Professional Documents
Culture Documents
Ma 1997
Ma 1997
© 1997ElsevierScienceLtd
Pergamon All rightsreservedl~mtedmGreatBritain
0738-0593/97117O0÷ 0 00
PII: S0738-0593(96)0005 I-X
XIN MA
Department of Mathematics and Science Education, The University of British Columbia, Vancouver, BC,
Canada V6T IZ4; and The Atlantic Center for Policy Research in Education, The University of New Brunswick,
Fredericton, NB, Canada E3B 6E3
Abstract- - The purpose of this study was to examine mathematics achievement of high school sen-
ior students in the Dominican Republic. The data (N = 1082) were collected in a national evaluation
project on the learning of high school mathematics. Multiple regression analyses were used to explain
the relationship of mathematics achievement to characteristics of student, family, school and region.
Sets of independent variables were used and their causal priority in explaining mathematics achieve-
ment was determined by putting child-centered learning into a sociologically specified context. The
main findings were: (a) among the significant predictors of mathematics achievement, student char-
acteristic, school type and geographical region were the most important, while attitude toward
mathematics and father's educational level were the least important; (b) student age, mother's educa-
tional level and career aspiration had no significant effects on mathematics achievement; (c) student
characteristics were still most responsible for mathematics achievement in the hierarchical sociologi-
cal context of family, school and region. © 1997 Elsevier Science Ltd
313
314 XIN MA
the priority of family, school and region in put into this sociological context. Given this
explaining academic achievement be determined hierarchical sociological context of family, school
based upon the closeness of their relations with and region, can student characteristics still be
individual learners. The influence of family on most responsible for mathematics achievement?
students' academic achievement has been recog- A more general research question was: Given the
nized by many researchers (e.g. Anyon, 1983; hierarchical sociological context, what is the rela-
Eccles et al., 1985; Fennema and Peterson, 1985; tionship of mathematics achievement to charac-
Fuller and Heyneman, 1989; Theisen et al., 1983). teristics of student, family, school and region?
Socio-economic status (SES) and home environ- Using data from the Dominican Republic, this
ment variables usually explain a substantial study attempted to investigate these research
amount of variance in academic achievement. questions that have been barely explored in the
Educational sociologists have theorized that fam- developing world.
ily characteristics shape values of individual learn-
ers which may lead to frustration and failure when METHOD
confronted with different values inherent in the The Dominican Republic setting
schools (Cloward and Ohlin, 1960; Cohen, 1955; The Dominican Republic is located in the
Coleman et al., 1966; Jencks et al., 1972; Merton, Caribbean Sea, neighboring Cuba, Puerto Rico
1968). As a result, it has been argued that family and Haiti. The country has an area of about
background plays a central role in determining 49,000 km 2 which is divided into 26 provinces,
students' academic achievement (Fuller and Hey- and one national district where the capital city,
neman, 1989). Santo Domingo, is located. The population is
Research also supports the special place of approximately 7 million, of which nearly 75% are
school in the explanation of students' academic under 25 years of age (Fern/mdez, 1990). The
achievement (e.g. Anyon, 1983; Crespo, 1990; average annual rate of population increase is
Eccles et al., 1985; Fennema and Peterson, 1985; about 2.3%. The form of government is repre-
Fennema and Sherman, 1977; Leder, 1992; The- sentative democracy with traditional division of
isen et al., 1983). Different types of schools have powers between the executive, legislature and
been found to produce different levels of aca- judiciary. The economy is typical of that of many
demic outcomes (Coleman et al., 1982; Hoffer et developing countries, mainly agriculture, mining
al., 1985; Lee, 1986; Lee and Bryk, 1988). It is also and recently tourism.
observable in the literature that family character- Primary and secondary education is compul-
istics often precede school characteristics in sory in the Dominican Republic. The Ministry of
accounting for academic achievement (see Cole- Education controls and provides curricula for
man et al., 1982; Hoffer et al., 1985; Lee, 1986). public and private education at the pre-school,
Academic outcomes of schooling depend on primary and secondary levels. There are three
cultural characteristics of students that differ not types of schools in the country: public, private
only from country to country in the world but also 61ite and private non-61ite, depending upon their
from region to region in a country (Lennon, status with the Ministry of Education and SES of
1988). Crespo (1990) and Theisen et al. (1983) their students (see Luna et al., 1995). Public
contend that it is not possible to ignore the effect schools are run and financed by the government.
of geographical regions on students' academic Students attending public schools are of low SES
achievement, especially in developing countries. because public schools are free of charge. Private
Theisen et al. (1983) further suggest that region ~lite schools are autonomous and authorized to
and school together have some combined effects administer their own examinations. Students
on students' academic achievement. Note also attending these schools are of middle to high SES.
that characteristics of family and school often Private non-~lite schools operate under the super-
precede those of region, because families and vision of the Ministry of Education and are not
schools in a certain region often reserve and authorized to administer their own examinations.
reflect considerable characteristics of the region Students attending these schools are of low to
in terms of culture and educational expectations. middle SES.
Based upon the above review, family, school
and region, in this order, were assigned priorities D a t a source
in accounting for the variance in mathematics The data used in the present study were col-
achievement. Child-centered learning was then lected during the 1988-1989 school year, in a
MATttEMATICS ACHIEVEMENT 315
national evaluation project on the learning of good validity for a follow-up independent vari-
high school mathematics. The stratified random able. Often, sets of independent variables are used
sampling procedure enables this national assess- in hierarchical regression. Similarly, seven sets of
ment to have a fair representation of students independent variables were involved in this study.
from three types of schools (public, private ~lite Before explaining these sets, a brief description is
and private non-~lite) in three densely populated necessary on categorical independent variables.
geographical regions (Santo Domingo, Santiago Many independent variables contain exclu-
and the next ten major cities). The target popula- sively qualitative levels or categories. These kinds
tion is high school senior students. The original of categorical independent variables can not be
national sample contains over 1200 high school entered directly into any regression equation.
senior students from 49 schools. In the present Coding must be used as a method of creating
study, students with missing data were deleted vectors to represent the levels of a categorical
listwise from the analyses. The deletion procedure independent variable. The number of vectors
resulted in a sample of 1082 high school senior needed to do so is one less than the number of
students for analyses in this study. levels in the categorical independent variable. One
Mathematics achievement was measured of the coding methods is dummy coding in which
through two test forms. A total of 70 multiple- the values entered in the vectors are Is and 0s.
choice items (35 in each form) cover four major Each level is represented by one vector in which
curriculum strands in mathematics: arithmetic, that level is assigned a value of 1 and other levels
algebra, geometry and trigonometry. Items are are assigned a value of 0. The vector for the last
drawn from existing item banks of the Second level is a zero (or base-line) vector against which
International Mathematics Study (SIMS) and other vectors are compared. There were five cat-
British Columbia Grade 12 Provincial Mathemat- egorical independent variables in this study,
ics Examinations. Students need three kinds of namely gender, school type, provision of text-
cognitive skills (computation, comprehension and books, desire for future career and region. Dummy
application) to solve test items successfully. In coding was used to represent these categorical
addition to mathematics achievement tests, there independent variables. The first three columns in
is a 15-item student questionnaire that was used Table 1 present the seven sets of independent
to obtain information about students, their fam- variables and the coding of categorical independ-
ily background, career aspiration, attitude toward ent variables which was used in the regression
mathematics and some school characteristics. analyses.
The student questionnaire was administered to Note that, in a hierarchical regression analysis,
the same national sample that wrote mathematics a causal priority of sets of independent variables
achievement tests. in explaining the dependent variable is specified
(see Cohen and Cohen, 1983). Thus. a set of
Data analysis independent variables entered at a given step
There are a number of ways in which independ- should not be presumed to have any effect on
ent variables are entered into the regression equa- another set of independent variables entered at an
tion, known as simultaneous, stepwise and earlier step. The causal priority of sets of inde-
hierarchical regression I (Cohen and Cohen, 1983). pendent variables was decided upon several theo-
The first two statistical techniques are suitable for retical considerations that were discussed in the
prediction. Because this study was mainly explana- literature review. Student, family, school and
tory in nature, hierarchical regression was utilized region, in this order, were assigned causal priori-
with the purpose of: (a) examining mathematics ties in accounting for the variance in mathemat-
achievement, the dependent variable, under a ics achievement. There were four sets of
certain causal priority of independent variables; independent variables related to school. Their
and (b) controlling potentially confounding inde- causal priority in explaining mathematics achieve-
pendent variables. Cohen and Cohen (1983) and ment was determined as follows: (a) school types
Pedhazur (1982) provide the rationale for design- decide school characteristics; (b) attitude toward
ing a hierarchical regression analysis. Hierarchi- mathematics is affected by school t3pes and
cal regression is usually considered as theoretically school characteristics: and (c) desire for future
conservative and statistically rigorous, because it career is influenced by school types, school char-
ensures that the influence of preceding independ- acteristics and attitude toward mathematics. In a
ent variables is statistically controlled to gain hierarchical regression analysis, the proportion of
316 XIN MA
Set o f s t u d e n t c h a r a c t e r i s t i c s
Gender 1 = male I 1.625 5.663
2 = female 0 -- --
Age -- -- 18.073 1.097
N u m b e r o f b r o t h e r s a n d sisters -- - - 3 962 1.097
Set o f p a r e n t a l e d u c a t i o n a l levels
Father's e d u c a t i o n a l level -- -- 3.019 1.684
M o t h e r ' s e d u c a t i o n a l level -- -- 2.809 1.613
Set o f s c h o o l types I = public 10 0.268 0.443
2 = private 61ite 01 0.419 0.494
3 = private non-61ite 00 -- --
Set o f school c h a r a c t e r i s t i c s
Frequency of homework -- -- 1.608 0.871
Dally time for h o m e w o r k -- -- 3.004 1.068
Provision o f t e x t b o o k s 1 = yes 1 1.337 0.473
2=no 0 -- --
Set o f a t t i t u d e t o w a r d m a t h e m a t i c s -- -- 45.745 8.750
Set o f desire for future c a r e e r 1 = university s t u d y 10000 0.570 0.495
2 = full-time j o b 01000 0.007 0.086
3 = p a r t - t i m e j o b a n d university study 00100 0.339 0.474
4 = rest a n d then j o b 00010 0.003 0.053
5 = o t h e r plans 00001 0 039 0.193
6 = n o decision 00000 -- - -
variance explained by a set of independent vari- dents in the Dominican Republic. Independent
ables at a certain step over and above that explained variables were classified into different sets: the set
by previous sets of independent variables is the of student characteristics, including student gen-
most important statistic. Similarly, in this study, der, age and the number of brothers and sisters;
the significance of a set of independent variables the set of parental educational levels, including
was tested by examining the increment of R 2 for father's and mother's educational levels; the set of
the set over and above the R 2 for those sets entered school types; the set of school characteristics,
earlier. Most interpretations were based upon including frequency of homework, daily time for
these significance tests. homework and provision of textbooks; the set of
The final multiple regression equation is a attitude toward mathematics; the set of desire for
weighted linear composite of independent vari- future career; the set of geographical characteris-
ables. In that equation, non-significant regression tics. Categorical independent variables of gender,
coefficients are considered to have only trivial school type, provision of textbooks, desire for
effects on the dependent variable. These regres- future career and geographical region were all
sion coefficients, consequently their correspond- dummy coded for regression analyses. The descrip-
ing independent variables, are usually deleted to tive statistics of all the variables are listed in
achieve parsimony of a regression model (see Table 1.
Shelton, 1987). The simplified regression model is The results of the two multiple regression
then re-run to obtain necessary statistics. This analyses (the original and simplified models) are
procedure was followed in this study. presented in Tables 2 and 3. In the original
regression model, the set of desire for future
career was the only non-significant predictor set,
RESULTS F(5, 1066) = 1.614,p = 0.154. The percentage of
The dependent variable in this study was math- variance in mathematics achievement accounted
ematics achievement of high school senior stu- for uniquely by this set was 0.62%. This predictor
MATHEMATICS ACHIEVEMENT 317
Table 2. R squared, change m R squared, F ratm for the change, significance of the change (br the original model (N = 1082)
Notes: *p < 0.05. **p < 0.01. Set of student characteristics = (gender. age. number of brothers and sisters). Set of parental
educational levels = (father's education level, mother's education level) Set of school types = (public, private ,~lite,private
non-~lite). Set of school characteristics = (frequency of homework, dady time for homework, provision of textbooks). Set
of attitude toward mathematics = (attitude toward mathematics). Set of desire for future career = (university study, full-t~me
job, part-time job and university study, rest and then job, other plans, no decision) Set of geographical characteristics =
(Santo Domingo, Santiago, the next ten major cities). Gender, school t?pe. proviston of textbooks, desire lbr future career
and geographical characteristics were dummy coded.
Table 3. R squared, change in R squared, Fratio for the change, significanceof the change for the simplified model (N = 1082)
Notes: *p < 0.05. ** p < 0.01, Set of student characteristics = (gender, number of brothers and sisters) Set of parental
educational levels = (father's educational level). Set of school types = {pubhc. private 6hte, private non-~lite). Set of school
characteristics = (frequency of homework, daily time for homework, pro~~sion of textbooks). Set of attitude toward
mathematics = (attitude toward mathematics). Set of geographical characteristics = (Santo Domingo, Santiago, the next ten
major cities). Gender, school type, provision of textbooks and geographical characteristics were dummy coded
Table 4. Beta weights, correlations, part correlations, and F ratios tbr the orig,nal and simplified models
Notes: *p < 0.05. **p < 0.01. School type vectors, desire for future career vectors, or region vectors were entered as a set.
Dummy coding was used for each categorical variable. For gender, males (17 and females (07. For school type: public (10),
private glite (01) and private non-6lite (00). For provision of textbooks: yes (1) and no (0). For desire for future career:
university study (10000), full-time job (01000), part-time job and university study (00100), rest and then job (000107, other
plans (00001) and no decision (000000). For region: Santo Domingo (10), Santiago (017 and the next ten major cities (00).
the Dominican Republic. Data show that the level accounted for a significant amount of vari-
range of ages is from 12 to 24 years. This study ance in mathematics achievement. The explana-
shows that student age, as a predictor variable, tion, according to the implication outlined above,
had only a trivial effect on mathematics achieve- is that father's SES varies dramatically and,
ment in grade 12. Because of the lack of special therefore, stimulates father's educational level to
programs for slow and gifted students in develop- be effective in explaining mathematics achieve-
ing countries, students of different ages usually ment. But the effect of father's SES on mathemat-
attend the same grade. The finding implies that ics achievement cannot be exaggerated. Table 4
letting gifted students 'jump' grades and slow shows that one unit change in father's educational
students repeat grades is probably a workable level causes a small change of 0.09 in the stand-
coping strategy to the lack of educational ardized score of mathematics achievement
resources in developing countries. (equivalently 2 percentage points). What was
Another interesting finding is that mother's found in this study on the relationship between
educational level was not statistically reliable in parental educational levels and mathematics
explaining students' mathematics achievement. achievement of their children may be typical in
Note that, in the Dominican Republic, as in many many developing countries that share a similar
developing countries, although society has begun situation to that of the Dominican Republic
to change, most mothers still work at home. regarding mother's and father's SES.
Therefore, almost all students have a similar Educators often ask whether students' plans
mother's SES. The finding implies that mother's for future careers play an important role in their
educational level affects mathematics achieve- study of mathematics. The answer seems negative
ment mainly through mother's SES. Stated differ- in the Dominican setting. In the original model,
ently, mother's educational level is strongly desire for future career was the only non-
moderated by mother's SES. If mother's SES significant predictor set. In Table 4, part correla-
becomes constant, it constrains mother's educa- tions for this set also indicate that the unique
tional level from actively affecting mathematics contribution made by vectors of this set to
achievement. In contrast, father's educational mathematics achievement was trivial. The finding
MATHEMATICS ACHIEVEMENT 319
is in disagreement with not only people's common ciaily in terms of educational policies and expec-
sense but also some studies conducted in devel- tations. Therefore, school characteristics may
oped countries on the relationship between career affect individual learning more directly than region
aspiration and mathematics achievement (e.g. characteristics.
Bieschke and Lopez, 1991,; Cooper and Robin- Interestingly, in the Dominican setting, family
son, 1991; Pedersen et al., 1986). Note that, when fails to show significant contribution to math-
entering the set of desire for future career into the ematics achievement in the sociological context.
regression equation, there had already been five Note, however, that only parental educational
sets of independent variables in the equation. levels were used to present family characteristics
Therefore, it is still possible for the set of desire for in this study. Perhaps, they are inadequate to fully
future career to have some power to explain reflect family characteristics and, therefore, con-
mathematics achievement, but it shares the power strain the importance of family background in
with those sets preceding it in the equation. One explaining mathematics achievement. Further
way to rule out this speculation is to examine the studies are recommended to involve more family
correlation between career aspiration and math- characteristics, such as SES and home environ-
ematics achievement. Because correlations ment variables, in the examination of child-
between mathematics achievement and vectors in centered learning in the sociologically specified
the set of desire for future career are very small in context.
Table 4, it is less possible for this set to share a Finally, I discuss the individual predictor vari-
meaningful power to explain mathematics achieve- ables in the simplified model. As mentioned
ment with those sets of independent variables before, the final regression equation is a weighted
preceding it. Therefore, the conclusion is sustain- linear composite of individual variables. The
able that students' desire for future career has no weighted standardized regression coefficient (13)
significant contribution to their mathematics
for each individual variable indicates its impor-
achievement in the Dominican setting.
tance in the regression equation. Because the
Recall that, in the present study, child-centered
sample size was large in this study, the practical
learning was put into a hierarchical sociological
meaning or significance of the statistically signifi-
context of family, school and region in which the
cant results (Table 4) are taken into account.
set of student characteristics was assigned an
exclusive priority for the explanation of math- First, gender differences were statistically signifi-
ematics achievement. The set of student charac- cant in mathematics achievement, but may not be
teristics did continue to be the most important practically meaningful. From male to female,
predictor set for mathematics achievement. Child- mathematics achievement dropped about 2 per-
centered learning seems applicable to the hierar- centage points. Thus, there were no evident gen-
chical sociological context of family, school and der differences in mathematics achievement
region: Although individual learning interacts among the high school seniors.
with characteristics of family, school and region, Students with fewer brothers and sisters had
individual learners are still most responsible for better mathematics achievement. However, one
their mathematics achievement. It can, therefore, more brother or sister caused mathematics
be speculated that the influences from family, achievement to decrease by about 1 percentage
school and region have to be meaningfully inter- point. As a result, there were no substantial dis-
nalized by individual learners before they are advantages regarding mathematics achievement
capable of affecting individual learning. School for students from large families. It is undeniable
and region also seem to have important functions that students with a large number of brothers and
in the hierarchical sociological context, being sisters at home may receive less care in terms of
significant predictors for mathematics achieve- their living and learning from their parents than
ment. Note that sets of school type and school students with fewer brothers and sisters. But it
characteristics together explained twice as much seems that the care from'educators may compen-
variance in mathematics achievement as the set of sate for the lack of care from parents with respect
geographical characteristics, indicating that school to the learning of mathematics. As long as edu-
characteristics are much more important than cators pay enough attention to students from
region characteristics. One speculation is that large families, there may not be evident differ-
school characteristics are capable of reflecting a ences in mathematics achievement between stu-
sizable portion of region characteristics, espe- dents with and without a large number of brothers
320 XIN MA
Education and National Systems of Educatton pp. 236-238 Lennon, O. (1988) Cultural variations, cogmtwe styles, and
Pergamon, Oxford. educatton m Latin America. Prospects 28, 413--420
Fuller. B. and Heyneman, S P. (1989) Third world school Luna, E., Gonz~tlez, S., Robltadle, D. Crespo. S. and Wolfe,
quality: Current collapse, future potential Educational R. (1995) Improving the teaching and learning of math-
Researcher 18, 12-19. ematlCS in the Dominican Republic Journal of Curriculum
Goldstein. H (1987) Multdevel Models in Educational and Studtes 27, 67-79
Social Research. Oxford University Press, Oxford. Merton, R K. (Ed) (1968) Socsal Theory and Soctal Struc-
Hoffer, T.. Greeley, A. M. and Coleman. J. S. (1985) Achieve- ture. Free Press, New York.
ment growth in public and private schools. Sociology of Miller, J P and Seller, W (1990) Currtculum Perspectives and
Education 58, 74-97 Practice Copp Clark Pitman, Toronto.
Hungwe, K. (1989) Culturally appropriate media and technol- Pedersen, K. Elmore, P. and Bleyer. D (1986) Parent attitude
ogy: A perspective from Zimbabwe. Tech Trends 34, 22-23 and student career interests in junior high school Journal
Jencks, C., Acland, S. H., Bane, M. J., Cohen, D., Gintls, H , for Research m Mathematics Educatton 17, 49-59
Heyns, B. and Michelson, S. (1972). Inequality: A Reassess- Pedhazur, E J. (1982) Multtple Regresston in Behavtoral
raent of the Effect of Family and Schoohng in Amertca. Research Explanation and Predtctton Holt. Rinehart &
Harper & Row, New York. Winston, New York.
Leder, G. C. (1992) Mathematics and gender: Changing per- Shelton. E A (1987) Using regression analysis: A guided tour
spectlves. In (Edited by Grouws, D. A.), Handbook of Computers and Education 11, 95-111.
Research on Mathemattcs Teaching and Learning. pp. 597- Stromquist. N. P (1989) Determinants of educational partici-
622. Macmillan, New York. pation and achievement of women in the third world: A
Lee, V. E. (1986) Catholic school minority students have review of the evidence and a theoretical critique. Revtew of
'reading proficiency advantage'. Momentum 17, 20-24. Educattonal Research 59, 143-183.
Lee, V. E. and Bryk, A. S. (1988) Curriculum tracking as Theisen, G. Achola, P and Boakari. F (1983) The undera-
mediating the social distribution of high school achieve- chievement of cross-national studies of achievement Corn-
ment Sociology of Education 61, 78-94. parattve Education Review 27, 46-68