Article Uncovering interactions in multivariate contingency tables: ...
Article Age, sex, qualifications and voting at recent English genera...
Article Using Shrinkage in Multilevel Models to Understand Intersectionality
The approach is like CHAID but is model based and uses random effects not fixed which is protective of finding unimportant unreliable results. Unreliable results are down-weighted or shrunk.
This working paper explain this shrinkage property:
Article Do multilevel models ever give different results?.