Short Acculturation Scale for Hispanics (SASH-Y and SASH) Reliability and Validity in Low-Income,Mexican-Heritage Youth and Parents

Abstract

The Short Acculturation Scale for Hispanics (SASH) is a popular acculturation measure. Although SASH has been validated, conflicting information has been published about the youth version (SASH-Y). We examined the psychometric properties (i.e., validity and reliability) of SASH-Y in low-income, Mexican-heritage youth (aged 8-18; n = 100) and SASH with their parents (n = 100; total N = 200). Confirmatory factor analyses comparing the three published structural variations of the SASH-Y revealed the three-factor structure originally proposed for SASH-Y fit the data best, and the frequently used unidimensional structure did not fit. The adult SASH fit parent data well. For researchers interested in comparing parent and child scores (e.g., acculturation gap research), we examined the 10 identical parent and child items for invariance. Invariance was established for parent and child data with the traditional SASH structure. However, internal consistency was inadequate for the child data regardless of variable structure. Implications are discussed.

Keywords

acculturation psychometric validity acculturation gaps children parent-child

The Hispanic/Latino(a) population (henceforth, Hispanic) is the largest ethnic/racial minority in the United States (U.S. Census Bureau, 2015), with Hispanic children representing the largest ethnic/racial minority group of children (National Kids Count, 2017). Hispanic families’ adaptation to United States’ mainstream culture associates with parents’ and children’s mental health, social and emotional adjustment and development, and family functioning, especially for immigrant families (Kuo, 2014; Zambrana & Carter-Pokras, 2010). Acculturation describes the cultural adaptation process, referring to the change in people’s practices, values, and cultural identifications that stem from their contact with a dissimilar culture (Berry, 1997; Schwartz, Unger, Zamboanga, & Szapocznik, 2010). How adult and child immigrants cope with acculturation processes influences their psychological health and well-being (for reviews, see Kuo, 2014; Lara, Gamboa, Kahramanian, Morales, & Bautista, 2005; Thomson & Hoffman-Goetz, 2009), self-reported health, and psychological health outcomes (Todorova et al., 2013).

A major limitation in connecting acculturation research to the Hispanic population’s health outcomes is the lack of well-validated measures, especially for children (for reviews see Thomson & Hoffman-Goetz, 2009; Wallace, Pomery, Latimer, Martinez, & Salovey, 2010). Using acculturation measurements that have not been psychometrically examined poses a severe threat to the validity of findings and methodological rigor of acculturation and health research (Chaffee, 1991). To partially address these issues, the present study seeks to confirm the factor structure of a widely used measurement of acculturation—the Short Acculturation Scale for Hispanics (SASH; Marín, Sabogal, Marín, Otero-Sabogal, & Perez-Stable, 1987)—and more importantly to validate its modified version for Hispanic youth (SASH-Y; Barona & Miller, 1994).

The adult SASH has 12 items that factor into three sub-domains of acculturation and has demonstrated validity in multiple Hispanic subgroups (Marín et al., 1987; Yamada, Valle, Barrio, & Jeste, 2006). The first factor (i.e., language use) includes items that measure English and Spanish language use when reading, thinking, and speaking with family and friends. The second factor (i.e., media language use) taps preference of language when watching television or listening to the radio, and the third factor (i.e., ethnic social relations) assesses preference of ethnicity of one’s social contacts (Marín et al., 1987). When the SASH was modified for youth (Barona & Miller, 1994), two of the items were replaced with items that were more appropriate for adolescents (see Table 1 for items of SASH-Y and Table 2 for items of SASH).

Table 1.

Child SASH-Y CFA Standardized Item Loadings.

		Model 1	Model 2						Model 3
		Original	Original		Item 1 double loaded		Item 1 dropped		Original
SASH-Y items	M (SD)	β	β	Factor	β	Factor	β	Factor	β	Factor
1. What languages do you read and speak?	3.41(.73)	.24ns	.24	2. Extra	.25ns, .15ns	1 and 2	—	Dropped	.27	1. Pers
2. What languages do your parents speak to you in?	1.57(.67)	.19ns	.53	1. Fam	.42	1. Fam	.54	1. Fam	.18ns	1. Pers
3. What languages do you usually speak at home?	2.49(.08)	.27	.58	1. Fam	.74	1. Fam	.56	1. Fam	.33	1. Pers
4. In which languages do you usually think?	3.63(.11)	.53	.54	2. Extra	.54	2. Extra	.52	2. Extra	.60	1. Pers
5. What languages do you usually speak with your friends?	4.09(.94)	.57	.55	2. Extra	.56	2. Extra	.57	2. Extra	.57	1. Pers
6. In what languages are the T.V. programs you usually watch?	3.88(.03)	.67	.67	2. Extra	.67	2. Extra	.65	2. Extra	.66	2. External
7. In what languages are the radio programs you usually listen to?	3.28(.23)	.59	.58	2. Extra	.58	2. Extra	.60	2. Extra	.60	2. External
8. In what languages are the movies, T.V., and radio programs you prefer to watch or listen to?	3.83(.96)	.61	.64	2. Extra	.64	2. Extra	.64	2. Extra	.65	2. External
9. In what languages do your parents speak with their parents?	1.07(.26)	.09ns	.37	1. Fam	.29	1. Fam	.38	1. Fam	.09ns	2. External
10. Your close friends are:	2.28(.00)	.05ns	.58	3. Social	.58	3. Social	.57	3. Social	.58	3. Social
11. You prefer going to parties at which the people are:	2.35(.79)	.24ns	.57	3. Social	.57	3. Social	.57	3. Social	.57	3. Social
12. The persons you visit or who visit you are:	1.96(.93)	.17ns	.64	3. Social	.63	3. Social	.64	3. Social	.64	3. Social
Factor 1 correlation with Factor 2			.33*		.31*		.32*		.86***
Factor 2 correlation with Factor 3			.10		.20		.21		.17
Factor 1 correlation with Factor 3			.20		.05		.10		.20

Note. Standardized loadings reported. Model 1 = unidimensional model. Model 2 = Barona and Miller (1994) 3-factor model. Model 3 = Serrano and Anderson (2003) 3-factor model. Extra = Extrafamilial language use. Fam = Familial language use. Social = Ethnic social relations. Pers = Personal language use. External = External language use. ns = nonstatistically significant item loading.

p < .05. ***p < .001.

Table 2.

SASH CFA Item Loadings for Parent Only and Dyadic Scale.

	Parent data only			Parent and child invariant scale
	M (SD)	β	Factor	B	Factor	Child item
1. In general, what language(s) do you read and speak in?	1.60 (.64)	.62	1: Lang	.35	1: Lang	Item 1
2. What was the language(s) you spoke when you were a child?	1.00 (.00)	—	~~1: Lang~~	—	—
3. In general, what language(s) do you speak at home?	1.48 (.71)	.53	1: Lang	.22	1: Lang	Item 3
4. In general, in what language(s) do you think?	1.33 (.67)	.71	1: Lang	.46	1: Lang	Item 4
5. In general, in what language(s) do you speak with your friends?	1.31 (.65)	.74	1: Lang	.54	1: Lang	Item 5
6. In general, in what language(s) are the television programs that you watch?	1.89 (.97)	.80	2: Media	.71	2: Media	Item 6
7. In general, in what language(s) are the radio programs that you listen to?	1.65 (.92)	.59	2: Media	.58	2: Media	Item 7
8. In general, in what language(s) do you prefer to hear and watch movies, and radio and television programs?	1.81 (.93)	.74	2: Media	.70	2: Media	Item 8
9. Your closest male and female friends are:	1.51 (.68)	.74	3: Social	.53	3: Social	Item 10
10. You prefer to go to meetings/social events at which people are:	1.62 (.72)	.62	3: Social	.47	3: Social	Item 11
11. The people that you visit or that visit you are:	1.40 (.59)	.67	3: Social	.42	3: Social	Item 12
12. If you could choose your children’s friends, you would like them to be:	1.58 (.74)	.32	3: Social	—	—
Lang correlation with Media		.73***
Media correlation with Social		.60***
Lang correlation with Social		.71***

Note. Parent items were assessed in Spanish, although the English translation is given. Standardized loadings reported for the parent data, unstandardized loadings reported for the invariant scale. Invariant loadings are with scalar invariance.

***

p < .001.

Although the scale factor structure for the adult version (i.e., SASH) has been psychometrically examined, the factor structure and reliability of the youth version (i.e., SASH-Y) remains unclear despite its wide use. In the original proposal of the youth version, Barona and Miller (1994) conducted an exploratory factor analysis (EFA) on the 12-item SASH-Y scale and identified three factors. The authors called for a confirmatory factor analysis (CFA) on the proposed structure, which they later conducted in an unpublished manuscript (Miller & Barona, 2001). Serrano and Anderson (2003) also performed an EFA on the 12 SASH-Y items but found a different three-factor structure. The factor structures reported in Barona and Miller (1994) and Serrano and Anderson (2003) both included an ethnic social relations factor, composed of items 10 to 12, but the two models differed on how the nine language items were divided (see Table 1 for item wording and factors). Although the EFAs produced multidimensional results, both Barona and Miller (1994) and Serrano and Anderson (2003) suggested the scale should be used as a composite measure (i.e., a unidimensional scale). Most published studies using SASH-Y have summed all 12 items (e.g., Olvera, Dempsey, Gonzalez, & Abrahamson, 2013; Schofield, Parke, Kim, & Coltrane, 2008; Weisskirch, 2005). Psychometric analysis of the SASH-Y is warranted given the popularity of the scale and the unresolved issues concerning its factor structure.

The current investigation aimed to psychometrically examine the SASH and SASH-Y with a sample of low-income, Mexican-heritage parent-child dyads (N = 100 dyads, 200 participants). The SASH was administered to the parents and the SASH-Y was given to the children. For the children, we compared three SASH-Y measurement models using CFA: (a) the unidimensional model (henceforth “Model 1”), (b) the three-factor model identified in Barona and Miller (1994; henceforth “Model 2”), and (c) the three-factor model identified in Serrano and Anderson (2003; henceforth “Model 3”). For the parents, we examined the previously published three-factor model of SASH. Finally, CFA allows researchers to assess whether measures function similarly in both members of a dyad (e.g., parents and children) or are invariant between dyad members (Claxton, Deluca, & Van Dulmen, 2015). Research comparing parent and child acculturation scores (see Birman, 2006) assume that parent and child measures are invariant and comparable. Thus, we also examined the measure for invariance between parents and children. Our findings seek to provide evidence for appropriate acculturation measures for low-income Hispanic families, how researchers should form scales from the acculturation items, and therefore conduct research on acculturation and health with more rigorous measures.

Method

Participants and Procedure

We recruited 100 low-income parent-children dyads in the United States (N = 200). Children ranged in age from 9 to 18 years (M = 13.22, SD = 2.07), with corresponding grade levels (M = 7.72, SD = 2.31, range =3 to 12). Children of both genders filled out the survey (females = 55, males = 45). Children were born in the United States (n = 70), Mexico (n = 23), and El Salvador, Honduras, and Venezuela (n = 3 each). Four did not report birthplace. Most children were of Mexican descent (n = 86). Parents were predominantly mothers (n = 87) and were born outside of the United States (n = 90, seven declined to answer). Parents were between 25 and 56 years old (M = 37.96, SD = 5.66). On average, parents lived in the United States for 15.41 years (SD = 6.66, range = 3 to 40) and had an eighth grade education (range = no education to 2 years in college).

Two bilingual employees of a university extension program recruited participants from the Gary, Indiana, and East Chicago area. Families with primarily Spanish-speaking parents and children bilingual in Spanish and English were recruited. The two employees were familiar with the targeted population, as their job was to recruit low-income families for family programs run through extension (e.g., nutrition programs for low-income families). Study recruitment procedures followed extension and state guidelines for defining low-income (i.e., at or below 100% of the poverty line); thus, families who were participating in or eligible for extension programs were also eligible for the current study. For families not associated with extension, the employees assessed current study eligibility in the same way they assessed extension program eligibility (e.g., asking questions that indicated a certain level of poverty, such as “Are you currently using SNAP?” [Supplemental Nutrition Assistance Program, formerly known as food stamps, which requires a 165% federal poverty level for eligibility]). Each dyad was compensated US$50. A bilingual assistant was present to answer any questions. Given the low education level of some parents, the bilingual assistant gave parents the choice of having the survey read to them. Although these data were not officially recorded, between 20 and 30 parents opted to take the survey in this manner. No surveys were read to the child participants. A university institutional review board approved the study protocol.

Measures

Children answered the 12-item Short Acculturation Scale for Hispanic Youth in English (SASH-Y; Barona & Miller, 1994; see Table 1 for item wording and descriptive statistics). Items 1 through 9 had a 5-point response: 1 (only Spanish), 2 (Spanish better than English), 3 (both equally), 4 (English better than Spanish), 5 (only English). Items 10 through 12 were assessed on a similar 5-point scale from 1 (all Hispanic) to 5 (all White).

Parents answered the 12-item Short Acculturation Scale for Hispanics in Spanish (SASH; Marín et al., 1987; see Table 2). Parents responded to the four questions on language use and four questions on media preferences on a 5-point scale from 1 (only Spanish) to 5 (only English). The four questions about ethnic social relations were assessed on a different 5-point scale from 1 (only Latinos) to 5 (only Americans).

Children and parents reported on their English and Spanish ability (speaking and listening) by responding to Tse’s (1996) language proficiency scale (1 = not at all or a little, 4 = very well). They also reported demographic questions (e.g., sex, age, and time living in the United States).

Results

Confirmatory factor analyses (CFAs) were performed with full information likelihood (less than 1% of data were missing) and robust maximum likelihood (MLR), given data nonnormality. We conducted CFAs in Mplus 7.3 (Muthén & Muthén, 2012). We evaluated model fit by considering whether χ² p > .05, comparative fit index (CFI) > .90, Tucker–Lewis index (TLI) > .90, root mean square error of approximation (RMSEA) ≤ .08, (Brown, 2015; Hu & Bentler, 1999). Internal consistency was assessed with Cronbach’s alpha and a CFA-based reliability coefficient (Brown, 2015). CFA-based reliability “represents the proportion of true-score variance to total observed variance in the measure” and addresses known issues with Cronbach’s alpha (see Brown, 2015, p. 309).

Child CFA Models: SASH-Y

Table 1 includes item loadings for all three CFA models. Model 1, the unidimensional model, did not fit the data well, χ² = 95.98, p < .001, χ²/df = 1.78, CFI = .67, TLI = .59, RMSEA = .088, 90% confidence interval (CI) = .059 to .117. Items’ internal consistency was low (Cronbach’s alpha = .68; CFA-based reliability = .68).

Model 2 produced adequate fit, χ² = 60.14, p = .18, χ²/df = 1.18, CFI = .93, TLI = .91, RMSEA = .042 (90% CI = .00 to .080). However, Item 1 loaded poorly at .24 (see Table 1). Barona and Miller originally proposed Item 1 loaded on both extrafamilial and familial factors; we reran the model with this variation, but the item did not statistically load on either factor (external = .15, family = .25), χ² = 58.26, p = .20, χ²/df =1.17, CFI = .93, TLI = .91, RMSEA = .041 (90% CI = .00 to .079). Thus, Item 1 was dropped (Brown, 2015) and the CFA rerun, χ² = 41.63, p = .44, χ²/df = 1.02, CFI = .99, TLI = .99, RMSEA = .012 (90% CI = .00 to .070). The revised Model 2 fit the data well. However, internal consistency statistics were low (Cronbach’s alpha for extrafamilial = .72, familial = .41, and social = .61; CFA-based reliability extrafamilial = .73, familial = .50, and social = .62).

Model 3 did not fit the data well, χ² = 71.99, p = .03, χ²/df = 1.41, CFI = .83, TLI = .78, RMSEA = .064 (90% CI = .022 to .097). Internal consistency statistics were low (Cronbach’s alpha for personal language = .52, external language = .62, and social = .61; CFA-based reliability personal language = .51, external language = .67, and social = .62).

Parent Model: SASH

All 100 parents scored Item 2 as a one (i.e., all spoke only Spanish as a child, see Table 2): Thus, the item had no variation and was excluded from all CFAs with parent data. The three-factor CFA adequately fit the parent data (see Table 2), χ² = 54.98, p = .07, χ²/df = 1.34, CFI = .93, TLI = .91, RMSEA = .059 (90% CI = .00 to .097). Items’ internal consistency statistics were adequate, although low for the social factor (Cronbach’s alpha for personal language use = .74, media = .73, and social = .66; CFA-based reliability personal language use = .74, media = .76, and social = .67).

Measure Invariance Between Parents and Children

To examine measurement invariance between parents and children, we assessed the 10 items with identical wording (i.e., removed Item 12 from the parent scale and Item 9 from the child scale; we also removed Item 2 because all parents gave the same answer; see Tables 1 and 2). Following the recommendations of Claxton and colleagues (2015), we examined the model for configural χ²(64)= 76.14, p = .14, CFI = .96, TLI = .94, RMSEA = .044 (90% CI = .000 to .078); metric χ²(71)= 83.23, p = .15, CFI = .96, TLI = .95, RMSEA = .042 (90% CI = .000 to .075); and scalar invariance χ²(78)= 88.71, p = .19, CFI = .96, TLI = .96, RMSEA = .037 (90% CI = .000 to .070). With the first item loading of each factor constrained to one, model fit did not worsen when constrained at different levels: configural against metric: χ²(7)= 7.07, p = .42, metric against scalar: χ²(7)= 6.11, p = .53. Scalar invariance remained when the second item loading of each factor was constrained to one instead of the first item loading. Because scalar invariance was achieved, we compared latent means between parents and children. Not surprisingly, latent means significantly varied, with lower parent unstandardized latent means on language (–1.79 lower than children, p < .001), media (–2.01 lower than children, p < .001), and social factors (–0.78 lower than children, p < .001). Residuals and latent variable covariances were not invariant between parents and children (i.e., structural invariance was not present as the two populations demonstrated heterogeneity; Muthén & Muthén, 2012). Thus, parent and child scores can be compared on the modified 10-item SASH because the measures meet invariance requirements. Internal consistency for the invariant model data ranged (Language use, media, and social variable Cronbach’s alpha were .53, .67, and .61 for children, and .74, .76, and .70 for parents; and CFA-based reliability were .51, .67, and .62 for children and .74, .76, and .72 for parents).

Correlations With Other Indicators of Acculturation

To examine convergent validity of the SASH-Y and SASH scales, we assessed the correlations between these scales and other indicators of acculturation. As shown in Table 3, the composite score of SASH-Y was negatively associated with adolescents’ Spanish proficiency and positively correlated with their time spent in the United States. More nuanced patterns of associations were revealed when decomposing the SASH-Y scale into different factors (see Table 3). For example, whereas familial language use was positively associated with English ability, extrafamilial language use and ethnic social relational tended to be inversely related to Spanish ability. Similarly, we assessed the correlations between parents’ reports on SASH and other indicators of acculturation (see Table 4). As would be expected, the three factors of acculturation were positively associated with English ability, with language use demonstrating the strongest association. Two of the factors in SASH were also positively related to parents’ time living in the United States.

Table 3.

Descriptive Statistics.

	Range	M (SD)
Individual data
Parent SASH
Language use^a	0.93 to 3.00	1.43 (.50)
Media use^b	0.92 to 4.00	1.77 (.74)
Social ethnic relations	1.00 to 3.50	1.79 (.48)
Child SASH-Y (Model 2)
Familial communication	1.00 to 3.00	2.14 (.45)
Extrafamilial communication	1.67 to 4.83	3.69 (.64)
Social ethnic relations^c	1.00 to 5.00	2.20 (.73)
Dyadic data (invariant scale)
Parent 10-item scale
Language use^a	0.93 to 3.00	1.43 (.50)
Media use^b	0.92 to 4.00	1.77 (.74)
Social ethnic relations	0.94 to 3.33	1.51 (.53)
Child 10-item scale
Language use	1.50 to 4.50	3.41 (.62)
Media use	1.67 to 5.00	3.66 (.84)
Social ethnic relations^c	1.00 to 5.00	2.20 (.73)
Demographics
Child demographics
Age	9 to 18	13.22 (2.07)
Time living in the United States	3 to 18	11.61 (3.15)
English ability	1.00 to 4.00	3.67 (0.58)
Spanish ability	1.00 to 4.00	3.47 (0.63)
Parent demographics
Age	25 to 56	37.96 (5.66)
Time living in the United States	3 to 40	15.41 (6.66)
English ability	1.00 to 3.50	1.82 (0.71)
Spanish ability	1.50 to 4.00	3.86 (0.38)

Note. Scales with the same superscript letter are identical. Age and Time living in the United States are in years. Spanish and English ability are speaking and understanding only, measured on a 4-point scale.

Table 4.

Correlations Between SASH and SASH-Y and Demographic Characteristics.

	Dyadic data (invariant scale)						Individual data
	Parent 10-item scale			Child 10-item scale			Parent SASH			Child SASH-Y Model 2
	Lang	Media	Social	Lang	Media	Social	Lang	Media	Social	Familial	Extrafam	Social
Demographics
Child demographics
Sex (0 = Male, 1 = Female)	−.11	−.18^{^†}	−.14	.16	−.10	−.24*	−.19^†	−.11	−.16	.12	.01	−.24*
Age	−.12	−.12	−.09	−.03	−.09	−.17^{^†}	−.12	−.12	−.13	−.01	−.11	−.17^†
Time living in the United States	.09	.03	−.11	.34**	.14	−.09	.03	.09	−.12	.34***	.24*	−.09
English ability	.15	.20*	.09	.21*	.11	−.06	.20*	.15	.12	.17^†	.17	−.06
Spanish ability	.03	.02	.12	−.25*	−.24*	−.32**	.02	.03	.17^†	−.23*	−.27**	−.32**
Parent demographics
Sex (0 = Male, 1 = Female)	−.25*	−.42***	−.40***	.01	−.10	.04	−.41***	−.25*	−.33***	.00	−.05	.04
Age	−.05	−.04	.20*	.06	.08	−.06	−.04	−.05	.13	.03	.09	−.06
Time living in the United States	.22*	.26*	.18^†	.29**	.22*	−.07	.26*	.22*	.09	.29**	.26*	−.07
English ability	.24*	.61***	.29**	.07	−.03	.02	.61***	.24*	.26**	.26**	−.03	.02
Spanish ability	−.11	.02	.08	−.16	.02	.04	.02	−.11	.10	−.05	−.07	.04
Children invariant scale
Child language use	.01	.03	−.13	—
Child media use	.05	.01	−.02	.45***	—
Child social relations	−.02	.08	.01	.06	.15

Note. Lang = language use; Media = media language use; Social = ethnic social relations.

†

p < .10. * p < .05. **p < .01. ***p < .001.

Discussion

The current study examined SASH and SASH-Y factor structures in 100 low-income, parent-child Mexican-heritage dyads (N = 200). The adult SASH measure, with the three factors of language use, media language use, and ethnic social relations, fit the data well although the internal consistency of the social factor was low. Three competing factor structures for the youth measure (i.e., SASH-Y) have been proposed in the literature: a unidimensional scale (Model 1) and two three-factor multi-dimensional scales (Models 2 and 3). When examining the child data only, the originally proposed 3-factor structure best fit the data (Model 2). However, researchers who want to use dyadic data (e.g., for acculturation gap research) and compare parent and child data need a measure that is invariant between parents and children. Examination of the dyadic data showed that a modified 10-item measure had strong factorial invariance (i.e., scalar invariance) between parents and children. However, the child data did not have strong internal consistency statistics for any factor structure, although the invariant scale variables had the highest internal consistency for both children and parents.

SASH-Y Factor Structure

When examining the child data only, and comparing the three models, the current findings did not support a unidimensional model. Although researchers initially proposed the SASH-Y should be used as a composite scale until the factor structure was confirmed, the current CFAs, the lack of statistically significant correlations between the factors, and the previously published EFAs indicate that the scale is multi-dimensional in nature.

Model 2 (Barona & Miller, 1994) best fit the child data in the current study. However, issues emerged in terms of item loadings and internal consistency. In the original EFA that proposed the factors in Model 2, Barona and Miller noted that Item 1 double loaded on the two language factors of familial and extrafamilial language use; additionally, they confirmed this double loading in an unpublished CFA (Miller & Barona, 2001). Item 1 is a double-barreled item (DeVellis, 2003), as the question has two parts with potentially different answers: What languages do you read (Part 1) and speak (Part 2)? Immigrant children in the United States often continue speaking their native language at home but are enrolled in the English education system and discontinue their native language formal education. Children may differ on what language they read well (based on education; i.e., extrafamilial language use), and what languages they speak well (based on familial language use), explaining why the item previously loaded on both factors. However, the item did not statistically load on either factor in the current analyses, and was ultimately dropped. Reporting spoken language use and reading and writing is a common way to assess acculturation (Wallace et al., 2010), so the removal of this item may be problematic.

Comparing Parent and Child Acculturation Scores

Research that compares scores between two individuals or two populations on a measure first needs to ensure that the measure structure is invariant, or comparable, between the populations (Claxton et al., 2015). A modified version of the SASH (10 items; see Table 2) demonstrated strong factor invariance between parents and children, indicating that parent and child means can be compared. As would be expected given the recruited sample, parents scored significantly lower than children on all three acculturation variables (language use, media use, and social ethnic relations). Although the factor structure that best mirrored the parent SASH was not the best-fitting factor structure when examining the child data only, the modified SASH did fit the parent and child data adequately and thus could be used when examining parents and children simultaneously. Although internal consistency was still an issue for the child variable data (e.g., Cronbach’s alpha ranged from .53 to .67), the internal consistency scores were highest for the parent and child data with the 10-item invariant variables compared to item groupings suggested by other factor structures.

Implications of Low Internal Consistency

In the current manuscript, we presented two types of internal consistency: Cronbach’s alpha and a CFA-based reliability coefficient (Brown, 2015). Internal consistency refers to the homogeneity of item responses, or “the degree to which the items on a test jointly measure the same construct” (Henson, 2001, p. 177). Rules of thumb state that internal consistency statistics equal or higher than .70 are acceptable (e.g., Kline, 2011; Streiner, 2003), indicating the range of internal consistency scores for the child scales were problematic. The SASH and SASH-Y items (particularly for language) are heterogenous, in that they assess different aspects of the concept of acculturation and therefore likely will not be highly consistent for a sample of bilingual children who use both languages. For example, children reported predominantly speaking Spanish at home (Item 3) because although children speak English with their siblings, parents find it disrespectful (De Ment, Buriel, & Villanueva, 2005). However, children reported thinking in both Spanish and English (skewed more toward English; Item 4) and predominantly spoke to their friends in English (Item 5). Likely, children make and socialize with friends at school, where English is the dominant and preferred language. Thus, when examining bilingual children who move between two linguistic contexts (e.g., English at school and Spanish at home), the scores for these items are expected to be heterogeneous. Populations such as first-generation native Spanish-speakers in the United States (Yamada et al., 2006) and individuals who have not yet acquired complete fluency in one language (such as the parents in this sample) may demonstrate greater internal consistency in their scores because their lives are more predominantly centered in one linguistic context. Internal consistency scores are specific to a sample, not a measurement, and, therefore, researchers should conduct internal consistency tests for their data. For example, Serrano and Anderson (2003) reported acceptable internal consistency Cronbach’s alpha from their sample for the Model 2 factor in a predominantly Euro-American sample of fourth and fifth graders (extrafamilial = .86, familial = .86, and social = .74). The low internal reliability in the current study children scores may be replicated in samples with similar characteristics to ours, but the scale may demonstrate higher reliability in other samples.

Given the lack of high item homogeneity in the current SASH-Y variables (regardless of which factor structure is used), clinicians should recognize that the composite scores are not an accurate reflection of all of the test’s items (Henson, 2001, p. 178) and therefore will likely lead to inappropriate decisions if used as such (e.g., using a certain scale “cut-off” point to determine child eligibility for a program). For researchers, the invariance parent and child SASH measures provide one advantage in research, as parent and child scores can be compared using invariant measures. However, because effect size magnitude and statistical significance are intrinsically tied to reliability of scores (Henson, 2001), when internal reliability test scores are low, power for statistical tests is attenuated and likelihood of Type II error increases (i.e., failing to reject a null hypothesis; Kline, 2011).

However, scholars also caution against overprioritizing item homogeneity:

Even a set of indicators of “poor” psychometric quality (e.g., low reliabilities, little common variance) can produce accurate estimates of the relationships among constructs provided that they (a) are spread out across the construct domain sufficiently to capture the centroid, (b) yield sufficient variability on the construct, and, again, (c) are analyzed by confirmatory analysis. (Little, Lindenberger, & Nesselroade, 1999, p. 207)

These authors note that the “true construct centroid” or “construct domain centroid” is the underlying, conceptual abstraction of the construct attempting to be assessed. When assessing acculturation, particularly in samples that have heterogeneity on the items, only using items that are highly correlated (and have high internal consistency) may lead to a measure well-defined by psychometric indicators, but that strays from the underlying concept (Little et al., 1999, p. 207). Thus, prioritizing high internal consistency may lead to a measurement assessing too narrow a construct that misses the true construct centroid. In sum, researchers using the current measures should (a) determine the internal reliability statistics for their sample, (b) verify that parent and child scores are invariant for their sample if using family data (and conduct CFAs), and (c) understand the implications of internal consistency scores and confirmatory factor analyses for research findings.

Implications and Future Directions

The current participants may have characteristics that are not generalizable to all populations and thus the SASH-Y findings and internal consistency issues may be unique. Our sample were all low-income families, at or below 100% of the poverty line. Per our recruitment procedures, all parents predominantly spoke Spanish and children were bilingual in Spanish and English. Thus, our recruitment procedures focused on families who likely have acculturation gaps between children and parents. Because the scale relies on some generational markers of acculturation (i.e., what language do your parents speak to you in, what language did your grandparents speak to your parents in), the scale may not function well in families in which generational acculturation is not consistent. Children whose parents speak to them in Spanish and who grew up speaking Spanish may still be highly acculturated, thus causing a lack of consistency between these items. Furthermore, the sample size of the current investigation (i.e., 100 parent-child dyads) was relatively small. Replication with larger samples will help validate our findings.

Researchers examining acculturation-related issues should utilize acculturation measures that are structurally valid and reliable in their target child demographic sample. Findings of the current study indicate that the 12 SASH-Y items should not be used as a unidimensional scale. These findings underscore the need for additional psychometric validation of child acculturation measures, particularly across levels of socioeconomic status and ethnic/cultural heritage (Guo, Suarez-Morales, Schwartz, & Szapocznik, 2009). Moreover, parent and child scales need to be compared for measurement invariance when examining acculturation gaps (Claxton et al., 2015).

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Author Biographies

Lisa M. Guntzviller, PhD (Purdue University, 2013), is an assistant professor in the Department of Communication at the University of Illinois at Urbana-Champaign. Her research examines interpersonal communication in health and family contexts, specifically in parent-child relationships, Spanish-speaking families, supportive interactions, and discussions of health issues. She studies individual and dyadic goals, messages, and conversation outcomes of interactions in which identities are challenged and contested. She has focused on language brokering interactions, in which a bilingual child culturally and linguistically mediates between a primarily Spanish-speaking parent and an English speaker, and advice interactions, such as parent-to-child advice on exercise.

Ningxin Wang, PhD (University of Illinois at Urbana-Champaign, 2017), is an assistant professor in the Department of Communication Studies at Bloomsburg University of Pennsylvania. Her research primarily focuses on interpersonal and family communication. Specifically, her research program addresses how parent-child communication, including supportive communication, technologically mediated communication, and communication about health-related behaviors (e.g., sexuality, substance uses), affects children’s well-being during adolescence and emerging adulthood.

References

Barona

Miller

J. A.

(1994). Short Acculturation Scale for Hispanic Youth (SASH-Y): A preliminary report. Hispanic Journal of Behavioral Sciences, 16, 155-162. doi:10.1177/07399863940162005

Berry

J. W.

(1997). Immigration, acculturation, and adaptation. Applied Psychology, 46, 5-34. doi:10.1111/j.1464-0597.1997.tb01087.x

Birman

(2006). Measurement of the “acculturation gap” in immigrant families and implications for parent-child relationships. In Acculturation and parent-child relationships: Measurement and development (pp. 113-134). Mahwah, NJ: Lawrence Erlbaum.

Brown

T. A.

(2015). Confirmatory factor analysis for applied research (2nd ed.). New York, NY: Guilford.

Chaffee

S. H.

(1991). Communication concepts 1: Explication. Newbury Park, CA: SAGE.

Claxton

S. E.

Deluca

H. K.

Van Dulmen

M. H. M.

(2015). Testing psychometric properties in dyadic data using confirmatory factor analysis: Current practices and recommendations. Testing, Psychometrics, Methodology in Applied Psychology, 22, 1-17.

De Ment

T. L.

Buriel

Villanueva

C. M

. (2005). Children as language brokers: A narrative of the recollections of college students. In Hoosain

Salili

(Eds.), Language in multicultural education (pp. 255-272). Greenwich, CT: Information Age.

DeVellis

R. F.

(2003). Scale development: Theory and applications (2nd ed.). Thousand Oaks, CA: SAGE.

Guo

X. L.

Suarez-Morales

Schwartz

S. J.

Szapocznik

(2009). Some evidence for multidimensional biculturalism: Confirmatory factor analysis and measurement invariance analysis on the Bicultural Involvement questionnaire–Short Version. Psychological Assessment, 21, 22-31. doi:10.1037/a0014495

10.

Henson

R. K.

(2001). Understanding internal consistency reliability estimates: A conceptual primer on coefficient alpha. Measurement and Evaluation in Counseling and Development, 34, 177-189.

11.

L. T.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55. doi:10.1080/10705519909540118

12.

Kline

R. B.

(2011). Principles and practice of structural equation modeling (3rd ed.). New York, NY: Guilford Press.

13.

Kuo

B. C. H.

(2014). Coping, acculturation, and psychological adaptation among migrants: A theoretical and empirical review and synthesis of the literature. Health Psychology and Behavioral Medicine, 2, 16-33. doi:10.1080/21642850.2013.843459

14.

Lara

Gamboa

Kahramanian

M. I.

Morales

L. S.

Bautista

D. E.

(2005). Acculturation and Latina health in the United States: A review of the literature and its sociopolitical context. Annual Review of Public Health, 26, 367-397. doi:10.1146/annurev.publhealth.26.021304.144615

15.

Little

T. D.

Lindenberger

Nesselroade

J. R.

(1999). On selecting indicators for multivariate measurement and modeling with latent variables: When “good” indicators are bad and “bad” indicators are good. Psychological Methods, 4, 192-211. doi:10.1037/1082-989X.4.2.192

16.

Marín

Sabogal

Marín

B. V.

Otero-Sabogal

Perez-Stable

E. J.

(1987). Development of a short acculturation scale for Hispanics. Hispanic Journal of Behavioral Sciences, 9, 183-205. doi:10.1177/07399863870092005

17.

Miller

J. A.

Barona

(2001). Structural invariance of the short acculturation scale for Hispanic youth. Unpublished manuscript. Retrieved from http://sites.google.com/site/drjeffmiller/files/SASHY2001Man.pdf

18.

Muthén

L. K.

Muthén

B. O.

(2012). Mplus user’s guide (7th ed.). Los Angeles, CA: Author.

19.

National Kids Count. (2017). Child population by race. Population Division, U.S. Census Bureau. Available from http://datacenter.kidscount.org/

20.

Olvera

Dempsey

Gonzalez

Abrahamson

(2013). Weight-related teasing, emotional eating, and weight control behaviors in Hispanic and African American girls. Eating Behaviors, 14, 513-517. doi:10.1016/j.eatbeh.2013.06.012

21.

Schofield

T. J.

Parke

R. D.

Kim

Coltrane

(2008). Bridging the acculturation gap: Parent-child relationship quality as a moderator in Mexican American families. Developmental Psychology, 44, 1190-1194. doi:10.1037/a0012529

22.

Schwartz

S. J.

Unger

J. B.

Zamboanga

B. L.

Szapocznik

(2010). Rethinking the concept of acculturation: Implications for theory and research. American Psychologist, 65, 237-251. doi:10.1037/a0019330

23.

Serrano

Anderson

(2003). Assessment of a refined short acculturation scale for Latino preteens in rural Colorado. Hispanic Journal of Behavioral Sciences, 25, 240-253. doi:10.1177/0739986303025002006

24.

Streiner

D. L.

(2003). Starting at the beginning: An introduction to coefficient alpha and internal consistency. Journal of Personality Assessment, 80, 99-103. doi:10.1207/S15327752JPA8001_18

25.

Thomson

M. D.

Hoffman-Goetz

(2009). Defining and measuring acculturation: A systematic review of public health studies with Hispanic populations in the United States. Social Science & Medicine, 69, 983-991. doi:10.1016/j.socscimed.2009.05.011

26.

Todorova

I. L. G.

Tucker

K. L.

Jimenez

M. P.

Lincoln

A. K.

Arevalo

Falcón

L. M.

(2013). Determinants of self-rated health and the role of acculturation: Implications for health inequalities. Ethnicity & Health, 18, 563-585. doi: 10.1080/13557858.2013.771147

27.

Tse

(1996). Language brokering in linguistic minority communities: The case of Chinese- and Vietnamese-American students. The Bilingual Research Journal, 20, 485-498. doi:10.1080/15235882.1996.10668640

28.

U.S. Census Bureau (2015). Facts for features: Hispanic Heritage Month 2015. Retrieved from https://www.census.gov/newsroom/facts-for-features/2015/cb15-ff18.html

29.

Wallace

P. M.

Pomery

E. A.

Latimer

A. E.

Martinez

J. L.

Salovey

(2010). A review of acculturation measures and their utility in studies promoting Latino health. Hispanic Journal of Behavioral Sciences, 32, 37-54. doi:10.1177/0739986309352341

30.

Weisskirch

R. S.

(2005). The relationship of language brokering to ethnic identity for Latino early adolescents. Hispanic Journal of Behavioral Sciences, 27, 286-299. doi:10.1177/0739986305277931

31.

Yamada

A. M.

Valle

Barrio

Jeste

(2006). Selecting an acculturation measure for use with Latino older adults. Research on Aging, 28, 519-561. doi:10.1177/0164027506289721

32.

Zambrana

R. E.

Carter-Pokras

(2010). Role of acculturation research in advancing science and practice in reducing health care disparities among Latinos. American Journal of Public Health, 100, 18-23. doi:10.2105/AJPH.2008.138826