Unemployment And Mortality: Evidence From The Great Recession
World Bank – Development Research Group (DECRG)
World Bank – Africa Region
March 16, 2016
Did unemployment in the Great Recession hurt people’s health? The broad answer is no: job losses have statistically insignificant impacts on mortality. The exogenous sources of job losses in a U.S. county is the tradable job losses driven by external demand collapses during the Great Recession. The insignificant relationship holds for males and females, for all age groups, and for almost all categories of mortality. Three important exceptions are Alzheimer’s, poisoning, and homicide.
Unemployment And Mortality: Evidence From The Great Recession – Introduction
The Great Recession is the most serious and painful economic downturn in the world economic history since the 1930s. Aggregate unemployment reached as high as 10% in the U.S. It is important to understand the causes and impacts of unemployment in the recession. In this paper, we focus on its health impact: did unemployment in the Great Recession create adverse health problems in the population?
The question is relevant, as temporary financial and psychological tensions from job losses could lead to more permanent damages, such as family fighting, divorces and stress-induced health issues. These could have lasting impacts on adults, and more importantly, on children. Furthermore, tensions could have repercussions on unemployment itself. Health issues, and more broadly, social instability, crime and community deterioration could discourage future employers and entrepreneurs to settle, establish businesses and create jobs. If the social cost channels of unemployment are indeed operative, this provides the basis for the government to combat unemployment and assist vulnerable groups before unemployment-induced social problems take root.
Establishing a causal relationship between unemployment and health is inherently difficult, because of the endogeneity problem. We are generally unsure if unemployment causes or is caused by deteriorating health. In addition, omitted variables can be problematic: a third factor can drive both unemployment and health. For example, the Great Recession was characterized by a great house price collapse that led to a vast number of foreclosures. This might have degraded health conditions and hurt construction jobs at the same time. In that case, both health and job losses are driven by the house price collapse.
To add to the complexity, while health is commonly thought to worsen during recessions, findings have surprisingly been mixed. Ruhm (2000, 2015a) are among the most prominent studies on this topic. They showed that recessions are actually good for health: health improves when the economy temporarily deteriorates. Ruhm (2000) used panel data for 50 states in the U.S. and D.C. and found consistently that most causes of deaths are actually pro-cyclical. He provided four reasons for this pro-cyclicality, which we will go into in detail in the literature review. That trend, however, has attenuated over time (Ruhm, 2015a).
This paper overcomes the identification difficulty by using a Bartik instrument. As will be clearer in the identification strategy section, the Bartik instrument captures a county’s tradable job losses that are only driven by declines in U.S. tradable aggregate demand, and not by county-specific issues. Since there are more than 3,000 counties in the U.S, the U.S.’s tradable aggregate demand is largely exogenous to a county, that is, it is little affected by county-specific fundamentals. This implies that reverse causality and county-specific omitted variable problems, such as productivity shocks, are not likely at play.
Using demand-driven tradable job losses as an instrument for total job losses in a county, we examine the impacts of unemployment on different types of mortality. We find that when county-characteristics control variables are included, unemployment has statistically insignificant impacts on mortality. In other words, aggregate mortality in a county is not significantly affected by the county’s job losses, ceteris paribus. This is consistent with Ruhm (2015a), where he found that mortality is becoming acyclical in recent years.
When disaggregating mortality by types, we find that most types of mortality have statistically insignificant relationship with unemployment. The only exceptions are Alzheimer’s, poisoning, and homicide. We find that while unemployment increases mortality by Alzheimer’s and poisoning, it decreases homicide.
Unemployment seems to have no impact on either males or females, and for most age groups. The only exception is the 25-34 age group. We find that mortality of the 25-34 displays a clear negative relationship with unemployment. In other words, during the Great Recession, counties that have higher job losses actually have smaller numbers of deaths in the 25-34 age group. We will discuss the potential reasons for this phenomenon in section 4.
This paper is organized as follows: Section 2 provides a literature review; section 3 details our methodology and data sources; section 4 presents our regression analyses; section 5 discusses robustness checks and the possibility of reverse causation. Finally, section 6 concludes.
See full PDF below.