A data-based model to predict postsecondary educational attainment of low-socioeconomic-status students.

The National Educational Longitudinal Study database was used to examine the educational development of students of low socioeconomic status (SES). A path analysis was conducted to determine the effects of student background, psychological, and behavioral variables on postsecondary educational attainment of low-SES students. The results show that high school math scores were the most powerful predictor of postsecondary educational attainment, followed by effects of academic expectations, locus of control, reading scores, problem behavior, and classroom behavior. Implications for school counselors are discussed.

**********

School counselors are charged with promoting students' development in three domains--academic, career, and personal/social (Myrick, 2003). The ASCA National Model[R] (American School Counselor Association, 2005) describes professional school counselors' role in supporting the school's overall mission by promoting the academic development, career development, and personal/ social development of all students. Although their importance relative to each other has fluctuated over time, the legitimacy of the three domains for school counselor involvement has remained constant (Paisley & Borders, 1995).

Given the advent of standards-based educational reform, particular attention is currently being given to the domain of academic development. Along with other school personnel, school counselors are being held increasingly accountable for creating school contexts where all students can be academically successful (Dimmitt, 2003). Obviously, school counselors have an important role in ensuring students' academic success. Because they have a school-wide perspective on serving the needs of every single student, including at-risk students (i.e., low-socioeconomic-status [SES] students), school counselors are in an ideal position to serve as advocates for all students, and as agents for removing systemic barriers to academic success (Paisley & Hayes, 2003).

Despite the recent emphasis on students' academic development, practicing school counselors and counselor educators have shown little concern for how school counselors address the academic achievement of students (Lee, Yu, & Lee, 2006). To optimize learning environments and to maximize the potential of at-risk student groups who continue to achieve below the level of other student groups, school counselors must understand which factors contribute to academic success and learn to identify the factors that interfere with academic success. Furthermore, to help ensure successful outcomes for the increasing number of students seeking postsecondary education, it is crucial for school counselors to understand the factors associated with degree completion.

Whether a person attends a postsecondary school (and the type of school he or she attends) has a major impact on individual development, occupational status, and wealth. Enrollment in higher education has expanded dramatically in the United States during the past century. While enrollment in college and universities increased by 66% between 1970 and 1993, the total U.S. population increased by only 27%. As access to higher education has expanded, the importance of postsecondary schooling in determining life changes, occupational status, and wealth has increased (U.S. Department of Education, 1995).

Although overall enrollment in postsecondary institutions has increased, low-SES, academically able students are less likely than are their high-SES counterparts to attend colleges and universities directly after high school (Jordan & Plank, 2000). According to the U.S. Department of Education's National Center for Education Statistics (NCES, 2002a), the percentage of low-SES students who had bachelor's degrees when they were 8 years beyond high school (6.9%) is much smaller than the corresponding percentage for high-SES students (51.0%). Policy makers in the United States have responded to this discrepancy by setting a national education goal to raise the rate of enrollment, retention, and attainment of low-SES minority students in postsecondary education (National Education Goals Panel, 1999).

Because education may be the primary means of social mobility for individuals from disadvantaged groups (Trusty, 2002), researchers and practitioners should devote attention to the educational development of low-SES students. Unfortunately, current understanding of the factors that explain the achievement patterns of low-SES students is limited. Presently, most research focuses on differences between low-SES and middle- and high-SES students; much less attention has been paid to variation among low-SES students, a group of particular interest to educational researchers. Analyses of low-SES student within-group differences are necessary to understand why some students fail, some barely survive, and still others thrive in this low-SES, high-risk environment. What factors make some low-SES students succeed?

The factors exerting long-term influence on low-SES students' success in attaining postsecondary degrees are not well known. Therefore, the purpose of our study was to analyze the longitudinal direct, indirect, and total effects of low-SES students' background, psychological, and behavioral variables on postsecondary degree attainment using a large, nationally representative, longitudinal database (the National Educational Longitudinal Study database).

SOCIOECONOMIC STATUS AND EDUCATIONAL ATTAINMENT

Previous research indicates that socioeconomic status is the best predictor of academic attainment, and that low SES forecasts low attainment (Caldwell & Ginther, 1996; Hobbs, 1990). The NCES (2002a) has estimated that students' likelihood of completing a bachelor's degree or higher increases based on their SES: Seven percent of low-SES students, 24% of middle-SES students, and 51% of high-SES students completed such a degree by 2000. Even among high-achieving students, attainment of a college degree is impacted by SES. For example, among students who scored in the highest mathematics test quartile in eighth grade, the likelihood of earning a bachelor's or higher degree increased with SES: from 29% among those from low-SES families, to 47% among those in the middle two quartiles, to 74% among those with the highest SES.

While low SES is highly correlated with low educational attainment, some low-SES students successfully attain a postsecondary degree. According to data from the NCES's (2002b) Statistical Analysis Report, strong academic preparation and achievement in school increase the likelihood that low-SES students will finish college. Achieving higher test scores was associated with higher rates of college completion, and the association was stronger for low-SES students than for others in this cohort (NCES). Among low-SES students, high achievers on eighth-grade mathematics tests were about 10 times more likely than low achievers to complete a degree by the year 2000. In contrast, among high-SES students, high achievers were only 2.4 times more likely than low achievers to complete a degree.

FACTORS INFLUENCING EDUCATIONAL ATTAINMENT

According to the NCES (2002b), there are different rates of education attainment by gender and race. For example, a slightly higher percentage of females than males completed high school in 2001 (89% compared to 87%). Higher proportions of Whites and Asian Americans than African Americans or Hispanic Americans completed schooling at each level of education. Among those aged 25 to 29, the percentage of White Americans who attained at least a bachelor's degree in 2001 was four times that of Hispanic Americans (36% compared to 9%) and twice that of African Americans (18%). While fewer females than males received at least a bachelor's degree (13% compared to 21%) in 1971, significantly more females than males completed some college (63% compared to 54%) and received at least a bachelor's degree (31% compared to 26%) in 2001. The increase for females in educational attainment over the past 25 years was very substantial--almost 12 percentage points--whereas the increase for males was only 2 percentage points (NCES).

Theoretically, locus of control, self-concept, educational aspirations, and related student psychological factors are useful in explaining educational attainment (Huang, Salvucci, Peng, & Owings, 1996). In various educational situations, for example, a positive self-concept is considered a desirable trait that serves as a mediating variable in facilitating attainment (Marsh, 1990). Several studies in the 1960s and 1970s (Coopersmith, 1967; Purkey, 1970; Wylie, 1979) examined the association between self-concept and educational attainment. These research studies found that the students' positive self-concept is part of the foundation for educational progress. For example, research consistently shows that low self-concept is concomitant with unproductive behavior (Lehr & Harris, 1988; Silvernail, 1986). There is general agreement that the combination of underachievement and low self-esteem is one of the primary reasons for dropping out of school (Brodinsky, 1989). Conversely, healthy self-esteem and successful learning experiences have been touted as critical factors for successful postsecondary educational attainment (Waitley, 1987). Some studies, however, have indicated no significant relationship between self-concept and academic attainment (Stupnisky et al., 2007; van Laar, 2000).

Research generally shows that internal-external locus of control is a powerful variable both in predicting and explaining educational achievement. Rotter (1966) reported significant relationships between higher socioeconomic status and internal locus of control on the one hand and lower socioeconomic status and external locus of control on the other. Several researchers (Parker, 1994; Shepherd, Owen, Fitch, & Marshall, 2006) have suggested that locus of control is an important psychological variable that relates to academic success in schools, with internals tending to attain greater academic success. Research evidence has indicated that internal locus of control also is associated with higher self-motivation, superior academic performance, and greater educational attainment (Nelson & Kristi, 1995; Parker). In addition to earning higher test scores, people with an internal locus of control tend to attribute their success to internal factors rather than to fate, luck, or powerful others (Hass, 1989).

Several studies have shown that students' academic expectation is an important variable both in predicting and explaining educational attainment (Hanson, 1994; Mickelson, 1990; Trusty, 2002). Academic expectation means the level of education that students expect to achieve, while academic aspiration means the education they hope to achieve (Hanson). Some students may have the aspiration to earn a college degree, yet they may not expect to earn that degree. Academic expectations are a more concrete variable than academic aspirations for predicting and explaining educational attainment (Trusty).

According to previous research, one of the most important predictors of postsecondary educational attainment is high school academic performance (Anderson & Keith, 1997; Parkerson, Lomax, Schiller, & Walberg, 1984). Academic performance can be measured through various indicators. The most widely used indicator is a standardized test score, deemed an appropriate measure of academic ability because it measures differences in terms of cognitive skills and abilities.

Several studies report that students' school behavior is also an important variable in predicting educational attainment (Anderson & Keith, 1997; Trusty, 2000; Wong, 1990). A number of research studies have indicated that students who show problem behavior score lower on academic achievement tests than those who do not. More recently, Trusty (2002) reported on the effects of two high school behavior variables--high school involvement and behavior problems. He indicated that the high school students' involvement directly influenced students' educational expectations. Students' problem behavior also was significantly negatively related to educational expectation. These results are consistent with previous research (Connell, Spencer, & Aber, 1994; Ogbu, 1991). Additionally, further research on the relationship between low-SES students' school behavior and educational attainment should be conducted to determine which student school behaviors need to be considered and to help in assessing the effectiveness of school-based prevention and intervention efforts aimed at children who are at risk.

Thus, the purpose of this study was to test a conceptual model of low-SES students' educational attainment 8 years post--high school. Specifically, the study focused on an examination of the direct and indirect longitudinal effects of student demographic characteristics (race and gender), psychological characteristics (locus of control, self-concept, and student academic expectation), and behavioral characteristics (math score, reading score, student homework behaviors, student classroom behaviors, and student problem behaviors) on the postsecondary educational attainment of low-SES students.

METHOD

Participants and Data Set

Participants totaled 2,460 low-SES students from the National Educational Longitudinal Survey (NELS:88) base year through the fourth follow-up panel sample. For the present study, participants consisted of students in the panel sample who were at or below the 25th percentile in SES in the 1988 survey and were resurveyed in the 2000 survey. The SES variable was based on five equally weighted, standardized components: father's education, mother's education, family income, father's occupation, and mother's occupation (NCES, 2002b).

This study utilized data from three waves of NELS:88 data collection: (a) the Base-Year Student Questionnaire, which was administered when students were 8th graders; (b) the First Follow-Up Student Questionnaire, which was administered when students were 10th graders; and (c) the Fourth Follow-Up Student Questionnaire, administered when most participants were 8 years beyond high school. Independent variables are from Base-Year and First Follow-Up waves of data collection. The dependent variable is from the Fourth Follow-Up (NCES, 2002b).

NELS:88 has a two-stage clustered sampling design. That is, U.S. schools were first sampled; then students within schools were sampled (see Ingels et al., 1994). Therefore, in NELS:88 samples, standard errors are spuriously small (Trusty, 2002). A number of statistical packages take into account complex sampling designs in the calculation of standard errors. For this study, the AM software was used for variance estimation. Root design effects (DEFTs) were employed with path analysis. The DEFTs in this study ranged from 1.16 to 2.14, with a mean DEFT of 1.73.

Forty-four percent of the participants were male and 56% were female. Participants were 53% White, non-Hispanic; 15% Black, non-Hispanic; 4% Asian or Pacific Islander; 1.5% American Indian or Alaska Native; 25% Hispanic or Latino (any race); and 1.5% more than one race. Approximately 39% were from the Southern United States; about 15% were from the Northeast; 26% were from the North Central region; and about 20% were from the Western United States. At the final follow-up, approximately 67% of these low-SES participants did not have any postsecondary degree; 21% had received a certificate or license; about 11% had received an associate's degree; 11% had received a bachelor's degree; and roughly 1% had received a master's or higher degree.

Research Variables

All variables in this study were measured by a questionnaire developed by the National Opinion Research Center. The main variable, educational attainment, was derived from a set of questions (i.e., "What type of high school qualification did you receive?" and "What is the highest level of education you completed?"). Originally, participants indicated the various types of postsecondary degrees or certificates they had attained. For the purpose of simplicity in this study, four final categories were generated for this variable: (a) had no postsecondary experience, (b) had some postsecondary education but did not attain a bachelor's degree, (c) attained a bachelor's degree, and (d) attained a master's degree or higher. This information was collected in 2000, which was 8 years after high school graduation for most respondents. This educational attainment variable was rescaled (see "Data Analysis" section) to better represent the true differences between the ordinal scale degrees.

Student demographic variables included gender (male and female) and race (African American, Latino, White, American Indian, and Asian American). Race was coded as a dichotomous variable, with students who checked "White" or "Asian" race/ethnicity on the base-year survey coded into one group and those who checked "Black," "Hispanic," and "Native American" coded into another group. White and Asian students were grouped together because these two ethnic groups are not generally associated with educational disadvantage even though Asian students represent a minority in American schools (Anderson & Keith, 1997).

The three student psychological variables were locus of control, self-concept, and academic expectation during high school. These three variables were measured on the NELS:88 Base-Year Student Questionnaire, administered when the participants were eighth graders. These psychological variables were selected based on significant findings in the educational development literature (Cassidy & Eachus, 2000; Zsolnai, 2002). Participants answered a total of six questions related to locus of control (e.g., "Every time I try to get ahead, somebody or something stops me"). The format was a 4-point Likert-type scale ranging from "strongly agree" to "strongly disagree." The alpha coefficient for the locus-of-control scale was 0.68, which suggests a moderate degree of consistency across items.

Self-concept consists of the strength of a student's sense of self-worth and overall self-esteem. Participants responded to a total of seven items related to self-concept (e.g., "I feel good about myself"). The NELS:88 self-concept variable is inferred from the positively worded items that were adapted from the Rosenberg (1979) scale for inclusion in the NELS:88 database (National Education Longitudinal Study, 2002). Like the locus-of-control variable, the format was a 4-point Likert-type scale ranging from "strongly agree" to "strongly disagree." The alpha coefficient achieved was 0.78, which suggests an acceptable degree of consistency across items.

Academic expectation is the perception a student holds about his or her future academic achievement. Participants responded to the question "What is the highest level of education you ever expected to complete?" by selecting one of the following categories: (a) less than high school, (b) high school diploma or equivalent, (c) vocational school, (d) college/associate's degree, (e) bachelor's degree, or (f) master's degree or higher. The ordering of categories was based on academic intensity, with expectations for college achievement ranked higher than vocational school achievement. Like the dependent variable, this variable also was rescaled (see Data Analysis section) to better represent the true differences between the ordinal scale degrees using the categorical regression method.

Student behavioral variables include academic performance and school behavior. Most research abiding by the status attainment model has used a one-dimensional measure of early academic performance (Eccles, 1993; Trusty, Robinson, Plata, & Ng, 2000). However, Trusty (2002) reported that for high school students, mathematics test scores have more influence on males' educational development, while language-related test scores have a greater influence on females' educational development. Therefore, for purposes of this study, both mathematics and reading scores were used as indicators of academic performance. While in the 10th grade, all students completed both mathematics and reading tests--standardized measures administered in NELS:88 (Curtin, Ingels, Wu, & Heuer, 2002).

High school behavior was added to the present model, which is consistent with the models used by Connell et al. (1994) and Trusty (2002). Several items from the First Follow-Up Student Questionnaire assessed in-school student behaviors. The researchers developed these variables through factor analysis with varimax rotation. This procedure is used to select appropriate variables of study already present in the NELS:88 data set. From this analysis, three factors/variables were selected. The first factor indicates students' class preparedness and has a total of three items (e.g., "How often do you go to class without pencil/paper?"). The format was a 4-point Likert-type scale ranging from "usually" to "never." The Cronbach's alpha for these three items was .69. The second factor/variable was student homework behavior in school. This factor also is composed of three items (e.g., "How much time do you spend on homework in school each week?"). The Cronbach's alpha for these three items was .72. The third factor/variable was student problem behavior in school. This factor is composed of three items (e.g., "How many times were you late for school in the first half of the current school year?"). The Cronbach's alpha for these three items was .73.

Data Analysis

To examine the factors influencing educational attainment, categorical regression (also known as an optimal scaling procedure) was used to analyze data, as the data were unsuitable for "standard" statistical procedures. The variables in the present study were inappropriate for analysis that involves a linear relationship due to the ordinal nature of several of the variables (e.g., educational attainment and academic expectation). Based on how variables are defined, Trusty (2002) stated that categorical regression analysis helps to rescale (i.e., reformulate) variables to maximize the strength of the relationships between a set of independent variables and a dependent variable. After relationships of the variables were rescaled and linearized, using the AMOS program, a path analysis model of educational attainment was performed. Path coefficients for the model were calculated using a maximum likelihood estimation procedure. Residual values specifying the difference between the observed and expected values for the path coefficients, as well as for several goodness-of-fit omnibus measures, were produced. Modifications of the results were made when the analysis was inconsistent with the original model structure.

RESULTS

Categorical Regression

A categorical regression method was used for linearizing relationships in a regression equation using an optimal scaling approach. The initial categorical regression equation for low-SES students indicated that all variables in the present study were linear or nondecreasing curvilinear. Because no variables exhibited quadratic tendencies, variables were defined as ordinal or interval in the following equations. Only small changes in the rescaled values across the first two levels of the original levels (no degree and associate's degree) were noted. However, there were large differences across the original values 2, 3, and 4 (associate's degree, bachelor's degree, and master's degree or above). These results indicate that the predicting variables (i.e., gender, race, locus of control, self-concept, academic expectation, academic performance, and school behavior), taken together, differentiated among an associate's degree, a bachelor's degree, and a master's degree or higher more effectively than they differentiated between no postsecondary degree and an associate's degree.

The transformation for the student academic expectation variable was very similar to the transformation for the educational attainment variable. There were small changes in the rescaled values across the first three levels of the original values (less than high school, high school diploma or equivalent, and vocational school). However, there were large differences across the original values 4, 5, and 6 (associate's degree, bachelor's degree, and master's degree or above). For the three student behavior variables, relationships were curvilinear. Changes for students with extremely low to low levels of behavior problems and high to extremely high levels of behavior problems showed a stronger impact on the educational attainment variable than did changes in students with middle levels of problem behaviors. The latter had very little effect on the educational attainment variable.

Path Analysis

The direct, indirect, and total effects on educational attainment for low-SES students are presented in Table 1. Error terms were adjusted for design effects. As noted in the table, several variables exerted a significant total effect on the educational attainment of low-SES students. The total effect of students' 10th-grade math scores was by far the most powerful. The path coefficient ([beta] = .248, p = .001) of the math score variable indicates that the students who scored high on the math tests were consistently more likely to attain a higher degree later on. The second strongest influence on educational attainment was the student academic expectation variable ([beta] = .215, p = .001). These results indicate that low-SES students who expected a higher degree when they were in eighth grade were more likely to attain a higher degree in later years. Locus of control provided the next best predictor of educational attainment. The path coefficient was significant ([beta] = .110, p = .005) and indicates that students with a higher internal locus of control tended to have higher educational attainment.

The effect of the reading score ([beta] = .100, p = .008) also was statistically significant. The effect of gender ([beta] = .063, p = .022) was small yet significant and indicated that girls tended to attain higher degrees than did boys. Finally, student problem behavior ([beta] = -.068, p = .018) and student classroom behavior ([beta] = .067, p = .019) had small but significant effects on educational attainment. While student problem behavior was negatively related to the educational attainment variable, student classroom behavior was positively related to the educational attainment variable.

As shown in Figure 1, the total effect of student academic expectation consisted of its direct effect on educational attainment ([beta] = 143, p = .003) plus its indirect influence on educational attainment as mediated by the math score, the reading score, and student problem behavior ([beta] = .072, p = .016). While race did not exert a significant influence on educational attainment directly, the race variable exerted a significant negative indirect effect ([beta] = -.066, p = .020) on educational attainment through the math score and the student problem behavior variable, indicating that the majority group (White and Asian students) tended to have higher scores in math and fewer problem behaviors than did the minority group (Hispanic, African American, and Native American students).

One important goal of research when using a path analysis procedure is achieving parsimony, or deriving a model that explains the most variance with the fewest indicators (Maruyama, 1997). This model was constructed with all statistically nonsignificant paths at a probability level of less than .05 removed (see Figure 1). The revised model for low-SES students excluded two variables (self-concept and student homework behavior) and several paths between variables that did not have significant effects on the educational attainment variable.

With regard to the fit of this model, a significant chi-square, [chi square](11) = 163.5, p = .015, indicated a significant difference between input and implied matrices. That is, this model was not shown to fit the data well according to the chi-square test result. It is often found, however, that with large sample sizes, even small differences between input and implied covariance matrices can lead to the rejection of the null hypothesis (Cool & Keith, 1991). Therefore, other indexes of fit also were examined to evaluate the model. These include the goodness of fit (GFI) and the adjusted goodness of fit (AGFI) indexes as well as the root mean square residual (RMSK). The GFI index (.986) and the AGFI (.962), being close to 1 and interpreted like correlation coefficients, represent a good fit of the data to the model. Likewise, the RMSR of .023, being low and representing the average discrepancy between elements of the input and the implied covariance, represents a good fit. Regarding of effect size, all variables together accounted for approximately 20% of the variability in educational attainment.

[FIGURE 1 OMITTED]

DISCUSSION

The purpose of this study was to examine the long-term academic development of low-SES students using the National Educational Longitudinal Study database. The results of the current study indicated that the influence of gender on educational attainment was small yet significant, with girls tending to attain higher degrees than boys. This finding raises a pertinent issue. The coefficient from gender to educational attainment, in the present study, is (a) higher than the path coefficient calculated in some studies, and (b) significant, where in other studies it was not. For example, Anderson and Keith (1997) reported a path coefficient from gender of .003 for at-risk students who were of non-Asian minority race or ethnicity or whose family SES fell within the lowest quartile. Based on this finding, Anderson and Keith concluded that gender had little effect on educational attainment. It should be noted that Anderson and Keith's study used a 1980s database, the NCES's (1983) High School and Beyond Longitudinal Study, while the current study used a 1990-2000 database (NELS:88).

The path coefficient from gender to math score also was significant, indicating that boys tended to achieve a higher math score on standardized tests than did girls. Anderson and Keith (1997) reported that gender did influence the achievement of at-risk students in math and science (boys scored higher) and in reading score (girls scored higher). Trusty (2002) also found that gender influences on academic performance are domain specific, with language-related skills being stronger for women and mathematics skills being more important for men. However, unlike previous research findings (Anderson & Keith; Trusty) with different sample populations, gender did not exert any effect on the reading score in the present study. The inconsistency in findings regarding the effect of gender on academic performance underscores the need for more research using different measures of academic performance in addition to standardized tests.

The students' gender in this study was significantly related to classroom preparedness behavior, with female students tending to be more prepared for class than male students. Furthermore, student classroom preparedness influenced educational attainment in later years. The finding of the present study is partially consistent with other studies (Etxebarria, Apodaca, Eceiza, Fuentes, & Ortiz, 2003). Etxebarria et al. found that girls are more pro-social and less aggressive with peers and more cooperative with teachers than are boys. In addition, over time inhibitory control increased in girls and decreased in boys, whereas aggressive behavior increased in boys and decreased in girls (Etxebarria et al.). This finding draws attention to the need for more research regarding the effect of gender on low-SES students' classroom preparedness behavior.

The results of this study indicate that the race variable did not exert any direct effect on educational attainment. There are tentative explanations for the absence of a relationship between race and educational attainment. First, this finding could reflect the recent trend that the effect of race on educational attainment has decreased over time. Using a 1980s database, Anderson and Keith (1997) found that race did exert a direct as well as indirect effect on low-SES students' educational achievement. However, according to the NCES (2002b), non-Asian minorities over the past 30 years have made significant gains in high school completion and have substantially reduced the majority-minority difference for degree completion, with 82% of majority group students and 52% of non-Asian minority group students having completed high school in 1971 and 93% and 85%, respectively, having completed high school in 2001. Second, it is likely that the differences identified between majority students and minority students in educational attainment could be due to socioeconomic differences rather than to the unique characteristics of a given racial group. Therefore, it could be argued that the race factor in the model is likely not to show much effect on educational attainment due to the homogeneity of the sample (all low-SES students).

Even though there was no direct effect of race on educational attainment, the race variable exerted a significant negative indirect effect on educational attainment through the math score, the reading score, and the student problem behavior variable. Minority racial status was negatively related to 10th-grade math scores and reading scores. Race also significantly related to student problem behavior. The data indicated that majority group (White and Asian) students had higher scores in math and reading tests and had fewer problem behaviors than did the minority groups (Hispanic, African American, and American Indians). In turn, it was found that higher math and reading scores and fewer problem behaviors tended to result in greater postsecondary educational attainment 8 years post-high school.

An extensive empirical and theoretical body of literature documents the relationships between students' locus of control and their academic attainment (Ford, 1994; Weiner, 1986). Previous research generally shows that internal-external locus of control is a powerful variable both in predicting and explaining educational achievement (Shepherd et al., 2006). Consistent with the literature, in our sample, locus of control had a direct effect on postsecondary educational attainment, with students who had an internal locus of control in the 8th grade achieving higher postsecondary educational attainment. Internal locus of control in 8th grade also was directly related to math scores and reading scores in 10th grade. This finding is also consistent with earlier research (Hass, 1989; Nelson & Kristi, 1995). Consequently, it can be surmised that regardless of socioeconomic status, locus of control is an important variable in later academic degree attainment.

The data from this study indicated that self-concept at Grade 8 was not directly or indirectly related to later educational attainment. While some studies have indicated no significant relationship between self-concept and academic attainment (Stupnisky et al., 2007; van Laar, 2000), others (Brodinsky, 1989; Waitley, 1987) have reported that healthy self-esteem and successful learning experiences are critical factors for successful postsecondary educational attainment (Lee et al., 2006). The finding of the present study is consistent with Collins' (1992) study. Collins found that unlike other groups, the low-SES students group had a high self-concept that was relatively independent of academic achievement.

Significant direct effects as well as indirect effects on educational attainment were found from the student academic expectation variable. In other words, low-SES students who expected a higher level of educational degree when they were in 8th grade were more likely to attain a higher degree in later years. This finding is consistent with Hanson's (1994) model. Hanson proposed that academic expectations are a more concrete variable than academic aspirations for predicting and explaining educational attainment. Several other researchers also found that self-perception variables, such as educational expectation, are related to academic achievement (Trusty, 2002). Data from this study also revealed that students' academic expectations in 8th grade were significantly related to math scores and reading scores in 10th grade. This contradicts some previous research indicating that low-SES students have positive attitudes and beliefs about education but low academic performance (Trusty, 2000). In contrast, the low-SES students in this sample who had high early educational expectations also had relatively high educational attainment in later years.

The result of the present study indicated that low-SES students who expected a higher degree were less likely to act out problem behaviors in the school. The significant effect of student academic expectations on problem behavior is extremely important, as it suggests that students' expectations about their future have a unique impact on their actual behavior and performance above and beyond the impact of the school environment, students' prior levels of achievement, students' SES, and their gender.

Academic performance in 10th grade was the strongest predictor of later educational attainment for the low-SES students in this study. That is, educational attainment was significantly affected by the low-SES students' skills and efforts. This finding is consistent with previous research. For example, the zero-order correlations between academic performance (math and reading scores) and educational attainment for low-SES students found in this study were almost the same as those found by Trusty (2004), using high school students in general.

Among the three categories of school behaviors, student problem behavior and student classroom behavior had a small but significant influence on educational attainment. In other words, high school students' classroom preparedness behaviors were positively related to later educational attainment, and students' problem behavior was significantly negatively related to educational attainment. This is consistent with previous research that found that students' school behavior is an important variable in predicting educational attainment (Anderson & Keith, 1997; Trusty, 2002).

Implications for School Counselors

The results of the present study present several implications for school counselors working with low-SES students. It seems that patterns of educational attainment (and conversely of low performance) are determined as early as middle school for low-SES students. Because early academic performance is important for low-SES students' long-term educational development, school personnel could assist low-SES students in their long-term development by helping them acquire academic skills during early childhood. Federal legislation such as the No Child Left Behind Act also places emphasis on and funds programs that will provide resources to help low-SES students achieve a state-defined standard of proficiency. Therefore, counselors need to design and deliver a classroom guidance curriculum that potentially impacts students' achievement earlier in their academic development (e.g., upper grades of elementary school). In addition, counselors need to systematically collect critical data on student placements and academic success and failure of all students (Lee et al., 2006) to allow for early intervention for students who are struggling academically.

Low-SES children frequently live in highly stressful and disadvantaged school environments characterized by multiple risk factors for the development of psychosocial adjustment problems, including high rates of poverty, exposure to violence, and drug use (Jones & Forehand, 2003). Student school behavior seems to be a suitable goal, or entry point, for educational support and intervention for low-SES students (Trusty, 2002). Therefore, further research on the relationship between low-SES students' school behavior and educational attainment should be examined to determine which low-SES students' school behaviors need to be considered and to help in assessing the effectiveness of school-based prevention and intervention efforts aimed at children who are at risk. For example, mentoring programs in which school personnel provide social support for low-SES students have demonstrated that mentoring can enhance positive behavior changes of students and increase positive attitudes toward school of most participants in such programs (Richman, Rosenfield, & Bowen, 1998). In addition, counselors should use these contexts for conceptualizing students and designing appropriate interventions and programming to help students become more engaged in school activity (Trusty, 2000).

Limitations and Suggestions for Future Studies

The present study overcomes many of the limitations of previous research in that it was conducted with a longitudinal design and a large, nationally representative sample. However, the large sample could be one of the limitations of the current study. As stated earlier, the chi-square is biased by large sample sizes. This probably resulted in finding some differences between groups that were statistically different but not meaningfully different. In addition, the total variance explained regarding postsecondary educational attainment of low-SES students in this study is relatively small (20%). Further studies are needed to explain the rest of the variance (80%) of educational attainment. One of the possible factors could be peer influence in high school. Viadero (2003) studied 141,000 students in San Diego public schools over 3 years and teased out some of the factors that were keys to their academic success. According to Viadero, students' peer factors had more consistent and higher effects on academic success than did parent and teacher factors.

The results of this study also provided evidence as to the nature and extent of the influence of several factors cited previously as being important to low-SES students' educational attainment. Several issues raised by this study warrant further investigation. Additional empirical research is needed to investigate whether degree of educational attainment is affected by factors existing prior to Grade 8 for low-SES students. This research would examine the influence on educational attainment of elementary grades variables. Further research also is needed to investigate later adolescent age factors (i.e., college factors) that influence degree attainment.

While the model presented here was designed specifically to address low-SES students as a group, it will be important to determine whether the effects noted here also apply to subgroups of the low-SES populations. Previous research indicated that influences of academic performance are domain specific, with language-related skills being more important for females and mathematics skills being more important for males (Trusty, 2002). Therefore, similar models may be developed and examined for subgroups (race and gender) to determine whether there are differences in factors that affect educational attainment for different groups.

CONCLUSIONS

To date, most research has focused on differences in educational attainment between low-SES and middle- and high-SES students; much less attention has been paid to variation among individuals within the low-SES group. This study focused on analyses of low-SES students' within-group differences to understand why some of these students fail, some barely survive, and still others thrive in the high-risk environment. The current research results provide strong evidence that postsecondary educational attainment of low-SES students can be facilitated by greater internal locus of control and higher student academic expectations during early adolescence. These two psychological variables were associated with greater academic performance (i.e., math and reading scores), greater student classroom preparedness behavior, and fewer student problem behaviors during early adolescence that, in turn, were associated with greater postsecondary educational attainment. Student demographic variables also influenced postsecondary educational attainment directly (by gender) as well as indirectly (by race).

In conclusion, it is vital to identify factors influencing the educational attainment of low-SES students, and the present research is an effort in that direction. Investigating these sources of variation, school counselors and counselor educators could be in a better position to provide appropriate educational interventions for low-SES students and, eventually, to modify guidance service delivery models toward enhancing the learning and development of low-SES students as well as all students at every educational level.

References

American School Counselor Association. (2005). The ASCA national model: A framework for school counseling programs (2nd ed.). Alexandria, VA: Author.

Anderson, E. S., & Keith, T. Z. (1997). A longitudinal test of a model of academic success for at-risk high school students. Journal of Educational Research, 90, 259-268.

Brodinsky, B. (1989). Students at risk. Arlington, VA: American Association of School Administrators.

Caldwell, G. P., & Ginther, D. W. (1996). Differences in learning styles of low socioeconomic status for low and high achievers. Education, 117, 141-147.

Cassidy, S., & Eachus, P. (2000). Learning style, academic belief systems, self-report student proficiency and academic achievement in higher education. Educational Psychology, 20, 307-320.

Collins, M. A. (1992). School social context, self-esteem, and locus of control among White, Black, and Hispanic youths. Unpublished doctoral dissertation, University of Maryland, College Park.

Connell, J.P., Spencer, M. B., & Aber, J. L. (1994). Educational risk and resilience in African-American youth: Context, self, action, and outcomes in school. Child Development, 65, 493-506.

Cool, V., & Keith, T. Z. (1991). Testing a model of school learning: Direct and indirect effects on academic achievement. Contemporary Educational Psychology, 16, 28-44.

Coopersmith, S. (1967). The antecedents of self-esteem. San Francisco: Freeman.

Curtin, T. R., Ingels, S. J., Wu, S., & Heuer, R. (2002). User's manual: NELS:88 base-year to fourth follow-up: Student component data file (Publication No. NCES 2002-323).Washington, DC: U.S. Department of Education, Office of Educational Research and improvement.

Dimmitt, C. (2003). Transforming school counseling practice through collaboration and the use of data: A study of academic failure in high school. Professional School Counseling, 6, 340-350.

Eccles, J. (1993). School and family effects on the ontogeny of children's interests, self-perceptions, and activity choices. In J. Jacobs (Ed.), Nebraska symposium on motivation: Developmental perspectives on motivation (pp. 145-208). Lincoln, NE: University of Nebraska Press.

Etxebarria, I., Apodaca, P., Eceiza, A., Fuentes, M. J., & Ortiz, Y. M. (2003). Gender differences in emotions and social behavior at school age. Infancia y Aprendizaje, 26, 147-162.

Ford, D.Y. (1994). Nurturing resilience in gifted Black youth. Roeper Review, 17, 80-85.

Hanson, S. L. (1994). Lost talent: Unrealized educational aspirations and expectations among U.S. youths. Sociology of Education, 67, 159-183.

Hass, M. H. (1989). Relationship between locus of control and self-efficacy in freshman probation and non-probation college students. Unpublished master's thesis, University of Utah, Salt Lake City.

Hobbs, D. (1990). School based community development: Making connections for improved learning. In S. Raferty & D. Mulkey (Eds.), The role of rural schools in community development (pp. 57-64). Mississippi State, MS: Southern Rural Development Center.

Huang, G., Salvucci, S., Peng, S., & Owings, J. (1996). National Education Longitudinal Study of 1988: Research framework and issues. Washington, DC: National Center for Education Statistics.

Ingels, S. J., Dowd, K. L., Baldridge, J. D., Stripe, J. L., Bartot, V. H., & Frankel, M. R. (1994). Second follow-up: Student component data file user's manual (Publication No. NCES 94-374).Washington, DC: U.S. Department of Education, Office of Educational Research and Improvement.

Jones, D. J., & Forehand, G. (2003). The stability of child problem behaviors: A longitudinal analysis of inner-city African American children. Journal of Child and Family Studies, 12, 215-227.

Jordan, W. J., & Plank, S. B. (2000). Talent loss among high-achieving poor students. In M. Sanders (Ed.), Schooling students placed at risk: Research, policy, and practice in the education of poor and minority adolescents (pp. 65-82). Mahwah, N J: Lawrence Erlbaum.

Lehr, J. B., & Harris, H.W. (1988).At-risk, low-achieving students in the classroom. Washington, DC: National Education Association.

Lee, S. M., Yu, K. L., & Lee, S. H. (2006). Helping academic achievement of low-SES students: The role of school counselors in Korea. Asian Journal of Counselling, 12, 47-77.

Marsh, H.W. (1990). Causal ordering of academic self-concept and academic achievement: A multiwave, longitudinal panel analysis. Journal of Educational Psychology, 82, 646-656.

Maruyama, G. M. (1997). Basics of structural equation modeling. Thousand Oaks, CA: Sage.

Mickelson, R. A. (1990). The attitude-achievement paradox among Black adolescents. Sociology of Education, 63, 44-61.

Myrick, R. D. (2003). Developmental guidance and counseling: A practical approach (4th ed.). Minneapolis, MN: Educational Media.

National Center for Education Statistics. (1983). High School and Beyond 1980 sophomore cohort first follow-up (1982): Data file user's manual. Washington, DC: U.S. Government Printing Office.

National Center for Education Statistics. (2002a). Coming of age in the 1990s: The eighth-grade class of 1988 12 years later. Washington, DC: U.S. Government Printing Office.

National Center for Education Statistics. (2002b). Statistical Analysis Report. User's manual: NELS:88 Base-year to fourth follow-up: Student component data file. Washington, DC: U.S. Government Printing Office.

National Education Goals Panel. (1999). The National Education Goals report: Building a nation of learners, 1999. Washington, DC: U.S. Government Printing Office.

National Education Longitudinal Study. (2002). National Education Longitudinal Study: 1988-2000: Dam files and electronic codebook system [CD-ROM data files and documentation].Washington, DC: National Center for Education Statistics.

Nelson, E. S., & Kristi, E. M. (1995). The relationships among college students' locus of control, learning styles, and self-prediction of grades. Education Research and Perspectives, 22, 110- 117.

Ogbu, J. U. (1991). Minority coping responses and school experience. Journal of Psychohistory, 18, 433-456.

Paisley, R 0., & Borders, L. D. (1995). School counseling: Evolving specialty. Journal of Counseling and Development, 74, 150-154.

Paisley, P. O., & Hayes, R. L. (2003). School counseling in the academic domain: Transformations in preparation and practice. Professional School Counseling, 6, 198-204.

Parker, A. (1994). Locus of control, demographics and mode of delivery as predictors of dropout from distance education. Unpublished doctoral dissertation, Arizona State University, Tempe.

Parkerson, J. A., Lomax, R. G., Schiller, D. P., & Walberg, H. J. (1984). Exploring causal models of educational achievement. Journal of Educational Psychology, 76, 638-646.

Purkey, W. W. (1970). Self-concept and school achievement. Englewood Cliffs, N J: Prentice-Hall.

Richman, J. M., Rosenfield, L. B., & Bowen, G. L. (1998). Social support for adolescents at risk of school failure. Social Work, 43, 309-323.

Rosenberg, M. (1979). Conceiving the self. New York: Basic Books.

Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs, 80, 1-28.

Shepherd, S., Owen, D., Fitch, T. J., & Marshall, J. L. (2006). Locus of control and academic achievement in high school students. Psychological Report, 98, 318-322.

Silvernail, D. L. (1986). Teaching styles as related to student achievement. Washington, DC: National Education Association.

Stupnisky, R. H., Renaud, R. D., Perry, R. R, Ruthig, J. C., Haynes, T. L., & Clifton, R. A. (2007). Comparing self-esteem and perceived control as predictors of first-year college students' academic achievement. Social Psychology of Education, 10, 303-330.

Trusty, J. (2000). High educational expectations and low achievement: Stability of educational goals across adolescence. Journal of Educational Research, 93, 356-365.

Trusty, J. (2002). African Americans' educational expectations: Longitudinal causal models for women and men. Journal of Counseling and Development, 80, 332-345.

Trusty, J. (2004). Effects of students' middle-school and high-school experiences on completion of the bachelor's degree (Research Monograph No. 1). Amherst, MA: University of Massachusetts, Center for School Counseling Outcome Research.

Trusty, J., Robinson, C., Plata, M., & Ng, K. (2000). Effects of gender, socioeconomic status, and early academic performance on postsecondary educational choice. Journal of Counseling & Development, 78, 463-472.

U.S. Department of Education. (1995). Digest of educational statistics, 1995. Washington, DC: U.S. Government Printing Office.

van Laar, C. (2000). The paradox of low academic achievement but high self-esteem in African American students: An attributional account. Educational Psychology Review, 12, 33-61.

Viadero, D. (2003). Study evaluates peers' effect on achievement. Education Week, 23, 10-12.

Waitley, D. (1987). Self-esteem: The transferable skill. Cedar Falls, IA: Advance Learning.

Weiner, B. (1986). An attributional theory of motivation and emotion. New York: Springer-Verlag.

Wong, M. G. (1990). The education of White, Chinese, Filipino, and Japanese students: A look at high school and beyond. Sociological Perspectives, 33, 355- 374.

Wylie, R. (1979). The self-concept: Theory and research on selected topics. Lincoln, NE: University of Nebraska.

Zsolnai, A. (2002). Relationship between children's social competence, learning motivation and school achievement. Educational Psychology, 22, 317-333.

Sang Min Lee, Ph.D., is an assistant professor in the Department of Education, Korea University, Seoul, South Korea. E-mail: [e-mail]

M. Harry Daniels, Ph.D., is a professor and Ana Puig, Ph.D., is a research director in the Department of Counselor Education, University of Florida, Gainesville.

Rebecca A. Newgent, Ph.D., is an associate professor in the Counselor Education Program, University of Arkansas, Fayetteville.

Suk Kyung Nam is a doctoral student in the Department of Education, Korea University.

Portions of this article are based on a doctoral dissertation by Sang Min Lee completed in August 2004 at the University of Florida. This research is based on work supported by the Association for Institutional Research, the National Center for Education Statistics, and the National Science Foundation, under Association for Institutional Research Grant 03-305 and National Science Foundation Grant SRS-0086139.

Table 1. Direct, Indirect, and Total Effect on Educational
Attainment in Models for Low-SES Students

                                Direct     Indirect    Total
Factor                          Effect      Effect    Effect

Race                             .012       -.066 *    -.055
Gender                           .066 *     -.003       .063
Locus of control                 .069 *      .040       .110
Self-concept                     .035       -.004       .031
Student academic expectation     .143 **     .072 *     .215
Math score                       .248 **                .248
Reading score                    .100 **                .100
Student homework behavior       -.045                  -.045
Student problem behavior        -.068 *                -.068
Student classroom behavior       .067 *                 .067

* p<.05. ** p<.01.