The recent drop in rates has created some interesting situations in the market, especially for lenders.   First of all, I'm not aware of any consensus on how to calculate a "current coupon" rate in this environment.  The current coupon is calculated by interpolating between coupons that are above and below par, adjusting for the delay days associated with the securities in question.

Let's take an example from the old days when coupons traded below par.  Let's assume the following, assuming (for simplicity's sake) that we're calculating the current coupon for February settlement, with FN 3.0s at 99.00 and FN 3.5 at 101.25.  First, you have to adjust for the delay days.  Fannie Mae pools pay on the 25th of the month following the record date, which results in a 24-day delay.  (The delay results from all the accounting and financing complications involved with managing the vast numbers of loans in the MBS universe.)  The prices can be adjusted for the delay by adding 24 days of coupon payments.  For a 3% pool, the price is adjusted higher by 0.20 (i.e., 3.0 x 24/360), resulting in a 99.20 adjusted price; the 3.5% pool has an adjusted price of 101.2333.  You would then interpolate between the two prices to get the rate that equates to par.  In this case, it is 3.1967%.  The last adjustment is to convert it from monthly yield (since MBS pay monthly to a semi-annual bond equivalent yield, which result in a current coupon rate of 3.218%.

However, we are in a world where the lowest tradable coupon (30-year 3.0s) is both highly illiquid and well above par.  In past periods of low rates, the practice would be to extrapolate (rather than interpolate) to par.  This looks like what some people are doing; however, it gives you some very bizarre numbers if you try to track this number (or look at the current coupon spread over Treasuries or swaps).  A major provider shows the current coupon rate rising on Thursday from 2.52% to 2.70%, even though MBS prices were higher on the day.  This in turn means that the spread of the current coupon over the 10-year Treasury yield, a closely-watched benchmark, has fluctuated this week between +65 bps and +88 bps with minimal change in MBS relative value.  As they say...go figure.

The huge run-up in MBS prices has impacted the market in other ways.  Matt Graham wrote about the liquidity (or lack of liquidity) in 30-year 3.0s.  As he noted, some lenders are originating loans that would be securitized as 30-year 3.0s (as well as 15-year loans that would go into Dwarf 2.5s), although it's unclear what's being done with the loans.  (They could be sold to the GSEs' cash window.)  With rates pushing down, a 3.75% loan can still be pooled into a 3.5% security (with a proviso-see below); however, the poor execution on 3.0s, and lenders' unwillingness to short the coupon, has been an impediment to rates moving even lower.  For example, the spread between the Freddie Mac survey rate and the 10-year Treasury yield is at +204 basis points, versus an average (over the last two years) of +162.

The "stickiness" of rates at current levels is, in my mind, largely a function of having limited outlets for loans with note rates of 3.625% and lower.  The biggest problem is that there is no natural buyer for 30-year MBS with 3% coupons.  I've recently written that the Fed should buy all outstanding 3% pools, which would do more good than just "buying the market."  In any case, markets for these very low coupons need to develop for rates to move decisively lower.

Another complicating factor is the impact of the recent tax on mortgages, paid as a 10 basis point addition to a loan's guaranty fee.  Consider the above example on pooling 3.75% loans into 3.5% pools.  It's almost certain that the new g-fee can be bought down entirely (although there has not yet been a definitive statement to that effect from Freddie or Fannie), leaving 25 basis points of servicing to be held by someone.  A question that the GSEs are grappling with, however, is the cap on agency buy-ups.  Most contracts are written such that the total amount that can acquired by the GSEs on any loan (including both the g-fee and servicing) is capped at 37.5 basis points.  This means that the 10 basis point tax limits the amount of servicing that the GSEs can buy as part of the pooling transaction.  While buy-ups have not been a big factor in the past (since most big lenders just held excess servicing, rather than sell it to the GSEs at puny multiples) this could be a factor in the future, especially in light of the shrinking number of players willing to take down servicing.   Supposedly, the GSEs are looking at increasing the caps, but it's unclear whether the contracts will (or can) be revised.