Well-being researchers have developed an impressive and extensive literature on many aspects of the level and trajectory of subjective well-being during adulthood, but the transitional period from late adolescence to early adulthood has received less attention. In this article, we examine patterns of association among a number of different factors that may contribute to differences in the level of subjective well-being among young people in transition to adulthood. This study uses a nationally representative sample of young people in Germany from the German Socio-Economic Panel. In the German Socio-Economic Panel, young people aged 17 are viewed as being on the cusp between adolescence and young adulthood. For this reason, when young people in each household attain the age of 17, they are formally treated as adults and are asked to respond to all the questions that are posed to adult household members. Recognition of the transitional period between adolescence and adulthood is recognized, however, as these young people are also asked to respond to a detailed and extensive series of questions in a separate youth questionnaire. This questionnaire provides researchers the opportunity to assess a wide range of characteristics for each young person, including personality traits and future expectations, which may continue to have a long-term influence on events and experiences throughout adulthood.

This current study combines data from this youth questionnaire with data from the adult questionnaire for the young person, the household questionnaire, and the parents’ questionnaires. Specifically, the study examines how demographic and socio-economic characteristics of the young persons and their parents, personality traits of the young persons, quality and quantity of relationships, the parent’s level of life satisfaction, and other measures of satisfaction for the young person are related to the initial assessment of life satisfaction by the individual at the critical point of transition from adolescence to adulthood. To the extent that such factors are associated with variations in self-reported levels of life satisfaction, the potential exists to formulate practices and policies that can have both short and long-term consequences for influencing individual subjective well-being. By examining the nature and patterns of these associations, our long-term goal is to identify potential areas where strategies may be developed—at the individual, micro, and macro level of policy—to influence the determinants of the level of subjective well-being.

1 Past Research on Predictors of Life Satisfaction Among Adolescents and Young Adults

A substantial body of literature exists that examines psychological correlates of life satisfaction among adolescents. (See Huebner 2004 for an overview of this literature.) Based primarily upon small and/or non-random samples from very limited geographical areas, past research has found predictors of life satisfaction among adolescents to be similar to predictors of life satisfaction for adults. Among adults, researchers have found that demographic factors, such as marital status, income, social class, ethnicity, and unemployment; personality factors, such as personal control, extraversion, and neuroticism; and social contexts, such as religious involvement, social involvement and uses of leisure, are consistent but relatively weak predictors of life satisfaction. See, for example, Argyle (1999), Clark and Oswald (2002), Diener and Lucas (1999), Easterlin (2002), Oswald 2002), and Warr (1999) for reviews of this literature.Footnote 1

1.1 Personal Relationships and Family Composition

Links between close relationships and quality of life are well-established (Myers 1999). In addition, one of the most robust findings in subjective well-being research concerns the relationship between family composition and subjective well-being (Lucas et al. 2003, 2004; Fujita and Diener 2005). Furthermore, considerable empirical evidence exists linking adult outcomes, including personality level outcomes, to childhood and adolescent environments and life cycle events experienced before adulthood. Parental divorce, for example, is associated with a range of negative adult outcomes, including life satisfaction. On average, results from studies in the United States have shown that adults who experienced a parental divorce have lower psychological well-being, more behavioral problems, less education, lower job status, lower standard of living, lower marital satisfaction, heightened risk of divorce, heightened risk of being a single parent, poorer physical health—all of these outcomes are also correlates of subjective well-being. When the divorce is also associated with violence, the effects are even more pronounced (See Amato and Cheadle 2005 Amato and Sobolewski 2001, Amato 2000, 1999; McNeal and Amato 1998 for reviews of this literature).

The effects of family composition have been shown to be more dramatic when some periods of time in childhood/adolescence were spent in households where neither parent was present. Zullig et al. (2005) used a statewide sample of 4,758 public high school students in South Carolina in the United States to investigate the associations among family structure, demographics and adolescent perceived life satisfaction. The authors found that adolescents who lived with other relatives, non-relatives, or guardians appeared to be at risk for experiencing a greater likelihood of dissatisfaction with life. The extent to which other types of living arrangements were associated with low, middle, and high levels of life satisfaction varied by race, age and gender.

Many studies of the correlates of adolescent well-being have examined the role played by the quality of adolescent-parent relationships and/or the quality and quantity of relationships with peers. Nickerson and Nagle (2004) examined the relationships between life satisfaction and parent and peer attachments among 303 fourth to sixth graders in one suburban school district in the southeastern United States. They found that relationship variables, such as levels of parental trust and alienation and levels of peer alienation and delinquency, accounted for 51% of the variance in reported levels of total life satisfaction. Similar findings were reported by Gilman and Huebner (2006), who investigated the characteristics of adolescents with different levels of life satisfaction among 485 adolescents in two schools in the southeastern United States. Gilman and Huebner divided the adolescents into three groups of life satisfaction, ranging from low to high, and examined differences in school outcomes; interpersonal variables; and intrapersonal variables. More positive levels of both parent and interpersonal relations were observed for adolescents in the high life satisfaction group. Cheng and Furnham (2003) also found that self-esteem and positive relationships with parents were positively related to happiness within a sample of 234 students (mean age 18.23, range 15–35) attending schools and colleges in the London area. They observed similar relationships among 90 senior students (mean age 17.23, range 16–18) in three schools in the UK (Cheng and Furnham 2002).

Based on a large, convenience sample of 2,722 British adolescents aged 14–18, Flouri and Buchanan (2003) examined the relative strength of the relationships between psychological well-being and the adolescents’ involvement with their parents, self-efficacy, self-reported levels of depression, parental conflict, socio-economic status, and family structure. Levels of maternal and paternal involvement were both positively related to well-being, but the strength of the relationship was stronger for paternal than for maternal involvement.

1.2 Psychological Factors and Attitudes

The second major set of correlates of adolescent and young adult well-being examined in past research comprises psychological factors and attitudes, including factors and attitudes such as optimism, neuroticism, self-esteem, self-efficacy, attributional style, extraversion, introversion and perfectionism. Observed associations between overall life satisfaction and other psychological factors have been well-established in the literature on subjective well being for both adolescents and adults. According to Diener and Lucas (1999), research has shown personality traits and adult subjective well-being to be strongly and robustly correlated. Diener and Lucas conclude their review of the literature with the observation that “SWB ratings reflect a stable and consistent phenomenon that is theoretically and empirically related to personality constructs” (1999: 226). However, they also note that the direction and strength of these relationships and the pathways from personality traits to subjective well-being are not unequivocally known.

In a review of the empirical and theoretical literature linking personal control and adult well-being, Peterson (1999) found a consistent correlation between personal control and well-being. He emphasized, however, that although some theorists have equated personal control and well-being, the empirical evidence does not support such an equivalency. Instead he argued that both theoretically and empirically it is more useful and accurate to view personal control as an enabling condition for well-being.

Gilman and Ashby (2003) conducted a small study of the link between perfectionism and life satisfaction among a sample of 132 middle school children enrolled in three rural middle schools in the southeastern United States. They found that children who set high standards for themselves also reported higher levels of global satisfaction, but that this relationship was mediated by the children’s ability or inability to achieve these standards. Gilman and Huebner (2006) also found strong positive associations between high levels of life satisfaction and indices of behavioral and psychological functioning and adjustment. Flouri and Buchanan (2003) found depression was negatively related while self-efficacy was positively related to well-being.

Fogle et al. (2002) used a sample of 160 middle school students in a mid-size city in the southeastern United States to investigate the relationship between life satisfaction and a range of personality traits. In their study, social self-efficacy was directly and positively related to life satisfaction, extraversion was indirectly and positively related through the pathway of self-efficacy, and neuroticism was negatively related through both direct and indirect pathways. They also examined teacher-rated social competence and found it to be unrelated to life satisfaction in their sample. Rigby and Huebner (2005) sought to examine the extent to which students’ attributional style (adaptive or nonadaptive) and their levels of emotional stability and extraversion predicted their subjective life satisfaction. Based on a sample of 212 high school students in Pennsylvania, they estimated a structural model where emotional stability and extraversion were hypothesized to affect life satisfaction directly and indirectly through the students’ attributional style. Their results provided support for the hypothesis that emotional stability is associated with higher levels of life satisfaction, but extraversion was not associated with life satisfaction for students in their sample.

Based on a convenience sample of 142 adolescents in an urban middle school in the United States, Yarcheski et al. (2001) tested whether self-esteem and hopefulness mediated the relationship between social support and subjective well-being. Their results indicated that social support and subjective well-being were strongly correlated, but the strength of the direct relationship diminished when self-esteem and hopefulness were also included in the estimation. They concluded that social support was both a direct and indirect predictor of life satisfaction through the pathway of hopefulness and self-esteem—both of which were directly and positively associated with life satisfaction.

Ben-Zur (2003) examined the relationships between subjective well-being, mastery, and optimism with a sample of 121 Jewish adolescents and their parents in Israel. The study included a measure of parent-adolescent relationships as well as assessments of subjective well-being, mastery and optimism for both the adolescents and their parents. Demographic variables were also included. Positive associations were observed between parent-adolescent relationships, levels of mastery and optimism reported by the adolescents and adolescent subjective well-being.

1.3 Satisfaction with Other Domains

Research on subjective well-being for adults consistently finds considerable correlation among different domains of subjective well-being, for example, between life satisfaction and job satisfaction (Warr 1999). In one of the only studies to directly assess cross-cultural differences in the level of life satisfaction among adolescents, Park and Huebner (2005) compared correlations among different domains of life satisfaction for a sample of 472 Korean and 542 U.S. adolescents. The authors found that satisfaction with school was strongly correlated with global life satisfaction for Korean adolescents but not for U.S. students. In contrast, the domain of self-satisfaction was a far stronger predictor of global satisfaction for U.S. than for Korean adolescents.

1.4 Income and Economic Hardship

Based on decades of research, several stylized facts exist concerning the relationship between income, economic hardship, and subjective well-being. In cross-sectional analyses both at the country and interpersonal level, higher income is associated with higher subjective well-being. The measured effects are strongest at low and very low levels of income, where increases in income that are used to provide for basic human needs lead to more substantial increases in subjective well-being than increases in income for those who are above poverty-level income. Across time, however, rising real incomes show a much weaker association with changes in subjective well-being, leading researchers to formulate theories concerning the importance of relative versus absolute income. See Easterlin (2002), Frey and Stutzer (2002), and Oswald (2002) for reviews of this literature. Considerable research also documents long-term consequences of poverty and social exclusion during childhood and adolescence (Duncan and Brooks-Gunn 1997; Danziger and Haveman 2002).

1.5 Employment and Unemployment

Separate from any direct effects on income and economic well-being, unemployment is strongly correlated with decreases in adult subjective well-being (Clark and Oswald 2002; Di Tella et al. 2002; Lucas et al. 2004). For adolescents, recent literature also suggests that maternal employment is associated with negative outcomes for adolescents under certain conditions (Trzcinski et al. 2006; Brooks et al. 2001).

1.6 Life Activities

A major theme within subjective well-being research that has received considerable empirical and theoretical attention concerns the uses of leisure and the level of satisfaction attached to different kinds of participation in life tasks. This literature spans philosophical dimensions, such as Scitovsky’s The Joyless Economy (1992) and Amartya Sen’s (2002) extensive writings on capability and well-being to examinations of the association of hours of television viewing with subjective well-being (see Cantor and Sanderson 1999 for a review of how leisure and participation in life tasks affect well-being).

Gilman (2001) investigated the association between social interest and involvement and a range of life satisfaction domains among 321 adolescents from 2 urban high schools in the southeastern United States. Strong positive associations were observed between the domains of life satisfaction and the adolescents’ social interest and between life satisfaction and actual participation in extracurricular activities, however, the relationship between social interest and actual participation in extracurricular activities was weak. They also found that different domains of satisfaction, such as satisfaction with family and friends and overall level of satisfaction, were mutually reinforcing.

1.7 Other Studies

Brown and Kasser (2005) conducted a small study of 206 students attending two middle and high schools in the United States where they examined the relationships between subjective well-being, materialism and generosity, and ecologically responsible behavior. They found positive associations between subjective well-being and ecologically responsible behavior, but that materialism and generosity were significant mediators of the relationship between subjective well-being and ecologically responsible behavior. They concluded that their results supported past research linking higher levels of subjective well-being to intrinsic value orientations.

Finally, Valois et al. (2002) used the same statewide sample of public high school students in South Carolina as Zullig et al. (2005) to examine the association between life satisfaction and sexual risk-taking behaviors among adolescents. Risk-taking behaviors included having been “beaten up” by a date, having “beat up” a date, forcing someone to have sex, age of first intercourse less than age 13, been pregnant/caused a pregnancy. For adolescents in South Carolina who engaged in (or who were forced to engage in) such behaviors, the negative associations with life satisfaction were substantial—ranging from an increase in the odds of reporting dissatisfaction from 1.17 to 3.08 for white females; from 1.32 to 5.01 for white males; and from 1.53 to 3.53 for black males. The authors did not report the incidence of these behaviors—and the relevance of such violence, early sexual activity, and teenage pregnancies for other Western industrialized countries may be somewhat limited.

2 Conceptual Model

Despite the existence of an extensive body of empirical literature on life satisfaction, a grand theory that accounts for the many well-known stylized findings observed in cross-sectional and longitudinal studies has yet to be developed. Nor does the literature provide a well-developed theory that attempts to account for the development of life satisfaction throughout childhood and adolescence into adulthood. As Huebner (2004) has noted in his review of the literature on the measurement of life satisfaction in childhood and adolescence, few theory-based studies are found among this work. Hence our own conceptual model is primarily guided by past empirical findings that have been observed for adolescents and adults and includes constructs that pertain to: (1) the role of relationships and family composition; (2) observed associations between life satisfaction and other psychological constructs and attitudes; (3) associations between overall life satisfaction and other domains of satisfaction; (4) the role of income and economic hardship, (5) engagement and participation in life activities; and (6) demographic factors.

2.1 Personal Relationships and Family Composition

Our conceptual model posits that the life satisfaction of youth at the point of transition to adulthood will be associated with the quality and quantity of interpersonal relationships in the young person’s life. The richness of the GSOEP data allows us to test a number of different dimensions pertaining to relationships and to examine the strength and direction of these relationships. Hence we included a number of measures to assess the level and quality of an adolescent’s relationships. As noted above, empirical evidence linking adult outcomes to family composition during childhood and adolescence is also strong and consistent. We include variables that allow us to test the following hypotheses:

  1. (1)

    Increases in the number of relationships deemed important by the youth will be associated with higher levels of reported well being;

  2. (2)

    higher quality of relationships between the young person and his/her mother and father will be associated with higher levels of self-reported subjective well being; and

  3. (3)

    the presence of a father and a childhood spent with both parents will be associated higher levels of well being at age 17.

We also hypothesize that positive associations will exist between the young person’s life satisfaction and the parental life satisfaction.

2.2 Psychological Factors and Attitudes

In our model, we include a measure of personal control. Consistent with past research, we hypothesize that young people who report that they have a greater level of control over life circumstances will report higher levels of life satisfaction. We include a second construct that deals with the young person’s expectations concerning the future. Here we hypothesize that individuals who express more positive expectations about the future school and job outcomes will also report higher levels of well-being. The final construct in a model concerns self-agency. We hypothesize that young people who view personal effort as the key to success will report higher levels of subjective well-being.

2.3 Satisfaction with Other Domains

For young people, satisfaction with grades represents an important dimension of satisfaction because school is a more central activity for most adolescents than work. Thus we hypothesize that young persons who report higher levels of satisfaction with their grades will also report higher levels of subjective life satisfaction.

2.4 Income and Economic Hardship

In our model, we hypothesize that the higher levels of household income will be associated with higher levels of life satisfaction. We also hypothesize the young adults who live in households experiencing economic hardship will report lower levels of satisfaction than youth in other households. The conceptual distinction between income and economic hardship is maintained because of the large body of research that indicates that social exclusion and economic hardship affect well-being differently from absolute income levels.

2.5 Parental Employment and Unemployment

We hypothesize that young persons in households where parents are unemployed will report lower levels of subjective life satisfaction. Although few studies have directly investigated that role of parental unemployment on the subjective well-being of young persons in the same household, the strong associations observed for subjective well-being and own unemployment may result in negative associations for the children as well as for the parent who is directly experiencing the unemployment. Given that maternal employment has been shown to be associated with a range of outcomes for older adolescents, we also hypothesize that full-time maternal employment will be associated with lower levels of life satisfaction of young persons. The rationale is that the stress associated with full-time employment for women in a society where women are still expected to assume the major responsibility for care giving and for home production could result in lower levels of subjective life satisfaction for young persons still residing in the same household. In the case of Germany, full-time employment of the mother can in some cases also serve as a proxy for a lower level of household resources.

2.6 Engagement and Participation in Life Activities

Young persons who are more active physically and intellectually are hypothesized to report higher levels of subjective well being than youth who do not spend as much time in these kinds of activities. In contrast, for young people who report that they frequently spend time “do nothing,” it is hypothesized that subjective well-being will be lower than for young persons for whom “doing nothing” is not a part of their daily routine.

2.7 Demographic Factors and Educational Aspirations

We hypothesize that immigrants and young persons living in the states formerly in the DDR will report lower levels of life satisfaction than young persons who are German citizens living in the western part of Germany. This hypothesis is based on the lower level of social and economic opportunities confronting immigrant youth and youth who live in the states formerly in the DDR compared with the level of opportunities available to youth who are citizens living in western Germany.

Finally, educational aspirations are also included in our model. For adults, subjective life satisfaction tends to increase with educational level. For young people who have not yet completed their education, the hypothesized direction of any associations is more difficult to predict. In Germany, a person aged 17 with aspirations to complete a university degree will not yet have completed the requirements necessary to enter university and may be under considerable stress or living under a high degree of uncertainty concerning his/her future. A person with lower educational aspirations, on the other hand, may already be enjoying a degree of independence and freedom not yet available to those with higher aspirations. Those with higher educational expectations, however, are more likely to have more positive expectations about future income and life opportunities. Hence at this stage of the life cycle, educational aspirations may be associated with a number of counterbalancing effects.

3 Methodology

3.1 Sample

The German SOEP is a wide-ranging representative longitudinal study of private households. It provides information on all household members, consisting of Germans living in the old and new German states, foreigners, and recent immigrants to Germany. The panel was started in 1984. In 2004, there were nearly 12,000 households and more than 22,000 persons sampled. Some of the many topics include household composition, occupational biographies, employment, earnings, health, and satisfaction indicators. Since the year 2000, themes related specifically to children and teenagers were integrated into the SOEP by a youth questionnaire. It contains, in particular, retrospective type questions on school career, music, education and sport, as well as on the current life situation. In addition numerous prospective questions inquire about educational and further training plans and about expectations on future career and family. This data on teenagers can be connected to the life history of the parents, because they (as well as all other household members aged 17 and older) are also surveyed in SOEP. This dataset allows a variety of analyses, including intergenerational studies. In 2000 a pre-test was carried out with 232 teenagers who lived in SOEP-households. An expanded and revised questionnaire was completed by 618 teenagers between the ages of 17 and 19 in 2001.Footnote 2 From the year 2002 on each 17-year-old who is personally surveyed for the first time in a SOEP-household will begin his/her “survey career” by completing the youth questionnaire as well as the standard adult questionnaire. By 2004, the data contained 1,202 people. In our analyses, we use all 17 year olds who responded to youth questionnaire during the period 2001–2004.

3.2 Variable Definitions

In this section, we provide specific descriptions of the variables that correspond to the constructs in our conceptual model. Table 1 presents descriptive statistics for each of the variables in our conceptual model; Table 2 presents zero order correlations for the variables that are included in the final estimated equation for subjective well-being.

Table 1 Means, standard deviation, or percentages of variables in model
Full size table
Table 2 Pearson correlation matrix, variables in subjective life satisfaction of 17 year olds, Part 1 and Part 2
Full size table

3.2.1 Subjective Life Satisfaction

In each interviewing year of the SOEP, every adult household member is asked to rank their overall life satisfaction using an 11-point scale. The level of life satisfaction is based on the question: “Finally, we would like to ask about your overall level of life satisfaction. Please answer again according to the following scale, “0” means completely and totally dissatisfied; “10” means completely and totally satisfied. How satisfied are you at the present time, all things considered, with your life?” We used this measure for subjective life satisfaction for the adolescent and for his/her mother.

3.2.2 Personal Relationships

The first measure relates to the number of relationships that the adolescent judges to be important or very important. The adolescent was given a list of nine individuals (mother, father, girlfriend, boyfriend, etc.) who could potentially be “especially important in your life at this time”. We constructed this variable by counting the number of individuals for whom the adolescent checked important or very important; the variable range was from “0” to “9” important to very important persons.

We also included measures to assess the quality of the adolescents’ relationship with their mothers and fathers. For the relationship with mother, we used an index constructed from 7 items with 5 potential responses ranging from very often to never. Specific items included How often does your mother talk to you about things you do or experience; ask you for your opinion before she decides something that affects you, etc. Because of missing data for absent fathers, we used a different question for relationship with father that assessed level of conflict. We constructed a set of dummy variables based on the question How often do you argue or fight with the following people (your father)? A dummy variable was included with the following categories (1) Fight very often or often, (2) Fight sometimes (3) No such person in my life, with the excluded category of fight seldom or never. This specification allowed us to investigate the effects of quality of relationship with father in comparison with no relationship with father. Finally, we included a set of dummy variables that indicated whether the adolescent fought often or very often about grades with both parents, mother only, or father only; the excluded category was did not fight often or very often with either parent about grades.

3.2.3 Family Composition

The German SOEP provides information on the adolescent’s living situation throughout the entire period of childhood and adolescence. Based on this information, we constructed a set of dummy variables to measure family composition during the first 15 years of the young person’s life. The first group included a set of four mutually exclusive dummy variables—living with both parents entire life, living some time with stepparent, living some time in non-traditional setting (other relatives, foster parents, in a home), excluded variable: other. The second variable indicated whether the father or the mother of the adolescent was deceased and the final one indicated whether the adolescent had contact with his or her father.

3.2.4 Personality Traits and Attitudes

The adolescent’s assessment of personal control was based on a set of four responses to a 6 item Likert scale designed to ascertain which factors the adolescents considered the most important for achieving success and improving one’s social situation in Germany (Cronbach’s alpha = .63). Examples of items include I have control over my own destiny, Success or failure in life is largely a question of fate or luck, and I have often had the experience that others have control over my life.

We included a measure of the adolescent’s assessment of future success on number of different domains related to school and work. This measure was based on index of 4 items for which the adolescent was asked to assess the probability that that they would receive a training or university slot in their preferred field, successfully finish training or university studies, find a job in their field, be successful and “get ahead.” Eleven choices were provided, ranging from “0” to “100” percent probability. This variable can be interpreted as a measure of self-efficacy.

Our third index measured what the adolescent views as keys to success and social mobility, which can be viewed as a measure of self-agency. Factor analyses indicated that this construct had separate positive and negative components. The first component consisted of five items assessing the importance of effort, such as success through working hard, success through specialized training, success through school grades, results, and success through initiative (Cronbach’s alpha = .62). The negative component contained six items that measured the importance of exploitation, ruthlessness and personal connections and wealth, such as success through exploiting others, success through family background, success through being tough, ruthless (Cronbach’s alpha = .76). For both components, each item was ranked on a four point scale ranging from agree completely to do not agree at all. The second component was used in the analyses presented below.

3.2.5 Life Participation

We included three sets of dummy variables to measure how the adolescent spends time when not working or in school: (1) the adolescent engages in sporting activities at least once a week; (2) the adolescent spends some time reading each day, and (3) the adolescent spends some day each day doing nothing. Each of these variables was entered as a separate dummy variable.

3.2.6 Satisfaction with Grades

We used three components to construct the overall index of satisfaction with grades (school grades in general, grades in German, and grades in mathematics), for each item responses ranged from 0 (unsatisfied) to 11 (satisfied).

3.2.7 Income and Economic Hardship

We used three types of variables to model income and economic hardship. The first was household income specified as the natural log of the yearly household equivalent income/1000.Footnote 3 The second two sets measured economic hardship with dummy variables—one set for the condition of the apartment (apartment in poor condition, apartment needs some repair, excluded category other) and the other for financial worries (major worries, some worries, excluded category other). The natural log of income and the condition of the apartment were household level variables; the assessment of financial worries was reported by the adolescent’s mother.

3.2.8 Maternal Employment and Unemployment

In order to test whether employment and unemployment are associated with subjective life satisfaction, we included three employment variables: number of months that the mother was unemployed in the previous year, a dummy variable indicating that the mother was employed full-time for 12 months in the previous year, and a dummy variable indicating that the mother was employed part-time for 12 months in the previous year.Footnote 4

3.2.9 Demographic Variables

We included a set of dummy variables that indicates whether the adolescent is a German citizen living in western German, a German citizen living in eastern German (excluded category), or a non-German citizen. We also included a dummy variable for gender.

3.2.10 Education Variables

Finally, we also included a set of variables to indicate the adolescent’s school aspirations. This variable was constructed based on the highest degree received, if the adolescent indicated that he or she had no additional aspirations for education and on the highest degree expected, in cases where the adolescent intended to pursue a higher degree. Hence this variable is a mixed measure of educational achievement and aspirations. Because our research only focuses on the adolescent’s educational achievement and aspirations at age 17, it is impossible for us to separate out the effects of educational aspirations, educational achievement, and any effects associated with success or failure in achieving one’s educational aspirations.

3.3 Estimation Techniques

In the initial stages of our estimation, we used ordinary least squares estimation to refine our model. We proceeded by including all the variables listed in our conceptual model and eliminated variables based on low levels of achieved significance. We also used stepwise regression techniques to confirm which variables should be eliminated from our model. We next used full information likelihood estimation with EQS 6.1 to estimate a set of simultaneous equations, where three of the predictor variables in the equation for the young person’s subjective well-being were treated as endogenous—subjective life satisfaction of the mother; personal control; and overall satisfaction with grades.

The rationale for this technique is that an ordinary least squares (OLS) estimation technique is limited in the case when the error term of a dependent variable is expected to be correlated with measurement error in a predictor variable. For each of these three predictor variables, the correlations among these explanatory variables and our primary dependent variable could occur at least in part because the error term associated with the subjective well-being of the youth could be associated with measurement error in these predictor variables. In such cases, the estimated coefficients will be neither consistent nor unbiased. As Table 2 indicates, the zero order correlations between subjective life satisfaction of the youth and life satisfaction of the mother; personal control; and satisfaction with grades are relatively high: .222, .225, and .229, respectively. In addition, it is highly likely that the error term of the youth’s subjective satisfaction may indeed be correlated with measurement error in the mother’s life satisfaction and with the measurement error in the other two constructs as well. As long as it is possible to find some variables that are highly correlated with these predictor variables, but uncorrelated with or only weakly correlated with, the dependent variable under investigation, one can use full information maximum likelihood simultaneous equation estimation to obtain consistent and unbiased estimators (Johnston and Dinardo 1997; Kmenta 1997; Gujarati 2003). Table 3 presents the zero order correlations between subjective life satisfaction of the youth, the three predictor variables, and the variables that were used to identify the three equations associated with each of these predictor variables. For each of the equations for these predictor variables, we were able to find instruments (that is, dependent variables) where the correlations with the predictor variables were relatively high, but relatively low with the dependent variable, subjective well-being of the youth. For example, the magnitude of the correlation between personal control and self-agency (.317) is four times greater than the correlation between self-agency and youth subjective well-being (.074); likewise the correlation between overall level of satisfaction with grades and grade in German (−.322) is nearly four times greater than the corresponding zero-order correlation for youth subjective well-being (−.081).

Table 3 Pearson correlation coefficient for additional variables in three equations in Table 5
Full size table

4 Discussion of Results

Table 4 presents the ordinary least squares and maximum likelihood estimates for subjective life satisfaction for the adolescents. Overall we found that the direction and magnitude of the estimated coefficients were robust across the two specifications. Table 5 presents the results for the estimation of the three equations for the predictor variables that are treated as endogenous in our four system model of equations.

Table 4 Ordinary least squares and maximum likelihood simultaneous equation estimation of subjective life satisfaction of adolescent
Full size table
Table 5 Maximum likelihood estimates of endogenous variables using endogenous predicted variables for subjective life satisfaction of mother, personal control, and satisfaction with grades
Full size table

4.1 Personal Relationships

Our results for adolescents are consistent with past research on the importance of relationships, particularly high quality relationships, in predicting levels of subjective well-being for adults. The variable, Subjective life satisfaction of mother, achieved a high level of statistical significance and was positively related to the life satisfaction of the adolescent (Table 4). Based on the estimated coefficient of .124 in the simultaneous equation model, the difference in predicted life satisfaction for an adolescent with a mother whose life satisfaction was one standard deviation above the mean versus one standard deviation beneath the mean was approximately .46 points.Footnote 5

Adolescents who reported a higher quality of relationship with their mothers reported higher levels of subjective well-being than did adolescents with a lower quality of relationship with their mothers. Quantity of relationships was also positively associated with subjective life satisfaction. In the set of dummy variables measuring conflict with father, the only category that was marginally significant in reference to the excluded category (does not argue with father) was the category “fights a lot with father.” Although this association was only marginally significant and negative, it does suggest that for the young person more contentious relationships with fathers are more problematic than a complete lack of contact.

Although none of the individual coefficients reported in Table 4 are large in magnitude, the cumulative overall difference in predicted level of subjective well-being for a young person with strong and positive personal relationships compared with a youth with weak ones is substantial. For example, if we calculate the predicted well being for a youth with a score of “4” on relationship with mother; with 8 important relationships, and who does not argue with his/her father and compare this value for a youth with a score of “2” on relationship with mother; only 2 important relationships, and who reports frequent arguments with his/her father, the difference in predicated well-being is over one full point. When added to the difference calculated above for associations between mother’s and young person’s well being, the predicted difference is nearly 1.5 points on the 11 point scale, that is, approximately one standard deviation from the mean subjective well being.

4.2 Family Composition

Despite strong associations that have been found in past research between adolescent outcomes and family composition, we found no statistically significant association between family composition and subjective well-being for the young person. In our preliminary regressions, we included a full set of measures to account for different types of family composition, including some period of residence in a group home. None of these variables achieved even a marginal level of statistical significance. Nor did we find any effect associated with the category “no contact with father.” In the case of Germany, a failure to find any statistically significant associations between family composition and subjective well being may be connected to the relatively small percentage of adolescents who live in non-traditional arrangements. As Table 1 indicates, more than 75% of the 17 year olds in our sample spent their entire youth with both parents; only 7% reported they have no contact with father; and less than 2% spent any time in households where neither parent was present. Germany also has relatively more stringent child support regulations and a far stronger system of family supports than the United States, where a large percentage of the research on the effects of family composition has been undertaken. Hence negative social and economic consequences for young persons living in households where both parents are not present may be less severe in a country with a well-developed system of social supports in contrast with a country such as the United States, where single parenthood is often associated with severe economic hardship and where non-traditional arrangements, including foster homes and residence with non-relatives, are far more common.

4.3 Personality Traits and Attitudes

Consistent with past research, adolescents who scored higher on the domain of personal control over life circumstances had higher predicted levels of subjective well-being than those who reported less control over their own destiny. The size of this association was modest—a predicted difference of .38 points in the simultaneous equation model for an adolescent who scored one standard deviation beneath the mean compared with an adolescent who scored one standard deviation above the mean. The results presented in Table 5 show that personal control was positively associated with adolescent engagement in sport and daily reading, with household composition, and with self-agency. Males also reported higher levels of personal control than females. Negative associations were observed for adolescents who reported that they did nothing daily and who lived in families where mothers report some or major worries about the family’s financial situation.

Among adolescents in our study, those with higher assessments of their probability of success in education and work had higher levels of subjective well-being and higher levels of satisfaction with grades, a result that is consistent with findings concerning the link between self-efficacy and subjective well-being for adults. In our preliminary models, we found no direct association between self-agency measured as the young person’s assessment of the keys to success and subjective well-being, but we did find a strong, positive association between self-agency and personal control (Table 5).

4.4 Satisfaction with Grades

The third and final endogenous predictor variable in the simultaneous equation model was the adolescent’s assessment of satisfaction with grades. Here the association was also positive and statistically significant, with the size of the predicted effect approximately the same as for mother’s subjective well-being. A two standard deviation difference in the reported level of satisfaction with grades was associated with a difference of .38 points in subjective well being in the simultaneous equation model. According to the results in Table 5, adolescents who were foreigners or immigrants reported higher satisfaction with grades than did West German adolescents, who in turn reported higher levels of satisfaction than their East German counterparts. A positive association also existed between actual grades and satisfaction with grades, with the absolute size of the coefficients considerably higher for grades in mathematics (.559) than in German (.469).Footnote 6 Although conflicts between the young person and his/her father were relatively unimportant predictors in the equation for subjective well-being, conflicts with mother and/or father over grades were strong and statistically significant predictors of an adolescent’s satisfaction with grades (Table 5).

4.5 Income and Economic Hardship

Surprisingly we did not find strong associations between adolescent subjective well-being and our measures of household income. In the subjective well-being equation for the adolescent, increases in income were marginally statistically associated with increases in well-being. However one measure of economic hardship, the mother’s report of major financial worries, was associated with decreases in the predicted level of well-being for the adolescent. None of the other variables measuring economic hardship reached even marginal levels of statistical significance in our preliminary analyses; hence these variables were not included in the final model.

Far stronger associations were observed between the full set of variables measuring income and economic hardship in the estimation of maternal well being. The variables measuring whether the apartment was in poor condition or in need of repairs were strongly and negatively associated with mother’s subjective well-being. In addition, household income was far more strongly associated with mother’s well-being than with adolescent’s well-being.

4.6 Life Participation

In contrast to the subjective well-being literature on adults, we did not find significant associations between our measures of life participation and adolescent subjective well-being in any of our preliminary estimations. Hence we excluded these variables in our final model as explanatory variables in the equation for subjective well being of the 17 year olds. These variables were, however, statistically significant predictors in our estimation of personal control over life circumstances—adolescents who reported that they were active in sport and who reported that they read daily reported higher levels of personal control, whereas those who reported they did nothing on a daily basis had lower reported levels of personal control. These results suggest that how adolescents spend their time may indirectly affect subjective well-being through influencing other important personality constructs, such as personal control, that in term may affect levels of subjective well-being.

4.7 Maternal Employment and Unemployment

As noted above, unemployment is a critical variable in predicting both level and changes in adult subjective well-being. For adolescents, maternal unemployment was strongly negatively associated with subjective well-being. Adolescents whose mothers worked the entire year also reported lower levels of subjective well-being—for full-time work, a decrease of −.283 points and for part-time work, a decrease of −.203 compared with adolescents whose mothers were not employed the entire year. The negative association of maternal unemployment and a 17 year old’s well being is not surprising. Potential explanations for negative associations with both year round employment are somewhat less evident. As noted above, however, discussions about the effects and appropriateness of maternal employment continue to be an issue of intense political and popular importance in Germany. In contrast to most other Western countries, most schools in Germany, including gymnasiums, only take place for half a day. Hence 17 year olds in Germany finish their school day by 12 or 1 p.m.; it is not uncommon for families (or at least mothers and their children) to eat the mid-day meal together. Such culture factors may potentially play a role in the relationship observed here between subjective well being of the 17 year olds and maternal employment.

4.8 Demographic and Education Variables

In the direct estimates of adolescent well-being, demographic variables only achieved marginal significance, with immigrants reporting higher levels of well-being than West German and East German adolescents—a finding that is counter to results observed for adults in general in Germany. Males also had marginally higher levels of subjective well-being than females. Stronger associations were observed, however, in the estimated equations for the endogenous variables. Our final set of variables assessed whether associations existed between the adolescent’s educational aspirations and subjective well-being. The results indicated that for 17 year olds those planning to attain an intermediate level degree or to complete vocational school had higher levels of subjective well-being than did youth with lower or higher educational expectations. The pressures facing young people who intend to pursue a university degree, but who have not yet completed all the necessary requirements may be a factor in the lower subjective life satisfaction reported by these youth compared with those with midlevel expectations.

5 Discussion and Conclusions

Our purpose in this study was to begin to expand the range of possibilities for understanding subjective well-being by examining links between experiences in the household and family of origin and the initial measurement of subjective well-being as adolescents make the transition to early adulthood. Although our methodology does not allow us to make any claims regarding causality, our results do suggest that links may exist between the formation of subjective well-being and experiences before adulthood. Our results also indicate that the same factors that are associated with subjective well-being throughout adulthood are associated with the level of subjective well-being at the point of transition from adolescence to adulthood. These associations seem to operate both directly and indirectly via the association of life satisfaction of the parents with the life satisfaction of the young person. For example, unemployment, which has emerged as one of the most important predictors of both level and change in adult subjective well-being, has both a direct association with adolescent well-being and an indirect one via its negative association with maternal life satisfaction.

In most cases, factors that predict adult well-being also predict the level of well-being reported by the adolescents in our study. We found consistency existed across different domains of satisfaction, specifically satisfaction with life and satisfaction with grades. Based on a wide range of measures, we found a strong pattern of association between the subjective well-being of the adolescents and variables that measured different dimensions of the quality and quantity of interpersonal relationships, including relationships with parents.

One finding that was somewhat surprising centered on the role of full and part-time maternal employment. Although the finding regarding unemployment was in the expected direction, the negative and relatively strong direct association between full-time adolescent well-being may possibly reflect internalization by adolescents of the norms of the German model regarding the balance of work and family for women with children. We strongly caution, however, against drawing any strong conclusions regarding why these findings were observed and whether they would be dissipated or eliminated if we were to control for economic hardship, patterns of parental labor market behavior, and parental life satisfaction during early and middle childhood.

6 Limitations

In terms of our long-term goal, which is to learn how experiences during early childhood, middle childhood and adolescence affect the level of subjective life satisfaction in adulthood, our current study has limitations. Our design is cross-sectional in nature; the alpha coefficients on a number of variables are relatively low; and our model focuses more extensively on variables that examine links between maternal variables rather than paternal variables and current subjective life satisfaction for the 17 year olds in our sample. For the more modest goal of this article, these limitations are less problematic. For example, a cross-sectional design is likely to understate rather than overstate the strength of associations between the subjective well-being of adolescents and our predictor variables. As noted above, previous research on long-term effects of the socio-economic and psychological characteristics of family of origin consistently find the strongest associations between adolescent outcomes and the duration of different patterns of income and economic hardship that occurred earlier in the life cycle. Hence, if an adolescent lived in a family where the parents experienced frequent and/or long spells of unemployment or where the parents reported major financial worries for many years during the early and middle childhood period for the adolescent, the effects of such an experience may far exceed the effects of a temporary hardship experienced at one point during late adolescence. In the cross-sectional evidence, for example, we cannot differentiate between an adolescent whose parents experienced many periods of economic hardship, but where the current situation is less problematic, and an adolescent who experienced little economic hardship during early and middle childhood, but whose parents are temporarily expressing a one time spell of economic difficulties. Relatively low alpha coefficients generally produce difficulties for analyses because of the lack of precision in measuring the underlying concept—a problem that results in less efficiency and in understated levels of statistical significance. The greater emphasis on maternal variables in our study also understates potential effects of parents on the adolescent’s well-being because we have more limited predictors for the fathers. At this point, however, we did not want to separately analyze adolescents in two-parent households because of well-documented effects of separation and divorce on child and adolescent outcomes. In future research, we intend to develop more complex models that enable us to exploit the longitudinal features of the German SOEP and to address more comprehensively our core long-term research question.

Overall, despite the cautions and limitations noted above, our results suggest that considerable opportunities may exist for the design of micro and macro policy level interventions that affect the level of subjective well-being for young people in transition from youth to adulthood. Hence policies that attempt to lower unemployment levels and that seek to minimize economic hardship and uncertainty for parents may influence not only the level of subjective well-being experienced by parents, but also the formation of the subjective well-being of their children. Programs that teach adolescents how to achieve greater personal control and/or provide direct opportunities in which youth can exercise personal control may lead to greater reported satisfaction with life. Similarly, programs that encourage parents to develop strong and positive relationships with their children and that provide the psychological and socio-economics supports necessary to facilitate the development of these relationships may also lead to long-term effects on subjective well-being of youth that may persist into adulthood.

In our future work, we intend to make use of the longitudinal feature of the German SOEP and to link patterns of income, economic hardship, timing of family dissolution, and patterns and levels of parental life satisfaction over time with the subjective well-being of adolescents. We also intend to investigate whether factors that are associated with the set point of life satisfaction are also associated with adult patterns of adaptation. Such questions are essential in order to deepen our understanding of how individual, family and policy efforts may help to shape subjective well-being. Our current study strongly suggests that such efforts may yield promising results.