a quantitative exploration of student

19
Accounting Education 12 (1). 15-32 (2003) A quantitative exploration of student performance on an undergraduate accounting programme of study LOUISE GRACIA'- and ELLIS JENKINS IJntvsrsity of Glamorgan. S. Wales, UK Received: September 2001 Prevised: May 2002: Septembei- 2002 Accepted- Octobc>' 2002 Abstract This paper explores tmdergraduate performance at second and final year levels on a degree programme in Accounting and Finance at a Welsh university using a blend of data from demographic, attitudinal and behavioural sources. It considers issues including gender, prior yeai" performance and students' application to study based upon classroom observation, and makes use of multiple regres- sion and principal components analysis. Eindings appear to highlight that both prior year results and apphcation to studies are positively associated with performance as measured by grade points. .A,t second year level, gender is also positively associated with performance whilst age is negatively asso- ciated. The findings with regard to gender and age are not repeated in the final year but the primai-y rtinking of this university by students is positively associated with performance in that final year, pnailv. there is atso some evidence that students who undertake a year of supervised work experience do better m the final year Finally, the implications of the empincai research are discussed amd further research is outlined m terms of undertaking a longitudinal study based upon application and the adop- tion of qualitative approaches to explonng the reasons for differences in undergraduate performances. Ke\n'onls: academic performance, accounting education, behavioural observation, application, multiple regression Introduction Academic success is of primary-- importance to students and is also significant for the staff and the higher education insdtutions (HEl) within which they study. Academic failure creates emotional and finaticial costs for individual students. Tlieir HEI may also suffer as its retendon rates fall, adversely lmpacdng on its league table placing with consequential finan- cial penaldes. First, failure reduces the number of students progressing through the under- graduate programme which directly reduces the head count on which fees and gi^ants may be claimed; secondly, withdi-awal during tlie academic year reduces the head count for that year. [Tliis paper explores undergraduate academic performance on the second and final years of study. It adopts a quantitative approach and attempts to identify those factors that appear to be most closely associated with success and failure in the formal assessments undertaken by Accounting and Finance degree students at a Welsh HEI in the academic years 1999/2000 and 2000/01 ^ Address for correspondence: Mrs. Lomse Gracia, Business School. L'nrversity of Glamorgan. Llantvvil Road. Trefores:, S. Wales. CF37 lDL. UK. E-mai; igraciaS'glam ac uk Accouming Education ISSN 0963-9234 prinLiSSX 1468-4489 online © 2003 Taylor & Francis Ltd http //www tandf co.uk/joumals DOI 10 1080/0963928032000049375

Upload: mellz-cheery

Post on 26-May-2017

216 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: A Quantitative Exploration of Student

Accounting Education 12 (1). 15-32 (2003)

A quantitative exploration of studentperformance on an undergraduateaccounting programme of studyLOUISE GRACIA'- and ELLIS JENKINS

IJntvsrsity of Glamorgan. S. Wales, UK

Received: September 2001Prevised: May 2002: Septembei- 2002Accepted- Octobc>' 2002

Abstract

This paper explores tmdergraduate performance at second and final year levels on a degreeprogramme in Accounting and Finance at a Welsh university using a blend of data from demographic,attitudinal and behavioural sources. It considers issues including gender, prior yeai" performance andstudents' application to study based upon classroom observation, and makes use of multiple regres-sion and principal components analysis. Eindings appear to highlight that both prior year results andapphcation to studies are positively associated with performance as measured by grade points. .A,tsecond year level, gender is also positively associated with performance whilst age is negatively asso-ciated. The findings with regard to gender and age are not repeated in the final year but the primai-yrtinking of this university by students is positively associated with performance in that final year,pnailv. there is atso some evidence that students who undertake a year of supervised work experiencedo better m the final year Finally, the implications of the empincai research are discussed amd furtherresearch is outlined m terms of undertaking a longitudinal study based upon application and the adop-tion of qualitative approaches to explonng the reasons for differences in undergraduate performances.

Ke\n'onls: academic performance, accounting education, behavioural observation, application,multiple regression

Introduction

Academic success is of primary-- importance to students and is also significant for the staffand the higher education insdtutions (HEl) within which they study. Academic failurecreates emotional and finaticial costs for individual students. Tlieir HEI may also suffer as itsretendon rates fall, adversely lmpacdng on its league table placing with consequential finan-cial penaldes. First, failure reduces the number of students progressing through the under-graduate programme which directly reduces the head count on which fees and gi ants may beclaimed; secondly, withdi-awal during tlie academic year reduces the head count for that year.

[Tliis paper explores undergraduate academic performance on the second and final yearsof study. It adopts a quantitative approach and attempts to identify those factors that appearto be most closely associated with success and failure in the formal assessments undertakenby Accounting and Finance degree students at a Welsh HEI in the academic years 1999/2000and 2000/01

^ Address for correspondence: Mrs. Lomse Gracia, Business School. L'nrversity of Glamorgan. Llantvvil Road.Trefores:, S. Wales. CF37 lDL. UK. E-mai; igraciaS'glam ac uk

Accouming EducationISSN 0963-9234 prinLiSSX 1468-4489 online © 2003 Taylor & Francis Ltd

http //www tandf co.uk/joumalsDOI 10 1080/0963928032000049375

Page 2: A Quantitative Exploration of Student

1" Gracia and Jenkins

Literature review

The aim of the research is to explore the reasons for differences in student performance. Thisaim may be related to a body of literature in the journals deahng with higher educadongenerally and accounting education specifically.

A number of studies have examined the impact of prior accounting knowledge on aca-demic performance in first year university accoundng studies. Baldwin and Howe (1982)found no difference between the overall performance of students irrespective of their levelof prior accoundng knowledge. Bergin (1983) confirms this finding, concluding that overallcourse performance is tmrelated to prior high school accounting exposure. Schroeder (1986)also confirms this finding. Keef (1992) reports on studies in Austraha and New Zealand thatexamine the stadstical reladonship between prior accounting educadon and the performancein first year undergraduate courses. He reports a range of outcomes extending from noreladonship to a posidve relationship that is stadsdcally valid.

In the UK, Bartlett et al. (1992.1993) carried out a study of students speciahzing inaccoundng at the University of Wales College of Cardiff over the period 1989-1992. Theycarried out tests of the relationship between a large range of explanatory vanables includ-ing, age, gender, region, prior study, and educational attributes. They found that the onlybackground charactenstic that appeared to have a significant effect on performance was theprior study of Economics at Advanced level (externally set examinations taken at the age of18 + and that help determine admission to university) and that die best predictor of third yearperformance was, in fact, performance in the first yeai" examinations.

However, in an earKer study in the UK, Mitchell (1985.1988) found that students widiprior accounting or Advanced level mathemadcs performed statisdcaUy better in the com-putational and quantitadve aspects of accounting courses. Gul and Fong (1993) examinedthe factors that affept the performance of first yeai- undergraduate accoundng students inHong Kong. They found a range of significant predictors of performance including person-ality type, previous knowledge of accoundng and school cerdficate grades in accoundng andmathematics. Similai' results by Tho (1994) found the pnor study of high school accounting,mathematics and grades achieved in high school economics to be significant and positivelyassociated with academic performance. The predictor value of the prior study of accountingand economics is also consistent with the findings of Eskew and Faley (1988), Farley andRamsey (1988). However, Doran et at (1991) produce conflicting conclusions concerningthe impact of prior accoundng knowledge on academic performance in two accountingcourses {JAPI and APII). They found the prior study of high school bookkeeping to bepositively related to performance in API but negadvely related to performance in APII.

In the USA there have been a number of studies of the impact of gender. Fraser et al.(1978) and Mutchler et al (1987) fotmd that female students outperformed their male coun-terparts. Tyson (1989) supports this view and concludes that female students outperformmale students in all secdons of an introductory accoundng class. However, Lipe (1989) failedto find any evidence of a gender effect whilst the smdy by Doran et al. (1991) found con-fiicdng restilts; men outperform women in API but not in APII. Carpenter et al. (1993) foundno gender differences in the dropout from introductory courses but found evidence of sig-nificant under-performance by students from racial minority groups. In an Australian study,Auyeung and Sands (1994) invesdgated gender differences in different deep/surface leamingcomponents of first year examinations. They found no difference. However, in contrast. Kohand Koh (1998) in a study in Singapore found gender, previous working experience,academic aptitude and age to be significant posidve predictors of academic perfonnance.

Page 3: A Quantitative Exploration of Student

Quantitative exploration of student petformance 17

Naser and Peel (1998) report on a study at Beimt Universit>'. They found that studentperceptions of factors associated wilh class size, the attributes of the teacher, student studyindependence and the complexity- of tlie course aie associated with student perfonnancein a first level principles of accounts course. Informadon about the attitudinal vaiiableswas obtained by aslting students to complete questionnaires after they had taken dieirexaminations.

Other stadies have exarmned the impact of entr>^ qualificadons on academic perfonnance.Johnes and Taylor (1989) carried out a cross-sectional comparison of the perfonnance ofuniversity students in the UK over the period 1979-1980. They found that the inter-universitydifferences in the non-completion rates of students were related to Advanced level scores, theproportion of smdents in hall and the proportion of students reading languages or businessstudies. Tlie smdy was concerned with athition. a much more blunt measure than performance,albeit much more clear-cut as well. Since then there have been other stadsdcal studies of inter-upiversity differences in pe.rFormance by Chapman (1994, 1997) and that undertaken for theHigher Educadon Quality Council (Chapman, 1996). Chapman's studies primarily demon-strate the vaiiability of standards in UK university departments using Advanced level gradesas a measure of input quality.

A stiidv by Paton-Salzberg and Lindsay (1993) at Oxford Brookes University comparedthe academic performance of students who undertook paid work witli those who did not.The}' found that paid work had a significant and negative effect on progress and increasedthe rate at which modules were failed.

Thus, tliere have been a number of studies of the performance of higher education studentsgenerally and accounting students specifically, most focusing on first year performance. TheStudies by Johnes and Taylor (1989), Chapman (1994, 1997). Mitchell (1985, 1988) andBartlett et at ( 1992, 1993) are all based in the UK. However, they may all have becomeoutdated because of tlie change in the education environment in the last ten yeai:s (the expan-sion; in Highec Educadon (HE) in the UK, a significant increase in pardcipation within the18-22 year old cohort, the increasing presence of more socially disadvantage^ or olderstudents witldn tlE, increasing cost pressures on insdtimons as government funding per stu-dent has been reduced in real terms, the disappearance of government awards to students andthe introducdon of fees charged by universides to theit students), ITie other studies are fromthe USA, the Middle East, tlie Far East and Australia/New Zealand and it is lilcely that theseare culturally specific and the lessons may not be readily transfeired to tlie UK. The aca-demic systems are diffei'sitit and there may be significant cultural differences between thestudents of, say, the Far East the USA and the UK.

Reseai-ch aims

This studv seeks to explore the reasons for differences in academic performance xvithin thecunent dynamic environment using a hybrid of demographic, attitudinal and observationaldata collected from second and final year students. FoRowing previous research studies itconsideis the impact on academic performance of: various background variables, previousacademic perfoimancs, gender, age and lQ. In addidoti, it also considers the reladonshipbetween individual applicadon to study and academic perforniance. Individual apphcationto study was measur&d by classroom obser\-adon along four dimensions. First students"attendance at class was recorded. Second, it was noted whether students had undertaken theprior preparadon for the class - this w as not assessed work, but regular weekly work. Third,the class pajticipatioa of individual students was recorded!; most students had at least

Page 4: A Quantitative Exploration of Student

18 Gracia and Jenkins

a year's exposure to the university system so it was expected that they would be confidentenough to pardcipate. Finally, it was noted whether students got on and completedobjectives set within the class, i.e. whether they took an acdve approach to learning in theclassroom.

The study addresses the deficit of research using behavioural observations within theeducadon literature and also focuses on the reladvely neglected ai-ea of academic performanceof second and final year students in higher educadon. Specifically, the research quesdons thatthe study addresses are the extent to which the data is consistent with the following issues:

1. Is there a positive relationship between 'application' and academic performance?2. Is there a positive reladonship between prior year perfonnance. a range of back-

ground variables and academic performance?3 Is there a posidve reladonship between a range of attitudinal variables and academic

performance "

Empirical research

The data collected related to all 102 second year students (49 in 1999/2000 cohori and 53in 2000/2001 cohort) and all 118 final year students (60 in 1999/2000 cohort and 58 in the2000/2001 cohort), all following the undergraduate Accounting and Finance degree. Thedata was collected over a two-year period.

The degree is a three-year programme although this may be extended to four years bythose students who elect to take a year of supervised work experience (SWE). The degreepermits direct entry to the second year of the programme - the second year stadstics include20 such students. All such students had studied accoundng prior to joining the undergradu-ate programme. The opdonal year of supervised work experience is between the secondyear and the final year. It is available'to all students and the finaf jeai" statistics include36 students who had tmdertaken this period of SWE (about 30% of the cohort).

The nature of the collected data was as follows:

(i) Data obtained from the Student Administration System for each student compnsingdate of birth and age at the start of the year {AGE), total grade points achieved inall modules studied by each student in the previous academic year {GPPREVIOUS)^nd in the current academic year (GPCURRENT). Further feormadon about gradepoints is contained in Appendix 1. The choice of variables was influenced by theresults of the literature search and the search for reasons for differences in academicperformance as defined by GPCURRENT. Additionally, the programme admits asmall number of students directly to the second year of the degree. These tend tobe students who have proceeded to this level affer under-performing at advancedlevel examinations Such students are not expected to do better than students whohave progressed from the first year.

(ii) The results of an IQ test given on a voluntary basis to second and third yearstudents. Sevent}'-five second year students (73.5%) and 80 final year students(67 8%) undertook the test. The test was administered in order to explore the rela-donship between IQ and performance measured by GPCURRENT

(iii) Atdtudinal data derived from a questionnaire administered on a voluntary basisto 94 second year students (92.2%) and 101 final year students (85.6%). The

Page 5: A Quantitative Exploration of Student

Quantitative exploration of student performance 19

questionnaire was intended to increase informadon m order to understand reasonsfor differences in performances.

(iv) All second and third year students were obsen'sd in class to record classroombehaviour in order to establish a proxy measure of tlie responsibility that studentsdemonstrate tow.ards dieir studies. Non-parddpant overt observ-adons were under-taken over 10-week periods covering the first semester of both academic years1999/2000 and 2000/2001. As has been indicated above, observadons were madealon? foTj;r dimensions: attendance, evidence of prior preparadon for the class, par-dcipadoH, and compledon of target work for the class. Each dimension was meas-ured via a scoring grid, weighted equally and summed. This variable is termedAPPLICATION.

Data specification

The data consisted of five scalar vaiiables, six binary' variables and nine nominal v^ariablesconcerned with hsstablishing students' views of accounting and the accounting profession. Asummary of the variables is contained in Appendix 2.

Data analysis

This secdon outlines the results of the statisdcal analysis. It first considers the second yearcohort and then the final year cohort of students. The methods employed are the same forboth cohorts as follows:'

1. Descriptive statisdcs and a statistical test of the diff'erences between groups.2. Riesults of conelation tests berw een the scalar variables and GPCURRENT.3. Influencs of a range of variables on student performance as defined by grade points

achieved through multiple regression.4. Invesdgation of the interaction between atdtudinal vaiiables and performance using

factor analysis in this investigation

SECOND YEAR STUDENTS

Describing the data

In Table 1, the lower number of second yeai" students with GPPREVIOUS is due to the directentry of 20 sttiidents into the second year. Details of how pardcular grades translate into

Table 1.. A summary of the scalar variables

Scalar variables

GPPREVIOUSGPCURRENTIQAPPLICATIONAGE (velars at

start o":' vear)

A'

8210275

102102

Mean

77.1581.02

109.6347.9822.33

Standarddeviation

217529 1720 422.533 15

Minimum

392398.51020

Maximum

120126141 39435

It IS recognized tiat stages 2-3 could be completed using multivariate analysis

Page 6: A Quantitative Exploration of Student

20 Gracia and Jenkins

grade points are contained in Appendix 1. The lower number of IQ returns is due to theabsence of 27 students when the test was administered. With regard to the first four vari-ables, the range and standard deviation of each show a high variability. GPPREVIOUSrepresent the first year results of this cohort and perhaps one would expect a range ofoutcomes for the first year experience of higher educadon. GPCURRENT represent! thesecond year results; these are on average slightly better than the first year but demonsuate ahigher variabihty. In this particular course, the second year contains a high technicalcontent of management accoundng, financial accounting, finance and taxadon and anumber of students have commented on the step up in difficulty from the first year. The vari-ability in the case of APPLICATION is particularly high with a coefficient of variadon of47%; this may be explained by the diversity of performance encapsulated by the variable.

Table 2 contains die result of r-tests on the mean GPCURRENT scores of each of thebinaiy variables. Thep-values indicate that four variables demonstrate a significant differencebetween the means at the 1% level They are the superior performance of: females, studentswho secured direct entry to the course at second year level, students who had studied account-ancy before joining the course and students who had made the study of accountancy theirpreferred choice in the tmiversity application process. The second year of the course is tech-nically demanding and many of the direct entrants are technically very competent. All thesecond year direct entiy students had studied accountancy prior to enrolling on the degreeprogramme. Some of ±ese students are from Hong Kong and they do well in the technicalaspects of accoundng. There is also better performance by those who have studied account-ancy before commencing on the degree programme, and this may be explained by the longerperiod of exposure to accountancy that such students have experienced. The better perform-ance of students who made accountancy their first choice subject may be due to the clearersense of their career goals and acadetnic direction tliat such students have in comparison tothose who perhaps 'drift' into accountancy programmes.

Correlation

Table 3 presents details of the conelation between each of the scalar variables. It indicatesthat the strongest correladon is between GPPREVIOUS and GPCURRENT. This is as

Table 2. Differences in the mean GPCURRENT of the binary variables

Dichotomous variables

Males (0)Females (1)Yr 1 at Glamorgan (0)Yr2entr>'(l)Studied accounting before (0)Not studied before (1)University 1st choice (0)Not 1st choice (1)Accounting 1st choice (0)Accounting not (1)Live at home (0)Not(l)

iV

544882203460544086

83262

MeanGPCURRENT

72.7490.3477.1097.1096.1575.0487 6875 8584.9757.5273.0487.82

StandarddeviationGPCURRENT

27 5427.3129 3823.9727.5128.5525.3135.0031.8023.2837 7726.36

p-value

0.001

0.000

0 002

0.08

0012

0 232

Page 7: A Quantitative Exploration of Student

0.783*-1

0.1810.340*'-1

0 252*0.437*"0 0411

-0.309*-0.293*-0.160

0.0421

Quantitative exploration of student performance 21

Table 3., Pearson's correlation coefficient second year

Correlation GPCURRENT GPPREVIOUS IQ APPLICATION AGE

GPCURRENT 1GPPREVIOUSIQAPPLICATIONAGE

Signiflcar-ce at * "I**, ' ' 5%

expected as both GPPREVIOUS and GPCURRENT are measuring performance in assess-ments and this accords with much of the published work refen-ed to above. Somewhatsurprising is the sigiiifi:cant but lower correladon o[APPLICATION v/ith GPPREVIOUS and(at the 5% level) with GPCURRENT. APPLICATION includes class participadon but themajority of direct entrants to the second year made verj' little contribudon to discussion:sotne of these students were from the Far East and m-any of these students present as morereserved and deferendal in the classroom. The direct entrants ai'e included in tlie correladonof GPCURRENT and APPUCATION but are excluded from correladon between GPPRE-VIOUS wA APPLICATION, which is significant. This is quite interesdng and suggests thatclass paiticipatioo is not an important component of performance. AGE demonstrates a low,significant and negadve correladon with both GPCURRENT and GPPREVIOUS: in recentyears the course has recruited a significant minority of mature students \vho are not gener-ally academically strong. This trend of recruidng mature and other non-tradidonal studentsis set to increase as Government policies to widen access to HE in the UK are maintained.

A furdier investigadon of the correladon between GPCURRENT and the elements ofAPPUCATION reveals that a significant correladon wtth GPCURRENT is obtained if classpardcipadon is removed from the aggregated .4i*/'L/CAr/C>A'measure. The revised APPLf-CATION vaiiable is tenned APPLICATIONS and the following correladon results areobtained:

GPCURRENTGPPREVIOUS

IQAGE

APPUCATI0N30.505**0.601**0.092

-0 066

This increased significance is not panicularly surprising because, in addition to tlie abovecomments, it does appear that many students are very diligent and able without pardcipat-ing sn-ongly in classroom discussions. Because APPLICATIONS has a higher coneladonv/ith GPCURRENT than APPLICATION, it was decided to utilize the potendally higherexplanatory po\^ er of APPLICATIONS in preference to APPUCATION in the further statis-tical tests that follow.

Multiple regression

The above analysis leveals that the following seven independent variables were significantin explaining variations in GPCURRENT at the 0.01 significance level: GPPREVIOUS,APPUCATION3. AGE, GENDER, direct entr\' students {YEARIX students who had smdiedaccountancy before joining tlie programme {STUDIEDAC), and students who made

Page 8: A Quantitative Exploration of Student

22 Gracia and Jenkins

accoundng their first choice of undergraduate study (ACCHOICE). GPCURRENT and theseven independent variables were entered in a Stepwise Regression model. The model withdie highest R- contains four independent variables plus the constant and is summarized inTable 4.

From Table 4, the p-value of the F score shows that the model is sigmficant in explainingthe variation in the results measured by GPCURRENT. The independent variables in themodel explain 78.4% of the variadon in GPCURRENT. The variables GPPREVIOUS,APPUCATI0N3, GENDER and AGE each have a significant effect on GPCURRENT. Thetolerance stadsdcs indicate that muldcolfinearitj' is unhkely to be a problem with this model.

The categorical variables

Following Field (2000) the rule of thumb that there should be at least ten cases per variableis sausfied as there are 94 cases and nine variables and factor analysis can be undertaken.

The nine categorical variables included in Appendix 1 provide information aboutstudents' attitudes towards accountancy. Clearly, categorical variables are being used as ifthey were scalar in nature. This requires the assumpdon that questionnaire respondentsassumed the same differendal distance between each number on the scale. Factor analysisis being used in what Sharma (1996) refers to as an exploratory fashion.

The alpha of die data was 0.5219 indicadng that the set of variables does not representa single unidimensional construct. Nunnally (1978) has indicated that 0.7 may be consid-ered to be an acceptable reliabilit>' coefficient. The data appears to be multidimensional.Factor analysis was then applied to the questionnaire data in order to see which items loadhighest on which dimensions. The analysis did not highlight any problems with lack ofcorrelation between the variables (Bardett's test of sphericity produced a/?-value < 0.05) andthe value of the determinant (0.111) indicated that muldcoUinearity was not a problem. Themeasures of sampling adequacy in the and-image matrix of coneladon did not yield valueslow enough to discard any vaiiables from further analysis.

Both orthogonal and oblique rotadons were effected in order to ease lntei-pretation of theoutput Consideradon of the component correlation matrix produced by the oblique rotationwas examined and. as none of the resulting conelation factors were large, it was decidedto proceed with the orthogonal rotadon and this is the basis of the output summarized inTable 5 The three principal components explain 63.85% of the vaiiation in the data withEigen values => 1. Following Stevens (1992) for this sample size, loadings >0.512 can be

Table 4. Regression analysis results for the second year

ConstantGPPREVIOUSAPPLICATION3GENDERAGE

Unstandardizedcoefficients

B

54.1200 5380.467

13.621-2.314

SE

12 6940 0870.0832.9480.457

Standardizedcoefficients

B

0 4300 3680 246

-0.280

t

4 2636 1685 5894.620

-5.056

P

0 0000 0000.0000 0000 000

Tolerance

0 5510 6200 9460.878

Dependent variable GPCURRENT(,V = 102. ft = 0 892, fi- = 0 796, Adjusted R- = 0 784, F = 70 164 (P = 0 000))

Page 9: A Quantitative Exploration of Student

Quantitative exploration of student performance 23

Table 5. Rotated component matrix

Compcnent

RESPECTRULESALONEINTERESTPRESTIGENONCONCEPTCRUNCHINTERACTLIKEEigen valuesVariance explained %Cumulative variance explained %

1

_--

0 8830 5390.612

-0 7350.7522 912

32 3630.35

2

--

0 796---

0.767--

1.72819 251.56

3

0.8110.781

-------

1.10612 2963.85

considered significant and loadings less than this have been excluded from the table. TheKaiser-Meyer-Olin Measure of Samphng Adequacy is 0 654.

The alpha stadsdcs for each component are as follows:

Component 1 0.7768Component 2 0.5755Component 3 0 4988

Component 1 is associated with positive attitudes to accountancy and the accountancyprofession. The alpha stadstic is above 0.7 indicadng a sufficient level of reliabihty in theconstruct. Beanng in mind the hmitadons posed by the use of categorical variables, thefactor scores for component 1 were obtained and then conelated with each of the scalarvanables. However, none of the conelation coefficients are significant.

Component 2 contains variables associated v.ith the nodon that accountants work aloneand spend their time calculadng. Component 3 is associated with the idea that accountantscommand respect whilst they are bound by rules and reguladons. However, neither constructis sufficiently reliable to allow further analysis.

FINAL YEAR STUDENTS

Describing the data

Table 6 contains a sumrnar}' of the scalar variables for final year students. Only 80 of thefinal year were present when the IQ test was administered. The table indicates that final yearperfotmance of this group of students, measured by GPCURRENT, was better than theirsecond yeai' perf-ormance, measured by GPPREVIOUS. In comparison to the second yearcohort (see Table 1), the final year cohort displayed a higher average level of applicadon thanthe second year, witli a lower standai'd deviation. This is not surprising as in this degreescheme die final year alone determines the class of degree to be awarded so smdents tend tobe more focused on their studies in the final year than in the second year.

Table 7 contains the results of f-tests on die mean GPCURRENT scores of each of thebinary vaiiables. interesdngly, the mean GPCURRENT of those students who undertook ayear of supe,r\ised work experience (SWE) in 1998/1999 is significantly higher than those

Page 10: A Quantitative Exploration of Student

24 Gracia and Jenkins

Table 6. A summan- of the scalar vanables

Scalar variables N MeanStandarddeviation Minimum Maximum

GPPREVIOUSGPCURRENTIQAPPLICATIONAGE

11811880

118118

70 2675.11

117.1569.9023 53

23.2723.7211.7523.803.47

101092.5

320

12913014710041

Table 7. Differences in the mean GPCURRENT of the binary vanables

Dichotomous variables NMean gradepoints 2000

Standarddeviation gradepoints 2000 p-value

Males (0)Females (1)No SWE (0)Undertook 5'1 ''£'(1)Studied accounting before (0)Not studied before (1)University 1st choice (0)Not 1st choice (1)Accounting 1st choice (0)Accounting not (1)Live at home (0)Not (1)

615782363911704893253186

76 5273.6169.1288 7776 7674.9483.4263.0078.4462 7669.277.32

27.1419.5423.4018.3617.2525 64189124 9718.0236.0125.6923.03

0.508

0 000

0 688

0 000

0.003

0.095

who did not do so, at the 1% level. This may indicate that supervised work experiencecontributes to an intellectual maturing of students or it may be that more able studentsundertake supervised work experience. However, a ?-test on the second year results of SWEand non-SWE students, prior to the period of SWE, indicates no significant differencebetween the means of the two groups. As such it appears likely that a period of SWE has apositive lnfiuence on students in terms of subsequent academic performance. In addition,students who attend their first choice of institution also did better than other students-, thismay refiect a greater satisfaction with the undergraduate programme than is the case forstudents whose preference was for other universities. As with the second year students, thosewho made accountancy their first choice subject did better than students who had wanted tostudy a different subject at university. Again, this may be due to increased motivation andsatisfacdon with the chosen programme, which posidvely influences performance. No othervariable in Table 7 exhibits a significant difference between their mean GPCURRENTscores. In the second year, the previous study of accountancy was significant but by the finalyear there aie three important differences. First, all students will have the benefit of an extrayear's maturing in the study of accoundng so the advantage from a prior study will bereduced. Second, some students wiU have benefited from a year of supervised work experi-ence. Third, at the final year level, the curriculum is much more conceptual and much lesstechnical than in the first and second years.

Page 11: A Quantitative Exploration of Student

0 659'=-=1

0.334*=^0.559**1

0.514**0.366*^=0.0681

-0 032-0 053-0.085-0.102

Quantitative exploration of smdent performance 25

Table 8. Pearson's correlation coefficient final year (?j =118)

'O^miatims 'GPCURRENT GPPREVIOUS IQ APPLICATION AGE

GPCURRENT 1GPPREVIOUS

IQAPPUCATIONAGE 1

Signiticance at '** 1 %.

Correlation

Table 8 presents details of the correlation between each of the scalar variables. The cone-ladon matrix reveals that die highest coneladon is between GPCURRENT and the secondyear grade points {GPPREVIOUS). There is also a significant conelat ion betweenGPCURRENT and the variables IQ and APPLICATION. IQ is also significantly correlatedwith GPPREVIOUS.

As witJi the second year results, it was also found that, when 'Participadon' was taken outof the Applicadon data, the resulting variable, APPUCATION3, was found to demonstratea higher coneladon with GPCURRENT as is shown below:

GPCURRENTGPPREVIOUS

IQAGE

APPUCAT10N3

0.720'''*0.486=^=*0.158

-0.099

As with the second year results, it was decided to progress the analysis using APPLICA-TI0N3 rather than APPLICATION.

Multiple regression

The foregoing analysis reveals that the following six mdependent variables were significantin explaining variadons m GPCURRENT at the 0.01 significance level: GPPREVIOUS,APPUCATI0N3. IQ, SWE. this University being the student's first choice (UNP). andAccountancy as the first choice subject {ACCHOICE).

GPCURRENT and the six independent variables were entered in a Stepwise Regressionmodel The model with the highest R~ contains three independent vanables plus the constantand is summanzed in Table 9. From this table, the p-value shows that the model is sigm-ficant in explaining the variation in the results as measured by GPCURRENT. The inde-pendent variables in the model explain 67 .3% of the variadons in GPCURRENT. Each ofthe variables APPUCATI0N3, GPPREVIOUS and UNI has a significant effect on the van-adon in GPCURRENT. The tolerance stadstics indicate that multicollinearity is unlikely tobe a problem with this model SWE, UNI and IQ are excluded in the stepwise regressionmodel when the inter-reladonship of all variables is considered.

Categorical variables

With the final yeai". there are 101 cases and nine quesdons so factor analysis may beundertaken.

Page 12: A Quantitative Exploration of Student

26 Gracia and Jenkins

Table 9. Regression analysis results

Unstandardizedcoefficients

B SE

of final year

Standardizedcoefficients

B T P Tolerance

Constant 6.602APPLICATIONS 0,440GPPREVIOUS 0,427UNIVERSITY 10 135

4.6480.0650.0622,724

0.4380.4190.211

1.4206 7696.9143.696

0,1580.0000.0000.000

0,6670,7600.860

Dependent vaiiable GPCURRENT{N = 118, /e = 0 825, R^ = 0 681, Adjusted R- = 0 673, E = Sl 189 P = 0 000)

The alpha of the categorical variables relating to the nine categorical variables for the finalyear students was 0.4805 indicating that the construct is multidimensional. As with thesecond year data, Factor Analysis was applied to the attitudinal variables. The analysis didnot highlight any problems with lack of correlation between the variables (Bartlett's test ofsphericity produced a p-value of 0.000) and the value of the determinant (0.0829) indicatedthat multicollineaiit>' was not a problem. The measures of sampling adequacy in the anti-image matrix of correlation did not yield values low enough to discard any variables fromfurther analysis.

As with the second year data, it was decided to proceed with the orthogonal rotation andthis is the basis of the output summarized in Table 10. The two principal components explain51.27% of the variation in the data with Eigen values => 1. Variable loadings < 0,512 havebeen excluded from the table. The Kaiser-Meyer-Olin Measure of Sampling Adequacyis 0.695.

The alpha statistics of each component are as follows:

Component 1 0.7872Component 2 0.5585

The rotated components matrix makes interesting reading. Component 1 brings togetherattitudinal values associated with an enjoyment of and an interest in accounting together with

Table 10. Rotated component matiix

Component

RESPECTRULESALONEINTERESTPRESTIGENONCONCEPTCRUNCHINTERACTLIKEEigen valuesVariance explained %Cumulative variance explained %

1

0.704——

0.7990.766

--

0.5610 7712 97

33.0033.00

2

_

0.6860,646

—_-

0.757-_

1.5218.2751.27

Page 13: A Quantitative Exploration of Student

Quandtattve explomtton of student petformance 27

a positive attitude to the work of accountants Component 2 is associated with attitudes thataccounting is rule dominated, that accountants work in a sol'itaiy fashion and spend their timecalculating. This may reflect students' experience of bookkeeping and financial accounting.

As Component 1 has an alpha above 0.7. SPSS was used to generate factor scores for thiscomponent and these were correlated with the scalai" variables Significant correlation coef-ficients were obtained \mth APPLICATION (0.270**). APPUCATION3 (0.302^=*) and AGf(0.325**). These results are not surprising: it is expected ti'iat students with posiuve attitudesto apply themselves to studying. The results also demonstrate the positive attitude of olderstudents for whom the degree programme represents a chosen change of career. However,these results muFit be treated with a degi-ee of caution^ as the underlying data is calfigorical.

Conclusions

In both second and final year cohorts there is an expected correlation between GPCURRENTand GPPREVIOUS. It indicates that there may be common and persistent factors that influ-ence academic performance such that students who are good or poor in formal assessmentsone yoffl- tend to repeat this in the following year. In order to improve overall performance,it may be valuable to identify students completing yeai" 1 with relatively poor marks oracademic profiles arid focus academic counselling or stipport on this group.

Much more interesting is the relationship between APPUCATI0N3 and GPCURRENT.First, removing the participation element increases Ilie strength of the relationship withGPCURRENT for both years. Obsen'ations reveal that students who actively demonstratecommitment arid self-responsibility towards their studies tend to do well in the formalassessments. Such an approach is revealed by regular and consistent patterns of attendance,completion of set work outside of the classroom, high levels of preparedness and readinessV hen attending class, and an active ability to pi;ogress their study during classes tend to dowell in the formal assessments. This aspect of the research is consistent with the commentsof Bartlett et al ( 1992, 1993) that interv^ening variables, rather than demographic variables,piav be important determinants of student performance ia university accounting examina-tions. The lack of influence of paiticipation is consistent with die research reported by Naserand Peel (199S) in which students who revised on theiin own significmtly outperfomnedstudents who stated that they revised witli other students.

The study also indicates that there is significant difference between the performance ofmale and female smdents. in favour of the latter, as revealed by the GPCURRENT scores atsecond year only.

ID the second year there is a low but significant negative correlation between AGE andGPCURRENT. The UK government policy of widening access to higher education, neces-sarily leads to the recruitment of more mature entrants - especially within the post-1992institutions In order to facilitate such access the entry qualifications into higher educationhave become ciiversified, for example, through Accredited Prior Learning (APL), whichrecognizes a wide range of prior smdy activities. Consequently non-traditional students,sometimes with sparse academic backgrounds, are admitted which in turn may impact onthe academic performajice of such students. The data seems to indicate the necessityto establish proactive student support structures aimed at this potentially significant andacademically vulnerable group of smdents,.

The research also indicates that students who undertake a period of SWE aippear toperform significa:atly better than students who proceeded directly from year two to the final

Page 14: A Quantitative Exploration of Student

^^ Gracia and Jenkins

year. There is a variety of possible reasons for this. First, it may be that the more ambitiousand career-oriented students undertake the option. Secondly, the year represents an oppor-tunity to earn a salary (salaries range from £9000-£14000). This imay allow students to gettheir finances in some sort of order before the final year, reducing the need to eai-n moneyfrom paid work in that year, thereby reducing financial and time management pressures.Thirdly, the sandwich year appears to have a maturing influence on students resulting froma range of exposures from complex computerized accounting systems to communicatingaccounting information to professionals.

A further finding is that smdents who entered the programme at second year level as directentry students perform significantly better than the remaining students at that level. Thesetend to be students who initially failed to gain admittance to a bachelor's undergraduatecourse and joined the second year of our programme after completing a tw-o-year diploma.About 60% of the smdents are drawn from Hong Kong with the remamder from the UK. Itmay be that such students, having already experienced failure, work harder to prevent itsre-occuiTence, and hence their motivation is greater. Alternatively, as direct entrants to thesecond year, their higher edlication pathway is more fragmented, thus reducing the depth ofsocial integration - and hence distraction - that they face amongst their peers.

Lastly, the students whose primary objective was to study accounting at university didbetter at both secon^ and final year levels. In line with this, students for whom this was theirfirst choice of university also did better at the final year level. Both these variables seem toindicate that students whc^se first choice is reahzed are more focussed and successful in theirstudies.

The findings of tins research may provide an opportunity to identify students who are 'atrisk' of academic failure and hence target and develop appropriate support programmes forsuch students. As such, based upon the results of the multiple regression analysis, it is hypoth-esized that the variables in Table 11 indicate an 'at risk' profile. As compared with the earheranalysis, the number of variables is significantly reduced when stepwise multiple regressionis used because of the perturbation resulting fi-om the impact of variables on each other.

The research hypotheses generated by this study are that there is a positive relationshipbetween observed fipplication and academic performance. Secondly, at second year, thatfemales out-perform males. Thirdly, at final year, that those students who attend the univer-sity of their first choice out-perform other students.

Limitations

There may be an element of bias in both IQ and questionnaire results. Students were invitedto take part - in line with ethical principles - and this may mean that students who feltuncomfortable with completing either, or who lacked the interest to do so. will not havetaken part.

Table 11. Variables that indicate an 'at nsk' profile

Second year Final year

GPPREVIOUS GPPREVIOUSAPPLICATIONS APPLICATIONSGENDER UNn'ERSITY NOT FIRST

CHOICE

Page 15: A Quantitative Exploration of Student

Quantitative exploration of student perfomiance 29

In terms of the behavioural observations, students were informed and aware that theywere being observed. This may have led to a 'Hawthorne effect' with performance improv-ing simply' because they knew that observation was taking place. How^ever, the obser\^ationstook place regularly over a ten-week period so students became accustomed to and acceptedthe presence C'f an observer. Such acceptance mitigates to a large extent die possible biasinginfluence of the observer on the students' behaviour.

The observations were canied out by one of the researchers in order to maintain consis-tency but it is accepted that one will possess expectations of students and this may have ledto unintentional bias. Finally, the academic year is organized in two semesters and the gradepoints represeui: performance in both semesters. However, the observations were made overa ten-week period in the first semester only.

Further and future research

The above research has developed predictive models for academic failure on an undergrad-uate Accounting and Finance programme. The models will be tested m the academic year2002/03.

The autliors have contemporaneously undertaken a quahtative,, interview-based groundedtheory exploration of the reasons for differences in students' performance. The authors havealso refined the APPLICATION measure for behavioural observations and are currentlyundertaking a longitudinal study of its relationship with student performance. It is hopedthat this will provide deeper insights into the reasons for differences in student performance.It is planned to use attitudinal data to inform die development of the model.

Finally, both the quantitative and qualitative work are underpimied by die aim to build apredictive model that may signal tliose students who are susceptible to academic failure,thereby providing tlie opportunity to develop and focus support structures on students whomay be academically 'at risk".

Acknowledgements

^he authors ai e indebted to the universitj' students who took part in the research and to themany helpful comments of the anonymous reviewers.

References

Auyeung, PK. and Sands, D.R (1994) Predicting success in first year university accountmg usinggender-based learning analysis. Accountmg Education: an intemational journal 2(1) . 259-72.

Baldwin. B.A. and Howe. K.R. f 1982) Secondai7-level study of accounting and subsequent perform-ance in the first college course, Tlie Accounting Review 57 (3), 619-26

Bartlett, S . Peel, M.J., "Pendlebury, M.W, and Groves. R.E.V (1992) An .Analysis of SmdentPerformance in Utidergraduate Accounting Courses. ACCA Occasional Paper 11. London.Association of Chartered Certified Accountants,

Bartlelt, S,, Peel. M.J and Pendlebtiry, M W (1993) From fresher to finalist: a three year analysis ofstudent performance on an accounting degi-ee programme. Accounting Education: an interna-tional journal 2 (2). 111-22

Bergin, J.L. (1983) The effect of previous accounung study on student performance in first college-level financial accountmg course. Issues m Accounting Education 19-28.

Page 16: A Quantitative Exploration of Student

Gracia and Jenkins

Carpenter V.L., Friar. S. and Lipe, M G. (1993) Evidence on the performance of accountmg students.Issues in Accounting Education 8 (1), 1-17

Chapman, K, (1994) Vadahility of degree results in geography in UK universities. Studies in HigherEducation 19 (1), 89-103.

Chapman, K. (1996) inter-Institutional Variability of Degree Results. London: HEQC.Chapman. K. (1997) Degrees of difference: variability of degree results m UK universities Higher

Education 33, 137-53.Doran, B.M,, BeniUon, M.L. and Smith, C.G. (1991) Determinants of student performance in

accounting principles I and II Issues In Accounting Education 6(1), Spring, 73-84,Eskew, N. and Faley. R.H. (1988) Some determinants of student performance in the first college-level

financial accounting course. The Accounting Review 66 (1), \2,1^1.Farley. A, and Ramsay, A.L. (1988) Student performance in first yeai" tertiary accounting courses and

its relationship to secondary- accounting education. Accounting and Finance 28 (1), 29-44Field, A (2000) Discovering Statistics Using SPSS for Windows. London: Sage Publications.Fiaser. A.A., Lyttle. R. and Stolle, C. (1978) Profile of female accounting majors- academic perform-

ance and behavioral characteristics. TM Womar^ CPA October, 18-21.Gul, FA. and Fong, S.C, (1993) Predicting success for introductory accounting students- some

further Hong Kong evidence. Accounting Education: an internattonal journal 2 (1). 33-42.Johnes, J. and Taylor. T (1989) Undergraduate non-completion rates: differences between universi-

ties. Higher Education 18, 209-25Keef, S.P. (1992) The effect of prior accounting education: some evidence from New Zealand.

Accounting Education: an international journal 1 (2). 63-68Koh, M.Y. and Koh, H.C. (1998) The determinants of performance in an accounting programme.

Arcoimting Education: an intentattonal journal 8 (1), 13-29.Lipe, M.G, (1989) Further evidence on the performance of female versus male accounting students.

Issues i?i Accounting Education 4 (1). Spring, 144-52Mitchell, F. (J1985) School accounting qualifications and student performance in tlie first level

university accounting examination. Accounting and Business Research 15 (58). Spring. 81-86.Mitchell, F. (1988) High school accounting and student performance in the first level umversity

accounting course: a UK study. Journal of Accountmg Education 6 (2), 279-91,Mutchler, J F , Turner, J.H. and Williams, D.D. (1987) The performance of female versus male

accounting students. Issues in Accounting Education 1 (1), Mai-ch. 63-68Naser, K. and Peel. M.J. (1998) An exploratory study of the impact of lnterv-emng variables on

student performance in a Principles of Accounting Course. Accountmg Education an interna-tional journal 7 (3), 209-23.

Nunnally. J,C, (1978) Psychometric Theory (2nd edn) New York- McGraw HillPaton-Salzberg, R. and Lhidsay, R.O. (1993) The Effect of Paid Employment on the Academic

Performance of Full-Time Students in Higher Education. Oxford: Oxford Brookes UniversitySchroeder. N.W. (1986) Previous accounting education and college-level accountmg examination

performance. Issues in Accounting Education 1 (1). 37-47.Sharma. S (1996) Applied Multivariate Techniques. New York- Wiley.Stevens, J.R (1992) Applied Multivariate Statistics for the Social Sciences (2nd edn) New York:

Academic Press.Tho, L.M. (1994) Some evidence on the detemunants of student performance in the University of

Malaya Introductory Accounung Course. Accounting Education: an international journal 3 (4)331-40.

Tyson. T. (1989) Grade performance in introductory accounting courses. Why female studentsoutperform males. Issues in Accounting Educatton 3 (1). 153-60.

Page 17: A Quantitative Exploration of Student

Quantitative exploration of student performance

Appendix 1: A summary of grade points

31

Grade

AlA2A3BlB2B3ClC2C3DlD2D3ElE2

Grade pointsper module

1312111098'7

654->J21

Grade points forten modulesper academic year

l-iO13012090

iOO

noso-70605040302010

Appendix 2: V'adables used in the analysis

Data

Grade points in prior academicyear assessments

Grade points m current academicyear assessments

AgeIQ test resultsObsen'atioos of students'

application in classGender

Year 1 at this University orYear 2 enti-y'The student spent last year in

college or undertook SWEHas student studied accountancy

before entry''This university \vas the student's

first choiceAccounting was the students' first

choice of subjectLived at parental home last year'^

Identifier

GPPREVIOUS

GPCURRENT

AGEIQAPPLICATION

GENDER

YEAR]

SWE

STUDIEDAC

UNI

ACCHOICE

HOME

Type

Scalar

Scalar

ScalarScalarScalar

Binary

Binaiy

Binary

Binary

Binary

Binary

Binaiy

Range

0-140

0-140

—95-1450-95

Male = 0Female = 1Yr 1 entry = 0Yr 2 entry = 1Last yr in = 0SWE':^ 1Not studied = 0Studied accy = 1Not first choice = 0First choice = 1Not first choice = 0First choice = 1Lived at home = 0Lived elsewhere = 1

(Continued)

Page 18: A Quantitative Exploration of Student

32 Gracia and Jenkins

Appendix 2: (Continued)

Data

The accounting profession iswell-respected

Accounting is a lot ofrale-memonzing

Accountants work alone morepften than they work withpeople

Accounting does not involveconceptual skills or judgment

Accounting is interestingBeing an accountant has a

lot of prestigeAccounting is a lot of number

crunchingI like accountancyProfessional accountantsinteract with lots of people

Identifier

RESPECT

RULES

ALONE

NONCONCEPT

INTERESTPRESTIGE

CRUNCH

LIKEINTERACT

Type

Categorical

Categorical

Categorical

Categorical

CategoricalCategorical

Categoncal

CategoricalCategoiical

Range

Agree strongly = 4Agree slightly = 3Disagree slightly = 2Disagree strongly = 1Do

Do

Do

DoDo

Do

DoDo

Page 19: A Quantitative Exploration of Student