The Impact of the Great Recession on Fertility in Europe

1. The Impact of the Great Recessionon Fertility in Europe Anna Matysiak1 – Tomáš Sobotka1– Daniele Vignoli2 1Wittgenstein Centre (IIASA, VID/ÖAW, WU), Vienna Institute of Demography/Austrian Academy of Sciences 2University of Florence, DiSIA – Department of Statistics, Informatics, Applications

2. Changes in TFR in 2000-12(13),main European regions Source: Own computations based on Eurostat 2013 & national statistical offices

3. Birth timing: accelerated postponement? Changes in age-specific fertility three years before (2005-8) and three years into the recession (2008-11) Source: Own computations based on Eurostat 2013 & national statistical offices

4. Why and how is the recent recession likely to have affected fertility? • Massive unemployment in some countries • Strongly affects young adults, further exacerbates the previous trend of their rising economic and employment uncertainty • Delayed home leaving, econ. independence (Aassve et al. 2012) • Rise in the share of NEETS & workless households • Falling incomes, rise in negative equity on housing (mortgages “under water”), foreclosures (US) • Massive cuts in government budgets, also for family support (double-dip effect on fertility?) • Prolonged duration of the recession; loss of hope in the future (Southern Europe) Source: OECD 2014: Society at a Glance 2014. The crisis and its aftermath

5. Past research • The effect of unemployment is unclear and depends on whether unemployment is measured at individual level or aggregate level • Aggregate level unemployment usually depresses fertility (Simó Noguera et al. 2005, Berkowitz King 2005, Aaberge et al. 2005: 150, Adsera 2005, 2011, Neels et al. 2012, Currie and Schwandt 2014), the effects of individual unemployment are less clear • The effects are sex- and age-specific and differentiated by social status / education (Kreyenfeld 2009, Pailhe and Solaz 2012, Neels et al. 2012, Currie and Schwandt 2014) • Other aggregate-level factors found important in some studies: GDP change, consumer confidence, housing foreclosure rate, self-employment rate, fixed-term contracts

6. Limits of previous research • Only few studies on the effects of the recent recession on fertility in Europe (Goldstein et al. 2011, overview by Eurostat / Lanzieri 2013) • Lack of suitable (panel) data for sound multi-country studies • Little or no use of regional data • US: wider range of suitable surveys & research underway to study wide-ranging effect of the Great Recession on families (e.g., Guzzo 2012, Cherlin et al. 2013)

7. Aims, data, methods

8. Goals Initial aim: Studying the impact of age, parity, education and aggregate-level conditions on first and second births NUTS-2 regions; EU-SILC • Data problems, especially in the recession period (2011)

9. Goals Initial aim: Studying the impact of age, parity, education and aggregate-level conditions on first and second births NUTS-2 regions; EU-SILC • Data problems, especially in the recession period (2011) Revised aim: Using “macro” data in 2000-12 for NUTS-2 regions to study the impact of aggregate-level employment conditions on fertility change • Main contribution: using recent data covering extended period of the recession, using regions as a main unit • Main drawback: losing individual-level dimension.

10. Data • Coverage: 2000-12: EU, Switzerland, Norway; 276 NUTS-2 units • Fertility: Age-specific fertility rates, cumulated into age groups (15-19, 20-24, 25-29, 30-34, 35-49) and Total Fertility Rates • Employment conditions: • Unemployment rates (ages 15-24, 25-64, 20-64), • long-term unemployment (% of unemployed), • % self-employed, • young adults NEETs (not in employment, education, training, age 18-24) • Other variables considered: Regional GDP change (not available > 2010), % with higher education (non-stationary), indicators on poverty, social exclusion (based on EU-SILC, high % missing, unstable);

11. Method • Time series tested for stationarity: • Unemployment rates (UNMP), ASFR(20-24), ASFR(30-34) – first difference stationary • Remaining fertility indicators, Long-term unemployment (LTUNMP), NEETs, Self-employment (SELFEMPL) – level stationary • Random-effects linear regression (regions nested within countries) with a time trend (3 periods) • Dependent variable: ΔTFRt or ΔASFRt • Explanatory variables: • ΔUNMPt-1 for ages 20-64, 15-24 or 25+ • ΔLTUNMPt-1 • Δ NEETt-1 • Δ SELFEMPLt-1

12. Main results

13. All countries & regions combined How a 10 pp. annual increase in • unemployment rate • the share of long-term unemployed • in the % of young adult NEETs • in the % self-employed predicted to change fertility rates?

14. All countries & regions combined Effects on TFR Effects on age-specific fertility Insignificant results (p>0.1) shown by patterned fill

15. Country groups: effects ofunemployment and long-term unemployment Effects of 10pp increase on fertility (TFR & by age)

16. Country groups: effects of NEETs and self-employment Effects of 10pp increase on fertility (TFR & by age)

17. Effect of the period: 2009-12 Additional (unexplained) period effect on the observed TFR: 2009-12 compared with 2005-8

18. “Predicting” TFR change since 2008 How much of the observed TFR change since 2008 “predicted” by our recession indicators?Unemployment + self-employment + NEETS Explaining the difference:Other important factors omitted? Need for an improved model fit?

19. TFR change in Latvia 2007-13 How much of the observed annual TFR change “predicted” by our recession indicators?Unemloyment + self-employment + NEETS The model predicts well fertility reversals, but not the magnitude of changes

20. Conclusions • Clear effect of the recession on fertility • Reflected both in unemployment and less standard proxies of economic uncertainty, also additional negative effect of the period 2009-12 • The role of uncertainty indicators varies by age, country groups / institutional settings (e.g. the importance of NEETs and self-employment in Southern Europe)

21. Future plans • Extend the analysis by introducing country-level covariates and regional deviations from the country levels • Considering other indicators of uncertainty at regional level (e.g. temporaryemployment) • Elaborating the model, trying different specifications & interactions, conducting sensitivity tests

22. A. Matysiak and T. Sobotka’s 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). EURREP website: www.eurrep.org