These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

In this concluding chapter we shall reflect on a series of issues of both a methodological and substantive nature encountered in this research project. Firstly, we must realize that the use of individual census records not only opened vast possibilities, but also entails a number of limitations. Secondly, the very large sample sizes allowed for the disaggregation of national trends into far more detailed spatial, ethnic and educational patterns. This, in its turn, allowed us to adopt a “geo-historical” view of the rise of cohabitation for almost the entire American continent, from Alaska to Tierra del Fuego. Such an approach is an indispensable ingredient in understanding settings in which older and newer pattern of cohabitation meet and intermingle. Furthermore, another crucial feature is that statistical analyses could be performed at the individual and contextual levels simultaneously. Individuals have histories, but regions have much longer histories. Therefore contextual analyses are of paramount importance.

This volume is but a starting point for much more in-depth studies of partnership formation in the Americas, and particularly in Latin America and the Caribbean. Indeed, there is ample room for studies that follow the life course longitudinally (Bozon et al. 2009; Grace and Sweeney 2014) and for qualitative studies probing into the motivations for preferring cohabitation over marriage. Nevertheless, as the more detailed conclusions below will illustrate, a statistical analysis of the vast body of census information since the 1970s or 1980s is a necessary stepping stone. These analyses bring out unexpected variations, intriguing patterns of diffusion, and intricate interactive effects. And by doing so, pre-existing theories and expectations could be challenged, adapted or refined.

2 Data and Analyses

The vast majority of the data used in this volume stem from the large samples of individual census records as compiled and archived by the Minnesota Population Center. This unique and vast data set is known as the Integrated Public Use Microdata Series or IPUMS for short (Minnesota Population Center 2014). In all Latin American sources, there were direct questions as to the presence and nature of partnerships, including the category of consensual union. In Mexico, we could even make use of such information for the 1930 census, thanks to the recovery efforts made by the Mexican Instituto Nacional de Estadística y Geografía (INEGI). In Canada there is a direct question since 1986. In the US, where unmarried cohabitation was uncommon and theoretically illegal, such a straightforward question was absent, and as a result, indirect procedures had to be used, which presumably underestimated the true incidence of the phenomenon (Kennedy and Fitch 2012). In addition, several chapters were also able to use information stemming from large scale surveys, such as the Demographic and Health Survey (DHS) or the pooled annual American Community Surveys for the period 2007–2011. It should be noted that the DHS surveys do not permit a more detailed spatial decomposition and are best used for entire countries.

The reader will note that the present project bears some resemblance to the well-known “Princeton European Fertility Project” of the 1970s studying the spatial aspects of the European fertility transition (Coale and Watkins 1986). This was equally a census-based investigation, but of regional patterns of fertility control and their economic and cultural determinants. The main criticism of the Princeton project pertained, obviously for the lack of better, to its exclusive use of aggregate data only. The availability of individual census records in the IPUMS files has entirely removed that barrier. The net outcome is that the present analyses of patterns of cohabitation can be performed both at the individual and the contextual levels simultaneously.

The spatial disaggregation of the national data sets not only pertains to entities such as large provinces and states but very frequently also to much smaller spatial units such as cantons, meso-regions or even municipalities. The outcome is that this project is unique in having information for over 19,000 such spatial units. Obviously the study of contextual effects is considerably enhanced by the availability of such smaller units. For instance, for Mexico, a very detailed disaggregation has been highly instrumental in documenting the diffusion pattern of consensual unions, which we certainly would have missed if our information would have been restricted to the Mexican states only. Very much the same would have happened in the US if the analysis were conducted at the level of the states, instead of at the currently used much finer grid of the Public Use Microdata Areas (PUMAs).

A major drawback of census data is the lack of retrospective information concerning the process of union formation. In other words, we only know the current type of partnership, i.e. married or cohabiting, but we do not know how the union was initiated. Obviously a simple question of having ever experienced a period of living in a consensual union would have gone a long way in splitting up the large category of married respondents into those who ever and those who never cohabited. As a result, we have to be careful when interpreting the lower figures of cohabitation for somewhat older women, as these can result from either a straightforward cohort effect (older generations cohabiting less) or from a life cycle effect (the differential conversion of cohabitation into marriage as age advances). Similarly, when considering a negative education profile of cohabitation in the age group 25–29, we do not know whether the better educated have a lower incidence because they were less prone to initiate a partnership via cohabitation at the onset, or whether they started out in the same way as the others but more frequently converted their consensual union into a marriage later on. We presume that it is likely that the latter pattern becomes more frequent as the stigma against cohabitation is lifted and as the incidence of cohabitation is rising among new cohorts. In this instance, marriage is not a pledge of commitment for the future, but the outcome of a tested stable existing relationship (Furstenberg 2014). This conundrum could be solved partially by considering younger women, but then many have not yet initiated a partnership of any kind, and those who have are a self-selected subsample at any rate. Furthermore, with advancing education, more permanent partnerships are commonly being initiated later as well. In the balance, our frequent focus on the 25–29 age group is a compromise, but it is not without drawbacks. Therefore, whenever possible, we have reconstructed the full cohort profiles by age and education.

But there are also limitations on the independent variables side. Most censuses have information on the level of education. This is a crucial variable, but it has many meanings and is therefore a proxy for both economic and cultural dimensions (e.g. income, social class, openness to the world, political awareness and cultural modernity). Also, the rise in education over the years may not have altered the relative social position of the younger generations compared to the older: literate daughters can still be as poor as their illiterate mothers. And this may hold in particular in societies with large class differentials and ethno-racial stratification.

Language and ethnicity are also important variables commonly recorded in censuses. But very often only the first language is recorded. Most respondents in Hispanic countries state that they are Spanish speakers, but they may also use indigenous languages which remain unrecorded in several censuses. As such, the relative sizes of indigenous populations tend to be underestimated.Footnote 1 Religious denomination is another important variable, but denomination alone falls short of measuring religiosity or the importance of religion in a person’s meaning-giving system. It is well known that Evangelical Christians and Mormons strongly oppose cohabitation, but for the large category of Catholics, denomination alone falls short of what is wanted. Actual practice of Sunday Mass attendance would also be needed. Also, censuses provide no information about the importance of syncretic religions which mix Christianity and older native religions. These syncretic religions are very important in Brazil and in the Andean region. Furthermore, the category without religion is probably a more mixed bag and does not only capture agnostics.

Finally, and very importantly, censuses provide no clues whatsoever on cultural shifts. More specifically, we have to infer the de-stigmatization of cohabitation from the mere rise of this form of partnership, but we cannot link it to related dimensions of changes in ethics at the individual level and to patterns of secularization at the contextual level. All that can be done is to use illustrations with data from other sources, such as the successive rounds of the World Values Surveys.Footnote 2 In other words, crucial cultural changes in attitudes toward politics, religion, and ethics are flying under the radar, which will inevitably lead to the underestimation (or worse, even negation) of their effect.

With these caveats in mind, we can now turn to the substantive findings.

3 The Pertinence of Historical Factors and Contexts

Indigenous populations, European immigrants and African slaves all had their distinct systems of partnership formation, but over the centuries, religious conversion, colonial reorganization, and marked ethno-racial social stratification frequently resulted in new sui generis partnership patterns as well.Footnote 3 During the twentieth century, and possibly even earlier, the general tendency was that consensual unions would eventually be replaced by the standard European pattern of marriage. But large pockets would remain, mainly among Afro-Americans and selected indigenous groups, in which the tradition of forming consensual unions would be maintained. So far, this summary of the situation would have been accurate until about 1970.Footnote 4 After that date the pattern of union formation turns around with cohabitation gaining greater prominence and even becoming the modal form in many places. We shall refer to this later period as the “reversal phase”, which in fact is not yet completed, as further rises in cohabitation are to be expected in areas with a later take-off.

What are the salient characteristics of the reversal phase? First and foremost, the effects of social stratification, religion and ethnicity are continuing to be of major importance. In other words, the historical “pattern of disadvantage” is still in evidence, virtually everywhere in the Americas. Only in Canada are these effects strongly attenuated since this is a much more egalitarian society with only small Indian, Inuit and Métis populations. Aside from the Canadian case, if one is black or belonging to an indigenous group, not very religious, and poorly educated, then the odds of starting and remaining in a consensual union are largest. If one is white, well educated, and religious, then the odds are totally reversed. This not only holds at the individual level, but at the contextual level as well. Hence, if one is black, uneducated and not very religious, and one furthermore resides in an ethnic, poor and not particularly religious area, then the odds for entering and staying in cohabitation increase even more. Also, residence in an area with more immigrants systematically increases the odds for cohabitation. Conversely, the odds shrink further for white educated and religious persons residing in areas with similar contextual characteristics. In all countries for which contextual analyses could be performed with a finer spatial resolution, it was found that the contextual effects were highly significant and, even more importantly, entirely robust for controls for individual characteristics.Footnote 5 In other words, area or region of residence matters a great deal over and above the effects of individual characteristics.

There are major exceptions to this basic rule. Several indigenous populations must have lost their preference for cohabitation much further in the past or had a pattern with more monogamous marriage at the onset.Footnote 6 For instance, among the Mayan groups in both Mexico and Guatemala monogamous marriage is the preferred form of entering a union, even if marriages take place at young ages (see also Grace and Sweeney 2014). Similarly, several Andean native populations in Columbia, Ecuador, Bolivia and Peru do not stand out as having a higher prevalence of consensual unions either. In fact, the maps in Chap. 1 show that there is an Andean Altiplano ridge of low cohabitation. The Bolivian, Ecuadorian and Peruvian censuses reveal that the two largest ethnic groups (i.e. the Quechua and Aymara) have, controlling for other characteristics, the lowest incidence of cohabitation.Footnote 7 By contrast, for Afro-American populations, we have not encountered any exceptions in the present set of country studies. Whether in the US, in the Caribbean, or along the Pacific coast of Colombia, the odds for cohabitation among women ages 25–29 are always higher for black descendants of slaves than for whites or for most indigenous or mixed populations.

The dichotomy sketched above merely capture the two extremes of the continuum. Decades, if not centuries, of mestizaje or miscegenation have blurred the ethno-racial factor. Mass migration to urban areas and megalopolis has created new patterns of segregation. And the growing Evangelical adherence has produced a reaction against the prevailing demographic and ethical trends. As a consequence, there are various combinations of factors that produce intermediate results. In order to illustrate these interactions between conditioning factors, our contextual variables are being constructed as combinations of categories. This leads to interesting insights. Here are a few examples.

In the US, the effect of the “pattern of disadvantage” on cohabitation completely disappears for the Black population when residing in areas with a large Evangelical presence and it is also attenuated when there is a strong presence of Afro-Protestant churches. Conversely, the odds for cohabitation increase with increasing proportions Catholic and Mainstream Protestants in the US PUMA areas. Also residence in a PUMA with a strong Democrat political composition increases the odds for cohabitation for everyone.Footnote 8

Another example of an interactive effect pertains to Mexican areas with a high concentration of educated women. In these upper social strata municipalities the odds for cohabitation were not lower, as expected, but significantly higher. Furthermore, this puzzling feature remained robust for all sorts of controls. A further scrutiny revealed that it was not women with more than secondary education that produced the positive contextual effect, but the least educated women residing in these areas. A plausible explanation for this is that women with no more than primary education find employment in the larger service sector in better off municipalities, and on the basis of their earnings can maintain a cohabiting household. Moreover, in such settings, the de-stigmatization of cohabitation could have advanced further than in the more homogeneous municipalities.

In the example of Brazil, individual membership of an indigenous or Black population is indicative of a higher risk for cohabitation, but the effect of this individual characteristic is either strongly attenuated or reinforced depending on the strength of Catholicism in the various meso-regions. In this interaction, a higher than average percentage of Catholics in the area substantially reduces the incidence of cohabitation, also for Blacks. Furthermore, the importance of religion in Brazil equally shows up at the individual level, with Lutheran Protestants (mostly whites), Baptist and Evangelicals (mostly Pardo or non-whites) having much smaller odds than Catholics, whereas women 25–29 in a union reporting no religion have a much higher incidence of being in a consensual union. Another striking feature for Brazil is that the educational contrasts are very substantial at the individual level, but much less so at the contextual one.

In Colombia 2005, the most striking effects in favor of cohabitation at the individual level are found for education, with the classic negative gradient, and for membership of the Afro-Colombian group. This population is concentrated along the Caribbean and Pacific coasts and the northern mining regions. By contrast, membership of an indigenous population compared to the majority of the mixed race population reduces the incidence of cohabitation. This is, along with the Mayas of Mexico and the Quechua and Aymara of Peru and Bolivia, another example of the fact that the correlation between ethnicity and consensual union formation is weaker for the indigenous Americans than for the Afro-American populations. Furthermore, as in Brazil, the contextual effect of education is weak, but that of the local strength of Catholicism much more important in reducing the incidence of cohabitation. Hence, Colombia is a typical case of continued heterogeneity according to social class and race (essentially Afro-Columbian versus others), but also of persisting regional differentiation according to the historical strength of Catholicism.

In Ecuador 2010, the negative gradient with education has been maintained during the reversal phase, in tandem with the impact of the ethnic factor. As expected, Black and mulatto populations have considerably higher proportions of women in consensual unions, whereas Quechua speakers maintain their strong tradition of moving into marriage. The populations on the Amazonian side such as the Shuar (Jivaro) fit the pattern with widespread cohabitation. At the contextual level, being a resident in a predominantly Quechua speaking area decreases the incidence of cohabitation even more. A similar, but weaker, effect in the same direction is also found when resident in areas of less immigration.

The Peruvian findings for 2007 are more attenuated. The education gradient remains negative, but the ethnic differentiation is less pronounced. The Quechua speakers are not standing out anymore, and it is the Aymara that now have the lower incidence of cohabitation. By contrast, the small groups on the Amazonian side, such as the Ashaninka, have much higher levels. The other dominant trait in Peru is the impact of Evangelical proliferation. The strong negative effect on cohabitation associated with being Evangelical Christians emerges mainly at the individual level, and not so much at the contextual level of the provinces. In fact, the Peruvian contextual effects as measured here are of secondary importance to the individual ones. The reasons for this are not only the weaker contrasts at the individual level, but equally the greater homogeneity of the 176 Peruvian provinces than in the neighboring Andean countries. It should also be stressed that Peru and Colombia have had a more rapid expansion of cohabitation than Bolivia and Ecuador, and that this could have contributed to a leveling of contrasts.

In Bolivia 2001, the education related gradient is steep, with less cohabitation among young women with secondary education or and much less among those with university degrees. Also at the individual level, Aymara, Quechua and Chiquitano speakers have again considerably lower relative odds for being in a consensual union, whereas the Guarani and other indigenous populations exhibit the reverse pattern. The contextual effects among the 84 provinces are more pronounced than in Peru. In addition to the individual effect of ethnicity, residence in areas with mainly Quechua and Aymara speakers significantly reduces the odds for cohabitation. The same holds for residence in areas with fewer immigrants. By contrast, the educational composition of the provinces produces no extra contextual effect.

In Central America, the evolution in the prevalence of consensual unions over the past five decades has shown different paces of change across countries and an increasing convergence in cohabitation levels. In general, countries which already had high levels of cohabitation in the 1960s (e.g., El Salvador, Honduras, Panama) have experienced small to moderate increases whereas countries with traditionally low levels of cohabitation, such as Costa Rica, have undergone large increases. Guatemala is the only country where a downward trend can be observed during the second half of the twentieth century, although recent survey data from 2011 suggest that the decline in cohabitation has halted and is possibly reversing. The recent increase in cohabitation in Central America has been largely concentrated among women with secondary and higher education, for whom cohabitation was negligible in the past. As elsewhere in Latin America, the historically negative educational gradient of cohabitation remains largely in place, but differentials in union patterns by educational level have narrowed considerably in the past two decades. The spread of cohabitation among the middle and upper classes has probably been facilitated by the wide social recognition conferred on consensual unions in the lower strata, but it challenges the traditional strong association between cohabitation, poverty and social disadvantage.

4 Indigenous Latin American Marriage and Cohabitation in a Global Perspective

It is frequently stated that consensual unions are common among indigenous people in Latin America and that this is the main reason for the expansion of cohabitation. Such a general formulation is invalid for major parts of the continent. In fact, our scrutiny of late twentieth and twenty-first century demographic data reveals the existence of a high degree of heterogeneity among native populations, and not only between whites and others. The Zapotec of Mexico, the Mayas of Mexico and Guatemala, and the Quechua and Aymara of the central Andean Altiplano stand out by considerably lower levels of cohabitation. The Nahuatl group in Mexico who are considered to be the direct descendants of the Aztecs have intermediate levels of cohabitation, but the adjacent civilizations in Central Mexico, i.e. the Mazahua, Otomi and Purepecha had the lowest incidence of consensual unions in 1930 and still are at the lower end of the distribution in 2010. These pre-Hispanic civilizations were based on intensive agriculture often with irrigation and terracing, advanced architecture and technology, state formation and central control, priestly and military castes, and local tribal nobilities. At the other extreme were hunter-gatherer populations and groups that engaged in shifting agriculture (slash and burn). These societies had much simpler forms of organization with only local heads, or occasionally in South America, even without any clear fixed pattern of authority structure.

This duality fits the Boserup-Goody typology of global patterns of partnership formation (Goody 1976). According to these authors, populations that reached the stage of intensive and technologically advanced forms of agriculture also tend to form larger states, develop a system of social stratification with social classes or castes, and have appropriation of agricultural land. If land belongs to a corporate kinship group or to smaller individual families, marriages need to be controlled to avoid misalliances resulting in devolution of property. In this situation, there is much less room for free partnership formation, shifting partnerships, polyandry, sister exchange etc. Instead, marriage becomes a firm institution under parental or kinship control, and marriages are furthermore ritualized. This commonly involves a public and elaborate ceremony (or even a sequence of ceremonies). A further distinction is made by Goody concerning the direction of the exchange of goods. In systems with “diverging devolution” women alienate property upon marriage through their dowry (bridewealth). In the opposite systems, exchanges are either bilateral or are at the expense of the male kinship group (brideprice). In the former system women are “a loss” to their brothers, and societies with diverging devolution tend to be strongly “patriarchal” with endogamous and arranged marriages, and various sorts of discriminations against women. Most Asian societies exhibit these characteristics. In the type without diverging devolution of property, such “patriarchal” control is much milder, and in the European setting the Catholic Church further limited the control of marriages by the parents and kin (Goody 1983). Unless altered by Islam, most sub-Saharan African populations have the system of bridewealth and of exogamous marriages. They also had slash and burn agriculture, lacked irrigation and plough, and had no individual appropriation of land. They are at the opposite end of the Boserup-Goody typology.

The Goody-Boserup reasoning goes a long way in describing the present duality concerning the incidence of cohabitation. Several Central Mexican, Zapotec, Maya, Quechua, and Aymara populations all seem to have maintained systems of stronger marriage control by parents and kin. Moreover, the Quechua-Aymara group is known for the lavish marriage ceremonies and other celebrations associated with rites of passage (births, puberty, deaths, fertility rites).

The duality between populations of pre-Hispanic organized empires and others was of direct relevance for the Spanish conquerors and missionaries. The Catholic monogamous marriage and its ceremony fitted the indigenous forms much better for populations such as the Quechua, Aymara or Maya. Hence, over time, cohabitation did not become the rule among them. By contrast, for most of the other indigenous populations without complex state formation, Christian marriage was not only an alien concept, but ran entirely against the much more free forms of courtship and partnership. This is very well illustrated by Livi-Bacci (2010) who describes the Jesuit efforts to eradicate widespread “promiscuity” in ChiquitanoFootnote 9 and Guarani populations around their seventeenth century missions. Today, according to the current Bolivian census figures, marriages are considerably more prevalent among the Chiquitano than among the Guarani.

As stated in the introduction to this chapter, regions have much longer histories than individuals. As is clear by now, the current picture of partnership formation is still influenced by the historical structuration of centuries ago. The impact of Christianization is undeniable, but older patterns of consensual union formation commonly prevailed. The “dessous des cartes” is at least five centuries deep.

Nevertheless, an entirely new wave of change started rolling over the pre-existing patterns from the 1970s onward. That wave is commonly referred to as the “Second Demographic Transition”.

5 The Trend Reversal and the Second Demographic Transition (SDT) Factors

The core thesis of the SDT-theory is the Maslowian principle that the nature of needs changes as populations become wealthier and, by extension, more educated. As the material needs are better satisfied, more non-material needs tend to be accentuated, and populations become more vocal in articulating them. This mechanism also translates into cultural changes, with individuals stressing the right to make decisions autonomously, i.e. independently of religious or older moral codes, and furthermore in articulating expressive needs: freedom of choice, self-actualization and emancipation, maintenance of a more open future and flexibility, gender equity etc.Footnote 10 The manifestations at the macro-level are the growth of emancipation movements claiming equal rights for women or for ethnic or sexual minorities, further secularization, and concomitant de-stigmatization of a number of moral issues such as divorce, abortion, euthanasia, homosexuality and suicide. Obviously, the moral stigma against the formation of a sexual union outside marriage belongs to that values dimension as well.

The SDT factors work in three different ways. Firstly, the pattern of union formation now belongs to the domain of individual choice (i.e. autonomy) and not any longer to that of a corporate or collective normative regulation. If choices are open, then the cost-benefit evaluation as perceived by individuals (rather than families) applies to a greater extent, and these evaluations may not be the same for men and women respectively. Also, the elements in the calculation must not of necessity be of a mere material nature. Fidelity and trust, for instance, may score equally high on the priority list, and if not guaranteed at the onset, a period of cohabitation could be preferred over marriage. Translated into the “Ready, Willing, and Able” (RWA) framework of preconditions for the adoption of new forms of behavior (Coale 1973), the opening up of wider choices and the evaluation of advantages and disadvantages constitute the “Readiness”-factor.

“Willingness” refers to the normative, i.e. the religious or moral acceptability of forms of behavior. The SDT operates via the “Willingness”-factor through the aforementioned de-stigmatization of a number of hitherto negatively sanctioned forms of conduct. There is often a positive recursive relationship at work: as religious or moral objections to a given form of conduct weaken, then the practice of that behavior will spread, and as that occurs, then the religious and moral objections will weaken even further.

“Ability” refers to the technical or legal constraints or possibilities for the new form of behavior to materialize. In the context of cohabitation, mainly the legal context is of relevance.Footnote 11 Typically, as a new form of behavior spreads, the legal system tends to adapt, but this frequently involves time lags of varying durations. In most Latin American countries, the legal impediments to consensual unions were not of a prohibitive nature, but that was not so in Canada and especially not in the US. On the whole, except for the Canadian chapter, the legal situations and their changes have been completely underexposed in this volume, as this requires specialists’ competence in what is often a complex and diverse subject matter.

An essential implication of the RWA-model is that these preconditions need to be met jointly for the outcome to materialize. However, these three components do not change at the same speed. Contrary to intuition, the slowest of the three conditions at the aggregate level does not set the ultimate pace of change of the outcome feature. This ultimate pace is slower still. The reason for this is that the conditions change at the individual level, and that the slowest condition in the aggregate is not of necessity uniformly the slowest for all the individuals. It is the remaining smaller group of individuals with lower scores on the other factors who slow down the process to an extra degree (Lesthaeghe and Vanderhoeft 2001). The implication of this form of change is that bottlenecks become even more important than they seem at first sight. For instance, even if cohabitation would have many advantages over marriage (high R) and if both forms of unions were legally equivalent (neutral A), then this form of partnership would emerge even slower than the pace set by the gradual removal of the religious or moral objections. This illustrates that the rapid rise in cohabitation as witnessed in so many parts of the Americas could not have taken place without the very fast removal of the moral and religious stigmata against it. In other words, the rapid rise of cohabitation required nothing less than an “ethical revolution”, similar to the “cultural revolution” that occurred in Western Europe in the 1960s and 1970s.

The cost-benefit evaluation of the marriage-cohabitation duality (Readiness) can obviously not be addressed with census records. More qualitative studies are required for that. Most of the sources that shed light on the motivations stem from US or European sources, and point to a multitude of elements being involved (e.g. Liefbroer 1991). Crucial factors cited in focus-groups in eight European countries seem to be centered around “commitment”, “the testing of a relationship” and “freedom” (Perelli-Harris et al. 2014).Footnote 12 In the Latin American context, there is, to our knowledge, no equivalent set of studies based on focus groups or in-depth interviews that probe into the motivations plethora. One can imagine that the three key dimensions found among European motivations would be relevant for the Latin American context as well, but we have no comparable investigations that would document this point or bring up other dimensions (e.g. “respect for tradition”, “affordability of marriage”, “weak employment prospects”, to name a few possibilities).

Also the dynamics of the process of partner formation help in interpreting the various meanings of cohabitation. With respect to the process of partnership formation over time, several surveys are by now available that have retrospective information on the sequences of events (e.g. DHS), but, with a few very recent exceptions (e.g. Covre-Sussai et al. 2015; Grace and Sweeney 2014), these data have remained underexploited on this topic.

Information on the “willingness”-factor can be gleaned from the World Values Surveys (WVS) as they measure attitudes in the domain of ethics. In fact, the WVS rounds that often started in the 1990s in Latin American countries are capable of documenting the “ethics revolution” in several cases. At the individual level, no link can be established between the ethics attitudes and type of partnership for the lack of a retrospective probe among married women about a possible prior cohabitation experience, but the WVS does provide aggregate trends on the ethics changes and de-stigmatization. We shall provide some further details on these issues in the next section.

6 The Education Gradient and the “Ethics Revolution”

As indicated, the historical negative gradient of cohabitation with respect to the level of education is a well-nigh universal Pan-American pattern, which, furthermore, remains largely in place during the reversal phase so far.Footnote 13 If the only process at work would be a composition change with respect to the considerable increases in education levels for men and women, then, given the negative gradient, consensual unions would have yielded further to marriage. In other words, no trend reversal would have taken place and the old trend towards more marriage would have been reinforced. Yet, the trend reversal is as universal as the negative cohabitation-education gradient itself. As depicted in all the previous chapters, since the 1970s the share of cohabitation among women in a union 25–29 has increased at all levels of education. Moreover, this holds for the adjacent age groups as well. Apparently, within an SDT-context, rising education must have spurred on the degree of autonomy of young adults in making crucial decisions, and must have de-stigmatized the formation of consensual unions among population segments, such as urban educated whites, that had hitherto exhibited a strong preference for marriage. Moreover, one could argue that autonomy in decision making and the de-stigmatization could have been initiated by the better educated in American societies. Data on the “ethics revolution” are supportive of this conjecture.

In Figs. 10.1 and 10.2 use is made of WVS-data pertaining to the “ethics revolution” for selected countries for which there are multiple measurements in time. Divorce is not so much of an ethical issue anymore, and suicide is only at the very beginning of de-stigmatization in the Americas.Footnote 14 More specifically, we have plotted the percentages of respondents (18+, both sexes) that are of the opinion that homosexuality and euthanasia can never be justified, and we show the results for three education levels and for two periods, the 1990s and the years 2005–06. These trends by education in acceptability of euthanasia and homosexuality document very clearly that the inferred de-stigmatization of cohabitation is matched by the explicitly measured de-stigmatization of the other two ethics issues. The results of Figs. 10.1 and 10.2 plainly indicate that for each period considered there is a clear education gradient, with the rejection of euthanasia and homosexuality weakening with advancing education. Conversely, the degree of de-stigmatization increases with education. In addition, the rejection of euthanasia and homosexuality rapidly weakens over time, with much smaller percentages taking a negative view in the twenty-first century measurements compared to those of the 1990s. In other words, these findings are in line with the interpretation that the de-stigmatization started in the higher education strata and, in tandem with advancing education, spread to the society as a whole.Footnote 15 So far, that positive gradient of ethical tolerance and education has remained intact. Hence, with respect to the “ethics revolution” there is no contradiction between the upward cohabitation trend, the education related gradient, and the shifting educational composition. The top to bottom diffusion of the de-stigmatization and the increasing levels of education operate in the same direction, and probably reinforce each other in accelerating the trend.

Fig. 10.1
figure 1

Percentages of population 18+ of the opinion that homosexuality is never justified, by education and period (Source: Authors’ elaboration based on World Values Surveys)

Fig. 10.2
figure 2

Percentages of population 18+ of the opinion that euthanasia is never justified, by education and period (Source: Authors’ elaboration based on World Values Surveys)

It is also interesting to note that the de-stigmatization profiles by level of education are at present indicative of more permissiveness in a number of Latin American countries than in the US. By 2005–06, the percentages never accepting homosexuality are lower in Brazil, Argentina, Chile and Uruguay, i.e. the countries with the largest white populations. With respect to the de-stigmatization of euthanasia, the US is still in the vanguard, but matched by Uruguay and Chile.

To sum up, there are two strong arguments that are in favor of the hypothesis that the trend reversal phase since the 1970s is fuelled by SDT factors as in Northern America and Europe. First, cohabitation very clearly increased among the middle and upper education groups, meaning that consensual unions are breaking loose from their ethnic and economically disadvantaged substratum. And, secondly, the de-stigmatization or “willingness”-factor operated entirely in the expected direction, both with respect to the positive tolerance gradient and the concurrent upward compositional shift in education. In other words, the reversal phase since the 1970s is largely induced by factors that are congruent with the SDT theory.

7 The Cohabitation Boom in Settings Without a Major Ethno-Racial Component

A widespread view of the rise of cohabitation in Latin America is that it should not come as a surprise, since these countries “always had it”. This standard view is evidently oblivious to the steeply upward trends of cohabitation in Southern Brazil and the Conosur (“Southern Cone” composed of Uruguay, Argentina and Chile), i.e. the four areas that have only small indigenous or mixed populations, and are largely made up of descendants of European immigrants. In this large Southern Cone, the ethno-racial component of the negative cohabitation gradient by education is largely absent. However, the negative gradient with education is equally in evidence, but then mainly connected to pure social class distinctions among whites. Moreover, due to their European origins, cohabitation was much less common in these regions than in the rest of Latin America during the 1960s and 1970s. In other words, there was no model at the onset that justified cohabitation, and the result was a strongly negative view of it. Despite internal differences among the Conosur countries in terms of their educational expansion, the development of welfare state provisions, political stability, and economic shocks, these areas have had the largest increases in cohabitation since the 1970s of the entire American continent. Particularly striking is the rise of cohabitation in Uruguay which trumps that in all other countries. Moreover, the negative gradient with education in Uruguay had almost disappeared in 2010. Hence, the classic argument that the ethno-racial component in Latin America triggers off the cohabitation boom is incorrect: white populations of European descent equally experienced the phenomenon, and even to a more marked degree.

The other region with a strikingly steeply upward trend in cohabitation is Quebec province in Canada. This is another area with a dominant white majority, who had in addition guarded its French language and its strong allegiance to the Catholic Church till the “Quiet Revolution” of the late 1960s. Even more striking is that the education related profile of cohabitation in Quebec did not display the negative gradient during the entire “reversal phase”. In fact, in 1986, the highest incidence of cohabitation existed among women with a university education (see Figs. 3.2a and 3.2b in the Canadian), and the gradient becomes essentially flat thereafter as the new behavior spreads very rapidly to the rest of the Quebec population. This occurred concurrently with a major secularization wave and the demise of Catholic authority. Furthermore, Quebec did not experience any major economic setbacks as the Conosur countries did, so that the “crisis” hypothesis has no empirical grounding in this part of Canada. The case of Quebec is a perfect, if not an extreme example of a Western European pattern of the SDT. But one could also argue that, in terms of cohabitation levels and lack of social stratification differentials, “Uruguay became the Quebec of Latin America”.

8 Patterns of Entry into Cohabitation and Mixed Types

At various points it has been stressed that traditional patterns of cohabitation with either an ethno-racial or a plain social class origin and the new SDT-type of cohabitation have also produced blended types. Such intermediate types can be studied from different angles. Esteve et al. (2012) use the characteristic of residence in an extended household, as opposed to the formation of a nuclear household, as a criterion for evaluating the maintenance of traditional form of marriage and cohabitation. Covre-Sussai et al. (2015) use DHS surveys to construct a three-way typology of cohabiting women depending on the maternity paths followed prior to the union and after cohabitation. Grace and Sweeney (2014) focus on the onset of sexual activity of adolescents and young adult women in Central America and the consequences for entering into a consensual or marital union.

The Esteve et al. study compares the percentages of women 25–29 in extended or composite households (as opposed to nuclear households) for cohabiting couples, married couples, cohabiting mothers, married mothers and single mothers. Again census data archived in IPUMS files are used. Of the 13 countries considered, three Andean ones, i.e. Bolivia, Ecuador and Peru, had the highest co-residence with parents or others for both cohabiting and married women 25–29. In these countries the percentages were similar for these two categories and situated between 50 and 60 %. Also for women with children, co-residence with parents or others remained high at around 30 %, and again with little difference between cohabiting and married mothers. Evidently, in these countries traditional co-residence in extended households is still very common, and there is no distinction between cohabiting and married women. The next group is made up of Cuba, Panama, Puerto Rico, Venezuela and Colombia, with 40–50 % of cohabiting women 25–29 residing in extended households. However these countries exhibit more diverging figures for percentages of married women in extended households. In Cuba, more married women than cohabiting women live together with parents or others (51.3 % vs. 44.7). By contrast, in Puerto Rico, co-residence is much more common for childless cohabiting women than for married ones (41.9 % vs. 14.6). In the other countries of the group, there are also more cohabiting women in extended households, but the difference with the married women are less pronounced (around 10 percentage points). Apparently in these countries the economic situation plays a prominent role in determining the outcome for childless cohabitors, with more precarious situations for them leading to prolonged residence with parents or others. For cohabiting and married mothers, however, the marital status distinction vanishes. Evidently, cohabitors split off from the extended family a bit later and upon the birth of a child. In the remaining countries, i.e. Mexico, Costa Rica, and Chile, co-residence in an extended household for childless cohabiting women drops below 40 % and in Brazil and Argentina even further below 30 %. In all these instances, co-residence with parents or kin for married women is lower, thereby again illustrating that the more precarious situations of cohabiting women are to some degree compensated by prolonged residence in the family of origin.Footnote 16 Hence, there is again a geographic clustering of the patterns with (i) an Andean form in which both cohabitation and marriage are most commonly occurring with prolonged co-residence with kin, (ii) a Central American and Caribbean one with lower overall co-residence, and with more cohabitating than married women staying in the extended family, and (iii) a more diluted pattern with less co-residence with kin among cohabitors and much less among married women.

In these respects, the contrast with European patterns of residence is striking. The Western and Northern European cohabitors and single mothers rarely derive support from co-residence in extended families, since the European historical pattern is overwhelmingly that of neolocal residence of nuclear families. Hence, there is a major type of cohabitation with co-residence in extended families in Latin America that is completely distinct from the European or US and Canadian pattern. This contrast is plainly rooted in the different historical patterning of household formation, spanning at least over several centuries.

The Covre-Sussai study is based on the 2005-2010DHS surveys in eight countries, and uses latent class analysis and retrospective data to construct a typology of cohabiting women (all ages). The classification criteria are the age at the start of cohabitation, the number of children and the ages at motherhood (1st birth), pre-cohabitation pregnancy or not, and currently living together with partner or not. Controls are introduced for age and education. The results indicate that between a traditional form and a modern form there is also a mixed group. The traditional group has the earliest age at the start of cohabitation (typically before age 19), and had children before the age of 20. They are concentrated among the younger women (younger than 26), women with primary education only and resident in the Dominican Republic, Nicaragua and Honduras. In other words, the typology also picks up the Caribbean and Central American pattern of cohabitation. The contrasting group (“the innovative group” according to the author) has a later age at entrance in cohabitation and of motherhood (over 20), had no pre-union pregnancy, and the highest incidence of still being childless. This type is most common among women with secondary education. Brazil has the highest proportion of this “innovative” type (43 %), but in all the other countries the incidence is between 30 and 38 %. The intermediate type in Covre-Sussai’s analysis resembles the more modern one. The main difference is that they all had a pre-union pregnancy and no childlessness, but otherwise their profiles are similar to the “innovative” group. This intermediate group has the smallest occurrence in the Central American and Caribbean countries, and also a smaller presence in Brazil,Footnote 17 but was more common (again about a third) in Bolivia, Colombia, Peru and Guyana. Besides capturing an educational difference, the typology also identifies an Andean pattern as being distinct from the Central American-Caribbean one.

A further study of the life-course unfolding in Central America (Grace and Sweeney 2014) focuses on the adolescent and young adult stages (ages 12–24) in Guatemala, Honduras and Nicaragua. Data stem from the DHS and RHS surveys from 2001 to 2009. The authors use an event history analysis of competing risks for entering a consensual union or marriage. At this point we must recall that the Central American region harbors many populations that already had a high to very high incidence of cohabitation to start with and still have the earliest ages for women at entering a union (Bozon et al. 2009). Hence, it comes as no surprise that the new SDT-form of cohabitation adds little to the already high percentages in consensual unions. Also, as expected, the analysis brings out that the start of a sexual relationship and potential pregnancy spur on the formation of a union at very young ages, i.e. before age 18 (Ibidem). However, by staying in school longer, the onset of sexual relations is delayed, and later on, further education is again linked to a higher probability of entering a marriage. But there is also an important ethnic effect: Mayan women in Guatemala have a greater likelihood of entering marriage, even at young ages, than women in the other two countries. As already indicated, this matches the much lower incidence of cohabitation of the Mayas of Yucatan in Mexico. In Honduras and Nicaragua, by contrast, the early onset of sexual activity strongly increases the probability of entering a consensual union and has little impact on the likelihood of marrying.

These three examples clearly bring out the heterogeneity within Latin America in patterns of partnership formation. In addition, they elucidate the differences with respect to family context and possibilities for co-residence with parents and kin. And, thirdly, historical factors associated with ethnicity are emerging again, even within much smaller regions such as Central America.

9 The Unfolding of a Latin American Duality: Expanding SDT and Persistence of the Pattern of Disadvantage

The original conceptualization of the SDT three decades ago (Lesthaeghe and van de Kaa 1986) was essentially the description of a Northern and Western European phenomenon. The SDT-theory had two central components: the “non-conformist” aspect, referring to the non-marital union formation and parenthood, and the “postponement” aspect, referring to the postponement of marriages and parenthood to much later ages than recorded in Europe during the 1960s.Footnote 18 In this European conceptualization, effective contraceptive methods disconnected the link between the start of sexual activity and marriage, and also the rise of cohabitation lead to the postponement of parenthood. Hence, the “non-conformist” and the “postponement” parts were very closely linked in time in that part of the world. The same was also observed in the US and Canada. Later on, however, it became more obvious that these two dimensions did not necessarily have the same determinants. The ideational changes in emancipation or expressive values and in ethics were more strongly predictive of the “non-conformist” part than of the “fertility postponement” part of the SDT.Footnote 19 In fact the relationship with values orientations operated the other way: it was parenthood that systematically altered these values in the conservative direction (Surkyn and Lesthaeghe 2004). Moreover, as the SDT spread beyond the Northern and Western European sphere, it became even more evident that the two aspects could be disconnected in time as well (Lesthaeghe 2014).

The Southern European pattern constitutes a second variant of the SDT. These countries had started their fertility postponement and fertility levels dipped far below replacement level without any signs of emerging cohabitation. The initial reactions in Spain and Italy to the SDT-theory was “not us, we’re different”, and after the fall of Communism, identical reactions were voiced in Central and Eastern Europe. Also there, fertility dropped precipitously as a result of massive postponement. After the turn of the Century, however, cohabitation did rise in these parts of Europe as well. The outcome is that with very few exceptions, European populations have had sub-replacement fertility for up to four decades, and rising levels of cohabitation as well, with the Nordic countries presumably reaching an upper ceiling. Marriage has obviously not disappeared, but when it occurs, it is at a later stage in the life-cycle and no longer of necessity at the occasion of a first birth.

As in Southern Europe, also in Japan the postponement transition of marriage and of fertility had already ran much of its course prior to the first signs of emerging cohabitation. It is only several years after the turn of the Century that demographers realized that Japan too was witnessing the emergence of consensual unions (Tsuya 2006; Raymo et al. 2009). Furthermore, also data for Taiwan illustrated the same phenomenon (Lesthaeghe 2010). Admittedly, the incidence of cohabitation is still lower than in Europe or in Latin America, but these examples nevertheless illustrate that much more strongly “patriarchal” societies in the Far East are not immune to the manifestation of the “non-conformist” part of the SDT.Footnote 20 It should be noted that maternity without marriage is still exceptional in Japan, whereas this is no longer so in Southern Europe and particularly not in Spain and Portugal.

In the Latin American situation, the sequence is reversed: as documented in this volume the cohabitation boom developed without the postponement effect of union formation and of fertility. This constitutes a third variant of the SDT. Fertility levels declined substantially since the 1970s in a number of countries, but this occurred without a major shift in its timing. In 1970, total fertility rates were above three children in all Latin American countries with Uruguay as the sole exception. By 2010 all countries but Bolivia, Guatemala, and Haiti were below three children per woman. A number of countries even dipped below replacement fertility: Chile, Costa Rica, Cuba, El Salvador and Uruguay (CELADE 2013). Despite such significant declines in fertility levels, women’s mean ages at first union and at first birth remained quite stable across cohorts and time. This has been a puzzling characteristic of Latin American family systems that sets them apart from western countries in which the “non-conformist” and “postponement” transitions occurred simultaneously.

This feature of stable mean ages at union formation and ages at maternity has attracted a fair amount of interest (e.g. Fussell and Palloni 2004; Esteve et al. 2013; Castro-Martín and Juarez 1995). Also, improvements in education were not accompanied by an expected overall tempo shift in fertility. Rather, opposite tendencies occurred at the extremes of the education spectrum. Recent analyses of census and survey data indicate that the women with tertiary education tend to postpone their first birth in a number of more developed regions, but also that teenage fertility is rising in the lower and middle education groups (Rosero Bixby et al. 2009; Esteve et al. 2013). Chile and Uruguay show the largest increases in childlessness among the best educated women, followed by Brazil and Mexico. In the Sao Paulo state of Brazil an increase in fertility among women 30+ is being noted among the wealthier strata (Berquo et al. 2014), which equally points in the direction of postponement and subsequent recuperation at later ages. These features are in line with the SDT scenario.

High teenage fertility, even before the age of 18, constitutes the other side of the coin and points in the direction of a persistent pattern of disadvantage. The DHS survey data for 12 countries show high and stable proportions of women with children by age 18 across cohorts (Esteve and Florez 2014). For some authors high teenage fertility is regarded as the main reason for the stable and low mean age at maternity (Rodríguez Vignoli 2008). Early ages of starting sexual activity in combination with deficient contraception among young women account for this to a significant degree. However, all indicators show that the use of contraception has increased throughout Latin America, before and after controls for factors such as education, age at sexual début, current age, among others. Therefore, there should be additional explanations as well. Rodríguez points to three additional factors: (i) weak autonomy for young women, (ii) a lack of economic opportunities for them and hence low opportunity costs associated with early maternity, and (iii) the availability of family support (e.g. co-residence).

The overall outcome for Latin America is the duality with increasing postponement of first births among an educated elite and high and often rising adolescent and teenage fertility among the most disadvantaged parts of the population (López-Gay and Esteve 2014). Among the former, the full SDT pattern is currently unfolding, whereas the latter have increased cohabitation in combination with very early fertility schedules. It remains to be seen to what extent the central category with secondary education will be following the elite. If they do so, a top-down pattern of fertility postponement would be followed, leading to lower period rates of total fertility in a number of better educated countries. However, a tenacious persistence of high teenage fertility pattern is highly likely, even when overall educational levels continue to increase. In fact, despite the striking advances in contraceptive technology, such a history of high teenage fertility has been observed in the US until the recent turn of the Century. Hence, high teenage fertility is an additional feature which sets the Latin American and Caribbean countries apart from most of Europe and the Far East.

10 Final Note

This entire volume deals with evolutions in partnership formation which are still in full progress. Admittedly, in some countries that evolution advanced with a big leap, whereas in others the trends have been more gradual. But in all cases these trends are following a firm course, irrespective of the economic ups and downs. What we are witnessing is not just “a temporary aberration” but a genuine systemic alteration covering an entire continent. The Americas, as opposed to many Asian societies and Africa, are now following in the European footsteps, be it with their own distinct and path-dependent characteristics associated with regionally varying historical antecedents.