Asset-backed securities Credit IO’s: Don’t be a slave to your data
In this article, I’m going to address a common complaint we’ve seen ABS investors have: that when building systems, too much automation creates a “black box” that then doesn’t allow the user to adjust the data. in the way they see fit.
Let’s face it, traders are on the front lines evaluating complex securities like ABS bonds and the more you can allow users to take the data and create useful models that don’t “lock them into a particular view” of what’s being traded, the better. Will Most of the time, traders create their own spreadsheets and generally do a great job. However, the lack of ability to dynamically communicate with a database of securities information can cause a host of problems in the ABS market, if only when next month’s data set comes out of the trustees and they are found. struggling to manually update your spreadsheets.
Furthermore, IT departments blanch at the thought of those overly flexible, manipulable spreadsheets that defy “systematization.” In this article we will discuss a specific example and how to meet the needs of both areas: IT and the Dealing Desk.
Let’s go back to the topic of “credit IO”.
Definition: A Credit IO is an ABS bond that is low enough in the capital structure of an ABS deal that, based on the level of collateral default and the severity of losses that the market is currently experiencing, makes a investor DO NOT expect any major payment.
Assumption: The principal of the bond will be reduced to zero at some point. The investor expects NOT to recover any capital. However, until that time, the bond can generate interest cash flows, therefore it is an “interest only” bond.
Key factor: Synchronization of losses. Between now and precisely WHEN the bond is fully amortized, the bond will earn interest. Those monthly cash flows are worth something. The faster the bond will be redeemed, the less cash interest will be received. The longer the bond exists, the longer it will receive cash flows. The trick is to figure out when the losses will affect the bond. Therefore, the timing of the losses will have a dramatic effect on the price that an investor should be willing to pay for the bond. Less time to fully scored point = lower price.
So let’s take a look at some of the items related to the data side of this. These are some of the relevant points:
1. Delinquency
2. Foreclosure and REO Deadlines
3. Loss Severity to be used to determine how much of each loan will be lost due to defaults.
4. Credit enhancement levels: primarily overcollateralization (OC) and the current level of credit enhancement for each tranche (how much of the capital structure supports the particular tranche or tranches we are evaluating).
In a Bloomberg, you can show a simplistic method to test this by typing an ABS cusip followed by the Mortgage (F3) key and then typing “MTCS”. This gives you the ability to take the current level of the 60 and 90 day delinquency agreement and apply a particular percentage to each that you expect to go into default. It is assumed that the loan amounts in Foreclosure (FC) and Real Estate (REO) are 100% delinquent. So we have as an example:
Table %% that will be the default default quantity
% of offer more than 60 days past due 8% 60% 4.8%
% of offer More than 90 days past due 5% 70% 3.5%
% trading in FC 3.5% 100% 3.5%
% Offer in REO 2.5% 100% 2.5%
For a total of 14.3%, we expect it to end in full default and therefore experience a loss.
Add those numbers (14.3%) and multiply by a single loss severity entry and you get the approximate amount of the deal you will experience as a loss. Let’s say we use 50% Loss Severity. That will give us 7.15% of the outstanding balance of collateral in the deal which we expect to impact the capital structure of the deal in the form of losses. Compare that amount to the credit enhancement of the particular bond you’re evaluating, and if you have a ratio (called a “Coverage Ratio” at Bloomberg), that’s less than 1.00, then chances are that that bond will disappear entirely because it just won’t. there is enough support for the link to survive. Anyone with access to a Bloomberg can do the above. The above is not actually trying to predict WHEN losses will occur, just that they are expected to occur at some point in the future. It also doesn’t allow you to consider future loans that are current on their mortgage payments or that are 30 days past due that will go down the “pipeline” to more severe delinquent statuses and ultimately to realized losses. It also doesn’t try to tell you what it all means in terms of a “price” you might be willing to pay for the bond.
So let’s give this a push.
Delinquency information at the loan level
First, suppose we have access to information at the loan level and that we know not only the current delinquency status of each loan, but exactly when it entered that status. Intex provides good loan level data for deals from 2006 onwards. Loan Performance provides loan-level information for all transactions: Loan-level information is generally what Loan Performance is known for (but they don’t have very good data on capital structures nor can they generate good cash flows in the bonds like Intex does). ). The point is that loan-level delinquency information is available.
So, let’s retrieve all the loans for a particular deal into a spreadsheet from our database of loan-level information. Ideally this should be automated from within the spreadsheet so that we can always update the data whenever we need to ensure it is representative of the most current data in our database.
We now have in our hands which loans are in which delinquent condition. Now, if we simplistically project the maximum terms that all loans in FC and REO will experience before they reach their point of loss, we can derive a table of months into the future and WHEN those losses will be experienced.
For example, we can state the following:
A. Let’s say a loan has been in FC for two months: Let’s allow 6 months for the total “normal” amount of time a loan is going to be in FC so that an additional 4 months of FC time is expected for this particular loan. Then allow 6 more months for the full REO process. This means that month 10 is WHEN we expect the loss to occur.
B. Let’s say a loan is currently in REO and has been that way for 4 months. Allowing 6 months of full REO time suggests we have 2 more months left. So 2 months from now is when we think we’ll make a loss on this loan.
C. Let’s say a loan has just become 90 days delinquent for the first time. They will probably be in FC very soon, but perhaps we feel we should allow an additional month of 90 days of delinquency. Then we would have 1 month more than 90 days of delinquency. 6 full months of FC and 6 months of REO, so we expect the loss to come in month 13.
We can continue to do the above for loans 60 days past due and loans 30 days past due. And possibly take some current loans based on the idea that some of these will collapse as well.
Assume an overall “severity of loss” of 60%. According to some market participants, the 60% is becoming more and more real. This means that, given a loan amount of $100,000, you expect to lose $60,000. Apply the loss severity entry to each of the loan balances and add those loss amounts in each of the months that you have projected into the future.
The result is that you end up with a table of months into the future within which losses can be summarized, month by month. At that point we have a relatively simple table that gives us WHEN we expect losses to hit. These losses will be applied to the outstanding balance of the bond and will eventually “pay off” the principal of the bond, through amortizations, until it reaches zero. In each month, calculate how much interest the bond should receive. We then apply that month’s loss amount and decrease the outstanding principal balance of the bond so that in the next month, of course, less interest is earned. We hold this until the bond balance has been reduced to zero, at which point you no longer earn any more interest on the bond. At that point, the link has disappeared. Then add up the interest payments you received during the time the bond was still “live” and you have the amount of cash you will receive for this bond. Divide that by the currently outstanding principal of the bond and you have a price that might be indicative of what you’d be willing to pay. Note that this last award does not take into account the time value of money. It can be a “present value” (PV) enhancement of those interest cash flows and then sum the PV-ed cash flows to get a more accurate price.
It should be noted that if there are any “OCs” left at the bottom of the equity structure in the deal, you must first allocate the loss amounts to the OC before they begin to affect the bond you are evaluating. Similarly, if there are bonds BELOW the one you are evaluating, due to the fact that losses are allocated from the bottom of the capital structure up, then each of those bonds below your bond should be reduced to zero. before the loss. the amounts begin to affect your particular bond. The point is that your spreadsheet application must recover all bonuses and any OC BELOW your bonus and apply the loss amounts to EACH of your outstanding principal amounts BEFORE losses begin to affect your particular bonus. This of course means that ALL of the bonds below the one you are evaluating are also each a “Credit IO” bond.
some other observations
I want to emphasize that decreasing the FC and REO terms in the model will have the impact of decreasing the amount of time that the bonds will survive and therefore will decrease the amount of time that the bonds will earn interest, resulting in a lower price than one would be. willing to pay for the bonus. Obviously, if you’re buying, you want to pay as low as possible, so underestimating the terms will help. If you’re selling, you’ll probably want to consider longer terms so you can sell it at a higher price. These are the normal competitive interests in the market.
This represents a simplistic model but one that provides a much greater degree of flexibility than Bloomberg’s MTCS function. If done correctly, it also allows the user to adjust timelines and severities to those they are comfortable with when evaluating “Credit IOs.”
Also, by taking all of the above factors into account, the user/merchant can still perform the analysis in the way that they see fit for the environment they are in. They are not “locked” in a “black box” that they cannot see inside. There are, of course, much broader features that can be incorporated into such a model that are not within the scope of this article.
