- This event has passed.
CUPC Research Presentation – Martha Bailey
November 3, 2016 @ 12:30 pm - 1:30 pm
Martha Baily, University of Michigan, Population Studies Center
How Do Alternative Linking Methods Perform? Evidence from the LIFE-M Project
Newly assembled, large-scale linked historical datasets are transforming the narratives of American economic history (Abramitzky et al. 2012, 2013, Ferrie and Long 2013, Abramitzky et al. 2014, Collins and Wanamaker 2014, 2015, Feigenbaum 2015a, Nix and Qian 2015, Aizer et al. 2016). Inferences using these data, however, depend upon the quality of the data linkages. This paper uses a clerically-reviewed sample from the LIFE-M project to provide new evidence on the performance of four automated linking algorithms. We also examine how two commonly used phonetic name-cleaning methods, SOUNDEX and NYSIIS, affect match quality. Our results show high link rates for each algorithm, and each algorithm generates reasonably representative samples. However, we document high rates of type I errors, ranging from a minimum of 17 percent to 54 percent. The incidence of type I errors increases with the use of name cleaning algorithms such as SOUNDEX or NYSIIS. These findings indicate that a large share of links used in historical analyses are likely erroneous, indicating that measurement error may play a large role in some of the literature's new findings. We conclude with constructive suggestions for improving the quality of machine algorithms to link data.
Dr. Bailey’s research focuses on issues in labor economics, demography and health in the United States, within the longer-run perspective of economic history. Her research has examined the implications of the diffusion of modern contraception for women’s childbearing, career decisions, and the convergence in the gender gap. Her most recent projects focus on evaluating the shorter and longer-term effects of Great Society programs, including a recently published book (co-edited with Sheldon Danziger) on the legacies of the War on Poverty. Bailey is an NBER Faculty Research Fellow and in 2007 was an RWJ Health Policy Research Scholar.
