teaching and learning conditions in schools with a high proportion of students from single- parent...

1

Teaching and learning conditions in schools with a high proportion of students from single-

parent families as explanation of the negative effect of single-parent family composition of

schools.

Jaap Dronkers,1 Gert-Jan M. Veerman & Suet-Ling Pong

Maastricht University, University of Amsterdam & Pennsylvania State University

Paper presented at the Twelfth Meeting of the European Network for Sociological and

demographic Study of Divorce, October 2-4 in Paris, France.

Introduction

Prior research on the consequences of divorce for children’s educational performance is mainly

restricted to the family context. In this paper, however, we focus on the school context. More

specifically, we study how the single-parent family composition of schools affects the

educational performance of children from single-parent and two-parent families. A similar

approach was applied by Pong (1997, 1998), indeed finding a negative contextual effect of

schools with high numbers of students from disrupted families in the USA. De Lange, Dronkers

& Wolbers (2014) found that the share of single-parent families at school effects on children’s’

educational performance within 25 OECD countries, including the USA.

In the literature, two explanations can be distinguished for the effect of school’s single-parent

family composition on children’s educational performance: i.e. the decline of the community

network of the school and the lower amount of teaching and learning time at school and at home.

According to Pong (1997) and Sun (1999) parental influence on children extends beyond their

own child and reaches the communities in which they live and the schools belonging to these

communities. As previous research has shown, the type of student attending the school appears to

be one of the most important factors influencing the effectiveness of the school (Pong 1997; Pong

1998). Schools with a large concentration of children from single-parent families are usually

characterized by a lower socioeconomic status and by less social capital (i.e. indicated by

parents’ social relations and networks with other parents). Therefore, all children attending such

schools will perform less well, compared to children at schools with a smaller concentration of

single-parent families.

In addition to this community network explanation for the negative contextual effect of family

disruption on children’s educational performance, Dronkers (2010) emphasizes the more difficult

teaching and learning conditions in schools with a high proportion of students from single-parent

families. The effectiveness of education depends on the amount of time that is available for both

teaching and learning, which can be greatly diminished in schools where children have problems

inside or outside the home that interrupt the teaching and learning process. As previously

described, children of divorced parents have on average more emotional and other problems

related to their parents’ divorce. If there are more students in a class with such problems, more

learning and teaching time of the whole class might be used for non-academic goals. Garriga

(2010) found that children of single-parent families are more often too late at school. A higher

percentage of pupils from single-parent families in a school might thus lead to more pupils

arriving too late at school and thus disturbing teaching and learning of the whole class. As a

consequence insufficient learning and teaching time needed to reach a certain educational

performance by all pupils might remain. Conversely, in student populations with none or few

1 Corresponding author: [email protected]

2

children from single-parent families, there might be less loss of teaching and learning time and

thus a higher chance on sufficient time. In fact, the real learning and teaching time might differ in

these two situations, despite identical class schedules, and thus educational performance will

differ between both situations.

The two research questions that we address are the following: 1) Does the school’s

composition of students from single-parent families affects the conditions of teaching and

learning (measured as classroom disruption) in these schools? 2) Do these conditions of teaching

and learning mediate the negative effect of the school’s composition of students from single-

parent families on students’ educational performance?

If we can answer both research questions positively, the teaching and learning conditions is an

additional or the best explanation of the negative effect of the school’s composition of students

from single-parent families on students’ educational performance. If we cannot answer both

research questions positively, the community network explanation is a better explanation of the

negative effect of the school’s composition of students from single-parent families on students’

educational performance.

Data and variables

Data

The analyses have been carried out using the cross-national Program for International Student

Assessment 2009 (PISA). The cross-national PISA contains both social economic background

and lesson behavior information of 15-year-old students from OECD and other developed

countries (OECD, 2012). We focus on the Western countries. Thus, our dataset contains

information on 232701 students in 28 countries (Australia, Austria, Belgium, Canada,

Switzerland, Czech Republic, Germany, Denmark, Spain, Estonia, Finland, France, United

Kingdom, Greece, Hungary, Ireland, Iceland, Italy, Luxembourg, Netherlands, Norway, New

Zealand, Poland, Portugal, Slovak Republic, Slovenia, Sweden, and United States). We omit 1.8

percent of our students due to student with no information about their home situation.

Furthermore, we omit 13.5 percent of these students due to missing data on other independent

variables or because these students are in schools with less than 8 students per school. Surveyed

pupils reported with whom they regularly live at home, and they were offered six possible

persons (mother, father, brother, sister, grandparents, others) whom they could all tick. We only

analyzed students who lived in a two-parent family or a single-mother family. We deleted all

pupils living in other possible family forms, like single-father and grandparents-family. These

possible family forms are rare in a number of involved countries, which might bias our analysis

due to selectivity bias. Finally we also deleted all migrant pupils, both first and second

generation, because single-parent might have a different meaning for migrants (Dronkers &

Kalmijn, 2013). Consequently we use the data of 183982 students in 8001 schools in 28 OECD

countries.

Variables

Dependent Variables

The first dependent variable in this study is classroom disruption as perceived by the individual

student. Classroom disruption is measured by the categorical question whether there is noise and

disorder in the lessons. PISA 2009 contains five possible questions that could measure the

classroom disruption with the following topics: ‘students don’t listen’, ‘wait for quiet’, ‘cannot

work well’, ‘long time to start’, and ‘noise and disorder’. The questionnaire for students contains

3

four possible answers: ‘never or hardly ever’, ‘some lessons’, ‘most lessons’, or ‘all lessons’.

Categorical Principal Components Analysis (CATPCA) in both in cross-national data and

country data show factor loadings above 0.72 for all questions except for ‘wait for quiet’. The

factor loading of ‘wait for quiet’ is in most countries near 0.7 but in Greece even 0.4. (see

appendix 1). Consequently, we created for our cross-national analysis a latent variable for

classroom disruption that contains all possible questions from PISA except ‘wait for quiet’ for

each student.

We also computed the mean classroom disruption using the disruption score of all

students in the school as a second dependent variable. Although the answers of the students refer

to the experience of the students and the interpretation of the question, we will refer in this paper

to individual disruption and class disruption to make the text more readable.

The third dependent variable math performance is a score on the math test developed by

PISA. To measure school skills accurately would make the test too long to be feasible. Hence,

PISA created a large number of very similar but shorter tests. Because such different tests can

never offer exactly the same degree of difficulty, Item Response Modeling (IRM) was used to

achieve comparable results between students who took different tests. We computed our

regressions for every plausible value and averaged the parameter estimates in order to take into

account the variance between these five plausible values. The skills scores were standardized for

the Organization for Economic Co-operation and Development (OECD) countries using an

average of 500 and a standard deviation of 100.

Table 1 gives the descriptive statistics of the variables we use in our analyses.

Individual level

Family forms. A disadvantage of PISA is that it lacks information about the cause of single

parenthood or guardianship of one the parents. Although we assume that in most OECD countries

divorce or separation is the most common reason for single parenthood of parents of 15-year old

students, there might be other reasons for growing up in a single-parent family (with or without a

guardian), i.e., birth out of wedlock without a following marriage or cohabitation, and death of

one of the parents. However, parents of 15-year old students are generally still too young to die,

and the number of people who (intentionally or unintentionally) become a single parent already

prior to childbirth will be rather low.

An important advantage of the measurement of family form in PISA is that students were

asked with whom they regularly live at home, and they were offered a number of possible

persons, whom they could all tick. This way, the real family form in the eye of the students is

measured instead of the formal situation, as reported by interested parents or authorities. Parents

who separated after cohabitation (instead of marriage) before the child reaches the age of 15 are

measured in the same way as formally divorced parents. Since separation after cohabitation has

more or less the same effect on children as compared to divorce after marriage (Dronkers &

Härkönen 2008; Härkönen & Dronkers 2006), the PISA data provide a more accurate picture in

countries where cohabitation with children is common. Married parents, who stopped living

together before the 15-year old student participates in the PISA survey, are also treated in the

same way as formally divorced parents. This feature is especially relevant for catholic countries

like Italy, Ireland, Portugal and Spain, where a formal divorce is still difficult to obtain. A

disadvantage is that some children may live without a parent temporarily (e.g. fishermen, fathers

working in the origin country). We believe, however, that this risk is small, as some students still

2 Only Greece shows a factor loading of 0.6 for ‘long time to start’.

4

will indicate that they live with both parents usually. We created a dummy variable indicating a

single-mother family.

Parental ESCS. The ESCS index of the parents is a composite index created within the

PISA dataset of the parents’ occupational status, measured with the International Socio-economic

Index of Occupational Status (ISEI) scale (Ganzeboom et al, 1992), the educational level of the

parents, measured with the ISCED (International Standard Classification of Education)

classification (UNESCO, 2006), and the presence of any material or cultural resources at the

students’ homes.

Female. We computed a dichotomous variable to classify gender. Boys are the reference

group.

Higher track refers to the track levels 2A and 3A of the International Standard

Classification of Education (ISCED). The 2A and 3A programmes ultimately lead to tertiary

education (OECD, 1999). This control variable takes into account the possible early selection of

children of single parents into a lower educational level, as a consequence of lower earlier

performance. The result of controlling for educational level might be that the relationship

between family form and school’s percentage of singleparent families is underestimated.

However, we prefer this risk of underestimation above a

too easy confirmation of our hypotheses. We include the dummy “track missing”, representing

1.1 percent of the students. Other tracks are the reference category.

School level

All these school characteristics are computed with all deleted pupils included.

The percentage pupils from single-parent families per school (either father or mother).

The mean ESCS per school was calculated using the ESCS score of all students in the

school.

Percentage of females. We computed the percentage of females using the number of

female students in the schools.

Percentage of immigrants was calculated using the number of immigrant students in

schools.

Results

All analyses are multi-level analysis with three levels: students, school and countries.

Individual classroom disruption

Table 2 shows the analysis with individual classroom disruption as dependent variable.

The equation of model 1 contains only a few variables: living in a single mother instead

of a two-parent family, the socio-economic background of the student (ESCS) and the school

ESCS. The higher the individual ESCS and school ESCS the lower the classroom disruption

perceived by the pupil. But students who live in a single mother family perceive more often

classroom disruption.

In model 2 we add other school- and individual characteristics to the equation of model 1,

especially the school percentage single parents. This addition hardly changes the results for living

in a single mother family: they perceive more class disruption. But interestingly we also find that

the higher the school percentage single parents, the higher the level of perceived classroom

disruption. This suggests that a higher percentage of single-parent families in a school increases

the individual perception of classroom disruption, even controlled for other relevant individual

and school characteristics.

5

We test this result with model 3, in which we includes the mean school level of classroom

disruption. As we might expect, we find that the higher the mean level of classroom disruption of

a school, the higher the level of classroom disruption as perceived by the student. But the

significant effect of the school percentage single parents has become insignificant by this

inclusion of the school level of classroom disruption. Even this addition hardly changes the

results for living in a single mother family: they perceive more class disruption, irrespectively of

the school level of classroom disruption or school percentage of single parents. We draw as a

conclusion that a higher percentage single-parent family per school increases the school level of

class disruption. In an equation, not shown in table 2, we added the interaction term ‘single

mother family * level of school classroom disruption’, but the parameter of that interaction term

is not significant.

Table 3 shows the same outcome at a school level analysis. The higher the school

percentage of single parents, the higher the school level of schoolroom disruption, irrespective

the ESCS school composition and the school percentages migrants or girls.

Math score

We analyse in table 4 whether classroom disruption can explain the negative effect of school

percentage single parents on educational performance.

Model 1 of table 4 shows the well-known outcome that the school percentage single

parents has a negative effect on the educational performance of all pupils in that school,

irrespective whether they live in a single mother family or in a two-parent family. Living in a

single mother family has no significant effect on performance and the same holds for the

interaction between single mother family and school percentage single-parent families. With

other words it is more the single parent context, which influences all students, instead of the

individual situations within the separate single mother families.

We include in model 2 the individual classroom disruption as perceived by the student.

That variable has the expected negative effect on the educational performance, but the significant

effect of school percentage single-parent families or the insignificant effects of single mother

family and the interaction-term hardly change by this addition.

The inclusion of the school level of class disruption in model 3 lowers the effect of school

percentage single-parent family, without making it insignificant. This means that a part of the

effect of percentage single-parent family on math score is indirect, mediated by classroom

disruption. We estimated a cross-level indirect path of percentage single-parent family to math

performance through classroom disruption, using the Structural Equation Model software package Mplus

(Muthen and Muthen, 2012). We estimated a Structural Equation Model that is a combination of table 3

and model 3 of table 4 without cross-level interaction terms. Figure 1 shows parameters from this

Structural Equation Modelwhich are comparable to those in table 3 and table 4. The indirect relationship

that the proportion of single parents have with math performance through disorder have a parameter

estimate of -0.055 and is significant at p<0,05. In model 4 we add the interaction between individual perception of classroom disruption

with the school level of classroom disruption. Interestingly, the opposite parameters of individual

perception variable and the interaction term have more or less the same strength and thus

neutralize each other. We interpret the positive effect of the interaction term in the following

way: pupils in schools with high levels of classroom disruption and who also experience a high

level of disruption are better able to avoid the most negative consequences of the high level of

classroom disruption. However the school level of classroom disruption has a strong negative

effect on educational performance, overriding all other effects. This does not mean that there is a

6

change in the negative effect of school percentage single-parent families: it continues to have a

negative and significant effect on educational performance (although it is substantially smaller

than in model 1).

Conclusion

Given our results, we can answer our two research questions positively.

First, the school’s composition of students from single-parent families affects the

conditions of teaching and learning in schools, in our case indicated by classroom disruption. We

also show that this negative effect cannot be fully explained by the individual situation of the

single mother family, because the school percentage single-parent families has still a significant

effect on classroom disruption, also after controlling for living with a single mother.

Second, these lower conditions of teaching and learning (as measured by the level of

school classroom disruption) mediate the negative effect of the school’s composition of students

from single-parent families on individual educational performance of all students. This mediation

is only partly, because the school level of classroom disruption cannot explain the full effect of

school percentage single-parent families: there remains a significant negative effect of school

percentage single-parent families on individual educational performance of all students. This

remaining effect gives support to the social capital explanation: schools with a large

concentration of children from single-parent families are usually characterized by less social

capital (i.e. indicated by parents’ social relations and networks with other parents), as proposed

by Pong (1997) and Sun (1999). Another explanation of this remaining effect might be that

classroom disruption does not fully measure the quality of the conditions of teaching and learning

in schools, as influenced by the school percentage single-parent families.

Pupils living in a single mother family have not a significantly lower educational

performance if we control for the school percentage single-parent families. This means that the

different context (in this case school composition) of the single-mother pupil is a more important

explanation of low educational performance than the ling in a single mother family. However,

living in a single mother family increases the individual level of classroom disruption and also

the chance to attend a school with higher percentages of single-parent families. These factors in

their turn influence negatively educational performance of pupils, irrespectively whether they live

in a single mother family or in a two-parent family.

The analyses of these contexts of single-parenthood are also important because that will

learn us more the processes which bring about the relative strong variation in negative outcomes

of various family forms (Pong, Dronkers & Hampden-Thompson, 2003).

But this analysis is the first step to unravel this interesting and important effect of school

percentage single-parent families, which is a modern form of a negative school composition,

independent of the socio-economic school composition. It is important to analyze this effect

further, because it illustrates that single-parenthood is not only a divorce or separation decision,

taken by two individual partners; it is an decision which effect also the life chances of other

children, and thus of their society.

Literature De Lange, M.. Dronkers, J. & Wolbers, M. (2014). Single-Parent Family Forms and Children’s

Educational Performance in a Cross-Comparative Perspective: Effects of School’s Share

of Single-Parent Families. School Effectiveness and School Improvement 25 (3): 329-350.

Dronkers, J. (2010). "Features of educational Systems as Factors in the Creation of Unequal

Educational Outcomes." pp. 299-328 in Quality and Inequality of Education. Cross-

7

National Perspectives, edited by J. Dronkers. Dordrecht /Heidelberg/ London/New York:

Springer

Dronkers, J. & Harkonen, J. (2008). "The intergenerational transmission of divorce in cross-

national perspective: Results from the Fertility and Family Surveys." Population Studies-

A Journal of Demography 62:273-288.

Dronkers, J. & Kalmijn, M. (2013). Single-parenthood among migrant children: Determinants

and consequences for educational performance. CReAM Discusion Paper Series 09/13

Garriga, A. (2010). Consequences of parental divorce and family structure on children's

outcomes in European societies: individual, cohort and country explanations. PhD

dissertation, University Pompeu Fabra (Barcelona).

Härkönen, J. & Dronkers, J. (2006). "Stability and Change in the Educational Gradient of

Divorce. A Comparison of Seventeen Countries." European Sociological Review 22:501 -

517.

Muthén, L.K. and Muthén, B.O. (1998-2012). Mplus User’s Guide. Seventh Edition. Los

Angeles, CA: Muthén & Muthén

Organization for Economic Co-operation and Development. (1999). Classifying Educational

Programmes. Manuel for ISCED-97 Implementation in OECD countries. Paris: OECD

Publishing

Organization for Economic Co-operation and Development (2012). PISA 2009 Technical Report,

PISA. Paris: OECD Publishing. doi: 10.1787/9789264167872-en

Pong, S.-L. (1997). Family Structure, School Context, and Eighth-Grade Math and Reading

Achievement. Journal of Marriage and the Family, 59(3), 734-746.

Pong, S.-L. (1998). The school compositional effect of single parenthood on 10th-grade

achievement. Sociology of Education, 71(1), 24-43.

Pong, S.-L, J. Dronkers en G. Hampden-Thompson, (2003). "Family Policies and Children's

School Achievement in Single- Versus Two-Parent Families." Journal of Marriage and

Family 65:681-699.

Sun, Y. (1999). The Contextual Effects of Community Social Capital on Academic Performance.

Social Science Research, 28(4), 403–426.

8

Figure 1: Direct and indirect effects of percentage single parents family per school on math

performance of pupils.

9

Table 1: Descriptive Statistics

N Minimum Maximum Mean Std. Deviation

Individual classroom Disruption 181916 -1,12 3,26 -,01 ,99

Student ESCS 183982 -3,82 3,41 ,16 ,91

School ESCS 183982 -1,90 1,72 ,11 ,51

School % single parents* 183982 ,00 75,00 11,93 7,69

Mean school classroom disruption 183982 -1,02 1,91 -,01 ,39

Math score 183982 20,96 864,32 511,34 84,65

Single parent 183982 ,00 1,00 ,11 ,32

Female 183982 ,00 1,00 ,50 ,50

% school migrants* 183982 ,00 96,77 8,23 11,93

% school Female* 183982 ,00 100,00 50,14 18,19

Higher track 183982 ,00 1,00 ,77 ,42

Track missing 183982 ,00 1,00 ,01 ,10

*Mean centered at country-level in analysis

10

Table 2: The effects of individual characteristics and the school compositions on individual classroom disruption

Model 1 Model 2 Model 3

B S.E. B S.E. B S.E.

Constant 0.028 0.031 0.148 0.033 0.045 0.006

Single mother 0.058 0.007 0.054 0.007 0.053 0.007

Individual ESCS -0.016 0.003 -0.019 0.003 -0.020 0.003

School ESCS -0.175 0.010 -0.133 0.010 0.023 0.005

School % single parent 0.005 0.001 -0.001 0.000

Female -0.105 0.005 -0.105 0.005

Higher track -0.102 0.014 -0.008 0.005

Track missing -0.035 0.037 0.007 0.021

% migrants 0.001 0.000 0.001 0.000

% Female -0.003 0.000 0.001 0.000

Mean school classroom disruption 1.009 0.006

Variance

Country 0.025 0.007 0.026 0.007 0.000 0.000

School 0.105 0.002 0.098 0.002 0.000 0.000

Student 0.864 0.003 0.861 0.003 0.831 0.003

-2*loglikelihood 500.001.413 499.081.274 482.684.675

Source: own computation of PISA wave 2009; n-countries = 28; n-schools = 8001; n-students = 181916

11

Table 3: The effects of the school compositions on school classroom disruption (school level

analysis)

B S.E.

Constant 0.031 0.030

School percentage single parents 0.005 0.001

School ESCS -0.165 0.009

Percentage school migrants -0.000 0.000

Percentage school female -0.004 0.000

Variance

Country level 0.025 0.007

School level 0.132 0.002

-2*loglikelihood: 6.589.861

Source: own computation of PISA wave 2009; n-countries = 28; n-schools = 8001.

12

Table 4: The effects of the school composition and classroom disorder on math score.

Model 1 Model 2 Model 3 Model 4

B S.E. B S.E. B S.E. B S.E.

Constant 480.061 3.648 481.332 3.575 482.021 3.472 481.046 3.462

Single mother -0.741 0.559 -0.372 0.558 -0.363 0.558 -0.323 0.558

School percentage single parent -0.465 0.059 -0.440 0.058 -0.371 0.058 -0.365 0.058

Single mother* school % single parent -0.104 0.061 -0.096 0.061 -0.101 0.061 -0.101 0.061

Individual ESCS 18.330 0.201 18.163 0.201 18.172 0.201 18.128 0.201

Mean school ESCS 53.753 0.995 52.710 0.985 50.601 0.995 50.374 0.996

Female -17.466 0.323 -18.229 0.323 -18.200 0.323 -18.366 0.323

Higher track 36.423 1.127 35.986 1.130 35.551 1.127 35.519 1.127

Track missing 20.029 3.091 20.731 3.081 20.376 3.071 20.428 3.069

Percentage migrants -0.130 0.033 -0.121 0.033 -0.123 0.033 -0.125 0.033

Percentage Female 0.149 0.024 0.127 0.024 0.073 0.024 0.072 0.024

Individual Disruption -5.770 0.163 -5.516 0.165 -6.682 0.177

Mean school disruption -13.196 1.181 -14.052 1.182

Individual Disruption*Mean disruption 7.310 0.405

Variance

Country 341.637 93.425 326.835 89.424 306.648 83.994 304.525 82.631

School 1.268.552 23.412 1.234.229 22.866 1.211.858 22.509 1.213.275 22.521

Student 4.117.288 13.878 4.060.795 13.768 4.060.841 13.768 4.053.118 13.742

-2*loglikelihood: 2.069.659.769 2.043.907.270 2.043.783.271 2.043.457.666

Source: own computation of PISA wave 2009; n-countries = 28; n-schools = 8001; n-students = 181916

13

Appendix 1: Factor structure of classroom disruption per country

Australia:

Dimension

1 2

Lessons - Students don’t listen ,848 ,321

Lessons - Noise and disorder ,837 ,389

Lessons - Cannot work well ,860 -,349

Lessons - Long time to start ,860 -,346

Austria

Dimension

1 2





Belgium

Dimension

1 2

Lessons - Students don’t listen ,831 -,387

Lessons - Noise and disorder ,840 -,317

Lessons - Cannot work well ,815 ,336

Lessons - Long time to start ,811 ,388

Canada

Dimension

1 2





Switzerland

Dimension

1 2





Czech Republic

Dimension

1 2



14



Germany

Dimension

1 2





Denmark

Dimension

1 2





Spain

Dimension

1 2





Estonia

Dimension

1 2





Finland

Dimension

1 2





France

Dimension

1 2


15




United Kingdom

Dimension

1 2





Greece

Dimension

1 2





Ireland

Dimension

1 2





Iceland

Dimension

1 2





Italy

Dimension

1 2





Luxembourg

Dimension

1 2

16





Netherlands

Dimension

1 2





Norway

Dimension

1 2





New Zealand

Dimension

1 2





Poland

Dimension

1 2





Portugal

Dimension

1 2





Slovak Republic

Dimension

17

1 2





Slovenia

Dimension

1 2





Sweden

Dimension

1 2





USA

Dimension

1 2





teaching and learning conditions in schools with a high proportion of students from single- parent...

Documents