Citation
Exploring Reading and Math Masteries

Material Information

Title:
Exploring Reading and Math Masteries Influences of Classroom and Family
Creator:
Donaldson, Kristi
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (45 p.)

Thesis/Dissertation Information

Degree:
Master's ( M.A.)
Degree Grantor:
University of Florida
Degree Disciplines:
Sociology
Committee Chair:
Shehan, Constance L.
Committee Members:
Ardelt, Monika
Graduation Date:
8/8/2009

Subjects

Subjects / Keywords:
Academic achievement ( jstor )
Child psychology ( jstor )
Classrooms ( jstor )
Demography ( jstor )
Fathers ( jstor )
Mathematics ( jstor )
Mothers ( jstor )
Multilevel models ( jstor )
Parents ( jstor )
Schools ( jstor )
Sociology -- Dissertations, Academic -- UF
achievement, classroom, education, family, math, reading
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Sociology thesis, M.A.

Notes

Abstract:
Using data from the public-use Early Childhood Longitudinal Study: Kindergarten Class of 1998-1999, Fifth Grade wave, this study looked at the influences of demographic, classroom and family characteristics on children's elementary-level mastery skills of reading and math. To examine these relationships, five hypotheses were tested using hierarchical linear modeling. Female students were found to be more likely to have higher average reading achievements than males, where the opposite occurred in mathematics. For both subjects, White respondents had higher average achievement scores than Black or Hispanic children. Parental education, household income and expected educational achievements were all positively associated with both reading and math achievement scores. Classroom size, but not teacher experience or gender composition of classroom, had a significant, positive effect with both math and reading achievement scores. Findings highlight the need for continued emphasis on education by parents, educators and the media. Limitations and suggestions for future research are discussed. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (M.A.)--University of Florida, 2009.
Local:
Adviser: Shehan, Constance L.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-08-31
Statement of Responsibility:
by Kristi Donaldson.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Donaldson, Kristi. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
8/31/2010
Resource Identifier:
489240960 ( OCLC )
Classification:
LD1780 2009 ( lcc )

Downloads

This item has the following downloads:


Full Text





Hyde, Janet S., Elizabeth Fennema, and Susan J. Lamon. 1990. "Gender Differences in
Mathematics Performance: A Meta-Analysis." Psychological Bulletin 107:139-55.

Ireson, Judith, Susan Hallam, Sarah Hack, Helen Clark, and Ian Plewis. 2002. "Ability Grouping
in English Secondary Schools: Effects on Attainment in English, Mathematics and
Science." Educational Research and Evaluation 8:299-318.

Jackson, Carolyn, and Ian D. Smith. 2000. "Poles Apart? An Exploration of Single-Sex and
Mixed-Sex Educational Environments in Australia and England." Educational Studies
26:409-22.

Kandel, Denise B. 1980. "Drug and Drinking Behavior Among Youth." Annual Review of
Sociology 6:235-85.

Kowaleski-Jones, Lori, and Greg J. Duncan. 1999. "The Structure of Achievement and Behavior
across Middle Childhood." Child Development 70:930-43.

Laureau, Annette. 2003. Unequal Childhoods: Class, Race and Family Life. Berkeley, CA:
University of California Press.

Lee, Jihyun, Wendy S. Grigg, and Gloria S. Dion. 2007. The Nation's Report Card. : [ItAlhe,1tii
2007. Washington, DC: U.S. Department of Education, National Center for Education
Statistics.

Lee, Jihyun, Wendy S. Grigg, and Patricia L. Donahue. 2007. The Nation's Report Card:
Reading 2007. Washington, D.C.: U.S. Department of Education, National Center for
Education Statistics.

LoGerfo, Laura, Austin Nichols, and Duncan Chaplin. 2006. Gender Gaps in Math and Reading
Gains during Elementary and High School by Race and Ethnicity. Washington, DC: Urban
Institute.

Muller, Chandra. 1998. "Gender Differences in Parental Involvement and Adolescents'
Mathematics Achievement." Sociology of Education 71:336-56.

Nye, Barbara A., Larry V. Hedges, and Spyros Konstantopoulos. 2004. "Do Minorities
Experience Larger Lasting Benefits from Small Classes?" The Journal of Educational
Research 98:94-100.

O'Connell, Ann A., and D.B. McCoach (Eds). 2008. Multilevel Modeling of Educational Data.
Charlotte, NC: Information Age Publishing Inc.

Painter, John. 2002. Designing Multilevel Models Using SPSS 11.5 Mixed Model. Chapel Hill,
NC: Jordan Institute for Families.

Parsons, Jacquelynne E., Terry F. Adler, and Judith L. Meece. 1984. "Sex Differences in
Achievement: A Test of Alternative Theories." Journal ofPersonality and Social
Psychology 46:26-43.










Table 3-1. Continued
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6


Classroom Characteristics
Years Teaching 0.075 / 0.078
Number of Students 0.319 / 0.216 "
Percent Male -0.047 / -0.046
Intercept 49.557 49.589" 52.408
Classroom Variance 69.417 64.572 58.985
Change in Classroom Variance 4.845 10.432
Class. Proportion of Variance 0.07 0.15
Student Variance 37.734 37.771 37.528
Student Proportion of Variance -0.001 0.005
X2 21442.988' 21168.712' 18799.270'
N 10874 4265 10874
Source: Early Childhood Longitudinal Survey, Kindergarten Class of 1998-1999, Fifth Grade Wave
Unstandardized / Standardized Coefficients
* p <.05; p .01;* p<.001


51.990*
40.626
28.791
0.415
35.019
0.072
15308.517
10874


52.175*
37.794
31.623
0.456
33.411
0.115
15407.958 "
10874


0.028 / 0.029
0.230 / 0.155 *
-0.017 /-0.017
52.258 "
36.006
33.411
0.481
33.346
0.116
15103.600 "
4265

































To all of the people who have put up with my seclusion while working on this project...you
know who you are










Table 3-2. Hierarchical linear modeling results of coefficients of between-classroom models of fifth-grade math achievement
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6


Demographic Variables
Gender (male)
Race (White)
African American
Hispanic
Asian
Other
Household Characteristics
Mother's Age
Mother's Education
Mother Employed
Father's Age
Father's Education
Father Employed
Marital Status (Married)
Single
Separated
Divorced
Widowed
Income
Number in Household
Time for Homework
Parental Involvement
Expected Education


1.592 / 0.085


-5.652 / -0.175
-3.290 / -0.133
0.504/ 0.013
-2.312 / 0.059


1.506 / 0.080 *


-3.391/-0.112*
-0.897 / -0.037
1.065 / 0.027
-0.990 / -0.025


-0.001 /-0.001
0.845 / 0.162 *
0.044/ 0.002
0.052/ 0.040
0.448 / 0.100 *
-0.757 /-0.023


-1.157 /-0.034
-0.383 /-0.008
-0.611 /-0.021
-1.097 /-0.014
0.446 / 0.150 *
-0.291 /-0.043


1.774 / 0.094


-3.972 / -0.123
-1.590 / -0.064
0.140 / 0.004
-1.297 / -0.033


0.006/ 0.004
0.668 / 0.128 "
0.139 / 0.007
0.058 / 0.044*
0.346 / 0.077
-0.770 / -0.023


-0.866 / -0.026
-0.131 /-0.003
-0575 / -0.020
-1.285 /-0.016
0.394 / 0.133 *
-0.251 /-0.037
-0.002 / -0.007
0.099 / 0.017
1.592 / 0.176


1.806/0.096


-3.943 / -0.122
-1.916 / -0.077
-0.172 /-0.004
-1.355 /-0.035


0.001 / 0.001
0.652 / 0.125 *
0.211 /0.010
0.066/ 0.051 *
0.361 / 0.081
-0.701 /-0.021


-0.825 / -0.025
-0.048 / -0.001
-0.568 / -0.019
-1.353 /-0.017
0.368 / 0.124 *
-0.251 /-0.037
-0.003 / -0.011
0.113 /0.019
1.532 / 0.170










LIST OF FIGURES


Figure


2-1. Conceptual M odel............ .... ............................................................................................. 23


page









Finally, for the last model, the average classroom reading achievement was 52.26, while

the average classroom math achievement was 50.88, controlling for gender, race, family

characteristics and some classroom characteristics as well. About 48.1 percent of the explainable

variation in between-classroom average reading achievement scores, and 62.2 percent of

variation in between-classroom average math achievement scores could be explained by all

predictors: gender, race, family and classroom characteristics. Finally, only 11.6 percent of the

within-classroom reading achievement variation, and 4.3 percent of the within-classroom math

achievement variation could be explained by the combination of all predictors. Similar to the last

model, the parental expectations of educational achievement had the strongest effect on

achievement scores.

In this model, the sample size decreased from 10,874 to 4,265 for reading and to 2,730 for

math. This decrease again accounted for the smaller number of classroom observations than

student observations. Here, HLM averaged the student and family characteristics for each

classroom to use for analysis.

For reading, male respondents (y10=-.94, p<.01) were found to have lower average

expected reading scores than females, by .94 points. Black (y20=-3.99, p<.001), Hispanic (y30=-

2.42, p<.001), Asian (y40=-2.41, p<.001) and other race (y50=-1.91, p<.001) respondents were

found to have lower average expected reading scores than White respondents.

Higher education for mothers (y70=.69, p<.001) and fathers (y100=.41, p<.001), older

father's age (y90=.13, p<.001), and higher household incomes (y160=.39, p<.001) were

significantly associated with higher average expected reading scores. At the same time,

household size (y170=-.65, p<.001) was negatively associated with reading scores, where with

each person added to the household, the expected average reading score decreased by .65 points.









TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ......... .................................................................................................4

L IST O F T A B L E S ........... ... ........................................................ ................................. 6

LIST OF FIGURES .................................. .. .... ... ...............7

A B S T R A C T ...................................................................................................... . 8

CHAPTER

1 IN TR O D U C TIO N .............................................. .............. ...........................

Specific A im s........................................................ 9
B background and Significance ......................................................................... ......... ......... 10
H ypotheses....... .........................................................15

2 RESEARCH DESIGN AND METHODS ............................................................... ........ ..16

P ro c e d u re ................... ...................1...................6..........
S a m p le .............. ..... ..............................................................1 6
M e a su re s ................... ...................1...................7..........
Analysis ........................................ 19
Equations .......................................19

3 R E S U L T S ..............................................................................................2 4

4 CON CLU SION ... .................................................36

Findings .............................. ........ .......................... ......... 36
Limitations and Suggestions for Future Research ........................................................39

R E F E R E N C E S ..........................................................................42

B IO G R A PH IC A L SK E T C H .................................................................................................... 45









Raudenbush, Stephan W., and Anthony S. Bryk. 1986. "A Hierarchical Model for Studying
School Effects." Sociology of Education 59:1-17.

2002. Hierarchical Linear Models: Applications and Data Analysis Methods (2nd ed.).
Thousand Oaks, CA: Sage.

Raudenbush, Stephen, Anthony Bryk, and Richard Congdon. 2005. HLM6. Lincolnwood, IL:
Scientific Software International.

Roberts, J.K. 2004. "An Introductory Primer on Multilevel and Hierarchical Linear Modeling."
Learning Disabilities: A Contemporary Journal 2:30-8.

Sadker, Myra, and David Sadker. 1995. Failing at Fairness: How Our Schools Cheat Girls. New
York: Charles Scribner's Sons.

Singer, Judith D. 1998. "Using SAS PROC MIXED to Fit Multilevel Models, Hierarchical
Models, and Individual Growth Models." Journal of Educational and Behavioral Statistics
24:323-55.

Van de gaer, Eva, Heidi Pustjens, Jan Van Damme, and Agnes DeMunter. 2004. "Effects of
Single-Sex Versus Co-Educational Classes and Schools on Gender Differences in Progress
in Language and Mathematics Achievement." British Journal of Sociology of Education
25:307-22.

2006. "The Gender Gap in Language Achievement: The Role of School-Related Attitudes
of Class Groups." Sex Roles 55:397-408.

Vilenius-Tuohimaa, Piia M., Kaisa Aunola, and Jari-Erik Nurmi. 2008. "The Association
Between Mathematical Word Problems and Reading Comprehension." Educational
Psychology 28:409-26.









ACKNOWLEDGMENTS

First, I thank my chair, Connie Shehan, for her support, both academically and personally,

throughout the last three years. I also thank Monika Ardelt for her attention to detail and

insistence on "doing things right," both of which have helped improve this paper. I am grateful

for Graham, who helped ease tensions by making me laugh and providing breaks away from my

work, and Becky and Jackie for being my cheerleaders now and always.

I thank my Dad for supporting me even though he does not always understand my goals or

intentions, my sister who tells me I can do anything, and my mom for giving me space when

needed and encouraging me to push forward. Finally, I thank Brewer for providing relief from

stress and always making me smile. I also appreciate all others my roommates, colleagues,

coworkers who have supported me along the way.









normally distributed with mean 0 and between classroom variance) and a random error (rij,

normally distributed with mean 0 and within classroom variance) associated with the ith student

in the jth classroom (Singer 1998).

DVij = yOO + u0j + rij (2-1)

The next equation (2-2) represents model 2, where level-2 variables (classroom

characteristics) were modeled against the dependent variables to determine how classroom

characteristics alone affect academic achievement. Variables with a 'GM' were grand-mean

centered for analysis, meaning the explanatory variables were centered on the overall mean of

that variable (Raudenbush & Bryk, 2002). Centering is frequently used in HLM to aid in

interpretations of results. In this equation for example, the intercept y00 represents the expected

outcome for a student who is in a classroom with a teacher of average experience (average

number of years teaching), an average number of students and an average percentage of boys. If

the variables were left in their original metric, the intercept would represent the expected

outcome for a student in a classroom where the teacher has no previous experience (since teacher

questionnaires were given toward the end of the school year instead of the beginning, no teacher

had 0 experience), without students (which is impossible a classroom then would not exist) and

without boys (which is not included in this data). See Table 2-2 for a listing of the means and

standard deviations of all centered variables. The following is the model-2 equation with level-2

variables grand-mean centered:

DVij = y00 + y01*(GM of YRSTEACHij) + y02*(GM of NUMBERKIDSij)+ y03*(GM
of PERCENTBOYSij)+ uOj + rij (2-2)

The next three models represent the level-1 equations run for analyses. In model 3, the

dependent variables were modeled as a function of gender and race (White omitted). Following

previous studies, gender and race were first examined, as they are frequently cited as potential









(and often main) predictors of academic achievement. By including these variables first, I was

able to examine how influence of gender and race might vary depending on additional predictors

included in the models. Dummy variables were not centered in the analyses due to interpretative

results; by not centering, I examined the difference in average achievement scores between males

and females, White and Black students, etc. Equation 2-3 represents the HLM analysis for

gender and race, cited frequently as potential predictors of academic achievement:

DVij = yOO + ylO*MALEij + y20*BLACKij + y30*HISPANICij + y40*ASIANij +
y50*OTHERij + uOj + rij (2-3)

In model 4, the dependent variables are modeled as a function of gender, race, mothers' &

fathers' ages, educations and employment statuses, marital status (married omitted), household

income, and the number of people living in the household (Equation 2-4). Basic demographic

characteristics of the family were added to this model to see how family characteristics affected

achievement scores and also the effects of race and gender:

DVij = yOO + ylO*MALEij + y20*BLACKij + y30*HISPANICij + y40*ASIANij +
y50*OTHERij + y60*(GM of MOMAGEij) + y70*(GM of MOMEDUCij) +
y80*(MOMEMPLOYij) + y90*(GM of DADAGEij) + ylOO*(GM of DADEDUCij) +
yl 10*(DADEMPLOYij) + yl20*(SINGLEij) + yl30*(SEPARATEDij) + yl40*(DIVORCEDij)
+ yl50*(WIDOWEDij) + yl60*(GM of INCOMEij) + yl70*(GM of NUMinHHij) + uOj + rij
(2-4)

In model 5, the dependent variables are modeled as a function of gender, race, mothers' &

fathers' ages, educations and employment statuses, marital status (married omitted), household

income, the number of people living in the household, the time spent on homework, parental

involvement and the degree expected of the child (Equation 2-5). Here, three variables

representing non-demographic family variables were added:

DVij = yOO + ylO*MALEij + y20*BLACKij + y30*HISPANICij + y40*ASIANij +
y50*OTHERij + y60*(GM of MOMAGEij) + y70*(GM of MOMEDUCij) +
y80*(MOMEMPLOYij) + y90*(GM of DADAGEij) + ylOO*(GM of DADEDUCij) +
yl 10*(DADEMPLOYij) + yl20*(SINGLEij) + yl30*(SEPARATEDij) + y140*(DIVORCEDij)


































2009 Kristi Lynn Donaldson









LIST OF TABLES


Table page

2-1. Participant characteristics, weighted (N=10,874) ...................................... ............... 22

2-2. M eans and standard deviations of centered variables ................................. ............... 23

3-1. Hierarchical linear modeling results of coefficients of between-classroom models of
fifth-grade reading achieve ent ............................................... ............................. 32

3-2. Hierarchical linear modeling results of coefficients of between-classroom models of
fifth-grade m ath achieve ent ................................................. ............................... 34









sectional nature of the data, it is unknown if these differences have been increasing or decreasing

since the beginning of school.

Higher incomes were associated with higher subject achievement scores, perhaps due to

an increasing availability of resources outside of school to aid in academic advancement (Lareau

2003). At the same time, larger household sizes were associated with lower reading scores, but

had no significant effects on math scores. Familial motivation also had significant impacts on

average academic achievement scores, with high parental educational expectations associated

with high average achievement scores. In the last two models, parental educational expectations

had the strongest effects on reading and math achievements. Also, across models 4 through 6,

mother's education had stronger impacts on achievement than father's education did. This

demonstrates that continued emphasis on education is important for the future of our nation's

children, and perhaps even more so for young females.

Though there were many similarities among the predictors of high average achievement

scores for both reading and math, there were also some differences. For example, why did

household size have a negative effect on reading scores but not on math ones? Why was parental

involvement significantly associated with high reading scores but, again, not math ones? What is

it about the subject of reading, and perhaps the homework assigned for it, that is significantly

affected by these family characteristics? Math and reading are two dissimilar subjects, and there

are often different class activities, homework exercises and demonstrations in the learning of

each. Perhaps the differences in how people learn and process math and reading could account

for some of these conflicting findings between the two subjects. Additionally, the observed

differences between the subjects aid in cautioning parents and educators about "one size fits all"

policies aimed at improving achievement levels.









CHAPTER 1
INTRODUCTION

Specific Aims

According to the National Center for Education Statistics in 2007, 39 percent of fourth-

grade students were proficient in their math level, and 33 percent were proficient in the reading

level. Of all math scores, Black males and females, Hispanic males and females, as well as white

females scored lower than white males. Of all reading scores, Black males and females, Hispanic

males and females, as well as white males scored lower than white females. What are some of

the causes of these observed differences in proficiencies?

Questions of achievement gaps between boys and girls have been widely covered in

mainstream media over the course of the last twenty years. Within a similar time period, there

has been an introduction of same-sex classrooms in public schools to help bridge the gap in

achievement and proficiency between boys and girls in different subjects. The idea behind these

classrooms is that boys and girls learn differently, and separating the genders will benefit

students in both achievement and confidence.

To explore these potential differences, I used data from the Early Childhood Longitudinal

Survey, Kindergarten Class of 1998-1999 Fifth Grade wave. Although data on same-sex

classrooms was not available in this study, there were still a number of measures that could be

used to examine the differences in children's academic achievements or proficiencies. I aimed to

study characteristics of the child's environment, including the classroom and family, to explore

differences in children's reading and math achievement scores.

The quality of education children receive may be a direct result of the resources and

opportunities afforded unto them. Children living in higher-income areas may have access to

schools and classrooms with newer technologies and greater learning resources. Not all










Table 3-1. Hierarchical linear modeling results of coefficients of between-classroom models of fifth-grade reading achievement
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6


Demographic Variables
Gender (male)
Race (White)
African American
Hispanic
Asian
Other
Household Characteristics
Mother's Age
Mother's Education
Mother Employed
Father's Age
Father's Education
Father Employed
Marital Status (Married)
Single
Separated
Divorced
Widowed
Income
Number in Household
Time for Homework
Parental Involvement
Expected Education


-1.169/-0.061


-5.721 /-0.186
-4.328 / -0.177
-2.669 / -0.067
-3.255 / -0.078


-1.325/-0.069" -0.995/-0.052


-3.281 /-0.106 *
-1.261 /-0.051 *
-1.405 /-0.035 *
-1.187 /-0.029


-0.067 /-0.046
0.922 0.175
-0.126 /-0.006
0.115 /0.088
0.533 /0.118
-0.435 /-0.013


-0.372 /-0.011
-0.040 /-0.001
-1.123 / -0.038 *
-0.157 /-0.002
0.498 / 0.166
-0.699 / -0.102


-3.905/ -0.127
-2.001 /-0.082
-2.151 /-0.054
-1.835 / -0.044


-0.066 / -0.046
0.703 / 0.134"
-0.084 /-0.004
0.115 / 0.087
0.416 / 0.092
-0.541 /-0.016


-0.071 /-0.002
-0.070 / -0.001
-0.907 / -0.030
-0.377 / -0.005
0.417 / 0.139
-0.658 / -0.096
0.005 / 0.021
0.213 / 0.036*
1.795 / 0.197


-0.938 / -0.049


-3.994 / -0.130
-2.421 / -0.099
-2.407 /-0.060
-1.906 / -0.046


-0.069 / -0.048
0.686/0.131
-0.048 / -0.002
0.125 / 0.095
0.409 / 0.090 *
-0.654 / -0.019


-0.015 /-0.001
-0.035 / -0.001
-0.960 / -0.032
-0.392 / -0.005
0.390 / 0.130
-0.653 / -0.095 *
0.004 / 0.018
0.204 / 0.034*
1.717 / 0.189









at home? If so, there may be subject-area differences in the type of homework or interaction

between parents and children.

The fourth hypotheses (4a and 4b), relating gender composition to academic achievement,

were not supported in either subject. Interaction-level effects of gender and classroom gender

distribution might be useful in future analyses to determine the impact, if any, the combination of

child's gender and classroom composition of gender has on overall classroom achievement (and

to aid in determining the 'minority' and 'majority group (Dryler 1999)).

The last hypothesis that class size is negatively associated with reading and math

achievement scores, was not supported, contradicting the findings of Funkhouser (2009) and

Nye, Hedges, and Konstantopoulos (2004). For both subjects, larger classroom sizes were

associated with higher achievement scores. These findings could be contradictory due to the

difficulty in separating the effects of classroom size with other characteristics, such as teacher

experience, which often have interactive effects with one another (Funkhouser 2009; & Nye,

Hedges and Konstantopoulos 2004). The relationship of classroom size to student achievement

may be affected (and perhaps misleading) by not taking these interaction-level effects into

account.

Results also indicated a racial gap in academic achievement, supporting the Monitoring the

Future statistics in the Specific Aims. Across all models, Black students consistently lagged

behind their White counterparts in both reading and math scores, while Hispanic and Asian

students fell behind White counterparts in reading. Across all models, Black students scored

lower than the national average in both reading and math, highlighting the continued need for

specialized attention and resources targeted at this at-risk group. Again though, due to the cross-









and Sadker 1995), especially in asking and answering questions (Hyde, Fennema, and Lamon

1990).

When testing the effects of single-sex and coeducational Dutch classrooms, Van de gaer,

Pustjens, Van Damme, and De Munter (2004) found there to be gendered differences in

achievements. Boys' reading scores were improved by coeducational classes, with speculation

that the addition of girls aids in socializing boys to be more attentive and less accepting of

deviant or disruptive behaviors (Jackson and Smith 2000). At the same time, although no

differences were found in girls' achievements in coeducational and same-sex classrooms, Van de

gaer, et al. (2004) found there to be improvements in mathematics scores in same-sex schools.

Additionally, studies have found that when controlling for household characteristics, such as

income, parents' educational levels and the like, the differences between coeducational and

same-sex classrooms and schools are reduced (Harker 2000).

Although largely cited in context with deviance, social learning theory can also be used to

look at the primary institutions of education and family to examine their effects on a child's

educational achievement. In this theory, behavior is learned through the operant conditioning of

"groups which compromise or control [an] individual's major sources of reinforcement"

(Burgess and Akers 1966:140). According to this theory, parents who place a high value on

involvement in school, and have high academic expectations may pass on these values to their

children, resulting in higher achievement. Children observe their parents' behaviors, either

through the encouragement for higher education, stories of their educational pursuits and

outcomes, or actual observations of their parents attending school. After internalizing these

behaviors, children may begin to exhibit behaviors approved and valued by their significant

others, conditioned through social reinforcement (Kandel 1980); parents may encourage high









REFERENCES


Benbow, Camilla P., and Julian C. Stanley. 1980. "Sex Differences in Mathematical Ability:
Fact or Artifact?" Science 210:1262-64.

Burgess, Robert L., and Ronald L. Akers. 1966. "A Differential Association: Reinforcement
Theory of Criminal Behavior." Social Problems 14:128-47.

Cooper, Carey E., and Robert Crosnoe. 2007. "The Engagement in Schooling of Economically
Disadvantaged Parents and Children." Youth & Society 38:372-91.

Crosnoe, Robert. 2001. "Academic Orientation and Parental Involvement in Education during
High School." Sociology of Education 74:210-30.

Davis-Kean, Pamela E. 2005. "The Influence of Parent Education and Family Income on Child
Achievement: The Indirect Role of Parental Expectations and the Home Environment."
Journal of Family Psychology 19:294-3 04.

Dee, Thomas S. 2006. "The Why Chromosome." Education Next 6:69-75.

Ding, Cody S., Kim Song, and Lloyd I. Richardson. 2006. "Do Mathematical Gender
Differences Continue? A Longitudinal Study of Gender Difference and Excellence in
Mathematics Performance in the U.S." Educational Studies 40:279-95.

Dryler, Helen. 1999. "The Impact of School and Classroom Characteristics on Educational
Choices by Boys and Girls: A Multilevel Analysis." Acta Sociologica 42:300-18.

Funkhouser, Edward. 2009." The Effect of Kindergarten Classroom Size Reduction on Second
Grade Student Achievement: Evidence from California." Economics of Education Review
28:403-14.

Guo, Guang, and Kathleen M. Harris. 2000. "The Mechanisms Mediating the Effects of Poverty
on Children's Intellectual Development." Demography 37:431-47.

Halle, Tamara G., Beth Kurtz-Costes, and Joseph L. Mahoney. 1997. "Family Influences on
School Achievement in Low-Income, African American Children." Journal of Educational
Psychology 89:527-37.

Hanushek, Eric A., John F. Kain, Jacob M. Markman, and Steven G. Rivkin. 2003. "Does Peer
Ability Affect Student Achievement?" Journal ofApplied Econometrics 18:527-44.

Harker, Richard. 2000. "Achievement, Gender and the Single-Sex/Coed Debate." British Journal
of Sociology of Education 21:203-18.

Haveman, Robert, and Barbara Wolfe. 1995. "The Determinants of Children's Attainments: A
Review of Methods and Findings." Journal of Economic Literature 33:1829-78.









EXPLORING READING AND MATH MASTERIES: INFLUENCES OF CLASSROOM AND
FAMILY




















By

KRISTI LYNN DONALDSON


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF ARTS

UNIVERSITY OF FLORIDA

2009









Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Arts

EXPLORING READING AND MATH MASTERIES: INFLUENCES OF CLASSROOM AND
FAMILY

By

Kristi Lynn Donaldson

August 2009

Chair: Connie Shehan
Major: Sociology

Using data from the public-use Early Childhood Longitudinal Study: Kindergarten Class of

1998-1999, Fifth Grade wave, this study looked at the influences of demographic, classroom and

family characteristics on children's elementary-level mastery skills of reading and math. To

examine these relationships, five hypotheses were tested using hierarchical linear modeling.

Female students were found to be more likely to have higher average reading achievements than

males, where the opposite occurred in mathematics. For both subjects, White respondents had

higher average achievement scores than Black or Hispanic children. Parental education,

household income and expected educational achievements were all positively associated with

both reading and math achievement scores. Classroom size, but not teacher experience or gender

composition of classroom, had a significant, positive effect with both math and reading

achievement scores. Findings highlight the need for continued emphasis on education by parents,

educators and the media. Limitations and suggestions for future research are discussed.









CHAPTER 4
CONCLUSION

Findings

The results support the first three hypotheses (la, b and c). Gender did have a significant

association with achievement scores of both reading and math. Supporting current research and

speculations from the field, females' average scores were higher in reading, and males' were

higher for math. In both subjects however, girls and boys achieved at or above the national

average. Though results may appear small males and females were within two points of one

another for average expected reading and math scores due to the cross-sectional analyses, it is

unknown if these differences have been decreasing or increasing since kindergarten.

Interestingly, for reading, as more predictors were added to the model, the average score

difference between males and females decreased, but increased for math scores (outside of the

addition of family characteristics). It would appear as if math classes with more heterogeneous

characteristics are more beneficial to girls' advancements, while reading classes with more

homogeneous characteristics benefit males.

The second hypothesis is supported, with parental education positively associated with

subject achievement scores. For both math and reading, both mothers' and fathers' education

levels were positively associated with average achievement scores across models 4 through 6.

This finding supports social learning theory and previous work of Vilenius-Tuohimaa et al.

(2008) and Hayeman and Wolfe (1995).

The third hypothesis was partially supported. Parental involvement had a positive,

significant effect for reading subject area (supporting Crosnoe 2001), but not for math. Future

research might look at an interaction effect with parental involvement and homework time,

gender, race, etc. For example, are parents who are more involved at school also more involved









Hypotheses


Based on the literature review, the following hypotheses were formulated for this project:

* Hla: Gender will have a significant association with academic achievement scores of both
reading and math subjects.

* Hlb: Males are expected to be more likely to achieve higher achievement scores in math,
compared to females.

* H1c: Females are expected to be more likely to achieve higher achievement scores in
reading, compared to males.

* H2: Parental education is expected to be positively associated with academic achievement
scores in both reading and math.

* H3: Parental involvement is expected to be positively associated with academic
achievement scores in reading and math.

* H4: The classroom gender distribution will have a significant association with academic
achievement scores of both subjects.

* H4a: Equal gender distributions in reading classrooms are expected to be positively
associated with reading scores for males.

* H4b: Male enrollment in math classrooms is expected to be positively associated with
math scores for females.

* H5: Classroom sizes are expected to be negatively associated with both reading and math
academic achievement scores.









averages. Females had higher average math GPAs than males from middle school through high

school.

Older studies have found early female advantages in math proficiency, but the advantages

dissipate as the child progresses through school (Parsons, Adler, and Meece 1984). However,

more recent work with the ECLS-K has found that boys and girls begin kindergarten with similar

math scores, but boys' math advantage appears in first grade, and increases over the course of

schooling (LoGergo, Nichols, and Chaplin 2006). At the same time, girls have higher reading

scores upon beginning kindergarten, and their achievement growth is higher than that of their

cohort males. With respect to both subjects, gender differences have been found to be less

pronounced than those in race (LoGerfo, Nichols, and Chaplin 2006).

These trends and inconsistencies continue to demonstrate a need for research into this area

of education, in order to continue to bridge the gap between male and female students and

understand if, and why these gender gaps persist. With this, knowledge, school, classroom and

family policies may be amended to aid in the improvement of child achievement in math and

reading subjects. Within this area of research, it is important to look at the larger environment of

the child, from the classroom to the home, because all factors can interact to aid or detriment a

child's achievement.

The media and government have reported and pushed for smaller classroom size

enrollments, claiming these smaller classes improve students' performance by more

individualized attention from teachers. Funkhouser (2009) somewhat supported these results, by

studying the after-effects of a California initiative to reduce elementary-level classroom sizes by

providing funding for classrooms of reduced size of 20 or less. While there was a small,

significant effect of class-size reduction on achievement, other changes, such as standardized









achievement through praise and rewards. While this paper is not a direct test of social learning

theory, it does aid in shaping the second and third hypotheses, which explore the connections

between familial achievement and goals and children's mastery levels.

Numerous studies have looked at the effects of parents' education in relation to children's

achievements. Both mothers' and fathers' education levels have been positively associated with

children's math and reading performance in school (Vilenius-Tuohimaa, et al. 2008; Hayeman

and Wolfe 1995). Along the same lines, parental educational expectations (Davis-Kean 2005;

Halle, Kurtz-Costes, and Mahoney 1997) and behaviors (Guo and Harris 2000) have been found

to have an indirect link to child outcomes. In this sense, parental expectations for high

achievements or education may have an effect on other factors, such as the amount of time

parents spend with their children reading, doing homework, or volunteering for school.

Parental involvement, at all levels of achievement, grades and ages, has been associated

with benefits to students (Crosnoe 2001). Cooper and Crosnoe (2007) found that in economically

disadvantaged families, parental involvement and children's academic achievement were

positively associated. However, for nondisadvantaged families, parental involvement and

children's achievement had a negative association. Cooper and Crosnoe (2007) reasoned that

parents of children who are not performing well academically may be required to spend more

time involved in school, attending teacher conferences, in order to increase their child's

academic achievement and orientation. At the same time, if the child begins to do well

academically, the parent may become less involved. Muller (1998) was one of the first to look at

gender differences in academic achievement relating to parental involvement. She found that

parental involvement may aid in decreasing gender differences, because when it is controlled for,

the gap in test scores between boys and girls increases.









p<.001) and fathers (y100=.45, p<.001), and higher household incomes (yl60=.45, p<.001) were

also associated with higher average expected math scores.

In model 5, the average classroom reading achievement was 52.18, while the average

classroom math achievement was 50.80, controlling for gender, race and family characteristics.

Additionally, 45.6 percent of the explainable variation in between-classroom average reading

achievement scores, and 60.6 percent of variation in between-classroom average math

achievement scores can be explained by gender, race and family characteristics. At the same

time, looking at within-classroom variation, 11.5 percent of the variation in reading achievement,

and 3.8 percent of the variation in math achievement could be explained by the independent

variables. This increase from the previous model indicates that differences in family

characteristics account for some of the observed within-class variation in scores. For both

subjects, the parental expectations of educational achievement had the strongest effect on

achievement scores.

For reading, male respondents (y10=-1.00, p<.01) were found to have lower average

expected reading scores than females, by 1 point. Black (y20=-3.91, p<.001), Hispanic (y30=-

2.00, p<.001), Asian (y40=-2.15, p<.001) and other race (y50=-1.84, p<.01) respondents were

found to have lower average expected reading scores than White respondents. Here, the addition

of familial involvement characteristics increased the differences between students of white and

other races, seen by the change in significance from model 4 to model 5.

Higher education for mothers (y70=.70, p<.001) and fathers (y100=.42, p<.001), older

fathers' age (y90=. 12, p<.01), and higher household incomes (y160=.42, p<.001) were again

significantly associated with higher average expected reading scores. However, children from

divorced and married families were not significantly different from one another in this model as









they were in model 4. The addition of familial involvement characteristics to the model reduced

the differences between these two groups of children, indicating that when some forms of

parental involvement are controlled for, children from divorced and married families are no

different in their achievement scores.

At the same time, household size (yl70=-.66, p<.001) was negatively associated with

reading scores, where for each person added to the household, the expected average reading

score decreased by .66 points. Additionally, parents who had higher involvement levels

(y190=.21, p<.05) and expected higher educations (y200=1.80, p< .001) for their children were

significantly associated with higher average expected reading scores.

For math, male respondents (y10=1.77, p<.001) were found to have higher average

expected math scores than females, this time by 1.77 points. Black (y20=-3.91, p<.001) and

Hispanic (y30=-1.59, p<.01) respondents were found to have lower average expected math

scores than White respondents. In looking at the change in significance from model 4 to this

model, the addition of familial involvement characteristics seems to increase the differences in

achievement between white and Hispanic students.

Asian (y40=. 14, p>.05) and other race (y50=-1.30, p>.05) respondents were not

significantly different from White respondents in their average expected math scores. Higher

education for mothers (y70=.67, p<.001) and fathers (y100=.35, p<.001), older father's age

(y90=.06, p<.05), higher household incomes (y160=.39, p<.001), and parents who expected

higher educations (y200=1.59, p<.001) were also significantly associated with higher average

expected math scores. In this model, father's age became significantly associated with

achievement scores when controlling for the three familial involvement characteristics.









At the same time, the selection of IRT scores for the dependent variable may have biased

classroom-level effects on students' achievement. The fifth-grade IRT scores were calculated

using the fifth-grade proficiency exams, as well as those from previous waves. These

calculations assume consistency among the students in their answering patterns i.e., answering

more questions right in one wave but not on the next might indicate (under IRT parameters) that

the child had 'guessed' at previous answers and adjust the students' scores accordingly. In this

way, the IRT calculations may average improvements in scores, and therefore reduce the effects

of teacher- or classroom-level characteristics.

Additionally, teacher and child measures of peer achievement would aid in examining the

effect of peer groups on a child's achievement (Van de gaer, et al, 2006) and what their

achievements are, relative to their peers. It would also be useful to include the teachers' gender

and race in the analysis, to see how these variables affect children's scores, perhaps depending

on the gender and race of the child. Although these two variables are asked on the ECLS-K, they

are excluded from the public-use sample and were not available for this research. Furthermore,

exploring different pedagogies or teaching styles might be useful in an attempt to understand the

differences between the two subjects how students learn them and what thought processes are

best suited for each subject area.

These results indicated that very little within-classroom variation could be explained by the

included variables. This indicates that classrooms are somewhat homogenous with respect to

student and family characteristics, and school-level effects may provide a more complete picture

of student achievement. Future studies should include school characteristics to examine student

and classroom differences between and within schools. Three-level hierarchical linear models

could be run to determine the effects of not only students clustered within schools, but









6. Parent acted as a school volunteer
7. Parent participated in fundraising efforts for school

Another variable in this section was the level of schooling the parent expects the child to

achieve, an ordinal measure, with less than high school diploma the lowest (coded 1), followed

by graduating from high school (coded 2), attending two or more years from college (coded 3),

finishing a four- or five-year degree (coded 4), earning a master's degree or equivalent (coded 5),

and finishing a PhD, MD or other advanced degree as the highest (coded 6). Additionally, the

amount of time set aside each day for the child to do homework (measured in minutes) was also

included.

Classroom variables consisted of characteristics of the actual classroom, and a

characteristic of the teachers, both of which could vary, depending on if the child had a different

teacher for reading than for math. The classroom characteristics consisted of: the total number of

students and the gender composition (broken into percentage of boys in the classroom). One

variable, the number of years teaching, measured teacher characteristics.

The dependent variables were standardized scores of math and reading achievement,

measured with separate models for each subject. Students were given short proficiency exams,

and their answers were analyzed using Item Response Theory. IRT calculates the probabilities of

correct answers by examining the pattern of right, wrong and omitted questions to establish a

score for each child, estimating the probability of the child answering all questions correctly (had

the child taken the entire 185-question exam). The fifth-grade scores reflected a combination of

the proficiency exam taken during the fifth-grade wave, as well as those in earlier waves, using

the answer patterns in all waves to estimate scores. These IRT scores were then standardized

(ranging from 0 to 96) to have a mean of 50 (50.3 after weighing due to attrition) and a standard

deviation of 10.









testing, were implemented at the same time as classroom-size reduction, so it was difficult to

determine the individual effect of class size on achievement (Funkhouser 2009). At the same

time, Nye, Hedges, and Konstantopoulos (2004) also found small classes to have a positive

relationship with student achievement, resulting in long-lasting positive academic achievement.

Again, because the relationship between classroom size and achievement was so small, results

suggest that there are other factors that play a larger role in variations in student achievements.

So with respect to classroom size, though there are literature supporting smaller classrooms and

their benefits, it may be difficult to separate these benefits from other classroom or teacher

characteristics.

Along with this, peers may also significantly affect a child's achievement. The

composition of students in a classroom, be it by gender or race, can significantly affect resulting

achievement through conformity to the hegemony (Dryler 1999). In other words, those in the

minority may adapt to characteristics of the majority, resulting in higher or lower achievement in

one subject (Dryler 1999; Hanushek, Kain, Markman, and Rivkin 2003; Ireson, Hallam, Hack,

Clark, and Plewis 2002). For example, if there is a majority of boys in a classroom who like and

excel in math, girls (minority) in the classroom might come to like math as well, and as a result,

the entire class might achieve higher in math overall. On the other hand, if the majority of the

class does not excel or make an effort in a subject area, the minority who does excel may

decrease their efforts, resulting in lower scores overall. In both situations, it all depends on what

the reference group is, and their level of commitment or enjoyment in the subject.

At the same time, distractions and social relationships within the classroom can impact

one's academic achievement. There are arguments for single-sex education for girls, citing gaps

in teachers' attentions, which tend to favor more time spent on boys and their behaviors (Sadker









Analysis

Since the data come from a nested data set, hierarchical linear modeling (Raudenbush and

Bryk 2002) was used to study the intersection of students within classrooms on reading and math

proficiencies and also account for similarities among the data. For example, in a classroom, each

student may have different familial backgrounds (mother's education, father's education,

income, etc.), but all have the same classroom characteristics (gender of teacher, number of

students in the classroom). This nesting needs to be taken into account when analyzing the data,

for the observations are not independent of one another violating one of the assumptions of

OLS regression.

The conceptual model (Figure 2-1) illustrates how the analyses were run. First, an

unconditional model (one without predictors) was run for each dependent variable to serve as a

baseline by which subsequent models could be evaluated (Painter 2002). Next, a model was run

with level-2 predictors to assess how classroom characteristics alone affected academic

achievement. Next, level-1 characteristics of the child and family were run, followed by level-2

characteristics of the classroom and teacher to determine how classroom characteristics influence

children's mastery levels of both subjects. The effects of all level-1 independent variables were

treated as being independent of level-2 variables (fixed effects in HLM terminology). Here, the

coefficient for the slope of each level-1 variable was an average of the effect of this variable

across all classrooms (Roberts 2004).

Equations

The following are HLM equations for each level of analysis. The first equation (2-1)

represents the unconditional multilevel models, whereby the dependent variable (of both reading

and math) associated with the ith student in the jth classroom (DVij ) is a linear combination of

the overall classroom-level mean (y00), a series of random deviations from that mean (u0j ,









There was a fairly even distribution of gender, with males comprising 51.9 percent of the

weighted sample, and a majority of respondents (59 percent) were white. Table 2-1 contains

information about the demographics and corresponding percentages that characterize the sample.

Measures

Demographic variables included gender, recorded as a dummy variable Male, and race,

recorded into a number of dummy variables (using White as the reference group in analysis).

Familial or home variables are measured in two successive models. The first includes basic

demographic characteristics about the household, including: mother's age, highest educational

attainment (an ordinal variable, coded: 1 less than 8th grade, 2 9th-12th grade, 3 high

school diploma or equivalent, 4 vocational/technical program, 5 some college, 6 bachelor's

degree, 7 graduate/professional school without degree, 8 master's degree, 9 doctorate or

professional degree) and employment status recordedd to a dummy variable for employed and

not); father's age, highest educational attainment (coded same as mother's education) and

employment status recordedd to a dummy variable for employed and not); parents' marital status

(single, married (excluded in analyses), widowed, divorced, separated). Household income and

the total number of persons in the household were also used.

Other household variables included information about parental involvement in school and

expected achievements of children. An index of parental involvement (ranging from 0 to 7,

where higher scores indicate a higher level of parental involvement) included the following

items:

Parental Involvement:

1. Parent contacted the school
2. Parent attended open house
3. Parent attended a PTA meeting
4. Parent attended a parent-teacher conference
5. Parent attended a school event









classrooms as well. Previous work has focused on this variation of achievement within and

between schools, and this added level of analysis would be useful in future studies with this, past

and future waves of the ECLS-K.

Finally, once more of these variables and levels of analysis have been examined, and we

have a more holistic understanding of what is affecting achievement, future research would want

to model change across waves. Results of this study indicated that there are significant

differences in the mastery achievements of male and female fifth-grade students, with gender,

race, income and parental education all having significant impacts on mastery levels. Modeling

change (Eccles, Adler & Meece, 1984; LoGergo, Nichols & Chaplin, 2006) could indicate if

these gaps are widening or getting smaller as children age and progress through the educational

system, and allow policy makers to better target programs of at-risk youth.









+ yl50*(WIDOWEDij) + yl60*(GM of INCOMEij) + yl70*(GM of NUMinHHij) +
yl80*(GM of HWTIMEij) + yl90*(PARENTINVOLVEij) + y200*(GM of DEGEXPECTij) +
u0j + rij (2-5)

In model 6, both the level-1 and level-2 variables were included together (Equation 2-6):

DVij = yOO + yl0*MALEij + y20*BLACKij + y30*HISPANICij + y40*ASIANij +
y50*OTHERij + y60*(GM of MOMAGEij) + y70*(GM of MOMEDUCij) +
y80*(MOMEMPLOYij) + y90*(GM of DADAGEij) + yl00*(GM of DADEDUCij) +
yl 10*(DADEMPLOYij) + yl20*(SINGLEij) + yl30*(SEPARATEDij) + yl40*(DIVORCEDij)
+ yl50*(WIDOWEDij) + yl60*(GM of INCOMEij) + yl70*(GM of NUMinHHij) +
yl80*(GM of HWTIMEij) + y190*(PARENTINVOLVEij) + y200*(GM of DEGEXPECTij) +
y01*(GM of YRSTEACHij) + y02*(GM of NUMBERKIDSij)+ y03*(GM of
PERCENTBOYSij)+ u0j + rij (2-6)

Table 2-1. Participant characteristics, weighted (N=10,874)
Demographic Valid percentage
Sex
Female 48.1
Male 51.9
Race
White 59.0
Black 15.1
Hispanic 18.7
Asian 2.6
Other 4.6
Family Income
$10,000 or less 6.8
$10,001-15,000 5.5
$15,001-20,000 6.4
$20,001-25,000 7.6
$25,001-30,000 6.8
$30,001-35,000 6.4
$35,001-40,000 7.3
$40,001-50,000 9.1
$50,001-75,000 16.9
$75,001-100,000 12.7
$100,001-200,000 11.0
$200,001 or more 3.5









For classrooms with average number of students, average percentage of boys and the

teacher with an average number of years teaching, the estimated overall average reading

achievement was 49.589 and the average math achievement was 49.681. For reading, the more

years teaching (y01=.075, p<.01) and more students in the classroom (y02=.319, p<.001) were

significantly associated with higher average reading achievement scores. For math, higher

classroom enrollments (y02=.271, p<.001) was significantly associated with higher average math

achievement scores. For both subjects, the number of students in the classroom had the strongest

effects on achievement scores.

Model 3 shows how demographic characteristics of gender and race affect academic

achievement. These predictors aid in reducing the between-classroom variance (from the fully

unconditional model) by 15 percent for reading and 21 percent for math. Also, the predictors

reduce the within-classroom variance by only .55 percent for reading and .40 percent for math,

indicating that very little variance within-classrooms can be explained by gender and race. The

average classroom reading achievement was 52.41, while the average classroom math

achievement was 50.60, controlling for gender and race.

In reading, male respondents (yl0=-1.17, p<.01) were found to have lower average

expected reading scores than females, by 1.17 points. Black (y20=-5.72, p<.001), Hispanic

(y30=-4.33, p<.001), Asian (y40=-2.67, p<.001) and respondents of other races (y50=-3.26,

p<.001) were found to have lower average expected reading scores than White respondents.

In math, male respondents (y10=1.59, p<.001) were found to have higher average expected

scores than females, by 1.59 points. Black (y20=-5.72, p<.001) and Hispanic (y30=-4.33,

p<.001) respondents were found to have lower average expected math scores than White









classrooms or family situations are equal, and looking at achievement across these categories

provided insight as to the status of our educational system, as well as highlighted areas of

concern for families and educators.

In all: how do classroom and family characteristics affect a child's math and reading

achievement scores? What are the characteristics of high-achieving students and their

environments? Are there any variations between the subjects regarding achievement and gender?

Background and Significance

In school, children learn the skills needed to make their way in the world: how to read,

write effectively, perform mathematical functions and think critically and logically. Schools,

classrooms and teachers vary depending on region, median income of a city or town, and a

variety of other factors. However, within even the same classrooms, there are variations in

achievement scores. Is this due to a lack of ability on the child's part, learning differences, life

circumstances, or a combination of factors?

Since 1980, gender differences in math scores have been researched and publicized

(Benbow and Stanley 1980; Hyde, Fennema and Lamon 1990), along with concerns of their

future implications. According to the National Center for Education Statistics, although math and

reading scores have steadily been increasing, there are still gender gaps in achievement in these

subject areas. As late as 2007, males scored higher than females on all math content, with the

exception of geometry (Lee, Grigg, and Dion 2007). Additionally, females continue to outscore

males in reading, both for literary experience and for information (Lee, Grigg, and Donahue

2007; Vilenius-Tuohimaa, Aunola, and Nurmi 2008).

However, not all studies have found support for the gendered achievement gaps in

mathematics. Ding, Song and Richardson (2006) found no significant difference in standardized

test scores between boys and girls, but did find there to be significant differences in grade point









Additionally, parents who had higher involvement levels (yl90=.20, p<.05) and expected higher

educations (y200=1.72, p< .001) for their children, and classrooms with more students (y02=.23,

p<.001) were significantly associated with higher average reading achievement scores. The

effect of teacher experience observed in model 2 was eradicated by the addition of demographic

and family demographic and involvement characteristics included in model 6.

For math, male respondents (yl0=1.81, p<.001) were found to have higher average

expected math scores than females, by 1.81 points. Black (y20=-3.94, p<.001) and Hispanic

(y30=-1.92, p<.01) respondents were found to have lower average expected math scores than

White respondents. Asian (y40=-.17, p>.05) and other race (y50=-1.36, p>.05) respondents were

not significantly different from White respondents in their average expected math scores.

Higher education for mothers (y70=.65, p<.001) and fathers (y100=.36, p<.001), older

fathers' age (y90=.07, p<.05), higher household incomes (y160=.37, p<.001), and parents who

expected higher educations (y200=1.53, p<.001), and classrooms with higher student

enrollments (y02=.21, p<.001) were also associated with higher average expected math scores.









average expected reading scores than White respondents. Respondents of other races (y50=-1.19,

p>.05) were not significantly different from White respondents in their average expected reading

scores, changing from the previous model. Here, the addition of family demographics reduced

the differences between white children and those of other races.

Higher education for mothers (y70=.92, p<.001) and fathers (y100=.53, p<.001), older

fathers' age (y90=.12, p<.01), and higher household incomes (y160=.50, p<.001) were

significantly associated with higher average expected reading scores. At the same time,

household size (yl70=-.70, p<.001) was negatively, though significantly, associated with reading

scores, where for each person added to the household, the expected average reading score

decreased by .70 points. Only children whose parents were divorced (y140=-1.12, p<.05) were

found to have lower average expected reading scores than children whose parents were married.

Children whose parents were single (y120=-.37, p>.05), separated (y130=-.04, p>.05) or

widowed (yl50=-. 16, p>.05) were not significantly different than children whose parents were

married in their average expected reading scores.

For math, male respondents (y10=1.51, p<.001) were found to have higher average

expected math scores than females, by 1.51 points. Black (y20=-3.39, p<.001) respondents were

found to have lower average expected math scores than White respondents, by 3.39 points.

Hispanic (y30=-.90, p>.05), Asian (y40=1.07, p>.05) and other race (y50=-.99, p>.05)

respondents were not significantly different from White respondents in their average expected

math scores.

In this model, the difference between Hispanic and White children's achievements was

eradicated, indicating that when family demographics are equal, White and Hispanic children are

no different in their average achievement scores. Higher education for mothers (y70=.85,









Furthermore, besides the second model (where the level-2 predictors explained similar

percentages), the proportion of explainable variance in between-classroom achievements was

smaller for reading than for math subjects. This suggests that predictors affect subject

achievements differently, and there are other, additional characteristics besides child and family

demographics and classroom predictors that are affecting the differences in between-classroom

reading achievement scores. Along this line, programs aimed at reading reform or improvement

should be targeted accordingly and treated differently than ones for math.

In all, results highlighted the family characteristics of high achievement in both subject

areas. In reading, being female, White, high mothers' and fathers' education, older fathers' ages,

high income, low total household number, high parental involvement and high parental

educational expectations were positively associated with high average reading achievement

scores. In math, being male, White or Asian, high mothers' and fathers' education, high income

and high educational expectations were positively associated with high average math

achievement scores.

Limitations and Suggestions for Future Research

Although there are a number of interesting variables in the ECLS-K, many of them are not

applicable for the overall scope of this research. One definite limitation of the selected variables

is that they may not provide a complete picture of the child's overall achievement in math and

reading. For example, there are no measures of child achievement, measured by the parentss.

This assessment could be useful, for it may influence a variety of home factors, like the amount

of time set aside for homework each day. Parents of children who are not yet proficient in a

particular skill or knowledge area may impose longer mandatory homework time in order to aid

the student in skill mastery.









CHAPTER 3
RESULTS

Results of HLM reading analyses are included in Table 3-1 and math analyses are in Table

3-2; both are broken down by model. Changes in classroom variance from model to model

decreased, indicating that the addition of each set of predictors helped to explain more and more

of the explainable variation between classrooms.

Model 1 depicts the fully unconditional models for both reading and math subjects. These

models provide a benchmark by which future models can be compared, describing how much

classrooms vary in their subject achievements. The estimated overall classroom average on

reading scores was 49.56, and the average on math scores was 49.61; both were below the

national average of 50.3 for both subjects. Intraclass coefficients, the observed variations in

achievement scores attributable to classroom-level characteristics, were significant, with 65

percent for reading and 47 percent for math. This indicated that classroom-level characteristics

account for more between-classroom differences in reading than in math classes.

Next, model 2 gives information about how classroom characteristics alone affect

academic achievement. Comparing the intercepts, or classroom, variance between model 1 and

model 2 indicated that the level-2 predictors alone explained some of the between-classroom

variance observed in the fully unconditional model (1).

About 7 percent of the between-classroom variance in reading achievement is accounted

for by the three classroom characteristics, and 6.2 percent of the explainable variation in

classroom average math achievement scores could be explained by the combination of classroom

size, teacher experience and gender composition. Finally, because there was more than one

student per classroom, there are fewer classrooms than there are students, which accounts for the

decrease in sample size between models 1 and 2.











Level 1 Level 2


Figure 2-1. Conceptual Model

Table 2-2. Means and standard deviations of centered variables
Variables Mean S.D.
Household Characteristics
Mother's Age 39.77 6.65
Mother's Education 4.60 1.82
Father's Age 42.34 7.27
Father's Education 4.46 2.11
Income 8.38 3.19
Number in Household 4.59 1.39
Time for Homework 37.53 41.17
Parental Involvement 4.80 1.61
Expected Education 4.03 1.05
Reading Classroom
Years Teaching 13.83 9.98
Number of Students 22.06 6.47
Percent Male 50.67 9.24
Math Classroom
Years Teaching 14.06 10.07
Number of Students 22.55 6.27
Percent Male 50.97 9.24


Outcome









CHAPTER 2
RESEARCH DESIGN AND METHODS

Procedure

The data were taken from the public-use Early Childhood Longitudinal Study:

Kindergarten Class of 1998-1999, Fifth Grade Wave. The survey collects information from a

number of sources: from children, their families, teachers and school administrators on an

assortment of topics using a variety of methods, including: field visits, face-to-face interviews

and assessments, and mail-in and computer-assisted phone surveys. Employing multistage

probability sampling procedures, the base sample (kindergarten) was a nationally representative

sample of more than 21,000 children, enrolled in about 1,000 different kindergarten programs

during the 1998-1999 school year. Since the fifth grade wave did not add students who did not

have the opportunity to be sampled in kindergarten, and does not include students who were lost

(due to lack of parental participation, death, or the child moving out of the country), the data are

not representative of all fifth graders. Instead, this sample is representative of the cohort of

kindergartners enrolled in schools in the fall of 1998. Despite this, the data are still useful, for it

is one of the largest, comprehensive measures of education in the United States, and there are a

wide range of measures and topics of interest included.

Sample

In the public-use sample, only children with complete data files for all years of the ECLS-

K, including kindergarten and/or first grade and third grade, were eligible for the fifth-grade

wave, reducing the sample size to 11,820 potential respondents. Of those, 946 respondents did

not have complete data files (due to refusals, movers, etc.) for fifth grade, further reducing the

final sample size to 10,874 children in 4,265 reading classrooms and 2,730 math classrooms.










Table 3-2. Continued
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6


Classroom Characteristics
Years Teaching 0.048 / 0.052
Number of Students 0.271 / 0.181 "
Percent Male 0.045 / 0.044
Intercept 49.607 49.681 50.602
Classroom Variance 45.989 43.135 36.297
Change in Classroom Variance 2.854 9.692
Class. Proportion of Variance 0.062 0.211
Student Variance 51.516 51.560 51.310
Student Proportion of Variance -0.001 0.004
X2 9193.533' 9437.227 8012.716
N 10874 2730 10874
Source: Early Childhood Longitudinal Survey, Kindergarten Class of 1998-1999, Fifth Grade Wave
Unstandardized / Standardized Coefficients
* p <.05; p .01; p<.001


50.565 "
20.352
25.637
0.557
50.725
0.015
6228.384 "
10874


50.797
18.138
27.851
0.606
49.565
0.038
6031.640 "
10874


0.004 / 0.004
0.206 / 0.137 "
0.019 / 0.019
50.875 "
17.372
28.617
0.622
49.307
0.043
6214.000 "
2730









respondents. Asian (y40=.50, p>.05) and respondents of other races (y50=-2.31, p>.05) were not

significantly different from White respondents in their average expected mathematics scores.

In model 4, the average classroom reading achievement was 51.99 for white, female

students, in households with average income, average number of people, whose parents were

married, did not work, and had average educational attainments and ages. For math, the average

classroom math achievement was 50.57 for white, male students, in households with average

income, average number of people, whose parents were married, did not work, and had average

educational attainments and ages. For both subjects, mother's education had the strongest impact

on achievement scores.

In this model, about 41.5 percent of the explainable variation in between-classroom

average reading achievement scores, and 55.8 percent of variation in between-classroom average

math achievement scores could be explained by gender, race and family demographic

characteristics. At the same time, only 7.2 percent of the explainable variation within-classroom

average reading achievement scores, and 1.5 percent of variation in average math scores could be

explained by the predictors.

Additionally, the addition of family demographics reduced the between-classroom variance

by the largest amount. This indicates that family demographic characteristics account for larger

amounts of the observed between-classroom variance than demographics alone, or the

combination of child and family demographics and familial involvement, or the combination of

child and family demographics, familial involvement and classroom characteristics.

For reading scores in model 4, male respondents (y10=-1.33, p<.001) were found to have

lower average expected reading scores than females, by 1.33 points. Black (y20=-3.28, p<.001),

Hispanic (y30=-1.26, p<.05), Asian (y40=-1.41, p<.05) respondents were found to have lower









BIOGRAPHICAL SKETCH

Kristi Lynn Donaldson was born in New Jersey but spent the majority of her childhood in

Clearwater, Florida. She graduated high school in 2001 from Palm Harbor University High

School, earning an International Baccalaureate diploma. She then attended the University of

Florida for her undergraduate work, majoring in public relations, and later, sociology. After

earning two bachelor's degrees and graduating with honors in both subjects, Kristi took a year

off from school to work and travel. In 2006, she returned to UF to begin graduate studies, and

will graduate with her M.A. in sociology in August 2009. Following graduation, Kristi will move

to Indiana to continue on to her Ph.D. studies at the University of Notre Dame, where she plans

to research educational inequalities within the Center for Research on Educational Opportunity.




Full Text

PAGE 1

EXPLORING READING AND MATH MASTERIE S: INFLUENCES OF CLASSROOM AND FAMILY By KRISTI LYNN DONALDSON A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS UNIVERSITY OF FLORIDA 2009 1

PAGE 2

2009 Kristi Lynn Donaldson 2

PAGE 3

To all of the people who have put up with my seclusion while working on this projectyou know who you are 3

PAGE 4

ACKNOWLEDGMENTS First, I thank my chair, Connie Shehan, for her support, both academically and personally, throughout the last three years. I also thank Monika Ardelt for her attention to detail and insistence on doing things right, both of which have helped improve this paper. I am grateful for Graham, who helped ease tensions by making me laugh and providing breaks away from my work, and Becky and Jackie for bein g my cheerleaders now and always. I thank my Dad for supporting me even though he does not always understand my goals or intentions, my sister who tells me I can do anything, and my mom for giving me space when needed and encouraging me to push forward. Fi nally, I thank Brewer fo r providing relief from stress and always making me smile. I also apprec iate all others my roommates, colleagues, coworkers who have supported me along the way. 4

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS ...............................................................................................................4LIST OF TABLES ...........................................................................................................................6LIST OF FIGURES .........................................................................................................................7ABSTRACT .....................................................................................................................................8 CHAPTER 1 INTRODUCTION ................................................................................................................ ....9Specific Aims ............................................................................................................................9Background and Significance .................................................................................................10Hypotheses ..............................................................................................................................152 RESEARCH DESIGN AND METHODS ..............................................................................16Procedure ................................................................................................................................16Sample ....................................................................................................................................16Measures .................................................................................................................................17Analysis ..................................................................................................................................19Equations ................................................................................................................................193 RESULTS ..................................................................................................................... ..........244 CONCLUSION .................................................................................................................. .....36Findings ..................................................................................................................................36Limitations and Suggesti ons for Future Research ..................................................................39REFERENCES ..............................................................................................................................42BIOGRAPHICAL SKETCH .........................................................................................................45 5

PAGE 6

LIST OF TABLES Table page 2-1. Participant character istics, weighted (N=10,874) .................................................................222-2. Means and standard devi ations of centered variables ...........................................................233-1. Hierarchical linear modeli ng results of coefficients of between-classroom models of fifth-grade reading achievement ........................................................................................323-2. Hierarchical linear modeli ng results of coefficients of between-classroom models of fifth-grade math achievement ............................................................................................34 6

PAGE 7

LIST OF FIGURES Figure page 2-1. Conceptual Model ..................................................................................................................23 7

PAGE 8

8 Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Arts EXPLORING READING AND MATH MASTERIE S: INFLUENCES OF CLASSROOM AND FAMILY By Kristi Lynn Donaldson August 2009 Chair: Connie Shehan Major: Sociology Using data from the public-use Early Childhood Longitudinal Study: Kind ergarten Class of 1998-1999, Fifth Grade wave, this study looked at the influences of demographic, classroom and family characteristics on childrens elementary-l evel mastery skills of reading and math. To examine these relationships, five hypotheses were tested using hierarchical linear modeling. Female students were found to be more likely to have higher average read ing achievements than males, where the opposite occurred in mathema tics. For both subjects, White respondents had higher average achievement scores than Black or Hispanic children. Parental education, household income and expected educational achie vements were all positively associated with both reading and math achievement scores. Classr oom size, but not teacher experience or gender composition of classroom, had a significant, positive effect with both math and reading achievement scores. Findings highlight the need for continued emphasis on education by parents, educators and the media. Limitations and sugge stions for future research are discussed.

PAGE 9

CHAPTER 1 INTRODUCTION Specific Aims According to the National Center for Educati on Statistics in 2007, 39 percent of fourthgrade students were proficient in their math level, and 33 percent were proficient in the reading level. Of all math scores, Black males and females, Hispanic males and females, as well as white females scored lower than white males. Of all reading scores, Black males and females, Hispanic males and females, as well as white males scored lower than white females. What are some of the causes of these observed differences in proficiencies? Questions of achievement gaps between boys and girls have been widely covered in mainstream media over the course of the last tw enty years. Within a similar time period, there has been an introduction of same-sex classrooms in public schools to help bridge the gap in achievement and proficiency between boys and girl s in different subjects. The idea behind these classrooms is that boys and girl s learn differently, and separa ting the genders will benefit students in both achievement and confidence. To explore these potential differences, I used data from the Early Childhood Longitudinal Survey, Kindergarten Class of 1998-1999 Fifth Grade wave. Although data on same-sex classrooms was not available in th is study, there were still a number of measures that could be used to examine the differences in childrens aca demic achievements or proficiencies. I aimed to study characteristics of the child s environment, including the cl assroom and family, to explore differences in childrens reading and math achievement scores. The quality of education children receive ma y be a direct result of the resources and opportunities afforded unto them. Children livin g in higher-income areas may have access to schools and classrooms with newer technologies and greater learni ng resources. Not all 9

PAGE 10

classrooms or family situations are equal, and looking at achievement across these categories provided insight as to the status of our educational system, as well as highlighted areas of concern for families and educators. In all: how do classroom and family characte ristics affect a childs math and reading achievement scores? What are the characteri stics of high-achieving students and their environments? Are there any variations between the subjects regarding achievement and gender? Background and Significance In school, children learn the skills needed to make their way in the world: how to read, write effectively, perform mathem atical functions and think crit ically and logically. Schools, classrooms and teachers vary depending on region, median income of a city or town, and a variety of other factors. Howeve r, within even the same classr ooms, there are variations in achievement scores. Is this due to a lack of ability on the chil ds part, learning differences, life circumstances, or a combination of factors? Since 1980, gender differences in math scores have been research ed and publicized (Benbow and Stanley 1980; Hyde, Fennema and Lamon 1990), along with concerns of their future implications. According to the National Center for Education Statis tics, although math and reading scores have steadily been increasing, there are still gender gaps in achievement in these subject areas. As late as 2007, males scored higher than females on all math content, with the exception of geometry (Lee, Grigg, and Dion 2007). Additionally, females continue to outscore males in reading, both for literary experience and for information (Lee, Grigg, and Donahue 2007; Vilenius-Tuohimaa, Aunola, and Nurmi 2008). However, not all studies have found suppor t for the gendered achievement gaps in mathematics. Ding, Song and Richardson (2006) found no significant difference in standardized test scores between boys and girls, but did find th ere to be significant differences in grade point 10

PAGE 11

averages. Females had higher average math GP As than males from middle school through high school. Older studies have found early female advantages in math proficiency, but the advantages dissipate as the child progresses through school (Parsons, Adler, and Meece 1984). However, more recent work with the ECLS-K has found that boys and girls begin kindergarten with similar math scores, but boys math advantage appears in first grade, and increa ses over the course of schooling (LoGergo, Nichols, and Chaplin 2006). At the same time, girls have higher reading scores upon beginning kindergarten, a nd their achievement growth is higher than that of their cohort males. With respect to both subjects, gender differences have been found to be less pronounced than those in race (LoGe rfo, Nichols, and Chaplin 2006). These trends and inconsistencies continue to demonstrate a need for research into this area of education, in order to con tinue to bridge the gap between male and female students and understand if, and why these gende r gaps persist. With this, knowledge, school, classroom and family policies may be amended to aid in the improvement of child ac hievement in math and reading subjects. Within this area of research, it is important to look at the larger environment of the child, from the classroom to the home, because all factors can interact to aid or detriment a childs achievement. The media and government have reported and pushed for smaller classroom size enrollments, claiming these smaller classe s improve students performance by more individualized attention from teachers. Funkhous er (2009) somewhat supported these results, by studying the after-effects of a Calif ornia initiative to reduce elemen tary-level classroom sizes by providing funding for classrooms of reduced size of 20 or less. While there was a small, significant effect of class-size reduction on achievement, other changes, such as standardized 11

PAGE 12

testing, were implemented at the same time as classroom-size reduction, so it was difficult to determine the individual effect of class si ze on achievement (Funkhouser 2009). At the same time, Nye, Hedges, and Konstantopoulos (2004) also found small classes to have a positive relationship with student achievement, resulti ng in long-lasting positive academic achievement. Again, because the relationship between classroo m size and achievement was so small, results suggest that there are other factor s that play a larger role in va riations in student achievements. So with respect to classroom size, though there are l iterature supporting sma ller classrooms and their benefits, it may be difficult to separate these benefits from other classroom or teacher characteristics. Along with this, peers may also significantly affect a childs achievement. The composition of students in a classroom, be it by ge nder or race, can significantly affect resulting achievement through conformity to the hegemony (Dryler 1999). In other words, those in the minority may adapt to characteristics of the majo rity, resulting in higher or lower achievement in one subject (Dryler 1999; Hanushek, Kain, Ma rkman, and Rivkin 2003; Ireson, Hallam, Hack, Clark, and Plewis 2002). For example, if there is a majority of boys in a classroom who like and excel in math, girls (minority) in the classroom might come to like math as well, and as a result, the entire class might achieve higher in math ove rall. On the other hand, if the majority of the class does not excel or make an effort in a subject area, the minority who does excel may decrease their efforts, resulting in lower scores ov erall. In both situations, it all depends on what the reference group is, and their level of co mmitment or enjoyment in the subject. At the same time, distractions and social re lationships within the classroom can impact ones academic achievement. There are arguments for single-sex education for girls, citing gaps in teachers attentions, which te nd to favor more time spent on boys and their behaviors (Sadker 12

PAGE 13

and Sadker 1995), especially in asking and an swering questions (Hyde, Fennema, and Lamon 1990). When testing the effects of single-sex and co educational Dutch classrooms, Van de gaer, Pustjens, Van Damme, and De Munter (2004) found there to be gendered differences in achievements. Boys reading scores were improve d by coeducational classes, with speculation that the addition of girls aids in socializing boys to be more attentive and less accepting of deviant or disruptive behaviors (Jackson a nd Smith 2000). At the same time, although no differences were found in girls achievements in coeducational a nd same-sex classrooms, Van de gaer, et al. (2004) found there to be improvements in mathematics scores in same-sex schools. Additionally, studies have found that when controlling for household characteristics, such as income, parents educational levels and the li ke, the differences between coeducational and same-sex classrooms and schools are reduced (Harker 2000). Although largely cited in context with deviance, social learning theory can also be used to look at the primary institutions of education and family to ex amine their effects on a childs educational achievement. In this theory, behavi or is learned through th e operant conditioning of groups which compromise or control [an] i ndividuals major sources of reinforcement (Burgess and Akers 1966:140). According to this theory, parents who place a high value on involvement in school, and have high academic e xpectations may pass on th ese values to their children, resulting in higher achie vement. Children observe their parents behaviors, either through the encouragement for hi gher education, stories of th eir educational pursuits and outcomes, or actual observations of their pare nts attending school. Afte r internalizing these behaviors, children may begin to exhibit behavi ors approved and valued by their significant others, conditioned through social reinforcement (Kandel 1980); parents may encourage high 13

PAGE 14

achievement through praise and rewards. While this paper is not a direct te st of social learning theory, it does aid in shaping the second and third hypotheses, which explore the connections between familial achievement and goals and childrens mastery levels. Numerous studies have looked at the effects of parents educ ation in relation to childrens achievements. Both mothers and fathers educati on levels have been positively associated with childrens math and reading performance in sc hool (Vilenius-Tuohimaa, et al. 2008; Hayeman and Wolfe 1995). Along the same lines, parental educational expectati ons (Davis-Kean 2005; Halle, Kurtz-Costes, and Mahoney 1997) and be haviors (Guo and Harris 2000) have been found to have an indirect link to child outcomes. In this sense, parental expectations for high achievements or education may have an effect on other factors, such as the amount of time parents spend with their ch ildren reading, doing homework, or volunteering for school. Parental involvement, at all levels of achie vement, grades and ages, has been associated with benefits to students (Crosnoe 2001). Coope r and Crosnoe (2007) found that in economically disadvantaged families, parental involvemen t and childrens academic achievement were positively associated. However, for nondisadvant aged families, parental involvement and childrens achievement had a negative associa tion. Cooper and Crosnoe (2007) reasoned that parents of children who are not performing well academically may be required to spend more time involved in school, attendi ng teacher conferences, in orde r to increase their childs academic achievement and orientation. At the same time, if the child begins to do well academically, the parent may become less involved. Muller (1998) was one of the first to look at gender differences in academic achievement re lating to parental invol vement. She found that parental involvement may aid in decreasing gender differences, because when it is controlled for, the gap in test scores betw een boys and girls increases. 14

PAGE 15

Hypotheses Based on the literature review, the following hy potheses were formulated for this project: H1a: Gender will have a significant association with academic achievement scores of both reading and math subjects. H1b: Males are expected to be more likely to achieve higher achievement scores in math, compared to females. H1c: Females are expected to be more likely to achieve higher achievement scores in reading, compared to males. H2: Parental education is expected to be pos itively associated with academic achievement scores in both reading and math. H3: Parental involvement is expected to be positively associated with academic achievement scores in reading and math. H4: The classroom gender distribution will have a significant association with academic achievement scores of both subjects. H4a: Equal gender distributions in reading cl assrooms are expected to be positively associated with reading scores for males. H4b: Male enrollment in math classrooms is e xpected to be positively associated with math scores for females. H5: Classroom sizes are expected to be negati vely associated with both reading and math academic achievement scores. 15

PAGE 16

CHAPTER 2 RESEARCH DESIGN AND METHODS Procedure The data were taken from the public-use Early Childhood Longitudinal Study: Kindergarten Class of 1998-1999, Fift h Grade Wave. The survey collects information from a number of sources: from children, their familie s, teachers and school administrators on an assortment of topics using a variety of methods, including: field visits, fa ce-to-face interviews and assessments, and mail-in and computer-assisted phone surveys. Employing multistage probability sampling procedures, the base sample (kindergarten) was a nationally representative sample of more than 21,000 children, enrolled in about 1,000 different kindergarten programs during the 1998-1999 school year. Since the fifth grade wave did not a dd students who did not have the opportunity to be sampled in kindergar ten, and does not include students who were lost (due to lack of parental partic ipation, death, or the child moving out of the country), the data are not representative of all fifth gr aders. Instead, this sample is re presentative of the cohort of kindergartners enrolled in schools in the fall of 1 998. Despite this, the data are still useful, for it is one of the largest, comprehensive measures of education in the United States, and there are a wide range of measures and t opics of interest included. Sample In the public-use sample, only children with co mplete data files for all years of the ECLSK, including kindergarten and/or first grade and third grade, were eligible for the fifth-grade wave, reducing the sample size to 11,820 potential respondents. Of those, 946 respondents did not have complete data files (due to refusals, movers, etc.) for fifth grade, further reducing the final sample size to 10,874 children in 4,265 read ing classrooms and 2,73 0 math classrooms. 16

PAGE 17

There was a fairly even distribution of gende r, with males comprising 51.9 percent of the weighted sample, and a majority of respondents (59 percent) were white Table 2-1 contains information about the demographics and corresponding percentages that characterize the sample. Measures Demographic variables included gender, rec oded as a dummy variable Male, and race, recoded into a number of dummy variables (usi ng White as the reference group in analysis). Familial or home variables are measured in two successive models. The first includes basic demographic characteristics about the household, including: mothers age, highest educational attainment (an ordinal variable, coded: 1 le ss than 8th grade, 2 9th-12th grade, 3 high school diploma or equivalent, 4 vocational/tech nical program, 5 some college, 6 bachelors degree, 7 graduate/professional school without de gree, 8 masters degree, 9 doctorate or professional degree) and employme nt status (recoded to a dummy variable for employed and not); fathers age, highest educational attain ment (coded same as mothers education) and employment status (recoded to a dummy variable for employed and not); parents marital status (single, married (excluded in analyses), widow ed, divorced, separated). Household income and the total number of persons in the household were also used. Other household variables included information about parental involvement in school and expected achievements of children. An index of parental involvement (ranging from 0 to 7, where higher scores indicate a higher level of parental involvement) included the following items: Parental Involvement: 1. Parent contacted the school 2. Parent attended open house 3. Parent attended a PTA meeting 4. Parent attended a parent-teacher conference 5. Parent attended a school event 17

PAGE 18

6. Parent acted as a school volunteer 7. Parent participated in fundraising efforts for school Another variable in this section was the level of schooling the parent expects the child to achieve, an ordinal measure, w ith less than high school diploma the lowest (coded 1), followed by graduating from high school (coded 2), attendi ng two or more years from college (coded 3), finishing a fouror five-year de gree (coded 4), earning a masters degree or equivalent (coded 5), and finishing a PhD, MD or other advanced de gree as the highest (coded 6). Additionally, the amount of time set aside each day for the child to do homework (measured in minutes) was also included. Classroom variables consisted of charact eristics of the actual classroom, and a characteristic of the teachers, both of which c ould vary, depending on if the child had a different teacher for reading than for math. The classroom ch aracteristics consisted of: the total number of students and the gender composition (broken into percentage of boys in the classroom). One variable, the number of years teaching, measured teacher characteristics. The dependent variables were standardized scores of math and reading achievement, measured with separate models for each subject. Students were given sh ort proficiency exams, and their answers were analyzed using Item Res ponse Theory. IRT calculate s the probabilities of correct answers by examining the pattern of ri ght, wrong and omitted questions to establish a score for each child, estimating the probability of the child answering all questions correctly (had the child taken the entire 185-question exam). The fifth-grade scores refl ected a combination of the proficiency exam taken during the fifth-grade wave, as well as those in earlier waves, using the answer patterns in all waves to estimate scor es. These IRT scores were then standardized (ranging from 0 to 96) to have a mean of 50 (50.3 after weighing due to attr ition) and a standard deviation of 10. 18

PAGE 19

Analysis Since the data come from a nested data set, hierarchical linear modeling (Raudenbush and Bryk 2002) was used to study the in tersection of students within cl assrooms on reading and math proficiencies and also account for similarities am ong the data. For example, in a classroom, each student may have different familial backgrounds (mothers education, fathers education, income, etc.), but all have the same classroom characteristics (gender of teacher, number of students in the classroom). This nesting needs to be taken into account when analyzing the data, for the observations are not independent of one another violating one of the assumptions of OLS regression. The conceptual model (Figure 2-1) illustra tes how the analyses were run. First, an unconditional model (one without predictors) was run for each depe ndent variable to serve as a baseline by which subsequent models could be evaluated (Painter 2002). Next, a model was run with level-2 predictors to assess how classr oom characteristics alone affected academic achievement. Next, level-1 characteristics of the child and family were run, followed by level-2 characteristics of the classroom and teacher to determine how classroom ch aracteristics influence childrens mastery levels of both subjects. The effects of all level-1 independent variables were treated as being independent of level-2 variable s (fixed effects in HLM terminology). Here, the coefficient for the slope of each level-1 variable was an average of the effect of this variable across all classrooms (Roberts 2004). Equations The following are HLM equations for each leve l of analysis. The first equation (2-1) represents the unconditional multilevel models, whereby the dependent variable (of both reading and math) associated with the ith student in the jth classroom (DVij ) is a linear combination of the overall classroom-level mean (y00), a series of random deviations from that mean (u0j 19

PAGE 20

normally distributed with mean 0 and between classroom variance) and a random error (rij, normally distributed with mean 0 and within clas sroom variance) associated with the ith student in the jth classroom (Singer 1998). DVij = y00 + u0j + rij (2-1) The next equation (2-2) represents mode l 2, where level-2 va riables (classroom characteristics) were modeled against the de pendent variables to determine how classroom characteristics alone affect academic achieveme nt. Variables with a GM were grand-mean centered for analysis, meaning the explanatory va riables were centered on the overall mean of that variable (Raudenbush & Br yk, 2002). Centering is frequently used in HLM to aid in interpretations of results. In this equation for ex ample, the intercept y00 re presents the expected outcome for a student who is in a classroom w ith a teacher of average experience (average number of years teaching), an average number of students and an average percentage of boys. If the variables were left in their original metr ic, the intercept would represent the expected outcome for a student in a classroom where the te acher has no previous experience (since teacher questionnaires were given toward the end of the school year instead of the beginning, no teacher had 0 experience), without students (which is impo ssible a classroom then would not exist) and without boys (which is not included in this data ). See Table 2-2 for a listing of the means and standard deviations of all centered variables. The following is the model-2 equation with level-2 variables grand-mean centered: DVij = y00 + y01*(GM of YRSTEACHij) + y02*(GM of NUMBERKIDSij)+ y03*(GM of PERCENTBOYSij)+ u0j + rij (2-2) The next three models represent the level-1 equations run for analyses. In model 3, the dependent variables were modeled as a functi on of gender and race (White omitted). Following previous studies, gender and race were first examin ed, as they are frequently cited as potential 20

PAGE 21

(and often main) predictors of academic achieveme nt. By including these variables first, I was able to examine how influence of gender and race might vary de pending on additional predictors included in the models. Dummy variables were not centered in the analyses due to interpretative results; by not centering, I examined the differenc e in average achievement scores between males and females, White and Black students, etc. Equation 2-3 represents the HLM analysis for gender and race, cited frequently as potenti al predictors of academic achievement: DVij = y00 + y10*MALEij + y20*BLACKij + y30*HISPANICij + y40*ASIANij + y50*OTHERij + u0j + rij (2-3) In model 4, the dependent variables are modeled as a function of gender, race, mothers & fathers ages, educations and employment status es, marital status (married omitted), household income, and the number of people living in the household (Equation 24). Basic demographic characteristics of the family were added to this model to see how family characteristics affected achievement scores and also th e effects of race and gender: DVij = y00 + y10*MALEij + y20*BLACKij + y30*HISPANICij + y40*ASIANij + y50*OTHERij + y60*(GM of MOMAGEij) + y70*(GM of MOMEDUCij) + y80*(MOMEMPLOYij) + y90*(GM of DADAGEij) + y100*(GM of DADEDUCij) + y110*(DADEMPLOYij) + y120*(SINGLEij) + y130*( SEPARATEDij) + y140*(DIVORCEDij) + y150*(WIDOWEDij) + y160*(GM of INCOMEij) + y170*(GM of NUMinHHij) + u0j + rij (2-4) In model 5, the dependent variables are modeled as a function of gender, race, mothers & fathers ages, educations and employment status es, marital status (married omitted), household income, the number of people living in the household, the time spent on homework, parental involvement and the degree expected of the child (Equation 2-5). He re, three variables representing non-demographic fa mily variables were added: DVij = y00 + y10*MALEij + y20*BLACKij + y30*HISPANICij + y40*ASIANij + y50*OTHERij + y60*(GM of MOMAGEij) + y70*(GM of MOMEDUCij) + y80*(MOMEMPLOYij) + y90*(GM of DADAGEij) + y100*(GM of DADEDUCij) + y110*(DADEMPLOYij) + y120*(SINGLEij) + y130*( SEPARATEDij) + y140*(DIVORCEDij) 21

PAGE 22

+ y150*(WIDOWEDij) + y160*(GM of INCOME ij) + y170*(GM of NUMinHHij) + y180*(GM of HWTIMEij) + y190*(PARENTINVOL VEij) + y200*(GM of DEGEXPECTij) + u0j + rij (2-5) In model 6, both the level-1 and level-2 variab les were included together (Equation 2-6): DVij = y00 + y10*MALEij + y20*BLACKij + y30*HISPANICij + y40*ASIANij + y50*OTHERij + y60*(GM of MOMAGEij) + y70*(GM of MOMEDUCij) + y80*(MOMEMPLOYij) + y90*(GM of DADAGEij) + y100*(GM of DADEDUCij) + y110*(DADEMPLOYij) + y120*(SINGLEij) + y130*( SEPARATEDij) + y140*(DIVORCEDij) + y150*(WIDOWEDij) + y160*(GM of INCOME ij) + y170*(GM of NUMinHHij) + y180*(GM of HWTIMEij) + y190*( PARENTINVOL VEij) + y200*(GM of DEGEXPECTij) + y01*(GM of YRSTEACHij) + y02*(GM of NUMBERKIDSij)+ y03*(GM of PERCENTBOYSij)+ u0j + rij (2-6) Table 2-1. Participant charac teristics, weighted (N=10,874) Demographic Valid percentage Sex Female Male Race White Black Hispanic Asian Other Family Income $10,000 or less $10,001-15,000 $15,001-20,000 $20,001-25,000 $25,001-30,000 $30,001-35,000 $35,001-40,000 $40,001-50,000 $50,001-75,000 $75,001-100,000 $100,001-200,000 $200,001 or more 48.1 51.9 59.0 15.1 18.7 2.6 4.6 6.8 5.5 6.4 7.6 6.8 6.4 7.3 9.1 16.9 12.7 11.0 3.5 22

PAGE 23

Figure 2-1. Conceptual Model Table 2-2. Means and standard de viations of centered variables Variables Mean S.D. Household Characteristics Mothers Age 39.77 6.65 Mothers Education 4.60 1.82 Fathers Age 42.34 7.27 Fathers Education 4.46 2.11 Income 8.38 3.19 Number in Household 4.59 1.39 Time for Homework 37.53 41.17 Parental Involvement 4.80 1.61 Expected Education 4.03 1.05 Reading Classroom Years Teaching 13.83 9.98 Number of Students 22.06 6.47 Percent Male 50.67 9.24 Math Classroom Years Teaching 14.06 10.07 Number of Students 22.55 6.27 Percent Male 50.97 9.24 23

PAGE 24

CHAPTER 3 RESULTS Results of HLM reading analyses are included in Table 3-1 and math analyses are in Table 3-2; both are broken down by model. Changes in classroom variance from model to model decreased, indicating that the addi tion of each set of predictors he lped to explain more and more of the explainable variat ion between classrooms. Model 1 depicts the fully unconditional models for both reading and math subjects. These models provide a benchmark by which future models can be compared, describing how much classrooms vary in their subject achievements. The estimated overall classroom average on reading scores was 49.56, and the average on math scores was 49.61; both were below the national average of 50.3 for both subjects. Intraclass coefficients, the observed variations in achievement scores attributable to classroom-l evel characteristics, were significant, with 65 percent for reading and 47 percent for math. This indicated that classroom -level characteristics account for more between-classroom differences in reading than in math classes. Next, model 2 gives information about how classroom characteristics alone affect academic achievement. Comparing the intercepts or classroom, variance between model 1 and model 2 indicated that the level-2 predictors alone explained some of the between-classroom variance observed in the fully unconditional model (1). About 7 percent of the between-classroom va riance in reading achievement is accounted for by the three classroom characteristics, and 6.2 percent of the expl ainable variation in classroom average math achievement scores coul d be explained by the combination of classroom size, teacher experience and gender composition. Finally, because there was more than one student per classroom, there are fewer classrooms th an there are students, which accounts for the decrease in sample size between models 1 and 2. 24

PAGE 25

For classrooms with average number of stude nts, average percentage of boys and the teacher with an average number of years teaching, the estimated overall average reading achievement was 49.589 and the average math achievement was 49.681. For reading, the more years teaching (y01=.075, p<.01) and more stud ents in the classroom (y02=.319, p<.001) were significantly associated with higher average reading achievement scores. For math, higher classroom enrollments (y02=.271, p<.001) was significan tly associated with higher average math achievement scores. For both subjects, the number of students in the clas sroom had the strongest effects on achievement scores. Model 3 shows how demographic characteristics of gender and race affect academic achievement. These predictors aid in reducing the between-classroom variance (from the fully unconditional model) by 15 percent for reading a nd 21 percent for math. Also, the predictors reduce the within-classroom variance by only .55 pe rcent for reading and .40 percent for math, indicating that very little vari ance within-classrooms can be explained by gender and race. The average classroom reading achievement wa s 52.41, while the aver age classroom math achievement was 50.60, controlling for gender and race. In reading, male respondents (y10=-1.17, p< .01) were found to have lower average expected reading scores than females, by 1.17 points. Black (y20=-5.72, p<.001), Hispanic (y30=-4.33, p<.001), Asian (y40=-2 .67, p<.001) and respondents of other races (y50=-3.26, p<.001) were found to have lower average expected reading scores than White respondents. In math, male respondents (y10=1.59, p<.001) were found to have highe r average expected scores than females, by 1.59 points. Black (y20=-5.72, p<.001) and Hispanic (y30=-4.33, p<.001) respondents were found to have lower av erage expected math scores than White 25

PAGE 26

respondents. Asian (y40=.50, p>.05) and responden ts of other races (y50=-2.31, p>.05) were not significantly different from White respondents in their average expected mathematics scores. In model 4, the average classroom reading achievement was 51.99 for white, female students, in households with average income, average number of people, whose parents were married, did not work, and had average educationa l attainments and ages. For math, the average classroom math achievement was 50.57 for white, male students, in households with average income, average number of people, whose parent s were married, did not work, and had average educational attainments and ages. For both subjects, mothers educ ation had the strongest impact on achievement scores. In this model, about 41.5 percent of the explainable variation in between-classroom average reading achievement scores, and 55.8 per cent of variation in be tween-classroom average math achievement scores could be explaine d by gender, race and family demographic characteristics. At the same time, only 7.2 percen t of the explainable variation within-classroom average reading achievement scores, and 1.5 percent of variation in average math scores could be explained by the predictors. Additionally, the addition of family demographics reduced the between-classroom variance by the largest amount. This indicates that family demographic ch aracteristics account for larger amounts of the observed between-classroom variance than demographics alone, or the combination of child and family demographics and familial involvement, or the combination of child and family demographics, familial involvement and classroom characteristics. For reading scores in model 4, male res pondents (y10=-1.33, p<.001) were found to have lower average expected reading scores than females, by 1.33 points. Black (y20=-3.28, p<.001), Hispanic (y30=-1.26, p<.05), Asian (y40=-1.41, p< .05) respondents were found to have lower 26

PAGE 27

average expected reading scores than White respondents. Respondents of other races (y50=-1.19, p>.05) were not significantly diffe rent from White respondents in their average expected reading scores, changing from the previ ous model. Here, the addition of family demographics reduced the differences between white childr en and those of other races. Higher education for mothers (y70=.92, p<.001) and fathers (y100=.53, p<.001), older fathers age (y90=.12, p<.01), and higher household incomes (y160=.50, p<.001) were significantly associated with higher average ex pected reading scores. At the same time, household size (y170=-.70, p<.001) was negatively, though significantly, associated with reading scores, where for each person added to the hous ehold, the expected average reading score decreased by .70 points. Only children whose pa rents were divorced ( y140=-1.12, p<.05) were found to have lower average expected reading scor es than children whos e parents were married. Children whose parents were single (y120= -.37, p>.05), separated (y130=-.04, p>.05) or widowed (y150=-.16, p>.05) were not significantly di fferent than children whose parents were married in their average expected reading scores. For math, male respondents (y10=1.51, p<.001) were found to have higher average expected math scores than females, by 1.51 poi nts. Black (y20=-3.39, p<.001) respondents were found to have lower average expected math scores than White respondents, by 3.39 points. Hispanic (y30=-.90, p>.05), Asian (y40=1.07, p>.05) and other race (y50=-.99, p>.05) respondents were not significantl y different from White responde nts in their average expected math scores. In this model, the difference between Hispanic and White childrens achievements was eradicated, indicating that when family demographics are equal, White and Hispanic children are no different in their average achievement scores. Higher education for mothers (y70=.85, 27

PAGE 28

p<.001) and fathers (y100=.45, p<.001), and high er household incomes (y160=.45, p<.001) were also associated with higher average expected math scores. In model 5, the average classroom readi ng achievement was 52.18, while the average classroom math achievement was 50.80, controlling for gender, race and family characteristics. Additionally, 45.6 percent of the explainable va riation in between-cla ssroom average reading achievement scores, and 60.6 percent of vari ation in between-classroom average math achievement scores can be explained by gender, race and family characteristics. At the same time, looking at within-classroom variation, 11.5 percent of the vari ation in reading achievement, and 3.8 percent of the variation in math achieve ment could be explaine d by the independent variables. This increase from the previous model indicates that differences in family characteristics account for some of the observe d within-class variation in scores. For both subjects, the parental expectations of educa tional achievement had the strongest effect on achievement scores. For reading, male responde nts (y10=-1.00, p<.01) were f ound to have lower average expected reading scores than females, by 1 point. Black (y20=-3.91, p< .001), Hispanic (y30=2.00, p<.001), Asian (y40=-2.15, p<.001) and othe r race (y50=-1.84, p<.01) respondents were found to have lower average expected reading sc ores than White respondents. Here, the addition of familial involvement characteristics increased the differences between students of white and other races, seen by the change in significance from model 4 to model 5. Higher education for mothers (y70=.70, p<.001) and fathers (y100=.42, p<.001), older fathers age (y90=.12, p<.01), and higher household incomes (y160=.42, p<.001) were again significantly associated with higher average exp ected reading scores. However, children from divorced and married families were not significantly different from one another in this model as 28

PAGE 29

they were in model 4. The addition of familial in volvement characteristics to the model reduced the differences between these tw o groups of children, indicatin g that when some forms of parental involvement are controlled for, children from divorced and married families are no different in their achievement scores. At the same time, household size (y170=.66, p<.001) was negatively associated with reading scores, where for each person added to the household, the expected average reading score decreased by .66 points. Additionally, pa rents who had higher involvement levels (y190=.21, p<.05) and expected higher educations (y200=1.80, p< .001) for their children were significantly associated with higher average expected reading scores. For math, male respondents (y10=1.77, p<.001) were found to have higher average expected math scores than females, this time by 1.77 points. Black (y20=-3.91, p<.001) and Hispanic (y30=-1.59, p<.01) respondents were f ound to have lower average expected math scores than White respondents. In looking at the change in significance from model 4 to this model, the addition of familial involvement charact eristics seems to increase the differences in achievement between white and Hispanic students. Asian (y40=.14, p>.05) and other race (y50=-1.30, p>.05) respondents were not significantly different from White respondents in their average expected math scores. Higher education for mothers (y70=.67, p<.001) and fa thers (y100=.35, p<.001), older fathers age (y90=.06, p<.05), higher household incomes ( y160=.39, p<.001), and parents who expected higher educations (y200=1.59, p<.001) were also significantly associated with higher average expected math scores. In this model, fath ers age became significantly associated with achievement scores when controlling for the three familial involvement characteristics. 29

PAGE 30

Finally, for the last model, the average classroom reading achievement was 52.26, while the average classroom math achievement was 50.88, controlling for gender, race, family characteristics and some classroom characteristic s as well. About 48.1 percent of the explainable variation in between-classroom average reading achievement scores, and 62.2 percent of variation in between-classroom average math achievement scores could be explained by all predictors: gender, race, family and classroom ch aracteristics. Finally, only 11.6 percent of the within-classroom reading achievement variation, and 4.3 percent of the wi thin-classroom math achievement variation could be explained by the comb ination of all predictors. Similar to the last model, the parental expectations of educa tional achievement had the strongest effect on achievement scores. In this model, the sample size decreased from 10,874 to 4,265 for reading and to 2,730 for math. This decrease again accounted for the sm aller number of classroom observations than student observations. Here, HLM averaged the student and family characteristics for each classroom to use for analysis. For reading, male responde nts (y10=-.94, p<.01) were f ound to have lower average expected reading scores than females, by .94 points. Black (y20=-3.99, p<.001), Hispanic (y30=2.42, p<.001), Asian (y40=-2.41, p<.001) and other race (y50=-1.91, p<.001) respondents were found to have lower average expected r eading scores than White respondents. Higher education for mothers (y70=.69, p<.001) and fathers (y100=.41, p<.001), older fathers age (y90=.13, p<.001), and higher household incomes (y160=.39, p<.001) were significantly associated with higher average ex pected reading scores. At the same time, household size (y170=-.65, p<.001) was negatively a ssociated with reading scores, where with each person added to the household, the expected average reading score decreased by .65 points. 30

PAGE 31

31 Additionally, parents who had higher involvement levels (y190=.20, p<.05) and expected higher educations (y200=1.72, p< .001) for their childre n, and classrooms with more students (y02=.23, p<.001) were significantly associated with hi gher average reading achievement scores. The effect of teacher experience observed in model 2 was eradicated by the addition of demographic and family demographic and involvement characteristics included in model 6. For math, male respondents (y10=1.81, p<.001) were found to have higher average expected math scores than females, by 1.81 points. Black (y20=-3.94, p<.001) and Hispanic (y30=-1.92, p<.01) respondents were found to have lower average expected math scores than White respondents. Asian (y40=-.17, p>.05) and other race (y50=-1.36, p>.05) respondents were not significantly different from White respondents in their aver age expected math scores. Higher education for mothers (y70=.65, p<.001) and fathers (y100=.36, p<.001), older fathers age (y90=.07, p<.05), higher househol d incomes (y160=.37, p<.001), and parents who expected higher educations (y200=1.53, p< .001), and classrooms with higher student enrollments (y02=.21, p<.001) were al so associated with higher av erage expected math scores.

PAGE 32

Table 3-1. Hierarchical linear modeling resu lts of coefficients of betw een-classroom models of fift h-grade reading achievement Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Demographic Variables Gender (male) -1.169 / -0.061 -1.325 / -0.069 -0.995 / -0.052 -0.938 / -0.049 Race (White) African American -5.721 / -0.186 -3.281 / -0.106 -3.905 / -0.127 -3.994 / -0.130 Hispanic -4.328 / -0.177 -1.261 / -0.051 -2.001 / -0.082 -2.421 / -0.099 Asian -2.669 / -0.067 -1.405 / -0.035 -2.151 / -0.054 -2.407 / -0.060 Other -3.255 / -0.078 -1.187 / -0.029 -1.835 / -0.044 -1.906 / -0.046 Household Characteristics Mother's Age -0.067 / -0.046 -0.066 / -0.046 -0.069 / -0.048 Mother's Education 0.922 / 0.175 0.703 / 0.134 0.686 / 0.131 Mother Employed -0.126 / -0 .006 -0.084 / -0.004 -0.048 / -0.002 Father's Age 0.115 / 0.088 0.115 / 0.087 0.125 / 0.095 Father's Education 0.533 / 0.118 0.416 / 0.092 0.409 / 0.090 Father Employed -0.435 / -0.013 -0.541 / -0.016 -0.654 / -0.019 Marital Status (Married) Single -0.372 / -0.011 -0.071 / -0.002 -0.015 / -0.001 Separated -0.040 / -0.001 -0.070 / -0.001 -0.035 / -0.001 Divorced -1.123 / -0.038 -0.907 / -0.030 -0.960 / -0.032 Widowed -0.157 / -0.002 -0.377 / -0.005 -0.392 / -0.005 Income 0.498 / 0.166 0.417 / 0.139 0.390 / 0.130 Number in Household -0.699 / -0.102 -0.658 / -0.096 -0.653 / -0.095 Time for Homework 0.005 / 0.021 0.004 / 0.018 Parental Involvement 0.213 / 0.036 0.204 / 0.034 Expected Education 1.795 / 0.197 1.717 / 0.189 32

PAGE 33

33Table 3-1. Continued Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Classroom Characteristics Years Teaching 0.075 / 0.078 0.028 / 0.029 Number of Students 0.319 / 0.216 0.230 / 0.155 Percent Male -0.047 / -0 .046 -0.017 / -0.017 Intercept 49.557 49.589 52.408 51.990 52.175 52.258 Classroom Variance 69.417 64.572 58.985 40.626 37.794 36.006 Change in Classroom Variance 4.845 10.432 28.791 31.623 33.411 Class. Proportion of Variance 0.07 0.15 0.415 0.456 0.481 Student Variance 37.734 37.771 37.528 35.019 33.411 33.346 Student Proportion of Variance -0.001 0.005 0.072 0.115 0.116 X221442.988 21168.712 18799.270 15308.517 15407.958 15103.600 N 10874 4265 10874 10874 10874 4265 Source: Early Childhood Longitudinal Survey, Kindergarten Class of 1998-1999, Fifth Grade Wave Unstandardized / Standardized Coefficients p .05; p .01; p .001

PAGE 34

Table 3-2. Hierarchical linear modeling resu lts of coefficients of betw een-classroom models of fi fth-grade math achievement Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Demographic Variables Gender (male) 1.592 / 0.085 1.506 / 0.080 1.774 / 0.094 1.806 /0.096 Race (White) African American -5.652 / -0.175 -3.391 / -0.112 -3.972 / -0.123 -3.943 / -0.122 Hispanic -3.290 / -0.133 -0.897 / -0.037 -1.590 / -0.064 -1.916 / -0.077 Asian 0.504 / 0.013 1.065 / 0.027 0.140 / 0.004 -0.172 /-0.004 Other -2.312 / 0.059 -0.990 / -0.025 -1.297 / -0.033 -1.355 /-0.035 Household Characteristics Mother's Age -0.001 / -0.001 0.006 / 0.004 0.001 / 0.001 Mother's Education 0.845 / 0.162 0.668 / 0.128 0.652 / 0.125 Mother Employed 0.044 / 0.002 0.139 / 0.007 0.211 / 0.010 Father's Age 0.052 / 0.040 0.058 / 0.044 0.066 / 0.051 Father's Education 0.448 / 0.100 0.346 / 0.077 0.361 / 0.081 Father Employed -0.757 / -0.023 -0.770 / -0.023 -0.701 / -0.021 Marital Status (Married) Single -1.157 / -0.034 -0.866 / -0.026 -0.825 / -0.025 Separated -0.383 / -0.008 -0.131 / -0.003 -0.048 / -0.001 Divorced -0.611 / -0.021 -0575 / -0.020 -0.568 / -0.019 Widowed -1.097 / -0.014 -1.285 / -0.016 -1.353 / -0.017 Income 0.446 / 0.150 0.394 / 0.133 0.368 / 0.124 Number in Household -0.291 / -0.043 -0.251 / -0.037 -0.251 / -0.037 Time for Homework -0. 002 / -0.007 -0.003 / -0.011 Parental Involvement 0.099 / 0.017 0.113 / 0.019 Expected Education 1.592 / 0.176 1.532 / 0.170 34

PAGE 35

35Table 3-2. Continued Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Classroom Characteristics Years Teaching 0.048 / 0.052 0.004 / 0.004 Number of Students 0.271 / 0.181 0.206 / 0.137 Percent Male 0.045 / 0.044 0.019 / 0.019 Intercept 49.607 49.681 50.602 50.565 50.797 50.875 Classroom Variance 45.989 43.135 36.297 20.352 18.138 17.372 Change in Classroom Variance 2.854 9.692 25.637 27.851 28.617 Class. Proportion of Variance 0.062 0.211 0.557 0.606 0.622 Student Variance 51.516 51.560 51.310 50.725 49.565 49.307 Student Proportion of Variance -0.001 0.004 0.015 0.038 0.043 X29193.533 9437.227 8012.716 6228.384 6031.640 6214.000 N 10874 2730 10874 10874 10874 2730 Source: Early Childhood Longitudinal Survey, Kindergarten Class of 1998-1999, Fifth Grade Wave Unstandardized / Standardized Coefficients p .05; p .01; p .001

PAGE 36

CHAPTER 4 CONCLUSION Findings The results support the first th ree hypotheses (1a, b and c). Gender did have a significant association with achievement scores of both r eading and math. Supporting current research and speculations from the field, females average scor es were higher in reading, and males were higher for math. In both subjects however, gi rls and boys achieved at or above the national average. Though results may appear small male s and females were within two points of one another for average expected reading and math scores due to the cross-sectional analyses, it is unknown if these differences have been decr easing or increasing since kindergarten. Interestingly, for reading, as more predictors were added to the model, the average score difference between males and females decreased, but increased for math scores (outside of the addition of family characteristics). It would app ear as if math classes with more heterogeneous characteristics are more beneficial to girls' advancements, while reading classes with more homogeneous characteristics benefit males. The second hypothesis is supported, with parent al education positively associated with subject achievement scores. For both math and reading, both mothers an d fathers education levels were positively associated with average achievement scores across models 4 through 6. This finding supports social le arning theory and previous wo rk of Vilenius-Tuohimaa et al. (2008) and Hayeman and Wolfe (1995). The third hypothesis was partially supporte d. Parental involvement had a positive, significant effect for reading subject area (supporting Crosnoe 2001), but not for math. Future research might look at an interaction effect with parental involvement and homework time, gender, race, etc. For example, are parents who are more involved at scho ol also more involved 36

PAGE 37

at home? If so, there may be subject-area differe nces in the type of ho mework or interaction between parents and children. The fourth hypotheses (4a and 4b), relating gender composition to academic achievement, were not supported in either s ubject. Interaction-level effect s of gender and classroom gender distribution might be useful in fu ture analyses to determine the im pact, if any, the combination of childs gender and classroom composition of gender has on overall classroom achievement (and to aid in determining the minority and majority group (Dryler 1999)). The last hypothesis that cla ss size is negatively associat ed with reading and math achievement scores, was not supported, contradicting the findings of Funkhouser (2009) and Nye, Hedges, and Konstantopoulos (2004). For bo th subjects, larger classroom sizes were associated with higher achievement scores. Th ese findings could be contradictory due to the difficulty in separating the effect s of classroom size with other characteristics, such as teacher experience, which often have interactive eff ects with one another (Funkhouser 2009; & Nye, Hedges and Konstantopoulos 2004). The relationship of classroom size to student achievement may be affected (and perhaps misleading) by not taking these interaction-level effects into account. Results also indicated a racial gap in academ ic achievement, supporting the Monitoring the Future statistics in the Specific Aims. Across all models, Black student s consistently lagged behind their White counterparts in both readi ng and math scores, while Hispanic and Asian students fell behind White counterparts in readin g. Across all models, Black students scored lower than the national average in both readi ng and math, highlighting the continued need for specialized attenti on and resources targeted at this at-risk group. Ag ain though, due to the cross37

PAGE 38

sectional nature of the data, it is unknown if these differences ha ve been increasing or decreasing since the beginning of school. Higher incomes were associated with higher subject achievement scores, perhaps due to an increasing availability of resources outside of school to aid in academic advancement (Lareau 2003). At the same time, larger household sizes were associated with lower reading scores, but had no significant effects on math scores. Famili al motivation also had significant impacts on average academic achievement scores, with high parental educational expectations associated with high average achievement scores. In the last two models, parental educational expectations had the strongest effects on reading and math achievements. Also, across models 4 through 6, mothers education had stronger impacts on achi evement than fathers education did. This demonstrates that continued emphasis on educati on is important for the future of our nations children, and perhaps even more so for young females. Though there were many similarities among th e predictors of high average achievement scores for both reading and math, there were also some differences. For example, why did household size have a negative effect on reading sc ores but not on math ones? Why was parental involvement significantly associated with high reading scores but, again, not math ones? What is it about the subject of re ading, and perhaps the homework assi gned for it, that is significantly affected by these family characteristics? Math and reading are two dissim ilar subjects, and there are often different class activities, homework ex ercises and demonstrations in the learning of each. Perhaps the differences in how people learn and process math and reading could account for some of these conflicti ng findings between the two subj ects. Additionally, the observed differences between the subjects ai d in cautioning parents and educat ors about one si ze fits all policies aimed at improving achievement levels. 38

PAGE 39

Furthermore, besides the second model (where the level-2 predicto rs explained similar percentages), the proportion of explainable va riance in between-clas sroom achievements was smaller for reading than for math subjects. Th is suggests that predic tors affect subject achievements differently, and there are other, addi tional characteristics besides child and family demographics and classroom predictors that are affecting the differences in between-classroom reading achievement scores. Along this line, prog rams aimed at reading reform or improvement should be targeted accordingly and treated differently than ones for math. In all, results highlighted the family charac teristics of high achievement in both subject areas. In reading, being female, White, high mothers and fathers educati on, older fathers ages, high income, low total household number, high parental involvement and high parental educational expectations were positively associated with high average reading achievement scores. In math, being male, White or Asian, high mothers and fathers education, high income and high educational expectations were posi tively associated with high average math achievement scores. Limitations and Suggestions for Future Research Although there are a number of in teresting variables in the EC LS-K, many of them are not applicable for the overall scope of this research. One definite limitation of the selected variables is that they may not provide a complete picture of the childs overall achievement in math and reading. For example, there are no measures of child achievement, measured by the parent(s). This assessment could be useful, for it may influence a variety of home factors, like the amount of time set aside for homework each day. Parents of children who are not yet proficient in a particular skill or knowledge area may impose longer mandatory homework time in order to aid the student in skill mastery. 39

PAGE 40

At the same time, the selection of IRT scores for the dependent variable may have biased classroom-level effects on students achievement. The fifth-grade IRT scores were calculated using the fifth-grade proficiency exams, as well as those from previous waves. These calculations assume consistency among the students in their answering pa tterns i.e., answering more questions right in one wave but not on the next might indicat e (under IRT parameters) that the child had guessed at previous answers and adjust the students scores accordingly. In this way, the IRT calculations may average improvements in scores, and therefore reduce the effects of teacheror classroom-level characteristics. Additionally, teacher and child measures of p eer achievement would aid in examining the effect of peer groups on a child s achievement (Van de gaer et al, 2006) and what their achievements are, relative to their peers. It woul d also be useful to include the teachers gender and race in the analysis, to see how these variab les affect childrens scores, perhaps depending on the gender and race of the ch ild. Although these two variables are asked on the ECLS-K, they are excluded from the public-use sample and were not available for this research. Furthermore, exploring different pedagogies or teaching styles might be useful in an attempt to understand the differences between the two subj ects how students le arn them and what thought processes are best suited for each subject area. These results indicated that ve ry little within-classroom varia tion could be explained by the included variables. This indicates that classrooms are somewhat homogenous with respect to student and family characteristic s, and school-level effects may pr ovide a more complete picture of student achievement. Future studies should in clude school characteristics to examine student and classroom differences between and within sc hools. Three-level hierar chical linear models could be run to determine the effects of not only students clustere d within schools, but 40

PAGE 41

41 classrooms as well. Previous work has focused on this variation of achievement within and between schools, and this added leve l of analysis would be useful in future studies with this, past and future waves of the ECLS-K. Finally, once more of these variables and leve ls of analysis have been examined, and we have a more holistic understanding of what is af fecting achievement, future research would want to model change across waves. Results of th is study indicated that there are significant differences in the mastery achievements of male and female fifth-grade students, with gender, race, income and parental education all having significant impacts on mastery levels. Modeling change (Eccles, Adler & Meece, 1984; LoGer go, Nichols & Chaplin, 2006) could indicate if these gaps are widening or get ting smaller as children age an d progress through the educational system, and allow policy makers to be tter target programs of at-risk youth.

PAGE 42

REFERENCES Benbow, Camilla P., and Julian C. Stanley. 1980. Sex Differences in Mathematical Ability: Fact or Artifact? Science 210:1262-64. Burgess, Robert L., and Ronald L. Akers. 1966. A Differential Association: Reinforcement Theory of Criminal Behavior. Social Problems 14:128-47. Cooper, Carey E., and Robert Crosnoe. 2007. T he Engagement in Schooling of Economically Disadvantaged Parents and Children. Youth & Society 38:372-91. Crosnoe, Robert. 2001. Academic Orientation and Parental Involvement in Education during High School. Sociology of Education 74:210-30. Davis-Kean, Pamela E. 2005. The Influence of Pa rent Education and Family Income on Child Achievement: The Indirect Role of Parental Expectations and the Home Environment. Journal of Family Psychology 19:294-304. Dee, Thomas S. 2006. The Why Chromosome. Education Next 6:69-75. Ding, Cody S., Kim Song, and Lloyd I. Richardson. 2006. Do Mathematical Gender Differences Continue? A Longitudinal Study of Gender Difference and Excellence in Mathematics Performance in the U.S. Educational Studies 40:279-95. Dryler, Helen. 1999. The Impact of School a nd Classroom Characteristics on Educational Choices by Boys and Girls: A Multilevel Analysis. Acta Sociologica 42:300-18. Funkhouser, Edward. 2009. The Effect of Kinde rgarten Classroom Size Reduction on Second Grade Student Achievement: Evidence from California. Economics of Education Review 28:403-14. Guo, Guang, and Kathleen M. Harris. 2000. The Mechanisms Mediating the Effects of Poverty on Childrens Intellectual Development. Demography 37:431. Halle, Tamara G., Beth Kurtz-Costes, and Joseph L. Mahoney. 1997. Family In uences on School Achievement in Low-Income, African American Children. Journal of Educational Psychology 89:527. Hanushek, Eric A., John F. Kain, Jacob M. Ma rkman, and Steven G. Rivkin. 2003. Does Peer Ability Affect Student Achievement? Journal of Applied Econometrics 18:527. Harker, Richard. 2000. Achievement, Gender and the Single-Sex/Coed Debate. British Journal of Sociology of Education 21:203-18. Haveman, Robert, and Barbara Wolfe. 1995. The Determinants of Childrens Attainments: A Review of Methods and Findings. Journal of Economic Literature 33:1829. 42

PAGE 43

Hyde, Janet S., Elizabeth Fennema, and Susa n J. Lamon. 1990. Gender Differences in Mathematics Performance: A Meta-Analysis. Psychological Bulletin 107:139-55. Ireson, Judith, Susan Hallam, Sarah Hack, Hele n Clark, and Ian Plewis. 2002. Ability Grouping in English Secondary Schools: Effects on Attainment in English, Mathematics and Science. Educational Research and Evaluation 8:299. Jackson, Carolyn, and Ian D. Smith. 2000. Poles Apart? An Exploration of Single-Sex and Mixed-Sex Educational Environmen ts in Australia and England. Educational Studies 26:409-22. Kandel, Denise B. 1980. Drug and Drinking Behavior Among Youth. Annual Review of Sociology 6:235-85. Kowaleski-Jones, Lori, and Greg J. Duncan. 1999. The Structure of Achievement and Behavior across Middle Childhood. Child Development 70:930-43. Laureau, Annette. 2003. Unequal Childhoods: Class, Race and Family Life Berkeley, CA: University of California Press. Lee, Jihyun, Wendy S. Gri gg, and Gloria S. Dion. 2007. The Nation's Report Card: Mathematics 2007. Washington, DC: U.S. Department of Ed ucation, National Center for Education Statistics. Lee, Jihyun, Wendy S. Grigg, a nd Patricia L. Donahue. 2007. The Nation's Report Card: Reading 2007 Washington, D.C.: U.S. Department of Education, National Center for Education Statistics. LoGerfo, Laura, Austin Nichols, and Duncan Chaplin. 2006. Gender Gaps in Math and Reading Gains during Elementary and High School by Race and Ethnicity Washington, DC: Urban Institute. Muller, Chandra. 1998. Gender Differences in Parental Involvement and Adolescents Mathematics Achievement. Sociology of Education 71:336-56. Nye, Barbara A., Larry V. Hedges, and Spyros Konstantopoul os. 2004. Do Minorities Experience Larger Lasting Be nefits from Small Classes? The Journal of Educational Research 98:94-100. OConnell, Ann A., and D.B. McCoach (Eds). 2008. Multilevel Modeling of Educational Data. Charlotte, NC: Information Age Publishing Inc. Painter, John. 2002. Designing Multilevel Models Using SPSS 11.5 Mixed Model Chapel Hill, NC: Jordan Institute for Families. Parsons, Jacquelynne E., Terry F. Adler, and Judith L. Meece. 1984. Sex Differences in Achievement: A Test of Alternative Theories. Journal of Personality and Social Psychology 46:26-43. 43

PAGE 44

44 Raudenbush, Stephan W., and Anthony S. Bryk. 1986. A Hierarchical Model for Studying School Effects. Sociology of Education 59:1-17. ___. 2002. Hierarchical Linear Models: App lications and Data Analysis Methods (2nd ed.). Thousand Oaks, CA: Sage. Raudenbush, Stephen, Anthony Bryk, and Richard Congdon. 2005. HLM6 Lincolnwood, IL: Scientific Software International. Roberts, J.K. 2004. An Introductory Primer on Multilevel and Hierarchi cal Linear Modeling. Learning Disabilities: A Contemporary Journal 2:30-8. Sadker, Myra, and David Sadker. 1995. Failing at Fairness: How Our Schools Cheat Girls New York: Charles Scribners Sons. Singer, Judith D. 1998. Using SAS PROC MIXE D to Fit Multilevel Models, Hierarchical Models, and Individual Growth Models. Journal of Educational and Behavioral Statistics 24:323-55. Van de gaer, Eva, Heidi Pustjens, Jan Van Damme, and Agnes DeMunter. 2004. Effects of Single-Sex Versus Co-Educational Classes a nd Schools on Gender Differences in Progress in Language and Mathematics Achievement. British Journal of Sociology of Education 25:307-22. ___. 2006. The Gender Gap in Language Achievem ent: The Role of School-Related Attitudes of Class Groups. Sex Roles 55:397-408. Vilenius-Tuohimaa, Piia M., Kaisa Aunola, and Jari-Erik Nurmi. 2008. The Association Between Mathematical Word Probl ems and Reading Comprehension. Educational Psychology 28:409-26.

PAGE 45

BIOGRAPHICAL SKETCH Kristi Lynn Donaldson was born in New Jersey but spent the majority of her childhood in Clearwater, Florida. She gra duated high school in 2001 from Pa lm Harbor University High School, earning an International B accalaureate diploma. She then attended the University of Florida for her undergraduate work, majoring in public relations, and la ter, sociology. After earning two bachelors degrees a nd graduating with honors in both s ubjects, Kristi took a year off from school to work and travel. In 2006, she re turned to UF to begin graduate studies, and will graduate with her M.A. in sociology in August 2009. Following graduation, Kristi will move to Indiana to continue on to her Ph.D. studies at the University of Notre Dame, where she plans to research educational inequali ties within the Center for Research on Educational Opportunity. 45