Home | Research & Policy Library | Promises, Promises

Promises, Promises

The Impact of School Sector on Formation and Stability of Marriage and Cohabitating Relationships

December 15, 2017

Jonathan Schwarz

David Sikkink


Research Report

What is the impact of schooling experiences on the formation, quality, and stability of marriages and other romantic relationships of young adults? This report brings evidence to bear on the hypothesis that schools contribute to family formation and flourishing.

What is the impact of schooling experiences on the formation, quality, and stability of marriages and other romantic relationships of young adults? This report brings evidence to bear on the hypothesis that schools contribute to family formation and flourishing. We would expect that evangelical Protestant schools spend a good deal of time and effort socializing their students to respect and seek out a traditional marriage relationship. And Catholic schools both in practice and teaching are likely supportive of a traditional marriage and family ideal—at least in comparison to public schools. Previous analyses of Cardus Education Survey (CES) data revealed evidence that evangelical high schoolers marry younger and have children at an earlier age. This report completes the portrait of religious schoolers’ family life by providing an in-depth analysis of whether and why religious-school experiences contribute to a marriage life marked by stability and flourishing. It leverages the existing CES data on North American young adults to explain the relationship outcomes of religious-school graduates. The influence of schooling on longer-term flourishing in families is an important “non-cognitive” contribution of schools to society that takes its place alongside schooling contributions to civic engagement. By answering questions of family formation and stability, we provide some evidence on the question of whether and how religious schools contribute to the public good through their impact on family.

Data and Methods

We use the Cardus Education Survey (CES)—a biannual nationally representative sample of young adults (ages twenty-three to thirty-nine) in the United States and Canada—to test for lasting effect of high school sector on family formation. So far, there have been two cross-sectional waves of data collection in each country. For these analyses, we use the date of survey administration, respondent’s self-reports of age (in years), the year in which one was married for the first time, and the year in which the respondent’s first child was born to calculate a consistent measure of the age at which important benchmarks in family formation occur across all four waves of the CES. To account for differences in the structure of schooling and social norms between the United States and Canada, we combine data from both countries only to test for differences or similarities at the national level. For all other analyses in this report, we run separate models for each country.

For the analysis we fit a series of “nested” regression models to estimate each outcome. This strategy allows us to (1) establish a baseline of “average” effects of school sector on family formation, practices, and outlooks, and (2) identify other factors that might explain why we observe sector difference. In the end, any “unexplained” differences in our final (full) model can be theoretically attributed to lasting effects of the type of high school from which one graduates.

Figure 1 provides a visual representation of the analytical framework on which this research is designed. The “bars” increase in size from left to right as we incorporate more in- formation about the respondent and his or her family back- ground in our estimation of the effect of school sector on family formation. The models are cumulative in nature. For instance, model 2 includes information on school sector from model 1. Model 3 includes school sector from model 1, back- ground characteristics from model 2, and adds information about the respondent’s parents.


Marital Status and Relationships

Across all models, Canadian young adults are consistently less likely to presently be or previously have been married relative to their counterparts in the United States. In our full model, young adults in Canada are 38 percent less likely to be married than twenty-three- to thirty-nine-year-olds in the United States. We find that Canadians are just as likely as observably similar peers in the United States to be in a romantic relationship but are 52 percent more likely to be “single,” and more than twice as likely to live with a romantic partner (cohabiting).


In the United States, graduates of independent Protestant high schools are consistently more likely to have been, or presently be, married than public high school graduates (by nearly 60 percent in our full model). Certain characteristics of parental religiosity are also meaningful indicators of whether a young adult gets married. Independent of the effect of school sector, those who were raised by parents whom they describe as evangelical, Pentecostal, fundamentalist, or charismatic are 50 percent more likely to marry. On top of that, children whose parents frequently attended religious services while they were growing up are 40 percent more likely to get married.

Conversely, US graduates of Protestant high schools are less likely to be single and never married. When compared to being married, Protestant graduates are about 35 percent less likely to be single than public high school graduates, on average.

However, we don’t find school-sector differences in the likelihood of being currently separated or divorced relative to married. The relatively small sample sizes may explain the lack of findings on this score.

In some models, home-schoolers are about 70 percent more likely to be single than public high school graduates in our final model. This finding holds strong until the respondent’s educational level is included, which appears to indicate that home-schoolers’ greater likelihood of being single rather than married—on average and in comparison to public schoolers—is due to the lower levels of educational attainment of home-schoolers.

Next, we look at cohabitation, or sharing a residence with a romantic partner. Across all models, US graduates of independent Protestant schools are significantly less likely to live with a boyfriend or girlfriend than public high school graduates. Specifically, they are about 45 percent less likely to be cohabitating rather than married compared to public school graduates in the United States. For many, this Protestant-school impact will be amplified by the effects of parental religiosity.

Beyond the effect of attending an independent Protestant high school, the young adult children of conservative Protestants are 46 percent less likely to cohabit, and 37 percent less likely if their parents attended church on a regular basis.

Finally, we ask whether school sector influences singles toward involvement in romantic relation- ships, that is, dating with perhaps a view toward marriage or cohabitation. After accounting for all of the observable characteristics described in figure 1, we find that people in the United States who were home-schooled are less than half as likely to be in a “dating” relationship than public high school graduates. It is possible that this reflects skepticism about conventional dating norms and practices among some home-schooling families in the United States.


Among Canadian young adults, we find that graduates of independent Protestant high schools are, on average, more likely to be married and less likely to be single, dating, or living with a romantic partner. Unlike the United States, how- ever, these observed differences are explained by family background, particularly religious conservatism among Protestant parents and frequency of church attendance. Home-schooling graduates are less likely to cohabitate, but again this effect is due to family background.

Independent Catholic school effects are starkly different in Canada than the United States. US Catholic schoolers are not significantly different from public schoolers on marital-status outcomes. On average, graduates of Canadian independent Catholic schools are more than twice as likely to have married and are over 60 percent less likely to be “single.”

Age at First Marriage

Across all models, young adults in Canada delay marriage longer than young adults in the United States. In our full model, we show that Canadians are over a year and a half older when they first get married. In the United States, but not in Canada, one’s level of education is a strong indicator of the age at which he or she marries. Our full model indicates that the typical married college graduate in the United States “tied the knot” at age twenty-five. Those with less education get married younger, while higher credentials correspond with delays in first marriage.

When considering school sectors in North America, the results show that independent Catholic and private non-religious (PNR) are older when they are married compared to public schoolers, and this is explained in part by the higher average levels of education for these graduates. Home-schooling graduates marry at younger ages, and this seems to be related to average educational levels.

However, the home-school finding is tentative since it is complicated by generational effects. When accounting for generation, the younger age of marriage for the home-schoolers is explained by family background variables, especially the religion of the family. This home-schooling effect also doesn’t apply in Canada, since the negative effect does not hold in Canada after controlling for demographic variables.


In Canada there are essentially no observed sector differences in the age of first marriage, but the type of high school one attends in the United States does predict how old one is on their wedding day. On average, graduates of Catholic and PNR high schools are a year and a half to two years older when they get married, while young adults who were home-schooled are over one year younger even after controlling for family religious background. Among home-schoolers, the observed difference is attributable to one’s personal background.

After accounting for age, sex, and race, there is no difference in age at first marriage between young adults who attended public school and those who were home-schooled. Interestingly, the home-schooling finding is complicated by generation. Without controlling for generation, the home-schooling effect remains significant even after religion variables are added to the model.

Individual background, parental characteristics, and level of education explain a large portion of the observed difference in age of first marriage for graduates of Catholic and PNR high schools (relative to public school graduates). Yet the marital age differences are still statistically significant. It appears that the college preparatory environment in many Catholic and PNR schools combines a push toward completing a “normal” life trajectory, in which marriage comes after educational career and perhaps also after establishing a career. Access to and expectations for university admission and completion likely increase the age of marriage.


School experiences may influence the likelihood of divorce through instilling a traditional morality, modelling the place of stable marriages, or shaping the characteristics of spouses, among other mechanisms. Yet it is also possible that divorce is a response to a plethora of other factors and is largely outside the bounds of high school influence. Some important patterns emerge in the results when we limit the sample to those who were ever married.

To begin, the pooled results reveal a strong negative effect of PNR school attendance on the likelihood of divorce; however, this is explained by demographic and family background variables. There is evidence of a strong negative effect of home-schooling on divorce as well, which is consistent after including family background controls. The negative effect of Protestant school- ing on divorce is also very strong, and consistent across all models. Protestant school attendance cuts the likelihood of divorce almost in half in some of our complete models.


In the United States, the findings are very clear both for PNR and Protestant schoolers, which have much lower likelihoods of divorce (odds of divorce are over 60 percent less for Protestant school graduates). The Protestant school effect seems reasonably chalked up to the place of the evangelical community in supporting marriage and discouraging consideration of divorce. The PNR finding is more difficult to ex- plain, but perhaps parallels the general effect of higher levels of education, which has been associated with lower divorce rates in younger generations in the United States. A greater degree of homophily, such as similar education levels, stability of income and job prospects, and similarity of expectations for marriage, among other things, may also explain the PNR effect. In our results, the educational level of respondents appears to ex- plain why PNR attendees have lower likelihood of divorce.


The Canadian findings contrast sharply. School sector has little to do with the likelihood of divorce in Canada. The results show a higher likelihood of divorce among independent Catholics, but this is accounted for by family background variables. In some models, home-schooling is associated with substantially lower likelihood of divorce. Other than that, divorce is not related to school sector in Canada.

Importance of Marriage

Besides influencing relationship status, religious schools may affect how singles think about the importance of getting married in the future. In the full sample of those who were never married, we find that Canadian respondents are less likely to think marriage is important. Those whose parents were divorced are also less likely to support marriage, while the experience of affection between parents increases the extent that North Americans see marriage as important. Respondents who grew up in religious homes, especially religiously conservative homes, are also much more supportive of marriage. High school graduates are less supportive, while those with a graduate degree are more supportive of marriage.


Before considering the results for Canada, note that the never-married young adults are a select group. Especially in the Protestant and in- dependent Catholic sector, many graduates have already married, and those who have not may have made an explicit decision to delay marriage, even in the face of norms for marriage among their schooling peers. Perhaps that explains why among the never-married respondents, independent Catholic schoolers are not significantly different from public schoolers on the importance of marriage. Similarly, the Protestant-schooler effect, while significantly positive in early mod- els, is explained by family religious background. Conversely, the public Catholic schoolers are significantly and positively related to the importance of marriage across all models. Interestingly, then, the public Catholic effect on views of marriage plays out not in relationship status so much as the importance placed on marriage by their graduates.


Similarly in the United States, Catholic schoolers who have never been married—which is a relatively high number given the delay in marriage for this group—are quite strongly and significantly positive on the importance of marriage. Catholic schoolers who delay marriage are not doing so due to a lack of interest in or support for marriage. The findings for Protestants and home-schoolers are equally counterintuitive, though again these results may reflect the unique sample of never-married respondents. Protestant schoolers are not significantly different from public schoolers on the importance of marriage. Home-schooling is strongly and significantly related to lower levels of support for the importance of a future marriage.

Age at First Birth

In our fully adjusted model, we find that young adults in Canada are more than two years older, on average, than their counterparts in the Unit- ed States when they become parents. Despite this gap, there are also noteworthy similarities in predictors of how old one is when their first child is born. In both countries, young adults whose par- ents attended church frequently while growing up become parents about seven months earlier than parents whose own parents did not regularly attend religious services.

As with age of first marriage, those with lower educational credentials become parents earlier in life than individuals with higher educational degrees. Unlike age of marriage, this pattern is statistically significant in both countries. High school graduates are about two and a half (Canada) to nearly four (United States) years younger than college graduates when their first child is born. Graduate degree recipients who are parents are about one year (Canada) to sixteen months (United States) older than college graduates when they become a mother or father.

The North American models reveal that independent Catholic and PNR schoolers are older when their first child is born. For example, PNR schoolers are about one and a half years older than public schoolers when their first child is born. Protestant school graduates are older at first birth than public schoolers, but this is accounted for by family religious background.


Sector comparisons of age at first birth in Canada are similar to our findings for relationship status. On average, graduates of independent Protestant high schools are older at the birth of their first child than are non-religious public high school graduates, but this difference is explained by pa- rental background (education and religiosity). The same can be said for home-schoolers in Canada.


In the United States we find persistent sector effects. On average, graduates of Catholic, Protestant, and PNR high schools are one to two years older when they become parents, while young adults who were home-schooled are over one year younger when their first child is born. Interestingly, there is some interaction be- tween school sector and generation that influences age at first birth. When we do not adjust for generation, we find that the Protestant school effect is a function of differences in parental background, while level of education explains differences ob- served between PNR and home-schooled parents relative to parents who graduated from public high schools. Thus the PNR effect in particular depends on the extent that this sector influences the educational level of their graduates. But when we do adjust for generation, the Protestant schoolers are on average older than public schoolers at first birth by about three-quarters of a year. The reverse is true for the home-schoolers: they are significantly younger when we do not adjust for generation but no different from the public schoolers when we do account for generation. Our tentative conclusion is that home-schooling does not per se reduce the age of childbearing, while Protestant schooling does increase age at first birth compared to US public schoolers. In the final model, whether we adjust for generation or not, graduates of Catholic schools are over a year older than pub- lic school graduates when they become parents. Again, the disciplined, college preparatory orientation of many Catholic schools, perhaps especially within girls schools, creates a well-worn path to higher education, including graduate work, which is associated with a delay in child-bearing.

Number of Children

Among those who are married, young adults in Canada have two fewer children than young adults in the United States. The magnitude of this finding likely reflects a combination of the CES sample and our finding regarding age of first fertility. As mentioned earlier, Canadian young adults become parents later in life. This factor combined with the restriction of the CES sample to respondents who are under forty years old limits the window of child-rearing years. It is possible that Canadian families have just as many children as families in the United States, but among young adults, we find that Canadian families are smaller.

As with the age at which one becomes a parent, level of education significantly predicts the number of children one will have as a young adult in both countries. Respondents under forty with low- er levels of education tend to have more children than those with higher credentials. As with the relative fertility between countries, it is important to note that we are attempting to extrapolate only to young adults (twenty-four to thirty-nine years old). We find that young adults who have spent more time in higher education delay becoming parents, and ultimately have fewer children. It is possible that parents with graduate degrees have just as many children, but do so later in life.

Across North America, we find evidence that Protestant school graduates have slightly more children on average than public schoolers. PNR schoolers have fewer children than public schoolers. (This finding is strongest when controlling for whether the respondent was ever married.)


In Canada, independent Catholic schoolers have significantly larger families, and that effect is not accounted for by any of our control variables. The strength of Catholic identity and religious teachings on marriage and family nurtured within independent Catholic schools, in contrast to Catholic public schools, appears to have a longer-term impact on number of children in the respondent’s family.


In contrast, Catholic schoolers in the United States have fewer children than pub- lic schoolers, and this is explained by the higher levels of educational attainment among Catholic schoolers. Perhaps in conjunction with or as an unintended by-product of Catholic schools’ success in fostering higher education degrees, Catholic schoolers in the United States tend to have smaller families. Protestant school attendance, however, has a strong and positive effect on family size, and this is in addition to the positive effect of growing up in a religious family.

Marital Satisfaction

While there is considerable variation in the CES sample regarding self-reported satisfaction with one’s marriage, our analysis reveals that happiness in one’s marriage is unrelated to nation of origin, school sector, parental characteristics, or education. In other words, marital satisfaction is not influenced by the type of school one attend- ed, the characteristics of the parents who raised them, level of education, or where she or he grew up. It is possible that religious schools generate greater expectations for marriage, which tends to lower marital satisfaction scores relative to public schoolers. All told, it is a wash. Nothing consistently predicts satisfaction with one’s marriage.

Support for Traditional Gender Roles

Complementarianism is the belief, often rooted in religion, that men and women have distinct but complementary roles in a household or romantic relationship. This usually implies support for traditional gender roles, including a division of labor in the home and male authority in decision-making. CES respondents were asked to rate their level of agreement with the following statements: “It is better if the man earns the main living and the woman takes care of the house,” and “if a husband and wife disagree about something important, the wife should give in to her husband.” Our mea- sure is the average level of agreement with these two statements, standardized. By standardizing this measure, we are able to better compare the strength of the impact of school sector and statistical controls on complementarianism.

Married young adults in Canada subscribe to complementarianism less than observably similar Americans, even in our final model. While the difference is statistically significant, the gap is very small and lacks substantive importance in these samples.

In the results for North America, the strongest and most consistent school-sector effects are evident for Protestant and home-schoolers, which are more likely to support complementarian views.

United States

In the United States, school sector has a lasting and independent role in predicting one’s agreement with complementarianism. Graduates of private Catholic high schools are less likely to believe that husbands and wives have different roles in the household than those who attended public schools, and this effect is particularly consistent among married respondents.

Graduates of Protestant high schools and young adults who were home-schooled are both consider- ably more likely to hold complementarian beliefs (relative to public school graduates). In addition to the religious school effects, people raised by parents they describe as evangelical, fundamentalist, Pentecostal, or charismatic are much more likely to believe in separate roles for husbands and wives, while having been raised by parents who attend church frequently increases this belief slightly.

Level of education plays a moderating role in complementarianism. Young adults in the Unit- ed States with lower levels of education feel more strongly about “traditional” roles for husbands and wives. Those with graduate degrees, on the other hand, are considerably less traditional in views on gender roles within a marriage.


In Canada, graduates of independent religious schools—both Catholic and Protestant—as well as home-schoolers more strongly agree with complementarianism than public school graduates, on average. Independent Catholic schoolers in particular are strongly and significantly related to traditional gender roles. Interestingly, the relationship of Protestant schoolers and complementarianism in Canada is explained by other factors. The support for traditional roles among graduates of independent Protestant high schools is explained by the individual’s age, sex, or ethnic identity. Home-schoolers in Canada, though less complementarian than independent Catholic schoolers, remain consistently in favour of traditional gender roles across all models. However, if we limit our sample to married respondents, home-schoolers in Canada are more supportive of complementarianism, but this is explained by family religious background.

The relationship between level of education and complementarianism is not as straightforward in Canada as it is in the United States. While young adults with lower levels of education feel more strongly about traditional gender roles, there is no difference at other levels of educational attainment. While we only observe an education-level effect in one direction, the difference in magnitude is noteworthy. The gap between high school graduates and college graduates in terms of their belief in complementarianism is twice as large in Canada as it is in the United States.

Moderating the Effects of Family Instability

A further interest of this report is whether school sector mitigates the effects of family instability while growing up on marriage and family out- comes in young adulthood. In theory, a close-knit religious school may become a surrogate family for the student who is experiencing a divorce or separation of their parents. The holistic relationship between students and teachers in religious schools (Bryk et al.) could provide valuable and timely social support for teenagers dealing with disruptions at home. Religious schools may also influence how children respond to the parental divorce. Perhaps a divorce in the religious-school context changes long-term orientations of graduates toward up- holding the traditional family, partly as a reaction to one’s experience of family disruption.

In order to test for these possibilities, we included an interaction between the variable for whether the respondent’s parents were divorced or separated when they were a teenager and the school sector that the respondent attended. Admittedly, this is a rather conservative test given the relatively small sample sizes in some school sectors. And it is more difficult to detect a longer-term, indirect influence on marriage and family outcomes.

Perhaps not surprisingly, then, we find very few significant school-sector differences in moderating the effects of divorce during childhood. The only significant findings emerge for Catholic school graduates, whether government-funded or independent. Catholic school graduates who experienced divorce during childhood are more likely to take up complementarian views of gender roles in marriage as well as to incorporate religion into family life. Although highly tentative, per- haps this finding provides some credence to the argument that reactions to divorce are molded by religious-school experiences.

School Effects for Men and Women

One of the key claims in existing literature is that traditional gender norms are upheld in religious schools, especially evangelical schools (Peshkin 1982). We would expect, then, that marriage and family outcomes for female graduates of religious schools would be very different than for male graduates of the same. To test this, we included an interaction between gender and school sector in a separate set of models. The difference between males and females in the public schools is the comparison in these analyses. What is surprising in our analysis is that we have found very few gender differences in the effect of religious-school experiences on marriage and family outcomes. There are some notable differences in comparison to the effect of public schools, which we detail below.

We find a positive effect of home-schooling, in which females are more supportive of tradition- al gender roles in marriage than males relative to the public school gender effect. The gender difference among PNR graduates differs from public schools, such that PNR women report more religious activities with the spouse or family than do PNR men. Interestingly, PNR women are more complementarian in gender roles than men, but these differences with public schools are not quite statistically significant. Regarding divorce, the effect of home-schooling on females compared to males differs from the public school effect. Home- schooled females have a higher likelihood of divorce than males.

In Canada, the picture is somewhat different. Here we estimate that in comparison with public schools PNR women are older when married than PNR men, but this difference is not quite statistically significant. The Protestant school effect on age at first birth differs by gender compared to public schools. Perhaps surprisingly, Protestant school women graduates are older at age of first birth than are Protestant school men. Consistent with this find- ing, Protestant school women report fewer number of children than Protestant school men.

The effect of schooling on male and female differences regarding gender roles does vary between independent Catholic and public schools. In this case, independent Catholic women tend to be more complementarian in their views of gender roles than independent Catholic men. For reasons that are not clear, public Catholic school females report less religion at home in comparison with the gender effect of public non-religious schoolers. Less surprising is the finding that home- school women compared to men are more likely to report the practice of religion in family life.

There are important gender differences between public, public Catholic, and Protestant schools on the likelihood of divorce. First, public Catholic females are less likely to divorce than males. Second, Protestant school females are less likely to divorce than males. When it comes to a school effect on the likelihood of divorce, then, females compared to males in public Catholic and in- dependent Protestant schools are more strongly influenced toward intact marriages by their experiences in high school.

Religious Practices at Home

We would expect that the influence of religious schools on marriage, family formation, and family size would extend to the place of religion within the family, especially since we expect that religious schoolers would frame marriage and family decisions in more strongly religious terms.

Across the models we estimate, Canadians participate in fewer religious activities per week with their families. While statistically significant, the size of the difference requires careful consideration. In our final model, we find that Canadian families participate in 1.3 fewer religious activities (praying, read- ing the Bible, or talking about religion) per week.

Parental religiosity appears to have a lasting effect on young adults in both countries. In both the United States and Canada, young adults who were raised by Pentecostal, evangelical, fundamentalist, or charismatic parents participate in over four more religious activities with their family per week than similar peers who do not describe their parents as religiously conservative Protestants. Young adults who were raised by frequent church attenders engage in an additional three (United States) to three and a half (Canada) religious activities at home per week. In Canada, we also find a significant effect of being raised by a parent that the respondent described as conservative Catholic: as young adults, these individuals participate in an additional three and a half religious activities per week on average.

In North America we find that public Catholic schoolers report a higher number of religious activities in the family, though this is explained by educational level of the respondent. Still, there is a potentially indirect positive effect of public Catholic schoolers on family religious activities. When grouped across the United States and Canada, in- dependent Catholic schoolers report fewer activities, though this is not statistically significant in all models. Protestant schoolers report much higher levels of family religious practices across all models. Home-schoolers report even higher levels of family religious activities. PNR schoolers report lower frequency of family religious activities, an effect that holds after accounting for whether respondents were ever married.

United States

In the United States, there are clear, lasting relationships between the type of school one attended and the number of religious activities they engage in with their spouse or family. Table 2 presents the differences in number of religious activities with one’s spouse or family be- tween each US sector of schooling relative to public school graduates. On average, PNR graduates engage in the fewest religious activities, followed by Catholic school graduates, then public school graduates, followed by home-schoolers. On aver- age, young adults who attended Protestant high schools engage in the most religious activities per week by far. After accounting for the influence of demographics, parental characteristics, and education, the home-school effect disappears, and the gaps between each of the other sectors and public schools narrow, but remain significant. To put these findings in context, we note that home-schoolers are very high on family religious activities, but these differences with public schoolers are largely accounted for by family religious background (thus the relatively low estimate after controls are introduced).


In Canada, graduates of every major sector except PNR participate in more religious activities at home per week, on average, than non-religious public school graduates. The effect of Catholic school attendance is explained by demographics and especially family religion for independent Catholic graduates and education for graduates of public Catholic schools. While education, background, and family characteristics explain about half the protestant/public and home- school/public gaps, these sectors have a lasting relationship with how frequently their graduates pray together, read religious texts, or talk about religion with family members.

Home-schoolers in Canada are strikingly high on family religious activities.


Across the dimensions of marriage and family examined here, the influence of school sector is substantial and sometimes surprising. In several ways, the outcomes for religious schools are consistent with school missions. Still, in each religious sector, not all the findings are consistent with expectations.

First, cross-national differences on school-sector effects cannot be ignored. Religious school outcomes regarding marriage and family vary so strongly between the United States and Canada that it is difficult to gain much insight from the pooled effects. The cross-national contrasts are illuminating, however. Most obviously, the lack of a public Catholic sector in the United States appears to result in a very diverse US Catholic school sector. In Canada, the more conservative or traditional approaches to Catholic schooling are solely represented in the independent sector, while the Catholic school sector in the United States includes traditional and progressive orientations to Catholic schooling. Future research should attempt to separate out the graduates of more traditional Catholic schools. In addition, the differing historical contexts of Catholic schooling likely explains why Catholic school outcomes vary substantially across the border.

In general, the school-sector effects are much less crisp in Canada. The most interesting is the independent Catholic school findings, which show some strong, expected effects. The traditional orientation embedded within independent Catholic schools in Canada leads to stronger likelihood of being married, having more children, and supporting traditional gender roles. Public Catholic results showed some surprising findings, though the overall impact of this sector is minimal. Still, the positive public Catholic school effects on views of the importance of marriage and incorporation of religion in family life run against the common notion that public Catholic schools are no different from non-religious public schools. Also surprising is the relative lack of findings for independent Protestant schools in Canada. Other than weak evidence for a stronger likelihood of ever being married and higher levels of incorporation of religion in the home, Protestant schools in Canada are not distinct from public schools. We caution, however, that the sample size for this sector is relatively small, making it difficult to detect anything but fairly large differences. Home-schoolers in Canada are not consistently different from public schoolers on marriage and family outcomes, but they do have substantially lower likelihood of divorce as well as much more traditional views of gender roles and higher levels of integration of religion in home life.

The US school-sector differences are much more distinct. Catholic schools, for example, influence marriage and family outcomes net of demographic and family background differences. Catholic school singles are much more committed to marriage than are public schoolers. Other effects do not follow standard views of a conservative marriage and family effect of Catholic education. Catholic schoolers are not more likely to be married, marry much later, have fewer children, and are less likely to integrate religion into family life. They are also very skeptical of traditional gender roles.

Protestant schoolers have perhaps the most distinctive outcomes, which are almost entirely consistent with the mission and orientations of evangelical schools. They are more likely to be married than single or cohabitating, and they are more likely to have married at some point in their lives and are less likely to divorce. Protestant schoolers have more children and are more likely to integrate religion into family life. They are more supportive of traditional gender roles as well. Overall, and net of the religiously conservative families that attend evangelical Protestant schools, Protestant school communities appear to instill and support traditional marriage outcomes.

The private non-religious school effects are expected in several respects, especially the positive effect of PNR schooling on age of marriage and age at birth of first child, and negative effect on the integration of religion into family life. More surprising is the negative effect of PNR school attendance on divorce, which might reflect the educational trajectory of these graduates.

Home-schooling definitely affects marriage and family outcomes, as we would expect given the uniting of home and education in this sector. But the direction of the effects is not always predictable. Home-schoolers are more likely to be single and less likely to have ever been married, and those who do not marry are less likely to say that marriage in the future is important. There is some evidence of marriage at younger ages as well.

Finally, these findings do not vary a great deal by gender of respondent. In general, the gender differences point to more openness of females to religious school socialization, especially toward traditional gender roles and the integration of religion into the home.