Health Endowments and Unemployment during Macroeconomic Crises

Abstract

A large literature in economics has examined the causes and consequences of macroeconomic fluctuations. Given the importance of health human capital for labor market outcomes, an important facet of the literature on consequences of economic fluctuations has examined whether and how events like recessions, job displacements and business cycles affect health outcomes (Sullivan and Von Wachter 2009, Ruhm 2000, Currie and Tekin 2011, Stillman and Thomas 2008). Some of this work has focused on how such events affect early childhood health or even health at birth (see for example Paxson and Schady (2005), Chay and Green- stone (2003) and Dehejia and Lleras-Muney (2004)); this research is especially important given the research highlighting the long term economic implications of health in utero and during infancy (Almond and Currie 2011, Heckman 2007). While examining the consequences of macroeconomic shocks on health is extremely important, it is also critical to understand whether people with poorer health ex ante are more vulnerable to job loss during a crisis. The research in this area has largely examined how business cycles affect labor market outcomes across a wide range of demographic characteristics such as age, gender, sex, race and education (Engemann and Wall 2009, Cho and Newhouse 2012, Hoynes, Miller, and Schaller 2012). However, despite the large body of important work in this area, there appears to be a dearth of studies examining whether pre-determined health dictates the degree to which one is affected during economic downturns. In this paper, we build on the literature documenting the long term consequences of early childhood health and on the literature examining who is affected during a crisis to show that pre-determined health endowments, such as health at birth, are an important marker for labor market vulnerability during economic downturns. We study the economic crisis in Sweden in the early 1990s when unemployment went from less than 2% to around 8% in just two years. This crisis is referred to as one of the “Big Five” downturns along with that of Spain, Norway, Finland and Japan according to Reinhart and Rogoff (2008). Many observers of the Great Recession in 2008 compared it to the Swedish crash of the 1990s, and they especially noted the ways in which Sweden recovered from the crisis (New York Times, September 22, 2008; Time, September 24, 2008). While much has been written about the causes and consequences of the crisis in Sweden during the early 1990s (e.g., Englund 1999, Jonung, Kiander, and Vartia 2009), the main import from these studies appears to be that the crisis was a result of a combination of various factors including monetary policies in the 1980s, budget deficits, financial deregulation and collapse of trade. Using birth weight as an indicator of health at infancy1, we examine how adults who were born with lower birth weight fare during the Swedish crisis. Recognizing that birth weight likely represents nutritional inputs and other attributes of the mother that might confound such long term analysis, we examine plausibly exogenous variation in birth weight within twin pairs similar to prior studies (Black, Devereux, and Salvanes 2007, Almond, Chay, and Lee 2005, Royer 2009). Examining this relationship requires rather unique data. Most electronic birth records, even in countries known for their excellent administrative records (for example Norway), start in the late 1960s. For this reason, examining how pre-determined health endowments affect job attachment during major crises has been underexplored since subjects for whom we have reliable birth data are generally too young to be observed for a substantial period in the labor market before and after the crisis. In the case of Sweden, we use a unique source of twin birth records collected for nearly the entire population of births between 1929-1956. Hence, most of our sample is observed while they were active in the labor market for ten years before and after the crisis. These unique birth records are then matched to individual yearly income (including income from sources such as unemployment insurance, disability, sickness records from 1981-2005. Twins studies examining the role of birth weight typically also rely on the idea that parental investments or behaviors do not react to birth weight differences. However, we need even weaker assumptions about the role of parental investmentscompared to prior settings, since we examine twin pairs before and after the crisis and use panel data on twin pairs (Bharadwaj, Eberhard, and Neilson 2011, Figlio, Guryan, Karbownik, and Roth, forthcoming). Under the assumption that all birth weight-specific investments are made prior to the crisis, our estimates would effectively hold these differential investments constant. The fact that OLS results are small and largely insignificant underscores the need for controlling for unobserved characteristics that might otherwise be correlated with birth weight and labor market outcomes. We find that health endowments at birth can have important repercussions for labor market outcomes during severe econmoic crises, that is, adults who were born with poorer health at birth were much more likely to face job loss and go on unemployment insurance (UI) during the crisis. Importantly, we find that these adults tended to stay on UI for a while after the crisis, indicating a persistent impact on their labor market outcomes. These effects are especially pronounced in the private sector, which is also the sector being hit the hardest by the crisis. Lower birth weight adults also have lower earnings after the crisis, although this result is not statistically significant. Importantly, we find that the relationship between health at birth and UI take up during the crisis is not determined by intermediating factors like educational attainment or pre-crisis selection into certain occupations. Since some occupations were hit more extensively by the crisis than others, one hypothesis might be that lower birth weight twins have selected into occupations that just happened to be hit harder by the crisis. While birth weight determines occupational sorting to some degree, our results hold even when we examine twin pairs who worked in the same 3-digit, and even within the same 5-digit, occupation code. Also, since the effects seem to be especially pronounced for older workers, job protections laws like “last in - first out” principle is not a likely explanation to the results either. Last, the results are not explained by people simply moving in between different social benefits systems before and after the crisis. Finally, recent work has shown the importance of social assistance programs in improving early childhood health as well as long run effects of early exposure to social safety nets (Bitler and Currie 2005, Hoynes, Schanzenbach, and Almond 2012). Our paper adds to this literature by showing that better health at infancy can be particularly protective during periods of economic fluctuations. Hence, we add another dimension by showing early childhood health-related spillovers of safety net programs become important because children born with better health as a result of these programs are themselves less likely to take up social assistance later in life. This study is also important for highlighting the role of social assistance during 3 a crisis. One of the fundamental questions about the design of optimal insurance policy is the extent to which it can mitigate morally arbitrary misfortunes of nature. By exploiting random variation in birth weight, we are able to show that social assistance, at least in the case of Sweden, appears to come to the aid of those who were born at a health disadvantage.