GRANDPARENTING AND HEALTH IN LATER LIFE: EVIDENCE FROM THE UNITED STATES, SOUTH KOREA, AND CHINA By Seung - won Choi A DISSERTATION Submitted to Michigan State University in partial fulfillment for the requirements for the degree of Sociol ogy Doctor of Philosophy 2019 ABSTRACT GRANDPARENTING AND HEALTH IN LATER LIFE: EVIDENCE FROM THE UNITED STATES, SOUTH KOREA, AND CHINA By Seung - won Choi A significant increase in life expectancy over the past decades has changed intergenerational rel ationships and the role of older adults in aging families. Increasingly, older adults have been involved in caring for grandchildren (i.e., grandparenting), a role emerging in later life. In light of the role strain/enhancement theories, this dissertation investigates how grandparenting is related to the physical and mental health of older adults, and how this association varies by sociocultural context. I adopt a three - essay format to address this research question. My first study draws from the Health and Retirement Study (1988 - 2014) to examine the linkage between grandparenting and mortality risk, and its racial/ethnic variation in the U.S. The results from the event history models reveal different racial/ethnic patterns in the effects of grandparenting o n mortality risk. The mortality advantage of providing grandparenting is robust for white grandparents, whereas the mortality disadvantage of grandparenting is mainly found among black grandparents. In the second study , I investigate how grandparenting aff ects depressive symptoms among older women in South Korea, using the Korean Longitudinal Study of Aging (2008 - 2012 ). The results from growth curve models indicate that caregiving grandmothers in multigenerational households experience a decline in depressi ve symptoms ove r time . The mental health gap between the multigenerational household grandparenting and non - caregiving groups decreases with age and reverses after age reaches the mid - 60s. Drawing from the China Health and Retirement Longitudinal Study (20 11 - 2015), the third study uses growth curve models to assess how grandparenting influences depressive symptoms in China. The analyses show that the level of depressive symptoms increases over time among older adults who reside in rural regions and provide multigenerational household grandparenting. The provision of full - time noncoresident grandparenting has an initially protective effect on the depressive symptoms among rural older adults. However, socioeconomic status partially accounts for the association between grandparenting and depressive symptoms. Taken together, the findings of this dissertation confirm that grandparenting plays a significant role in physical and mental health in later life. The effects of grandparenting on health and the underlying mechanisms which explain these relationships vary by sociocultural context. My dissertation advances our knowledge about intergenerational relationships by examining racial/ethnic and social/cultural differences in the consequence of grandparenting for hea lth and well - being in later life. Copyright by SEUNG - WON CHOI 2019 v ACKNOWLEDGEMENTS It has indeed been a long journey. I would like to express deep appreciation to my advisor, Zhenmei Zhang, for her enormous and con stant encouragement and guidance throughout my graduate career. She is an outstanding mentor, role model, and one of the most brilliant but caring people I have ever met. I have enjoyed every minute of time spent working with her. I am grateful to my com mittee members, Clifford Broman, Hui Liu, and Amanda Woodward, for their constructive feedback and insight that markedly improved the quality of this dissertation. They are a wonderful mentor group who have always supported me with confidence. My grat itude goes out to my old friends, Tse - Chuan, Aggie, Eunlyung, Li - Chen, and Rebekah, for their endless love and support , regardless of where we are, what we do, and whatever happens to us. Thank you to my fellow Spartans, Xiaomeng, Shannon, Jerry, and Yan, for their heartwarming friendship and all of the great memories in Berkey Hall. Last, but never least, this dissertation is dedicated to my parents, Dong - il Choi and Bohee Kim, and my brother Yeonsu with all my love and thanks. You raised me up whenever I fell down and wanted to give up. Without your unconditional love and belief in me, I would not be the person I am today and this achievement would not have been possible. I am truly blessed to have you as my loving and supportive family. vi TABLE OF CONTEN TS LIST OF TABLE S ................................ ................................ ................................ ...................... viii LIST OF FIGURES ................................ ................................ ................................ ..................... ix CHATER 1 ................................ ................................ ................................ ................................ ..... 1 INTRODUCTION ................................ ................................ ................................ ......................... 1 CHAPTER 2 ................................ ................................ ................................ ................................ .. 3 GRANDPARENTING AND MORTALITY: H OW DO RACE AND ETHNICITY MATTER? ................................ ................................ ................................ ................................ ......................... 3 Introduction ................................ ................................ ................................ ................................ . 3 Background ................................ ................................ ................................ ................................ . 5 Grandparenting, Health and Later Life, and Mortality ................................ ........................... 5 Race and Ethnicit y, Grandparenting, and Health and Mortality ................................ ............ 9 Data and Methods ................................ ................................ ................................ ..................... 12 Data ................................ ................................ ................................ ................................ ........ 12 Measures ................................ ................................ ................................ ................................ 13 Analytic Strategy ................................ ................................ ................................ ..................... 15 Re sults ................................ ................................ ................................ ................................ ....... 16 D escriptive Statistics ................................ ................................ ................................ .............. 16 Results from Event History Models ................................ ................................ ........................ 18 Discussion ................................ ................................ ................................ ................................ . 20 CHAPTER 3 ................................ ................................ ................................ ................................ 26 PTOMS I N SOUTH KOREA ....... 26 Introduction ................................ ................................ ................................ ............................... 26 Background ................................ ................................ ................................ ............................... 28 Theoretical Perspectives ................................ ................................ ................................ ........ 28 Empirical Evidence on Grandparenting and Mental Health ................................ ................. 29 Korean Grandmothers and Mental Health ................................ ................................ ............ 32 Present Study ................................ ................................ ................................ .......................... 33 Data and Methods ................................ ................................ ................................ ..................... 34 Data ................................ ................................ ................................ ................................ ........ 34 Measures ................................ ................................ ................................ ................................ 35 Analytic Strategy ................................ ................................ ................................ .................... 37 Resul ts ................................ ................................ ................................ ................................ ....... 39 Sample Characteristics ................................ ................................ ................................ .......... 39 Grandparenting and Depressive Symptoms Trajectories ................................ ...................... 40 Discussion ................................ ................................ ................................ ................................ . 43 CHAPTER 4 ................................ ................................ ................................ ................................ 47 F DEPRESSIVE SYMPTOMS IN CHINA ................................ ................................ ................................ ............. 47 vii Introduction ................................ ................................ ................................ ............................... 47 Background ................................ ................................ ................................ ............................... 48 Grandparenting and Mental Health ................................ ................................ ...................... 48 Mental Health Implications of Grandparenting in China ................................ ..................... 51 Data and Methods ................................ ................................ ................................ ..................... 53 Data ................................ ................................ ................................ ................................ ........ 53 Measures ................................ ................................ ................................ ................................ 54 Analytic Strategy ................................ ................................ ................................ .................... 57 Results ................................ ................................ ................................ ................................ ....... 58 Descriptive Statistics ................................ ................................ ................................ .............. 58 Grandpa renting and Trajectories of Depressive Symptoms ................................ .................. 60 Discussion ................................ ................................ ................................ ................................ . 62 CHAPTER 5 ................................ ................................ ................................ ................................ 69 CONCLUSION ................................ ................................ ................................ ............................ 69 APPENDICES ................................ ................................ ................................ ............................. 71 APPENDIX A: Chapter 2 Tables ................................ ................................ ............................. 72 APPENDIX B: Chapter 3 Tables ................................ ................................ .............................. 74 APPENDIX C: Chapter 4 Tables ................................ ................................ .............................. 76 APPENDIX D: Chapter 2 Figure ................................ ................................ .............................. 8 2 APPENDIX E: Chapter 3 Figures ................................ ................................ ............................. 83 APPENDIX F: Chapter 4 Figures ................................ ................................ ............................. 85 REFERENCES ................................ ................................ ................................ ............................ 87 viii LIST OF TABLES Table 2 - 1. Weighted Descriptive Statistics for Grandparents by Race, Health and Retirement Study, 1998 - 2014 ( N = 13,705) ................................ ................................ ................................ ... 72 Table 2 - 2. Hazard Ratios for Death by Grandparenting and Race ( N = 13,705) ......................... 73 Table 3 - 1. Weighted Descriptive Sta tistics for Grandmothers Aged 47 - 80, KLoSA, 2008 ( N = 2,184 ) ................................ ................................ ................................ ................................ ........... 74 Table 3 - 2. Growth Curve Estimates of Grandparenting on Trajectories of Depressive Symptoms among Grandmothers, KLoSA , 2008 - 2012 ( N = 2,814) ................................ ............................. 75 Table 4 - 1. Weighted Descri ptive Statistics, C HARLS , 2011 ( N = 5,691) ................................ .. 76 Table 4 - 2. Growth Curve Estimates of Grandparenting on Trajectories of Depressive Symptoms among Rural Grandparents , CHARLS, 2011 - 2015 ( N = 4,638 ) ................................ ................. 78 Table 4 - 3. Growth Curve Estimates of Grandparenting on Trajectories of Depr essive Symptoms among Urban Grandparents , CHARLS, 2011 - 2015 ( N = 1,053 ) ................................ ................ 80 ix LIST OF FIGURES Figure 2 - 1. Predicted Probabilities of Morta lity by Grandparenting and Race ........................... 82 Figure 3 - 1. Depressive Symptoms by Grandparenting, KLoSA, 2008 ................................ ....... 83 Figure 3 - 2. T rajectories of Depressive Symptoms by Grandparenting: Growth Curve Model Estimates ................................ ................................ ................................ ................................ ...... 84 Figure 4 - 1. Depressive Symptoms by Grandparenting and Region, CHARLS, 2011 ................. 85 Figure 4 - 2. Trajectories of Depressive Symptoms by Grandparenting among Rural Grandparent s : Growth Curve Model Estimates ................................ ................................ ................................ ... 86 1 CHAPTER 1 INTRODUCTION The health implications of intergenerational relationships in aging families have attracted attention given the significant increase in li fe expectancy over the past decades. Living longer and h ealthier than in the past enables older adults to actively engage in intergenerational relationships, including caring for grandchildren (i.e., grandparenting) (Margolis 2016). A fast - increasing number of older adults have been providing grandparenting eit her as a custodial or secondary caregiver. This phenomenon is attributed to sociodemographic transitions including economic recession (Fuller - T homson, Minkle r and Driver 1997; Silverstein and Giarrusso 2010; Casper et al. 2016). The growing trend of grandparenting is not limited to the U.S. Rather, it is widely found in various sociocultural contexts including Asia, where older adults have long supported grand child care, stemming from traditional family values and strong lineage solidarity (Bak er, Silverstein and Putney 2008; Mehta and Thang 2012). There has been little research investigating the consequences of grandparenting on health in later life, despite t he increased prevalence of grandparenting among older adults. Although the body of studies examining the association between grandparenting and later health is growing, the findings are inconclusive. Grandparenting has both positive and negative effects on older provided (B lustein, Chan, and Guanais 2004; Chen et al. 2014; Hughes et al. 2017). However, little research uses nationally representative samples of older a dults to comprehensively measure grandparenting utilizing both family structure (i.e., in what circumstance the caregiving is 2 provided) and care intensity (i.e., hours spent on the caregiving). Even less is known about how sociocultural background influenc es the grandparenting - health relationship (Choi and Zhang 2018; Cong and Silverstein 2008). In this dissertation, I present three essays that address three separate, but related, research questions to inform how grandparenting affects physical and mental health in later life and how the health implications of grandparenting vary by sociocultural context. In the first study, I not only examine how grandparenting is associated with mortality risk in the United States , but also assess how the link between gra ndparenting and mortality differs by race/ethnicity. My second study asks how grandparenting affects depressive symptoms in older adults in South Korea. Specifically, I focus on the experience of grandmothers, who make up the majority of caregiving grandpa rents, based on the gender gap in caregiving roles in the Korean context. I also investigate the extent to which potential mechanisms explain the association. In the third study, I assess how grandparenting is linked to depressive symptoms in older adults in China. Given the substantial rural - urban gap in various socioeconomic dimensions in the Chinese context, I consider the experience of grandparents in both rural and urban regions to identify the rural - urban differences in the health consequences of gran dparenting. 3 CHAPTER 2 GRANDPARENTING AND MORTALITY: HOW DO RACE AND ETHNICITY MATTER? Introduction The active roles of older adults in families and intergenerational relationships have attracted attention as life expectancy has substantially increased in the U.S. Caring for grandchildren (hereafter, grandparenting) has been one of the emerging roles of older adults in - force participation and single parent families in the adult children generation (Hofferth 1996) . One of the foremost issues in terms of grandparenting is the increasing number of grandparents who raise their grandchildren as custodial caregivers. Approximately 7.3 million older adults lived with grandchildren and over 35% of them wer e the primary caregivers for their grandchildren in 2015 (U.S. Census Bureau 2017) . These grandparents perform the primary caregiving on behalf of their adult children who are incapable as a result of the drug epidemic, the rise in the female incarceration rate, and the economic recession over the past few decades (Fuller - T homson, Minkler and Driver 1997; Keene and Batson 2010) . Adding to this, in terms of the older adults who provide temporary noncoresident grandparenting (i.e., babysitting), it is apparen t that a considerable number of older adults have engaged in grandparenting in later life (Harrington Meyer and Kandic 2017). Despite this growing phenomenon, the health implications of grandparenting for older adults have been underexplored. Moreover, sur prisingly little literature has addressed the consequences of grandparenting for older adult mortality despite its strong association with health. 4 It is a notable demographic transition that the older population in the U.S. is becoming more racially/ethn ically diverse. Given the growth of the overall minority population and greater longevity, the proportion of minority older adults, which comprised 17.5% of the older population in 2003, is projected to increase to 28.5% in 2030 (Administration on Aging 20 14) . Racial/ethnic minorities present fairly different grandparenting patterns than white grandparents. Grandparenting is a more prevalent and normative role for minority older adults as one way of supporting their family (Burton and Devries 1992; Yancura 2013) . Coresidence with grandchildren and having primary responsibility for the grandchildren are also more prevalent among minority grandparents (Minkler and Thomson 2005). In addition to the lower socioeconomic status (hereafter, SES) on average, grandpa renting may have different meanings and implications for health and mortality for those groups. Yet research on racial/ethnic variations in grandparenting and later health has been sparse. To my knowledge, this study is a first attempt to assess racial/eth nic differences in the link between grandparenting and older adult mortality in the U.S. using population - level longitudinal data. This study aims to investigate how caregiving experience as a grandparent influences older adult mortality using data from nine waves of the Health and Retirement Study (1998 - health condition and behaviors, contribute to mortality differences among grandparents. The second objective of this study is to examine racial/ethnic differences in these associations. I compare the effects of grandparenting on mortality for white, black, and Hispanic grandparents and highlight the importance of social/cultural contexts in understanding racial/ethn ic variations in intergenerational relationships and later health. 5 Background Grandparenting, Health in Later Life, and Mortality In recent years, researchers have attempted to understand the implications of grandparenting for physical and mental health in later life as an increasing number of older adults have become involved in caregiving for grandchildren. However, little is known about the association between grandparenting and older adult mortality. Given that health condition significantly contribut es to mortality, it is plausible that grandparenting also influences older adult mortality. Only a few studies have examined the relationship between grandparenting and mortality risk, and most of these studies were carried out in Europe. Hilbrand et al. (2017) recently found that grandparents who provide non - custodial care for grandchildren are more likely to survive than non - caregiving grandparents and non - grandparents in Germany. A mortality advantage is also found among those who provide child care to non - family members. Christiansen (2014) has suggested that, in Norway, entering into grandparenthood at an early age has an adverse effect on mortality for men. For women, becoming grandmothers at an early age increases mortality risk, whereas those who b ecome grandmothers after middle age (i.e., age 50) report lower mortality. However, it would be inappropriate to apply the prior findings to the U.S. context because the studies are based on Western Europe only. Moreover, the measure of grandparenting in t hese studies is either simply non - custodial child care versus no care (Hilbrand et al. 2017) or not directly related to caregiving for grandchildren, but rather being a grandparent per se (Christiansen 2014) . More importantly, neither of these studies addr essed racial/ethnic variations in the link between grandparenting and older adult mortality, which is crucial for the U.S. context because minority older adults in the U.S. tend to take care of grandchildren more 6 intensively than whites and live with grand children more often due to lower SES (Minkler 2005; U.S. Census Bureau 2014) . Grandparenting, Social role, and H ealth A growing body of literature has investigated the health consequences of grandparenting in later life to understand the new or changing role of older adults in families. Two competing theoretical perspectives on social roles serve to account for the effects of grandparenting on health. The role enhancement theory supports a positive relationship between grandparenting and lth. The theory suggests that individuals feel more satisfaction and experience greater well - being when they carry out various social roles and get a sense of achievement and social support from those activities (Moen, Robison and Dempster - McClain 1995) . I n the case of grandparenting, older adults get the opportunity to feel a sense of achievement from caring for their grandchildren as well as increased physical and emotional interactions with family members. While performing an active grandparental role, o lder adults are more likely to be physically active and have feelings of reward, life satisfaction, and purpose (King, Rejeski and Buchner 1998; Pruchno and McKenney 2002; Rozario, Mo rrow - Howell and Hinterlong 2004; Szinovacz and Davey 2006) . An alterna health. The role strain theory argues that managing multiple role obligations results in a strain on individuals, and that limited resources (e.g., time, goods, emotional capa city, etc.) exacerbate the difficulty (Goode 1960) . A high level of stress stemming from role strain leads to psychological distress (Barnett and Baruch 1985; Pearlin 1989) . Over - demanding role obligations related to grandparenting may cause stress and fat igue for grandparents and possibly increase the likelihood of negative health outcomes. Grandparenting may also act as a source of stressful 7 disagreement over child care or insufficient time and resources to share. Furthermore, it is plausible that a lack of personal time for healthy behaviors (e.g., regular exercise) and social h (Choi and Zhang 2018) . Empirical studies on grandparenting and later health have been few and presented mixed psychological health. The measures of grandpa renting have been various, but primarily classified family structure, grandparenting arrangement has three main types: noncoresident, skipped - generation househ old, and multigenerational household grandparenting, which correspond respectively to grandparents living alone as a couple, a grandparent - headed family without adult children, and an extended family with three or more generations of household members. Old er adults, especially grandmothers, who provide noncoresident grandparenting (i.e., babysitting) experience better self - rated health and a lower level of depressive symptoms (Hughes et al. 2007) . The health benefits of noncoresident grandparenting are attr ibuted to healthy behaviors (e.g., exercise), emotional supports, and a feeling of role fulfillment while caring for grandchildren, consistent with the role enhancement theory. On the other hand, the association between skipped - generation household grandp arenting and health is well known to be negative (Silverstein and Giarrusso 2010) . Researchers have found that, compared to non - caregiving grandparents, older adults providing grandparenting in a skipped - generation household experience a health deficit, in cluding worse self - rated health (Hughes et al. 2007) , a higher level of depressive symptoms (Hughes et al. 2007; Szinovacz, 8 DeViney and Atkinson 1999) , and greater frailty (Chen et al. 2014) . Older adults providing skipped - generation household grandparenti ng are most likely doing so because of the absence of the adult children generation for voluntary or involuntary reasons (e.g., illness, death, incarceration, marital dissolution, or unemployment). Elevated stress and physical burden as a custodial caregiv er account for the detrimental effect of skipped - generation household grandparenting, the scant existing research reports that it is associated with greater frailty for whi te older adults since the extended family is not a common family structure in the U.S. or Western societies (Chen et al. 2014). g randparenting intensity , measured as total hours of child care, are also mixed. Some amount of caring for grandchildren has a protective effect on cognitive functioning for verbal fluency (Arpino and Bordone 2014) and the frailty index (Chen et al. 2014) . In particular, a moderate level of grandparentin g is beneficial for health. Grandmothers who provide moderate grandparenting suffer less from functional limitations and depressive symptoms (Grundy et al. 2012; Hughes et al. 2007) . Supporting the role strain perspective, however, intensive grandparenting is inversely associated with health. Providing highly intensive grandparenting over time predicts deteriorating physical health and increasing stress levels for grandmothers due to the demanding nature of the responsibility (Musil et al. 2011) . It is imp ortant to point out that previous literature has mixed the concepts of grandparenting intensity and grandparenting arrangement. For instance, most studies treat intensive grandparenting and skipped - generation household grandparenting interchangeably. Yet, an intensive level of grandparenting is possible in combination with other arrangements, 9 including noncoresident grandparenting (Choi and Zhang 2018). Some studies differentiate grandparenting arrangement and intensity, but include both concepts as separat e indicators (Chen et al. 2014; Hughes et al. 2007) . Guided by the theoretical perspectives, I combine the concepts of grandparenting arrangement and intensity together and propose the following hypotheses: Hypothesis 1 : Grandparents who provide light or m oderate noncoresident grandparenting have a lower risk of mortality than non - caregiving grandparents. Hypothesis 2 : Grandparents who provide intensive noncoresident grandparenting have a higher risk of mortality than non - caregiving grandparents. Hypothesis 3 : Grandparents who provide skipped - generation household grandparenting have a higher risk of mortality than non - caregiving grandparents. Hypothesis 4 : Grandparents who provide multigenerational household grandparenting have a higher risk of mortality tha n non - caregiving grandparents. Race and Ethnicity, Grandparenting, and Health and Mortality Systematic discussion on racial/ethnic variations in grandparenting and later health and mortality has been sparse. There have been, however, recent attempts to st udy grandparents from more diverse racial/ethnic backgrounds (Grundy et al. 2012; Pruchno and McKenney 2002) . The patterns and meanings of grandparenting are different for minority older adults than for whites. Minority grandparents more often live with g randchildren (and/or adult children) in the same household and even take the role of primary caregiver for grandchildren (Minkler and Thomson 2005). This is attributed to the fact that, for minority grandparents, caring for grandchildren is regarded as mor e normative, given their cultural practices, and as more 10 , given their lower SES (Burton and Devries 1992; Yancura 2013) . For instance, black families have a long tradition of shared childrearing across generations in t he face of external adversity, such as subjugation, racism and discrimination, poverty, and single parenthood (Uhlenberg and Kirby 1998). A notably high percentage of black grandparents, compared to other racial/ethnic groups, provide skipped - generation ho usehold grandparenting (Bryson and Casper 1999) . Hispanics also tend to exchange more frequent and direct support and transfers among generations based on their strong family values and ties (Kataoka - Yahiro, Ceria and Caulfield 2004) . Hispanic older adults are more likely to adopt a multigenerational family arrangement (Keene and Batson 2010), and interactions between Hispanic grandparents and grandchildren reproduce Hispanic family culture (Facio 1996; Raphael 1989) . Most prior research on grandparenting and health has included race/ethnicity as only one of the control covariates to predict later health (Hughes e t al. 2007; King and Elder 1998; Luo et al. 2012b) . Other literature has examined how grandparenting, especially grandparenting arrangement, as op posed to care intensity, affects health for particular racial/ethnic groups. Studies on black grandparents have heavily concentrated on their practice of skipped - generation household grandparenting. In general, black older adults who provide skipped - genera tion household grandparenting experience deteriorating health conditions, including depressive symptoms (Burton 1992) and a higher level of frailty (Chen et al. 2014) . The adverse consequences of skipped - generation household grandparenting for black older be due to a high level of stress from the caregiving burden (Pruchno and McKenney 2002) and lower SES (Chen et al. 2014) . 11 There is a dearth of studies on Hispanic grandparents and their health. Past literature has suggested that grandpa renting is generally beneficial for the later health of Hispanic older adults. Hispanic grandmothers report a lower level of depressive symptoms when providing noncoresident grandparenting (Grundy et al. 2012) . Compared to white and black grandparents who live with and care for grandchildren, caregiving Hispanic grandparents do not experience a health deficit. Multigenerational household grandparenting has no detrimental impact on - family ho useholds directly or indirectly take care of grandchildren on a daily basis (Chen et al. 2014) . The positive and normative value of grandparenting and the cultural norm of a multigenerational family setting explain the positive health implications of grand parenting for Hispanic o lder adults (Grundy et al. 2012; Chen et al. 2014). In light of the current growth of the minority older population in the U.S., especially Hispanics, it is necessary to take diverse racial/ethnic backgrounds into account and explor e their implications for well - being in later life. The consequences of grandparenting for mortality risk or health presumably vary across racial/ethnic groups because of racial/ethnic differences in family culture and behaviors (Keene and Batson 2010; Mink ler and Fuller - Thomson 2005; Mouzon 2013) . Taking these factors together, I expect that: Hypothesis 5 : Among black, grandparents providing skipped - generation household grandparenting have a higher risk of mortality than non - caregiving grandparents. Hypothe sis 6: Among Hispanic, grandparents providing multigenerational household grandparenting have a lower risk of mortality than non - caregiving grandparents. 12 Data and Methods Data I used the nine waves (1998 - 2014) of the Health and Retirement Study (HRS) data to examine the relationship between grandparenting and mortality and how the link differs by race. The HRS is a longitudinal study of a nationally representative sample of 21,384 Americans over age 50 in the 1998 wave. The survey has biennially collected d changes in physical and mental health factors including mortality, health behaviors, family conditions, labor force participation, and financial situation. Data for these analyses came from raw HRS data, RAND HRS data, and the Tracker file of the HRS. The RAND data are the cleaned and streamlined versions of HRS data developed by the RAND Center for the Study of Aging. The Tracker file is a cleaned version which facilitates the use of HRS data across waves. The HRS oversampled two racial/ethnic minority groups, blacks and Hispanics. Asians and other races were not included due to insufficient sample size. The analytic sample in this study includes 13,705 non - Hispanic whites (hereafter, whites), non - Hispanic blacks (hereafter, b lacks), and Hispanics. I restricted the sample to respondents aged 50 80 who had grandchildren and who completed questions on grandparenting and race/ethnicity in 1998. ts starting at age 50 not only to utilize more cases in the data but also to take into account the childcare experience of young grandparents. I set the upper age limit at 80 years because the oldest older adults are less likely to take care of grandchildr en because of declining health (Hughes et al. 2007; Ku et al. 2013) . 13 Measures Mortality . The outcome variable used in this study is older adult mortality, which indicates whether a respondent died from any cause in the follow - up surveys (death = 1). Death information was extracted from the Tracker file. I considered only respondents having valid data for the time of death. Of the 13,705 respondents aged 50 80 in 1998, 3,351 about 25 % were reported to have died by 2014. Grandparenting. Grandparenting, the k ey independent variable, assesses older their family structure and caregiving intensity to take into account the caveat from prior studies using only one of these information and the total hours of child care provided for grandchildren that respondents reported for the previous two years, I measured grandparenting as a time - varying categorical variable. T he cate gories are no grandparenting ( reference), skipped - generation household grandparenting, multigenerational household grandparenting, and noncoresident grandparenting. I classified older adults who reported providing no child care for grandchildren int o the no - grandparenting group regardless of family structure. Among older adults who provided care, those living with grandchildren only (without adult children) were categorized as providing skipped - generation household grandparenting. Caregiving grandpar ents living with adult children as well as grandchildren were categorized as providing multigenerational household grandparenting. Grandparents who cared for grandchildren but lived in a separate household were classified as providing noncoresident grandpa renting. I further categorized noncoresident grandparenting into two sub - groups based on caregiving intensity: light/moderate noncoresident grandparenting (0 499 hours during the 14 previous two years) and intensive noncoresident grandparenting (over 500 hour s). I did not apply this sub - categorization to the skipped - generation household grandparenting and multigenerational household grandparenting groups due to small sample sizes. Only a small number of coresident grandparents reported their time spent on chil d care, possibly because living together with grandchildren makes it hard to count the exact number of hours of caregiving during the previous two years. Health condition . Health condition includes three sets of time - varying predictors. Self - rated health is coded 1 for fair or poor health. I utilized a short version of the Center for Epidemiologic Studies - Depression (CES - D) scale, ranging from 0 to 8, to measure depressive symptoms. Chronic disease is included as a dichotomous variable (yes = 1), indicatin g whether respondents had been diagnosed with one or more chronic conditions (i.e., hypertension, diabetes, stroke, lung disease, heart disease, or cancer) among the leading causes of death in the U.S. (Heron 2013) . SES . SES is measured using education, h ousehold income, net household wealth, employment, and health insurance. Education is a continuous variable constructed from the - varying continuous variables, adjusted as l ogged. I also created dummy variables for employment (working = 1) and having long - term health insurance (yes = 1). All of these variables except for household income and net household wealth were measured at baseline. Health behaviors . I include three dimensions of health behaviors. I dichotomized smoking (currently smoking = 1) and drinking (currently drinking = 1), and both are time - varying variables. Exercise was coded as 1 for those who did not participate in vigorous physical activity three or more times a week in the previous year ( reference) and 0 for those who did. 15 Exercise was measured at baseline only because the question on exercise was inconsistent over the waves. Controls . I controlled for sociodemographic characteristics, including gender, race, age, marital status, and nativity. Gender is a dichotomous variable (women = 1). I coded self - reported race into three categories: white ( reference), black, and Hispanic. Age is measured as a time - varying continuous variable. Marital status (married = 1) and nativity (foreign born = 1) are dummy variables. Analytic Strategy This study employs a complementary log - log model to investigate the relationship between grandparenting and older adult mortality over sixteen years, 1998 - 2014. The complementary l og - log model can specify how the discrete - time hazard depends on time and explanatory variables. In addition, the model is also directly equivalent to the Cox proportional hazards model in continuous time since the complementary log - log function builds in a proportional hazards assumption (Allison 2014) . The complementary log - log model is specified as: where is the conditional probability of death for individuals at wave (1 - 9). represents the complementary log - log transformation of the baseline hazard. indicates the effects of time - invariant covariates, and is the effects of time - varying covariates. Using person - year record files f or two - year intervals from 1998 to 2014, I estimated a series of nested complementary log - log models to test my hypotheses. I first estimated the main 16 effect of grandparenting on mortality, net of sociodemographic characteristics. I then introduced health condition to adjust for the selection effect that healthier older adults are more likely to care for grandchildren. Next, I added multiple indicators (SES and health behaviors, respectively) to assess whether and to what extent they explained mortality ris k as a function of grandparenting. Finally, I included the interaction term of race and grandparenting, net of all covariates, to evaluate the racial gap in mortality explained by grandparenting. Missing data for the independent variables ranged from none to approximately 17% for grandparenting. In order to retain all cases, I imputed missing values with multiple imputation by chained equations in Stata 14 (Young and Johnson 2015) . Results were based on estimates from 10 imputed data sets. I applied weights to adjust for the complex sampling design of the HRS. Results Descriptive Statistics I provide weighted descriptive statistics for the older adults in Table 1. The results are reported by race/ ethnicity to show the differences among whites, blacks, and Hi spanics. Significant differences among racial groups are indicated with an asterisk. Table 2 - 1 shows that 24.89% of white grandparents and 24.87% of black grandparents died between 1998 and 2014 . The proportion of death is significantly smaller for Hispani c grandparents (18.78%) in comparison with white grandparents. In terms of grandparenting, the majority of older adults do not take care of any grandchildren. The proportion of non - caregiving grandparents is especially high among whites (80.32%), although not significantly higher than among blacks or Hispanics. However, caregiving grandparents show different grandparenting patterns by race/ethnicity. For whites, the 17 second largest group are those who provide light/moderate noncoresident grandparenting (10. 54%), followed by intensive noncoresident grandparenting (4.87%), multigenerational household grandparenting (2.81%), and skipped - generation household grandparenting (1.47%). By contrast, among blacks, 8.23% of grandparents provide multigenerational househ old grandparenting the second most common care type. The light/moderate noncoresident grandparenting group accounts for 7.52%, followed by skipped - generation household grandparenting (4.66%) and intensive noncoresident grandparenting (3.84%). Compared to w hite grandparents, significantly more black grandparents live with either adult children and grandchildren or with grandchildren only. As for Hispanics, 7.62% provide light/moderate noncoresident grandparenting. The next largest groups are multigenerationa l household grandparenting (6.85%) and intensive noncoresident grandparenting (4.11%). The skipped - generation household grandparenting group represents 2.80% of Hispanics. Like black grandparents, Hispanic grandparents tend to reside with younger generatio ns including grandchildren more often than whites, and the difference is statistically significant. The general health conditions (i.e., self - rated health, CES - D, and chronic disease) of blacks and Hispanics are significantly worse than those of whites. SE S also differs by race in my sample. Black and Hispanic older adults, as expected, report lower SES than whites in all dimensions, including education, household income, net household wealth, and long - term health insurance. With respect to health behaviors , the racial gaps are mixed: blacks and Hispanics are less likely to exercise vigorously than whites, but currently drink less than whites. Blacks currently smoke more than whites. Last, looking at the control variables, significantly more black grandparen ts are female. Blacks and Hispanics are more likely to be younger and less likely to be married than whites. More black and Hispanic grandparents are foreign born compared to whites. 18 Results from Event History Models As a next step, I estimated a series o f complementary log - log models to investigate the association between grandparenting and older adult mortality and whether and how the linkage differs by race/ethnicity. Table 2 - 2 presents results from the six models. The results from Model 1 of Table 2 - 2 indicate that older adults who provide any level of noncoresident grandparenting have significantly lower mortality than older adults who provide no caregiving, net of control variables. Specifically, the hazard of death for those who provide light/moderat e noncoresident grandparenting is approximately 20% lower than for non - caregiving grandparents. Grandparents who provide intensive noncoresident grandparenting have a 19% lower hazard of death compared to their non - caregiving counterparts. Model 2 includ es health condition. The significant effect of grandparenting on mortality holds even after controlling for various health statuses. The lower mortality risk for those who provide any noncoresident grandparenting remains, and the magnitude of the hazard ra tios slightly increases. I add SES to the association between grandparenting and mortality in Model 3 of Table 2 - 2 . The results show that the hazard of death for older adults who provide light/moderate noncoresident grandparenting slightly increases (from .83 to .89), while the hazard for those who provide intensive noncoresident grandparenting remains similar (changing from .84 to .85), net of health condition, SES, and controls. The decreasing effect of noncoresident grandparenting on mortality mainly de rives from its positive association with several SES indicators: education, household income, net household wealth, and currently working, which are negatively related to mortality as predicted. 19 When health behaviors are introduced in Model 4, the result s indicate that the protective effects of noncoresident grandparenting on mortality still exist. The hazard of death for the light/moderate and intensive noncoresident grandparenting groups (.85 and .85, respectively) remains similar to that in Model 2, co ntrolling for health condition and behaviors as well as controls. I next include all variables in the full model (Model 5). The impacts of noncoresident grandparenting are still significant, net of all covariates. Grandparents providing a light/moderate level of noncoresident grandparenting have a 10% lower hazard of death compared to non - caregiving grandparents. The hazard of death for grandparents providing intensive noncoresident grandparenting is 14% lower than that of non - caregiving grandparents. T he final model (Model 6 of Table 2 - 2 ) adds an interaction term to further examine whether and how the association between grandparenting and mortality varies by racial/ethnic groups. The results show that several types of grandparenting interact significan tly with being black. The mortality risk is significantly increased among black grandparents who provide skipped - generation household grandparenting (OR=1.51*0.87=1.31), multigenerational household grandparenting (OR=1.50*0.84=1.26), and intensive noncores ident grandparenting (OR=1.31*0.82=1.07). For Hispanics, however, only the mortality difference between intensive noncoresident grandparenting and non - grandparenting is marginally significant. To better illustrate the joint effects of grandparenting and race on mortality, Figure 2 - 1 presents the predicted probabilities of mortality, adjusting for all covariates in Model 6 of Table 2 - 2 . The patterns are different among races. Providing any type of grandparenting except for skipped - generation household gran dparenting The probability of mortality for whites is the highest among those without grandparenting 20 experience but the lowest among those who provide intensive noncoresident grandparenting. In contrast, f or blacks, the mortality probability is the highest for the skipped - generation household grandparenting group and the lowest for the light/moderate noncoresident grandparenting group. hose who provide skipped - generation household grandparenting and multigenerational household grandparenting. The mortality of Hispanic grandparents is consistently the lowest in comparison with white and black grandparents, but does not significantly diffe r by grandparenting type. Finally, I note that all basic control variables are significant in the expected directions in all six models. Women have a lower hazard of death than men. As age increases, the hazard of death also increases. Those who are marri ed and foreign - born have a lower hazard of death than non - married and U.S. - born individuals, respectively. Discussion The present study is among the first to examine the association between grandparenting and older adult mortality using population - level l ongitudinal data. The first aim of this study is to examine how grandparenting experience influences mortality. The second aim is to further investigate whether and how the linkage between grandparenting and mortality differs by race/ethnicity, to better u nderstand the consequences of contemporary grandparenting for various racial/ethnic groups. Six hypotheses are proposed on the basis of previous literature that argues for both positive and negative health implications of grandparenting depending on caregi ving In terms of the general mortality implications of grandparenting, I first hypothesize that grandparents who provide light/moderate noncoresident grandparenting have a lower mortality 21 risk than their non - grandparenting counterparts. The results support this hypothesis. This finding is consistent with a recent study that finds a beneficial impact of noncoresident grandparenting (Hilbrand et al. 201 7) . The result is also broadly in line various physical and mental health outcomes (Chen et al. 2014; Hughes et al. 2007; Ku et al. 2013) . The role enhancemen t theory may explain why grandparenting plays a protective role in - being by boosting positive feelings, such as self - efficacy, self - esteem, and life satisfaction (Moen, R obison and Dempster - McClain 1995) . Accordingly, grandparents may obtain emotional reward from taking on an additional caregiving role, grandparenting, in their later life. Grandparenting - being because of their inc reased interactions with grandchildren and adult children during child care (Pruchno and McKenney 2002; Rozario, Mo rrow - Howell and Hinterlong 2004; Szinovacz and Davey 2006) . A decent level of healthy behavior, such as physical activity and a regular and n utritious diet, while caring for grandchildren also decreases health and mortality risks (King, Rejeski and Buchner 1998) . Hypothesis 2 stated the contrasting expectation that grandparents who provide intensive noncoresident grandparenting have a higher m ortality risk than non - caregiving grandparents. I do not find any supporting evidence. The results instead show that the mortality risk for older adults providing intensive noncoresident grandparenting is lower than that of their non - grandparenting counter parts. Although very little is known about the mortality implications of intensive noncoresident grandparenting, the finding here is inconsistent with extant literature that suggests an inverse relationship between highly intensive grandparenting and older (Musil et al. 2011) , which may ultimately predict mortality. The inconsistency is perhaps due to the 22 buffering effect of the family structure of these grandparents. Compared to older adults who provide skipped - generation household and multi generational household grandparenting, it is possible for older adults providing noncoresident grandparenting to have a respite from child care and have time for their own needs and leisure (Choi and Zhang 2018 ) . Thus, noncoresident grandparents may experi ence less physical burden and stress from their caregiving despite its high intensity and still benefit from grandparenting with a lower mortality risk. In addition, grandparenting intensity seems to affect mortality risk little, compared to grandparenting arrangement, given that both light/moderate and intensive noncoresident grandparenting have effects of the same direction and similar magnitude. The third hypothesis in this study is that older adults providing skipped - generation household grandparenting experience a higher risk of mortality than their non - caregiving counterparts. I do not find any significant evidence to support the hypothesis. The result is surprising in light of many previous studies reporting an inverse relationship between skipped - gen eration household grandparenting and health in later life. The inconsistency could be because of the prior conditions of caregiving grandparents rather than the grandparenting practice itself. mic resources explain older (Hughes et al. 2007) . I also hypothesize that grandparents who offer multigenerational household grandparenting have a higher mortality risk than non - caregiving grandparents (Hypothes is 4), but no significant supporting evidence is found. Next, I turn to racial/ethnic variations in the linkage between grandparenting and older adult mortality. In light of the previous literature on grandparenting and health, I expect higher mortality fo r black grandparents who provide skipped - generation household grandparenting than 23 for non - caregiving grandparents (Hypotheses 5). I also expect lower mortality for Hispanic grandparents providing multigenerational household grandparenting than for their no n - caregiving counterparts (Hypotheses 6). The findings are mixed. Supporting Hypothesis 5, black older adults providing skipped - generation household grandparenting have a higher mortality risk than the non - caregiving group. Moreover, multigenerational hous ehold grandparenting and intensive noncoresident grandparenting are also associated with a higher risk of mortality for black grandparents. As for Hispanics, however, I did not find any significant mortality differences among grandparenting groups. I also note that the results indicate that white grandparents providing multigenerational household grandparenting and light/moderate or intensive noncoresident grandparenting have a lower mortality risk relative to their non - caregiving counterparts. In sum, the results indeed suggest significant racial/ethnic variations in the association between grandparenting and mortality. The robust mortality benefits of grandparenting throughout the models mainly from light/moderate and intensive noncoresident grandparenting are driven by the white grandparents, the majority group, even net of SES and health condition and behaviors. White grandparents enjoy significantly longer periods of healthy grandparenthood than minority grandparents (Margolis and Wright 2017) . Furthermo re, direct or indirect support from/to family members, a healthy lifestyle, and the psychological satisfaction derived from consistent intergenerational relationships through grandparenting without the stressful responsibility of being the primary caregive r may contribute to the lower mortality risk for white grandparents (Hays lip, Blumenthal and Garner 2014; Pruchno and McKenney 2002) . On the other hand, most types of caregiving for grandchildren leads to mortality disadvantages for black older adults. In addition to the physical burden and stress from child care (Pruchno and McKenney 2002) , this is 24 caregiving grandparents considerably reflect their SES and prior hea lth, rather than being a consequence of grandparenting itself (Chen et al. 2014; Hughes et al. 2007) . Black grandparents tend to be poor and unhealthy compared to whites ( Burton and Devries 1992 ). The lower SES of black older adults and their families is o ne of the primary reasons why black grandparents are more likely to take care of grandchildren intensively in either a skipped - generation household or multigenerational household setting, or even when not coresiding with their adult children or grandchildr en ( Bryson and Casper 1999; Burton and Devries 1992; Uhlenberg and Kirby 1998 ) . Thus, black caregiving grandparents on the whole seem to be unable to enjoy the health benefits of grandparenting due to their disadvantaged social/individual status. I conduct ed several sensitivity analyses to ensure that the findings are robust (results not shown but available upon request). First, I further categorized the intensity of caregiving for noncoresident grandparenting into three groups (light, moderate, and intensi ve care) and implemented the same models. The analytic results support the main findings above. Next, I analyzed whether social support (e.g., social activities with family and friends, or in a community) affects the relationship between grandparenting and mortality. Social support has no significant effect, and the effect of grandparenting still holds. Lastly, given the high percentage of foreign - born individuals in the Hispanic sample (approximately 53%), I tested whether nativity moderates the racial/eth not show any statistically significant three - way interaction effects. The present study is not without limitations. First, I have small samples of minority older adults for some grandparentin g groups. Although the HRS oversamples black and Hispanic populations, particular categories (i.e., the skipped - generation household grandparenting and 25 intensive noncoresident grandparenting groups) still contain a relatively insufficient number of observa tions. This data limitation may prevent me from detecting a significant impact of such grandparenting types on mortality for Hispanics. Second, I may be unable to entirely eliminate a selection effect for caregiving grandparents. Healthier grandparents are more likely to take care of grandchildren and survive longer than their non - caregiving counterparts (Hilbrand et al. 2017), and there may be racial/ethnic differences in those patterns of grandparents. Although I control for various health indicators (hea lth condition and health behaviors) at baseline and as time - varying, the results may need to be interpreted with caution. Despite these limitations, this study is important as one of the first to investigate how grandparenting and older adult mortality ar e associated and whether and how the linkage differs by race/ethnicity. I contribute to the current discussion on grandparenthood by finding mixed effects of caring for grandchildren on mortality by racial/ethnic groups using a nationally representative lo ngitudinal data set. The mortality advantage from grandparenting is robust for white older adults. A mortality disadvantage from grandparenting is mainly found among black grandparents. Given the increasing diversity of the older population in the U.S., th is study emphasizes the importance of understanding how social/cultural contexts for each racial/ethnic group shape their intergenerational relationships and well - being in later life. 26 CHAPTER 3 Introduction Over the past few decades , researchers have increasingly sought to understand how significantly increased life expectancy affect s role s in aging families. As older adults live longer and sta y healthier than they had in the past, their grandparenthood, a period when older adults can interact with their grandchildren, has also been extended (Margolis 2016) . An increasing number of grandparents have become engaged in caring for grandchildren (i. e., grandparenting) during later life (Baker, Silverstein a nd Putney 2008; Silverstein and Giarrusso 2010) . Some older adults even provide very intensive care for their grandchildren depending on the family situation, especially their eds increasing work hours (Casper et al. 2016; Cherlin 2010) . Th e se phenomena are found in various sociocultural settings around the world since grandparents, specifi cally grandmothers, have long played an important role in supporting child care with in their families (Baker, Silverstein and Putney 2008) . South Korea (hereafter, Korea) is one of the countries that ha s achieved a rapid increase in life expectancy over the past several decades. Indeed, the life expectancy at birth for Korean women is 84.6 years old , which rank ed eighth in the world in 2012 (World Health Organization 2014) . As their healthy later life has been extended, Korean older women have been more a ctively engaged in their family care as caregiver s . As in the U.S. or other western countries, one of the popular caregiving roles of Korean older women is grandparenting. An increasing number 27 of grandmothers have provided an intensive level of daily child care for their grandchildren , even when their grandchildren and adult children do not reside with them (Korea Institute of Child Care and Education 2015) . The unique facet of Korean grandparenting is derived from the situation wherein older women usually take care of grandchildren to support their dual - income adult children. Traditional family culture emphasiz ing intra - family support and the absence of sufficient quality daycare centers are the primary reasons why Korean older women are more frequently ass ociated with the grandparenting role. There have been few studies that investigate the consequences of grandparenting for - being and health in Korea despite the significant growth of the grandparenting trend. Little is known ab out the health consequence s of Korean grandparenting, especially for mental health, though the findings from other cultural contexts have reported both (Hughes et al. 2007; Ku e t al. 2013; Szinovacz, DeViney and Atkinson 1999) . In addition, the majority of prior research on Korean grandparenting has focused only on grandparents who provide skipped - generation household grandparenting (i.e., care in grandparent - grandchildren only h ouseholds) and their excessive caregiving demands (Kim 2009 ; Lee and Han 2008; Park 2010) . It is unknown whether other types of grandparenting have significant effects on health. Finally, most prior studies have utilized either cross - sectional regional dat a (Bae 2007; Baek 2009; Choi and Cha 2013; Kang 2011) or qualitative data (Kim and Seo 2007) . The aim of the present study is to untangle the association between grandparenting and mental health, particularly depressive symptoms , in the Korean context. F irst, drawing on the Korean Longitudinal Study of Aging (2008 - 2012), I examine whether grandparenting influences the trajectory of depressive symptom s over time among older women. Second, I investigate the 28 extent to which socioeconomic status (hereafter, S ES), health behaviors, and social support account for th at association. Background Theoretical Perspectives Grandparenting is one of the new social roles for older adults living longer and healthier later li ves . In order to understand how grandparenting two competing theoretical perspectives are often used: the role enhancement theory and the role strain theory. The role enhancement theory suggests that executing multiple social roles simultaneously contributes - being, which also promotes mental health. Those who have different social roles enjoy a greater sense of fulfillment and satisfaction cumulated from carrying out each role (Moen, Robison and Dempster - McClain 1995) . In light of this per spective, grandparenting is associated with better mental health. Older adults who provide care for their grandchildren, a new additional role in life, obtain positive health - promoting emotions, including the sense of purpose in life, self - efficacy, and li fe satisfaction, while interacting with their grandchildren and their parents, the adult children (Pruchno and McKenney 2002; Rozario, Mo rrow - Howell and Hinterlong 2004; Szinovacz and Davey 2006) . Caregiving grandparents are likely to be more physically ac tive in tak ing care of grandchildren, which in turn ha s beneficial effects on mental health (King, Rejeski and Buchner 1998) . The role strain theory argues that, in cont rast to the role enhancement theory, simultaneously holding - being. The more role obligations one has, the more struggling and stress 29 property, have limits (Goode 1960) . As such, overburden ing tasks lead to stress and more psychological distress (Barnett and Baruch 1985; Pearlin 1989) . Based on the role strain theory, grandparenting can be a source of stress and poor mental status for o lder adults. Given that grandparenting is not a mandatory responsibility of grandparents, older adults who care for grandchildren can experience strain from fulfilling th is additional role. Caregiving grandparents may find increasing role strain from manag ing concurrent social roles, such as spouse, parent, grandparent, friend, and other social positions. These over - demanding role obligations and strain s strain deteriora conflicts over child rearing with their adult children, less time with their spouse or other family members, and loss of time for leisure and participation in social activities (Bl ustein, Chan and Guanais 2004) . Both of these two theoretical perspectives explain the association between That said, the health consequences of grandparenting may depend on whether the strains from grandchil d care outweigh the physical and mental benefits of the caregiving experience. Empirical Evidence on Grandparenting and Mental Health Research on grandparenting has been limited and the majority of the studies are predominantly based on cases in the U.S . and European countries. Given the scarcity of Korean studies on grandparenthood, I first address the previous findings from other countries in this section then discuss the Korean context. The extant literature on grandparenting has suggested that grandp arenting has both positive and negative impacts on mental health in later life. The mixed evidence depends on what types of grandparenting older adults provide for their grandchildren. Researchers ha ve mainly measured grandparenting using either family str ucture 30 (i.e., in what family circumstance the caregiving is provided) or care intensity (i.e., how many hours are spent on the caregiving) of older adults. As for the beneficial effect s of grandparenting on mental health, prior studies have found that prov iding noncoresident grandparenting (i.e., maintaining a separate household from grandchildren but still babysitting) is associated with Caregiving g randmothers are also more likely than non - caregiving gran dmothers to report lower rates of depressive symptoms when providing moderately intensive grandparenting (i.e., providing 200 to 500 h ours of care over two year s, which is approximately 8 - 19 hours per week ) (Hughes et al. 2007) . A moderate level of grandpa renting , especially on a regular basis , is linked to less depression among grandmothers (Grundy et al. 2012) . A moderate level of role obligation and interaction with grandchildren and/or adult children via caregiving would thus contribute to the older adu health as explained by the role enhancement theory. Other empirical studies, by contrast, suggest that grandparenting is detrimental to mental health in later life. As the role strain theory argues, grandparenting adversely affects menta l health, especially when the intensity of grandchild care is excessive . An increasing number of older adults live with and take care of grandchildren as a custodial caregiver on behalf of their incapable adult children. Adult children in the se case s are m ostly absent in the household (i.e., skipped - generation households) for various reasons, such as divorce, unemployment, death, and incarceration (Silverstein and Giarrusso 2010) . Previous research has found that an intensive level of grandchild care, such as with skipped - generation household grandparenting, is significantly associated with depressive symptoms among grandparents ( Blustein, Chan and Guanais 2004; Hughes et al. 2007; Szinovacz, DeViney and Atkinson 1999) . Stress and physical 31 burden from a subs tantial amount of grandparenting seems to explain the adverse effects of The implications of grandparenting for later mental health may vary by sociocultural context, given the cultural differences in the mea ning of grandparenthood and the social expectations for grandparental roles (Hayslip et al. 2012) . The provision of grandparenting is more common and normative for older adults in Asia to support their adult children. Strong family values and ties within t he Asian culture encourage such direct family support across generations (Kataoka - Y ahiro, Ceria and Caulfield 2004; Yancura 2013) . A growing number of - being and health in the As ian context (Mehta and Thang 2012) . Those studies have often focused on grandparenting in multigenerational households an extended family setting, which is a more traditional and common family structure in Asia than in western countries. For instance, Taiw anese older adults experience lower rates of depressive symptoms when providing long - term grandparenting in a multigenerational household, relative to non - caregiving and short - term caregiving grandparents (Ku et al. 2013) . In addition, numerous Chinese gra ndparents in rural areas are taking care of grandchildren as the primary caregiver to support their adult children, who are working in urban areas. Those rural grandparents providing skipped - generation household grandparenting have fewer depressive symptom s when migrant adult children provide financial support as a reward (Cong and Silverstein 2008) . The inconsistent evidence suggests that sociocultural differences in the grandparenting experience and its health consequences exist even in the same regional context. 32 Korean Grandmothers and Mental Health It is inappropriate to apply the evidence on grandparenting and mental health to the Korean context because most of the previous studies are from the U.S. and Europe, where populations largely come from ver y different cultural backgrounds. Although there is some research on Asian grandparenting, the se studies have often focused on multigenerational household grandparenting, which is a fairly rare family structure in Korea (Korea Institute for Health and Soci al Affairs 2014) . Moreover, the mixed findings from the Asian context indicate that country - specific studies are necessary for a more comprehensive understanding of the effects of grandparenting on mental health. More and more grandparents in Korea have ta ken care of their grandchildren over the past decade. Korean grandparenting is unique because a sizable number of older adults offer intensive child care (e.g., full - time) for their grandchildren despite not liv ing in the same household with grandchildren and/or adult children (Korea Institute of Child Care and Education 2015) implications remain underexplored. Especially with regard to mental health, only a few studies have exam ined the association with grandparenting in the Korean context. Some studies have found that caring for grandchildren has a protective effect on mental health in later life. Grandparents who provide part - time grandparenting (i.e., less than 40 hours of car e per week) are more likely to experience life satisfaction than full - time caregiving grandparents (Choi and Cha 2013) . Most prior literature, however, has reported the harmful impacts of grandparenting on ive symptoms. As found in the U.S. and Europe, certain family structures and intensive care are factors that diminish depressive symptoms. For 33 instance, older adults providing skipped - generation household grandparenting are prone to experience a higher lev el of depressive symptoms (Bae 2007) . Moreover, providing full - time caregiving is connected to more severe depressive symptoms among grandparents compared to those who are part - time caregivers (Bae 2007; Baek 2009) . It is important to point out that p revious findings do not fully represent the experience of Korean grandparents. The majority of studies have focused only on excessive caregiving demand, such as skipped - generation household grandparenting and full - time grandparenting (Kim and Kim 2004; Lee and Han 2008) . Those types of grandparenting, however, are not common practice among older adults. It is still unclear whether and how other types of grandparenting affect mental health. In addition, researchers do not exclusively utilize the concept of f amily structure and care intensity to measure grandparenting; most studies consider only one of the two dimensions (Baek 2009; Choi and Cha 2013) . Furthermore, it is unknown how other individual characteristics of older adults, such as SES, health behavior s, and social support, influence the linkage between grandparenting and mental health. Finally, most prior research has used either qualitative methods (Kim and Seo 2007) or cross - sectional or regional data (Bae 2007; Baek 2009; Choi and Cha 2013; Kang 201 1) , which makes the potential causal relationship between grandparenting and mental health much less clear . Present Study The current study aims to examine the association between grandparenting and mental health, specifically the trajectory of depressive symptoms of caregiving grandmothers in Korea. Guided by the role strain/enhancement theories and previous literature, I expect that the effect of grandparenting on depressive symptoms may be more profound among grandmothers who provide more intensive type s of care, such as skipped - generation household grandparenting and 34 full - time noncoresident grandparenting. Given the low prevalence of multigenerational households in Korea, I hypothesize that multigenerational household grandparenting may develop grandmot I also expect that the trajectory of depressi ve symptoms by grandparenting may be partially explained by the underlying mechanisms of SES, health behaviors, and social support . This study goes beyond previous studies in several ways. First of all, I incorporate both family structure and care intensity for a more thorough measurement of grandparenting among older adults . This approach also captures various types of contemporary Korean grandparenting and compare s the effect of eac h type of grandparenting on mental health . Second, this study is among the first to investigate potential mechanisms underlying the association between grandparenting and mental health. Third, I use a nationally representative longitudinal survey to better understand the causal relationship of grandparenting with mental health and the changes in th is link over time. Data and Methods Data This study uses data provided by the Korean Longitudinal Study of Aging (KLoSA) to examine the relationships between gran dparenting and depressive symptoms among Korean grandmothers. The KLoSA is a longitudinal study of a nationally representative sample of 10,254 Koreans aged 45 and older in 15 major cities and provinces. The primary purpose of the survey is to collect data retirement, health status , and social welfare. The KLoSA has been biennially conducted since 2006, the first survey year. 35 This study uses Waves 2, 3, and 4 (2008, 2010, and 2012) o f the KLoSA. I exclude Wave 1 (2006) because the baseline wave does not contain sufficient information household members, which is important for measur ing relevant grandparenting type. Of the 8,688 respondents at baseline (2008), 5,682 respondents are grandparents who have at least one grandchild. I limit my sample to 3,457 grandmothers ; grandfathers (n = 2,225) are dropped from the sample due to the low number of caregiv ing grandfathers. I also exc lude respondents who are older than 80 (n = 385) since the oldest older adults are less likely to provide child care due to their health decline (Hughes et al. 2007; Ku et al. 2013) . The range of missing da ta in the sample varies from less than 1% for most variables to 6.2% for household asset. I use listwise deletion to handle the missing data. The final analytic sample includes 2,814 grandmothers, contributing to 7,657 observat ions across three survey waves with an average of 2.7 observations per responde nt . Measures Depressive Symptoms . Depressive symptoms are measured using the Center for Epidemiologic Studies - Depression (CES - D) scale, which rang es from 1 to 10. The depressive symptoms are constructed as a time - varying variable across the three waves. Grandparenting. Grandparenting in this study is measured as a time - varying categorical variable using two characteristics of older Korean women: family structure and care intensity. First, I classify grandmothers into caregiving and non - caregiving groups b ased on the question of structure) and b) the hours of child care for grandchild , which I qualify as 36 care intensity , grandparenting variab le includ e no grandparenting ( reference), multigenerational household grandparenting, part - time noncoresident grandparenting, and full - time noncoresident grandparenting. Grandmothers who are in an extended - family setting (i.e., living with both adult children and g randchildren) and provide care for the grandchildren are categorized under head their own household (i.e., not living with either adult children or grandchildren) are classified into - ll - classification reflects the trend of highly intensive grandparenting among Korean grandparents who report approximately eight hours of daily child care for five days or more per week, on average (Korea Institute of Child Care and Education 2 015) . I exclude the group of custodial grandmothers who live with only grandchildren (i.e., skipped - generation household grandparenting) due to very low sample size (0.4% of the sample). Moreover, I do not subcategorize the multigenerational household gran dparenting group based on care intensity because there are insufficient numbers of respondents for each sub group. SES . SES comprise s four predictors. Education is a categorical variable coded using e: element ary school or less ( reference), middle school, and high school diploma or more. Household income is a time - varying continuous variable to which I appl y the natural logarithm transformation in order to reduce the skewness. Household asset, which is standard ized and time - varying, is measured using the total amount of 37 assets owned by all household members. Employment status is measured as time - varying and dichotomous (1 = currently working, 0 = currently non - working). Health behaviors . I include three sets of time - varying dummy indicators: currently exercis ing (1 = yes ), currently smoking (1 = yes ), and currently drinking (1 = yes ). Social support . Social support is assessed with three time - varying dummy variables: social activity, support from adult childre n, and support to adult children. Social activity is measured as how often respondents actively participate in any organizations, clubs, or societies (1 = more than monthly). In addition, I include financial and non - financial support from adult children (1 = yes) and financial and non - financial support to adult children (1 = yes). Controls . I include age, marital status, past grandparenting experience, self - rated health , and chronic condition as control variables. Age is a continuous variable and centered at 47 . A preliminary analysis show s no evidence that the age effect has a nonlinear pattern. Marital status is coded as time - varying dummy (1 = married). Past grandparenting experience is measured at baseline as dummy (1 = yes). I also control for self - rat ed health (1 = fair or poor health) and whether respondents were ever diagnosed with any major chronic diseases (e.g., cardiovascular disease, heart disease, diabetes, cancer, high blood pres sure) by a physician (1 = yes ) as time - varying dummy variable s to avoid potential bias from reverse causality. Two yes), are included in all models. Analytic Strategy I first present weighted descriptive statistics of the sam ple in this study. I then apply growth curve models to examine how grandparenting affects the trajectory of depressive 38 symptoms. The growth curve model s ha ve an advantage that distinguishes two different types of variance (i.e., within - individual level and between - individual level) in estimating population average difference in order to remove possible bias from repeated measures. Taking advantage of three waves of the KLoSA and the growth curve analysis, I evaluate the development of depressive symptoms as a function of age . The growth curve model s I use take the following forms : Level 1 model: Level 2 model: In the level - 1 model for within - individual change of depressive symptoms over time, of depressive symptoms is a function of time and other time - varying covariates. Y ij denotes the dependent variabl e, the depressive symptoms of individual i at the j th wave, and j = 1, 2, or 3 indicating KLoSA waves 2 (2008) through 4 (2012); Age ij is the age of individual i at the j th wave . G kij represents the grandparenting of individual i at the j th wave ; TVC mij de notes all other time - varying covariates; i and i are the i intercept and age slope. j is the coefficient for the effects of grandparenting for individual i at the j th wave, which do not vary for individuals (fixed effects); ij is the level 1 residual. The level - 2 model specifies heterogeneity in change across individuals and incorporates the association between time - of 39 depressive symptoms. 0i and 1i represent all other time - invariant covariates. 0i and 1 i are individual - specific residual terms. The analyses from the growth curve models estimate five models. The base model include s grandparenting and basic controls (i.e., marital status, past grandparenting experience, self - rated health , chronic condition, death and dropout ). Next, I assess how potential mechanisms explain the association between grandparenting and depressive symptoms. Indicators of SES, health behaviors, and social support are added in a series of nested models (Model 2 - 4). Finally, the fu ll model considere s all covariates. Results Sample Characteristics Table 3 - 1 presents weighted descriptive statistics for Korean grandmothers at baseline (2008) in this study. The average CES - D score for all grandmothers is 4.2 7 out of 10. The average age of grandmothers is 65.09 years old. Most grandmothers are not taking care of grandchildren. Among caregiving grandmothers, full - time noncoresident grandparenting is the largest group accounting for 3% of all grandmothers. Grandmothers who provide part - tim e noncoresident gran d parenting and multigenerational household grandparenting account for 2.37 % and 1.53%, respectively. Figure 3 - 1 shows the average CES - D scores by grandparenting type. Non - caregiving grandmothers ha ve the highest level of depressive symp toms, followed by grandmothers who provide part - time noncoresident grandparenting and multigenerational household grandparenting. Those who provide full - time noncoresident grandparenting report the lowest level of depressive symptoms. 40 In terms of SES, t he majority of grandmothers report an educational attainment of elementary school or less (69. 52 %). The average household income and household asset of grandmothers are 1978.44 and 18303.83 , respectively. M ore than 2 7 percent of grandmothers are currently working. With respect to health behaviors, more than 30 percent of grandmothers report regular exercise. Only 3.35 percent of grandmothers are smokers and approximately 17 percent were drinkers. As for social support, most grandmothers engage in frequent s ocial activities (8 8.02 %) and receive any financial or non - financial support from their adult children (8 3.30 %). Approximately 2 6 percent of grandmothers provide any type of financial or non - financial support to adult children. Lastly, t wo - thirds of grandm others ( 67.95 %) are married and 13.34% of grandmothers ha ve grandparenting experience in the past. 38. 70 % of grandmothers rate their health as poor and slightly more than 65% of them ha ve been diagnosed with at least one major chronic disease . Grandpar enting and Depressive Symptoms Trajectories Table 3 - 2 reports growth curve estimates of depressive symptoms by grandparenting between ages 47 and 80 . Results from the base model show that multigenerational household grandparenting is marginally significan t and positively associated with the initial depressive symptoms. The marginally significant positive coefficient for multigenerational household grandparenting on the depressive symptoms intercept ( 1.615 ) indicates that grandmothers who provide multigener ational household grandparenting ha ve 1.615 - point higher CES - D scores than non - caregiving grandmothers at age 47 , net of marital status, past grandparenting experience, poor self - rated health , chronic condition, and attrition . However, the significant nega tive coefficient for multigenerational household grandparenting on the age slope ( - .098) indicates that 41 they ha ve a slower rate of increase of depressive symptoms compared to those who d o not provide grandparenting. Figure 3 - 2 illustrates the significant association between grandparenting and the trajectories of d epressive symptoms (CES - D score ) among grandmothers on the basis of M odel 1 in Table 3 - 2 . Overall, non - caregiving grandmother - D scores increase by roughly over 1 point (3.119 to 4.208) b etwe en age 47 and 80. On the other hand, grandmothers who provide multigenerational household grandparenting exhibit more severe depressive symptoms at age 47 (4.734) than non - caregiving grandmothers, but experience a decline in depressive symptoms over time. The trajectories of depressive symptoms between the two groups converge by age in the mid - 60s, then reverse until age 80. Grandmothers providing multigenerational household grandparenting, relative to non - caregiving counterparts, experience a steeper reduc tion in depressive symptoms and better mental health after their mid - 60s than non - caregiving counterparts. Next, I add SES, health behaviors, and social support to Model 1 of Table 3 - 2 , respectively, to test whether those mechanisms m ay explain the grandp arenting - depressive symptoms relationship. After adding SES (Model 2), the coefficient for multigenerational household grandparenting on the depressive symptoms intercept and age slope decrease and the age slope remain s statistically significant. Model 3 s uggests that the marginally significant association between multigenerational household grandparenting and depressive symptoms at baseline still holds after adding health behaviors to Model 1. The magnitude and shape of the effect also decrease . Yet there is no evidence that any particular health behavior explain s the multigenerational household grandparenting - depressive symptoms relationship over time. 42 In Model 4, I add social support to Model 1. The results indicate that the effect s of multigenerational household grandparenting on the intercept and the age slope decrease ( 1.470 and - .095, respectively) and the age slope remains statistically significant. Social support, however, d oes not explain the trajectory of depressive symptoms between age 47 and 80 for the multigenerational household grandparenting group. The full model (Model 5 of Table 3 - 2 ) shows that, after controlling for all covariates, multigenerational household grandparenting is still marginally associated with a higher level of depressive symptoms at age 47 , relative to non - caregiving counterparts. The effects of multigenerational household grandparenting on the initial depressive symptoms decrease by approximately 1 2 % (= (1.420 1.615) / 1.615) , compared to the base model. T he remaining s ignificant coefficient for the age slope of the multigenerational household grandparenting group suggest s that none of the proposed mechanisms aff differences in depressive symptoms over time . Other covariates are significant in the expected directions. For example, being married, providing grandparenting in the past, being highly educated, currently working, having a higher household income, currently exercising and drinking, actively participating in social activities, and providing financi al or non - financial support to adult children are beneficial to mental health by reducing depressive symptoms. In contrast, being in poorer health , having chronic conditions , attrition, currently smoking, and receiving financial or non - financial support fr om adult children are significantly associated with a higher degree of depressive symptoms. 43 Discussion Studies on how grandparenting affect s mental health in later life are limited and even the findings in the limited literature are inconsistent. Moreover, the majority of the se studies have been conducted in the U.S. and Europe, although the trend of grandparenting is growing in other sociocultural contexts including Asia (Mehta and Thang 2012) . Using a nationally representative sample from Korea, this stud y extends the extant literature by investigating whether grandparenting affects the trajectory of depressive symptoms among older women. I also examine the extent to which potential mechanisms, including SES, health behaviors, and social support, explain t he association between grandparenting and depressive symptoms. Results from this study reveal that a particular type of grandparenting shapes depressive symptoms trajectories in later life. Specifically, multigenerational household grandparent i ng signific antly decreases depressive symptoms as they age, although it is only marginally significant that grandmothers providing multigenerational household grandparenting have a higher level of depressive symptoms initially, at age 47, tha n that of non - caregiving counterparts. In contrast, non - increase between ages 47 and 80. Thus, the mental health gap between grandmothers with multigenerational household grandparenting and non - caregiving grandm others decreases over time and even reverses when they reach their mid - 60s, suggesting that grandmothers who provide multigenerational household grandparenting enjoy better mental health in later life. The association between multigenerational household gr andparenting and depressive symptoms over time is stable across all models. Th ese results contribute one of the new findings to the literature on Korean grandparenthood because previous studies have overlooked multigenerational household 44 grandparenting giv en the low prevalence of th is caregiving type (Kim and Kim 2004; Lee and Han 2008) . The results are in line with the previous literature that found fewer depressive symptoms among older adults providing multigenerational household grandparenting in China a nd Taiwan (Cong and Silverstein 2008; Ku et al. 2013) . However, the results are inconsistent with the studies in the U.S., which find no relationship between multigenerational household grandparenting and depressive symptoms (Hughes et al. 2007; Musil et a l. 2013) . These findings suggest that grandmothers who provide multigenerational household grandparenting may psychologically benefit from stable intergenerational relationships and grandchild care. The coresidence setting with both adult children and gra ndchildren allows grandmothers to have a sufficient level of interaction with family members to receive physical and psychological support through the experience. Moreover, supported by the role enhancement theory, those grandmothers may have more self - eff icacy, sense of achievement, and life satisfaction derived from the additional care giving role (Chen et al. 2014; Choi and Zhang 2018) . S imultaneously , grandmothers who provide multigenerational household grandparenting may experience less stress and sense of responsibility through the role because they are still not a custodial grandparent with primary care obligations . The traditional Confucian value s in Korea, which emphasize the concepts of filial piety in an extended family (Yee et al. 2009) , may also contribute to the better mental health of grandmothers because they live in the family structure that the older generation conventionally favors. Based on the role strain theory and prior findings, I expected that there may be more profound association be tween intensive grandparenting (e.g., skipped - generation household grandparenting and full - time noncoresident grandparenting) and depressive symptoms, with the grandparenting - depressive symptoms relationship being contingent on age. However, I did not 45 find an y significant evidence for depressive symptoms at baseline or a linear age slope in relation to any intensive types of grandparenting . The results differ from the previous literature in that earlier studies find that providing full - time care is adversel y associated with depressive symptoms among Korean older adults (Bae 2007; Baek 2009) . In addition , although this study anticipated that the trajectory of depressive symptoms by grandparenting can be explained in part by SES, health behaviors, and social s upport , none of those mechanisms significantly account s for the mental health change over time in relation to grandparenting. This study is not without limitations. First, I have only small sample of caregiving grandparents in the data. Due to th is constra int , I was unable to include some types of grandparenting (e.g., skipped - generation household grandparenting) and more elaborately categorize caregiving groups by caregiving intensity (e.g., low, medium, and high intensity). As such , the small sample size may prevent me from detecting significant effects of particular types of grandparenting on depressive symptoms. Second, it is plausible that I was unable to completely eliminate the possibility that healthier grandparents are more likely to take care of gr andchildren, although I control for health condition and health behaviors at baseline and as time - varying. Third, future research needs to explore the gendered implications of grandparenting for mental health. Although this study excludes grandfathers from analyses due to very small sample sizes, an increasing number of grandfathers are engaged in grandchild care. Given a traditional gender norm which regards child care more as a (De Vos and Lee 1993; Kamo 1998) , the meaning of grandparenting a nd its impacts on health may be different for grandfathers. Despite these limitations, the current study contributes to the literature on grandparenting and mental health in later life. Using longitudinal data on a nationally representative sample in 46 Ko rea, this study examines how grandparenting is associated with the trajectory of older is link varies in a different sociocultural context. I find that multig enerational household grandparenting . The gap in depressive symptoms between grandmothers who provide multigenerational household grandparenting and non - caregiving grandmothers decreases over time and reverses in later life. The findings in this study exte nd previous research on grandparenting and health. This study also calls for future studies that explore the social/cultural differences in grandparenthood and well - being. 47 CHAPTER 4 EPRESSIVE SYMPTOMS IN CHINA Introduction As people are now living longer and healthier lives, o lder adults have been increasingly involved in caring for grandchildren (i.e., grandparenting) as either secondary or primary caregivers (Silverstein and Giarru sso 2010; Margolis 2016) . This phenomenon is particularly prevalent in China . A recent study using the nationally representative sample s reports that approximately 34 percent of older adults care for grandchildren (Sun 2017) . Chinese older adults regard ac tive grandparenting as their normative role in the family based on the traditional values of strong family ties and lineage solidarity (Yancura 2013; Mehta and Thang 2012) . Another notable aspect about Chinese grandparent ing is urban - rural difference s . Inc reased number of grandparents in rural regions intensively care for grandchildren as a custodial caregiver to help support their adult children, the parents of those grandchildren, who migrate to urban regions for employment (Lou et al. 2013) . Despite rapi d growth of the grandparenting trend in China, few studies have examined the consequences of grand child care on later health . The mental health implication s of grandparenting remain especially underexplored . Some research has found that grandparenting is a ssociated with a lower level of depressive symptoms and greater life satisfaction (Cong and Silverstein 2008; Ku et al. 2013; Xu et al. 2012; Silverstein, Cong, and Li 2006) . Most of the findings, however, capture the grandparenting experience of older adu lts who reside in rural regions only (Cong and Silverstein 2008; Xu et al. 2012; Silverstein, Cong, and Li 2006) . In 48 addition, the literature only utilizes information on either family structure or care intensity to measure grandparenting (Xu 2018; Ku et a l. 2013) . Furthermore , little is known about the potential mechanisms that buffer or exacerbate the effects of grandparenting on later mental health. The aim of the current study is thus to examine how grandparenting is associated with l health, specifically depressive symptoms, in China. This study extends previous research on grandparenting and mental health by addressing two research questions. First, I assess whether and how grandparenting affects a trajectory of depressive symptoms among older adults in both rural and urban regions using a nationally representative survey. Second, I examine the extent to which individual and household characteristics (i.e., socioeconomic status (SES), health behaviors, and social support) account for the link between grandparenting and depressive symptoms. Background Grandparenting and Mental Health Research on grandparenting and mental health in later life has been underexplored. Most literature has focused on the experience of U.S. and European gra ndparents while few studies have been done regarding Chinese grandparenting. Thus , I first address previous findings based on those western countries before then discuss ing the Chinese context. Previous literature has measured grandparenting using eithe r family structure (i.e., family composition and living arrangement) or care intensity (i.e., hours spent on grandchild care) of the grandparents. In general, grandparenting is both positively and negatively associated with mental health among older adults (Silverstein and Giarrusso 2010) . The mixed findings are driven by the 49 different types of care older adults provide for their grandchildren. With respect to the protective effects of grandparenting on mental health, a moderate level of noncoresident grand parenting (i.e., 200 - 500 hours of babysitting over two years , that is about 8 - 19 hours per week ) is linked to milder depressive symptoms in later life in the U.S (Hughes et al. 2007) . The regular provision of moderate grandparenting also has a positive rel ationship with depression in Chile (Grundy et al. 2012) . Multigenerational household grandparenting is beneficial for subjective psychological well - being (Goodman and Silverstein 2002) . Yet, the causal relationship of grandparenting with mental health has been unclear due to the cross - sectional nature of some prior studies. The role enhancement theory is often used to account for the positive impacts of grandparenting on mental health in later life. It argues that those who hold various social roles are l ikely to have better mental health. Carrying out different social roles and the compounded - being (Moen, Robison, and Dempster - McClain 1995) . In this sense, g randparenting, an emerging to achieve greater life satisfaction, self - efficacy, and feelings of reward through the caregiving role (Pruchno and McKenney 2002; Rozario, Mor row - Howell, and Hinterlong 2004; Szinovacz and Davey 2006) . In addition to the emotional benefits, grandparenting helps older adults stay physically active (King, Rejeski, and Buchner 1998) and maintain closer family ties and intergenerational support (Mahne and Huxhold 2014) while interacting with their grandchildren and adult children. On the other hand, some literature suggests that grandparenting negatively affects mental health in later life. For example, i ntensive types of grandparenting have an adverse association with the mental health of older adults. Specifically, custodial grandparenting in skipped - 50 generation households an increasing vulnerable family type which consists of grandparent generation and grandchildren only is linked to el evated depressive symptoms in several western countries (B lustein, Chan, and Guanais 2004; Minkler et al. 1997; Szinovacz, DeViney, and Atkinson 1999) . The role strain theory provides an explanation for the negative consequences of grandparenting on mental health. The theoretical perspective suggests that carrying out multiple - being. Given the limited nature of resources including time, energy, and goods, individuals experience difficult ies execut ing different roles concurrently (Goode 1960) . The i ncreased obligations of meeting the se additional expectations serve (Barnett and Baruch 1985; Pearlin 1989) . From this perspective, grandparent ing can lead to further role strain and stress. Considering the various social roles that older adults must already fill (e.g., spouse, parent, child, or any other position) , the additional role as a grandchild care provider can be a greater source of heal th risks for older adults. The burden of grandparenting deteriorates older , especially when the care work is considerably intensive. Other grandparenting - related stressors, such as intergenerational conflicts over child rearin g, a shortage of private time for self - care, leisure and social engagement, and financial burden, are adversely related to later mental health (Minkler 1999 ; Bake r, Silverstein, and Putney 2008; B lustein, Chan, and Guanais 2004; Jendrek 1993) . Pre - existing disadvantages in SES and health condition also undermine the mental health status of older adults, given that grandparents with elevated levels of deprivation tend to provide more intensive grandparenting (Burnette, Sun, and Sun 2013; Silverstein and Giar russo 2010) . When all the strain and burdens from 51 grandparenting outweigh the benefits of caregiving, the health consequences of grandparenting are more likely to be negative. Mental Health Implications of Grandparenting in China The perception and expec tations for grandparental roles vary by social/cultural contexts (Hayslip et al. 2012) . In Asia, active grandparental roles are normative and prevailing as one form of family support based on strong family values and intergenerational ties (Kataoka - Yahiro, Ceria, and Cau lfield 2004; Yancura 2013; Mehta and Thang 2012) . As such, in China, a large number of older adults have provided v arious types of grandparenting as support for their adult child ren, in addition to the grandchildren themselves . Accordingly, increased attention has been recently paid to the health implications of grandparenting in later life. Previous studies have mainly found that grandparenting is positively associated with physical and mental health, although some negative or null effects o f grandparenting have also been reported (Xu 2018) . In terms of Chinese grandparenting, a primary issue found in the literature is rural - urban differences. Chinese populations are identified as rural and urban residents by the hukou , the official househol d registration system (Chan 2009) . There are significant socioeconomic disparities between the rural and urban regions in China. In comparison with urban residents, rural residents have limited access to quality public resources, facilities, and services ( e.g., schools, pension benefits, health care access, etc.), which result in their greater risk of poverty in later life. Thus, older adults living in rural regions are more likely to develop depressive symptoms (Li et al. 2016) . The majority of resear - being have explored caregiving grandparents in skipped - generation households who reside in 52 rural regions. Numerous grandparents left behind in rural regions take care of grandchildren as cus todial caregiver s to support their migrant adult children who work in urban centers (Lou et al. 2013) . Previous findings have suggested that , despite the intensive care burden, skipped - generation household grandparenting can have mental health benefits , de pending on the rewards from adult children, such as financial support (Cong and Silverstein 2008; Silverstein, Cong, and Li 2006) , and care intensity (Xu et al. 2012) . Prior studies have also focused on multigenerational household grandparenting, which r has been common and preferred in China and other Asian countries based upon strong family solidarity, filial piety, and collective interest (Mehta and Thang 2012) . Research from China, Taiwan, and Hong Kong has found t hat g randparents who provide multigenerational household grandparenting report fewer depressive symptoms and greater life satisfaction , relative to non - caregiving grandparents (Ku et al. 2013; G uo, Pickard, an d Huang 2008; Silverstein, Cong, and Li 2006) . Findings on the link between grandparenting and mental health in China are still sparse and inconclusive. The mixed evidence suggests that the health implications of grandparenting are likely t o vary by not only caregiving experience itself but also regional context and underlying individual characteristics (Xu 2018) . Moreover, some significant limitations exist in the extant literature. P rior studies have focused on the grandparenting experienc e of older adults who reside in rural regions only (Cong and Silverstein 2008; Silverstein, Cong, and Li 2006; Xu et al. 2012) . Furthermore, previous research has not successfully take n both family structure and care intensity into consideration to capture the various types of older adults respective health consequences (Cong and Silverstein 2008; Guo, Pickard, and Huang 2008) . 53 Taken together, the current study aims to examine how grandparenting influences ntal health, specifically the trajectory of depressive symptoms. To extend the literature, this study also detects various potential mechanisms underlying the relationship of grandparenting with depressive symptoms. I include grandparents in both rural and urban regions using data from a nationally representative longitudinal study to assess the rural - urban differences. Grandparenting is more thoroughly measured by combining both family structure and care intensity. In light of the role enhancement/strain t heories and prior findings, I expect that highly intensive types of grandparenting skipped - generation household grandparenting and full - time noncoresident grandparenting have more pro nounced depressive symptoms with age. In terms of the rural - urban disparities in China, I hypothesize that rural grandparents who provide skipped - generation household grandparenting and multigenerational household grandparenting report fewer depressive symptoms relative to non - caregiving counterparts. I also expect that some underlying mechanisms, such as SES, health behaviors, and social support, partially explain the depressive symptoms trajectory of caregiving grandparents. Data and Methods Data The current study utilizes the China He alth and Retirement Longitudinal Study (CHARLS) to investigate the association between grandparenting and depressive symptoms. The CHARLS has tracked a nationally representative sample of Chinese adults aged 45 years and older to collect data on older adul styles . The baseline sample includes 54 approximately 10, 000 households and 17,500 individuals in 150 counties/districts and 450 villages/resident committees in 2011. The respondents have been followed up every two years. I use three wave s of the CHARLS data (2011 - 2015). The analytic sample of this study includes 5,691 older adults who have grandchildren and complete the questions on grandparenting, depressive symptoms, and region. I exclude grandparents who are over 80 years old given tha t few of the oldest older adults are engaged in grandparenting due to worsening health (Hughes et al. 2007; Ku et al. 2013) . Measures Depressive Symptoms . The dependent variable of this study is depressive symptoms. I measure depressive symptoms using the Center for Epidemiologic Studies - Depression (CES - D) scale. Ten items in the CHARLS ask how respondents felt and behaved in the past week on a 4 - of the time (five to seven days). - varying continuous variable ranging from 1 to 30, where 30 indicates the worst level of depressive symptoms. Grandparenting. Grandparenting is a key independent variable in this study. Grandparenting is a tim e - varying categorical variable: no grandparenting (reference), skipped - generation household grandparenting, multigenerational household grandpar enting, part - time noncoresident gr andparenting, and full - time noncoresident grandparenting. I first use the ques tion to identify caregiving and non - caregiving grandparents. To categorize the caregiving grandparents, I then utilize two items, h ousehold member information and a g rand parenting - 55 related question ( taking family structure and care intensity, respectively. Using the family structure, I classif y caregiving grandparents who live with grandchildren , but not adult children , - Caregiving grandparents who live in an extended family with both grandchildren and adult grandparents who live without both grandchildren and adult children are categorized as - - per week). The part - time group includes grandparents who spend less than 40 hours per week on noncoresident grandparenting. Grandparents spending 40 or more hours per week on noncore sident grandparenting are labelled as the full - time group. This study includes three potential mechanisms to explain the association between grandparenting and depressive symptom s: SES, health behaviors, and social support. SES . Four indicators represent from baseline, including illiterate (reference), less than elementary school, elementary school, and middle school or higher. In order to measure household - level economic status, I use an asset index, rather than household income, because the asset index could be a better proxy of wealth or standard of living in the context of developing countries, which is less subject to measurement errors (Bollen, Glanville, and Stecklov 2002, Filmer and Pritchett 20 01) . I measure the asset index as a time - varying continuous variable using the household ownership of 17 luxury items (e.g., automobile, electric bicycle, motorcycle, refrigerator, washing machine, TV, computer, etc.), ranging from 0 - 17. Current employment status is coded as time - varying dichotomous 56 (working = 1). Pension receipt is a time - varying dichotomous variable indicating whether respondent received any pension income (yes = 1). Health behaviors . I include three time - varying categorical predictors. Exercise is measured based on the question whether respondents do any vigorous physical activity for at least 10 minutes every week: no (reference), yes, and missing report. With respect to current smoking and drinking status, respondents are also categori zed into three groups: no (reference), yes, and missing report. Social support . I assess social support using three time - varying dummy variables. Social activity is constructed using the item on whether respondents participate in any social activities, inc luding spending time with friends; playing ma - jong, chess, or cards with others or going to a community club; sport, social, or other types of club activities; activities of community - related organizations; volunteer or charity work; and educational or tra ining courses. Respondents who have engaged in any such activities are coded as 1 (= yes). I also include whether older adults have received any financial support from adult children (yes = 1), and whether older adults give any financial support to adult c hildren (yes = 1). Controls . This study controls for basic demographic characteristics and health status of older adults. Gender is a dummy variable (female = 1). Age is treated as continuous and centered at 55, a common reti rement age in China. A prelim inary analysis reveals that there is no significant nonlinear pattern of age effect s . Marital status is a time - varying dummy variable (married = 1). I also include whether older adults live in rural regions (rural = 1, urban = 0). The rural residence is me hukou status, which is the official household registration system indicating legal rural and urban populations in China (Chan 2009) . In terms of health status, I control for self - rated health (poor health = 1) and wheth er respondents 57 have been diagnosed with any chronic conditions among hypertension, diabetes, heart disease, 1) are included to take into account attrition in th e data. Analyti c S trategy I first present weighted descriptive statistics from the baseline sample of this study. Next, I estimate growth curve models to investigate the effects of grandparenting on depressive symptoms initially and over time. The growth c urve analysis takes advantage of the longitudinal nature of the CHARLS and the development of depressive symptoms is evaluated as a function of age in this study. The two - level growth curve model is specified as: Level 1 model: Level 2 model: The Level 1 model characterizes within - individual change in depressive symptoms over time . In this model, an other time - varying covariates. Y ij is the depressive symptoms of individual i at the j th wave, where j = 1, 2, or 3 , indicating the waves 1 - 3 of the CHALRS; Age ij represents the age of individual i at the j th wave; G kij denotes the grandparenting of individual i at the j th wave; TVC mij indicates other time - varying covariates included in the model; i and i are the i intercept and the rate of change in depressive symptoms with age ; k denotes the coefficient for 58 the impacts of grandparenting for individual i at the j th wave, which is fixed over individuals; the error term ij is the l evel 1 residual. The Level 2 model specifies heterogeneity in change across individuals and determine s the relationship between time - invariant covariates and the shape of each individual s growth trajectory of depressive symptoms. 0i and 1i represent all other time - invariant predictors in the models. 0i and 1i are individual - specific residual terms. The analyses of this study estimate five models. Model 1 includes grandparenting, the key independent variable , to assess how grandparenting is associated with depressive symptoms over time. Model s 2 - 4 e ach add predictors of SES, health behaviors, and social support, respectively, to Model 1 in order to examine whether and how those mechanisms account for the relationship between grandparenting and depressive symptoms. Model 5 is the full model including all covariates. Given the importance of rural - urban differences in the Chinese context found in the literature, I estimate those series of models separately for grandparents liv ing in rural versus urban regions to find regional differences in the mental he alth implications of grandparenting. Results Descriptive Statistics Table 4 - 1 presents weighted descriptive statistics from baseline (2011) for all analyzed variables. I display t he results by region to show the differences between rural and urban grandp arents. The results show that rural grandparents report a significantly higher level of depressive symptoms (9.04) compared to urban grandparents (6.69). The average age of rural grandparents is 58.83, which is younger than that of urban grandparents (61.4 5). 59 In terms of grandparenting, approximately half of all grandparents take care of grandchildren last year . The patterns of grandparenting differ by region among caregiving grandparents: rural grandparents report higher proportions of grandchild care in coresident family structures compared to urban grandparents. Among caregiving rural grandparents, the largest group is those who provide multigenerational household grandparenting (28.82%), followed by skipped - generation household grandparenting (7.11%), p art - time noncoresident grandparenting (7.46%), and full - time noncoresident grandparenting (4.60%). Note, the proportion of skipped - generation household grandparenting is significantly higher among rural grandparents than urban grandparents as previously re ported in the literature. Whereas urban grandparents tend to provide grandparenting in noncoresident settings. For urban grandparents, although multigenerational household grandparenting still makes up the largest group (25.75%), part - time noncoresident gr andparenting ( 11.55%) and full - time noncoresdient grandparenting (7.33%) are the next large r groups. Skipped - generation household grandparenting represents 4.91%. Figure 4 - 1 presents depressive symptoms by grandparenting type for rural and urban grandparen ts. Overall, rural grandparents show worse depressive symptoms relative to urban grandparents. Among rural grandparents, the level of depressive symptom s is the highest, especially for the skipped - generation household grandparenting and no - grandparenting g roups . Urban grandparents report the worst depressive symptoms with the provision of n o - grandparenting and multigenerational household grandparenting. In comparison to urban grandparents, rural grandparents have, on average, lower SES, including lower edu cation levels, less household asset s , higher current employment , and less pension receipt . Rural grandparents are more likely to exercise, smoke and drink, and have less social activity and more financial support from adult children , compared to urban gran dparents. 60 Rural grandparents have a lower proportion of being married but better self - rated health and fewer chronic condition s than urban grandparents. Grandparenting and Trajectory of Depressive Symptoms Table 4 - 2 shows the estimated coefficients of dep ressive symptoms (CES - D) from growth curve models for rural grandparents. The results of Model 1 in Table 4 - 2 suggest that multigenerational household grandparenting and full - time noncoresident grandparenting have lower level of initial depressive symptoms compared to non - caregiving rural grandparents . The negative coefficient ( - 0.493) on the depressive symptoms intercept indicates that rural grandparents who provide multigenerational household grandparenting report 0.493 - point lower CES - D scores at initial time (age 55) compared to non - caregiving rural grandparents, controlling for gender, marital status, self - rated health, chronic condition s , and attrition. In addition, rural grandparents providing full - time noncoresident grandparenting have 0.508 - point lo wer scores on CES - D initially than their non - caregiving counterparts. As for age slope, the significant positive coefficient for multigenerational household grandparenting (0.057) suggests that, in comparison to their non - caregiving counterparts, rural gra ndparents who provide multigenerational household grandparenting have a faster increase in depressive symptoms with age. Figure 4 - 2 displays the significant association between grandparenting and the trajectories of depressive symptoms based on Model 1 in Table 4 - 2. The figure shows that depressive symptoms marginally increase between age 45 and 80 for non - caregiving rural grandparents. In contrast, the depressive symptoms of rural grandparents who provide multigenerational household grandparenting increase with age. Due to the fast er rate of increase, the depressive symptoms of rural grandparents providing multigenerational household grandparenting surpasses that of non - caregiving counterparts after the early - 60s, although their 61 level of depressive symptoms is lower initially. Meanwhile, rural grandparents providing full - time noncoresident grandp arenting consistently have fewer depressive symptoms than their non - caregiving counterparts , both initially and over time. Full - time noncoresident grandparenting is also more beneficial for depressive symptoms than multigenerational household grandparenting after the mid - 50s. Model s 2 - 4 of Table 4 - 2 add SES, health behaviors, and social support, respectively, to the base model to assess how those potential mechanism s may account for the association between grandparenting and depressive symptoms. In Model 2, the initial differences between multigenerational household grandparenting and full - time noncoresident grandparenting are no longer statistically significant afte r including SES. However, the significant age slope for multigenerational household grandparenting remains significant. Model 3 shows that the coefficients of the intercept and age slope for multigenerational household grandparenting and full - time noncor esident grandparenting remain similar and significant when adding health behaviors. Model 4 suggests that including social support makes no change in the significant relationship between multigenerational household grandparenting and the trajectory of depr essive symptoms. Yet full - time noncoresident grandparenting loses its significant effect on the intercept of depressive symptoms. Model 5, the full model, indicates that none of the grandparenting types are significantly associated with rural grandpare covariates. However, the effect of multigenerational household grandparenting on the age slope still holds. The results suggest that the proposed mechanisms (i.e., SES, health behaviors, and soc ial support) affect the initial status of depressive symptoms, but not the rate of change in depressive symptoms with age. 62 The effects of other covariates are significant in the expected directions. Rural grandparents who are women, report poor health , ch ronic condition s or death, are more likely to experience worse depressive symptoms. On the other hand, higher education levels, being currently working, pension receipt, engaging in social activit ies , and providing financial support to adult children lower the level of depressive symptoms for rural grandparents. Lastly, I find no significant evidence for the association between grandparenting and depressive symptoms among urban grandparents (Table 4 - 3). In addition to the rural - urban difference, I test whe ther the consequences of grandparenting for mental health differ by gender as a sensitivity analysis (results not shown) . The results indicate no significant gendered implications in the relationship. Discussion Research on grandparenting and mental heal th is limited. It is unclear how There is a dearth of studies that examine whether and how grandparenting is related to Chinese spite the ever - increasing trend of grandparenting (Mehta and Thang 2012) . In order to extend the existing literature, this study aims to examine the association between grandparenting and depressive symptoms over time in China. Using nationally representat ive data, I further advance the literature by assessing the extent to which three potential mechanisms SES, health behaviors, and social support explain the relationship between grandparenting and depressive symptoms. I also test the rural - urban difference s in grandparenting patterns and their mental health implications , which are important given the fast - growing number of grandchildren left behind with their grandparents in rural regions. 63 This study has several major findings. Overall, the results sugges t that for Chinese older adults, grandparenting has a significant impact on depressive symptoms over time and th at association varies by urban - rural context. In particular, grandparenting significantly shapes the trajectory of depressive symptoms among rur al grandparents. First, with regard to grandparenting types, rural grandparents who provide multigenerational household grandparenting have a lower level of depressive symptoms at initial time (age 55) compared to non - caregiving rural grandparents. Multig enerational household time. In contrast to a minute increase in depressive symptoms among non - caregiving rural grandparents, rural grandparents providing multigen erational household grandparenting experience an accelerated rate of inc rease in depressive symptoms with age . Consequently, rural grandparents engaging in multigenerational household grandparenting have worse depressive symptoms after reaching their early - 60s than non - caregiving counterparts, despite their lower initial level of depressive symptoms. In the long - term, multigenerational household between multigeneration al household grandparenting and non - caregiving groups converges , then widens over time with a negative implication . The adverse effect of multigenerational household grandparenting on depressive symptoms in this study is inconsistent with two previous st udies using cross - sectional data with regional samples, which report fewer depressive symptoms in older adults providing multigenerational household grandparenting in China (Guo, Pickard, and Huang 2008; Silverstein, Cong, and Li 2006) . T he increase in dep ressive symptoms over time in relation to providing multigenerational household grandparenting adds a new finding to the literature on 64 Chinese grandparenting. Moreover, the results are in line with prior research which suggests a rapid decline in self - rate d health among caregiving Chinese grandparents in multigenerational households (Chen and Liu 2012) . The role strain theory supports the finding on the negative mental health implication s of multigenerational household grandparenting. Older adults who prov ide multigenerational household grandparenting may suffer physical and/or psychological burden s as a result (Pruchno and McKenney 2002) , as this is a new role that they are asked to additionally carry out in later life. Executing multiple roles with the in clusion of grandparenting increases older adults level given their limited time and resources (Goode 1960; Barnett and Baruch 1985; Pearlin 1989) . Older adults who provide multigenerational household grandparenting may have less time and resources to invest in their own health and well - being since the coresident setting precludes any full respite from the caregiving (Choi and Zhang 2018) . Family relationships can be another possible stressor for caregiving grandparents in multigenerational househol ds . Negative interactions and relationships among family members result in worse psychological well - being, including elevated depressive symptoms (Lai 1995; Rook 1984) . M ore fami ly members from different generations living together means that more c onflict s and te nsions over family life or child - rearing philosophy and practices are possible. Specifically in China, older adults mostly live with their sons and daughter - in - law, and it is culturally allowed that paternal grandmothers actively intervene in child rearing (Cong and Silverstein 2008) . Considering that those norms remain stronger in rural area (Xu 2018) , it is plausible that caregiving rural grandparents have greater risk of stress and depressive symptoms due to conflicts with their adult children, e specially daughter in - law. Although older adults providing multigenerational household grandparenting live in a culturally preferred family type 6 5 in China (Logan, Bian, and Bian 1998; Mehta and Thang 2012) , the cumulative health risks from the care responsi bility and coresident family structure offset the advantages of grandparenting and may lead to a faster decline in mental health . Second, this study finds that rural grandparents who provide full - time noncoresident grandparenting have a lower level of de pressive symptoms at initial time (age 55), relative to non - do not significantly increase over time with the provision of full - ti me noncoresident grandparenting; t he mental health ga p between the t wo groups is consistent . The mental health benefit of full - time noncoresident grandparenting is a new contribution to the literature, which has mostly focused on skipped - generation household grandparenting or multigenerational household gran dparenting. Moreover, the result is consistent with earlier studies showing the positive effects of grandparenting on psychological well - being (Xu 2018; Cong and Silverstein 2008; Silverstein, Cong, and Li 2006) . Despite the in tensive level of caregivin g , full - time no ncoresident grandparenting has less harmful influences on depressive symptoms unlike multigenerational household grandparenting. The psychological burden of o lder adults who provide full - time noncoresdient grandparenting is perhaps less than that of older adults with other intensive care arrangements because they are still not custodial caregiver s with the sole or primary responsibility for the grandchildren. It is also plausible that the availability of entirely private leisure time , which i s possible due to noncoresidence with grandchildren and/or adult children, outweighs the adverse effect of intensive full - time grandparenting on mental health (Choi and Zhang 2018) . Less psychological and physical commitment may help older adults enjoy the various rewards of grandparenting, such as a sense of achievement, self - efficacy, an active life style, and a healthy diet (Chen et al. 66 2014; Silverstein, Cong, and Li 2006; Pruchno and McKenney 2002) . Furthermore, given the Chinese context emphasizing st rong family bonds and intergenerational support for collective well - being (Mehta and Thang 2012; Burnette, Sun, and Sun 2013) , older adults who live apart from grandchildren and adult children but still provide grandparenting are likely to gain more self - w orth and family involvement. Additionally, they are less likely to feel the shortcomings of living apart (e.g., loneliness and physical hardship) through the provision of stable noncoresident grandparenting. Ultimately, the emotional support and health ben efits from regular caregiving Lastly, I find that SES is the key underlying mechanism that partially explains the association between grandparenting and depressive symptoms. Unlike other mechanisms including health behaviors and social support, SES accounts for a substantial portion of and full - time noncoresident grandparenting . Adding SES nullifies the significant effects of those grandparenting types on depressive symptoms, although the impact of multigenerational household grandparenting on the change in depressive symptoms over time is robust regardless. The results echo previous literature, which shows a considerable rol e of individual factors (e.g., prior health, sociodemographic characteristics, and SES) over grandparenting itself in mental health (Hughes et al. 2007) . In China, economic resources especially play an indispensable role in explaining the health consequences of grandparenting (Chen a nd Liu 2012; Silverstein, Cong, and Li 2006; Cong and Silverstein 2008) . Note, it is important that this study finds SES to be a significant underlying mechanism that serves only rural grandpar ents. Chinese older adults in rural regions are more vulnerable than older adults in urban regions in terms of SES. Rural older adults tend to rely heavily on their 67 adult children for financial support and health care due to the lack of quality health serv ice and social security available in th ose regions (Lee and Xiao 1998) . As a return, rural older adults provide grandchild care based on time - for - money (Cong and Silverstein 2008, 2011) , even if the caregiving is heavily intensive , suc h as with skipped - generation household grandparenting and full - time grandparenting. Given the unique rural context, the mental health of rural grandparents is largely contingent on SES , rather than a sole consequence of grandparenting. Previous literatu re has suggested that there are gendered implications of grandparenting. Grandmothers and grandfathers show different patterns of grandparenting and , as such, their health consequences also vary (Hughes et al. 2007). Although it is beyond the scope of this study, the results from a sensitivity analysis indicate that the association between grandparenting and depressive symptoms does not differ by gender among both rural and urban Chinese grandparents (results not shown). In light of the more gendered divisi on of care work and nurturing roles in Asian contexts (Mehta and Thang 2012), however, it is still plausible that some health implications of grandparenting are also gendered in China (Xu 2018). Future research is necessary to further examine how gender pl This study has several limitations. First, I had to exclude some categories of grandparenting types because of data limitation s . T he groups of skipped - generation household grandparenting and multig enerational household grandparenting were not further classified utilizing c are intensity information as the noncoresident grandparenting groups were because the sample size was insufficient . It is also possible that I was unable to detect significant rela tionships between grandparenting and depressive symptoms among urban grandparents due to the small sample size of caregiving older adults in urban regions . Next, I call for future 68 research which takes into account the quali ty of grandparenting, rather than being based solely on the amount of grandparenting. Even if older adults provide the same amount of time for grandchild care, the impacts of grandparenting on health can differ according to various detail s of the care (e.g., the age or special needs of gr andchildren). Future studies should explore further understand contemporary grandparenting and its health implications. Finally, this study cannot entirely rule ou t the possibility that the identified grandparenting differences in depressive symptoms are in part driven by selection. For example, it is more likely that healthier older adults engage in grandparenting compared with older adults who are frailer. Althoug h I control for some health condition s and behaviors as time - varying, the results of this study are conservative. Despite these limitations, the current study makes important contributions to the research on grandparenting and mental health. Using nationa lly representative longitudinal data in China, this study adds to the literature the finding that the level of depressive symptoms in rural grandparents who provide multigenerational household grandparenting inc reases over time. The provision of full - time noncoresident grandparenting is associated with fewer depressive symptoms among rural grandparents relative to their non - caregiving counte rparts . However, SES partially accounts for the relationship between grandparenting and depressive symptoms. All in al l, the findings in this study advance our understanding of the sociocultural differences in the motive, process, and health consequences of grandparenting in later life . This study also suggests that effective intervention in assisting rural grandparents w ith low er SES need s to be implemented to help reduce the mental health disparities brought on by grandparenting in China. 69 CHAPTER 5 CONCLUSION The health implications of intergenerational relationships , specifically grandparenting, in aging families h ave long been underexplored. This dissertation contributes to the body of knowledge on intergenerational relationships by examining how grandparenting influences health in later life as well as how that association varies by sociocultural context. Drawin g from three longitudinal surveys of nationally representative samples of older adults in the U.S., South Korea, and China, my findings confirm that sociocultural differences exist in the consequences of grandparenting on physical and mental health. The fi rst study suggests that there are racial/ethnic variations in the linkage between grandparenting and mortality in the U.S. White grandparents gain a mortality advantage from grandchild care, whereas black grandparents have a mortality disadvantage from the caregiving experience. The second study South Korea. The findings show that grandmothers who provide multigenerational household grandparenting enjoy its long - term protective ef fect on depressive symptoms. The third study assesses the relationship between grandparenting and mental health in China. This project reveals that multigenerational household grandparenting has an adverse effect on Chinese s over time, although socioeconomic status partially accounts for this association. This dissertation sheds light on the importance of grandparenting to health and well - being in later life. The findings also add a deeper understanding to the current schol arship that the significant role of grandparenting in explaining later health differs among societies based on 70 their unique socioeconomic and cultural contexts. This dissertation calls for greater attention to the sociocultural differences in the concept, motivation, and practice of grandparenting to better understand the health implications of its provision. 71 APPENDICES 72 APPENDIX A: Chapter 2 Tables Table 2 - 1. Weighted Descriptive Statistics for Grandparents by Race, Health and Retire ment Study, 1998 - 2014 ( N = 13,705) White Black Hispanic Variable ( n = 10,616) ( n = 1,978) ( n = 1,111) Deceased 1998 - 2014 (%) 24.89 24.87 18.78* Grandparenting (%) No grandparenting (ref ) 80.32 75.75 78.62 Skipped - generation household grandpare nting 1.47 4.66* 2.80* Multigenerational household grandparenting 2.81 8.23* 6.85* Light/moderate noncoresident grandparenting 10.54 7.52* 7.62* Intensive noncoresident grandparenting 4.87 3.84* 4.11 Health Condition Fair/poor self - rated healt h (%) 28.07 41.37* 50.08* CES - D (0 - 8) 1.43 (.04) 1.93 (.07)* 2.31 (.08)* Chronic disease (%) 86.63 88.02 80.60* Socioeconomic Status Years of schooling 12.60 (.06) 11.15 (.12)* 8.47 (.39)* Household income (Ln ) 10.50 (.02) 9.94 (.03)* 9.83 (.06 )* Net household wealth (Ln ) 11.93 (.05) 10.22 (.09)* 10.38 (.13)* Working (%) 47.20 48.13 46.52 Long - term health insurance (%) 10.44 7.65* 2.71* Health Behaviors (%) No vigorous exercise 49.86 59.55* 56.66* Drinking 32.72 18.91* 24.94* Smo king 14.11 18.25* 15.72 Controls Women (%) 58.16 63.46* 58.05 Age 70.61 (.18) 68.38 (.24)* 68.57 (.27)* Married (%) 73.68 46.68* 63.96* Foreign born (%) 3.74 5.14* 52.88* Note : Standard deviations are in parentheses. *Statistically signifi cant difference between white and black or white and Hispanic at the .05 level. 73 Table 2 - 2 . Hazard Ratios for Death by Grandparenting and Race ( N = 13,705) 74 APPENDIX B: Chapter 3 Tables Table 3 - 1. Weighted Descriptive Statistics for Grandmothers Aged 47 - 80, KLoSA, 2008 ( N = 2,184) Variable Mean or % SD CES - D (1 - 10) 4.27 2.96 Age 65.09 8.05 Grandparenting No grandparenting (ref ) 93.10 Multigenerational household Grandparenting 1.53 Part - time n oncore sident grandparenting 2.37 Full - time n oncore sident grandparenting 3.00 Socioeconmic Status Elementary school or less (ref ) 69.52 Middle sch ool 15.77 High school diploma or more 14.71 Household income (10,000 Won) 1978.44 2698.28 Household asset (10,000 Won) 18303.83 30669.63 Working 27.67 Health Behaviors Exercise 32.28 Smoking 3.35 Drinking 17.47 Social Support Social activity (more than monthly=1) 88.02 Support from adult children 83.30 Support to adult children 26.31 Controls Married 67.95 Past grandparenting experience 13.34 Poor self - rated health 38.70 Chronic con dition 65.41 Death 4.01 Dropout 10.42 75 Table 3 - 2. Growth Curve Estimates of Grandparenting on Trajectories of Depressive Symptoms among Grandmothers, KLoSA, 2008 - 2012 ( N = 2,814) Variable Model 1 Model 2 Model 3 Model 4 Model 5 Fixed Effect s Intercept 3.119*** 4.641*** 3.421*** 3.986*** 5.716*** Grandparenting (r ef = No grandparenting) Multigenerational household grandparenting 1.615+ 1.599+ 1.557+ 1.470+ 1.420+ Part - time noncoresident grandparenting - 0.726 - 0.793 - 0.676 - 0.5 84 - 0.611 Full - time noncoresident grandparenting 0.056 - 0.004 0.015 - 0.040 - 0.127 Linear slope (age) 0.033*** 0.017** 0.026*** 0.026*** 0.008 Multigenerational household grandparenting - 0.098* - 0.095* - 0.096* - 0.095* - 0.090* Part - time noncoresident grandparenting 0.035 0.039 0.032 0.027 0.029 Full - time noncoresident grandparenting - 0.024 - 0.025 - 0.023 - 0.016 - 0.017 Socioeconomic status Middle school (r ef = Elementary school or less ) - 0.382** - 0.298* High school diploma or more - 0.574 *** - 0.432*** Working - 0.374*** - 0.329*** Household income - 0.139*** - 0.148*** Household asset - 0.034 - 0.039 Health behaviors Exercise - 0.446*** - 0.399*** Smoking 0.763** 0.649** Drinking - 0.386*** - 0.291** Social s upport Social activity (more than monthly=1) - 1.012*** - 0.999*** Support from adult children 0.314*** 0.286** Support to adult children - 0.357*** - 0.281*** Controls Married - 0.421*** - 0.336*** - 0.424*** - 0.403*** - 0.326*** P ast grandparenting experience - 0.326* - 0.332* - 0.285* - 0.398** - 0.361** Poor self - rated health 1.433*** 1.356*** 1.390*** 1.392*** 1.281*** Chronic condition 0.257** 0.231** 0.285** 0.254** 0.257** Death 0.984*** 0.892*** 0.892*** 0.907*** 0.753** Dropout 0.438** 0.437** 0.441** 0.398* 0.408** Random Effects Level 1 residual 4.554*** 4.539*** 4.540*** 4.558*** 4.530*** Level 2 intercept 2.861*** 2.704*** 2.806*** 2.615*** 2.465*** Level 2 slope 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** - 2 Log likelihood 18166 18131 18136 18095 18039 *** p<0.001, ** p<0.01, * p<0.05, + p<0.1 76 APPENDIX C: Chapter 4 Tables Table 4 - 1. Weighted Descri ptive Statistics, CHARLS , 2011 ( N = 5,691) Mean or % Variable Rural Grandparents (n = 4,638) U rban Grandparents (n = 1,053) CES - D (1 - 30) 9.04* (6.39) 6.69 (5.64) Age 58.83* (8.03) 61.45 (7.52) Grandparenting No grandparenting (ref ) 52.01 50.46 Skipped - generation household grandparenting 7.11* 4.91 Multigenerational household g randparenting 28.82 25.75 Part - time noncoresident grandparenting 7.46* 11.55 Full - time noncoresident grandparenting 4.60* 7.33 Socioeconomic Status Illiterate (ref ) 30.07* 10.10 Less than elementary school 21.63* 12.91 Elementary school 25.27* 21.79 Middle school or higher 23.03* 55.29 Household asset (1 - 17) 3.92* (2.22) 4.83 (2.43) Working (ref = no) Yes 74.16* 31.52 Missing 0.45* 0.20 Pension (ref = no) Yes 16.39* 64.66 Missing 0 .21* 0.34 Health Behaviors Exercise (ref = no) Yes 17.66* 6.43 Missing 55.52 56.41 Smoking (ref = no) Yes 32.19* 26.12 Missing 0.01 0.00 Drinking (ref = no) Yes 19.53* 16.52 Missing 4.92 6.72 Note . Standard deviations are in parentheses. *Statistically significant difference at the .05 level. 77 Table 4 - Mean or % Variable Rural Grandparents (n = 4,638) Urban Grandparen ts (n = 1,053) Social Support Social activity (ref = no) Yes 44.96 58.04 Missing 0.06* 0.00 Support from adult children (ref = no) Yes 42.94* 27.72 Missing 0.14* 0.06 Support to adult children (ref = no) Yes 18.02* 26.45 Missing 0.17* 0.14 Controls Women 51.23 48.86 Married 73.47* 79.21 Poor self - rated health (ref = no) Yes 52.61 57.25 Missing 28.10* 23.96 Chronic condition Yes 34.88* 50.46 Missing 1.41* 0.42 Death 5.12* 4.03 Dropout 13.02* 28.00 Note . Standard deviations are in parentheses. *Statistically significant difference at the .05 level. 78 Table 4 - 2 . Growth Curve Estimates of Grandparenting on Trajectorie s of Depressive Symptoms among Rural Grandparents, CHARLS, 2011 - 2015 ( N = 4,638) Variable Model 1 Model 2 Model 3 Model 4 Model 5 Fixed Effects Intercept 6.257*** 8.918*** 6.083*** 6.791*** 9.044*** Grandparenting (ref = No grandparenting) Skipped - generation household grandparenting 0.027 0.173 0.048 0.132 0.218 Multigenerational household grandparenting - 0.493** - 0.113 - 0.496** - 0.483** - 0.123 Part - time noncoresident grandparenting - 0.151 - 0.033 - 0.143 - 0.027 0.037 Full - time noncoresi dent grandparenting - 0.508* - 0.320 - 0.504* - 0.406+ - 0.283 Linear slope (age) 0.001 - 0.028* 0.004 0.001 - 0.029* Skipped - generation household grandparenting 0.036 0.031 0.034 0.035 0.029 Multigenerational household grandparenting 0.057** 0.057** 0.056** 0.059** 0.058** Part - time noncoresident grandparenting - 0.020 - 0.015 - 0.021 - 0.023 - 0.018 Full - time noncoresident grandparenting 0.033 0.043 0.031 0.030 0.039 Socioecon o mic Status Less than elementary school (ref = Illierate) - 0.182 - 0.163 Elementary school - 0.944*** - 0.930*** Middle school or higher - 1.313*** - 1.263*** Household asset (1 - 17) - 0.368*** - 0.348*** Working (ref = no) Yes - 0.281+ - 0.296* Missing 0.909 0.917 Pension (ref = no) Ye s - 0.426** - 0.358* Missing - 0.938** - 0.926** Health Behaviors Exercise (ref = no) Yes 0.295 0.296 Missing 0.155 0.105 *** p<0.001, ** p<0.01, * p<0.05, + p<0.1 79 Table 4 - 2 Variable Model 1 Model 2 Model 3 Model 4 Model 5 Smoking (ref = no) Yes 0.184 0.046 Missing - 0.300 - 0.134 Drinking (ref = no) Yes - 0.165 - 0.096 Missing 0.447 0.181 Social Support Social activity (ref = no) Yes - 0.699 *** - 0.564*** Missing 1.361 1.149 Support from adult children (ref = no) Yes - 0.089 0.072 Missing - 1.051 - 0.913 Support to adult children (ref = no) Yes - 0.477*** - 0.299* Missing 0.321 0.600 Controls Women 2.041*** 1.650*** 2.067*** 2.070*** 1.670*** Married - 0.922*** - 0.744*** - 0.923*** - 0.963*** - 0.784*** Poor self - rated health (ref = no) Yes 2.742*** 2.665*** 2.747*** 2.724*** 2.656*** Missing 2.185*** 1.722*** 2.154** * 2.044*** 1.667*** Chronic condition Yes 0.938*** 1.058*** 0.955*** 0.963*** 1.071*** Missing 0.455 0.494 0.466 0.493 0.512 Death 1.643*** 1.262** 1.625*** 1.576*** 1.248** Dropout 0.260 0.196 0.254 0.252 0.196 Random Effects Level 1 residual 21.177*** 21.174*** 21.169*** 21.193*** 21.182*** Level 2 intercept 15.716*** 14.414*** 15.668*** 15.365*** 14.206*** Level 2 slope 0.000*** 0.000*** 0.000*** 0.000*** 0.000*** - 2 Log likelihood 30,548 30,432 30,543 30,520 30,414 *** p <0.001, ** p<0.01, * p<0.05, + p<0.1 80 Table 4 - 3. Growth Curve Estimates of Grandparenting on Trajectories of Depressive Symptoms among Urban Grandparents, CHARLS, 2011 - 2015 ( N = 1,053) Variable Model 1 Model 2 Model 3 Model 4 Model 5 Fixed Effects I ntercept 5.391*** 8.866*** 5.666*** 6.135*** 9.452*** Grandparenting (ref = No grandparenting) Skipped - generation household grandparenting - 0.312 - 0.074 - 0.408 - 0.213 - 0.129 Multigenerational household grandparenting 0.063 0.365 0.058 0.108 0.344 Part - time noncoresident grandparenting - 0.702 - 0.635 - 0.722 - 0.671 - 0.662 Full - time noncoresident grandparenting - 0.495 - 0.293 - 0.528 - 0.415 - 0.281 Linear slope (age) 0.016 0.001 0.015 0.015 - 0.002 Skipped - generation household grandparenting - 0.055 - 0.060 - 0.044 - 0.052 - 0.051 Multigenerational household grandparenting - 0.053 - 0.047 - 0.048 - 0.050 - 0.043 Part - time noncoresident grandparenting 0.055 0.068 0.058 0.061 0.075 Full - time noncoresident grandparenting 0.027 0.021 0.029 0.028 0.022 Soci oecon o mic Status Less than elementary school (ref = Illierate) - 0.687 - 0.626 Elementary school - 1.785*** - 1.683** Middle school or higher - 2.798*** - 2.684*** Household asset (1 - 17) - 0.253*** - 0.216*** Working (ref = no) Yes - 0.029 - 0.106 Missing - 0.363 - 0.093 Pension (ref = no) Yes - 0.474+ - 0.378 Missing - 0.231 - 0.317 Health Behaviors Exercise (ref = no) Yes 0.562 0.491 Missing - 0.132 - 0.135 *** p<0.001, ** p<0.01, * p<0.05, + p<0.1 81 Table 4 - 3 Variable Model 1 Model 2 Model 3 Model 4 Model 5 Smoking (ref = no) Yes - 0.209 - 0.329 Missing - 0.569 - 0.422 Drinking (ref = no) Yes - 0.134 - 0.075 Missing - 0.4 32 - 0.356 Social Support Social activity (ref = no) Yes - 1.123*** - 0.854*** Missing - - Support from adult children (ref = no) Yes 0.075 0.007 Missing - 1.974 - 2.158 Support to adult children (ref = n o) Yes - 0.463* - 0.247 Missing 0.972 1.461 Controls Women 1.284*** 0.974** 1.112** 1.290*** 0.803* Married - 0.947** - 0.589+ - 0.972** - 0.905** - 0.611* Poor self - rated health (ref = no) Yes 2.251*** 2.233*** 2.2 57*** 2.253*** 2.240*** Missing 1.615*** 1.330*** 1.577*** 1.473*** 1.234** Chronic condition Yes 0.561* 0.609* 0.539+ 0.583* 0.596* Missing - 0.797 - 0.959 - 0.764 - 1.008 - 1.081 Death - 0.270 - 0.769 - 0.223 - 0.389 - 0.769 Dropout - 0.5 24 - 0.312 - 0.503 - 0.530 - 0.317 Random Effects Level 1 residual 15.742*** 15.867*** 15.769*** 15.850*** 15.930*** Level 2 intercept 14.473*** 12.618*** 14.304*** 13.618*** 12.049*** Level 2 slope 0.001*** 0.000*** 0.001*** 0.001*** 0.001*** - 2 Log likelihood 6,705 6,661 6,702 6,689 6,651 *** p<0.001, ** p<0.01, * p<0.05, + p<0.1 82 APPENDIX D: Chapter 2 Figure Figure 2 - 1. Predicted Probabilities of Morta lity by Grandparenting and Race Note . ** indicates significant mortality differences as rep orted in Model 6 controlling for all covariates . 83 APPENDIX E: Chapter 3 Figures Figure 3 - 1. Depressive Symptoms by Grandparenting , KLoSA, 2008 84 Figure 3 - 2 . Trajectories of Depressive Symptoms by Grandparenting: Growth Curve Model Estimates Note . Figure 3 - 2 is based on Model 1 in Table 2 controlling for basic demographic characteristics and health status . 85 APPENDIX F: Chapter 4 Figures Figure 4 - 1. Depressive Symptoms by Grandparenting and Region, CHARLS, 2011 86 Figure 4 - 2 . Trajectories of Dep ressive Symptoms by Grandparenting among Rural Grandparents : Growth Curve Model Estimates Note . Figure 4 - 2 is based on Model 1 in Table 2 controlling for basic demographic characteristics and health status . 87 REFERENCES 88 REFEREN CES U . S . Department of Health and Human Services Rockville, MD. Retrieved April 21, 2019 ( https://acl.gov/sites/default/files/Aging%20and%20Disability%20in%20America/2014 - Profile.pdf ). Allis on, Paul D. 2014. Event Histo ry and Survival Analysis . Thousand Oaks, CA: SAGE Publications, Inc. Arpino, Br Does Grandparenting Pay Off? The Effec t of Child Journal of Marriage and Family 76(2):337 - 51. Bae, Jin - hee. Factors Affecting the Psychological Well - Being of Grandparents Raising Grandchildren - Focusing on the Comparison between Parenting Grandparents and Co - Social Welfare Policy 29(2):67 - 94. Baek, Kyung - th: Physical and Women Health 10(2):87 - 112. Baker, Lindsey A . , Merril Silverste Grandparents Raising Grandchildren in the United Stat es: Changing Family Forms, Stagnant Social P Journal of Societal and S ocial policy 7:53. Barnett, Rosali Women's Involvement in Multiple Ro les and Journal of Personality and S ocial P sychol ogy 49(1):135. Blustein, Jan, Sewin Chan , and Fred Elevated Depressive Symptoms Health Services R esearch 39(6p1):1671 - 90. Bollen, Kenneth A . , Jennifer L . Glan Economic status proxies in studies of fertility in developing coun Population S tudies 56 (1):81 - 96. Bryson, Ken and Lynne M. Casper. 1999. Coresident Grandparents and Grandchildren . Washington, DC: Census Bureau. Burnette, Denise, Ju are o Ageing International 38 (1):43 - 57. Black Grandparents Rearing Children of Drug - Addicted Parents: Stressors, Outc omes, and S The Gerontologist 32(6):744 - 51. 89 Burton, Lind Challenges and Rewards: African American Gra Generations: Journal of the American Society on Aging 16(3):51 - 54. Casper, Lyn ne M., S andra M. Florian, C. Brady - 32 in Grandparenting in the United States , edited by Madonna Harrington Meyer and Ynesse Abdul - Malak. Amityville, NY: Baywood Press. Eurasian Geography and E conomics 50 (2):197 - 221. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 67(1):99 - 112. The Journals of Gero ntology Series B: Psychological Sciences and Social Sciences 70(5):793 - 803. Demographic Trends in the United States: A R eview of Research in the Journal of Marriage and Family 72(3):403 - 19. Choi, Seun g - won and Zhenmei Z Grandparenting and Self - Rated H ealth among Older Research on Aging 40(10):911 - 32. Cho The Effects of Caregiving Characteristics on the Perceived Life Satisfaction of the Grandparents Who Providin Social Welfare Policy 40(3):183 - 206. Christ The Association between Social Science & Medicine 118:89 - 96. Cong, Zhen Time - for - Money Exchanges in Rural Research in Human Development 5 (1):6 - 25. Cong, Zhen l Exchange b etween Parents and Migrant and N Journal of Marriage and Family 73 (1):93 - 104. Currie, Ja Mortality Inequality: The Good New s from a County - The Journal of Economic Perspectives 30(2):29 - 52. De Vos , Susan and Yean - Change in Extended Family Living among Elderly Pe ople in South Korea, 1970 - Economic Development and Cultural Change 41(2):377 - 93. 90 Facio, Elisa. 1996. Understanding Older Chicanas: Sociological and Policy Perspectives , Vol. 14. Thousand Oaks, CA: SAGE Publications, Inc. Data Demography 38 (1):115 - 132. Fuller - Thomson, Es me, Meredith M A Profile of Grandparents Raising Gran The Gerontologist 37(3):406 - 11. Goode, William J. American S ociologic al R eview 25(4): 483 - 96. Goodman, C Structure and Well - The Gerontologist 42 (5):676 - 689. Grundy, Emily M . , Cecilia Albala, Elizabeth Allen, Alan D . Dangour, Diana El bour ne, and Ricardo Grandparenting and Psychosocial Health among Older Chileans: A Longitudinal Analysis." Aging and Mental H ealth 16(8):1047 - 57. Guo, Baorong, Joseph erspective on Health O utcomes of Car egiving G ran Journal of Intergenerational Relationships 5 (4):25 - 40. Innovation in Aging 1(2):igx023 - igx23. Hayslip, Bert, Heidemarie B lumenthal, and Ashley Garner Grandchild Well - Being One - Journal of Aging and Health 26(4):559 - 82. Great - Grandchild - 212 in Handbook of Families and Aging , edited by Rosemary Blieszner and Victoria H. Bedford. Santa Barbara, CA: Praeger. Heron, Melonie P. 2013. Deaths: Leading Causes for 2010 . National Vital Statistics Reports, 62, 1 - 96. Hilbrand, Sonja, David A. Coall, Denis Ger Caregiving within and Beyond the Family Is Associated with Lower Mortality for the Caregiver: A Prospective Evolution and Human Behavior 38(3):397 - 403. Hofferth, Child Care in the United States The Future of Children 6(2): 41 - 61. 91 Hughes, Mary Elizabeth, Linda J . Waite, Trace All in the Family: The Impact of Caring for Grandch The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 62(2):S108 - S19. randc hildren: Effects on Lifestyle Journal of Marriage and the Family 55(3): 609 - 621. Johnson, N The Racial Crossover in Comorbid Demography 37(3):267 - 83. - 112 in Handbook on Grandparenthood , edited by Maximiliane Szinovacz. Westport, CT: Greenwood Press/ Greenwood Publishing Group. The Psychosocial Well - Being of Grandparents Raising Their Grandchildren: The Role of Resources, Type of Care , and Perception of Korean Demography 34(1):73 - 97. Kataoka - Yahiro, Merle, Clementin a C Grandparent Caregiving Role Journal of Pediatric N ursing 19(5):315 - 28. Kawachi, Ichiro and Lisa F . Berkman. 20 Journal of Urban H ealth 78(3):458 - 67. Under One Roof: A Review of Research on Intergenerational Coresidence and Multigenerational H Sociology Compass 4(8):642 - 57. Kim, E un - Ju and Young - A Qualitat ive Study on Child - Korean Journal of Child Studies 28(2):175 - 92. Kim, Hye - The Fostering Experience of Custodial Grandmother in Low - Income Grandparent - Korean J ournal of Elderly Welfare 43:61 - 92. Ki m, Mi - Hye and Hye - A Study on Factors Influencing Grandparent`S Life Satisfaction in Low Income Gr andparent - Korean Journal of Gerontology 24(3):153 - 70. King, ical Activity Interventions Targeting Older Adults: A Critical Am erican Journal of Preventive M edicine 15(4):316 - 33. King, V Perceived Se lf - The Journals of Gerontol ogy Series B: Psychological Sciences and Social Sciences 53(5):S249 - S57. 92 Korea Institute for Health and Social Affairs. 2014. National Report on Older Adults 2014 . Seoul, South Korea: Congress. Korea Institute of Child Care and Education. 2015. A Study o f Grandparents Raising Young Grandchildren and Support System . Seoul: South Korea: Korea Institute of Child Care and Education. Ku, Li - Jung E., Sally C. Stearns, Courtney H. Van Houtven, Shoou - Yih D. Lee, Peggye Dilworth - Anderson, and Thomas R. Konrad. 20 The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 68(6):1009 - 1021. sychologi cal Well - Being in Urb Journal of Health and Social Behavior 36(1): 11 - 37. LaVeist, Thomas A. 2005. Minority Populations and Health: An Introduction to Health Disparities in the United States , Vol. 4. Hoboken, NJ: John Wiley & Sons. Lee, Ju nghwa and Kyung - hye Han. 20 Intergenerational Triads of Grandparent - Headed Families and Psychological Well - Being of Custodial Grandmothers in R Korean Journal of Gerontology 28(1):177 - 96. Lee, Y ean - Urban and R ural China: Journal of Cross - Cultural G erontology 13 (1):39 - 62. Li, Lydia W . , Jinyu Liu, Hongwe - Urban Differences in Depressive S ympto ms among Older Adults in C Journal of Aging and H ealth 28 (2):341 - 362. Logan, John R . , Fuqi Social Forces 76 (3):851 - 882. Lou, Vivian W. Q., Nan Lu , Ling Xu, and - Grandchild Family C a pital and Self - Rated Health of Older Rural Chinese Adults: The Role of the Grandparent - Parent The Journals of G eront ology. Series B: Psychological Sciences and Social Sciences 68(4):599 - 608. Luo, Ye, Louise C. Hawkley, Linda J. Wait Loneliness, Health, and Mortality in Old Age: Social Science & Medicine 74( 6):907 - 14. Luo, Ye, Tracey A. LaPierre, Mary Elizabeth Hug hes, and Lind Grandparents Providing Care to Grandchildren: A Population - Based Journal of Family Issues 33(9):1143 - 67. 93 Mahne, Kathar ubjec tive Well - Being: Moderating Journals of Gerontology Series B: Psychological Sciences and Social Sciences 70 (5):782 - 792. Journal of Marriage and Family 78(3):610 - 622. Margolis, Rach Healthy Grandparenthood: How Long Is It, and How . Demography 54(6):2073 - 99. Mehta, Kalyani K . and Leng Leng Thang. 2012. Experiencing Grandparenthood: An Asian Perspective , Vol. 47. Dordrecht , Netherlands: S pr inger . Miech, Richard, Fred Pampel, Jinyoung Kim , and Richard G. Rogers The Enduring Association between Education and Mortality: The Role of Wide ning and Narrowing American Sociological R eview 76(6):913 - 34. Minkler, Meredith. 1999 Intergenerational Households H eaded by Grandparents: Contexts, R ealities , and Implications for P Journal of Aging Studies 13 (2):199 - 218. Minkler, Meredith and Esme Fuller - Grandchildren: A Na The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 60(2):S82 - S92. Minkler, Meredith, Esme Fuller - in G Archives of Family Medicine 6(5):445 - 52. Moen, Phyllis, Julie Robison , and Donna Dempster - Caregiving and Women's Well - Journal of Health and Social Behavior 36(3): 259 - 73. Can Family Relationships Explain the Race Paradox in Mental Journal of Marriage and Family 75(2):470 - 85. Musil, Carol M., Nahida L. Gordon, Camille B. Warner, Jaclene A. Zauszniew ski, Theresa Standing , The Gerontologist 51(1):86 - 100. Musil, Carol M., Alexandra B. Jeanblanc, Christopher J. Burant, Jaclene A. Zauszn iewski , and Nursing Outlook 61(4):225 - 34. e2. Palloni, Albe Paradox Lost: Explaining the Hisp anic Adult Mortality Demography 41(3):385 - 415. 94 Park, Chung - A Study on the Formation Process of Grandparents - Grandchildren Families, Raising Experiences, and Lif Korean Journal of Family Welfare 15(2):105 - 25. The Sociological Study Journal of Health and Social Behavior 30(3): 241 - 56. Pruchno, Rachel A Psychological Well - Being of Black and White Grandmothers Raising Grandchildren Exa mination of a Two - The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 57(5):444 - 52. Grandparents: A Study of T Physical & Occupational Therapy In Geriatrics 6(3 - 4):31 - 62. ide of Social Interaction: Impact on Psychological Well - Journal of Personality and Social Psychology 46(5):1097 - 1108. Rozario, Philip A . , Nancy Morrow - Howell , and James E . Hinterlong. Role Enhancement or Role Strain Assessing the Impact of Multiple Productive Roles on Older Caregiver Well - Research on Aging 26(4):413 - 28. Arrangements of Older People in Rural China: Consequences for Psychological Well - The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 61(5):S256 - S266. Journal of Marriage and Family 72(5):1039 - 58. of Older People Innovation in Aging 1 (Suppl 1):1295. Szinovacz, Maximiliane E. and The International Journal of Aging and Human Development 62(1):1 - 20. Pa - The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 54(6):S376 - S88. Measuring America: An Overview to Grandparent and Grandc hildren ieved April 21, 2019 ( https://www.census.gov/content/dam/Census/data/training - workshops/recorded - webinars/grandparentgrandchildren - nov2014.pdf ). 95 U.S. Census (https://www.c ensus.gov/content/dam/Census/newsroom/facts - for - features/2017/ cb17 - ff14.pdf ). World Health Organization. 2014. World Health Statistics 2014: A Wealth of Information on Global Public Health. World Health Organization. tal Health of Chinese Grandparents Caring for Grandchildren and Great - Social Science & Medicine . doi: https://doi.org/10.1016/j.socscimed.2018.05.047. Xu, Ling, Bei Wu, Iris Chi, and Hsin - Intensity of Grandparent Caregiving and Life Satisfaction Among Rural Chinese Older Adults: A Longitudinal Study Using La tent Family & Community Health 35 (4):287 - 299. Justifications for Caregiving in White, Asian American, and Native Hawaiian Gran The Journals of Gerontology Series B: Psychological Sciences and Social S ciences 68(1):139 - 144 . Pacific Islander Famil - 316 in Asian American Psychology: Current Perspectives , edited by Nita Tewari and Alvin N. Alvarez. New York, NY: Psychology Press, Taylor & Francis Group. Young, Rebek Handling Missing Values in Longitudinal P anel Journal of Marriage and Family 77(1):277 - 94.