A Look Beyond Cognitive Predictors of Academic Success:Understanding the Relationship Between Academic Self-Beliefs and Outcomes
The purpose of this study was to examine the relationship between academic self-beliefs (i.e., self-efficacy and degree aspirations) with various academic outcomes. Based on previous findings, it was hypothesized that students with more positive academic self-beliefs would perform better in school. The results supported prior research as students with higher academic self-beliefs also had higher SAT scores, grades, and second-year retention rates. Students with more negative writing and math self-efficacy beliefs were more likely to state that they would desire help with improving those skills. Suggestions for those in college counseling positions to intervene and provide assistance are discussed.
A recurrent trend in higher education research has been to identify additional predictors of college success beyond the traditional measures of high school grade point average (HSGPA) and standardized scores (e.g., Camara & Kimmel, 2005; Willingham & Breland, 1982). A major impetus for this research is due to the resulting performance disparity among ethnic groups on these traditional measures. Most pointedly, it has been shown that African American students score, on average, one standard deviation below White students on cognitive measures (Sackett, Schimitt, Ellingson, & Kabin, 2001). As such, much research has been devoted to developing new measures that are valid but also result in smaller subgroup differences (e.g., Oswald, Schmitt, Kim, Ramsay, & Gillespie, 2004).
Additionally, it has been recognized that, even though HSGPA and standardized test scores are strongly related to college success, there is still additional variance to be explained, especially depending on how college success is operationally defined. For example, a meta-analysis on the predictive validity of the SAT by Hezlett et al. (2001) found multiple correlations ranging from 0.44 to 0.62 between the verbal and mathematics sections of the SAT and first-year GPA (FYGPA). Results are similar for the relationship between HSGPA and FYGPA, with corrected validity coefficients in the 0.50s. The multiple correlation of HSGPA and SAT scores with FYGPA is usually in the 0.60s (e.g., Camara & Echternacht, 2000). Despite the strong link between both SAT scores and HSGPA with FYGPA, over 60% of the variance still remains unexplained. This percentage becomes even larger when examining other indicators of college success such as cumulative GPA, persistence, and graduation rate (e.g., Hezlett et al.; Robbins et al., 2004). Both issues—score disparities by subgroup and the desire to improve the prediction of college success—have led to an increased focus on the identification of additional predictors, specifically noncognitive measures, of college success.
Most notably, self-efficacy beliefs can often explain approximately one quarter of the variance in the prediction of academic performances (Pajares, 2006). According to social cognitive theory advanced by Bandura [End Page 665] (1986, 1997), human achievement depends on interactions between an individual's behaviors, personal factors, and environmental conditions. Individuals hold self-efficacy beliefs that enable them to exert control over their thoughts, feelings, and actions. These beliefs affect cognitive, motivational, affective, and decisional processes and determine whether individuals will view themselves as capable or incapable, whether or not they are motivated to persevere in the face of hardships and barriers, their emotional well-being, as well as the choices they make at crucial points in time (Bandura, 1997; Bandura & Locke, 2003). Additionally, self-efficacy beliefs are context-specific, or linked to a particular domain (Klassen, 2002; Zimmerman, 1995).
These beliefs are largely formed through mastery experiences, vicarious experiences, verbal messages and social persuasions, and interpretations of physiological and emotional states (Bandura, 1995). For most students, past performance is the most reliable guide for gauging self-efficacy (Schunk & Meece, 2006). Successful experiences will generally raise self-efficacy, whereas failures will generally lower self-efficacy. However, an intermittent failure among many successful experiences is not likely to significantly alter a person's self-efficacy beliefs. Self-efficacy beliefs can also be formed by watching the performance of people who one believes to be similar to oneself. Vicarious experiences, however, are not as powerful as personal experiences in forming self-efficacy beliefs because personal performance failure will usually take precedence over vicarious success. Self-efficacy can also be raised or lowered through comments or feedback from others. Finally, self-efficacy beliefs can also be influenced by physiological reactions such as increased heart rate or feelings of anxiety that can communicate to individuals that they are lacking in skill or ability to accomplish something.
Although self-efficacy beliefs are highly predictive of performance, they are not always accurate or truly representative of the student's true capabilities. Not surprisingly, students lacking confidence in skills they actually possess will be less likely to engage in tasks requiring those skills and put forth the needed effort and will be more likely to quit when the task becomes challenging (Bandura, 1993, 1997; Linnenbrink & Pintrich, 2003; Pintrich & Schunk, 1996; Schunk, 2003). Although low self-efficacy can be detrimental to learning, overly high levels of self-efficacy can be just as problematic (Schunk, 2003). Students with overly high levels of self-efficacy may feel overconfident and not exercise the appropriate effort needed to be successful at something (Bandura, 1989; Linnenbrink & Pintrich; Salomon, 1984; Schunk). Bandura and Schunk (1981) noted that students' inaccurate self-efficacy estimates may develop from erroneous task analysis or lack of self-knowledge, both of which, Klassen (2006) highlighted, are problems commonly associated with learning disabled students. What seems apparent is that students benefit from holding accurate and appropriate self-efficacy beliefs, and teachers and counselors also benefit from knowing and understanding their students' self-efficacy beliefs.
Recently, research on the relationship between self-beliefs, including self-efficacy and college outcomes, have received a great deal of attention. For example, a meta-analysis by Robbins et al. (2004), which integrated the literature on educational persistence and motivational theory models, examined the relationship between psychosocial and study skills factors and college outcomes. Among the constructs examined, academic self-efficacy and academic goals (i.e., one's degree expectations) were among those most highly related to college outcomes, even after controlling for traditional cognitive predictors of college success. [End Page 666]
Specifically, Robbins et al.'s (2004) meta-analytic findings showed that the correlation between academic goals and retention, after correcting for criterion unreliability, was .21 based on 20,010 students across 33 studies. For academic self-efficacy, the correlation with retention was .26 based on 6,930 students across 6 studies. In addition, they examined these relationships with cumulative GPA and found positive support. For degree goals, the correlation was .16 based on 17,575 students across 34 studies. The correlation between academic self-efficacy and cumulative GPA was .38 based on 9,598 students across 18 studies. In terms of incremental validity, hierarchical regression analyses showed that academic self-efficacy still had a positive relationship with both retention and cumulative GPA after controlling for socioeconomic status (SES), HSGPA, and standardized test scores. Academic goals provided incremental validity to the prediction of retention after controlling for the same traditional measures. In sum, students with higher self-efficacy and degree goals tended to have more positive college outcomes.
The purpose of the current study is to extend the findings of Robbins et al. (2004). Specifically, this study examines the relationship of academic self-efficacy with other motivational indicators including willingness to seek academic help and admission test score-sending behavior as well as traditional high school and college level academic outcomes. Help-seeking and score reporting are included as outcomes in this study based on their association with academic achievement. For example, help-seeking is associated with active learning, higher grades, and greater strategy use (Karabenick & Knapp, 1991). Sending one's scores to a greater number of colleges and universities may also be thought of as strategic or proactive behavior by casting a wide net of institutions to which to apply and hopefully get accepted. (It should be pointed out that score sending represents a proxy of college application behavior in that we only know where the students sent their scores but do not know to which institutions they actually applied.) Recent research showed that the number of institutions that students sent SAT scores to was significantly related to HSGPA as well as to the number of AP tests taken (Shaw, Kobrin, Packman, & Schmidt, 2009).
Students' degree goal aspirations were also examined in this study. Given that the pursuit and attainment of higher education in the United States is a highly valued and rewarded commodity (Baum & Ma, 2007), further examination of the relationship between this variable and other related academic and motivational student characteristics is warranted. Previous research findings have been mixed in terms of the relationship between performance and degree goal aspirations (e.g. Centra, 1980; Robbins et al., 2004). More research on the topic is warranted.
In order to make the results of the current study more comprehensible and meaningful, mean performance by academic indicators was provided for each self-belief group studied. Additionally, whether students high in academic self-beliefs varied systematically in terms of gender, ethnicity, and best language compared to students low in academic self-beliefs was examined. That is, before these measures are offered as a solution to some of the adverse impact often attributed to traditional measures, research must examine how underrepresented students perform on them. More sophisticated methodology in the examination of the study variables and subgroups was purposefully not employed in this research. The intention of this study is to provide practical information to higher [End Page 667] education professionals who play a role in assisting students who are academically at risk. These individuals could benefit from learning more about the role of self-efficacy and help-seeking attitudes in order to better serve their students and ensure their success in college.
As part of a larger research endeavor, 4-year colleges and universities across the United States were contacted in order to provide first-year performance data from the fall 2006 entering cohort of first-year, first-time students. Participating institutions were offered a stipend for the work involved in creating the files with first-year performance data, which were uploaded to a secure online portal. The data collected from each institution included students' coursework and grades, FYGPA, and whether or not they returned for the second year. These data were matched to College Board databases that included SAT scores, self-reported HSGPA from the SAT questionnaire, and other demographic information for these students. The sample consisted of individual level data on 196,364 students from 110 colleges and universities across the United States. Students in the sample who did not have complete information on all study variables were excluded from the analyses, resulting in a final sample size of 107,453 students. The sample of institutions is diverse in terms of region of the country, selectivity, size, and control (i.e., private vs. public).
Demographic information, including gender, race/ethnicity, and best language, was self-reported by the students and obtained from SAT questionnaire responses. For best language, students were asked, "What language do you know best?" and the responses options included: English, English and another language about the same, and another language.
Self-Estimate of Math Ability.
On the SAT questionnaire, one item asks students to rate themselves in terms of perceived math ability relative to other people their age ("How do you think you compare with other people your own age in math ability?") with the following response options: (a) among the highest 10 percent in this area of ability, (b) above average in this area, (c) average in this area, (d) below average in this area.
Self-Estimate of Writing Ability.
Similar to self-estimate of math ability, students rated themselves in terms of perceived writing ability relative to other people their age ("How do you think you compare with other people your own age in writing ability?") with the following response options: (a) among the highest 10 percent in this area of ability, (b) above average in this area, (c) average in this area, (d) below average in this area.
An item on the SAT questionnaire asks students to indicate their educational degree aspiration ("What is the highest level of education you plan to complete beyond high school?") with the following response options: (a) specialized training or certificate program, (b) two-year associate of arts or sciences degree, (c) bachelor's degree, (d) master's degree, (e) doctoral or related degree, (f) other, (g) undecided.
Official SAT scores, obtained from the 2006 College-Bound Senior Cohort database, were used in the analyses to examine the relationship between performance in specific academic domains (e.g., math) with self-beliefs in the same or similar domains. This database comprises scores of students who participated in the SAT program and graduated from high school in 2006. Students' most recent SAT scores were used in the analyses. The SAT comprises three sections, Critical Reading, [End Page 668] Math, and Writing, and the score scale ranges from 200 to 800 for each section.
HSGPA was self-reported and obtained from SAT questionnaire responses. Students' HSGPAs were scored on a 12-point scale ranging from a maximum of A+ (4.33) to a minimum of F (0.00).
Score Report Sending.
The number of colleges and/or universities to which a student sent his/her SAT score report was obtained from College Board records. Students receive four free score reports when they take the SAT. The number of score reports sent by students ranged from 1 to 30 (M = 5.85, SD = 3.40).
Another item on the SAT questionnaire asked whether students may want help in college to improve their mathematics skills and writing skills, in addition to other areas. The items were scored dichotomously with a 1 indicating that they did want help and a 0 indicating that they did not want help for each academic area, respectively.
First-Year GPA (FYGPA).
Each participating institution supplied FYGPA values for their 2006 first-time, first-year students. The range of FYGPA across institutions was 0.00 to 4.27.
Each participating institution supplied second-year retention data for their 2006 first-time, first-year students. Students were assigned a value of 1 if they returned for their second year and a value of 0 if they did not return.
[End Page 669]
Self-Belief in Math Ability
Of the 107,453 students in the sample, 33.5% believed they were in the highest 10% of math ability, 42.3% believed they were above average, 23.0% believed they were average, and only 1.2% believed they were below average. Compared to the total sample, students in the highest 10% in math ability were more likely to be male, Asian-American, and White. For gender, the difference was most pronounced. The total sample was 45.6% male, but the group believing they were in the highest 10% was 57.7% male. The demographic characteristics of the group believing they were above average largely resembled the total group. The average and below average groups comprised more females, African American students, and Hispanic students as compared to the total sample. There was little variability in the percentage of students by best language groups across perceived math ability groups. See Table 1 for more details.
In Table 2, the means and standard deviations of the academic outcomes of HSGPA, SAT scores, FYGPA, desiring math and writing skills help, number of SAT score reports sent, and retention rates by perceived math ability groups are provided. One-way ANOVA by perceived math ability groups was run for each outcome; results reveal [End Page 670] significant group differences for each outcome, all Fs(3, 100749) = 10.45, all ps < .001. Specifically, most academic indicators were positively related to math ability beliefs. Compared to students who stated they were below average in math ability, students who stated they were in the highest 10% in math ability had higher HSGPAs, SAT Critical Reading scores, SAT Math scores, SAT Writing scores, FYGPAs, and second-year retention rates. They also sent their SAT scores to more colleges.
As for the two items related to desiring help in math and writing, a small percentage of the highest 10% math group (6.6%) stated they wanted help to improve their math skills, whereas 65.2% of the students believing they were below average in math desired help. The percentage of students across the math self-beliefs groups that desired help in writing was relatively constant. This is not surprising given the mismatch of academic domains under examination (Bandura, 1986; Klassen, 2002; Zimmerman, 1995).
Despite the general above-average effect, students were somewhat accurate in their math perceptions with a difference of nearly 200 points on the SAT Math section between the highest 10% group and the below average group. The SAT differences for the Critical Reading and Writing sections were much smaller (90 to 95 points). Such patterns provide additional support for the domain-specific nature of self-efficacy beliefs.
[End Page 671]
Self-Belief in Writing Ability
Similar to the results for math self-beliefs, the majority of students believed that they were above average in writing ability. Specifically, 29.4% believed that they were in the highest 10% of writing ability, 44.7% believed they were above average, 24.9% believed they were average, and only 1.0% believed they were below average. Compared to the total sample, students in the highest 10% group were more likely to be White and report their best language as English. For gender, there was a slightly larger percentage of females in the highest 10% group (55.8%) as compared to the total sample (54.4%). Similar to the math results, the demographic characteristics of the above average group largely resembled the total group. The average and below average groups had more males, Asian American students, students whose best language was English and another language, and students whose best language was another language as compared to the total sample. See Table 3 for more details.
In Table 4, the means, standard deviations, ANOVA results of the academic outcomes of HSGPA, SAT scores, FYGPA, desiring help in math and writing, number of SAT score reports sent, and retention rates by perceived writing ability groups are provided. Significant group differences for each [End Page 672] outcome existed, all Fs(3, 100749) ≥ 29.35, all ps < .001. Most academic indicators were positively related to writing ability beliefs. For example, compared to students who stated they were below average in writing ability, students who stated they were in the highest 10% in writing ability had higher HSGPAs, SAT Critical Reading scores, SAT Math scores, SAT Writing scores, FYGPAs, and second-year retention rates. They also sent their SAT score report to more colleges.
As for the two items regarding help with writing and math skills, a small percentage (9.9%) of the highest 10% group stated that they wanted help to improve their writing skills, whereas 59.4% of those believing they were below average in writing desired help. For the math help item, the percentage of students desiring help across writing self-beliefs groups was relatively constant, similar to the math results.
With the caveat that the results reveal an overwhelming above-average effect, students were still somewhat accurate in their writing ability perceptions with a difference of around 150 points on both the Critical Reading and the Writing section of the SAT between the highest 10% group and the below average
[End Page 673]
group. Furthermore, perceived writing ability was positively related to HSGPA, SAT scores, FYGPA, and retention, all of which are critical academic outcomes. The SAT Math difference between the two groups was much smaller (60 points).
The majority (79.4%) of students in the sample aspired to complete a bachelor's or higher (see Table 5), whereas 19.7% of the sample was undecided. Less than 1% indicated the desire to earn a certificate, associate, or other. Compared to the total sample, students who stated that their degree aspiration was a doctoral degree were more likely to be female. Additionally, they were more likely to be Asian, African American, and Hispanic and less likely to be White, suggesting that White students have lower degree aspirations than do those in other ethnic/racial subgroups. In fact, White students made up 69.0% of the total sample but only 57.9% of the doctoral aspiration goal group. The demographic characteristics of the master's degree group largely resembled the total group. The bachelor's degree group comprised more males, White students, and students whose best language was English, when compared to the total sample. For the undecided group, [End Page 674] there were more White students and students whose best language was English and fewer African American and Hispanic students as compared to the total sample.
In Table 6, the means, standard deviations, and one-way ANOVA results of HSGPA, SAT scores, FYGPA, desiring help in math and writing, number of SAT score reports sent, and retention rates by degree aspirations are provided. Higher degree aspirations tended to be related to higher academic outcomes with significant mean differences across groups, all Fs(6, 100746) ≥ 36.10, all ps < .001. The one notable exception was the undecided group, which performed similarly to students who selected either a doctoral or master's degree. Furthermore, students who selected "other" degree performed similarly to students who selected a bachelor's degree. Compared to students who stated they wanted to obtain an associate degree, students who stated they wanted to obtain a doctoral degree had higher HSGPAs, SAT Critical Reading scores, SAT Math scores, SAT Writing scores, FYGPAs, and second-year retention rates. Additionally, they sent their SAT scores to more colleges, though the differences are smaller when compared to the math and writing self-belief results. The percentage of students across degree goal groups who desired help in math and writing was relatively constant, ranging from 18% to 24%.
Congruent with a substantial body of literature, students' academic self-beliefs are positively related to academic outcomes. With regard to demographic characteristics of the self-beliefs group, students in the highest 10% in math ability were less likely to be female, African American, or Hispanic, as compared to the total group. For the highest 10% in writing ability, there were smaller percentages of African American, Asian, and Hispanic students as well as students whose best language was not English as compared to the total group. These results suggest that using these measures in the admission process would not eliminate the current problem of adverse impact. In other words, despite the fact that there appears to be a strong link between efficacy measures and academic outcomes (such that students with higher efficacy demonstrate higher performance), students who may be disadvantaged by traditional measures tend to also hold lower self-efficacy beliefs.
A different pattern emerges for degree goals, however. Students with degree aspirations of a doctorate were more likely to be African American, Asian, or Hispanic and also indicate that their best language is not English as compared to the total group. Furthermore, White students comprised a smaller percentage of the doctoral degree aspiration group (57.9%) as compared to the total group (69.0%). Other studies have found similar patterns, with traditionally underrepresented students having higher degree goal aspirations than do White students (Centra, 1980; Laanan, 2000; Phillips, 1972). The potential of degree aspirations in the admission process as a valid predictor of college success resulting in less adverse impact should be further explored.
An additional contribution of this study is the investigation of the relationship between self-beliefs and desiring help to improve skills in the respective content areas. The results showed that students who believed they were below average in either math or writing ability indicated that they also desired help in those areas at a much higher rate (65.2% and 59.4%, respectively) as compared to the total sample (18.6% and 20.3%, respectively). This is in line with the consistency principle in the help-seeking literature, which notes that students with more negative self-beliefs are more likely to seek help, believing they are in need of help (Karabenick & Knapp, 1991). [End Page 675] These students also perform substantially worse in college in terms of FYGPA and retention. This begs the question of whether students desiring help actually receive help once matriculated in college. If they are not receiving help, is it because the students ultimately chose not to seek help or because there was no help offered? Or perhaps they are receiving ineffective help. Another possibility is that the transition from high school to college changes these students' beliefs about seeking help, and they may feel less comfortable with seeking help in their larger colleges than in their smaller high schools.
As a suggested area for exploration, colleges and universities could identify first-year students with weaker HSGPAs or SAT scores than most enrolled students at their institution and determine who among these students has also indicated desiring help in a particular area based on their SAT questionnaire response. These identified students can then be contacted in order to receive special resources and/or courses to improve particular skills. Females and minority students, specifically, should receive special attention given the results of the current study. Efforts focused on these subgroups may help minimize group disparities in self-beliefs but may also help reduce academic performance differences, which continues to be one of the most pressing concerns among the educational community.
Related to this notion, it is important to establish a campus culture which is receptive and looks positively upon help-seeking. The way that professors and fellow students respond to requests for assistance is a central determinant in whether students will seek resources to improve their knowledge and performance in a particular area (Karabenick, 2004). Therefore, the general encouragement and offering of a wide availability of helpful resources will aid in creating a more receptive environment for those open to and considering seeking help. Faculty professional development should incorporate information on the importance of developing an academic culture where seeking help is encouraged.
Even though the majority of students with low academic self-efficacy stated that they wanted help with improving their skills, a large percentage did not. This is likely related to the notion found in prior research where students with lower academic self-esteem view seeking help as threatening because it signifies inadequacy, also called the vulnerability principle (Karabenick & Knapp, 1991). Research should examine whether self-efficacy interventions by high school teachers and counselors could benefit students in their first year of college by working to raise students' self-efficacy beliefs, thereby increasing strategy use and openness to seeking help.
There have been numerous recommendations offered by researchers for appropriately increasing students' self-efficacy. Schunk (2003) emphasized using instructional methods that incorporate modeled strategies in the content area, progress feedback to students, goal-setting, and self-evaluations of progress. Peer tutoring programs in high school, where higher ability students are modeling particular homework or study strategies to weaker students, may work to increase students' self-efficacy and comfort with seeking help in college. If such interventions proved helpful, higher education institutions may want to collaborate with partner high schools in creating instructional programs geared to increasing student's self-efficacy. Research should also examine whether this particular group of students who hold low self-efficacy beliefs and were not planning to seek help to improve their skills would be more resistant to help if it was provided to them. Future research on altering students' beliefs so that they are more apt to seek help if they are not confident in their skills in order to achieve their academic aspirations should be a priority. [End Page 676]
Finally, this study contributed to the current literature of self-efficacy and academic behavior by examining the relationship between self-beliefs and score sending. It was hypothesized that students with higher self-beliefs would send their scores to more institutions, which is what the current study found. However, it could be argued that the number of scores sent could also be influenced by the student's SES. Though students have the option to send up to four score reports for free, future research should determine whether controlling for SES factors (e.g., parental income and parental education) accounts for the differences among self-belief groups and score sending behavior.
In sum, this study found that students' beliefs about their ability as well as their degree goals were related to numerous academic outcomes. Furthermore, this study sheds light on the characteristics of students with positive self-beliefs. Though this study addressed questions similar to those explored in previous research, by displaying the results as mean performance by self-belief groups, the effects are more easily interpretable and informative for readers—especially college counselors and admission officers, who may not have a background in psychometrics or advanced statistical methods. [End Page 677]
Krista D. Mattern and Emily J. Shaw are both Associate Research Scientists in Research & Development at The College Board.