October 29, 2013
According to Wikipedia:
“A mortgage-backed security (MBS) is a type of asset-backed security that is secured by … mortgages. The mortgages are sold to a financial institution … that "securitizes", or packages, the loans together into a security that can be sold to investors.”
In simplest scenario, MBS is a pass-through security where every month borrowers make principal and interest payments on their loans. These payments are collected by the financial institution that issued the MBS (or by a loan servicer) and investors in that MBS receive part these payments.
Mortgage-loans are similar to coupon paying bonds and hence are sensitive to interest rate environment. However, there are main two differences that make MBS interesting. Firstly, face values of these loans amortize over time and secondly, because of early payment option with the borrower, cash flows of these loans are not fully deterministic. The second difference, i.e. early payment option, makes pricing and valuation of MBSs challenging.
Here we will try to learn how an MBS is priced. The framework I have chosen in minimal to keep things simple. To price an MBS, at the minimum we need two components: an interest rate model and a prepayment model. Let’s dive into framework now.
There is full range of interest rate model commonly categorized into short rate model and term-structure model. A short rate model, as name implies, forecasts short term rate (1-month for example). On the other hand, term-structure models for interest rates forecast all the points on the term-structure (1-month, 3-month, 1-year upto 30-years, for example).
Here we will use short-rate model called Vasicek model for interest rate (for details). The model is described by the following equation:
drt = a(b-rt)dt + σdWt
While prepayment modeling often involves quite complex and sophisticated modeling, often at the loan level, here we are going to use a slightly modified approach based on Richard and Roll. The Richard and Roll prepayment model involves the following factors:
Richard and Roll propose a multiplicative model of the following:
CPR = RefIncentive * SeasonongMultiplier * SeasonalityMultiplier * BurnoutMultiplier
We will ignore Burnout Multiplier here and assume it to be 1.
Refi = .2406 – .1389 * arctan [ 5.952 * ( 1.089 – CouponRate / MortgageRate ) ]
month 1= 1/30, month2=2/30…month30=30/30=1 and for all other months also factor=1
For months Jan to Dec = [.94 .76 .73 .96 .98 .92 .99 1.1 1.18 1.21 1.23 .97];
Although as a first step we are using a simple prepayment model but let take some time and see what factors influence prepayments in details.
i) Loan-level variables: These are static variables decided at the time of origination like:
ii) Macro-economic variables: These are dynamic variables that are updated monthly like:
iii) Derived variables: These are also dynamic and are calculated monthly like:
Each of these will be explained in the subsequent blogs in ‘R’ language. Stay tuned for more!
Get a detailed analysis and great insights into FICO score, SATO, LTV and audit chart relating to Mortagage Backed Securities in the next part of this blog.
Our counsellors will get in touch with you with more information about this topic.
* Mandatory Field