The association between the directional accuracy of self-efficacy and accounting course performance.
This study investigates the relation between students' beliefs about their course-specific abilities and their classroom performance. Stone (1994, 452) defines self-efficacy as "judgments of task-specific, expected performance." (1) We extend this definition to focus on the directional accuracy of students' judgments about their expected performance. This focus means that in addition to examining students' ability to accurately gauge their current standing in the course, we also investigate the direction of inaccuracy and the consequences associated with pessimism relative to optimism regarding performance outcomes. Bandura (1986) argues that rather than simply predicting future behavior, individuals' beliefs actually determine future behavior. However, if beliefs determine behavior, then a tension arises between early studies that treated beliefs as though they were accurate predictors of performance and later evidence (Carpenter et al. 1993) that overly optimistic students perform poorly. We examine how th e directional accuracy of students' beliefs is associated with their classroom performance. We find that being inaccurate in one direction (optimism) is associated with poorer performance, while being inaccurate in the opposite direction (pessimism) is associated with better performance. This study contributes to the educational literature by documenting a relation between the accuracy of expectations and course performance in a college-level accounting context. Furthermore, it provides evidence regarding the differential effects of optimism and pessimism in determining classroom performance.
Educators with shared experience of watching poor-performing students miscalculate, misgauge, and misread their performance, despite all of the cues available to them, feel a natural interest in exploring how self-efficacy can be measured and how it relates to performance. The concept of self-efficacy is important because students with unrealistic expectations (especially overly optimistic expectations) may have difficulty aligning efforts with desired performance levels and, as a result, perform more poorly than if their self-efficacy assessments were reasonable. (2 )Therefore, an understanding of the relation between the directional accuracy of students' beliefs and course performance is important because it can guide educators in working with students with overly optimistic or pessimistic expectations.
The specific question of whether a student's self-efficacy is a determinant of subsequent outcomes is not an entirely new concept in the accounting literature. Although Keef and Roush (1997, 326-328) focus mainly on gender and race effects, they provide some evidence that a sense of self-concept (3) is associated with accounting student performance. However, Keef and Roush suggest (1997, 326) that specific self-concepts are an under-explored avenue for future research.
Path analysis results based on a sample of 214 students from two universities and nine classes suggest that students who more conservatively (that is, less optimistically or more pessimistically) assess their abilities tend to perform better on a second exam and receive higher final course grades. This relationship holds after controlling for cumulative GPA in accounting courses, average exam performance during the course, trajectory of achievement (that is, whether the student is improving or deteriorating in exam performance), number of accounting classes already completed, and the extent of involvement in extracurricular activities. Path analysis results further suggest that conservatism of self-efficacy is an intervening variable through which student characteristics have an indirect relationship to performance. We infer from these results that better students tend to be pessimistic, which apparently leads to more study time and greater improvement in actual outcomes, while poorer students tend to be ove rly optimistic, which apparently leads to inadequate self-regulatory activities and worse performance. (4)
We organize our paper by first elaborating on the theoretical construct of self-efficacy and associated research, including its implications for research design. Next, we set forth the hypothesis and methodology. We then describe the results, limitations, and implications of our findings for accounting educators, as well as related opportunities for further research.
RELATED RESEARCH AND HYPOTHESIS DEVELOPMENT
Self-Efficacy and Achievement
The stream of literature most relevant to this study examines students' academic self-efficacy beliefs (5) and their correlation with motivation constructs and academic performance (Pajares 2000; Bandura 1997). Collins (1982) finds that self-efficacy beliefs are related to effort, persistence, and perseverance. Zimmerman et al. (1992) report that students with a better sense of self-efficacy make better use of cognitive strategies and self-regulatory practices (such as the approach to learning (6) and time taken to study a particular topic). Similarly, Schraw (1994) finds that college students' self-knowledge of strengths and weaknesses is correlated with test performance. Pintrich and De Groot (1990) find evidence that self-regulatory efficacy (such as awareness of how preparation time is used) is related to academic efficacy.
Several other studies find evidence that performance in different contexts is associated with self-efficacy beliefs. (7) For example, Bouffard-Bouchard et al. (1991) present evidence that regardless of cognitive ability, children with a high sense of self-efficacy are more successful in solving conceptual problems than those of equal ability with lower self-efficacy. More self-efficacious students manage work time better, are more persistent, and are less likely to reject correct solutions prematurely.
Pajares and Kranzler (1995) report that individuals' judgments of their own capabilities to accomplish specific tasks strongly influence their motivation and behavior. They also find a general bias in the inaccuracy of self-efficacy, in that most students are over-optimistic about their mathematics capability. This links to prior research demonstrating that providing children who are deficient in mathematics and reading comprehension with strategy knowledge (i.e., knowledge of what areas need their attention), as well as instruction on how to take corrective action when necessary, significantly enhances their performance (Bandura 1997). These findings link back to that element in the definition of self-efficacy related to individuals' personal capability to exercise control over their own level of functioning (Bandura 1993).
Accuracy of Self-Efficacy and Importance of Task Specificity
Bandura (1986) emphasizes that successful functioning in virtually all human endeavors (e.g., achievement of a goal or completion of a task) requires reasonably accurate efficacy appraisals. Bandura (1986) observes that it is common for individuals to over- or under-estimate their abilities and to suffer consequences from these errors of judgment. However, these consequences are a part of the continual process of efficacy self-appraisals. Bandura (1986, 21) refers to self-reflective capability as "distinctly human." Individuals reflect on and evaluate their own experiences and thought processes, potentially altering their own thinking and subsequent behavior.
Bandura (1986) theorizes that judgments of capability that are more task-specific are better predictors of related performance than are more generalized judgments. Relich et al. (1986) and Pajares and Miller (1997), report that domainspecific self-efficacy (e.g., item-specific math self-efficacy beliefs) is more predictive of mathematics problem-solving performance than of more general self-efficacy assessments. Graham et al. (1981) argue that exposure to sources of self-efficacy information varies across different domains of activity, such as in the undergraduate environment (Ancis and Phillips 1996). This highlights the importance of context. In this study, we operationalize self-efficacy accuracy as the directional difference between expectations and actual outcomes, using variables tailored to accounting. Specifically, we follow Stone (1994, 452), who promotes the accuracy of expectations as a crucial dimension of self-efficacy, by focusing on the correspondence between direction and magnitude of inaccur acy in expectations and actual exam and course-performance criterion variables.
Based on prior literature, we predict that more accurate self-efficacy beliefs will lead to better student performance. However, prior literature suggests that differences in performance depend on the direction of observed inaccuracies. Our method for measuring the accuracy of self-efficacy (as distinct from self-efficacy or self-confidence) quantifies both the magnitude and direction of prediction error. This tailored "measure of self-precept" (Bandura 1986, 396) addresses under-and overconfidence regarding performance, since the direction of expectations is theoretically expected to lead to differing self-regulatory practices. For example, the more underconfident individuals are, the more they may prepare. We expect that better preparation, in turn, will lead to higher performance. Accuracy in either direction is arguably preferable to inaccuracy. However, one might analogize to notions of efficiency and effectiveness, whereby a goal of good performance would be inefficiently achievable with underconfidenc e, yet ineffectively achievable (if achievable at all) by overconfidence. Our measure permits exploration of differences between over- and underconfident self-beliefs. As a result, we hypothesize that both the accuracy and direction of inaccuracy of self-efficacy are related to academic performance. We thus state our hypothesis (in alternate form) as follows:
Ha: The directional accuracy of self-efficacy is related to academic performance.
DATA AND METHODOLOGY
Figure 1 provides a timeline of key events in the cost accounting course from which data for this study were obtained. On the first day of class, students filled out a survey from which we gathered demographic information. During the first two-thirds of the course, students received daily feedback on quizzes and homework. In addition, they took the first of two exams and received prompt feedback on their performance. Then (as will be described in more detail later), we measured students' accuracy of self-efficacy immediately after taking the second exam, but before they received feedback. (8)
Wenning (2000) suggests that correlating student grades (9) with self-assessment scores can be helpful to triangulating data. As a result, we incorporate this approach in our study in examining the relation between self-efficacy and academic performance. Further, Pajares (1996a) suggests that self-efficacy is more effectively assessed when related to a particular task outcome. We follow his suggestion by focusing on specific course components (see Figure 1) in assessing the directional accuracy of self-efficacy.
Past literature has often used overall semester grades (across all courses taken) and achievement tests as outcome tasks. But these suffer from generality, whereas task-specificity and context-specific measurement of self-efficacy (Langenfeld and Pajares 1993) generate more precise judgments of capability when they are matched to a specific outcome (Bandura 1986; Gobeil and Phillips 2001). Our focus on a particular course ensures sufficient breadth to assure external validity and practical relevance (Lent and Hackett 1987), and our use of differencing (outcome less prediction) ensures an easy-to-compute measure for evaluating the accuracy and direction of self-efficacy beliefs (Lent et al. 1993).
We collected data from 214 students enrolled in nine cost accounting classes (taught by one of the authors) at two different universities. This design was necessary so that instructor effects could be controlled as completely as possible. (10) We employ data from two universities to enhance our ability to generalize the results: a small, private northern school (118 students), and a large, public southern school (96 students). The sample includes 130 men (61 percent) and 84 women (39 percent). Accounting majors comprise 80 percent of the sample, with the remainder either M.B.A.s or undergraduate management students. (11) Approximately 75 percent of the students are at the junior level.
We hypothesize that both the direction and accuracy of students' self-efficacy are related to academic performance. Figure 2 summarizes our predictions regarding (1) the relation between the directional accuracy of self-efficacy and course performance, and (2) the relations among student characteristics and course performance both directly and indirectly through the mediation of self-efficacy. We test the theory depicted in Figure 2 using path analysis. Then we use regression analysis in robustness testing. The following subsections explain how we operationalize each construct described in Figure 2.
Explanatory and Control Variables
In testing our hypothesis, we employ explanatory variables representing student characteristics that prior research has found relevant to self-efficacy and course performance. Pajares and Valiante (1996) identify three important predictors of self-efficacy: (1) skill, (2) ability, and (3) previous accomplishments. Each of our explanatory variables relates to at least one of these three constructs.
Previous Accomplishments in Accounting
Our first explanatory variable, grade point average in accounting courses taken (ACCTGPA), relates to skill, ability, and previous accomplishments in accounting courses. (12) We expect (as Figure 2 shows) that ACCTGPA will be positively associated with our self-efficacy and academic performance measures. We collected this information from official student records whenever possible, since self-reported GPA tends to be less reliable (Wilson et al. 1997). However, since GPA data for one class could not be obtained from university records, we used self-reported data in that instance.
Cost Accounting Skill Level
We measure task-specific skill and accomplishments by averaging student scores on the first two exams (AVEREXAM). (13) Since students had taken two exams prior to self-efficacy measurement, both should enter into this proxy for cost accounting (i.e., task-specific) competency level. In addition, the averaging of two test observations should reduce the measurement error that could result from students having an exceptionally good or bad day when they took a single exam. AVEREXAM conceptually maps to students' experiences to date that are an element in formulating their beliefs. We predict that AVEREXAM will be positively associated with our accuracy of self-efficacy and course performance measures.
Trajectory of Achievement
In examining a student's relative optimism/pessimism in predicting his/her performance, the student's perceived "momentum" in the class will likely affect his/her beliefs about future performance. Since a student's trajectory could affect his/her confidence in making self-efficacy assessments, we include an explanatory variable that tracks the change in current performance relative to previous accomplishments. Our proxy for a student's direction and momentum, GRADECHG, is the change in exam scores (the students' score on the second exam minus the student's first exam score). Thus, positive numbers indicate ascendancy (improvement), and negative numbers the opposite. This variable allows for the possibility that these improvements or declines may be a separate element from the level of competence captured in the previous variables. We expect that GRADECHG will be positively related to both the directional accuracy of self-efficacy and course performance.
Experience in Accounting Classes
Experience in accounting classes represents another facet of previous accomplishments, which should relate positively to our self-efficacy and academic performance measures. Therefore, we also include an explanatory variable, NUMCLASS, for the number of accounting classes a student has already completed. We collected this number in a survey on the first day of the semester, as Figure 1 shows. A spot check of transcripts indicated no major discrepancies from self-reported information.
In addition to academic accomplishments, previous experience also includes activities outside of the classroom. These activities could affect a student's sense of self-efficacy and course performance. We asked students to self-report the extracurricular activities in which they had participated while attending the university. They were also asked whether they had a job while attending school. We coded this measure (EXTRACUR) 1 if the student had participated in campus activities (such as sports, clubs, fraternities, sororities, etc.) or had a job. We coded EXTRACUR 2 if the student had participated in campus activities and had a job. Students reporting neither a job nor campus activities were coded 0. We expect higher levels of extracurricular involvement to divert students' attention from schoolwork, therefore resulting in lower course performance. Such activities could have both a positive and negative association with students' ability to assess their exam and course performance accurately. Therefore, we make no prediction as to the sign of the relation between EXTRACUR and our self-efficacy measures.
Control for Differences across Classes
While no theoretical reason exists for expecting different results at a class or university level, given a single instructor, our sample comprises nine classes at two universities. Therefore, we also include indicator variables (T2 through T9) in our robustness checks using regression analysis (reported following our path analysis) to ensure consistency across classes and that no single class dominates our results. (14, 15)
Self-Efficacy with Respect to Exam Score
We use two empirical measures of the directional accuracy of self-efficacy, which we predict to relate positively to academic performance. The first measure captures a student's belief as to his/her performance on the second of two interim examinations. As Figure 1 depicts, we measured this variable in a survey taken during the first class following the second exam. Students were asked to estimate the number of points (out of 100) they earned on that exam (before the graded exams were returned to them). The difference between a student's actual exam score (EX2SCORE) and the student's estimated exam score measures the accuracy of self-efficacy, including direction of inaccuracy with respect to their exam performance (EXAMERR). (16)
Self-Efficacy with Respect to Course Grade
Our second self-efficacy measure captures a student's belief about his/her final (letter) grade in the course. We took this measurement in the same survey in which we measured self-efficacy with respect to Exam 2. The exact phrasing of the questions included on the experimental instrument is as follows:
* How many points do you think you earned on Exam 2?
* What grade do you anticipate receiving in this class?
The difference between a student's actual grade in the course (FINGRADE (17)) and the student's estimated final grade measures the student's accuracy of self-efficacy (GRADERR), including direction of inaccuracy, with respect to overall performance in the class. As Figure 1 shows, both measures (EXAMERR and GRADERR) were taken at the same time--approximately 11 (8) weeks into the semester (quarter). This point of measurement is appropriate, since students have already participated in one complete feedback cycle (graded and returned first exams) from which they can anticipate the difficulty of the second exam. (18) It is also sufficiently distant from the end of the course to permit adaptive behavior, such as increased effort or strategy corrections. (19)
This section provides descriptive statistics and a correlation matrix for our variables of interest and presents both path analysis and multiple regression results of our hypothesis tests. (20)
Panel A of Table 1 provides descriptive information about the variables used in the study for the entire dataset. The mean (median) self-efficacy inaccuracy with respect to the second exam, EXAMERR, is 3.59 (4.00). This suggests that students under-estimate their second exam grade by approximately four points, on average, which implies pessimism. The largest positive prediction error (pessimism) in the data set is 51, associated with a student who earned a 76 on the second exam and a 3.3 (i.e., B+) final course grade. The largest negative prediction error (optimism) in the data set is -25, associated with a student who earned a 45 on the second exam and a 2.3 (i.e., C+) final course grade. This suggests that the association of pessimism and optimism, are not monotonically associated with the best and poorest student, respectively, for the specific task. Yet the potential consequences of corrective or destructive self-regulatory behavior would seem apparent in these extremes.
The mean (median) self-efficacy inaccuracy with respect to a student's final grade, GRADERR, is -0.07 (0.00). This indicates that by the time students took the second exam, they had a good sense of their final grade in the class. The mean (median) score on the second exam, E2SCORE, is 76.24 (76.00). The average (median) grade in the class, FINGRADE, is 3.11 (3.30). Thus, while the average student achieves a B+ in the course, the distribution is skewed slightly negatively.
As suggested earlier, signs and magnitudes of prediction errors capture potentially important differences in students. To explore this dimension, we created three distinct groups within our data set: (1) pessimistic students--those who performed better than they predicted on both the second exam and the final course grade (94 of the 214 subjects); (2) overly optimistic students--those who performed worse than they predicted on both the second exam and the final course grade (47 students); and (3) mixed accuracy students--those who were both optimistic and pessimistic in predicting their exam score and course grade (numbering 39 with pessimistic/optimistic and 34 with optimistic/pessimistic prediction errors in anticipating the outcome of the second exam and course grade, respectively). Panel B of Table 1 reports mean and median statistics for the second exam and final course grade for each subgroup. These results are consistent with the notion that, in terms of performance on the second exam and final course grades, better students are more modest or pessimistic in their expectations, whereas the poorer students are more optimistic.
Correlations among Variables
Table 2 provides pairwise correlations among our variables of interest for the full sample. We find that our directional accuracy of self-efficacy measures (EXAMERR and GRADERR) are positively correlated with course performance measures (E2SCORE and FINGRADE) and accomplishment measures (ACCTGPA and AVEREXAM). This suggests that students with more conservative (i.e., less optimistic or more pessimistic) self-efficacy beliefs have better performance in cost accounting and stronger prior accomplishments. EXAMERR is positively correlated with trajectory of achievement (GRADECHG), suggesting that students who more conservatively predict their exam score have greater improvement from the first exam to the second. This is consistent with the notion that students with more conservative expectations may spend more time preparing for the second exam. GRADERR is positively correlated with NUMCLASS, suggesting that students with more experience in accounting classes have a more conservative sense of self-efficacy with respect to overall course grade. In addition, EXAMERR is negatively correlated with extracurricular activities (EXTRACUR). However, the fact that EXTRACUR is not significantly correlated with the directional accuracy of self-efficacy with respect to overall course performance (GRADERR) suggests that the effects of extracurricular activities can be mixed. In addition to the possibility of distracting students' attention, such experiences may improve skills and abilities (such as time management and work ethic) and thereby enhance students' ability to assess accurately their current standing in their coursework. We find strong positive correlations among performance (E2SCORE and FINGRADE) and competency measures (ACCTGPA and AVEREXAM). However, the correlation between E2SCORE and AVEREXAM is high by construction since AVEREXAM is the simple average of the first and second exams. Finally, we find a negative correlation between the number of classes taken (NUMCLASS) and extracurricular activities (EXTRACUR). This suggests that newer students tend to be involved in more extracurricular activities and that students closer to graduation tend to focus more time on their schoolwork.
Figure 3 presents path analysis results that examine the predicted relations summarized in Figure 2. We hypothesize that our directional accuracy of self-efficacy measures (EXAMERR and GRADERR) are related to accounting-student performance on the second exam (EX2SCORE) and the overall course (FINGRADE), respectively. The results indicate that task-specific prediction error with respect to the exam (EXAMERR) is positively related to exam performance (p < 0.01). This suggests that, consistent with results presented in Panel B of Table 1, students with more conservative (i.e., less optimistic or more pessimistic) expectations perform better on exams. In addition, the positive coefficient on GRADERR (p < 0.01) is consistent with the notion that a less optimistic (more pessimistic) prediction error about one's standing in the course is associated with better overall performance. While the pairwise correlations presented in Table 2 suggest that the more conservative a student's self-efficacy appraisals, the better the performance, the path analysis results suggest that this positive relation holds even after controlling for student characteristics that prior research has found to be related to self-efficacy and student performance.
Consistent with prior research, we find that several student characteristics related to skill, ability, and previous achievements are positively related to directional self-efficacy. Figure 3 shows that average exam score (AVEREXAM), trajectory of achievement (GRADECHG), and number of accounting classes taken (NUMCLASS) are positively related to our self-efficacy measures. (21) This suggests that students with better task-specific abilities, positive momentum, and more experience in accounting classes tend to have a more conservative outlook on their abilities. We also find evidence that extracurricular activities (EXTRACUR) are negatively related to our task-specific self-efficacy with respect to exam performance measure (EXAMERR). This is consistent with the notion that students involved in extracurricular activities tend to be more optimistic in their self-efficacy appraisals.
We also find that student characteristics are directly related to student performance. Figure 3's path analysis results indicate that grade point average in prior accounting classes (ACCTGPA), average exam score (AVEREXAM), and trajectory of achievement (GRADECHG) are positively related to student performance in at least one of the two models. These results suggest that students with higher achievement in prior classes, better task-specific course skills, and a positive trajectory perform better in the cost accounting class. Overall, the path analysis results provide support for the plausibility of pre-specified causal hypotheses depicted in Figure 2. (22)
This section reports the results of several sensitivity tests. First, we examine the robustness of our path analysis results by investigating the two specific questions embedded in our path analysis using separate regression analyses: (1) how do student characteristic explanatory variables relate to self-efficacy measures and (2) how do self-efficacy measures relate to student performance (after controlling for student characteristics). While our research design uses a single instructor, which should mitigate systematic differences across classes, we also examine the robustness of our results after controlling for class effects in these regressions. Second, we examine whether these results are consistent across optimistic, mixed, and pessimistic student subsamples. Finally, we investigate potential differences between accounting and nonaccounting majors.
Regression analysis--full sample. Table 3 presents regression results regarding the relation between the directional accuracy of self-efficacy and various student characteristics (the left half of Figure 2). (23) In addition to examining the tested relations in our path analyses, regression results include indicator variables to control for systematic differences across the nine classes included in the study. The only differences between Table 3's first regression and the first half of the path analysis results are that the coefficients on NUMCLASS and EXTRACUR are not statistically significant. This suggests caution in drawing inferences regarding the relations between these variables and the directional accuracy of selfefficacy. In addition, we find that the coefficients on T5, T6, and T7 are significantly positive, suggesting that these classes, on average, are more conservative in predicting their exam scores than the baseline group, Ti, which is represented by the intercept. This could suggest either th at the student composition and class personality produced an environment that fostered less optimistic self-efficacy with respect to exam performance or that different feedback was given following the first exam in other classes.
Table 3's second regression measures the relation between student characteristics and GRADERR. While regression 2 still reports a positive relation between task-specific ability, AVEREXAM, and directional self-efficacy, the coefficients on GRADECHG and NUMCLASS are not statistically significant, again suggesting cautious interpretation of these particular significant relations in the path analyses.
The fact that none of the class indicator variables (T1-T9) is statistically significant in this regression suggests that there are no significant differences across classes in the directional accuracy of overall course performance beliefs.
Table 4 examines our hypothesis that the accuracy of self-efficacy, including direction of inaccuracy, is related to accounting-student performance, both with respect to the second exam (EX2SCQRE) and overall course performance (FINGRADE), even after controlling for student characteristics and systematic differences across classes (the right half of Figure 2). The first regression examines the relation between the accuracy of self-efficacy with respect to exam performance (EXAMERR) and the actual outcome of the exam (EX2SCORE). (24) The results are consistent with those presented in the path analysis. We find that the more conservative a student's self-efficacy beliefs with respect to the exam, the better the student's performance on that exam. In addition, we also find that two of the course indicator variables (T5 and T7) are statistically significant. This suggests that students in these two courses performed better on the second exam than did students in the baseline course. The second regression examines whether the conservatism of self-efficacy with respect to overall course grade (GRADERR) is associated with final course grade (FINGRADE). The results are completely consistent with the path analysis results. In addition, since none of the course indicator variables is significant, we conclude that there are no systematic differences, on average, across classes in overall course grades. (25)
Regression analysis--pessimistic, mixed, and optimistic subsamples. To assess the robustness of our results across the different categories of students based on performance prediction errors, we repeat the analyses reported in Table 4 for each of our prediction error subgroups. Replicating Table 4 for the pessimistic subset of students produces results consistent with those reported for the full sample (with comparable adjusted [R.sup.2] values of 0.797 and 0.792, respectively). Consistent with our combined results, we find that students with higher average exam scores tend to be more pessimistic about their performance on exams. We find only two significant differences. First, consistent with path analysis results, NUMCLASS is statistically significant relative to EX2SCORE. This suggests that pessimistic students with more experience in accounting courses tend to score better on the second exam. Second, results for the second model are consistent with our full-sample regression results except that, similar t o our path analysis results, GRADEOHG is positively significant, suggesting that pessimistic students with a positive trajectory tend to perform slightly better in terms of final grades. Replicating Table 4 for the mixed directional error subsample produces results consistent with those reported for the full sample, though the adjusted [R.sup.2] values of 0.513 and 0.781 are lower. (26) Replication of Table 4 for the overly optimistic student subsample produces results consistent with those reported tor the tull sample with adjusted [R.sup.2] values of 0.64 and 0.88, respectively. (27)
These results are consistent with the notion that self-regulatory behavior varies with the direction of prediction error. In commonsense terms, if students become overconfident due to higher average exam scores, then their optimistic prediction error increases and their course grade is lower--suggesting that they take less time to prepare. Similarly, pessimistic students, encouraged by grade improvement from the first to the second exams, experience better outcomes in course grade. This is consistent with increased preparation due to an underrated self-concept.
Regression analysis--excluding nonaccounting majors. To ensure that our results are not attributable to the sample's nonaccounting majors, we replicate Tables 3 and 4, concentrating on the 80 percent of our students who are accounting majors. The replication of our results on the (more homogeneous) subset of accounting majors merely strengthens the results reported herein. (28) Since the research is designed for classes and not intended to apply only to accounting majors, the core results presented in the Tables 3 and 4 relate to our full sample.
CONCLUSION, LIMITATIONS, AND FUTURE RESEARCH OPPORTUNITIES
We hypothesize that both the accuracy and direction of inaccuracy of self-efficacy are related to academic performance. Our results suggest that specific academic ability, as well as trajectory in the course, coincide with more conservative (that is, less optimistic or more pessimistic) self-efficacy expectations. In addition, we find that the relative conservatism of self-efficacy, holding the control variables constant, is related to exam scores and overall performance in the classroom. Likewise, the control variables of previous accomplishments in accounting and course-specific competency add to the explanatory power of a model that describes course grade. We find that pessimistic students, ceteris paribus and if encouraged by a trajectory of improvement, receive higher course grades. Optimistic students tend to increase their inaccurate exuberance regarding course grades when their average exam scores are higher. Path analysis demonstrates the indirect effect of student characteristics associated with ski lls, ability, and prior achievements on performance, through the intervening variable of self-efficacy. Consequently, we argue that more rigor in the initial exam, to the extent that it encourages self-regulatory behavior to improve by the second exam, can improve trajectory in a manner that may provide incentives for both pessimistic and optimistic students. In other words, if the first exam causes poorer students to be less optimistic and better students to be more pessimistic, then they should all perform better on the second exam. Hence, we also argue that educators should actively participate in regular feedback to enhance students' ability in order to deter overly optimistic expectations in our courses. (29)
The grade scale used in this study limits our measure of self-efficacy accuracy. (30) In addition, the non-experimental nature of our study also prevented certain variables from being directly manipulated. This constrains our findings to the reporting of associations rather than causal relationships. Our one-point-in-time assessment, as opposed to a sequential process, limits our ability to explore varied interactions or alternative paths to performance. While we propose that actions influencing averages and trajectory, along with feedback from the instructor, may enhance the accuracy of self-efficacy, and in turn performance, future research is required to test more formally such a process model. Path analysis can evaluate causal hypotheses, but cannot establish the direction of causality. If feedback loops become a part of the hypotheses, then path analysis will not be of use, since there must be a steady causal progression across a path diagram. Our work builds on previous path-analysis research in educat ion and psychology (e.g., Pajares and Miller 1997) that permits reduced-form testing of self-efficacy relative to performance. Moreover, feedback is embedded in our design in the form of the completed first exam cycle. This approach to an embedded feedback loop preceding the measures of accuracy of self-efficacy permits our use of path analysis. However, future research may wish to explore methodologies capable of a more robust treatment of feedback. (31)
Past performance and successful experience are the most influential sources of academic self-efficacy beliefs and the accuracy of those beliefs. This finding, evidenced in past research and this study, implies that students benefit when intellectually challenged and provided an opportunity to acquire necessary skills in a manner that creates self-awareness of their own achievements. The self-efficacy paradigm implies that the effects produced by the actions of others also matter. This includes, for example, the influence of an instructor who instills self-beliefs that ultimately affect academic accomplishment (Pajares 2000; Skinner and Belmont 1993). Bandura (1997) observes how teachers' evaluative reactions can influence students' judgments of their capabilities and scholastic performance. Indeed, many studies suggest that students' appraisals of their academic capabilities relate closely to teachers' judgments of them in ratings, comments, comparisons, and even subtle actions, such as setting standards, fo rming instructional groups, assigning tasks with varying difficulty, and according differential attention to certain students (e.g., Brazelton 1998). Feedback that credits attainments to personal capabilities rather than to effort appears to have more predictive effects on self-efficacy. The problem with the role of effort is that its influence on individuals relates in large part to their belief in whether ability is built through sustained effort or whether it is an inherent aptitude not under personal control.
Future Research Opportunities
Future research can explore the process of self-efficacy beyond the one-point-in-time measures used in this study. Ng (1998) uses the diary as a research tool to assess students' learning and contextual motivation, finding differential motivations by mastery goals, task value, and mastery-oriented learning climate. Since motivation is a process, an ongoing student diary could be helpful evidence. However, the diary approach proved problematic for low achievers; Ng (1998) finds the subjects who are low achievers are ineffective at maintaining their diaries. Other means by which this process can be addressed, beyond snapshots of self-efficacy, also merit exploration. Thomas and Mathieu (1994) find that the goals of college students evolve over time. Moreover, how feedback can be more informative in the university context of accounting instruction in producing accurate self-efficacy should be investigated, including the effects of(1) timeliness, (2) frequency, (3) individual and group grading, and (4) attributi on to skill and effort. Future work could compare assessments before and after an examination is administered and returned with feedback. Use of two measures could isolate the separate surprise element of a testing instrument and permit more of a process orientation in modeling feedback. Path analysis, with intermittent self-efficacy measures, could be usefully employed. The self-efficacy literature has demonstrated that instruction that provides experience in mastering a particular skill or assignment most effectively builds self-efficacy, particularly relative to instruction on self-esteem (Pajares 2000, 39). Such experience ties to pedagogy in that once a student has the experience of mastering an idea or concept, that experience can serve as a benchmark in both setting goals and aligning efforts.
Our evidence is consistent with prior research, which reports that overly optimistic students perform poorly (e.g., Carpenter et al. 1993), while adding the more unique insight that pessimistic students tend to perform well within a university context. This result should be subjected to replication, as well as sensitivity analysis. This might address situations, for example, in which pessimism turns to self-defeat-a possibility not reflected in this data set. The similarity of our results concerning a relationship between self-efficacy and performance to those reported for other domains implies the possibility of using students' self-efficacy information to identify those with inaccurate judgments. This could facilitate appropriate interventions in an effort to alter beliefs. In some cases, inaccurate self-efficacy perceptions, rather than lack of capability or skill, have been responsible for avoidance of courses and careers. While optimism can enhance, just as pessimism can defeat, either extreme essential ly distorts. Yet, our documentation of a bias in student inaccuracy suggests that when the direction of prediction error reflects underconfidence, improved performance is observed. Theory would suggest that such a result is related to self-regulatory behavior, as long as the underconfidence is not so extreme as to result in self-defeat. Theory suggests that only a realistic expectation can align efforts toward an effective performance level. Yet empirically, we observe a systematic tendency of better students to be pessimistic and poorer students to be optimistic. This is consistent with observations in past literature that positive attitudes, which are wholly unrealistic, prompt inadequate preparation and failure (Carpenter et al. 1993). Future research can pursue the extent of teacher influence on self-efficacy of students, particularly through the directional accuracy of expectations. Moreover, the finding that successful past experience is the most influential source of self-efficacy information has impli cations for self-enhancement of academic achievement (Pajares 2000, 39). In other words, we conclude that the question of how to raise competency and confidence through authentic learning experiences is an important area for focus by theorists, empirical researchers, and practitioners.
We appreciate helpful comments and suggestions from anonymous reviewers, Sue P. Ravenscroft (associate editor), and David E. Stout (editor) on this paper and on a predecessor paper related to this research, as well as Donald E. Wygal, participants at the 1999 Ohio Regional AAA Meeting and at the 1999 AAA Annual Meeting in San Diego. We especially thank Scott L. Summers for his contribution. We also thank Brian Holbrook for his capable research assistance. All data (except for the GPA variable) are available on request from the first author.
Editor's Note: This paper was received, processed, and accepted by the previous editor, David E. Stout.
Theodore E. Christensen is an Assistant Professor at Brigham Young University, Timothy J. Fogarty is a Professor at Case Western Reserve University, and Wanda A. Wallace is a Professor at the College of William & Mary.
(1.) The literature also provides similar definitions of self-efficacy and other closely related terms. For example, Bandura defines self-efficacy as individuals' beliefs, thoughts, and feelings about their personal capabilities that affect how they exercise control over their own level of functioning and, in turn, their performance (Bandura 1977, 1986, 1993). Self-referent thought or self-reflection is related to knowledge, skills, self-efficacy beliefs, and outcome expectations (Bandura 1986). Self-beliefs influence behavior by (1) affecting the choice of behavior, determining how much effort is to be expended and for how long, and (3) encouraging thought patterns and emotional behaviors. Self-beliefs help individuals to recognize that humans produce rather than foretell behavior (Bandura 1986).
(2.) We thank the associate editor for this insight and for other useful recommendations.
(3.) Keef and Roush (1997,319) define "self-concept" as the way one perceives oneself and includes self-assessments of ability, psychological traits, and physical characteristics. They argue that an academic self-concept shares strong similarities with a measure of expectations in performance in academic pursuits. Keef and Roush (1997, 323) operationally define self-concept by applying factor analysis to mathematics, verbal, and general academics scores.
(4.) We use "apparently" here because we do not specifically test this notion. While our results are consistent with this explanation, future research should specifically test whether students' self-regulatory mechanisms follow these patterns.
(5.) The term "self-efficacy beliefs" is the common verbiage in the education, counseling, and psychology literature and is used herein to properly reflect the language of the underlying theory and empirical work to which this study ties. Arguably "self-efficacy beliefs" is implicitly redundant and is adequately captured by the term "self-efficacy."
(6.) For example, one approach is for students to avoid looking at correct answers during self-study in order to avoid the "knew-it-all-along effects," which may worsen performance (Helleloid 1989, 94).
(7.) See Pajares (1994), Lent et al. (1984, 1986), Hackett and Betz (1989), Multon et al. (1991), Hackett et al. (1992), Pajares and Kranzler (1994, 1995), Pajares (1996a, 1996b), Pajares and Johnson (1996), Pajares and Valiante (1997), and Ancis and Phillips (1996).
(8.) Prior literature recognizes that if individuals are uncertain about the nature of their task, then their efficacy judgments can mislead them (Bandura 1986), which is the reasoning for eliciting students' assessments after taking the second exam but prior to feedback on performance on the second exam. This eliminates the "surprise effect" of the second exam. By delaying measurement of students' expectations until this point in the course, the students have had an opportunity to move up the learning curve regarding course content, nature of examinations, grading process, and feedback from the cycle associated with the first exam, as well as periodic feedback on daily quizzes and homework. We analogize to the concept of not using a first trial run in an experiment as the basis for evaluating the variable of interest. Moreover, prior research has found students to have little ability to predict their test performance prior to completing a test (Glenberg et al. 1987; Pressley and Ghatala 1990).
(9.) We observe the students' grades in the course directly, reducing the need to rely on students' self-reports (Pajares 1996a, 1996b).
(10.) Prior research demonstrates that different grading approaches by individual instructors can have differing effects on student performance (Ravenscroft and Buckless 1992). Moreover, Brownell and Pajares (1996) cite potential interaction problems if instructors vary in their own self-efficacy. Therefore, we restrict the treatment to a single instructor. The downside to this design choice, however, is that it decreases external validity.
(11.) Replication of our results excluding nonaccounting majors produces qualitatively similar results, as reported following our description of the full-sample results.
(12.) Schunk (1990, 1991) explains that level of cognitive ability and prior educational preparation and attainment influence academic performance alongside efficacy beliefs. Pressley et al. (1987) is likewise relevant.
(13.) Reed and Holley (1989, 327, 332) average interim test scores to represent a subject's past learning and find it to be a primary predictor of final exam grade in intermediate financial accounting.
(14.) Indicator variables T2 through T9 represent classes two through nine and are coded 1 if a student belongs to that particular class and 0 otherwise Therefore, the regression coefficients on indicator variables represent the incremental slope for a particular course relative to the excluded class, T1.
(15.) In lieu of class indicator variables T2 through T9, we also repeat our regression analysis using a single indicator variable to control for differences between universities. The results are qualitatively similar, though controlling for class differences rather than merely for which university students attend enhances descriptive power. Therefore, we retain the class indicator variables in the tables.
(16.) The second exam (as well as the first) consisted exclusively of objective questions (50 percent multiple-choice and 50 percent work-out and short-answer questions); all classes included in the sample were given identical exams. In this way, our design eliminates exam differences and subjectivity in the grading process. These attributes also maximize the base-level predictability of performance above which personal differences gain expression. Pajares and Miller (1997) specifically report that multiple-choice tests have higher scores and better calibration of students' ability in a study of 327 middle school students in mathematics. The use of numerical scores allows greater precision than can be obtained with grade levels (Tine 1998). Such scores are used for the examinations, though they must be translated to letter grades for the overall course.
(17.) The final course grade, FINGRADE (also used to compute GRADERR), corresponds to possible letter grades from A to F (scored from 4 to 0). One of the two universities uses minus or plus grades (ranging from A- to D-), and we incorporate this information by mapping A- to 3.7, B+ to 3.3, a B to 3.0, B- to 2.7, etc.
(18.) Porcano (1984) finds that feedback is related to student performance. Accuracy of feedback or truth telling by instructors is both desired by students and is important in facilitating strategic choices (Graham and Pajares 1997). In our research, accurate feedback was provided regarding the first exam. We believe that this similar information environment for the second exam strengthens our research design.
(19.) Tracking these adjustments to students' behavior, beyond observing actual outcomes, is an area for future research.
(20.) A survey of the data reveals no broad departures from the necessary assumptions for path analysis or ordinary least squares regression analysis.
(21.) We removed average exam score, AVEREXAM, from the EXAMERR model because it is so highly collinear with E2SCORE.
(22.) We can tell which of our proxies is better supported by the data by examining the larger path coefficients in the output path diagram (Figure 3), and we show that both direct and indirect relationships exist between student characteristics and performance, with the directional accuracy of self-efficacy acting as a significant intervening variable.
(23.) We cannot reject the null hypothesis of correct model specification, based on the test of first and second moment specification ("spec" test in SAS), as demonstrated by the following Chi-square statistics: Table 3 EXAMERRC model 64.01 with 64 degrees of freedom (df) (p > 0.48); GRADERR 56.59 with 63 df (p > 0.70); Table 4 EX2SCORE 55.26 with 64 df (p > 0.77); FINGRADE 57.59 with 78 df (p > 0.96). Multicollinearity is not problematic, based on condition indices, as they lie below the usual benchmark of 30 (Belsley et al. 1980). The stability of the regression models further supports that multicollinearity is not a problem, as does the resilience of the results when we use the subset of observations relating only to accounting students.
(24.) Similar to the path analysis, we also excluded average exam score, AVEREXAM, from the EXAMERR regression model due to its high collinearity with E2SCORE.
(25.) We obtain similar results to those reported in Tables 3 and 4 when we repeat our analyses using a third self-efficacy measure constructed by applying factor analysis to EXAMERR and GRADERR.
(26.) Replication of Table 3 models for the mixed student group demonstrates that students with lower GPAs, higher average exam performance, and a positive trajectory tend to be less optimistic (more pessimistic) in predicting their second exam scores. Mixed accuracy students with higher GPAs and higher average exam scores tend to be more conservative predictors of final course grade. Replication of Table 3 models produces improved adjusted [R.sup.2] values of 0.4455 and 0.2740, respectively.
(27.) Replication of Table 3 results indicates that optimistic students with higher average exam scores are significantly more optimistic in predicting exam scores than are either the pessimistic students or those with mixed direction prediction errors.
(28.) For the EXAMERR model in Table 3, all coefficients have the same direction and significance without exception, with an improvement in the adjusted [R.sup.2] to 0.2988. The GRADERR model is likewise similar to our reported results, except for an insignificant T9 indicator value, improving adjusted [R.sup.2] to 0.1988. Turning to Table 4, the EX2SCORE model has consistent signs and significance to our reported results and an adjusted [R.sup.2] of 0.7497. The FINGRADE model likewise is very similar except for the statistically insignificant GRADECHG variable, which has a coefficient of -0.0005 (i.e., the opposite sign but still a small and insignificant coefficient) and an adjusted [R.sup.2] of 0.8753. These results indicate added variability due to the inclusion of nonaccounting majors, but robustness in results. Increased measurement error from some variance in student composition is to be expected at many universities. Therefore, we feel the inclusion of these students in our main results adds external validity.
(29.) Note that the schism between academic and nonacademic pursuits is a point of debate. Whereas it may be true that extracurricular activities provide means to self-discovery and may fuel the development of important competencies, their net relationship to the academic beliefs tested in this task setting is an empirical question. Note that different attributes and characteristics are involved in each type of task, which can produce mixed results, or the fact that students extensively involved in extracurricular efforts do not have sufficient time to adapt to their self-efficacy in a manner that enhances accurate self-efficacy and outcomes (at least relative to their academic results) is another possibility. If their identity is heavily invested elsewhere, then self-efficacy of the sort that we have studied may vary as to personal priority.
(30.) The measurement scale for course performance presents methodological challenges because it constrains behavior actions for students at the ends of the distribution that can likewise affect the scale's informativeness. For example, very good students have a ceiling effect on their estimates of performance (because A students can only err on the negative side) and very poor students have a floor effect (because F students can only overstate their grades). In addition, poor students may drop the course and remove themselves from consideration. Nonetheless, a 4.0 grading scale dominates higher education and therefore provides a relevant performance metric.
(31.) Other limitations relate to potential proxy errors in our attempts to map our variables to the constructs used in the prior literature.
Ancis, J. R., and S. D. Phillips. 1996. Academic gender bias and women's behavioral agency self-efficacy. Journal of Counseling and Development 75 (2): 131-137.
Bandura, A. 1977. Social Learning Theory. Englewood Cliffs, NJ: Prentice Hall.
-----. 1986. Social Foundations of Thought and Action: A Social Cognitive Theory. Englewood Cliffs, NJ: Prentice Hall.
-----. 1993. Perceived self-efficacy in cognitive development and functioning. Educational Psychologist 28: 117-148.
-----. 1997. Self Efficacy: The Exercise of Control. New York, NY: Freeman.
Belsley, D., E. Kuh, and R. Welsch. 1980. Regression Diagnostics: Identifying Influential Data and Sources of Collinearity. New York, NY: John Wiley & Sons, Inc.
Bouffard-Bouchard, T., S. Parent, and S. Larivee. 1991. Influence of self-efficacy on self-regulation and performance among junior and senior high-school-aged students. International Journal of Behavioral Development 14: 153-164.
Brazelton, J. K. 1998. Implications for women in accounting: Some preliminary evidence regarding gender communication. Issues in Accounting Education 13 (August): 509-530.
Brownell, M. R., and F. M. Pajares. 1996. The influence of teachers' efficacy beliefs on perceived success in mainstreaming students with learning and behavior problems: A path analysis. Florida Educational Research Council Research Bulletin 27 (3, 4): 1-23.
Carpenter, V. L., S. Friar, and M. G. Lipe. 1993. Evidence on the performance of accounting students: Race, gender, and expectations. Issues in Accounting Education 8 (Spring): 1-17.
Collins, J. L. 1982. Self-efficacy and ability in achievement behavior. Paper presented at the annual meeting of the American Educational Research Association, New York, NY.
Glenberg, A. M., T. Sanocki, W. Epstein, and C. Morris. 1987. Enhancing calibration of comprehension. Journal of Experimental Psychology: General 116: 119-136.
Gobeil, J., and F. Phillips. 2001. Relating case presentation style and level of student knowledge to fact acquisition and application in accounting case analyses. Issues in Accounting Education 16 (2): 205-222.
Graham, L., G. Hackett, and N. E. Betz. 1981. A self-efficacy approach to the career development of women. Journal of Vocational Behavior 18: 326-339.
-----, and F. Pajares. 1997. The psychologizing of language arts instruction: Teachers' and students' beliefs about what it means to care. Paper presented at the Annual Meeting of the American Educational Research Association, Chicago, IL.
Hackett, G., and N. Betz. 1989. An exploration of the mathematics self-efficacy/mathematics performance correspondence. Journal of Research in Mathematics Education 20: 261-273.
-----, -----, J. Casas, and I. Rocha-Singh. 1992. Gender, ethnicity, and social cognitive factors predicting the academic achievement of students in engineering. Journal of Counseling Psychology 39: 527-538.
Helleloid, R. T. 1989. Providing answers to self-study questions: An experimental investigation of possible effects. Issues in Accounting Education 4 (1): 94-108.
Keef, S., and M. Roush. 1997. New Zealand evidence on the performance of accounting students: Race, gender and self-concept. Issues in Accounting Education 12 (Fall): 315-330.
Langenfeld, T. E., and F. Pajares. 1993. The mathematics self-efficacy scale: A validation study. Paper presented at the Annual Meeting of the American Educational Research Association, Atlanta, GA.
Lent, R. W., S. D. Brown, and K. C. Larkin. 1984. Relation of self-efficacy expectations to academic achievement and persistence. Journal of Counseling Psychology 31: 356-362.
-----, -----, and -----. 1986. Self-efficacy in the prediction of academic performance and perceived career options. Journal of Counseling Psychology 33: 265-269.
-----, and G. Hackett. 1987. Career self-efficacy: Empirical status and future directions. Journal of Vocational Behavior 30: 347-382.
-----, F. G. Lopez, and K. J. Bieschke. 1993. Predicting mathematics-related choice and success behaviors: Test of an expanded social cognitive model. Journal of Vocational Behavior 42: 223-236.
Multon, K. D., S. D. Brown, and R. W. Lent. 1991. Relation of self-efficacy beliefs to academic outcomes: A meta-analytic investigation. Journal of Counseling Psychology 38: 30-38.
Ng, C. 1998. A diary study of students' classroom learning and motivation. Paper presented at the Annual Conference of Australian Association for Research in Education. Adelaide, Australia.
Pajares, F. 1994. Inviting self-efficacy: The role of invitations in the development of confidence and competence in writing. Journal of Invitational Theory and Practice 3 (1): 5-11.
-----, and J. Kranzler. 1994. Self-efficacy, self-concept, and general mental ability in mathematical problem-solving. Florida Educational Research Council Research Bulletin 26 (1/2).
-----, and -----. 1995. Self-efficacy beliefs and general mental ability in mathematical problem-solving. Contemporary Educational Psychology 20: 426-443.
-----. 1996a. Assessing self-efficacy beliefs and academic outcomes: The case for specificity and correspondence. Paper presented at the annual meeting of the American Educational Research Association, New York, NY.
-----. 1996b. Self-efficacy beliefs in academic settings. Review of Educational Research 66 (4): 543-578.
-----, and M. J. Johnson. 1996. Self-efficacy beliefs in the writing of high school students: A path analysis. Psychology in the Schools 33: 163-175.
-----, and G. Valiante. 1996. Predictive utility and causal influence of the writing self-efficacy beliefs of elementary students. Paper presented at the annual meeting of the American Educational Research Association, New York, NY.
-----, and -----. 1997. Influence of writing self-efficacy and related beliefs about writing on the writing performance of elementary school students. Paper presented at the meeting of the American Educational Research Association, Chicago, IL.
-----, and M. D. Miller. 1997. Mathematics self-efficacy and mathematical problem solving: Implications of using different forms of assessment. Journal of Experimental Education 65 (3): 213-228.
-----. 2000. Information on self-efficacy. Available at: http://www.emory.edu/EDUCATION/mfp/effpage.html (12/11/01)--contact information for the author is provided on his web site http://www.emory.edu/EDUCATION/mfp.
Pintrich, P. R., and E. V. De Groot. 1990. Motivational and self-regulated learning components of classroom academic performance. Journal of Educational Psychology 82: 33-40.
Porcano, T. 1984. An empirical analysis of some factors affecting student performance. Journal of Accounting Education (Fall): 111-126.
Pressley, M., and J. G. Borkowski, and W. Schneider. 1987. Cognitive strategies: Good strategies users coordinate metacognition and knowledge. In Annals of Child Development, Vol. 5, edited by R. Vasta, and G. Whitehurst, 89-129. Greenwich, CT: JAI Press.
-----, and E. S. Ghatala. 1990. Self regulated learning: Monitoring learning from text. Educational Psychologist 25: 19-33.
Ravenscroft, S., and F. Buckless. 1992. The effects of grading policies and student gender on academic performance. Journal of Accounting Education 10 (1): 163-179.
Reed, S. A., and J. M. Holley. 1989. The effect of final examination scheduling on student performance. Issues in Accounting Education 4 (2): 327-344.
Relich, J. D., R. L. Debus, and R. Walker. 1986. The mediating role of attribution and self-efficacy variables for treatment effects on achievement outcomes. Contemporary Educational Psychology 11: 195-216.
Schraw, G. 1994. The effect of metacognitive knowledge on local and global monitoring. Contemporary Educational Psychology 19: 143-154.
Schunk, D. H. 1990. Introduction to the special section on motivation and efficacy. Journal of Educational Psychology 82: 3-6.
-----. 1991. Self-efficacy and academic motivation. Educational Psychologist 26 (3): 207-231.
Skinner, E. A., and M. J. Belmont. 1993. Motivation in classroom: Reciprocal effects of teacher behavior and student engagement across the school year. Journal of Educational Psychology 85 (4): 571-581.
Stone, D. N. 1994. Overconfidence in initial self-efficacy judgments: Effects on decision processes and performance. Organizational Behavior and Human Decision Processes 59 (3): 452-474.
Thomas, K., and J. Mathieu. 1994. Role of causal attributes in dynamic self-regulation and good processes. Journal of Applied Psychology 79 (6):812-818.
Trine, J. 1998. Student performance in financial accounting in relation to information processing, psychological and ability factors. Paper presented at the Midwest Accounting Society meetings, Chicago, IL.
Wenning, C. J. 2000. Effect of cooperative learning on individual perceptions of self-efficacy as problem solvers in college physics students. Working paper, Illinois State University. Please contact the author at: http://www.phy.ilstu.edu/~wenning/hmpg.html.
Wilson, T., S. Ward, and D. Ward. 1997. Empirical evidence regarding the use of self-reported student data in accounting education research. Accounting Educators' Journal 9 (Spring):50-68.
Zimmerman, B. J., A. Bandura, and M. Martinez-Pons. 1992. Self-motivation for academic attainment: The role of self-efficacy beliefs and personal goal setting. American Educational Research Journal 29: 663-676.
TABLE 1 Descriptive Statistics Panel A: Descriptive Statistics for All Students (n = 214) Standard 25th 75th Variables Mean Deviation Percentile Median Percentile EXAMERR 3.59 12.18 -4.10 4.00 11.10 GRADERR -0.07 0.58 -0.30 0.00 0.30 E2SCORE 76.24 14.44 66.00 76.00 86.00 FINGRADE 3.11 0.70 2.70 3.30 3.70 ACCTGPA 3.18 0.60 2.80 3.20 3.70 AVEREXAM 78.88 11.59 71.00 79.00 87.00 GRADECHG -5.28 12.24 -13.00 -5.00 3.00 NUMCLASS 5.05 2.62 3.00 5.00 7.00 EXTRACUR 1.07 0.64 1.00 1.00 1.50 Panel B: Descriptive Statistics by Prediction Error Subsamples Variables Prediction Error Subsample (a) n Mean Median E2SCORE Optimistic Predictions 47 68.61 68.40 E2SCORE Mixed Predictions 73 73.94 74.00 E2SCORE Pessimistic Predictions 94 81.83 80.35 FINGRADE Optimistic Predictions 47 2.66 2.70 FINGRADE Mixed Predictions 73 3.06 3.30 FINGRADE Pessimistic Predictions 94 3.38 3.30 (a)Based on whether EXAMERR and GRADERR are positive (pessimistic) or negative (optimistic). EXAMER = student's second exam score (0-100) minus the student's predicted second exam score GRADERR = student's actual final grade (0-4) minus the student's predicted final grade (including +/- grades) E2SCORE = student's score on the second exam (100-point scale) FINGRADE = final grade in the class, where A = 4, B = 3, C = 2, D = 1, and F = 0 (including +/- grades) ACCTGPA = grade point average in previous accounting classes taken (4.0 scale) AVEREXAM = student's average first and second exam score (100-point scale) GRADECHG = student's second exam score minus the student's first exam score NUMCLASS = the number of accounting classes the student has already completed; and EXTRACUR = coded 1 if the student has work experience or if the student participates in campus activities, 2 if the student has both work experience and campus involvement, and 0 otherwise. TABLE 2 Correlations among Variables (two-tailed p-values) Variables GRADERR E2SCORE FINGRADE ACCTGPA AVEREXAM GRADECHG EXAMERR 0.3557 0.3843 0.2900 0.1399 0.3446 0.2538 (0.0001) (0.0001) (0.0001) (0.0731) (0.0001) (0.0002) GRADERR 0.1931 0.4000 0.2630 0.2318 0.0161 (0.0048) (0.0001) (0.0007) (0.0007) (0.8162) E2SCORE 0.7824 0.5579 0.9123 (a) 0.6309 (0.0001) (0.0001) (0.0001) (0.0001) FINGRADE 0.7043 0.8772 0.1840 (0.0001) (0.0001) (0.0069) ACCTGPA 0.6240 0.1473 (0.0001) (0.0590) AVEREXAM 0.2578 (0.0001) GRADECHG NUMCLASS Variables NUMCLASS EXTRACUR EXAMERR 0.0192 -0.1266 (0.7800) (0.0646) GRADERR 0.3077 -0.0923 (0.0001) (0.1805) E2SCORE -0.2461 0.0718 (0.0003) (0.2956) FINGRADE 0.0228 -0.0054 (0.7407) (0.9372) ACCTGPA 0.0256 0.0655 (0.7438) (0.4035) AVEREXAM -0.0716 0.0317 (0.2969) (0.6446) GRADECHG -0.4448 0.1094 (0.0001) (0.1106) NUMCLASS -0.1388 (0.0426) (a)This correlation is high by construction since AVEREXAM is the simple average of the first two exams. EXAMERR = student's second exam score (0-100) minus the student's predicted second exam score; GRADERR = student's actual final grade (0-4) minus the student's predicted final grade (including +/-grades); E2SCORE = student's score on the second exam (100-point scale); FINGRADE = final grade in the class, where A = 4, B = 3, C = 2, D = 1, and F = 0 (including +/-grades); ACCTGPA = grade point average in previous accounting classes taken (4.0 scale) AVEREXAM = student's average first and second exam score (100-point scale); GRADECHG = student's second exam score minus the student's first exam score; NUMCLASS = the number of accounting classes the student has already completed; and EXTRACUR = coded 1 if the student has work experience or if the student participates in campus activities, 2 if the student has both work experience and campus involvement, and 0 otherwise. TABLE 3 The Relation between the Directional Accuracy of Students' Self-Efficacy and Student Characteristics (t-statistics reported in parentheses) Independent Dependent Variable Variable (Predicted Sign) EXAMERR GRADERR Intercept -27.81 (***) -1.36 (***) (-4.46) (-3.93) ACCTGPA (+) -3.17 0.14 (-1.82) (1.44) AVEREXAM (+) 0.48 (***) 0.01 (**) (5.52) (1.80) GRADECHG (+) 0.25 (***) 0.01 (3.30) (1.27) NUMCLASS (+) 0.05 0.05 (0.09) (1.40) EXTRACUR -0.69 -0.04 (-0.53) (-0.57) T2 1.83 -0.20 (0.60) (-1.21) T3 1.55 -0.22 (0.46) (-1.19) T4 -1.58 -0.17 (-0.42) (-0.82) T5 8.85 (***) -0.08 (2.63) (-0.40) T6 6.54 (**) -0.16 (2.01) (-0.88) T7 7.02 (*) 0.20 (1.72) (0.90) T8 -5.81 0.09 (-0.53) (0.14) T9 0.85 -0.03 (0.20) (-0.13) Adjusted [R.sup.2] 0.2705 0.1456 (*,**,***)Statistically significant at the 0.10, 0.05, and 0.01 levels, respectively (one-tailed test if a sign is predicted, otherwise two-tailed). EXAMERR = student's second exam score minus the student's predicted second score; GRADERR = student's actual final grade (0-4) minus the student's predicted final grade (including +/- grades); ACCTGPA = grade point average in prior accounting classes taken (4.0 scale); AVEREXAM = student's average first and second exam score (100-point scale); GRADECHG = student's second exam score minus the student's first exam score (100-point scale); NUMCLASS = the number of accounting classes the student has already completed; EXTRACUR = coded 1 if the student has work experience or if the student participates in campus activities, 2 if the student has both work experience and campus involvement, and 0 otherwise; and T2-T9 = class indicator variables. TABLE 4 The Relation between Academic Performance Measures and Directional Accuracy of Self-Efficacy Measures (t-statistics reported in parentheses) Independent Dependent Variable Variable (Predicted Sign) EX2SCORE FINGRADE Intercept 44.13 (***) -0.88(***) (10.25) (-4.54) EXAMERR (+) 0.35 (**) (5.52) GRADERR (+) 0.26 (***) (5.97) ACCTGPA (+) 11.38 (***) 0.31 (***) (9.71) (5.73) AVEREXAM (+) 0.04 (***) (14.50) GRADECHG (+) 0.49 (***) 0.00 (7.40) (0.03) NUMCLASS (+) 0.41 0.00 (0.87) (0.03) EXTRACUR 0.57 -0.02 (0.53) (-0.45) T2-T9 T5 (**), T7 (***) n.s. Adjusted [R.sup.2] 0.7012 0.8377 (**,***)Statistically significant at the 0.05 and 0.01 levels, respectively, (one-tailed test if a sign is predicted, otherwise two-tailed). n.s. Not statistically significant at conventional levels. EX2SCORE = student's second exam score; FINGRADE = student's actual final course grade (where A = 4, B = 3, C = 2, D = 1, F = 0)(including +/- grades); EXAMERR = student's second exam score minus the student's predicted second exam score (including +/- grades); GRADERR = student's actual final grade (0-4) minus the student's predicted final grade; ACCTGPA = grade point average in prior accounting classes taken (4.0 scale); AVEREXAM = student's average first and second exam score (100-point scale); GRADECHG = student's second exam score minus the student's first exam score (100-point scale); NUMCLASS = the number of accounting classes the student has already completed; EXTRACUR = coded 1 if the student has work experience or if the student participates in campus activities, 2 if the student has both work experience and campus involvement, and 0 otherwise; and T2-T9 = class indicator variables.
|Printer friendly Cite/link Email Feedback|
|Author:||Christensen, Theodore B.; Fogarty, Timothy J.; Wallace, Wanda A.|
|Publication:||Issues in Accounting Education|
|Date:||Feb 1, 2002|
|Previous Article:||Editor's Report July 1, 1998 through June 30, 2001.|
|Next Article:||Resolving difficult accounting issues: a case study in client-auditor interaction.|