Informal care and long-term labor market outcomes

Abstract

In this paper we estimate long-run effects of informal care provision on female caregivers’ labor market outcomes up to eight years after care provision. We compare a static version, where average effects of care provision in a certain year on later labor market outcomes are estimated, to a partly dynamic version where the effects of up to three consecutive years of care provision are analyzed. Our results suggest that there are significant initial negative effects of informal care provision on the probability to work full-time. The reduction in the probability to work full-time by 4 percentage points (or 2.4–5.0 if we move from point to partial identification) is persistent over time. Short-run effects on hourly wages are zero but we find considerable long-run wage penalties.

Introduction

The effects of informal care provision on caregiver's labor force outcomes have been subject to a large literature in the previous two decades. Labor supply reactions (of females, mostly) have been studied by, e.g., Carmichael and Charles (1998), Heitmueller (2007), Ciani (2012), Casado-Marín et al. (2011), Bolin et al. (2008), Ettner, 1995, Ettner, 1996, Crespo and Mira (2014), Heger (2014), Meng, 2012a, Meng, 2012b, where the effects range from small to very large (up to 30 percentage points) reductions in the probability to work for pay.² The effect on working hours, as studied by, e.g., Wolf and Soldo (1994), Casado-Marín et al. (2011), Bolin et al. (2008), Ettner (1996), Johnson and Sasso (2000), and Van Houtven et al. (2013) are quite mixed, while wage penalties are more consistently found (Van Houtven et al., 2013, Carmichael and Charles, 2003, Heitmueller and Inglis, 2007).

All of these studies have in common that they look at the contemporaneous effect of caregiving on labor market outcomes. As many societies aim at increasing female labor force participation, one often reported policy implication is to set up more flexible work-arrangements to facilitate informal caregiving while keeping the job (Heitmueller, 2007). However, one might argue that negative short-term effects do not pose severe problems – both for the caregivers and societies as a whole – if they are not persistent. Caregiving spells typically last only a couple of years and as soon as caregivers who put their labor force participation on hiatus return to the labor market after cessation of their caregiving spell, the life-time opportunity costs of caregiving might not be too large. However, caregivers are often in the age of 50+ and might have problems to return into the labor force once they left it – either because they voluntarily decide to stay absent or because they cannot return due to labor market frictions.³ This would imply negative consequences that potentially add up over many years after their caregiving period. Thus, to draw conclusions about the holistic costs of care, it is necessary to turn to a longer-run perspective since the cost of caring might be more complex than forgone income for the time spent caring.

This study looks at longer term labor market effects up to eight years after care provision. Evaluating the persistence of effects is the main contribution of this paper. As far as we are aware, only three papers explicitly move away from the contemporaneous perspective. Fevang et al. (2012) use Norwegian data to study labor market outcomes up to around 10 years before and 5 years after the death of a lone parent and do find notable effects on labor market participation (for women, not for men) around the death which, however, are not persistent. On the other hand, reliance on social assistance increases persistently for men. Although the authors do not observe actual care provision these effects can largely be ascribed to informal care obligations. Skira (2015) explicitly takes into account the dynamic effects on labor supply as one of the first papers in this literature. She estimates a dynamic discrete choice model that is underpinned with a theoretical framework. Her results highlight existing labor market frictions for caregivers as their reduced labor supply in the US due to caregiving persists over time. Michaud et al. (2010) also estimate a structural model in order to look at dynamic effects of caregiving on employment. Yet, the authors do not explicitly look at long-run effects and effects beyond three periods after caregiving are not reported.

We use a representative German data set to assess short- and longer-term effects of care provision on labor market outcomes such as the probability to work full-time, to be in the labor force, the number of weekly working hours (conditional on working) and hourly wages. In Germany, the largest European economy, there were 2.6 million people in need of care in 2013 (Statistisches Bundesamt, 2015) and this number is estimated to increase steadily to outnumber 3.4 million people demanding care services by 2030 (Augurzky et al., 2013). Even according to these official – and probably underestimating – numbers, 1.9 million received care in their private home and 1.3 exclusively received informal care (typically by close relatives) making this the most important pillar of the German long-term care system. Thus, not only due to its size as the largest European labor market, Germany is an interesting country to study: it is rapidly aging and already now has a large informal care sector which is even going to increase in the future.

Apart form the longer-term perspective we, as another contribution to the literature, also take the dynamic nature of caregiving spells into account and use sequential inverse probability weighting (IPW) estimators as suggested by Lechner (2009) and Lechner and Miquel (2010) to estimate effects of up to three consecutive years of care provision. A further, if minor, contribution comes from the methodological side where we offer an identification strategy that relies on less functional form assumptions than the previous literature on short-run effects but also than Skira (2015) and Michaud et al. (2010).⁴ Our strategy rests on (sequential) conditional independence assumptions (CIA) which we justify by exploiting cross-sectional but also longitudinal information from our rich household survey, the German Socio-Economic Panel (SOEP). In auxiliary analyses, we relax the CIA to identify effect bounds under weaker assumptions. Other sensitivity tests such as placebo regressions imply that remaining time-invariant unobservables are unlikely to lead to an upward bias of our estimates.

Our main finding is that female caregivers reduce the probability to work full-time by 4 percentage points (at a baseline probability of 35 per cent). The effect is persistent over a period of eight years and seems to be mainly driven by switches to part-time work. High care intensities and longer episodes, however, also increase the long-run probability to leave the labor force. When we move away from point identification to effect bounds, the reduction in full-time work changes to an interval of 2.4–5.0 ppts. As another finding, wages seem to be unaffected contemporaneously but are significantly lower 8 years after the start of a care episode.

The paper proceeds as follows. Section 2 gives a brief introduction into the German long-term care system. Section 3 presents the data and how we exploit the panel structure. Section 4 lays out the estimation strategy and reports results of the baseline (static) model. Section 5 scrutinizes the identifying assumptions and allows for deviations. Results of the dynamic model are reported in Section 6, while some alternative specifications are carried out in Section 7. Section 8 concludes.