Contemporary union formation in Bulgaria. The emergence
Elena Koytcheva∗ ∗ ∗ ∗
Paper to be presented at the XXV International Population Conference of the
International Union for the Scientific Study of the Population,
Tours, France, July 18-23, 2005
Do not cite or quote
The demographic changes that were observed in all the Eastern European countries in
the last 15 years have not surpassed Bulgaria. Only for 15 years (from 1986 until
2001) the population has decreased with one million, which is a result from the
negative natural increase and the high emigration. The drop of birth rates and the
decline of marriage rates have started already in the 1980s, but these changes
increased in speed after 1990 and reached values never observed before in the history
of Bulgaria. After 1997 a slight stabilization appears as the fertility recovers and the
negative values of the natural increase get lower. However, the values of these
coefficients are far away from the ones observed before the start of the transition of
the country towards a market economy.
∗ Max Planck Institute for Demographic Research. Contact: Koytcheva@demogr.mpg.de
Amongst all of these changes, the recent developments regarding the interrelated
processes of union formation and childbearing in Bulgaria constitute a particular
interest for demographers. Some scholars have argued that what we observe today is a
consequence of economic crisis, and the postponement of life course events will
decrease as the economic situation of the country improves (economic interpretation).
Others claim that also cultural aspects such as the adoption of liberal western values
by the young generations will lead to a profound change in the Bulgarian family and
fertility demography (cultural interpretation). However, due to a prior lack of data and
analytical methods, the empirical basis that is required to support or refuse one or the
other interpretation has remained weak. This paper draws on results of a recent
dissertation project (Koytcheva, forthcoming) and examines the recent changes in
marriage and cohabitation and their relation to childbearing in detail. In particular, we
study women’s behavior regarding entering a direct marriage, forming a cohabitation
and later transferring it into a marriage, and becoming a mother. We pay particular
attention to women’s ethnic group affiliation and education level as cultural indicators
of the individuals.
The Bulgarian family and fertility demography has been massively affected by
postponement processes. Bulgaria has traditionally been a country in which the birth
of first child occurred at very young ages compared to other European countries. For
instance, for a very long period (at least from the 1950s), the average age of first birth
for women stayed around 22 years. The tendency in the last 10 years shows a gradual
increase and in 2001 it reaches the level of 23.8 years. In addition, the difference
between this measure and the mean age at birth becomes smaller during the last years.
This can be explained with a rising share of women who stay with one child, that is,
for many women the average age at birth and first birth is identical. Since the 1980s,
also the number of marriages has decreased substantially. The delay of marriages has
been confirmed by the literature (Sougareva, 1995). More and more people enter
marriage at later ages, compared to the years before. Before 1990, the mean ages at
first birth and first marriage for women have been quite stable. The mean age at
marriage was about 21.4 years and the mean age at birth at about 22.0 years. The
difference between them was always around 0.6 to 0.8 years, and until 1994 the
average age at first marriage has always been lower than the one at first birth. After
1994, this tendency has reversed and a new phenomenon appeared: the mean age at
first birth has fallen below the mean age at first marriage, and the share of out-of-
wedlock births has increased. While in the 1980s approximately every tenth child was
born out of marriage, in 2001 it is almost every second. A similar increase is observed
in all the Eastern European countries but Bulgaria is one of the leaders in this respect.
This process of births outside marriage is closely connected with another new
phenomenon in the models of family formation – the so-called cohabitation. While in
Western Europe this model of family is very popular for decades (van de Kaa, 1987),
it is still very new for the countries of the Eastern block. In the Bulgarian scientific
literature the new terminology of cohabitation is introduced by Kostova (2000) and
defined as “the living together of two people from different gender, in which they live
like married people without having an official marriage for different reasons such as
impossibility or unwillingness, temporarily or in principle”. The latest data show
(Belcheva, 2003) that 13.1 of the population in ages 15-59 live together without being
married. According to the data of the last census, the share of cohabiting people at
ages 15-29 is 17.6%, at ages 30-44 it is 12.1 and at the older ages (45-59) it is 10.4%.
This is a new tendency of family formation in Bulgaria and this process has not been
deeply investigated up to now. What we know, however is that the high percentage of
out-of-wedlock births “is not necessarily synonymous with children being born
outside a family union of some type” (Council of Europe, 2001).
The most common explanation for the observed changes in the countries from the ex-
socialist block is the impact of the economic crisis and the process of ideational
changes. We want to mainly investigate the possible impact of the value changes in
Bulgaria on the union formation patterns. Of course, we do not want to neglect or
minimize the impact of the economic changes. We fully agree that the transition to
market economy has an enormous impact on the demographic behavior of the
population. However, we want to put more stress on the ideational changes as
according to our opinion, it is not sufficiently investigated in the existing scientific
The basis of the theories of ideational change is the proposition that values and
traditions of people change with time. These theories search for an answer to the
question how values change and shape fertility by affecting the choice of the people
when to have a child or if to have a child at all, when to marry or to marry at all. The
idea of changes in the value system of the societies and their impact on the fertility
behavior of the people is most often associated with the notion of a Second
Demographic Transition which was introduced for the first time by Lesthaeghe and
van de Kaa (1986). Their theory is based on the observed tendencies in the western
countries. Lesthaeghe and Surkyn (1988) outline two main mechanisms, which lead to
changes in the values and aims and the resulting of it preferences of the people. The
first mechanism refers to the economic growth and its influence on the value changes
and needs from “irreducible needs” to “higher order needs”. The second mechanism
deals with the role of the social stratification and education in the process of the
cultural transmission (Lesthaeghe and Surkyn 1988).
One of the main ideas in the classical theories of Tard (1890) and Sorokin (1947) is
that the cultural changes start from the higher strata in the society as a result of the
privileges, education and concentration of means and opportunities; the lower social
strata perceive the new preferences through imitation. In this connection diffusion
theory is a potential explanatory model for demographic changes. The diffusion of
ideas, behavior and techniques is often considered to follow the routine established
from the social-cultural forces like language, ethnicity, living quarter, working place
or canals of communication and exchange (De Bruijn 1999, Bernardi, 2003, Kohler et
al., 2001). According to Kirk (1996) diffusion is not only a residual effect, it is an
active factor in the increase or slowing down of the birth control.
This paper argues that these theoretical frameworks may be a crucial tool for the
understanding of Bulgaria’s changing family and fertility demography. The analyses
we present in this paper aim at contributing to the empirical evidence on the ongoing
demographic change in Bulgaria and put these theories to an empirical test with recent
individual-level data and hazard regression models. As a proximity for the value
difference in the different segments of the population we use ethnicity and education
level. The main ethnic groups in Bulgaria are: Bulgarians (about 80 % of the
population), Turks (about 10 %) and Roma (about 6-8 %) and they differ largely in
their traditions, religion and values. That is why we suppose that each of the ethnic
groups in Bulgaria has a different reaction to the political and societal change in the
country when considering the fertility and family formation patterns.
We also consider education level as an indicator for different social status in the
society and having impact on the fertility and family decisions of the individuals.
Women with higher education may have more liberal views regarding family values
and thus be more inclined to live in consensual unions and bear children in them. On
the other hand, we consider the lowly educated women as more traditional and
tending to legalize a relationship as soon as pregnancy is recognized. We divide the
educational levels in three groups: primary, secondary or higher.
Data and methods
For the complimentary analyses of family formation patterns, we use a new data set
coming from the Survey “The Young People – Partnership, Marriage and Children”1
in Bulgaria. The survey took place from June to September 2002. The sample of the
survey was based on the data from Census 2001 and includes 10,009 participants aged
18 to 34 at the time of the interview. We restrict our study only to the female
participants in the survey. This data set is a first of its kind in Bulgaria that includes
union histories of the respondents. As little is known about the changes in the family
formation patterns in Bulgaria, analyzing the transition to first union formation will
substantially contribute to understanding the demographic changes in Bulgaria. We
can analyze the family formation patterns only for the period after 1985 as the
respondents in the study are from very young cohorts. The oldest cohort was aged 22
in year 1990. This period for the analyses is well suited because we know that most of
the changes that appear in the transition to first union formation took place since the
end of the 1980s. This data allows us to follow the changes throughout the 1990s and
to account for the effect of cohabitation on fertility.
1 This survey is better known as “Social Capital Survey”. We also will adopt this
name for convenience and for the sake of brevity.
In the present paper we study the transitions to direct marriage, to cohabitation, to
marriage after cohabitation and to first birth. On the analyses on the transition to
union formation we take into account the motherhood status of the women (if they are
pregnant or have children) and on the analyses on the transition to first birth we take
into account the union status of the women. We use event history analyses as in our
view it is the most suitable method of studying the individuals’ life-course transitions.
For the analyses of each of the processes we use separate model. The mathematical
expression of our models read:
ln h(t) = γ T(t) +
where ln h(t) is the logarithm of the risk of occurrence of the event in moment t,
γ T(t) covers the risk duration of the event and
β X(t) represents the (time-varying)
covariates, which affect the risk of occurrence of the event. The baseline hazard
duration dependence,γ T(t), in our models is always a piecewise-linear spline. Each of
the covariates included in the models contributes to the shift of the baseline.
For the transitions to direct marriage, cohabitation and first birth we start to observe
the women as soon as they become 13 years old and for the transition to marriage
after cohabitation the starting point is the time at forming a cohabitational union. The
time-varying covariates are calendar time, age of the woman, education level,
education enrolment, motherhood status and union status. Time-constant variables are
ethnic group, number of siblings, place of residence till age 15 and level of religiosity.
For the estimation, we use the statistical software aML, version 2.0 developed by
Lillard and Panis (2003). The data preparation is done with the help of the Stata
software package, version 7.0
Results and Discussion
We are not going to discuss the impact of each included variable on family formation
and transition to motherhood. Rather, we want to concentrate on the effects of ethnic
group and education level and discuss the changes in time that have occurred. To start
with, we will have a look at the development of the intensity of union formation and
first birth through the calendar years. We plotted the results in Figure 1. Our results of
the full model for each transition under study we present at the Appendix (Tables 3 to
Figure 1: Intensities of direct marriage, cohabitation and marriage after cohabitation
by calendar year
(1) The starting point of the intensities for marriage after cohabitation and
direct marriage are changed in order to achieve a comparable scale for all
The first clearly visible trend in the union formation pattern is that marriage
intensities (both direct and after cohabitation) go down and at the same time the
cohabitation risk is strongly rises. This proves the decrease in the rate of marriage
transitions for women in Bulgaria during the 1990s. Since the start of the transition of
the country, women are less susceptible of conducting marriages. The number of
people who are prone to start their union formation as a consensual union gets higher.
The tendency is a reduction of formation of marriages (direct or after having a
cohabitation)– trend that is observed through the last twelve years and does not have a
sign of a slowing down. However, to big part this is compensated from the newly
emerged family formation pattern – the cohabitation. There is a clear rise in the
intensity of forming a cohabitation unions, especially after 1995.
There could be two major reasons for the change in the marriage behavior of the
women. The one is the economical deprivation. For instance, it has long been
recognized that marriage rates increase in times of prosperity and decrease in times if
recession (Bracher and Santow, 1998). Also, marriage is viewed as a long-term
commitment and people usually consider well their action before getting married.
They try to marry the most suitable partner according to their requirements. An
individual without job and good future prospects does not have good positions on the
marriage market2. Thus, we suppose that fewer people are ready to start a marriage
before making sure they have a prospective job and some security in life. Marriage
rates could be also affected by worsening expectations about future living standards as
a whole (Cornia and Paniccia, 1996).
The other reason could be the emergence of new family formation pattern, namely
cohabitation. The emergence of such a new pattern shows either a change in the
societal norms or a less importance of these norms or, mostly probable, both. The
societal pressure for living together only when married is loosing its strength.
Moreover, even people who tend to transform their consensual union into a marriage
are getting less. This shows that for many people it is not of great importance if their
union is a legal marriage or consensual union. This is a sign for clear changes in the
value orientations of people in general and less social pressure for marriage. Also, this
could be an indication that converting cohabitation into marriage is independent of
any direct measure of economic independence (Bracher and Santow, 1998). There
have hardly been any signs for such development during the state socialism and this
novel behavior has its strong onset only after start of the political and economical
transformations in the country.
In another estimation, we find a steady decline in the tendency of giving birth to a
child (see Figure 2). Till 1990 we observe a rise in the first birth intensity and after
that a steady decline with a trend of recovery after 1995. However, the tendency of
decrease still remains till the end of the observation period.
2 This argument, according to us, refers not only to the attractiveness of men on the marriage market,
but also to women. In the Bulgarian society, the role of women is visioned not only as a mother and
housewife, but also as an active participant in the labor market with a possible career orientation.
Figure 2 First birth intensity by calendar year.
Our results for this trend through time show that the pro-natalistic policies that were
operating in the country had influence till the end of the 1990s. After the fall of the
Berlin wall, the government could not control the fertility and give the strong support
to the mothers any more. Additionally, people faced many other changes in their lives.
New cultures and views came from the western countries, new opportunities appeared
in the life course of the individuals and contributed to changes in the life styles. Also,
the economic transition affected people’s lives as they had to cope with situation of
uncertainty which they have never experienced before. All this contributed to the
postponement of childbearing. The decision to have a child was replaced to a later
stage in life of the Bulgarian woman.
It is interesting to see if all these trends are observed for each of the main ethnic
groups in Bulgaria. In order to have a better view on our results, we summarize the
results obtained for each transition under study in Table 1 (the full results are
presented in the Appendix).
Table 1: Relative risks of transitions to direct marriage, cohabitation, marriage after
cohabitation, and first birth according to ethnic group.
to Transition to
***: p≤ 0.01 **: 0.01<p≤0.05 *: 0.05<p≤0.10.
Sig. Relative risk
Sig. Relative risk
Sig. Relative risk
Our results show that from all the ethnic groups, the Bulgarians are the least prone to
enter a direct marriage or cohabitation, that is to form a union. On the contrary, the
Roma population has the highest risk of forming a family, especially pronounced in
the case of entering a cohabitation. Once a cohabitation is formed, women coming
from the Bulgarian ethnical group tend to transform it into marriage, while the Roma
group for instance is the least susceptible of doing so. These results show than in
general, the Bulgarian women postpone the formation union at most, compared to the
other ethnic groups and they favor at most marriage as a form of union.
There could be several possible explanations for the high risk of cohabtation for the
Roma group. One of them is that the Roma usually start their sexual life in very early
ages (Yachkova, 1998) and it is not unusual for them to have a child before age 163.
In this age group it is not that easy to get married in Bulgaria as many authorities are
involved (including court). So, it can be considered that many Roma simply do not get
married or delay marriage for this very reason (Kaltchev, personal communication).
Another explanation could be that usually the Roma group is lowly educated, suffers
strongly from the unemployment and lives usually on social help. A mother gets
higher social help if she is a “lone” mother, that is, not married. However, the most
common and plausible explanation goes into a different direction. The reasoning
comes from the cultural and anthropological studies. According to some studies in
Bulgaria (Pamporov 2003), after 1990 the Roma population has returned back to their
old customs and morals and lives according to their own traditions. This suggests that
they conduct marriages according to their customs, which does not include visiting
the town hall. In other words, it is possible that the Roma population form marital
unions, but not according to the “official rules”, which leads to this “bias” in the
statistical results. Whatever the true reasons for the low intensity of entering a
3 The available data is very scarce on this issue. According to NSI, at the time of the Census, of all the
Bulgarian women aged 15 and below, about 0.01 % had a child. The equivalent percent for the Roma
women is 0.9 (NSI, 2001).
marriage is, we want to underline that, according to us, what we observe is not a
change in the values and ideas as seen in the theory of the Second demographic
transition, but rather economical motives or ethnic-cultural peculiarities (Ilieva 1995).
The results for the transition to motherhood show that the highest disposal for first
birth has the group of the Roma population. Except for the high fertility intentions, the
Roma group is characterized by early age of start of childbearing. Additionally, we do
not find very strong differences between the reproductive behavior of the Bulgarians
and the Turks. Similar results have been obtained in many other studies too (Zhekova
2001, Philipov 2000, etc.).
The large differences between the Roma group and all the other ethnic groups has
drawn the attention of many scientists. Sougareva (1995) argues that the Roma people
traditionally get married in earlier age groups and in comparison to the other ethnic
groups the transition to first child occurs much earlier. According to Zhekova (2001)
the high fertility in the Roma group is a result of the cultural and value differences
and of the higher number of unwanted births. The Roma have higher percentage of
unwanted births as they have a lower level of use of effective contraceptive means
(Yachkova, 1998) as well as low family planning when compared to the other ethnic
groups in Bulgaria (Zhekova, 2001).
In addition, we can return again to the study of Pamporov (2003) where he argues that
the Roma population returns back to the traditions, values and social strategies that
were prevailing in the times before the start of the communist regime. This return to
the old values is considered to be a part of the copying strategies of the Roma that
they perform in order to overcome the difficulties in the new economic and political
system. As a result of this, the demographic behavior of this ethnic group is very
similar to the one that was characteristic before the start of the first demographic
transition, namely extremely high mortality and high fertility. The different behavior
of the Roma group can also be regarded as a kind of ethnic identification, which
becomes more and more substantial on the Balkans as a whole.
The other very important question is if the differences in the education level are
substantial for the family formation trends and the transition to motherhood in
Bulgaria. In Table 2 we have summarized our result for the education level and each
transition under study.
Table 2: Relative risks of transitions to direct marriage, cohabitation, marriage after
cohabitation, and first birth according to education level of the women.
Sig. Relative risk
(3) ***: p≤ 0.01 **: 0.01<p≤0.05 *: 0.05<p≤0.10.
Sig. Relative risk
Sig. Relative risk
Our results show that the women with the lowest education have the lowest transition
to direct marriage. However, the trend in the cohabitation is much more different. We
find out that women with higher education are the least susceptible to forming a
consensual union, while between the secondary and primary educated women we do
not find any difference. On the other hand, women with higher education tend to
transform their cohabitation into a legal union, while women with primary education
are the least prone to enter a marriage after being in cohabitation.
Contrary to the expectations that women with higher education are the heralds of new
ideas and the ones who first accept the non-marital cohabitation (Lesthaeghe, 1995)
here we see that this is not the case in Bulgaria. It turns out that women with primary
education also have low risk of entering a direct marriage and women with high
education are the least prone to form a cohabitation. There could be several reasons
for this finding. We suppose that women with primary education belong to the group
of people having no good position on labor market and thus the delay of marriages is
caused by financial difficulties. Cohabitation requires less investment and does not
involve long-term commitment. Therefore it might be a preferred replacement of
marriage for the lowly educated women (Thornton et al., 1995). Furthermore, the
finding that the higher educated women are highly inclined to enter a marriage does
not support the neoclassical economic theory (Becker 1991). This theory states that
the higher the education of women the lower the women’s gains from marriage.
However, this hypothesis is based on the traditional division of labor in the
household. Other authors have already stated that the economic theory does not catch
all the gains that one has in a marriage like for instance psychological or social gains
(Berrington and Diamond 2000). Another reason for the higher proneness of the
higher educated women to enter a marriage could be the longer time that they invest
in education. Usually they postpone the union formation activities till they finish
education and after that within a short time they form a family. This trend is known as
the time-squeeze effect (Kreyenfeld, 2002, Bracher and Santow1998). Similarly,
Billari and Philipov (2003, p.214) find for the case of the Eastern European countries
that “entry into first unions is much more linked to end of education than to the
achieved level of education”. Coppola (2003) also finds out that the human capital
investment seems to accelerate rather to delay the process of union formation. Higher
propensity to marry after being on cohabitation for the higher educated women is also
found for the case of Sweden (Duvander, 1999). One of the explanations for this trend
is that the couples with more economic resources have higher gain from marriage.
In our analyses we were not able to find any significant differences between the
education level of the women on the risk of first birth. The lack of any difference
between the education levels could be due to the fact that the transition to first birth in
Bulgaria is still a very universal process – more than 90 % of the women have at least
one child. The women, who nevertheless stay childless, do not differ obviously by
education level. In other words, the education does not influence the transition to
motherhood. We would assume, that staying childless is either unwanted or is a
decision that is not influenced by the education attainment.
We also find that the enrolment in education has a strong impact on the transition to
first union formation (See Appendix). Our results show that being in education leads
to significantly lower level of willingness to form a family – no matter if we are
talking about marriage or cohabitation. Additionally, it turned out that the education
enrolment does not have any impact on the transition from cohabitation to marriage.
This comes to show that being in school matters only for the first union formation
process. If a woman has formed already an union, then the education enrolment does
not play any role in the transformation of this union. The negative association
between education enrolment and cohabitation or marriage is found in other studies
too – Hoem, 1996; Thornton et al, 1995; Bracher and Santow, 1998; Goldscheider et
al, 2000; Baizan et al, 2003; Coppola, 2003; Nazio and Blossfeld, 2003, to name a
few. A spread view for this trend is that the school enrolment delays women’s
transition to adulthood, in line with the normative expectations that when women are
in school, they are still not ‘ready’ for marriage and motherhood (Blossfeld and
Huinink, 1991). Also, it is regarded that a woman in studies is economically
dependent on her parents (Blossfeld and Huinink, 1991), which affects her ability to
marry or forma a union. In general, students have less money, time and inclination for
the commitments of marriage (Hoem, 1986). Additionally, this causality could be
operating in both ways. From one side, women in school are unlikely to feel that they
have the financial resources to get married, but also for women who are facing
economic constraints is much more difficult to become a student (Rindfuss and Van
den Heuvel, 1990). Sander (1992) also suggests that schooling could be correlated
with unobservables (for instance, the rate of time preferences) that affect the marital
An important question is the influence of childbearing on the decision and timing of
family formation as well as the influence of family formation on the decision to have
a child. Our results show (see Appendix) that there is a strong connection between the
Although we find that the cohabitation is gaining more popularity among the
Bulgarian population, it is also true that when it comes to raising children, still many
women prefer this to happen in a legal marriage. The results show a great importance
of the first pregnancy on the transition to first marriage. That is, women who
experience premarital pregnancy are highly inclined to get married (Blossfeld and
Huinink, 1991; Billari and Kohler, 2000, Goldsheider et al., 2000). The high
proneness of single women to get married when they become pregnant could be
connected to their high desire to offer their child the social and economic environment
and protection that normally accompanies a union (Baizan et al., 2003). Other
explanations for this trend is that the social norms that support the marital fertility still
prevail in the Bulgarian society. In the last years there would be many couples that
cohabit, but obviously, when it comes to children, most of them prefer to have the
children in an official, legal marriage. Similar results are received in other studies in
Bulgaria, for instance by Mirchev (1998).
Not surprisingly, the results on the influence of union formation on the
childbearing decisions showed that the married women have the highest proneness of
having a child, compared to women who have never been married or to divorced and
widowed women. A woman gets married when she plans to have a child or just gets a
child as soon as she marries. The strong link between bearing children and the
decision to get married is found to be important in other studies too (Hoem and
Selmer, 1984; Blossfeld and Huinink, 1991; Thornton et al., 1995; Billari and Kohler,
2000; Buber, 2001 to name a few). However, further investigation is needed to
estimated the strength of the relationship between the marriage and first birth, which
is out of the scope of this paper.
The results on union formation by ethnic group allow us to conclude that there are
different patterns of forming a family for each of the ethnic groups. The Bulgarians
start with the family formation process later than the other ethnic groups and most of
their unions are marriages. If the union starts as cohabitation marriage usually
follows shortly. The Roma group is highly prone of starting a union either by entering
a direct marriage or by starting a cohabitation. In the case of cohabitation, the chance
that it is transformed into a marriage is very low. The Turkish ethnic group stays in its
trends somewhere between the Bulgarians and the Roma. They have a comparatively
early start in family formation, and they tend to form direct marriages and at the same
time are highly inclined to form a consensual union. If a cohabitation is formed it is
not much probable either that it will turn into a marriage.
In our analyses on the effect of the different cultures on the first birth process we
found out that the Roma ethnic group has the highest proneness of childbearing. This
shows that the different ethnic groups respond in different way to the changes in the
economic situation on the macro level. This proves that except for the economic
reasons, we should pay attention to the cultural and value differences when we study
the fertility behavior in a country like Bulgaria.