The Development of Science, Technology, Engineering, and Mathematics (STEM) Efficacy and Identity Among Mexican-Origin Youth Across Latino/a Destinations

Abstract

Persistence in high school curricula leading to science, technology, engineering, and mathematics (STEM) careers is structured by complex institutional systems, but developmental processes underlie how young people navigate these systems. This study examined differences in the development of STEM identity and efficacy during high school among Mexican-origin youth—a large and fast-growing demographic group that shows developmental assets and risks. Contextualizing development within larger community structures, this examination focused on the diverse array of destinations throughout the United States where Mexican-origin youth are living as contexts for their STEM identity and efficacy development. Drawing on a dataset integrating the High School Longitudinal study of 2009; Civil Rights Data Collection, decennial U.S. censuses, and the American Community Survey, multilevel models revealed variability in Mexican-origin math/science identity and efficacy development across destinations. Mexican-origin youth in established destinations had higher net growth in math identity but lower net growth in science efficacy than Whites in established destinations. Mexican-origin youth in new destinations followed similar trajectories as their Mexican-origin peers in established destinations but had lower net growth in science identity. Additionally, these patterns varied by immigrant generation. Mexican-origin youth who were the U.S.-born children of immigrants in established destinations had higher net growth in math identity than Whites in established destinations, but this generational group in new destinations had lower net growth in math identity, science identity, and science efficacy than these peers. These findings highlight the importance of communities and their embedded ecological contexts in shaping STEM identity and efficacy among Mexican-origin youth.

The growth of science, technology, engineering, and mathematics (STEM) educational fields and occupational sectors holds promise for U.S. global competitiveness and for individual and family socioeconomic mobility. The development of a strong STEM identity (seeing oneself as a math or science “person”) and sense of self-efficacy (believing in one’s ability to accomplish tasks in math or science) is intertwined with achievement in STEM subject areas in K-12 education, and forecasts STEM achievement and attainment at later points in the life course. (Chemers et al., 2011; Fast et al., 2010; Hazari et al., 2010; Kitsantas et al., 2011; Syed et al., 2011). Developing positive math and science identities and efficacy is particularly important for groups that are underrepresented in STEM majors and occupations (Chemers et al., 2011; Hazari et al., 2010; Lewis et al., 2012; Merolla & Serpe, 2013; Riconscente, 2014; Syed et al., 2011).

This study examined the development of STEM identity and efficacy among Mexican-origin Latino/a youth across geographic areas of residence. Latino/a youth have both assets and risks that contribute to their STEM identity and efficacy development. Both Latino/a parents and their children are invested in education and hold positive attitudes toward school (Kao & Tienda, 2005; Kuperminc et al., 2009). Despite these assets, Latinos/as are less likely than non-Latino/a Whites to receive STEM degrees and enter the STEM labor force (Crisp & Nora, 2012; Landivar, 2013). Prior research suggests that psychological dimensions of STEM development—such as lower math and science self-concept and math self-efficacy—could explain these differences in STEM outcomes (Kitsantas et al., 2011; Riegle-Crumb et al., 2011).

This current study provides more insight into STEM identity and efficacy development among Mexican-origin youth by focusing on geographic location of residence. Mexican-origin Latino/a youth are now developing their math and science identities and efficacy in a wider array of geographic areas of residence than ever before. In the last 3 decades, youth of Mexican origin and their families have dispersed beyond communities in U.S.-Mexico border states and Illinois, and they are increasingly living in “new” destination states such as North Carolina and Nebraska (Massey & Capoferro, 2008; Zúñiga & Hernández-León, 2005).

Geographic variability in the communities where Mexican-origin youth live could shape their math and science identity and efficacy development for several reasons. First, the web of ecological contexts that influence the STEM outcomes of underrepresented groups vary across geographic areas, including opportunities to participate in STEM curricula, access to skilled and caring STEM teachers, and exposure to STEM role models and mentors (Britner & Pajares, 2006; Fast et al., 2010; Lewis et al., 2012; O’brien et al., 1999; Riconscente, 2014). Second, such ecological differences are associated with variation in assets for youth development in general and for Latino/a youth development in particular (Benson, 2003; Kuperminc et al., 2009; Rodriguez & Morrobel, 2004). Finally, Latino/a destinations have varying Latino/a group sizes, institutional resources, and intergroup relations (Waters & Jiménez, 2005), and these aspects of destinations are connected to community resources for math/science identity and efficacy development, external assets for youth development, and positive ethnic identity (Benson, 2003; Bronfenbrenner & Morris, 2006; Kuperminc et al., 2009; Rodriguez & Morrobel, 2004).

By focusing on Mexican-origin youth, destinations, and math/science identity and efficacy development, this study highlights the importance of ecological contexts for Latino/a youth development (Bronfenbrenner & Morris, 2006), and it has policy implications for increasing Latino/a representation in STEM fields and occupations.

Identity and Efficacy in Math and Science

Psychological development in the STEM domain is a key part of the STEM attainment process (Aschbacher et al., 2010; Chemers et al., 2011; Hazari et al., 2010; Osborne & Jones, 2011; Stevens et al., 2004; Zeldin & Pajares, 2000). Developing a strong math/science identity entails recognizing oneself and being recognized by others as a math “person” and/or a scientist (Aschbacher et al., 2010; Boaler et al., 2000; Chemers et al., 2011; Hazari et al., 2010; Merolla & Serpe, 2013). Math/science self-efficacy refers to judgments about one’s abilities to accomplish subject-specific tasks and to solve math/science problems (Chemers et al., 2011; Fast et al., 2010; Zeldin & Pajares, 2000).

Math/science identities and efficacy are bidirectionally related with a range of relevant academic indicators. They predict general academic performance (Britner & Pajares, 2006; Chemers et al., 2011; Fast et al., 2010; Lewis et al., 2012; Stevens et al., 2004), STEM major choice, aspirations, and career interest (Syed, 2010; Syed et al., 2011), and commitment to and persistence in STEM education and graduate school matriculation (Hazari et al., 2010; Merolla & Serpe, 2013). STEM achievement and attainment are also predictors of math/science identity and efficacy, such as when early math achievement facilitates the development of a strong math identity over time (Stevens et al., 2004). Math/science identities and efficacy can also mediate links between STEM enrichment programs and positive STEM outcomes (Chemers et al., 2011; Merolla & Serpe, 2013), and a strong science identity can positively moderate links between college GPA and STEM graduate school enrollment (Merolla & Serpe, 2013).

Few studies focus on math and science identity and efficacy among Mexican-origin youth or the broader Latino/a population (but see Lewis et al., 2012; Riconscente, 2014). Some that do have found that Latino/a youth have lower math self-efficacy and lower perceptions of doing well in math and science than non-Latino/a Whites, which helps to explain corresponding disparities in math achievement and math/science aspirations (Kitsantas et al., 2011; Riegle-Crumb et al., 2011; Stevens et al., 2004, 2006). Others have identified more positive assets for math/science identity and efficacy development for Latinos/as relative to Whites, such as a higher interest in and enjoyment of math (Riegle-Crumb et al., 2011; Stevens et al., 2006).

Contexts and the Development of Math/Science Identity and Efficacy

Math/science identity and efficacy development for youth in general, and Latinos/as in particular, is influenced by key features of community ecologies (Britner & Pajares, 2006; Bronfenbrenner & Morris, 2006; Lewis et al., 2012; Merolla et al., 2012; Riconscente, 2014). Strong math/science identities develop in contexts where there are relationships based around STEM activities, opportunities to participate in STEM curricula, instrumental guidance in STEM pursuits, and access to STEM enrichment programs (Chemers et al., 2011; Merolla & Serpe, 2013). For youth from underrepresented groups, exposure to mentors and role models of similar racial/ethnic backgrounds who are in STEM-related professions and teaching positions can also be instrumental for math/science identity development (Boaler et al., 2000; Syed et al., 2011).

Relatedly, positive math/science efficacy development occurs in community contexts where youth have access to math/science course work, can engage in math/science enrichment activities in school and in out-of-school settings, and where they can have vicarious experiences of math and science enjoyment and mastery, which are predicated on the STEM pursuits of parents, peers, and teachers (Britner & Pajares, 2006; O’brien et al., 1999; Stevens et al., 2004; Williams & Williams, 2010). Perhaps not surprisingly, within communities, teachers play an outsized role in facilitating the development of math/science efficacy in young people, especially teachers who display a strong sense of mastery of STEM content, who communicate to youth that STEM subjects can be learned, and who demonstrate that they “care” about youth (Fast et al., 2010; Friedel et al., 2010). Notably, this teacher role is evident in the math/science efficacy development of Latino/a youth (Lewis et al., 2012; Riconscente, 2014).

Ecological contexts also vary in external assets for youth (Benson, 2003), particularly Latinos/as (Rodriguez & Morrobel, 2004). Many factors that facilitate math/science identity and efficacy development (e.g., mentors, role models, caring teachers, and enrichment programs) are external developmental assets (Benson, 2003; Search Institute, 2020). A number of other external assets for development—not specifically linked to math/science identity and efficacy development but related to development more broadly-are also embedded within communities, such as receiving support from nonparent adults, having a caring neighborhood and school climate, and living in a community that values youth and sees them as resources (Benson, 2003; Search Institute, 2020). Importantly for this study, many external assets for positive Latino/a youth development also vary across communities, such as supportive neighbors and other adults, safe neighborhoods, and schools with positive climates (Rodriguez & Morrobel, 2004).

Latino/a Destinations and the Development of Math/Science Identity and Efficacy

For Mexican-origin youth, factors related to the history and growth of the Latino/a population within a community may also affect their math/science identity and efficacy development. Newer Latino/a destinations—places that have had noticeable growth of the Latino/a population only in the last 3 decades-differ from established destinations in ways that could be consequential for math/science identity and efficacy development among Mexican-origin youth. Differences between new and established destinations include Latino/a group size, intergroup relations, and institutional supports (Waters & Jiménez, 2005). Variation in these contextual factors could differentiate math/science identity and efficacy development pathways among Mexican-origin youth in several ways. First, destinations may shape predictors of math/science identity development such as academic experiences, exposure to STEM curricula, and access to STEM teachers and mentors. Second, destinations may have different external assets for youth development, such as access to nonparent adults who provide support, caring neighborhoods, and positive school climate. Finally, destinations may provide different resources for Latino/a youth to develop positive ethnic identities, which are linked to positive academic identities and efficacy (Berkel et al., 2010; O’brien et al., 1999; Rivas-Drake et al., 2014).

Prior research yields mixed findings about differences in ecological contexts across destinations. Relative to established Latino/a destinations, new destinations have some contextual attributes that could facilitate the development of strong math/science identity and higher levels of math/science efficacy. Schools in new destinations, for example, have characteristics that are more amenable to positive schooling outcomes in general, such as lower student poverty rates and pupil-to-teacher ratios (Dondero & Muller, 2012). New destinations impose some ecological challenges, however, to youth educational development. Many of the places where Blacks have large test score disparities with Whites (see Reardon et al., 2019) include areas in the U.S. South (Chapel Hill, NC; Charleston, SC; Atlanta, GA) and the Midwest (Minneapolis, MN; Madison, WI)—these places are all new destinations for the Mexican-origin population (Lichter & Johnson, 2009; Zúñiga & Hernández-León, 2005). Mexican-origin youth in these areas could face similar barriers to STEM identity and efficacy development as Black youth in these places. Latinos/as also tend to be more residentially segregated in new destinations (Hall, 2013; Lichter et al., 2010), which could lead to limited exposure to STEM mentors for Mexican-origin youth in new destination areas if these mentors live in different neighborhoods.

Established destinations may also have more contextual assets for immigrant-origin youth development. Relative to new destinations, established destinations have more institutional resources specifically for immigrant-origin youth, such as such as linguistic support in schools (Dondero & Muller, 2012). A larger Latino/a group presence in established destinations may also foster more parent involvement in schools and programs (Klugman et al., 2012) which is an asset for youth development (Search Institute, 2020). Due to their larger Latino/a populations, established destinations may also offer more opportunities for Mexican-origin youth to cultivate a positive ethnic identity and ethnic pride, which could benefit their math/science identity and efficacy development (Berkel et al., 2010; O’brien et al., 1999; Rivas-Drake et al., 2014).

There is also mixed evidence about educational development across destinations. Mexican-origin youth are less likely to be enrolled in school in new destinations (Ackert 2017; Fischer, 2010). In high school, children of immigrants in new destination states have higher reading and math scores in new versus established destinations, but these advantages are ameliorated (for reading) or reversed (for math) when demographic, family, school, and neighborhood characteristics are taken into account (Potochnick, 2014). This finding suggests that such children in new destinations may enjoy assets in their families, schools, and neighborhoods that give them an observed advantage in math. Finally, a study of a cohort of Latino/a youth who were in grade 10 in 2002 found that Latinos/as were less likely than White youth in the same school to take rigorous upper-level advanced math courses (Dondero & Muller, 2012). This finding suggests that, even when they attend the same school, Latinos/as in new destinations may not be able to access rigorous STEM curricula to the same extent as their peers. A number of studies have also reported that associations between new destinations and Mexican-origin educational outcomes are stronger and more negative among children and youth in newly-arrived immigrant families (Ackert et al., 2019; Ackert, 2017; Fischer, 2010). Thus, there may be heightened sensitivity among Mexican-origin immigrant newcomers to ecological factors embedded within destinations that influence math/science identity and efficacy.

Hypotheses

The following hypotheses were developed through empirical patterns in the prior literature and insights from ecological theory.

Hypothesis 1: Mexican-origin youth in new destinations would have significantly different math/science identity and efficacy development patterns than peers in established destinations due to variability in embedded ecological contexts within destination communities.

Hypothesis 2: Developmental differences in math/science identity and efficacy across destinations would be explained by sociostructural factors related to youth, family contexts, and schooling and community contexts, including student and family background characteristics, prior educational and STEM experiences, and school and community attributes.

Hypothesis 3: For Mexican-origin youth, these patterns would vary by generational status, with the largest differences in math/science identity and efficacy development across destinations observed among immigrant youth (1^st generation) and the smallest differences among U.S.-born Mexican-origin children of Latinos/as (3^rd and higher generation).

Method

Data and Sample

Sponsored by the National Center for Education Statistics (NCES; Duprey et al., 2018), the High School Longitudinal Study (HSLS) is a nationally representative sample of approximately 25,000 grade 9 youth in 940 U.S. schools who were enrolled in fall 2009 (note: all sample sizes from the HSLS have been rounded to the nearest 10, per NCES restricted-use license guidelines). In a two-stage sampling design, schools were sampled via a stratified probability proportional to size, then about 30 youth were sampled randomly within each school. This study used data from youth’s Wave 1 (grade 9) and Wave 2 (grade 11) interviews during the 2009–10 and 2011–12 school year, respectively.

We also merged data on school characteristics from the 2009 and 2011 Civil Rights Data Collection (CRDC) into the HSLS, based on a school identifier. The CRDC data allowed us to assign youth in the HSLS to the county where they attended school using the school zip codes in the CRDC. Because CRDC only reported school zip codes for public education agencies (National Forum on Education Statistics, 2018), we restricted the sample to public school youth in three racial/ethnic groups: Mexican-origin Latinos/as (the focal group), non-Latino/a Whites (the most historically advantaged group), and non-Latino/ a Blacks (one of the most historically marginalized groups). For brevity, the non-Latino/a descriptor is dropped when referring to White and Black youth, and the Latino/a descriptor is dropped when referring to Mexican-origin youth.

Comparing Mexican-origin youth to other groups, especially to White peers, can be problematic because it may perpetuate a deficit-oriented rather than asset-oriented perspective of Latino/a youth development (Rodriguez & Morrobel, 2004). We think comparing Mexican-origin youth to White and Black comparison groups is useful, however, for several reasons. First, few studies of racial/ethnic differences in math/science identity and efficacy development use nationally representative data. Given pervasive inequalities in STEM outcomes later in the life course, documenting whether Mexican-origin youth experience lower, comparable, or higher math/science identity and efficacy development as White and Black peers is important. Second, including these two reference groups allowed us to determine whether destination differences in math/science identity and efficacy development (Hypothesis 1) were global (i.e., observed for all three racial/ethnic groups), specific to groups that have historically faced discrimination and segregation (i.e., observed only for Mexican-origin and Black youth), or specific only to a large and fast-growing immigrant population (i.e., observed only for Mexican-origin youth). Finally, including Whites allowed for a comparison of racial/ethnic differences in these STEM outcomes across destinations to corresponding differences in other STEM outcomes across destinations that have been found in previous work (Dondero & Muller, 2012).

The final analytical sample included approximately 10,680 public school youth from the three focal racial/ethnic groups. Table 1 presents the distribution of the sample by race/ethnicity (65% White, 18% Black, and 17% Mexican-origin).

Table 1.

Sample Characteristics for Analysis of Math/Science Identity and Efficacy Across Destinations

Variable	M	SE
Math/science measures
Math identity–grade 9	−0.008	0.016
Math identity–grade 11	−0.013	0.015
Math identity change–grade 9 to grade 11	−0.005	0.014
Math efficacy–grade 9	−0.005	0.018
Math efficacy–grade 11	−0.004	0.016
Math efficacy change- grade 9 to grade 11	0.001	0.017
Science identity–grade 9	−0.022	0.017
Science identity–grade 11	−0.009	0.013
Science identity change–grade 9 to grade 11	0.013	0.016
Science efficacy–grade 9	−0.003	0.019
Science efficacy–grade 11	0.012	0.017
Science efficacy change–grade 9 to grade 11	0.016	0.020
Race, ethnicity, immigrant generation
Mexican origin (all generations)	0.173	0.010
Mexican origin first-generation (foreign-born parents)	0.021	0.003
Mexican origin second/2.5-generation (U.S.-born; 1–2 FB parents)	0.063	0.005
Mexican origin third and higher generation (U.S.-born; U.S.-born parents)	0.033	0.003
Mexican origin unknown generation (parent or student place of birth NA)	0.056	0.005
Black (non-Latino)	0.179	0.012
White (non-Latino)	0.648	0.013
Latino/a destination
New destination	0.244	0.015
Other destination	0.496	0.020
Established destination	0.260	0.016
Individual-level characteristics
Youth age	14.6	0.0
Youth gender (female)	0.496	0.008
Stepparent family structure	0.169	0.005
Mother only family structure	0.195	0.008
Other family structure	0.081	0.004
Both parents family structure	0.554	0.009
Household SES	−0.104	0.016
School mobility (1 time) prior to grade 9	0.160	0.006
School mobility (2+ times) prior to grade 9	0.178	0.007
Unknown school mobility prior to grade 9	0.329	0.007
Changed schools but not counties between grade 9 and grade 11	0.054	0.004
Changed counties between grade 9 and grade 11	0.050	0.003
Unknown school or county changes between grade 9 and grade 11	0.032	0.004
Took Algebra I or above in grade 9	0.820	0.009
Math theta–grade 9	−0.094	0.019
School-level characteristics
Charter or magnet school	0.096	0.019
Total enrollment	1,386	30
School SES	−0.126	0.015
School percent White	60.5	1.3
County-level characteristics
County percent White	68.6	0.7
County Mexican-origin-White dissimilarity	0.389	0.006
County population (in thousands)	875.2	59.1
County per-capita income (in thousands)	26.1	0.4
n	10,680
Number of schools	680
Number of counties	480

Note. Data source was the U.S. Department of Education, National Center for Education Statistics, High School Longitudinal study of 2,009, 2,013 Update and High School Transcripts Restricted-use Data File.

Measurement

Math/Science Outcomes

Math identity, math efficacy, science identity, and science efficacy were measured in grades 9 and 11. NCES created each variable using principal components factor analysis of selected items measured on a four-point Likert scale (disagree to agree). For the math and science identity scales, in fall 2009 students reported how much they agreed or disagreed with two statements: 1) “You see yourself as a math (science) person”; and, 2) “Others see you as a math (science) person.” These questions were repeated in spring of 2011.

The math and science efficacy scales had four items. Students were asked how much they agreed or disagreed with four statements about their fall 2009 math or science course: 1) “You are confident that you can do an excellent job on tests in this course”; 2) “You are certain that you can understand the most difficult material presented in the textbook used in this course”; 3) “You are certain that you can master the skills being taught in this course”; and, 4) “You are confident that you can do an excellent job on assignments in this course.” These items were repeated in spring 2011 concerning the math or science course that students were enrolled in at that time.

For each scale, the coefficient of reliability (alpha) was at least .65. We used the final, standardized variable (centered at zero with a standard deviation of 1.0), each norm-referenced to the full HSLS population for both grades (grade 9 or grade 11). Although all four measures were positively correlated (ranging from .16 between math identity and science efficacy to .58 between math identity and math efficacy), sufficient variation suggested separate evaluation. The grade 11 identity and efficacy measures were the focal outcomes, with multivariate models controlling for the corresponding grade 9 identity or efficacy variable to measure growth. Because of the standardization by grade-level cohort (grade 9, grade 11), multivariate models captured a change in a youth’s relative standing on the measures. As shown in Table 1, by design, the mean math and science identity and efficacy levels were near zero in both grades 9 and 11.

Destinations

Prior research has measured destinations at multiple levels, including the Public-Use Microdata Area (PUMA), place, county, metropolitan area, and state (Ackert, 2017; Fischer, 2010; Hall, 2013; Lichter et al., 2010; Potochnick, 2014). This study used the county level, as all youth in the restricted-use HSLS sample were living in identifiable counties. The HSLS included school identifiers that could be linked to the CRDC. The CRDC provided school addresses, including zip codes, that could be matched to counties using a zip-code to county crosswalk for March 2010 from the U.S. Department of Housing and Urban Development. Thus, counties were assigned to schools based on the location of the school that the student attended, which was not necessarily the county in which the student lived. For cases where zip codes spanned multiple counties, we assigned the zip code to the county with the highest proportion of residential addresses from the zip code.

Following prior research (Ackert et al., 2019; Ackert, 2017; Hall, 2013), this study employed a group-specific typology (i.e., Latino/a presence and growth) to measure destinations. Counties were separated into three Latino/a destination types using data from the 1990 and 2010 decennial censuses: 1) Established Latino/a destinations; 2) new Latino/a destinations; and 3) other Latino/a destinations. Established destinations were counties that had a Latino/a population of 9% or higher in 1990 (the national average in that year). New destinations were counties that had a Latino/a population below 9% in 1990 and met one or both of the following criteria: 1) the county had a Latino/a population above 16% in 2010 (the national average for that year), and/or 2) the county’s Latino/a population growth from 1990 to 2010 was at or above the median for all nonestablished counties (272%) and the county was at least 5% Latino/a by 2010. Other destinations were counties that were less than 9% Latino/a in 1990 and did not experience high Latino/a growth from 1990 to 2010 (less than 272%), or that did experience high Latino/a growth but had very small Latino/a populations in 2010 (less than 5% Latino/a). About 50% of the analytic sample lived in other destinations, with 26% and 24% living in established and new destinations, respectively. This pattern differed in expected ways for the Mexican-origin subgroup, in which 74% lived in established destinations, 15% in new destinations, and 11% in other destinations.

Generational Status

Mexican-origin adolescents were categorized into generational groups based on the country of birth of themselves and their parents: (a) first-generation (youth born outside of the United States); (b) second-generation and 2.5-generation (youth born in the United States to at least one foreign-born parent); 3) third and higher generation (youth born in the United States to only U.S.-born parents); and, 4) youth with unknown generational status. Each group consisted of at least 120 youth, with 1st-generation Mexican-origin youth accounting for 2% of the sample (12% of the Mexican-origin subgroup), 2nd/2.5-generation Mexican-origin 6% (37% of the Mexican-origin subgroup), 3rd and higher generation 3% (19% of the Mexican-origin subgroup), and unknown generation 6% (33% of the Mexican-origin subgroup). HSLS did not measure citizenship, and native language was strongly correlated with immigrant generation. For instance, a non-English language was the native language of 96% of 1st-generation Mexican-origin youth but only 25% of 2nd/2.5- and 3rd and higher generation Mexican-origin youth.

School-Related Covariates

One particular need was to account for cases in which youth moved in general and moved between destinations in particular, as school mobility can affect identity and efficacy development by disrupting peer networks and continuity of learning environments (Rumberger & Larson, 1998). Because of our interest in exposure to schools within destinations over time, two covariates measured mobility between schools between grades 9 and 11. The first considered movement prior to high school. Youth were categorized as: 1) not changing school for reasons other than promotion; 2) switching schools one time for a reason other than promotion; 3) changing schools two or more times for reasons other than promotion; and, 4) unknown school mobility prior to grade 9. The second considered school changes between grades 9 and 11. Rather than focus on the number of transfers, we looked at whether youth switched schools and counties between grades 9 and 11 with a four-category variable: 1) did not change schools between grades 9 and 11; 2) changed schools but not counties between grades 9 and 11; 3) changed counties between grades 9 and 11; and, 4) unknown school or county change between grades 9 and 11. Unknown school or county change resulted from HSLS school identifiers not being linked to CRDC identifiers, most likely because the student transferred to a nonpublic school. Table 1 shows that the majority of youth were in the same school in grades 9 and 11. Because math/science identity and efficacy may reflect prior mastery of these subject areas (Britner & Pajares, 2006), we measured grade 9 math background as: 1) a dichotomous variable of having taken a math course at or above Algebra I in grade 9, and, 2) a continuous variable measuring performance on a standardized assessment in grade 9. The theta score measured achievement relative to the population of youth in grade 9 in the fall of 2009.

School Covariates

Given that some magnet schools may have STEM-focused programming and curricula (Bottia et al., 2018; Dixon et al., 2020); we included a dichotomous variable measuring whether the school was a charter or magnet school versus a traditional public school. Total enrollment of the school was a continuous variable equal to the number of youth attending the school in 2009 (grade 9). We also included a measure from CRDC of the percent of the school that was (non-Latino) White in 2009, as schools with more White youth may have more access to high-level math courses like calculus (National Science Board, 2018; U.S. Department of Education, Office for Civil Rights, 2014) and more experienced math and science teachers (Jackson, 2009). When schools did not report this figure, we used the 2011 CRDC measure or the school administrator reports from HSLS (if neither CRDC measure was available). The correlations between the 2009 and 2011 CRDC measures and HSLS administrator reports of percent White were high (.97 and .96, respectively) among youth who had information from all three sources. Because socioeconomically advantaged schools offer many of the benefits that schools with higher percentages of White students offer, and are linked to higher student attainment (Palardy, 2013), school SES was controlled in multivariate models. School SES was calculated using in-sample student-level data aggregated to the school they attended in grade 9. Although within-school sample weights were not included in HSLS, we used the base-year weight for each student when averaging the student SES to the school level. Thus, for each student i in school j, the average SES level of the school was equal to the following:

$$\begin{gathered} [({\text{Σ(Base–year SES composite for Student}}_{\text{ij}}) \\ \times \left({\text{Base–year Weight for Student t}}_{\text{ij}}\right)/ \\ \text{Σ}\left({\text{Base–year Weight for Student}}_{\text{ij}}\right)] \end{gathered}$$ (1)

Community Covariates

County characteristics from the 2007–2011 American Community Survey (ACS) and 2010 decennial census captured specific attributes of Latino/a destinations that could be related to institutional supports, group size, and intergroup relations (Waters & Jiménez, 2005). Borrowing from the neighborhood effects on health literature, these measures also holistically capture levels of advantage/disadvantage in county contexts (Ross & Mirowsky, 2001). A measure of county percent White was calculated from ACS. With 2010 decennial census data, we also calculated the index of dissimilarity for each county in order to measure residential segregation between Mexican-origin Latino/a and White populations across census tracts within the county (Massey & Denton, 1988). ACS also allowed measurement of county population size and per-capita income (note: both were logged in multivariate models).

Sociodemographic Covariates

Multivariate analyses controlled for a number of student and household characteristics that past research has reported to be correlated with the math/science variables as well as student sorting across communities and schools. We included a standardized index of household socioeconomic status from grade 9 in the HSLS, which NCES created by averaging parental educational attainment, family income, and parental occupation measures. Family structure was a four-category variable: 1) lived with both biological or adoptive parents; 2) lived with one biological or adoptive parent and one stepparent; 3) lived with only their biological or adoptive mother, and 4) another family structure. The adolescent’s gender was dichotomous (with 1 = female), and student age was continuous (in months).

Analytical Strategy

All analyses were weighted by the HSLS panel weight (grade 9 to grade 11) and HSLS school weight to account for sample design and survey attrition. Hypothesis-testing used mixed effects (i.e., multilevel) linear regressions in Stata 15.0. The multilevel models were two-level random intercept models that nested youth (level 1) within schools (level 2), thereby accounting for variation in math/science outcomes attributed to schools. We explored nesting youth in schools and then in counties (level 3), but many counties had insufficient numbers of schools (we cannot give a precise number because of restricted-use NCES data license regulations). Models nesting youth in counties instead of schools revealed similar results to those presented. County characteristics in this analysis were thus considered attributes of schools.

Four separate two-level mixed effects models were estimated for each outcome. Model 1 included race/ethnic/immigrant generational groups (with Whites as the reference group) and destinations (with established destinations as the reference group), with no covariates. Model 2 added a two-way interaction between destination and race/ethnicity to see if destination associations with the focal outcomes varied by race/ethnicity. Model 3 added all covariates, including individual-level characteristics, school and county characteristics, and prior background in the subject area of interest. Finally, model 4 included a three-way interaction to see if net results from model 3 varied further for Mexican-origin immigrant generational groups.

Additional models excluded the grade 9 identity or efficacy measures to examine grade 11 outcomes rather than changes in outcomes. These results, available upon request, generally showed no significant differences in grade 11 math/science identity or math efficacy between Mexican-origin and White youth and no significant variations in these differences across destinations. For science efficacy in grade 11, Mexican-origin youth who were U.S.-born children of immigrants (2nd/2.5-generation) had slower growth than Whites in new versus established destinations, and this result was consistent with the multivariate estimates of changes in science efficacy from grade 9 to grade 11 that are presented below.

Research Ethics Committee Approval

This study is part of the broader project “Family and School Contexts of Young People’s Health and Wellbeing”, which was declared exempt human subjects research by the Institutional Review Board at the University of Texas at Austin, Office of Research Support and Compliance, under study number 2016-05-0080. The study was determined to be exempt because it obtained information from publicly available data sets and did not meet the criteria for human subjects research defined in the Common Rule (45 CFR 4) or FDA Regulations (21 CFR 56).

Results

Describing Math/Science Identity and Efficacy Across Destinations

Table 2 shows a sample of weighted mean levels of math/science outcomes for the three racial/ethnic groups of interest. We present focal outcome measures for grade 11, with one-sample t-tests determining whether within-person changes in these measures between grades 9 and 11 significantly differed from zero. The up arrows in Table 2 indicate when the grade 11 outcome was significantly higher than in grade 9, and the down arrows indicate when the grade 11 outcome was significantly lower than in grade 9. T-tests evaluated whether values on these outcomes in grade 11 differed by race/ethnicity/immigrant generation group, destination type, and race/ethnicity/immigrant generation group by destination. Table 2 shows significant differences in grade 11 between groups, designated with superscripts, but that adolescents’ (relative) level of math/science identity and efficacy in grade 11 did not differ appreciably from their grade 9 level for most groups. For this reason, we focus this initial discussion on differences in levels of math/science identity and efficacy in grade 11 instead of on changes between grades.

Table 2.

Math and Science Identity and Efficacy by Race/Ethnicity/Immigrant Status and Destination in Grade 11

	Math identity			Math efficacy			Science identity			Science efficacy
Subsample	M	SE		M	SE		M	SE		M	SE
Race/ethnicity/immigrant generation
White	−0.013	0.016		−0.023	0.015		0.059	0.015		0.036	0.018
Black	0.023	0.044		0.134a	0.059		−0.078a	0.035		0.130	0.050
Mexican-origin (all)	−0.052	0.035		−0.066	0.043		−0.192a	0.040		−0.206a	0.043
Mexican-origin first generation	0.033	0.082		0.002	0.091		−0.239a	0.098		−0.318a	0.090
Mexican-origin second/2.5 generation	0.057	0.066		0.006	0.076		−0.109a	0.071		−0.155a	0.065
Mexican-origin 3+ generation	−0.101	0.083		−0.075	0.107		−0.216a	0.064		−0.074	0.101
Mexican-origin unknown generation	−0.178a	0.055		−0.169	0.076		−0.255a	0.086		−0.320a	0.069	↓
Destination
Established destination	−0.028	0.033		−0.013	0.032		−0.083	0.031		−0.054	0.037
New destination	−0.048	0.026	↓	−0.018	0.024		−0.023	0.025		0.030	0.032
Other destination	0.012	0.020		0.007	0.023		0.037a	0.018		0.036b	0.024
Race/ethnicity/immigrant generation by destination
White–Established destination	−0.020	0.058		0.031	0.043	↓	0.030	0.042		0.107	0.060
White–New destination	−0.040	0.028		−0.022	0.025		0.060	0.024		0.063	0.031
White–Other destination	0.002	0.021		−0.035	0.021		0.066	0.020	↓	0.008	0.022
Black–Established destination	−0.157	0.088		−0.050	0.111		−0.100	0.071		−0.043	0.064
Black–New destination	0.038	0.064	↓	0.085	0.076		−0.169c	0.081		0.112	0.102
Black–Other destination	0.095	0.059		0.239c	0.075		−0.025	0.042		0.218	0.057
Mex. origin first gen.–Established destination	0.200	0.114		0.020	0.143		−0.077	0.184		−0.311c	0.136
Mex. origin first gen.–New destination	−0.034	0.137		0.235	0.135		−0.395c	0.106		−0.378c	0.153
Mex. origin first gen.–Other destination	−0.258c	0.102	↓	−0.172	0.130		−0.434c	0.072		−0.246	0.221
Mex. origin second/2.5 gen.–Established destination	0.115	0.077		0.043	0.085		−0.059	0.085		−0.143c	0.077
Mex. origin second/2.5 gen.–New destination	−0.153	0.137		−0.248	0.232		−0.359c	0.136		−0.374c	0.222
Mex. origin second/2.5 gen.–Other destination	−0.153	0.140		0.035	0.178		−0.190	0.136		0.063	0.224
Mex. origin 3+ gen.–Established destination	−0.073	0.096		−0.074	0.132		−0.251c	0.077		−0.031	0.124
Mex. origin 3+ gen.–New destination	−0.277	0.189		0.024	0.143		−0.201	0.184		−0.250	0.237
Mex. origin 3+ gen.–Other destination	−0.127	0.139		−0.156	0.160	↑	−0.029	0.096		−0.225c	0.140
Mex. origin unknown gen.–Established destination	−0.119	0.065	↑	−0.114	0.097		−0.234c	0.108		−0.308c	0.088	↓
Mex. origin unknown gen.–New destination	−0.404c	0.100		−0.299c	0.122		−0.331c	0.137		−0.352c	0.130
Mex. origin unknown gen.–Other destination	−0.258	0.117		−0.444c	0.115		−0.291c	0.088	↓	−0.365c	0.093

Note. ↑ = grade 11 math/science identity/efficacy was significantly higher than in grade 9 at p < .05; ↓ = grade 11 math/science measure was significantly lower than in grade 9 at p < .05.

^asignificantly different from White youth at p < .05.

^bsignificantly different from youth in established destinations at p < .05.

^csignificantly different from White youth in established destinations at p < .05.

Source was the U.S. Department of Education, National Center for Education Statistics, High School Longitudinal study of 2009, 2013 Update and High School Transcripts Restricted-use Data File.

Table 2 reveals that differences between Mexican-origin youth overall and by generational subgroup with Whites were mainly in science. Differences with Whites for Mexican-origin youth overall and of any generational status were nonsignificant for grade 11 math identity and efficacy. White youth had higher science identity in grade 11 than Mexican-origin youth overall (t = 5.85, p < .001) and each of the generational subgroups. Science efficacy in grade 11 was higher for White than for Mexican-origin youth overall (t = 5.36, p < .001) and for Mexican-origin 1st- (t = 3.96, p < .001) and 2nd/2.5-generations (t = 2.84, p < .01). Notably, Whites’ science efficacy in grade 11 did not significantly differ from 3rd and higher generation Mexican-origin youth, suggesting that Mexican-origin youth with higher generation status may have greater assets for science efficacy than their 1st- and 2nd/2.5-generation peers.

Table 2 also shows differences in the outcomes by destination and by race/ethnicity/immigrant generation groups within destinations. In the pooled sample, no grade 11 outcome significantly differed between new and established destinations. Between grades 9 and 11, however, youth in new destinations experienced significant decreases in their relative standing in math identity. Table 2 also shows that the main differences in grade 11 math/science identity and efficacy for race/ethnicity/immigrant generation groups in new versus established destinations were observed for science rather than for math outcomes. In general, across all destinations and among different generational subgroups, Mexican-origin youth showed parity in their math identity and math efficacy with their White peers in established destinations.

Some Mexican-origin generational subgroups in new destinations, however, had lower science identity and efficacy levels in grade 11 than White peers in established destinations. For example, Mexican-origin youth in the 1st- and 2nd/2.5-generation groups in new destinations had lower science identity in grade 11 (t = −3.71, p < .001 and t = —2.73, p < .01, respectively) than Whites in established destinations. Blacks in new destinations showed a similar disadvantage for science identity in grade 11 (t = −2.18, p < .05) but Whites in new destinations did not. In both new and established destinations, Mexican-origin youth in immigrant households (1st- and 2nd/2.5-generation groups) also had lower science efficacy in grade 11 (t = −2.95, p < .01 and t = —2.09, p < .05, respectively) than Whites in established destinations. These results can be interpreted in many ways, including as pointing to assets related to being a higher-generation Mexican-origin youth (for science efficacy).

Based on a subset of values from Table 2, Figure 1 shows differences in mean levels of math/science identity and efficacy in grade 11 for Mexican-origin youth of all generations and White youth by destination. There were no significant differences in the focal outcomes among White youth across the three destination types. The only outcome where Mexican-origin youth in new destinations differed significantly from their Mexican-origin peers in established destinations was math identity, which was higher in established destinations. Disparities in these four outcomes between Mexican-origin youth and their White peers also differed by destination. In established destinations, Mexican-origin youth had significantly lower science efficacy than their White peers, but they had statistically similar levels of math identity, math efficacy, and science identity. In contrast, in new destinations, Mexican-origin youth had lower levels of three out of the four outcomes (math identity, science identity, and science efficacy) compared to Whites in these destinations. Mexican-origin youth in other destinations also had lower levels of math/science identity than Whites in these areas. These patterns suggest that Mexican-origin youth in established destinations may have more assets for positive math/science identity.

Figure 1. Math/Science Identity and Efficacy Levels Among Mexican-Origin and White Youth Across Destinations.

Note. Data source was the U.S. Department of Education, National Center for Education Statistics, High School Longitudinal study of 2009, 2013 Update and High School Transcripts Restricted-use Data File.

Math Identity Growth

Table 3 shows the results for math identity growth. Results from model 1 indicate that, net of destination of residence, Mexican-origin and Black youth had comparable growth in math identity with White peers. Net of race/ethnicity, however, youth in new destinations had significant declines in their math identity relative to youth in established destinations (b = −.11,p < .05). Destination associations with math identity, however, varied by race/ethnicity (Table 3, model 2). For Whites and Blacks, there was virtually no significant variation in math identity change across destinations. For Mexican-origin youth, however, math identity growth varied significantly across destinations. Mexican-origin youth in established destinations had significantly higher growth in math identity than Whites in established destinations (b = .17, p < .05). This math identity advantage for Mexican-origin youth in established destinations, was attenuated in new destinations (b = −.30, p < .01) and other destinations (b = −.31, p < .05).

Table 3.

Growth in Math Identity by Race/Ethnicity/Immigrant Status and Destination

	Model 1		Model 2		Model 3		Model 4
Variable	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
Race/ethnicity/immigrant generation
Black	0.035	0.049	−0.114	0.090	−0.011	0.091	−0.003	0.091
Mexican-origin (all)	0.033	0.058	0.170	0.075*	0.220	0.090*
Mexican-origin first generation							0.136	0.207
Mexican-origin second/2.5 generation							0.265	0.102**
Mexican-origin 3+ generation							0.125	0.096
Mexican-origin unknown generation (ref. White)							0.211	0.123+
Destination
New destination	−0.108	0.053*	−0.043	0.068	−0.080	0.080	−0.069	0.078
Other destination (ref. Established destination)	−0.050	0.051	−0.013	0.063	−0.027	0.084	−0.012	0.081
Interactions
Black × New Destination			0.095	0.106	0.078	0.110	0.068	0.110
Black × Other Destination			0.193	0.111+	0.215	0.113+	0.187	0.108+
Mexican-Origin (all) × New Destination			−0.298	0.090**	−0.189	0.106+
Mexican-Origin (all) × Other Destination			−0.312	0.145*	−0.242	0.170
Mexican-Origin First Generation × New Destination							−0.239	0.264
Mexican-Origin First Generation × Other Gestination							−0.861	0.390*
Mexican-Origin Second/2.5 Generation × New Destination							−0.249	0.126*
Mexican-Origin Second/2.5 Generation × Other Destination							−0.177	0.187
Mexican-Origin 3+ Generation × New Destination							−0.247	0.189
Mexican-Origin 3+ Generation × Other Destination							0.044	0.276
Mexican-Origin Unknown Generation × New Destination							−0.198	0.155
Mexican-Origin Unknown Generation × Other Destination							−0.265	0.144+
Grade 9 math identity	0.542	0.018***	0.541	0.018***	0.478	0.015***	0.476	0.016***
Immigrant generation	No		No		Yes		No
Individual level characteristics	No		No		Yes		Yes
School characteristics	No		No		Yes		Yes
County characteristics	No		No		Yes		Yes
Math background	No		No		Yes		Yes
Constant	0.021	0.049	−0.018	0.059	−1.606	0.831+	−1.540	0.849+
Random effects parameters
School variance	0.017	0.006	0.017	0.006	0.015	0.005	0.015	0.005
Residual variance	0.647	0.018	0.645	0.018	0.604	0.018	0.604	0.018
n	10,440		10,440		10,440		10,440
Number of schools	740		740		740		740

Note. Results were also estimated without grade 9 math identity and are available upon request. Data source was the U.S. Department of Education, National Center for Education Statistics, High School Longitudinal study of 2009, 2013 Update and High School Transcripts Restricted-use Data File.

⁺p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

Model 3 in Table 3 indicates, however, that much of the variation in math identity growth among Mexican-origin youth across destinations was due to differences in background and contextual factors. Mexican-origin youth in established destinations had significantly higher growth in math identity than Whites in new destinations (b = .22, p < .05), and this association was not moderated significantly (p < .05) for Mexican-origin youth in the other destinations. Model 4 in Table 3 further shows that the net Mexican-origin advantage in math identity in established destinations was only observed for the 2nd/2.5 generation (b = .27, p < .01), not for all Mexican-origin generational groups in these areas. The 2nd/2.5 generation in new destinations, however, did not exhibit this math identity advantage; their math identity growth was attenuated (b = −.25, p < .05) relative to their 2nd/2.5 generation peers in established destinations.

To summarize the results for math identity growth, Mexican-origin youth exhibited significant variation in math identity growth by destination, whereas White and Black youth did not. This contextual variation in math identity among Mexican-origin youth was due to individual, household, school, and community background factors. In the full model, Mexican-origin youth in established destinations had significantly higher math identity growth in these destinations, and this association was not moderated for other Mexican-origin groups by destination of residence. The full model with interactions for generational status revealed a math identity growth advantage relative to Whites only for 2nd/2.5-generation Mexican-origin youth in established destinations. Overall, these results suggest that math identity growth is an asset for Mexican-origin youth overall, especially for the 2nd/2.5-generation in established destinations.

Math Efficacy Growth

Results for math efficacy growth are displayed in Table 4. Mexican-origin youth did not significantly differ from Whites in math efficacy growth (model 1). Math efficacy growth did not vary significantly among Mexican-origin youth across new versus established destinations, in models without and with all covariates (model 2, model 3). Furthermore, net math efficacy growth did not vary among Mexican-origin immigrant generational subgroups across destinations (model 4). A few differences in math efficacy growth by race/ethnicity and destination were notable. Mexican-origin youth in “other” destinations showed significantly lower math efficacy growth relative to peers in established destinations (model 2; b = −.35, p < .05), but this disadvantage was largely explained by background and contextual factors (model 3). In sum, math efficacy growth is a domain where Mexican-origin youth—regardless of destination of residence and across immigrant generation groups within destinations—were similar to Whites. Additionally, unlike for math identity growth, Mexican-origin youth in established destinations did not have higher growth in math efficacy than peers in new destinations.

Table 4.

Growth in Math Efficacy by Race/Ethnicity/Immigrant Status and Destination

	Model 1		Model 2		Model 3		Model 4
Variable	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
Race/ethnicity/immigrant generation.
Black	0.223	0.081**	−0.052	0.085	0.066	0.093	0.075	0.090
Mexican-origin (all)	−0.109	0.074	−0.024	0.094	0.007	0.120
Mexican-origin first generation							−0.232	0.431
Mexican-origin second/2.5 generation							0.112	0.160
Mexican-origin 3+ generation							−0.144	0.144
Mexican-origin unknown generation (ref. White)							0.071	0.133
Destination
New destination	−0.096	0.048*	−0.066	0.064	−0.067	0.089	−0.061	0.086
Other destination (ref. Established destination)	−0.100	0.051+	−0.102	0.056+	−0.088	0.090	−0.082	0.085
Interactions
Black × New Destination			0.051	0.128	0.018	0.127	0.014	0.125
Black × Other Destination			0.371	0.128**	0.339	0.144*	0.325	0.132*
Mexican-Origin (all) × New Destination			−0.092	0.152	−0.069	0.172
Mexican-Origin (all) × Other Destination			−0.354	0.169*	−0.339	0.194+
Mexican-Origin First Generation × New Destination							0.470	0.451
Mexican-Origin First Generation × Other Gestination							−0.668	0.632
Mexican-Origin Second/2.5 Generation × New Destination							−0.238	0.209
Mexican-Origin Second/2.5 Generation × Other Destination							0.082	0.189
Mexican-Origin 3+ Generation × New Destination							0.164	0.183
Mexican-Origin 3+ Generation × Other Destination							−0.036	0.206
Mexican-Origin Unknown Generation × New Destination							−0.256	0.223
Mexican-Origin Unknown Generation × Other Destination							−0.544	0.339
Grade 9 math identity	0.328	0.024***	0.329	0.024***	0.273	0.025***	0.274	0.024***
Immigrant generation	No		No		Yes		Yes
Individual level characteristics	No		No		Yes		Yes
School characteristics	No		No		Yes		Yes
County characteristics	No		No		Yes		Yes
Math background	No		No		Yes		Yes
Constant	0.057	0.037	0.051	0.043	0.500	0.862	0.679	0.903
Random effects parameters
School variance	0.026	0.009	0.027	0.009	0.022	0.008	0.023	0.008
Residual variance	0.858	0.032	0.853	0.032	0.815	0.028	0.811	0.029
n	9,200		9,200		9,200		9,200
Number of schools	740		740		740		740

Note. Results were also estimated without grade 9 math efficacy and are available upon request. Data source was the U.S. Department of Education, National Center for Education Statistics, High School Longitudinal study of 2009, 2013 Update and High School Transcripts Restricted-use Data File.

⁺p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

Science Identity Growth

Table 5 displays results for science identity growth. Model 1 in Table 5 shows that, net of destination of residence, Mexican-origin youth had decreased growth in science identity relative to Whites (b = −.13, p < .05). This pattern varied by destination. Mexican-origin youth in established destinations did not significantly differ from White youth in established destinations in changes in science identity (model 2), even in the full model (model 3). Interactions were significant and negative, however, for Mexican-origin youth in new (b = −.32, p < .01) and other (b = −.38, p < .01) destinations (model 2). This pattern of shallower science identity growth in new and other destinations persisted in the full model (model 3). Notably, science identity growth also varied among White and Black youth across destinations, in models without and with covariates (models 2–3). Science identity growth in new and other destinations declined for Blacks relative to their peers in established destinations, but it increased for Whites in new and other destinations relative to their peers in established destinations (model 3).

Table 5.

Growth in Science Identity by Race/Ethnicity/Immigrant Status and Destination

	Model 1		Model 2		Model 3		Model 4
Variable	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
Race/ethnicity/immigrant generation
Black	0.005	0.050	0.264	0.115*	0.219	0.121+	0.219	0.122+
Mexican-origin (all)	−0.133	0.059*	0.061	0.068	−0.035	0.077
Mexican-origin first generation							0.134	0.182
Mexican-origin second/2.5 generation							−0.007	0.070
Mexican-origin 3+ generation							−0.102	0.084
Mexican-origin unknown generation (ref. White)							0.071	0.173
Destination
New destination	0.006	0.057	0.150	0.063*	0.124	0.062*	0.125	0.061*
Other destination (ref. Established destination)	0.047	0.059	0.166	0.060**	0.162	0.065*	0.165	0.064*
Interactions
Black × New Destination			−0.424	0.131**	−0.390	0.131**	−0.385	0.131**
Black × Other Destination			−0.273	0.131*	−0.255	0.132+	−0.257	0.133+
Mexican-Origin (all) × New Destination			−0.318	0.101**	−0.255	0.101*
Mexican-Origin (all) × Other Destination			−0.376	0.129**	−0.308	0.128*
Mexican-Origin First Generation × New Destination							−0.346	0.204+
Mexican-Origin First Generation × Other Gestination							−0.826	0.226***
Mexican-Origin Second/2.5 Generation × New Destination							−0.294	0.111**
Mexican-Origin Second/2.5 Generation × Other Destination							−0.214	0.107*
Mexican-Origin 3+ Generation × New Destination							−0.071	0.141
Mexican-Origin 3+ Generation × Other Destination							0.098	0.148
Mexican-Origin Unknown Generation × New Destination							−0.314	0.232
Mexican-Origin Unknown Generation × Other Destination							−0.493	0.259+
Grade 9 science identity	0.400	0.018***	0.401	0.018***	0.375	0.019***	0.377	0.019***
Immigrant generation	No		No		Yes		Yes
Individual level characteristics	No		No		Yes		Yes
School characteristics	No		No		Yes		Yes
County characteristics	No		No		Yes		Yes
Math background	No		No		Yes		Yes
Constant	−0.033	0.057	−0.144	0.054**	−0.450	0.849	−0.299	0.853
Random effects parameters
School variance	0.016	0.006	0.016	0.006	0.010	0.005	0.010	0.005
Residual variance	0.758	0.029	0.755	0.030	0.745	0.029	0.745	0.029
n	10,380		10,380		10,380		10,380
Number of schools	740		740		740		740

Note. Results were also estimated without grade 9 science identity and are available upon request. Data source was the U.S. Department of Education, National Center for Education Statistics, High School Longitudinal study of 2009, 2013 Update and High School Transcripts Restricted-use Data File.

⁺p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

Science identity growth across destinations was moderated by immigrant generational status within the Mexican-origin group (model 4). In established destinations, all Mexican-origin immigrant generational subgroups had comparable science identity growth with Whites in these areas (model 4). In contrast, Mexican-origin youth who were 1st and 2nd/2.5 generation had declining science identity growth in new and other versus established destinations.

In summary, Mexican-origin and Black youth in new and other destinations had lower net science identity growth than their peers in established destinations, but this pattern was reversed for Whites, who had higher science identity growth in new and other destinations than in established destinations. Mexican-origin 1st and 2nd/2.5-generation youth in new destinations had lower science identity growth than their peers in established destinations.

Science Efficacy Growth

Science efficacy growth results are displayed in Table 6. Mexican-origin youth in established destinations had larger decreases in science efficacy from grade 9 to grade 11 relative to Whites in established destinations (model 1; b = −.21, p < .001). In the full model (model 3), science efficacy declines persisted for Mexican-origin youth in established destinations (b = −.28, p < .001). This decline for Mexican-origin youth relative to White youth in established destinations did not differ significantly for Mexican-origin youth in new destinations. It was attenuated for Mexican-origin youth in other destinations (b = .20, p < .05). Thus, Mexican-origin youth in other destinations had smaller declines in science efficacy than their peers in established destinations. Three-way interaction models revealed differences in science efficacy growth within the Mexican-origin subgroup by immigrant generation (model 4). In established destinations, Mexican-origin 3rd and higher generation youth had lower science efficacy growth than Whites in these destinations (b = −.27, p < .01), but 1st- and 2nd/2.5-generation Mexican-origin youth had parity with Whites in science efficacy growth these areas. Mexican-origin 2nd/2.5-generation youth in new destinations had more negative growth in science efficacy (b = −.44, p < .05) relative to their peers in established destinations.

Table 6.

Growth in Science Efficacy by Race/Ethnicity/Immigrant Status and Destination

	Model 1		Model 2		Model 3		Model 4
Variable	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
Race/ethnicity/immigrant generation
Black	0.178	0.067**	−0.128	0.138	−0.104	0.139	−0.104	0.139
Mexican-origin (all)	−0.209	0.041***	−0.237	0.047***	−0.277	0.072***
Mexican-origin first generation							−0.631	0.439
Mexican-origin second/2.5 generation							−0.133	0.112
Mexican-origin 3+ generation							−0.267	0.103**
Mexican-origin unknown generation (ref. White)							−0.365	0.095***
Destination
New destination	0.013	0.052	−0.013	0.069	0.016	0.078	0.019	0.078
Other destination (ref. Established destination)	0.005	0.051	−0.055	0.064	0.004	0.080	0.005	0.080
Interactions
Black × New Destination			0.229	0.175	0.165	0.170	0.171	0.169
Black × Other Destination			0.384	0.159*	0.335	0.153*	0.341	0.151*
Mexican-Origin (all) × New Destination			−0.097	0.102	−0.040	0.106
Mexican-Origin (all) × Other Destination			0.093	0.108	0.202	0.093*
Mexican-Origin First Generation × New Destination							0.506	0.482
Mexican-Origin First Generation × Other Destination							0.697	0.512
Mexican-Origin Second/2.5 Generation × New Destination							−0.443	0.176*
Mexican-Origin Second/2.5 Generation × Other Destination							0.166	0.222
Mexican-Origin 3+ Generation × New Destination							−0.185	0.250
Mexican-Origin 3+ Generation × Other Destination							0.196	0.147
Mexican-Origin Unknown Generation × New Destination							0.161	0.167
Mexican-Origin Unknown Generation × Other Destination							0.114	0.144
Ninth grade Outcome Grade 9 science efficacy	0.327	0.021***	0.327	0.021***	0.306	0.024***	0.307	0.024***
Immigrant generation	No		No		Yes		Yes
Individual level characteristics	No		No		Yes		Yes
School characteristics	No		No		Yes		Yes
County characteristics	No		No		Yes		Yes
Math background	No		No		Yes		Yes
Constant	0.013	0.043	0.060	0.056	0.430	0.905	0.541	0.890
Random effects parameters
School variance	0.028	0.009	0.028	0.009	0.024	0.009	0.024	0.009
Residual variance	0.839	0.031	0.837	0.031	0.817	0.032	0.817	0.032
n	8,240		8,240		8,240		8,240
Number of schools	740		740		740		740

Note. Results were also estimated without grade 9 science efficacy and are available upon request. Data source was the U.S. Department of Education, National Center for Education Statistics, High School Longitudinal study of 2009, 2013 Update and High School Transcripts Restricted-use Data File.

^*p < .05.

^**p < .01.

^***p < .001.

Within established destinations, therefore, 3rd and higher generation Mexican-origin youth had the lowest science efficacy growth relative to White peers, a pattern that extended to the other two destinations. Within new destinations, the 2nd and higher generation had lower science efficacy growth than peers in established destinations. These disadvantages were not observed for Black peers, who had comparable science efficacy development with Whites in established destinations and new destinations and higher science efficacy development in other destinations. Science efficacy growth among White youth did not vary across destinations.

Conclusion

This study focused on math and science identity and efficacy development from grades 9 through 11 among Mexican-origin adolescents in established, new, and other destination counties. Youth of Mexican origin are the largest Latino/a subgroup, and they have many assets for positive math/science identity and efficacy development, but also face some challenges. Such youth are increasingly living in diverse array of communities across the United States, and the multilayered and interconnected ecological contexts in these communities could be creating differential pathways—either capitalizing on developmental assets or increasing challenges for math/science efficacy and identity development.

We found some support for the first hypothesis that Mexican-origin youth math and science efficacy and identity development differed between new and established destinations. Multivariate models without covariates revealed that Mexican-origin youth in new destinations had declining levels of math identity and science identity relative to Mexican-origin youth in established destinations. Mexican-origin youth in established destinations had higher growth in math identity relative to Whites in these locations and comparable changes in science identity. In other words, Mexican-origin youth in established destinations had more positive developmental patterns for math and science identity than their Mexican-origin peers in new destinations.

The second hypothesis—that these differences across destinations would be explained by observable factors related to general development and differential selection into destinations—was supported for math identity. Differences between Mexican-origin youth in new and established destinations persisted, however, when looking at science identity. Thus, factors not measured in this analysis contributed to an apparent disadvantage for Mexican-origin youth in new destinations for science identity development relative to those in established destinations.

Finally, we found support for the third hypothesis that generational status would moderate associations between destinations and the focal outcomes among Mexican-origin youth. Unexpectedly, however, we did not always find that these associations were strongest among the 1st generation and weakest among the 3rd and higher generation. Models revealed that the 2nd/2.5-generation in established destinations were more advantaged than their peers in new destinations for math and science identity and science efficacy. The 1st generation experienced a similar advantage in established destinations for science identity, but not other outcomes. Some subgroups in established destinations had more positive outcomes than Whites, such as increased math identity growth among the 2nd/2.5 generation relative to Whites in these areas.

These results reveal patterns of developmental assets and risks for Mexican-origin youth in their math/science identity development across the places where they live. Mexican-origin youth in established destinations had several positive outcomes, such as more advantaged math identity and comparable math efficacy and science identity growth relative to Whites in these areas. One exception did appear, as science efficacy development declined for Mexican-origin youth relative to White youth in established destinations. Further, findings show the importance of accounting for both generational status and place of residence when studying STEM development among Mexican-origin youth. We found that Mexican-origin youth had lower observed levels of science identity and efficacy in grade 11 than White youth, which is consistent with prior studies (Kitsantas et al., 2011; Riegle-Crumb et al., 2011), but these observed patterns varied by location and generational status. For example, science efficacy development was an area of disadvantage relative to Whites for 3rd and higher generation Mexican-origin youth in established destinations and 2nd/2.5 generation youth in new destinations. Thus, location and generation interactively shape STEM development for Mexican-origin youth.

The common narrative of Mexican-origin disadvantage in much of the literature was also countered by our results. This study found areas of strength among Mexican-origin youth such as positive math identity growth in established destinations. Additionally, with the exception of science efficacy growth, there were no significant differences between 3rd and higher generation Mexican-origin youth and Whites in established destinations. Mexican-origin youth whose families have been in the country for generations, therefore, largely experienced comparable outcomes to Whites in these destinations.

The disadvantaged outcomes for Mexican-origin in new destinations found in this study, especially among the 2nd/2.5 generation, are consistent with prior research showing disadvantaged outcomes in educational domains in these locations among Latino/ a and Mexican-origin children and youth who are immigrants or children of immigrants (Ackert et al., 2019; Ackert, 2017; Dondero & Muller, 2012; Fischer, 2010). What factors might explain a disadvantage in new destinations? Multivariate models suggest that factors specific to Mexican-origin and/or non-White populations, rather than to all youth in these areas, may be salient. For example, math identity growth, math efficacy growth, and science efficacy growth did not differ among Whites or Blacks across new versus established destinations. For the outcome of science identity growth, Whites in new destinations had increased growth, whereas Black and Mexican-origin youth in new destinations had decreased growth. Thus, White youth in new destinations did not exhibit the same types of disadvantages as Mexican-origin youth in these areas. Vulnerabilities among Mexican-origin youth in new destinations could also be due to unmeasured attributes such as living in a mixed-status household with undocumented family members or discrimination.

Reframing these results, why did Mexican-origin youth have more advantaged outcomes in established versus new destinations? Established destinations could have more external assets and resources for math/science identity and math efficacy development, such as teachers and/or mentors of similar backgrounds and/ or access to programs that promote Latino/a involvement in STEM. It may also be that established destinations have attributes that foster positive ethnic identity development among Mexican-origin youth, which can spill over to academic identity domains (Berkel et al., 2010; O’brien et al., 1999; Rivas-Drake et al., 2014).

These findings have several policy implications. For example, policies to improve STEM outcomes among Mexican-origin youth need to be geographically tailored, with particular emphasis on Mexican-origin youth from immigrant families in new destinations. Policies geared toward improving STEM outcomes, such as access to STEM mentors and/or caring teachers (Fast et al., 2010; Lewis et al., 2012), are assets for development. Additionally, this study shows the potential for sharing best practices across destinations. Schools in new destinations, for instance, could learn from schools in established destinations where Mexican-origin youth show positive STEM identity and efficacy development. Finally, across both new and established destinations, science efficacy development was a domain of disadvantage for Mexican-origin youth relative to White youth. More science curriculum, programming, teaching, and mentoring may, therefore, benefit Mexican-origin youth across all types of destinations.

This research was not conducted without limitations. Even though the HSLS includes a large, nationally representative sample, it does not contain comprehensive measures of math/science identity and efficacy, which are multifaceted constructs. A more comprehensive survey instrument would inform if the results from this study are consistent with more precise measures of math/science identity and efficacy among subsamples of Mexican-origin youth in new and established destinations. Additionally, this study focused on math/science, but research should investigate variation in other areas of development in educational and other domains across destinations. Finally, this study took a generalized approach to testing linkages between destinations and math/science identity and efficacy development. Examining specific mechanisms, such as access to mentors and youth programs, will tell us more about these youth.

This study also sets an agenda for future research on differential pathways of development among Mexican-origin youth across geographic contexts. Future research should identify how external assets for development vary across Latino/a destinations. Relatedly, future research could investigate how the interplay of family, school, and community contexts explains differential development outcomes among Latino/a youth across destinations. Finally, examining variation in linkages between ethnic identities and STEM identities across Latino/a communities could be another fruitful area for future inquiry.

As the Mexican-origin youth population grows, researchers must continue to investigate the ecological factors related to assets and risks for their development, including processes leading to attainment in STEM education and occupations. Developmental research can play an important role in understanding variability in STEM pathways among Mexican-origin youth and their peers, given comprehensive research in this area on the domains of identity and efficacy (Bandura, 1989; Merolla et al., 2012; Merolla & Serpe, 2013; Syed, 2010) and a focus on multiple and interconnected ecological systems (Bronfenbrenner & Morris, 2006). This study showed that geographic contextual factors related to immigration history and growth, and the interplay of community contexts and immigrant generation, were important determinants of math/science identity and efficacy development processes among Mexican-origin youth.