Acculturation and parental attachment in Asian-American adolescents’ alcohol use

INTERNATIONAL ARTICLE

Acculturation and Parental Attachment inAsian-American Adolescents’ Alcohol Use

HYEOUK C. HAHM, Ph.D., MAUREEN LAHIFF, Ph.D., AND NEIL B. GUTERMAN, Ph.D.

Purpose: To test whether the degree of acculturationpredicts subsequent alcohol use among Asian-Americanadolescents, and to test the moderating effect of parentalattachment.

Methods: This was a prospective study using a sub-sample of the National Longitudinal Adolescent Healthdata set. A nationally representative sample of 714 Asian-American boys (n � 332) and girls (n � 382) in grades7–12 was analyzed. In-home self-report data were col-lected on two types of acculturation status, alcohol use,demographics, and parental attachment. After control-ling for acculturation status and background variables atWave I, logistic regression analysis was used to estimatethe odds ratios to assess the association between accul-turation and alcohol use at Wave II for adolescents.

Results: Asian-American adolescents with the highestlevel of acculturation (English use at home, born in theUnited States) were identified as the highest risk group.For adolescents with low parental attachment, the oddsof alcohol use were 11 times greater in the highlyacculturated group than in the least acculturated group.However, the odds of alcohol use for adolescents withmoderate or high levels of parental attachment did notvary across acculturation groups.

Conclusions: Overall, a greater level of acculturationwas associated with greater alcohol use. However, whenparental attachment was taken into account, highly ac-culturated adolescents with moderate or high parentalattachment had no greater risk than adolescents withsame levels of parental attachment who were less accul-turated. Thus, it appears that acculturation per se was nota risk factor unless it was accompanied by a low level of

parental attachment. © Society for Adolescent Medicine,2003

KEY WORDS:AcculturationAlcohol useAsian-American adolescentsRisk and protective factorsParental attachment

Adolescent alcohol use is a significant problem in theUnited States [1]. Suicide, homicide, and motor ve-hicle injuries are related to adolescent alcohol use,and the prevalence of adolescent substance use hasbeen growing rapidly [2].

Although numbers of studies have focused onadolescents’ alcohol use, studies that have focusedon Asian-American adolescents’ alcohol use havebeen extremely limited [3]. Those studies [1,3,4] thatinvestigated this issue have demonstrate that Asian-American adolescents’ substance use is a complexand significant problem. Asian-American adoles-cents who drink are often involved in heavy drink-ing, and those who binge drink consume morealcohol than white adolescents who binge drink [4].

One of the important factors that elucidates ethnicminority adolescents’ substance use is the role ofacculturation [5–8]. Acculturation is the process bywhich an ethnic group gradually changes its behav-iors and attitudes to be more like those of the hostsociety [9]. Acculturation is also described as theprocess by which foreign-born individuals and theirfamilies learn and adopt the language, values, be-liefs, and behaviors of the new sociocultural environ-ment [6].

From Columbia University School of Social Work, New York, NewYork (H.C.H., N.B.G.); and University of California, Berkeley, School ofPublic Health, Berkeley, California (M.L.).

Address correspondence to: Hyeouk C. Hahm, Ph.D., University ofCalifornia, School of Social Welfare, 120 Haviland Hall, Berkeley, CA94720. E-mail: [email protected].

Manuscript accepted January 31, 2003.

JOURNAL OF ADOLESCENT HEALTH 2003;33:119–129

© Society for Adolescent Medicine, 2003 1054-139X/03/$–see front matterPublished by Elsevier Inc., 360 Park Avenue South, New York, NY 10010 doi:10.1016/S1054-139X(03)00058-2

The most developed theory of acculturation isfound in the work of Berry [10] who proposed fourpossible strategies of acculturation (integration, as-similation, separation, and marginalization) on thebasis of how individuals wish to maintain theiroriginal culture and the extent to which individualswish to have day-to-day interactions with the newcultural group. Berry argued that integrationistshave better mental health than those who are assim-ilated (or acculturated), separated, or marginalized.The process of acculturation takes place over time,and it involves two independent cultural groupswith different characteristics [7,10]. Immigrant Asianparents are more likely to have “collective andinterdependent” cultural values that put more em-phasis on interdependence and cooperation. In con-trast, American-born Asian-Americans have greaterexposure to American cultures, which value “inde-pendence,” “self-actualization,” and “expressingone’s unique configuration of needs, rights, andcapacities” [11,12]. To succeed in U.S. work andschool environments, immigrants and their childrenmust develop an ability to function in an individu-alistic culture [12]. Being able to shift frameworksfrom a collective-oriented culture to an individualis-tic culture poses great challenges to immigrant fam-ilies [7,13,14].

Ying et al [13–15] have articulated why intergen-erational and intercultural conflicts are inevitableamong parents and adolescents who are at differentstages of the acculturation process. These conflictsmay develop owing to the contrasting cultural values[10,13–15] and incompatibility of language compe-tence between parents and adolescents [7]. Accultur-ated adolescents, who feel more comfortable withEnglish, must deal with ongoing frustration andobstacles in expressing their thoughts and feelings totheir parents who are less fluent in English. As aresult, the lack of meaningful communication be-tween parents and adolescents is often perceived byadolescents as an absence of parental interest andinvolvement in their lives [16]. According to Ying[14], highly acculturated Chinese-American adults(second-generation) reported that they felt less un-derstood by their mothers than unacculturated Chi-nese-American adult children (first generation).

In addition to a lack of communication andattachment to their parents, highly acculturatedadolescents also face struggles to choose betweenthe values of their parents and the values ofAmerican society and their peers without neces-sarily getting assistance to negotiate these contra-dictions [13]. As a result, more acculturated ado-

lescents develop a strong sense of separation fromtheir parents and their parents’ cultural values,which may lead to an increased reliance on peersas a source of behavior codes and values [17].Therefore, more acculturated adolescents who areweakly attached to their parents and their parents’cultural values may have an increased susceptibil-ity to substance use [7]. Delinquent behaviorsduring adolescence are relatively common inWestern cultures compared with Asian cultures,even among those individuals who were well-behaved during childhood and who will becomelaw-abiding adults [17]. Oetting and Beauvais’scultural identification theory [18] also implies thatlow levels of cultural identification are associatedwith negative psychosocial characteristics, includ-ing lower self-esteem, poor school performance,negative personal adjustment, and fewer personaland social resources; hence, this is likely to in-crease the vulnerability to drug use among minor-ity children.

If acculturation is a risk factor for Asian-Americanadolescents’ alcohol use, what factor could moderatethe relationship between acculturation and alcoholuse? The family interaction theory (FIT) by Brook etal [19] asserts that a strong parent–child attachmentis a powerful protective factor for preventing adoles-cents’ drug use because a strong attachment resultsin the parents having a greater influence on theadolescent. The FIT also suggests that acculturationdifficulties among adolescents are associated withdifficulty in the parent–child attachment relation-ship, and this also increases the adolescents’ devel-opment of substance use. Furthermore, Brook et al[19] proposed that although a sociocultural riskfactor (such as an acculturation gap) may increaseadolescents’ development of substance use, a strongparent–child attachment can buffer the effect of asociocultural risk factor.

This longitudinal study represents the first sys-tematic and prospective investigation examiningboth acculturation and a parental protective factorassociated with alcohol use among Asian-Americanadolescents. On the basis of the acculturation theory,cultural identification theory, and FIT, we hypothe-sized that (a) highly acculturated Asian-Americanadolescents have greater rates of alcohol use; and (b)highly acculturated Asian-American adolescentswho have a strong parental attachment drink lessthan highly acculturated Asian-American adoles-cents who do not have a strong parental attachment.

120 HAHM ET AL JOURNAL OF ADOLESCENT HEALTH Vol. 33, No. 2

MethodsData Source and Sample

This study used data from the National LongitudinalStudy of Adolescent Health (Add Health). AddHealth data were collected by the University ofNorth Carolina (UNC) School of Public Health, andthe UNC Institutional Review Board on ResearchInvolving Human Subjects approved all proceduresfor the Add Health study. A stratified nationallyrepresentative sample of all public and private highschools in the United States was collected by QualityEducation Data, Inc. A sample of 80 pairs of eligiblehigh schools and feeder schools were stratified onthe basis of region, urbanicity, school type (public,private, and parochial), and racial composition toensure that the sample would be representative ofU.S. schools. An in-school questionnaire was givento students in grades 7–12 between September 1994and April 1995.

Approximately 90,000 adolescents completed thein-school questionnaire. All students who completedan in-school questionnaire, as well as those who didnot complete a questionnaire but who were listed ona school roster, were eligible for selection into the“core sample.” This core sample was gathered bychoosing 17 students for each grade and gender whowere randomly selected from each grade-genderstratum, so that a total of about 200 adolescents wereselected from each of the 80 pairs of schools. Of thosein the core sample, 12,105 participated in the Wave Iin-home interview, for a response rate of about 75%[20].

Wave II in-home interviews took place betweenApril and August 1996 (1 year later). Wave II surveyswere completed by 9278 of Wave I respondents(77%). The same sample of Wave I respondentsparticipated in Wave II, except for high school se-niors who had aged out of the study [20].

The subjects for this present study were Asian-American adolescents who participated in bothWave I and Wave II interviews. The total number ofAsian-American adolescents who participated inWave I was 1584. This number dropped to 1048 inWave II, owing to the “aging out” of Wave I highschool seniors. Of this 1048, about one-third hadmissing values owing to skipping questions; there-fore, the final sample size for the analyses totaled714. No difference was noted in gender distributionbetween study participants and non-study partici-pants. Study participants were 6 months youngerthan non-study participants. The study participants’socioeconomic status (SES) was 0.3 greater than

nonparticipants on a scale of 1–11. Study participantshad a mean self-esteem score that was 0.7 greater ona scale of 0–55.

Measurement

All the variables were selected from a variety ofstandardized, validated measurements used in na-tional and state surveys of adolescents’ mental healthand psychosocial functioning. The designers of theAdd Health study selected the instruments using aresiliency framework [20]. Resnick et al [21] pro-vided the evidence of the construct validity of thescales and internal consistency indexes (Cronbach �).

Predictors

The background variables included age, gender, SES,self-esteem, and school attachment. Age was mea-sured in years at the time of the Wave I homeinterview. Gender was coded “0” for male and “1”for female. Parental occupational status and educa-tional attainment were used to assess the SES. Acomposite scale for SES was created by coding occu-pation and education. Parents’ occupations werecoded according to a five-point ordinal scale usingoccupational categories ranging from blue collar/laborer (low) to white collar/professional (high). Thescores of the education and occupation scale weresummed, and finally, the SES score was assignedaccording to the higher SES score of either parent,with a Cronbach � coefficient of .73 in Wave I. Thesemeasures have been used successfully in a study byFord et al [22] and our previous study [23]. Self-esteem was measured using the mean score of 11items collected at an in-school interview. A five-point scale (“strongly agree” to “strongly disagree”)was used by asking how strongly the student agreedor disagreed with each of the statements: “I have alot of energy”; “I seldom get sick”; “when I do getsick, I get better quickly”; “I am well-coordinated”; “Ihave a lot of good qualities”; “I am physically fit”; “Ihave a lot to be proud of”; “I like myself just the wayI am”; “I feel like I am doing everything just aboutright”; “I feel socially accepted”; and “I feel lovedand wanted.” Resnick et al [21] demonstrated thatthis scale has construct validity. The Cronbach �coefficient for this study sample was .84. Schoolattachment was measured by summing the followingsix items: “I feel close to people at my school”; “I feellike I am a part of my school”; “students at my schoolare prejudiced”; “I am happy to be at my school”;

August 2003 ASIAN-AMERICAN ADOLESCENTS’ ALCOHOL USE 121

“the teachers at my school treat students fairly”; and“I feel safe in my school.” A five-point scale (“strong-ly agree” to “strongly disagree”) was used. This scalewas based on the concept of an individual’s interac-tion with their school environment [24], and also hasan acceptable construct validity [21], with a Cron-bach � coefficient of .70.

Study Predictor Variables

Acculturation. Acculturation was measured usingtwo separate variables: the use of English at home(“English at home” � 1, “no English at home” � 0)and the place of the adolescent’s birth (“U.S. born” �1, “foreign born” � 0). According to these variables,participants were categorized into four differentgroups: Group 1, English at home, U.S. born; Group2, English at home, foreign born; Group 3, no Englishat home, U.S. born; and Group 4, no English at home,foreign born.

Studies [7,25] cited in the literature review havedemonstrated that these groups lie on a continuumfrom most acculturated to the least. Those in the firstgroup (English at home, U.S. born) are most familiarwith the American culture; they were considered themost acculturated group in this study. Phinney [8]emphasized that language use is a more accurateassessment tool for acculturation than place of birth.Therefore, this study assumed that those who speakEnglish at home and are foreign born (Group 2) werethe second most acculturated group, and those whodo not speak English at home, but were born in theUnited States (Group 3) were the third acculturatedgroup. Finally, those who do not speak English athome and were born outside of the United States(Group 4) were assumed to be the least acculturatedgroup.

Parental attachment. This was examined in theAdd Health by asking adolescents to agree or dis-agree on a five-point scale with the following state-ments: (a) “Most of the time, my mother is warm andloving toward me”; (b) “I am satisfied with the waymy mother and I communicate with each other”; (c)“How close do you feel to your mother?”; (d) “Howclose do you feel to your father?”; (e) “How much doyou think your mother cares about you?”; (f) “Howmuch do you think your father cares about you?”; (g)“How satisfied are you with your relationship withyour mother?”; (h) “How satisfied are you with yourrelationship with your father?”; and (i) “How muchdo you feel that your parents care about you?” Thesenine items were summed and used as a composite

scale of parental attachment. This scale was adoptedby Resnick et al [21] and was also successfully usedin our previous study [23]. The Cronbach � coeffi-cient for this sample was .90.

Criterion Variable

The question from Add Health, “During the past 12months, on how many days did you drink alcohol?”was recoded as “1” for one or more days of drinkingand “0” for none. This criterion variable was used byHarris [17].

Table 1. Number and Percent Distribution of Asian-American Adolescents, by Background Characteristics,National Longitudinal Study of Adolescent Health, 1995(n � 714)

CharacteristicUnweighted

(%)Weighted

(%)

GenderGirls 382 (53.5) 412 (58)Boys 332 (46.5) 301 (42)

Age (yr)12–14 116 (16.2) 200 (28)15–18 578 (80.9) 507 (70.9)19–21 20 (2.9) 7 (1.1)

EthnicityFilipino 316 (44.3) 276 (38.6)Chinese 221 (31) 151 (21.1)Korean 35 (4.9) 51 (7.1)Japanese 52 (7.3) 70 (9.6)Vietnamese 22 (3.1) 60 (8.4)Asian Indian 20 (2.8) 32 (4.5)Other Asians 117 (16.4) 152 (21.3)

Family SESUpper 129 (18.1) 174 (24.4)Middle 482 (67.5) 452 (63.3)Low 103 (14.3) 88 (12.3)

Birth placeU.S. 417 (58.4) 369 (51.7)Other 297 (41.6) 345 (48.3)

Years in the U. S.� 6 104 (17.1) 015 (17.3)6–10 99 (16.3) 113 (18.5)11–15 186 (30.6) 222 (38.5)16–20 218 (35.8) 154 (25.7)

English use at homeYes 469 (65.7) 445 (62.3)No 245 (34.3) 269 (37.7)

Acculturation groupGroup 1 English, U.S. born 335 (46.9) 292 (40.9)Group 2 English, foreign born 134 (18.8) 153 (21.2)Group 3 No English, U.S. born 82 (11.5) 78 (10.9)Group 4 No English, foreign born 163 (22.8) 191 (26.8)

On the basis of the sampling design, the observations wereweighted to create estimates of the population proportions; per-centages do not always add to 100 because of rounding, andnumbers do not always add to total owing to missing data.

SES � socioeconomic.

122 HAHM ET AL JOURNAL OF ADOLESCENT HEALTH Vol. 33, No. 2

Statistical Analysis

Add Health used a stratified cluster sampling de-sign. It is important to understand the samplingmethod of Add Health because a stratified clustersampling design produces different estimates ofstandard errors and confidence intervals comparedwith simple random sampling [26]. Adjusting thestandard errors and confidence intervals is criticalfor precision of data interpretation [27].

To handle the design effects of Add Health, all thestatistical analyses were performed using Stata, ver-sion 6.0, software. This statistical package has anumber of commands designed to handle complexsurvey data. Stata allows the estimation of confi-dence intervals that account for cluster sampling.Stata also handles probability sampling weights andstratification, which were essential for analyzing theAdd Health data set [28].

Stata’s logistic command allows users to computeodds ratios from parameter estimates and linearcombinations of parameter estimates, which is re-quired for models that include interactions. Stata’slincom command was used to compute the oddsratios in Table 5. Statistical significance was definedat an � level of .05. The results reported here werebased on the most rigorous and conservative ap-proach that considers clustering and stratification,producing the highest standard errors.

ResultsTable 1 presents the number and percent distributionof Asian-American adolescents by background char-acteristics in this study. The estimated populationproportions were weighted by the region of theschool, grade, and gender. The mean age of allAsian-American adolescents included in this studywas 16.1 years (SD 1.5). Filipinos represented thehighest proportion of Asian-American adolescents inthis study. A significantly greater proportion of

Asian-American adolescents spoke English ratherthan other languages at home. Most Asian-Americanadolescents were categorized as having middle SES.

The most acculturated group, Group 1, repre-sented the greatest proportion among Asian-Ameri-can adolescents. The second largest group wasGroup 4, representing 23% of Asian-American ado-lescents.

The proportion of Asian-American adolescents’alcohol use varied according to their acculturationstatus in both Wave I and Wave II. The results ofChi-square tests indicated that systematic differencesexist (Wave I Chi-square � 35.6, df � 3, p � .01;Wave II Chi-square � 28.7, df � 3, p � .05). Group 1(English use at home, U.S. born) had substantiallyand consistently greater rates of alcohol use (45% inWave I; 42% in Wave II). Group 3 (No English athome, U.S. born) had the lowest use of alcohol use inboth Wave I and Wave II (19.7% in Wave I; 22% inWave II).

Table 2 presents the mean and standard deviationfor the covariates among Asian-American adoles-cents by their acculturation groups. For each vari-able, analysis of variance was conducted to assess thedifference, if any, among the four acculturationgroups. For self-esteem, parental attachment, andschool attachment, no differences were found. En-glish speakers had higher family SES (Groups 1 and2); Group 4 had the lowest SES.

Table 3 shows the adjusted odds ratios and 95%confidence intervals for a hierarchy of logistic regres-sion models for alcohol use in the past 12 months.Likelihood ratio Chi-square tests to compare the fitof model 1 with model 2 and model 2 with model 3are also given.

Model 1, a baseline model, showed that age andself-esteem were significant predictors of alcohol use.Acculturation status was included in model 2. Model2 indicated that the most acculturated group (Group1) was associated with greater alcohol use, indicating

Table 2. Mean and Standard Deviations of Predictors Among Asian-American Adolescents by Acculturation Group

English, U.S.-Born(Group 1, n � 335)

English, Foreign-Born(Group 2, n � 134)

No English, U.S.-Born(Group 3, n � 82)

No English, Foreign-Born(Group 4, n � 163) F

Age (yr) 15.9 (1.5) 16.5 (1.5) 15.9 (1.4) 16.5 (1.4) 8.0***Family SES (1–11) 7.3 (2.4) 7.5 (2.0) 6.3 (2.7) 5.9 (2.6) 17.5***Self-esteem (0–55) 43.5 (6.3) 44.0 (5.7) 43.5 (5.6) 42.1 (5.5) 2.4Parent attachment (0–65) 55.2 (7.2) 54.2 (8.6) 56.0 (6.4) 55.4 (7.6) 1.4School attachment (0–30) 21.8 (3.2) 21.6 (3.6) 22.1 (3.1) 21.8 (2.8) 0.4Friends substance use (0–6) 1.7 (1.8) 1.4 (1.8) .96 (1.6) 1.0 (1.4) 6.5***

*** p � .001.SES � socioeconomic status.

August 2003 ASIAN-AMERICAN ADOLESCENTS’ ALCOHOL USE 123

that the odds of those in Group 1 being involvedwith alcohol use was 3.2 times greater than for thosein Group 4 (no English at home, foreign born).Therefore, hypothesis 1 was supported by the anal-ysis.

Model 3 presented the interaction effects to testthe moderating effect of parental attachment. Theoverall likelihood ratio Chi-square test indicated thatmodel 3 fit better than model 2 (Chi-square � 10.4, df� 3, p � .002). The interaction term “Group 1(English use at home, U.S. born) * parental attach-ment” was statistically significant, evidence that thedifference between Groups 1 and 4 varied with thedegree of parental attachment. The odds ratio for the

risk of alcohol use for the most acculturated adoles-cents with a higher degree of parental attachmentwas lower than the odds ratio for the most accultur-ated adolescents with a lower degree of parentalattachment. Because parental attachment was in-cluded in model 3 as a continuous variable, theadjusted odds ratio of .85 for “Group 1 * parentalattachment” indicates that for every unit increase inparental attachment, the odds for Group 1 wasreduced by a factor of .85.

Because the cross products for Groups 2 and 3were not statistically significant, no evidence wasfound that parental attachment has a moderatingeffect when these groups were compared withGroup 4. Group 2 (English use at home, foreign born)and Group 3 (no English use at home, U.S. born) didnot have significantly greater odds of alcohol usethan Group 4, adjusting for all the other covariates.

To explore the moderating effect of parental at-tachment in more detail, parental attachment wasdivided into three categories: high (27.3%), medium(45.6%), and low (27.1%). Table 4 shows the logisticregression analysis predicting Wave II alcohol usewhen using the three levels of parental attachment.As in Table 3, Group 4 (no English use at home,foreign born) was the reference group for accultura-tion. For parental attachment, the lowest level wasthe reference group. Thus, the adjusted odds ratio forGroup 1 (English use at home, U.S. born), 11.27,compared the odds of alcohol use for the mostacculturated adolescents at the lowest level of paren-tal attachment with the least acculturated adoles-cents at the lowest level of parental attachment.Because the 95% confidence interval for this oddsratio did not include 1, the data clearly indicate thathighly acculturated adolescents with low parentalattachment are at a higher risk of alcohol use.

The cross product (“Group 1 * high parentalattachment”) was statistically significant, showingthat the odds ratios comparing Group 1 (English useat home, U.S. born) and Group 4 (no English use athome, foreign born) varied with level of parentalattachment. Because the cross product odds ratio foralcohol use in Group 1 was less than 1, the odds ratiofor alcohol use in Group 1 vs. Group 4 was lower forthose with high parental attachment than for thosewith low parental attachment.

Because none of the cross products for Group 2(English use at home, foreign born) and Group 3 (noEnglish use at home, U.S. born) was statisticallysignificant, no significant changes occurred in theodds ratios for alcohol use between those two groups

Table 3. Logistic Regression Analysis of PredictingWave II Alcohol Use Adjusting for Parental Attachmentas a Continuous Scale (n � 684)

PredictorVariable

AdjustedOddsRatio

(95% CI)

LikelihoodRatio

Chi-squaretest

Model 1 (baseline)Gender 1.11 (.64–1.93)Age 1.25 (1.09–1.42)**Family SES 1.13 (.99–1.28)Self-esteem .95 (.91–.99)*School attachment .97 (.89–1.06)

Model 2Gender .95 (.48–1.89)Age 1.32 (1.12–1.57)**Family SES 1.07 (.94–1.24)Parental attachment .99 (.94–1.03)Self-esteem .97 (.91–1.02)School attachment .99 (.89–1.09)

Acculturation statusGroup 1, English, U.S.-born 3.17 (1.34–7.43)**Group 2, English, foreign born 1.82 (.67–.98)Group 3, no English, U.S.-born .53 (.22–2.34)Group 4, no English, foreign

bornReference group 31.1***

Model 3Gender .92 (.45–1.90)Age 1.30 (1.10–1.54)**Family SES 1.07 (.94–1.24)Parental attachment .96 (.93–1.02)Self esteem .98 (.89–1.02)*School attachment 1.00 (.89–1.11)Group 1, English use, U.S.-born 3.47 (1.56–7.70)**Group 2, English use, foreign

born1.92 (.72–5.09)

Group 3, no English use, U.S.-born .75 (.22–2.61)Interaction

Group 1, * parental attachment .85 (.77–.94)**Group 2, * parental attachment .91 (.81–1.04)Group 3, * parental attachment .91 (.75–1.09) 10.4**

* p � .05; ** p � .01; *** p � .001.CI � confidence interval; SES � socioeconomic status.

124 HAHM ET AL JOURNAL OF ADOLESCENT HEALTH Vol. 33, No. 2

and Group 4 as the level of parental attachmentvaried.

Table 5 provides the 95% confidence interval forthe odds ratio that compares Group 1 (English use athome, U.S. born) and Group 4 (no English use athome, foreign born) for adolescents with low paren-tal attachment. To obtain the corresponding oddsratios and confidence intervals for moderate andhigh levels of parental attachment, which involvedthe Group 1 term and the respective cross products,Stata’s lincom command was used.

For Group 1 (English use at home, U.S. born) andGroup 4 (no English use at home, foreign born)adolescents with low parental attachment, the oddsof alcohol use were 11.27 times greater in Group 1than in Group 4. For adolescents from Groups 1 and4 with a moderate level of parental attachment, thepoint estimate of the odds ratio for alcohol use was2.66. However, because the 95% confidence intervalincluded 1 for adolescents with a moderate degree ofparental attachment, no evidence was found that themost acculturated adolescents (Group 1) had anygreater odds of alcohol use than the least accultur-

ated adolescents (Group 4). Likewise, for the adoles-cents with a high level of parental attachment, noevidence was found that the most acculturated ado-lescents (Group 1) and the least acculturated adoles-cents (Group 4) had any different odds of alcoholuse.

DiscussionThe results of this study provide evidence that themost acculturated group (English speaking at home,U.S. born) had the highest risk of using alcohol inboth Wave I and Wave II. Logistic regression analy-sis also indicated that after controlling for the back-ground variables, the most acculturated group wasthree times more likely to be involved in the subse-quent use of alcohol compared with the least accul-turated group (no English at home, foreign born).This finding was similar to the findings of Unger et al[6] that Asian-American adolescents who spoke En-glish only had a greater level of cigarette use com-pared with Asian-American adolescents who spokean Asian language or who spoke both English and anAsian language.

In bivariate analyses, the proportion of health riskbehaviors was the lowest among adolescents whobelonged to Group 3 (no English at home, U.S. born).For instance, about one-half of the most acculturatedgroup reported using alcohol in both Wave I andWave II, and only about one-fourth of Group 3reported alcohol use. In the multivariate analysis, therisk of using alcohol for Group 3 was similar to thatof Group 4 (no English at home, foreign born). Thelower rates of alcohol use among Groups 3 and 4highlight that for Asian-American adolescents, use oftheir ethnic language at home is associated withimportant health benefits. A number of studies havefound that language is an important symbol of ethnicidentity [29, 30]. The acquisition of ethnic language

Table 4. Logistic Regression Analysis of PredictingWave II Alcohol Use, Three levels of ParentalAttachment, Group 4 (No English Use at Home, ForeignBorn) and Low Parental Attachment as Baseline (n �684)

Predictor VariableAdjusted Odds Ratio

(95% CI)

Model 4Gender .97 (.46–2.04)Age 1.30 (1.10–1.53)**Family SES 1.08 (.94–1.24)Self esteem .97 (.91–1.03)School attachment 1.00 (.90–1.11)

Acculturation statusG1 (English use, U.S. born) 11.27 (3.69–34.40)***G2 (English use, foreign born) 5.13 (.84–31.26)G3 (No English use, U.S. born) 1.98 (.20–19.44)G4 (No English use, foreign born) Reference group

Three levels of parental attachmentHigh parental attachment 2.33 (.49–11.24)Moderate parental attachment 2.44 (.51–11.58)Low parental attachment Reference group

Interaction(G1) * High parental attachment .07 (.01–.63)*(G1) * Moderate parental attachment .24 (.04–1.30)(G2) * High parental attachment .18 (.01–2.68)(G2) * Moderate parental attachment .20 (.03–1.72)(G3) * High parental attachment .38 (.02–6.29)(G3) * Moderate parental attachment .19 (.02–2.12)

* p � .05; ** p � .01; *** p � .001.SES � socioeconomic status; CI � confidence interval; G �

group.

Table 5. Odds ratios from Model 4 Comparing the MostAcculturated Group, Group 1 (English at Home, U.S.Born) With Least Acculturated Group, Group 4 (NoEnglish Foreign Born) for Three Levels of ParentalAttachment

ParentalAttachment

Adjusted Odds Ratio(Group 1 vs. Group 4) 95% CI

High .89 .19–4.10Moderate 2.66 .76–9.31Low 11.27 3.69–34.40***

*** p � .001.CI � confidence interval.

August 2003 ASIAN-AMERICAN ADOLESCENTS’ ALCOHOL USE 125

by adolescents also indicates parents’ active roles ininstructing them in specific cultural and social infor-mation. The linguistic similarity between adolescentsand their parents may be associated with a healthierpsychological status and a more supportive familyenvironment because healthy family interactions arecreated and maintained through a sense of sharedunderstanding of culture, given that language andculture are deeply tied to one another [31].

An essential contribution of this study was thediscovery of the moderating effect of parental attach-ment. For highly acculturated Asian-American ado-lescents, the rates of alcohol use were much greateramong those with low parental attachment. How-ever, when the level of parental attachment wasmedium or high, the most acculturated adolescentsdid not have significantly greater alcohol use. Thus,acculturation was not a risk factor per se unlessaccompanied by low parental attachment. This resultstrengthens the findings of Brook et al [32]. In theirstudy with Colombian adolescents, they found thatalthough violence was endemic and illegal drugswere easily available, a close parent–child bond wasa protective factor that contributed to lower mari-juana use and delinquency.

However, it is important to note that in our study,parental attachment was a significant predictor onlyfor Group 1, but it did not show any interaction effectfor the rest of the groups. To explore this finding,additional analysis was conducted to determinewhether the level of attachment was different accord-ing to the adolescents’ drinking status in each group.Those who drank had a lower mean parental attach-ment only in Group 1, but not in Groups 2–4. Inother words, except for Group 1, parental attachmentdid not differ according to drinking status.

Why is parental attachment a protective factoronly for Group 1? We hypothesized that for Groups2–4, alternative protective mechanisms may play arole. Drawing on the research on cross-cultural par-enting styles by Chao [33], we considered the possi-bility that different levels of acculturation requiredifferent parenting styles. The concept of “parentalattachment” is based more on a Western parentingstyle that involves children “feeling loved” and “spe-cial.” Hence, for the adolescents in Group 1, the mostacculturated and westernized, parental attachment isan effective and meaningful parenting style to pre-vent them from engaging in health-risk behaviors.However, parental attachment may not be culturallyrelevant to the other groups, because they are moreembedded in an Asian culture. In an Asian culturalcontext, parents exercise a “training parenting style”

that is described as parental control, prioritizingobedience and continuous monitoring and guidanceof children’s behaviors [33]. Future research couldshed light on the effective parenting style dependingon the adolescents’ acculturation status.

It is notable that no difference was found inalcohol use among girls and boys. This finding issimilar to the results from the National HouseholdSurvey on Drug Abuse in that between youths aged12 and 17 years, boys and girls in the United Stateshad comparable rates of alcohol use in 2001 [34].However, Asian-American boys had higher rates ofbinge-drinking and higher rates of admission tosubstance abuse treatment facilities [35].

This study had some limitations. First, accultura-tion was only measured by the use of English athome and place of birth. Although the use of Englishat home and place of birth have been viewed asrough indicators, they have been used extensively asproxy measurements of acculturation in numerousempiric studies of Asians [5,36,37] and Hispanics[38,39] and have been shown to be predictive ofimportant outcome variables. Language use is themost frequently used variable for measuring of ac-culturation [5,6]. Epstein et al [40] reported thatlanguage use has been proven to correlate highlywith more comprehensive acculturation scales andthat it also accounts for a large proportion of thevariance among individuals.

However, more specific variables, such as read-ing skills, writing skills, ethnic interaction, ethnicidentification, ethnic activities, knowledge aboutcultural heritage, and social affiliation and pride,reflect and capture the various components ofacculturation [41]. The following questions stillneed to be addressed in future research: Howshould acculturation best be measured? Is theremore than one dimension to acculturation? Howcan acculturation measures with good psychomet-ric properties be developed? How should research-ers ascertain the equivalence of acculturation mea-surements across different cultural groups?Psychometrically appropriate scales that satisfythe standards of reliability and validity must ac-count for the multiple linkages between accultur-ation and the psychological and behavioral pro-cesses of Asian-Americans.

Second, now that this study has demonstratedthat acculturation is associated with alcohol use forAsian-American adolescents, it would be helpful tostudy specific Asian heritage groups. This wouldyield more precise estimates of the association be-tween acculturation and parental attachment. Be-

126 HAHM ET AL JOURNAL OF ADOLESCENT HEALTH Vol. 33, No. 2

cause of the diversity among subgroups and theirimmigrant experiences, the role played by thesevariables may differ. However, to achieve adequatestatistical power, different ethnic groups were notcontrolled for in this study.

Third, the findings from this research project werebased on self-reported data. Therefore, the findingsare dependent on the participants’ willingness tocomplete the measures thoroughly and honestly. Toovercome the shortcoming of underreporting sensi-tive risk behaviors associated with self-reporteddata, Add Health was administered using both com-puter-assisted self-interviewing (CASI) and audio-CASI systems for the sensitive health risk behaviorquestionnaires. CASI and audio-CASI have beenproved to be highly effective in eliciting disclosure ofspecific activities that have legal and social conse-quences [42]. For instance, O’Reilly [43] found thatusing both CASI and audio-CASI significantly in-creased reports of marijuana and cocaine use com-pared with self-administered questionnaires (SAQs).Waterton and Duffy [44] also found that reportedalcohol consumption was 30% greater with CASIthan with SAQs.

Fourth, only one dependent variable was used tomeasure alcohol use among Asian-American adoles-cents in this study. Future studies should explorequestions such as binge-drinking and the level ofalcohol consumption to provide a more comprehen-sive understanding of Asian-American adolescents’alcohol use.

Finally, there were potential biases owing tocultural and linguistic issues. For instance, theinstruments used in Add Health were not origi-nally designed for cross-cultural research. All thequestionnaires were in English. Considering thatthese students were all participating in high schoolin English, it is reasonable to assume a certaindegree of fluency in English. However, it may havebeen more difficult for those who did not useEnglish at home to understand completely some ofthe questions about attitudes, feelings, and rela-tionships. Also response bias related to culturalissues may have been present, because a set ofpsychological processes is shaped by culturallysignificant narratives and culture-specific social-ization patterns [12]. For instance, a positive self-evaluation is encouraged and the norm in NorthAmerican culture, but a self-critical orientation ismore accepted in Japanese social contexts [12].

These findings are more relevant to alcohol useamong adolescents than to alcohol abuse. However,given that Asian-American adolescents who do

drink consume more and have greater rates of binge-drinking [3,35,45], these findings have implicationsfor alcohol prevention programs designed for Asian-American adolescents. The assessment of accultura-tion and parental attachment must be integrated intoclinical evaluations and treatment when workingwith Asian-American adolescents. Because the mostacculturated adolescents with low parental attach-ment are much more likely to drink, these adoles-cents are a target for “selective preventive interven-tion” [46].

This study provided a fuller picture of the dynam-ics of acculturation by taking into account the role ofparental attachment. The findings of this study raisenew questions about the context in which accultura-tion is a risk factor. This study also provides evi-dence that the bond between parents and adolescentsis a vital factor in adolescents’ lives that serves as aprotective shield to ward off potential alcohol use foracculturated adolescents. Interventions that fosterincreased attachment between parents and adoles-cents would counteract the risk of alcohol use asso-ciated with acculturation.

Finally, policy makers and mental health profes-sionals need to place more emphasis on helpingAsian-American adolescents establish a balance be-tween learning and respecting traditional languageand values and simultaneously promoting the acqui-sition of skills to facilitate healthy integration into theAmerican culture.

This study was funded by a dissertation grant (Grant 1 RO3MH64341-01, Primary investigator: H.C.H.) from the NationalInstitute of Mental Health Office of AIDS Research and theCouncil on Social Work Education NIMH Minority ResearchFellowship Program (#3-T32-MH-16089-20SI). We thank Drs.Steven Schinke, Ellen Lukens, Raymond Arons, and MeredithHanson for reading and commenting on the manuscript. AddHealth is a program designed by J. Richard Udry (primaryinvestigator) and Peter Bearman and funded by Grant P01-HD31921 from the National Institute of Child Health and HumanDevelopment to the Carolina Population Center, University ofNorth Carolina at Chapel Hill, with cooperative funding partici-pation by the National Cancer Institute; National Institute ofAlcohol Abuse and Alcoholism; National Institute on Deafnessand Other Communication Disorders; National Institute on DrugAbuse; National Institute of General Medical Sciences; NationalInstitute of Mental Health; National Institute of Nursing Research;Office of AIDS Research, NIH; Office of Behavior and SocialScience Research, NIH; Office of the Director, NIH; Office ofResearch on Women’s Health, NIH; Office of Population Affairs,DHHS; National Center for Health Statistics, Centers for DiseaseControl and Prevention, DHHS; Office of Minority Health, Cen-ters for Disease Control and Prevention, DHHS; Office of MinorityHealth, Office of Public Health and Science, DHHS; Office of theAssistant Secretary for Planning and Evaluation, DHHS; and theNational Science Foundation.

August 2003 ASIAN-AMERICAN ADOLESCENTS’ ALCOHOL USE 127

References1. Ellickson PL, Morton S. Identifying adolescents at risk for hard

drug use: Racial/ethnic variations. J Adolesc Health 1998;25:382–95.

2. Swadi H. Individual risk factors for adolescent substance use.Drug Alcohol Depend 1999;55:209–24.

3. Makimoto K. Drinking patterns and drinking problemsamong Asian Americans and Pacific Islanders. Alcohol HealthRes World 1999;22:270–84.

4. Barnes GM, Welte JW. Patterns and predictors of alcohol useamong 7–12th grade students in New York State. J StudAlcohol 1986;47:53–62.

5. Chen X, Unger JB, Cruz TB, et al. Smoking patterns ofAsian-American youth in California and their relationshipwith acculturation. J Adolesc Health 1999;24:321–28.

6. Unger JB, Cruz TB, Rohrback LA, et al. English language useas a risk factor for smoking initiation among Hispanic andAsian American adolescents: Evidence for mediation by to-bacco-related beliefs and social norms. Health Psychol 2000;19:403–10.

7. Szapocznik J, Kurines WM. Bicultural involvement in His-panic American youths. In: Padilla AM (ed). Acculturation:Theory, Models, and Some New Findings. Boulder, CO:Westview Press, 1980:139–57.

8. Phinney JS. Ethnic identity in adolescents and adults: Reviewof research. Psychol Bull 1990;108:499–514.

9. Gordon MM. Assimilation in American Life: The Role of Race,Religion, and National Origins. New York: Oxford UniversityPress, 1964.

10. Berry J. Acculturation as varieties of adaptation. In: PadillaAM (ed). Acculturation: Theory, Models, and Some NewFindings. Boulder, CO: Westview Press, 1980:9–26.

11. Triandis HC. Individualism, collectivism and personality. JPers 2001;69:907–24.

12. Heine SJ, Lehman DR, Markus HR. Is there a universal needfor positive self-regard? Psychol Bull 1999;106:766–94.

13. Ying YW, Lee PA, Tsai JL. Relationship of young adultChinese Americans with their parents: Variation by migratorystatus and cultural orientation. Am J Orthopsychiatry 2001;1:342–9.

14. Ying YW. Chinese American adults relationship with theirparents. Int J Soc Psychiatry 1994;40:35–45.

15. Ying YW, Coombs M, Lee PA. Family intergenerational rela-tionship of Asian American adolescents. Cult Divers EthnicMinority Psychol 1999;5:350–63.

16. Beiser M, Dion R, Gotowiec A, et al. Immigrant and refugeechildren in Canada. Can J Psychiatry 1995;40:67–72.

17. Harris KM. The health status and risk behaviors of adolescentsin immigrant families. In: Hernandez D (ed). Children ofImmigrants: Health, Adjustment, and Public Assistance.Washington, DC: National Academy Press, 1999:287–347.

18. Oetting ER, Beauvais F. Orthogonal cultural identificationtheory: The cultural identification of minority adolescents. IntJ Addict 1990;25:655–85.

19. Brook JS, Brook DW, Gordon AS, et al. The psychosocialetiology of adolescent drug use: A family interaction ap-proach. Genet Soc Gen Psych Monogr 1990;116:113–267.

20. Bearman PS, Jones J, Udry JR. The National LongitudinalStudy of Adolescent Health: Research Design. Available at:http://www.cpc.unc.edu/projects/addhealth/design.html.Accessed February 1, 2002.

21. Resnick MD, Bearman PS, Blum RW, et al. Protecting adoles-cents from harm: Findings from the National LongitudinalStudy of Adolescent Health. JAMA 1997;278:823–32.

22. Ford CA, Bearman PS, Moody J. Foregone health care amongadolescents. JAMA 1999;282:2227–34.

23. Guterman NB, Hahm HC, Cameron M. Adolescent victimiza-tion and subsequent use of mental health counseling services.J Adolesc Health 2002;30:336–45.

24. Hawkins JD, Catalano RF, Miller JY. Risk and protectivefactors for alcohol and other drug problems in adolescenceand early adulthood: Implications for substance abuse preven-tion. Psychol Bull 1992;12:64–105.

25. Rogler LH, Cortes DE, Malgady RG. Acculturation andmental health status among Hispanics. Am Psychol 1991;46:585–97.

26. Rubin A, Babbie E. Research Methods for Social Work. PacificGrove, CA: Brooks/Cole Publishing, 1993.

27. Chantala K, Tabor J. The Longitudinal Study of AdolescentHealth: Strategies to perform a design-based analysis usingthe Add Health data. Available at: http://www.cpc.unc.edu/projects/addhealth/design.html. Accessed January 17, 2001.

28. Stata Press. Stata 6.0: Reference Manual Extract. CollegeStation, TX: Stata Corporation, 1999.

29. Giles H, Johnson P. Ethnolinguistic identity theory: A socialpsychological approach to language maintenance. Int J SocLanguage 1987;68:69–100.

30. Tong Y, Hong Y, Lee S, et al. Language use as a carrier of socialidentity. Int J Intercultural Relations 1999;23:281–96.

31. Schieffelin B, Ochs E. Language socialization. Ann Rev An-thropol 1986;15:163–91.

32. Brook JS, Brook DW, Mario DL. The role of parents inprotecting Colombian adolescents from delinquency and mar-ijuana use. Arch Pediatr Adolesc Med 1999;153:457–64.

33. Chao RK. The parenting of immigrant Chinese and EuropeanAmerican mothers: Relations between parenting style, social-ization goals, and parental practice. J Appl Dev Psychol2000;21:233–48.

34. The Substance Abuse and Mental Health Services Adminis-tration. The National household survey on drug abuse. Avail-able at: http://www.samhsa.gov/mhsda/2k1nhsda/pdf/cover.pdf. Accessed January 3, 2003.

35. The Substance Abuse and Mental Health Services Adminis-tration. Asian and Pacific Islander adolescents in substanceabuse treatment: 1999. Available at: http://www.samh-sa.gov/oas/2k2/AsianTX/AsianTX.htm. Accessed December29, 2002.

36. Schuster MA, Bell R, Nakajima GA, et al. The sexual practicesof Asian and Pacific Islander high school students. J AdolHealth 1998;23:221–31.

37. Wortley PM, Metler RP, Hu DJ, et al. AIDS among Asians andPacific Islanders in the United States. Am J Prev Med 2000;18:208–14.

38. Griffith J. Relationship between acculturation and psycholog-ical impairment in adult Mexican Americans. Hispanic JBehav Sci 1983;5:431–59.

39. Samaniego RY, Gonzales NA. Multiple mediators of theeffects of acculturation status on delinquency for MexicanAmerican adolescents. Am J Psychol 1999;27:189–210.

40. Epstein JA, Botvin GJ, Dusenbury L, et al. Validation of anacculturation measure for Hispanic adolescents. Psychol Rep1996;79:1075–79.

41. Tsai JL, Ying YW, Lee PA. Cultural predictors of self-esteem: A study of Chinese American female and maleyoung adults. Cult Divers Ethnic Minority Psychol 2001;7:284 –97.

42. Lessler JT, O’Reilly JM. Mode of Interview and Reporting ofSensitive Issues: Design and Implementation of Audio Com-

128 HAHM ET AL JOURNAL OF ADOLESCENT HEALTH Vol. 33, No. 2

puter-assisted Self-Interviewing. Research Triangle Park, NC:Research Triangle Institute Publications, 1995.

43. O’Reilly JM. Audio and video computer-assisted self-inter-viewing: Preliminary tests of new technologies for data col-lection. J Official Stat 1994;10:197–214.

44. Waterton JJ, Duffy JC. A comparison of computer interview-ing techniques and traditional methods in the collection of

self-report alcohol consumption data in a field survey. Int StatRev 1984;52:173–82.

45. D’Avanzo CE, Frye B, Froman B. Culture, stress, and sub-stance use in Cambodian refugee women. J Stud Alcohol1994;55:420–6.

46. Mrazak PJ, Haggerty RJ. Reducing Risks for Mental Disorders:Frontiers for Preventive Intervention Research. Washington,DC: National Academy Press, 1994.

August 2003 ASIAN-AMERICAN ADOLESCENTS’ ALCOHOL USE 129