10.18128/MPC2014-1
Luo, Liying
Liying
Luo
Hodges, James
James
Hodges
Winship, Christopher
Christopher
Winship
Powers, Daniel
Daniel
Powers
The Sensitivity of the Intrinsic Estimator to Coding Schemes: A Comment on Yang, Schulhofer-Wohl, Fu, and Land
Minneaplis, MN: Minnesota Population Center
2014
Text
2014
Minnesota Population Center Working Paper Series, 2014-1
The Intrinsic Estimator (IE) has been proposed to address the age-period-cohort problem and is believed by many to yield robust and reliable estimates. We, however, show that IE estimates are highly sensitive to one’s choice of coding scheme or model parameterization. We reanalyze data from published articles to demonstrate that estimation results using one coding scheme (e.g., the zero-to-sum coding) can be dramatically different from those obtained using a different coding scheme (e.g., reference group coding). The results are so different that a researcher would sometimes reach opposite conclusions about the effects of age, period, and/or cohort depending on the seemingly innocuous choice of coding scheme. We provide a nontechnical explanation for this sensitivity; an appendix provides a mathematical proof.
Minnesota Population Center Working Paper Series