Boomerang generation in Europe: the effect of the Great Recession

Fernanda Mazzotta and Lavinia Parisi

University of Salerno

Department of Economics and Statistics

Celpe

Abstract The paper analyses the effect of the Great Recession on the decision of returning home for young Europeans aged 20-36. Moreover, we take into account the effect of partnership and employment on the decision to go back home estimating a trivariate probit model, using the panel structure of EU- Silc. The decision of returning home emerges as a new phenomenon during the crisis in Southern and Baltic countries. Given the effect of and separation on the probability of returning home, in those countries the economic crisis worsening labor market condition and intention to marry increasing the probability of return. Residential independence is strongly linked to family formation and dissolution (more than employment status) albeit with differences across welfare regimes.

1

Introduction

The last two decades have seen a continuously changes in family structure and in the pathway to adulthood (the so called Second Demographic Transition - Billari and Liefbroer, 2010). This can be due either to a delayed independence of adult children (i.e. delaying leaving home) or to the ‘boomeranging’ i.e. the situation of young adults going back to the parental home after a period of independence (Jenkins and Allen, 1998). This paper aims to analyze the effect of Great Recession (GR) on the emancipation model in six group of European Countries (namely Continental, Southern, Anglo-Saxon. Scandinavian, Eastern and Baltic countries). In particular, we would analyze the returning home decision of young Europeans (aged 20-36) in the period 2006-2012 (i.e. across the economic crisis) using European Union Statistics on Income and Living Conditions. The hypothesis we aim to test is that the economic crisis would increase the probability of returning home (Dunne, 2012; Eurofound, 2014; Őzdemir et al., 2014; Painter, 2010); in other words we expect to find that after 2009 the probability of returning is higher. The already variegated and complex situations get more difficult for the current economic crisis, which is having a major adverse impact on the economic performance of North America and Europe causing many countries to enter into recession. The economic crisis affects the leaving and returning home decision directly and also through the effect that it has on employment and partnership. In fact, the current economic situation increases the fragility of the school-to-work transition the unemployment and the financial hardship (Plantenga et al. 2013). This is more evident for vulnerable groups in the population (Scarpetta et al. 2010) increasing the rate of co-residence with parents and hence lengthens the transition to adulthood (Bell et al. 2007; Morgan et al. 2011; Mykyta 2012; Painter 2010 ). The crisis has hit the young population very hard: during the recession, has increased disproportionately with respect to the overall unemployment level (Bell et al. 2011; Cho et al. 2012). Declining real earnings and poor employment prospects 1 result both in a “failure to launch” into economic independence (Bell et al. 2007; von Wachter et al. 2013) and also in the ‘boomeranging’, with increasing numbers of young people returning to their parents’ home after leaving (Plantenga et al., 2013). Usually young children leave home to live with a partner just as they are more likely to return after a separation (DaVanzo & Goldscheider 1990). The economic crisis has had very real effects on marital intentions in Europe and in North America. On average across these countries, the difference between those considering the economic crisis made marriage less likely and those not is 10%; the wealth losses engendered by the economic crisis appear to play the most important role in explaining these changes in marital intentions (Schneider 2011). Considering this framework, we argue that the decision of returning to the parental home is driven by employment and partnership status during the transition and other demographic variables usually included in the empirical literature (i.e. education, health, age, personal and parental income). The main hypothesis is that young adults are more likely to return to the parental home if home is perceived to be a strong, reliable base and this perception intensifies during an economic crisis. We consider employment and partnership as endogenous variable when studying emancipation model, in other words they are often jointly determined given that young people are more likely to live independently just as they are likely to have a partner or a job. This hypothesis justifies our research strategy that is a multivariate probit model developed to take into account the fact that employment, partnership and living arrangements are usually simultaneously determined.

1 There is a lot of evidence suggesting that high unemployment or precarious job encourages young people to remain living with their parents (Billari and Liefbroer, 2007, 2010; Blossfeld et al., 2005).

2

Two main results have to be pointed out: the first one is the probability of returning home emerges as a new phenomenon during the crisis in Southern and Baltic countries. Given the effect of unemployment and separation on the probability of returning home, in those countries the economic crisis worsening labor market condition and intention to marry (Plantenga et al., 2013) increasing the probability of return. Finally, residential independence is strongly linked to family formation and dissolution (more than employment status) albeit with differences across welfare regimes. The paper is structured as follows: the next section describes findings of previous studies that provide the motivation for this piece of work. Section 3 defines the sample and the variables of interest. Section 4 provides some descriptive statistics on the observed rate of returning also taking into account employment and partnership status. Section 5 provides the empirical analysis and results. The final section summarizes and concludes, contributing to the economic, sociological and demography literature on household formation and dissolution by focusing on the effect of the economic crisis while controlling for the role of employment and partnership; we provide new insights on the concept of the “boomerang” generation in European Union providing new empirical evidence using panel data.

Empirical and theoretical background

This section aims to highlight the main determinants of returning home focusing on the effect of welfare regimes and time periods. There are several reasons why young people may choose to co-reside with parents as living arrangements 2. The decision to stay at parental home could be driven by institutional and structural factors such as labor market condition (Card & Lemieux 1997) high housing prices (Ermisch 1999, 2003), as well as cultural factors (e.g. Settersten & Ray 2010; Billari 2004). Economic theory (Ermisch (1999; 2003) 3 have emphasized the fact that living in the parental home may increase the utility of both parents and children: on one hand, children may prefer to live in their parental home because of the support provided by their parents; on the other hand, parents may greatly value having children at home longer and so offer transfers to keep their own children at home as long as possible (Manacorda and Moretti, 2006). Economic theory also analyzed the costs associated to co- residence: co-residence may lower marital satisfaction (DaVanzo & Goldscheider 1990) and housing situation (Kluve 2004), there may be also a loss of autonomy and control on ones’ own space and activities, finally a delayed sense of achieved independence for young people (for instance, Italian children value their independence and, everything else equal, prefer to live on their own, Manacorda & Moretti 2006). The decision of co-reside may be a strategy to support kids for the parents, but at the same time young adults may stay at home in order to help to reduce the poverty risk of their parents (Cantó and Mercader-Prats, 2001). With regard to the empirical literature on returning home, studies are rare and often researchers, focusing on return back home, analyze migration 4. Thanks to data availability, several studies can be found for USA and Canada (e.g. Goldscheider & Goldscheider 1999; DaVanzo & Goldscheider 1990; Gee et al. 2003; Mitchell 2006a). European research literature is relatively scarce neither comparative, in fact there are some studies for the UK and Sweden (Kerckhoff & Macrae 1992; Jones 1995; Stone et al. 2014; Nilsson & Strandh 1999), as well as for Germany (Konietzka & Huinink 2003; Konietzka 2010). However, some of key factors to explain the so

2 For a comprehensive review on determinant of home leaving and returning see Burn & Szoeke (2016) 3 Usually theoretical research, following McElroy (1985), adopt a comparative utility framework in which the decision to co-reside depend on the comparison of the indirect utility when parents live with their adult children and when they do not). 4 See Masso et al., chap. 12 in this book, (Galgóczi et al., 2012; Le Mare et al., 2015)

3

called “boomerang generation” can be highlighted. First the enrollment in educational system: it makes young people more likely to leave at younger age as just it makes them more likely to return to the parental home at the end of formal education. Stone et al. (2014) indicate students who get their final degree as a key turning points 5 in determining returning home; also returning is likely to increase as a result of rising levels of student debt 6 and a weaker graduate jobs market (Andrew 2010, Clapham et al. 2012). Another important determinant is the condition of youth in the labor market: a precarious and temporary employment situation can increase the likelihood of returning home (Goldscheider & Goldscheider 1999). Moreover several study underline the importance of the changing in economic activity status (mainly becoming unemployed) that increase the need for intergenerational support, fostering a potential return to parental home (Stone et al., 2012). The same studies find that partnership breakup can often result in the decision to go back to parental home: separation and divorce foster returning to parental home (Mitchell et al. 2000; DaVanzo & Goldscheider 1990; Gee et al. 1995). The association between partnership dissolution and returning home is moderated by gender and parenthood (Stone et al., 2012). Overall, men are more likely than women to return to the parental home upon marital/cohabiting dissolution (Ongaro et al. 2009; Sullivan 1986). One explanation for this is that women are more likely to be responsible for any dependent children, and the presence of children will tend to increase the likelihood that women will stay in the home previously shared by the couple (Gram-Hanssen & Bech-Danielsen 2008; Mulder & Wagner 2010). From a psychological perspective, researchers have highlighted some of the important implications of returning home such as the risk of conflict between parents and children, opportunities for shared leisure time (Aquilino and Supple, 1991),consumption-smoothing mechanism for weathering periods of financial stress (Dettling and Hsu, 2014), the negotiation of adult roles and identities, the extent to which the is able to interact with their parent(s) from a position of equality rather than dependence (Sassler et al., 2008) but also reflecting personal and collective failure and creating deep frustration. This paper contributes to the existing literature providing barely new results on returning home especially analyzing the phenomenon in a dynamic and comparative perspective (completely new for returning home in European Union) and above all before and after the Great Recession with panel data covering a period between 2006-2013. There is a lively debate about the causes of different pattern of late home-leaving across Europe and periods. With regards to difference across countries previous finding highlight broad, geographically based patterns in the transition to adulthood (Cavalli & Galland, 1993). Some (Jones, 1995; Reher, 1998) define Southern and Eastern European cultural roots as reasons for the late home-leaving and the strong synchronization between leaving and first marriage. Others have emphasized the importance of worse economic conditions for young adults in Southern Europe (Saraceno, 2015). Esping Andersen (1999) focused on the peculiarities of the Southern European welfare state and on the absence of support for young people who are unemployed in those countries and for the role played by the family in helping them (sometime southern countries are also defined familistic countries). On the other side Northern and Continental Europe emerge as countries where a low percentage of young people live with their parents and time spent living as a single is prolonged. Finally, Cavalli and Galland (1993) identify also a British pattern, with early transitions from school to work and delayed – but clustered – household and family formation. As a result, following the aggregation suggested by the European Commission (EC 2006 and 2007) modified by Boeri (2001) we group countries in Continental (Austria, Belgium, France and Luxembourg), Southern European (Spain, Italy, Greece and Portugal) Eastern (Czech Republic,

5 The “turning points” is a key concept within life course theory, referring to an event, an experience, or a change in circumstances that significantly alters the individual’s subsequent life course trajectory (Stone et al., 2012) 6 Dettling and Hsu (2014) show that debt is an important determinant of co-residence, even after controlling for labor and housing market conditions, shedding light on the fact that parental co-residence is used to smooth utility when young adults have exhausted their ability or willingness to borrow.

4

Poland, Bulgaria and Hungary) Baltic (Estonia, Latvia and Lithuania), Nordic (Norwey Finland, Denmark and Island) and Anglo-Saxon (United Kingdom, Cyprus,) countries. 7

With regard to whether there are differences on returning across periods, and in particular during the Great Recession, previous studies have stresses the fact that the economic crisis affects the emancipation model mainly through employment and wealth loss. Youth unemployment has increased disproportionately with respect to the overall unemployment level during the Great Recession (Bell et al. 2011; Cho et al. 2012). In particular the crisis has worsened the labor market conditions more for young men (particularly those aged 15-24) than for young women, thus reducing the existing gender gaps (Plantenga et al., 2013 p. 34). This increases the rate of co- residence with parents and hence it lengthens the transition to adulthood (Bell et al. 2007; Morgan et al. 2011; Mykyta 2012; Painter 2010). Eurofound (2014) states that the strong decrease in employment security and availability of jobs is likely to delay the transitions in and out parental home even further. Őzdemir et al.(2014) find that in countries where jobs for young people were particularly scarce, (i.e. Ireland, Cyprus, Spain and Greece) the proportion of young people aged 18-24 living with parents went up significantly, they also find that in Baltic countries (together with Slovenia) the proportion of people living independently from their parents increased, despite the labor marked deterioration. However Őzdemir et al. (2014) also show that the relative number living with parents also increases in country less affected by the economic crisis (i.e. France, Austria, Luxemburg and Germany). One of the reason could be the decrease of young student living outside the family or in student housing (Ward et al., 2012). In addition, the wealth losses engendered by the economic crisis appear to play the most important role in explaining the reduction in marital intentions (Schneider 2011), thus the decision of leaving home may also reduce for this reason. A number of studies found that the percentage of young people in parental home increases in the early years of the crisis (Aassve et al., 2013; Eurofound, 2014; Ward et al., 2012), with young men, in particular, being more likely to do so than young women. There is, in addition, evidence of a ‘boomerang’ phenomenon in France, Spain and the UK, with increasing numbers of young people returning to their parents’ home after leaving during the crisis (Plantenga et al., 2013).

Data and sample selection

The data used in this chapter are drawn from Eurostat, European Union Statistics on Income and Living Conditions (henceforth EU-SILC). 8 We select the fourth database up to the eighth, for a period that covers from the year 2006 to the year 2013. For each panel we considered couples of years. For instance, for the panel 2006-2009 we have 3 sub-samples: 2006-2007, 2007-2008, 2008- 2009. Thus, when talking about year 2006 we consider changes occurring between 2006–2007, as just year 2012 considers changes occurring between 2012 and 2013. The sample consists of young people aged 20-36 years old the first time they are observed. To improve interpretation of results we group countries according to their socio-economic system. Esping-Andersen (1990) carried out the seminal work on the taxonomy of socio-economic systems 9, a lot classifications developed since that one (including the present). Several studies (Muffels wet al. (2002), Wilthagen (2004), Auer (2005), and the European Commission (EiE 2006,

7 Different hypothesis on which group of countries include them could be made for some European countries i.e. Malta and Cyprus. We follow Stovicek and Turrini (2012: 9-10) which place Malta and Cyprus to the Anglo-Saxon group on the basis of unemployment benefit. 8 The responsibility for all conclusions drawn from the data lies entirely with the authors. 9 Esping-Andersen classified the welfare systems of developed economies into three models: Liberal, Conservative and Social Democratic.

5

Chapter 2)a< covering a wide range of subjects (such as welfare, labor market, innovation and healthcare) seem to confirm the original taxonomy. For the purpose of this analysis it is important to focus on classification based on flexibility, work security and the diffusion of unemployment benefit which seem linked to good and bad performance in the labour market, which can in turn influence the decision to leave and returning to parental home specially during an economic crisis. As a consequence, we follow the aggregation suggested by the European Commission (2006 and 2007) modified by Boeri (who include Greece among the Mediterranean countries and not in the Eastern countries). Moreover, this classification has been largely confirmed by the findings of Eurofound (2014) about the different dynamics among the EU member states in young people’s transition to adulthood. We selected 21 countries which reflect the peculiarities of 6 welfare regimes: Continental welfare regimes which provide benefits targeted to individuals who belong to specific categories, such as families or a specific type of worker (Austria, Belgium, France and Luxembourg). Southern European regime (Spain, Italy, Greece and Portugal) where welfare coverage is often residual and left to the family, with limited social benefits (Billari and Liefbroer, 2010), Ferrera, 1996), Nordic (Norway Finland, Denmark and Island), Anglo-Saxon (United Kingdom, Cyprus), Eastern (Czech Republic, Poland, Bulgaria and Hungary) and finally Baltic regimes (Estonia, Latvia and Lithuania). 10

Descriptive statistics

The method to analyze returning to the parental home is itself a contribution to the empirical literature given it is dynamic exploiting the nature of panel data, i.e. the possibility to observe the same person for up to four consecutive years. In this section we provide some descriptive statistics on key variables focusing on the role of the economic crisis across welfare regime. The first variable we consider is the observed probability of Returning Home (R). R describes whether young Europeans, who were living on their own at time t, are still living without their parents at t+1. If the individual is not living with his\her parents at time t+1 we know whether he/she has returned in the parental home or his\her parents have returned home or whether he/she is not in the panel anymore (attrition). Also in this case we distinguish those that at t+1 are returning in the parental home with a partner or not, without a job or with a job. With regard to the observed rate of returning to the parental home (Table 1) all the European Countries show very low rates of returning to the parental home (around 1% on average). We find different reaction to the economic crisis according to welfare regime. While in Continental and Anglo-Saxon countries we observe an increase already at the very beginning of the economic downturn (2009-2010, from around 0.5% to 1.3%) in Southern European countries the increase is postponed to 2011-2012 with a jump in 2011-2012 (2%). However, overall in the very last period (2012 – 2013) the rate of returning decrease for all the welfare system expect for Anglo-Saxon and Southern countries (Table 1). 11

Table 1 Observed rate of returning to parental home for young people aged 20-36 by welfare.

10 European Commission (2006 and 2007) include the Baltic countries together with the European countries, we decide to separate them here because according to the family formation these countries show different pattern compare to Easter, more closest with Continental countries with 50% of young people leave home in line with the EU average, only to move in very quickly with a partner and then become parents (Eurofound, 2014) 11 Between 2011 and 2012, the decreasing is significant different from zero at 1% for Continental and Southern and for the total sample .

6

Period of Anglo- transition Nordic Saxon Continental Southern Eastern Baltic Totale 2006-2007 0.98% 0.50% 0.99% 1.05% 0.65% 0.90% 0.91% 2007-2008 0.91% 0.27% 0.85% 1.41% 0.88% 0.64% 0.91% 2008-2009 0.65% 0.53% 0.56% 1.42% 0.62% 1.70% 0.75% 2009-2010 0.35% 1.28% 1.34% 1.78% 0.67% 1.54% 1.32% 2010-2011 1.48% 0.70% 0.72% 1.36% 0.69% 0.96% 0.86% 2011-2012 0.45% 0.60% 1.28% 2.08% 0.50% 0.57% 1.21% 2012-2013 0.85% 1.33% 0.41% 1.74% 0.23% 0.48% 0.80% Totale 0.82% 0.71% 0.91% 1.56% 0.64% 1.03% 0.98% N.of Obs. 25,442 25,142 87,818 73,865 82,235 27,024 321,526 Note: Percentage are calculated each years as the number of young people who have returned home at time t+1 divide for the total number of young people (including students) living independently at time t.

As we have already seen in the previous paragraphs, when studying the effect of the GR on leaving and returning home decision we should not disregard both the decision to form a family and the probability of finding a job. Consequently, figures 2a and 2b reveal the patterns of employment and partnership distinguished between those who did not return (that we labeled alone or living independently) and those who returned home. Looking at Figure 2a it seems that being not employed is more associated to time periods than to living arrangements. In fact, even though individuals who return home are on average more likely to not be employed compare to the one who live independently 12 we find a very huge difference among time periods. There is a very low proportion of people not-employed at the beginning of the period (with no differences between living arrangements except for Continental countries) while a very big proportion of people not employed in the last periods above all for those who return home in Baltic, Continental, Nordic and Southern countries (around 20 for the first two and around 60% for the latter).

Figure 2a: Share of young people not employed at time t+1 distinguishing between those who lived independently (top part) and those who returned home (bottom part) in the period under considerations and by welfare regimes.

12 Differences statistically significant at 1% level.

7

Nordic Anglo-Saxon Continental

2006 2006 2006 2007 2007 2007 2008 2008 2008 0 2009 0 2009 0 2009 2010 2010 2010 2011 2011 2011 2012 2012 2012 2006 2006 2006 2007 2007 2007 2008 2008 2008 1 2009 1 2009 1 2009 2010 2010 2010 2011 2011 2011 2012 2012 2012

Southern Eastern Baltic

2006 2006 2006 2007 2007 2007 2008 2008 2008 0 2009 0 2009 0 2009 2010 2010 2010 2011 2011 2011 2012 2012 2012 2006 2006 2006 2007 2007 2007 2008 2008 2008 1 2009 1 2009 1 2009 2010 2010 2010 2011 2011 2011 2012 2012 2012

0 .2 .4 .6 .8 0 .2 .4 .6 .8 0 .2 .4 .6 .8 Share of young people not employed at t+1

On the contrary, the effect of marital or partnership dissolution is statistically significant for almost the countries (Eastern being the exception): young people without a partner return home more often. A very high share of young people (around 90% in some countries) return home without having a partner. This pattern is less strong for Eastern countries, where (in line with Iacovou, 2010) young people are found living with a partner, in the same house as their parents. Overall, the differences in the percentage of those returning with or without a job are less strong compare to the difference between being or not being in a partnership. It seems that being in a partnership is more important than having a job on the chances of returning back home. Across period, there are nor clear neither significant patterns in Continental and Baltic countries, whereas in Southern countries the percentage of dissolution for those who return to parental home increases while the opposite is true in Eastern countries (these effects are statistical significant).

8

Figure 2b Share of young people not in a partnership at time t+1 distinguishing between those who lived independently (top part) and those who returned home (bottom part) in the period under considerations and by welfare regimes.

Nordic Anglo-Saxon Continental

2006 2006 2006 2007 2007 2007 2008 2008 2008 0 2009 0 2009 0 2009 2010 2010 2010 2011 2011 2011 2012 2012 2012 2006 2006 2006 2007 2007 2007 2008 2008 2008 1 2009 1 2009 1 2009 2010 2010 2010 2011 2011 2011 2012 2012 2012

Southern Eastern Baltic

2006 2006 2006 2007 2007 2007 2008 2008 2008 0 2009 0 2009 0 2009 2010 2010 2010 2011 2011 2011 2012 2012 2012 2006 2006 2006 2007 2007 2007 2008 2008 2008 1 2009 1 2009 1 2009 2010 2010 2010 2011 2011 2011 2012 2012 2012

0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1 Share of young people not in a partnership at t+1

Econometrics analysis: methodology

The method used to estimate both the probability of returning home is a trivariate probit model. It is a simulation method to maximum likelihood estimation of the multivariate probit regression model. The model controls for unobservable factors that influence both the probability to return home, to not be employed and to not be in a couple. Not considering the mutual correlation between the three outcomes could produce biased results 13 . Moreover, it is a type of first-order Markov approach. It takes into account pairs of observations in two consecutive years namely t and t + 1 for each individual. In year t the young person lives independently (without parents) and in years t + 1 she\he has returned at home. The model estimates simultaneously the probability of returning home (R t+1 ), the probability of not being in partnership (UP t+1 ), and the probability of not being employed (NE t+1 ). We include the following variable to identify the model: the crowding index and whether having children at time t+1 in equation R t+1 , whether having children and not employment status at t+1 in equation UP t+1 , not be employed at time t in equation NE t+1. The variables of main interest are: the probability of not being employed at t+1, the probability to not be in a relationship at t+1 and time dummies to disentangle the effect of the crisis. Other control variables Xs (i.e. gender - whether

13 The maximum likelihood estimates of the implied trinomial probit model differ sharply from those obtained when either market work or household membership is taken as exogenous (McElroy, 1985).

9

male; age and age squared; education - two dummies for secondary education and tertiary education, reference category is compulsory education; general health status - whether he\she has good health; tenure status – two dummies indicating whether owner or tenant with reference category being reduced rent or for free; degree of urbanization – whether living in a thinly populated area respect to densely populated area; dwelling type – two dummies indicating whether accommodation has more than 10 dwellings or less than 10 dwellings, reference category is whether it is detached and semi-detached house; financial burden of total housing cost – two dummies indicating whether the burden is heavy or medium reference category is not a burden; financial burden of the repayment of debt from hire purchases or loans – two dummies indicating whether the burden is heavy or medium reference category is not a burden ) have been chosen according to the literature. The model is written as follows:

* R it+1 = α 0 + α1YearDum + α2Hasn’tEmploy t+1 + α 3Hasn’tPartner t+1 + α 4CrowdIndex t + α5IsParent t + α5CountryDum + α6Xs + ei (1)

* I UP it+1 = γ0 + γ 1YearDum + γ 2Hasn’tEmploy t+1 + γ 3 sParent t + γ 4 Xs + u i (2)

* NE it+1 = µ 0 + µ 1YearDum + µ 2Hasn’tEmploy t + µ 3Xs + v i (3)

where Rit+1 = 1 if (R *it+1 > 0), Rit+1 = 0 otherwise (4) UPit+1 = 1 if (U *it+1 > 0), Uit+1 = 0 otherwise (5) NEit+1 = 1 if (NE *it+1 > 0), Eit+1 = 0 otherwise (6)

Corr(e, u, v) = ρ (7) Corr(e, u) = ρ21 (7a) Corr(e, v) = ρ31 (7b) Corr(u,v) = ρ32 (7c)

The three outcomes are allowed to be freely correlated. The trivariate normal distribution for the three outcomes that are allowed to be freely correlated has zero mean and σ2 variance. Correlations relate to unobservable traits such as ability, intelligence, personality traits, ambition, quality of the relationship with parents, family background and so on. We estimate the correlation among the three error terms: if positive unobservable individual factors determining the outcome of primary interest (i.e. leaving or returning) are also positively associated to the other two (being employed and having a partner for leaving and not employed and not in a partnership for returning) ρ21 summarises the association between unobservable individual factors determining returning home and not being in partnership. If ρ 21 is positive (negative) it means that those individuals more likely to be return home are also more (less) likely to not be in a relationship. Furthermore, ρ 31 accounts for unobserved heterogeneity between not being employed and returning home. When positive (negative) it means that unobservables that make young people more likely to not be in employment make them more (less) likely to return. And, finally, ρ 32 relates unobserved heterogeneity between not employment and not in a partnership.

Results and discussion

This section presents and discusses results, the model is estimated separately for six different welfare regimes. We present estimates that include country dummies (within each welfare - not reported) and year dummies.

10

Table 2 presents estimates for a trivariate probit model for the probability of returning home, while Figures 3a plot the marginal predicted probabilities across time periods, and Figure 3b the joint predicted probabilities for the three outcomes being equal to 1. Seven columns represent results for pooled model and 6 different welfare regimes (country dummies are included accounting for variations within each regime). To disentangle the effect of the economic crisis, we include year dummies excluding 2009 and we calculate predicted probabilities for each year plotted in Figures 3a and 3b The correlation between the error terms 14 (ρ) is significantly different from zero, thus the three equation are strongly related: the same unobservable factors affect positively the probability of returning, of not being employed and of not having a relationship. This indicates that a trivariate probit technique is appropriate in this context.

Looking at the dummies that explain the difference between time periods, we find that there is a time effect only in Continental and Anglo-Saxon countries: the probability of returning home is always lower with respect to 2009. Differently from Stone et al. (2012) the period is still statistically significant when we include what they call the turning points (such as separation or unemployment), meaning that the economic crisis have a direct effect on returning home given it produces uncertainty about the future. However, when we plot predicted marginal probability of returning home (Figure 3a) we observe that also in Southern countries the probability of returning home constantly increased for all the period under consideration. We can argue that this increase observed already in 2007 (before the crisis) may be due to structural or cultural aspects and not entirely to the Great Recession. Conversely, Continental, Anglo-Saxon and Nordic countries have the lowest percentage of individuals returning home with a jump respectively in 2009-2010 (for the first two groups of countries) and 2010-2011 (for the latter). Eastern countries has almost stable and somewhat decreasing percentage of individuals returning home while it decreases during the crisis for Baltic countries. The low rate of returning, (specially in Nordic and Continental countries) has been related to relatively generous welfare state benefits and to cultural factors given that both young people and their parents value greatly independence compare to their southern counterparts (Iacovou, 2010) . Just as partnership had a strong effect on leaving home, separation has a very strong effect on returning home: the more likely young people are to be without a partner, the more likely they are to return home. Union dissolution is a key determinant of returning home. With regards to control variables, except for Nordic, Continental and Eastern countries, not being employed increases the probability to return. However, parents (both mothers and fathers) are less likely to return to the parental home than individuals without children. The parenthood decrease the probability to return back home given that the presence of children will tend to increase the likelihood to stay in the marital home (Gram-Hanssen and Bech-Danielsen 2008; Stone et al., 2014) even if the children are no longer co-resident with their parents (Mulder and Wagner 2010. In the same way not being alone and living in crowded family (usually with more than one children or other relatives) reduce the likely to go back to their parental home. Men are more likely to return home in all the countries excluded Anglo–Saxon countries, this results confirm the fact that the crisis has worsened the labor market conditions more for young men (particularly those aged 15-24) than for young women, thus reducing the existing gender gaps (Stone et al., 2012). In fact, in 2008, as the crisis hit mainly male-dominated sectors (such as the construction sector) in some countries (as Belgium, Spain, Finland and Uk) the male youth unemployment becoming, for the first time, higher than female youth unemployment. From 2012, as the crisis started to bite again, male youth unemployment, once more, increased more sharply

14 Accordingly the overall likelihood ratio test of rho21 = rho31 = rho32 = 0, is always not accepted with Prob > chi2 = 0.0000.

11

than the female rate. Then more difficulty in the labour market for young men, could explain more returning. The incidence of returning drops with age with a minimum 32-33 yrs old for Southern and Baltic countries and 29 yrs old for Eastern countries. Higher educated people are less likely to return home in almost all countries, this is because those youth have higher probability to be employed in high wage job than lower educated people. Education is not significant for Southern while the opposite is true for Anglo–Saxon countries where other financial and affordability constraints, such as the cost of higher education and the associated student debts, will be a further barrier to self-sufficiency of young higher educated adult in the English speaking countries As has been found in previous studies, young individual in Nordic and Anglo-Saxon countries that are students at time t are more likely to return home at time t+1, this is also due to the fact that in those countries there is a high percentage of youth leaving for studying respect to the rest of Europe. This result is also in line with the findings that returning to the parental home at the end of formal education is likely to increase as a result of rising levels of student debt and a weaker graduate jobs market (Andrew 2010, Clapham et al. 2012).. Health does not have any effect on the probability that a young person returns home. For Nordic, Anglo –Saxon and Continental countries we register a positive relation between the probability to return and the burden of the housing cost. Presumably, the student debt together with a weaker graduate jobs market and with increased borrowing constraints imposed on first-time buyers by lenders, could lead to delayed transitions into first home-ownership by graduates (Andrew, 2010) accelerating the returning back home. In these countries we do not find any effect of tenure status variables thus there is not difference among ownership, tenant at market rate and tenant at reduced rate or free. In Southern and Eastern countries the burden of other debts is positively correlated with the probability to return and then co-residence is used to smooth utility when young adults have exhausted their ability or willingness to borrow (Dettling and Hsu, 2014). Moreover tenants at reduced rate in Southern and Baltic countries have a lower probability to return respect the home owner that maybe suffer more of other expenses. On the contrary in the Continental are the tenant at reduced rate to return more than home owner.

Table 2: Trivariate probit model for probability of returning home by welfare regimes 1)Probability of Returning Home Pooled Nordic Anglo_saxon Continental Southern Eastern Baltic 2006-2007 -0.178*** 0.019 -0.485** -0.248** -0.098 -0.05 -0.166 2007-2008 -0.167*** 0.068 -0.552*** -0.278*** 0.024 0.056 -0.236* 2008-2009 -0.213*** -0.021 -0.259* -0.404*** 0.013 -0.041 0.164* 2010-2011 -0.176*** 0.494*** -0.305** -0.351*** -0.022 0.001 -0.086 2011-2012 -0.027 0.081 -0.184 -0.053 0.126* -0.143* -0.285*** 2012-2013 -0.197*** 0.387* 0.032 -0.518*** 0.026 -0.246** -0.315** Male 0.097*** 0.001 -0.028 0.134** 0.087** 0.058 0.262*** Age at t -0.207*** -0.196 -0.032 -0.186** -0.190*** -0.172*** -0.199** Age squared at t 0.003*** 0.003 0 0.002 0.003** 0.003*** 0.003* Tertiary education at t -0.238*** -0.529*** 0.468* -0.407*** -0.094 -0.251*** -0.305** Secondary education at t -0.063 -0.082 0.528** -0.176** -0.023 -0.116* -0.013 Parenthood -0.285*** 0.117 -0.619*** -0.207** -0.377*** -0.099 -0.206*** Good health at t -0.032 0.009 -0.087 -0.007 -0.046 0.027 0.114 House crowded at t -0.152*** 0.016 -0.166** -0.109*** -0.215*** -0.197*** -0.207*** Not in partenship at t+1 0.922*** 0.794*** 1.124*** 1.040*** 1.195*** 0.701*** 0.892*** 12

Not in employed at t+1 0.024 -0.112 -0.314* 0.052 0.258*** -0.053 0.201*** Student at t 0.024 0.361** 0.455*** -0.108 0.098 0.017 -0.459*** Owner tenure status -0.029 -0.051 0.196 -0.194*** 0.166*** -0.012 0.258*** Tenant at market rate -0.049 0.114 0.246 -0.107 0.004 -0.011 -0.265* Thinly populated area -0.038 -0.145 -0.332 -0.092 0.069 0.004 -0.292*** apartment or flat in a building -0.258*** -0.497*** -0.686*** -0.258*** -0.064 -0.331*** -0.353*** Housing cost burden – heavy 0.111** 0.357** 0.851*** 0.032 0.151 0.101 0.01 Housing cost buden – medium 0.076 -0.043 0.440*** 0.134** 0.087 0.05 -0.189** Debt repayment – heavy 0.043 -0.339* -0.194 -0.09 0.214*** 0.133** -0.036 Debt repayment – medium 0.173*** 0.288*** -0.176 0.254*** 0.178*** 0.142** -0.226** Constant 1.441** 1.065 -2.304 1.549 0.432 0.405 1.007 2)Probability of not being in partnership at t+1 2006-2007 -0.034 -0.016 -0.200*** -0.001 -0.067** 0.046 -0.063 2007-2008 -0.033** 0.034 -0.170*** 0.004 -0.102*** 0.04 -0.037 2008-2009 -0.039*** 0.024 -0.206*** -0.009 -0.058*** 0 -0.049 2010-2011 0.015 -0.076* -0.156*** 0.064*** 0.035 -0.028 -0.043 2011-2012 -0.019 -0.171*** -0.207*** 0.003 0.089*** -0.041 -0.013 2012-2013 -0.022 -0.068 -0.243*** 0.004 0.075** -0.027 0.03 Male -0.033** 0.06 -0.032 -0.048* 0.105*** -0.191*** -0.326*** Age at t -0.312*** -0.181*** -0.348*** -0.277*** -0.507*** -0.494*** -0.375*** Age squared at t 0.005*** 0.003*** 0.006*** 0.005*** 0.008*** 0.008*** 0.007*** Tertiary education at t -0.111*** -0.195*** -0.329*** -0.169*** 0.085*** -0.087** -0.229*** Secondary education at t 0.005 -0.108* -0.250*** -0.038 0.115*** -0.062 -0.066 Parenthood -1.271*** -1.475*** -0.840*** -1.309*** -1.430*** -1.464*** -1.260*** Good health at t -0.201*** -0.212*** -0.212*** -0.217*** -0.125*** -0.146*** -0.075* Not in employed a t+1 0.340*** 0.549*** 0.834*** 0.374*** 0.041 0.154*** 0.081 Constant 4.429*** 2.778*** 5.167*** 3.925*** 7.770*** 7.103*** 5.448*** 3)Probability of not being employed at t+1 2006-2007 -0.103*** -0.097 -0.061 -0.099** -0.164*** -0.052* -0.406*** 2007-2008 -0.098*** -0.161*** 0.002 -0.083** -0.156*** -0.090*** -0.291*** 2008-2009 0.047** 0.007 0.189*** 0.042 0.04 -0.015 0.013 2010-2011 -0.066*** 0.071 -0.024 -0.154*** 0.016 0.061** -0.169*** 2011-2012 0.042** 0.212*** 0.197*** -0.034 0.113*** 0.063** -0.141*** 2012-2013 0.005 0.039 0.024 -0.035 0.085** 0.021 -0.141*** Male -0.417*** -0.263*** -0.483*** -0.358*** -0.436*** -0.644*** -0.261*** Age at t -0.084*** -0.283*** 0.007 -0.038 -0.226*** -0.085*** -0.116*** Age squared at t 0.001*** 0.004*** 0 0 0.003*** 0.001** 0.002*** Tertiary education at t -0.499*** -0.339*** -0.575*** -0.499*** -0.448*** -0.662*** -0.567*** Secondary education at t -0.242*** -0.077 -0.278*** -0.256*** -0.212*** -0.347*** -0.263*** Good health at t -0.231*** -0.355*** -0.375*** -0.253*** -0.103*** -0.242*** -0.191*** Not in employed at t 1.928*** 1.640*** 1.934*** 1.970*** 1.892*** 1.882*** 1.614*** Constant 0.669*** 3.892*** -0.603 0.158 2.983*** 1.245*** 1.475*** N. of Observations 321,526 25,442 25,142 87.818 73,865 82,235 27,024 Log likelihood -1.43E+09 -5.04E+07 -1.53E+08 -7.37E+08 -2.97E+08 -1.55E+08 -2.07E+07 Rho21 0.028** -0.024 -0.013 -0.033 0.013 0.018 -0.043

13

Rho31 0.031** -0.001 -0.034 0.02 -0.017 0.007 0.004 Rho32 -0.079*** -0.211*** -0.333*** -0.056*** -0.032** -0.092*** -0.056** Note: Coefficients are presented in the table; Nordic include: Finland, Norway, Denmark and Island, Anglo- Saxon include: United Kingdom and Cypro: Continental countries include Austria, France, Belgium and Luxemburg; Southern countries include Spain, Italy and Portugal, Eastern countries include Czech Republic, Poland and Slovenia; Baltic countries include Estonia, Latvia, Lithuania. Likelihood ratio test for the hypothesis: rho21=rho31=rho32=0 is statistically different from zero at 1% level. P-value at ***1%**5%*10%.

Figure 3a: Predicted probabilities of returning home by welfare regimes and across periods. .02 .015 .01 .005 0

2006-2007 2007-2008 2008-2009 2009-2010 2010-2011 2011-2012 2012-2013

Nordic Anglo Saxon Continental

Southern Eastern Baltic

Note: Bands around lines show confidence intervals at 95%.

With regard to the joint predicted probability (Figure 13b) we can see that the percentage of people that have returned home, are not employed and have not a partner increases during the period under consideration for all welfare regimes except for Eastern European; Anglo-Saxon and Continental countries. The sharpest increase has to be found in Baltic countries immediately after 2008 and in Southern Countries throughout all the period under consideration.

Figure 3b: Joint probabilities of returning home, not being employed and not being in partnership by welfare regimes and across periods.

14

.004 .003 .002 .001 0

2006-2007 2007-2008 2008-2009 2009-2010 2010-2011 2011-2012 2012-2013

Nordic Anglo Saxon Continental

Southern Eastern Baltic

Note: Bands around lines show confidence intervals at 95%.

Conclusion

The chapter examines whether the Great Recession have influenced the probability of returning to the parental home in Europe. It, particularly, did it controlling for both the employment and the partnership of young people. We, in fact, analyze whether the economic crisis has affected the emancipation model (i.e. the patterns of returning home) directly and also through the effect it has on employment and partnership. The shares of young people that live without their parents exhibit two distinct characteristics. First, in all country groups the share of young individuals that live without their parents has decreased over time. The decrease accelerates in the period 2009-2010, particularly in Continental European countries, although from sensibly lower levels. Second, there are significant differences across country groups: higher in Continental countries, lower in Southern and Eastern countries. Across all welfare regimes the share of co-residence increases with the crisis and declines afterwards, but remains above the pre-crisis levels. In Southern, Eastern and Baltic countries there is a strong cultural root of emancipation where young people are hesitant to live independently. Continental European countries characterized by higher level of people living independently before the economic crisis, have shown an increase because of the fragility of the school-to-work transition and the unemployment and financial hardship may have affect the uncertainty of the economic situation lowering the probability to leave. Moreover, Eurofound (2014) shows that all the European countries experience a rise in youth unemployment rate (except Germany), given that continental countries have an intermediate level of flexibility with high level of employment protection this may contribute to not recover completely with the share of young people leaving home decreasing after the crisis, remaining below the pre-crisis levels afterwards. 15

All countries experienced an increase of the percentage of people returning home, apart from Eastern and with some exception about timing. Southern European countries have registered an increase throughout all the period, while Continental and Anglo-Saxon only in the very first period after the crisis while Nordic countries the effect was postponed at 2011. Returning home emerges as a new phenomenon in Southern countries after the economic crisis. In southern Europe, young persons are less often entitled to welfare benefits and living with the parents is probably more socially acceptable. Results on the effect of the Great Recession hold also after controlling for partnership and employment. The first has a strong effect on both the probability of returning home. Young not (or no longer) in a partnership are more likely to return home, while losing a job increases the probability of returning home only in Continental and Southern countries (the result holds also if we exclude inactive). The consequences of not being employed on the opportunity for young people to establish their own families and economic independence is critical to understand the barriers encountered in the transition to independent adulthood. In Continental and Southern countries not having an employment play a role in explaining the probability of returning home. Adulthood transitions is highly variable among European countries. Family support economically and emotionally young people for the transitions to adult life. However formal sources of help should complement informal ones. In particular, affordable housing, local housing allowances, social and unemployment benefits that support financial need in period of crisis, formal care should all be available for young people to foster emancipation. Some authors (Jones 1995) have also underlined the importance of temporary “respite care” for both parents and children as a mean of supporting young people and their parents during periods of conflicts when co-reside. The availability of such kind of support together with the availability of council and local housing could also help people already living independently in period of financial distress to make them less likely to return back home. Even though it seems that still there are cultural root implying late home leaving in some countries and thus postponing emancipation, specific attention should devote to income support and housing policies to try to change such trends. Late home leaving as great negative consequences also on fertility rate, as just returning home can have particular consequences from a psychological point of view, such as the risk of conflict between parents and children, negotiation of adult roles and identities, and also reflecting personal and collective failure creating deep frustration.

Reference Aassve, A., Cottini, E., and Vitali, A. (2013). Youth Vulnerability in Europe during the Great Recession ( No. 57). DONDENA WORKING PAPERS. Milano. Aquilino, W. S., and Supple, K. R. (1991). Parent-Child Relations and Parent’s Satisfaction with Living Arrangements When Adult Children Live at Home. Journal of Marriage and Family , Vol. 53 (1), pp. 13–27. Billari, F. C., and Liefbroer, A. C. (2007). Should I Stay or Should I Go? The Impact of Age Norms on Leaving Home. Demography , 44 (1), 181–198. Springer-Verlag. doi: 10.1353/dem.2007.0000 Billari, F. C., and Liefbroer, A. C. (2010). Towards a new pattern of transition to adulthood? Advances in Life Course Research , 15 , 59–75. doi: 10.1016/j.alcr.2010.10.003 Blossfeld, H.-P., Klijzing, E., Mills, M., and Kurz, K. (2005). , Uncertainty and Youth in Society . (H.-P. Blossfeld, E. Klijzing, M. Mills, and K. Kurz, Eds.). London : Routledge, 2005. Burn, K., and Szoeke, C. (2016). Boomerang families and failure-to-launch: Commentary on adult children living at home. Maturitas , 83 , 9–12. doi: 10.1016/j.maturitas.2015.09.004

16

Cantó, O., and Mercader-Prats, M. (2001). The Dynamics of Child Poverty in Industrialised Countries . (B. Bradbury, S. P. Jenkins, and J. Micklewright, Eds.). Cambridge: Cambridge University Press. doi: 10.1017/CBO9780511522147 Dettling, L. J., and Hsu, J. W. (2014). Returning to the Nest: Debt and Parental Co-residence Among Young Adults . Finance and Economics Discussion Series. Dunne, T. (2012). Household Formation and the Great Recession ( No. 12). Clevelend. doi: issn 04281276 Esping-Andersen, G. (1999). Social Foundations of Postindustrial Economies . (Oxford University Press, Ed.). Oxford New York. Retrieved from http://www.oupcanada.com/catalog/9780198742005.html Eurofound. (2014). Mapping youth transitions in Europe . Luxembourg. Galgóczi, B., Leschke, J., and Watt, A. (2012). Eu Labour Migration in Troubled Times: Skills Mismatch, Return and Policy Responses . Ashgate Publishing. Iacovou, M. (2010). Leaving Home: Independence, togetherness and income in Europe. Advances in Life Course Research, 15 (4), 147–160. Jenkins, H., and Allen, C. (1998). Relationship between staff burnout/distress and interactions with residents in two residential homes for older people. Int J Geriatr Psychiatry , 13 (No. 7), 466– 472. Jones, G. (1995). Leaving home . Open University Press. Retrieved from https://books.google.co.uk/books/about/Leaving_home.html?id=NawoAAAAYAAJ&pgis=1 Manacorda, M., and Moretti, E. (2006). Why do Most Italian Youths Live with Their Parents? Intergenerational Transfers and Household Structure. Journal of the European Economic Association , 4(4), 800–829. doi: 10.1162/JEEA.2006.4.4.800 Le Mare, A., Promphaking, B., and Rigg, J. (2015). Returning Home: The Middle-Income Trap and Gendered Norms in Thailand. Journal of International Development , 27 (2), 285–306. doi: 10.1002/jid.3064 Őzdemir, E., Ward, T., and Zolyomi, E. (2014). The effect of the crisis on the young people’s ability to live independently ( No. 5). Rsearch Notes. Painter, G. (2010). What Happens to Household Formation in a Recession? ( No. 29). SSRN Series. Plantenga, J., Remery, C., and Samek Lodovici, M. (2013). Starting Fragile Gender Differences In The Youth Labour Market . doi: 10.2838/53392 Reher, D. S. (1998). Family ties in western Europe: Persistent contrasts. Population and Development Review , Volume: 24 (Issue: 2), Pages: 45–76. doi: 10.2307/2807972 Saraceno, C. (2015). Il lavoro non basta. La povertà in Europa negli anni della crisi (Collana de.). Feltrinelli. Sassler, S., Ciambrone, D., and Benway, G. (2008). Are They Really Mama’s Boys/Daddy’s Girls? The Negotiation of Adulthood upon Returning to the Parental Home. Sociological Forum , 23 (4), 670–698. doi: 10.1111/j.1573-7861.2008.00090.x Stone, J., Berrington, A., and Falkingham, J. (2014). Gender, Turning Points, and Boomerangs: Returning Home in Young Adulthood in Great Britain. Demography , 51 (1), 257–276. Springer US. doi: 10.1007/s13524-013-0247-8 Stone, J., Berrington, A., and Jalkingham, J. (2012). Is the boomerang generation of young adults a real phenomenon? Returning home in young adulthood in the UK . Ward, T., Ozdemir, E., Gáti, A., and Medgyesi, M. (2012). Young people in the crisis ( No. 5). Research notes.

17