What effect did the global financial crisis have upon youth wellbeing? Evidence from four Australian cohorts.

Macrocontext and Wellbeing

There has been growing interest in recent years of the effects of macrocontext (national or international conditions or events) on individual factors in psychology ( Fletcher, 2015 ). However, the idea that dramatic changes in the global environment can have meaningful influence on individual psychology is not a new one. C. Wright Mills (1959/2000 , p.3) laid the groundwork for this area of inquiry, stating “neither the life of an individual nor the history of a society can be understood without understanding both”. In a pioneering study, Glen Elder’s (1999) research on children growing up in the Great Depression prompted consideration of not only the influence of macrolevel conditions on progress and frustration in development, but also how such effects filter through to young people via links with local institutions, social ties, and family networks. Elder (1999) noted effects of the Great Depression on social wellbeing, psychological health, and hope and optimism for the future; particularly among those who were younger and thus less cognitively developed. In addition, Elder drew attention to the effect of economic downturns on populations of youth as a whole, in addition to those suffering abject and persistent deprivation (see Elder & Caspi, 1988 ). Thus, one needs to consider the effects of economic downturns on factors such as wellbeing across whole cohorts ( Jahoda, 1988 ).

Recent research by Di Tella et al. (2006) found that a country’s economic position has significant effects on wellbeing. Indeed, Di Tella et al. indicated that rising unemployment that results from economic hardship has a critical effect not only on those who lose their job, but for the population as a whole. These effects were observed across a range of macroeconomic events including recessions, changes in GDP, inflation, and the relative generosity of the welfare system. Schoon (2006) considered cohorts of British people born in 1958, thus growing up in a golden age of economic stability and prosperity, and those born in 1970, thus growing up in more economic vulnerable times. Schoon reported that growing up in times of economic prosperity seems to be a protective factor against psychological distress and promotes wellbeing. Conger, Rueter, and Conger (2000) , studying the effects of the severe economic downturn in the rural midwest of the United States found that economic distress affected young people’s wellbeing via its impact upon parents’ mood and parenting behavior. Finally, Forkel and Silbereisen (2001) considered the effect of the reunification of Germany on development. Using a family stress model framework, they found that economic uncertainty had an effect on child wellbeing via parents depressed mood in the West, but less so in the more significantly altered society in the East.

The GFC and Wellbeing

In relation to the GFC, a review by Clark and Heath (2014) found dips in trends in happiness and social wellbeing, including trust and experiences of prosocial behavior in the United Kingdom and the United States. In Australia, Sargent-Cox et al. (2011) focused on the influence of the GFC on Australian seniors, suggesting that this group was at particular risk due to vulnerability in retirement savings as well as fear spread by the Australian media. They also found significant increases in depression and anxiety. Likewise the recent UNICEF Innocenti report ( Fanjul, 2014 ) found that in 29 of the 41 OECD and non-OECD EU countries wellbeing decreased and experience of everyday stress increased from 2007 to 2013. They attributed this impact as likely due to the GFC.

Taken together, the literature to date suggests three important considerations. First, changes in macrocontexts, and economic conditions in particular, can have meaningful impacts on wellbeing. Second, these may have an impact upon everyone (i.e., those directly and indirectly affected). Third, consideration of general wellbeing should be supplemented by consideration of domain specific measures within multiple life domains, given findings that social domains of life appear to be vulnerable to economic conditions.

Youth and Vulnerability

Although Elder (1999) focused on the effect of the Great Depression on youth, recent research has tended to focus on the elderly as a group of particular vulnerability. Although the elderly were particularly exposed to the GFC (e.g., Sargent-Cox et al., 2011 ), there are important reasons to also consider the developmental period ranging from the transition from high-school to the mid-20s. Here we explore the biological, social, and economic reasons for this.

Steinberg (2009 , 2013 ) has highlighted convincing biological, behavioral, and neurological evidence to extend the definition of adolescence up to mid-20’s. Steinberg’s (2014) argument is both social, noting that youth are now becoming financially and socially independent at later ages, and biological, with evidence of continued and significant brain plasticity well into the mid-20s. Steinberg noted that this malleability means that young people are particularly vulnerable to toxic contexts that can lead to lifelong negative impacts. Cummins (2014) likewise noted that wellbeing is particularly volatile during adolescence due to heightened biosocial change. This is consistent with the work of Elder (1999) who noted that age was negatively related with impact of the Great Depression, hypothesizing that ongoing cognitive development meant that hardship had a more severe and long lasting impact.

Socially, not only is the post-high-school period defined by identity formation and uncertainty in social and occupational roles ( Arnett, 2000 ) but it is a period in which developmental transitions are both plentiful and of considerable importance to long-term status attainment ( Guo, Parker, Marsh, Morin, 2015 ; Parker, Lüdtke, Trautwein, & Roberts, 2012 ; Parker et al., 2012 ; Parker, Thoemmes, Duinveld, & Salmela-Aro, 2015 ). The life span theory of control indicates that those making the transition from formal schooling to tertiary education or the labor market are particularly at risk of contextual events and influences ( Heckhausen, Wrosch, & Schulz, 2010 ; Heckhausen & Schulz, 1995 ; see also Dietrich, Parker, & Salmela-Aro, 2012 ). Such a period is defined by the convergence of developmental tasks from multiple life domains (educational, occupational, social, family, romantic, and values) and, as such, is one of the most critical developmental periods ( Zarrett & Eccles, 2006 ). From the perspective of life span theory of control ( Heckhausen & Schulz, 1995 ) the particular danger of macroeconomic events, like the GFC, would be the potential to knock youth off a typical developmental track; delaying transitions, interfering with increasing independence from parents, and extending periods of career and educational uncertainty. For example, research on transition delays provides evidence that even a relatively short delay can have ongoing consequences for status attainment well into adulthood (see Haase, Heckhausen, & Köller, 2008 ; Heckhausen & Tomasik, 2002 ; Parker et al., 2015 ).

Economically, not only is unemployment particularly high during this developmental period, but in Australia, the United Kingdom, and the United States the jump in unemployment levels during the GFC for those aged 16 to 24 was notably larger than for the working population as a whole; youth unemployment in Australia jumped from 8.9% to 13.8%, whereas overall unemployment grew from 4% to almost 6%, in the period of 2008 to 2011 (our calculations are based on Australian Bureau of Statistics data). As noted above, both unemployment and the risk of unemployment has a particularly detrimental effect on wellbeing ( Clark, Georgellis, & Sanfey, 2001 ). The risk of unemployment can cause young people to make different choices about their educational and occupational plans than they otherwise would, which can put them at a distinct disadvantage when competing with their near age peers who entered this developmental period at a more economically advantageous time (see Kahn, 2010 ). Finally, at the post-high-school transition young people are increasing independence via entry into the labor market or tertiary education, yet they also remain strongly connected to parents ( Parker, Lüdtke et al., 2012 ). As such, the wellbeing of young people may suffer from both their own exposure to economic downturns but also that of their parents as suggested from a family stress model perspective ( Conger et al., 2000 ).

Multidomain Wellbeing

Psychologists, economists, and sociologists have all been interested in the influence of both micro- and macrolevel conditions on wellbeing. A common thread across much of this research is general or aggregated wellbeing (e.g., life satisfaction). There is, in contrast, relatively little attention given to how such events might differentially affect multiple life domains. Part of the reason is that it is difficult to determine how many and which life domains to cover. As Cummins (1996) noted, if every human action is considered a life domain, true multidimensional measurement becomes impossible.

Derived from the work of Cummins and colleagues, however, youth surveys of the Australian population have covered between 12 to 14 life domains focusing on achievement, social life, community engagement, perspectives on the future, and living standards. These domains are derived from empirical research on what most participants consider to be important and have been used over long periods of time, across countries, and age groups. This provides strong evidence of validity and utility of multiple dimensional measures of wellbeing in these areas (see Cummins, 2014 ; Tomyn, Fuller Tyszkiewicz, & Cummins, 2013 , for a review). As Cummins (2014) noted, there is value in a parsimonious multidomain approach, and the domains that are used here capture the domains that are relevant for the majority of young people ( Tomyn et al., 2013 ).

Thus, taking a multidimensional perspective, we consider the degree to which there are differential impacts of events like the GFC on wellbeing measured in different domains. As noted above, there is some evidence to suggest that social wellbeing and optimism for the future is particularly at risk during economic hard times ( Clark & Heath, 2014 ; Elder, 1999 ; Lau et al., 2008 ), yet research in this area has been relatively limited in the number of domains explored.

Hypotheses

Empirical research suggests economic conditions can lead to significant changes in wellbeing. This literature, however, has tended to use cross-sectional studies without the ability to follow individuals over time. Here we make use of the unique opportunities afforded by the LSAY datasets, which follow young people from four birth cohorts for up to 10 years. The nature of the LSAY data, four birth cohorts measured roughly three years apart, allows us to compare the influence of the GFC at two distinct ages in the post high-school transition period (i.e., age 19 and 22). As can be seen in Table 1 , the 19-year-old age group captures much of the movement of young people from high-school to tertiary education or the labor market. At age 22, young people appear to have mostly made this transition. The comparison of these age groups is opportunistic (i.e., due to the possibilities afforded by the data), however, and thus we have little evidence on which to assume the GFC would have differential effects. On this basis, we put forward the following hypotheses:

Hypothesis 1: The GFC will have a negative impact upon young people’s wellbeing across the major domains of importance to late adolescents.

Hypothesis 2: We expect the influence of the GFC to differ by life domain, with particular impact on domains related to social life and long-term prospects.

Hypothesis 3: As existing research base is not yet large enough on which to make a strong hypothesis, we do not anticipate that there will be differences in the size of the effect of the GFC at age 19 compared to 22.

GFC

The GFC is generally considered to have begun during 2008. However, the impact on Australia and the individuals in the study likely came later. Sargent-Cox et al. (2011) made the case that the impact of the GFC on Australians, and particularly the psychological impact, should be dated to 2009. We thus consider the GFC to have occurred when participants were aged 19 in the 1990 cohort and 22 for the 1987 cohort. Marking the GFC at 2009 is both consistent with previous research, captures both the dramatic jump in unemployment levels that centered on this period and the zenith of media reporting on the GFC where there was a particular environment of heightened “panic, anxiety, and insecurity” ( Sargent-Cox et al., 2011 , p. 1105).

Counterfactual reasoning

In addition to concerns relating to isolating period effects, we were also concerned with providing estimates of the effect of the GFC that were as close to causal as the data would allow. To do this, we aimed to find counterfactual conditions that serve as an indication of what would have occurred to a variable of interest had a given event not occurred ( Morgan & Winship, 2014 ). Put simply, in the case of the current research, we ask the question “What if the GFC never happened?” In the current research a birth cohort that experienced the GFC at a particular age serve as the exposed group (i.e., experienced the GFC at age 19 or 21) and the closest earlier cohort at the same age serves as the nonexposed group (i.e., did not experience the GFC at age 19 or 21). To increase our confidence that the control group acts as a sufficient counterfactual for the treatment group we used two approaches common in sociology and economics; namely a matching and a difference-in-differences (DID) technique.

Propensity score matching

Matching aims to find strategic subsamples of individuals in the exposed and nonexposed groups that either match participants across groups exactly on a small number of critical confounding variables, match approximately on a large number of confounding variables, or some combination of the two ( Morgan & Winship, 2014 ). In the current research we used a mixture of exact and approximate matching via a propensity score matching (PSM) approach. Here participants in the exposed and nonexposed groups were matched exactly on exogenous demographic variables (gender, state of residence, social class [Erickson-Goldthorpe-Portocarero schema; Ericson, Goldthorpe, & Portocarero, 1979 ], and Indigenous status) and postschool pathway variables (number of years of high-school completed, labor market status [employed, unemployed, not in labor market], and tertiary education status [enrolled, completed, dropped out, not relevant] measured at age 18 for the 19-year-old comparison and 21 for the 22-year-old comparison). Participants were also propensity matched on age in days and all wellbeing variables up to the year prior to the GFC.

The aim of PSM is to create samples of exposed and nonexposed individuals who are similar (or balanced) on a wide range of potentially biasing covariates. Initial analysis consisted of modeling the relationship between the covariates and presence in either the exposed or nonexposed groups. We used logistic regression to estimate the propensity score and, based on these scores, we used nearest neighbor matching with matches allowed when participants were within .20 of the standard deviation of the logit of the propensity score. As noted above, exact matching was used for several demographic, educational, and occupational status variables. One-to-one matching was used, without replacement (see Stuart, 2010 ; Thoemmes & Kim, 2011 , for a review). Propensity score estimation and matching were done with the MatchIt package in R ( Ho, Imai, King, & Stuart, 2011 ) and regression with clustered standard errors for school membership was conducted with the survey package ( Lumley, 2011 ). Hypotheses were tested using equation 1 .

Here γ represented the effect of the wellbeing variable P R E _ Y before the GFC (age 18 for the 19 year-old comparison and 21 for the 22-year-old comparison), β is the parameter of interest—the difference in Y between the GFC exposed cohort (coded 1) and control cohort (coded 0). Subscript j was the school that individual i was in at wave 1. Importantly, PSM allowed us to match participants on both grade in school and age in days, thus ensuring participants were similar in both biological and social developmental stage at the comparison point.