I'm interested in papers which consider mathematical models of risks of different portfolios of retail credit. This is not my area of research, so I may be misusing some terms. The idea is simple: I have different sets of credit portfolios with aggregate information known (no personal level detail of each borrower) and want to decide which portfolios to keep and which to sell.
The retail credit risk management is generally based on models that try to discriminate between good (people that probably will be able to pay back the debt) and bad customers (people that probably will not).
Particularly, as the question explicitly asks for, you want to some references to allow to decide which customers, already acquired, to keep and which not keep in portfolio; this field in retail risk management refers to LGD (loss given default) models and you should focus your studies in this kind of model to deepen the field (try to google "LGD model", to look something for).
As regards you questions particularly, I suggest you to read:
Anolli, M., Beccalli, E., Giordani, T., 2013, Credit Risk Management, Palgrave Macmillan (Studies in Banking and Financial Institutions)
It is a good recent book about retail credit risk management and examine pretty in depth the models you need for.