New fertility behaviour in Eastern Europe: How it has emerged and future prospects

Case studies of Bulgaria, Hungary and Georgia

Marion Burkimsher LaboDémo, University of Geneva in conjunction with the Population Activities Unit, United Nations Economic Commission for Europe

Abstract This project looks at the number of children people have had and expect to have in Bulgaria, Hungary and Georgia. Trends in the changing proportions of those who have remained childless, or who have had 1, 2, 3 or 4 and more children were analysed for women born between 1925 and 1974. The data source was the Generations and Gender Programme (GGP) individual-level database. Changes in the age bands at which women and men have had their first and second child were also studied. The transition from communist regimes to market economies, and the associated disruption to social support systems, had far-reaching effects on those societies, including childbearing. Births are being increasingly postponed, and the age at first birth continues to rise, though it is still younger than in Western Europe. The number of children being born to couples is declining, and in Bulgaria and Hungary an increasing number of couples are staying at a single child, rather than moving on to the previously strong 2-child norm. The total fertility rate (TFR) is being systematically deflated by the increasing age at childbearing for all the countries studied. For Bulgaria, the combination of this, together with low actual fertility, a relatively low desire to have a second child and very low desire for third or higher parity children, means that it is likely to stay in the 'lowest-low' fertility category for some time to come. In contrast, in Georgia, cohort fertility could remain over two children per woman for the near future. In Hungary, fertility rose during the 1970s and 1980s before falling in the 1990s, when births were increasingly postponed. A number of variables were examined for their effect on fertility. The relationship of higher educational attainment of women, smaller family of origin, higher age at first birth and urban residence with lower fertility were as expected. However, the fertility of women who more frequently attend religious services was found to be lower than for those who attend rarely in Bulgaria and Georgia, in contrast to that found in western countries. The difference in number of children between the Roma minority and native populations in Bulgaria and Hungary was particularly striking. In conclusion, the countries of Eastern Europe have traditionally had quite different fertility behaviour patterns to Western Europe: women had their first child much younger, then waited for longer before having a second. Childlessness was very rare, and if fertility desires are achieved, then this is likely to continue. Larger families of three or more children have been a rarity in Bulgaria, and the 2-child norm is being eroded by an increasing number of 1-child families. In both timing and number of children, there was less diversity within each of the countries studied than seen in the western countries. Georgia lags behind the other two countries studied in its fertility trends, but with its higher traditional fertility and greater diversity of behaviour patterns, it would appear to have less of a fertility crisis than the other two countries. Key words: total fertility rate, cohort fertility, lowest-low fertility, GGP, fertility postponement, age at first birth, determinants of fertility

Contents Introduction Goals and outline of research project The place of this study within the Generations and Gender Programme Advantages of the GGP data

Literature review Period or cohort fertility measures: which to use? Convergence or divergence of fertility behaviour? The value of parity-by-parity analyses General fertility determinants Childlessness – significance and determinants First to second child transition Second to third child transition

Data source Variables examined

Comparison of GGP sample results with other published data TFR Band of years for which a valid TFR could be calculated Comparison of TFRs derived from male data and from female data Comparison of TFRs – Bulgaria Comparison with TFR derived from Bulgarian FFS data Comparison of TFRs – Hungary Comparison of TFRs – Georgia Period mean age at childbirth (all parities) Comparison of mean age at childbearing – Bulgaria Comparison of mean age at childbearing – Georgia Cohort mean age at first birth Completed cohort fertility (CCF) Comparison of CCF – Bulgaria Comparison of CCF – Hungary Comparison of CCF – Georgia Why do older cohorts have too low a CCF in the GGP sample? Differences between realised CCF and expected CCF Differences between male and female CCF Proportion of childless women by cohort Parity structure: comparison with Frejka and Sardon Comparison of completed cohort fertility from GGP and FFS Conclusion

1 1 2 2 3 3 5 7 8 10 13 13 14 15 17 17 17 18 19 20 20 21 21 21 22 23 23 24 24 25 26 26 27 27 28 30 31

Changes in timing of childbearing Why is timing so important? Influence of change in parities on mean age of childbearing Variations in shape of curve of age at first birth Measures of central tendency – mode, median and mean Changing proportions of ‘young’ and ‘old’ first-time mothers Bulgaria – changes in timing of first birth Hungary – changes in timing of first birth Georgia – changes in timing of first birth Inter-quartile range of age at first birth Difference between median age at first and at second births Difference between median age at first birth for men and for women Effect of changes in age at childbearing on fertility rates Association of age at first birth and total number of children At what ages do women consider that a child is ‘early’ or ‘late’? Causes in delay in childbearing Summary of implications of changes in timing of childbearing

Changes in parities Parity distribution analysis Trends of changes in parities – Bulgaria Trends of changes in parities – Hungary Trends of changes in parities – Georgia Country-to-country comparisons Childlessness 1-child families 2-child families 3-child families Families with 4 and more children Is the data on proportion of parities reliable? Measures of concentration of reproduction Fertility plans of women Attitude of pregnant women depending on existing parity Falling fertility/rising fertility: which cohorts experienced these? Comparison of TFR and CCF Determinants of fertility Rural/urban residence Number of siblings Education Religiosity Ethnicity Bulgarian parity distributions by ethnicity Hungarian parity distributions by ethnicity Summary of fertility trends and determinants

32 32 32 32 33 34 36 37 38 39 41 43 43 45 47 49 49 51 51 51 52 54 55 55 56 56 57 58 58 59 63 64 65 65 66 66 67 68 69 71 72 73 73

Overview of discussion topics Fertility predictions Increasing diversity – is it a ‘good thing’? Education, urban living, family of origin, religion The ethnic challenge Policy applications Where more research is needed End note Acknowledgements

References

75 75 76 77 77 78 79 79 79 80

1 _____________________________________________________________________

Introduction The countries of Eastern Europe have had a tumultuous history over the past sixty years. The traumas of the Second World War were succeeded by the spread of communist ideology into all sectors of life, followed by its abrupt collapse in 1989. These social and political upheavals have greatly influenced fertility behaviour, and continue to do so. The number of children born in each year and to each cohort of women is, in effect, a fossilised reflection of the social situation of that time. Changes in a country's social and economic outlook, government policies, the prevailing belief systems, lifestyles and contraceptive availability all have a profound influence on fertility outcomes. This study looks at the changes in fertility behaviour – the timing of childbearing and the number of children - of women and men born after the mid-1920s. These correspond to those cohorts who have gone through their fertile life primarily in the post-war years. The effects of the transition from communist to capitalist systems is of particular interest, as is a prospective view of what may be expected in the coming years. The three countries studied - Bulgaria, Hungary and Georgia - have seen dramatic falls in their total fertility rates (TFRs), reaching ‘lowest low’ levels of under 1.3 in the case of Bulgaria and Hungary. They share major concerns about declining absolute population numbers caused by a combination of negative net migration, falling life expectancy in the years immediately after 1989, and very low birth rates. A major concern is that too few babies are being born. Too few for what? Too few to balance the number of people dying, so the population in Bulgaria and Hungary is declining. Too few to grow up and supply the work force with skilled labour, a problem that is likely to get significantly worse in the years to come. Too few to grow up and financially support the bulge of middle-aged people who will be retiring in the coming decades. The population is currently enjoying the ‘demographic bonus’ of a large working age population relative to young and old dependents, but this will turn into a top-heavy population pyramid in the coming decades as the bulge ages, and social security systems will struggle to support their ageing population. Goals and outline of research project The primary aim of this research is to elucidate current fertility behaviour in the light of past trends, and to suggest scenarios of what the future might hold. In particular, the intention is to highlight where research and policy interventions may best be focussed to support women in their fertility intentions. The information gained from this analysis will be foundational data on which to build a sound longitudinal analysis of the second wave data of the Generation and Gender Programme (GGP). There are six main sections in this report. The first gives an overview of the literature covering previous research of various relevant fertility issues. The second describes the source of data used for the analysis, the individual-level databases of the GGP. _____________________________________________________________________ New fertility behaviour in Eastern Europe

2 _____________________________________________________________________ The third assesses the representativeness of the GGP sample data by comparison with demographic indicators derived from vital statistics. The fourth section looks at changes in timing of childbearing and of age at first birth in particular, and to see how these influence both period and cohort fertility rates. The fifth section investigates the variations in the number of children born by cohort and between the three countries studied. The aim is to diagnose which changes in the parity distributions have been influential in the recent fall in fertility and so pinpoint where further research is needed. A number of different determinants of variations in fertility levels are investigated as foundational information in preparation of a more comprehensive longitudinal study into the barriers of progression from one parity to the next. The sixth and final section presents an overview of the most pertinent topics presented in the report, with a particular emphasis on what the future may hold and where policymakers and future researchers may wish focus their attention. The place of this study within the Generations and Gender Programme There is currently a general shortfall between the number of children that people in developed countries say they want to have and the number they actually end up having. One of the central purposes of the Generations and Gender Programme (GGP) is to investigate the barriers that people encounter when trying to attain their fertility goals – and consider if these can be surmounted by policy interventions. For this reason the GGP is a longitudinal study with at least three waves, so that the decisionmaking and implementation processes can be analysed in ‘real time’. In particular, it was designed to answer the question of what determinants at the micro- and macrolevel may act as barriers or incentives to fertility at an individual level. The purpose of this particular study is to lay the groundwork for more detailed longitudinal analysis of the second wave of data. Some of the determinants of fertility are time-dependent and change during a person’s fertile life (eg. age, employment situation, partnership status) while others are generally time independent during that period of life (eg. number of siblings in family of origin, educational level). This study will focus primarily on time independent variables in order that their influence is known and so they can be control variables for a subsequent analysis of second wave data. Advantages of the GGP data The GGP data has several advantages that other data sources pertaining to fertility do not. Firstly, it is designed for country-to-country comparisons, having a standardised questionnaire and data base. Therefore, time-consuming analysis of different determinants for one country can easily be replicated for other countries.

_____________________________________________________________________ Marion Burkimsher

3 _____________________________________________________________________ Secondly, as cohort fertility is our interest, the fact that the respondents range in age from 18 to 79 is particularly valuable. This gives a wide band of cohorts who have passed through a variety of welfare regimes and life experiences. Thirdly, the GGP sample data describes the fertility outcomes of residents of the country who are currently living there. Cohort fertility measures can be calculated from birth registration records if the data records are sufficiently long and are accessible, but they will naturally include women who have borne children in the country but who have subsequently moved away or died. Similarly, the fertility of immigrants who gave birth outside the country cannot be known from birth registration data. Fourthly, although ascertaining of individuals’ short-term fertility plans is extremely valuable in itself, it will be even more enlightening to discover in the second wave whether these plans were fulfilled. If a large proportion of people are, in fact, planning to have a first or subsequent child in the coming years, then the outlook for completed fertility will be completely different than if we just look at current fertility levels. It allows predictions to be made about fertility, even for cohorts just reaching their 30s. Fifthly, there is useful information to be gleaned from pregnant women and their partners, who were asked specific questions on whether or not their pregnancy was desired or not, and whether the timing was sooner or later than planned. This can give useful additional insights about societal attitudes to the timing and number of children. Finally, the questionnaire includes a wide range of questions on family background, life history, current situation, opinions and attitudes (including those of the respondent’s family and friends) that can be investigated as possible determinants.

Literature review This section looks at a number of pertinent questions relevant to an in-depth study of fertility, issues that are important for understanding the results presented later in the report. The advantages and limitations of the two primary measures of fertility, the TFR and completed cohort fertility, are discussed, together with the benefits of analysing fertility on a parity-by-parity basis. Both the variation in age of giving birth and the number of children borne can become narrower or wider over time, and this is of particular interest, as the fall in communism saw U-turn from convergence in behaviour to divergence. Fertility choices are determined by many different factors, and some of these are discussed in the latter part of this section. Period or cohort fertility measures: which to use? There are two ways of looking at fertility, the period (year to year) perspective and the cohort (generation of women) perspective. There has been animated debate about the advantages and shortcomings of each in the demographic literature (eg. Bongaarts and Feeney 2005, Sobotka 2004, Ni Bhrolchain, 1992). The problem is the distortion, in _____________________________________________________________________ New fertility behaviour in Eastern Europe

4 _____________________________________________________________________ particular the deflation, of the TFR because of the current ongoing postponement of births; in contrast, in the baby boom years, the TFR was inflated because women were bearing children successively younger ages. To quote Sobotka from his in-depth treatise on the postponement of childbearing in Europe: “The TFR may give misleading signals about trends in cohort fertility for a long period of time”. Bongaarts and Feeney (2005), in their Working Paper for the UN Population Council, give a similar warning about erroneous interpretation of the TFR: “Distorted views of past levels and trends in the quantum and tempo of lifecycle events may lead to misleading projections and to the adoption of sub-optimal social and health policies”. However, Sabotka (2004) adds: “While the TFR constitutes a distorted, potentially misleading indicator of fertility, each of its alternative counterparts also has shortcomings, and none of them represents an unambiguous measure of fertility quantum”. The cohort perspective more accurately reflects the experience of individuals as they pass through their different seasons of life in parallel with events happening around them. From a fertility perspective, a woman may choose to postpone having a child if the conditions are not conducive to childbearing (such as during the tumultuous time of transition for the eastern bloc countries in their move to market economies), or to have one sooner if there are special incentives. These can be financial or other incentives offered by the state (or others), or the attraction of having a child in a special year (eg. auspicious years in some cultures, or even the year 2000 for those in the western world). It might seem to be a rather tacky way of ‘buying’ babies, but to quote one recent example: “The governor of Ulyanovsk region in Russia is offering prizes to couples who have babies in exactly nine months - on Russia's national day on 12 June. Sergei Morozov wants couples to take the day off work to have sex. If a baby is born on national day, they will receive cars, TVs or other prizes”. (BBC 2007a). An advantage of cohort fertility measures is that they tend to change more slowly than period indicators. To quote Sabotka (2004): “All period indicators fluctuate more than the relatively stable cohort fertility rate. This is consistent with the fact that period measures reflect period influences that are often temporary and less stable than cohort trends, which we may think of as an aggregate result of changes over a long period of time. Some period fluctuations may be interpreted in terms of plausible explanations (population policy measures, economic influences, etc)”. The problem with analysing completed cohort fertility (CCF) is that it can only fairly be analysed when a woman’s reproductive life is complete, usually considered to be at age 50. However, few women have children when over 40, so that age is often considered as the earliest for satisfactory CCF comparisons. Sabotka, looking at European birth registration data, found that less than 1.5% of first births take place over age 40, and so estimates of these additional births can be incorporated into the CCF index. Therefore, in 2002, we can have a reasonable idea of the CCF of the 1962 cohort of women – but less reliably for those born after then. For men their fertile period is never definitively ended, at least until a vasectomy – or death (perhaps even longer if their sperm were frozen!). This may have been one reason why men’s fertility has been examined less often than women’s, as it is more difficult to estimate reliably their CCF rates. _____________________________________________________________________ Marion Burkimsher

5 _____________________________________________________________________ A potential new source of information for improving estimates of cohort fertility is becoming available with some recent surveys asking respondents about their shortand medium-term fertility intentions. The GGP is one such survey, and it is planned to exploit this information in the analysis. Whether fertility intentions translate into actual births, however, remains to be seen. Convergence or divergence of fertility behaviour? Is behaviour becoming more similar (convergent) or more varied (divergent) in the realm of fertility? This is an interesting question. It can be looked at either by comparing countries (are all countries heading in the same direction in their fertility behaviour or is there a widening gulf in the patterns observed in different countries?) or looking at behaviour within countries. In general, the debate on whether the European countries are becoming more similar or more varied in their fertility behaviour is favouring those who perceive divergence. Although the trends may all be in a similar direction (older childbearing, more births outside marriage), the different rates of change between countries mean that they currently have widely differing norms. The jury is still out on whether they will ultimately approach similar outcomes. To quote a Eurodata report on the diverging patterns of household and family types across the European countries after a long period of convergence “Not only for the European Union but also for sociological research in general the question of divergence or convergence of national household and family structures and of national demographic developments is of great importance. Until the 1960s a growing similarity of patterns in household and family structures could be observed in the industrialised countries of Western Europe. Since that time some family patterns have been showing a tendency to diverge in the European countries” (Rothenbacher, 1994). Measures of convergence/divergence pertaining to two distinct aspects fertility are relevant to this study: age at childbearing and number of children borne. Concerning the age at childbearing, generally it is the age at first birth that is most commonly examined. A widening of the curve of first birth rate against age indicates divergence in behaviour norms of childbearing. Some countries, such as the United States and the United Kingdom have experienced this most markedly (Sabotka, 2004). As an example, the interdecile range of first birth timing has increased from nearly 12 years in 1980 to approaching 16 years in 2000 in the UK. By contrast, other demographers have hypothesised increased ‘rectangularization’ or ‘concentration of childbearing within an increasingly narrow age interval’ between the late 20s and early 30s, particularly in countries with lowest-low fertility (Kohler, Billari and Ortega, 2002). The second attribute that is of particular interest is the number (or quantum) of children borne to a woman (or man). Again the general verdict is that growing divergence of family sizes is being seen within some countries. A growing gulf between women who have children and those who do not has been noted; for example to quote Pinnelli, Hoffmann-Nowotny and Fux (2001): “The last mentioned group of countries, (namely the Netherlands or Switzerland) is already characterised by the _____________________________________________________________________ New fertility behaviour in Eastern Europe

6 _____________________________________________________________________ comparatively pronounced polarization between the family and the non-family sector…”. Several measures of convergence/divergence can be used. The interquartile or interdecile range was used by Sabotka (2004) to describe change in age at first birth. Two other measures have been used to describe variation in family sizes: ‘standardized variance’ (which is variance divided by the mean, as opposed to coefficient of variation, which is standard deviation divided by the mean), applied in an early study by Preston (1976); and concentration (or Gini) coefficient, used by Vaupel and Goodwin (1987) and Shkolnikov (2004). The coefficient of variation (an alternative way of expressing this is as a percentage, in which case it is called the relative standard deviation) has also been used by the author to investigate variations in parity in Switzerland (Burkimsher, 2006). Preston’s analysis (1976) is novel in that he looked at the different perspective of family size experienced by mothers and by children. Only Shkolnikov (2004) has followed up on this potentially important aspect of fertility. The wider the diversity of number of children borne by women, the larger the average family size experienced by children will be. He cites a simple example: “if half of women have four children and half have none, the average family size for a woman will be two but for a child it will be twice as large”. He also says that in most western countries, there had been a significant contraction in the standardized variance of family sizes (up to that point in 1976), although in Ireland there was still ‘substantial dispersion… a product of high proportions who never married combined with high fertility for those who did’. Vaupel and Goodwin (1987) look at the question of family size variation from the angle of ‘the concentration of reproduction’, in other words “What proportion of a cohort of women has what proportion of the children?” They used a standard economic measure of concentration coefficient, also known as the Gini coefficient; this can be plotted on a graph as a Lorenz curve. There are two associated measures with this type of analysis, ‘Have-half’ and ‘Half-have’ – in other words the proportion of women who have half the children, and half the women have what proportion of children. They looked only at past data from the United States, and their conclusion is significant: “Populations are heterogeneous. Averages hide that heterogeneity. Knowing that the cohort of US women who completed their fertility in 1980 had 3.16 children on average tells only part of the story. The 36 percent of these women who had four children or more accounted for 63 percent of the children. Equivalently, 63 percent of the children were born in families with four or more children”. Both Preston and Vaupel and Goodwin came from an era in which there was concern that certain segments of their community (Blacks in the United States) were in danger of producing an excess of children in comparison to other sectors (particularly educated whites). However, the importance of targeting policy measures to women of certain parities still holds true: “concentration analyses may be relevant to policy decisions, especially those relating to the targeting of an intervention. If one-fifth of women are bearing half the children, perhaps policies to reduce (or to increase) births should be directed toward this group” (Vaupel and Goodwin, 1987). Perhaps in Europe today there may be more emphasis on measures to increase fertility wherever possible.

_____________________________________________________________________ Marion Burkimsher

7 _____________________________________________________________________ Most recently Shkolnikov et al (2004) have done a thorough examination of the concentration ratio (or Gini coefficient) for cohorts of women (as well as looking at just mothers and family size of children) for many European countries plus the United States. Their studies show that: “Concentration of reproduction was high in cohorts of the beginning of 20th century due to high proportions of childless women combined with high proportions of women with many children. Concentration of reproduction declined across younger US cohorts. Its highest values were observed in the US cohorts born around 1910 and the lowest value was observed in the cohort of 1931”. Comparing European and American trends of convergence/divergence, their conclusions were as follows: “in a majority of countries inter-individual differences in fertility have begun to increase from older to younger cohorts. The increase started from cohorts of the mid-1930s in the USA and cohorts of the mid- or late 1940s in Western Europe. In Eastern Europe the change was less pronounced and began in cohorts of the 1950s. Thus, population reproduction becomes more unevenly distributed among women and their smaller proportions produce greater proportions of children. Comparison of trends in concentration of reproduction for all women with the equivalent trends for mothers suggest that growing childlessness is the greatest contributor to the general increase in western countries”. The value of parity-by-parity analyses Most studies of fertility look at average values: a few, such as those examined in the section above, also look at the spread of family sizes. However, as children come only in integer units, then it is also possible to look at each parity separately. The need to look at the evolution of each parity separately has been discussed for a long time (Ni Bhrolchain, 1992, discusses the history and emphasises this point), though few studies have actually done so. Most recently it was underlined by Sabotka (2004): “Whenever possible, fertility scenarios should utilise parity-specific data”. Now that families in developed countries come in just a few different sizes, the delicate balance between the proportions of women who have no children or 1, 2, 3 or more can have a marked impact on total fertility rates. Two recent studies have looked at this: one by the United Nations Population Division (2003) - a wide-ranging study on ‘Partnership and reproductive behaviour in low-fertility countries’ - and the second, already cited study by Shkolnikov et al (2004) on ‘The concentration of reproduction in cohorts of US and European women’. The first of these looks at four influences which can determine the overall fertility level in a country: age at first birth; prevalence of childlessness; propensity to have more than one child; and propensity to have three or more children. Some countries, such as the United States, have a high prevalence of childlessness, but as the age at first birth is quite young and there are high propensities of women to have more than one child and to move on to three or more children, then the overall fertility rate is approaching replacement level. France also has a relatively high fertility level, even though the age at first birth is high, because few women remain childless, many have more than one child and larger families are still quite common. It is the conjuncture of high age at first birth, a high prevalence of childlessness, many mothers staying at just one child and larger families being uncommon, that can cause fertility levels to be extremely low: the report cites the countries of eastern Asia, southern Europe, Austria, Canada and Germany to have just this combination of fertility deflators. From a cohort perspective, they suggest that “the average completed family size of women _____________________________________________________________________ New fertility behaviour in Eastern Europe

8 _____________________________________________________________________ born in the mid-1960s will probably vary from 1.6 children in Austria, Italy, Germany and the Russian Federation to 1.9 to 2.1 children per woman in a number of countries from Eastern Europe and Northern Europe, France and the United States”. The UN Population Division report (2003) gives an excellent example of how the composition of fertility can vary, even if the total fertility rate is the same: “The structure of female cohorts by the number of children varies considerably, even though completed fertility may be similar. For example, women born circa 1960 in France, Japan, and the United States had on average 1.9 to 2.1 children. Yet, women who had three or more children constitute 31 per cent of the American cohort and 33 per cent of the French cohort, while in Japan such women represent only 13 per cent of the 1960 cohort. On the other hand, 3 per cent of Japanese and 8 per cent of French remained childless, while this proportion attained 20 per cent in the United States”. Another example shows how childlessness does not, of itself, predetermine a low overall fertility rate: “Many women remained childless in the United Kingdom (21 per cent) and the United States (15 per cent), but childlessness was compensated by higher propensity of women who already had one child to bear a second, and of those with two children to have more. On the contrary, in Italy, Spain and the Russian Federation only 7 to 11 per cent of women remained childless, but many fewer women who had one child had a second, and of those with two children a much smaller number proceeded to higher parities”. Shkolnikov et al (2004) has made the most detailed study to date of the proportions of each parity by cohort in European countries and the United States. Looking at cohorts of women born in the late 1950s and early 1960s, the country with the highest level of childlessness was West Germany (24% of women) while the lowest was Bulgaria (5%). The country with the highest proportion of 1-child women was also West Germany (27%), closely followed by Russia and Spain; Ireland had the lowest proportion (10%). In Bulgaria the 2-child norm was strongest with 57% of women in that country falling into that category. Only 29% of Irish women, by contrast, fell into the 2-child category. There was a wide spread for 3-child women: from a high of 26% in Norway to a low of 10% in West Germany and Bulgaria. As for large families of four or more children, Ireland again was noticeably higher at 18% than its next nearest rival, Romania, at 14%. Slovenia and Spain were in the lowest bracket of large families with around 3% falling in this category. West Germany, perhaps surprisingly, was not in the lowest group of countries with large families, having a higher proportion than Italy, Denmark, Russia, Czech Republic and Bulgaria, as well as Slovenia and Spain. General fertility determinants The discussion above has described the effects of changing distributions of parity on overall fertility rates. But what factors determine a particular fertility level and how do changes in these factors cause changes over time? In these days of reliable modern contraceptives (replacing abortion as the main means of birth control), childbearing has moved into the realm of requiring a positive choice rather than simply the default if one has a partner, while delaying the decision about such a life-changing transition has become easy.

_____________________________________________________________________ Marion Burkimsher

9 _____________________________________________________________________ The influence of age has already been mentioned, but this needs to be discussed in more depth. An increasing age at childbearing naturally causes a deflation effect in period measures such as the TFR: however, if women simply delay their childbearing at younger ages and this is recouped at higher ages, then completed cohort fertility (CCF) may not be affected. However, postponement may also cause cohort fertility to fall for several other reasons, which have been examined in detail by Kohler, Billari and Ortega (2002) and by Beets (2006). The problem with delaying childbearing is that it can become harder to conceive with increasing age after 30 and most markedly after 35. Economic uncertainty, particularly in the labour market, can “provide an incentive to delay decisions that imply long-term commitments, such as the decision to have children” (Kohler, Billari and Ortega, 2002): the problem is that people can wait too long, until the decision about having a child or not is taken out of their hands because of age. More and more women are seeking assisted reproduction technology (ART), but this is not a solution for many individuals and comes at a high price for countries (Beets, 2006). There are also self-sustaining social multiplier effects in the increasing age at childbearing, so it is probably likely that childbearing ages will continue to rise for some time still, especially in eastern Europe, where childbearing is still young in comparison to western European countries (Kohler, Billari and Ortega, 2002). The effect of religious adherence and participation in religion (religiosity) has been focussed on by Frejka and Westoff (2006) to potentially explain the difference in fertility levels between the United States and European countries. At an individual level they found that higher fertility is commonly found in people with a higher frequency of attendance at religious services, both in the U.S. and throughout Europe ( Frejka and Westoff, 2006). It might be hypothesized that participation in religion, both as a lifestyle and as a group activity, provides the security in the face of uncertainly that the outside world does not. A previous study by the author (Burkimsher 2007b) confirmed that religious belief and practice was a primary explicative variable of individual childlessness in Switzerland, those participating very rarely in religion being much more likely to remain childless (this was the same for men and for women). Here lies a conundrum, however. Those European countries currently seeing higher fertility levels (France and the Scandinavian countries) have a higher proportion of individuals who are non-religious. Clearly factors that work at the national level and at the individual level need not be the same – the ecological fallacy. The same holds true for women’s participation in the labour force and divorce: although both of those decrease an individual’s fertility outlook, countries which currently have high rates also have relatively higher fertility compared to countries where divorce is uncommon and women working outside the home is less common (Billari and Kohler, 2004; Engelhardt and Prskawetz, 2004). Fertility levels have wide variations within countries and it is common to find that cities have much lower fertility rates and much higher proportions of childless people compared to rural areas (eg, the study of Switzerland by Burkimsher, 2006; the study of Japan by Hirosima and Mita, 1995). Some of the lowest TFRs today are in city states, eg. Hong Kong, Singapore and Macau have estimated TFRs for 2007 of around 1 child per woman (CIA, 2007). Another interesting factor that has been examined recently is the role of the couple in fertility decisions (Thomson, 1997). To express it in laymen’s terms, ‘It takes two to _____________________________________________________________________ New fertility behaviour in Eastern Europe

10 _____________________________________________________________________ tango!’ Thomson states: “The results show clearly that husbands' desires and intentions influence couples' births, with approximately equal force to that of wives' desires and intentions. When couples disagreed about wanting a child, each partners' intentions were shifted toward not having a child; and disagreement in desires or intentions were reflected in birth rates that were lower than average”. Major changes in attitudes and behaviour have been used to explain the fall in fertility in developed countries. One specific example is Japan – a developed country which has similar fertility patterns to some European countries (Atoh, 2001). He (she?) attributes the fall in fertility in that country to the increase in the proportion of young people who never marry (in Japan the number of births outside marriage is still very low). This is explained by several parallel developments: a trend towards individualism; a change in attitude towards the status of women; an increasingly positive attitude to girls giving rise to their increased enrolment in higher education; and major changes of attitude towards aged parents, whether a woman should marry, divorce, sexuality, etc. Finally, an emphasis on economic and social structure determinants of fertility is commonly seen in explanations of country-to-country variations. The advantage of this approach is that policy makers can see if they can influence fertility by improving the economic or societal support of families. The problem is that societies are complex, with a nation’s inhabitants bonded into behaviour patterns and norms that are not quickly or easily changed. There is no doubt that policies can influence shortterm birth rates – individuals can choose economically opportune moments to have a child – but whether this changes the longer term cohort fertility rates is less easy to see. Policies need to be planned for the long haul and need to be self-sustaining. Otherwise a short-term baby boom is likely to be followed by a ‘baby bust. This would not necessarily affect cohort fertility outcomes, but general long-term policies, which are effectively pro- or anti-natalist in practice, can promote an environment that is more or less conducive to having children. Some eminent demographers see little likelihood of cohort fertility falling very far below replacement level in most countries, “unless circumstances change drastically” because most women, young and older, plan on having two children (Bongaarts, 1999); while others, such as Coleman (1998), wonder “why enlightened people ever choose to have any children at all”. Childlessness – significance and determinants The number of people who remain childless throughout their life is of interest for two reasons. The first is that if the proportion is large, then it naturally pulls down the fertility rate of the country: a higher proportion of women would need to have 3 or more children for the fertility rate to approach replacement level. As Shkolnikov et al (2004) state: “childlessness makes an especially important contribution to temporal change of the concentration of reproduction. Therefore, it is useful to distinguish between two parts of diversity: diversity due to proportion of childless and diversity due to variability in number of children among mothers”. The second reason that the proportion of the population remaining childless is important is that, as they age, they will tend to be more dependent on the state for _____________________________________________________________________ Marion Burkimsher

11 _____________________________________________________________________ their welfare. In the past, such as the early 20th century, when childlessness was also common, the extended family, particularly siblings, nieces and nephews, would care for old spinsters/bachelors and childless widow(er)s (Rowland, 1998). However, with shrinking family sizes, the chance of having surviving relatives – particularly those interested in caring for an elderly aunt or uncle – is diminishing. The full burden of care is then left to the social services. From time to time the high level of childlessness in some western countries hits the news, eg “Germany agonises over 30% childless women” (Harding, 2006). Some people, from policy makers to some demographers, have interpreted very low period total fertility rates (TFRs) as a direct indication of increased levels of childlessness. The two do not necessarily go hand in hand. The TFR measure has, over most European countries, been deflated for years or sometimes decades by the increasing age at childbirth for each parity (Sabotka, 2004, has analysed this in great depth). As an example, Italy has experienced lowest-low TFRs of less than 1.3 children per woman in recent years, but Italian women have less chance of remaining childless than, for example, English women, for whom the TFR is considerably higher (Frejka and Sardon, 2006). In England and Wales, there are more larger families to compensate for those who have no children, and this then raises the mean number of children. People who remain childless fall into two quite distinct groups: one is the individuals who have never been in a stable partnership (and formal marriage is still a prerequisite in many western cultures, even today); the other is those who have been in a partnership but who have either chosen not to have children or who have delayed it too long. The Australian newspaper, The Age, would blame partnering difficulties as the main reason for low fertility and childlessness in particular. They headline an article “Fertility crisis: why you can’t blame the blokes” (Birrell, 2003). Several reasons for childlessness are cited. Educated women have trouble finding equally qualified male partners as female education rates exceed those of young men. Men on low or insecure incomes are not desirable to women who have jobs themselves and are “are likely to be wary of taking on the responsibility of marriage and parenthood”. He concludes that men are the ones suffering: “In these circumstances, single males might be casualties, not beneficiaries, of the marriage marketplace. Sadly, the prospects of partnership do not improve with age. For men in their 30s nearly half the single women of the same age are single parents. This situation naturally complicates the partnering process. Change will not be easy. Low partnering levels are no longer the sole province of educated men and women” (Birrell, 2003). Other researchers would say childlessness is a choice, particularly for some women eg. to quote Hakim (2004): “childlessness is most frequently voluntary and a result of a lifestyle that prioritises careers, personal development and material wellbeing over family life. Thus, rising childlessness may be the result of a polarisation of women’s choices into either a work-oriented or a family-oriented life, and the increasing prevalence of women choosing the former lifestyle”. Others would consider it less so, particularly for women with a partner, as proposed by Toulemon in his paper entitled “Very few couples remain voluntarily childless” (Toulemon, 1996). Most people would agree that the first requirement for having a child is having a partner (not _____________________________________________________________________ New fertility behaviour in Eastern Europe

12 _____________________________________________________________________ necessarily a spouse any more): the second is to be out of education (Gonzalez and Jurado-Guerrero, 2006). Higher educational achievement has been associated with higher risk of childlessness in several studies, though some would say that this is mainly because of postponement – waiting too long until finally a woman has passed the age when she can (easily) become pregnant. Sabotka (2004) suggests: “childlessness is typically an outcome of sequential postponement and prolonged indecision about parenthood… an increasing proportion of women remaining childless past age 30 may, for a variety of reasons, remain childless until the end of their childbearing age.” Similarly, Gonzalez and Jurado-Guerrero (2006) say: “We maintain that uncertainty is partly responsible for the progressive delay of motherhood and the eventual increase in unintended childlessness”. They cite France as an example of a country where childlessness is relatively rare: “In France, life transitions happen earlier. That also lessens risk of childlessness”. From their comparison of childlessness in France, West Germany, Italy and Spain, they consider economic issues - employment stability and income - to be foremost determinants of who does and doesn’t become a parent, and country-tocountry variations to be because of different levels of institutional support for young couples. Their conclusion is: “national institutions influence women’s life transitions, in particular partnership and motherhood. For coupled women, we find 2 alternative modes out of childlessness. In countries with high direct and indirect child costs, like Spain and Italy, entering a male-breadwinner couple or occupying a stable and highincome position facilitates motherhood, while in the French context motherhood is most likely in a dual-earner partnership”. In a previous study by the author on determinants of childlessness in Switzerland (Burkimsher 2007b), from a longitudinal analysis on Swiss Panel data, the following variables were found to have a significant influence: cohort (with more recent cohorts being more likely to remain childless); place of residence (region of Switzerland and, even more importantly, whether rural or urban); level of education; level of religious participation; and whether or not she grew up with a sibling. There were also small effects of background, with higher or lower than average social backgrounds producing a lower propensity to have a child. Women who refused to declare a political leaning were also significantly less likely to have a child. The likelihood of remaining childless was also investigated for men, something rarely investigated. The main factors - in particular cohort and religious participation were the same. The effect of region was somewhat more significant; but conversely the effect of rural or urban residence was much smaller for men than for women. One hypothesis was that perhaps men stay put, while women move to the city if they remain single. Educational level had a quite different effect on men than on women: although lesser educated men were more likely to marry younger than more highly educated ones, it had no effect on child-bearing. Well-educated men may have their children later in life, but they are less likely in the long run to end up childless. Sabotka (2006) has made predictions about childlessness in developed countries and discusses these in the light of past trends: “Lifetime childlessness in the highchildlessness regions like England and Wales, West Germany, and Austria is likely to come close to 25%, and almost certainly remain below 30%, while the more ‘common’ childlessness levels will range between 15 and 22%. Historical estimates of _____________________________________________________________________ Marion Burkimsher

13 _____________________________________________________________________ childlessness reveal that a large proportion of women born in the second half of the 19th and first quarter of the 20th century remained childless – 19% white women, 25% non-white in USA, 25% in France, 26% in Germany and the Netherlands, 30% in Australia for women born beginning of 20th century”. The causes of those high levels were a “combination of a high proportion of women never marrying and high childlessness in marriage, partly attributable to negative economic conditions during the economic crisis of the 1930s”. He would agree that economic factors are important in determining country-to-country variations: “If we assume a strong childbearing motivation as a given, the question remains how to explain increasing differences in childlessness levels between countries….the policies facilitating a flexible combination of work and childrearing appear to play a strong role”. First to second child transition Just a few studies have been made on the transition from one child to two from an event history perspective. Rieck (2006) has recently done this for Russian women using GGP data, and the following variables proved to be significant: time period of the second birth (with 1985-1988 being the time period when state pro-natalist policies in Russia were encouraging women to have more children); age at first birth; marital status; education (being in education lowered the chance significantly, but otherwise there was no difference in likelihood between more or less highly educated women having a second child); and number of siblings (if she had none, there was significantly more likelihood of her staying with just a single child). Rieck comments: “Whereas in the 1980s Russia tended to be a two-child-society (50 percent of women had two children), fertility dropped sharply after the political and economic transition, especially owing to a reduction in second order births”. Torr and Short (2004) also examine the transition from one child to two children in the United States in the period 1987-1994, using the multi-wave National Survey of Families and Households. They looked only at couples where the woman had one child at the time of the first survey and she was employed outside the home. In particular they looked at whether the division of housework between the husband and wife influenced which of the couples went on to have a second child. Couples were considered to be ‘modern’ if the housework was shared relatively equally (the wife doing less than 54%), ‘traditional’ if the woman did the vast majority of the housework (over 84%) or ‘partial sharers’ for couples lying somewhere between the two extremes. Their analysis showed that: “the probability of a second birth varies substantially by division of housework. Some 81 percent of modern couples and 74 percent of traditional couples would be expected to have a second child. By contrast, only 55 percent of couples who fall in the middle 50 percent of the distribution would be expected to have a second child. Thus, ‘modern’ and ‘traditional’ couples seem less likely than ‘partial-sharers’ to delay or forgo second births”. Another factor which significantly increased the likelihood of a second child was if the woman had a college degree, while factors that decreased the probability were (not surprisingly) being over age 35 or it being five or more years since the first child. Second to third child transition Again, this transition has been studied in just a few instances. Alich (2006) comments “Taking into account the widespread two-child norm, we assume that three-child _____________________________________________________________________ New fertility behaviour in Eastern Europe

14 _____________________________________________________________________ mothers form a special group with different motivations and fertility ideals”. He compared women in two countries using Fertility and Family Survey (FFS) data and observed: “The third-birth propensity in Norway and West Germany differs remarkably. The gap between the two countries is even larger when we take a look at women’s perceptions of the ideal family size. In the datasets analyzed, nearly half (48 percent) of all Norwegian women perceive a family with at least three children to be an ideal size, in contrast to only 19 percent in West Germany”. He then looked at ten potential variables which may affect the intensity of the transition, which were of three types: “1. Demographic determinants, that is, the importance of age and timing; 2. Socio-psychological determinants and family composition; 3. Socio-economic and welfare determinants”. The factor that he found was the most significance in determining whether a woman would go on to have a third child in both West Germany and Norway was entering into a second marriage after the birth of her second child. Having a highly educated partner increased the likelihood of having a third child. Other potential determinants which were investigated, and which varied between the two countries were: age at first birth; duration between 1st and 2nd child; gender of the first two children (having two the same, especially two boys significantly increased the likelihood of having a third child in Norway, but the effect was not significant in West Germany); urban or rural residence before 15; education level of the mother (low and highly educated women had somewhat higher likelihood than the middle band); having worked between the 1st and 2nd birth; having worked after the 2nd birth; and birth cohort. Interestingly, in West Germany, cohorts of women born after 1958 were significantly more likely to have a third child than those born between 1952 and 1957: in Norway, later cohorts were slightly less likely to have a third child, though the effect was not significant. Comparing the two countries, Alich says: “It appears that the impact of different demographic factors is very similar across the two countries…Differences in thirdbirth behaviour between West German and Norwegian women are primarily connected to socio-economic determinants. We ascribe these differences to differences in the German and Scandinavian welfare-regimes and family policies”.

Data source The data source used for this study is that of the Generations and Gender Programme (GGP). This is a multi-country and multi-wave survey covering many issues, including fertility, pertaining to relationships between men and women and between parents and grandparents, children and grandchildren. It follows on from three previous surveys, which focussed more specifically on fertility questions – the Comparative Fertility Surveys of the 1965-1972, the World Fertility Survey (WFS) of 1975-1981 and the Fertility and Family Surveys (FFS) of 1988-1999. Several weaknesses of the FFS were addressed when designing the GGP: these include sampling a wider age band of respondents to study historical trends (the GGP includes 18-79 year olds); covering males and females roughly equally and including adequate data about partner attributes; having a multi-wave design, so that fertility _____________________________________________________________________ Marion Burkimsher

15 _____________________________________________________________________ decisions and outcomes can be studied in ‘real time’; including more issues that can investigated as potential explanatory variables; and including several pertinent questions to pregnant women and those planning a child in the near future. The aim is to be able to answer the question as to what the real determinants of individual fertility decisions and outcomes are today (micro-level and macro-level) and thus discover what barriers may exist to people in achieving their fertility desires. Previous surveys have found that in the developed countries there is a considerable gap between the number of children people say they would like, and the number they actually end up having (eg. van Peer, 2002). If stated fertility wishes were achieved, then in most countries fertility levels would be at least at replacement level. Finding the root causes of this imbalance is one of the primary goals of the GGP, and thus being able to make proposals to policymakers as to what measures may help to facilitate people fulfil their fertility goals. Micro-level data of the first wave of the GGP is just becoming available for intercountry analysis, after cleaning and harmonisation of the databases by the Population Activities Unit (PAU) of the United Nations Economic Commission for Europe (UNECE). This study covers Bulgaria, Hungary and Georgia, three countries which are quite different, yet which share the common path of experiencing the rise and fall of the communist system. The number of respondents in each country’s survey and month and year of the survey is shown in the following table. The Hungarian survey data is a little different in that it an amalgam of two waves of surveys: however, the fertility data that is relevant to this analysis is from the second wave of 2004/2005. Bulgaria Hungary Georgia Months of survey Oct-Dec Nov-May March-July Year of survey 2004 2004-5 2006 Male 5851 6315 4405 Female 7007 7223 5595 Year of birth, oldest respondents 1925* 1926 1926 Year of birth, youngest respondents 1986* 1983 1988 Age of 1970-1974 cohort at interview 29-34 30-35 31-36 * a fewer older/younger respondents were included, but these were discounted as the year’s sample was not larger enough to carry out a meaningful analysis

Variables examined To calculate the precise age of the respondent at the birth of each child, then the year and month of birth of respondent and year and month of birth of all biological children (resident and non-resident) was required. For certain analyses then the respondents were grouped into 5-year cohort groups. The youngest cohort group analysed for total (and expected) fertility was the 1970-1974 cohort. The table above shows that this cohort was aged in their early thirties at the time of the first wave survey. As these countries still have early childbearing patterns (quite unlike Western Europe) then the majority of women respondents will have completed their reproductive careers at this age.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

16 _____________________________________________________________________ The total number of biological children for each respondent was calculated by summing those that were resident in the household plus non-resident children. Stepchildren, foster children and adopted children were not included. Future fertility expectations for the coming three years were estimated from adding a maximum of one to the current number of biological children if any of the following three variables was positive: the resultant variable was named ‘Expected kids’. The first variable was if the respondent (or partner) was pregnant. These can be expected to be achieved with a fair degree of certainty (though some may be lost because of miscarriage, stillbirth or abortion). If a respondent (or partner) were pregnant, then they were not asked about any further fertility intentions over the next three years. The second variable was if a positive response was given to the question “Do you yourself want to have a/another baby now?”. The third variable was in answer to the question “Do you intend to have a/another child during the next three years?”. If the respondent answered “Probably Yes” or “Definitely yes”, then one (extra) child was added to ‘Expected Kids’. Of course, some, perhaps many, will not succeed in having that planned child. And some will have not only one, but two in the next three years (especially possible for those who are currently pregnant). On the other hand, some respondents who have said they are not planning a child will end up having one. Therefore, these estimations can only be considered approximate wishes on a cohort-wide basis. It will be of great interest to see what proportion of those wishes and plans were realised when the second wave data is available – and for which ages and parities of respondents. There were two other questions that could have been used to predict ultimate fertility, but it was considered that the results may be less reliable than plans for the next three years, particularly considering that the youngest cohorts to be studied were in their early 30s. The questions that were not used were: “Supposing you do not have a/another child during the next three years, do you intend to have any (more) children at all?” and “How many (more) children in total do you intend to have?”. A number of variables were also selected as potential explanatory variables: these are used as illustrations in this report rather than being discussed in depth. They were selected as they are generally time-independent (although some are more subject to change than others). These include the following: whether the place residence is rural or urban; the religiosity of the respondent (frequency of attendance at religious services); educational level of respondent; number of siblings; and ethnicity. The age at first birth was also derived as an explanatory variable. Several variables were looked at but either appeared to have no significant effect or the subgroups were not large enough for this particular comparative analysis where the emphasis was on cohort-to-cohort variations: whether or not parents had ever split up, and attitude of trust of others and whether the respondent considered others are generally fair or not. Two of the three countries studied asked respondents about their attitudes to a current pregnancy, including opinion as to its timing: Bulgaria and Georgia included these questions, but Hungary did not. Answers from these questions, although not involving a large number of respondents were informative and are discussed in the relevant sections.

_____________________________________________________________________ Marion Burkimsher

17 _____________________________________________________________________

Comparison of GGP sample results with other published data Before using survey data to extrapolate to population-wide conclusions, it is important to examine whether the survey sample is representative of the population. A number of different demographic indicators can be calculated from the GGP data and these can be compared with whole-population data derived from vital statistics. The GGP’s contextual data base contains the available demographic indicators for each country and the publication by Frejka and Sardon (2004) includes others. The measures which were compared were the following: total fertility rate (TFR); period mean age at birth for all births; cohort mean age at first birth; completed cohort fertility (CCF); proportion of childless women by birth cohort; and the parity structure by cohort. Not all these measures were available for comparison for all countries; only TFR and CCF were available for all three (and the CCF for Georgia covered only a limited number of cohorts). Comparisons can also be made between data collected in the FFS and the GGP surveys. The question of weighting should also be discussed. The Hungarian and Georgian GGP surveys assigned weights to individual respondents to (theoretically) compensate for over- or under-sampling of types of individuals. However, in this analysis, the weights were not applied; only the raw unweighted distribution was analysed. No weights had been assigned to the Bulgarian data, as the GGP sample data was considered by the Bulgarian GGP team to fairly reflect the Bulgarian population distribution. For Hungary and Georgia, the analysis (to follow) shows that the sum of individuals sampled is representative of the population as a whole. TFR Band of years for which a valid TFR could be calculated In previous fertility surveys, one of the weaknesses was that only women (generally) of fertile ages were interviewed and therefore TFRs could only be calculated for a very limited range of years. This was attempted for the FFS in the evaluation of the project (Festy and Prioux, 2002). With the age span of respondents being much greater with the GGP, then valid TFRs can be calculated for a wider band of years. This band will now be discussed. For women, their fertile lifespan is considered to be from 15 to 50. However, the proportion of births (particularly in the Eastern European countries in question) taking place after the age of 40 is very small; in fact in these countries, there are only a few births to women over 35 (this is discussed in more depth in the following section). For men, of course, the potential age range is wider. For each year that we are trying to calculate the TFR, we need to have had respondents from as near the full range of fertile ages as possible. The table below shows what this band is for each country. With only interviewing people over the age of 18, then a few births in very recent years will not have been counted as the individuals will have been too young to have been included in the sample, yet will already have had a child. Therefore, TFRs calculated for the most recent 2-4 years will be slightly under-estimated.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

18 _____________________________________________________________________

Oldest cohort Youngest cohort Earliest valid TFR* Last year all births recorded

Bulgaria 1925 1986 1965 2004**

Hungary 1926 1983 1966 2004***

Georgia 1926 1988 1966 2005

* This is given as the year of birth of oldest cohort + 40 years. ** As interviewing was done in October-December 2004, then the majority of births for that year had taken place *** Most interviews were carried out in November-December 2004, though where people had removed then they were carried out in the following March-July.

TFRs calculated for before 1965-66 will also be under-estimated as the full range of fertile ages is truncated, as children born to people in their 40s, ie. cohorts born before 1925/6, are not included, and there could be a little under-estimation even up to 1975 (this is more probable for TFRs calculated from male data). The age band in which people have children varies quite markedly from country to country, a subject discussed in more depth in the following section. Women are more focussed in time than men, and in Bulgaria and Hungary the age span of childbearing is narrower than in Georgia. Comparison of TFRs derived from male data and from female data As far as I know, the TFR has never been calculated from male data before. The assumption could be that males may under-report number children or misreport their years of birth. There is also the problem, described above, of defining the fertile life of males: however, most men do in fact have their children before the age of 40 in Eastern European countries. However, this analysis shows that TFR values calculated from male data do not differ markedly from that derived from female data. For Georgia and Hungary there is no noticeable difference for any time period. For Bulgaria, there may be a slight tendency for the TFR derived from women’s data to be larger than that derived from men’s data for recent years. This could be as expected – women with small children are more often at home, while the men are out of the house; this can lead to women with younger children being over-represented.

_____________________________________________________________________ Marion Burkimsher

19 _____________________________________________________________________ Comparison of TFRs – Bulgaria

This shows good agreement between the TFRs derived from population-wide vital statistics and those calculated from the GGP sample for the years after 1989. Before that date the GGP-derived TFR is significantly lower. Why could this be? The following three potential explanations are proposed (although it is also possible, perhaps even likely, that they may all be partly true). 1. There were mis-estimates either of the number of births (over-estimated) before 1989 or the resident population (under-estimates of number of fertile women). Unless there is a good reason to suppose otherwise, it would generally be considered that the number of births in each year has been recorded accurately. But could the number of women aged between 15 and 50 have been underestimated for the years before 1989? One possible source of error is in estimating the number of Roma (or Turks) in the population. In the official census of 2001, 371,000 Roma were recorded - around 5% of the population - and this proportion is accurately reflected in the GGP sample. However, unofficial estimates suggest the total Roma population in Bulgaria may be as high as 700,000-800,000 (Marushiakova and Popov, 1995). As the Roma (and Turks) have completely different fertility characteristics to the native Bulgarian population (see later section of this report), then this could be a plausible explanation. 2. Another cause for the discrepancy could come from migration. The population remaining in Bulgaria to be sampled in 2004 may be different from the original population, from which vital statistics were derived. This could have happened because people with larger families have emigrated or died, leaving singles and small families behind. Emigration of ethnic Turks, who are the largest ethnic minority in Bulgaria (over 9% of the population in 2001), and who have significantly higher fertility than native Bulgarians, could be an explanation. Another possibility is immigration into Bulgaria in recent years of singles (workers) and people with small families. 3. The GGP sample is biased, giving over-representation to childless individuals and small families compared to large families. Although the opposite situation has been seen more often before in other countries, it is not inconceivable that people _____________________________________________________________________ New fertility behaviour in Eastern Europe

20 _____________________________________________________________________ less involved with immediate family needs may have more time and inclination to participate in a survey of this kind. Comparison with TFR derived from Bulgarian FFS data On completion of the FFS project an evaluation was carried out on the validity of the data collected in different countries and although a comparison with TFRs was only possible for a few years because of the limited range of ages of women sampled, some conclusions were drawn. This is what they reported: “In most cases, the FFS overestimates the fertility levels as calculated from vital statistics. Bulgaria and Lithuania are exceptions to this rule” (Festy and Prioux, 2002). The fact that the GGP and FFS both encountered a similar mismatch for Bulgaria between survey-derived TFRs and population TFRs is very interesting, especially as it is in the opposite direction to that encountered in most other countries. It would tend to give one suspicions that the vital statistics derived data may not be accurate. Unfortunately, however, this argument does not completely follow through when we look at the FFS comparison. The TFR was only able to be compared with FFS data for the years 1989-1996 , and it was only for the years 1989 to 1993 that the FFS-derived TFR was lower than the population TFR. However, what we have seen from the GGP-derived TFRs is that there is good comparability for those particular years, in fact all years after 1988. It is for the years before 1988 that the discrepancy is seen. Is there some characteristic of the Bulgarian population that would tend to favour over-sampling of certain groups over others, and which is different to that encountered in other countries? Have any inconsistencies ever been found in censuses, when the population predicted from balancing births, deaths and migration is compared to that found in the new census totals? These are questions that can only be answered, perhaps, by someone with deep knowledge of the specificities of that country, together with their data collection system for both population and vital statistics data, and survey sampling methods. Comparison of TFRs – Hungary

_____________________________________________________________________ Marion Burkimsher

21 _____________________________________________________________________ There is good agreement over most of the time period, although there may be a slight over-estimation for the period 1986-1995 from the GGP sample – perhaps because of over-sampling of families with children. Otherwise the trends are reflected very well. Comparison of TFRs – Georgia

This shows very good agreement, except for the period between 1988 and 1995, when the GGP data gives a higher estimate than the data derived from vital statistics. It is interesting that two different data sources for the national TFRs are given in the contextual data base: from the graph above, it would appear that the one from the Demographic Yearbook of Georgia, 2005 (source 2) matches the GGP data better than that from the publication “Recent demographic developments in Europe, 2005” (source 1). Period mean age at childbirth (all parities) The mean age at birth is influenced by three factors: changes in the age at which a woman gives birth to her first child, second child, etc.; changes in the spacing between children; and changes in the total number of children born. Higher parity births naturally take place at higher ages; therefore as higher parities become rarer, the mean age at childbirth may fall solely from this reason, even if the mean age at first birth (for instance) remains steady. The following sub-sections look at the period changes in the mean age of childbearing for births of all parities. The CDB only has data for Bulgaria and Georgia for this measure. Comparison of mean age at childbearing – Bulgaria There are marked differences in the age at childbirth as derived from the GGP data and from vital statistics, although the trends are similar: for both, the mean age starts to increase abruptly after 1992/1993 after a period of stability. An increase of around 4 years has occurred since then.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

22 _____________________________________________________________________

The reason for a lower mean age at childbearing for GGP respondents for years before 1994 could be that larger families were under-sampled, or they may have emigrated since the births occurred. If women with higher parity births were not included in the GGP sample, then the mean age at childbearing would be less. Why should the difference be the inverse for the most recent years? It could be because more educated women are more likely to participate in a survey such as the GGP – and they have a higher age at first (and subsequent) birth (Betts, 1996). Comparison of mean age at childbearing – Georgia

The match between the three data sets is good after around 1985. For the years before then, perhaps under-sampling of larger families would explain the fact that the GGP data gives a lower age at childbearing than both sources cited in the CDB. _____________________________________________________________________ Marion Burkimsher

23 _____________________________________________________________________ The fall in mean age at childbearing before 1980 could well be (partially) explained by a fall in higher parity births. The recent increase in age at childbearing has been less severe for Georgia than for Bulgaria or Hungary; however, it has still increased by around 18 months since 1993/1994. Cohort mean age at first birth This indicator was only available for Hungary in the contextual data base, and a comparison with the same measure derived from GGP individual-level data is shown below.

The difference between the two is in the opposite sense to that for period data of mean age at childbearing, and is particularly noticeable for older cohorts. As the influence of declining higher parities is not an influence on this measure, then it could be because of some over-sampling of more highly educated women. In contrast, for the younger cohorts, the slight mismatch could be because of the greater accessibility to interviewers of women with young children. Completed cohort fertility (CCF) This is an important measure and will be discussed in more depth in a later section. In particular, the parallels and differences in the TFR and CCF trends will be discussed, as this is a central theme to the research. The following graphs include cohorts up to those born in 1974, ie. women who were older than around 30 at the time of survey. The data for men’s completed fertility is included too, for interest, although comparable information from other published sources is not available. A short discussion about the similarities and differences of the curves for men and women is given at the end of this section.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

24 _____________________________________________________________________ Comparison of CCF - Bulgaria

As seen in the TFR measure, there is a marked difference between the CCF derived from vital statistics and that from GGP sample data. For cohorts born before 1960 the CCF from population-wide statistics gives estimates around 0.2-0.5 higher than the GGP CCF, with the difference widening the older the cohorts. Even for most recent cohorts, the GGP is still lower by around 0.1 child per woman. The overall trends seen in the two data sets are also different. The CCF derived from GGP data would suggest a small rise in fertility from the oldest cohorts, who have a completed fertility of under 1.5, up to those born between 1948 and 1960, when CCF was up to 1.8. For actual fertility of cohorts born after 1960, then there has been a fall, though if the fertility expectations of these currently fertile women are realised, then the fall will actually be quite small. In contrast, population-wide data suggests stability of cohort fertility for cohorts born between 1930 and 1958 at a level of just over 2 children per woman, after which there was a fall below replacement level. Further discussion of the mismatch is in a later subsection on a comparison with Frejka and Sardon’s decomposed parity analysis. Comparison of CCF – Hungary There is a much closer match between the CCF derived from population data and from GGP data for Hungary than for Bulgaria. Older cohorts show the same direction of mismatch, with the CCF derived from vital statistics being somewhat higher.

_____________________________________________________________________ Marion Burkimsher

25 _____________________________________________________________________

Both GGP sample and CDB show that cohort fertility was under 2 for cohorts between 1934 and 1956, with the trough being for the cohorts born at the end of the Second World War. The subsequent rise in cohort fertility, reaching a peak with the late 1950s - early 1960s cohorts, is more pronounced in the GGP sample than in the CDB trend. Since then cohort fertility has been falling, and even if fertility expectations are realised, then women who are currently in their early 30s are unlikely to have a completed fertility of more than 1.6-1.7. Comparison of CCF - Georgia

The number of years for which comparable data is available is rather small for Georgia (Frejka and Sardon did not include Georgia in their study of low-fertility countries). As with Hungary, and to a greater extent with Bulgaria, there does appear to be some mismatch between GGP data and population data. For cohorts 1953-1955, _____________________________________________________________________ New fertility behaviour in Eastern Europe

26 _____________________________________________________________________ the most recent years for which population data is available, the match is almost perfect; before then the CDB data shows a higher CCF. Why do older cohorts have too low a CCF in the GGP sample? As all three country comparisons show a similar mismatch between population data and GGP sample data for older cohorts, one might ask why this should be the case. One possibility is differential mortality. If women with more children have a higher mortality, then they would be unavailable for sampling in the GGP survey. Higher fertility is associated with some sub-populations such as the Roma, who do also experience higher mortality. Having more children is associated with rural residence and lower education (see later section for discussion on this): so were these women either unavailable because of death or incapacity or under-sampled because of their unwillingness to participate or inaccessibility to the interviewers? It is also possible that older women, who have lived through certain levels of oppression under the communist system, are more reticent about participating in a very personal survey such as the GGP – and perhaps especially the less educated ones. Differences between realised CCF and expected CCF In the three countries studied, early childbearing is the norm – this will be discussed in more depth in the next section. Therefore, the majority of women at age 30 have completed their fertility career. However, a significant proportion of women in the samples do intend to have a first or subsequent child in the 3 years after the survey. The differences between achieved fertility and expected fertility are illuminating. Bulgaria Hungary Georgia Actual fertility women aged ~35 1.53 1.64 1.76 Expected fertility women aged ~35 1.73 1.94 2.09 Difference actual and expected at ~35 0.20 0.30 0.33 Actual fertility women aged ~30 1.21 1.21 1.42 Expected fertility women aged ~30 1.59 1.73 1.93 Difference actual and expected at ~30 0.38 0.52 0.51 The values for the 3 year moving averages of CCF was used For Bulgaria the 1974 cohort of women were used to represent ~30 year old women and the 1969 cohort for ~35 year old women For Hungary the cohorts used were 1974 and 1969 For Georgia the cohorts used were 1976 and 1971

The first comment is that if Georgian women achieve their fertility plans, then CCF for women over 35 will exceed 2 and will almost reach that level for women aged 30. It should be remembered that in the calculations done for this study, it was only possible to move up to one higher parity in the ‘expected fertility’ calculations, so some women will ultimately exceed this, particularly in Georgia where later childbearing and higher parities are still more common. Therefore, it is quite possible that the CCF for the younger Georgian women could exceed 2.

_____________________________________________________________________ Marion Burkimsher

27 _____________________________________________________________________ Expectations for additional children are very similar for 30 and 35-year old women in Hungary and Georgia. Both sets of women anticipate adding 0.5 of a child to their current fertility if aged 30, and 0.3 of a child if aged 35. However, as Hungarian women are starting from lower actual fertility, their ultimate fertility is expected to be lower. How realistic these expectations were remains to be seen in the second wave of the GGP. Both the levels of actual fertility and future fertility expectations are lower in Bulgaria than in the other two countries. Even if fertility goals are achieved, it is unlikely that eventual cohort fertility will reach more than 1.6 to 1.7 for the cohorts of women born after 1969 – though it must be remembered that this is a similar level to that seen for all the cohorts surveyed in the Bulgarian GGP. Differences between male and female CCF As men have children at higher ages than women, then lower current fertility is naturally expected for the younger cohorts of men compared to women of comparable ages. The main interest is in comparing the CCF for men and women born before around 1960, and so around their mid-40s and older at the time of survey. For Bulgaria there appear to be no systematic differences between the CCF of men and women of the older cohorts. The same holds true for Hungary up to the cohorts born around 1950. Men born after then would appear to have a slightly lower CCF than women do. One wonders whether younger men with small families (or childless) have been somewhat over-sampled in Hungary, or if men with larger families have been under-sampled. The differences between men’s and women’s CCF in Georgia appears to be somewhat larger for some cohorts, although the differences do not appear to be systematic. It is also noticeable for Georgia that men’s fertility expectations for the next 3 years are somewhat higher than for women, unlike in the other countries. Do Georgian men still value children as a symbol of success? Will their expectations be met? Proportion of childless women by cohort This was the final indicator that was available for some countries in the contextual data base that could be compared with GGP sample data. It was not available for Bulgaria, but was available for Hungary and Georgia. Naturally the proportion of childless depends on the exact age when the respondent was observed. As seen in above, a significant proportion of women want to have a child, even if they are over 30 or 35. Therefore, whether the proportion of childless ultimately demonstrates an upward trend depends on whether the fertility ambitions of these women are realised. The agreement between CDB data for childlessness and the GGP sample for Hungary appears to be good, except for a potential over-sampling of childless women in the oldest age brackets (see graph below). Childlessness has traditionally been below around 10%, dipping even below that for the cohorts born in the late 1950s and early 1960s. _____________________________________________________________________ New fertility behaviour in Eastern Europe

28 _____________________________________________________________________

The GGP sample for Georgia (graph below) shows some quite erratic variations in the proportion of childless, even applying a 3-year moving average, which are not reflected in the CDB data. We can only assume these are a result of relatively small samples. The absolute level of childlessness is slightly higher than for Hungary, hovering somewhat above the 10% level. As with Hungary, there are some indications that the proportion of women remaining childless was falling for earlier cohorts, but lately this level has started to rise again.

Parity structure: comparison with Frejka and Sardon Frejka and Sardon (2002) analysed birth registration data and decomposed total cohort fertility by each parity (0, 1, 2, 3 and 4+). Comparing their data with the proportions of each parity derived from GGP data can give indications of the validity of the latter. _____________________________________________________________________ Marion Burkimsher

29 _____________________________________________________________________ It should be noted that the time period available for comparison is somewhat shorter than for the CCF measure discussed earlier, being in 5-year cohort groups from 1930/1934 to 1965/1969.

The mismatches between Frejka and Sardon’s data and the GGP derived data give some cause for concern. If the former are reflections of the true population, then it would appear that childless and 1-child women have been significantly over-sampled and women with 2, 3, 4 and more children have been under-sampled in the GGP. The differences hold across all the cohort groups analysed. Clearly this explains why the CCF derived from GGP data is too low. While there appear to be a systematic differences in level between the two data sets, at least the trends in each parallel each other quite well. Childlessness fell for earlier cohorts before reaching a low and stable level. The proportion of 1-child families has started to rise for more recent cohorts, while the proportion of 2-child families reached a peak before falling. The proportion of 3-child families has fallen slightly (from an already low level) for recent cohorts, while large families continue to be a rarity but at a steady level. One observation that may put into question Frejka and Sardon’s data is the extremely low level of childlessness they record, reaching a minimum of just 1.6% for the 1950/1954 cohort. Because some women (5-10%) always have physical limitations in their fertility, this level would seem to be unrealistically low. A level of 10% actual childlessness, as seen in the GGP data, would seem to be more likely to reflect reality, as there will always be some women who do not marry (even if only a few in the Eastern European countries) and some who cannot have children. Without more information we cannot judge which of the data sets is more accurate, but further investigation would be valuable.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

30 _____________________________________________________________________

There is considerably better agreement for Hungary than for Bulgaria, with no systematic differences in the data sets. The trends are also reflected in both (these will be discussed in more detail in a later section), although the peak in 2-child families would appear to happen for later cohorts according to Frejka and Sardon compared to the GGP sample. Comparison of completed cohort fertility from GGP and FFS The fertility outcomes have been sampled before in the FFS, and so it is also possible to compare GGP sample data with that from the FFS. As the FFS took place several years before the GGP, then the CCF of later cohorts will, of course, diverge, as the younger FFS women had not completed their reproductive life at the time of sampling. One of the main advantages of the GGP sampling a wide age band of respondents is that comparisons can be made with previous surveys.

_____________________________________________________________________ Marion Burkimsher

31 _____________________________________________________________________ The Bulgarian FFS was carried out in December 1997 and included only women aged between 18 and 45, ie. those born between 1952 and 1979 (UNECE, 2001). There is a small difference between the CCF of the oldest cohort sampled (1.7 for the FFS compared to 1.75 for the GGP) but this is considered good comparability. It is possible that the additional fertility was actually achieved, as the women in the FFS survey aged 45 and over could have had a few extra children in the years between the FFS and GGP surveys. What is interesting is that although the FFS and GGP survey data agree well, the published CCF rate for the whole population for the 1952 cohort (and for many cohorts of a similar generation) was 2.04 (see earlier sub-section), quite substantially higher than 1.75. Together with the doubts expressed about the childlessness rate derived from population by Frejka and Sardon, discussed above, this may indicate that there are weaknesses in the fertility data derived from vital statistics.

The Hungarian FFS survey was carried out between November 1992 and December 1993 and included women aged 18-41 (as well as a smaller sample of men) (UNECE, 1999). The agreement between the FFS and GGP data for the oldest cohort of women is exact – which is perhaps a little surprising, as a few additional children could have been expected to have been born to women aged only 41-42 at the time of the FFS sampling. However, it is reassuring to see such good agreement between the two data sets. There is also excellent agreement with the population data for these same cohorts. Conclusion To conclude, the data validation analysis of this section would appear to give confidence that the GGP sample data is representative of the populations of Hungary and Georgia. There are some concerns about the Bulgarian GGP sample, although trends in fertility would appear to be correctly portrayed. It is also possible that there are weaknesses in the population data of Bulgaria and so this could be the reason for the apparent non-comparability. _____________________________________________________________________ New fertility behaviour in Eastern Europe

32 _____________________________________________________________________

Changes in timing of childbearing Why is timing so important? Changes in timing of childbearing is important because this affects the standard measure of mean number of children per woman, the total fertility rate. In periods when childbearing is getting older – the current situation in many countries in Europe – then the TFR is automatically deflated. As a simple example, if the mean age of childbearing for all parities increases by one month over a calendar year, then one month’s worth of babies are effectively ‘missing’ in that year, and the TFR is one twelfth lower than it would be if there had been no delay. Changes in timing cause a mismatch in period measures of fertility and cohort measures. In the above example, where there is a delay, then cohort fertility – the number of children actually borne by a woman over her lifetime – will exceed period fertility measures. Therefore, if the TFR is to be correctly interpreted, then a study of changes in timing is critical. Influence of change in parities on mean age of childbearing As mentioned in the previous section, changes in the number of children being borne by women can influence the mean age of childbearing. If higher parities become rarer, but women continue to have their first and second child at the same age, then the overall mean age of childbearing will go down. For this reason, for any robust study of changes in timing, it is better to look at timing of a single parity, eg. age at 1st birth, 2nd birth, etc… Variations in shape of curve of age at first birth The distribution of age at first birth is most simply visualised as a graph plotting birth rate (total number of births divided by the population in question) against age. One example, for the GGP sample of the 1960-1964 cohorts in Bulgaria, is given below.

A number of points can be made about this graph. _____________________________________________________________________ Marion Burkimsher

33 _____________________________________________________________________ Firstly, although the area under the curves for men and for women are equal (for this example, only those men and women who had experienced a first birth were included in the calculation of the birth rate), the curve is wider for men than for women. This is a common pattern: men have their first child over a wider age span than do women, partly for physical reasons, but mostly for cultural ones. Secondly the curves are skewed to the left, ie. there is a steep rise to the peak (modal) age of first birth, followed by a longer tail to the right. This pattern is quite different from many Western European countries where age at first birth is much higher. For example, in Switzerland, the curve is almost symmetrical about the modal age of 30. The shape of the curve varies over time and from country to country. If there is a narrowing in the curve to a sharp peak, then there is convergence in behaviour to a well-defined most popular age at first birth; if it is getting wider, then behaviour patterns in age at first birth are becoming more flexible. The changing norms of society are reflected in the shape of these curves. Measures of central tendency – mode, median and mean At what age do women and men have their first child? As seen from the curve above, there is not a simple answer to this question. Three measures of central tendency can be used - the mode, median and mean - but each has its weaknesses, particularly when the distribution is skewed, as in our specific example. As mentioned above, in some countries where the age at first birth is around 30 then the curve is almost symmetric, and the mean, mode and median can have equal values. In that case, the peak birth rate at 30 is midway in the ‘normal’ age band of first births of 15-45. However, in the Eastern European countries, the curves are skewed to younger ages. Therefore, in those cases the modal age at first birth is less than the median, which in turn is less than the mean. A specific problem of calculating the mean age of first birth for cohorts is that the full band of possible childbearing years is required to obtain a fair estimate: for women this would be considered to be 15-49, while for men there is no upper age limit. Therefore, for the younger cohorts in the GGP sample, the mean is bound to be an under-estimate, as they have not completed their fertile life-span and so some will still have a first child in the future – which would push up the mean age. The mean age can be calculated as described above by simply averaging all the ages at which first births have occurred for that cohort, or a better estimate may be obtained by using a Kaplan-Meier survival analysis. The latter should give a fairer estimate in that all individuals are considered, even those who have not (yet) experienced a first birth. But then there is a problem of defining an upper limit to a possible age at first birth: is it defined rigidly as 49 for women and no limit for men, or should a de facto limit of 40 for women and 44 for men used? Because of this difficulty, the estimates of the mean used in this study are the simple means of children actually born – and it should be understood that for younger cohorts this estimate will be an under-estimate of age at first birth.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

34 _____________________________________________________________________ Because of this weakness of the mean, rather more emphasis has been placed on analysing the median age at first birth. That cannot change with behaviour in the future: once 50% of the population in question has had a first birth, then the median age at first birth is defined. The same holds true for the first quartile (when the first 25% have experienced a first birth) and the third quartile (when 75% have experienced a first birth). The problem with the third quartile (in particular) was that no value could be calculated if 75% of the population did not experience that event: this was the case for second births for Bulgaria and Hungary. The mode, or most common age for having a first child, is a crude measure that simply shows the age of peak birth rate. The following table shows the mode for men in Bulgaria. Clearly some cohorts have a bi-modal or ill-defined peak, which is not useful for indicating trends in childbearing ages. The mode is also subject to quite rapid changes, generally greater than the median or the mean which take the full population into account. 1927-1929 1930-1934 1935-1939 1940-1944 1945-1949 1950-1954 1955-1959 1960-1964 1965-1969 1970-1974 1975-1979

17 0 0 0 0 0 0 0 0 0 0 1

18 0 0 0 0 0 1 1 0 0 2 0

19 0 0 0 0 0 1 1 0 3 0 2

20 0 2 2 1 2 1 1 0 4 4 3

21 0 4 1 1 1 2 1 2 3 6 5

22 1 3 3 2 2 2 0 9 10 8 9

23 0 4 5 3 5 9 11 8 18 19 19

24 2 8 9 10 12 23 14 28 32 21 19

25 5 12 9 14 16 25 17 41 41 28 29

26 4 14 21 17 30 31 27 50 46 33 35

27 8 14 25 25 30 31 27 58 46 38 18

28 12 22 24 28 39 24 29 58 46 49 24

29 8 15 23 28 29 35 32 47 34 34 22

30 15 21 26 29 36 35 33 40 33 42 8

31 5 22 26 32 29 26 31 39 32 49 0

32 9 16 19 24 26 22 19 25 26 39 0

33 4 17 18 21 25 17 21 25 21 13 0

34 5 9 20 14 21 22 24 15 29 14 0

35 7 16 12 13 11 16 13 6 17 7 0

36 4 9 8 14 11 10 9 12 12 0 0

37 4 9 5 7 8 8 5 11 8 0 0

38 4 11 5 3 12 7 4 13 3 0 0

39 3 11 11 2 8 4 8 6 2 0 0

40 2 8 7 5 2 8 5 5 4 0 0

Frequency of births to men by age and cohort in Bulgaria showing a multi-modal pattern (mode highlighted in green)

Changing proportions of ‘young’ and ‘old’ first-time mothers As mentioned already, the distribution of age at first birth is very different in Eastern Europe from Western Europe. In many countries teenage mothers are considered to be a ‘problem’ and Britain in particular (which has the highest rate in Western Europe) is trying hard to reduce the number of teenage pregnancies (BBC, 1999). Young parenthood is seen as one factor leading to long-term social deprivation and to quote Tony Blair “Teenage pregnancy is not right, it is not in anyone's interests and does not work for mothers, fathers, children or society”. In Eastern Europe the pattern has been very different until very recently. The modal age for recent cohorts of women has been as low as 18 in Georgia and 20 in Bulgaria and Hungary. Young motherhood across the region is the norm rather than the exception, and the trend has been towards younger ages until recently. By contrast, becoming a first time mother at older ages is also not uncommon in Western Europe, even, perversely, in Britain, which has the highest proportion of women giving birth to their first child over the age of 40 (Sobotka, 2004). In the Eastern European countries having a first birth over the age of 30 is quite a rarity, generally involving less than 10% of women. The following graphs show the trends cohort-to-cohort in the proportion of ‘young’ (under 20 years of age) and ‘old’ (over 35 years of age) first time mothers for the countries under study.

_____________________________________________________________________ Marion Burkimsher

35 _____________________________________________________________________

_____________________________________________________________________ New fertility behaviour in Eastern Europe

36 _____________________________________________________________________ The problem in looking at ‘old’ childbearing is that only the older cohorts can be studied who have past that age in life: hence the above graphs only go up to the 19601964 cohort, and even that precludes possible future births to women who were only 40 at the time of the GGP survey. Therefore the bottom line, showing proportion of births to women over 35, may ultimately not be quite as low as it is on these graphs. The trends however, for the three countries, however, are quite clear. There was a rise in the proportion of women bearing their first child under age 20 from the oldest cohorts to the cohorts born in the early 1950s, after which the proportion stabilised at somewhat over 20% for Bulgaria and Hungary, and approaching 20% for Georgia. The proportion of women bearing their first child after 35 is very low – well under 5% and declining in Bulgaria and Georgia. It had reached over 5% for some of the oldest cohorts in Georgia, but there too the level has been falling for younger cohorts. Clearly if greater integration with Western Europe encourages Eastern European women to emulate their Western counterparts in fertility timing behaviour, then there is the potential for enormous changes. The following sub-section suggests that this may have started. Bulgaria – changes in timing of first birth The following graphs show the changing pattern in age at first birth for women and men in Bulgaria. The effect of the skewed pattern of birth rates with age is clear, with the modal age being less than the median, which in turn is less than the mean. Similarly, the number of years between the first quartile and median is less than between the median and third quartile. The mode has not been plotted on the graphs for men, as men have a broader range of behaviour, sometimes with a multi-modal pattern being seen for some cohorts (as seen on the table above).

The main change over time that can be seen from these graphs is that age at first birth became younger from the older cohorts up until those born in the early 1970s, when it suddenly started to rise again. The third quartile started to rise earlier and more _____________________________________________________________________ Marion Burkimsher

37 _____________________________________________________________________ abruptly than the first quartile, suggesting more postponement for those who would naturally choose to have their first child later in life – commonly the most educated women. There was clearly a convergence of behaviour from the oldest cohorts to those born between the 1940s and 1960s, with the breadth of behaviour, as seen in the interquartile range, becoming very narrow. Starting with the cohorts born in the latter part of the 1960s, the breadth of behaviour started to widen again. This equates to children being born in the turbulent 1990s. However, the breadth of behaviour is still not as wide as it was for the earliest cohorts surveyed, when the births occurred in the immediate post-war period.

The pattern for men parallels that for women, although median and mean age at first birth is higher and the age range between the median and third quartile is wider. The widening of behaviour patterns relating to babies born in the 1990s affects men born earlier than women, as they have a higher age at first birth; hence there is a widening of behaviour for the cohorts of men born after the 1960s. Hungary – changes in timing of first birth The pattern of timing of first birth in Hungary is broadly similar to that seen in Bulgaria. However, the cohort with the youngest first time mothers was that of 19551959, reflecting births around 1980. Since then the age at first birth has been getting older and the age band has been broadening – a pattern seen for men as well as for women. Why the sudden fall in the median and third quartile for the youngest cohorts of women? Probably the most likely explanation is a sampling effect (and also a small sample size for the youngest cohort band). Women in their early 20s who do not have children are difficult to access for interviewers as they are generally out at work or in

_____________________________________________________________________ New fertility behaviour in Eastern Europe

38 _____________________________________________________________________ education during the day and out socialising in the evening, while those who do have young children are at home, and thus are much more easily accessible.

The graph for men again parallels that for women, although the inter-quartile range is markedly less than in Bulgaria – even narrower than for women for the older cohorts.

Georgia – changes in timing of first birth The graphs below show that in Georgia women have a wider range of behaviour in timing of first birth compared to women in Bulgaria and Hungary. Although the median age of first birth declined somewhat from the older cohorts to those born in the early 1970s, there was not the same level of convergence in the age band of giving birth to a first child as was seen in the other two countries. However, the most recent _____________________________________________________________________ Marion Burkimsher

39 _____________________________________________________________________ cohorts do similarly show a rapid broadening and ageing in age at first birth. This again parallels the occurrence of first births taking place in the turbulent 1990s.

The graph for men in Georgia shows a narrower inter-quartile range for men than for women for the earlier cohorts, though once again a broadening of behaviour towards the younger cohorts.

Inter-quartile range of age at first birth The following two graphs show the evolution in the inter-quartile range in age at first birth for women and for men in the selected countries. Looking first at women, Georgia has traditionally had a wider spread of behaviour than Bulgaria and Hungary, though the latter are catching up with the youngest _____________________________________________________________________ New fertility behaviour in Eastern Europe

40 _____________________________________________________________________ cohorts. There was a long period of convergence in timing for women of all three countries, some three decades, with a minimum inter-quartile range being achieved of just over 5 years in Bulgaria and Hungary. The most recent cohorts have seen a divergence of behaviour again, particular abrupt in Hungary. This can be explained by the choice of many women in these countries to delay their first child during the uncertain period of the 1990s. This narrow age band of having a first child is in marked contrast to the inter-quartile age range in some Western European countries and the United States. From Sobotka’s analysis (2004), the inter-quartile range in England and Wales, Ireland and the United States was around 9 – 9.5 years in 1998/2000 (as a period measure rather than a cohort measure). It is, however, worth remarking that the inter-quartile range was widening for all 18 countries studied by Sobotka – perhaps a general sign of loosening norms in childbearing timing?

For men, the changes in behaviour have been somewhat different. The maximum convergence of timing of first births in Bulgaria and Georgia was with the 1930-1934 cohort, with a steady broadening of behaviour with later cohorts. In contrast, men in Hungary reached maximum convergence with the 1940-1944 cohort with an interquartile range of just over 5 years, similar to that seen in later cohorts of women in that country. All three countries now have an inter-quartile range of first birth for men of between 10 and 11 years, an increase of 4 years or more from their maximum convergence. It is interesting that the men of Bulgaria and Georgia have followed closely similar paths, while the women in those same countries have had quite different behaviour patterns and trends. The women of Bulgaria and Hungary are roughly similar, while Georgian women are the outliers.

_____________________________________________________________________ Marion Burkimsher

41 _____________________________________________________________________

Difference between median age at first and at second births The graphs for mean, median and modal age of second births for men and women in each country are not included here. However, an examination of timing of second births is also valuable and informative. In this section we will look at the difference in median age at first birth and second birth, in other words the spacing between the first two children.

This graph shows the very different spacing of children in Georgia and the other two countries. In Georgia the spacing between first and second child has hovered around the 3 year mark, while in Bulgaria and Hungary the gap has generally been at least 5 years. As with comparisons of age at first birth and inter-quartile range, this spacing is very different from countries in Western Europe. In Switzerland, for instance, the _____________________________________________________________________ New fertility behaviour in Eastern Europe

42 _____________________________________________________________________ difference between mean age at first birth and at second birth has declined from around 3 years to 2 years in the last 35 years. It is interesting that Bulgaria and Hungary followed an almost identical trend until the most recent cohorts, when they have diverged widely. However, all three countries saw some increase in the spacing with the 1970-1974 cohort, followed by a fall for the later cohort. One suggestion of why they may have gone up (Bulgaria’s dramatically), and then plummeted is the selection effect: women born in 1970-1974 were having their children in the 1990s and postponement of a second child was increasingly common. More than a quarter of mothers with one child decided not to progress on to a second child (and so a third quartile of age at second birth could not be calculated). The sudden dip for the most recent cohort analysed could be because women who want a second child more commonly decide to do so quickly (to reduce their time out of the workforce perhaps) or else they now decide against having a second child at all. It is worth noting the increase in spacing in the earlier cohorts – to 7 years between first and second child for the cohort born 1935-1939. These women would have had their first child around 1960 and then delayed their second child for a long time. It would be interesting to know what was the cause of this and why Bulgarian and Hungarian women followed such similar paths. The graph for men below shows less variations than for women (although the younger cohorts are not included as they had not had time to have their second child at the time of the survey). Again the gap between first and second child is less in Georgia than the other two countries, though the difference between Georgia and the other countries is lower than for women. In Georgia and Hungary, the spacing for men has been declining overall – Hungary more than Georgia, while Bulgaria has hovered around 5 years for most of the period.

_____________________________________________________________________ Marion Burkimsher

43 _____________________________________________________________________ Difference between median age at first birth for men and for women The difference in age at first birth between men and women depends, naturally, on the average age difference of the partners, with the man normally being the older partner. The following graph shows very similar patterns and trends between the three countries. It would appear there has been an increase in the age difference between the partners, from around 3 years to around 6 years for the youngest cohorts. The 19701974 cohort shows a steep increase from the next older cohort group: is this because young men are becoming more averse in entering into partnerships, and particularly into parenthood?

Effect of changes in age at childbearing on fertility rates As mentioned before, delay in childbearing depresses the TFR, although it does not (necessarily) lower the cohort fertility rate. This section examines exactly how much the TFR may have been depressed and for how long this could continue. Bongaarts and Feeney (2005), following on from earlier work by Ryder, have devised a straightforward correction to approximate the TFR to the CCF for delay in childbearing. Thus the corrected TFR – the value expected if there had been no change in age at childbearing – is the observed TFR divided by (1-rp) where rp denotes the rate of change in the period mean age at childbearing in year t. This value is termed the tempo-adjusted TFR. However, because changes in the proportions of each parity can also influence changes in age at childbearing, the correction should be made to the fertility rates decomposed for each parity. A combined corrected TFR can then be estimated by summing the corrected fertility rates for parity 1, parity 2, parity 3, etc. The tempo-adjusted TFR is not necessarily a perfect estimate of CCF because the shape of the curve – convergence or divergence in timing – can change over time. _____________________________________________________________________ New fertility behaviour in Eastern Europe

44 _____________________________________________________________________ What the GGP data gives us is cohort fertility measures, although it is also possible to use the data to estimate period TFRs (as done in an earlier section). The previous subsections have looked at changes in timing from a cohort perspective: this section looks at the changes in timing (of first births) from a period perspective. As already mentioned, delaying first and subsequent births only really started in the 1990s, a result of the social upheavals in the countries in question. The following graph shows this quite clearly.

The delay started earliest in Bulgaria, but it has been most severe in Hungary, where the period mean age at first birth rose over 3 years in the space of just 7 years between 1995 and 2002. Georgia has seen a less dramatic rise, though it still rose by over 2 years between 1994 and 2000. These unprecedented delays have had a major influence in depressing the TFR. If we assume that all parities experienced the same delay, then the following table shows the effect on the TFR.

Bulgaria Hungary Georgia

Year of start of delay

MAFB start of delay

Latest year of data

1991 1991 1994

21.8 22.7 22.4

2003 2002 2003

Delay in MAFB Number of Delay per TFR in 1989 1st births latest year years year from CDB (years) 25.3 26.2 24.0

12 11 9

3.5 3.5 1.6

0.29 0.32 0.18

1.9 1.78 2.231

Annual Number of depression years of of TFR at delay until 1989 level 30 0.79 15.96 0.84 11.75 0.49 33.19

Note: Georgia TFR for 1989 is that from CDB source 2

This shows clearly that the ‘lowest-low’ fertility experienced by Bulgaria and Hungary at this time is primarily caused by the TFR having being depressed by the order of 0.8 child/woman because of the timing effect. Completed cohort fertility of women who have passed their fertile lives during this period can be expected to be that much higher. Could delay in childbearing continue? It would seem to be very possible, even likely: greater integration with Western European norms would encourage this; higher education and longer education of women would encourage this; ongoing social and economic uncertainty may well encourage women to continue postponing a family. At _____________________________________________________________________ Marion Burkimsher

45 _____________________________________________________________________ the start of the period, age at first birth was very low in these countries by European standards, so there is ample opportunity for raising the age at childbearing. The final column in the table above considers the possibility that the delay could continue until the mean age for women having their first child is 30 – the age currently seen in Switzerland. At the current rate of delay, then this could be achieved in 12-33 years, depending on the rate of change. That will mean TFRs being depressed by these considerable factors for the foreseeable future. By comparison, Switzerland has seen a steady, albeit slower, rise in mean age at first birth over the last 35 years at a rate of 0.12 year of delay per year and this has depressed its TFR by the order of 0.2 over this whole period. If these Eastern European countries continued their progression in delaying childbearing at that rate, they would continue to have depressed TFRs for the next half century! This scenario has also been predicted by Sobotka (2004). Eventually, of course, the delay in timing must stop: it will either hit the physical constraints of high-age fertility, or (perhaps more likely) social pressures and norms will change again and either stability in age of childbearing will ensue, or there will once again be advancement of timing, as was seen in the Baby Boom years of Western Europe, or for several recent periods in the countries of Eastern Europe. When timing of childbearing does reach a platform, then the TFR will automatically recover and will then approximate the long-term CCF. This has already been seen in the Netherlands and the United States. However, for Bulgaria and Hungary (and perhaps Georgia if it follows the Western European pattern), this could be a long time in the future. Although this section has focussed on the effect of delayed timing, advancement of timing of childbirth has the opposite effect on TFR, inflating it above long-term CCF. From the graph above, it can be seen that there was advancement during the periods 1966-1978 and 1986-1994 in Georgia and from 1975-1979 in Bulgaria. The resultant ‘bulge’ in the TFR can be seen for Georgia (see earlier section) although this is less obvious for Bulgaria. Association of age at first birth and total number of children Changes in the timing of childbearing automatically affect the TFR, but do not necessarily affect completed cohort fertility. We will now look at whether a delay in childbearing could cause a fall in cohort fertility – which would, then, have a multiplicative effect in causing the TFR to fall. Potential fertility (fecundity) starts to fall significantly after a woman reaches around 35 years of age. In some of the Western European countries, this is a major concern, where postponement is causing an increasing number of women to face involuntary infertility because they have ‘left it too late’. In the Eastern European countries studied here, where childbearing has been very young, then these physical limits are not being approached yet. However, social acceptance of older motherhood is, perhaps, a different matter. As discussed in an earlier section, the norms for age at first birth have been very tight until very recently.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

46 _____________________________________________________________________ We will now look at whether the age of bearing a first child has an influence on the total number of children born. Because childbearing has been so young (to western eyes), then the dichotomy between a ‘young’ first-time mother (under age 23) and ‘older’ mother (first birth at age 23 or above) may seem surprising. The example of Georgia is shown, although the patterns seen in Bulgaria and Hungary are similar. The first graph shows how the proportion of ‘young’ mothers has increased from the older to younger cohorts.

The following graph looks just at women with a completed fertility of just one child, and to see what proportion of these women had their first child when they were ‘young’ or ‘older’. It is quite clear that an ultimate outcome of one child is much more likely if a woman has a first child at higher age than if she had a first child when very young. Roughly a quarter of women with an ultimate fertility of just one had that child when they were over 23, while only 5-10% of one-child mothers had that child when under 23: most had gone on to have further children. On the whole the older first-time mothers tend to be better educated women, who have stayed in education longer or who want to work in the labour force before starting a family. It is interesting to note that many of the ‘older’ first time mothers do plan to have another child, and hence move into the 2-child group in the next 3 years – hence the big dip of the dotted blue line. The ‘young’ mothers show less of a desire to add another child in the coming years.

_____________________________________________________________________ Marion Burkimsher

47 _____________________________________________________________________ The following graph plots the proportion of mothers with 4 or more children by age of first birth. Not surprisingly, the earlier a woman started her child-bearing career, the more likely she is to end up with a large family, although the difference is widest for the oldest cohorts (perhaps an influence of modern contraception meaning that a long period with a spouse during the fertile years would naturally lead to a larger family for the older generation, but not necessarily so for younger generations). Similar patterns can be seen for 3-child families in all the countries studied. It is interesting to note that older mothers in Georgia are still considering moving into the 4-child bracket (up from 3), and if they succeed, they will equal the proportion of young first-time mothers.

To conclude, there is ample evidence that starting a family early usually leads to having more children, and conversely starting later tends to limit family size. This could be an example of Hakim’s preference theory (2003) of a differentiation between ‘family-oriented women’, who enter motherhood early and non-family oriented women, who postpone or avoid motherhood altogether. It could also simply be a reflection of chance: if a woman starts a family early, she will have many more fertile years in which to have a second or subsequent children than if she starts late. In conclusion, it seems entirely possible that the postponement of childbearing seen recently in these Eastern European countries may not only depress the period measure of fertility, the TFR, but also reduce the completed cohort fertility. However, that may depend on whether they deem childbearing at certain ages ‘too late’ or not, and whether the desires of older women to have a subsequent child will be thwarted by age-induced infertility. At what ages do women consider that a child is ‘early’ or ‘late’? The GGP questionnaire included a question to pregnant women (and men, though they are not considered in this section) as to whether, when they got pregnant they had wanted a child ‘at some time’. If they answered ‘yes’ to that, then there was a followup question on whether they considered the pregnancy to be ‘sooner than desired’, ‘later than desired’ or ‘about the right time’. These questions were not included in the Hungarian questionnaire: in Bulgaria it was asked to 62 women; in Georgia to 50 women. _____________________________________________________________________ New fertility behaviour in Eastern Europe

48 _____________________________________________________________________ The following tables summarise the frequencies of each response, depending on age of the woman and her current number of children. Bulgaria Timing of pregnancy

Cohort

Age at survey

sooner (than R wanted)

1970-1974 1975-1979 1980-1984 1985-1989

30-34 25-29 20-24 18-19

later (than R wanted)

1965-1969 1970-1974 1975-1979 1980-1984

35-39 30-34 25-29 20-24

0 2 1 2

1 1 0 0

about the right time

1965-1969 1970-1974 1975-1979 1980-1984 1985-1989

35-39 30-34 25-29 20-24 18-19

1 4 14 11 3

3 10 11 3 0

Georgia Timing of pregnancy

Age at survey

Total number of biological children already born 0 1 2 3 1 0 0 0 0 6 1 0

0 0 0 1 0

5 0 1 0 0

0 1 0 0 0

Total number of biological child already born 0 1 2 3 0 1 1 0 4 0 1 0 0

Sooner than planned

1975-1979 1980-1984 1985-1988

26-30 21-25 18-20

Later than planned

1970-1974 1975-1979 1985-1988

31-35 26-30 21-25

3 6 3

0 1 1

2 0 0

About right time

1965-1969 1970-1974 1975-1979 1980-1984 1985-1988

36-40 31-35 26-30 21-25 18-20

0 1 7 8 6

3 1 8 11 1

0 1 2 0 0

0 1 0 0 0

Most women considered their pregnancy to be ‘about the right time’. Very few considered their first child to be sooner than planned; in each country just one in the 18-19 year old age bracket, which in western eyes would be considered ‘early’ (there were 6 ‘about the right time’ in Georgia and 3 in Bulgaria to 18-19 year old girls). This confirms that youthful motherhood is till very acceptable in these countries. It is perhaps surprising that so many young women under the age of 30 consider their first child to be ‘later than planned’ – 3 in Bulgaria, 9 in Georgia. Second children, however, seem to be more often considered ‘sooner than planned’ – 6 in Bulgaria, 5 in Georgia. Is this because there are quite wide spacing norms in these countries, especially Bulgaria? Of the women who were pregnant and over the age of 30, 19 in Bulgaria considered the pregnancy to be ‘about the right time’, while 7 in Georgia did: in Bulgaria 4 said their pregnancy was ‘later than planned’, with 5 in Georgia doing so. Not surprisingly, in the two countries combined, only one woman over the age of 30 thought her pregnancy was ‘earlier than planned’.

_____________________________________________________________________ Marion Burkimsher

49 _____________________________________________________________________ This study, while lacking a large number of cases, would tend to confirm that in Bulgaria and Georgia, the attitude is still supportive of having a first child while young, while a second child should be delayed a while. Although bearing a child over the age of 30 is acceptable, for many it would be considered later than desirable. Causes in delay in childbearing The turmoils of the 1990s have been pinpointed as the main reason for the sudden increase in age at childbearing in the three countries under study. However, other forces may also have had an influence and may become the main engines of future changes. Integration with Western Europe, with Hungary having joined the EU in 2004 and Bulgaria in 2007, is likely to open up those countries to Western European norms of timing in childbearing as migration of people and ideas takes place. The desire of young people to emulate and follow the trends of their richer western neighbours could play a part in ideational change and attitudes to parenthood. The concentration of behaviour patterns in timing of fertility which had led to rather rigid norms may well be breaking down, as the spirit of free enterprise takes over from the constraints of communist ideals. In the current climate of uncertainty – global, national and individual – then postponement of decisions which have long-term consequences is the safest option. Childbearing is a long-term commitment of resources, and with the increased availability of effective contraception, then postponement is an attractive option. The domains of uncertainty particularly affecting young adults are twofold: employment difficulties – finding and keeping a stable job; and relationship challenges – finding a partner and having the ability to maintain the relationship/marriage over the long term. In many other developed countries, young people are staying in education for ever longer periods, and this trend could well be followed in Eastern European countries. As women reach higher educational levels, then their fertility falls – an effect seen in many other countries and discussed further in the following section of this report. Summary of implications of changes in timing of childbearing Women (and men) are having children at higher ages than they did in the 1980s. As described in detail earlier, this is having a deflationary effect on the TFR measure of fertility. The classification of Bulgaria and Hungary in the ‘lowest-low’ fertility bracket is because of this deflationary effect. Cohort fertility – the number of children actually borne by women over their lifetime – is not falling as fast as the TFR would imply. So should the TFR be ignored altogether? No, because it also defines how many new babies are coming in to the bottom of the population pyramid. It also indicates how the ‘popularity’ of childbearing waxes and wanes over time, and this can reflect whether or not the overall conditions at a particular time are seen as conducive or not to childbearing. But only some of these conditions are amenable to adjustment by policymakers, eg. the economic state of the country and how it affects unemployment. _____________________________________________________________________ New fertility behaviour in Eastern Europe

50 _____________________________________________________________________ It could be more helpful to use an adjusted measure of TFR to account for the effects of changes of timing in childbearing as proposed by Bongaarts and Feeney. However, solid statistics are needed for that: true biological parity for children being born (not just parity in the current marriage, with births outside marriage being unclassified); and accurate estimates of the population of each cohort of women. An adjusted TFR will then give a better impression of whether fertility really is falling or whether it is simply births that are being postponed. As age at first birth could continue rising for a considerable period – it may be decades, though it could be much less – then a better period measure of fertility than the TFR would be helpful. Postponement of births is not, of itself, a cause for concern and may be of benefit to society. Older mothers (and fathers) are more likely to be more highly educated and have a more secure financial status. They are more independent of needing help from their families or the state. Partners who are more mature have a better chance of a stable long-term relationship, which benefits their children and society. The normal age of childbearing can become a lot older in the Eastern European countries before hitting the potential health problems and age-related infertility that some western countries may be facing. Widening of behaviour patterns is happening and this is likely to continue: this should be embraced, not feared. Georgia has had a wider band of age at first birth than Bulgaria and Georgia, though a shorter gap between first and second births, and its fertility rate is higher. In the past, it would seem that the societal constraints on ‘normal’ timing of childbearing were quite tight. Loosening of these constraints may help more women to consider first-time motherhood or have a further child, even if they have passed the traditional time period. Perhaps the leaders of these Eastern European countries could daringly follow in the footsteps of Britain’s leaders? The wife of the country’s prime minister, Tony Blair, had a fourth child at the age of 45, and the wife of Gordon Brown had her third child when she was 42 whilst he was Chancellor of the Exchequer. The positive impact of increasing fertility levels in women over 30 has been highlighted in the media in Britain: a recent BBC report states “The UK's population is growing as more women have babies in their 30s” (BBC, 2007b). It would seem that in Bulgaria and Hungary, there has generally been quite a long gap between the first and second child, of the order of 5 years. This need bring no disadvantages, except that it would seem that postponement of a second child is turning more and more into abandonment of the idea. This will be discussed in more detail in the next section.

_____________________________________________________________________ Marion Burkimsher

51 _____________________________________________________________________

Changes in parities Parity distribution analysis This section of the report looks at the proportions of women by parity, how these proportions have changed from one cohort to another, and the country-to-country variations in parity distributions. In the initial analysis men’s parities were studied too, but as they essentially paralleled those of women, then no useful additional information was gained by studying men. The advantage of focussing on women is they have a fairly well-defined fertile period, whereas for men it is never certain what their completed fertility will be. As described earlier, predicted fertility could be deduced by including those children planned in the next 3 years. The following graphs show a solid line for current actual fertility, with dotted lines showing planned future fertility. How closely the predicted fertility matches the subsequent outcome is one of the results eagerly anticipated with the second round of the GGP. It may be presumed that the final outcome will be somewhere between the current and predicted lines, but which it will be nearer remains to be seen. The following graphs look at the distribution of parities by woman and also by mother, a mother being a woman who has had at least one child. Therefore women who have remained childless are not included in the calculations of parities of mothers. The reason for this is that the causes of childlessness are rather different from the reasons determining exactly how many children a woman has. In Eastern European countries childlessness is almost always because a woman has not married (or entered a long-term partnership) or because she has impaired fecundity; very few couples voluntarily choose to remain childless. The reasons are broader in Western Europe and childlessness levels are higher there. Trends of changes in parities – Bulgaria

_____________________________________________________________________ New fertility behaviour in Eastern Europe

52 _____________________________________________________________________ This graph shows that the 2-child norm has been dominant over the whole period. After a rise to a peak with the cohorts born between 1940 and 1964, its dominance has just started to fall. Bulgaria is noted for having the highest predominance of 2-child families of any country in the world. The most marked recent trend has been the rise in 1-child families, starting with the 1960-1964 cohort. The level of childlessness fell from an initial level of 20% to a current level of 10%, which may fall even lower if the currently fertile cohorts fulfil their reproductive plans in the coming years.

If we look at the distribution of parities of just mothers, then the change in balance of 1- and 2-child families becomes even clearer. The changes in the proportion of women having 3 children have been slight, although recent cohorts have seen a small fall (from an already low level). Over the whole period studied, larger families of 3 or more children have never exceeded a level of 15%. Moving up from a second to a third child has clearly been a large barrier which has not declined over time, but reinforced. The possibility that 1-child families could overtake 2-child families in popularity in the cohorts born after 1974 is not beyond the bounds of possibility, if current trends continue. Trends of changes in parities – Hungary The pattern for Hungary is not dissimilar to that of Bulgaria, although larger families of 3 and more children have been more popular in Hungary. Once again the 2-child norm has been strongly predominant, with over 50% of women having this family size until recently. Once again there has been a recent sudden fall of the 2-child model, with 1-child families moving up sharply in popularity. However, this started 5 years later than in Bulgaria, beginning only with the 1965-1969 cohort. As in Bulgaria, childlessness levels fell cohort-to-cohort to reach under 10% in recent cohorts, and will continue to fall to around 5% if the women surveyed achieve their fertility plans. The difference between current childlessness and expected levels is very wide, so it will be particularly interesting to see what the outcome is in the next wave of the GGP. _____________________________________________________________________ Marion Burkimsher

53 _____________________________________________________________________

One interesting trend to note is the increase in 3-child and higher parities for women born between 1940 and 1964. At the same time there was a marked fall in 1-child families and childless women. It would appear that family sizes were growing during this period, presumably encouraged by social policies or economic growth. Since reaching a (recent) maximum, families of 4 and more children have again been losing popularity since the 1960-1964 cohort of women, but 3-child families have lost less ground. It is interesting to see that a reasonable proportion of women in Hungary are planning to have a 3rd child (especially) or, to a lesser extent, a 4th child in the coming years: in Bulgaria it was minimal. _____________________________________________________________________ New fertility behaviour in Eastern Europe

54 _____________________________________________________________________ Trends of changes in parities – Georgia The trends of parities are quite different for Georgia than for Bulgaria or Hungary. For the earliest cohorts, there were approximately equal proportions (in the 15-25% bracket) for each parity (0, 1, 2, 3 and 4+). During the latter half of the 20th century, Georgia was essentially just completing its first demographic transition, with fertility rates falling from a reasonably high level by European standards. As seen in Bulgaria and Hungary, childlessness started at a level close to 20% and fell to around 10% with the 1955-1959 cohort. This level will continue to fall for more recent cohorts if the women surveyed fulfil their short-term fertility ambitions.

If we look just at mothers, then has been essentially stability in the proportion of women who have a completed fertility of 1 child or 3 children. Between 15 and 20% _____________________________________________________________________ Marion Burkimsher

55 _____________________________________________________________________ of mothers have had one child, while around 25% have had three. The big changes have happened in the proportions who have had large families of 4 or more children, with a steady fall over the whole period, and a rise in 2-child families, which have risen in popularity, reaching around the 50% level in recent cohorts. Country-to-country comparisons There will now be a brief overview of the differences in the trends of each parity from country to country. In each of the following graphs the Y-axis (proportion of women with each parity) covers a band of 40%, although the base level varies from one graph to another. This allows the relative changes to be more easily compared. Childlessness

The level and trends of childlessness in the three countries of Eastern Europe studied are remarkably similar. There was a steady fall from the earliest cohorts, and this fall will continue to a level of just 5% if childbearing plans are fulfilled in the younger cohorts surveyed. Whether this is achieved remains to be seen. It is clear that childlessness is shunned by women in these countries as an undesirable outcome. In the earlier cohorts, childlessness was probably associated with health problems and these have largely been overcome. It should be remembered that the dotted lines represent plans for the short period of 3 years following the survey, so they could be considered to be reasonably realistic. The large rise in current childlessness in Hungary for the 1970-1974 cohort is likely to be due to the rise in age at first birth – women are leaving it until later to start a family, and that country currently has later childbearing than in the other two countries. The 1970-1974 cohort was only 30-35 at the time of survey, which is not many years beyond the mean age at first birth (see previous section). A later subsection will discuss fertility expectations of childless women in more depth.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

56 _____________________________________________________________________ 1-child families Unlike the proportion of childless women, the proportion with just one child shows much greater country-to-country variation, as well as changes over time. Georgia has the lowest proportion of 1-child families, and the level has remained roughly stable at between 15 and 20% of mothers across the cohorts in question. Bulgaria and Hungary started with a reasonably high level – a third of women approximately, and this proportion then fell. Bulgaria saw the proportion of 1-child families start to increase again first, a trend followed recently – and strongly – by Hungary. It is likely that, of the 1970-1974 cohort, the proportion of mothers with one child will (once again) be around a third.

2-child families As mentioned earlier, Bulgaria was the country with the highest proportion of 2-child families, and this is confirmed in the graph above. Hungary, however, was not far behind in its establishment of a 2-child norm. For most of the cohorts studied, over half of the women who had children in those two countries ended up with two. Georgia is now following suit, though it is lagging behind the other two countries by several decades. However, the 2-child norm is finally starting to be eroded in Bulgaria and Hungary, and it is the 1-child family that is on the ascendance.

_____________________________________________________________________ Marion Burkimsher

57 _____________________________________________________________________

3-child families If we look at the proportion of mothers who have had 3 children, there has been relative stability across the cohorts, although the absolute level for the three countries is quite different. In Georgia, around a quarter of mothers have three children, while in Bulgaria it has been under 10% over the full period studied. Concerning fertility plans – shown by the difference between the solid lines and the dashed lines - clearly many more women are planning to move on to a third child in Georgia than in the Hungary and Bulgaria, with very few women in Bulgaria contemplating a third child.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

58 _____________________________________________________________________ The other interesting trend, already mentioned, is the rise in 3-child families in Hungary for the cohorts 1940-1964, equivalent to births which happened in the 1970s and (even more) in the 1980s. The social and economic shocks of the 1990s seem to have curtailed the growth in 3-child families for all the countries concerned. Families with 4 and more children

Finally, we look at large families, those of 4 children and more. Georgia shows an initially much higher proportion of large families compared to Hungary and Bulgaria, but there has been a sustained fall to reach levels comparable with the other two countries. Once again there is the interesting growth in larger families in Hungary for the cohorts of women born 1940-1964, reflecting births in the 1970s and 1980s. The number of women planning to have a 4th child is low in Hungary, a little higher in Georgia and minimal in Bulgaria. The proportion of mothers of the 1970-1974 cohort with a final parity of four or over is likely to be between 2 and 6% in all three countries studied. Is the data on proportion of parities reliable? This was discussed in an earlier section, but it is important to point out where there may be weaknesses. For Georgia, the comparison of various indices derived from whole population data and GGP sample data showed a good match, which gives confidence that the GGP sample is representative. For Hungary and Bulgaria a detailed comparison of distribution of parities was possible with Frejka and Sardon’s population-derived data. For Hungary, there is good agreement in the proportions of each parity, especially for more recent cohorts – for the earliest ones, there is some suggestion that women with smaller families could have been over-represented in the GGP sample. The important trends, however, are reflected in both Frejka and Sardon’s data and the GGP data: that there was a rise in 3-child families between the 1940-1944 cohort up to the 1960-1964 cohort and a parallel fall in the proportion of 1-child families. The change in proportion of families _____________________________________________________________________ Marion Burkimsher

59 _____________________________________________________________________ of 4+ children is smaller in Frejka and Sardon’s data than in the GGP data (the latter show a larger rise up to the 1960-1964 cohort), but the differences are not large. For Bulgaria, the proportions of each parity show less agreement between the statistics derived from the whole population and those derived from the GGP sample. In the GGP sample, there is a higher level of childlessness and 1-child families, and a lower proportion of 2, 3 and 4+ child families. In Frejka and Sardon’s analysis of Bulgaria, childless levels are very low indeed, with levels significantly lower than observed in Hungary and Georgia (sowing seeds of doubt as to the validity of their analysis). The proportion of 1-child families in Bulgaria would be lower if Frejka and Sardon’s figures are true, and the trend was stability for earlier cohorts, but a rise in the proportion among more recent cohorts. The trend for 2-child families is the same – a rise to a peak followed by a fall since the 1960-1964 cohort. The trends for 3-child and larger families are the same for the sample data and population data, but the levels of each would be closer to those seen in Hungary if Frejka and Sardon’s data are correct. However, in Bulgaria there was no rise in the proportion of 3- and 4-child families for the cohorts up to 1960-1964. To conclude, we can have confidence that the GGP data correctly reflect the trends of changes in parities over the last half century. There is some doubt as to the actual levels of each parity in Bulgaria, and further confirmation as to the validity of the GGP and/or the whole-population data would be valuable. Measures of concentration of reproduction The previous graphs have shown that there had been an increasing focus in earlier cohorts to the 2-child norm, followed, in Bulgaria and Hungary, by a subsequent decline. This subsection will consider the question of whether fertility behaviour has shown convergence or divergence within each country over time. Convergence means increasing similarity of behaviour between individuals; divergence means increasing heterogeneity or a wider spread of behaviour patterns. Two measures of dispersion (or variation within a population) are used in this analysis: the coefficient of variation (normalised standard deviation, used when the mean is changing); and the Gini coefficient, a measure of inequality, traditionally used as an indicator of inequality of income. The Gini coefficient has been used in previous demographic studies, though not the coefficient of variation. Shkolnikov et al (2004) and Vaupel and Goodwin (1987) have used the expression ‘concentration of reproduction’ to show country-to-country variations in describing what proportion of women bear what proportion of children: in countries with a high concentration of reproduction (defined by a high Gini coefficient) then a small proportion of women have a large proportion of children. When the concentration of reproduction is declining, the Gini coefficient is (by definition) going down and convergence of behaviour is happening. Thus convergence and divergence are terms that signify change in the measures of dispersion over time. Convergence is happening when the indicators are getting smaller; divergence when the indicators are increasing.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

60 _____________________________________________________________________ In terms of income or wealth disparity, convergence is commonly considered to be a ‘good thing’, as it is likely to reflect a society with increasingly equal access of individuals or households to resources. A society in which wealth is concentrated in the hands of a few individuals while the masses live in poverty is considered to be inherently unstable. The guiding principle of communism was to bring equality to the people. In this they were largely successful: on the Wikipedia page describing the Gini coefficient, the measure for income variability of a sample of countries was lowest for Bulgaria for most of the period of the 1960s, 1970s and 1980s until it suddenly started to rise in the 1990s (Wikipedia, 2007a). Since the fall of the communist regimes, there has been rapid divergence in the distribution of wealth in those countries. However, this new-found ‘freedom for diversity’ has spread not only to wider income distributions, but also to behaviour norms. As discussed at length in the previous section, there has been a widening in the age of childbearing (as defined by the inter-quartile age at first birth). This section looks at changes in the overall distribution in number of children born. In the earlier discussions it was proposed that there was the potential for good for relaxing the norms in childbearing timing. Whether widening or narrowing of behaviour patterns in the realm of number of children can be considered a ‘good thing’ or not will be discussed in the conclusion of this report.

_____________________________________________________________________ Marion Burkimsher

61 _____________________________________________________________________ The two graphs above show the changing variation in number of children; it is interesting and reassuring to see that the two measures - Gini coefficient and coefficient of variation - show essentially the same patterns, though with some minor differences. It must be stated here that for the calculation of both Gini coefficient and coefficient of variation, women having 4 or more children were grouped together and considered to have had exactly 4 children. Therefore the calculated indicators are an under-estimate of the true variation (especially for the earlier cohorts when larger families were somewhat more common). However, it is unlikely that there would be a great deal of change in the indicators if all family sizes had been included, as the proportions of women having over 4 children is very low indeed (except perhaps for Georgia in the earliest cohorts). With a small sample size of larger families, then a few extreme values can bias the whole calculation, so it was not thought unwise to make the simplification, even on theoretical grounds. Comparing the Gini values calculated from the GGP sample and those published by Shkolnikov et al (2004), derived from population data, then the agreement is quite good. They showed Bulgaria’s Gini falling from around 0.30 for the 1930-1931 cohort to a minimum of 0.20 with the cohorts of women born between 1947 and 1957 (with a small peak between those dates). This study would suggest values somewhat higher, but with a similar trend. For Hungary, their Gini values were higher than for Bulgaria – around 0.34 for the 1936-1937 cohort, falling to 0.27 for the cohorts born in the mid-1950s, before rising again. These are in line with those found in this study. The decline in variation of family sizes, ie. convergence, was a pattern noted in all the countries studied by Shkolnikov et al, and is clearly seen in these graphs. They noted the strong influence of the proportion of childless women on the Gini measure, with a fall leading to convergence, especially when accompanied by a fall in larger families. It is interesting that the three countries studied have followed a similar trajectory, all reaching a maximum point of convergence with the cohorts born 1955-1964. Shkolnikov et al did not consider future potential fertility of the cohorts born 19601961 (the youngest analysed in their study). However, the graphs above show that potential future fertility can change the level of variation widely – because of the strong influence of childlessness on the indicator. If the fertility plans as stated in the GGP survey are accomplished, then convergence will continue; if these planned children are not born (especially to currently childless women), then there will be divergence. The graphs suggest that even if fertility plans are achieved, divergence is probably likely to be commencing in Hungary with the youngest cohorts, and Bulgaria may see this too. The levels of variation in family sizes as seen by these indicators are very low by western European standards. For comparison, in Switzerland, the coefficient of variation for cohorts of women born in the early 1920s was 0.75, and this fell to a minimum of 0.64 with the cohorts born at the end of the 1930s (Burkimsher, 2006). Since then there has been a rise to around 0.7 for the cohort born in 1960. The higher level of variation is due to the greater proportion of childless women in Switzerland, but also a higher proportion with three and more children.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

62 _____________________________________________________________________ The following two graphs show the variation in family sizes looking just at mothers. Obviously the degree of variation is cut down considerably if only parities 1, 2, 3 and 4 are considered and childless women are not included. The breadth of variation seen is extremely limited, with their increasing focus on the 2-child norm. At the minimum, the coefficient of variation value was between 0.35 and 0.37 for all three countries: compare this to a value of around 0.45 in Switzerland for recent cohorts. The values found from studying mothers are again broadly similar with those found by Shkolnikov et al. They found a trend in the Gini coefficient from 0.24 down to 0.19 for Bulgaria, , while for Hungary it fell from 0.27 to 0.20: both these are a little higher than those derived from the GGP sample. The recent trends show that divergence of behaviour has started to happen between mothers in Hungary, and depending on the outcome of the women’s fertility plans, the same may happen for Bulgaria. In Georgia, the trend is even more dependent on future fertility outcomes, whether convergence on the 2-child norm is accentuated and so becomes like Bulgaria of 20 years previously, or whether diversity will increase.

_____________________________________________________________________ Marion Burkimsher

63 _____________________________________________________________________ Fertility plans of women The GGP, in enquiring about planned future children, offers a wealth of useful information on the desirability of each parity in each country. The previous graphs have shown the predicted distribution of parities if everyone has the number of children they desire (though up to a maximum of one extra child). The following short analysis shows the proportion of women by parity who plan to have a(nother) child in the coming three years. In the first graph below, we can see here the strong desire to have a first child, especially among women in their late 20s and early 30s. As women get past their mid30s, the proportion of women planning their first child diminishes. However, even in their early 40s, 30-40% of childless women are still hoping for a child in the coming 3 years. For the cohort of women in their late 20s, the desire for a first child is highest in Hungary, perhaps because a larger proportion of women there are still childless. At higher ages, there is little difference in the fertility desires between the three countries. Clearly the desire for at least one child is very strong in these countries. In the graph looking at women who currently have one child, it is interesting to see that the desirability of a second child is over 10% lower in Bulgaria for the younger cohort groups compared to Hungary and Georgia, which are very similar.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

64 _____________________________________________________________________ Looking at women currently with two children, Bulgaria women again show that they have much less desire for a third child than women in Hungary and Georgia. Only 57% of Bulgarian women in their main childbearing years are considering having a third child, in contrast to over 20% of Hungarian and Georgian women. It is notable that the proportion of women wanting a second or third child is similar in Hungary and Georgia. Looking at the final graph of women with three children who are considering a fourth, the number is quite low, ranging from 10% to under 20%. Not surprisingly, Georgian women (at least those in their early 30s) have most desire for a fourth child, and Bulgarian women least. Except for the final graph above, the cohort group 1975-1979 has been included - that is women in their late 20s at the time of survey (for the final graph, few women have already had three children by this age). This suggests that the countries are unlikely to see any major changes in their fertility patterns in the near future. In particular, the desire for children higher than parity two is low in Bulgaria, and even the desire to move on to a second child is waning there. In Hungary, however, there is still a healthy potential desire for children, on a similar level to that seen in Georgia. Attitude of pregnant women depending on existing parity We can also investigate the attitude to having a(nother) child by looking at the attitudes of pregnant women. They were asked in the GGP questionnaire “Just before this pregnancy began, did you yourself want to have a/another baby at some time?” to which they could reply “Yes”, “No”, or (in Bulgaria) “Not sure”. This question was not asked in the Hungarian survey. The following tables summarise the responses. Bulgaria

Existing number of children

Georgia

Existing number of children

R wanted a baby before pregnancy began

0 1 2 3 5

yes

no

not sure

% not wanted or not sure

40 31 1 1 0

1 0 2 1 0

0 4 0 0 1

2.4 11.4 66.7 50.0 100.0

R wanted a baby before pregnancy began

0 1 2 3

yes

no

% not wanted

35 31 6 1

7 4 5 2

16.7 11.4 45.5 66.7

In both countries, the vast majority of first and second children were desired (although some of those were earlier or later than planned, as discussed in the previous section). However, it is notable that 4 out of 35 women in Bulgaria pregnant with their second child were “Not sure” whether they had really wanted that child. Subsequent children, with increasing ambivalence the higher the parity, were a lot less desired, particularly _____________________________________________________________________ Marion Burkimsher

65 _____________________________________________________________________ in Bulgaria. It is perhaps surprising the relatively high proportion of unwanted first and second children in Georgia (but perhaps this is because women did not have the option of a “Not sure” answer there). Falling fertility/rising fertility: which cohorts experienced these? Overall the trend of completed fertility has been downwards for the cohorts born in the 20th century. However, this generalisation masks some significant variations. Childlessness has been falling, and this may continue if the respondents fertility plans are realised. The levels seen in Eastern Europe are considerably lower than in Western Europe. From the additional information that the GGP data can give on fertility plans of childless women and the attitudes of pregnant women expecting their first child, it would seem that having at least one child is the desire and aim of the vast majority of women (and men). The only marked rise in fertility seen in the three countries studied was in Hungary for the cohorts born between 1945 and 1964. In that period, the proportion of 1-child families fell and that of 3- and 4-child families increased. More recently, in Bulgaria and Hungary, there has been a large rise in the proportion of 1-child families at the expense of 2-child families. This has brought down the completed cohort fertility level down to significantly below replacement level. It is quite possible that the CCF for the cohorts born in the early 1970s will be around 1.5 children per woman. In Georgia, the CCF may slip below 2 for the youngest cohorts. Comparison of TFR and CCF The TFR is affected by actual changes in fertility and also changes in timing of fertility. The effects of the delays which have occurred since 1989 in childbearing and the associated deflation of the TFR has already been discussed. Let us now look at trends in the TFR before 1989 and see how they were reflected in the CCF. For Bulgaria, there are indications that there were small peaks in the TFR in 1968 and 1974, with a dip in between. However, these would appear to have had no impact on CCF as seen from both the GGP sample data and population data. Short-term rises and falls of TFR, caused by economic or social factors, are usually smoothed out in the CCF, and this would seem to be an example. In Hungary there was a marked peak in the TFR in the mid-1970s, followed by a trough and then a more gentle peak in the latter part of the 1980s. The latter peak, although less abrupt, does seem to have influenced the CCF, raising it above 2 for the cohorts of women born around 1960. The influence of this favourable climate for childbearing in Hungary and its influence on changing family sizes has already been discussed. In Georgia, TFR and CCF have been falling from around 2.5 to 2. There was a spike in the TFR around 1990, but as it was short-lived, this is unlikely to have any influence on the CCF of women going through their fertile life at that time. Although _____________________________________________________________________ New fertility behaviour in Eastern Europe

66 _____________________________________________________________________ the TFR has dropped very sharply since then, CCF is unlikely to fall to the low levels seen in Bulgaria and Hungary. Since the regime change of 1989, TFRs have fallen sharply because of the combined factors of delay in childbearing added to an actual fall in fertility, in particular a drop in the proportion of families of 3 or more children, and the rise in 1-child families at the expense of 2-child families. Changes in the proportion of childless women are relatively unlikely to have an impact on fertility rates in these countries, if the fertility desires of childless women are fulfilled to a large extent. Determinants of fertility We will now look at a (small) number of determinants of fertility. In the previous section we saw how age at first birth influenced the total number of children born, and also that there was a potential multiplicative effect – that older mothers tend to have fewer children, but also an increasing proportion of first-time mothers in the younger cohorts are falling into the ‘older’ category. We will now look at four other variables: these also have the potential to have a multiplicative effect in causing fertility to fall. They are place of residence, whether urban or rural; number of siblings; educational level; and religious participation. Ethnicity is then discussed in a later subsection. These are used as illustrations of variables that have an influence on fertility levels across all three countries studied. As such, they would be used as control variables when a rigorous longitudinal analysis is carried out to determine potential temporal determinants of fertility. Although place of residence and religiosity can change over time, for most people they are time-independent variables, at least in the short to medium term. Educational level and number of siblings are (almost) certainly time independent variables after an individual has reached their early 30s. Rural/urban residence

The first graph illustrates that an increasing proportion of younger women are living in urban areas compared to rural areas. The second two graphs show that the women in urban areas across all cohorts were more likely to have just one child and less likely to have three children compared to women in rural areas.

_____________________________________________________________________ Marion Burkimsher

67 _____________________________________________________________________

The variable of place of residence was not included in the data base for Hungary, but it was in Georgia. The same trends were seen in Georgia as in Bulgaria: younger women are more highly concentrated in urban areas (presumably having migrated there for education or employment), and they have fewer children if living in urban areas than in rural areas. If we look just at childless women, the difference in proportion between rural and urban areas is small; it is in family size that the variations become apparent. In both Bulgaria and Hungary, larger families of 3 or more children are considerably more common in rural areas than urban areas, while 1-child families are more common in urban areas. This is not an unexpected result: the costs of housing and the pressure of space are greater in urban areas so the desire the restrict family size is stronger. Also the availability of contraception (and abortion) is likely to be greater in urban areas. Number of siblings

All three countries’ GGP questionnaires asked how many siblings the respondent had ever had. Respondents were split into those who had come from a 1- or 2-child family _____________________________________________________________________ New fertility behaviour in Eastern Europe

68 _____________________________________________________________________ (ie. had had no siblings or one sibling) and those who had come from larger families (of 3 or more children). Not surprisingly, in all the countries the number of siblings has gone down towards the younger cohorts. Similarly, it is no surprise that women who come from smaller families tend to have smaller families themselves, and women from larger families tend to have more children. The patterns were similar for each of the three countries, but Hungary is shown here as an example. In Hungary, the proportion of ‘smaller’ families first outnumbered the ‘larger’ families of origin with the 1955-1959 cohort, as seen where the lines cross in the graph below. In Bulgaria, the families of origin were smaller, with the crossover from ‘larger’ families to ‘smaller’ families occurring earlier (with the 1940-1944 cohort). In Georgia there was a declining proportion of larger families and increasing proportion of smaller families, but the smaller families of 1 and 2 children had not yet attained 50% even with the youngest cohorts of respondents. It should be commented that size of family of origin had less influence on childlessness than on the number of children a woman had. From that we may deduce that childlessness is probably not a choice, whereas number of children borne probably is. This result is in contrast to the results of a study on the determinants of childlessness in Switzerland, which found that having had no sibling significantly increased the likelihood of remaining childless for both men and women (Burkimsher, 2007). This present study grouped respondents coming from both 1-child and 2-child families, so the possible ‘inheritability’ of low fertility may have been attenuated by including both these. Education

The effect of increasing educational levels on reducing fertility in women has been discussed at length in previous literature. This study confirms the effect, and the _____________________________________________________________________ Marion Burkimsher

69 _____________________________________________________________________ example of Georgia is used for illustration. Educational achievement for girls (as well as men, but more markedly for women) has been rising over the past half century in all the countries studied. While place of residence or number of siblings were not associated with greater childlessness in women, higher educational attainment does appear to increase the likelihood, as seen in the graph for Georgian women above. The same association held true in Bulgaria and Georgia. It was not true, however, for men: men with low educational attainment were equally likely to remain childless than men with higher qualifications in Hungary and Georgia; in Bulgaria they were slightly more likely to stay childless. It should be noted, however, that many of the highly educated and currently childless women are planning to have a child in the next 3 years in all three countries. If they succeed, then their level would drop to be equal to that of less qualified women. It will be interesting to see if they meet their fertility expectations by the next round of the GGP. As seen on the graph for Georgia above, more highly educated women are planning to have their first child at higher ages than less educated women, who would seem to be resigned to their state of childlessness after the age of 35-40. Herein lies one of the nubs of the problem: women who are highly educated tend to postpone starting a family until later in life, believing it is still possible to have a child even after the age of 40 – but then the delay proves to have been too long and they end up involuntarily childless. Women with higher qualifications are also less likely to have large families; this goes with starting a family later, as discussed in the previous section. As seen for the graph of Georgia above, highly educated women have rarely had large families, even those in the oldest cohorts. In Bulgaria practically no women with higher educational attainment had more than three children, and well-educated women with 3 children were a rarity. In Hungary, however, there would appear to be a significant proportion of highly educated women with 2 children in the 1965-1969 cohort who are planning to have a third child in the coming 3 years. If they succeed then there will be little difference between the proportions with high education and low education in the 3child bracket. Families of 4 and more children are, however, the preserve of lesser educated women even in Hungary. Religiosity In Bulgaria and Georgia, respondents were asked about their frequency of participation in religious services as well as their religious affiliation, but this was not asked in Hungary. It was felt that the level of participation would be more likely to influence an individual than their nominal affiliation, and so this was what was studied for Bulgaria and Georgia. The dichotomy of higher versus lower religiosity was set at participation in religious services at least once a month. As mentioned in the literature review, higher religious participation has been previously been associated with higher fertility when comparing the United States with Western Europe (Frejka and Westoff, 2006).

_____________________________________________________________________ New fertility behaviour in Eastern Europe

70 _____________________________________________________________________

Firstly, it is interesting to look at levels of religiosity by cohort, and we see in the graph above that in Georgia, there in an increasing level of participation among the younger cohorts (a pattern quite different to that seen in Western Europe). In fact the trend continues upwards with the youngest cohorts (not shown on graph), and it also holds true, though at a slightly lower level, for men. For the youngest cohort interviewed, those born 1985-1988, more than 50% of men were attending religious services over once a month, and more than 60% of young women. It would appear that religious participation is an increasingly popular activity among the young in Georgia. In that country, the majority religion is Christian Orthodox (although this analysis does not differentiate between attendance at what type of church). If we then look at the proportions of childless women and the proportions with large families, we see that the more religious have lower fertility than the less religious. The ‘new’ behaviour in religion in paralleled with ‘new’ behaviour in fertility. Childlessness and 1-child families are more common for more religious women, while larger families are associated with less religious women. For men, however, there is very little difference in their fertility levels depending on religious participation in Georgia. Let us compare the situation in Georgia with that in Bulgaria. The majority religion in Bulgaria is also Christian Orthodox, though with sizeable minorities of Muslims and also those who claim no religion.

_____________________________________________________________________ Marion Burkimsher

71 _____________________________________________________________________

The levels of participation in religious services are much lower than in Georgia and there has been little change across the cohorts – around 25-30% for women, and around 10-15% for men, with a small decline in younger cohorts. However, the association of religiosity with lower fertility for Bulgarian women is similar to that seen in Georgian women. Religious women are more likely to remain childless or have one child, and less likely to have three or more children. It would be interesting to investigate further if higher religiosity is associated with any of the other factors known to lower fertility – higher education, urban living or later first birth (probably associated with later marriage). One also wonders whether there had been discrimination (or worse) against those actively involved in religion in communist days, and this had a depressing influence on fertility. Ethnicity The final determinant to be investigated is that of ethnicity. Respondents in Georgia were not asked their ethnicity and so this analysis covers only Bulgaria and Hungary. In Bulgaria, there are significant minorities of Turks and Roma (9% and 5% of the Bulgarian population as of the 2001 census), and these were compared to the ethnic Bulgarian population. In Hungary, the most significant minority was, again, the Roma (called Gypsies in the Hungarian survey) and ‘those of Roma origin’ who account for roughly 2% of the population (these two categories were differentiated in the questionnaire, but in this analysis were combined). Because the numbers involved in the minority groups were small, then the cohort groupings had to be much larger to be able to make valid comparisons. Therefore _____________________________________________________________________ New fertility behaviour in Eastern Europe

72 _____________________________________________________________________ there are just two cohort groups in this comparison (1935-1954 and 1955-1974). The parity distributions for the different ethnic groups are represented below as pie charts. Bulgarian parity distributions by ethnicity Bulgarian women distribution of parities for 1935-1954 cohorts

Bulgarian women distribution of parities for 1955-1974 cohorts

0

0

1

1

2

2

3

3

4+

4+

Turkish women distribution of parities for 1935-1954 cohorts

Turkish women distribution of parities for 19551974 cohorts

0

0

1

1

2

2

3

3

4+

4+

Roma women distribution of parities for 19351954 cohorts

Roma women distribution of parities for 19551974 cohorts

0

0

1

1

2

2

3

3

4+

4+

These show most dramatically how the different ethnic groups in Bulgaria have completely different fertility patterns. For ethnic Bulgarians families of 4 and more children have been almost non-existent over the whole period, while this was the norm for Roma women in the cohorts up to 1954. One-child families have been the fertility outcome for over a quarter of Bulgarian women across all cohorts. There has actually been little change in the distribution of parities from the older cohorts to the younger ones for ethnic Bulgarians, while family sizes have shrunk markedly for the Turkish and Roma minorities. The fertility patterns of the Turkish minority is between the two extremes of the native Bulgarians and the Roma population, and the 2-child norm is increasingly becoming established in their community.

_____________________________________________________________________ Marion Burkimsher

73 _____________________________________________________________________ Hungarian parity distributions by ethnicity Hungarian women parity distribution of 1955-1974 cohorts

Hungarian women parity distribution of 19351954 cohorts

0

0

1

1

2

2

3

3

4+

4+

Gypsy women parity distribution of 1935-1954 cohorts

Gypsy women parity distribution of 1955-1974 cohorts

0

0

1

1

2

2

3

3

4+

4+

If we look at the parity distributions in Hungary for ethnic Hungarians compared to those the Gypsy (Roma) population a similar contrast is found, though rather less stark than in Bulgaria (native Hungarians do sometimes have 4 children). Once gain, over half of the Gypsy women had families of over 4 children in the earlier cohorts, though this had declined with the younger cohorts (though they still had larger families than the Roma in Bulgaria). In Hungary, an increasing proportion of native Hungarians had 3 children comparing the younger cohorts with the older ones, with a slight increase in 4+ child families. This is a reflection of the rise in fertility in Hungary already discussed earlier. Hungary managed in the 1970s and 1980s to increase the fertility of its native population while at the same time decreasing the fertility of its Gypsy population. These graphs clearly demonstrate that two sub-populations can have completely different fertility behaviour: they are like different countries! The high fertility of the Roma population is seen as a problem, particularly in Bulgaria where they are more numerous, and this must colour the population policy for the whole of that society. If larger families (3 or more children) are inextricably associated with Roma behaviour, then it will be difficult for policy-makers to devise a way of encouraging higher fertility in the native population. Summary of fertility trends and determinants Fertility levels are determined by the proportions of women who have no children, one child, two children, three children or larger families. This analysis has dissected where and when the changes have occurred.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

74 _____________________________________________________________________ In all three countries the proportion of women remaining childless is low by western standards and has been declining across the cohorts. If women’s stated fertility expectations are met, then this decline may continue. The proportions of mothers having 1, 2, 3 or more children, however, varies quite widely from country to country. Larger families have traditionally been more common in Georgia, while small, 1-child families have been relatively common in Bulgaria across the whole time frame. The 2-child norm became increasingly firmly established in Bulgaria and Hungary for cohorts born during and after the Second World War, but in the most recent cohorts this has been giving way to an increasing proportion of women with just one child. It would seem that the delays in having a second child (discussed in the previous section) are increasingly turning into abandonment of doing so. The barriers perceived by couples in progressing from one to two children should be studied in more depth when the second wave of GGP data is available, as that is where the greatest changes are currently happening. This is most marked in Bulgaria, but Hungary appears to be catching up on the trend. Are the barriers at a personal level in the stability or quality of relations between the partners, or are economic and social constraints paramount? The only country of the three to achieve a sustained rise in cohort fertility at any time during the second half of the 20th century was Hungary, which it did in the 1970s and 1980s, though the fall of communism in 1989 brought this to an abrupt end. In that period there was an increasing number of families of 3 and more children and a fall in 1-child families. It would be instructive to know more about what was the main engine of this increased fertility. In the previous section, it was posited that greater acceptance of diversity in age at childbearing may help more women to feel free to have children. This section looked at diversity in number of children. From the measures of dispersion used, it would seem that the range of family size is starting to become more varied in Hungary and Bulgaria (if one discounts the childless), after decades of convergence on the 2-child norm. Unfortunately this increased variability stems only from the increase of 1-child families at the expense of 2-child families. Families larger than 2 children are becoming increasingly rare with the youngest cohorts. Georgia has had more variability in family sizes in the past, but it is now going down the path to a 2-child norm, a few decades behind Hungary and Bulgaria. Looking at a number of factors which have an influence on fertility levels, we have seen that these factors can (at least partially) explain the fall in fertility. More highly educated women have fewer children and there is now a higher proportion of more highly educated women than in the past. Urban couples have fewer children than rural ones, and more young people are choosing to live in the city. People are postponing having a first child until later in life – but if they do they tend to have fewer children. The influence of religious participation on fertility is interesting, as it is opposite in effect to that found in Western Europe and the United States. An increasing proportion of younger people are participating in religious services in Georgia, but their ‘new’ behaviour in the realm of religion is paralleled by ‘new’ fertility behaviour, ie, smaller families. It would be interesting to have more insight as to whether the increased likelihood of remaining childless (and not having larger families) in women across all cohorts who participate in religious activities in

_____________________________________________________________________ Marion Burkimsher

75 _____________________________________________________________________ Bulgaria and Georgia could be attributed to anti-religious discrimination in the communist era. Finally, the fertility patterns between the different ethnic groups must be highlighted, because the variations are huge. The Roma people have traditionally had considerably higher fertility in both Bulgaria and Hungary than the native population, although it has been falling across cohorts. To a lesser extent the same is true for the Turkish minority in Bulgaria. The Roma are an under-privileged and unloved people group, and mainstream policy-makers are unlikely to offer any incentives for families to have more children that may then (be seen to) favour the Roma. Roma children are discriminated against in the education system and are sidelined by society. Unless and until Roma children are valued as potentially valuable members of society, then it is difficult to see how policies favouring childbearing for all members of society can successfully and fairly be formulated. In addition, if families of 3 children are seen as ‘too large’ in Bulgaria, and having 4 children would be seen as ‘Roma behaviour’ then fertility levels cannot help but remain considerably below replacement level. However, the regime in Hungary in the 1970s and 1980s (either by policies or economic climate) did succeed in encouraging native Hungarians to have larger families while at the same time facilitating the Roma population to have smaller families. It would be worthwhile investigating if any lessons gained from that time can be applied today.

Overview of discussion topics This final section reviews the different strands of this research project and makes inferences on the current fertility situation in the three countries under study, suggesting what may be in store in the near future. With these insights, policymakers can then deduce the gravity of the population crisis specific to their country, and researchers can focus their efforts on the parity transitions which are currently changing most. Fertility predictions What can the fertility levels be expected to be in the three countries under examination in the foreseeable future? The preceding analyses have given pointers to where the trends may be leading; the following table presents predictions of the parity distributions - and thus the expected completed fertility - of the cohorts born in the late 1970s and early 1980s, and the TFR in the years up to 2010.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

76 _____________________________________________________________________ Bulgaria Hungary Georgia % childless 5 5 5 % 1 child 37 33 20 % 2 children 51 44 50 % 3 children 5 15 19 % 4+ children 2 4 6 CCF 1.62 1.82 2.01 Delay in childbearing 0.2 0.2 0.1 TFR 1.30 1.46 1.81 Delay in childbearing is the change in mean age of birth year-on-year applied equally to each parity TFR is the expected total fertility rate, calculated by multiplying the CCF by (1-the delay) Parities higher than 4 are considered as 4 exactly, as they are rare

Making predictions about the future is a challenging task. However, if we follow through on current trends and attitudes, we can suggest where those are leading – and then if that is perceived to be an undesirable outcome, evasive action can perhaps be taken. It is by considering the ‘What if?’ scenarios that policymakers can consider the gravity of the problem and see whether action is advisable. The expected proportion of childlessness is expected to remain low, although this could rise if age at childbearing continues to rise rapidly. The other proportions were deduced by selecting a value somewhere between current actual fertility and predicted fertility if fertility plans are fulfilled (usually a value closer to the latter). The changes in timing of fertility chosen for illustration are somewhat lower than the rates seen in recent years, but it can be expected that those precipitous rises will slow in the coming years. The delays are expected to be similar in Bulgaria and Hungary, but rather less in Georgia. It is expected that Bulgaria’s fertility rates (CCF and TFR) will be consistently low, bordering on ‘lowest-low’ levels, while Hungary’s will be somewhat higher and Georgia’s higher still. Completed cohort fertility for Georgia may remain over 2 children per woman, even though TFRs will be deflated below that level because of delays in childbearing. Increasing diversity – is it a ‘good thing’? It should first be stated that a ‘good thing’ is seen here as something that increases fertility. With fertility as low as it is, a trend which may continue for the foreseeable future particularly in Bulgaria and, to a lesser extent, Hungary, then increasing it is the main challenge. Past fertility behaviour in all three countries saw increasing convergence in both timing and parity distributions, until the far-reaching changes with the fall of communism started to break those norms. Timing of having a first child had been increasingly focussed into a 5 year period (the inter-quartile range) between the ages of 20 and 25 in Bulgaria and Hungary. More variability was seen in Georgia, although the spacing between first and second child was closer there. Since the beginning of the 1990s, there has been divergence in timing of births, and postponement of starting a family has been marked. Second births, especially, have been delayed, and in many cases the delay has turned into abandonment of the idea. _____________________________________________________________________ Marion Burkimsher

77 _____________________________________________________________________ Widening the time span when women feel comfortable in having their first or subsequent child may provide more freedom for older women to enter into motherhood or have an extra child. Several countries in Western Europe have a quite large proportion of children born to women in their 30s (and even 40s). However, having a first child later does tend to be associated with smaller overall family size, and it must be advertised – perhaps especially to highly educated women who are pursuing a career – that they should not leave starting a family ‘too late’, otherwise they may be disappointed. With the positive media coverage of ART, it is sometimes thought that it is as easy to choose to have a child when one wants, as using contraception works to avoid having a child – but it is not. Bulgaria and Hungary have gone through a period of tight 2-child norms in family size and Georgia is following them down the same path. In the first two, those strictures are finally easing – but unfortunately this is not leading to any increase in larger families of 3 or more children, it is only an increase in the proportion of 1-child families at the expense of 2-child families. Therefore, at the present time, it does not seem that the slight increase in diversity in family sizes seen in Hungary, and to a lesser extent in Bulgaria, will increase their fertility levels. Education, urban living, family of origin, religion These four factors were examined as potential determinants of fertility levels. For each of these – higher educational attainment of women; urban versus rural residence; smaller family of origin; and greater religious participation – lower fertility was observed across all cohorts and in all three countries. However, the proportion of women having each of those attributes is increasing with the younger cohorts: women are more highly educated; more young people are living in the cities; families of origin are smaller as fertility has declined; and, in Georgia, an increasing proportion of the young are regularly attending religious services (predominantly Christian Orthodox). Therefore, it is not surprising that fertility is falling among the younger cohorts. Unfortunately none of these factors is inside the realm of policymakers being able or willing to change. The true causative nature of these determinants is complex and beyond the remit of this study, but it would be useful if further studies investigated this, together with the correlations between each of them. The ethnic challenge One of the more remarkable results found from this analysis was the radical difference in fertility patterns seen in Bulgaria and Hungary when comparing the ethnic minorities of Roma and Turks with the ethnic majority population. The Roma (Gypsies in Hungary) still traditionally have large families, while, in Bulgaria in particular having 4 or more children is almost unknown among the native Bulgarian population. The fertility levels of the Roma have been falling with younger cohorts – but perhaps not as much as some people in those countries would like. The problem is that although the fertility rate is so low in Bulgaria, it is a challenge for policymakers to be seen to be encouraging larger families if they are seen as the _____________________________________________________________________ New fertility behaviour in Eastern Europe

78 _____________________________________________________________________ behaviour pattern associated solely with the Roma (and some Turks). The Roma suffer severe discrimination (Kirova, 2008) and until this diminishes and every new baby is seen as a potential asset to the country then it will be difficult to favour higher fertility. Every child must be seen as a wanted child when fertility levels are so low, and the potential of every child to be a productive member of society must be maximised. Another aspect of sidelining the Roma may have been observed in the population statistics relating to Bulgaria. The values of both the TFR and CCF derived from the GGP sample data are lower than the published historical rates based on the whole population. Could this be because the population of Roma is under-estimated? This has been hypothesised elsewhere – there may be twice as many Roma in Bulgaria as were counted at the last census (Marushiakova and Popov, 1995). The Roma do not like to be counted and native Bulgarians prefer they did not exist, so are happy to under-count them. As the Roma have considerably higher fertility rates than the native population, then an under-estimate of the resident Roma population may account for these differences in TFR and CCF. Policy applications In the latter half of the 20th century, just one of the three countries managed to raise its already low fertility for a period of time – that was Hungary in the 1970s and 1980s. There was not a sharp spike in the TFR, which can sometimes happen with short-lived policy interventions, but a 2-decade long period in which childless levels went down, and 3-child families slowly came to outnumber 1-child families. It was during this period that Hungary was termed “the happiest barrack in the socialist camp” as it was a period of social stability and relative economic prosperity (Wikipedia, 2007b). If policies are being sought that may favour increased fertility, then the factors and policy interventions that succeeded then in Hungary should be examined more closely. As children come in integer units, then it can be useful to see which transitions to a higher parity are changing most. This study pinpoints that the greatest change in parity distributions in Bulgaria and Hungary is the increase in number of women that are staying with one child. Therefore the barriers in progressing from one child to two children should be looked at in particular. Are they predominantly economic? Or is it inter-personal instability that is a root cause? Would additional social services to assist working mothers help – child care, more part-time work, or job security? The huge barrier in moving from a second to a third child in Bulgaria should also be examined. Can anything be done to help couples consider a third child? In Georgia the 2-child norm is still in ascendance as higher parity births are declining. One of the main recommendations for policymakers in the population sphere is to use a fertility indicator that is better than the TFR at representing the actual number of children being born to each woman. A simple improvement is to use the BongaartsFeeney correction for change in timing of childbearing. However, for this good statistics need to be collected for birth by biological party (not just parity in current marriage) and accurate population numbers need to be known for each cohort of (fertile) women. With those, then parity-specific birth rates, corrected for changes in

_____________________________________________________________________ Marion Burkimsher

79 _____________________________________________________________________ timing, can be calculated, and these can give a much more accurate picture of the component causes of fertility changes. Where more research is needed As mentioned several times in this report, this research project is foundational to a more in-depth longitudinal study of fertility transitions using the data available from the second wave of the GGP. The idea of the multi-wave GGP was to study fertility decisions and outcomes in ‘real time’ so that the true influences on fertility decisions can be ascertained. Vikat (2006) states: “To make causal inferences, the analyst needs data where the hypothesized cause is observed before the outcome in a person’s lifetime”. This report looked at just a few, essentially time-independent, determinants of fertility. A subsequent study should use these as control variables to look more deeply at inter-individual variations in fertility to see whether it is economic, social, ideational or personal factors that are paramount in determining fertility outcomes. The GGP is highly instructive in asking for fertility plans for the coming 3 years, and this study has exploited that information to make tentative predictions of future fertility. Follow-up of the same individuals in the second wave should yield a mine of information. For individuals who said they plan to have a(nother) child in the coming 3 years in the first wave, how many succeed in their plans? What are the differences between cohorts, between existing parities and between countries? And which women end up having a child when they were not planning to? End note This study has looked at the trends in fertility in three Eastern Europe countries, interpreted the current low fertility levels in the light of certain new behaviour patterns, and made cautious predictions about future trends. It is unfortunate that the author has very limited personal knowledge of the countries in question, which somewhat restricts the range of interpretations possible. The standpoint of someone from Western Europe (a native Brit and student of Swiss demography) looking at the quite different patterns to the east will also be clear to the reader, and it is hoped that this natural bias has not excessively coloured the discussions or conclusions. Acknowledgements I would like to thank my academic supervisor, Miroslav Macura, for advice before and during this study. The research work was carried out during an internship in the Population Activities Unit of the United Nations Economic Commission for Europe, under the supervision of André Kveder, who must also be thanked for help in data extraction and guidance on the direction of the project. Thanks to Rudolf Anich for help in cracking the Gini coefficient and to my husband for supporting me, morally and financially, during this return foray into academic research as a (very) mature student. _____________________________________________________________________ New fertility behaviour in Eastern Europe

80 _____________________________________________________________________

References Alich, D. 2006 The third child: a comparison between West Germany and Norway. Rostock, MPIDR Working Paper WP-2006-001. Atoh, Makoto. 2001. Very low fertility in Japan and value change hypotheses. Review of Population and Social Policy, No. 10, 1-21 BBC. 1999. Boys face teen pregnancy crackdown. http://news.bbc.co.uk/2/hi/health/367422.stm BBC. 2007a. Russian ‘sex day’ to boost births. http://news.bbc.co.uk/2/hi/europe/6990802.stm BBC, 2007b. Mothers in 30s boost population. http://news.bbc.co.uk/2/hi/uk_news/7139079.stm Beets, Gijs. 2006. An assessment of the size and cost of involuntary infertility that may be attributed to postponement. Research Note, European Commission. Betts, Katharine. 1996. Patriotism, immigration and the 1996 Australian election. People and Place, Vol 4, no. 4 (http://elecpress.monash.edu.au/pnp/free/pnpv4n4/bettsi.htm) Billari, Francesco C. and Hans-Peter Kohler. 2004. Patterns of low and lowest-low fertility in Europe. Population Studies, Vol. 58, No. 2, pp. 161–176 Birrell, R. 2003. Fertility crisis: why you can't blame the blokes. In The Age http://www.theage.com.au/articles/2003/01/16/1042520724087.html Bongaarts, John. 1999. Fertility Decline in the Developed World: Where Will It End? The American Economic Review, Vol. 89, No. 2, Papers and Proceedings of the One Hundred Eleventh Annual Meeting of the American Economic Association. (May, 1999), pp. 256-260. Bongaarts, J. and G. Feeney. 2005. The quantum and tempo of life-cycle events. Population Council Working papers, no. 207, Policy Research Division. Burkimsher, Marion. 2006. Convergence or divergence? A study of family sizes in Switzerland. Unpublished report, LaboDemo, University of Geneva, Switzerland. Burkimsher, Marion. 2007. What factors increase the likelihood that a woman will have no children in Switzerland? Are they the same as for remaining single? Are men influenced by the same factors? Unpublished report, LaboDemo, University of Geneva, Switzerland. CIA. 2007. Central Intelligence Agency. The World Factbook. Rank order – Total Fertility Rate. https://www.cia.gov/library/publications/the-world-factbook/rankorder/2127rank.html

_____________________________________________________________________ Marion Burkimsher

81 _____________________________________________________________________ Coleman, D. A. 1998. Reproduction and survival in an unknown world: what drives today’s industrial populations, and to what future? Presented as a lecture under the auspices of the Social Science Council (SWR) of the Royal Netherlands Academy for Arts and Sciences (KNAW). Amsterdam, 7 May 1998 Engelhardt, Henriette and Alexia Prskawetz. 2004. On the changing correlation between fertility and female employment over space and time. European Journal of Population 20: 35–62. Festy, Patrick and France Prioux. 2002. An evaluation of the Fertility and Family Surveys Project. United Nations, New York and Geneva, 2002. Frejka, T. and J.-P. Sardon. 2004. Childbearing trends and prospects in low-fertility countries. A cohort analysis. Kluwer Academic Publishers, Dordrecht, Netherlands. Frejka, T. and Sardon, J.-P. 2006 First birth trends in developed countries: a cohort analysis. Rostock, MPIDR Working Paper WP-2006-014. Frejka, T. and Westoff, C. F. 2006 Religion, religiousness and fertility in the U.S. and in Europe. Rostock, MPIDR Working Paper WP-2006-013. Gonzalez, Maria-José and Teresa Jurado-Guerrero. 2006. Remaining childless in affluent economies: a comparison of France, West Germany, Italy and Spain, 19942001. European Journal of Population, 22, 317-352 Hakim, Catherine. 2003. A new approach to explaining fertility patterns: preference theory. Populations and Development Review, 29 (3). Harding, Mark. 2006. Germany agonises over 30% childless women. Guardian Unlimited. http://www.guardian.co.uk/germany/article/0,,1695850,00.html Hobcraft J. 2002. Moving beyond elaborate description: Towards understanding choices about parenthood In M. Macura and G. Beets (eds.), The Dynamics of Fertility and Partnership in Europe: Insights and Lessons from Comparative Research, Vol. I. New York and Geneva, United Nations. Hirosima, Kiyoshi and Fusami Mita. 1995. Prefectural differentials in recent fertility. IPSS publication Jinkomon, Vol 50, no. 4, 29-30. Kirova, Iskra, 2008. The Decade of Roma Inclusion. Addressing Racial Discrimination Through Development. UN Chronicle, A magazine for the United Nations. http://www.un.org/Pubs/chronicle/2007/issue3/0307p36.html Kohler, Hans-Peter, Francesco C Billari and José Antonio Ortega. 2002. The Emergence of Lowest-Low Fertility in Europe During the 1990s. Population and Development Review 28 (4), 641–680.

_____________________________________________________________________ New fertility behaviour in Eastern Europe

82 _____________________________________________________________________ Marushiakova, Elena and Vesselin Popov, 1995. A History of the Roma of Bulgaria. http://www.geocities.com/~patrin/bulgaria-hstry.htm Ni Bhrolchain, Maire. 1992. Period paramount? A critique of the cohort approach to fertility. Population and Development Review 18 (4) (December 1992). Pinnelli, A., H. J. Hoffmann-Nowotny and B. Fux. 2001. Fertility and new types of households and family formation in Europe. Council of Europe, Strasbourg. Preston Samuel H. 1976. Family Sizes of Children and Family Sizes of Women Demography, Vol. 13, No. 1. (Feb. 1976), pp. 105-114. Rieck, D. 2006 Transition to second birth - the case of Russia. Rostock, MPIDR Working Paper WP-2006-036. Rothenbacher, F. 1994. Household and Family Trends in Europe: from Convergence to Divergence. Eurodata Newsletter No.1, p.3-9. Rowland, Donald T. 1998. Cross-national trends in childlessness. Working papers in demography, no.73, Australian National University. Sobotka, Tomas. 2004. Postponement of childbearing and low fertility in Europe. Dutch University Press, Amsterdam. Shkolnikov, V. M.; Andreev, E.M.; Houle, R.; Vaupel, J. W. 2004. The concentration of reproduction in cohorts of US and European women. Rostock, MPIDR Working Paper WP-2004-027. Thomson, Elizabeth. 1997. Couple Childbearing Desires, Intentions, and Births Demography, Vol. 34, No. 3. (Aug., 1997), pp. 343-354. Torr, Berna Miller and Susan E. Short. 2004. Second Births and the Second Shift: A Research Note on Gender Equity and Fertility. Population and Development Review 30(1): 109-130 (March 2004) Toulemon, Laurent. 1996. Very Few Couples Remain Voluntarily Childless. Population: An English Selection, Vol. 8. , pp. 1-27. UNECE. 1999. Fertility and Family Surveys in countries of the ECE region. Standard Country Report. Bulgaria. United Nations, New York and Geneva, 1999. UNECE. 2001. Fertility and Family Surveys in countries of the ECE region. Standard Country Report. Bulgaria. United Nations, New York and Geneva, 2001. UNECE. 2005. Gender and Generations Programme. Survey Instruments. United Nations Economic Commission for Europe and United Nations Population Fund. United Nations, New York and Geneva, 2005.

_____________________________________________________________________ Marion Burkimsher

83 _____________________________________________________________________ United Nations Population Division. 2003. Partnership and reproductive behaviour in low-fertility countries. http://www.un.org/esa/population/publications/reprobehavior/partrepro.pdf Van Peer, C. 2002. Desired and achieved fertility. In: E. Klijzing and M. Corijn (eds.) Dynamics of Fertility and Partnership in Europe: Insights and Lessons from Comparative Research. Volume II. New York and Geneva: United Nations, pp. 117-142. Vikat et al. 2006. Generations and Gender Survey (GGS): Towards a better understanding of relationships and processes in the life course. Presented at the EAPS European Population Conference, 21-24 June 2006, Liverpool. Vaupel, James W and Dianne G. Goodwin. The Concentration of Reproduction among US Women, 1917-80. Population and Development Review, Vol. 13, No. 4. (Dec., 1987), pp. 723-730. Wikipedia, 2007a. Gini coefficient. http://en.wikipedia.org/wiki/Gini_coefficient Wikipedia, 2007b. Goulash Communism. http://en.wikipedia.org/wiki/The_happiest_barrack

_____________________________________________________________________ New fertility behaviour in Eastern Europe