UCLA

Using a machine learning approach, this study examined how union context — including union status at the time of interview, at the time of sterilization, and post-sterilization affects sterilization regret among American women. Using data from the National Study of Family Growth (NSFG) 1995-2015, we utilized feature importance from the random forest model to identify the most important features in predicting women’s regret. Seven machine learning models were employed using the selected features. Logistic regression, random forest and kernel regularized least squares (KRLS) models out-perform others according to both accuracy and AUC. Examining the effect of union context using the three top-performing models, we found that women who formed new union relationships were at higher risk of regretting their sterilization decisions. Moreover, the effects of union status at the time of interview and of sterilization vanish when post-sterilization union formation was considered.