Social Status Polarization in the Timing and Trajectories to Motherhood

This paper examines the polarization by social status of Canadian women’s
timing and trajectories to motherhood. The study applies event history analysis on data gathered through the 2001 General Social Survey on Family History and focuses on women born from 1922 to 1980. Women with high social status are more likely to delay their entry into motherhood and to follow normatively preferred trajectories that include graduation from post-secondary education. In contrast, women with low social status are more likely to follow shorter routes, often bypassing graduation from post-secondary education, regular work, or marriage, and consequently become mothers at younger age.

finds that parental education, family income, and parenting practices are all significantly related to children's outcome and proposes that this may be connected to an emerging 'bifurcation of fertility models'.
Drawing upon the polarization and bifurcation hypotheses, we examine the timing and trajectories toward first birth or the onset of motherhood on the assumption that, not only family life or having children, but life courses encompassing events early in life such as attainment of education and work are polarized by social status.Further, we assume that one's socio-economic status, and consequently, the polarization in the demographic sphere, is largely influenced by parental social status through inter-generational transfers or parental investment on children.In this study, we focus on the influence of parents on the subjects' life courses rather than the subjects' influence on their own children's outcome.The study is done for cohorts of women born from 1922 to 1980 in order to detect changes over time.

B. Polarization of the Life Course
The onset of parenthood is linked to other early life course transitions that comprise the transition to adulthood.Birth of first child could be considered as the definitive benchmark of having reached adulthood.In a normatively ordered life course, the start of parenthood may be taken as the culmination of the transition to adulthood that includes other important early life course events such as completion of schooling, start of regular work, and entry into marital union.Factors that influence the timing of parenthood may be similar to those affecting the other transitions and may be mediated through these earlier transitions.
Canadians born from the mid 1960s made the transition to adulthood at later ages than those born earlier, which seems to have happened in other Western countries as well (Ravanera, Rajulton, and Burch, 1998;Ravanera et al, 2002;Fussell, 2002).Young Canadians complete a higher level of education, enter the work force and stay in parental homes longer, and delay their family formation, either through cohabitation or marriage, and start of parenthood (Lapierre-Adamcyk, Le Bourdais, and Lehrhaupt, 1995;Boyd and Norris, 1999;Ravanera, Rajulton, andBurch, 1995, 1998;Ravanera et al. 2002).However, the timing and trajectories to adulthood have varied within cohorts as these are influenced by factors such as individual and parental characteristics (Shanahan, 2000;Booth, Crouter, and Shanahan, 1999 and articles therein).Having already explored these differentials in our earlier papers (see for example, Ravanera, Rajulton, and Burch 1998;2003), we turn our attention to specifically examine by social status Canadian women's early life course transitions with particular focus on the onset of motherhood.
That the timing and life course trajectories differ by social status, most often measured by level of education, is not a recent phenomenon.Studies in the United States show, for example, that the order of transitions has varied by social class within cohorts (Hogan, 1981;Hogan and Astone, 1986;Marini, 1984a).The normative sequence (completing schooling before marrying, for example) is more likely experienced by those in high social class.Consequently, they are also less likely to experience negative consequences in later life such as marital instability (Hogan, 1980;Hogan andAstone, 1986, but see Marini, 1984b).In this study, rather than simply examining differentials by level of education, we focus on parental social status as a context in which to view the life courses of individuals.Acquiring education, while important particularly for establishing one's own social status, is taken here as just one of the events within a young person's life trajectory.
In the following sections, we first discuss the data and methodologies used in this study.We then discuss the results of our analysis in two parts.The first presents analysis of the timing differentials by social status and by cohort in the onset of motherhood and in other related life course events.The second presents the trajectories to motherhood passing through early life course transitions such as graduation from post-secondary education, start of regular work, and marriage.A discussion of possible explanations for the findings follows the presentation of results.The paper concludes with a brief discussion of the context of the recent interest in polarization and implications for policies.

C. Data and Methodology
The study uses the 2001 General Social Survey on Family History, a country-wide survey conducted by Statistics Canada with a representative sample of those aged 15 and older, excluding residents of Yukon, Northwest Territories, and Nunavut and full-time residents of institutions (Statistics Canada, 2003).The survey has 24310 respondents; however, we limit our study to women born from 1922 to 1980, or 11,780 (weighted) respondents.Information gathered by the survey includes various aspects of the family including parents, children, union histories through both common-law and marriage, fertility, and socioeconomic variables.The survey also collected education and work histories.In this study, we make use of retrospective information on age at birth of first child and age at experience of other events at early life.
The timings are obtained from questions on dates when the events occurred in conjunction with the date of birth of the respondent, yielding the ages at experience of these events.
The social status variable was derived from two parental variables, mother's education and father's occupation when the respondent was aged 15.This assumes that parental education and occupation are the most relevant for measuring the social status, which in turn is relevant to the respondents' early life transitions, when the respondents themselves are in the process of establishing their own social status.The mother's education and father's occupation were ranked into low, middle, and high and then combined to obtain the social status variable 1 .Where mother's education is missing, the measurement of social status is based only on father's occupation, which could shift a small percentage of respondents to a category lower than what would have been assigned had information on mother's education been available.
We analyzed the data using the following methodologies for event-history analysis.1.To explore the differentials in the onset of motherhood by social status (SS), we constructed single-decrement life tables of age at first birth using SPSS.As in the subsequent analyses, life tables were built separately for birth cohorts -1922-402 (with 2223 respondents), 1941-60 (4647), and 1961-80 (4908), as cohorts go through different historical events that impact on their life courses (Ryder, 1965;Elder, 1978).It would be ideal to study narrower birth cohorts, for example 5 or 10-year birth cohorts, but the necessity of working with adequate sample sizes does not allow this.In the discussion of the results from these life tables, we use mainly the median ages at birth of first child.
2. As a preliminary step to doing the trajectory analysis, we did life table analysis also for other early life course events -home-leaving, graduation from postsecondary education, start of regular work, first union, and first marriage.
3. The trajectories through four life course events -education, work, marriage, and first birth (also referred to as "states") are traced for different social status.For this procedure, we used the LIFEHIST program that computes the conditional probabilities of making specific trajectories to parenthood on the assumption that past history is important (that is, a non-Markovian assumption).(Rajulton, 2001).
Essentially, the procedure is a multiple-decrement life table technique that estimates the conditional probabilities of transition to each state and the mean duration of stay in each state.For our purpose, we focus on two specific results: (a) the probabilities of experiencing selected pathways or trajectories; and (b) the age at which the specific trajectory is completed.
In this study we use individual (fractional) sampling weights in all statistical procedures.This is necessary as Statistics Canada uses complex sampling procedures for its surveys (Statistics Canada, 2003).
The retrospective information used in the analysis has its limitations.There could be errors in recalling events that have occurred in the past; and the respondents in the sample Occupations Unique to Processing and Manufacturing, Occupations Unique to Primary Industry), Middle (Trades, Transport, and Equipment, Business, Finance, and Administrative Occupation, Artistic, Culture, Recreational, Sport, and Occupations in Social Sciences, Education) and High (Management Occupations, Natural and Applied Sciences, and Health Occupations).The two rankings were added and the final social status rank was assigned as follows: low (1,2), middle (3,4), high (5,6).A score of one is possible when information on mother's education is missing.may be selected.As mortality rates are generally higher among those in the lower status, the sample may overly represent those in the higher social statuses.The bias caused by the first limitation may be small because dates of important life events are generally well remembered.As for the second limitation, the effect would be mainly in the estimates of those in the oldest cohort.And, if such a bias does exist, it would be toward a more conservative estimate of the differentials; that is, had the sample been more representative, the differentials would probably be greater.

Change in timing of motherhood initiated by high social status women
As shown in Chart 1, high status women tended to start motherhood later than women of lower status in all cohorts.This tendency was accentuated with a dramatic increase among high social status women born in 1941-60.Whereas in the oldest cohort  there was only about a year that differentiated the high status women (25.8) from women of low status (24.3), in the mid-cohort (1941-60) the gap increased to 4.3 years (28.7 among high against 24.4 among those with low status).While there was a large increase in age from the oldest to the mid-cohort among high status women (that is, from 25.8 to 28.7 years), there was virtually no change among women with low social status.
Women with lower status did some catching up among the youngest cohort.The ages at onset of motherhood increased by two years among low (from 24.4 to 26.5) and middle status women (from 25.9 to 27.9).However, the age among the high status women increased as well (by about a year and a half; that is from 28.7 to 30.3) over those of the mid-cohort.Thus, the gap between high and low status women in the age at onset of motherhood remains huge (at 3.8 years) even among the youngest cohort.

Onset of motherhood embedded in the life course
The changes over cohorts and the differences by social status did not happen only for the timing of first birth but, as shown in Table 1, for other events that usually happen in early life.This indicates that the onset of parenthood is embedded in the life course and that its timing is closely related to that of other life events and, as will be shown in the next section, to the sequences of experiencing these events.
An important early life event is completion of schooling.As seen in Table 1, however, there is almost no difference in age at graduation from post-secondary education by cohorts or by social status.This is because the survey asked the age at completion of first episode of post-secondary education only from those who graduated.Those who did not pursue or did not complete post-secondary education were not asked the dates when they stopped schooling.Even with this limitation, however, we were able to still make use of this information in the analysis of trajectories to motherhood (see below).
Like the onset of motherhood, in general, there has been an increase over cohorts in the average age at experience of all the other early life course events from home-leaving to first marriage (Table 1).In all three cohorts, the order of the median ages of experiencing the events is almost similar.The only difference in sequence is in the timing of the start of regular work and home-leaving: in the two oldest cohorts, on the average, start of regular work happens before leaving the parental home whereas in the youngest cohort, home-leaving is earlier than work start.Moreover, in the youngest cohort, the average ages at start of work and first union are virtually the same, while the age at first marriage is much higher than the age at first union.The latter is an indication of the widespread cohabitation in the youngest cohort.
Women of high social status experience all the early life events at later ages than those from lower status (Charts 2A to 2C).In the oldest cohort, the median ages are not too different for the low and middle status women; the differences between these two statuses emerge only in the two younger cohorts.Moreover, in the two younger cohorts, the difference by social status in the age at birth of first child is greater than the differences in the other events, with home-leaving showing the least difference.This indicates that the difference by social status in age at onset of motherhood is only partly accounted for by the differences in ages at experience of prior events.
As shown above, insights can be obtained from the average ages of various early life events; however, this type of information is not very useful for understanding the relations among the various events.It is clear, for example, that not all women experience all the events and many do not go through the events in the sequence implied by the average ages.To get a better understanding of the inter-relation among the various events, we refine our analysis by tracing the pathways toward motherhood, the results of which are presented in the next section.

E. Results of Trajectory Analysis
A trajectory analysis follows members of a cohort through the various events that they experience (or "states" that they occupy).These states need to be judiciously chosen since a large number of states would invariably lead to unmanageable number of trajectories and would require a large number of cases for a proper analysis.This need is particularly difficult to meet when members of each cohort are categorized further, here, by social status.On the basis of the results of life table analysis discussed above, we select only three other life course events in addition to first birth: (a) graduation from first post-secondary education, (b) start of regular work, and (c) first marriage.We excluded first union in favour of first marriage because if a trajectory does not pass through the marriage state, it can be inferred that the birth occurred within a cohabiting union (except when the marriage dates are missing).For a similar reason, we included graduation from first post-secondary education as its absence in a trajectory implies the non-completion of tertiary education.
Tables 2A, 2B and 2C show the conditional probabilities of transitions from one state to another, the standard errors of these probabilities, and mean duration of stay in each state.These conditional probabilities have been corrected for censoring and thus provide the best possible estimates of true probabilities (unless there is a very heavy censoring).Multiplication of these conditional probabilities in a specific trajectory provides an estimate of the trajectory's final probability of transition to first birth.And, summing up the mean durations of stay in each state provides a good estimate of the mean age at transition to first birth (since the means are computed from the conditional probabilities that have been corrected for censoring).Tables 2A, 2B and 2C show only the first six most common trajectories to motherhood.

Preferred pathway to motherhood most common mainly among high status women
The normatively preferred pathway to motherhood, that is, graduation→ work→ marriage→ motherhood trajectory (A1 in Tables 2A, 2B and 2C) is mainly followed by women with high social status.Tables 2A and 2B show that the final probabilities of this trajectory are the highest for high status women in the mid and youngest cohorts at 0.21 and 0.24 respectively.(The small number of respondents belonging to high social status in the oldest cohort shown in Table 2C does not allow the tracing of trajectories beyond the first transitions.)The probabilities of this trajectory are much lower for lower status women but these have increased over cohorts.In the oldest cohort, the probability among low status women of going through this trajectory is only 0.05 (A1 in Table 2C) but the probability increased to 0.08 in the mid cohort (Table 2B) and to 0.12 in the youngest cohort (Table 2A).
That women of high social status have greater likelihood of graduating from postsecondary education before marriage is also seen from the next most common trajectory of high status women, particularly for the youngest cohort; working prior to completion of post-secondary education --that is, the work → graduation → marriage→ motherhood trajectory (B1 in Tables 2A-2C).Among the youngest cohort, this has a probability of 0.17 in high social status women but only 0.09 among those with low status.The combined probabilities of these two trajectories (graduation→ work→ marriage→ motherhood and the work → graduation → marriage→ motherhood) are twice as high among high status women as among women with low status; in the youngest cohort, the probability is 0.41 as against 0.20; and in the mid-cohort, 0.34 as against 0.15 (see next to last row of Tables 2A and 2B).

Different trajectories among lower status women
Most women belonging to low or middle status go for regular work without first finishing a tertiary education; they then marry and become mothers.This work→ marriage→ motherhood trajectory (B2 in Tables 2A-2C) is the most common among lower status women in all three cohorts; however, the probabilities, though high, have decreased.The probability of going through this trajectory among women with low social status, for example, is 0.35 in the oldest cohort, remains the same for the mid-cohort, but decreases to 0.20 in the youngest cohort.Not surprising, the probability of this trajectory among high status women is lower.
A trajectory that does not go through either education or work is also widespread among the lower status women.However, this marriage→ motherhood trajectory (C1 in Tables 2A-2C) has also significantly decreased over cohorts.Among the low status women, for example, a quarter of women in the oldest cohort go through this trajectory.This was almost halved in the mid-cohort (to 0.14) and cut by half again (to 0.07) in the youngest cohort.A similar trend occurred among middle social status women -with the probability decreasing from 0.19 in the oldest to 0.05 in the youngest cohort.The common pathway among women in the past of marrying and becoming mothers without going through post-secondary education or work is no longer the way to go among the younger cohorts.As expected, this trajectory has not been popular among women of high social status.

Becoming mothers without marrying most likely among low status women
While the proportion of women who graduate from post-secondary education before becoming mothers have increased over cohorts, the other pathways that have become more widespread among the younger cohort are those that do not go through marriage; that is, most likely, motherhood occurring in cohabiting unions.The work→ motherhood trajectory (B3 in Tables 2A-2C) among women with low social status, for example, has increased from 0.05 in the oldest cohort to 0.13 in the youngest cohort, which also occurred among women with middle status.In contrast, this trajectory has a very low probability (0.03 and 0.04 in the mid and youngest cohort respectively) among high status women.
This trend among lower status women becomes even more a cause for concern when the probability of becoming mothers without first completing post-secondary education, working, or marrying (row D in Tables 2A-2C) is considered.The probability of this trajectory has almost doubled among low status women (from 0.06 to 0.11) in the youngest cohort over those of the older cohorts.As one might expect, this transition's probability among high status women is low (0.02) in the oldest cohort though it increased as well among the youngest cohort (to 0.05).One might wonder what subsequently happens to those who go through this trajectory.As with the other results, our analysis shows that the transitions following motherhood vary by cohort and social status (results not shown here).For the two older cohorts, the most common transition after motherhood is marriage for all social statuses.In the youngest cohort, the most common transition in the two lower social statuses is to start of regular work; whereas for women of high social status, completion of post-secondary education is the most common transition after motherhood, though the probability of starting regular work comes very close.(Note however that the number of women who go through this trajectory, particularly among high status women, is small.)

Life courses have their own momentum
As noted by Rindfuss, Morgan and Swicegood (1988), a life course trajectory has its own momentum and carry with it opportunities and constrains that in turn influences the timing of first birth.This seems to be validated to a great extent by the results of our trajectory analysis.For instance, when the first transition is graduation from post secondary education, the most likely second transition is to start of regular work [see A1 (ii) in Tables 2A-2C] rather than to marriage or motherhood.Similarly, a transition to marriage is most likely followed by a transition to motherhood [see, for example, B2(iii) in Tables 2A-2C].However, this seeming inevitability of the life course has changed over cohorts.For example, women in the oldest cohort who did not complete postsecondary education but went directly for regular work were most likely to have marriage as their next transition.In the youngest cohort, however, more women move on to completing post-secondary education.Among the low status women, this probability of marriage is 0.74 in the oldest cohort but only 0.42 in the youngest [B2 (ii) in Tables 2A-2C].In contrast, the probability of post-secondary graduation is 0.14 in the oldest cohort and 0.29 in the youngest [B1 (ii) in Tables 2A-2C].The change in momentum of the life course is also seen in the total final probabilities of transition to motherhood through the six most common trajectories shown in Tables 2A to 2C (last rows).For low status women, for example, the total decreased from 0.81 in the oldest cohort to 0.72 in the youngest.The decrease in the totals for all social statuses is an indication of the greater flexibility and more complexities in the trajectories to motherhood among today's young women.
There are variations by social status as well.Women of higher status seem to have greater flexibility in their life course.For them, for example, marriage need not be necessarily followed by motherhood, particularly if they have not as yet completed post-secondary education or have had regular work.(See, B2 (iii) and C1 (ii) in Tables 2A and 2B).This differential by social status is also seen in the totals of final probabilities, which for the youngest cohort are 0.72 for the low but only 0.59 for high status women (last row of Table 2A).

Age at onset of motherhood is largely determined by the number of prior transitions
The fewer the number of prior early life events, the younger the age at onset of motherhood.Thus, for all cohorts and all social statuses, those whose first transition is to first birth starts motherhood the earliest, while those who go through the other three events of graduation from post-secondary education, start of regular work, and marriage become mothers the latest.[Compare, for example, the 'duration' column of D with A1 (v) in Tables 2A-2C].The differences range from 6 to 10 years.Obviously, time spent for other pursuits, mainly for education and work, delays the onset of motherhood.
Remarkable, however, is the time spent between marriage and motherhood in the two younger cohorts.The duration is longest among those who go through post-secondary education, mostly about 3 years; and shortest among those who directly marry, mainly about a year and a half.[Compare A1 (iv) with C (ii) in Tables 2A and 2B] There are dissimilarities by social status as well.In general, women with high social status start motherhood at older ages than low status women though they may have gone through the same trajectory.However, the differences are not large.In the youngest cohort, for example, women with low social status who go through the graduation→ work→ marriage→ motherhood trajectory start motherhood at 28.7, whereas high status women do so at 29.4.

F. Discussion
Before offering explanations, we first sum up some of our findings as follows: o The increase in age at motherhood over cohorts was led mainly by women of high social status in the 1941-60 birth cohort but women with lower status have somewhat caught up in the 1961-80 cohort.o The differences by social status in the ages at the onset of motherhood are traceable through other events that happen in early life.The delay in onset of motherhood could be partly accounted for by delays in such events as completion of schooling (though information on this from the survey is limited), start of regular work, first union, and first marriage.o Women with high social status are more likely to go through the normatively preferred trajectory that includes completion of post-secondary education.o Women with lower social status are more likely to go through shorter routes to motherhood, including the trajectory that bypasses post-secondary education, regular work, and marriage.o Life course trajectories have become more flexible, and consequently, more complex among younger cohorts; and, the flexibility is greater among women with high social status.o The timing of motherhood is largely influenced by the number of prior life course events experienced.However, though differences are not large, women of high social status tend to become mothers later than women of lower status even if they go through the same trajectory.
The literature abounds with explanations for the decline of fertility, some of which could be used to also explain the increasing age at childbearing, or in particular, the start of parenthood.The most common explanation proffered is that of economic rationality.Education requires investment in resources including time and money, which then requires well-paying jobs to recoup the investment.With higher education and paid employment, the opportunity cost of having children increases.The perceived high cost of 'high quality' children coupled with the decline in material benefits from children provide more incentives to delay parenthood.The differentials in timing and trajectories suggest that the economic rationale for delaying entry may be stronger among high status women.
Canadian women have increasingly acquired higher education and have entered the labour force in greater proportions from around the 1970s (Beaujot, 2000).However, our findings show that the delay in the start of parenthood started mainly among women with high social status in the 1941-60 birth cohort.Most women in this cohort are the baby boomers, many of whom got their post-secondary education in the late 1960s and in the 1970s.This is an indication that in the early years of expansion of opportunities for higher education and employment, the beneficiaries were mainly those belonging to high social status.That opportunities for higher education and work have expanded in subsequent decades to include those with lower status can be gleaned from the results of the trajectory analysis but the differences in the trajectories to motherhood by social status also show that the inequality of opportunities has not been eliminated.
The timing of fertility is not just a product of rational economic calculation but is also influenced by cultural factors such as attitudes and values.Underpinning the second demographic transition, for example, is the change of values, mainly toward individualism and desire for self-fulfillment (Lesthaeghe, 1995).One's background imparts shared values or attitudes regarding fertility and timing of parenthood through socialization (Rindfuss, Morgan, and Swicegood, 1988;Michael and Tuma, 1985).Our own hazards analysis of the onset of motherhood (not shown here) indicates that values do influence the start of parenthood.The age at onset of motherhood among those who profess a religion is earlier compared to those who do not belong to any religion.Furthermore, those who regard as important to happiness family-related values (such as having a lasting relationship, having a child, and being married) are more likely to parent early; whereas those who give importance to paying job (job-related value) are more likely to delay entry into motherhood.
Change of values is thus another plausible explanation for the increase in age at motherhood between cohorts; and value differences could partly account for the timing divergence by social status in the onset of motherhood.As shown in Appendix  1).However, the importance they place for paying job is also much higher than those of the two previous cohorts, which is probably why the delay in the onset of motherhood continues.
Certainly, we have not exhausted all the possible reasons for the increasing delays in the onset of motherhood and for the differentials by social status.Our hope however is that our attempt at situating the onset of motherhood in a life course framework and tracing the various trajectories would contribute to the continuing search for explanations of the decline in fertility in modern societies (for latest attempts at understanding fertility decline, see for example, Caldwell and Schindlmayr, 2003;Hakim, 2003;McDonald, 2000).

G. Conclusion
To the question of whether there is a polarization by social status of the life course leading to the onset of motherhood, the answer provided by our analysis is "yes".However, this polarization is not a recent phenomenon; the timing and trajectories of life course events have differed by social status for all cohorts included in our analysis, the biggest change having occurred in the 1941-60 birth cohort.Moreover, social status differences need to be viewed in the context of social mobility.There has been a shift towards higher status over cohorts in the population.For example, using this same survey data, we find that the proportion of women with low social status in the 1922-40 birth cohort is 45% while it is 17% in the 1961-80 birth cohort.The middle class expanded from 31% in the oldest to 47% in the youngest cohort; and the high status from 4% to 19%.
The recent resurgence of concern over bifurcation of fertility is possibly triggered, for a number of reasons, by the increasing conspicuousness of those who become parents early.The general affluence of the population and the greater social mobility through education make noticeable those who are "left behind".With high rates of divorce and separation, the negative consequences of early entry into family formation, specifically marital instability and lone parenthood, have become more widespread.And, the weakening of age norms and the increase in age at experience of family events (including the start of marital union and parenthood) have made the timing of transitions more variable (Settersten and Hagestad, 1996;Ravanera, Rajulton, and Burch, 2004) and thus made those who make the transition to parenthood at a young age more visible.
However, while placing the polarization of the life course in broader perspective, our study does indicate that inequality of opportunities accentuates the differences in the timing and trajectories to the onset of motherhood.Interventions that would diminish the inequalities say in the access to higher education and subsequently, to employment, would most likely have the effect of reducing the differentials in the timing of entry to parenthood as well.But, reducing the disparities in opportunities could mean, in the population level, a greater delay in the onset of motherhood, which might in turn lead to even lower fertility than the current rate.This brings to the fore the importance for interventions that facilitate the balancing of family and work life such as those related to family benefits and the provision of child-care services (for a detailed discussion of the various types of interventions that relate to fertility, see Beaujot, 2004;Pampel, 2001;Gauthier, 1996;Gauthier and Hatzius, 1997;Kaufmann et al, 2002).

Table 1 ,
the oldest cohort of women had the highest score on the indicator of family-related values, which changed among the mid-cohort, who have the lowest.Moreover, among these baby boomers, women of high social status have the lowest score on family values (excepting those whose social status is undetermined) and highest score on job-related values.If family values were the only influence on age at parenthood, the youngest cohort of women would start motherhood earlier as they seem to hold family values more strongly than the mid-cohort (Appendix Table