- This event has passed.
Did We Botch the Notch? Investigating the Social Security Notch Literature and Use of the Health and Retirement Study
March 1, 2019 @ 12:00 pm - 1:00 pm
On Friday, March 1, Jeremy Moulton, PhD, will present Did We Botch the Notch? Investigating the Social Security Notch Literature and Use of the Health and Retirement Study as part of the Carolina Population Center 2018-2019 Interdisciplinary Research Seminar Series.
Moulton is an Associate Professor in the Department of Public Policy at UNC-Chapel Hill. His research investigates the intended and unintended consequences of public policy on labor supply, consumption, retirement, self-employment, real estate, and the intergenerational transmission of wealth and education.
Moulton is hosted by Carolina Population Center Fellow and Center Director Elizabeth Frankenberg. Frankenberg, Professor of Sociology, has served as the Director of Carolina Population Center since 2017.
Friday, Mar 1
Carolina Square Room 2002
123 West Franklin Street
Location information is here.
The Social Security Amendments of 1972 and 1977, often referred to as the “Social Security Notch,” have been exploited as a natural experiment by researchers across numerous literatures to estimate causal relationships between income and an array of different outcomes. In this paper, we investigate whether a demographic or sampling shift potentially confounds results from Notch papers that use the Health and Retirement Study (HRS), a popular data source for many of these studies. Specifically, we find a large discontinuous shift in the proportion of non-white respondents in the HRS starting with the 1917 birth cohort (i.e., those adversely affected by the Social Security Amendments), which we conclude is due to sampling problems in the HRS. To determine whether this confounds key results in the literature, we replicate three papers with different empirical approaches that used the HRS in their analysis of the Social Security Notch (Moulton, 2017; Goda et al, 2010; and Moran and Simon, 2006) and test the extent to which the results hold after accounting for the sampling issue. Overall, the results imply that the Social Security Notch literature and other work using the HRS data should carefully consider this issue.