Parental Involvement and Students' Academic Achievement

8/10/2019 Parental Involvement and Students' Academic Achievement

1/22


2/22

2 Fan and Chen

INTRODUCTION

Society in general, and educational researchers in particular, have long

been interested in the positive effect that parental involvement may haveon students academic achievement (e.g., Christenson, Rounds, and Gorney,1992; Epstein, 1991; Keith, 1991; National Center for Education Statistics[NCES], 1997). The perception that parental involvement has positive ef-fect on students academic achievement/success is so intuitively apealingthat policy makers (Prindle and Resinski, 1989; Van Meter, 1994; Wagnerand Sconyers, 1996), school board administrators (Khan, 1996; Roach, 1994;Wanat, 1994), teachers (Allen, 1996; Clarke and Williams, 1992; Matzye,1995), parents (ECS Distribution Center, 1996; Dye, 1992; Lawler-Prince,

Grymes, Boals, and Bonds, 1994; Schrick, 1992), and even students them-selves (Brian, 1994; Choi, Bempechet, and Ginsburg, 1994), have agreed thatparental involvement is critical for childrens academic success (Akimoff,1996; Austin Independent School District, 1977; Deford, 1996; Edwards,1995; Mendoza, 1996; Mundschenk and Foley, 1994; Ryan, 1992). As a re-sult, there has accumulated what appears to be a voluminous body of lit-erature about parental involvement (Austin Independent School District,1977; Edwards, 1995; Egan, OSullivan, and Wator, 1996; Foster-Harrisonand Peel, 1995; Merttens and Vass, 1993; Patterson, 1994).

Although the appeal of parental involvement as part of a remedy forschool education has been strong in society as a whole, there remain somethorny issues related to research on parental involvement, because the re-search findings in this area have been somewhat inconsistent. Generallyspeaking, although some empirical studies have shown evidence of positiveeffect of parental involvement on school learning (e.g., Christenson, Rounds,and Gorney, 1992; Epstein, 1991; Singh etal.,1995),othershavefoundlittle,ifany, such measurable effect (e.g., Bobbett, 1995; Ford, 1989; Keith, Reimers,Fehrmann, Pottebaum, and Aubey, 1986; Natriello and McDill, 1989).

Research in this area has been somewhat fragmented for quite sometime, because the limited empirical research has been conducted withoutthe benefits of a guiding theoretical framework. This, however, appears tobe changing, because several promising theoretical frameworks for parentalinvolvement have appeared. Epstein (1987, 1992, 1994) suggested a widelyrecognized typology to account for different levels of parental involve-ment in their childrens education. In her early work, Epstein (1987) iden-tified four types of parental involvement in schools: (1) basic obligations,(2) school-to-home communications, (3) parent involvement at school, and(4) parent involvement in learning activities at home. More recently, Ep-stein (1992, 1994) expanded the typology and defined six levels (types) ofschool-related opportunities for parental involvement: (1) assisting parents


3/22

Parental Involvement and Achievement 3

in child-rearing skills, (2) school-parent communication, (3) involving par-ents in school volunteer opportunities, (4) involving parents in home-basedlearning, (5) involving parents in school decision-making, and (6) involving

parents in school-community collaborations. Epstein views this issue mainlyfrom the perspectives of schools, and her research is usually concernedabout what schools (teachers) can do to stimulate more active parentalinvolvement.

Hoover-Dempsey and Sandler (1995) put forth what appears to be acomprehensive theoretical framework about parental involvement. Theirtheoretical conceptualization about parental involvement focuses on threemain issues: (1) why parents become involved in their childrens education,(2) how parents choose specific types of involvement, and (3) why parental

involvement has positive influence on students education outcomes. Thistheoretical framework promises to be more than a typology for parental in-volvement, because it not only deals with specific types of parental involve-ment, but more importantly, it attempts to explain why parents choose tobe involved, and what the mechanisms are through which parental involve-ment exert positive influence on students educational outcomes. Althoughthe theoretical model is promising, it is still unclear how the major elementsin this model can be operationally defined and measured empirically.

Despite its intuitive meaning, the operational use ofparental involve-menthas not been clear and consistent. Parental involvement has been de-fined in practice as representing many different parental behaviors and par-enting practices, such as parental aspirations for their childrens academicachievement and their conveyance of such aspirations to their children (e.g.,Bloom, 1980), parents communication with children about school (e.g.,Christenson et al., 1992; Walberg, 1986), parents participation in schoolactivities (e.g., Stevenson and Baker, 1987), parents communication withteachers about their children (e.g., Epstein, 1991), and parental rules im-posed at home that are considered to be education-related (e.g., Keithet al.,1993; Keithet al., 1986; Marjoribanks, 1983). This somewhat chaotic state inthe definition of the main construct not only makes it difficult to draw anygeneral conclusion across the studies, but it may also have contributed to theinconsistent findings in this area.

Although parental involvement is often simplistically perceived as uni-dimentional, in reality, it is probably better to conceptualize this construct asbeing multifaceted in nature, because parental involvement subsumes a widevariety of parental behavioral patterns and parenting practices (e.g., Balli,1996; Brown, 1994; Snodgrass, 1991; Taylor, Hinton, and Wilson, 1995). Suchan approach has been adopted in several recent empirical studies (e.g., Keithet al., 1993; Singhet al., 1995). Furthermore, there is evidence indicating thatcertain dimensions of parental involvement may have more noticeable effect


4/22

4 Fan and Chen

than some other dimensions on students academic achievement (Singhet al., 1995).

In a similar vein, there are different indicators of academic achieve-

ment, ranging from global indicators, such as post-secondary attainmentand school GPA, to some specific indicators, such as standardized test scoresin a specific academic area (e.g., math), and even to such variables as stu-dents academic aspiration and students academic self-concept. It is possiblethat the measurableeffect of parental involvement on students academicachievement may be different depending on the degree of generality of theacademic achievement measure (Fan, 1997).

A direct result of these multifaceted dimensions of parental involve-ment and academic achievement is the inconsistency in the literature as to

the beneficial effect of parental involvement on students academic achieve-ment. As discussed previously, parental involvement and academic achieve-ment have been operationally defined differently by different investiga-tors. The inconsistent operational definitions of both parental involvementand academic achievement have probably led to some inconsistent findingsabout how beneficial parental involvement is to students academic achieve-ment, with some studies reporting positive empirical relationships betweenparental involvement and students academic achievement (e.g.,Christensonet al., 1992; Epstein, 1991; Singhet al., 1995), and others reporting no mea-

surable effect of parental involvement on students academic achievement(e.g., Bobbett, French, Achilles, and Bobbett, 1995; Ford, 1989; Keithet al.,1986; Natriello and McDill, 1989; Reynolds, 1992; Storer, 1995).

Because of the inconsistencies in the literature both about theexistenceof any measurable positive effect of parental involvement on students aca-demic achievement, and about the extentof such effect, there appears to be astrong need for conducting a meta-analytic synthesis of the literature aboutthe empirical relationship between parental involvement and students aca-demic achievement. Such a quantitative synthesis of the empirical findingsin this area has the potential of providing insights into the relevant issues re-lated to parental involvement research that would otherwise not be readilyavailable from individual studies. It is the purpose of this study to conductsuch a meta-analysis.

The body of literature related to parental involvement in students ed-ucation appears to be huge, replete with studies involving parental involve-ment as a factor in students academic achievement. A closer examinationof the literature, however, reveals that a very small number of these studiesareempiricallybased. Among those empirically based studies that are can-didates for being included in this meta-analysis, there are two different typesof empirical findings: (a) empirical findings in the form of bivariate correla-tions between indicators of parental involvement and students achievement


5/22


(e.g., Gonzalez and Blanco, 1991); and (b) empirical findings in the form ofregression coefficients from regression analysis, or path coefficients fromeither regressionbased path analyses or structural equation models (e.g.,

Keith, 1982; Patrikakou, 1997; Singhet al., 1995).Because a regression or path coefficient, which represents the direct

effect of parental involvement on students academic achievement obtainedfrom a path analysis or structural equation model, is necessarily influencedby other variables in the model in a complicated fashion, regression or pathcoefficients do not lend themselves easily to a meta-analysis, at least notwithin the current framework of meta-analysis. Because of this, we limitedour quantitative synthesis to the first type of empirical findings. We focusedon the bivariate relationship between parental involvement and students

academic achievement, and we conducted meta-analysis involving corre-lation coefficients between the two constructs. For this meta-analysis, wesought to address two general questions:

1. What is the strength of the general relationship between measuredparental involvement and students academic achievement?

2. What are some potential study features that have moderating ef-fect on the relationship between parental involvement and studentsacademic achievement?

METHODS AND PROCEDURES

Both the ERIC and PSYCHLIT data bases were searched using thefollowing key words either singly or in combination: achievement, academicachievement, parents, parental involvement. We initially identified some2,000 articles, papers, or reports spanning more than a ten-year period. Thesewere either published (e.g., in journals and as book chapters) or unpub-lished (e.g., conference presentations, technical reports). Based on abstracts

of these initial 2,000 plus articles and papers, we narrowed our search to sev-eral hundred studies as being relevant to our topic. We further examined thecontents of these several hundred articles, and only those that reported theirown empirical findings were kept as being potentially usable for this meta-analysis. All others were excluded from further consideration. It turned outthat the number of studies that reported empirical findings about the rela-tionship between parental involvement and students academic achievementwas very small.

Among those studies that reported empirical findings, we finally keptfor this meta-analysis only those from which Pearson correlations betweenany of the parental involvement indicators and any of the achievement out-come variables could be obtained. Twenty five studies met our inclusion


6/22

6 Fan and Chen

criteria and were subsequently used in this meta-analysis. From the twenty-five studies, ninety-two correlation coefficients between parental involve-ment and students academic achievement were collected. Although we had

anticipated that the majority of the articles and papers we initially identifiedwere nonempirical, we were still surprised that the number of usable empir-ical studies we were able to find from the literature was so small, becausethe overwhelming majority of articles and papers we initially identified werenonempirical.

Coding

It turned out that the operational definition of parental involvementin the literature was diverse and very different across individual studies.In some studies, the construct parental involvement was clearly defined,and the measurement of this construct was adequately described (e.g., Pengand Wright, 1994). In other studies, however, the descriptions and mea-surement of parental involvement were very ambiguous (e.g., Reynolds,1994). After careful consideration of the variety of definitions for parentalinvolvement described in different studies, we grouped the definitions forparental involvement into several broad dimensions of parental involve-

ment. Similarly, the definition for students achievement also varied fromstudy to study, although it was not as chaotic as that for parental involvement.Table I presents the commonly used indicator variables in the literature forboth parental involvement and for academic achievement. The commonlyused indicator variables are grouped into broader categories based on thecommonalities we identified.

Each effect size measure (i.e., the correlation coefficient betweenparental involvement and students academic achievement) was coded ac-cording to seven study features: (a) the study ID, a number assigned to each

study for identification; (b) sample size, a continuous variable indicating thesample size on which the correlation coefficient is based; (c) the subjectsapproximate average age; (d) ethnicity of the subjects used (five categories);(e) type of measure for academic achievement (three categories); (f) areaof academic achievement (six categories); and (g) parental involvement di-mension (five categories). The coding details for these study features werepresented in Table II, and these features were later used in both descriptiveand inferential analyses for the correlation coefficients between parentalinvolvement and students academic achievement.

Two types of meta-analyses were conducted. The first, which is basedon study features, included all correlation coefficients between parental in-volvement and students achievement, and ignored the fact that some studies


7/22


Table I. Commonly Used Indicator Variables of Parental Involvement and AcademicAchievement in the Literature

a. Parental involvement variables

Parental involvementgeneralParent-child communicationInterest in home/school work (e.g., Paulson, 1994a, b)Assistance with homework (e.g., Gonzales and Blanco, 1991; Peng and Wright, 1994)Discusses school progress (e.g., Yap and Enoki, 1995; Peng and Wright, 1994)

Home supervisionTime spent doing homework (e.g., Fehrmann, Keith, and Reimers, 1987; Peng and

Wright, 1994)Time spent watching TV (e.g., Fehrmannet al., 1987; Paik, 1995; Peng and Wright, 1994)Home surroundings conducive to studying (e.g., Yap and Enoki, 1995)Should come home after school (e.g., Ho Sui-Chu and Willms, 1996)

Educational aspiration for children

Educational expectations (e.g., Hess, Holloway, Dickson, and Price, 1984; Pengand Wright, 1994; Voelkl, 1993)Values academic achievement (e.g., Paulson, 1994a, b)

School contact and participationParents contact school and school contacts parents (e.g., Ho Sui-Chu and Willms, 1996)Parents volunteer at school (e.g., Ho Sui-Chu and Willms, 1996)Parents attend school functions (e.g., PTA meetings) (e.g., Paulson, 1994a, b)

b. Achievement outcome variablesOverall Grades (GPA) (Fehrmannet al., 1987; Steinberg, Elmen, and Mounts, 1989;

Steinberg, Lamborm, Dornbusch, and Darling, N., 1992)MathematicsReading

ScienceSocial Studies

Test scores in mathematics, reading (e.g., Reynolds, 1994),science, social studies (e.g., Keithet al., 1993),music (e.g., Zdzinski, 1992)

Grade Promotion vs. Retention (e.g., Marcon, 1993a, b)

had multiple effect size measures. In this meta-analysis, each study may con-tribute multiple correlation coefficients, and the search for variables thathave potential moderating effects on the relationship between parental in-

volvement and students academic achievement is conducted via a generallinear model analysis with each study feature as an independent variable,and the correlation coefficients between parental involvement and studentsacademic achievement as the dependent variable.

The second meta-analysis is a study effectsmeta-analysis in which, byaveraging multiple effect size measures within one study, each study onlycontributes one effect size measure to the analysis. Bangert-Drowns (1986)suggested this variation of meta-analysis commonly known as study-effectsmeta-analysis. Instead of using multiple effect sizes from one study, study-effects meta-analysis only uses one effect size from each study. In case a studythat has multiple effect sizes, they are typically averaged, and the averageeffect size is then used in the meta-analysis. This approach has the advantage


8/22

8 Fan and Chen

Table II. Coding of Study Features

Study 1 to 25, representing 25 studies used in this meta-analysisSample size A continuous variable representing the actual or estimated sample size

used for each correlation coefficient between parental involvementand students academic achievementAge A continuous variable representing actual or estimated average age of

subjects used in studiesEthnicity

1. Caucasian2. African-Americans3. Hispanics4. Asian-Americans5. Mixed/unknown

Measure of academic achievement1. School GPA

2. Tests3. Other (Teachers Rating, Educational Attainment, Grade Retention, etc.)Area of academic achievement

1. Math, Quantitative2. Reading, Language Arts3. Sciences4. Social Studies5. Other (e.g., music aptitude/achievement)6. General/Unspecified

Parental involvement dimensions1. Educational expectation/aspiration for children2. Communication with children about school-related matters

3. Parental supervision/ home structure related to school matters4. Parental participation in school activities5. Other/General parental involvement

of avoiding statistical dependence caused by multiple effect sizes from thesame study (Hunter and Schmidt, 1990), and it may reduce the potential biasin favor of a few studies that have many effect sizes.

Analyses

Searching for Moderator Variables

General linear model (GLM) analysis was used to assess the effectof each study feature in Table II on the correlation coefficients betweenparental involvement and students academic achievement. The effects ofstudy features on the correlation coefficients between parental involvementand academic achievement was assessed by partitioning the variance inthe correlation coefficients into different sources contributed by the studyfeatures. A common effect size measure, eta-squared [2 = (sum-of-squaressource/sum-of-squarestotal) 100], is used as the descriptive measurefor the effect of each study feature on the correlation coefficients between


9/22


parental involvement and students academic achievement, in addition tothe statistical inferential testing for each study feature. Also in our analy-ses, because the study features are not necessarily independent, the unique

sum-of-squares (Type III sum-of-squares) contributed by each source wereused for computing the2 and for statistical inferential testing.

Averaging Correlations

For each study feature that was revealed by the GLM analysis as havingstrong meaningful moderating effect on the correlation coefficients betweenparental involvement and students academic achievement, average correla-

tion coefficients were then obtained for each level of the study feature (e.g.,average correlation coefficient between parental involvement and studentsacademic achievement for the dimension of parental involvement definedas parental supervision vs. that for parental participation). This average isused as the best estimate for the relationship between parental involvementand students academic achievement for the specific condition.

To guard against the effect of skewness of sampling distributions of cor-relation coefficients (Glass and Hopkins, 1996), in our analyses, we appliedFishersz transformation to the correlation coefficients. [FisherZtransfor-

mation takes theform:Z= tanh1

r, or equivalently,Z= 0.5ln[(1+ r)/(1r)]. Back transformation from Fisher Zto Pearson rtakes the form: r=tahn Z. (for details, see Glass and Hopkins, 1996, pp. 355356).] Each in-dividual correlation coefficient was transformed to corresponding Fishersz, weighted by sample size, averaged by level of a study feature, and thenback-transformed to mean correlation coefficient on the scale of Pearsonr[r z= (nizi)/Ni , where ziis the Fishers z transformed r] (Wolf, 1986).As discussed in Glass and Hopkins (1996), average correlation coefficientsobtained through Fisher z transformation are statistically preferred over that

without Fishersztransformation.

RESULTS AND DISCUSSION

Effects of Study Features

Table III presents the GLM analysis for the potential effects of studyfeatures on the correlation coefficients between parental involvement andstudents academic achievement. In this analysis, both the original corre-lation coefficients between parental involvement and students academicachievement, and their counterparts in the form of transformed Fishers zs,were used as the dependent variables in two separate GLM analyses, and


10/22

10 Fan and Chen

Table III. Effects of Study Features on the Correlation BetweenParental Involvement and Students Academic Achievement (2)

Dependent Variable

Study Features Fisherzs Pearsonrs

Age 5.09 4.22Ethnicity 5.68 4.16Measure of Academic Achievement 1.13 1.06Area of Academic Achievement 27.89 32.13Parental Involvement Dimensions 26.60 35.17

ModelR2 .63 .68

Statistically significant at= .05.

five study features were used as independent variables in the general linearmodel. As explained previously, 2 associated with the each study featurewas used as the measure for the moderating effect of the study feature. 2

represents the percentage of variation in the correlation coefficients betweenparental involvement and students academic achievement that is accountedfor by the study feature in question. A very small2 for a study feature (e.g.,1.13 for Measure of Academic Achievement in Table III) indicates thatthe study feature in question is probably not a moderator variable for therelationship between parental involvement and students academic achieve-

ment, because the levels of the study feature have very similar average cor-relation coefficients, thus not accounting for any substantial proportion ofvariation in the correlation coefficients.

It is obvious from Table III that both Area of Academic Achievement(math, reading, science, social studies, etc.) and Parental Involvement Dimen-

sions(different operational definition of parental involvement) stand out tobe study features that have strong moderating effects on the correlationcoefficients between parental involvement and students academic achieve-ment;Area of Academic Achievementand Parental Involvement Dimensions

account for approximately 28% and 27%, respectively, of the variation inthe dependent variable when Fishers zs were used, and about 32% and35% of the variation when original correlation coefficients were used as thedependent variable.

On the other hand, Measure of Academic Achievement (test scores,school GPA, etc. used to represent academic achievement in individual stud-ies) has no moderating effect (2 = 1.13 and 1.06, respectively, for Fishersz and Pearson ras the dependent variable) on the relationship betweenparental involvement and students academic achievement.Age (2

=5.09

and 4.22 respectively for Fisherszand Pearsonras the dependent variable)andEthnicity (2 = 5.68 and 4.16, respectively, for Fishers z and Pearson


11/22


ras the dependent variable) showed relatively small, although statisticallysignificant, moderating effect on the relationship between parental involve-ment and students academic achievement.

The general linear modeling analyses indicate that the relationship be-tween parental involvement and students academic achievement shouldnot be generalized across different operational definitions of parental in-volvement, nor should it be generalized across different areas of academicachievement. Consequently, it became necessary to examine theaverage cor-relation coefficients between parental involvement and students academicachievement separately for different levels of these two study features. Forthe study feature of Measure of Academic Achievement, the relationshipbetween parental involvement and students academic achievement is obvi-

ously generalizable across the types of measurement for academic achieve-ment (test, GPA, etc.). Both EthnicityandAgehave very small moderatingeffects on the relationship between parental involvement and students aca-demic achievement; as a result, we considered it unnecessary to conduct anyseparate analyses for different levels for these two study features.

Average Correlations

Table IV presents the average correlation coefficients both across all

studies (Overall), and separately for the six levels of Area of AcademicAchievementand the five levels of Parental Involvement Dimensions, twostudy features identified in previous GLM analyses as contributing substan-tially to the variation of correlations between parental involvement andstudents academic achievement across studies.

The overall average correlation coefficient between parental involve-ment and students academic achievement is about .25, based on 92 correla-tion coefficients collected from 25 empirical studies with cumulative samplesize of about 133,577. Based on the guidelines suggested by Cohen (1988,Chapter 3) about the magnitude of correlation coefficient as an effect-sizemeasure, this average correlation coefficient of .25 represents approximatelyamediumeffect size in social sciences (small effect: r= .10, medium effect:r= .30, and large effect: r= .50), which is approximately correspondingto the more popular effect size measure ofd = .52 [d = 2r

1r2 (Wolf, 1986,p. 35)] (d: standardized mean difference between two groups). As suggestedby many researchers, a medium effect size typically represents a noticeableand apparent effect (Stevens, 1990, Chapter 3), and generally, it should notbe regarded as trivial.

This overall medium effect size of r= .25 in Table IV suggests thatparental involvement does have positive influence on students academic


12/22

12 Fan and Chen

Table IV.AverageCorrelation Between Parental Involvement and Students AcademicAchievement

Level of Study Features ka rb runweightedc ni d

Overall 92 .2533 (.2943) 133577Area of academic achievement

Math/quantitative 7 .1805 Ae (.2002) 19506Reading/language arts 8 .1793 A (.1801) 19522Science 6 .1538 A (.1658) 18523Social studies 5 .1768 A (.1794) 16382Other 7 .3424 B (.3973) 32872General/unspecified 59 .3286 B (.3291) 102321

Parental involvement dimensionsAspiration for child education 10 .3978 A (.3931) 24826Communication 10 .1929 B C (.1651) 26493Supervision 12 .0943 B (.1278) 69137Participation 7 .3177 A C (.3229) 56755Other 53 .2975 A C (.3305) 85888

a krepresents the number of correlation coefficients used to compute the mean.bAll correlation coefficients have been transformed to Fishers zs, weighted according

to sample size, averaged, and then back-transformed to their corresponding rs.cUnweighted(Fisher z transformation applied) average correlation coefficients are in

parenthesis.dThis refers to the cumulative sample size across studies used to arrive at this mean

correlation coefficient.eThese are post hoc multiple comparison results. Means with the same letter are not

statistically significant from each other at =

.05 level.

achievement. This finding confirms the intuition harbored by many edu-cators and researchers, that parental involvement and students academicachievement are positively related, although in individual studies, there hasbeen considerable inconsistency about the magnitude of such relationship.

The break-down analysis for the average correlation coefficients of thesix levels ofArea of Academic Achievementshows that, for the majority ofthe reported correlation coefficients (k= 59) between parental involvementand academic achievement, the academic achievement measure is either

very general (such as general school GPA or combined grades in severalacademic areas) or not clearly specified in the original articles. For this largegroup of correlation coefficients between parental involvement and aca-demic achievement, the average correlation is relatively high (r= .33). Butfor studies that focused on achievement in more specific academic areas (e.g.,math, science), the average correlation coefficients are obviously lower, butcomparable (approximatelyr= .18).

We are not entirely clear about the reasons for this observation. How-ever, we believe that general school achievement, such as that representedby school GPA, may be a better indicator for students overall academicachievement than those that focused on a specific academic area (e.g., mathgrade or reading test score). There are two reasons to support our belief.


13/22


First, obviously, general GPA is a more comprehensive indicator of aca-demic achievement than subject-specific indicators. Second, from the mea-surement perspective, GPA is a composite of multiple measurements, and

a composite is generally more reliablethan one of its sub-components. Asis generally known, variable unreliability has a tendency to attenuate thecorrelation coefficient between two variables (Hunter and Schmidt, 1990,Chapter 3); consequently, the correlation between parental involvementand academic achievement could suffer if the measurement of academicachievement is less reliable. If this is true, we have reasons to believe thatthe average correlation for the category of General/Unspecified ( r .33)is a better representation for the relationship between parental involve-ment and academic achievement than those when academic achievement

is represented by subject-specific indicators; thus the findings here suggestthat the relationship between parental involvement and students academicachievement may be slightly stronger than that represented by the overallaverage correlation coefficient of r .25.

The break-down analysis for the levels ofParental Involvement Dimen-sions is also interesting. Some previous research has suggested that somedimensions of parental involvement may have more noticeable effect on stu-dents academic achievement than others (e.g., Singh et al., 1995). The resultshere appear to suggest that parental involvement, as represented by parents

supervision of children at home (e.g., home rules for watching TV, for do-ing school work, etc.), has the weakest relationship with students academicachievement (r .09), whereas parents aspiration and expectation for chil-drens educational achievement appears to have the strongest relationshipwith students academic achievement (r .40). The considerable variationamong the average correlation coefficients between parental involvementand academic achievement contributed by the dimensions of parental in-volvement explains why this variable accounts for a large proportion of thevariance in the general linear model analysis presented in Table III.

The finding that parental supervision has a weak relationship with stu-dents academic achievement, whereas parental aspirationor expectation forchildrens educational achievement has a considerably stronger relationshipwith students academic achievement, confirms what some individual studieshave shown before. For example, Singh et al. (1995), by using a structuralequation modeling approach, presented evidence that parents aspirationfor childrens education is the strongest predictor for academic achievementamong all the dimensions of parental involvement examined in their study,andhomestructure(similartosupervision usedinthisstudy)actuallyshoweda very small negative effect on academic achievement.

The findings above, however, should not be interpreted simplisticallyas indicating that home supervision has very little to offer in enhancing


14/22

14 Fan and Chen

childrens education. One potential reason for the weak relationship be-tween home supervision and students academic achievement as observedhere is that closer parental supervision is implemented at home because stu-

dents are not doing well academically in school in the first place. If this is thecase, close parental supervision in many homes may be the result of poor aca-demic performance of the students. Consequently, parental supervision mayhave weak, or even negative, relationship with students academic achieve-ment. The findings here, however, do suggest that parental home supervisionis probably not a good general indicator for parental involvement.

It should be pointed out that some caution is warranted in interpret-ing the results for these moderator analyses. Because the number of usableempirical studies for this meta-analysis is relatively small in the first place,

break-down analysis for the levels of potential moderator variables (dimen-sions of parental involvement, area of academic achievement) further re-duced the number of correlation coefficients used to compute the averagefor each level of the moderator variable. As a result, the averages presentedin these moderator analyses may not be as stable as we desire them to be.

Unweighted Average Coefficients

In meta-analysis literature, the use of effect sizes (coefficients in this

study) weighted by their respective sample sizes (n) is generally consideredmethodologically critical for obtaining unbiased average correlation coef-ficients in meta-analysis, because unweighted coefficients may contain anunknown degree of bias. To assess the potential biasunweightedcoefficientsmay have, the mean unweighted coefficients are presented side-by-side (inparenthesis) with the sample size weighted coefficients in Table IV. Theunweightedmean coefficients show, in most cases, a slight upward bias com-pared with theweightedmean coefficients.

Unweighted coefficients, however, do serve some useful purposes. One

major use of unweighted coefficients is for describing the distribution ofthe coefficients graphically, such as in stem-and-leaf or bar graph displays.Figure 1 graphically presents the frequency distribution of the correlationcoefficients between parental involvement and students academic achieve-ment. As can be seen in Fig. 1, the majority of the original coefficientsare within the range of 0.1-0.4. But there appear to be two outliers in theneighborhood of 0.90. Upon close examination of the source of these twocoefficients (Reynolds, 1994), it seems that, although it is not entirely clearhow these coefficients were calculated, it appears that the investigator of thisstudy reported correlation coefficients based on group means (group as theunit of analysis), rather than those based on individual scores (individual asthe unit of analysis). As discussed by Glass and Hopkins (1996, pp. 127128),


15/22


Fig. 1. Frequency distribution of unweighted correlation coefficients between parental in-volvement and students academic achievement.

such change of unit of analysis may dramatically impact the correlation

coefficients, making these ecological correlations (Glass and Hopkins,1996) substantially higher than those based on individuals. But because weare not certain on this, we did not excludethese two coefficients from ouranalyses.

Study-Effects Meta-Analysis

Of the 25 studies used for this meta-analysis, a total of 92 correlationcoefficients were collected, because many studies had multiple correlation

coefficients between different aspects of parental involvement with differentmeasures of students academic achievement. As discussed previously, analternative approach to handling nonindependent multiple effect sizes is toconduct astudy-effectsmeta-analysis in which an average effect size is firstobtained from each study, followed by an average of all the effect sizes acrossstudies. This approach has its own advantages and disadvantages, however.

On one hand, this approach avoids the problem of data nonindepen-dence caused by multiple effect sizes from the same study, and it also mayreduce potential bias in favor of those studies with multiple effect sizes. Onthe other hand, it has the disadvantage of making it almost impossible toexamine the potential moderating effects of the study features on the effectsizes. For example, in our analysis, averaging effect sizes within one study


16/22

16 Fan and Chen

usually means obtaining an average effect size across dimensions of parentalinvolvement, or across different areas of academic achievement, or acrossthe levels of both study features. As a result of averaging the effect sizes

across the levels of the study features within each study, we lost the infor-mation about the study features, and break-down analysis for levels of studyfeatures became impossible.

For the reason stated above, from the 25 studies used for this meta-analysis, we were only able to obtain an overall average correlation co-efficient across all 25 correlation coefficients, many of which are averagecoefficients within each individual study. The overall average correlationcoefficient between parental involvement and students academic achieve-mentfromthisstudy-effects meta-analysis is r

=.33. Readers may notice that

this overall average correlation coefficient between parental involvementand students academic achievement from thisstudy-effectsmeta-analysis ishigher than the overall r= .25 presented in Table IV.

A closer look at the data revealed that the discrepancy is most likelycaused by a couple of studies with very large sample sizes and multiple cor-relation coefficients, but some correlation coefficients were quite low (e.g.,Keithet al., 1993). In thestudy-effectsmeta-analysis, each study contributedonly one average correlation coefficient which was then weighted by therespective sample size. In our previousstudy-featuresmeta-analysis, such a

study contributed multiple correlation coefficients, andeachcorrelation co-efficient was weighted by its sample size. In essence, a study with large samplesize and multiple effect sizes would be overweighted in the process of obtain-ing weighted averages. If such a study contains some small effect sizes, theywould bias the overall average effect size by pulling it downward. Thisstudy-effects meta-analysis reveals that the previous overall average correlationcoefficient of r= .25 is probably a slight underestimate of the relationshipbetween parental involvement and students academic achievement.

The overall relationship between parental involvement and students

academic achievement is close to r= .30. Although an average correlationof .30 may at first appear to be rather low, it should be pointed out that thisrepresents a medium effect size in social sciences. A medium effect size isusually considered a meaningful effect, one which is often readily noticeableto researchers (Stevens, 1990).

What difference can a medium effect size make in practical terms? Asshown by Rosenthal and Rubin (1982) and illustrated by Wolf (1986, pp. 32,33), if we characterize parental involvement as above or below median level,and characterize academic achievement as success (above median level)or failure (below median level), a correlation coefficient of .30 betweenthe two variables translates into increasing the success rate of academic


17/22


18/22

18 Fan and Chen

same study. This will provide evidence to verify if indeed the relationshipbetween parental involvement and academic achievement is stronger whenacademic achievement is measured by a global indicator than when it is

measured by a subject-specific indicator.Like many other studies, this meta-analysis has its own share of limi-

tations. The number of usable empirical studies for this meta-analysis wasmuch smaller than we had anticipated for the voluminous body of litera-ture related to parental involvement. The relatively small number of usableempirical studies has probably made the results from moderator analysis(break-down analysis for dimensions of parental involvement, and that forareas of academic achievement) unstable, because the number of effect sizesfor each level of a moderator variable became very small. For this reason,

there should be some caution in interpreting the results from the moderatoranalysis.

Another relevant issue is the relationship between SES and parentalinvolvement. The relationship between social economic status (SES) andparental involvement in students education has attracted the attention ofmany researchers. It is widely believed, and also supported empirically tosome degree, that SESandparental involvementare positively related (Balli,1996; Bracey, 1996; Brody, 1995). If SES does indeed influence parentalinvolvement, then it is very likely that the observed relationship between

parental involvement and students academic achievement in this meta-analysis reflects, to some degree, the relationship between SES and stu-dents academic achievement. This issue could not be addressed in thismeta-analysis, because such information is generally not available from theindividual studies used in this metaanalysis. It is desirable that future stud-ies should include both SES and parental involvement and should exam-ine the relationship between parental involvement and students academicachievement both before and after partialling out the influence of SES. Thisapproach will reveal the strength of relationship between parental involve-

ment and students academic achievement above and beyond the influenceand confounding effect of SES.

ACKNOWLEDGMENT

This research was partially supported by a grant from the American Ed-ucational Research Association which receives funds for its AERA GrantsProgram from the National Science Foundation and the National Center

for Education Statistics (U.S. Department of Education) under NSF Grant#RED-9452861. Opinions reflect those of the authors and do not necessarilyreflect those of the granting agencies.


19/22


20/22

20 Fan and Chen

ECS Distribution Center (1996).Listen, discuss, and act: Parents and teachers views on educa-tion reform.(ERIC Document Reproduction Service No.: ED 397-515)

Edwards, S. L. (1995). The effect of parental involvement on academic achievement in elementaryurban schools. (ERIC Document Reproduction Service No.: ED 398-331)

Egan, C. L., OSullivan, C., & Wator, V. (1996).Improving the reading skills of at-risk students.Unpublished Masters Thesis. St. Xavier University. (ERIC Document Reproduction Ser-vice No.: ED 398-545)

Epstein, J. L. (1987). Parent involvement: What research says to administrators. EducationUrban Soc.19 (2): 119136.

Epstein, J. L. (1991). Effects on student achievement of teachers practices of parent involve-ment. Advances in Reading/Language Research: Literacy Through Family, Community,and School Interaction, Greebwich, CT, JAI Press, Vol. 5, pp. 261276.

Epstein, J. L. (1992). School and family partnerships. In M. Aiken (Ed.), Encyclopedia ofeducational research(6th ed.), Macmillan, New York, pp. 11391151.

Epstein, J. L. (1994). Perspectives and previews on research and policy for school, family,and community partnerships. Paper presented at the National Symposium, Family-SchoolLinks: How Do They Affect Educational Outcomes? Pennsylvania State University, Oc-tober 31November 1, 1994.

Fan, X. (1997). Parental involvement: Its dimensions and longitudinal effect on academic achieve-ment of high school students. Grant proposal submitted to AERA Grants Program.

Fehrmann, P. G., Keith, T. Z., and Reimers, T. M. (1987). Home influence on school learning:Direct and indirect effects of parental involvement on high school grades.J. Ed. Res., 80:330337.

Ford, D. L. (1989).Parental participation and academic achievement. (ERIC Document Repro-duction Service No. ED 344-659)

Foster-Harrison, E. S., and Peel, H. A. (1995). Parents in the middle: Initiatives for success.Schools in the Middle5: 4547.

Glass, G. V., and Hopkins, K. D. (1996).Statistical Methods in Education and Psychology, Allyn& Bacon, Needham Heights, MA.

Gonzalez, R-A. M., and Blanco, N. C. (1991). Parents and children: Academic values and schoolachievement.Int. J. Ed. Res.15: 163169.

Griffith, J. (1996). Relation of parental involvement, empowerment, and school traits to studentacademic performance.J. Ed. Res.90: 3341.

Griffith, J. (1997). Linkages of school structural and socio-environmental characteristics toparental satisfaction with public education and student academic achievement. J. Appl.Social Psychol.27: 156186.

Grolnick, W. S., and Slowiaczek, M. L. (1994). Parents involvement in childrens school-ing: A multidimensional conceptualization and motivational model.Child Devel.65: 237252.

Hess, R. D., Holloway, S. D., Dickson, W. P., and Price, G. G. (1984). Maternal variables as pre-

dictors of childrens school readiness andlater achievement in vocabulary andmathematicsin sixth grade.Child Devel.55: 19021912.

Ho Sui-Chu, E., and Willms, J. D. (1996). Effects of parental involvement on eighth-gradeachievement.Sociol. Ed.69: 126141.

Hoover-Dempsey, K. V., and Sandler, H. M. (1995). Parental involvement in childrens educa-tion: Why does it make a difference? Teachers College Rec.97: 310331.

Hunter, J. E., and Schmidt, F. L. (1990). Methods of Meta-analysis: Correcting Error and Biasin Research Findings. Sage Publications, Newbury Park, CA.

Keith, T. Z. (1991). Parentinvolvement andachievement in high schools. In Silvern, S. (Ed.),Ad-vances in Reading/Language Research: Literacy Through Family, Community, and School

Interaction (Vol. 5), JAI Press, Greenwich, CT.

Keith, T. Z., Keith, P. B., Troutman, G. C., Bickley, P. G., Trivette, P. S., and Singh, K. (1993).Does parental involvement affect eighth-grade student achievement? Structural analysisof national data.School Psychol. Rev.22: 474496.

Keith, P. B., and Lichtman, M. V. (1994). Does parental involvement influence the academicachievement of Mexican-American eighth graders? Results from the National EducationLongitudinal Study.School Psychol. Quart.9: 256272.


21/22


Keith, T. Z., Keith, P. B., Troutman, G. C., Bickley, P. G., Trivette, P. S., and Singh, K. (1993).Does parental involvement affect eighth-grade student achievement? Structural analysisof national data.School Psychol. Rev.22: 474496.

Keith, T. Z., Reimers, T. M., Fehrmann, P. G., Pottebaum, S. M., and Aubey, L. W. (1986).Parental involvement, homework, and TV times: Direct and indirect effects on high schoolachievement.J. Ed. Psychol.78: 373380.

Khan, M.-B. (1996). Parental involvement in education: Possibilities and limitations. SchoolCommun. J.6: 5768.

Lawler-Prince, D., Grymes, J., Boals, B., and Bonds, C. (1994).Parent responses to public schoolprograms for three- and four-year-old children. Paper presented at the Annual Meetingof the Mid-South Educational Research Association. (ERIC Document ReproductionService No.: ED 328-361)

Marcon, R. A. (1993a).At-risk preschoolers: Early predictors of future grade retention. Paperpresented at the Annual Meeting of the Southeastern Psychological Association. (ERICDocument Reproduction Service No.: ED 357-880)

Marcon, R. A. (1993b).Parental involvement and early school success: Following the Class of2000 at year five. Paper presented at the Biennial Meeting of the Society for Research inChild Development. (ERIC Document Reproduction Service No.: ED 357-881)

Marjoribanks, K. (1983). The evaluation of a family learning model.Studies Ed. Eval. 9:343351.Matzye, C. (1995). Parental involvement in middle schools. (ERIC Document Reproduction

Service No.: ED 385-365)Mendoza, Y. (1996). Developing and implementing a parental awareness program to increase

parental involvement and enhance mathematics performance and attitude of at-risk seventhgrade students. Unpublished Masters Final Report. Nova Southeastern University. (ERICDocument Reproduction Service No.: ED 400-971)

Merttens, R., and Vass, J. (1993). Partnerships in maths: Parents and schools, The IMPACTProject. (ERIC Document Reproduction Service No.: ED 387-355)

Mundschenk, N. A., and Foley, R. M. (1994). Collaborative relationships between school andhome: Implications for service delivery.Preventing School Failure39: 1620.

National Center for Education Statistics (NCES). (1997). Fathers Involvement in Their Chil-drens Schools. U.S. Department of Education,Washington, DC.

Natriello, G., and McDill, E. L. (1989). Performance standards, student effort on homework,and academic achievement.Soci. Ed.59: 1831.

Paik, H. (1995). Television viewing and mathematics achievement. Paper presented at the annualmeeting of the International Communication Association. (ERIC Document Reproduc-tion Service No.: ED384-940)

Patrikakou, E. N. (1997). A model of parental attitudes and the academic achievement ofadolescents. J. Res. Devel. Ed.31: 726.

Patterson, S. R. (1994). Increasing parental involvement in grades one, four, and five in a ruralelementary school. Unpublished Ed. D. Dissertation. NovaSoutheastern University. (ERICDocument Reproduction Service No.: ED 389-480)

Paulson, S. E. (1994a). Parenting style and parental involvement: Relations with adolescentachievement.Mid-Western Ed. Researcher7: 611.

Paulson, S. E. (1994b). Relations of parenting style and parental involvement with ninth-gradestudents achievement.J. Early Adol.14: 250267.

Peng, S. S., and Wright, D. (1994). Explanation of academic achievement of Asian Americanstudents.J.Ed.Res.87: 346352.

Prindle, C., and Rasinski, K. A. (1989). The National Education Longitudinal Study of 1988:Data collection results and analysis potential. Paper presented at the Annual Meeting ofthe American Educational ResearchAssociation. (ERIC Document Reproduction ServiceNo.: ED 308-215)

Reynolds, A. J. (1994). Effects of a pre-school follow-on intervention for children at risk.Devel. Psychol.30: 787804.Reynolds, A. J. (1992). Comparing measures of parental involvement and their effects on aca-

demic achievement.Early Childhood Res. Quart. 7: 441462.Roach, V. (1994). The superintendents role in creating inclusive schools. School Admin. 51:

2025, 27.


22/22

22 Fan and Chen

Rosenthal, R., and Rubin, D. (1982). Comparing effect sizes of independent studies. Psychol.Bull.92: 500504.

Ryan, T. E. (1992). Parents as Partners Program.School Commun. J.2: 1121.Schrick, J. (1992).Building bridges from school to home: Getting parents involved in secondary

education. Unpublished Masters Thesis, Dominican College. (ERIC Document Reproduc-tion Service No.: ED 349-519)

Sealover,I.E.(1988).The relationshipbetweenparental involvementand academic achievementof high school students. (ERIC Document Reproduction Service No.: ED 321-191)

Singh, K., Bickley, P. G., Trivette, P., Keith, T. Z., Keith, P. B., and Anderson, E. (1995). Theeffects of four components of parental involvement on eighth-grade student achievement:Structural analysis of NELS-88 data.School Psychol. Rev.24: 299317.

Snodgrass, D. M. (1991). The parent connection.Adolescence26: 8387.Steinberg, L., Elmen, J. D., and Mounts., N. S. (1989). Authoritative parenting, psychosocial

maturity, and academic success among adolescents.Child Devel.60: 14241436.Steinberg, L., Lamborm, S. D., Dornbusch, S. M., and Darling, N. (1992). Impact of parenting

practices on adolescent achievement: Authoritative parenting, school involvement, andencouragement to succeed.Child Devel.63: 12661281.

Stevens, J. (1990).Intermediate Statistics: A Modern Approach. Lawrence Erlbaum Associates,Publishers, Hillsdale, NJ.

Stevenson, D. L., and Baker, D. P. (1987). The family-school relation and the childs schoolperformance. Special issue: Schools and development.Child Devel.58: 13481357.

Stevenson, D. L., and Baker, D. P. (1987). The family-school relation and the childs schoolperformance. Child Devel.58: 13481357.

Storer, J. H. (1995). Increasing parent and community involvement in schools: The importanceof educators beliefs.Commun. Ed. J.22: 1620.

Taylor, R. D. (1996). Adolescents perceptions of kinship support and family managementpractices: Association with adolescent adjustment in African American families. Devel.Psychol.32: 687695.

Taylor, L. C., Hinton, I. D., and Wilson, M. N. (1995). Parental influences on academic perfor-mance in African-American students.J. Child Fam. Studies4: 293302.

Twillie, L. D., Petry, J. R., Kenney, G. E., and Payne, R. (1991).Attitudes of parents and teacherstowards improving academic achievement in inner-city schools. Paper presented at theAnnual Meeting of the Mid-South Educational Research Association. (ERIC DocumentReproduction Service No.: ED 340-802)

Uguroglu, M. E., and Walberg, H. J. (1986). Predicting achievement and motivation. J. Res.Devel. Ed.19: 112.

Van Meter, E. J. (1994). Implementing school-based decision making in Kentucky. NASSPBull., 78: 6170.

Voekl, K. E. (1993). Academic achievement and expectations among African-American Stu-dents.J. Res. Devel. Ed.27: 4255.

Wagner, T., and Sconyers, N. (1996). Seeing the school reform elephant: Connecting policymakers, parents, practitioners, and students. (ERIC Document Reproduction Service No.:ED 400-078)

Walberg, H. J. (1986). Synthesis of research on teaching. In Wittrock, M. C. (Ed.), Handbookof Research on Teaching(3rd ed.), New York: MacMillan pp. 214229.

Wanat, C. L. (1994). Effect of family structure on parental involvement: Perspectives of princi-pals and traditional, dual-income, and single parents.J. School Leadership4: 631648.

Wolf, F. M. (1986). Meta-analysis: Quantitative Methods for Research Synthesis. Sage, NewburyPark, CA.

Yap, K. O., and Enoki, D. Y. (1995). In search of the elusive magic bullet: Parental involvementand student outcomes.School Commun. J.5: 97106.

Zdzinski, S. F. (1992). Relationships among parental involvement, music aptitude, and musicalachievement of instrumental music students. J. Res. Music Ed.40: 114125.

Documents

Parental Involvement and Students' Academic Achievement