I am currently drafting something on the same general topic, so I will add comments based on my current thinking. The folks at CBO discuss three options--purely public, purely private, and a hybrid. They find drawbacks to every option. They explain the hybrid option:
Many proposals for the future of the secondary market involve providing federal guarantees of certain mortgages or MBSs that would qualify for government backing. ...However, a hybrid approach would depart from the precrisis model in three main ways: A potentially different set of private intermediaries would be established to securitize federally backed mortgages, the federal guarantees on those mortgages would be explicit rather than implicit, and their subsidy cost would be recorded in the federal budget.
The prospect of "a potentially different set of private intermediaries" with federal backing should strike terror into your heart. I think that if we are going to have a hybrid option, I would rather go back to Freddie and Fannie than try something new. Just because Freddie and Fannie have earned bad names does not mean that you will gain something by creating a new structure. Based on my experience at Freddie 20 years ago, I would say that the GSE's have a lot of organizational capital built up, and if you start up a new institution from scratch you are creating a whole new learning curve for both the institution and its regulator. That would be both costly and risky.
I would rather tweak a model that failed than build a new model. My first choice would be to tweak the pre-1980 model, in which depository institutions held most of the mortgages. That failed because of
1. Too much maturity mismatch between short-term deposits and thirty-year fixed rate mortgages. I would hope that the market instead would charge such a premium on thirty-year fixed rates that borrowers would prefer shorter-term mortgages, such as the five-year rollover that seems to work well in Canada.
2. Complete absence of risk-based capital or risk-based pricing for deposit insurance insurance. We now know better.
3. Failure to mark balance sheets to market. Ironically, regulators used historical-value accounting when that was least appropriate (with very old mortgages in portfolios) and regulators used market-value accounting where it was most troublesome (on securities with a lot of market volatility). But if we go back to depository institutions holding mortgages, some reasonable (i.e., quarterly, not daily) mark to market would be good.
My second choice would be to tweak the Freddie-Fannie model of 1990 or so.
1. Put the regulator under Treasury (not HUD) and focus regulation on safety and soundness (not affordable housing goals).
2. Forbid the GSE's from getting involved with anything other than long-term, fixed-rate mortgages. We don't need them to play around with adjustable-rate products or all the other stuff that fed the sub-prime boom.
3. Forbid the GSE's from in any way supporting loans with down payments under 10 percent. We don't need them doing that, either.
4. More rigorous capital requirements. The GSE's counted some stuff as capital that really wasn't, and you should combine stupid and clever capital requirements.
The S&L model and the GSE model each failed catastrophically. But I would rather build on a model that failed than build a new model. There are always new risks to be discovered, with either an old model or a new one. But a new model has greater potential to create undiscovered risks than an old one.