educational gradients in fertility across the developed countries Tom - - PDF document

Reversals, diminishing differentials, or stable pattern? Long-term trends in educational gradients in fertility across the developed countries

Tomáš Sobotka, Éva Beaujouan, and Zuzanna Brzozowska Wittgenstein Centre for Demography and Global Human Capital (IIASA, VID/ÖAW, WU), Vienna Institute of Demography (Austrian Academy of Sciences), Vienna. Version 1 October 2017 Work in progress, please do not cite

Intr trod

• ducti

tion

• n

Education is a key marker of social status and economic potential, but also of values and preferences, including those pertaining to partnership and family life. In the course of the 20th century, higher education was almost universally linked to lower and later fertility among women (Skirbekk 2008). Having a small family was a typical way in which better educated people accumulated resources and enhanced social status and educational chances of their offspring (e.g. Goodman et al. 2012). Furthermore, highly educated women in the past remained frequently unmarried and childless; in fact non-marriage was often the most important factor behind of the observed negative relationship between women’s education and fertility (e.g. Reher and Requena 2014, Van Bavel 2014). In the last two decades, a growing number of studies have suggested that the negative education- fertility gradient among women in low-fertility countries may eventually diminish or even reverse. Factors that could bring about this change include the rapid expansion of high education, changing mating patterns, expanding public childcare, wider and more flexible employment opportunities for women, and more egalitarian gender relations (e.g. Kravdal and Rindfuss 2008, Esping‐Andersen and Billari 2015). Indeed, in some countries the negative education gradient in completed fertility among women has been diminishing (e.g. Wood et al. 2014); in the Nordic countries and Belgium the educational differences in fertility not only have been diminishing, but they have been also small (e.g. Andersson et al. 2009, Kravdal 1994, Neels and De Wachter 2010). However, a comprehensive cross- country analysis of long-term trends in fertility educational gradient is still lacking as most of the existing evidence comes from studies on single countries, notably the Nordic countries which in many respects differ substantially from other societies. Moreover, the debate is often clouded/distorted/biased by the reverse ecological fallacy: conclusions on aggregate-level trends are drawn from results of micro-level studies (based on individual-level regression models).

2

Aim ims

The aim of this paper is to provide a systematic review of long-term trends in women’s education- specific fertility across the developed world. When examining whether fertility gradients tend to diminish or to persist when family size declines to low levels, we pay particular attention to how the trends among the lower educated and the highly educated women compare to their peers with medium education. To gain better understanding of the possible role of wider institutional context, we conduct our analyses not only by country, but also by broad regions (see Data section for details). We look particularly closely at the differences between the East and the West of Europe that formed two competing political blocs between the late 1940s and 1989. Further, by inspecting the educational differences in childlessness, we assess their role in explaining the observed fertility gradients and their changes over time. We expect that in countries with higher childlessness rates the educational gradients in fertility are likely to be fuelled by childlessness among better educated

• women. Consequently, the educational gradient among mothers should be smaller and diminishing

across cohorts. Our findings about the changes in fertility gradients give important clues about the likely future trends in fertility as the share of women with university degree continues to expand. Wide and persistent gaps in fertility between women with high and medium education level would give a strong signal about the potential for further fertility declines. This is particularly important in the view of concerns many governments in Europe and East Asia express about low fertility.

Data a

We have collected comparative data on fertility, childlessness and parity distribution by education of women born between 1916 and 1970 and aged between 40 and 76 at the time of the data collection. They were in their prime childbearing years between 1930 and 2005 and had experienced a continuous expansion of education. Our analyses include 27 low-fertility countries in Europe, North America, Australia and East Asia, which are grouped in the following broad regions:  English-speaking countries: Australia, Ireland, New Zealand and the United States;  Northern Europe: Denmark, Norway, Finland and Sweden;  Western Europe: Belgium and France;  German-speaking countries: Austria, the territory of former West Germany (former FRG) and Switzerland;  Southern Europe: Greece, Italy and Spain;  Central Europe: Croatia, Czech Republic, Hungary, Slovakia and Slovenia;  Eastern Europe: Bulgaria, Romania, Russia and Serbia;

3

 East and south-eastern Asia: Singapore and the Republic of Korea (South Korea) To assure cross-country comparability, three broad educational categories are used: primary and below, secondary education and university education. The data come mostly from population censuses, but also from population registers and large-scale

• surveys. They were collected in different periods between 1970 and 2013 and originally covered

different cohort groups (both 1-year and 5-year cohorts) and education categories. They were collected within the EURREP project (www.eurrep.org), and are mostly part of the Cohort Fertility and Education database (CFE database, www.cfe-database.org; see Zeman et al. 2014). More details about the data, cohorts and education categories used are provided in Appendix 1.

Se Selected ted results ts

When comparing the completed fertility rate (CFR) by level of education, a clear regional distinction shows up (Figure 1). In cohorts born in the early 1930s the CFR was above 2.5 children per women in

• nly a few countries. This was the case in all education groups in the English-speaking countries, and

in Singapore and South Korea. The change in the overall CFR for all women was partly influenced by compositional changes of the

• population. While the CFR was dropping in many countries, the CFR by level of education remained

relatively stable in some countries or even increased in several countries of Central and Eastern

• Europe. In many countries, however, the fall in CFR was largely driven by the strong decline in CFR

among the low educated women, partly converging towards the CFR of the highly educated women. This pattern is clearly visible for the two Asian countries studies, Singapore and South Korea. The key finding is that there is no uniform trend in education gradient in fertility between countries. In most countries this gradient persisted across the cohorts studied. In Denmark and New Zealand it was narrow across the cohorts analysed, while it almost disappeared among the younger cohorts in Belgium and Finland. At the other end of the spectrum countries including the United States, Austria, Poland, Slovakia or Romania displayed wide educational differentials even among the younger cohorts born in the 1960s.

4

Figure 1 Cohort fertility rate by country and level of education, by broader regions

Figure 2 shows the relative changes in education gradient across regions and countries. Overall, the fertility differences between high- and medium-educated women are considerably narrower than the differences between low- and medium-educated. In countries with wide fertility differentials it is the low-educated women who stand out as a distinct group with elevated fertility. In countries where this relative gap was largest, including Singapore, South Korea, and many countries of Central and

5

Eastern Europe, it was narrowing among women born between 1920s and 1950s, often stabilising or growing among the youngest cohorts (with the main exception of South Korea, where a full convergence took place). In contrast, the smaller gap between high- and medium-educated women proved remarkably persistent and stable in most countries and broader regions, with the main exception of Nordic countries where it disappeared and Belgium, where a cross-over took place.

Figure 2 Relative difference in CFR for low versus medium educated and for high versus medium educated. Countries grouped by broader regions Interpretation of Figure 2: Values around 0 mean that the CFR among the low- or high-educated women was similar (0 means identical CFR levels) to the CFR among the medium-educated. Values above and below 0 denote higher and lower levels, respectively, of fertility among the low- and high-educated than among the medium-educated, e.g. 0.2 and -0.2 signify CFR 20% higher and lower, respectively, than among the medium- educated.

Preli limin inar ary y conclu lusion

• ns

Our analysis reveals a great variety in educational gradients of fertility. There does not seem to exist a general tendency for fertility gradients to diminish when family size declines to low levels. The gap between low-educated women and all the others largely diminished with time, but in some countries

6

is started to grow again in the 1960s cohorts. Generally, low-educated women typically still have considerably more children than medium- and high-educated women. As research on fertility ideals and intentions among women reveals that in Europe and the United States they do not significantly differ by education in most low-fertility countries (Musick et al. 2009, Berrington and Pattaro 2014, Testa 2014, Sobotka et al. 2015, Beaujouan and Berghammer 2017), our findings might suggest that education differentials in fertility signal “excess” unplanned fertility among the lower educated women (Musick et al. 2009) rather than unrealised fertility intentions among the higher educated women.

Ackn know

• wledgements

ts

We are very grateful to Ivan Čipin, Sam Hyun Yoo, Anneli Miettinen, and Glenn Sandström for preparing and providing the data for Serbia, South Korea, Finland and Sweden, respectively. This research was funded by the European Research Council under the European Union’s Seventh Framework Programme (FP7/2007-2013) / ERC Grant agreement n° 284238 (EURREP project).

Re Refe ferences

Andersson, G., Rønsen, M., Knudsen, L.B., Lappegård, T., Neyer, G., Skrede, K., Teschner, K. and Vikat,

• A. (2009). Cohort Fertility Patterns in the Nordic Countries. Demographic Research 20(14):

313–352. Andersson, G., Rønsen, M., Knudsen, L.B., Lappegård, T., Neyer, G., Skrede, K., Teschner, K. and Vikat,

• A. (2008). Cohort Fertility Patterns in the Nordic Countries. MPIDR Working Paper 2008(8).

Available at: http://www.demogr.mpg.de/papers/working/wp-2008-008.pdf. Berrington, A. and Pattaro, S. (2014). Educational differences in fertility desires, intentions and behaviour: A life course perspective. Advances in Life Course Research 21: 10–27. Esping‐Andersen, G. and Billari, F.C. (2015). Re‐theorizing Family Demographics. Population and Development Review 41(1): 1–31. Goodman, A., Koupil, I. and Lawson, D.W. (2012). Low fertility increases descendant socioeconomic position but reduces long-term fitness in a modern post-industrial society. Proceedings of the Royal Society B: Biological Sciences 279(1746): 4342–4351. Kravdal, Ø. (1994). The Importance of Economic Activity, Economic Potential and Economic Resources for the Timing of First Births in Norway. Population Studies 48(2): 249–267. Kravdal, Ø. and Rindfuss, R.R. (2008). Changing Relationships between Education and Fertility: A Study of Women and Men Born 1940 to 1964. American Sociological Review 73(5): 854–873. Musick, K., England, P., Edgington, S. and Kangas, N. (2009). Education Differences in Intended and Unintended Fertility. Social Forces 88(2): 543–72. Neels, K. and De Wachter, D. (2010). Postponement and recuperation of Belgian fertility: How are they related to rising female educational attainment? Vienna Yearbook of Population Research 8: 77–106. Reher, D. and Requena, M. (2014). The mid-twentieth century fertility boom from a global

• perspective. The History of the Family. Available at:

http://www.tandfonline.com/doi/abs/10.1080/1081602X.2014.944553 [Accessed: 10 November 2015].

7

Sandström, G. (2014). The mid-twentieth century baby boom in Sweden – changes in the educational gradient of fertility for women born 1915–1950. The History of the Family 19(1): 120–140. Skirbekk, V. (2008). Fertility trends by social status. Demographic Research 18: 145–180. Sobotka, T., Zeman, K., Potančoková, M., Eder, J., Brzozowska, Z., Beaujouan, É. and Matysiak, A. (2015). European Fertility Datasheet 2015 [Online]. Available at: www.fertilitydatasheet.org [Accessed: 25 November 2015]. Testa, M.R. (2014). On the positive correlation between education and fertility intentions in Europe: indivudual- and country-level evidence. Advances in Life Course Research 21: 28–42. UNESCO (2006). International Standard Classification of Education: ISCED 1997 (Reprint). Montreal, Canada: UNESCO Institute for Statistics. Van Bavel, J. (2014). The mid-twentieth century Baby Boom and the changing educational gradient in Belgian cohort fertility. Demographic Research 30: 925–962. Wood, J., Neels, K. and Kil, T. (2014). The educational gradient of childlessness and cohort parity progression in 14 low fertility countries. Demographic Research 31(46): 1365–1416.

8

Appendix 1: Data used in this study: sources, comparability, education categories and cohorts covered

We have collected census and large-scale survey data for women born between 1916 and 1970 and aged between 40 and 76 at the time of the data collection in 27 low-fertility countries in Europe, North America, Australia and East Asia. The age 40 was chosen as a threshold when fertility was alsmot completed among the women analysed. The countries and census/survey years included are as follows:  Australia: 2011 Census  Austria: 1991 and 2001 censuses  Belgium: 2001 census,  Bulgaria: 2001 census  Croatia: 1991, 2001 and 2011 censuses;  Czech Republic: 1980, 2001, and 2011 censuses;  Denmark: register-based data for 2008 provided by Statistics Denmark  Finland: register-based data, 10-percent sample of the Finnish population resident in Finland during 1970-2010;  France: 1982, 1990, 1999, and 2011 census-based surveys;  western Germany: 2008 and 2012 micro-census surveys; data for East Germany were not included due to the low numbers of respondents in some educational and cohort categories;  Hungary: 1990, 2001 and 2011 censuses;  Italy: Family and Social Subjects surveys of 2003 and 2009;  Ireland: 2006 and 2011 censuses;  New Zealand: 2013 census;  Norway: population registers for cohorts born 1940 to 1964 as described in Kravdal and Rindfuss 2008;  Poland: a survey accompanying the 2002 census;  Romania: 1992 and 2002 censuses;  Russia: 2010 census;  Serbia: 1991, 2002 and 2011 censuses;  Singapore: 2010 census; data on children ever born among married women by education and age groups were combined with the data on the family structure of women by age, assuming that all never-married women were childless;  Slovakia: 1991 and 2001 censuses,

9

 Slovenia: 2002 census:  South Korea: 1985-2010 censuses. data on children ever born among married women by education and age groups were combined with the data on the family structure of women by age, assuming that all never-married women were childless;  Spain: 2011 census;  Sweden: 1960 population census and the Multi-generation Register drawn from the Linnaeus Database (see Sandström 2014); for the 1945-59 birth cohorts we used register data as described by Andersson et al. (2009));  Switzerland: 2000 census. The data for Hungary 1990 and 2001 (5 per cent census sample), Romania, and Slovenia (10 per cent census samples) were derived from IPUMS International (Minnesota Population Center 2015). To ensure comparability, we used a three-category classification of education: low, medium, and

• high. In most cases these categories correspond to ISCED-97 levels 0-2, 3-4, and 5-6; respectively (see

UNESCO 2006 for a description of the International Standard Classification of Education). For data collected within the EURREP project, the translation of national school levels to the ISCED-97 levels used is described in the country documentation files to be found at www.cfe-database.org. More information about the data provided by the Cohort Fertility and Education database as available in Zeman et al. (2014). In addition to the quality checks we made during the construction of the database, we undertook specific checks that showed that the general and education-specific trends in fertility, including parity distribution and parity progression ratios, in the countries covered were very consistent. The data for the United States were validated against the National Survey of Family Growth (waves 1995 to 2011- 13) and the Human Fertility Database. The following data issues and adjustments in the data processing for individual countries should be mentioned.  Birth cohorts analysed: data for Slovenia, Australia and New Zealand were given in five-year cohorts, starting from the 1927-31, 1937-67 and 1938-42 birth cohorts and finishing with 1957-61, 1967-71 and 1968-72 ones, respectively. Thus, the analysed five-year birth cohorts for these three countries are one to two years younger than the respective birth cohorts for

• ther countries. For the Nordic countries, the analysed cohorts are by one year older than

five-year cohort categories used for other countries; e.g. the birth cohort 1946-50 encompasses women born between 1945 and 1949.

10

 Specific data sources: data for Norway were taken from Kravdal and Rindfuss (2008); data for Sweden come from two data sources: up to the 1944 birth cohort we used detailed data from the Linnaeus Database (courtesy of Glenn Sandström) and for the birth cohort 1945-59 we based our calculations on the data given in Andersson et al. (2008) and the 2001 Swedish population of women aged 51-55 to 61-65 as published by Eurostat. For Denmark, the available education-specific register-based data covered the parity distribution only up to the parity 3 and above, so in order to compute the total number of children we used the average number of children among women in Denmark with three and more children as given in the Human Fertility Database.  Education groups analysed: in France the ISCED grouping that we adopted did not seem to match the grouping usually displayed in OECD studies: the share of women who were highly educated was half the usual size (OECD 2014). However, the definition and levels displayed were comparable to those of the other countries in this study.  Data pertaining to ever-married women only: for Singapore and South Korea the parity distribution was reported only for ever-married women. Given that non-marital fertility is rare in these countries, we added the never-married women to the population of childless women.  Missing records on education. In the Australian and New Zealand data, there was a considerable proportion of cases with unknown education (around 11%). In New Zealand, the fertility trends in this group indicated that it consisted of very low educated women. Therefore, we recoded the unknown education category into low education. In Australia, we deleted the cases of unknown education as they seemed to be randomly distributed. In the analyses for individual countries, we cover the whole birth cohort range (1916-20 to 1966-70)

• r any of these cohorts available in the source data. In the regional comparisons, we start with the

1936-40 birth cohort and finish with the 1956-60 one, as data for the two oldest and the two youngest five-year cohorts are present only in few countries.