Weight Gain During the Transition to Adulthood among Children of Immigrants: Is Parental Co-residence Important? Elizabeth Baker

Transcription

1 Weght Gan Durng the Transton to Adulthood among Chldren of Immgrants: Is Parental Co-resdence Important? Elzabeth Baker Abstract Immgrants tend to be healther than ther natve born peers, despte ther low socoeconomc status and often precarous lvng stuatons. One common explanaton for ths s that mmgrants brng cultural norms wth them that protect themselves from ther health hazardous envronments. I examne BMI trajectores from adolescence to young adulthood, usng the NLSY 97, and examne whether patterns of parental co-resdence moderate or medate the relatonshp between BMI and generaton, focusng on Mexcan-Amercans. Home leavng may take chldren of mmgrants away from the cultural protecton of ther parents and neghborhoods, but may also be assocated wth ncreases n young adult socoeconomc status through college attendance and employment. I fnd that home leavng s assocated wth ncrease n BMI for all generatons, but ths effect s stronger for the frst and second generaton. Addtonally, only among the thrd or hgher generaton s ths effect explaned by famly transtons, partnerng and chldbearng. Lastly, the reason for home leavng (partnerng, college attendance, or other) and ts assocaton wth BMI s examned. 1

2 INTRODUCTION In ths project, I focus on the effect of home leavng on BMI durng the transton to adulthood, whle smultaneously consderng the other transtons young adults are makng at ths tme. One of the most mportant transtons that many young adults experence durng ths tme perod s leavng the parental home. Between the ages of 25 to 29 roughly 75% of young adults have left ther natal home (Portes 2003) and ths s seen as an mportant ndcator of achevng adult status among young adults themselves (Arnett 2003). Leavng the parental home may result n ether ncreases or decreases n obesty. On the one hand, leavng home s assocated wth hgher levels of self effcacy and mental health (Shanahan 2000; Fulgn and Peterson 2002) and may be ndcatve of a successful transton to adulthood (Schnaberg and Goldenberg 1989), and healther lfestyle choces. On the other hand, when gven greater autonomy over food choces young adults may be more lkely to chose less healthful foods and decrease ther consumpton of more healthful tems. Indeed, the transton to young adulthood s assocated wth a sgnfcant decrease n frut and vegetable consumpton and young adults consume more fast food than any other age group (Nelson et al 2005; Len et al 2001; Larson et al 2007; Paeratakul et al 2003). Beyond explorng the relatonshp between home leavng and BMI, I examne whether ths relatonshp vares by mmgrant status. Chldren of mmgrants, both the frst and second generaton, are less lkely to experence pre-martal resdental ndependence than chldren of natves (Goldscheder and Goldscheder 1987). Furthermore, the effect of home leavng on other health outcomes vares by generaton (Fulgn and Peterson 2002). Cultural factors are often ponted to as an explanaton for mmgrant s better health and ther lower lkelhood to leave home relatve to natves (Kanajanpan 1989; Goldscheder and Goldsheder 1989; Aqulo 1990; Pallon and Aras; Hummer et al 2007; Lara et al 2004). Among chldren of mmgrants, resdng wth one s parents may result n slower acculturaton and hence a slower uptake of obesty promotng behavors. In sum, I examne the extent that home leavng s assocated wth changes n weght durng the transton to adulthood and whether ths effect vares by generaton. In dong so, I focus on dfferences among the frst generaton (foregn born), second generaton (has mmgrant parents), and the thrd generaton or hgher (both parents were born n the U.S.), n both the pooled and Mexcan-Amercan, whte, and black sub-samples. In the followng sectons I descrbe the transton to adulthood as a partcularly mportant phase for understandng the development and mantenance of obesty and overweght, the context and changes that take place durng ths phase, followed by a dscusson on how home leavng may be assocated wth weght by generaton. Eatng more and exercsng less: young adulthood and mplcatons for health Det, physcal actvty, and weght change drastcally durng the transton from adolescence to young adulthood (Gordon-Larsen et al 2004; Mokdad et al 1999; Nelson et al 2008; Tucker et al 2005). Experences n adolescence have a sgnfcant mpact on later adult health outcomes ether drectly (becomng less physcally actve, quttng smokng, etc.) or ndrectly through ther effect on educatonal and occupatonal outcomes (Lnk and Phelan 1995; Nelson et al 2008). The transton to adulthood has also been dentfed as a crtcal pont for the adopton of negatve health behavors such as smokng, bnge drnkng, and maruana use (Tucker et al 2005 (RAND). Understandng the pathways that lead to healthy or unhealthy behavor then becomes mportant for the development of early preventon programs. Ths age perod s marked by an ncrease n poor detary habts (Len, Lytle, and Klepp 2001; Larson et al 2007). Daly frut and vegetable consumpton drops precptously as adolescents move nto adulthood. In one study, roughly half of boys and grls reported daly frut 2

3 and vegetable consumpton at age 14, but ths dropped to about 20% when they were rentervewed at age 21 (Len et al 2001). Adolescents and young adults consume more fast-food than any other age group (Paeratakul et al 2003). Furthermore, Larson and colleagues (2007) report that adolescents decrease ther consumpton of frut and vegetables as they approach adulthood. Even more troublng s that ths pattern has worsened overtme. Frut and vegetable consumpton among adolescence decreased sgnfcantly from 1999 to Whle frut and vegetable consumpton decreased, other poor detary habts, such as fast-food consumpton ncreased (Wane et al 2010; Len et al 2001; Nelsen et al 2002). Though t s wdely understood that parents nfluence ther chldren's weght n chldhood and adolescence, parents may contnue to mpact the health of ther young adult chldren f they contnue to lve wth them. For example, past research has found that famly meals durng adolescence are assocated wth ncreased consumpton of fruts and vegetables and greater frequency of daly breakfast consumpton as young adults (Larson et al 2007). In other words, greater parental montorng of det s assocated wth a better det at the tme and benefcal det patterns nto adulthood. Addtonally, past research has extensvely examned whether leavng for college and subsequently the parental home, s assocated wth weght gan. Commonly referred to as the freshman 15, attendng college s often assocated wth ndvdual weght gan, especally durng the frst year of college. Ths ncrease s attrbuted to lower vegetable and frut consumpton, decreased physcal actvty, and ncreased consumpton of fast food (Wane et al 2010). Ths research suggests that poorer det may be assocated wth ncreased autonomy n food decsons that young adults experence as they move away to college. The lack of drect parental montorng through famly meals and ndrect parental montorng through ther control of the food choces avalable to young adults may be at least partally responsble for the weght gan that ndvduals experence as they transton to a unversty settng. However, ths research does not provde a comparson group of young adults who do not attend college. As such, t s dffcult to tell whether college attendance s assocated wth more or less weght gan than other actvtes and whether resdng wth one's parents has an mportant medatng or moderatng effect. Overall, pror research suggests that contnued co-resdence wth parents may be benefcal to det and hence result n slower weght gan among adults. However, a much broader tenant of lterature dscusses the complextes nvolved n the transton to adulthood, and how the role of home leavng nfluences feelngs of self-effcacy, autonomy, mental health, and educaton. Moreover, these factors may vary for mmgrant chldren and chldren of mmgrants. Understandng how co-resdence may be assocated wth health and health behavors requres a more thorough examnaton of the transton to young adulthood, the meanng that co-resdence wth parents has for young adults, and the mplcatons ths may have for adult outcomes. THE TRANSITION FROM ADOLESCENCE INTO YOUNG ADULTHOOD: A PERIOD OF UNCERTAINTY The hstorcal ncrease n lfe expectancy and contnued educaton, now well nto adulthood, has resulted n an ncrease of the number of dstnct lfe phases. In the early part of the 20th century, adolescence was a relatvely short tme perod between chldhood and adulthood, at the end of whch ndvduals were seen economcally, socally, and personally as adults. They were, for the most part, ndstngushable from other adults n ther 30s and 40s (Furstenburg, Rumbaut, and Settersen 2004; Hogan and Astone 1986). However, the transton from adolescence to adulthood has become more ambguous and prolonged. It tends to occur n a gradual, complex, and less unform fashon. At the same tme, early adulthood has become ncreasngly structured by nsttutons outsde the famly (Furstenburg et al 2004; Hogan and Astone 1986). 3

4 The perod between the late teens and the late twentes s now often referred to as transtonal or emergng adulthood. Ths perod s dstnct from the adolescent years where ndvduals are hghly relant on ther natal famles for support and socalzaton and are not legally seen as adults. However, ths perod s also dstnct from adulthood. Many ndvduals have yet to experence the role transtons hstorcally assocated wth markers of adult status, such as marrage, chldbearng, career entry, economc ndependence, and educaton completon (Arnett 1998, 2003). Also, many young adults do not feel as f they have entrely reached adulthood (Arnett 2003; Shanahan 2004). Usng the Youth Development Survey, Shanahan and colleagues (2004) fnd that only about 60% of the 25 to 26 year olds n the study feel entrely lke an adult most of the tme. The context of ths prolonged perod of adulthood has been shown to have mportant effects on many outcomes for young adults, such as labor force partcpaton, educaton, and health. One of the most commonly ndcated requrements for achevng adult status s ndependence from one s parents. Arnett (1998) found that 78% of young adults beleve that developng belefs and values ndependent from one s parents, 74% beleve that achevng fnancal ndependence from one's parents, and 60% beleve achevng resdental ndependence from one s parents are necessary pre-cursors to obtanng adult status. Ths s qute hgh gven that only 17% of young adults beleve marrage, full-tme employment, or ownng a home are necessary precursors of adult status. However, whle t s clear that young adults vew ndependence as necessary for attanment of adult status, past lterature demonstrates mxed effects of home leavng on many outcomes for young adults. Falure to acheve resdental ndependence and returnng to one s natal home as an adult has been referred to as ncompletely launched young adults by Schnaberg and Goldenberg (1989). The famly lterature suggests that ths may be due to ncomplete socalzaton of youth for adult roles, especally by parents (Schnaberg and Goldenberg 1989). Adult chldren who return to ther parents home and chldren who never leave ther parents home have a lower educatonal attanment than those who left ther parent's home and dd not return (Whte and Lacy 1997). Addtonally, young adult chldren who never left ther parents home have hgher rates of depresson, are more lkely to feel lonely, and have lower rates of self-effcacy compared to those who do not lve at home (Whte 1994; Mallett 2004; Fulgn and Peterson 2002). Contnued resdence wth parents may stall the transton to adulthood, not allowng youths to develop the tools necessary to ensure mantanng or achevng a healthy weght, such as self effcacy, educaton, and mental health (Goldman and Whtaker 2002; Ross and Mrkowsky 1999; Strecher, DeVells, Becker, and Rosenstock 1986). However, other recent research suggests that ths perspectve on home-leavng may be msguded. Returnng home and lower rates of resdental ndependence may be an adaptve strategy to an ncreasngly more demandng labor market and the reducton of jobs (Whte 1994; Booth, Crouter, and Shanahan 1999). Whle those who never left home or returned home had a lower educatonal attanment than those who left home permanently, the age at whch one left home s also mportant. Whte and Lacy (1997) report that young adults who contnue to resde wth ther parents untl the age of 21 had a hgher educatonal attanment than those who left ther parents home before the age of 21. Addtonally, many parents now beleve that ther adult chld wll resde wth them for a perod of tme and most parents are receptve to acceptng ther adult chldren n ther homes (Both et al 1999). Contnued co-resdence means that parents have more nfluence over the behavors of ther adult chldren than those who gan resdental ndependence (Matras 1990). Ths may explan the lower rates of bnge drnkng and maruana use (Whte et al 2006) and the hgher educatonal attanment, at least for those who move out by the age of 22 (Whte and Nancy 1997). Past lterature s relatvely slent about the effects of home leavng on det, exercse, and 4

5 weght. Nevertheless, research concernng the transton to college and lterature on bnge drnkng and famly meals suggest that leavng the parental home may be assocated wth weght gan. Not all are chosen: Varaton n the transton nto adulthood by race/ethncty and natvty status Besdes havng mportant mplcatons for health, emergng adulthood experences dffer across groups and therefore may contrbute to health dspartes. Though much of the early research on emergng adulthood was carred out on largely whte natve-born samples (Mollenkampf et al 2007), recent research has attempted to account for ths obvous and large dearth n the lterature by specfcally examnng how race/ethncty and generaton status nfluence pathways nto adulthood (Arnett 2003; Fussell and Furstenberg 2004; Mollenkopf et al 2004; Osgood et al 2004; Portes and Rumbaut 2001; Rumbaut 2004; Fulgn and Pedersen 2002; Fulgn and Hardway 2004). However, most of ths research s confned to specfc regonal areas or derved from a non-random sample (see Fussel and Furstenberg 2004 for an excepton). There are many reasons to thnk that the pathways for chldren of mmgrants or mnortes may dffer from the experences of the largely whte mddle class. Often these ndvduals receve less help or money from ther natal famles, are more lkely to face blocked opportuntes (such as poor schools or neghborhoods), and are more lkely to experence dscrmnaton than whte ndvduals from natve-born famles (Arnett 2003; Portes and Rumbaut 2001; Pong and Hao 2007; Telles and Ortz 2008). As dscussed earler n the secton on segmented assmlaton theory, mmgrant parents tend to have fewer resources, work longer hours, and provde less supervson to ther chldren durng adolescence, and ths may mpact ther chldren s pathways nto adulthood. For example, chldren who receve less parental supervson and nvolvement are more lkely to drop out of hgh school and have lower educatonal aspratons (Astone and McLanahan 1991). Unfamlarty wth Englsh and the lower educatonal attanment of most mmgrant parents also makes t dffcult for them to help ther chldren wth ther homework. Chldren from low SES backgrounds and chldren wth less parental supervson are also at an ncreased rsk for an earler pregnancy than ther peers, whch n turn s assocated wth lower educatonal attanment (Arnett 2003; Fergusson and Woodward 2000; Rndfuss and St. John 1983). Lastly, the lower materal resources of mmgrant parents mean that ther chldren have less of a parentally provded safety net to ad them durng ths tme of rapd change. Schone and Ross (2005) fnd that ths form of support s especally mportant for young adults as parents contnue to support them well nto early adulthood. They estmate that parents spend about $38,000 for housng, educaton expenses, food, or drect cash assstance durng the ages of 18 to 34, although assstance does decrease wth age. For many, young adulthood s marked by movng out of the parental home. But resdental patterns for chldren of mmgrants dffer greatly from those observed for chldren of natves, especally for those of Hspanc or Mexcan descent (Balstrer 2007; Goldscheder and Goldscheder 1987; Van Hook and Glck 2007; Fulgn and Pedersen 2002). Aqulo (1990) fnds that 54% of Mexcan-Amercan parents resde wth ther adult chldren compared to 40% of black and 28% of non-hspanc whte parents. Addtonally, dfferences n marrage patterns accounted for the black-whte dfferental, but not the Mexcan-Amercan whte dfferental. The greater co-resdence of Mexcan-Amercan young adult chldren wth ther parents has been attrbuted to cultural preferences concernng famly arrangements. Indcatve of ths cultural preference, further research demonstrates that the extended co-resdence wth parents decrease wth the duraton of U.S. resdence (Kanjanpan 1989). Whle contnued co-resdence wth parents s greatest among the frst generaton and newly-arrved mmgrants, ths preference s also apparent among more settled mmgrants and 5

6 the second generaton. Young adults among the second generaton and the 1.5 generaton (foregn born ndvduals who arrved to the U.S. as chldren) are more lkely to reman n ther parents home than ther smlarly stuated natve born peers (Goldscheder and Goldscheder 1987; Fulgn and Pedersen 2002), especally among Mexcan and Cuban women (Goldscheder and Goldscheder 1987). Ths research also fnds that the reasons for and mplcatons of later famlal home leavng dffer by generaton and race/ethncty. For young adults of European orgn (largely natve born), remanng n the natal home s assocated wth depresson and may represent a faled attempt of strkng out on one's own (Fulgn and Pedersen 2002). But, among Latnos and Asan Amercans, stayng n one's natal home s not negatvely assocated wth emotonal well-beng. These young adults often provde resources to ther parents and contnue to lve wth ther parents out of feelngs of famlal oblgatons rather than desperaton (Fulgn and Pedersen 2002). Home leavng may also have mportant mplcatons for chldren of mmgrants n the assmlaton process. Lvng wth one's parents may be assocated wth slower assmlaton and greater selectve acculturaton (Portes and Rumbaut 2001). Contnued lvng wth parents may mean contnued use of parents natve tongue, greater partcpaton n more cultural tradtons, and contnung to lve n communtes of co-ethncs. Of course, when chldren of mmgrants do leave ther natal home, t s possble that they move to areas of smlar ethnc concentraton, as suggested by evdence on the hghly moble 1.5 generaton (Goodwn-Whte 2007). In addton, not all mmgrants lve n ethnc enclaves. Immgrant parents may acheve mddle class status and resde n the largely whte suburbs (Alba 1999), so ther chldren may never have lved n an ethnc neghborhood and would therefore not leave one when they leave the parental home. Ether way, Portes and Rumbaut (2001) demonstrate that chldren who reman n co-ethnc communtes or wth ther parents are more lkely to experence selectve acculturaton than chldren who leave the parental home and the ethnc communty. THE PRESENT STUDY Young adulthood s a perod marked by dramatc changes, such as home leavng, attendng college and partnerng. Past research has also demonstrated that transton to adulthood s assocated wth rapd gans n weght (Wane et al 2010 Mokdad et al 1999; Ogden et al 2002; Gordon-Larsen et al 2004) and changes n det and physcal actvty patterns (Len et al 2001; Larson et al 2007; Wane et al 2010). Parental montorng of det s assocated wth hgher levels of frut and vegetable ntake (Larson et al 2007), lower uptake of other negatve health behavors, and transtonng to a unversty s assocated wth weght gan (Wane et al 2010). Together, ths evdence suggests that those who contnue to lve at home may have a healther det than those who leave home. For mmgrant youth, leavng home may also mean losng the cultural protecton provded by ther parents. These deas lead to the followng hypotheses: H1: Home leavng s assocated wth weght gan. The benefcal mpact of parental montorng s lost and changes n det are assocated wth weght gan. H2. Ths assocaton s greater for chldren of mmgrants, n general, and largest for the 1.0 generaton compared to chldren of natves. H3. The greater lkelhood of frst generaton mmgrants (both generaton 1.0 and 1.5) to resde wth ther parents partally explans why mmgrants tend to wegh less than chldren of natves. Home-leavng s also assocated wth other mportant predctors of adult status such as college attendance and partnerng. In addton to home leavng, these factors may have mportant mpacts on weght change. For example, pror research had demonstrated that unversty transtons are assocated wth weght gan. However other research, demonstrates that those who attend college have a lower BMI and growth n BMI than those who do not attend 6

7 college (Yang et al 2008). In order to understand ths contradcton, I wll examne college attendance usng a growth curve model lookng at both whether college attendance s assocated wth more or less weght gan compared to those who do not attend college, as well as whether the assocaton between weght gan and college attendance s the result of leavng the natal home. Gven the lterature on college attendance and weght gan I hypothesze that college attendance may be assocated wth ntal weght gan (Wane et al 2010). However, those who attend college may have a slower rate of change n the growth of BMI. Ths hypothess s based off of the large body of lterature that fnds worse health outcomes among adults wth less educaton (Ross and Mrkowsky 2003). Lastly, home leavng s often assocated wth famly transtons, such as marrage or cohabtng. In general, marred ndvduals tend to have better health than unmarred ndvduals on many health ndcators. However ths s not true for overweght or obesty. Marred ndvduals tend to be heaver than never marred ndvduals, especally among men, and marrage s assocated wth an ncrease n weght (Kahn, Wllamson, and Stevens 1991; Schoenborn 2004). Leavng the parental home may be assocated wth weght gan for those who leave to move n wth partners. As such, the potental role of partnerng as a moderatng factor wll be explored. METHODS The data for ths project contnues to use the Natonal Longtudnal Study of Youth (NLSY), 1997 cohort. I use the longtudnal fle from wth annual waves of data assessed from The orgnal cohort conssted of roughly 8,894 respondents. Attrton vared by waves, wth the hghest at 19% n wave 10. Roughly, 11% of the respondents attrte at each wave, though many were followed up at later waves. The data are organzed nto a person perod fle (N = 78,005 records) wth one record contrbuted by respondents for each year they were ntervewed between the ages of 12 and 16 durng the frst wave and 24 to 28 by the last wave. The analytc sample excludes 13% of the total possble records from waves n whch respondents were not ntervewed and an addtonal 15% who had nvald responses for the analytc varables for that wave. Tme s measured as age n months. However, for ease n nterpretablty annual changes n BMI are presented n the tables (.e. coeffcents and standard errors are multpled by a factor of twelve). Dependent Varable BMI. Body mass ndex (BMI) s a commonly used measure to examne adposty among ndvduals and s calculated as weght/heght 2. I use reported heght and weght. Though somewhat controversal I use BMI n ts raw form as opposed to transformng t nto percentle BMI or BMI z-scores. Research that examnes adolescents and chldren converts BMI nto percentles and z-scores to reduce bas due to dfferences n developmental phase (based on age n months) and dfferences n weght gan by developmental phase for boys and grls. However, the BMI percentle and z-score measures created by the Center of Dsease Control were created usng cross-sectonal samples of chldren and adolescents. As such t masks normal changes n weght related to peaks n heght and non-lnear peaks n BMI (Cole et al 2003; Berkey and Coldtz 2006). The varablty wthn ths measure s large and represents a wde range of actual change n adposty. For example, a one z-score ncrease n BMI over the course of a year s assocated wth an ncrease of 2.2 kg/m2 to 8.9 kg/m2 or a weght gan of 13.3 to 53.6 pounds, for a ffteen year old grl holdng heght constant (Berkey and Coldtz 2006). Also the metrc n tme for ths project s age n months, the same measure used by the CDC to calculate BMI z- scores, and estmates are collected on a yearly bass over the course of ten years. Another concern s that percentle BMI standardzes across sex. Boys and grls enter puberty at dfferent ages and boys tend to lose weght durng ths phase whle grls gan weght (Wang 2002). To examne any potental gender bas ntroduced by usng raw BMI I conduct all analyses on both 7

8 the entre sample and then separately by gender. Reported heght and weght are generally very close to each other, but ssues arse f msreportng heght and weght systematcally dffers by generatonal/ethnc groups. For example, Antecol and Bedard (2006) found n ther study of natvty dfferences n BMI that mmgrant women msreported ther weght by a factor of 1.2%, whle natve women ms-reported by 2.3%. There were no natvty dfferences for men. In addton, when usng reported heght and weght to calculate BMI, t s better to leave t contnuous. Usng cut-offs to make classfcatons leads to serous mss-classfcaton errors (obese vs. non-obese), but only small dfferences n BMI (29.5 vs. 30.0) (Stewart et al 1987). Whle there were dfferences n ms-reportng by generaton results usng ether the corrected heght and weght data or the un-corrected heght and weght data produced smlar results. As such, I use reported heght and weght wthout correctng for the error between reported and measured heght and weght. Independent Varable The focus ndependent varable n ths project, besdes generaton and ethncty, s resdental status. At every wave the NLSY has detaled nformaton concernng the household roster. From ths nformaton I am able to determne who the respondent s lvng wth at each wave and whether he/she changed resdental status. Unfortunately, the way the household roster was measured changed between waves. Waves 1 through 6 refer to the respondent s permanent household, rather than current household. Respondents away at college, away for temporary employment, n the mltary, or ncarcerated may have ndcated that they were lvng wth ther parents, when they were actually lvng away from home. To address ths problem, several measures were taken to measure parental coresdence for waves 1-6. Frst, I examned the type of dwellng the respondents occupy. If they ndcated that they lve n a dorm, fraternty/sororty house, or mltary barracks, they were coded as not lvng wth ther parents. Second, I examned whether they reported lvng n a dorm at college for at least three terms. These respondents are coded as lvng away from home. Ths captures those who may have been at home on break when they were ntervewed, though most ntervews were conducted from September to May. Thrd, f they reported they were currently enrolled n college, consdered ther current resdence temporary, and reported ther current dwellng s rented, they were coded as lvng away from home. Lastly, those who prevously reported that ther past dwellng was permanent, ther current dwellng s temporary, and they had not experenced a move snce the last ntervew are coded as lvng away from home. The mgraton questons for these earler waves refer to whether there has been a change n address n ther permanent dwellng. By examnng mgraton I am able to exclude ndvduals who may have moved nto temporary housng wth ther parents, from those who moved nto temporary housng wthout ther parents. If ther household moved wth them nto temporary housng than ther mgraton varable captures ths move, but f they moved wthout ther parents and dd not consder ther current housng as permanent than ther mgraton varable does not reflect ths move. Ths captures ndvdual who may have been away from home for work or other reasons, but dd not consder ther current resdence permanent. Addtonally, those who reported that they were ncarcerated are coded as lvng away from home. In subsequent waves (waves 7-12), the household roster was changed to reflect current household composton rather than permanent, so t was possble to drectly measure whether the respondent lved wth hs/her parents. The tme-varyng controls ncluded n ths analyss are self-reported health, smokng, number of chldren, martal status, pregnancy, and employment. The non-tme varyng control varables are mother's educaton, parent's weght status, famly status at wave 1, and gender. Mother's educaton and parent's weght status are nteracted wth tme because they have a 8

9 sgnfcant effect on the slope of BMI growth. Methodologcally, the queston of home leavng poses a few problems. As stated earler, the perod of young adulthood s a state of flux. Indvduals often move out of ther parent s house multple tmes and some may never move out. To allow for ths complexty I modeled home leavng three dfferent ways. Frst, I examne home-leavng as a tme-varyng dummy varable and a varable that ndcates whether ths s the frst wave that the respondent reported movng out. Ths allows me to examne whether there s an ntal ncrease n BMI when the respondent frst leaves home and whether ths elevaton s mantaned whle they resde away from home. Second, I examne the amount of tme that has passed snce the respondent frst left ther parents house (measured n months) Ths allows me to examne whether movng out s assocated wth a change n the rate of growth n BMI. Thrd, I examne the reason the respondent moved out of ther parent s house, ether to attend college or move n wth a partner. In ths analyss, I focus on the shft n BMI and the change n rate of growth n BMI as a functon of home leavng. I descrbe these three approaches n greater detal below. These questons are examned usng nested growth curve models. The frst model ncludes all analytc varables except home leavng. The next model ncludes the home leavng varables. Ths allows me to examne whether patterns n home leavng explan dfferences n BMI trajectores by generaton (hypothess 2). Fnally, model 3 ncludes the nteractons between home leavng and generaton. Ths allows me to see whether the effect of home leavng dffers by generatonal status (hypothess 3). Approach #1: Parental Coresdence and Home Leavng Past research suggests that the frst year of resdental ndependence may be especally assocated wth weght gan (Wane et al 2010). By examnng the frst wave that the respondent left home I am able to see f the frst year the respondent leaves home s an especally susceptble to weght gan. Fgure 1 llustrates ths hypothess by comparng two ndvduals, one who never moved out of her parents home wth one who moved out at age 18. The ndvdual who moved out ganed weght durng the frst wave she left home and contnued to wegh more than the person who dd not leave home. To capture ths possble pattern n my models, I examned whether the effect of home leavng s assocated wth an ntal shft n BMI and whether ths shft n BMI s mantaned. To do ths, I use a tme-varyng dchotomous varable ndcatng whether the respondent resdes outsde the parental home (1=currently resdes away from home, 0=lves wth parents) (referred to below as non-parental resdence ), and another dummy varable ndcatng that the respondent left the parental home n the gven wave (referred to below as just left home ). Because respondents can have multple home leavngs, the just-left-home varable s set equal to 1 for the frst wave the respondent reported movng out of the home for that epsode of resdental ndependence. To llustrate, Table 1 presents the codng scheme for all home leavng varables for a hypothetcal respondent who left home n wave 6, came back n wave 8 and subsequently moved out agan n wave 10. The non-parental resdence varable s set to 1 for waves 6 and 7 and waves 10 through 12, and the just-left-home varable equals 1 n waves 6 and 10. Ths codng scheme allows for multple home leavngs, but t does not dstngush between the frst or second tme the respondent left home, nor does t dstngush between those who have moved back n wth ther parents from those who never moved out. Addtonal analyses examned whether movng back n wth one s parents was also assocated wth a change n BMI. Ths s examned by creatng two tme varyng varables. The frst ndcates the frst wave that the respondent reports movng back n wth her parents (1=frst wave moved back n 9

10 wth parents, 0=otherwse). The second varable ndcates whether the respondent contnued to lve wth hs/her parents after movng back home. For the hypothetcal respondent shown n Table 1, the frst varable would equal 1 for wave 8 and the second varable would equal one for waves 8 and 9. Ths allows me to compare those who do not lve wth ther parents, those who do lve wth ther parents but had prevously moved out, and those who have never left ther parents home (the reference category). Fnally, to assess whether the effects of home leavng vary across generatons, I ncluded nteractons between the home-leavng varables and generatonal status. Prelmnary analyses ndcated that the growth curve model does not converge when both home leavng varables are added n the level one model. The goodness of ft statstcs (AIC and BIC) ndcate that the non-parental resdence varable produces a better model ft than the homeleavng varable. Includng ths varable n the level 1 equaton results n the followng expresson: Y = π + π Tme + π Tme π non-parental resdence + ε 3 [Eq 4.1]. I ncluded fxed effects n four dfferent level 2 equatons to determne ther nfluence on the level 1 coeffcents, where: π = γ + γ Famly of Orgn SES + γ Generaton + γ just-left-home + γ * just-left-home Generaton π = γ + γ Generaton π = γ + ζ π 3 = γ 30 + γ 31 Famly of Orgn SES * just-left-home Generaton + ζ 3 + Z 0 + γ 12 + Z 1 +ζ 0 Generaton +ζ 1 + γ just-left-home + γ Or n composte equaton (by substtutng the level-2 equatons nto the level-1 equaton): Y = [ γ + γ Famly of Orgn SES + γ Generaton + γ just-left-home + γ (Generaton * just-left-home) [γ + γ just-left-home) [γ + [γ 30 [ε 20] *Tme 2 + γ 31 + ζ 0 + Z ] + 0 Famly of Orgn SES + Z ]*Tme 1 Generaton ]* Non-parental resdence + ζ 1 + *Tme + ζ 2*Tme 2 + γ Generaton + γ just-left-home + γ (Generaton * ζ 3 + * Non-parental resdence I estmated ths model for the whole sample and for the whte, black and Mexcan-Amercan subsample. ]. 10

11 Approach #2: Duraton Effects I next examne whether the tme snce leavng home s assocated wth a change n the slope of BMI growth. Ths s mportant because home leavng may be assocated wth a change n the slope of the BMI trajectory (.e., the rate of weght gan), beyond any ntal change n elevaton of BMI. Fgure 2 llustrates ths hypothess for two respondents: one who never moved out and one who moved out at age 18. At age 18, there s the ntal shft n BMI. The nonparental resdence varable captures ths effect; movng out s assocated wth a change n the shft or elevaton of the BMI trajectory. However, movng out of the parental home s also assocated wth a change n rate of BMI growth, as demonstrated by the contnued dvergence n BMI over tme between the respondent who dd move out and the one who dd not. To model ths possble pattern, I ntroduce a new varable ndcatng the tme n months snce the respondent left home (referred to below as duraton ). To create ths varable, I subtract respondents current age from ther age when they frst left home. The frst wave the respondent moves out s equal to zero and ncreases by one for every month the respondent contnues to resde outsde the home. If the respondent moves back home wth a parent, the duraton varable s recoded to zero and the orgnal slope of BMI s assumed. Ths codng scheme s dsplayed n table 1 for the hypothetcal respondent descrbed above. The duraton varable would equal 0 for wave 1 through 6, 12 for wave 7, 0 for waves 8, 9, and 10, 12 for wave 11, and 24 for wave 12, assumng that the waves are exactly a year apart. Addtonal analyses examned whether movng back n wth parents s also assocated wth a change n the elevaton and slope of BMI growth. Ths varable s created the same way as the duraton varable. For respondents who moved back n wth ther parents, ths varable ncreases wth tme as long as the respondent contnues to resde wth ther parents after they moved back home. For our hypothetcal respondent, ths would equal 12 for wave 7 and 0 for all other waves (see table 1). Prelmnary analyses ndcated that the model does not converge when both home leavng varables are added n the level one model. The goodness of ft statstcs (AIC and BIC) ndcate that the non-parental resdence varable produces a better model ft than the duraton varable. Thus, non-parental resdence s added n the level 1 model and tme snce left home ( duraton ) s ncluded only n the level 2 equaton. The level 1 equaton for ths model then becomes: Y = π + π Tme + π Tme 2 + π Non-parental resdence + ε [Eq 4.2]. I ncluded the fxed effects results n three dfferent equatons to determne the nfluence of the level 1 predctors. Where: π = γ + γ Famly of Orgn SES + γ Generaton + γ Duraton + γ (Generaton * ) + Z Duraton π = γ + γ Duraton π = γ + ζ ζ 0 Famly of Orgn SES ) + Z 1 +ζ 1 π = γ + γ Generaton + ζ γ Generaton + γ Duraton + γ (Generaton *

12 Or n composte equaton (by substtutng the level-2 equatons nto the level-1 equaton): Y = [γ + γ Famly of Orgn SES + γ Generaton + γ Duraton + γ (Generaton * ) + Z ] + Duraton [γ + γ Duraton [γ + γ [ε + ζ 0 0 Famly of Orgn SES ) + Z 1 Generaton + ζ 1 + γ Generaton + γ Duraton + γ (Generaton * ]*Tme + [γ ]*Tme 2 20 ]* Non-parental resdence + *Tme + ζ 2*Tme 2 + ζ 3 * Non-parental resdence]. Approach #3: Reasons for Leavng Home Lastly, I examne whether the effect of home leavng on weght s moderated by where the respondents go when they leave home, specfcally whether they lve wth a partner (lvng wth ether a spouse or boyfrend/grlfrend) or go to college. Among adults, college attendance s negatvely assocated wth BMI. However past research suggests that ntal college attendance s a perod of rapd weght gan. Those who attend college may gan weght ntally, but have a slower growth n BMI. Addtonally, whle the effect of home-leavng s unknown, past research consstently demonstrates that ndvduals wth partners tend to wegh more than people wthout partners (Kahn, Wllamson, and Stevens 1991; Schoenborn 2004). Therefore, the effect of home leavng, when t happens n the context of partnerng, may be especally assocated wth ncreases n weght. To examne these potentally medatng and moderatng effects of partnerng and college attendance, I focus on the ntal reason that the respondent left home, ether to attend college or move n wth a partner. Whle ths queston s not explctly asked n the survey, I nfer ths by examnng whether the respondent was enrolled n college or lvng wth a partner durng the frst wave they reported they were no longer lvng wth ther parents. I used the same codng scheme descrbed prevously to create duraton varables ndcatng the number of months snce the respondent left home. The only dfference n ths analyss s that I created three dfferent mutually exclusve duraton varables ndcatng the tme snce leavng home for college, tme snce leavng home for partner, and tme snce leavng home for other reasons. I also used a modfcaton of the non-parental resdence varable ndcatng the reason for non-parental resdence: non-parent resdence and left for college, non-parent resdence and left for partner, and non-parent resdence and left for other reasons (ref. = stll lvng wth parents). For example, f the prevous hypothetcal respondent reported she left home the frst tme for college and the second tme to lve wth a partner, then her person perod record would be as follows. Non-parent resdence and left for college would equal 1 for waves 6 and 7 and 0 for all other waves. The duraton varable for college would equal 12 n wave 7 and 0 for all other waves. Non-parent resdence and left for partner would equal 1 for waves 10, 11, and 12 and 0 for all other waves. The correspondng duraton varable would equal 12 for wave 11, 24 for wave 12 and 0 for all other waves. Only the ntal reason for leavng home for that perod of non-parent resdence s examned. For example, an ndvdual who leaves home to attend college would subsequently graduate and may contnue to resde outsde ther parents home after fnshng college. Ths ndvdual would always be coded as leavng home for college (as long as they dd not move back n wth ther parents), even f they eventually are no longer attendng college. If the 12

13 prevous hypothetcal respondent never moved back n wth ther parents after wave 6 they would always be coded as non-parental resdence and left home for college even after they moved n wth ther partner n wave 10. Addtonally, f a respondent leaves home for a partner and does not move back n wth ther parents, they are always coded as non-parent resdence and left for a partner even f that partnershp ends n subsequent waves. Ths analyss s conducted for the pooled sample and the Mexcan, whte, and black specfc sub-sample. However, due to sample sze lmtatons only nteractons wth the second generaton are tested. The frst generaton s dropped from ths model. Addtonally, I do not estmate model 3, whch ncludes the nteractons between home leavng and generaton, for the Mexcan, whte, and black specfc sub-sample due to sample sze constrants. Prelmnary analyses ndcated that the model does not converge when all home leavng varables are added n the level one model. Because the home-leavng varables n ths analyss are so nterrelated, I opted to nclude all home leavng varables n the level 2 porton of the model rather than ncludng some, but not others n the level 1 equaton. The level 1 equaton for ths model s therefore expressed as: Y = π + π Tme + π 0 1 2Tme 2 +ε [Eq 4.3]. I ncluded fxed effects n two dfferent equatons to determne ther nfluence on the level 1 coeffcents, where: π = γ + γ Famly of Orgn SES + γ Generaton + γ Non-parental resdence by ) + Z Reason π = γ + γ (Generaton π = γ + ζ γ (Generaton 04 Famly of Orgn SES * Duraton by Reason * Non-parental resdence by Reason + γ Generaton + γ ) + Z ζ 1 0 +ζ 0 Duraton by Reason Or n composte equaton (by substtutng the level-2 equatons nto the level-1 equaton): + γ 14 Y = [ γ + γ Famly of Orgn SES + γ Generaton + γ Non-parental resdence by (Generaton ) + Z ] + Reason [γ 10 + γ 11 (Generaton ]*Tme 2 + [γ 20 [ε RESULTS + ζ 0 + γ 04 Famly of Orgn SES * Duraton by Reason + ζ *Tme 1 * Non-parental resdence by Reason + ζ 2*Tme 2 ]. + γ Generaton + γ ) + Z + 1 ]*Tme 0 Duraton by Reason + γ 04 Descrptve Fndngs Table 1 presents summary measures on home leavng for the sample and by generaton.

14 In accordance wth past research, I fnd that the frst generaton s the least lkely to leave ther natal home durng ths tme perod (70% of the frst generaton ever leaves ther natal home) and ths s sgnfcantly dfferent than the later generatons. The second generaton s more lkely to leave home than the frst, but less lkely than the thrd or hgher generaton, 78% of the second generaton ever leave home. The thrd generaton s the most lkely to leave home. 85% of the thrd or hgher generaton ever moves out of ther parents home durng ths tme perod. Whle home leavng s less lkely for all chldren of mmgrants, especally the frst generaton, the age of home leavng does not vary greatly by generaton. Mean age when the respondent frst left home s roughly 20.6 for the 1.0, 1.5, and 2 nd generaton, and 20 for the thrd or hgher generaton; ths dfference s not sgnfcant. As mentoned earler, young adults often move n and out of the parental home. I allow for ths fluctuaton n lvng arrangements by measurng resdence at each wave. Of those who ever leave ther natal home, 40% return. The frst generaton s more lkely than ether the second or thrd or hgher generaton to return home (55% vs. 52% and 38%). These results are smlar to those found n other large natonally representatve data sets (Goldscheder and Goldcheder 1994; Whte 1994). Table 2 dsplays mean BMI for those who left home compared to those who dd not by age and generaton. Table 2 suggests that those who leave home early, before the age of 21, have a slghtly hgher mean BMI than those who reman n ther parents home. However, those who leave home later, after the age of 23, have a lower mean BMI than those who reman n ther natal home. However, these dfferences are not sgnfcant. In addton, ths weak assocaton between home leavng and BMI may be due to other confoundng varables that are assocated wth both home leavng and weght, such as famly of orgn SES, partnerng, and race/ethncty. In order to examne ths possblty, I estmate growth curve models to control for several confoundng varables. Approach #1: Non-parental resdence and home leavng I frst present the results of the analyss that estmates the effects of non-parental resdence and home leavng (approach #1). Table 3 presents the growth curve models based on equaton 4.1. Model 1 ncludes all the sample varables, except the home leavng varables. The second model adds the just-left-home and non-parental resdence varables. Ths model demonstrates that currently resdng away from home s assocated wth a sgnfcant shft n BMI. Those who lve away from home have a 0.06 hgher BMI than those who lve wth ther parents. However, the frst wave that the respondent moved out s sgnfcant and negatvely assocated wth BMI. Ths means that ndvduals do not experence the greatest shft n BMI when they frst move out. Rather, the weght they had pror to movng out s mantaned durng the frst wave they moved out ( =0.01), but there s an upward shft n BMI by the second wave the respondent has left home (0.06). Thus, home leavng has a lagged effect on BMI. The second wave that the respondent currently resdes away from home s assocated wth a 0.06 shft n BMI, whch s mantaned for the duraton they resde away from home. The effect of generaton remans relatvely unchanged once the home leavng varables are ncluded n the model. The thrd model adds the nteractons between non-parental resdence, home leavng and generaton. In accordance wth my prevous hypothess, home leavng s assocated wth more weght gan for the frst and second generatons than the thrd or hgher. Non-parental resdence s assocated wth an addtonal 0.29 ncrease n BMI for the frst generaton and an addtonal 0.20 ncrease n BMI for the second generaton (the nteractons and the total effects are sgnfcant). The pattern for the frst generaton s llustrated n fgure 3, whch plots the predcted BMI for a frst generaton ndvdual who never left home versus one who left home at 14

15 the age of 19. Includng the nteracton terms reduces the man effect of leavng home to nonsgnfcance. Because the thrd generaton s the reference category, ths means that leavng home s not assocated wth BMI for the thrd generaton. Addtonal analyses ndcate that among the thrd generaton, the effect of currently resdng away from home s postvely assocated wth BMI, but s medated by the controls for martal status and number of chldren. The postve effect of home leavng on BMI for the thrd or hgher generaton appears to operate through famly transtons undertaken durng ths tme perod. However, among the frst and second generaton, home leavng s sgnfcantly assocated wth weght gan even after famly transtons are taken nto account. Table 4 dsplays the results for the Mexcan-Amercan, NH black, and NH whte sample. The results for ths sample mrror those found for the pooled sample. Non-parental resdence s assocated wth an ncrease n BMI, but only for the frst and second generaton Mexcan- Amercans. For the frst generaton, non-parental resdence s assocated wth a 0.41 ncrease n BMI, and for the second generaton, t s assocated wth a 0.21 ncrease n BMI. The man effect s not sgnfcant once the nteractons are ntroduced to the model, ndcatng that the effect of non-parental resdence s not sgnfcant for the thrd generaton. Addtonally, the nteractons between the just-left-home varable and generaton are also sgnfcant and negatve for the second and thrd generaton Mexcan-Amercans and NH blacks. For the second generatons the nteracton s negatve, ndcatng that home leavng has a lagged effect on BMI and weght gan s not experenced untl the second wave the respondent has lved away from home. For thrd generaton Mexcan-Amercans, movng out of one s parents home s assocated wth a decrease n BMI. However, the total effect s not sgnfcant (total effect=(man effect for just-left-home + nteracton) + (man effect for non-parent resdence + nteracton), or ( ) + ( )= -0.14). Smlarly, the total effect of frst wave for NH blacks s not sgnfcant (b=0.06). Approach #2: Duraton Effects Overall, the results suggest that the ntal wave that the respondent moves out of ther parents home s not assocated wth a sgnfcant ncrease n BMI. Instead, BMI s hgher for those who contnue to resde outsde ther parents home after the frst wave compared to those who resde wth ther parents. However, equaton 4.1 does not consder the effects of duraton spent outsde of the parent s home on BMI. To consder ths possblty I use equaton 4.2. Introducng the duraton varable helps assess whether home leavng s assocated wth a change n the rate of growth of BMI. Table 5 presents the growth curve models for ths equaton. As n the prevous analyss, Model 1 ncludes all the sample varables, except the home leavng varables. The second model adds the home-leavng varables (non-parental resdence and duraton). None of the home leavng varables are sgnfcantly assocated wth BMI. Includng the controls martal status and number of chldren reduces these effects to nonsgnfcance. The effects of generatonal status reman relatvely unchanged after the addton of the home leavng varables. Fnally, Model 3 adds the nteracton terms. The only nteractons to approach sgnfcance are those for the second generaton. For ths group, non-parental resdence s assocated wth a sgnfcant shft n BMI. The nteracton between tme snce left home ( duraton ) and the second generaton s margnally sgnfcant and negatve; however the total effect (.e., the man effect + nteracton) s not sgnfcant. These results so far suggest that home leavng s assocated wth a shft n BMI after the frst wave the respondent lves away from home and s mantaned for the duraton the ndvdual resdes away from home, but there s no change n the rate of growth n BMI. Table 6 dsplays the results for the Mexcan-Amercan, NH black, and NH whte sample, 15