If you're like me, you don't like to wait for stuff, especially when the delivery of that "stuff" is entirely in the hands of someone else. I don't mind doing a little work to get something though-- but I'm lazy, so it has to truly be "a little". So when it comes to something like projecting mortgage prepayment speeds, I'm in a bit of a pickle. In theory, I should be doing monte carlo simulations or transition models-- but I don't want to spend all the time / effort that would take (I could buy someone else's model, but that would be cheating...). However, what I can get quickly are strats. Of all sorts and all cuts and all shapes and sizes.
Now, before I go on and on, if you want to skip to the chase, here's the picture I will eventually talk about in this post, which shows GSE sellers plotted on a chart with Rate on X and CPR on Y, based on a strat (which I ran in under a minute with Liquidaty [sorry, shameless plug I know]) on the 30-yr FRM public data released by Fannie and Freddie. Draw your own conclusions, but before doing so, notice:
- the coupon/CPR relationship seems pretty linear
- the horizontal dispersion at times is quite wide (e.g. Quicken vs Ditech = 60 basis points)
- the vertical dispersion is also pretty wide sometimes (e.g. Quicken vs any of the lenders in the most crowded part of the chart = 5+ CPR):
And now, back to going on and on...
Mortgage rates today are a bit higher than they were when 2013 started, and lower than when 2013 ended, so I thought it might be interesting to look at 2013 as a sort of proxy for what the future might hold. So I put Liquidaty's software to work on the public GSE loan performance datasets (note: I didn't include the new dataset released last week) to see if any interesting patterns might emerge.
First thing I noticed in the $2.8 trillion (roughly half each of Fannie and Freddie) was that the Fannie pool had a weighted average coupon nearly 30 basis points higher than Freddie's. Which makes you wonder, how much is that near-30 basis points worth? Which of course you can't answer without projecting prepayments.
Which makes you wonder how much that extra coupon incentivized borrowers to prepay faster. It looks like the answer is: not much. Looking at my handy-dandy Liquidaty CPR strat, I can see that the lifetime CPRs of the pools (through Fall 2015) was 9.76 and 9.23. On top of that, the weighted-average coupon difference only narrowed a tiny bit to 27 basis points. For a 3.8-ish coupon and a 9.23 vs 9.76 CPR, WALs would be around 3.75 to 4. So you're getting an extra 30 basis points for about 3.75 years, with a little WAC deterioration that brings it down let's say to a lifetime average of 27 (remember, we're already seeing it deteriorate only to 27 after 2 years, so let's just linearly project that). So that gets us to an undiscounted extra cashflow strip over the sub-4-year WAL of more than a full point. Not too shabby!
So then I figured, let's forget all about the theory for projecting CPRs and strip it down to just the Seller strat. Ran Liquidaty to get CPR by seller on 2013 vintage, stuck that into a bubble chart, added a green line to it and voila:
The trend is fairly consistently linear, but more interesting are the data points that are far from the line (btw, I intentionally did not consolidate similar seller names, because those small differences often indicate differences in products / programs). For example, I think if I was deciding whether to buy from Quicken or Ditech, then all else (other than rate) equal, I would be willing to bid plenty higher for the extra 60 basis points Ditech is getting, which comes along with about the same CPR (and a 2-year WAC deterioration of only 7 basis points). Sure, that was 2013 and now is 2016, and everything could be different this time around. But knowing what you know now, would you really bid those pools without considering how their older-sibling vintages performed?
There's a lot more cool stuff this data can tell you. Want to do this for yourself? You can! Liquidaty will crunch the numbers in just a few seconds for all the data, or any subset you want of it. Feel free to reach out for more info to firstname.lastname@example.org.
I am not a researcher and research is not my company's product (software is). I do not represent any of the information herein to be accurate or complete, and it should not be relied on as such. The information, opinions, estimates and forecasts contained herein are subject to change without prior notification. No part of this material may be (i) copied, photocopied or duplicated in any form by any means or (ii) redistributed without the prior written consent of Liquidaty.