Replication of “Do state‐customized TANF work policies actually reduce unemployment?”

Abstract

Background In Na Yeon Kim and Frances Stokes Berry's paper “Do State‐Customized TANF Work Policies Actually Reduce Unemployment?,” the researchers attempt to determine if the implementation of Worker Supplement Programs (WSPs) caused the unemployment rate of low‐income females to decrease. They found that states that enacted WSPs attained about 18 percent fewer unemployed low‐income females when compared with states that did not implement WSPs. Objectives The original purpose of this project was to expand on Kim and Berry's research regarding TANF policies. Methods I created a dataset that closely matched Kim and Berry's original dataset. I then applied the same difference‐in‐differences model they reported using and two supplementary difference‐in‐differences models. Additionally, I used synthetic control methods for four states to further investigate the causal claim. Results I closely matched Kim and Berry's summary statistics, but the ambiguity of their data collection methods hindered my ability to exactly match the data. When excluding fixed effects and trend variables, I found that states with WSPs had about 20 percent fewer unemployed low‐income females. The inclusion of fixed effects and trend variables, which Kim and Berry reported using in their model, caused the WSP variable to become about 0 percent and insignificant. The results of the synthetic control methods provided additional evidence that questions the validity of the original findings. Conclusions Kim and Berry's research seemingly contains significant flaws, including incorrect results. This replication project offers evidence that the significant decrease in unemployment among low‐income females is likely not present, highlighting the importance of documentation and replication.