1Over the last forty years, non-marital cohabitation has become widespread across Europe, including in central and eastern Europe, but the social characteristics of the couples concerned vary widely across countries. In Romania, the prevalence of cohabitation is low, especially among the most educated, who marry more frequently than the others, both directly or after a consensual union. Using data from the Romanian Generations and Gender survey 2005 (GGS), Jan Hoem, Cornelia Mure?an and Mihaela H?r?gu? measure the fertility of marital and non-marital unions by duration of union, using moving averages to limit random variations due to the small sample sizes. They propose marital and non-marital agreggate fertility rates for several population subgroups. The authors show that over a long period, fertility in consensual unions is similar to that of marital unions, although the fertility of low-educated persons and persons in consensual unions had already fallen sharply at the end of the 1980s following the collapse of the Communist regime.
2Until the late 1980s, time spent in non-marital cohabitation in Romania represented only a small fraction of the total time spent in unions every year, but since the late 1980s, consensual unions have become much more prevalent, and by 2005 the overall proportion of time spent in such unions had increased to some 10%, while the fraction of time lived in direct marriages (i.e. marriages not preceded by a consensual union between the two partners) had dropped from around 90% to some 70% (Figure 1). Similar developments were common across Eastern Europe, with an increase in cohabitation and a decline in direct marriage towards the end of the socialist period (Thornton and Philipov, 2009 ; Philipov and Jasilioniene, 2008 ; Spéder, 2005 ; Hoem et al., 2009ab). The change-over between family forms accelerated after the change of political regime in each country, though the pattern of transition was far from uniform. Compared with (say) Hungarians, Bulgarians, or Russians, Romanian women have consistently entered their first partnership much more often as a direct marriage and much less often as a cohabitational union (Hoem et al., 2009ab, 2010). More than in other post-Communist countries, Romanians tend to convert consensual unions into marriages relatively quickly.
Share of exposure time spent in consensual union, in direct marriage and in marriage after cohabitation, Romanian women, 1965-2005
Share of exposure time spent in consensual union, in direct marriage and in marriage after cohabitation, Romanian women, 1965-2005
3Births in consensual unions have increased throughout Europe in recent decades and their contribution to non-marital births is much greater than that of single women without a co-residing partner (Kiernan, 2004 ; Thomson, 2005 ; Sobotka and Toulemon, 2008). In other words, we have witnessed a decoupling of marriage and parenthood in many European countries, but not a separation of partnership and parenthood (Kiernan, 2004). [1] In older Romanian cohorts, births to non-partnered women dominated what little recorded non-marital fertility there was, but the pattern has changed in more recent cohorts. The importance of consensual unions for non-marital childbearing has risen and childbearing in consensual unions has become increasingly prevalent (Mure?an, 2008 ; Rotariu, 2009, 2010, 2011 ; Hoem et al., 2009a, 2009b ; H?r?gu?, 2010, 2011 ; Perelli-Harris et al., 2010).
4Cohabitation has remained quite strongly linked to marriage in Eastern Europe, and the pattern of conception in cohabitation followed by (shot-gun) marriage has been slower to change than elsewhere (Perelli-Harris et al., 2012 ; Potârc? et al., 2013). Non-marital cohabitation has been a prelude to marriage (with relatively short duration and low frequency of childbearing) or has appeared as a stage in the marriage process (usually leading to marriage, often after childbirth) (Heuveline and Timberlake, 2004). Even in countries with low shares of births in cohabitation (such as Romania), once women give birth in this union type, they do not later go on to marry : three-quarters still cohabit one year after birth, and a similar proportion still cohabits three years after birth (Perelli-Harris et al., 2012). In all countries, the main factor triggering a marriage for cohabiting couples has always been pregnancy (Berrington, 2001 ; Blossfeld and Mills, 2001 ; Hoem et al., 2009b).
5It is particularly attractive to study group-specific trends in fertility in Romania because its population was subject to unusually dramatic changes in family policies during the period of state socialism. The main purpose of the present article is to pay more attention than before to childbearing trends in consensual unions and marriages in Romania, and to study the effects of educational attainment and urban/rural origin. The rarity of cohabitational childbearing in the early years represents a particular challenge, however. Because of such difficulties we concentrate on the period 1985-2005 in what follows.
I – Background
Why choose cohabitation instead of direct marriage ?
6There has been much social-status heterogeneity in the trajectories of parenthood in Eastern Europe (Sobotka, 2008 ; Kiernan, 2004 ; Kiernan and Mensah, 2010). While the prevalence of cohabitation and non-marital childbearing increased between 1985 and 2005, there were pronounced differences across post-Communist countries (Billingsley, 2010) and across social groups (Sobotka and Toulemon, 2008). In investigations of such features, education has most often been used as a proxy for social and economic resources and opportunities, in a reflection of the social stratification that defines basic groups (Perelli-Harris et al., 2010).
7Good resources (stable employment, housing of a certain quality) are important for a marriage, as the institution of marriage includes expectations about economic roles. When people are confronted with instability and uncertainty in their lives, as has often been the case in Eastern Europe, they may choose the adaptive strategy of starting a union by cohabitation rather than marriage (Oppenheimer, 2003 ; Perelli-Harris et al., 2010). In such cases, non-marital cohabitation is seen as a temporary and reversible living arrangement that is less binding and more compatible with the uncertainties of life (Mills et al., 2005 ; Kalmijn, 2011 ; Perelli-Harris et al., 2010). Likewise, the severity and duration of uncertainty and economic instability have been found to influence the extent of fertility decline in Eastern Europe and the manner in which it occurs (Billingsley, 2010).
8In Western Europe, uncertainty in family life trajectories is connected mainly to the globalization process and to the changing nature of youth employment (Mills et al, 2005). In Eastern Europe, people have additionally been confronted with the crisis of transition from a centrally planned to a market economy and its social consequences, including falling standards of living and widespread impoverishment (Dorbritz, 2003). The change of the political regime and the economic transition that followed in Romania, as in other Eastern European countries, brought rising unemployment and the loss of many welfare provisions connected to the socialist regime, making life even more difficult for young people. Under socialism, young persons in Romania were allocated into jobs directly or soon after graduating from secondary or tertiary education, and almost all welfare benefits were linked to employment status (Popescu, 2004). After the change of the political regime, the state retrenched its support (Mure?an et al., 2008) and young adults’ lives became marked by economic insecurity. The least educated were the most vulnerable in this respect. Under these conditions, investments in human capital (education) became the primary means of protecting against the risks associated with unemployment (Kohler et al., 2002 ; McDonald, 2006).
9With fewer resources and fewer economic opportunities, low-educated women have continued to live in non-marital unions while higher-educated women, who have more resources and economic opportunities, have been able to choose the more stable lifestyle associated with marriage (Perelli-Harris et al., 2010).
The negative educational gradient of first entry into cohabitation
10The growing preference for cohabitation as the initial format of a first union in Europe in recent decades is most often interpreted in the framework of the second demographic transition (SDT) which predicts that the forerunners of new behaviour are the socioeconomically better off. High or prolonged education is a key element in the SDT account : Lesthaeghe and Surkyn (1988) found for Western Europe that educational attainment was positively associated with central notions of the SDT such as non-conformism, permissiveness in personal matters, post-materialism, protest-proness and so on, as well as with non-traditional family forms. A high level of education, and especially prolonged education, should act as a mechanism for value change and for choosing cohabitation as a framework for childbearing and childrearing. For former socialist countries, researchers like Perelli-Harris and Gerber (2011) have found that new family behaviours such as the postponement of marriage and childbearing (or even voluntary childlessness) were initiated by higher-educated women, just as in other European countries, while other aspects of behaviour (cohabitation and non-marital childbearing) more often originated among lower-educated individuals (Sobotka, 2008 ; Mure?an, 2008 ; Potârc? et al., 2013).
11Studies of partnership formation in Eastern Europe have shown a negative educational gradient for the choice of cohabitation as a first union, with a higher propensity towards direct marriage among highly educated women and a higher intensity of entering cohabitation as a first union among the low-educated. These patterns have been documented for Romania (Hoem et al., 2009b), Bulgaria (Kostova, 2007 ; Koytcheva and Philipov, 2008), and Hungary (Br?d??an and Kulcsar, 2008). A substantially different effect was found for Russia, in that Russian women with a tertiary education have higher rates of entry into cohabitation than women with a secondary or lower education (Kostova, 2007).
12Higher education is also associated with a greater inclination to transform a consensual union into marriage, while lower-educated women more often tend to stay in cohabitation without progressing to marriage (Hoem et al., 2009b, for Romania ; Br?d??an and Kulcsar, 2008, for Hungary).
The negative educational gradient of childbearing in cohabitation
13Several authors have found evidence of a negative educational gradient of childbearing in cohabitation for a number of European countries (Konietzka and Kreyenfeld, 2002, for Germany ; Kiernan, 2004, Kiernan and Mensah, 2010, for the UK ; Spéder, 2005, for Hungary ; Perelli-Harris and Gerber, 2011, for Russia ; Rotariu, 2009, 2010, 2011, H?r?gu?, 2010, 2011, and Hoem and Mure?an, 2011a, for Romania ; Perelli-Harris et al., 2010 for eight European countries). The negative association holds not only for post-socialist countries but also for Western European countries with a long tradition of cohabitation as a type of partnership. Perelli-Harris et al. (2010) found the strongest evidence for this pattern in Norway, Russia, the UK and the Netherlands, while for France, Austria, and West Germany the educational gradient, although negative, was not significant, which suggests a higher similarity between cohabitation and marriage in these countries. They also found that in countries with a longer experience of alternative forms of partnership, such as France, a positive educational gradient was observed in the early days of cohabitational childbearing, but as the phenomenon spread, a negative educational gradient became the predominant pattern.
14Other studies, like those reported by Toulemon and Testa (2005), have shown that after controlling for age, number of children, and fertility intentions, the probability of having a child was very similar for cohabiting or married people in France over the period 1998-2003. The negative educational gradient does not characterize births in cohabitation only, but is also well documented for births to non-partnered mothers (especially for the US, but even for Europe).
15Regarding fertility in general, building on Becker’s ideas about a quality-quantity trade-off and women’s opportunity costs of having children, many authors have argued that due to the accumulation of human capital, women with higher education are more prone to concentrate on their careers and on their earning power, which results in lower economic incentives to marry and have children (Balbo et al., 2013). Moreover, the opportunity costs of childbearing and childrearing become higher as human capital increases, so highly educated women should be more likely to postpone marriage and births. In many European countries, these women have the highest levels of childlessness and the lowest levels of fertility (Sobotka, 2008). Perelli-Harris et al. (2010) have also found a negative educational gradient for marital births, but their main result is that the gradient for births in cohabitation is much steeper than for marital births.
16At different levels of educational attainment, the reasons for adopting unconventional family behaviours may differ, and this may also be the case in comparisons between urban and rural environments. Such reasons reflect attitudes, opportunities, or constraints (Perelli-Harris et al., 2010). In Eastern Europe, younger, urban, and higher-educated women show the most positive attitudes towards new forms of partnership and to childbearing in different contexts, but they are the last to translate these attitudes into actual behaviour (Rotariu, 2006, 2009, H?r?gu?, 2008 for Romania ; Sobotka, 2008, for other countries in the region). Using vital statistics, Rotariu (2009, 2011) has shown that in Romania, non-marital births are more frequent among women who have a low level of education, who live in poverty, and who are from rural areas, and that their births occur at much lower ages than marital births. Higher-educated, professional women are less prone to give birth outside marriage : among children whose mothers have a tertiary education and are employed, and whose fathers work outside agriculture, only 4.7% are born outside marriage (Rotariu, 2011, based on data for, 2006-2009). H?r?gu?, (2011), using data from the Romanian Generations and Gender Survey (GGS), has also found a strongly negative association between educational attainment and first birth in cohabitation, and this association is much more visible than for first marital births.
II – Data, method and first results
Data
17As is evident from the literature quoted above, there is already much information about family developments in Romania. Our own contribution to this picture is based on data from the Romanian Generations and Gender Survey (GGS) of 2005 whose total sample consists of 11,986 respondents (5,977 men and 6,009 women) aged 18-79 at the time of interview. For our study, we focus on the 5,847 women who belong to the Romanian and Hungarian ethnic groups, and within these groups, on women at ages 15-40, since childbearing is very rare after age 40 in Romania.
18Our sub-sample excludes 162 women who were of other ethnicities (including Roma) because of their very special union formation and childbearing behaviour ; they are too few for a reliable separate analysis. We also eliminated 22 records because they had incomplete marital or educational histories, and also the record of the sole woman who was born and grew up outside Romania. Our final sample thus includes 5,824 women, which seems to be a habitual size order for a GGS survey. [2] In our experience, this sample size, while relatively large as sample surveys go in general, is still rather small for the analysis of cohabitational fertility, which is therefore a challenge for standard GGS data. Fortunately, a simple smoothing technique allowed us to reveal interesting patterns in the data nevertheless.
19The size and structure of the raw data available to us are shown in Appendix Table A.1, which gives occurrences and exposures recorded by calendar year for each union status, and Appendix Table A.2 which lists the number of consensual unions recorded at the beginning of each duration year for groups of calendar years. In Appendix Table A.2, “1st year” means the number of consensual unions initiated in the period, i.e., the number of such unions recorded at the beginning of the first union year. Similarly “2nd year” means the number of such unions remaining at the beginning of the second year of the union, “3rd year” means the number remaining at the beginning of the third year, and so on, up to the 15th duration year. In 1995-1999 no unions were recorded beyond the 11th duration year and in 2000-2005 no unions where recorded beyond the sixth year, in both cases because the data were collected in 2005.
Fertility measures and fertility trends in consensual unions
20To analyse fertility for a selected population subgroup such as women in a consensual union, we first computed partially-cumulated occurrence/exposure rates 
of fertility for women of all parities in the group, separately for each of a number of union duration intervals indexed by d. We did this for each single calendar year t in our period of observation (1985-2005), and also for each of the years 1980-1984. [3] We worked with combined duration intervals consisting of the first twelve months of the union (d = 1), the second and third year of the union (d = 2), the fourth through sixth year (d = 3), and the seventh through fifteenth year (d = 4). More precisely, we initially computed conventional single-year-duration occurrence/exposure rates then summed them to get interval-specific rates. Even with this broad grouping, the raw values initially computed for the years t turned out to be too strongly influenced by random variation. We therefore replaced the raw values by smoothed values 
, except in the right-hand tail of the curve, where we used fewer terms (down to five) in the unweighted centred moving average that this procedure represents. Such special treatment was not needed in the left tail because we have data going back to well before 1985. The resulting values of the smoothed rates 
for women in consensual unions are displayed in Figure 2, which essentially contains a decomposition (by intervals of union duration) of the aggregate fertility that we will introduce in the following Section.
21Moving average graduation is a classical technique used for smoothing curves to reveal their underlying structure. [4] We would expect smoothed curves to display general levels and trends in fertility but to give only a pale reflection of the sudden perturbations that followed the abrupt changes in family policies in Romania. Figure 2 gives a more precise picture. It shows a strong drop in fertility at long cohabitational durations from a level just below 1 to a level around one-half in the late 1980s ; fertility at those long durations remains quite stable after the fall of Communism. We also see a gradual fertility decline at shorter durations.
Cumulated-duration-specific rates of fertility in consensual unions, Romania, 1985-2005
Cumulated-duration-specific rates of fertility in consensual unions, Romania, 1985-2005
22Note that women in consensual unions live in a strongly dynamic situation : first of all the group of cohabiting women is eroded steadily both by union dissolution and by union conversion, i.e. in any month, cohabiting women may stop living in a consensual union or alternatively may convert their consensual union into a marriage. These decrements are reflected visibly in Appendix Table A.2. Second, at duration 0 the group of cohabiting women continuously receives an inflow of entrants who begin a (new) consensual union. Whenever a woman leaves a cohabitational or marital union and starts a new union (perhaps later), she is counted among the corresponding entrants at the start of the new union. One great advantage of our approach is that it allows such dynamics to take place smoothly without producing the complex structure of details that normally result from an event-history analysis accounting for both parity and partner dynamics.
Summation over durations
23To compare population subgroups and to study trends in a more compact format, it is convenient to also introduce fertility rates aggregated by summation over all durations, in the nature of 
. In our case, rates are aggregated up to the 15th duration year, except that aggregation stops at the duration reached when the data were collected, in 2005. Such aggregation will facilitate our later study of fertility differentials by educational attainment and by place of residence at age 15 (i.e. urban or rural). The aggregation is inspired by the practice of computing the usual total fertility rates (TFRs) by summing age-but-not-parity-specific fertility rates across all ages ; in fact Hoem and Mure?an (2011a) would call a quantity like Fg (t) a “duration-based” TFR to distinguish it from the usual “age-based” TFR. Hoem, Jalovaara, and Mure?an (2013) have adopted this terminology. However, using a name that contains a TFR designation for the present quantities may be confusing for some readers, so we have decided to avoid the TFR terminology in this paper.
24An aggregate like Fg (t) can also be seen as a development of the notions of parity-progression ratios. The basic idea behind a parity-progression ratio is, of course, to follow a group of individuals who have reached a given childbearing parity and to record the fraction of the group who ever attains the next higher parity. The parity-progression fraction can be computed from rates of parity progression added up over durations since the index parity was attained. In its simplest version, this is applied (say) to a real childbearing cohort of women, but more often the rates are aggregated over durations in a synthetic cohort. In our application, we avoid the concentration on a selected parity and count progressions across all parities in synthetic cohorts. [5] The whole computation can be restricted to married or to cohabiting women, in which case the first duration is counted from (type-specific) union formation ; otherwise the first duration is counted from age 15. We have used the latter practice for never-partnered women.
25Parity-progression ideas have a strong pedigree with roots going back to Ryder (1951), Henry (1953), and Brass (1974). Duration-specific rates and their sums were already pervasive in the work around the World Fertility Survey (Rodriguez and Hobcraft,1980 ; Hobcraft et al., 1982 ; Hobcraft and Casterline, 1983). The same notions have been used subsequently by a number of authors, such as Blayo (1986), Rallu (1986), Ní Bhrolcháin (1987), Feeney and Yu (1987), and Murphy and Berrington (1993), and again recently by Breton and Prioux (2005), Barkalov (1999, 2005), and Hosseini-Chavoshi et al. (2006), and surely many others. In most applications, the parity progression is computed for single-year intervals (say) in discrete time, but for us it was more natural to follow Hoem and Mure?an (2011a, 2011b) and use rates for intervals in continuous time. This variant has allowed us to take transitions between the various types of unions smoothly into account without getting involved in describing the complexities that characterize transitions between union forms (Hobcraft et al., 1982).
Trends in aggregate marital fertility
26Applying the above notions, we calculated fertility for directly married women and aggregated the rates over all marital durations. We did the same for women who married their cohabitant after some time in a consensual union, grouped separately by duration of pre-marital cohabitation. The outcome is shown in Figure 3. Curve B is for women who start with a consensual union which they convert to a marriage in their first year of cohabitation ; curve C is for women who married during their second or third year of cohabitation, and so on. These curves were smoothed (again by an 11-term centred unweighted moving average as explained above), but there seem to be just too few married women with most lengths of pre-marital cohabitation in our data to produce useful results. The aggregate-fertility values shown on Figure 3 are robust and can be interpreted for the directly married (which anyway is by far the largest group of those who ever marry) and for married women with up to one year of cohabitation before marriage, but this is hardly the case for women with longer durations of pre-marital cohabitation. [6] The curve for the group with seven or more years of pre-marital cohabitation was of no interest.
Aggregate marital fertility by duration of pre-marital cohabitation, married women, Romania, 1985-2005
Aggregate marital fertility by duration of pre-marital cohabitation, married women, Romania, 1985-2005
27For the rest of our discussion we concentrate on the directly married and ignore all other groups. [7] Our analysis also ignores the aggregate-fertility contributions of women outside of any (consensual or marital) union because they are very small and contribute little to women’s lifetime fertility computed across all union types. Figure 3 demonstrates that marital fertility has declined gradually over our period of observation.
Control variables
28In previous work about Romanian fertility (Mure?an and Hoem, 2010 ; Hoem and Mure?an, 2011a) the analysis contained control variables like starting age, starting parity, and union order, whose effects were given as relative risks. Our use of group-specific rates as the basis for aggregate fertility computations allows us to consider additional variables. For example, it is possible to compute indices similar to those denoted PDTFR (parity- and duration-specific fertility), PATFR (parity- and age-specific fertility), and PADTFR (parity-, age- and duration-specific fertility) by Rallu and Toulemon (1994) in their systematic overview of the methodology of period-specific fertility measures. The procedure goes as follows :
29Instead of letting the 
be partially-cumulated occurrence/exposure rates, one can first fit some piecewise-constant model (say) with covariates to the occurrences and exposures in the original data, using union duration as process time, and then add the fertility rates predicted by the model over the values of d to get initial aggregate fertility rates for each group g defined by the model regressors. The model-defined raw aggregates can then be smoothed as needed.
30We use this approach in our study of the effects of educational attainment and urban/rural residence of origin. We experimented with additional covariates and (not surprisingly) found that group-specific aggregate fertility declines systematically with an increasing age at union formation. Otherwise, findings reported in earlier work (including the paper by Hoem and Mure?an, 2011a) turned out to be quite robust against model or data re-specification. In the present report we pay little attention, therefore, to additional covariates.
III – Effects of education and origin
Effect of educational attainment
31Following up earlier work mentioned above, we now turn to one of the most interesting covariate effects that can be illuminated by our data, namely that of educational attainment. For the present purpose we distinguish between the three educational levels “low” (which means no formal qualifications beyond compulsory school), “middle” (which covers those who have completed high school or have taken a vocational certificate), and “high” (tertiary degree or higher). [8] We would also have liked to compute aggregate fertility for women who are still in education, but this status is too poorly recorded in GGS data for the most part, and we avoided including it in our analyses. [9]
32The patterns of aggregate fertility for married women appear in Figure 4, and for cohabiting women in Figure 5. There are too few cohabitants among the higher-educated women in our data for us to include a curve for them in Figure 5. These diagrams display the now well-known negative gradient by educational attainment : for both partnership types, the curve for each level of attainment lies below the corresponding curve for lower attainments. [10] Remarkably, fertility developments in the low-level educational group are much the same in consensual unions as in direct marriages (Figure 6), except that low-educated women in consensual unions must have been visibly more sensitive to political developments around the time of the fall of state socialism (this feature is visible in Figure 5 already). In fact these women were more strongly coerced by authoritarian pro-natalist policies in earlier years, since they had fewer resources to escape them. In addition, childcare allowances often were, and still are, the main source of income for this poor population segment, which has also been less successful on the marriage market. We observe a strong fertility decrease in this group of women from a level of around 2.3 before the fall of Communism to a new level of around 1.7 thereafter (Figures 5 and 6).
Aggregate fertility of directly married women by level of education, Romania, 1985-2005
Aggregate fertility of directly married women by level of education, Romania, 1985-2005
Aggregate fertility of women in consensual unions by level of education, Romania, 1985-2005
Aggregate fertility of women in consensual unions by level of education, Romania, 1985-2005
Aggregate fertility of women with low educational attainment by type of union, Romania, 1985-2005
Aggregate fertility of women with low educational attainment by type of union, Romania, 1985-2005
33In wealthy cities like Bucharest, Romanian society has recently started to accept modern behaviour like consensual unions and non-marital births (Dohotariu, 2010). Otherwise, consensual unions in Romania more frequently take the forms of “concubinage”, which is specific to couples with a low socioeconomic status and a low level of education, living most often in rural areas. [11] Among cohabiting women, those with the highest education, with the highest socioeconomic status, and with residence in urban areas tend rather to be childless (H?r?gu?, 2008).
34The share of women with a low level of education is (and always has been) relatively high in the Romanian population, even if this share is on the decrease. For example, in the female sample of the GGS, 31% of women have no qualifications, 58% have high school or a vocational certificate, and only 11% have a tertiary degree. The share of women of rural origin has changed in a similar manner. In the GGS, 71% of the women were born and grew up in rural areas.
Women of rural versus urban origin
35We have already indicated the importance of a rural origin for fertility behaviour in Romania. The large socioeconomic differences between urban and rural areas in the country motivate the use of the type of original residence as another criterion for social stratification. Urban-rural differences are partly due to compositional factors (Kulu, 2013) : fewer people with higher education live in rural areas, where economic opportunities are more scarce than in towns and cities. In line with our previous arguments about the negative educational gradient of births in cohabitation, where education was a proxy for resources and opportunities, we argue that the fertility of cohabiting women with similar levels of education should be higher in rural than in urban areas. This is true in our data, but aggregate fertility levels for women of rural and urban backgrounds have approached each other over time both for directly married women (Figure 7) and for cohabitants (Figure 8). Historical differences are gradually disappearing. Meanwhile, some urban/rural differentials seem to have survived, in that the sudden fertility drop among cohabitants after the collapse of state socialism was concentrated among those with a rural background (Figures 8 and 9). Figure 9 further shows that both cohabiting and directly married rural women have roughly the same fertility levels, a similarity which parallels that of women with a low level of education (Figure 6).
Aggregate fertility of directly married women of rural and urban origin, Romania, 1985-2005
Aggregate fertility of directly married women of rural and urban origin, Romania, 1985-2005
Aggregate fertility of cohabiting women of rural and urban origin, Romania, 1985-2005
Aggregate fertility of cohabiting women of rural and urban origin, Romania, 1985-2005
Aggregate fertility of women of rural origin by union status, Romania, 1985-2005
Aggregate fertility of women of rural origin by union status, Romania, 1985-2005
IV – Discussion
36Our current investigation has revealed or confirmed several aspects of fertility patterns in Romania.
37• In our data, duration-based aggregate fertility decreases monotonically as educational attainment increases, both for cohabiting and for directly married women, as noted before by Mure?an and Hoem (2010) for all women taken together. This covers births of all orders. Similar results for first births to cohabiting women have been documented by Perelli-Harris et al. (2010).
38• Women with a rural background have a consistently higher aggregate fertility than women with an urban background, both in cohabitational unions and in direct marriages, net of educational attainment. Urban women have been forerunners in the new fertility trends.
39• More remarkably, we have found that among women with a low educational attainment, as well as among women of a rural origin, aggregate fertility in marital and in cohabitational unions is largely of the same size order. For such women, marital status is less important for their fertility than the fact of having a partner.
40Our findings neatly complement those of official Romanian statistics : the 2002 census showed that couples in consensual unions have 1.97 children on average and married couples just 1.72.
41There remains the question of why Romanian women with a low socioeconomic status and low educational attainment more often choose alternative living arrangements and bear children outside marriage. Rotariu (2011) has emphasized that most of the growth in non-marital childbearing after the fall of the state-socialist regime is not due to the spread of post-modern values and attitudes, as posited by second-demographic-transition arguments. He maintains, instead, that they result from the revival of some behaviours that were also manifest during past centuries in Romania. He argues that in this cultural area, the modernization process was incomplete and that births outside marriage have always existed as a socially tolerated behaviour, at least in some communities. [12] The socioeconomic crisis that followed the recent change of political regime favoured the spread of this behaviour mainly in the cultural areas where it existed already. [13]
Acknowledgements
We are grateful to Kim Lindoff Jansson for early research assistance. Economic support for Cornelia Mure?an from the Max-Planck-Gesellschaft is gratefully acknowledged. The work of Mihaela H?r?gu? and in part that of Cornelia Mure?an was supported by a grant of the Romanian National Authority for Scientific Research, CNCS – UEFISCDI, project number PN-II-ID-PCE-2011-3-0145. We have been influenced by comments from four referees working for Population and are particularly grateful for the mapping of the parity progression literature that a referee made for us.Exposure time and number of births for women by calendar year and union status, Romania, 1985-2005*, **
Exposure time and number of births for women by calendar year and union status, Romania, 1985-2005
Number of consensual unions still intact at the start of the nth year since union formation, by 5-year periods of union formation
Notes
-
[*]
Stockholm University, Demography Unit, Sweden.
-
[**]
Babe?-Bolyai University, Faculty of Sociology and Social Work, Romania.
-
[***]
Babe?-Bolyai University, Centre for Population Studies, Romania.
-
[1]
For the most recent period (1995-2004), 12% of all births were to cohabiting women in Romania and 4% were to non-partnered mothers ; in Bulgaria the proportions were 22% and 8%, respectively ; in Hungary 18% and 4% (for the period 1995-2001) ; and in Russia 17% and 15% (Perelli-Harris et al., 2012).
-
[2]
In terms of size, the GGS compares unfavourably with the 12-14,000 used by Breton and Prioux (2005), the 40,000 of Rallu (1986), the 91,000 of Hosseini et al. (2006), and the, 200,000 used by Murphy and Berrington (1993). The latter study is only outdone by Scandinavian-type register data.
-
[3]
We also have data for 1965-1979 but have omitted all results for those early years because there may be problems with the reliability of partnership histories in such old survey data due to recall errors and to the massive scale of selective out-migration (but only after 1989) for the Romanian population. Note that the rates for 1980-1984 are used only in the smoothing computations for years 1985-1989 and are not otherwise reported here.
-
[4]
This quite standard general smoothing method has also been used before for parity progression by Breton and Prioux (2005) for a data set where the sample of women was about twice as big as ours, and by Hoem and Mure?an (2011a) for the data set used in this article.
-
[5]
Most previous investigations of Romanian fertility have only studied first births, as is common.
-
[6]
Over the period 1985-2005, we counted a total of 668 marriages with pre-marital cohabitation. We found union conversions as follows : 384 consensual unions were converted into marriages during the first year of cohabitation, 180 during the second or third year of cohabitation, 67 during the fourth to sixth year of cohabitation, and 37 during the seventh or later years of cohabitation. For the same period we recorded 2,795 direct marriages and 660 consensual unions not converted to marriage.
-
[7]
Figure 3 pertains to married women. The curves represent their aggregate fertility as produced in marriage only. Any childbearing before marriage is excluded. To determine a woman’s complete aggregate fertility her contribution from any premarital cohabitation must be added to the contribution derived for Figure 3. The two contributions are easily spliced together, as explained by Hoem et al. (2013), but we ignore the fertility in post-cohabitational marriages anyway. For the directly married there is, of course, no period of pre-marital cohabitation and no corresponding contribution to fertility.
-
[8]
For more exact definitions, see Mure?an and Hoem, 2010.
-
[9]
Since complete educational histories are not available, to include this status we would need to record a woman as in education even in periods where she has interrupted her studies, and this would be particularly problematic for later study periods among the highly educated.
-
[10]
Negative educational gradients are also reported by the 2002 census of Romania. At ages 50-54 it found an average number of children by the mother’s educational level as follows : no education 3.2 children ; primary school 3.1 ; lower secondary 2.6 ; high school 1.7 ; and tertiary education 1.4.
-
[11]
For this interpretation, see Rotariu (2009, 2010, 2011), who studied vital statistics, and H?r?gu? (2010), who worked with the same data set as us.
-
[12]
Kok (2009) also mentions the high level of illegitimacy at the turn of the twentieth century in provinces that form the present Romanian territory. He connects it with the family system in the region, arguing that the cohabitation and non-marital childbearing seen nowadays seem to be a return to old traditions.
-
[13]
Rotariu (2009, 2011) also points out that the incidence of non-marital births is high in the Roma population, whose living standards are very low, and whose social integration is deeply deficient. Due to its small size, the Roma population was excluded from our analysis.
