosocial variables identified as risk factors from Western studies showed similar influences on depressive symptoms in Hong Kong Chinese adolescents.However, because few epidemiological studies of adolescent depression and its correlates have been conducted in Japan, little is known about whether the psychosocial factors shown in previous studies of Western countries are associated with depressive symptoms in Japanese adolescents as well. resources

Table 1 .
Mean and S.D. of the CES-D scores by sociodemographic variables.

Table 2 .
Mean and S.D. for the two genders on the psychosocial variables.

Table 3 .
Hierarchical multiple regression models of the CES-D scores by gender.
and friends, health practices, self-esteem, locus of control, and social support) were simultaneously entered in step 2.Results from the hierarchical multiple regression analyses are presented in Table3.The findings for the regression predicting the depressive symptoms in males revealed that sociodemographic variables explained 2% of the variance in depressive symptoms.On step 2, the psychosocial variables social support were related to lower level of depressive symptoms.Among sociodemographic control variables, lower graders were likely to report more depressive