Determinants of Study Length in Higher
Education
An Empirical Study on Dutch Data
July 2014
Qin Yan
Bachelor Thesis Economics
Supervisor: MSc Francisco Gomez Martinez
University of Amsterdam, Faculty of Economics and Business
Abstract
Recent debates in the Netherlands about the ‘long study fine’, a law clause introduced by the government penalizing students for taking too long to attain their academic degree, were circulating along with a growing concern resulting from an OECD report, which stated that Dutch students were among the slowest to graduate compared to their peers in other industrialized countries. Attention therefore is drawn to the origins of the problem. What possibly determines how long it will take for students to attain a degree? This study’s primary purpose is to locate these determinants and to assess the extend to which they contribute to study length. Based on an extensive literature review, six key factors were found to affect time until completion of academic programs: ethnicity, gender, high school GPA, income of parents, time spent on part-time jobs and effort. Empirical analysis then were applied based on data retrieved from a survey conducted among students from the two Amsterdam universities (UvA and VU). After testing for
multicollinearity using Pearson correlation coefficients and variance inflation factors (VIFs), no significant correlations among the independent variables were detected. Using OLS-regression method an estimation of the effects of the cited variables on years until graduation (independent variable) was made. The model in general was found to be
significant, as well as the coefficients of the variables ethnicity, gender, high school GPA, income of parents and effort. Results were largely in line with expectations gained from prior studies in this field. Asian and western students seemed to attain their degree faster than African and Latin-American students. Higher high school GPA, higher income of parents and higher level of effort put into study all appeared to have an adverse impact on study length, supporting earlier literature where it is shown that they had a positive relation to academic performance. Furthermore, it is worth remarking that this study reported a clear male dominance in terms of degree attainment, whereas previous researches on academic performance were still inconclusive about the gender gap. We also found that time spent on part-time jobs did not significantly alter study length.
Reason are unclear, but a couple of potential explanations were stated. In the final section, implications of the empirical results are evaluated and suggestions for further research are given.
Introduction
Long study duration by students in higher education have led to debates in the Netherlands to whether undergraduates should be penalized for not completing their bachelor and master programs in a predetermined period of time. Advocates of the ‘long study fine’ (A law clause introduced in 2011 and abolished in 2012, according to which undergraduates had to pay 3000 euro’s on top of their tuition fee yearly when they take longer than the appropriate amount of time determined by law to complete an academic program) refer to the fact that according to an OECD report (Education at a glance, 2011), students at Dutch universities require on average 5.02 years to attain a tertiary degree, whereas the overall average of the 34 OECD-countries lies at 3.94 years. This significant gap could potentially become a burden for the Dutch social benefit system as
considerable parts of government budget are dedicated to foster a stimulating learning environment and accessibility in higher education.
In addition, academic attainment is commonly recognized as a fundamental component of human capital that is formed during adolescence and early adulthood. With the increasing concern over the ever expanding income gap and the polarization of the labor market that has materialized along with the progresses of information technologies, it is quite instinctive to be paying more attention to how human capital, a paramount determinant of income, is formed (Hojo, 2012).
So, how can there be such a striking delay in attaining a tertiary degree among students in Dutch universities? The driving factors behind length of study as well as the magnitudes of their influence should be determined and commonly learned. Any advancement in our knowledge about these determinants could substantially enrich our abilities in locating potential flaws of the educational system and/or unfavorable personal characteristics of undergraduates. In addition, recent studies have shown that academic attainment is crucial in comprising human capital (Heckman, 2006; Kohara & Ohtake, 2009), which in turn is affecting income. Therefore, understanding the determinants of academic attainment is also essential in contributing to solutions of the income inequality issue.
This study aims to identify the key factors affecting study length of undergraduates in Dutch universities and their magnitudes. The first part shall be
accomplished by reviewing existing literature, whereby the most important factors will be selected and adjusted regarding the situation in the Netherlands. Unfortunately, as this issue is quite contemporary, there is not much literature to be found on that exact subject. To circumvent this obstacle, we can use publications on determinants of academic
performance instead. The correlation between the two should be very high for a couple of reasons. First of all, high academic performance implies high intelligence, better task management, organizing skills and planning abilities. All of which are cardinal elements of completing an academic program within the prespecified amount of time. Furthermore, Dutch universities follow the European Credit Transfer System (ECTS), which awards students with study credits when they perform above a certain degree (grade). Higher academic performance would suggest satisfactory performances in the offered courses
and thus will result in shorter time span needed till completion of academic programs. As for the second part, a regression analysis using the key factors obtained from part one will be carried out based on data of the two universities in Amsterdam (UvA and VU),
retrieved through a survey conducted by students of the listed universities. Finally, the empirical results will be reported and a conclusion will be provided.
Literature Review
As clarified in the introduction, we are looking for determinants for study length in years of students pursuing an academic degree at an university. It is also mentioned that due to the recentness of our issue, there are little resources available and instead previous studies in the field of academic performance would make a good replacement for the reasons mentioned above. This being said, study length and performance are undoubtedly not the same thing, therefore not all the factors provided by preexisting studies are suitable for this research and we might add other factors along the way besides the ones we find in the literature about academic performances for supplementary purposes. We will keep that in mind as we proceed.
In the past few decades, academic performance of students in higher education has been thoroughly investigated and many researches on that issue were published. We stumbled upon a wide range of predictors for performance. McKenzie and Schweizer (2001) were the first to categorize these predictors into mainly four groups: academic, psychological, cognitive and demographic.
Studies on academic predictors were mainly concerned with prior academic performances, such as high school GPA. A large percentage of studies conducted within this field confirmed the positive relationship between high school GPA and academic performance in university (Lenning, 1975; Noble 1991; Cohn et al, 2004). Other studies on academic predictors found that learning skills and learning strategies were also explanatory for academic performance (Abbott-Chapman, Hughes, & Wyld, 1992; Sadler-Smith,1996). The most noteworthy finding concerning learning strategies was contributed by Pintrich (1986). After conducting a survey and reporting the results he suggested that effort was the only direct significant explanatory variable for performance. However, counter arguments were also found in the literature. The study of Plant et al.
(2005) showed that time spent studying had an adverse relationship with performance. With these opposing conclusions it’s captivating to include effort in our own research to acquire more information on its effect.
As for psychological factors, family background, financial status, social and emotional support by family and friends and motivation were found to be the most relevant factors. Gonzales et al. (1996) examined the effect of family background and neighborhood on performances of African American high school students. They found that family background was much less predictive of school performance than
neighborhood. A reason for this result could be economic segregation of people (Wilson, 1987). Highly educated parents are likely earning a high income. High income
households are more likely to live in an environment surrounded by households with similar levels of education and income. This should potentially affect students
performances at school. Indeed, in an U.S. nationwide survey of 17,000 school districts, Parish, Matsumoto and Fowler (1995) found that richer neighborhoods are strongly correlated to greater school related expenses per student and their performances.
Furthermore, the neighborhood a student lives in also greatly influences the cohort with whom he/she spends time and goes to school with, which also could be a factor of concern regarding academic performance. At this point, we should recall the fact that we are investigating determinants of study length and it might be a good time to introduce an additional predictor variable suited for our research. Part-time jobs are very common among average Dutch undergraduates, which not necessarily have great correlations with family background, but rather with the students own financial status. One would imagine that the more time a student spends on part-time jobs the less time is left to put an effort on studying and thus could have consequences for the amount of time he/she needs to complete the academic program. Although motivation is a widely used personal variable in psychological studies, due to level of difficulty to obtain accurate survey
measurements on that matter, we will leave this factor out of the equation.
According to Chen, Jing-Lin and Li (2010), the cognitive appraisal studies fall mainly into two streams: self-efficacy and attributional style. Self-efficacy refers to the students’ self-beliefs about their capabilities to initiate and successfully perform specific tasks at designated levels (Pajares, 1996). Attributional style means “the general tendency
of an individual to generate similar causal explanations across events” (Yee, Pierce, Ptacek, & Modzelesky, 2004, p. 359). These are both psychological concepts and we shall not go any deeper into them. For the same reason as motivation, the cognitive predictors will also not be part of this research.
In contrast to the other predictors, the last group, demographic features are known to have inconclusive relations with academic performance. Take for example the main demographic features age and gender. Even though most studies show a male advantage in exact science courses like mathematics (Anderssen, Benjamin, & Fuss, 1994), some studies found no significant gender dependence (Rhine, 1989), and others even found a female advantage in the same courses (Williams, Waldauer, & Duggal, 1992). The same is true for age as well. McInnis et al. (1995) found that students tend to perform better with age, whereas Clark and Ramsay (1990) found the opposite. With respect to this study age should not be a significant factor of concern, as the data sample consists mainly of students of similar ages, but it should be quite interesting to examine whether males or females outperform the other in a sample with several distinctive majors. Another
demographic we might want to add is ethnicity. Definition of ethnicity might cause ambiguity, as classification of an ethnic group tends to be associated with historical background, shared cultural heritage, language and many other aspects. Based on one’s conceptual perception, members of certain ethnic groups might be designated elsewhere depending on the source consulted. That is the reason why a lion’s share of studies do not use specific labeling of ethnic groups but instead refer to natives and minorities only. Although most researches were evident in their conclusions that natives outperformed minorities (e.g. Cooper, Baron, & Lowe, 1975), but as the majority of studies were conducted in the U.S. and a Dutch sample might not consequently produce the same outcome, attention devoted to this variable might enhance knowledge in this field in Europe.
In the thick of the mentioned predictors above, quite a few could potentially be intercorrelated. For example time spent on part-time jobs and financial status are likely to interact closely. It is straightforward to grasp that students with less financial resources to their disposal are more likely to actively seek part-time jobs and vice versa. As such, the effect of one predictor might be indirectly reflected in the effect of the other. Therefore,
great care should be taken to adopt the key factors for this study. Hence, based on empirical testing regarding multicollinearity later in the text, some predictors might be removed.
Research Methodology
The aim of this study, as mentioned in the introduction, is to gain more information on the determinants of study length among undergraduates in universities. In particular, Dutch universities are the subjects of concern as their performances in the recent years are perceived to be mediocre at best and provided an impulse for this study in the first place.
Based on preexisting literature, we adopted six key predictors of study length: 1.Ethnicity
2.Gender
3.High school GPA 4.Income of parents
5.Time spent on part-time jobs 6.Self-perceived amount of effort
The study used quantitative research methods (OLS-regressions) based on a survey conducted among UvA and VU students in 2014. Participation was entirely voluntary and was made clear to students in advance. Students did not need to provide their names, student-codes or any kind of identification. The time allotted for answering the questions is eight minutes. Furthermore, only students who have already completed a bachelor program or students who are in the middle of their bachelor thesis (completed all required courses) were approached. Also, to circumvent complications of potential gap-years taken by participants which would plainly bias the results, we made it explicitly apparent in the questionnaire that only years officially registered at the university should be incorporated in their answers. Moreover, income of parents were cited in intervals of 0-10.000, 10.000-20.000, 20.000-40.000, 40.000-60.000 and more than 60.000 euro’s. Likewise, time spent on part-time jobs was also cited in ranges of hours (0-5h, 5-12h, 12-20h and more than 20h per week). Self-perceived amount of effort was answered in a
scale from one to five, from low to high. As for ethnicity, in above section the
problematic approach of assigning people to an ethnic group has already been illustrated. Although by choosing to divide our sample into just two groups (natives and minorities) would mitigate the matter, we also would give up valuable extra insights of cultural effects on study length. By virtue of the limited scope of this study however, it is impossible to impart statistical meaning to each single group since the sample sizes associated with individual groups were simply too small. Hence, we settled on
differentiating between Western (European and North-American), Eastern (Asian) and Southern (African and Latin-American) groups only.
Based on data described above, we estimated a production function of study length using the OLS-method. The function takes on the following form:
1 2 3 4 5 6
YEAR= +C βETH +β GEN+β HSG+β IP+β PTJ +β EFT +ε
Where C is the estimated intercept, the betas are de coefficients of the corresponding variables and ε is the error-term.
Empirical Analysis
Descriptive statistics
In total, the sample consisted of 209 valid observations. Among them, 103 were females (49.28%) and 106 were males (50.72%) (see table 1). Ethnic groups were distributed as given in table 2.
Table 1. Sample Gender Composition
Gender #Observations Percentage
Male 106 50.72
Female 103 49.28
Table 2. Sample Ethnicity Composition
Ethnicity #Observations Percentage
Western 144 68.90
Eastern 30 14.35
Southern 35 16.75
Total 209 100.00
Table 3 shows the basic descriptions of the other key independent variables. In order to incorporate the ordinal nature of the variables income of parents and time spent on part-time jobs into the analysis, a ranking was assigned to each interval. For instance, an income of less than 10.000 corresponds to 1 and an income between 10.000 and 20.000 corresponds to 2, etc. When we examine the means we see that the average income of parents is 2.871, which approximates the third interval of 20.000-40.000. Average time spent on part-time jobs is 2.263, indicating that students spend about 5-12 hours on average per week working. Furthermore, the average high school GPA of the subjects is 6.868 and their self-perceived level of effort is 3.081. Glancing at the Pearson product-moment correlation coefficients between the key variables, we noticed no dissatisfactory high levels (>.35). Another point worth noticing is that among the used variables,
correlations between income of parents and high school GPA were relatively high. This seems quite in accordance with our expectations based on rational deliberation and earlier findings in the literature. In addition to simple correlation between the independent variables, the variance inflation factors (VIFs) were also computed. According to Myers (1990), VIF is a very reliable complementary diagnostic when it comes to tracing multicollinearity and on its own it is a considerably more productive approach than the simple Pearson correlation. The VIFs are shown in table 4. Ranging at around 1. Under the generally accepted guideline of VIF less than 10 (Myers, 1990), this suggests that all
key variables were suited for regression analysis and no eliminations of predictors were needed.
Regression Analysis with Multiple Variables
The OLS-regression method is the most widely used tool in determining relationships between a certain phenomenon and its potential predictors. This study applies this methodology to estimate significance and magnitude of the various factors mentioned above in predicting the study length of an university student. The independent variables of concern were ethnicity, gender, high school GPA, income of parents, time spent on part-time jobs and self-perceived level of effort. The dependent variable is the length of study period per student until attainment of bachelor degree in years. Results are
presented in table 5.
Table 3. Descriptives and Pearson Correlation of Key Variables
Variable Mean SD Income
parents
Time on
part-time job Effort
HS GPA 6.868 1.141 Income parents 2.871 1.113 0.229 Time on part-time job 2.263 1.080 -0.040 -0.184 Effort 3.081 1.347 0.126 0.058 -0.068
Table 4. VIFs of Key Variables
Variable VIF
HS GPA 1.09
Income Parents 1.17 Time on part-time job 1.05
Table 5 shows the results of the regression using the variable gender as dummy variable (female was coded 0 and male 1). In addition, the variable ethnicity was relabeled a categorical variable, where the betas of Eastern ethnicity and Southern ethnicity are a result of comparison between the cited ethnicities and the Western ethnicity (control group). First, when looking at overall performance of the model, we notice that the F-score is 23.82, indicating that the model as a whole is sufficiently significant. The R-squared of 0.4535 and the adjusted R-R-squared of 0.4344 suggest moderate explanatory power of the model, since only 45.35% or 43.44% respectively of the variance in study length can be disclosed from the variances of the listed independent variables. Even though the R-squared and adjusted R-squared may not be considered extremely high, they are moderately deemed adequate for the purpose of this very study, which is to evaluate some key determinants of study length and not the prediction of it. Next, we examine the individual determinants.
Table 5. Regression Analysis Summary
Variable Coefficient SE P-value
Ethn. Eastern -.4486209 .1755661 0.011 Ethn. Southern .7064862 .1608909 0.000 Gender -.6316016 .1194763 0.000 High school GPA -.2654599 .0537332 0.000 Income parents -.2492355 .0569527 0.000 Time on part-time jobs .0640437 .0555727 0.251 Effort -.0781562 .0442298 0.079 Intercept = 6.958518 .41566 0.000 #observations = 209 F-score = 23.82 R-squared = 0.4535 Adj R-squared = 0.4344
Ethnicity
As mentioned in the above section, the betas corresponding to the ethnicities Eastern and Southern should be interpreted as the effect of being a member of the cited ethnicity groups on study length compared with the effect of being a member of the Western ethnicity group (control group) on study length. Table 5 shows that the p-values for both Eastern and Southern group are rather low, 0.011 and 0.000 respectively, suggesting that both coefficients are statistically significant at the 5% level of significance. By examining the sign of the coefficients, we further learn that students from the Eastern ethnic group use less time to complete study programs than their Western peers, whereas the opposite is true for the Southern ethnic group. This result supports earlier findings in the U.S. that African-American students are structurally outperformed by their native-American and Asian counterparts.
Gender
Table 5 also reports that the variable gender is significant at the 5% significance level with a p-value of 0.000. Recall from the general assessment of the model section that males were coded with 1 and females with 0. The negative sign of the coefficient
therefore indicates that according to the analysis being a male notably shortens the length of study. This finding thoroughly confirms the conception of male dominance in
academic attainment, whereas existing literature was still inconclusive on that matter.
High School GPA
Also for the variable high school GPA, our analysis shows a statistically significant relationship at a 5% significance-level with p-value of 0.000. The negative regression coefficient for this variable signals the adverse correlation between high school GPA and study length. Everything else held equal, higher high school GPA predicts shorter period to degree attainment. This is greatly in line with published findings of performance in higher education being positively dependent upon prior academic achievements.
Income of Parents
In the literature review section, the importance of family background and its relationship with the environment of development and its impact on academic performance was discussed extensively. Eventually, we deemed income of parents as the easiest to measure and most adequate representation of the family background factor. Based on evidence in the literature, we expected a negative relationship between income of parents and study length. Indeed, the regression analysis confirmed the hypothesis with a negative
coefficient for that variable and a corresponding p-value of 0.000 which implied statistical significance.
Time spent on part-time jobs
Hinging on rational deliberation, we would expect length of study to widen with more time spent on jobs, since consequently less time would be available for scholarly
purposes. Table 5 shows a positive coefficient for this determinant, certifying our view. However, looking at the p-value of 0.251, its statistical significance cannot be warranted. Meaning there is not enough evidence that the effect of time spent on part-time jobs on study length is significantly different from zero. Possible explanation could be that time available to students in our sample pool is thus excessive that spending more time on working does not alter the amount of time he/she would use learning. Another cause for this result might be that working more hours could possibly benefit the learning process by improving the students ability to, for example, organize and cooperate, thereby offsetting the negative effects.
Effort
The regression coefficients associated with the variable effort have a p-value of 0.079, as is shown by table 5. They are therefore not significant at the 5% significance level, but if we would adopt a less strict significance level (10%), they would be significant. This indicates that some statistical connotation should be given to effort when assessing its link to the dependent variable. The sign of the coefficient predicts a negative relationship between effort and study length, which coincides with the expectation that by putting in more effort, the time till attainment would be shortened.
Summary of findings and side notes for interpretation
From the above analysis, it can be postulated that among the six independent variables included in this study, four of them (ethnicity, gender, high school GPA and income of parents) significantly affect study length at the 5% significance level. Effort was significant at the 10% level, while insignificant at the 5% level. Time on part-time jobs did not have an significant effect on study length in this study. Regression results were roughly in accordance with suggestions from the literature.
Out of the significant variables, gender appeared to be the most important predictor, whereas most previous studies in this field found prior academic achievement (high school GPA) to be the most influential determinant. Though it should be stated that the dependent variable in those studies was academic performance rather than study length. Thus slight differences in regression results were expected beforehand.
Furthermore, because of the ordinal nature of the variables time spent on part-time jobs and parents income, less statistical importance should be devoted to the
magnitudes of the coefficients corresponding to those determinants. Rather the sign of the coefficients should receive our attention.
Finally, it should be made clear that the scope of the study is highly limited by its sample: the subjects were drawn from two relatively similar institutions in Amsterdam. Making it difficult to generalize these findings to other areas of the world, Europe or even the Netherlands. Caution should be borne in mind when interpreting these results.
Conclusion
This study has examined the effects of various socio-economic and academic factors on the time needed for bachelor degree attainment of students in the two universities of Amsterdam, UvA and VU. The selected determinants, gender, ethnicity, high school GPA, time spent on part-time jobs, income of parents and self-perceived level of effort showed no significant correlation with each other when tested for multicollinearity. Empirical analyses were then conducted using the OLS regression method with data obtained through a survey questionnaire among the students of the cited universities. Several general conclusions stand out. First, the model in its whole was found statistically
significant with reasonable explanatory power. Second, all estimated coefficients of the determinants were statistically significant, except for time spent on part-time jobs, which were found to be not significantly different from zero. Exact reasoning behind it is still unclear, but potential causes could be excess spare time of the students and the bilateral effects of working which might even out benefits and losses. As for the other individual factors, gender seemed to have the most dominant effect in determining study length. Male students were predicted to need 0.63 years less than an otherwise identical female, while the literature was predominantly inconclusive. Previous academic achievement (high school GPA), on the other hand, showed a relatively smaller effect than was predicted by earlier studies. The empirical analyses also reported a significant difference in study length caused by ethnicity. It appeared that compared to western students, Asian students were faster in attaining their degree, whereas the opposite is true for African and Latin-American students. Moreover, the results indicated that family background also played a prominent role in determining study length. Higher income of parents appeared to be advantageous for undergraduates in completing their study program. Overall, the outcomes obtained from this study were broadly supportive of previous findings and produced useful insights, but at the same time great care should be taken in interpreting these results. Inevitably, the limited scope of the sample makes generalizations difficult. In addition, statistical meaning of several determinants were somewhat weakened by cause of their ordinal nature. Another limitation of his study was lacking of specific literature, as inferences were made building on previous works on academic performance instead of study length. Consequently, assumptions about the determinants of study length based on that literature might be slightly off and incomplete.
As suggestion for further research, it might be worthwhile to take samples from different areas and institutions for a more general picture. Also a refinement of statistical
techniques dealing with ordinal data in order to increase further understanding would be desirable. Finally, progresses in this field should lead to more efficiency in the
educational system by using our improved knowledge to enhance policy guidance. While today’s academic and socio-economic environments have drastically advanced compared to a couple of decades ago, much room is still left for improvement.
References
Abbott-Chapman, J., Hughes, P., & Wyld, C. (1992). Monitoring student progress: A
framework for improving student performance and reducing attrition in higher education. Hobart, Tasmania: National Clearinghouse for Youth Studies.
Anderson, B., Benjamin, H., & Fuss, M. A. (1994). The determinants of success in university introductory economics courses. Journal of Economic Education, 25, pp. 99-119.
Chen, W., Jing-Lin, D., & Li, G. (2010). Determinants of International Students’
Academic Performance: A Comparison Between Chinese and other International Students. Journal of Studies in International Education, 14, pp. 26-35.
Clark, E. E., & Ramsay, W. (1990). Problems of retention in tertiary education.
Education Research and Perspectives, 17, pp. 47-57.
Cooper, H. M., Baron, R. M., & Lowe, C. A. (1975). The importance of Race and Social Class Information in the Formation of Expectancies About Academic
Performance. Journal of Educational Psychology, 67 (2), pp. 312-319.
Gonzales, N . A., Cauce, A . M., Friedman, R. J., & Mason, C . A. (1996). Family, peer, and neighborhood influences on academic achievement among African-American adolescents: One-year prospective effects. American Journal of Community
Psy-chology, 24(3), pp. 365-387.
Heckman, J. (2006). Skill Formation and the Economics of Investing in Disadvantaged Children. Science, 312, pp. 1900-1902.
Hojo, M. (2012). Determinants of Academic Performance in Japan. The Japanese
Economy, 39-3, pp. 3-29.
Kohara, M., & Ohtake, F. (2009). Komodo no kyouiku seika no kettei youin
(Determinants Of the Educational Performance of Children). Nihon roudou kenyu
zasshi, 588, pp. 67-84.
Lenning, O. T. (1975). Predictive validity of the ACT tests at selective colleges. Report No. 69 [050269000]. Iowa City, IA: American College Testing.
McInnis, C., James, R., & McNaught, C. (1995). First Year on Campus: Diversity in the
Initial Experiences of Australia Undergraduates. Melbourne: University of
Melbourne Press.
McKenzie, K., & Schweitzer, R. (2001). Who succeeds at university? Factors predicting academic performance in first year Australian university students. Higher
Education Research and Development, 20, pp. 21-33.
Noble, J. P. (1991). Predicting college grades from ACT assessment scores and high
school course work and grade information. Report No. 91-3 [50291930]. Iowa
City, IA: American College Testing.
OECD (2011), Education at a Glance 2011: OECD Indicators, OECD Publishing. http://dx.doi.org/10.1787/eag-2011-en.
Pajares, F. (1996). Self-efficacy beliefs and mathematical problem-solving of gifted students. Contemporary Educational Psychology, 21, pp. 325-344.
Parrish, T. B., Matsumoto, C. S., & Fowler, W. J., Jr. (1995). Disparities in public school
district spending 1989-90: A multivariate, student-weighted analysis, adjusted for differences in geographic cost of living and student need. Washington, DC:
National Center for Education Statistics, U .S. Department of Education. Pintrich, P. (1986, July). Anxiety, motivated strategies and student learning. Paper
presented at the International Congress of Applied Psychology, Jerusalem, Israel. Plant, E. A., Ericsson, K. A., Hill, L., & Asberg, K. (2005). Why study time does not
predict grade point average across college students: Implications of deliberate practice for academic performance. Contemporary Educational Psychology, 30, pp. 96-116.
Rhine, S. L. (1989). The effect of state mandates on student performance. American
Economic Review, 79, pp. 231-235.
Sadler-Smith, E. (1996). Approaches to studying: Age, gender and academic performance.
Educational Studies, 22, pp. 367-380.
Williams, M. L., Waldauer, C., & Duggal, V. G. (1992). Gender differences in economic knowledge: An extension of the analysis. Journal of Economic Education, 23, pp. 219-231.
Wilson, W. J. (1987). The hidden agenda. In W. J. Wilson (Ed.), The truly disadvantaged:
The inner city, the underclass and public policy, pp. 140-164. Chicago:
University of Chicago Press.
Yee, P. L., Pierce, G. R., Ptacek, J. T., & Modzelesky, K. L. (2004). Learner helplessness Attributional style and examination performance: Enhancement effects are not necessarily moderated by prior failure. Anxiety, Stress, and Coping, 16, pp.359-373.
