study of comparison of performance of year ...pills : pacific island literacy levels (niveaux de...

63
STUDY OF COMPARISON OF PERFORMANCE OF YEAR 6 AND YEAR 10 STUDENTS IN ANGLOPHONE AND FRANCOPHONE SCHOOLS IN VANUATU BY ROY CARR-HILL 1 ALETTA GRISAY 2 (Study financed by UK Department for International Development and the French Government, carried out in collaboration with the National Examinations Office) 1. Institute of Education, London, UK December 1998 2. University of Liege, Belgium

Upload: others

Post on 07-Feb-2021

2 views

Category:

Documents


0 download

TRANSCRIPT

  • STUDY OF COMPARISON OF PERFORMANCE

    OF YEAR 6 AND YEAR 10 STUDENTS

    IN ANGLOPHONE AND FRANCOPHONE SCHOOLS

    IN VANUATU

    BY

    ROY CARR-HILL1

    ALETTA GRISAY2

    (Study financed by UK Department for International Development and the French Government,

    carried out in collaboration with the National Examinations Office)

    1. Institute of Education, London, UK December 1998

    2. University of Liege, Belgium

  • Vanuatu/1999/translate 15/01/99 15:51 1

    ABRÉVIATIONS USED IN THIS REPORT:

    BETC : Basic Education Training Centre (CFBT Centre de Formation pour l’Éducation de

    Base).

    DFID : Department for International Development (Département pour le développement

    international).

    IALS : International Adult Literacy Study (Etude internationale sur la littératie des adultes).

    IEA : International Association for the Evaluation of Educational Achievement

    (Association internationale pour l’Evaluation du rendement scolaire).

    IRT : Item Response Theory (Théorie de la réponse aux items).

    MCQ : Multiple Choice Questions (QCM Questions à Choix Multiple).

    NEO : National Examinations Office (BNE : Bureau National des Examens).

    PILLS : Pacific Island Literacy Levels (Niveaux de compétence en lecture des îles du

    Pacifique).

  • Vanuatu/1999/translate 15/01/99 15:51 2

    Prologue

    This is the final report of the Study comparing the performance of students in Years 6 and

    10 in Anglophone and Francophone schools in Vanuatu. The original text of the report was

    presented earlier this year to a workshop held in the USP campus on Monday 8th March

    1999. As a result of the discussion at that workshop and subsequently, we have made a

    small number of changes to the report; and included a number of appendices: the first

    includes the text of the speech made by the Minister at the opening of the workshop; the

    second an edited version of the questions that were put to us and of our replies (including

    where appropriate additional material); and the third includes reactions from two sources: a

    follow-up meeting held at the Curriculum Development Centre; and a statement from the

    French Cultural Attache.

  • Vanuatu/1999/translate 15/01/99 15:51 1

    BACKGROUND

    1. A distinctive characteristic of the Vanuatu educational system, reflecting the

    country’s colonial history is that there are two teaching mediums (in English and in

    French). The duality is not simply linguistic: the traditions, the “school culture”, teaching

    methods, differ in the two systems: moreover, the apparent disparities in performance

    between students in anglophone and francophone students, revealed by the results of the

    national examinations at Year 10, has been a source of concern for the government, the

    teaching profession and the parents.

    2. Are these disparities simply a consequence of the different levels of difficulty of the

    tests taken by the two groups of students, or of differential levels of severity used when

    correcting by the two sets of markers? Or are they a reflection of a real difference in the

    performance of the two populations? If there are real differences in performance, can they

    be attributed to the characteristics of the population over which the educational systems has

    little or no control - for example, more deprived social background in one of the groups?

    Or again, should one point to school based factors - such as, less effective pedagogy in the

    schools attended by one group of students?

    3. The current study was carried out in order to provide (some) answers to these

    various questions - which are important in order to evaluate the equity of the education

    system - at the request of the Ministry of Education in Vanuatu, with funds provided by

    DFID and the French Government.

    4. The study, carried out in 1997 and 1998, was directed by two bilingual independent

    consultants in close collaboration with the Vanuatu National Examinations Office who

    have throughout provided extremely effective support. The study had several components:

    • examination of available documents (both current and historical) about the Vanuatu

    school system;

    • interviews with the principal actors of the educational system;

    • visits to numerous primary and secondary schools, both in Efate and other islands,

    observing lessons and interviewing the teaching staff.

  • Vanuatu/1999/translate 15/01/99 15:51 2

    • on the basis of information collected in these ways, establishing a series of working

    hypothesis which might explain the observed differences

    • systematic collection and analysis of objective data in order to test these hypotheses.

    This report briefly describes these working hypotheses, the data that was collected to test

    the hypotheses, and the principal conclusions of the study. More detail will be provided in

    the follow-up technical report.

    Working Hypotheses

    5. On average, the results of pupils in francophone schools on the standardised Pacific

    Islands Literacy Levels tests taken at the end of primary cycle are not as good as those of

    pupils in anglophone schools. The same differences are observed with the national exams

    taken at the end of the primary cycle (at the end of Year 6) and during secondary cycle (at

    the end of Year 10). The differences could have several explanations.

    Hypothesis One: Bias Linked to the Exams Themselves

    6. It could be that the tests given to the two groups of pupils are not of the same

    difficulty either because of:

    H.1.1. their content;

    H.1.2 poor translation of some of the tests;

    H.1.3 differential familiarity of the students with format of tests items

    H.1.4 cultural biases;

    H.1.5 differential severity of the marking schemes and procedures used by

    examiners.

    Hypothesis Two: Differences Linked to the Social Origin of Students or their Initial

    Aptitudes.

  • Vanuatu/1999/translate 15/01/99 15:51 3

    7. It could be that the populations going to the francophone schools are different from

    those in anglophone schools because the students are:

    H.2.1 from a more deprived social origin;

    H.2.2 from a family environment less conducive to learning whether because

    of linguistic characteristics or because of cultural resources;

    H.2.3 (for secondary students) subject to less selection at the transition from

    Year 6 to Year 10, which would have enabled access to secondary school for

    a group of students who were slightly weaker than the average students

    promoted in the anglophone group.

    Hypothesis Three: Differences Linked to the Quality of Teaching Delivered

    8. It could be that the handicap of francophone schools in comparison to the

    anglophone schools is due to:

    • poorer material resources;

    • less effective teaching practices;

    • less opportunity to learn the material which is tested in the examinations;

    • less qualified or less well prepared teachers.

    Data Collection Procedures

    9. In order to examine these hypotheses, information was required on the following:

    (a) the examination tests taken by students and their psychometric qualities;

    (b) the real performance of students, measured by independent and objective tests which are

    of rigorously comparable difficulty;

    (c) the characteristics of students in the two groups, and of the learning environment they

    have experienced both in the family and at school;

  • Vanuatu/1999/translate 15/01/99 15:51 4

    (d) the relation between the content of the tests they have taken and the material they have

    learnt (opportunity to learn);

    (e) the characteristics of the teaching staff;

    (f) (for students in Year 10) the results they obtained when they took their Year 6

    examination.

    1. Population

    10. Given the relatively small school population in the country, it was judged preferable

    to carry out an exhaustive study (involving all the schools in the country and all the

    students in Years 6 and 10 in those schools) rather than a sample survey. Data was

    therefore collected for two groups of students finishing in 1997, respectively, Year 6 (about

    4,800 pupils) and Year 10 (about 1,150 students) and involved all schools in the country

    which included either a Year 6 or Year 10 class.

    2. Instruments

    A Examinations

    11. Thanks to the well ordered archives of the national Examinations Office, and the co-

    operation of the resident DFID personnel, the consultants were able to obtain a sample of

    the tests taken by pupils/students at the end of Year 6 and Year 10 between 1991 and 1996,

    as well as the PILLs tests taken during the primary cycle (in both Years 4 and 6). The

    examiners in charge of marking the national examinations for the 1997 cohort (which was

    the target year of the study), agreed to carry out the marking on an item-by-item basis for

    the examination for language and for mathematics in both Year 6 and Year 10 and for

    sciences in Year 10, in order to permit a more detailed statistical analysis of the results of

    these tests and to relate them to independent measures of student achievement. Finally, for

    the group of students who took the Year 10 examination in 1997, it proved possible to

    extract from the archives, their score on the examination taken by the same student four

  • Vanuatu/1999/translate 15/01/99 15:51 5

    years earlier, at the end of Year 6, in order to study the possibility that there was an effect

    due to the relative level of selectivity imposed by these tests.

    B. Tests of Achievement

    12. It was important, in the context of this study, that the instruments used to measure

    the achievement of the anglophone and francophone populations were strictly comparable

    and appropriate for the age and educational/school environment of students in Vanuatu.

    We obtained authorisation from the International Association for the Evaluation of

    Educational Achievement (IEA), to use some of the tests developed recently as part of the

    large scale international studies on reading (IEA/Reading Literacy Study 1991) and for

    mathematics and sciences (IEA/Third International Maths and Science Study, 1995).

    13. The comparability of the English and French version of the original tests

    instruments had been carefully verified in order that they could be used in those

    international studies. A national (Vanuatu) committee of teachers and examination

    moderators including both anglophone and francophone personnel was asked to select,

    from among the IEA test batteries available, a sub-sample of items which they considered

    most appropriate given the culture and curricula in Vanuatu. A small number of minor

    corrections have also been made to the items retained for use in order to adapt them to the

    Vanuatu situation.

    14. The tests developed as a result of this process include:

    • for Year 6 - a reading test (54 items);

    • for Year 10 - a reading test (41 items), a mathematics test (31 items), and a science test

    (31 items).

    15. We also obtained authorisation to use some of the tests from the International Adult

    Literacy Study (IALS, STATCAN/OECD, 1995) in order to evaluate the reading levels (30

    items) and capacity to manipulate numerical data (13 items) of the trainee student teachers

    at the Basic Education Training Centre (BETC) in Port Vila. Like the IEA test instruments,

    the IALS instruments already existed in English and French versions, the equivalence of

  • Vanuatu/1999/translate 15/01/99 15:51 6

    the versions having been strictly verified when they were used in the previous international

    study. The data available concern 108 future teachers (42 future anglophone primary

    school teachers, 36 future francophone primary school teachers, and 30 future francophone

    secondary school teachers).

    C. Assessment of Opportunity to Learn

    16. Copies of the versions of the IEA tests in mathematics and science which were to be

    used in Year 10 were given to the corresponding subject teachers in the schools where the

    test was already administered. For each of the test items, these subject teachers were asked

    to indicate, on a check list, whether the material evaluated by the test items had been

    covered or not by their students.

    D. Questionnaires

    17. In order to collect the information on the pupils and students on their learning

    environment and on the material and pedagogical resources at their school, the ‘Student’,

    ‘Teacher’ and ‘School’ questionnaires were developed in collaboration with the team at the

    National Examinations Office. These questionnaires were administered to all students in

    Year 6 and Year 10 who took the tests, to their teachers (in primary school, to the class

    teacher; in secondary school to the teachers of English in the anglophone schools, teachers

    of French in the francophone schools and to the teachers of mathematics and science in

    both schools), as well as to the directors and principals of the primary and secondary

    schools.

    3. Administration and Marking

    18. The various tests and questionnaires were piloted in May 1997, in order to check the

    feasibility of the exercise. Based on this pilot, there were a few revisions and the definitive

  • Vanuatu/1999/translate 15/01/99 15:51 7

    version of the year 10 instruments were administered in July-August 1997 by the

    Examination Office Team. The administration of the tests and questionnaires for Year 6

    took place at the same time as the national examinations (end October 1997) and profited

    from the infrastructure and staff who were organising and administering the national

    examinations. Finally the IALS tests for teacher trainees were also administered in October

    1997 by one of the consultants in collaboration with the personnel of the National

    Examination Office. In all three cases, the test administrators received a guidebook in

    order to standardise the procedures for taking the tests. Similarly, standard guidelines for

    the tests were developed for coding and correcting the questionnaire and tests; the large

    amount of work involved was managed and organised by a local assistant trained by the

    consultants and appointed by DFID. The data were entered into computer files at the

    University of York.

    4. Methods of Analysis

    19. For the analysis of the results of the IEA based tests and the various national

    examinations, the analysis was based on Item Response Theory (IRT). This technique

    allows one to scale both the difficulty of the test items and the proficiency of the groups of

    students who took tests including the same items, even in the particular situation where

    only some of the items had been administered to both groups. In the current study, the IEA

    test items served as the ‘anchor’ group of items which was common to the two populations

    which enabled us to estimate - at least for some of the tests (see below) - the relative

    difficulty, for the whole population, of the items which were only taken by either the

    anglophone or francophone group. In other words, the techniques makes it possible to

    obtain an estimate of the difficulty which the anglophone student population would have

    had if they had taken the francophone examination, and the difficulty the francophone

    student population would have had if they had taken the anglophone examination.

    20. The data collected through the questionnaire responses were analysed using

    standard univariate and bivariate techniques, but certain groups of variables were also

    manipulated so as to devise a series of composite indicators which would both be more

    robust and more easily interpretable in the multivariate analysis.

  • Vanuatu/1999/translate 15/01/99 15:51 8

    21. The more detailed analysis of the various factors associated with differences in

    student achievements, separating out effects associated with differences between students,

    between the schools they attended and between the two teaching medium groups, was

    carried out using multi-level statistical models (multivariate and multilevel regression).

    These models are much closer to the hierarchical structure of the data being analysed than

    the classic ordinary least squares regression analysis and have the advantage of providing a

    more accurate estimate of the partitioning between the different variance components, as

    well as more accurate estimation of the regression coefficients.

    RESULTS

    1. Differences in Achievement as Measured by Tests and Exams

    A. Differences in the performace of the two groups of students on the

    IEA tests

    22. Both in Year 6 and in Year 10 the IEA tests show that students in the francophone

    system do not achieve as well as students in the anglophone system

    23. In Year 6, the anglophone students obtain significantly higher scores than the

    francophone pupils in the IEA reading test. The difference is +0.420 of a standard

    deviation (a statistical measure of the variation in the scores) which means that an

    anglophone student getting the average score for her or his group has a result which is

    better than about three quarters of the students in the francophone group.

    24. In the three IEA tests given to students in Year 10, the anglophone students obtain

    significantly better scores than the francophone students (Table 2). The differences are

    particularly large in Reading and in Sciences, where the anglophone students on average

    score more than one standard deviation higher than the francophone students in Reading

  • Vanuatu/1999/translate 15/01/99 15:51 9

    and three quarters of a standard deviation higher in Sciences. The francophone ‘handicap’

    is smaller in Mathematics (about a quarter of a standard deviation)1.

    Table 1: Year 6 - scores in the IEA reading test broken down by teaching medium

    Teaching medium

    Raw score(number of items answered

    correctly out of 36)1

    IRT score

    Anglophone Francophone Altogether

    N of students Mean Standard deviation Reliability N of students Mean Standard deviation Reliability N of students Mean Standard deviation Reliability

    3194 24.6

    6.8 0.85

    1511 22.3 5.9

    0.79

    4705 23.85 6.60 0.84

    3194 1.56

    1.35 0.85

    1511 1.03 0.96 0.79

    4705 1.39 1.26 0.84

    Note: A few of the IEA items were excluded from this analysis (as in the following analyses) because they did not provide any substantial discrimination in the two populations, or because they behave differently according to the teaching medium, which would have implied an eventual bias in favour of either the anglophone or the francophone students (i.e. those items are significantly easier or more difficult for one of the two groups of students, even for those students who perform equally well on the remaining items in the test).

    1 The difference of a standard deviation in the reading score means that the average score of the anglophone group is better than about 85% of the scores of the francophone group; for the science and mathematics results, the average score in the anglophone group is better than about 75% and 60% respectively of the scores of the francophone group..

  • 012-VanuatuReport-ENG 17/03/2003 7:57 10

    25. For the different IEA tests, the IRT analysis shows that the hierarchy of difficulty of

    items and their fit indices are very close in the two populations, which means that the

    hypothesis that the content or the format of the items have any overall effect on the two

    groups, cannot be sustained (Hypothesis One).

    Table 2 Year 10: scores on IEA tests by teaching medium

    Teaching Medium Reading Mathematics Science

    Raw score

    IRT score

    Raw score

    IRT score

    Raw score

    IRT score

    Anglophone

    Number of students

    Mean

    Standard Deviation

    666

    29.5

    4.2

    666

    1.96

    1.05

    667

    14.8

    4.3

    667

    0.17

    0.81

    667

    15.6

    4.0

    667

    0.29

    0.75

    Francophone

    Number of students

    Mean

    Standard Deviation

    403

    23.3

    5.3

    403

    0.73

    0.88

    407

    12.7

    4.1

    407

    -0.22

    0.80

    407

    11.4

    3.5

    407

    -0.46

    0.64

    Together

    Number of students

    Mean

    Standard Deviation

    1069

    27.2

    5.5

    1069

    1.46

    1.10

    1074

    14.0

    4.3

    1074

    0.02

    0.83

    1074

    14.0

    4.4

    1074

    0.01

    0.80

    Reliability of the test 0.83 0.71 0.70

    Difference ENG/FRA S (0.0001) S (0.0001) S (0.0002)

    26. We have, however, tested the more specific hypothesis, sometimes proposed, that the

    Francophone students might be disadvantaged because they are unfamiliar with multiple

    choice questions (MCQ) and that they would be more at ease with open questions

    (Hypothesis 1.3).

    27. In general, with those who are not used to multiple choice questions, there is a

    substantial level of ‘multiple’ answers (the student does not follow the instruction to only tick

    the most appropriate of the four or five answers suggested). He (or she) ticks all those which

    are, at least in part, appropriate answers. In this study, there were very few such ‘multiple’

    responses, suggesting that the francophone students were not severely handicapped in this

  • 012-VanuatuReport-ENG 17/03/2003 7:57 11

    way - even though it is true that the level of ‘multiple’ answers is slightly higher among the

    francophone students as compared to their anglophone peers.

    28. If one calculates, for each of the three tests, separately the partial scores only for the

    multiple choice questions and only for the open questions, the differences observed with the

    global scores is repeated for each of the partial scores. The differences are however,

    significantly greater for the partial scores for multiple choice questions than for the partial

    scores for open questions, which does tend to confirm that the multiple choice questions

    impose a small handicap on the francophone students. Note however that this factor is in

    no way sufficient on its own to account for the size of the francophone deficit. Even if based

    only on open questions, the scores of francophone students are significantly inferior to those

    of anglophone students, particularly in science and in reading.

    B. Differences in Content between the Anglophone and Francophone National

    Examinations.

    29. The examination at the end of Year 6, which determines transition to the secondary

    cycle, and the examination at the end of Year 10, which marks the end of schooling for the

    majority of secondary school students, are the two major tests for the school career of the

    students. Verifying that the selection, which is a consequence of those examinations, is

    equitable (or, more precisely, assessing whether or not students of the same capacity are

    likely to obtain similar scores, whichever their teaching medium) was one of the important

    objectives of the study. From the point of view of the psychometric quality of the tests

    making up the examinations, the conclusion has to be negative. The English and French

    versions of the tests that we have examined are too far apart for us to consider that, overall,

    the examinations are operating so as to rank the two groups of students in the same way.

    30. The language tests both in Year 6 and Year 10 have not yet been unified. The

    content of the language tests, which reflect these curricula, consequently differ very

    significantly. It is true that real efforts have been made to ensure that the English and French

    versions of the tests give the same weight for the various partial scores (for example, the two

    versions of the tests give the same number of points for writing and for reading: the number

  • 012-VanuatuReport-ENG 17/03/2003 7:57 12

    of items and the time allowed is about the same in the two versions). However, these

    similarities are more apparent than real, as we can see from the language test in Year 10 in

    1997 (Table 3).

    Table 3: Content of Language Examination in Year 10

    English Examination French Examination

    Three tests of writing: a choice of four subjects (minimum 250 words); a precis/resume of one of the books read during the year; a letter. The three answer sheets given to the student each have 60 lines

    One writing test: a choice of three subjects (250 words). The one answer sheet given to the student has 28 lines

    A cloze test (ten items: principally articles, prepositions, etc. )

    A test of grammar, 30 items (conjugation, grammatical agreement, phrase structure)

    A comprehension test: text about 900 words; 15 multiple choice questions; 5 open ended questions. Two of these 20 questions refer to vocabulary.

    A comprehension test: text of about 900 words; 10 multiple choice questions, 5 open ended. Six of these 15 questions are on vocabulary; one on grammar

    31. As one can see, the two tests are not equivalent in terms of the type of competence

    that they are principally testing: the English version emphasises functional skills (written

    expression and comprehension): the French version gives a more important role to formal

    competencies (grammar, vocabulary).

    32. The same is true, to a lesser extent, for the mathematics examination in Year 6:

    although the two versions of the test do include nearly ten items with the same content,

    the wording of the items given to the student is often substantially different.

  • 012-VanuatuReport-ENG 17/03/2003 7:57 13

    33. In contrast, the mathematical and science exams of Year 10 are entirely parallel

    in terms of the content of the items - the curriculum having been unified for several years

    now. The test items are generally of excellent quality and their translation from one language

    to another cannot be faulted. In mathematics, a few differences can be observed in respect of

    the formulation of the questions and the prompts used to obtain replies (the wording of the

    items to the student tends to be formulated slightly more precisely and rigorously in the

    French version).

    34. Both for the English and French versions of the tests (and for whichever subject), the

    correction procedures for open-ended questions do not seem to be sufficiently

    standardised. One of the most problematic aspects is the coding of partially correct

    responses. Several of the francophone markers appear to operate a very analytic correction

    procedure, giving half or quarter points to responses for which the official marking scheme

    only allows for awarding whole number scores (one, two or three points). This procedure

    also occur, but less frequently, with the English markers. It has a number of disadvantages:

    • firstly, there are no strict guidelines for the procedure (which means that the frequency

    with which these decimal scores are used, vary significantly from one marker to another).

    • secondly, and for the same reason, there is no guarantee that the same incomplete or

    partially correct response given by two different students will receive the same mark.

    • finally, it makes the aggregation of the scores more complex (which multiplies the risk of

    error when making a manual calculation of total scores). For certain tests we have been

    able to recalculate by computer the total of the points obtained by the student and to

    compare this to the global score calculated by the marker. The number of discordant cases

    is far from negligible: it is rarely less than 7-8% and in some cases was as high as 15-

    20%, which seems too high for examinations of this importance.2

    C. Statistical Analysis of the National Exam Results

    2 Note however that these error rates are over estimated, because they include also the errors due to data entry. These latter errors are however obviously not relevant to the examination mark actually received by the student because that was calculated manually.

  • 012-VanuatuReport-ENG 17/03/2003 7:57 14

    35. The scaling of the various tests making up the examinations using IRT was made

    more difficult because of the inconsistencies in the content and marking procedures which we

    have just described. Whilst the IRT models themselves are robust, they can only be used

    reliably when the raw data meet several basic requirements: on the one hand, the two tests

    that one wants to calibrate on the same scale, should be measuring the same latent

    competence or trait; on the other hand, the same score awarded to the response of two

    students for the same item should, in principle, signify exactly the same thing. We have seen

    that the raw data do not strictly conform to either one of these two requirements, which

    necessarily limits (sometimes severely) our ability to make a comparison.

    36. For the language examinations both in year 6 and Year 10, it did not prove possible

    to scale the totality of the questions included in the English and French versions of the tests:

    the clearly multidimensional character of the competencies tested by each of these

    examinations made it too difficult to identify a common latent trait. We have therefore had to

    drop from the analysis all items concerning written expression as well as the majority of the

    cloze items (English examinations), grammar (French examinations), and of vocabulary (IEA

    primary tests), only retaining those items measuring reading comprehension. The

    comparison presented below therefore only concerns that part of the national language

    examination which measures the student competencies in the two teaching mediums in

    reading, and not on the other skills measured by the national exams.

    37. For all the items with a decimal code (which means in practice for all of the open

    questions in the tests in language, mathematics and in science) we had to recode by

    computer the replies in order to make them more coherent and manageable in the analytic

    framework. This means that the comparisons presented below are not based exactly on the

    way in which the students scripts were corrected by their markers, but rather on the basis of

    what can be considered to be a computerised rationalisation of the marking procedures

    which were actually used in the field.

    38. For the mathematics exams in year 6, we did not have available the item-by-item

    results of the PILLs tests, which would have provided us with a common anchor point for the

    two populations. The few items which were common to the English and French version of

    the examination proved to be insufficiently stable to provide this anchor point. It was not

  • 012-VanuatuReport-ENG 17/03/2003 7:57 15

    therefore possible to generate a common scale for the English and French versions of the

    mathematics examinations for Year 6.

    39. Taking account of these restrictions, the analysis presented in Table 4 below are much

    too partial to provide a firm response to the question: are the national examinations of

    similar difficulty or do they include a bias in favour of one or other of the two linguistic

    groups? They do however permit us to draw two important conclusions:

    • the data on examination tests which are based on unified curriculums and where the

    content is equivalent (the mathematics and science examinations of Year 10) confirm

    the significant superiority of the anglophone students over the francophone students.

    • both in Year 6 and in Year 10, the reading comprehension section of the language

    tests (the only section which could be modelled into a common scale for the two

    populations) similarly confirms that the anglophone group has better results.

    40. Each of the scores in Table 4 has been calculated by combining into a single scale all

    of the items available for the particular subject for that Year group (items from the IEA test,

    from the National Anglophone test and the National Francophone test) as if we were

    considering one ‘unique’ test taken by the whole of population (where the anglophone

    students had simply omitted to reply to the items on the francophone examination, and the

    francophone students had omitted to reply to the items on the anglophone examination). The

    metric of the scale is given by the ‘anchor’ items from the IEA tests which were common to

    the two populations. Both for the IEA test and for the two examinations, the only items

    included in calculating the students score where those with satisfactory psychometric

    properties (i.e: measuring the same competence for the whole population, and presenting no

    significant bias in favour of one or other of the two teaching medium groups).

  • 012-VanuatuReport-ENG 17/03/2003 7:57 16

    Table 4 Examination scores callibrated on the basis of IEA anchor items

    Year 6 IRT Reading

    Score1 English

    Number of students

    Mean

    Standard Deviation

    3040

    0.73

    1.63 French

    Number of students

    Mean

    Standard Deviation

    1555

    0.14

    1.16 Altogether

    Number of students

    Mean

    Standard Deviation

    4595

    0.56

    1.53 Fidélité 0.92

    Year 10 IRT Reading Score2

    IRT Maths Score3

    IRT Science Score4

    Anglophile Number of students

    Mean

    Standard Deviation

    683

    1.87

    0.82

    699

    0.16

    1.19

    700

    0.38

    0.78 Francophone Number of students

    Mean

    Standard Deviation

    429

    0.71

    0.77

    438

    -0.43

    0.95

    437

    -0.61

    -0.63 Altogether

    Number of students

    Mean

    Standard Deviation

    1112

    1.44

    0.98

    1137

    -0.07

    1.14

    1137

    0.00

    0.87

    Reliability of the test 0.88 0.86 0.84

    Difference ENG/FRA S (0.0001) S (0.0001) S (0.0002)

    Notes 1. Reading Score Year 6 based on 27 items IEA; 26 items ENG; 37 items FRA 2. Reading Score Year 10 based on 38 items IEA; 33 items ENG; 29 items FRA 3. Mathematics Score Year 10 based on 11 items IEA; 31 common ENG/FRE items 4. Science Score Year 10 based on 13 items IEA; 37 common ENG/FRE items.

  • 012-VanuatuReport-ENG 17/03/2003 7:57 17

    41. The examination scores calibrated in this way can therefore be considered ‘equitable’

    (or ‘non-biased’), in the sense that a francophone student and an anglophone student with the

    same level of competence would have obtained the same score: that is, they would have the

    same statistical probability of giving the right (or wrong) answer to items of equivalent

    difficulty included in the tests which each of them had taken.

    42. The differences estimated by this method are of the same order of magnitude as those

    suggested by the IEA test on its own: a little less than half a standard deviation in language

    in Year 6; more than a standard deviation of difference in language and in sciences in Year

    10. Mathematics in Year 10, where the difference is about half a standard deviation, is the

    area where the handicap of the francophone group of students at the end of secondary cycle,

    whilst still very significant, is the smallest.

    Conclusions about the Examinations

    43. It appears therefore that the handicap of students in francophone schools is not a

    simple artefact linked to the content of examinations, to the way in which the questions were

    posed, or to the marking schemes and procedures used. These various factors mean that it is

    not easy to compare directly the results of the examinations taken by the two groups (this is

    particularly true in respect of the language tests, but also, to a lesser extent, for the other

    tests). However, the performance of the students in the francophone groups remains

    significantly weaker when they are assessed using rigorously parallel instruments or when

    more detailed statistical treatments are applied in order to calibrate the examination scores

    on the same scale. The size of the observed differences is pedagogically worrying. A

    ‘handicap’ of the size of a standard deviation (which were observed for both reading and for

    sciences in Year 10) reflects a considerably slower rate of learning, probably equivalent to at

    least a school year, and possibly more.

    44. In respect of the functioning of the national exams themselves, the analysis has

    demonstrated the good psychometric properties and comparability between the two versions

    of the mathematics and science tests in Year 10 - although this is a little compromised by the

    fact that the marking schemes for the open ended questions are not sufficiently standardised.

  • 012-VanuatuReport-ENG 17/03/2003 7:57 18

    The language tests suffer from the same marking problems (even more severely in the sense

    that there are many more open ended questions), but also have differences of content and of

    design which makes their equivalence entirely problematic

    2. Differences in the characteristics of the populations

    2.1 There are only small differences between the anglophone and francophone students

    in terms of socio-economic background, although the latter tend to come from a more

    rural environment.

    45. On the whole, the two populations appear to be very similar (as we would expect given

    the historical pattern of growth of the Vanuatu school system). The differences that are

    observed are relatively small (see Table 5). Thus, whilst the numbers of boys and girls are

    similar in Year 6, there is a slightly higher ratio of boys to girls in the anglophone secondary

    system. Whilst the proportion of Year 6 students boarding at school is the same in both

    systems (at 15%), nearly three quarters of Year 10 students in the anglophone system board

    at school compared to under half of those in the francophone system.

    46. About 30% of Year 6 students had access to water from a tap inside the house, although

    the percentage was higher (at 43%) among secondary school students which is not

    unsurprising as they are an elite population. However substantial numbers were also likely

    to fetch water from a river or spring and this was more likely among students in the

    francophone system, suggesting they are from a rural environment. Similarly, the proportion

    eating a traditional breakfast (banana, yam or manioc) were all greater among students in the

    francophone system (although the differences were very small among Year 6 students).

  • 012-VanuatuReport-ENG 17/03/2003 7:57 19

    Table 5: Comparisons between Anglophone and Francophone populations in respect of selected characteristics and rank correlation of characteristics with Year 10 Fair IRT scores Distribution of the

    characteristic in Rank correlation of the variable with the IRT scores in …

    Year 6 Year 10 Year 6 Year 10 Anglo Franco Anglo Franco Lang. Lang. Maths Scien. Boys =0 Girls = 1 50.5 49.5 46.7 50.6 -.08 +.07 -.08 -.13Water from river or spring? 42.8 47.1 23.8 38.9 -.21 -.22 -.09 -.12Cook on gas? 35.8 35.6 58.4 53.0 -.02 +.27 +.14 +.14Eat ‘traditional’ foods? 53.5 41.6 16.1 31.9 -.23 -.19 -.16 -.16 Books at home? 80.1 50.5 82.5 74.1 +.31 +.31 +.19 +.21Parents to primary school? 74.2 58.2 86.6 80.7 +.22 +.14 +.11 +.04Active Reading? na na 81.2 43.9 na +.23 +.15 +.16Uses Teaching Medium 32.4 24.5 72.3 61.3 +.19 +.13 +.08 +.12 Textbooks? (2+) 31.6 35.5 52.1 25.9 +.37 +.32 +.38Separate Exercise Books 58.2 49.6 75.5 76.6 +.01 -.04 -.03 -.13Use ringbinder? na na 11.6 86.4 na -.50 -.29 -.55Own calculator and dictionary? 58.2 45.0 59.1 26.9 +.16 +.33 +.24 +.28Own geometrical insts?(2+) 50.0 71.9 38.8 78.2 +.06 -.32 -.15 -.26 Bored in language lessons? na na 71.7 49.7 na +.16 +.04 +.12(Most) Teacher(s) corrects work 29.7 55.6 46,6 21.4 -.01 +.19 +.03 +.22(Most) Teacher(s) has favourites 40.7 43.3 51,5 66.9 -.07 +.11 +.13 +.13Likes subjects? 93.5 85.7 37.4 19.3 +.01 +.18 +.22 +.24Student pays for materials etc.? 12.8 27.9 29.1 73.6 -.06 -.26 -.11 -.29

    47. Slightly more of the Year 10 anglophone students have a father who ‘works for the

    government’ or ‘does some other paid job’.

    2.2 The francophone students appear to be significantly handicapped in terms of their

    linguistic background

    48. When asked what language they usually spoke, whilst nearly half Year 6 students report

    speaking their own language with their friends outside school, and over three quarters with

    their families, nearly one in three say that they speak Bislama as the lingua franca with their

    friends outside class (and one in six with their families). Although the pattern of languages

  • 012-VanuatuReport-ENG 17/03/2003 7:57 20

    spoken, of course, varies across the archipelago (see Annexe), in every region there is a

    minority who are using Bislama as a lingua franca with their friends and are not using their

    ‘own’ language at home. Among Year 10 students the tendency is even more marked with

    three quarters of students saying that they speak Bislama as the lingua franca with their

    friends outside school, although nearly the same proportion also say that they speak their own

    language with their family. Although there are few differences in respect of the usual

    language spoken, when asked whether or not they often spoke Bislama in different situations,

    it was clear that Bislama is much more a lingua franca with Year 10 anglophone students and

    their families than for francophone students and their families.

    2.3. The available data on the Year 6 examinations passed in 1993 by the students

    enrolled in Year 10 in 1997 confirm that francophone students were subject to less

    selection at the end of Year 6 (Hypothesis 2.3)

    49. Using the student’s scores on the 1993 Year 6 examinations, we have compared

    the selectivity of francophone and anglophone educational systems with regard to the access

    of students to the secondary schools (see Table 6). Although the scores are on totally

    different tests, in terms of the proportions who are chosen, it is clear that the Anglophone

    system is more ‘selective’ (about 15% of the anglophone students, compared to over 35% of

    the francophone, were accepted in secondary schools).

    Table 6. Scores of current year 10 students when they took the ‘Selection’ Exam in year 6 (in 1993) compared to the scores of all Year 6 students taking the examination in 1993 Anglophone Francophone

    All N = 3236

    Selected N = 495

    All N = 1540

    Selected N = 382

    Language M Range 0-76 SD

    Mathematics M Range 0-49 SD General Studies M Range 0-47 SD

    37.3 17.3 14.7 10.0 27.1 10.5

    53.8 15.8 26.1 8.4 36.5 5.3

    44.0 17.6 21.6 10.5 20.2 9.2

    56.9 10.6 30.5 7.9 27.8 6.8

  • 012-VanuatuReport-ENG 17/03/2003 7:57 21

    50. In terms of their marks in the Year 6 examination relative to the populations from which

    they were drawn, there is a difference in that the mean scores of those selected tend to be

    relatively higher than the overall ‘parent’ means among the Anglophone than among the

    francophone but it is not strong.

    51. Indeed, compared to what is usually observed in similar studies of scores of the same

    pupils separated by a few years, the correlations between the Year 6 marks and Year 10 IRT

    scores are low, especially for francophone students (see Table 7). This observation, for

    which there is no obvious explanation, should be investigated further, especially as it leads

    one to query the predictive value of the assessment at the end of Year 6 and thence it’s

    usefulness as a means of selection of candidates for secondary school. In the francophone

    group in particular, the weakness of the correlations, suggests that there is very little

    continuity of the teaching-learning processes between primary and secondary, or that the

    assessment instruments do not correspond to what has been taught and learnt.

    Table 7 Correlation between scores of Year 10 students in the ‘selection exam in Year 6 (in 1993) and their score in the Year 10 exam Year 6 Exam Score Year 10 IRT Scores Anglophone Francophone Lang Maths Science Lang Maths Science Language Maths General Knowledge

    .53

    .37

    .43

    .46

    .63

    .37

    .36

    .39

    .38

    .31

    .24

    .21

    .06

    .31

    .12

    .08

    .17

    .09

    Conclusions about the characteristics of the two groups of students

    52. There are only small differences between the anglophone and francophone students in

    terms of socio-economic background although, among Year 10 students, those in the

    francophone system tend to come from a more rural environment. Year 6 and Year 10

    students in the francophone system tended to come from more deprived cultural home

    environments although they tended to be helped more often with their homework. The

    development of Bislama as a lingua franca amongst adolescents is perhaps an obvious

    reflection of what they hear being spoken among adults in the wider community, but almost

  • 012-VanuatuReport-ENG 17/03/2003 7:57 22

    certainly benefits students in the anglophone system, given that their teaching medium is

    closer to Bislama. Finally, it is clear that the transition from primary to secondary in the

    anglophone system is more ‘selective’; although in the absence of detailed comparison of the

    examinations used in 1993, it is difficult to say whether or not this means that the students

    entering the secondary system in the francophone system are less able. This uncertainty is

    compounded by the low correlations observed between the scores of the Year 10 students in

    1997 and the marks they obtained when they took the Year 6 examination in 1993.

    3. Differences in the school environment

    3.1. Less material resources (especially textbooks) are available in the francophone

    schools (Hypothesis H3.1)

    53. Among Year 6 primary students, in fact slightly more francophone than anglophone

    students had their own language and mathematics textbooks. However, among Year 10

    students, in each subject, significantly more students had their own textbook in the

    anglophone than in the francophone system (53%, 59% and 58% compared to 48%, 43% and

    27% respectively) with the greatest gaps in science; and, in contrast, many more students in

    the francophone system reported that there was only one or a few copies per class (9%, 14%

    and 12% for language, mathematics and science), this was the case for many more students in

    the francophone system (28%, 43% and 54% respectively). There is clearly a substantial

    disadvantage here even assuming the textbooks are equally appropriate.

    54. More than half of anglophone students in Year 6 compared to slightly less than

    half of francophone students reported separate exercise books for different subjects,

    compared to three quarters of both groups in secondary. However among Year 10 students,

    whilst very few of those in the anglophone system reported using a ring binder, this was true

    of most in the francophone system. Among both Year 6 and Year 10 populations, students in

  • 012-VanuatuReport-ENG 17/03/2003 7:57 23

    the francophone system were more likely to have geometrical instruments (compass,

    protractor…) whilst those in the anglophone system were more likely to have a calculator

    and dictionary.

    55. At primary level, nearly a third of those in the francophone system said they paid for

    materials compared to one in eight of those in the anglophone system. Similarly, at

    secondary level, whilst three quarters of those in the francophone system said they paid for all

    their school materials, this was true for less than a third of those in the anglophone system.

    56. At primary, more teachers in the anglophone system (53/124) reported shortfalls in basic

    teaching materials compared to the francophone system (17/57). However, there are fewer

    cases of severe shortage in that more teachers in the anglophone system said that there was

    at least one textbook between two students. In contrast, at secondary level, more teachers in

    the francophone system (10/36) reported shortfalls in basic teaching materials compared to

    six out of 70 in the anglophone system, and more said that they only had a few books in the

    whole class. More teachers in the francophone schools gave out lecture notes in the form of

    photocopies or stencils and many more asked students to copy lessons from the blackboard.

    3.2. Most of the differences evidenced in the instructional practices tend to confirm

    Hypothesis H3.2 (better pedagogy in anglophone schools).

    57. On the basis of our observations when visiting schools in Vanuatu, we expected to find

    more differences in the school environment than what we were able to substantiate through

    the students and teachers questionnaires. For example, there were fewer differences than we

    had anticipated in the students’ reported experience of lessons in the previous week.

    In Class

    58. Among Year 6 students there was no consistent patterns, although teachers in the

    francophone system were reported to be more likely to correct the pupils exercise books. In

    contrast, among Year 10 students, considerably more students in the anglophone system

  • 012-VanuatuReport-ENG 17/03/2003 7:57 24

    reported that their teachers corrected their homework (46% compared to 21% in the

    francophone system) and checked that they had copied down their homework (57% compared

    to 23% in the francophone system).

    59. An astonishing 86% of Year 6 pupils said that their teacher was sometimes difficult to

    understand and 61% that their teacher often complained about them but this was true in both

    groups. However, more Year 10 students in the francophone system reported that they did

    not understand their teachers all the time - but at the same time that they were never bored in

    class.

    60. Among Year 6 students, teachers in the anglophone system were reported to be more

    likely to help when the pupil did not understand and to give lessons that everyone liked.

    Similarly, among Year 10 students, teachers in the anglophone system were reported to be

    more likely to be concerned with their difficulties, wanting to make them succeed and giving

    interesting lessons. However, in contrast to Year 6 students, more Year 10 students in the

    anglophone students thought their teachers had favourites in class.

    61. Among Year 6 students, those in the anglophone system were tested most frequently,

    whilst Year 10 students in the francophone system said they were tested more often and

    answered questions more often. However, teachers in the anglophone system were more

    likely to use last year’s tests as revision.

    62. Finally, whilst among Year 6 students, there was little difference in the extent to which

    subjects were liked, considerably more Year 10 anglophone students claimed that they like a

    lot both Maths and Science lessons than was the case in francophone schools.

    Homework

    63. Similar proportions of Year 6 teachers in both groups said that, for homework, their

    students used a textbook, or an exercise book or worksheets. In contrast, among Year 10

    teachers, the homework given by anglophone teachers more often (51%) involved students

    using their textbooks compared to those in francophone schools (25%). Two thirds of

    anglophone teachers gave homework that involved students using their notebooks compared

  • 012-VanuatuReport-ENG 17/03/2003 7:57 25

    to half in francophone, and whilst anglophone teachers nearly always used previously

    prepared homework sheets, this was true of only a third in francophone schools.

    64. In terms of the content of the homework, Year 6 teachers in the francophone system were

    more likely to give lessons to memorise, although the proportions were about the same

    among Year 10 teachers; perhaps the major difference was that in both Year 6 and Year 10,

    more of the teachers in anglophone schools gave their students a passage to read compared to

    those in francophone schools.

    3.3 The francophone students received the same amount of mathematics, but less

    science instruction than the anglophone (Hypothesis 3.3 on Opportunity to Learn)

    65. The Maths and Science Opportunity to Learn check lists were completed by 55

    teachers ( 34 Anglophone, 21 Francophone; 26 Science and 20 Maths teachers). No

    significant differences were observed in Mathematics, where in both groups the average

    percent of teachers reporting that the IEA item content had been taught to their Year 10

    students was 86%. In Science, however, there is some evidence that the francophone

    students were offered less opportunities to learn the topics measured by the IEA sciences test

    (although all items used were part of the official curriculum). On average, the science items

    had been taught by 73% of the francophone teachers compared to 82% of the anglophone

    teachers.

    3.4 The proportion of untrained teachers is higher in the francophone schools

    (Hypothesis H3.4).

    66. Whilst nearly all the Year 6 teachers in the anglophone system had 2 or more years

    training, this was true for less than two thirds in the francophone system. Similarly, among

    Year 10 teachers, about twice as many in the francophone system had no training (31%)

    compared to those in anglophone schools (17%), although roughly the same proportion also

    had been on In Service Training (INSET). Clearly, these patterns are different from those

    portrayed in, for example, the Master Plan, Annexes 20-25. However, it is important to

  • 012-VanuatuReport-ENG 17/03/2003 7:57 26

    understand that our data refer to those teachers who were currently teaching in the schools

    and who replied to the questionnaire: some of them may not have had contracts with the

    Ministry of Education.

    67. Whilst a higher proportion of the francophone Year 6 teachers were female (43%

    compared to 29%), there was a significant male bias among Year 10 teachers in francophone

    schools (23M, 5F), compared to anglophone schools (47M 23F).

    68. Finally, it is noteworthy that more francophone teachers thought they “spoke French very

    well” compared to how the anglophone teachers thought they spoke English.

    3.5 The results of the IALS test show that the anglophone trainee teachers preparing

    for the primary teaching outperform the francophone trainee teachers in both the

    primary and secondary training sections.

    69. For both the IALS reading and quantitative scales, it is the group of primary anglophone

    primary teachers who score highest, followed by the francophone students taking the

    secondary course and then the francophone students taking the primary course. These

    significant differences in the level of competencies in reading and arithmetic relate, of course,

    to future teachers and not to those currently teaching. Nevertheless, they seem to confirm

    Hypothesis H3.4 (that francophone teachers are less well prepared).

    70. The ‘background’ questionnaire completed by the same future teachers suggests a

    different attitude to books: the francophone trainee teachers claim to use the BETC library

    much less, spend less time reading, and less often have books in their rooms than the

    anglophone trainee teachers. Finally, in contrast to the teachers currently teaching, the

    francophone trainees appear to be less confident of the mastery of the teaching medium (both

    spoken and written) than the anglophone trainees.

    Conclusion

  • 012-VanuatuReport-ENG 17/03/2003 7:57 27

    71. In Year 6, slightly more students in the francophone system had their own language

    and mathematics textbooks. However, in Year 10, significantly fewer students in the

    francophone system had appropriate material resources both in terms of textbooks and

    exercise books, and these reports were confirmed by the teachers. More students in the

    francophone system more often had to pay for their own materials. Although the differences

    in instructional practices were smaller than expected from our initial observations, the

    teachers in the anglophone system tended to adopt the more interactive style of teaching

    which is known to be more effective. Further, not only is the level of qualification of existing

    teachers in the francophone system lower among both Year 6 and Year 10 teacher

    populations, the scores of francophone teacher trainees, on an internationally comparable test

    of linguistic competence, were lower.

    Factors Associated with Success

    72. As can be seen, the examination of the hypotheses in the light of the evidence has

    enabled us to eliminate some and retain others. Thus, whilst there are some differences in the

    content and organisation of the examinations, these are not sufficient to account for the

    substantial differences in scores. The background characteristics of the two groups of

    students are similar and therefore unlikely to be a major factor in accounting for the

    differences between the groups. There are substantial differences in cultural environment (in

    particular in terms of books in the home) in favour of anglophone students; and, for

    anglophone Year 10 students, their linguistic environment is favourable. Whilst the students

    selected after the Year 6 exam are clearly from the top of their respective groups, relatively

    more francophone students are chosen. There are differences in instructional practices but

    they are relatively small and inconsistent between primary and secondary; although

    anglophone students have a clear textbook advantage in Year 10.

    73. The following results, based on more comprehensive analyses in the more extensive

    technical report, aim to assess the relative power of the various factors. They are based on

    the adjusted ‘fair’ scores for Year 6 students in the language examination, and on the adjusted

    ‘fair’ scores for year 10 students in language, mathematics and science. These adjusted

  • 012-VanuatuReport-ENG 17/03/2003 7:57 28

    scores have been derived from items which proved to be equally difficult (or fair) as between

    students of the same competence in the anglophone and francophone groups.

    74. The bivariate associations show that nearly all the factors identified as potentially

    important make a difference to the scores of individual pupils within both anglophone and

    francophone groups. Some examples of the correlations between the various explanatory

    factors and these IRT scores are given in Table 5 above. The most important variables (in

    terms of the size of the correlations) are:

    • background characteristics and environment (age, gender, whether or not eat traditional

    foods, access to water, gas and electricity, books at home, and parental education and

    help with homework).

    • selectivity (for Year 10 students only), their score on the Year 6 exams in 1993 is

    correlated with their Year 10 score although not as much as one might expect from

    international experience.

    • material resources (whether or not pupils has corresponding textbooks, separate exercise

    books, pays fees for materials).

    • instructional practices (whether or not pupil is bored in lessons, reads stories for

    homework, perceives the teacher as fair).

    75. Clearly however, these different factors interact and overlap: if there are more books

    in the home, the students linguistic environment will be different which also means that they

    can take more advantage of teaching: certain pedagogies will be favoured by greater/lesser

    reliance on textbooks, and so on. It is therefore important to assess the relative weight of

    each of these factors; and we do this by estimating equations relating the scores to the

    candidate variables that we have discussed above.

    76. The interpretations below are based on seven different models:

    • the first (null model A) aims to divide the total variance for each of the dependent

    variables (the IRT scores) into two components: the variance between the schools and the

    variance between students in the schools.

  • 012-VanuatuReport-ENG 17/03/2003 7:57 29

    • the second (model B) and the third (model C) allows one to estimate the fraction of the

    total variance between schools which can be ‘explained’, respectively by the teaching

    medium and by the mark obtained in 1993 when the student in Year 10 in 1997 took the

    Year 6 examination.

    • the fourth model (B + C) allows us to demonstrate the extent to which the two preceding

    factors overlap each other (and therefore to estimate the extent to which the disparities

    between secondary schools can be explained by the differential difficulty of the

    examinations taken in Year 6 by anglophone and francophone groups)

    • the fifth model D provides an estimate of the total variance in each of the IRT scores

    which we can account for by taking into account the background, cultural and school

    environment of the students which were described in paras 45-71 above. These variables

    explain, at the same time, part of the differences observed between schools (for example,

    the variables describing pedagogical practices), and part of the variance observed between

    students in the same school (for example the variables describing the cultural environment

    of the students).

    • the sixth model (B+D) helps to understand the extent to which the environmental variables

    described immediately above are associated with the differences identified between

    teaching mediums. The comparison between the results of models B, D and B+D allow us

    to draw the conclusion that the variance between schools, which is apparently explained

    by the teaching medium, is associated - at a statistically significant level - to the handicaps

    suffered by the francophone group of students in terms of their socio-economic status and

    cultural background and in terms of the quality of the education provided by the schools to

    which these students go.

    • finally the seventh model (B+C+D) allows us - for Year 10 - to make a better estimate of

    the bias introduced by the differential selectivity between anglophone and francophone

    groups, and therefore to estimate with greater precision the effect of the various other

    factors. In addition, it gives us an idea of the proportion of the total variance which we

    have not been able to account for with all the information we have collected. This ‘non-

    accounted-for’ variance is due to (a) factors other than those studied here which affect the

  • 012-VanuatuReport-ENG 17/03/2003 7:57 30

    students results; (b) the factors that we have been able to include in the model not being

    measured sufficiently precisely by the information we have collected; (c) the scores which

    are the dependent variables are, just as much as the independent variables, affected by

    measurement error.

    77. The first observation is that, apparently, a very large proportion of the variance

    between individual students can be ascribed to variation between schools: a little under a

    third (32%) of the variance in Year 6 IRT scores in language is variation between schools;

    the proportions are even higher for the analysis with the Year 10 IRT scores (a little under

    60% for language, 40% for mathematics and 50% for sciences - Table 8, model A). These

    proportions are much higher than those usually observed at the start of this kind of analysis

    suggesting that the differences between schools are much greater than one would

    typically meet in other school systems.

    78. Model B in Table 8 shows that a significant part of this between school variance can

    be ascribed to differences between anglophone and francophone schools: and this effect is

    particularly marked for the language and science Year 10 scores, where the introduction of

    the variable ‘teaching medium’ into the equation appears to explain about 35% of the

    variance in the language score (out of the 58% of the between schools variance) and 30% of

    the science score (out of the 49% of the between school variance).

    79. In mathematics, (where we only had data that was comparable between the

    anglophone and francophone group for the Year 10), the impact of the teaching medium is

    much less important (about 8% of the 39% of the between school variance can be attributed

    to the teaching medium). As we have seen with the results of the analysis of the IEA scores,

    the handicap of the francophone students is smaller in this subject than for the two others.

    The differences from one school to another however remain considerable; and they cannot be

    attributed exclusively to the contrast between the anglophone system and the francophone

    system. BOTH systems are characterised by large differences between schools.

    80. As we have seen from the analysis of the IRT scores derived from both the IEA tests

    and the National examinations, the differences linked to the teaching medium are much less

    strong in Year 6 as compared to Year 10: apparently only 2.4% of the variance of the

  • 012-VanuatuReport-ENG 17/03/2003 7:57 31

    language scores in Year 6 can be attributed to the teaching medium (Table 8, model B),

    although the between school variance is still very substantial at that level (31%). That

    suggests that, at least for language, the handicap of students in francophone schools is

    accentuated whilst they are at secondary school.

    81. This result could be, at least in part, linked to the less severe selection of students

    into the francophone systems. Models C and B+C tend however to show that the ‘selectivity’

    factor is complementary to the effects of the teaching medium rather than the effects

    overlapping with each other (the overlap between the two factors is weak, not only for

    language but also for the mathematics and the sciences). However it should be noted that the

    Year 6 score entered into these analysis is standardised to the average of the Year 6 scores in

    each language group.

    82. It is interesting to note that the Year 6 score is a better ‘predictor’ for the Year

    10 exam results in mathematics, than for either languages or for science (probably

    because the mathematics tests are more reliable) (Model C).

    83. Surprisingly, this transition score to the secondary system contributes more to an

    ‘explanation’ of the disparities between schools than to an explanation of the differences

    between the students within the schools (Model C). This suggests that the students who

    obtain better results in the Year 6 examination tend to congregate in certain secondary

    schools rather than others. This phenomenon is most marked in languages and in sciences,

    where the two variables ‘teaching medium’ and ‘Year 6 score’ by themselves, allow us

    to account for more than three quarters of the differences in achievement observed

    between the schools in Year 10 (Model B+C).

  • 012-VanuatuReport-ENG 17/03/2003 7:57 32

    Table 8 : Partitioning Variance between Schools and Students with Different Models

    Year 6 Language

    Year 10 Language

    Year 10 Maths

    Year 10 Science

    (A) Null Model Level 1 (School) Level 2 (Student)

    0.655 (31.9%) 1.396 (68.1%)

    5670 (57.9%) 4120 (42.1%)

    4954 (39.1%) 7711 (60.9%)

    3864 (50.6%) 3769 (49.4%)

    (B) Medium Level 1 (School) Level 2 (Student) Effect on School Variance

    0.603 (29.4%) 1.396 (68.1%) -2.4%

    2234 (22.8%) 4122 (42.1%) -35.1%

    4094 (32.3%) 7715 (60.9%) - 6.8%

    1433 (18.8%) 3782 (49.5%) - 31.8%

    (C) Previous Score Level 1 (School) Level 2 (Student) Effect on School Variance Effect on Student Variance

    na 4787 (48.9%) 3856 (39.4%) -9.0% - 2.7%

    3268 (25.8%) 6458 (51.1%) -13.3% - 9.8%

    3123 (40.9%) 3593 (47.1%) -9.7 -2.3%

    (B) Teaching Medium + (C) Previous Score Leve1 1(School) Level 2 (Student) Effect on School Variance Effect on Student Variance

    na 1282 (13.9%) 3862 (39.4%) -44.0% -2.7%

    2336 (18.5%) 6468 (51.1%) -20.6% -9.8%

    833 (10.9%) 3611 (47.3%) -39.7% -2.1%

    D) Background, Cultural and School Environment Leve1 1 (School) Level 2 (Student) Effect on School Variance Effect on Student Variance

    0.464 (22.6%) 1.203 (58.6%) -9.3% -9.5%

    3279 (33.5%) 3737 (38.2%) -24.4% -3.9%)

    3654 (28.9%) 6976 (55.1%) -10.2% -5.8%

    2384 (31.2%) 3397 (44.5%) -19.4% -4.9%

    (B) Teaching Medium + (D) Background, Cultural and School Environment Level 1 (School) Level 2 (Student) Effects on School Var. On Student Variance

    0.443 (21.6%) 1.203 (58.6%) -10.3% -9.5%

    1424 (14.5%) 3739 (38.2%) -43.4% -3.9%

    3539 (28.0%) 6974 (55.1%) -11.1% -5.8%

    1166 (15.3%) 3409 (44.7%) -35.3% -4.7%

    (B) Teaching Medium + (C) Previous Score + (D) Background, Cultural and School Environment Level 1 (School) Level 2 (Student) Effects on School Variance Effect on Student Variance

    na 742 (7.6%) 3536 (36.1%) -50.3% -6.0%

    2055 (16.2%) 5956 (47.0%) -22.9% -13.9%

    714 (9.4%) 3296 (43.2%) -41.2% -6.2%

    Notes:

    (a) Estimations are presented for seven separate models: the first simply partitions the variance between the schools and the students in those schools; the second and third show how much the teaching medium and the previous score separately account for the differences between the schools; the third shows the combined effect of the previous score and the teaching medium; the fifth shows the effect of all the environmental variables (both home and school); the sixth the combined effect of these environmental variable with the teaching medium; and the final model, the effect of also including the previous score on differences between both schools and students.

    (b) The percentages are calculated by reference to the total variance in the null model: for example, 3726 is 50% of 7453 not of 5508 (=1479+3829). (c) The ‘Effect on the Variance’ is the percentage amount by which the proportion of variance attributable to the school or to the student level is decreased by the introduction of the variable. 84. The same is true to an even greater extent for the collection of variables which

    describe the student and her/his educational environment (both at home and at school). Many

  • 012-VanuatuReport-ENG 17/03/2003 7:57 33

    of these variables are related to the personal characteristics of the student and her/his family

    milieu: one might therefore have expected that they would account for a much greater part of

    the individual between-student variance. In fact, they only ‘explain’ about 10% of the

    variance between students within the schools in Year 6 and only between 4% and 6% in Year

    10. In contrast they are associated with a substantial fraction of the between school variance,

    particularly in Year 10 for languages and sciences (Model D). This confirms the existence of

    non-negligible aggregation effects: the secondary schools in the country are significantly

    different, both in terms of the social, cultural and educational background of the

    students enrolled and the quality of the education environment that they provide for

    their students.

    85. If one compares the results of Models B, D and B+D, one can estimate the fraction

    of the between school variance which can be attributed specifically to the characteristics of

    the educational environment (i.e. the effect, whichever the teaching medium) and to

    distinguish it from that fraction which is linked to the joint effect of the teaching medium and

    the environmental characteristics. The identification of this latter effect was one of the most

    important objectives of the study, because it allows to verify hypothesis 2 and 3: are the

    weaker results of the students in the francophone schools due to a less favourable educational

    environment to which they are exposed, either at home or at school?

    86. The ‘joint’ effect is in fact negligible in Year 6 (only 1.4% of the between school

    variance in the language score); in contrast, it is highly significant in Year 10 (16% of the

    between school variance for both the language and science scores, and 5% for the

    mathematics score). The handicaps observed among the francophone students at

    secondary level in terms of socio economic and cultural background and the poorer

    school environment allow us to account for about half the differences of the results in

    languages and science observed between the anglophone and francophone groups (for

    the language score, 16% out of the 35% between school variance associated with the teaching

    medium; 16% out of 32% for the science score).

    87. Whilst these analyses are very important, it is also useful to examine the pattern of

    variables which retain statistical significance in the ‘fixed’ part of the model. For both Year

    6 and Year 10 students, after controlling for school membership (which we have to do in

  • 012-VanuatuReport-ENG 17/03/2003 7:57 34

    order to obtain unbiased estimates), whilst gender always and age nearly always has an

    effect, the other ‘background’ characteristics appear only occasionally as statistically

    significant in the analyses. When we further include either teaching medium or the previous

    score or both in the analysis, this tends to substitute to a limited extent for these background

    variables (and the effect of the previous score and the teaching medium are reduced), but not

    by as much as one might expect. The most robust effects are as follows: with the analysis for

    language in Year 10, there are effects attributable to whether or not the family draws water

    from a river (negative), whether they cook on gas, whether or not there are books at home,

    and active reading on the part of the child (all positive); with the analysis for mathematics in

    Year 10, there are effects for whether or not the family eats a traditional breakfast (negative),

    and whether or not the head of the household does paid work (positive); and for the analysis

    of science in Year 10, the only effect is active reading on the part of the child (see Table 9).

    88. For both Year 6 and Year 10 students moreover, even after controlling for school

    membership, some of the variables reflecting material resources or instructional practices

    remain statistically significant in all three analyses at Year 10; and these effects are also not

    affected substantially by the presence of teaching medium and/or previous score in the

    model. Thus, whether or not the student has a calculator of a dictionary is always significant

    (positively); in the analyses for language, whether or not the student has the textbook has a

    positive effect as does the possession of geometrical instruments (although this is a negative

    effect); in the analyses for mathematics and for science, whether or not the student uses a

    ringbinder has a large negative effect, whilst whether or not they like the subjects has a

    positive effect; in addition for the analysis with mathematics, whether or not the student pays

    for their materials appears to increase their score (see Table 9).

    Table 9 Coefficients and Standard Errors of ‘Best’ Models for whole population

    Language Mathematics Science Constant 174.9 (16.7) -16.9 (31.7) -45.5 (23.3) (0-1) Effect of teaching medium -103.8 (13.0) -37.0 (22.1)

    -73.0 (15.6)

  • 012-VanuatuReport-ENG 17/03/2003 7:57 35

    Score in Year 6 on that subject 19.9 (2.5) 39.4 (3.2) 8.0 (2.4) Gender Age

    --10.8 (2.2)

    -15.5 (5.6)-9.8 (2.9)

    -20.6 (4.2) -9.6 (2.2)

    Water from River Cook on Gas Books at Home Eat traditional breakfast Occupation of Head of Household Active Reading

    -10.8 (5.5)12.4 (5.1)

    6.1(3.1)--

    7.7 (3.1)

    ---

    -7.6 (3.1)13.8 (6.2)

    -

    - - - - -

    6.5 (3.0) Subject Manual Use Ringbinder Own calculator and Dictionary Own Instruments Pays for Materials Like Subjects

    5.4 (2.6)-

    4.6 (3.5)-8.8 (2.3)

    --

    --17.8 (9.0)12.8 (4.6)

    -16.8 (8.2)7.3 (2.1)

    - -18.5 (6.8)

    8.4 (3.4) -

    3.5 (1.6)

    RESET test (separately for anglophone) (separately for francophone)

    4.03.91.9

    4.73.21.3

    2.9 2.2 1.9

    Notes: (a) only variables that are statistically significant at the conventional 5% level have been retained (b) the RESET test is a general test for mis-specification of the model

    89. The equations have been tested for completeness and it appears that there are some

    systematic effects which have not been captured by the variables that we have been able to

    include in the model. It is important to recall here first, that it was not possible to include the

    teacher level variables into these analyses without losing too much data; and second, that we

    had ourselves been surprised, on the basis of our observations, that there were fewer

    quantitative differences in the questionnaire responses between the instructional variables. If

    we had been able to measure the differences that we thought we saw, we would doubtless

    have achieved a more complete model. Moreover, when separate models are estimated for

    the anglophone and francophone populations, even though similar (although not exactly the

    same) variables appear in both groups, the models for the francophone population and for the

    science analysis for the anglophone population appear to be well-specified, suggesting that

    there are different processes at work.

  • 012-VanuatuReport-ENG 17/03/2003 7:57 36

    90. Overall, therefore, the picture appears to be as follows:

    • younger children do better than older;

    • girls tend to do better than boys;

    • those from a rural or poorer background perform worse at language and mathematics (but

    not necessarily at science)

    • it helps if the children like subjects!

    • books at home and active reading are associated with better language performance

    • students with their own calculator and dictionary perform better at all three subjects

    • students are handicapped with a ringbinder

    These will be taken up in the recommendations.

    RECOMMENDATIONS

    91. On the basis of the preceding analyses and the results obtained, a series of

    suggestions can be made which might help to diminish the observed disparities and/or to

    improve the quality of the educational system in Vanuatu. We have distinguished, in these

    recommendations, between those which are short term (ST), medium term (MT) or long term

    (LT).: where short-term refers to changes requiring resources but little prior planning or

    preparation, medium term to changes requiring preparation and planning as well as resources,

    and long term to those reforms needing delicate negotiations between the different groups

    and a substantial political commitment.

    Material Resources

    92. Compared to many other developing countries, the children do have basic materials

    and textbooks, although the Master Plan paints perhaps too rosy a picture of the coverage

    (see section II.3(a) Availability of Textbooks and other teaching/learning materials). In

  • 012-VanuatuReport-ENG 17/03/2003 7:57 37

    particular, where the approach to teaching relies on those textbooks then even small shortfalls

    can become serious (ST); and in this respect the francophone students appear to be

    disadvantaged especially in science. Moreover, it seems clear that, at least in Year 10, many

    teachers in the Francophone system think that that the quality of the mathematics and science

    textbooks are not appropriate and need to be revised (MT). The replies to the Opportunity to

    Learn questions confirm, at least for science, that the subject matter taught to students in the

    francophone system is more limited and corresponds less to the official curriculum than is the

    case in the anglophone system..

    93. In addition to the issue of textbooks, the use of ring-binders - which are difficult to

    manage - should be discouraged; and serious efforts should be made to rectify the apparent

    discrepancy between the students in the francophone system, who have to pay for their

    exercise books, whilst these appear to be free - or subsidised - in the anglophone system

    94. More generally, the students who attend francophone schools have a more limited

    learning environment. Not only do they have fewer textbooks in Year 10 (although not in

    Year 6), but there are fewer books in their homes and in their school libraries (or they are less

    accessible). They are less frequently asked to read, and francophone student teachers also

    read less. The government could launch a national campaign to sensitise parents and teachers

    to the importance of reading; donors could support this through the provision of appropriate

    materials.

    Instructional Practices.

    95. The differences whether reported by students or by the teachers were not as large as

    we had originally surmised. However, international experience and some of the results

    observed in the present study, suggest that there would be advantages in moving towards a

    more interactive style of learning for both systems but especially for the francophone system.

    Whilst some short term progress could be made with in-service training (INSET), this also

    implies revision of the initial teacher training programme and substantial retraining over the

    medium term.

  • 012-VanuatuReport-ENG 17/03/2003 7:57 38

    96. At the same time, it is also evident that there are substantial differences in the

    qualifications and experience - even within their own systems - of the anglophone and

    francophone teachers. Although, administrative data show that ‘few teachers are unqualified’

    (Master Plan section II.3(c), Teachers and Teaching Conditions), our data suggest that,

    among those currently teaching in the schools, who may include some teachers not employed

    by the state, there is a substantial minority without basic teaching qualifications. Rectifying

    this would imply a substantial investment over the long term; although a comprehensive

    INSET programme would contribute to reducing these disparities.

    Selectivity

    97. The problem posed by the difference of selectivity of the examination at the end of

    Year 6 is probably the one which is the most politically sensitive to resolve. Should one

    maintain the current pragmatic and apparently ‘open’ approach which means that the rate of

    promotion/transition at the end of the primary cycle, can be adjusted according to the

    differential capacity of the anglophone and francophone secondary systems? Should one

    completely integrate the curriculum and the examination in the primary cycle, in order to

    guarantee that students of the same capacity have the same chance of being promoted, with

    the risk of serious instability in the secondary school systems, as the numbers of enrolments

    in the francophone system could drop rapidly whilst those in the anglophone system might

    grow rapidly. Should one instead increase the capacity of the anglophone secondary school

    system, in order to reach a better equilibrium - and at the same time reduce the potential

    tension because of the relatively severe selection in the anglophone examination at the end of

    Year 6? This latter has the advantage of opening up access to secondary schooling to a larger

    proportion of the country's juvenile population. However, even this 'equilibrium' is unlikely

    to be acceptable to the francophone popul