Putting the pieces together: 40 years of fertility trends across 19 post-socialist countries

Abstract

Demographic change has been a key consequence of transition, but few studies trace fertility trends across countries over time. We describe fertility trends immediately before and after the fall of state socialism across 19 Central and Eastern European and Central Asian countries. We found a few common patterns that may reflect economic and political developments. The countries that experienced the most successful transitions and integration into the EU experienced marked postponement of parenthood and a moderate decline in second and third births. Little economic change in the poorest transition countries was accompanied by less dramatic changes in childbearing behavior. In western post-Soviet contexts, and somewhat in Bulgaria and Romania, women became more likely to only have one child but parenthood was not substantially postponed. This unique demographic pattern seems to reflect an unwavering commitment to parenthood but economic conditions and opportunities that did not support having more than one child. In addition, we identify countries that would provide fruitful case studies because they do not fit general patterns.

Keywords:

• Post-socialism
• fertility
• postponement of parenthood
• Central and Eastern Europe
• economic transition

Acknowledgements

We thank Brienna Perelli-Harris for her previous work on the Harmonized Histories (www.nonmarital.org), which we followed closely. In addition, we thank the European Bank for Reconstruction and Development for the Life in Transition Survey, USAID’s Demographic and Health Survey Program for the Demographic and Health Surveys, UNECE for the Family and Fertility Survey, and the Generations and Gender Surveys: these data were obtained from the GGP Data Archive and were created by the organizations and individuals listed at http://www.unece.org/pau/ggp/acknowledge.htm. We are also grateful for helpful suggestions from Thomas Spoorenberg and anonymous reviewers. We gratefully acknowledge the Stockholm Centre on Health of Societies in Transition (SCOHOST) at Södertörn University for supporting Aija Duntava and Sunnee Billingsley’s efforts to create CLiCR.

Notes

1. The TFR is the average number of children women have according to age-specific fertility rates in that year.

2. The countries included in this study are Albania, Armenia, Azerbaijan, Bulgaria, Czech Republic, Estonia, Georgia, Hungary, Kazakhstan, Kyrgyzstan, Latvia, Lithuania, Moldova, Poland, Romania, Russia, Slovenia, Ukraine, and Uzbekistan.

3. We thank Juris Krumins for this information.

4. This term has been widely used since its first use by Kohler, Billari, and Ortega (2002) to denote a TFR below 1.3.

5. For example, the differences between the first birth rates of our 1965 cohort and those of Frejka and Sardon (2006) for Bulgaria, the Czech Republic, Hungary, Poland, Romania, Russia, and Slovenia ranged from 0.001 for the Russian and Romanian cohort to 0.072 for the Bulgarian cohort. Four rates showed a difference greater than 0.01, but they mostly hovered around this number or below, which we consider a negligible difference.

6. Poland is the only exception, and is represented by three surveys.

7. The implication of censoring at age 35 is that our parity transition calculations are an underestimation of transition rates. Were postponement to have already substantially begun or were we to study birth orders higher than the third, the choice to censor at age 35 would have more serious implications for our estimates. Our parity summary estimates miss the contribution of parity 4+ events, which limits how complete our estimate of fertility levels is in higher fertility countries. Finally, our parity summary estimates may be insufficient for a few countries (Czech Republic, Latvia, and Slovenia) due to the lack of enough respondents being surveyed late enough to get a reliable estimate of higher parity events, given the postponement of first births. In regards to the timing of our surveys, we miss an upswing in fertility rates for many countries that occurred in the last years of the 2000s. The TFR has not remained low and this has spurred new research on the forces behind this upswing (Goldstein, Sobotka, and Jasilioniene 2009; Spoorenberg 2015). Our study cannot contribute to that discussion. As regards the future of fertility trends in the region, Basten and Frejka (2015) oriented their study of select Central and Eastern European countries toward that question and argue that we have not seen the end of fertility decline in the region.

8. Some limitations must be noted here: the median age at childbearing could not be estimated for the 1975–1979 cohort for Slovenia because 50% of the cohort had not entered parenthood by the time of the latest survey (2006), which is evidence of continued strong postponement because this cohort had reached the ages of 27–31 at the time of the survey. For the same reason, we could not estimate the most recent median age (for the 1980–1984) cohort for Estonia, Latvia, the Czech Republic, and Hungary. These cohorts had reached the age of 22–26 by the time of the interview, which indicates that postponement had advanced beyond the other countries for which the median age could be estimated for the same birth cohorts with a 2006 survey. This is indicated with a solid gray bar (“unknown” in Figure 1) that extends across all the relevant years.

9. These estimates are derived from the Kaplan–Meier failure function.

10. Specifically, we observe parity transitions using a period approach: Kaplan–Meier failure estimates were calculated for different time periods. The estimates are therefore of synthetic cohorts that allow each woman to contribute to each time period’s intensity as she progressed through those years, rather than including only those who had their first child in 1991 or later, for example. In this way, we exploit the available data to the fullest degree and avoid introducing selectivity based on postponement of previous births into our estimates. Women with multiple births at parity 1 or 2 transitions were dropped for the estimation of the following parity event. Women were observed from the month of the previous birth and censored at the time of interview or turning 45 years old.

11. This very small decline for Slovenia may be due to the early year of the first survey used (1995 FFS) and insufficient time to calculate reliable post-transition second birth rates for the LiTS data (2006) due to postponement of parenthood.

12. Based on cohort parity progression ratios of first, second, and third births, which are not presented for reasons of space.

13. Earlier economic comparisons are not possible because of a lack of data.

14. No information on employment rates for Latvia was available to compare.

15. This is constructed from Kaplan–Meier parity-specific period estimates: we sum the share of women transitioning to the first three parities according to the share of women at risk of a parity event. If we take the example of Albania (92% of childless women had a first child, 76% of women with one child had a second child, and 54% of women with two children had a third child), the parity summary measure would be 2.03 = (0.92 + (0.92 × 0.76) + (0.76 × 0.54)). The summary measure 2.03 might then be interpreted as a TFR for the period, excluding birth orders of 4 or more. To be clear, this summary measure is not directly comparable to a TFR measure. Depending on how common higher order births of four or more are in a context, the summary measure will underestimate the number of children women have. According to Anichkin and Vishnevsky (1994, 49), focusing only on the first three births means that in 1988 we captured 93.2% of all births in Estonia, 85.4% in Azerbaijan, and only 62.1% in Tajikistan (which would be similar to Kyrgyzstan and Uzbekistan). The universal decline in fertility after 1988 means we should be capturing an even higher share of all births in the post-transition period. An additional reason the TFR may differ from our measure is that a decline in the age at first birth in many countries during the 1980s should have somewhat inflated the TFR measures at this time.

16. The estimate for the Czech Republic and Slovenia may be less accurate than others for this measure, as discussed previously.

17. The remarkably rapid advancement of age at first birth for the 1980–1984 cohort is very likely why Albania shows a drop in TFR to as low as 1.3 in 2007, followed by a quick rebound to 1.6 in 2009 (TransMonEE 2012), despite a high value on our 1–3 parity summary measure.

18. This sentence refers to the process of destandardization, which means growing diversity in the timing of parenthood within a country. So, within the same country, we saw the age at parenthood fall for the first 25% of the cohort to enter parenthood at the same time that the age at which 75% of a cohort reached parenthood came later. So, the earliest and latest first births were diverging from the median age.