The trajectory of adolescent mental health: The effects of parental divorce and marital conflict during childhood

全文

(1)國立臺灣師範大學人類發展與家庭學系博士論文. The Trajectory of Adolescent Mental Health: The Effects of Parental Divorce and Marital Conflict during Childhood. 指導教授：周麗端博士研究生：勞賢賢. 中華民國 101 年 6 月.

(2)

(3) ABSTRACT. Using a nationally representative dataset (Taiwan Education Panel Survey, TEPS), this study examined the effect of parental divorce and marital conflict during childhood on adolescent mental health, including happiness and depressed mood. The final sample comprised 3,886 adolescents. Results from Hierarchical Linear Regression (HLM) demonstrated that both parental divorce and marital conflict during childhood reduced the initial level (grade 7th) of adolescent happiness, and increased the initial level (grade 7th) of adolescent depressed mood. Furthermore, adolescents from families with parental divorce (occurring when they were 0~12 years old) with pre-divorce marital conflict have the lowest happiness and highest depressed mood, even worse than adolescents from two-parent families with marital conflict.. Keywords: Adolescent mental health, parental divorce, parental marital conflict, HLM.

(4) TABLE OF CONTENTS. CHAPTER 1 INTRODUCTION................................................................................................................................1 RESEARCH MOTIVATION .........................................................................................................................................1 RESEARCH PURPOSES.............................................................................................................................................5 DEFINITION OF TERMS............................................................................................................................................5 1. Mental health ..........................................................................................................................................5 2. Adolescents..............................................................................................................................................6 3. Parental divorce during childhood.........................................................................................................6 4. Parental marital conflict during childhood ...........................................................................................6 CHAPTER 2 LITERATURE REVIEW........................................................................................................................7 WHY USE LIFE COURSE THEORY AS THE FRAMEWORK OF THIS STUDY? ..........................................................................7 The Core Assumption: Life Trajectory........................................................................................................7 The Principle of Linked lives ........................................................................................................................8 The Principle of Timing ...............................................................................................................................8 MENTAL HEALTH .................................................................................................................................................11 1. A concept defined using multiple disciplines ......................................................................................11 2. The sociological approach to mental health ......................................................................................13 3. Measurements of mental health.........................................................................................................15 4. The need for longitudinal research on mental health .......................................................................18 PARENTAL DIVORCE AND PARENTAL MARITAL CONFLICT ...........................................................................................19 1.The Effects of Parental Divorce..............................................................................................................19 2.The Effects of Parental Marital Conflict ................................................................................................20 3.The Effects of Pre-divorced Parental Marital Conflict .........................................................................21 PARENTAL INVOLVEMENT......................................................................................................................................23 PARENTAL EDUCATION AND HOUSEHOLD INCOME ...................................................................................................24 GENDER DIFFERENCES..........................................................................................................................................25 SCHOOL PROGRAM TYPE, SCHOOL TYPE, AND SCHOOL URBANIZATION ......................................................................25 USING HIERARCHICAL LINEAR MODELING IN LONGITUDINAL RESEARCH......................................................................26 1.Why use Hierarchical Linear Modeling(HLM)? ...................................................................................26 2. Using HLM on longitudinal research...................................................................................................26 CHAPTER 3 METHODOLOGY.............................................................................................................................28 DATA..................................................................................................................................................................28 HYPOTHESES.......................................................................................................................................................30.

(5) DATA ANALYSIS....................................................................................................................................................31 1. Level-1 and Level-2 Variables ...............................................................................................................31 2. Definition of “Time” and its centering.................................................................................................33 3. Defining the linear and non-liner models ...........................................................................................34 3-1 The linear model ...............................................................................................................................35 3-2 The quadratic model ........................................................................................................................35 3-3 The Comparison of Linear and Quadratic Models ........................................................................36 4. Models with predictive variables: Full model .....................................................................................37 DEPENDENT VARIABLES ........................................................................................................................................39 Mental health ............................................................................................................................................39 DEPENDENT VARIABLES ........................................................................................................................................39 Mental health ............................................................................................................................................39 1. Global happiness...................................................................................................................................39 2. Depressed mood ...................................................................................................................................39 INDEPENDENT VARIABLES .....................................................................................................................................39 1. Parental divorce during childhood.......................................................................................................40 2. Parental marital conflict during childhood .........................................................................................40 3. Family types ...........................................................................................................................................41 (1) Four types .............................................................................................................................................41 (2) Six types ................................................................................................................................................42 4. Paternal and maternal involvement ...................................................................................................43 CONTROL VARIABLES............................................................................................................................................44 1. Parental education................................................................................................................................44 2. Household income ................................................................................................................................45 3. Respondent gender...............................................................................................................................45 4. School program type ............................................................................................................................45 5. School type.............................................................................................................................................45 6. School urbanization ..............................................................................................................................46 CHAPTER 4 RESULTS...........................................................................................................................................47 DESCRIPTIVE RESULTS...........................................................................................................................................47 HLM RESULTS ....................................................................................................................................................53 1. The Trajectory of Happiness.................................................................................................................53 (1) The Unconditional Model .................................................................................................................53 (2)The Effects of Parental Divorce..........................................................................................................55 (3)The effects of parental marital conflict .............................................................................................62.

(6) (4)The effects of family types .................................................................................................................68 2. The Trajectory of Depressed Mood .....................................................................................................76 (1) The Unconditional Model .................................................................................................................76 (2)The Effects of Parental Divorce..........................................................................................................78 (3)The effects of parental marital conflict .............................................................................................84 (4)The effects of family types .................................................................................................................90 CHAPTER 5 DISSCUSSION ..................................................................................................................................99 THE MAIN FINDINGS............................................................................................................................................99 RECOMMENDATIONS ........................................................................................................................................ 102 LIMITATIONS..................................................................................................................................................... 104 FUTURE DIRECTION........................................................................................................................................... 104 REFERENCES..................................................................................................................................................... 106 APPENDIX A...................................................................................................................................................... 114 APPENDIX B...................................................................................................................................................... 114 APPENDIX C...................................................................................................................................................... 115 APPENDIX D ..................................................................................................................................................... 115 APPENDIX E ...................................................................................................................................................... 116 APPENDIX F ...................................................................................................................................................... 117 APPENDIX G ..................................................................................................................................................... 120 APPENDIX H ..................................................................................................................................................... 121 APPENDIX I ....................................................................................................................................................... 122 APPENDIX J....................................................................................................................................................... 123 APPENDIX K...................................................................................................................................................... 126.

(7) LIST OF TABLES. 1. Data collection in TEPS .......................................................................................................................... 32 2. Time Centering........................................................................................................................................... 37 3. The Deviance of Linear and Quadratic Model for happiness ................................................................ 39 4. The Deviance of Linear and Quadratic Model for Depressed Mood ................................................... 40 5. Sample Selections of Four Family Types ................................................................................................ 45 6. Sample Selections for Six Family Types ................................................................................................. 46 7. Sample Characteristics: Level 2 variables ............................................................................................... 50 8. Sample Characteristics: Level 1 variables................................................................................................ 54 9. The Unconditional Model—Happiness................................................................................................... 57 10. HLM Result: The Effect of Parental Divorce on Adolescent Happiness ............................................. 59 11. HLM Result: The Effect of Parental Marital Conflict on Adolescent Happiness ................................ 66 12. HLM Result: The Effect of Family Types on Adolescent Happiness................................................... 72 13. The Unconditional Model—Depressed Mood ....................................................................................... 79 14. HLM Result: The Effect of Parental Divorce on Adolescent Depressed Mood.................................. 81 15. HLM Result: The Effect of Parental Marital Conflict on Adolescent Depressed Mood .................... 87 16. HLM Result: The Effect of Family Types on Adolescent Depressed Mood ....................................... 94. 1.

(8) LIST OF FIGURES 1. Wheaton’s concept of how life transition event affect mental health .................................................... 26 2. Study framework........................................................................................................................................ 36 3. The trajectory of adolescent happiness..................................................................................................... 58 4. The trajectory of adolescent happiness – by parental divorce (two groups)......................................... 64 5. The trajectory of adolescent happiness – by parental divorce (three groups)....................................... 65 6. The trajectory of adolescent happiness – by parental marital conflict (two groups)............................ 70 7. The trajectory of adolescent happiness – by parental marital conflict (four groups) ........................... 71 8. The trajectory of adolescent happiness – by four family types .............................................................. 77 9. The trajectory of adolescent happiness – by six family types ................................................................ 78 10. The trajectory of adolescent depressed mood.......................................................................................... 80 11. The trajectory of adolescent depressed mood – by parental divorce (two groups).............................. 86 12. The trajectory of adolescent depressed mood – by parental divorce (three groups)............................ 86 13. The trajectory of adolescent depressed mood – by parental marital conflict (two groups)................. 92 14. The trajectory of adolescent depressed mood – by parental marital conflict (four groups) ................ 93 15. The trajectory of adolescent depressed mood – by four family types ................................................... 99 16. The trajectory of adolescent depressed mood – by six .........................................................................100. 2.

(9) AKOWLEDGEMENTS. I have many key people to thank for the completion of my dissertation. The following is only a small token of my real appreciation to those have helped make this become a reality. First and foremost is my mother, whose wholehearted support and unwavering trust are the foundation of everything I do; my husband, Hung-Chin, and my daughter, Yung-En, whose unconditional love and smiles (especially my daughter) are the lights of my life; I am particularly appreciative my advisor, Dr. Li-tuan Chou, whose time and devotion in personal instruction were innumerable; I appreciate my dissertation commit members Dr. Ping-yin Kuan, Dr. Li-chen Cheng, Dr. Ming-yeh Wu, and Dr. Fu-mei Chen for giving me all the useful comments to make my research more completed. Special thanks are also given to Dr. Hsien-yuan Hsu for the instruction in research method, and all my dear friends and schoolmates (especially Ming-shan, Wei-e, Yu-min, Siao-ying and Siou-huei) for the consistent support and encouragement. Finally, I would like to thank my heavenly father, for giving me the strength and perseverance throughout the process. I thank you all.. 3.

(10) CHAPTER 1. INTRODUCTION Research Motivation Adolescent mental health has attracted global attention. According to World Health Organization (WHO, 2012), mental well-being is an important indicator of good quality of life, and is related to self-esteem, behavior, educational performance, social skills and life opportunities (Bully, 1991). Mentally healthy adolescents possess problem-solving skills, social competence and a sense of purpose, and all of which can help them overcome difficult circumstances and also avoid risky behaviors. (Scales, 1999). Furthermore, happy and confident adolescents are more likely to grow into a happy and confident adults able to contribute their health or well-being to the national good (WHO, 2012). For governments, policy makers and stakeholders everywhere, it is important to understand current adolescent mental health and investigate both positive and negative influences. Depression is the main cause of mental disability among adolescents. According to a report by the WHO (2012), in any given year, about 20% of adolescents experience a mental health problem, typically depression or anxiety. Moreover, according to a report by the Department of Statistics, Ministry of Interior (DSMI, 2003), 40% of Taiwanese youth experiences depression when facing serious life problem or pressure. Poor mental health is strongly related to other health conditions and development outcomes in young people, including low educational achievement, substance abuse, violence, and poor reproductive and sexual health (WHO, 2012). In the worst scenario, depression leads adolescents to attempt or commit suicide. With the increasing rate of depression among adolescents, the global suicide rate is increasing for both males and females (WHO, 2002). The global suicide rate has increased 40% in the last 45 years (WHO, 2009), 1.

(11) meaning that currently one person successfully commits suicide every 40 seconds. In Taiwan, according to Department of Health, Executive of Yuan (DHEY, 2009), suicide has ranked as the eighth or ninth leading cause of death since 2001, and the suicide rate has increased approximately 41% over the past decade. Furthermore, suicide is one of the leading causes of death among young people in Taiwan. According to the National Adolescent Health Information Center (2006), suicide was the third leading cause of death in the 10-24 year old age group in the United States in 2003. In Taiwan, suicide is the third and second leading cause of death among young people aged between 15~19 years old and 20~24 years old, with these two groups having suicide rates 85% and 50% higher than the rest of the population, respectively. Suicide is a multidimensional disorder and results from a complex interaction among different factors, [such as OR including] biological, genetic, psychological, sociological and environmental factors (WHO, 2000). However, mental disorders are a major factor associated with suicide (WHO, 2000). Negative life events or experiences, such as parental divorce/separation or frequent conflicts between parents, are risk factors for suicide in adolescents with depression (WHO, 2000). The negative impact of parental divorce on the mental health of offspring is well-documented. Individuals who experience parental divorce during childhood or adolescence are more likely to have poorer psychological well-being or mental health (Amato, 1988; Paul R. Amato & A. Booth, 1991; Ä ngarne-Lindberg & Wadsby, 2009; Chase-Lansdale, Cherlin, & Kiernan, 1995; Maier & Lachman, 2000; McLanahan & Bumpass, 1988; McLanahan & Sandefur, 1994). Furthermore, researchers have recently suggested that “divorce” is a continuous process rather than a single event, and should be considered in the context of marital conflict before the dissolution of a marriage. However, the effect of pre-divorced marital parental conflict is not clear yet. Studies suggested that parental divorce can cause serious negative effect on children whose parents have low levels of pre-divorced martial conflict (Amato, Loomis, & Booth, 1995; Booth & Amato, 2001; Hanson, 1999), 2.

(12) but for children of parents with high levels of pre-divorced parental marital conflict, Booth and Amato (2001) and Amato et al. (1995) suggested that parental divorce provides relief for children of high-discord families, and therefore can increase offspring well-being; but Hanson’s (1999) result suggest that parental divorce neither increase or decrease offspring well-being. Wheaton (1990) pointed out that the influence of life transition events (such as divorce, losing job etc.) depends on the context of the event occurs. If the event can relieves more existing stress than bringing the new stress, then the event can increase offspring mental health; on the other hand, if the event brings more new stress than relieving existing stress, then the event decreases offspring mental health. In other words, it is important to understand and identify the context of parental divorce occurs. Research has shown that parental involvement is correlated with positive adolescent outcomes, including better mental health (Astone & McLanahan, 1991; Crouter, MacDermid, McHale, & Perry-Jenkins, 1990), lower internalizing, externalizing, and substance use problems, and higher psychosocial competence (Steinberg, 2001). Moreover, study suggested that parental involvement is also a protective factor for children's adjustment after parental divorce (Leon, 2003). After reviewing twenty-four studies, Leon’s (2003) suggested that for adolescents of divorced families, parental monitoring and involvement are important protective factors because children spend more time in school and with peers. According to a survey by the Department of Statistics of the Ministry of the Interior, the crude divorce rate in Taiwan during 2010 was 2.46%, and has apparently remained stable for the past ten years. However, the low crude marriage rate, which has decreased 39.7% over the past decade, might limit the growth of the crude divorce rate. A study conducted in 2011, examining details of more divorced couples found that 28.7% of divorces occur within five years of marriage, while a further 28.4% occur within five to nine years of marriage, implying that for most offspring of single-parent families involved in divorce, their parents divorced before they reached adolescence. Furthermore, studies suggest that parental 3.

(13) divorce and marital conflict is more harmful to offspring when it occurs during childhood than later in life (Krein & Beller, 1988; Zill, Morrison, & Coiro, 1993). Despite the above statistics, the links between childhood parental divorce and adolescent mental health in Taiwan are relatively scarce, and most studies have focused on the effect of parental divorce on offsprting educational achievement (Hsieh, 2008; Lau, 2006; Lee & Yu, 2005; Wang, 2009). However, this study focuses on the trajectory of adolescent mental health rather than adolescent educational achievement. According to the WHO (2001), mental health is defined as a state of well-being in which every individual realizes their own potential, can cope with normal life stresses, work productively and fruitfully, and contribute to the community. That is, mental health is a multi-dimensional concept incorporating both negative and positive measurements. Not only can the positive aspects of mental state reduce mental illness in the future, but can also exert long-lasting positive influences that benefit individual functioning, and enhance quality of life and well-being for both individuals and communities (Parham, 2008; Williams, Saxena, & McQueen, 2005). Therefore, this study used both negative measurements (depressed mood) and positive measurements (happiness) . Moreover, the trajectories, processes and mechanisms associated with specific issues cannot be understood unless the analytical method considers “time” (Wu, Chang, & Chen, 2008). However, when including repeated measurements of the same individuals in an analysis, observations tend to be similar within individuals. The assumption of traditional linear regression violates the principle that observations of specific individuals are not related to observations of other individuals. Hierarchical Linear Regression (HLM) is a particular regression technique used to consider the hierarchical structure, including repeated measurement within individuals, and to model the nested data structure and provide effects for both individual and multiple levels. In short, longitudinal dataset and HLM are appropriate for observing changes over time, and individual outcome trajectories. 4.

(14) Overall, the current study hopefully can not only provide the missing piece in the puzzle of understanding the effects of parental divorce and parental marital conflict during childhood on adolescent mental health, but can also utilize longitudinal datasets to appropriately describe the trajectory of adolescent mental health during adolescence. Research Purposes The main goal of this study is to estimate the effect of parental divorce and parental marital conflict during childhood on adolescent mental health. This study thus has the following objectives: 1. Does the mental health of Taiwanese adolescents (including happiness and depressed mood) change over time? If so, what is the trajectory throughout adolescence? 2. Does parental divorce during childhood negatively affect adolescent mental health? 3. Does parental marital conflict during childhood negatively affect adolescent mental health? 4. Does the effect of parental divorce depend on pre-divorce parental marital conflict? 5. Does parental involvement can decrease the negative effects of parental divorce and marital conflict? The mechanism and process of the above research purposes was examined. A longitudinal dataset, Taiwan Education Panel Survey (TEPS), and Hierarchical Linear Regression (HLM) were used to measure changes in adolescent mental health over time.. Definition of Terms 1. Mental health Mental health is defined as a state of well-being in which every individual realizes their own potential, can cope with normal life stresses, can work productively and fruitfully, and can contribute to their community (WHO, 2001). That is, mental health should include multiple dimensions, and is not simply the absence of mental illness (WHO, 2001). Supporting the perspective of the World Health 5.

(15) Organization, several studies have addressed the importance of both positive and negative measurement of mental health (Keyes, Dhingra, & Simoes, 2010; Norrish & Vella-Brodrick, 2009; Tennant, Joseph, & Stewart-Brown, 2007). Therefore, in this current study, both negative and positive measurement of mental health will be used, the former one is depressed mood, and the later one is global happiness. 2. Adolescents The World Health Organization defines adolescents as young people aged between 10 and 19 years old. However, the 10 to 19 year old age range crosses four different school stages in Taiwan: the last three years of elementary school, junior high school, senior high school, and the first years of college. Different school stages have a different atmosphere and environment, and may confound the effects of parental divorce. Therefore, to decrease confounding factors, this study limits the sample to adolescents attending junior and senior high school because these two school stages have a similar atmosphere, including pressure to perform well to enter a good school for the next stages of education. 3. Parental divorce during childhood This is defined as a legal dissolution of the marriage contract of the biological parents of a child by a court or other body with competent authority before the child enters junior high school. 4. Parental marital conflict during childhood This is defined as frequent severe disagreements or fights between the biological parents of a child before that child entered junior high school.. 6.

(16) CHAPTER 2 LITERATURE REVIEW Why use Life Course Theory as the framework of this study? Life course is an emerging paradigm, and comprises both macro aspects of age stratification, cultural and intergenerational models, and micro aspects of developmental life span psychology. Elder (1974) drew on generation and age models and adopted life course thinking in a study of California children who grew up during the Great Depression. His study focused on two cohorts of children with different economic and status backgrounds during the Great Depression, traced their development to maturity, and examined how and why they faced the same difficulties while growing up but achieved different outcomes.. The Core Assumption: Life Trajectory The life course framework has developed over the years. In 2003, Elder et al. (2003) summarized five general principles that guide life course research, as follows: life span development, linked lives, timing, agency, and historical time and place. The core assumption linking these five principles of life course theory is that life trajectories, which comprise various transitions or turning points, shape developmental processes and outcomes. Trajectories are the long-term patterns and sequences of that characterize the lives of individuals, and naturally include transition or turning points. Transitions or turning points bring changes, and naturally present life trajectories over time. Social science researchers have tried to estimate the real effects of changes, build up causal models, and understand how prediction variables affect outcome variables. Life course theory highlights that the causal model or the changes displayed by individuals experiencing specific transitions or turning points cannot be determined unless life trajectories are described in detail, and the process of “change” is 7.

(17) appropriately observed and measured.. The Principle of Linked lives According to Elder, Johnson, and Crosnoe (2003), the principle of linked lives means “lives are linked interdependently and socio-historical influences are expressed through this network of shared relationships” (p.13). Moreover, the principle of linked lives emphasizes the intergenerational connection between parents and children, because it helps clarify how changing parental fortunes can affect the development of their children (Elder, 1994, 1998). Family is the primary setting in which children are born and grow. Despite extensive literature on the influence of family context (such as family structure, family dynamics etc) on mental health outcomes for offspring. Uhlenberg and Mueller (2003) noted that researchers should not consider the consequences of family context for particular life course outcomes to be universal, because the influences of family environment may vary significantly across societies, cultures, and times. Moreover, the process for handling marital discord varies among families. Some couples may exhibit serious conflicts or arguments throughout their marriages, while others may adopt less violent and confrontational ways of managing their marriage. Family context is embedded in social, cultural and historical context, and can have different influences on offspring. Regardless of the nature of the family environment in which they grow up, children are unable to choose it, and inevitably are strongly influenced by those around them.. The Principle of Timing The principle of timing refers to how “the developmental antecedents and consequences of life transition, events, and behavioral patterns vary according to their timing in a person’s life” (Elder, et al., 2003). Moreover, many studies suggest that childhood family experiences (including parental divorce 8.

(18) and marital conflict) strongly [impact OR influence] offspring outcomes (Kalter & Rembar, 1981; Krein & Beller, 1988; Palosaari & Aro, 1994; Zill, et al., 1993).. (1) The impact of childhood parental divorce on outcomes of offspring Krein and Beller (1988) merged three datasets from the National Longitudinal Survey of Labor Market Experience (NLS) to examine the effect of parental divorce on offspring educational attainment. Krein and Beller (1988) identified three periods of childhood spent in a single-parent family: preschool years (0-5.5 years old), elementary school years (5.5-13.5 years old), and high school years (13.5-18.0 years old). Their findings show that the negative effect of parental divorce is strongest during the preschool years among all age groups and for both males and females. Moreover, the second strongest negative effect occurs during the elementary school years and for males, even after controlling for income. Zill, Morrison and Coiro (1993) used the longitudinal dataset National Survey of Children (NSC) to investigate the long-term effect of parental divorce on offspring outcomes, such as adjustment and achievement. Zill, Morrison and Coiro (1993) divided the respondents into three groups: those whose parents divorced before the respondent was 6 years old (early divorce), those whose parents divorced when the respondent was aged 6-16 years old (late divorce), and those whose parents did not divorce. Their results suggested that the early divorce group showed higher rates of dropping out from high school and behavioral problems during adolescence. Kalter and Rembar (1981) showed that both male and female adolescents who experienced childhood parental divorce, including during the age ranges of 0~2.5 years old, 3~5.5 years old, and over 6 years old, are more likely to display aggression towards parents and peers, and are also more likely to have academic problems compared to those who underwent parental divorce during adolescence. Palosaari and Aro (1994) conducted a Finnish study investigating the timing of parental divorce 9.

(19) and its influence on youth depression. Palosaari and Aro (1994) divided the children into three groups based on their experiences of parental divorce: those who experienced parental divorce before reaching school age (under 7 years old), those who experienced it in latency (7~12 years old), and those who experienced it in adolescence (13~16 years old). Their results showed a statistical association between parental divorce occurring in latency (7~12 years old) and increased risk of subsequent depression.. (2) The influence] of childhood parental conflict on offspring outcomes Parental conflicts exert both short-term and long-term effects on outcomes in offspring (Enos & Handal, 1986; Stocker & Youngblade, 1999). In a study of the effect of marital conflict and hostility on the relationship of children with peers and siblings, Stocker and Youngblade (1999) suggested that marital conflict is associated with problematic sibling and peer relationships among offspring. Since children tend to interpret and understand parental conflicts rather than simply being passive observers, they are more likely to feel threatened by parental conflict and blame themselves for its occurrence. Their study also demonstrated that parental marital conflict is associated with parental hostility toward children, and parental hostility is negatively related with relationships between children and others. Moreover, Enos and Handal (1986) recruited adolescents aged between 13 and 18 years old to investigate the effect of parental conflict on their adjustment. The mean length of time since parental divorce was approximately 6 years, implying that most of the adolescents sampled had experienced parental divorce during childhood. Enos and Handal (1986) divided respondents into three groups based on the intensity of parental marital conflict namely low-conflict, medium-conflict and high-conflict groups. Their analytical results showed that the low-conflict group reported better adjustment than the middle- and high-conflict groups, and furthermore reported higher satisfaction with their social life than the high conflict group. Other studies also suggested that the quality of the parental marital relationship during childhood is an important indicator of offspring mental health (Paul R. Amato & Alan Booth, 10.

(20) 1991; Rodgers, Power, & Hope, 1997) Although different studies have focused on parental divorce and marital conflict occurring in different age groups, the limited cognitive ability of children compared to adolescents makes them less able to assess the causes and consequences of parental divorce or marital conflict, and potentially more likely to blame themselves for parental marital conflict or divorce. Furthermore, children are highly dependent on their parents and therefore less able to turn elsewhere, such as to school, peers or social welfare agencies, for protection against harm originating from their family environment (Hetherington, 1989; Zill, et al., 1993). Since the effects of parental divorce and marital conflict may affect offspring more during childhood than subsequently, and since more than half of divorces in Taiwan occur within ten years of marriage, it is important to investigate the effect of childhood parental divorce and marital conflict on adolescent outcomes of mental health. Overall, the timing principle of life course theory guides this study to stress the childhood experience; the principle of linked lives leads this study to stress parental divorce and marital conflict, and to examine outcomes in a specific structural and static context; and the core assumption, that developmental process and outcome are shaped by life trajectories, causes this study to emphasize the importance of dynamic changes over time. This study attempts not only to understand how parental divorce and marital conflict affect adolescent mental health, but also to describe the changes of the development trajectories of mental health. Therefore, life course theory, which incorporates both static and dynamic perspectives, provides a suitable framework for this longitudinal study.. Mental Health 1. A concept defined using multiple disciplines Mental health is a multi-discipline concept, and different academic disciplines define and 11.

(21) measure it differently. Switzer, Dee, and Bromet (1999) summarized the four main parent disciplines, namely psychology, sociology, environmental studies and biology, and the orientations of psychology, sociology, epidemiology, and psychiatry in relation to mental health related issues (Fig. 1).. Note. Major academic disciplines concerned with mental health issues. Adapted with permission from the Handbook of the Sociology of Mental Health (p. 82), by C. S. Aneshensel and J. C. Phelan (Eds.), 1999, New York: Kluwer Academic/Plenum. Copyright 2007 by the Springer Science + Business Media LLC.. Although each of the four main approaches, namely psychology, sociology, epidemiology, and psychiatry, covers all four disciplines, the relative weightings on each of the four disciplines differ among these approaches. Generally, psychology and psychiatry assign a heavier weighting to internal states (the upper part), and assign a lighter weighting to external factors that affect mental health; meanwhile, sociology and epidemiology assign a heavier weighting to external factors when examining mental health issues (the lower part), and a lighter weighting to individual internal states. In terms of examining internal states, compared to psychiatry, psychology more strongly emphasizes cognitive processes (upper-right), whereas psychiatry focuses on how physical factors relate 12.

(22) to mental disorders (upper-left). These two disciplines typically measure mental health [individually OR at the individual level], for example using clinical interviews, neurological examinations, or behavioral observation to understand the causes of mental disorders. On the other hand, in terms of examining the relationship between external factors and mental health issues, compared to epidemiology, sociology emphasizes societal-level structures and process (lower-right), whereas epidemiology stresses physical environmental factors, such as environmental conditions or hazards, in mental health issues (lower-left). These two disciplines are more likely to use surveys, interviews, or secondary datasets that are population-based. Each of the four disciplines provides and contributes overlapping and unique perspectives for understanding mental health issues, and none can exist independently of the others. No clear boundaries or guidelines exist among these four disciplines, the measurements of mental health used in previous lectures sometimes cannot distinguish clearly between psychological and sociological approaches, and the terms describing mental state, such as mental health, mental illness, psychological well-being or depression etc., are frequently interchangeable. While many previous studies (Huang & Lin, 2010) used well-developed questionnaires (e.g., Goodman Strengths and Difficulties Questionnaire 1997) to represent “mental health” content, and to detect symptoms of mental illness or disorder in individuals This study focuses instead on changes in teenager mental state over time, and how different social segments affect that mental state. Therefore, this study adopts a sociological approach, and adopts both positive (global happiness) and negative (depressed mood) measurements of mental health. The reasons this study adopts a sociological approach are addressed in more detail later.. 2. The sociological approach to mental health While all individuals have a mental health status, specific mental health traits are clustered within certain social strata rather than being randomly distributed throughout society (Aneshensel & Phelan, 13.

(23) 1999). That is, social status, including gender, race/ethnicity, or socioeconomic status, influence individual mental health, and also influence the expression of mental disorder, such as through depression, drug use or alcoholism, and ways of identification or treatment, such as females being more likely than males to recognize emotional problems (Yokopenic, Clark, & Aneshensel, 1983), or young females being more likely to exhibit sadness, suicidal ideation and suicide attempts, while young males are more likely to actually commit suicide (NAHIC, 2006). Differences in mental health among social groups are linked to corresponding differences in exposure to the social conditions that affect mental health. The sources of societal systematic differences in mental health do not exist by accidence or happenstance, but rather in the repetition and reproduction of the commonplace. The effect of parental divorce is well-documented in numerous countries, and thus the disadvantages affecting children growing up in families that have undergone parental divorce are concentrated in certain social segments rather than affecting children in all segments of society randomly. Parental divorce is associated with socioeconomic status. Single mothers are more likely to have lower occupational status, and offspring of single parent families are disadvantaged not only in terms of socioeconomic resources, but also in academic performance (Bogenschneider & Steinberg, 1994; Jeynes, 1999; Lee & Yu, 2005; Pallas, 2003; Riala, Isohanni, Jokelainen, Jones, & Isohanni, 2003; Sandefur & Wells, 1999; Sui-Chu & Willms, 1996; Wojtkiewicz, 1993), mental health (Amato, 1988; Paul R. Amato & A. Booth, 1991; Ä ngarne-Lindberg & Wadsby, 2009; Chase-Lansdale, et al., 1995; Maier & Lachman, 2000; McLanahan & Bumpass, 1988; McLanahan & Sandefur, 1994), or delinquency behaviors (Amato & Cheadle, 2008; Amato & Keith, 1991; Brown, 2006; Doherty & Needle, 1991). However, not everyone who experiences parental divorce experiences the same outcomes. Some suffer negative consequences from parental divorce while others remain unscathed. The sociological approach focuses on why the same cause can have different effects on different groups from different societal strata. To summarize, even mental health seems to be an issue within the development of individuals, 14.

(24) but when the cause, in this case parental divorce, of mental health status is stratified by societal segments, adopting a sociological approach to studying the issue is appropriate.. 3. Measurements of mental health Several terms have been used to describe similar concepts to mental health. In a study of the effect of pre-divorce parental relations on offspring outcomes, Amato (2001) and Amato, Loomis & Booth (1995) used the concept of “psychological well-being”. The study of Amato (2001) comprised global happiness, life satisfaction, and self-esteem; and the study of Amato, Loomis & Booth (1995) comprised overall happiness and psychological distress. Hanson (1999) also used the term “psychological well-being” to analyze the effect of parental conflict and divorce on children, where “psychological well-being” comprised behavior problems, global quality of life, and self-esteem. Other common used measures of mental health are depression (Aseltine, 1996; Videon, 2002) and emotional problems (Chase-Lansdale, et al., 1995; Cherlin, Chase-Lansdale, & McRae, 1998). In Taiwan, some scholars have measured mental health using the Taiwanese Educational Panel Study (TEPS). In a study exploring the relationship between educational achievement and adolescent mental health, Yang (2005) used 14 questions to measure mental health, including depression, insomnia, dizzy, loneliness, helplessness, suicidal ideation, etc. Huang & Lin (2010) used 15 questions of TEPS to develop a Mental Health Questionnaire to compare adolescent mental health between Taiwan and the United States, where the questionnaire included four subscales, measuring emotional symptoms, problematic behavior, attention/distraction, and peer relations. The most recent study using TEPS to measure mental health was the investigation of Wei (2008) exploring the influences on mental health. Wei used 16 questions, similar to the study of Yang, to measure mental health, including depression, loneliness, helpless, insomnia, anger, etc. Although the three domestic studies addressed above all use the term “mental health”, the 15.

(25) measurements employed more closely resemble measures of “mental disorder” or “mental illness”. That is, the measures employed focus only on the negative side of mental health, and do not sufficiently accurately measure overall mental health (Keyes, et al., 2010; Power, 2010; Tennant, et al., 2007). Keyes, Dhingra, and Simoe (2010) distinguished positive mental health from mental illness. Measurement of positive mental health include such measures as feeling cheerful, in good spirits, happy, calm or peaceful, satisfied, and full of life during the past 30 days. Keyes et al. (2010) compared changes of positive mental health and mental illness between 1995 and 2005 in adult population, and suggested that changes in mental health strongly predicted changes in mental illness. Westerhof and Keyes (2010) also distinguished the concepts of mental illness and mental health, and indicated that mental illness (namely distressed) and mental health (namely happy) are related but distinct dimensions of a single concept. Demonstrating the distinctiveness of these two concepts, age has more impact on mental illness, while gender and marital status have more impact on mental health. Tennant, Joseph, and Stewart-Brown (2007) also proposed the importance of separately measuring negative and positive mental health. Positive psychology is another term similar to positive mental health, and is defined as “seeking to create more understanding of human happiness and optimal functioning”(Norrish & Vella-Brodrick, 2009), and being based on the assumption that “a fulfilling and happy life consists of more than an absence of mental dysfunction” (Keyes, 2005). Norrish and Vella-Brodrick (2009) considered positive measurements of mental health is particularly important for adolescents, because adolescents tend to have moderate or average mental health rather than good mental health. While most adolescents do not fall within the clinical range for mental illness, a large portion of adolescents exhibit mental health issues (namely lack of confidence, feelings of insecurity, etc.), creating a need for measurements that involve more than just mental illness. The World Health Organization (2001) defines mental health as “a state of well-being in which every individual realizes their own potential, can cope with normal life stresses, can work productively 16.

(26) and fruitfully, and can make a contribution to their community”. The concept of mental health is broader than those of mental disorder or illness, and includes more dimensions than the mere absence of illness (Aneshensel & Phelan, 1999; WHO, 2001). Of course mental disorder or illness can decrease the likelihood of achieving good mental health, but measuring only mental disorder or illness does not provide a complete picture of mental health. Furthermore, recent studies demonstrate the need for positive measures of mental health (Keyes, et al., 2010; Norrish & Vella-Brodrick, 2009; Tennant, et al., 2007; Westerhof & Keyes, 2010). Johansson, Burnberg, and Eriksson (2007) investigated mental health from the perspective of adolescents, and the results obtained suggested that adolescents perceived mental health as an emotional experience incorporating both positive and negative aspects. Therefore, this study uses both negative and positive measures of mental health. The negative measure is depressed mood, while the positive measure is happiness. (1) Depressed mood Adolescent depression can be classified using three levels (Petersen et al., 1993): depressed mood, depression syndrome, and clinical depression. Depressed mood is a common emotion that can occur at certain points in life and is typically linked to problems such as anxiety and social withdrawal. Depressed mood is typically measured through adolescents providing self-reports of their emotions, and is the single most powerful symptom for differentiating clinically referred and non-referred youth. Depression syndrome is a constellation of problematic behaviors and negative emotions, including social problems, thought problems, attention problems, etc. Depression syndrome is diagnosed based on reports of adolescents, parents, and school teachers. Clinical depression is linked to a set of long-term emotion and behavior problems that impair functioning of affected adolescents. Clinical depression can be diagnosed using the categorization of mental disorders developed by the American Psychiatric Association or the World Health Organization. In this study, because it adopts a sociological approach and the main focus is an adolescent population, 17.

(27) and because only self-reported data from adolescents is used, depressed mood is not an appropriate indicator of depression in this study. The measurement of depressed mood is calculated from four items from the student questionnaires, including “do not want to deal with others”, “feeling upset”, “wanting to yell or throw things”, and “feeling lonely”.. (2) Happiness This study considers happiness an appropriate positive measure of mental health because it not only represents an important mental condition identified in numerous studies (Amato, 2001; Amato, et al., 1995; Westerhof & Keyes, 2010), but also has been found to reduce symptoms of psychopathology (Diener, 2002), improve physical health (Dillon, 1995), improve coping ability (Fredrickson, 2002), increase self-control (Aspinwall, 1998), enhance relationships with others (Harker, 2001), increase opportunities for success (Lyubomirsky, 2005), and even contribute to longer lifespan (Diener, 2011).. 4. The need for longitudinal research on mental health According to Life Course Theory, early life experiences or states significantly influence later outcomes. Mental health is an accumulative process, and has the potential to alter subsequent life-course. Furthermore, the mental health of individuals during their the life-course is not static, but rather a dynamic developmental process influenced by individual life experience. The trajectory of individual mental health cannot be understood unless it is possible to investigate “time” and “changes over time”. Cross-sectional studies provided valuable information regarding certain issues. However, longitudinal research is required to understand certain phenomena that change over time (namely mental state), and particularly how and why they change. According to Menard (2002), longitudinal research is defined in terms of both data and method of analysis, and the criteria used in its definition are as follows: (a) data are collected for each item or 18.

(28) variable over two or more different time periods; (b) the subjects analyzed are the same or at least comparable between or among periods; and (c) the analysis involves comparison of data between or among periods. The current study uses the same variables and subjects from four waves of data of the Taiwan Education Panel Survey (TEPS), thus meeting criteria (a) and (b); moreover, Hierarchical Linear Regression (HLM) can measure changes over time and enable the comparison of different time period, meeting criterion (c). Therefore, this study can be considered a longitudinal study, and thus is well suited to describe the trajectory of mental health.. Parental Divorce and Parental Marital Conflict 1.The Effects of Parental Divorce Adolescence is a unique life stage during which individuals experience major physiological and psychological changes. The mental health of adolescents strongly influences adult outcomes (Amato, et al., 1995; Glenn & Kramer, 1985; Pallas, 2003), and family is the most important determinant of mental health in young people (Johansson, et al., 2007). Parental divorce during childhood had both short- and long-term negative impacts on offspring mental health, including anxiety and psychological distress (Amato, 1988; Paul R. Amato & A. Booth, 1991; Ä ngarne-Lindberg & Wadsby, 2009; Chase-Lansdale, et al., 1995; Maier & Lachman, 2000; McLanahan & Bumpass, 1988; McLanahan & Sandefur, 1994). Individuals who experienced parental divorce were more likely than those who grew up in two-biological-parent families to have poorer mental health. Childhood parental divorce was associated with adult psychological state, including] overall happiness (Amato, et al., 1995; Glenn & Kramer, 1985) and depressive symptoms (McLeod, 1991). Cherlin, Chase-Lansdale, and McRae (1998) also suggested that parental divorce during childhood had a long-term effect. 19.

(29) Videon (2002) used longitudinal data to investigate the short-term effects of parental divorce, and found that adolescents with divorced parents are more likely to be depressed within two years of the divorce than are their peers from two-biological-parent families. Chase-Lansdale, Cherlin & Kiernan (1995) suggested that parental divorce had moderate long-term effects on young adults. Young adults who experience parental divorce during childhood have higher total Malaise Inventory scores. A recent Swedish study showed that during the 15 years following divorce, divorced women are more likely than non-divorced women to have lower mental health, including higher depression or anxiety (Ä ngarne-Lindberg & Wadsby, 2009). Maier and Lachman (2000) examined the influences of early parental experiences on middle-aged adults, and suggested that middle-aged men and women from families that have undergone parental divorce exhibit higher prevalence of physical health problems, and men show higher rates of depression. Using samples from Taipei City, Cheng (2001) analyzed the life adjustment of fifth to eighth graders from two-biological-parent and single-parent families, and found the latter group to be disadvantaged in certain dimensions of social adjustment, including social adjustment. She also noted that the causes of the disadvantaged situation in single-parent families involve more than just the absence of one parent, but rather involve a combination of numerous complex factors.. 2.The Effects of Parental Marital Conflict Studies suggest that that both parental divorce and parental marital conflict, negatively influence outcomes in children. Children and adolescents who experience harmful parental conflict are more likely to [have OR suffer] both internalizing and externalizing problems(Buehler et al., 1997; Lindsey, Chambers, Frabutt, & Mackinnon-Lewis, 2009; Schoppe-Sullivan, Schermerhorn, & Cummings, 2007; Wang & Chen, 2010), and exhibit poor academic performance (Mechanic & Hansell, 1989). Lindsey et al. (2009) used longitudinal data to analyze the effect of marital conflict on 20.

(30) adjustment of children, including internalizing and externalizing symptoms. The children in their study ranged from 8 to 16 years old, and the analytical results indicate that after controlling for parenting process and early adjustment of children, marital conflict still negatively impacts later internalizing symptoms, but not externalizing symptoms. Schoppe-Sullivan et al. (2007) also used longitudinal data and included 268 mother-adolescent (fifth grade) dyads in their study analyzing the influences of marital conflict on overt and relational aggression in adolescents. The results demonstrate that marital conflict both directly and indirectly influences adolescent aggression. Mechanic and Hansell (1989) also used longitudinal data to analyze the influence of marital conflict on adolescent well-being, including depressed mood, anxiety, self-esteem, and physical symptoms. Their results suggested that higher family conflict increased depressed mood, anxiety, and physical symptoms in adolescents. Amato et al. (1995) provided several potential mechanisms for the negative impact of marital conflict on children. First, conflict between parents is a direct stressor, and parental arguments can cause children to experience fear, anger, or aggression. Conflict between parents or between parents and children can cause physical violence and increase the risk of psychological and behavioral problems, and can also reduce child academic performance. The egocentricity of young children makes them likely to blame themselves for parental conflict, resulting in lower levels of psychological well-being. Moreover, children learn some of their social skills from the behavior of their parents, and so parents adopting conflict behaviors rather than communication or compromise, may lead to children learning inappropriate social skills.. 3.The Effects of Pre-divorced Parental Marital Conflict Despite the above literature addressing the influences of parental divorce or marital conflict on offspring, researchers increasingly consider “divorce” a continuous process rather than a single event, 21.

(31) and consequently consider it in the context of parental marital conflict. Studies suggest that martial conflict can contribute partly, but not completely, to subsequent offspring outcomes (Amato, et al., 1995; Booth & Amato, 2001b; Hanson, 1999). Hanson (1999) suggested that children exposed to high levels of pre-divorce marital conflict are neither better nor worse off following parental divorce, but parental divorce can seriously and negatively influence children exposed to low levels of pre-divorce marital conflict. Amato, Loomis and Booth (1995) and Booth and Amato (2001b) obtained similar findings, namely that parental divorce reduces the well-being of offspring in families with low levels of pre-divorce marital conflict, and increases the well-being of offspring in families with high levels of pre-divorce marital conflict. Cherlin (1998) also indicated that some of the negative influences of parental divorce on offspring during childhood or adolescence appeared before parental divorce, and the influences lasted into adulthood. It is important to distinguish pre-divorced martial conflict from general parental marital conflict is because for two-parent families, the negative influence of parental marital conflict on offspring is well-documented. However, for divorced single-parent families, the role of pre-divorced marital conflict is not clear yet. Amato et al. (1995), Booth and Amato (2001) and Hanson (1999) all suggested that parental divorce decreases offspring well-being of low levels of pre-divorced parental marital conflict families, and the prior two studies suggested that parental divorce increase offspring well-being of high levels of pre-divorced parental marital conflict families, but the Hanson’s (1999) study suggested that parental divorce neither increases or decreases offspring well-being of high levels of pre-divorced parental marital conflict families. Wheaton (1990) pointed out that life transition events (i.e. job loss, divorce, retirement etc.) is a stressor for individuals. However, when the prior chronic stress (before the events occur) exists, and more stress can be relief from the existing stress, and then the life transition event is actually beneficial for individual mental health. On the other hand, when the life transition event relieves less existing stress 22.

(32) than the new stress it brings, and then the life transition event is harmful for individual mental health. His argument can be expressed in Figure 1.. Increasing mental health existing stress. Life transition event Decreasing mental health. Figure 1. Wheaton’s concept of how life transition event affect mental health. The key factor of how life transition event, in this study is parental divorce, can affect individual mental health is depending on the context of the life transition event occurs, and the pre-divorced parental marital quality is one dimension of the context. However, the literatures of pre-divorced marital conflict on offspring mental health is relatively few when comparing with the literatures of marital conflict, and more studies need to be done to understand the context of parental divorce occurs. This study tends to broaden the existing literature of the influences of parental divorce and martial conflict, but also to explore the effects of pre-divorced marital conflict on offspring mental health in Taiwanese context.. Parental Involvement Research has shown that positive parent-child relations, such as parental caring, acceptance, open parent-child communication, and generally supportive parent-child relations, are correlated with 23.

(33) positive adolescent outcomes, including better mental health (Astone & McLanahan, 1991; Crouter, MacDermid, McHale, & Perry-Jenkins, 1990), lower internalizing, externalizing, and substance use problems, and higher psychosocial competence (Steinberg, 2001). Moreover, Flouri (2004) found that parental involvement with offspring at seven years old can predict psychological functioning feeling (namely unhappy, depressed, or under strain), psychological distress (namely feeling miserable and depressed, getting easily annoyed by others, and suffering a nervous breakdown), and life satisfaction (namely how life has turned out so far, and expectations for the next ten years) of offspring at 42 years old. Furthermore, parental involvement is also a protective factor for children's adjustment after parental divorce. After reviewing twenty-four studies, Leon’s (2003) suggested that parental warmth and responsiveness are important protective factors throughout childhood and adolescence, but parental monitoring and involvement becomes more important in middle childhood and adolescence because children spend more time in school and with peers. Therefore in this study, parental involvement serves as the protective factors to reduce the negative influence of parental divorce and marital conflict on adolescent mental health.. Parental Education and Household Income Studies have suggested that children or adolescents with higher household income and parental education particularly paternal education are less likely to exhibit emotional or behavioral problems (Bradley & Corwyn, 2002; Carlson & Corcoran, 2001; Morris, 2003). On the other hand, other studies have suggested that students with higher household income and parental education have worse mental health (Luthar & D'Avanzo, 1999). In Western countries, controversy continues regarding the effect of parental education and household income on offspring mental health state. Yang (2005) analyzed the mental health of students in Taiwan, and found that after controlling for educational achievement, neither 24.

(34) parental education nor household income exerted a significant effect.. Gender Differences Mental health state differs between males and females. Studies have suggested that females are more likely than males to recognize emotional problems (Fletcher, 2008; Yokopenic, et al., 1983), frequently report mental health difficulties, or internalize mental health difficulties (Andersson et al., 2010). According to the National Adolescent Health Information Center (2006), young females are more likely to exhibit sad feelings, suicidal ideation and attempts, while young males are more likely to actually commit suicide Moreover, Tick, Ende and Verhulst (2008) used samples of Dutch adolescents to analyze trends in emotional and behavioral problems between 1993 to 2003. Their results suggested that thought problems, internalization problems, somatic complaints, suicidal ideation and self-harm were more likely to increase among girls than boys during the ten year study period.. School Program Type, School Type, and School Urbanization In the East Asian region, the competitive educational system is an unique social background (Yi, Wu, Chang, & Chang, 2009). Educational achievement is highly valued by society, teachers, and parents highly value. Adolescents is expected to study hard and enter good high school and college (Hsu, 1971). Therefore, the pressure of having good educational achievement is high and might decrease adolescent mental health. However, due to the educational tracking system of high school in Taiwan, adolescents who can choose either academic-oriented or vocational-oriented programs or high schools, and the pressure of educational achievement might be decreased for adolescent who choose vocational-oriented programs or schools, because they are not expected to have high academic achievement anymore, but rather to have more practical skills or abilities. Moreover, school type (such as public or private) and 25.

(35) school urbanization (such as rural, sub-urban, and urban) are both correlated with the educational tracking system. In general, school with higher educational pressure tends to be academic-oriented, public, and normally located in urban areas. Therefore these factors need to be controlled for.. Using Hierarchical Linear Modeling in Longitudinal Research 1.Why use Hierarchical Linear Modeling(HLM)? In the social sciences, data structures are frequently hierarchical (Raudenbush & Bryk, 2002). This study uses variables to describe individuals, but the individuals are grouped into larger units, each comprising numerous individuals, and other variables describe these higher order units. Individuals within a population exist within clusters. For example, students are grouped into classes, classes are grouped into schools, schools are grouped into areas, and so on. Data are often nested within persons, organizational units, and/or communities (O'Connell & McCoach, 2008; Raudenbush & Bryk, 2002). Numerous studies neglect this type of data structure. When the traditional linear regression employs nested data, it violates the assumption that observations of any individual are not systematically related to observations of any other individual. HLM is a particular regression technique used to consider hierarchical structures, and thus can model the nested data structure and provide individual level and cross level effects.. 2. Using HLM on longitudinal research Longitudinal studies involve multiple observations of the same individuals over a period of time. Observations taken repeatedly across individuals tend to be similar, just as is the case for observations taken from individuals within a cluster (O'Connell & McCoach, 2008). Multilevel models, such as HLM, can not only model the nested data but can also examine how individuals change or grow over time, and identify factors related to such growth or change. HLM accommodates intra-individual 26.

(36) correlation among individuals within the longitudinal study to serve as the nesting unit or cluster for each individual (O'Connell & McCoach, 2008), and thus minimizes the treat of unit heterogeneity: one expects more similarity for observations of the same unit at different times than for simultaneous observations of different units (Halaby, 2003). Numerous traditional longitudinal approaches, such as repeated-measures MANOVA, cannot easily handle unbalanced longitudinal data, missing data, or uneven time points (Luke, 2004). Applying HLM to longitudinal research can not only accurately estimate causal effects, but can also describe how the trajectories of development and growth over time vary systematically across groups with different life course experiences (Halaby, 2003). Latent curve analysis (LCA) is another approach for modeling growth, and the two approaches of HLM and LCA are appealing because they both model individual growth as a function of time, compare different growth rates across different groups, and yield comparable results (Chou, Bentler, & Pentz, 2009). Both approaches have different advantages, and HLM can more easily specify models and yield statistical results (Chou, et al., 2009).. 27.

(37) CHAPTER 3 METHODOLOGY Data. Data were obtained from the Taiwan Education Panel Survey (TEPS). TEPS is a nationally representative longitudinal dataset containing data on Taiwanese adolescents from seventh grade to twelfth grade. TEPS was designed, administrated, and supported primarily by Academia Sinica, which is the highest academic institution in Taiwan. TEPS is intended to assess the educational performance of adolescents, including in mathematics, reading, science, and problem-solving. Students answered both the assessment questions and a questionnaire on their background, including learning habits, friends and family. Parents and teachers also completed a questionnaire on the socioeconomic status of their family and the performance of their children or students. TEPS included four waves of surveys - Wave 1 in 2001, Wave 2 in 2003, Wave 3 in 2005 and Wave 4 in 2007 (Table 1). Stratified random sampling was used. The strata were based on urbanization (namely urban, rural etc.), private or public school, and school level (namely junior high school, senior school etc.). The sample was divided into three groups during the administration of the first survey wave in 2001: junior high school student sample (n=20,077), senior high school student sample (n=14,610), and junior college student sample (4,679). Two waves (2001 and 2003) were conducted for the sample of senior high school student, and only part of junior high school student sample (n=4,000) were followed up to wave 3 and wave 4 surveys (core panel). New sample (n=16,000) were included in wave 3 and wave 4 surveys for the group of junior high school student sample to make the wave 3 and wave 4 samples more comparable with wave 1 and wave 2 samples. The completion rates for the senior high school students in the wave 1 and wave 2 surveys were 99.6% and 99.7%, respectively; meanwhile, the rates for the junior high school students in waves 1 to 4 28.