does divorce or remarriage have the greater negative impact on the academic achievement of children?

This article was downloaded by: [University of Utah]On: 11 October 2014, At: 18:46Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number:1072954 Registered office: Mortimer House, 37-41 Mortimer Street,London W1T 3JH, UK

Journal of Divorce &RemarriagePublication details, including instructionsfor authors and subscription information:http://www.tandfonline.com/loi/wjdr20

Does Divorce orRemarriage Have theGreater NegativeImpact on the AcademicAchievement of Children?William H. Jeynes PhD aa Department of Education , University ofChicago , USAPublished online: 12 Oct 2008.

To cite this article: William H. Jeynes PhD (1998) Does Divorce or RemarriageHave the Greater Negative Impact on the Academic Achievement ofChildren?, Journal of Divorce & Remarriage, 29:1-2, 79-101, DOI: 10.1300/J087v29n01_05

To link to this article: http://dx.doi.org/10.1300/J087v29n01_05

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy ofall the information (the “Content”) contained in the publicationson our platform. However, Taylor & Francis, our agents, and ourlicensors make no representations or warranties whatsoever as to the

http://www.tandfonline.com/loi/wjdr20

http://www.tandfonline.com/action/showCitFormats?doi=10.1300/J087v29n01_05

http://www.tandfonline.com/action/showCitFormats?doi=10.1300/J087v29n01_05

http://dx.doi.org/10.1300/J087v29n01_05

accuracy, completeness, or suitability for any purpose of the Content.Any opinions and views expressed in this publication are the opinionsand views of the authors, and are not the views of or endorsed byTaylor & Francis. The accuracy of the Content should not be reliedupon and should be independently verified with primary sources ofinformation. Taylor and Francis shall not be liable for any losses,actions, claims, proceedings, demands, costs, expenses, damages,and other liabilities whatsoever or howsoever caused arising directlyor indirectly in connection with, in relation to or arising out of the useof the Content.

This article may be used for research, teaching, and private studypurposes. Any substantial or systematic reproduction, redistribution,reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of accessand use can be found at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

http://www.tandfonline.com/page/terms-and-conditions

http://www.tandfonline.com/page/terms-and-conditions

Does Divorce or Remarriage Have the Greater Negative Impact

on the Academic Achievement of Children? William H. Jeynes

ABSTRACT. Over the past twenty-five years considerably more research has been done on the effects of divorce on academic achievc- ment than on the effects of remarriage following divorce on academic achievement. Much of this fact results from thc assumption of many social scientists that remarriagc must, on average, benefit children of divorce. Recent studies examining the effects of remarriage on various dimensions of childhood ability and psychology seriously question this assumption (Kurdek et al., 1994; Hetherington, 1992; Bay- dar, 1988; Amato and Ochitree;l987; Hetherington, Nagan, and Anderson, 1989). In fact, some researchers have even argued that the effects of remarriage following divorce may rival or even exceed the effects of divorce (Baydar, 1988; Amato and Ochitree, 1987; Hey- man, 1992). Using the NELS data set from the 1988-1992 period, the rcsults of this study indicate that how one views the relative effects of divorce and remarriage may depend largely on whether one controls for socioeconomic status (SES). [Article copies availuble for a fee from The Haworth Docrunent Delivery Service: 1-800-342-9678. E-moil address: getir~fo@ha worthpressirtc.con~]

During the last thirty-five years rates of both divorce and remarriage in the United States have soared. Demographers estimate that

William H. Jeynes, PhD, is on the faculty of the Department of Education, University of Chicago.

Address correspondence to: William H. Jeynes, 1805 W. Golf Road #107, Mt. Prospect, IL 60056 or ([email protected]).

Journal of Divorce & Remarriage, Vol. 29 (112) 1998 O 1998 by The Haworth Press, Inc. All rights reserved. 79

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

80 JOURNAL OF DIVORCE & REMARRIAGE

the odds of a divorce occurring in a household before the children become grown rest at about 50-50% (Ahlburg and DeVita, 1992). The vast majority of those adults who divorce will also remarry (Cherlin, 1992; Bumpass, 1984).

Over the last twenty-five years social scientists have repeatedly documented the negative influence that divorce has on academic achievement (Wallerstein, Corbin, and Lewis, 1988; McLanahan and Sandefur, 1994; Jeynes, 1996; Shim, 1978; Hetherington, 1992; Cher- lin, 1992). Research on stepfamilies has lagged considerably behind that of divorce and is still a young area of study (Ganong and Cole- man, 1994; Booth and D m , 1994; Jeynes, 1997; Furstenberg, 1988; Heyman, 1992). Few studies on stepfamilies have been done (Ganong and Coleman, 1994; Heyman, 1992). The primary reason for this fact has been that many researchers and Americans, as a whole, believed that parental remarriage generally benefited children. First, remarriage introduces an additional caregiver who can help raise the child. The assumption was that this additional caregiver would especially benefit children in a divorced home, if the stepparent was the same gender as the child. Second, remarriage generally substantially raises the socioeconomic status (SES) level of a family and this should benefit any children involved.

The assumption that remarriage benefits children of divorce was so strong that many studies during the past twenty-five years determined the effects for divorce on academic achievement by compar- ing the academic achievement of children of divorce from single- parent families to that of children from both intact families and families that had been reconstituted following divorce combined. Unfortunately, because of the structure of their research design, these studies can give us no information about the effects of remarriage following divorce on academic achievement (Milne, 1989; Heyman, 1992). In addition, until about twelve years ago, most studies that did distinguish children of divorce from reconstituted families from other family structures did not use a random sample of a definable population. Ganong and Coleman (1984) evaluated thirty-eight studies specifically dealing with the effects of remarriage on children. Of these thirty-eight studies, only six used a random sample of a definable population.

As a consequence of all these limitations in the research, little is

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

William H. Jeynes 81

known about the effect of remarriage on the academic achievement and emotional well-being of children (Ganong and Coleman, 1994 ; Furstenberg, 1988). Although there exists considerable consensus among family theorists regarding the negative effects of divorce upon children both psychologically and in terms of academic achievement, the same consensus does not exist regarding the effects of remarriage. A large degree of this lack of agreement stems from these problems in the literature on stepfamilies.

In recent years, however, social scientists have accumulated a sizable amount of evidence indicating that remarriage may have ill-effects on many children (Zill, 1994; Popenoe, 1994; Zill and Nord, 1994; Hetherington and Jodl, 1994). Many children demonstrate reluctance in accepting a new parental figure (Kelly, 1992; Visher and Visher, 1988). Children in reconstituted families often struggle with rivalries with their stepbrothers and step- sisters, as w e l l a s jealous feelings toward their new stepparent (Anderson and Rice, 1992). Stepchildren often feel that the stepparent monopolizes the time and energy of the natural parent (Kelly, 1992; Walsh, 1992; Amato, 1987). The presence of a stepparent often reduces the closeness of the relationship that the children have with the biological parent (Hetherington, 1994; Walsh, 1992). In addition, children often suffer because remarriage often produces an increased tension between the biological parents (Walsh, 1992).

All in all, the research indicates that children in reconstituted families may suffer because they often live in an unstable environ- ment of increased mobility, adapting to difficult new roles and relationships (McLanahan and Sandefur, 1994; Fast and Cain, 1988; Hetherington and Jodl, 1994) and frequently unstable remarriages (Booth and Edwards, 1992; Popenoe, 1994). As a result, many children from reconstituted homes become frustrated and Show a greater tendency to be aggressive, anxious, and unhappy than children from intact families (NuM, Parish, and Worthing, 1993; Wallerstein and Kelly, 1980). Given these tendencies, it is not surprising that Duber- man's (1975) often cited study, on the reconstituted family, found that remarried parents rate "childrearing" as the number one problem of living in a reconstituted family. There may exist some moder- ating influences on the challenges that stepchildren face such as the

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4


age of the child, the length of time that the child has lived in a reconstituted home, and the extent of the communication between the stepchild and the stepparent (Anderson, Linder, and Bennion, 1992; Walsh, 1992; White, 1994).

Some researchers even suggest that it may be remarriage following divorce, more than divorce itself, that most negatively affects children (Baydar, 1988; Amato and Ochitree, 1987). Other research, such as that undertaken by E. Mavis Hetherington, indicates that probably both divorce and remarriage following divorce have a negative impact on children (Hetherington, 1992; Hetherington and Jodl, 1994; Kurdek et al., 1994; Hetherington, Hagan, and Ander- son, 1989). Finally, some researchers insist that the effects of remarriage following divorce are fairly mild overall compared to those of divorce (McLanahan and Sandefur, 1994; Zill and Nord, 1994). Which one of these three possibilities is correct? While it may take years for the social scientists to address this question in the broadest sense, this study will attempt to answer this question as it pertains to the academic achievement of children of divorce from single-parent and reconstituted families.

One of the primary ways the studies cited in the preceding paragraph differ from one another is the extent to which they present their results in a way that considers the effects of SES or whether a basic data summary is given. For example, in the Amato and Ochitree (1987) study, the researchers control for SES, and present their results with that in mind. Zill and Nord (1994), on the other hand, focus on presenting basic data summaries. While Zill and Nord do recognize the influence of SES, their conclusions are founded on the basic data that resulted from their research. With this in mind, it may be the case that the results of past research on the relative effects on divorce and remarriage may be more consistent than it appears. The different results that emerge may be a result of the extent to which the effects of SES are taken into consideration.

This study tests the prediction that whether it appears that divorce or remarriage has the primary negative impact on academic achievement depends on whether one controls for SES.

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

William H. Jeyues

METHODS AND DATA SOURCES

Sample

The sample that this study draws from includes students who participated in the National Education Longitudinal Survey (NELS) for the years 1988,1990, and 1992. The NELS research project was sponsored by the U.S. Department of Education's National Center for Statistics. The National Opinion Research Center (NORC) and NORC subcontractors designed the study. The Educational Testing Service (ETS) designed the achievement tests used in this study.

In 1988 using a "two stage probability design," the researchers selected a nationally representative sample of schools and eighth grade students (U.S. Department of Education, 1992); 24,599 students from 1,052 schools participated in this study. Questionnaires, regarding a vast array of topics, were given to students, parents, and teachers. Achievement tests in mathematics, reading, science, and social studies (history, civics, and geography) were also given to the students. These tests were "curriculum based cognitive tests that used item overlapping methods to measure" academic achievement (U.S. Department of Education, 1992).

The goal of the NELS researchers was to followup on these eighth graders in two-year intervals in 1990 and 1992. The first follow-up in 1990 surveyed not only those students still in school, but those who had dropped out since the time of the "base year" (1988) study. To compensate for the decline in sample size from 1988 due to students who had transferred out of the schools in the sample and for other reasons, this "core" group was then added to through a process called "freshening." This process added 1,043 eligible students to the total sample size. Achievement tests were again administered and data on the schools involved were again collected. This brought the taught sample size to 20,706. Achieve- ment tests were again administered and data on the schools involved were again collected.

In this particular study the analysis undertaken for the follow-up study (for tenth grade students in 1990) was limited to the 17,874 students who participated in bolh the base year (1988) and the first follow-up studies. This was due to the fact that no parent questionnaire was included in the NELS first follow-up study; and therefore

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4


the marital status of the parents could only be determined for those students who participated in the 1988 base year study.

A similar two stage sampling technique was used in choosing the student sample for the 1992 follow-up study. The "core group" consisted of those students who were in the tenth grade sample in the first follow-up study. An additional 243 eligible twelfth graders were added to the "core group7' to produce a total sample size of 18,726. Achievement tests were again administered.

Because of the administration of a parent questionnaire for the second follow-up 1992 study, all students who participated in the

'

second follow-up were included in the analysis.

Family Structure Variables

Family structure variables were based on using the NELS variables for marital status, family composition, and determining whether a parent was dead or alive.

In determining the family structure variables, any responses that were contradictory (eg., responses that indicated the natural parents were married and that the natural father was dead) were deleted and included in a variable called "delete" to measure the effect for the observations that were deleted in various analyses. I will elaborate on this "delete" variable later in this section. Similarly, a variable "miss" was designed to account for the effect of missing data for the family structure variable. The primary variables used in this analysis were:

Divorce. Any child whose natural parents had been divorced, whether the custodial parent had since remarried or not had the value 1 for this variable. All other children had the value 0 for this variable.

Remarried. In these regression analyses which examine only the effects of divorce and any subsequent remarriage, the remarriage variable only includes the effect of remarriage following divorce. The dummy variable was coded as 1 when remarriage had taken place. Other children were coded 0 for this variable.

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

Academic Achievement Variables

Grade Point Average. Average of the students' self-reported grades covering four subjects (English, math, social studies, and science). For the 1988 data set the following numerical values were assigned each letter grade. "A" = 4; "B" = 3; "C" = 2; "D" = 1; and mostly below "D" = .5. The mean was taken of all the "nonmiss- ing values of these four variables" and was "equally weighted."

In the 1992 data set GPA is calculated rather differently. For this data set grades were tabulated on a 0 to 100 basis. Credit was also given for advanced placement courses. As a result, the actual range of scores was 0 to 108.98. The GPA data for 1992, unlike the 1988 figures, were not on a 4.0 scale.

Standardized Tests. Standardized test scores were obtained using tests developed by ETS. IRT scores (Item Response Theory scores) were obtained for the Reading Comprehension Test, the Mathemat- ics Comprehension Test, the Social Studies (History/Citizenship/ Geography) Comprehension Test, the Science Comprehension Test, and the Test Composite (Reading and Math test results combined).

Left Back. Coded with the value 1 if a student had been held back a grade at any time in his life. Other students were coded with the value 0.

Basic Core. Coded with the value 1 if a twelfth grade student had taken the basic core courses identified by the NAEP (National Association of Educational Progress). This basics program consisted of: 4 years of English courses; 3 years of Social Studies courses; 3 years of Science courses; 3 years of Math courses; 2 years of Foreign Language courses; and 112 year of Computer Science courses. Other students were coded with the value 0.

OTHER INDEPENDENT VAROiBLES

Variables Involving Socioeconomic Status

The socioeconomic status of a child's family was determined "using parent questionnaire data, when available." Five components composed the socioeconomic status variables: (a) father's

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4


level of education; (b) mother's level of education; (c) father's occupation; (d) mother's occupation; and (e) family income.

The occupational data were recoded using the Duncan SEI (So- cioeconomic Index) scale, which was also used in the High School and Beyond Survey. If any of the components were missing from the parent questionnaire, equivalent or related questions were used from the student questionnaire to determine SES. Three coded SES variables were used.

Variables Involving Race

Asian-coded with the value 1 for an Asian student. Other students were coded with the value 0.

Hispanic-coded with the value 1 for a Hispanic student. Other students were coded with the value 0.

Black-coded with the value 1 for a Black student. Other students were coded with the value 0.

American IndianlAlaskan Native-coded for a Native American! Alaskan Native student. Other students were coded 0.

Other Variables

Gender-coded with the value 1 for a female. Males were coded with the value 0 for a male.

Deleted. This includes any student's family structure that was not included in a particular analysis either because: (1) it was not being studied in a particular analysis or (2) because the student's answer was contradictory in some way, e.g., the student answered that his parents were presently married, but one of them was dead. Under either of these circumstances this variable was coded with the value 1. Otherwise, this variable was coded with the value 0.

Missing Data Variables. A missing data variable was included for family structure as a means of ascertaining the effect of students whose answers were missing, so as to make the determination of the student's family structure impossible. Generally, more than one NELS question was needed to ascertain a student's family structure. If a student's answer for any one of these questions was not answered, the family structure could not be determined and the family structure was coded as "missing."

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

William H. Jey tm 87

Missing data variables were also used for other student attributes such as race, if less than all or virtually all of the students answered the relevant question(s). Missing variables were coded with the value 1 when this information was missing; and were coded with the value 0 otherwise.

MODELS

Two models were used in analyzing the effect of divorce, remarriage, and other family variables on academic achievement. Al- though there were variations of these two models used at various points in the analyses, these two models set the framework for all the analyses undertaken in this study. Both models and their variations included variables for family structure as well as for both missing and deleted data. I will refer to the variables just listed as the central variables. The models and their variations differ in terms of whether they included variables for SES, race, and gender. (1) Both variations of the Full Model include variables for.race and gender, whereas neither variation of the Basic Model includes these variables. (2) Both the Basic and Full Model have two variations. One variation of each model includes variables for SES, while the other does not.

Model 1: The Basic Model

The two variations of the Basic Model include the central variables, but do not include variables for race or gender. The Basic Model takes two forms:

1 . Basic Model. This variation of the Basic Model includes only the central variables, i.e., it does not include variables for SES, race, or gender.

2. Basic Model wifh SES Variables Included. This variation of the BasicModel includes the central variables plus the SES variables. The use of the Basic Model is important for the following reasons:

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

JOURN,4L OF DIVORCE d REMARRIAGE

a. to obtain overall effects for divorce and remarriage. b. to use models that are the most commonly used by family

researchers so that the results of this study can be understood in reference to the work that other family researchers have one. If controls are used in a family research study, race, sex, and SES are the most common ones used.

Model 2: The Full Model

The two variations of the Full Model include the central variables and, unlike the Basic Model, include variables for race and gender.

1 . Full Model. This variation of the Full Model includes the central variables and variables for race and gender, but does not include variables for SES.

2. Full Model with SES. This variation of the Full Model includes the central variables and variables for race, gender, and SES.

Limitations

There arise a number of limitations common to examining large data sets, such as NELS, and virtually any study examining the effects of divorce. While none of these limitations are particularly serious, they should be noted.

As in almost any study, there are certain relevant characteristics of the sample that one would like to know about, but are not acces- sible. In this study the most important of these characteristics of the sample involves those students who dropped out of the study. A fair number of students dropped out of the NELS sample at some point during the 1988-1992 period for various reasons that included moving to another school or moving to a different location, etc. Past studies that have examined how such dropouts from a study differ from the remainder of a sample indicate that (1) this group of students generally performs less well academically than the remainder of a given sample and (2) that this group of students has a disproportionately high percentage of students coming from divorced or reconstituted homes. These facts combined will probably

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

cause the effects for divorce to be lower in this paper than they otherwise would be, if we had access to the groupof students that dropped out of this study. This is probably the most important limitation of this study.

It would also be nice to have reliable measures of parental conflict, the age of the child at the time of divorce, and the length of time since the divorce. The NELS study did not make this information available. Reliable measures of parental conflict are difficult to obtain, because of the subjective nature of such conflict. But information on the length of time since the divorce would be particularly interesting, because research indicates that the effects of divorce on academic achievement may or may not slowly decrease with time after the third or fourth year following the divorce (Wallerstein and Blakeslee, 1989). It would also be advan- tageous for every study on divorce to have more information about the pre-divorce families. Research indicates that some of the effect of divorce occurs prior to the actual act of divorce (Neighbors, Forehand, and ~ r m i s t e a d , 1992). Hence, some of thechildren in "intact" families may reside in families where separation or divorce is about to take place. This fact probably mildly biases the results of this study in the direction of understating the effects of divorce.

Also, for the purposes of this study the family income aspect of SES was defined in actual terms rather than in terms of Der capita income or living standards. Living standards were impossible to determine from the NELS questionnaire. Using a per capita income measure is not only difficult to determine, butthis measure fails to consider economies of scale. An ideal measure of family income as a component of SES might include a measure that considered the distribution of income among the household (Lazear and Michael, 1988). But even this model would have a hard time wrestling with the fact that many big ticket items are purchased with every family member in mind. For the benefits that arise from such items are not distributed equally. Given the limitations of the other measures, it would appear that actual family income is the preferable, albeit imperfect, measure for the family income component of SES.

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

JOURNAL OF DIVORCE & REMARRIAGE

RESULTS

Whether one included the control for SES had a notable influence on the relative effects that emerged for divorce and remarriage. Table 1 shows the effects for divorce and remarriage using the Basic Model both with and without the SES variables included for the eighth grade (1988) data set. When SES is not controlled for, it appears that it is divorce, not remarriage, that is exerting the main downward pressure on academic achievement. For all the academic measures, the regression coefficients for divorce indicated a negative relationship between divorce and academic achievement that was statistically significant at the .0001 level of probability. The absolute value of these regression coefficients ranged from .33 for

TABLE 1. Effects for Divorce and Remarriage in Standard Deviation Units for the Basic Family Structures Using the Basic Model with and without the Three SES Variables Added for Eighth Grade (1 988). N = 18,176

Effects for Divorce and Effects for Divorce and Remarriage Remarriage without SES variables with SES variables

Academic

measure

Divorce Remarriage Divorce Remarriage

Standardized tests

Math Reading Science

Social Studies Composite

Other measures GPA

Left backa

*p < .05; **p c .Ol; -p < ,001; -p < .0001 a, logistic regression analysis was used.

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

William H. Jeyrw 91

GPA to .15 for the Reading test. Although the direction of the regression coefficients for remarriage indicate some negative impact on academic attainment in all cases, except for GPA, only one of these regression coefficients is statistically significant. In the case of the Math standardized test, remarriage following divorce produced a regression coefficient of -.06 standard deviation units. This result was statistically significant at the .05 level of probability.

When we use the Basic Model with the SES variables added into the regression equation, the absolute value of effects for remarriage increase considerably for standardized test scores particularly. As a result, for five of the seven academic measures, it is remarriage, not divorce, that produces the statistically significant effects. The largest effect occurs for the Math test, which is statistically significant at the .0001 level of probability. The effects for all the other standardized tests are statistically significant as well, ranging in statistical significance from the ,001 to the .05 level of probability. For the two other academic measures divorce, not remarriage, produces the statistically significant results. For both GPA and being left back a grade even with the SES variables added, the regression coefficients remain statistically significant at the .0001 level of probability. Nevertheless, the inclusion of the SES variables in the analysis reduces the absolute value of the regression coefficients in each case. Overall, whether one controls for SES has a large bearing on whether it is divorce or remarriage that exerts the downward Dressure on academic achievement. When the Basic Model is used. when SES is not controlled for, it appears that divorce exerts a greater downward pressure on academic achievement than does remarriage. But when SES is controlled for, in most cases, it is remarriage that exerts the greater downward pressure on academic achievement.

When the Full Model (which includes variables for race and gender) is used, the influence of either including or excluding the SES variables becomes even more obvious. Table 2 shows the effects for divorce and remarriage using the Full Model both with and without the SES variables included for the eighth grade (1988) data set. Using this larger model that includes controlling for race (with the four race variables: Asian, Black, Hispanic, and American IndianIAlaska Native) and gender yielded similar values as were

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

92 .JOURNAL OF DIVORCE & REMARRIAGE

TABLE 2. Effects for Divorce and Remarriage in Standard Deviation Units for the Basic Family Structures Using the Standard Full Model with and without the Three SESVariables Addedfor Eighth Grade (1988). N = 18,176

Effects for Divorce and Effects for Divorce and Remarriage Remarriage without SES variables with SES variables

Math 1 -.1Inm - .09* + .02 -.13-

Academic

measure

Standardized tests

Reading I -.lorr** - .04 + .01 - .07*

Divorce Remarriage Divorce Remarriage

Science I - . lZnm - .06 - .01 - .09* Social Studies I - .llnm - .07* .OO -.lom‘

Composite - .07* + .02 - .llm*

Other measures GPA - .29- + .04 -.20m +.01

found in the last section regarding the effect of including or not including the SES variables. There was, however, one significant difference that we can observe by using the larger model. That is, even before we add the SES variables into the equation, the addition of the race and gender variables strengthens the effect of rernar- riage, while weakening the effect for divorce.

When SES is not controlled for, it appears that it is divorce, not

Left backa

remarriage, that is exerting the main downward pressure on academic achievement. For all the academic measures the regression coefficients for divorce indicated a negative relationship between divorce and academic achievement that was statistically significant at the .0001 level of probability. The absolute value of these regression coefficients ranged from .29 for GPA to .10 for the Reading

f.17" +.02 +.lZm* +.04

*p < .05; *p < .01; *p < ,001; -p < .0001 a, logistic regression analysis was used.

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4


test. As noted above, these regression coefficients for divorce were of a lower absolute value than those that emerged using the Basic Model. everth he less, when the Full Model without the SES variables is used we once again note that it is divorce, and not remarriage, that is exerting the primary downward pressure on academic achievement. While the Full Models produced smaller effects for divorce than did the Basic Model, it yielded larger effects for remarriage. When the SES variables were not included in the model, the Basic Model yielded only one statistically significant effect for remarriage (the Math test). But with the addition of the race and gender variables that distinguish the Full Model, the effects for the Composite test and the Social Studies test also become statistically significant. In addition, the effect for the Science test barely misses statistical significance with a p-value of .0515. These results indicate that whether one controls-for race and sex can have a measur- able impact on the size of the regression coefficient for remarriage. Controlling for race, it turns out, emerges as an important step in understanding the effect for remarriage. National siudies indicate that Whites have a much higher rate of remarriage after divorce than their Black and Hispanic counterparts (Chadwick and Heaton, 1992; McLanahan and Sandefur, 1994). This would mean that Whites tend to have a higher representation among remarrieds than one would otherwise expect. This fact should be considered when determining the effects of remarriage on academic achievement. Therefore, controlling for race is important.

When we use the Full Model with the SES variables added into the regression equation, the absolute value of effects for remarriage again increases considerably, especially for the standardized test scores. Once again, the result is that for the majority of the academic variables it is remarriage, and not divorce, that is exerting the primary downward pressure on academic achievement. For the standardized test scores the effects range from - .13 for the Math test (p c .0001) to - . 7 for the Reading test (p < .05). Throughout this study divorce and remarriage appear to impact the results of the Math test more negatively than they impact the Reading test. A possible reason for this is dealt with in the Discussion section. The effect for divorce also reached the .0001 level of statistical significance for the Composite test, .001 for the Social Studies test, and

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

94 .IOUXNAL OF DIVORCE & REMARRIAGE

.O1 for the Science test. Although remarriage exerted the primary downward pressure on academic achievement for most of the measures, divorce remained the primary agent exerting downward pressure on GPA and being left back a grade. The effects of divorce in each of these cases were statistically significant at the .0001 level of probability.

These results confirm those found using the Basic Model. That is, whether one controls for SES has a large bearing on whether it appears that it is divorce or remarriage that is exerting the primary downward pressure on academic achievement.

Although 1990 SES variables were not available for the tenth grade (1990) data set, one should note that the effects for divorce and remarriage, when the SES variables are not included, are very close to those yielded by the eighth grade (1988) data set. In most cases, the effects for divorce for a given measure were slightly stronger favoring children from intact families, with all of the effects statistically significant at the .0001 level of probability. This raises the possibility that the academic disadvantages associated with being a child of divorce may become larger over time. It should also be noted that although the effects of adding the SES variables could not be measured for the tenth grade (1990) data set, simply adding the race and gender variables increases the relative negativity of the remarriage regression coefficients.

The same general pattern found in the eighth grade (1988) and tenth grade (1990) data sets held for the twelfth grade (1992) data set as well. Table 3 indicates the results obtained using the Full Model both with and without the SES variables added for the twelfth grade (1992) data set. The NELS study did not differentiate between children whose parents were divorced and those who were separated for the twelfth grade students. Nevertheless, the pattern remained the same. Table 3 shows the effects for divorce/separation and remarriage using the Full Model without the SES variables included in comparison with the effects for the Full Model with the SES variables included. When the SES variables are not included in the analysis, divorce, not remarriage, emerges as the primary factor exerting a downward pressure on academic achievement. For all nine of the academic measures the regression coefficients for divorcelsep- aration were statistically significant. In fact, all but one of the aca-

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

TABLE 3. Effects for Divorce/Separation and Remarriage in Standard Devi- ation Units for the Basic Family Structures Using the Full Model with and without the Three SES Variables Added for Twelfth Grade (1 992). N = 13,002

Effects without the Effects with the SES variables included SES variables included

Academic

measure

Divorce/ Remarriage Divorce/ Remarriage

Separation Separation

Standardized tests

Math Reading Science Social Studies Composite

Other measures GPA

Left backa

Basic Corea

*p < .05; *p c .Ol; *p < ,001; -p < ,0001 a. logistic regression analysis was used.

demic measures yielded statistically significant results at the .0001 level of probability. The effect for Math was the largest at - .19. Once again, the effect for Math was larger than the effect for Read- ing. The effects for remarriage, on the other hand, produced statistically significant results for four of the nine academic measures.

When the Full Model is used including the SES variables the overall pattern of the regression coefficients changes considerably. Now it is remarriage that arises as the primary factor exerting downward pressure on academic achievement. All the regression coefficients for remarriage produce statistically significant results with the exception of GPA. The regression coefficients for the Math, Science and Composite tests are all statistically significant at

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4


the .0001 level of probability. For divorce/separation five of the eight regression coefficients are statistically significant. The regression coefficients for GPA and being left back a grade remain statistically significant for the twelfth grade (1992) data set just as they were for the eighth grade (1988) data set. Hence, the overall picture regarding the effects of divorce and remarriage appears extremely consistent. In addition, the regression coefficient for the basic core courses is also statistically significant. This academic measure is unique to the twelfth grade (1988) data set.

The results for the eighth grade (1988), tenth grade (1990), and twelfth grade (1992) grade sets show considerable consistency. While divorce appears to be the primary agent exerting a downward pressure on academic achievement when the SES variables are excluded from the analysis, remarriage appears to be the primary agent exerting downward pressure on achievement when the SES variables are included. here fore the hypothesis is supported, i.e., whether one controls for SES has a major impact on whether it appears that divorce or remarriage has the primary negative impact on academic achievement.

DISCUSSION

The results found in this paper clearly demonstrate that how one views the relative impact that divorce and remarriage following divorce each have on academic achievement depends largely on whether one controls for SES. The analyses that used models that did not control for the SES level of the custodial parent(s) produced results that indicate that divorce has a greater negative impact on educational achievement than does remarriage. On the other hand, the analyses that used models that did control for SES produced results that indicate that remarriage has a greater negative impact on educational achievement than does divorce. The results found in this paper shed a great deal of light of why various researchers have arrived at different conclusions regarding the relative impact of divorce and remarriage. Indeed, if one chooses to examine the literature carefully one will find that those who emphasize the negative impact of remarriage following divorce almost always control for SES, while those who emphasize the negative impact of divorce

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4


may or may not control for SES. Therefore, the enigma that has arisen in recent years, regarding the relative influence of divorce and remarriage, may have a simple explanation. The disagreements among researchers probably have to'do more with methodology than anything else.

The question that emerges, then, is which way of examining the effects of divorce and remarriage is most appropriate? Certainly in most studies, social scientists controlling for SES has become almost a reflex action. Social scientists generally assume that studies that control for SES are superior to those that do not. If we assume that controlling for SES is most appropriate in this case, then one would conclude that remarriage following divorce, more than the divorce itself, negatively impacts the academic achievement of children. Indeed, what the analyses in this paper communicate to us is that children of divorce from reconstituted families perform poorly academically, given their level of SES. Research indicates that the act of remarriage following divorce raises the SES level of a family almost to the same point as that of an intact family (McLanahan and Sandefur, 1994). Therefore, some'theorists would expect that this would produce a substantial increase in the academic achievement of those children whose parents have experienced divorce. But actually this study indicates that this is not what happens at all. Instead, the academic achievement of children from reconstituted homes is actually roughly the same as or slightly less than that of children of divorce from single-parent homes. These results do not deny that, in general, an increase in parental socioeconomic status may produce some benefit for children. Rather, what these results indicate is that any positive impact that we would normally associate with a rise in the SES level is more than cancelled out by other ways in which remarriage is apparently harmful to children. Clearly, the results of the analyses that controlled for SES indicate that children of divorce from reconstituted families do more poorly academically than one would expect under the assumptions of theorists that emphasize the benefits of increased family income.

The question still arises as to whether controlling for SES gives us the clearest picture of the relative effects of divorce and remarriage. Herzog and Sudia (1971) were the first to emphasize the importance of controlling for SES in every study done on divorce.

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4


They asserted that those studies that do not control for SES over- state the effects of divorce. More recently, however, this perspective has come under attack (Milne, 1989; McLanahan and Sandefur, 1994; Heyman, 1992). A growing number of researchers have asserted that controlling for SES using a post-divorce measure for SES is inappropriate, because the act of divorce causes a family's SES level to drop substantially. Milne (1989), for example, states: ". . . to control (for SES) in this situation would be to remove variance that is truly attributable to father absence but that is pathed through SES" (p. 45). As a result, if one uses a post-divorce SES measure as a control, part of the effects for divorce will be lost in the post-divorce SES measure, because the post-divorce SES measure is itself a result of the act of divorce.

With all these statements in mind, one can see how one could make a case for each of the three perspectives mentioned in the literature review, as representing the best summation regarding the relative effects of divorce and remarriage on academic achievement. For how one views the relative effects of divorce and remarriage may depend on whether one believes it is appropriate to control for SES and how it therefore should be done. Nevertheless, it is probably overly simplistic to assert that only divorce or only remarriage following divorce negatively impacts the academic achievement of children. The results of this study indicate that we, as a society, would be best off recognizing the challenges faced by both children of divorce from single-parent families and children of divorce from reconstituted families. Children of divorce from single-parent families and children of divorce from reconstituted families have some challenges in common and other challenges that are totally unique to their respective family structures. We would do well to increase our sensitivity to these children and increase our knowledge of just what unique challenges they face.

REFERENCES

Ahlburg, Dennis A . and Carol J . DeVita, New realities of the American family. Pop~rkatiotl Bulletin, 47 (1986), pp. 2-44.

Amato, Paul R., Family processes in One-parent, Stepparent, and Intact families: The child's point of view. Jourtlal ofMarriage o t d the Family, 49 (1987), pp. 327-337.

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4


Amato, Paul R. and Gay Ochitree, Child and adolescent competence in Intact, One-Parent, and Stepfamilies, an Australian study. Journal of Divorce, 10 (1987), pp. 75-96.

Anderson, Edward R., Marjorie S. Lindner, and Layne D. Bennion, The effect of family relationships on adolescent development during family reorganization. Monographs of the Society for Research in Child Developnrenl, 57 (1992), pp. 178-199.

Anderson, Edward R. and Alyson M. Rice, Sibling relatinnships during remarriage. Monographs of he Society for Research in Child Developnzenl, 57 (1992), pp. 149-177.

Baydar, Nazli, Effects of parental separation and reentry into union on the emotional well-being of children. Journal of Marriage and [he Family, 50 (1988), pp. 967-981.

Booth, Alan and Judy Dunn, Preface. In Alan Booth and Judy Dunn (Eds.), Stepfa- miliest Who benefifs? Who does not? (Hillsdale: Erlbaum, 1994), pp. ix-x.

Booth, Alan and John N. Edwards, Starting over: Why remarriages are more unstable. Journul of Family Issues, 13 (1992), pp. 179-194.

Bumpass, Larry, Children and Marital Disruption: A replication and update. De- mography, 21 (1984), pp. 71-82.

Chadwick, Bruce A. and Tim B. Heaton (Eds.), Statistical handbook on he American family (Phoenix: Oryx Press, 1992).

Cherlin, Andrew J., Marriage, divorce, and remarriage. (Cambridge: Harvard University Press, 1992).

Duberman, Lucille, The reconslituted family: A sliidy of remarried couples and their children (Chicago: Nelson Hall Publishers, 1975).

Furstenberg, Frank F., Jr., Good dads and bad dads: Two faces of fatherhood. In Andrew J. Cherlin (Ed.), The changing America12 family a d public policy (Washington DC: Urban Institute Press, 1988), pp. 193-218.

Ganong, Laurence H. and Marilyn Coleman, Remarried family relationships. (Thousand Oaks: Sage Publications, 1994).

Ganong, Laurence H. and Mary Coleman, The effect of Remarriage on Children: A Review, of the empirical literature. Family Relations, 33 (1984), pp. 389-406.

Herzog, Elizabeth and Cecilia Sudia, Boys in fatherless families (Washington DC: U.S. Department of Health Education and Welfare, 1971).

Hetherington, E. Mavis and Kathleen M. Jodi, Stepfamilies as settings for child development. In Alan Booth and Judy Dunn (Eds.), Stepfamilies: Who beneJ;ts? Who does not? (Hillsdale, NJ: Erlbaum Associates, 1994) pp. 55-79.

Hetherington, E. Mavis, Coping with marital transitions: A family systems perspective. Monographs of the Society for Research in Child Developme& 57 (1992), pp. 1-14.

Hetherington, E. Mavis, Margaret Stanley-Hagan, and Edward R. Anderson, Mar- ital transitions: A child's perspective. American Psychologist, 44 (1989), pp. 303-12.

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4


Heyman, James R., The relationship between family structure as defined by parental marital status, family structure histories, gender, and various academic outcomes for seventh grade students in private schools. Unpublished doctoral dissertation, University of San Francisco, San Francisco, 1992.

Kelly, Joan, Conflict and children's post-divorce adjustment: A closer look. Paper presented at the National Council for Children's Rights in Arlington, Virginia, 1992.

Kurdek, Lawrence A. et al., The relation between parenting transitions and adjustment in young adolescents: A multisample investigation. Journal 'of Early Adolescence, 14 (1994), pp. 412-431.

Lazear, Edward T. and Robert T. Michael, Allocation of income withitl the Aouse- hold (Chicago: University of Chicago, 1988).

McLanahan, Sara and Gary Sandefur, Growing up with a single parent: What hurts, what helps (Cambridge: Harvard University Press, 1994).

Milne, Anne M., Family structure and the achievement of children. I n William J. Weston (Ed.), Education in the American family: A research synthesis (New York: New York University Press, 1989), pp. 32-65.

Neighbors, Bryan, Rex Forehand, and Lisa Armistead, Is parental divorce a criti- cal stressor for young adolescents? Grade point average as a case in point. Adolescence, 27 (1992), pp. 639-646.

Nunn, Gerald D., Thomas S. Parish, and Ralph J . Worthing, Perceptions of per- sonal and familial adjustment by children from intact, single-parent, and reconstituted families. Psychology in the Schools, 20 (1993). pp. 165-174.

Popenoe, David, The evolution of marriage and the problem of stepfamilies: A biosocial perspective. In Alan Booth and Judy Dunn (Eds.), Stepfamilies: Who benejits? Who does not? (Hillsdale, NJ: Erlbaum Associates, 1994), pp. 3-27.

Shinn, Marybeth, Father absence and children's cognitive development. Psycho- logical Rulletin, 85 (1978), pp. 295-324.

United States Department of Education, National education longitudinal study, 1988:firstfollow-up, 1990. Vol. 1: student data (Washington DC: US. Depart- ment of Education, 1992).

Visher, E.B. and J.S. Visher, Old loyalties, new ties: Therapeutic strategies with stepfamilies (New York: BrunnerIMazel, 1988).

Wallerstein, Judith S. and Sandra Blakeslee, Second chances: Men, women, and children a decade after divorce (New York: Ticknor and Fields, 1989).

Wallerstein, Judith S., Shauna B. Corbin, and Julia M. Lewis, Children of divorce: A 10-year study. I n E. Mavis Hetherington and Josephine D. Arasteh (Eds.), Impact of divorce, single parenting, and step-parenting on children (Hillsdale, NJ: Erlbaum Associates, 1988), pp. 197-216.

Wallerstein, Judith S. and Joan Berlin Kelly, Surviving the breakup (New York: Basic Books, 1980).

Walsh, William M., Twenty major issues in remarriage families. JournalofCoun- seling artdDevelopment, 70 (1992), pp. 709-715.

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4


White, Lynn, Stepfamilies over the life course: Social support. I n Alan Booth and Judy Dunn (Eds.), Stepfamilies: Who benefits? Who does nol? (Hillsdale, NJ: Erlbaum Associates, 1994), pp. 109-137.

Zakariya, Sally Banks, Another look at the children of divorce: Summary report of the study of school needs of one-parent. Principul, 62 (1982), pp. 34-37.

Zill, Nicholas, Understanding why children in stepfamilies have more learning and behavior problems than children in nuclear families. I n Alan Booth and Judy Dunn (Eds.), Stepfamilies: Who betlefits? Who does nol? (Hillsdale: Erlbaum, 1994), pp. 97-106.

Zill, Nicholas and Christine Winquist Nord, R~ul t~ i t tg it t place (Washington, DC: Child Trends, 1994).

for taculty/professronals wrlh]ournal subscrrpfion recommendallon author~ty for thelr insl~tutronal hbrary . . .

If you have read a reprint or photocopy of this article, would you like to make sure that your library also subscribes to this journal? If you have the authority to recommend sub- scriptions to your library, we will send you a free sample cow for review with vour librarian. Just fill out the form below-and make sure that you type or write out clearly both the name of the journal and your own name and address.

( ) Yes, please send me a complimentary sample copy of this journal:

(please write in complete journal title here-do not leave blank)

I will show this journal to our institutional or agency library for a possible subscription. The name of my institutional/agency library is:

NAME:

INSTITUTION:

ADDRESS:

CITY: STATE: ZIP:-

Return to: Sample Copy Department, The Haworth Press, Inc., 10 Alice Street, Binghamton, NY 13904-1580

Dow

nloa

ded

by [

Uni

vers

ity o

f U

tah]

at 1

8:46

11

Oct

ober

201

4

does divorce or remarriage have the greater negative impact on the academic achievement of children?

Documents