Italy’s Non-Negligible Cohabitational Unions

La cohabitation hors mariage en Italie: un phénomène non négligeable
Open Access

DOI: 10.1007/s10680-009-9193-y

Cite this article as:
Gabrielli, G. & Hoem, J.M. Eur J Population (2010) 26: 33. doi:10.1007/s10680-009-9193-y


Italy has long been regarded as the country with negligible non-marital cohabitation par excellence, but lately the pattern has begun to change and entry into consensual unions has increased strongly in younger Italian generations. This article is devoted to a study of such features between 1980 and 2003 based on the data from the Italian variant of the Gender and Generations Survey, Round 1. We consider entry into marriage and entry into cohabitation as competing risks and show how the incidence of cohabitation consistently much lower but has increased by some 70% over the 20-odd years of our study, while the marriage rate has dropped by almost as much. We find great variation across major regions of the country. The rise in cohabitation is confined to Northern and Central Italy, while the risk of marriage formation has declined strongly all over the country. Unlike previous investigations, our data suggest that non-marital cohabitation may be taking over whatever minor role civil marriage has had in Italian union formation.


Union formation Marriage Cohabitation Italy Geographical pattern Event history analysis Cox model Competing risks Generation and Gender Survey 


L’Italie a longtemps été considérée comme le pays par excellence où la cohabitation hors mariage est exceptionnelle, mais la situation a récemment changé, compte-tenu de la forte progression de ce type d’union au sein des jeunes générations en Italie. Cet article, qui est consacré à l’étude du choix de la forme d’union entre 1980 et 2003, s’appuie sur les données de la première vague de l’Enquête italienne Générations et Genre. Nous considérons que le choix de débuter une union par une cohabitation hors mariage et le choix de la débuter par un mariage sont associés à des risques compétitifs, et démontrons que l’incidence de la cohabitation est toujours plus faible que celle du mariage, mais qu’elle a augmenté de près de 70 % sur la vingtaine d’années couvertes par notre étude, pendant que l’intensité du mariage baissait d’autant. Nous observons également d’importantes variations entre les grandes régions du pays: la hausse de la cohabitation est limitée à l’Italie du Nord et du Centre, alors que la nuptialité a fortement décru dans tout le pays. A l’opposé des travaux antérieurs, nos données suggèrent que la cohabitation hors mariage pourrait être en train de supplanter le mariage civil -au rôle toutefois réduit- comme forme d’union en Italie.


Mise en couple Mariage Cohabitation Italie Variations géographiques Analyse des biographies Modèle de Cox Risques compétitifs Enquête Générations et Genre 

1 Introduction

In line with general trends across Western Europe (Sardon and Robertson 2004), Italian marriage rates declined sharply after the mid-1970s. Initially, consensual unions spread more slowly in Italy than in other parts of Europe (Kiernan 1999; Bernhardt 2004), but recently there has been a strong increase in non-marital cohabitation among younger Italian generations also (Barbagli et al. 2003; Rosina and Fraboni 2004; Di Giulio and Rosina 2007). Moreover, consensual unions have stayed non-marital increasingly longer during recent years (ISTAT 2006). These developments have attracted considerable interest and have been studied from a number of angles (Billari et al. 2002; Nazio 2007; Castiglioni and Dalla Zuanna 2009; Kertzer et al. 2008; and others). In this article, we present a new take on these issues in that we contrast trends in the incidence of cohabitation explicitly with simultaneous trends in marriage formation rates and find that the latter have been the stronger in all periods studied. In this process, we replicate for Italy analyses carried out for comparable countries in Central and Eastern Europe (Hoem and Kostova 2008; Hoem et al. 2009a, b; Matysiak 2009) and focus on period developments instead of the changes between birth cohorts that have frequently been highlighted in other contributions about Italy. The focus on period developments allows us to pinpoint accurately changes in calendar time that are harder to locate in cohort studies (cf. Ní Bhrolcháin 1992). The predecessors just mentioned have found that standardized cohabitation incidences have exceeded corresponding marriage-formation risks in Russia, Hungary, and Bulgaria at least since a decade ago, but not at any time in Romania or Poland. The latter is the non-Mediterranean European country which should be most similar to Italy, given the dominance of the Roman Catholic influence in both the countries. Interestingly, we find a pattern in Italy very similar to that in Poland. This answers one of the questions that have induced us to carry out this study in the first place. We have been lead by our curiosity about behavior in Italy more than by any general theory or by a drive to formulate and answer theoretical hypotheses; in our opinion it is too early to attempt such an approach. The closest we have got to a general theory is to address the question whether in Italy we can find traces of the narrative of the Second Demographic Transition (Lesthaeghe and van de Kaa 1986; Lesthaeghe and Surkyn 2002; Sobotka 2008) similar to what our predecessors have found for a number of other countries, and as we shall see below, we can answer this question in the affirmative.

Another question that has attracted our attention is whether the well-established tradition of behavioral differentials between the North and the South of the country persist in union formation as well, differentials that are normally explained by the differences in the economic and cultural history of these major regions (De Sandre 1997; Dalla Zuanna and Righi 1999; Castiglioni and Dalla Zuanna 2009; Kertzer et al. 2008). We follow-up on that tradition by employing our analytical methods to Italy’s North and South, separately and jointly.

Taking Italian peculiarities into account, we also include the distinction between civil and religious marriages (Dittgen 1995; Kertzer et al. 2008), in the expectation that a civil marriage will prove to be an indicator of secularity which in some way competes with cohabitation, perhaps as a kind of half-way house. We have no a priori hypothesis about the relative role of civil marriage in this picture, but we do underline that a standardization procedure like ours (see below) is needed to avoid distortion by compositional effects; a distortion that can be present if one just computes straightforward percentages of civil marriages among all marriages.

2 Data and Method

The data used were retrieved from the Italian version of the Gender and Generations Survey and collected in November 2003.1 The data have a net random sample of 10,960 Italian women aged 15–60 at interview, and they contain retrospective histories that cover the period 1980–2003 and more.2 The event we study is self-reported first-marriage formation, alternatively entry into a first consensual union.3 In order to compare our results directly with those of previous articles, we have run three sets of Cox event-history models,4 namely for the following transitions: (1) entry into first union by type (marital and non-marital union formation separately); (2) competing transitions to cohabitation and marriage analyzed jointly; and (3) conversion of cohabitation into marriage. The procedures have been described completely by Hoem and Kostova (2008) and Hoem et al. (2009a, b, Appendix). When we run these analyses, we pay particular attention to trends over calendar periods and check whether the drop in the marriage hazard is compensated by a simultaneous increase in cohabitation.

To hedge against compositional effects we have standardized all analyses with respect to the respondent’s age attained, municipality size, parity-and-pregnancy status, father’s and mother’s educational attainment, number of siblings, the respondent’s own educational level, occupational status, and whether the respondent lived with both parents at age 15. (For details see Appendix Table 4.) Most of these control variables have been coded so as to repeat the elaborations made by Hoem and Kostova (2008) and Hoem et al. (2009a, b).5 Our analyses do not show any particular surprising risk patterns for our control variables, so we confine our findings to Appendix Table 5 and concentrate attention on our main topic, which is the period trend in union-formation behavior. The only exception will be the analysis of an interaction between the parity-and-pregnancy status and the type of union in our model for the hazard of the first-union formation.

In a final analysis catering to the particular case of Italy, we run the competing risk analyses mentioned above for Southern Italy separately from the rest of the country,6 a practice which will allow us to compare the diffusion of cohabitation in the various geographical contexts. Moreover, we include civil marriage separately from religious marriage because authors of recent research articles (Bernardi and Gabrielli 2006; Kertzer et al. 2008) have demonstrated differentials between people with civil and religious marriages in Italian fertility behavior and we are curious to see how this carries over to union-formation behavior.

3 Non-Marital Cohabitation as a Competitor to Formal Marriage

3.1 General Trends

In Fig. 1, trends in (standardized) risks of entry into first cohabitation and into first marriage appear in a conventional analysis where the two life course transitions are treated separately.7 The results are in line with other recent studies (e.g., Di Giulio and Rosina 2007) and show that, after a period of hesitation in the 1980s, the risk of entry into cohabitation almost doubled in Italy during the last 15 years. Conversely, the standardized risk of entry into a first marital union was more than cut in half in the same period. This demonstrates the increase in first consensual unions easily, but we need to run a competing-risk analysis if we want to compare the two hazard rates directly and see differentials in the period effect on each transition, standardized for the other covariates.
Fig. 1

Trends in (standardized) relative risks of union formation. Separate single-decrement analyses by type of union. Childless non-pregnant Italian women, 1980–2003. Source for all diagrams and tables: Our own computations from the Italian GGS data of 2003

3.2 Joint Analysis of the Competing Risks

The first step in our competing-risk analysis is to consider the parity-and-pregnancy-status covariate, which distinguishes between (i) non-pregnant childless women, (ii) pregnant childless women, and (iii) mothers, i.e., women at parities 1 and above. In Table 1, we have subdivided 59,384 person-years of exposure and observe as we would expect that women who have no children and who are not pregnant during a period of exposure to the risk of union formation completely dominate the picture.8 For this reason, we concentrate on this group in most of our analysis and censor the records for the never-partnered women at the occurrence of a pregnancy.
Table 1

Numbers of person-years of exposure in the GGS, by parity-and-pregnancy status

Parity-and-pregnancy status



Childless non-pregnant



Childless, pregnant



Parity 1+ (mother); women who entered a stable union in our data



Parity 1+ (mother); women who did not enter a stable union in our data






Italian women, 1980–2003

Meanwhile Table 2 provides the interaction between the type of union formation and parity-and-pregnancy status. The estimates represent a kind of risk average over the 20-odd years since 1980. The first (and expected) result is that the risk of entry into a non-marital union among childless non-pregnant women is very low in comparison to the corresponding risk of entry into marriage for the same group. The increase of cohabitation in Italy is so recent that it cannot affect this relationship yet. As shown by Hoem et al. (2009a, b) for some other countries, the entry risk for a marital union increases strongly when an Italian woman becomes pregnant. (For Italy the relative risk is 25.6.) There is also a strong increase in the entry risk for a consensual union; it increases by a factor of almost 12.9 These results confirm the persistence of shotgun weddings (and other early unions) in Italy. Finally, a woman at parity 1 or more10 unsurprisingly runs a much higher risk of entry into a marital union than into a consensual union (the risk is 2.8 times higher), and both risks are higher than for a childless non-pregnant woman. In this case, we cannot observe a negative impact of the arrival of the first child on union formation described for some other countries (Hoem et al. 2009a, b).
Table 2

Standardized relative risks of first-union formation, by parity-and-pregnancy status, for each type of union

Entry into

Childless, not pregnant

Childless, pregnant

Parity 1 + (mother)

Marital union (direct)




Non-marital union




Italian women, 1980–2003

All values are 99% significant

Focusing then on the majority group of childless, non-pregnant women,11 we estimate trends over calendar periods in the competing risks of entry into cohabitation and into marriage jointly to check whether a drop in the marriage risk is compensated by an increase in cohabitation of a similar size order or more. Figure 2 shows that such a pattern does not occur in the Italian case: the sharp decrease of marital union is only marginally compensated by an increase in the standardized entry risk for consensual unions and, above all, there is no cross over between the curves during our period of observation.
Fig. 2

Trends in (standardized) relative risks of union formation. Competing risks by type of union. Childless non-pregnant Italian women, 1980–2003

3.3 Conversion of Cohabitation into Marriage

Our predecessors have located traces of the Second Demographic Transition without attempting to establish a full-scale determination of the presence of its features. In a similar partial hunt in our Italian data, we would look for a pattern where ever-more widespread cohabitational unions last steadily longer. To check up on this possibility, we focus on the risk of conversion of a consensual union into a marriage, estimated according to cohabitational duration and calendar period. After considering various other specifications (not shown12), we have located two very clearly different patterns, namely one for the years before 1990 and one for later years (Fig. 3). For the period 1980–1989, we observe a peak of the conversion rate during the first year of the consensual union, and so little exposure after 6 years of union duration that there is no point in estimating a conversion risk. In this calendar period, a consensual union must largely have been a temporary phase before marriage. In the second period (1990–2003), the conversion rate loses its peak in the earliest months and is much the higher thereafter. This means that many conversions are postponed to longer durations, and in fact we get an increase in the mean union duration before conversion of around 16 months. The rise in the conversion rate fits with findings in official statistics (ISTAT 2006), and we take it as a mild manifestation of an element of the Second Demographic Transition.
Fig. 3

Trends in the (standardized) relative risk of conversion of cohabitation into marriage, by union duration. Childless non-pregnant Italian women, 1980–2003

3.4 Special Features in Southern Italy

Taking account of the particular structure of union-formation behavior in Italy, we run the competing risk analyses described above once more with Southern Italy13 counted separately from the rest of the country. We have also separated civil marriages from religious ones and have worked with three competing risks in this part of the analysis. The results are shown in Fig. 4, and as expected we find clear geographical differences in the patterns of union formation.14 Southern Italy had a noticeably higher tendency to marriage formation than the rest of the country during the 1980s and 1990s, but the standardized risks fell all over Italy and both regions reached about the same level after the turn of the century (Fig. 4c). The recent increase in entry into a consensual union has been confined to Northern and Central Italy and has not extended to Southern Italy, where the standardized risk of entry into a non-marital union has remained quite low and stable (Fig. 4a, b).
Fig. 4

Trends in (standardized) relative risks of union formation. Competing risks by type of union for each region of residence at birth. Childless non-pregnant Italian women, 1980–2003

The relative risk of entry into a first civil marriage has stayed very low so far, and according to our data it even decreased in both parts of Italy during our period of observation (1980–2003).15 In Northern and Central Italy, first consensual unions seem to substitute for first civil marriages in the same period. We are not aware that this detail has appeared in the literature before; no one before us seems to have discovered this partial conversion of first civil marriages into first cohabitations over time. Unfortunately, we cannot pursue this issue for we do not have sufficient data for a robust analysis.

Finally, the analysis of the conversion of a consensual union by region (not shown) suggests an increasing tendency for marriage to occur after 13–24 months in the South. This represents a moderate increase of union duration before conversion in this region in contrast to the greater stability of the North-Center.

4 Discussion

The results reported above show a clear increase in the incidence of consensual unions during the last 20 years in Italy, at least in the North and Center of the country. We also observe a sharp postponement of the first marital union in the same period (all over the country). Given the long-standing institutional disincentives against consensual unions in Italy (Nazio 2007; Barbagli et al. 2003; Mencarini and Tanturri 2006; Castiglioni and Dalla Zuanna 2009; and others), what we have found must be the reflection of a change of heart rather than, say, an improving labor market or changing housing costs and welfare conditions. Nazio and Blossfeld (2008, p. 4) describe the mechanism clearly and associate the rising incidence of the new cohabitations with “a better knowledge and understanding of” the phenomenon. As they claim, “[l]ater birth cohorts will … experience cohabitation as less deviant (or stigmatised) and more socially accepted right from the beginning.” The role of the old generation is particularly important in a country like Italy where there are such strong ties between parents and their grown children (Barbagli et al. 2003; Rosina and Fraboni 2004; Schröder 2008; Castiglioni and Dalla Zuanna 2009). In the words of Di Giulio and Rosina (2007, p. 460), “it is … strategically important for a young adult to receive parental support in the critical events of his/her life” in such a society. In this sense, the new-found acceptance of a practice that was unconventional among parents would lead to an increase in consensual unions among their adult daughters. If this observation is correct, it would confirm the changing values and increasing tolerance in family matters that is supposed to be part of the basis for the Second Demographic Transition.

On the other hand, we analyze entries into a non-marital union jointly with entries into a marital union in this article, and this procedure allows us to observe an inertial force in the early process of the diffusion of cohabitation. The sharp decrease in marital unions is only marginally compensated by the increase in consensual unions. Although the persistence of economic uncertainty among young Italians, the lack of an efficient welfare system in the country, and the rigidities of the local housing market undoubtedly continue to be obstacles to the diffusion of cohabitation in Italy, as they do in most Southern European Countries (Nazio and Blossfeld 2008), we believe that traces of a value change are part of the particular Italian pattern of a shift in behavior.

So far, non-marital unions have been entered mostly by a selective group of the Italian population, namely by secularized, educated, and working young people in the country’s northern urban areas (Di Giulio and Rosina 2007; also note the structure of relative risks in our Appendix Table 5). Cohabitation is still considered by most Italians as a temporary phase before marriage and not as a permanent alternative to it (cf. Fig. 3). This could facilitate its acceptance among parents of the young generation, but it would still encounter mental inertia because of the preponderance of marriage.

In our analysis, we have paid particular attention to the geographical pattern and showed once again that the recent increase in consensual unions appears mainly in Northern and Central Italy and not in the South, where non-marital unions still remain quite rare. Let us add that in her Ph.D. thesis based on qualitative data, Schröder (forthcoming) explains that “despite [the observation that] women in two different Italian regions [may] show similar behavior, namely cohabitation, their attitude differs towards this behavior: the Northern woman prefers to cohabit because she likes it; the Southern [woman] chooses cohabitation as a result of economic constraints (Chap. 6.3.3)”. Italy displays at least two different settings for new union forms.


For a description of the GGS program, see Vikat et al. (2007). For a description of the Italian survey, see ISTAT (2006). For some summary features of our data, see Appendix Table 3.


Among our respondents, 4,926 women had not experienced a first union (yet) and were censored at interview. We have dropped cases with missing or unacceptable information and thus lost as little as 0.24% of the cases.


This covers most unions since Italy (along with all other Southern European countries) have uniquely low levels of divorce and separation rates (Sardon and Robertson 2004). In all of our analyses, we consider respondents who have had an informal union before a first formal marriage as individuals who first entered a consensual union.


In all our analyses, we have used the program STATA 10 with a non-parametric baseline hazard. Previous authors have used piecewise-constant baseline hazards in their articles, but this difference causes no problem for the comparisons we make. In any case, the use of event-history analyses constitutes a form of indirect standardization.


We exclude the variable ‘ethnicity’, which was not collected in the Italian data, and include occupational status (employed versus unemployed) in our analysis. These changes are indispensable in the Italian context; anyway they do not affect our comparative purpose as they do not change our final results much. (We have also carried out the analyses without these deviations from previous practice but do not give any documentation here.)


Behavior in Northern and Central Italy was sufficiently similar for us to combine these two regions.


The effect of the covariates included in the models is shown in Appendix Table 5.


The percentage of non-marital births have increased from 4.2% of all births in 1980 to 12.2% in 2002, but this is still one of the lowest values in Europe (Sardon and Robertson 2004).


We get a relative risk of 2.09/0.18 = 11.6.


The number of these cases is low in our sample. (“Only” 165 women get a child before first union formation.) It might have been interesting to elaborate this result further, but we avoid doing so as it is not among the topics we focus on in this article.


Because of the exposure dominance of the non-pregnant childless women, the results would not have been much different if we had considered all women taken together.


We find no particular change in the pattern when we subdivide each of the two periods considered.


Following previous authors (Bernardi and Gabrielli 2006; Kertzer et al. 2008), we have included six regions in what we call Southern Italy, namely Molise, Campania, Puglia, Basilicata, Calabria and Sicilia.


In Fig. 4a, b, we have carried out separate analyses for the two regions. In Fig. 4c, we have made a single analysis for the whole country with region of residence at birth as an additional fixed covariate. For clarity of presentation Fig. 4c contains only two of the six curves that result from the latter exercise. The patterns left out just replicate Fig. 4a, b.


The relative risk dropped from 0.11 to 0.05 in the Southern region and from 0.16 to 0.06 in the North-and-Central region. Correspondingly, the percentage of civil unions (among all first unions) dropped in our sample from 9.9% in 1980-1984 to 7.8% in 2000–2003 (Appendix Table 3). Others have found an increasing percentage of civil marriages among all marriages (e.g., Barbagli et al. 2003, and particularly ISTAT 2007), but such percentages are subject to compositional effects that are removed by the standardization that we use. Barbagli et al. (2003) note that the slight rise which they found in first civil marriages in Italy only concerns marriages with at least one foreign spouse (see also Gabrielli et al. 2007). ISTAT (2007) has later found the same, but also note a corresponding rise between 1995 and 2004 if attention is restricted to first marriage only among Italians, excluding non-Italian citizens. The strength of the relationship in their study worries us for it may mean that civil marriages have been recorded inconsistently in the data sources. Given the potential usefulness of a civil marriage as an indicator of secularity, the issue is worthy of further investigation.



We thank Laura Bernardi, Aiva Jasilioniene, and Christin Schröder (now Löffler) warmly for their useful comments and suggestions. We have also benefitted from comments from the journal’s editor and two reviewers.

Open Access

This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.

Copyright information

© The Author(s) 2009

Authors and Affiliations

  1. 1.University of BariBariItaly
  2. 2.Max Planck Institute for Demographic ResearchRostockGermany

Personalised recommendations