phd student divorce

Dissertation Does NOT Mean Divorce

By  Liz Homan

You have / 5 articles left. Sign up for a free account or log in.

Wait! Don’t stop reading, grad student singles. Despite my potentially misleading title, this post is not only for the partnered among us.   It is not only for those of us with boyfriends, girlfriends, live-in partners, or spouses. No, this post is for every graduate student, single or otherwise, who sustains a relationship with someone who means the world to them – be it a parent, a best friend, or a partner. It is for all of us.

First, some background.  I can say from experience that graduate school is hard on relationships. My partner, Kristoff, and I have been together for nearly 11 years. Of those years, at least one of us has been in graduate school for ten of them. In that time, many of our friends have divorced. We have separated (and since reunited). I’ve heard it more than once: “get a dissertation, get a divorce.” Well, not so fast.

Here’s the honest truth.  Maintaining healthy relationships in graduate school is akin to training for an  Ironman .  On some days you’re terrible at one thing while being awesome at something else. Regular rest is necessary. You need to be able to switch gears quickly, and sometimes without much warning. And if you’re not careful, the stress will cause something to break.

Graduate students, it turns out, sort of suck at relationships .  Studies suggest that we’re generally a gloomy crew —mostly because we’re vulnerable to depression and we are highly driven to achieve (and thus struggle with the uncertainty of graduate school). Fellow GradHackers have discussed the  ramifications of mental health struggles  and suggested ways to combat these issues.  These personal struggles can become relationship troubles  if we’re not careful.

In light of these challenges, here are three uncomplicated yet sometimes difficult-to-follow guidelines that have helped me get a dissertation while avoiding a divorce (or permanently damaging any other important relationships).

(1)       Make time for the people who matter in your life, and  follow through.

Planned activities make all the difference here—I find it helpful to have “a thing” that is “our thing” (or, for the people I’m closest to, multiple “things”). For example, Kristoff and I have lots of “things” that are “ours”: running, cooking, our dog, home improvement, and football are among them. But I also have “things” with other people that I don’t share with him; I kayak with one friend, I talk about teaching with another friend, I eat sushi and shop with my mom.

But there’s an important catch:  no saying “we should do that” without following through. See, if you suggest that something could happen in the future, and then it doesn’t, this could lead to disappointment or resentment for your friend or partner— without you even knowing . So instead of “we should do that,” try “when can we do that?” And get your calendars out.

(2)      Communicate,  honestly  and  promptly .

Okay, fine, I know – everyone has told everyone since the beginning of time that the key to a good relationship is communication, right? Duh. Moving on.

This one has a catch too.  After all, it wouldn’t be repeated so often if we were good at it. I find that the key to good communication is  honest  and  prompt  communication. This means that when a special someone annoys, hurts, or angers me, I need to tell that person.  Right then . Even if it’s uncomfortable. Even if it’s upsetting. This kind of communication is the easiest to avoid, precisely because it is so uncomfortable. Be kind and calm, and rest assured that your loved ones will respect you more for your honesty than they would have for your passive-aggressive anger.

(3)      Be an individual,  but don’t isolate .

Graduate school changes people. I enjoy doing things now that I never thought I would do, nevermind enjoy (like running), and I have lost interest in other things that used to occupy much of my free time (like crocheting). I need to know who I am on my own before I can have healthy relationships with friends or family, and that person continually shifts and evolves. Find things that make you happy. Do them, unapologetically. Be who you are.

You guessed it. The catch:  It is so easy for us to isolate ourselves in graduate school. To become too “in our own heads,” too absorbed by the work of a thesis or project to invest ourselves in the interests and lives of others. So, do those things that matter to you, that make you an individual,  in the service of others.  In other words, adamantly be your own person so that you can be a better person when you’re with those who matter most to you. For me this means long runs on the weekends are non-negotiable, because I’m a better partner and person when I take that time to be just me.

As you can see, there’s always a catch . It sometimes takes an Ironman’s dedicated soul and versatility to make graduate school relationships work. And sometimes—despite our best efforts—they just don’t. And that’s okay too, especially if you have a healthy network of kindred souls to catch your fall.

How do you maintain healthy relationships in graduate school? Tell us below!

[Image by Flickr user  Jessica Rossi,  used under creative commons licensing.]

4 Ways to Reduce Higher Ed’s Leadership Deficit

Without good people-management skills, we’ll perpetuate the workforce instability and turnover on our campuses, warns

Share This Article

More from gradhacker.

5 Productivity Practices That Helped Me Finish My Dissertation

Summer Planning Strategies

Holding Pattern

• Become a Member
• Sign up for Newsletters
• Learning & Assessment
• Diversity & Equity
• Career Development
• Labor & Unionization
• Shared Governance
• Academic Freedom
• Books & Publishing
• Financial Aid
• Residential Life
• Free Speech
• Physical & Mental Health
• Race & Ethnicity
• Sex & Gender
• Socioeconomics
• Traditional-Age
• Adult & Post-Traditional
• Teaching & Learning
• Artificial Intelligence
• Digital Publishing
• Data Analytics
• Administrative Tech
• Alternative Credentials
• Financial Health
• Cost-Cutting
• Revenue Strategies
• Academic Programs
• Physical Campuses
• Mergers & Collaboration
• Fundraising
• Research Universities
• Regional Public Universities
• Community Colleges
• Private Nonprofit Colleges
• Minority-Serving Institutions
• Religious Colleges
• Women's Colleges
• Specialized Colleges
• For-Profit Colleges
• Executive Leadership
• Trustees & Regents
• State Oversight
• Accreditation
• Politics & Elections
• Supreme Court
• Student Aid Policy
• Science & Research Policy
• State Policy
• Colleges & Localities
• Employee Satisfaction
• Remote & Flexible Work
• Staff Issues
• Study Abroad
• International Students in U.S.
• U.S. Colleges in the World
• Intellectual Affairs
• Seeking a Faculty Job
• Advancing in the Faculty
• Seeking an Administrative Job
• Advancing as an Administrator
• Beyond Transfer
• Call to Action
• Confessions of a Community College Dean
• Higher Ed Gamma
• Higher Ed Policy
• Just Explain It to Me!
• Just Visiting
• Law, Policy—and IT?
• Leadership & StratEDgy
• Leadership in Higher Education
• Learning Innovation
• Online: Trending Now
• Resident Scholar
• University of Venus
• Student Voice
• Academic Life
• Health & Wellness
• The College Experience
• Life After College
• Academic Minute
• Weekly Wisdom
• Reports & Data
• Quick Takes
• Advertising & Marketing
• Consulting Services
• Data & Insights
• Hiring & Jobs
• Event Partnerships

4 /5 Articles remaining this month.

Sign up for a free account or log in.

• Sign Up, It’s FREE

Stack Exchange Network

Stack Exchange network consists of 183 Q&A communities including Stack Overflow , the largest, most trusted online community for developers to learn, share their knowledge, and build their careers.

Q&A for work

Connect and share knowledge within a single location that is structured and easy to search.

PhD student and postdoc statistics related to relationships

As a graduate student and postdoc, I remember more of my colleagues becoming single than entering new relationships. In many cases the problem was that their partners could not accept their unusual working hours and low wages. In some others it was the lack of solution to the "two body problem".

So, I am curious if there are compiled statistics related to number of students and postdocs in relationships and with families, number of children, divorces, and comparison to people outside academia.

• graduate-school
• united-states
• reference-request

• 7 I'm afraid to see the numbers. You should add the category 'married to student/married to advisor'. –  HEITZ Commented Jun 29, 2016 at 8:08
• 1 That is a really...funny question. My wife was the associate chair of our department. Now we've both taken our educational abilities and applied them to other lines of work. –  Raydot Commented Sep 21, 2016 at 17:39
• Please note the statistics can vary a lot by country. Most of the PhD students I know, mine or not, are not single. Could you please specify the country. –  Massimo Ortolano Commented Jan 26, 2017 at 19:40
• @MassimoOrtolano In my case, it was US. –  user21264 Commented Jan 26, 2017 at 19:42
• I've added the US tag to better specialize the answers, unless you're interested in a worldwide perspective. –  Massimo Ortolano Commented Jan 26, 2017 at 19:48

This study has "postsecondary teachers" divorce rate at around 14%, and "Other education, training, and library workers" at 15.6.

For reference, "Computer scientists and systems analysts" were at 15.64.

Both were below "verage US Divorce Percentage" of 16.35.

Most occupations typically associated with financially secure lifestyles (doctors, engineers, CEOs) seem to fall <= 10%.

Most occupations not typically associated with stability or prosperity fell >= 20%.

Predictably, it appears that financial stability, on average, brings an order of magnitude increase in the stability of marriage. So all the girls looking to marry a doctor or lawyer (or engineer??) really do seem to be statistically more likely to live happily ever after ;) Likewise for all the boys looking to marry up.

In my relationship, we went through a total of 3 graduate degrees, for 2 of which we were both in school at the same time.

That definitely helped, as each of us had first-hand experience of living below subsistence level while in grad school. We were able to mitigate the inevitable adverse effect of financial insecurity on a relationship by being understanding and patient with each other; developing judicious spending habits; living below our means; and understanding that, in the long run, a relationship of the kind that we have is more important than a job, a salary, or a lifestyle.

• 6 "So all the girls looking to marry a doctor or lawyer (or engineer??) really do seem to be statistically more likely to live happily ever after ;)" Hey, this is 2017. Girls can be doctors, lawyers and engineers now, and boys can be looking to marry them. Could you perhaps rewrite this part of your answer? –  Pete L. Clark Commented Jan 26, 2017 at 18:14
• Done. Thank you for pointing this out. If you have further suggestions, you are welcome to write your own complete response to the OP. –  A.S Commented Jan 26, 2017 at 18:45
• 2 @PeteL.Clark It is well known that girls are not looking to marry an engineer (sorry, as an engineer I couldn't resist :-) ). –  Massimo Ortolano Commented Jan 26, 2017 at 19:51
• @MassimoOrtolano Engineers come in different shapes, and while the rank-and-file, Dilbert-type specimens are probably not much of a catch, there are exceptions to every rule; case in point: Rex Tillerson, Exxon Mobil's CEO (and Trump's pick for Sec of State) began with the company as an engineer, after graduating with a civil engineering degree ( en.wikipedia.org/wiki/Rex_Tillerson ). –  A.S Commented Feb 5, 2017 at 18:11
• @Aymor I was just kidding, don't take my comment seriously. –  Massimo Ortolano Commented Feb 5, 2017 at 18:48

You must log in to answer this question.

• Featured on Meta
• Introducing an accessibility dashboard and some upcoming changes to display...
• We've made changes to our Terms of Service & Privacy Policy - July 2024
• Announcing a change to the data-dump process

Hot Network Questions

• Solar System Replacement?
• Is Heaven real?
• Countably compact Boolean algebras versus distributivity
• What does "No camping 10-21" mean?
• How to get InverseMatrix for large size Matrix in Mathematica?
• Four Year Old Pushing and Positive Discipline Techniques
• How does one find historic prices for goods?
• Why did Jesus give Pilate "no answer" to the question “Where are You from?” (John 19:9)?
• Circuit for switching between two AC voltage sources
• What's the difference between "Model detail" and "Texture detail" in Portal 1?
• Confused with "per āera" in "Familia Romana"
• A spaceship travelling at speed of light
• What type of concept is "mad scientist"?
• 4 X 4/0 or 3 X 4/0+2/0 for manufactured home service entrance
• What would "doctor shoes" have looked like in "Man on the Moon"?
• A finance broker made me the primary instead of a co signer
• My first cryptic crossword
• A simple reusable class for rational numbers
• How to proceed if my manager is ghosting me?
• cd 'old' 'new' with multiple "whole directory-name" substitutions
• Closed form of an integral using Mathematica or otherwise
• What's "unregulated baggage"?
• What purpose did the lower-right "Enter" key serve on the original Mac 128k keyboard?
• How could ocean liners survive in a major capacity after large jet airliners become common?

Link between divorce and graduate education a concern as more jobs require advanced degree

Posted Aug 28, 2018 10:30 am

AMES, Iowa – Children of divorce are less likely to earn a four-year or graduate degree, according to new research from Iowa State University.

The study, published in the Journal of Family Issues , is one of the first to look specifically at divorce and graduate education. Susan Stewart, professor of sociology , says it is important to understand this relationship as more jobs require a graduate or professional degree.   

Stewart and co-authors Cassandra Dorius, assistant professor of human development and family studies; and Camron Devor, lead author and Iowa State alumna, found 27 percent of children with divorced parents had a bachelor’s degree or higher, compared to 50 percent of those with married parents. The split was 12 percent versus 20 percent for those who had or were working toward a graduate or professional degree.

The researchers analyzed 15 years of data collected through the National Longitudinal Survey of Youth 1997. The survey followed thousands of youth as they transition from school to work in young adulthood. The last round of data used for this study was collected when youth were 26 to 32 years old.

The data allowed researchers to look at the influence of human (parental education and income) and social (parental social and emotional investment in children) capital. They found married parents were more educated than divorced parents, and there was a significant difference in income. Nearly half of the children with married parents were in the high income category (greater than $246,500/year) compared to 29 percent of children of divorced parents.   

“After divorce, for both men and women, incomes take a hit. It takes much longer for that income to recover and for women especially, it never does,” Stewart said. “You are essentially starting over and much of the income that would have gone to a child’s education is sucked up with all the transitions that are part of divorce.”

Time for change?

According to the Bureau of Labor Statistics, jobs requiring a master’s degree are expected to grow by nearly 17 percent between 2016 and 2026. This includes careers ranging from mental health counselors to librarians to elementary and secondary school administrators. Devor, who earned a master’s degree in sociology in 2014, says she wouldn’t have her job as a finance coordinator had she not gone to graduate school.

However, the findings were somewhat surprising to Devor based on her experience at Iowa State. She says several of her classmates in graduate school were children of divorce. Recognizing that this is not always the norm, Devor would like to see the research signal a change.

“This could affect divorce proceedings for child support and the amount that is factored in for college,” Devor said. “In most divorce proceedings, child support cuts off at 18. Just because a child turns 18, that does not mean they still do not need help financially from their family.”

Child’s age matters, to a degree

Children who were still at home or under age 18 when their parents divorced did not fare as well as children who were 18 and older. The research found the odds of those younger children earning a bachelor’s degree were 35 percent lower. However, there was no relationship between the child’s age at the time of divorce and the likelihood of getting a graduate or professional degree.

The researchers also found parents had similar educational expectations for their children, regardless of whether they were divorced or married. Parental expectations were positively associated with children earning a master’s degree. Dorius says children of divorce may feel less entitled to a college degree, so it should help them to know their parents have high educational aspirations for them. However, that encouragement is not enough to offset the relationship between divorce and graduate education.

“This suggests that parental divorce continues to have an effect on children’s graduate school success even after accounting for the encouragement parents give to their children,” Dorius said. “It’s important for future research to look at other inadequacies in social capital that may affect long-term educational success for these children.”

Susan Stewart, Sociology, [email protected] , 515-294-5912

Cassandra Dorius, Human Development & Family Studies, [email protected] , 515-294-1640

Camron Devor, Sociology alumna, [email protected]

Angie Hunt, News Service, [email protected] , 515-294-8986

Children of divorce are less likely to earn a four-year or graduate degree, according to new research from Iowa State University. Researchers says it is important to understand this relationship as more jobs require a graduate or professional degree.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List

Divorce and adolescent academic achievement: Heterogeneity in the associations by parental education

Sondre Aasen Nilsen

1 Regional Centre for Child and Youth Mental Health and Child Welfare, NORCE Norwegian Research Centre, Bergen, Norway

2 Department of Psychosocial Science, Faculty of Psychology, University of Bergen, Bergen, Norway

Kyrre Breivik

3 Department of Health Promotion and Development, Faculty of Psychology, University of Bergen, Bergen, Norway

Kristin Gärtner Askeland

Børge sivertsen.

4 Department of Health Promotion, Norwegian Institute of Public Health, Bergen, Norway

5 Department of Research & Innovation, Helse Fonna HF, Haugesund, Norway

6 Department of Mental Health, Norwegian University of Science and Technology, Trondheim, Norway

Mari Hysing

Tormod bøe, associated data.

Norwegian Health research legislation and the Norwegian Ethics committees require explicit consent from participants in order to transfer health research data outside of Norway. In this specific case, ethics approval is also contingent on storing the research data on secure storage facilities located in our research institution. Data are from the Norwegian dnaladroh@htuoy study whose authors may be contacted at on.inu@bib .

The link between parental divorce and adolescents’ academic achievement may depend on parental educational levels. However, findings have been inconsistent regarding whether the negative associations between parental divorce and adolescents’ academic outcomes are greater or smaller in highly educated families. The present study aimed to investigate the possible heterogeneity in the associations between divorce and adolescents’ academic achievement by parental educational levels, within the context of the elaborate Norwegian welfare state.

The population-based cross-sectional youth@hordaland study of adolescents aged 16–19 years conducted in Norway in 2012, provided information about parental divorce and was linked to national administrative registries ( N = 9,166) to obtain high-quality, objective data on the adolescents’ grade point average (GPA), and their parents’ educational qualifications and income.

The negative association between parental divorce and GPA was stronger among adolescents with educated or highly educated parents compared to adolescents with less educated parents. This heterogeneity was driven by maternal educational qualifications, whereby divorce was more strongly and negatively associated with GPA among adolescents with educated mothers compared to those with less educated mothers, independent of paternal educational levels and income measures.

Conclusions

Among adolescents whose parents have low educational qualifications, parental divorce is not associated with their academic achievement. Educated divorced mothers appear less likely to transfer their educational advantages onto their children than nondivorced equally educated mothers, perhaps due to a “double-burden” regarding work pressure and child-rearing responsibilities. There is a need for future studies to detail the mechanisms underlying this finding.

Introduction

Children and adolescents with divorced or separated parents are less well-adjusted on average across a spectrum of outcomes, including physical and mental health, and do less well in school compared to those who grow up with nondivorced parents [ 1 – 3 ]. Adolescents whose parents divorce have been found to experience a decline in overall grade point average (GPA) of one quarter to one-third of a letter grade and to fail more classes than those continuously living with both parents [ 4 ]. In general, recent studies suggest that the association between divorce and youths’ academic success is partly causal [ 5 , 6 ]. This link between divorce and poorer academic achievement is important, as successful schooling may have a long-term impact on later educational attainment, occupational and economic stability [ 7 ], and future physical and mental health [ 8 – 10 ].

To advance research on divorce and youth adjustment, we need to move beyond focusing on averages and try to elucidate for whom and under which circumstances divorce might be associated with adverse outcomes [ 1 ]. It is well established that parents’ educational attainment is a strong predictor of their children’s academic achievement [ 11 , 12 ]. Educated parents monitor their children’s academic progress more closely, have more realistic expectations of their academic abilities, and apply more optimal parenting strategies than less educated peers; all factors linked to positive school outcomes among youth [ 11 , 13 – 15 ]. Further, educated parents may be better able than less educated parents to cope with a divorce due to having greater financial resources and otherwise being more robust against the often-stressful situation that divorce entails [ 16 ]. Thus, at face value, it could be reasonable to assume that highly educated parents, on average, buffer their children more against possible negative consequences of divorce on their academic achievement.

Interestingly, two contrasting theoretical perspectives have been proposed in explaining how a divorce might differently impact children’s academic achievement by parental educational levels, and both have received empirical support. Building on the above argument, the compensatory class hypothesis posits that adverse life events (such as parental divorce or separation) are less harmful to children from higher class (i.e., highly educated) families [ 17 ]. Greater financial, social, and parental resources among educated parents might enable them to plan ahead and counteract possible adverse post-separation effects on their children [ 17 , 18 ]. In support of this perspective, a few studies have found that highly educated parents buffer their children against negative consequences of divorce on measures such as math and reading skills, GPA, and later educational attainment [ 17 , 19 , 20 ].

The floor effects hypothesis, on the other hand, states that children in less educated families have limited access to financial, social, and parental resources to begin with. Thus, they have less to lose from a divorce than peers from more affluent families, who may experience a more substantial reduction in financial and parental resources invested in them [ 16 , 18 , 21 ]. Several recent studies support this perspective, whereby the educational disadvantages associated with divorce are relatively larger among children with highly educated parents [ 16 , 21 – 24 ]. For example, Martin [ 23 ] found a larger negative association of divorce among children with highly educated compared to children with less educated parents across current math test scores, GPA, and later educational transitions (e.g., high school completion).

Recent efforts have been made in trying to reconcile these contradictory findings (see [ 21 , 22 ]). Firstly, they may stem from differences in measurements and analyses across studies. Importantly, including both parents’ educational qualifications in the analyses appears necessary to capture the parental and economic resources available to the child when one of the parents moves out of the home. Relying solely on maternal educational levels, for instance, may not capture the loss in resources experienced when an educated father moves out, and have been found to remove heterogeneity in the outcomes of divorce on adolescents’ educational attainment [ 22 ]. Moreover, the results might depend on the educational outcome; while findings regarding later educational attainment tend to favor the floor effects hypothesis [ 16 , 21 – 24 ], studies investigating current school performance (e.g., subject grades, GPA, test scores) have yielded more inconsistent results [ 17 , 20 , 21 , 23 ].

The diverging results may also stem from societal factors that vary across countries and time periods, possibly impacting the mechanisms driving heterogeneous outcomes. The economic cost of divorce has traditionally been higher for women than men in several western countries [ 25 , 26 ]. Thus, children with an educated father may lose relatively more financial resources following divorce (as the father often moves out) than peers with a less educated father. Supporting this notion, two studies utilizing data from the 1970 British Cohort Study found that the relatively larger educational consequences of divorce among children with highly educated parents primarily was driven by the lost access to the fathers’ resources [ 21 , 22 ]. However, as noted by the authors, only a minority of nonresident fathers paid child support during this period in Britain. Thus, the role of fathers’ resources might be less important in other contexts. Indeed, a study on a somewhat more recent US sample found that maternal education was relatively more important than paternal education, whereby highly educated divorced mothers were less likely to transfer their educational advantages onto their children [ 23 ]. This finding was partly explained through highly educated divorced mothers’ relatively lower academic expectations, involvement in school, and leisure activities, compared to nondivorced peers. As put by the author [ 23 ]; “ High status single mothers are accomplished , but frequently time constrained” .

Another potential source of diverging results may stem from the commonality of divorce across different educational strata across countries. A recent study found that parental divorce was more detrimental to the educational attainment among children whose parents had a low likelihood of divorce, compared to those with parents with a high likelihood of divorce [ 24 ]. A divorce might thus come as more of a shock among families unprepared for disruption, in turn negatively impacting the children from a perhaps otherwise privileged background. Among children with parents at a high risk of divorce, a divorce might rather be one of many adverse events faced during childhood.

Previous studies have primarily been conducted on British, US and German samples on cohorts from the 1970–1980s, and it is uncertain whether previous findings generalize to other contexts. In keeping with the stated need for studies on more recent samples and in other cultural contexts [ 21 , 22 ], this study sought to investigate the potential heterogeneity in the associations between parental divorce and adolescents’ academic achievement within the elaborate Norwegian welfare state.

The Norwegian context

The Norwegian welfare state is amongst the biggest spenders on welfare in the world [ 27 ]. It provides an elaborate social safety net through free access to the health care system and access to sickness, unemployment, and family-related benefits. Levels of absolute deprivation and income inequality in Norway is low [ 28 , 29 ], and the population is highly educated; 38.2% of all women and 30.1% of all men had completed some form of university-level education in 2019 [ 30 ]. Like in the other Nordic welfare states, the “dual-earner family” is strongly encouraged; public childcare and schools are highly subsidized, and generous parental leave rights (also for the father) have facilitated the combination of full-time employment and childcare among both mothers and fathers (see [ 31 ]). Perhaps, as a result, fathers’ housework and childcare time are generally high among most groups of fathers, although women still do more than their male partners [ 32 ].

The crude divorce rate in Norway has nevertheless more than doubled since the 1970s, and there were approximately two divorces per 1000 persons in 2016 [ 33 ]. The risk of divorce is higher for the less educated, whereby a couple where both have low levels of education run a risk that is more than four-fold in magnitude compared to couples where both have higher educational qualifications [ 34 ]. After a divorce, custodians are supported by tax deductions, law-regulated cash allowances, and child support, which is enforced by the public authorities. It is estimated that 50% of parents experience a drop in household income following a divorce in Norway [ 35 ]. Nevertheless, the Norwegian welfare state appears fairly successful in equalizing gender differences in the cost of divorce, whereby men and women experience an approximately equal 20% decline in disposable income [ 36 ]. This contrasts findings from many other countries where women often lose substantially more than men [ 25 , 26 ]. However, a divorce increases the sickness absence among women with children in Norway but less so for men [ 37 ]. This is in general keeping with the “double-burden” hypothesis [ 38 ], suggesting that high labor participation, coupled with high child-rearing responsibilities, could be an extra burden for divorced, educated women in Norway. Indeed, although the rates of families who share custody have risen in Norway in the last decades, approximately 65% still live in mother custody following divorce [ 39 ].

The link between parental divorce and academic achievement is well documented in Norway; divorce has been associated with having more problems in school [ 40 ], lower GPA [ 41 , 42 ], and lower probability of completing higher secondary education [ 43 ]. The rates of students not completing high-school are higher in Norway than in many other comparable countries [ 44 ]. Not receiving a high-school diploma is associated with a higher risk of later receiving medical and non-medical social insurance benefits [ 45 ], and thus represents a significant public health concern. Increased knowledge of the links between parental divorce, parental education, and adolescents’ school performance, could thus provide insights that can be utilized in efforts aimed at facilitating high-school completion among adolescents.

The present study

The main aims of the present study were twofold: Firstly, to investigate the association between parental divorce and adolescents' GPA. Secondly, to examine whether parental educational qualifications moderated this association. To aid these aims, we draw on high-quality register-based measures of adolescents’ GPA, parental educational qualifications, and household income, that were merged with a population-based study.

The present study contributes to this field by studying heterogeneous associations of divorce in a society that combines generous social benefits; a highly educated population; high levels of labor-force participation among women; gender equity in the cost of divorce; and high divorce rates and gender differences in sickness absence and health outcomes among divorced individuals.

The high divorce rate among the relatively few uneducated parents in Norway suggests that a divorce perhaps is one of many adverse events experienced in these selected and often socioeconomically disadvantaged families. This led us to expect that the negative association between divorce and adolescents’ GPA would be stronger in more highly educated families where a divorce might carry greater changes to children’s lives. However, as the Norwegian welfare state appears successful in equalizing the cost of divorce between men and women, we suspected that loss of fathers’ financial resources following divorce might be less important in understanding potential heterogeneous outcomes in Norway compared to previous studies conducted in other sociopolitical contexts.

Materials and methods

We used data from the youth@hordaland study, a population-based survey of adolescents aged 16–19, conducted in the spring of 2012 in Hordaland County, Norway. The youth@hordaland study aimed to assess mental health, family life, lifestyle, school performance, and health service use in adolescents. The adolescents received information about the study per e-mail, and one regular school hour was allocated to complete the questionnaire. Those not attending school on the day of the study could complete the questionnaire at their convenience, and some schools arranged catch-up days. A teacher organized the data collection and protected confidentiality. The adolescents themselves indicated if they consented to complete the entire survey or selected parts of it, as Norwegian regulations dictate that individuals aged 16 years and older are required to consent themselves. Their parents were informed about the study, and the study was approved by the Regional Committee for Medical and Health Research Ethics in Western Norway.

All adolescents born between 1993 and 1995 and residing in Hordaland at the time of the survey were invited (N = 19,439) to participate, and 10,257 agreed, yielding a participation rate of 53% for the entire study. The present paper is based on a subsample of 9,166 adolescents (47% of the invited population) who consented to register linkage. This subsample was nearly identical to the total sample with regards to age and gender distribution, and self-reported sociodemographics ( S1 Table ).

Measurements from registers

Age and gender.

Date of birth and gender were obtained through the adolescents' identity number in the Norwegian National Registry. Exact age was calculated from the date of participation in the youth@hordaland study and the birthdate of participants.

The adolescents’ GPA for each year in upper secondary education were obtained from the National education database in Norway (NUBD) that is owned and administered by Statistics Norway. NUBD contains educational statistics from elementary school through PhD-level. In Norway, each subject is graded on a scale ranging from 1 (failure) to 6 (excellent), and the GPA is thus calculated by taking the sum of all grades received in a given school year divided by the total number of subjects. The grades used in the current study stem from the school-year of 2011–2012. Thus, the grades correspond to the school year that the adolescents were in at the time of the youth@hordaland study. A previous publication found that the mean GPA in the current sample was quite similar to both regional and national statistics, indicating representativeness of the sample [ 46 ].

Parental education

The highest completed educational level of both parents when the adolescents were 16 years old were also obtained from NUBD. The International Standard Classification of Education (ISCED) 2011 coding-scheme was utilized to create three main measures of parental educational levels: 1: A combined measure of parents’ educational level indicating the highest completed education in the family by either the mother or the father. The categories were (1) both parents have no qualifications higher than lower secondary education (ISCED 0–2), (2) at least one parent has qualifications equivalent to ISCED 3–5 (upper secondary education, post-secondary non-tertiary education, short-cycle tertiary education), at least one parent has education on Bachelor’s level or equivalent (ISCED 6), and at least one parent has attained a Master's or Doctoral level of education (ISCED 7–8). This variable aimed at capturing the range of parental educational levels within a manageable and meaningful set of categories.

2: A combined measure of parents’ educational level used to investigate the relative importance of maternal and paternal educational levels. The categories were (1) both parents have no qualifications (ISCED 0–2), (2) only the mother has some qualifications (above ISCED 2), (3) only the father has some qualifications (above ISCED 2), and both have some qualifications (above ISCED 2). This operationalization or similar has been utilized by previous studies [ 21 , 22 ], and we report the results of this categorization in order to facilitate comparison with the pre-existing literature.

3: We also created separate variables for maternal and paternal educational levels to investigate how sensitive the estimates were based on the choice creating combined measures of parents’ education, and to further detail the relative contribution of maternal and paternal educational qualifications within a broader range of educational levels (i.e., ISCED 0–2, ISCED 3–5, ISCED 6, ISCED 7–8).

Measures of family finances

The Norwegian national income registry provided information on family finances. The Norwegian Government utilizes this data to estimate taxation, and it can be considered to be of high quality. We utilized three measures of income as covariates in the analysis: Mother’s and father’s net income (i.e., the sum of wages and salaries, income from self-employment, property income and transfers received minus total assessed taxes and negative transfers), and the equivalized disposable income (EDI) in the household occupied by the adolescents. EDI is a measure of income in a household that is adjusted by an equivalence scale. EDI has been documented in prior publications from the youth@hordaland study [ 47 , 48 ]. The current study utilizes the OECD modified scale, which gives the first adult in the household a weight of 1, subsequent adults are given a weight of 0.5, and each child below 14 years of age is given a weight of 0.3 [ 49 ]. The equivalence scale thus enables comparison between households of different sizes and compositions. All income measures stem from the year 2011.

Measures from the youth@hordaland study

Parental divorce or separation.

We coded experience of divorce or separation according to the adolescents’ answers to the following questions: “Do your biological parents live together?" and “Have your biological parents divorced or separated?”. Adolescents stating that their biological parents did not live together and that their biological parents had divorced or separated were categorized as having divorced parents, while those stating that their biological parents still lived together were defined as living in a nondivorced two-parent (i.e., nuclear) family. These items allowed us to separate between adolescents whose parents split apart, from adolescents whose parents never lived together, were separated due to death, illness or other reasons (which were removed from the analyses), and resulted in a dummy coded variable (0 = nondivorced/nuclear family, 1 = divorced family). The adolescents also reported year of parental divorce or separation allowing us to calculate a variable of years since the event of dissolution.

We had no means of determining whether the parents were legally married. Official statistics report that 73.5% of children and youth below the age of 18 in a two-parent household in Hordaland county in 2012 lived with married parents (the rest with cohabiting parents) [ 50 ]. Thus, the nondivorced group in the present study most likely contained a group of adolescents with parents that had cohabitated since their birth. As some cohabiting unions eventually marry, we find it likely that the proportion of cohabiting unions in the present sample was somewhat lower than regional estimates also including younger children. Similarly, the divorced group likely contained a group of adolescents whose parents split up from cohabitation. Unfortunately, no official statistics regarding dissolution from cohabiting unions in Norway exists. Our inability to exactly detail the adolescents’ family structure is not unique to the present study but has been rather common within this research field [ 2 ]. For ease of exposition, while keeping the aforementioned statistics in mind, we use the term divorce to refer to the dissolution of either cohabitating or marital unions.

Statistical analyses

Ordinary Least Squares (OLS) regression analyses were conducted to investigate the associations between parental divorce, parental education, and the adolescents’ GPA. In the first OLS models, we used the highest completed education in the family as a measure of parental education. The regression models were structured as follows: A baseline model estimating the associations between parental divorce and adolescents’ GPA, adjusted by gender and age; Model 1 included the measure of parents’ highest completed education; Model 2 added the interaction term between education and parental divorce to investigate the possible heterogeneity in the effects of divorce on the adolescents’ GPA; and Model 3 further included the equivalized disposable income in the household currently occupied by the adolescent, mother’s net income, and father’s net income. These income measures would thus shed light on the possible attenuating effects of both maternal and paternal income levels on the associations between divorce, parental educational levels, and their interactions.

Age and all income measures were centered on their respective means in the regression analyses to ease the interpretation of the regression coefficients. The income measures were divided by a factor of 100,000. Thus, the regression coefficients of the income measures indicate the predicted change in the adolescents’ GPA by an increase of 100,000 NOK above the mean.

To replicate the categorization of parental education used by several prior studies [ 21 – 23 ], the above models were re-run utilizing the second measure of parental education, separating families where either none, only the mother, only the father, or both parents had educational qualifications greater than ISCED 2.

Lastly, to test the sensitivity of the above models and to investigate further possible differential associations of maternal and paternal educational levels, the analyses were re-run with maternal and paternal education entered as separate predictors, while retaining the full range of educational levels.

In all regression models, the reference categories for the parental educational variables were set at the lowest parental educational level (i.e., ISCED 0–2). Checks were made for other differences between the educational levels in the association between divorce and the adolescents’ GPA by alternating the reference categories.

Incomplete responses were fairly low in the current sample, where the majority of missing values pertained to the divorce status variable (8.8%), followed by father’s net income (4.6%) and parental education (3.7%), whereas the remaining variables utilized in the current study had below 3% missingness. Due to the relatively low proportion of incomplete responses, missing values were handled by listwise deletion in the regression analysis.

Robustness and sensitivity analyses

Conditioning on measures of income and paternal and maternal education simultaneously may introduce overcontrol bias [ 51 ]. In the two first set of regression analyses, we try to avoid this problem by creating single measures combining information on parental education from both parents, and by entering income variables in the last set of models (as we were not interested in the main effects of income per se). In the last set of regressions, we have made robustness checks by entering maternal and paternal education in separate models.

The timing of divorce could potentially covary with parental educational levels and the adolescents’ GPA (e.g., if highly educated parents divorced later on, the estimates of divorce by parental education on the adolescents’ GPA could be influenced by the proximity to the event of dissolution). Moreover, association between timing of divorce and the adolescents’ GPA may depend on parental educational qualifications (i.e., that more time spent with highly educated divorced parents differ from time spent with lowly educated divorced parents). We investigate these issues by comparing years since divorce across parental educational levels, and by graphically plotting potential linear and non-linear relationships between timing of divorce and GPA by parental educational qualifications. Generalized additive models (GAMs) were used to investigate potential non-linear relationships. In brief, GAMs may be considered as a semi-parametric extension of the generalized linear model, with the strength of the ability to detect non-linear structures in data that otherwise might be missed [ 52 ].

Lastly, we performed checks utilizing the income measures as alternative indicators of the family’s socioeconomic resources. The income measures were divided into quartiles (i.e., into four equal parts representing the lowest 25% to the highest 25%), and the adolescents’ GPA was regressed on the interaction term between parental divorce and the income quartiles (similarly to the procedure described above).

All statistical analyses were conducted using R version 3.5.2 for Mac [ 53 ]. Figures were created with the packages “ggplot2” [ 54 ], “sjPlot” [ 55 ], and “ggstatsplot” [ 56 ]. The GAMs were plotted with aid from the “mgcv” package [ 57 ] within the “geom_smooth” function of the ggplot2 package. For brevity, statistical parameters are included in figures displaying pairwise comparisons. In Table 1 , the effect sizes for categorical variables were calculated from the Mahalanobis distance method and compared between groups [ 58 ].

Eff. Size = effect size, as represented by the standardized mean difference. For categorical variables, the effect sizes were calculated from the Mahalanobis distance method. NOK = Norwegian krone.

** p < 0 . 01 ; p -values derived from chi square tests for categorical variables, and Welch two-sample t test for continuous variables.

Characteristics of the sample

There were fewer boys in the divorced sample (43.6%) compared to the nondivorced sample (47.2%). Parents who divorced had lower education; almost twice as many divorced parents did not have higher than ISCED 2 qualifications (6.6%) compared to nondivorced parents (3.8%), and having qualifications equivalent to Bachelor's level (ISCED 6) or Master's or PhD-levels (ISCED 7–8) were more frequent among nondivorced parents compared to divorced parents. Divorced households had lower equivalized disposable income compared to nondivorced households. While nondivorced fathers had higher net earnings than their divorced counterparts, divorced mothers had higher net earnings than nondivorced mothers (see Table 1 for details).

Regression results

Highest education in the family.

The first tested OLS models utilizing the highest completed education in the family as a measure of parental education are displayed in Table 2 . The baseline model indicated that adolescents with divorced parents on average had 0.30 points lower GPA score (Cohen’s d = 0.34) compared to their peers with nondivorced parents. Statistically controlling for the highest completed parental educational level (Model 1), reduced the strength of the association between parental divorce and GPA by 0.06 GPA points (20%), indicating that the association between parental divorce and GPA were relatively robust to adjustments for parental educational levels. Independent of parental divorce, higher education in the family was associated with a higher GPA. The interactions between divorce and parental education were further added in model 2, while income measures were added in model 3. Taken together, the results from these models showed that the associations between having divorced parents and the adolescents’ GPA were significantly stronger among adolescents where the highest parental education was at secondary school levels (ISCED 3–5) or Bachelor's levels (ISCED 6), compared to those with parents that did not have higher than basic-level education (ISCED 0–2). Although the same trend was observed among adolescents with at least one parent with a Master’s or PhD-level education (ISCED 7–8), the interaction term was not significant. Including the income measures (Model 3) hardly changed these estimates. The interactions are visually depicted in Fig 1A .

The predicted values of GPA by ( A ) the highest educational qualifications obtained in the family, and ( B ) by the highest maternal and paternal education above/below ISCED 2 from the fully adjusted regression models (cf. Tables ​ Tables2 2 and ​ and3), 3 ), stratified by divorce status. Error bars represent 95% confidence intervals of b . A : ISCED 0–2 = up to lower secondary education, ISCED 3–5 = upper secondary education, post-secondary non-tertiary education, short-cycle tertiary education, ISCED 6 = Bachelor’s level, ISCED 7–8 = Master’s or Doctoral level. B : No parent > ISCED 2 = No parent with higher than lower secondary education, Father > ISCED 2 = Only father has above lower secondary education, Mother > ISCED 2 = Only mother has above lower secondary education, Both > ISCED 2 = Both parents have above lower secondary education.

b = unstandardized regression coefficient, S.E = standard error, ref. = reference group.

Age and income variables are centered on their respective means. All income measures are presented in 100,000 NOK.

EDI = Equivalized disposable income.

* p < 0.05

** p < 0.01

Maternal and paternal education

The OLS models with the parental education measure differentiating between families where either mother, father, or both had above ISCED 2 qualifications are displayed in Table 3 . The main findings from these models were that the associated reduction in GPA by having divorced parents was significantly larger if only the mother or both parents had above ISCED 2 qualifications, compared to if no parent had above ISCED 2 qualifications. If only the father had above ISCED 2 qualifications, however, no significant interaction with parental divorce was observed (see Fig 1B ).

To check whether the estimates from the above models were sensitive to the choice of combining the maternal and paternal educational levels into overall measures parental education, the analyses were re-run by entering paternal and maternal education as two separate and independent variables. These models revealed that the heterogeneity in the associations between divorce and GPA by parental education were driven by maternal educational levels; the estimated reduction in GPA by having divorced parents was statistically significantly higher when maternal educational levels were at secondary school levels (ISCED 3–5; b = -0.120, p < 0.05), at Bachelor’s levels (ISCED 6; b = -0.175, p < 0.05) and at Master’s or PhD-level (ISCED 7–8; b = -0.209, p < 0.05) compared to basic-level education (ISCED 0–2), after adjustments for paternal education and income measures. There were, however, no significant interaction effects between paternal educational levels and divorce on the adolescents’ GPA while holding the effects of maternal education and income constant (see Table 4 ; Fig 2A and 2B ).

The predicted values of GPA by ( A ) maternal educational qualifications, and ( B ) paternal educational qualifications from the fully adjusted regression models, stratified by divorce status (cf. Table 4 ). Error bars represent 95% confidence intervals of b . ISCED 0–2 = up to lower secondary education, ISCED 3–5 = upper secondary education, post-secondary non-tertiary education, short-cycle tertiary education, ISCED 6 = Bachelor’s level, ISCED 7–8 = Master’s or Doctoral level.

Alternating the reference categories of the parental education variables in the regression analyses did not reveal any further statistically significant differences in the links between divorce and GPA by parental educational qualifications (i.e., the main differences were between the ISCED 0–2 levels and the other ISCED levels).

Robustness checks

Entering maternal and paternal educational levels in separate models, in order to check for overcontrol bias, yielded approximately identical estimates. The only exception was that the difference in the relationship between divorce and GPA was slightly smaller and not statistically significantly different (at p < 0 . 05 ) between the highest maternal educational levels (ISCED 7–8) compared to the lowest maternal educational levels (ISCED 0–2) in the interaction analyses ( b = -0.163, p = 0.096).

Adolescents with highly educated parents experienced, on average, that their parents divorced somewhat later (see Fig 3A and 3B ). The mean difference in years since divorce among highly educated (i.e., ISCED 7–8) vs. lowly educated (i.e., ISCED 0–2) mothers was about 2.2 years, while the comparable figure among fathers was 2.8 years. Plotting the adolescents’ GPA as a function of years since divorce across the parental educational qualifications (see Fig 4 ) revealed a slight negative linear association between years since divorce and GPA across most of both maternal and paternal educational levels. The negative association between time since divorce and GPA was strongest among lowly educated mothers. As lowly educated mothers on average had most years since divorce, this finding highlights that time since divorce could not explain the heterogeneity in the associations between divorce and the adolescents’ GPA by maternal educational qualifications. Indeed, the plot suggests that holding years since divorce constant across maternal educational qualifications would slightly strengthen the difference in the negative association between divorce and GPA among adolescents with highly- compared to lowly educated mothers.

Pairwise comparisons of years since divorce across maternal ( A ) and paternal ( B ) educational qualifications. The plots comprise a mix of a violin plot (displaying the shape of the variable distribution) and a box plot (where the box is split by the median and bounded by the first and third quartiles of the distribution) along with the jittered raw datapoints. The red dot signifies mean values, also reported in text as μ ^ with accompanying 95% confidence intervals ( CI 95% ). Only significant pairwise comparisons are shown with accompanying p-values.

This figure shows the associations between years since divorce and the adolescents’ GPA by maternal ( A ) and paternal ( B ) educational qualifications. The blue lines represent the linear association, while the smoothed dotted black lines stem from GAMs The shaded area represents the 95% confidence intervals. Pearson's product-moment correlations of the linear associations are displayed in the top left quadrants with asterisks denoting statistically significant associations (* < 0.05, ** <0.01).

The plotted GAM curves show some variability around the linear functions for some of the parental educational levels. Overall, these trends do not give any strong indications that GPA is highly influenced by the timing of divorce in the present study.

Lastly, using equivalized disposable income (EDI) as an alternative indicator of socioeconomic resources, we found a similar but weaker pattern whereby the negative association between divorce and GPA was relatively stronger among adolescents in the second income quartile (Q 2 ; b = - 0.16, p < 0.01) and in the fourth quartile (Q 4 ; b = -0.15, p = 0.02) compared to those in the first quartile (Q 1 ). The difference between Q 1 and Q 3 was not statistically significant (see S1 Fig with further test statistics). Adjusting the analyses for parental educational qualifications attenuated and removed the significant difference between Q 1 and Q 4 . No heterogeneity in the associations between divorce and GPA by mother’s or father’s net income were found (results not shown).

Of note, it is highly likely that the potential heterogeneity by measures of household income in the links between divorce and academic outcomes is sensitive to how income is operationalized. As parental education was the main focus of interest in the present study, we did not examine this any further in the present paper (e.g., other ways of dividing income into categories).

The purpose of this study was to investigate the possible existence of heterogeneity in the association between divorce and adolescents’ academic achievement within the context of an elaborate welfare state such as Norway. As expected, adolescents with divorced parents had on average lower GPA compared to their peers with nondivorced parents. This difference was robust and only moderately reduced (from 0.30 to 0.24 points, about 20%) after adjustments for parental education. This finding fits well with previous studies [ 2 , 4 , 6 ], and suggest that divorce is associated with poorer school performance also among Norwegian adolescents.

The negative association between divorce and GPA was relatively stronger among adolescents where at least one of the parents had educational qualifications equivalent to upper secondary education or a bachelor's degree, compared to families without any educational qualifications. Further analyses revealed that this heterogeneity was primarily driven by maternal educational levels, whereby having divorced parents was more strongly and negatively associated with the GPA among adolescents with educated mothers (i.e., above ISCED 2) than among adolescents with less educated mothers (i.e., ISCED 0–2), after holding the effects of paternal education and income constant.

Overall, our findings thus lend support to the floor effects hypothesis [ 21 ], suggesting that the negative associations between divorce and adolescents’ academic achievement are relatively stronger among adolescents with educated parents, compared to those with less educated parents. Similar floor effects have been reported in terms of youths’ later educational attainment [ 16 , 21 , 22 ]. Our results also align with the study by Martin [ 23 ], which found that the link between divorce and subject grades was relatively stronger among adolescents with educated mothers compared to less educated parents. Other studies have, however, reported either no heterogeneity in GPA according to parental education [ 21 ], or a compensatory advantage of having educated parents on current school performance [ 17 , 20 ].

The heterogeneity in the associations between parental divorce and adolescents' GPA observed in the present study stem from two related findings: Firstly, a divorce was hardly related to the GPA among adolescents with two uneducated parents. Secondly, among adolescents with educated or highly educated mothers, a relatively larger negative association between divorce and the adolescents’ GPA was observed. We offer the following interpretation of how this pattern may come about within the Norwegian context:

The Norwegian population is highly educated, and families where both parents have low levels of educational qualifications are relatively rare. Social gradients in health and education are nevertheless well established also in Norway [ 59 , 60 ], and socioeconomically disadvantaged families have more frequent experiences of negative life events and family stresses (e.g., stress related to unemployment, work, and housing) including higher rates of marital dissolution than more affluent families [ 34 , 61 ]. The on average higher levels of family instability experienced by children in uneducated families might suggest that these children have come to expect adverse events in their lives. A divorce might thus be but one of several potential adversities experienced during childhood, rendering the independent effect of divorce less severe [ 24 ]. Moreover, the expected school performance among adolescents with parents with low levels of education are, on average, already weak to begin with. Thus, there is less room for their grades deteriorating further as a consequence of divorce. It is also conceivable that the elaborate welfare systems in Norway effectively buffer against further financial strain following divorce among less educated families, perhaps partly because there is less potential for their economic situation to worsen any further. Thus, although less educated families are more likely to experience stress related to poor family finances, a divorce might not exacerbate their financial situation.

We found that the negative association between divorce GPA was relatively larger among adolescents with educated compared to less educated mothers. Martin [ 23 ] reported a similar finding in the U.S., which in turn was partly explained by children of educated divorced/single mothers not receiving similar levels of positive parenting practices as peers with educated married parents. Such mechanisms were not explored in the current study. Thus, we can only speculate to whether they also apply to the Norwegian context. However, this explanation may fit the notion of a “double burden” experienced by educated, divorced single mothers in Norway, due to the strain of high workload combined with child-rearing responsibilities [ 37 , 38 ]. As it takes time and effort to engage in parenting practices that foster academic skills in children, educated divorced mothers are perhaps less able to continue providing such parenting practices to the same extent after the divorce relative to their equally educated married counterparts. Less educated mothers may, however, not have spent as much time fostering such skills to begin with.

Moreover, as the divorce might come as more of a shock among both parents and children in educated families [ 24 ], school-promoting activities might be more offset as the family tries to adjust to post-divorce family life.

Adjusting the analyses by measures of parental income hardly changed the estimated interactions between divorce and parental educational levels on the adolescents’ GPA. This contrasts the findings of Bernardi and Boertien [ 21 , 22 ], which found that the link between divorce and adolescents’ educational attainment among youth with highly educated parents was driven by the loss of access to father’s financial resources following divorce. These studies were, however, conducted on a cohort born in 1970 in Britain, a context with higher levels of single mother poverty and where divorced fathers often failed to pay child support, as noted by the authors [ 22 ]. The Norwegian welfare state, on the other hand, appears to be rather successful in equalizing the cost of divorce among men and women [ 35 , 36 ]. Hence, the departure of an educated father after a divorce may arguably be less economically detrimental to divorced mothers in Norway. Moreover, income is likely more important when considering later educational attainment than current school performance in countries where admission to higher education is costly. It is important to stress that our findings do not suggest that income is irrelevant to children’s post-divorce adjustment, and indeed, single mothers are also in Norway among the least well of in society. Instead, the results of this study suggest that other mechanisms might be more important in explaining the observed heterogeneity in the associations between divorce and GPA by parental educational qualifications.

Our findings were robust to adjustments of age, gender, and current income measures. Adolescents with highly educated parents experienced, on average, that their parents had divorced somewhat more recently. Time since divorce was overall weakly and negatively associated with the adolescents’ GPA, with some small observable variations across parental educational qualifications. These variations could, however, not explain the heterogeneity found in the present study. Overall, this is generally in keeping with studies that find that youth post-divorce adjustment is quite stable or gradually slightly worsen as time passes since the divorce [ 62 , 63 ].

Some previous studies have adjusted their analyses by different pre-divorce characteristics such as the child’s behavioral problems, cognitive abilities, and material resources [ 21 ], the child’s psychological well-being, academic ability, mothers’ distress, and their family’s pre-divorce economic resources [ 20 ], or utilized family fixed effects models [ 17 ]. Unfortunately, besides parental education, which is usually established before a divorce, other pre-divorce measures were not available in the current study. It is plausible that the differential inclusion of pre-divorce characteristics may account for parts of the diverging results. Other differences between the studies, such as the operationalization of dependent variables, differences in age groups, and cross-national differences, are also likely of importance. Compared to previous studies examining cohorts from the 1970s - 1980s [ 20 , 21 , 23 ], the current study is nonetheless unique by being based on a relatively recent cohort (born 1993–1995, and assessed in 2012). Due to the elaborate social welfare systems in Norway, where the school system is highly subsidized, higher education is common, and gender equity is high [ 44 , 64 ], direct comparisons with studies utilizing older samples from other countries should be made with care.

The following limitations of the current study should further be acknowledged. Firstly, due to the cross-sectional structure of our study, we have had no means of investigating potential changes from pre- to post-divorce family life. Proposed explanations such as changes in parenting among highly educated parents following divorce have, therefore, not been examined. Hence, this study is largely descriptive. Similarly, we did not have data to control for further differences between divorced and nondivorced families, which may induce selection effects [ 1 ]. For example, historical information that could both increase parents’ inclination to divorce, and possibly affect the adolescents’ GPA differently according to parental educational levels (e.g., mental health problems or levels of parental conflict) could be of importance and might explain the larger educational penalty observed in families with highly educated parents [ 22 , 65 ]. Of note, studies that have statistically adjusted their analysis by pre-divorce characteristics often find that it weakens the associations between divorce and educational outcomes [ 43 ]. The present study might thus overestimate the link between divorce and GPA. Nonetheless, divorce is generally understood as a process that gradually unfolds, rather than being a discrete point in time [ 1 ]. Adjustments of pre-divorce characteristics are thus not without problems either, as it may remove some of the effects that are intrinsically linked to the divorce process. Adjustments of pre-divorce characteristics should, therefore, be made with care [ 1 , 66 ].

The present study focused on the distinction between divorced and nondivorced families, whereas an investigation of other family structures or arrangements (e.g., single father families, stepfamilies, joint physical custody) were outside the scope of this study. Adolescents’ mental health and school engagement are found to vary across family structures [ 67 – 69 ]. It would be interesting for future studies to conduct a more detailed investigation of whether heterogeneous outcomes of divorce by parental education also depend on the post-divorce family structure. The findings of the present study might also depend on the adolescents’ gender, an issue that warrants further investigations.

Lastly, the participation rate in the youth@hordaland study was 53%, and the sample in the current study was further reduced to 47% of the total invited population. Previous investigations have found that the GPA in this sample were quite similar to and not statistically significantly different from both the regional and national averages [ 46 ]. Nevertheless, non-response is known to be related to lower socioeconomic status, and an earlier study on the former waves of the Bergen Child Study (which the youth@hordaland is nested within) found more psychological problems among those not participating [ 70 ]. This could limit the generalizability of the findings.

Despite these limitations, some strengths of the current study deserve mentioning. We utilized high-quality register-based information about the adolescents’ GPA, parental educational levels, and household income, which was merged with a large population-based study. Whereas previous studies have calculated the GPA by a subset of either subject grades, test scores, or exams, sometimes self-reported, the measure of GPA utilized in the current study is calculated from all graded subjects during a whole school year. As the measure of GPA used in the present study forms the primary basis for admittance into higher education in Norway, it may be considered highly reliable.

To conclude, the present study found that the association between parental divorce and adolescents' GPA is robust also within a Norwegian context. However, divorce was hardly associated with GPA among adolescents whose parents have low educational qualifications. In contrast, among adolescents from families with educated or highly educated mothers, parental divorce was associated with a lower GPA. These findings were robust to adjustments of measures of household income. Future studies are needed to investigate potential mechanisms (such as reduced parental monitoring or school-involvement), which might drive this finding.

The generalizability of these findings might be limited to a Norwegian context, as differences in both school systems and policies across nations may play an essential part in how parental divorce and parental education might affect adolescents' academic performance. Due to diverging results among existing studies examining this phenomenon, there is a need for future studies that can shed further light on the complex interactions between divorce, parental education, and outcomes among youths.

Supporting information

Funding statement.

This project has been made possible by the Norwegian ExtraFoundation for Health and Rehabilitation. SAN received funding (Grant number: 2017/FO149543)

Data Availability

• PLoS One. 2020; 15(3): e0229183.

Decision Letter 0

21 Oct 2019

PONE-D-19-20535

Divorce and adolescent academic achievement: The moderating role of parental education

Dear PsyD Nilsen,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Thank you very much for your patience. I really enjoyed reading your paper, as it brings many exciting results, and I am sure that it will enrich the current knowledge on the topic investigated in your study. 

Also, both reviewers found your paper interesting and acknowledge its potential. The reviewers are not entirely convinced that your analysis fully supports all your claims (particularly Reviewer 1). Reviewer 1 further suggests an alternative framework for capturing the effects of parental divorce on test scores. Could you please consider their suggestions? 

I further believe that addressing the rest of the reviewers' comments should not be too challenging. 

We would appreciate receiving your revised manuscript by Dec 05 2019 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

• A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.
• A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.
• An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Tomáš Želinský, Ph.D.

Academic Editor

Journal Requirements:

1. When submitting your revision, we need you to address these additional requirements.

Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. You indicated that you had ethical approval for your study. In your Methods section, please ensure you have also stated whether you obtained consent from parents or guardians of the minors included in the study or whether the research ethics committee or IRB specifically waived the need for their consent.

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Partly

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: No

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Dear authors,

Thank you for the opportunity to read and comment on your work. I will proceed by addressing the recommended points mentioned in PLOS reviewer guidelines.

What are the main claims of the paper and how significant are they for the discipline?

The paper claims to offer evidence of heterogeneous effects of parental divorce on adolescent GPA. This is a relevant question that has clearly connections to recently published peer reviewed work.

Are the claims properly placed in the context of the previous literature? Have the authors treated the literature fairly?

The claims are properly placed, and I feel that the authors treated the literature fairly.

Do the data and analyses fully support the claims? If not, what other evidence is required?

-I do not believe that the current analysis offers the evidence needed to support the stated claims. The present analysis interacts a binary indicator of parental divorce experience with various measures of parental education in OLS regressions predicting mean GPA of children. The results successfully show that the difference in GPA between kids from surviving marriages and kids of divorce tends to be larger when mothers are more educated. While I agree with this result, I do not believe that this is the result that you need to support your claim.

When considering the effects of parental divorce on test scores, I conceptualize a framework where a student has produced some GPA history up until divorce occurs at time t=0. The disruption occurs at time zero, and in subsequent periods the GPA realization may be heavily affected by the disruption, but decreasingly affected over time until the effect dies or becomes undetectable. In this framework, parental divorce should have a measurable effect on GPA in a limited time period after the event. You would then need to show that the negative disruption relative to the t≤0 levels is larger for kids with more educated mothers/parents to support your point. This type of analysis would require data on the timing of divorce, as well as some type of fixed effects model that produces estimates using within person variability in GPA between pre- and post-divorce periods.

This approach would be feasible if your two divorce indicator questions were asked consistently from year to year of your longitudinal survey. If this is the case, you can easily write code that identifies the year in which kids experience divorce allowing the formation of a clear partition between pre- and post-divorce GPA observations.

-I also have a question about the timing of study events. Were kids asked about divorce experience before, or after the GPA observations. If kids were asked about divorce experience before GPA realizations, is it possible to confirm that there are not kids in the non-divorce group who divorce during GPA observations? If GPA is observed before posing the divorce questions, how do you know which observations occurred before the divorce and which occurred after?

-Can you say anything about differences in the timing of divorce between more and less educated couples? I would expect more educated couples to divorce later, on average. If this is the case, the larger effects for more educated parents might be catching divorce experiences that are closer to the GPA measurement dates. This would suggest an alternate story where effects may not be stronger for more educated families. Instead, we may be measuring effects for more educated families when effects are largest, and measuring effects for less educated families when effects are already subsiding.

-This project seems somewhat limited by the OLS modeling choice, with the constraints of your data. The OLS approach likely over-estimates divorce effects by adding co-linear effects of the unobserved factors that lead to divorce. You would need a very rich set of controls to convincingly isolate the divorce effect in this framework. Unfortunately, your data do not appear to offer richer controls that would add more precision to the divorce estimates.

PLOS ONE encourages authors to publish detailed protocols and algorithms as supporting information online. Do any particular methods used in the manuscript warrant such treatment? If a protocol is already provided, for example for a randomized controlled trial, are there any important deviations from it? If so, have the authors explained adequately why the deviations occurred?

If the paper is considered unsuitable for publication in its present form, does the study itself show sufficient potential that the authors should be encouraged to resubmit a revised version?

I do not know. I believe that the authors would need to write a different paper with a different statistical analysis to answer their stated question. At the same time, I feel that the work that they have done thus far is interesting, useful, and publishable if the results were directed toward a different component of their original question.

I see more than one way of doing this, but my immediate suggestion would be to follow through more fully on your interest in understanding how the Norwegian context matters for these effect estimates. Given that your OLS model is limited in its ability to isolate divorce effects, compare your estimates to other work that faces the same challenge and make an empirically motivated theoretical contribution about the relevance of national and cultural context in determining divorce effects after comparing and contrasting your estimates with results from comparable studies in different contexts.

Are original data deposited in appropriate repositories and accession/version numbers provided for genes, proteins, mutants, diseases, etc.?

I believe so.

Are details of the methodology sufficient to allow the experiments to be reproduced?

Is the manuscript well organized and written clearly enough to be accessible to non-specialists?

Reviewer #2: Thank you for giving me an opportunity to read your paper. I think it is a strong study. In particular, I find the discusssion section strong. Nevertheless I have some comments, which I hope you find useful.

1. As the authors note themselves on page 27 their study is descriptive. Therefore any causal language (effects, impacts, penalty, etc.) should be avoided. The same goes for the "moderating" role of parental education in the title. Parental education could also pick up the effect of some other, unobserved variable. It is quite easy to think that socioeconomic differences in the effects of divorce/ separation on children's school grades may be very different from socioeconomic differences in the associations between divorce/ separation and children's school grades. (I also do not thin that proposensity score approaches should be called causal as the authors do on page 3; the crucial step in causal analysis is to control for selection on unobserved variables.)

2. The paper uses survey data on Norway, a country in which register data is available and has been previously used (e.g., Sigle-Rushton et al., 2014). Why not use data on the whole population?

3. It would be nice to see some robustness checks with other indicators of social origin, e.g. parental occupation, income, and wealth. Also, what happens if the reference category of parental education is changed? Currently it seems as differences are mainly due to ISCED 0-2 vs. ISCED 3-7 (Model 2 in Table 2). As the authors note on page 24, the ISCED 0-2 is a disdvantaged group. Therefore, ISCED 3-4 may be a better choice for a reference category.

4. Condition on household income introduces overcontrol bias (model 3). The authors may discuss this issue. (Although I understand as well that the results are rarely affected by this control/ mediator.) There is also a problem of overcontrol bias when conditioning on paternal and maternal education simultaneously (Model 2 in Table 4). Can the authors enter only maternal education and then only paternal education in one set of models and do they get the same results?

5. Did the authors look at gender differences in the associations?

Smaller points:

- The authors study the dissoultion of married and non-married couples. Maybe therefore better to use the term separation throughout the whole study? The authors note on page 13 that the research practice is to use divorce but it does not strike me too be a good practice.

- Given that grades are obtained in different school years, may it not make sense to rank or standardize the variable within school years?

- The authors should add sample sizes to all tables and figures.

6. PLOS authors have the option to publish the peer review history of their article ( what does this mean? ). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy .

Reviewer #1: Yes: Ravaris Moore

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/ . PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at gro.solp@serugif . Please note that Supporting Information files do not need this step.

Author response to Decision Letter 0

11 Nov 2019

Author Response to Reviewers’ Comments

Journal: PLOS ONE

Manuscript ID: PONE-D-19-20535

Dear Tomáš Želinský, Ph.D.

Academic Editor, PLOS ONE

Thank you for allowing us to revise and resubmit our paper. Below are the comments from the reviewers (numbered) and our responses. We have uploaded one version of the manuscript where these changes are identified using track changes (Revised Manuscript with Track Changes”) and one version that is clean (Manuscript).

We thank you for your consideration of our revised manuscript and responses, and we look forward to hearing from you.

Sondre Aa. Nilsen

(on behalf of all authors)

Response to Editorial comments:

1. Thank you very much for your patience. I really enjoyed reading your paper, as it brings many exciting results, and I am sure that it will enrich the current knowledge on the topic investigated in your study. Also, both reviewers found your paper interesting and acknowledge its potential. The reviewers are not entirely convinced that your analysis fully supports all your claims (particularly Reviewer 1). Reviewer 1 further suggests an alternative framework for capturing the effects of parental divorce on test scores. Could you please consider their suggestions? I further believe that addressing the rest of the reviewers' comments should not be too challenging.

Response: We thank the editor for these positive comments and for sharing our enthusiasm with regards to this research topic. We have made several changes to the manuscript in response to the suggestions made by the reviewers. Their comments and our point-by-point responses can be found below.

Response to Comments from Reviewer 1

We thank Reviewer 1 for the evaluation of the manuscript, and for providing suggestions for improvements.

Reviewer comments:

1. What are the main claims of the paper and how significant are they for the discipline?

Response: Thank you for this comment.

2. Are the claims properly placed in the context of the previous literature? Have the authors treated the literature fairly?

Response: Thank you for stating that our claims are properly placed, and that we have treated the literature fairly.

3. Do the data and analyses fully support the claims? If not, what other evidence is required?

a. -I do not believe that the current analysis offers the evidence needed to support the stated claims. The present analysis interacts a binary indicator of parental divorce experience with various measures of parental education in OLS regressions predicting mean GPA of children. The results successfully show that the difference in GPA between kids from surviving marriages and kids of divorce tends to be larger when mothers are more educated. While I agree with this result, I do not believe that this is the result that you need to support your claim.

Response: We fully agree with the reviewer that due to limitations with our data (i.e., the lack of repeated/longitudinal measures of GPA in the years before and after the divorce), we have had no means to try to approximate the causal relationship between divorce, parental educational qualifications, and the adolescents’ GPA. We do see that our manuscript contains some usage of causal language, which clearly should be avoided (as also pointed out by Reviewer 2, point 1). We have therefore made changes to the title and throughout the manuscript in order to stress that this is a descriptive association study, and that any offered explanations for our findings are based on what we consider theoretically or empirically relevant information from other studies. We have made it clearer throughout the manuscript that the main aim of the present study was to examine potential heterogeneity in the associations between parental divorce and adolescents’ GPA by parental educational qualifications and also note the descriptive nature of our study in the limitations section (page 28, lines 594 - 595).

The title now reads:

b. When considering the effects of parental divorce on test scores, I conceptualize a framework where a student has produced some GPA history up until divorce occurs at time t=0. The disruption occurs at time zero, and in subsequent periods the GPA realization may be heavily affected by the disruption, but decreasingly affected over time until the effect dies or becomes undetectable. In this framework, parental divorce should have a measurable effect on GPA in a limited time period after the event. You would then need to show that the negative disruption relative to the t≤0 levels is larger for kids with more educated mothers/parents to support your point. This type of analysis would require data on the timing of divorce, as well as some type of fixed effects model that produces estimates using within person variability in GPA between pre- and post-divorce periods.

Response: Thank you for this insightful comment. We agree that this would be a feasible way to analyze longitudinal data with repeated measures of GPA and information about the timing of divorce, in order to come closer to establish a causal relationship. This would be a very interesting study to conduct. To the best of our knowledge, very few studies investigating heterogeneity in the link between divorce and adolescents’ GPA by parental education have had such rich longitudinal data available (although studies have to various degrees been able to control for some pre-divorce characteristics). Unfortunately, our study is also limited in this sense. The youth@hordaland study, is a cross-sectional study that was merged with register-based information on parental education and subject grades obtained during the school year 2011-2012. As such, we had no means to track adolescents’ GPA from pre- to post-divorce times (as noted on page 27 – 28, lines 591-594).

To make this more explicit, we have updated both the Abstract and the Procedure sections of the manuscript, to highlight that this study draws on cross-sectional data.

c. -I also have a question about the timing of study events. Were kids asked about divorce experience before, or after the GPA observations. If kids were asked about divorce experience before GPA realizations, is it possible to confirm that there are not kids in the non-divorce group who divorce during GPA observations? If GPA is observed before posing the divorce questions, how do you know which observations occurred before the divorce and which occurred after?

Response: The youth@hordaland study was conducted during spring of 2012, while the GPA was calculated by all grades obtained during the first term of the school year 2011/2012. Although possible, we find it unlikely that many experienced that their parents divorced during this short period, at least not in sufficient number to lead to any substantial change in the divorce – non-divorced categorization.

As shown in Table 1, the mean year since experiencing parental divorce was 10.58 years. Only16 respondents stated that their parents had divorced during the current school year. The rest reported that at least one year had gone since their parents divorced. As further elaborated in our response to reviewer comment 3d, a very weak relationship between the timing of divorce and the adolescents’ GPA was found, and accounting for the year of parental divorce did not alters the link between parental educational qualifications and the adolescents’ GPA among adolescents with divorced parents.

d. -Can you say anything about differences in the timing of divorce between more and less educated couples? I would expect more educated couples to divorce later, on average. If this is the case, the larger effects for more educated parents might be catching divorce experiences that are closer to the GPA measurement dates. This would suggest an alternate story where effects may not be stronger for more educated families. Instead, we may be measuring effects for more educated families when effects are largest, and measuring effects for less educated families when effects are already subsiding.

Response: Thank you for raising this interesting point. Indeed, adolescents with educated parents, did, on average, report that their parents had somewhat more recently divorced than less educated parents. However, most adolescents generally report quite some time since parental divorce, which is not too surprising given that we examine rather old adolescents aged 16 – 19. If considering the highest completed education in the family variable (i.e., either by the mother or father), mean years since divorce was 9.1 years for adolescents from the most educated families (i.e.,ISCED 7-8), while mean years since divorce was 12.18 for adolescents with the least educated parents (i.e., ISCED 0 -2). For those with highest education equivalent to ISCED 3-5 and ISCED 6, the average years since divorce was 10.58 and 10.23, respectively (Significant differences found between all educational levels except ISCED 3-5 and ISCED 6 in years since divorce).

In our data, we find a very weak and slightly curvilinear relationship between time since divorce and the adolescents GPA; the GPA rises slightly from 0 to about 8 years since divorce, then gradually declines again from 8 to 19 years since divorce. Overall, this trend nonetheless suggests that the lower GPA among adolescents with divorced compared to nondivorced parents appear rather stable irrespectively of time since divorce. This finding is in general keeping with previous literature suggesting that children’s adjustment following divorce is rather stable but somewhat worse than their peers with non-divorced parents [see e.g., 1].

1. Härkönen J, Bernardi F, Boertien D. Family dynamics and child outcomes: An overview of research and open questions. Eur J Popul. 2017;33

Adding time since divorce in subgroup analyses on adolescents with divorced parents did not in any substantial way alter the link between parental educational qualifications and the adolescents’ GPA. Thus, time since divorce does not appear to account for the heterogeneity in the associations between divorce and GPA by parental education found in the present study. We thank the reviewer for raising this interesting point, and we have added an extra paragraph to the statistics section:

The timing of divorce could potentially covary with parental educational levels and the adolescents’ GPA (e.g., if highly educated parents divorced later on, the estimates of divorce by parental education on the adolescents’ GPA could pick up on the proximity to the event of dissolution). Additional checks were therefore made to investigate whether the estimates were influenced by the timing of divorce (as measured by years since the divorce).

And added some extra results to the result section:

Adolescents with highly educated parents experienced, on average, that their parents divorced somewhat later (mean years since divorce across highest parental education in the family; ISCED 7-8 = 9.1 years, ISCED 6 = 10.2 years, ISCED 3-5 = 10.6 years, ISCED 0-2 = 12.2 years). Subgroup analyses showed a very weak and slightly curvilinear association between years since the divorce and the adolescents GPA; the GPA rose slightly from 0 to about 8 years since divorce, then gradually declined from 8 to 19 years since divorce. Overall, this trend nonetheless suggested considerable stability in the negative association between divorce and adolescents’ GPA. The association between parental education and GPA among adolescents with divorced parents hardly changed when adjusted by years since divorce (results not shown).

e. -This project seems somewhat limited by the OLS modeling choice, with the constraints of your data. The OLS approach likely over-estimates divorce effects by adding co-linear effects of the unobserved factors that lead to divorce. You would need a very rich set of controls to convincingly isolate the divorce effect in this framework. Unfortunately, your data do not appear to offer richer controls that would add more precision to the divorce estimates.

Response: We agree that our data limits our ability to draw causal conclusions, and that one would need a very rich set of controls to isolate the divorce effects. The fact that the OLS approach might over-estimate the link between divorce and the adolescents’ GPA have been acknowledged as a limitation in the manuscript (page 28, line 603).

f. I believe that the authors would need to write a different paper with a different statistical analysis to answer their stated question. At the same time, I feel that the work that they have done thus far is interesting, useful, and publishable if the results were directed toward a different component of their original question.

g. I see more than one way of doing this, but my immediate suggestion would be to follow through more fully on your interest in understanding how the Norwegian context matters for these effect estimates. Given that your OLS model is limited in its ability to isolate divorce effects, compare your estimates to other work that faces the same challenge and make an empirically motivated theoretical contribution about the relevance of national and cultural context in determining divorce effects after comparing and contrasting your estimates with results from comparable studies in different contexts.

Response: Thank you for stating that our work is interesting and useful, and for providing further suggestions on how to structure this manuscript. We have updated the aims of the study to highlight that this study sought to investigate heterogeneity in the associations between divorce and GPA by parental educational levels, and made several changes to the manuscript including the title and the abstract, to stress that this is a descriptive, associative study. Finally, we have also done our best to contrast our findings against previous studies in terms on different national and cultural context,

Reviewer #2: Thank you for giving me an opportunity to read your paper. I think it is a strong study. In particular, I find the discussion section strong. Nevertheless I have some comments, which I hope you find useful.

Response: We thank Reviewer 2 for the thorough evaluation of our manuscript, the positive feedback stating that this is a strong study, and for providing several constructive means to improve the manuscript.

1. As the authors note themselves on page 27 their study is descriptive. Therefore any causal language (effects, impacts, penalty, etc.) should be avoided. The same goes for the "moderating" role of parental education in the title. Parental education could also pick up the effect of some other, unobserved variable. It is quite easy to think that socioeconomic differences in the effects of divorce/ separation on children's school grades may be very different from socioeconomic differences in the associations between divorce/ separation and children's school grades. (I also do not think that propensity score approaches should be called causal as the authors do on page 3; the crucial step in causal analysis is to control for selection on unobserved variables.)

Response: Thank you for highlighting this, we fully agree that all causal language should be avoided. We have now made changes throughout the manuscript to address this. We have removed the “propensity score” part in the introduction so to not give the impression that propensity score matching sufficiently addresses causality. We have also updated the title of the manuscript which now reads:

Response: This study is part of a larger project financed to investigating outcomes of divorce among adolescents utilizing data from the youth@hordaland study. We managed to link the youth@hordaland to official registries, and saw the possibility to capitalize on this linkage in order to investigate heterogenous associations between divorce and adolescents’ GPA by parental educational qualifications. We agree that it would be very interesting to get hold of data of the entire population. In this study, we have, however, been limited to the data at hand from the youth@hordaland study.

Response: Unfortunately, we did not have information about parental occupation or wealth that would have enabled us to investigate heterogeneity in the associations between divorce and adolescents’ GPA on these indicators. The existing literature has primarily focused on parental education as a measure of social origin, and our aim with this study was to continue this research tradition within a Norwegian context. We agree that it would be interesting to examine this further with other indicators of social origin in future studies.

Alternating the reference category of the parental educational variables highlight that the significant differences lies between ISCED 0-2 and the other educational categories. Thank you for noticing that we have not explicitly mentioned this in the manuscript. We have added information about this in the result section:

Alternating the reference categories of the parental education variables in the regression analyses did not reveal any further statistically significant differences in the links between divorce and GPA by parental educational qualifications (i.e., the main differences were between the ISCED 0-2 levels and the other ISCED levels).

Response: We thank the reviewer for raising this interesting point. We agree that conditioning on household income could produce overcontrol bias when entered simultaneously with parental education in the same model. Controlling for maternal and paternal education might be problematic for similar reasons, as these are usually dependent (e.g., through “assortative mating”).

We have checked whether we get the same results by entering maternal education and paternal education separately. The results from these analyses are very similar to those reported in the paper. The only notable difference is that maternal education at ISCED 7-8 is not statistical significantly different compared to the ISCED 0-2 level (b = -0.163, p = 0.096), although the estimate is only slightly weaker than when paternal educational qualifications is also added to the model (by 0.04 GPA points).

We have made the following changes in the statistics section in order to address the issue of overcontrol bias:

Conditioning on measures of income and paternal and maternal education simultaneously may introduce overcontrol bias [51]. In the two first set of regression analyses, we try to avoid this problem by creating single measures combining information on parental education from both parents, and by entering income variables in the last set of models (as we were not interested in the main effects of income per se). In the last set of regressions (cf. Table 4), we have made robustness checks by entering maternal and paternal education in separate models.

We have also added some information in the result section:

Entering maternal and paternal educational levels in separate models yielded approximately identical estimates. The only exception was that the difference in the relationship between divorce and GPA was slightly smaller and not statistically significant different (at p < 0.05) between the highest maternal educational levels (ISCED 7-8) compared to the lowest maternal educational levels (ISCED 0-2) in the interaction analyses (b = -0.163, p = 0.096).

Response: We agree that it would be of interest to examine potential gender differences in the associations between divorce/separation on GPA by parental education. However, due to limited statistical power, we did not investigate this further in the current study. We see that gender differences remains largely unexplored within this research field, and it would be interesting for future studies to look into this, perhaps using register data with larger samples.

We have added a sentence in the limitations section of the paper to highlight that we have not investigated gender differences:

The findings of the present study might also depend on the adolescents’ gender, an issue we did not investigate due to power restrictions.

- The authors study the dissolution of married and non-married couples. Maybe therefore better to use the term separation throughout the whole study? The authors note on page 13 that the research practice is to use divorce but it does not strike me too be a good practice.

Response: We see that recent studies perhaps tend to utilize the term parental separation rather than divorce. One could argue that the term “parental separation” is not completely satisfying either, as it may be used to denote the status preceding a legal divorce (i.e., in several countries including Norway, one has to be separated a given time period before filing for a legal divorce).

This paper is part of a larger project whereby the term divorce is consistently used to denote couples who split up from marriage or cohabitation, and we therefore hope to keep this term also in this paper. We have been explicit in this paper with regards to how we operationalize divorce, and we hope that the reviewer might reconsider. If the reviewer feels very strongly about this point, we will nonetheless change the term to separation.

Response: We have added the adolescents’ age as a covariate in all regression models, a variable that indirectly captured the school year that the adolescents were in at the time of the study (at least for the vast majority of the adolescents). It would make more sense to rank/standardize the variable within school years. However, besides the adolescents age and the year (i.e., 2011/2012) of which the grades were given, we do not have information regarding which specific school year the adolescents were in at the time of the study. We are therefore unable to further rank the variable within school years.

Response: Thank you for noticing, sample sizes have now been added to all tables and figures.

Submitted filename: Response to Reviewers.docx

Decision Letter 1

15 Jan 2020

PONE-D-19-20535R1

Once again, I would like to thank you for your patience. Both referees were happy with your adjustments, and they only suggest minor revisions, which I believe, reflects the quality of the improvements you made. 

Both referees made a point related to the robustness checks, although from a different perspective. While Reviewer 1 would like to see whether adding "a control measure for time since separation/divorce" could affect your main findings, Reviewer 2 would like to see the results for an additional robustness check using household income as an alternative indicator of a family's socioeconomic resources. 

I believe that these suggestions will not affect the main results of your paper, and I will be happy to recommend your paper for acceptance. 

We would appreciate receiving your revised manuscript by Feb 29 2020 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: (No Response)

Reviewer #2: (No Response)

2. Is the manuscript technically sound, and do the data support the conclusions?

3. Has the statistical analysis been performed appropriately and rigorously?

4. Have the authors made all data underlying the findings in their manuscript fully available?

5. Is the manuscript presented in an intelligible fashion and written in standard English?

6. Review Comments to the Author

Reviewer #1: Dear Authors,

Thank you for your diligence in addressing comments from both reviewers. I hope you find the minor recommendations below helpful.

1. In the sample section you mention that your analytical sample is similar to the study's full sample in terms of age and gender distribution. It would be helpful to know if the samples are also similar in terms of parental education and income. If the data allow, please mention similarities and differences along other relevant dimensions of the data.

2. In the paragraph before the start of the "Statistic Analysis" section, you mention that the divorce group likely includes parents who split up after cohabiting. Please include some estimate from either national or regional data that gives the reader an idea as to what proportion of unions in this study were likely not marriages.

3. Your results table presents income measures in units of 1000NOK, while Table 1 describes income in units of 1NOK. You may want to present Table 1 income in units of 1000NOK as well to be consistent with results.

4. In the robustness checks you explore the correlation between parent's education and timing of divorce. You show an average difference of three years in the timing of divorce between the most and least educated parents. This translates into a difference between experiencing divorce around age 5 versus experiencing divorce around age 8. This difference may be important from a developmental framework. It would be worthwhile to control for these differences in your main regressions. Please add update your main models with a control measure for time since separation/divorce. The measure would equal zero for kids who did not experience parental divorce/separation, and it should equal time since event for kids who experienced divorce/separation. This should be a low cost update to models and figures.

Thank you for your work on this research.

Reviewer #2: I thank the authors for taking into account most of my comments and for further improving a already good study. However, I believe that they could conduct one robustness check. In response to my third point, the authors wrote that they do not have information on parental occupation and parental wealth. I see that they cannot then use these variables for a robustness check. However, they can use household income, a variable which is already present in their analysis as a control. I would like to see therefore at least a robustness check using household income as an alernative indicator of a family's socioeocnomic resources than parental education. It is precisely one shortocoming of the previous literature that they only used parental education and this study has the potential to address this shortcoming.

7. PLOS authors have the option to publish the peer review history of their article ( what does this mean? ). If published, this will include your full peer review and any attached files.

Reviewer #1: Yes: Ravaris L Moore

Author response to Decision Letter 1

30 Jan 2020

Manuscript ID: PONE-D-19-20535R1

Thank you again for allowing us to revise and resubmit our paper. Below are the comments from the reviewers (numbered) and our responses. We have uploaded one version of the manuscript where these changes are identified using track changes (Revised Manuscript with Track Changes”) and one version that is clean (Manuscript).

1. Once again, I would like to thank you for your patience. Both referees were happy with your adjustments, and they only suggest minor revisions, which I believe, reflects the quality of the improvements you made.

Both referees made a point related to the robustness checks, although from a different perspective. While Reviewer 1 would like to see whether adding "a control measure for time since separation/divorce" could affect your main findings, Reviewer 2 would like to see the results for an additional robustness check using household income as an alternative indicator of a family's socioeconomic resources.

I believe that these suggestions will not affect the main results of your paper, and I will be happy to recommend your paper for acceptance.

Response: Thank you for your availability and swift handling of this review process. We have made some additional changes based on the reviewer comments;

- Detailed associations of “time since divorce” on the adolescents’ GPA

- across parental educational levels

- Conducted additional robustness checks using household income

- We have also made some small changes based on reviewer 1 points 1-3, and fixed some typos and grammatical errors that we detected.

The reviewers’ comments and our point-by-point responses can be found below.

We thank the reviewer for the efforts of evaluating our manuscript a second time, and for providing constructive means to further improve the paper.

We agree that additional similarities/differences could be useful to add. Unfortunately, we only have register based information about parental income and education on the sample that consented to this register linkage (i.e. the sample that we investigate in this paper). We have, however, compared age, gender, and self-reported parental education and perceived economic well-being between the total sample and the sample consenting to the register linkage. We have uploaded this table as a supplementary file, as we think this information is best suited as a supplement to the paper. As shown in the table the samples were nearly identical across these dimensions between the two samples.

We have updated the sample section which now reads;

All adolescents born between 1993 and 1995 and residing in Hordaland at the time of the survey were invited (N = 19,439) to participate, and 10,257 agreed, yielding a participation rate of 53 % for the entire study. The present paper is based on a subsample of 9,166 adolescents (47 % of the invited population) who consented to register linkage. This subsample was nearly identical to the total sample with regards to age and gender distribution, and self-reported sociodemographics (see S1 Table).

Response: Thank you for this suggestion. We agree that additional information about marriage/cohabitation from regional data are useful to the paper. We have added the following based on available regional statistics to the paragraph just before the statistics section;

Official statistics from Hordaland county in 2012 report that 73.5 % of children and youth below the age of 18 in a two-parent household lived with married parents (the rest with cohabiting parents). Thus, the nondivorced group in the present study most likely contained a group of adolescents with parents that had cohabitated since their birth. As some cohabiting unions eventually marry, we find it likely that the proportion of cohabiting unions in the present sample was somewhat lower than regional estimates also including younger children. Similarly, the divorced group likely contained a group of adolescents whose parents split up from cohabitation. Unfortunately, no official statistics regarding dissolution from cohabiting unions in Norway exists. Our inability to exactly detail the adolescents’ family structure is not unique to the present study but has been rather common within this research field [2]

Response: Thank you for noticing this inconsistency. We have updated Table 1 so that the units match.

Response: Thank you for this interesting suggestion. We have spent quite some time debating how adding a measure that equals zero for kids who do not divorce, and time since event for those who do divorce, may affect our models – and especially the interpretations of the models. Firstly, - we have checked this following your advice. The results when adjusting for this variable slightly strengthens all our main findings (i.e., makes the heterogenous effects slightly larger (by the second decimal point)), and do not change the interpretation of the significance level of any of the reported estimates. However, although this might be a mathematically sound method of adding an adjustment to a single level of a dichotomous categorical predictor in a regression (i.e., the “divorced-level” of the “divorce status” variable”), this also appear to permit calculations of impossible/paradoxical values based on the coefficients reported. For instance, the estimates of having non-divorced parents who divorced four years ago.

We therefore prefer to not add this in our main reported analyses as it does not alter the main findings, and may induce some confusions to the interpretations of the estimates. We have, however, conducted further checks of the links between time since divorce, parental educational qualifications, and the adolescents’ GPA. We believe that adding this to the paper is perhaps more illuminating to the question of whether time since divorce is related to the adolescents’ GPA, and whether time since divorce potentially could influence our findings given the slight differences in time since divorce across parental educational qualifications.

We have described this in detail in the statistics section;

The timing of divorce could potentially covary with parental educational levels and the adolescents’ GPA (e.g., if highly educated parents divorced later on, the estimates of divorce by parental education on the adolescents’ GPA could be influenced by the proximity to the event of dissolution). Moreover, it is possible that associations between timing of divorce and the adolescents’ GPA depend on parental educational qualifications (i.e., that more time spent with highly educated divorced parents differ to time spent with lowly educated divorced parents). We investigate these issues by comparing years since divorce across parental educational levels, and by graphically plotting potential linear and non-linear relationships between timing of divorce and GPA by parental educational qualifications. Generalized additive models (GAMs) were used to investigate potential non-linear relationships. In brief, GAMs may be considered as a semi-parametric extension of the generalized linear model, with the strength of the ability to detect non-linear structures in data that otherwise might be missed [52].

Added a paragraph to the results-section;

Adolescents with highly educated parents experienced, on average, that their parents divorced somewhat later (see Figs 3A and 3B). The mean difference in years since divorce among highly educated (i.e., ISCED 7-8) vs. lowly educated (i.e., ISCED 0-2) mothers was about 2.2 years, while the comparable figure among fathers was 2.8 years. Plotting the adolescents’ GPA as a function of years since divorce across the parental educational qualifications (see Fig 4) revealed a slight negative linear association between years since divorce and GPA across most of both maternal and paternal educational levels. The negative association between time since divorce and GPA was strongest among lowly educated mothers. As lowly educated mothers on average had most years since divorce, this finding highlights that time since divorce could not explain the heterogeneity in the associations between divorce and the adolescents’ GPA by maternal educational qualifications. Indeed, the plot suggests that holding years since divorce constant across maternal educational qualifications would slightly strengthen the difference in the negative association between divorce and GPA among adolescents with highly- compared to lowly educated mothers.

And added a paragraph to the discussion;

Adolescents with highly educated parents experienced, on average, that their parents had divorced somewhat more recently. Time since divorce was overall weakly and negatively associated with the adolescents’ GPA, with some small observable variations across parental educational qualifications. These variations could, however, not explain the heterogeneity found in the present study. Overall, this is generally in keeping with studies that find that youth’s post-divorce adjustment is quite stable or gradually slightly worsen as time passes since the divorce [61,62].

Response to Comments from Reviewer 2

1. Reviewer #2: I thank the authors for taking into account most of my comments and for further improving a already good study. However, I believe that they could conduct one robustness check. In response to my third point, the authors wrote that they do not have information on parental occupation and parental wealth. I see that they cannot then use these variables for a robustness check. However, they can use household income, a variable which is already present in their analysis as a control. I would like to see therefore at least a robustness check using household income as an alernative indicator of a family's socioeocnomic resources than parental education. It is precisely one shortocoming of the previous literature that they only used parental education and this study has the potential to address this shortcoming.

Response: We thank the reviewer for assessing our paper a second time, and for the positive comments and encouragements to further improve the manuscript

We have followed the reviewer’s advice and added additional checks utilizing household income as an alternative indicator of socioeconomic status. Utilizing the equivalized disposable income measure divided into quartiles, we find a similar but weaker pattern than in the models with parental education. We have added a figure summarizing the regression results and uploaded it as a supplementary figure.

We have updated the statistics section;

Lastly, we performed checks utilizing the income measures as alternative indicators of the family’s socioeconomic resources. The income measures were divided into quartiles (i.e., into four equal parts representing the lowest 25 % to the highest 25%), and the adolescents’ GPA was regressed on the interaction term between parental divorce and the income quartiles (similarly to the procedure described above).

And made changes to the result section;

Lastly, using equivalized disposable income (EDI) as an alternative indicator of socioeconomic resources, we found a similar but weaker pattern whereby the negative association between divorce and GPA was relatively stronger among adolescents in the second income quartile (Q2; b = - 0.16, p < 0.01) and in the fourth quartile (Q4; b = -0.15, p = 0.02) compared to those in the first quartile (Q1). The difference between Q1 and Q3 was not statistically significant (see S1 Fig. with further test statistics). Adjusting the analyses for parental educational qualifications attenuated and removed the significant difference between Q1 and Q4. No heterogeneity in the associations between divorce and GPA by mothers’ or fathers’ net income were found (results not shown).

Of note, it is highly likely that potential heterogeneity by measures of household income in the links between divorce and academic outcomes is sensitive to how income is operationalized. As parental education was the main focus of interest in the present study, we did not examine this any further in the present paper (e.g., other ways of dividing income into categories).

Decision Letter 2

PONE-D-19-20535R2

Dear Dr. Nilsen,

We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/ , click the "Update My Information" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at gro.solp@gnillibrohtua .

If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact gro.solp@sserpeno .

With kind regards,

Additional Editor Comments (optional):

Once again, I would like to thank you for addressing reviewers' comments, which, I believe, contributed to improving the quality of your paper. 

Acceptance letter

Dear Dr. Nilsen:

I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact gro.solp@sserpeno .

For any other questions or concerns, please email gro.solp@enosolp .

Thank you for submitting your work to PLOS ONE.

PLOS ONE Editorial Office Staff

on behalf of

Dr. Tomáš Želinský