Asset securitisation has moved through a number of phases over the last decade. From an exciting new tool to allow the disbursement of risk through a period where more novel assets were becoming securitised. We explored in Bounded Rationality the way in which asset securitisation became so complex that it was all too easy to lose sight of risk which was the very reason for their introduction in the first place. We all know how that ended... But securitisation hasn't died, it lives on and will continue to form a large part of the financing world so it's worthwhile spending some time getting re-acquainted with our old friend, getting back to basics and perhaps uncovering where the traps await the unwary investor...
Lets revisit the basic mechanics of a simple mortgage asset backed security. There's three parties to a securitisation: -
The company who has provided the 'loan' to an individual (lets call them 'The Bank');
The person who took out the mortgage from The Bank (lets call them the 'Customer', although you'll normally see them called the Obligor in securitation speak);
The entity who exchanges cash now for a share of the future profits on that mortgage ('The Investor')
Provided all three parties keep their part of the bargain, everyone is happy. The Bank was accurate in their prediction over the income on the Mortgage, The Customer repays The Bank on time and until the mortgage is complete and The Investor pays the cash upfront and is happy to wait until the end of the Securitisation for his money back.
Now as we know, things don't go to plan... Not every customer will repay their loan and certainly not every customer will repay their loan to the end of the mortgage (when was the last time you heard of someone paying their mortgage for 25 years without switching to a cheaper product, moving house or advancing more money). With this in mind then, the predictions over mortgage repayments are the single most important factor in understanding how profitable the Securitation vehicle is going to be. Now before we get into the analysis methods, it is important to recognise one final element of the many variables that go into the assessment of the Mortgage Pool - the quality of the customer...
There's much rhetoric about the stereotypical 'unemployed American' who was given $100,000 to buy his own home. Clearly, extreme examples of poor lending and even poorer securitisation 'practice' have their part to play in the story of the credit crisis, but the more run of the mill lending is where we need to focus our attention. Much like the concept of insurance where many people pay a premium for loss protection and losses are paid from that pool of money, the securitisation 'pool' of assets (mortgages for example) will see some losses (defaults, house price value decline) and those losses are in effect covered by the payments on other assets in the pool. The issue for insurance and securitisation is where the losses exceed the capability of the pool to repay it. The quality of the underlying customer is the first place we should look for signs of confidence the pool will be repaid.
Customers are complex animals, they take all shapes and sizes and the outward appearance is only one thing we should consider when reviewing their ability to repay. How old are they, what do they do for a living, how long have they been doing that job and how much are they paid. But under the skin we need to think about their existing outgoing expenses (big salaries can lead to big expenditure - the more you earn the more you spend) so gross income is not correlated to disposable income. What about their past credit history and more recent changes in their behaviour (did the customer recently take out a few extra credit cards around the time of taking the mortgage, does this indicate a credit problem that has yet to surface?). With a large pool of perhaps tens or even hundreds of thousands of mortgages it is impossible to answer these questions for every customer individually so the wary investor is reliant on understanding the controls the company has put around its underwriting and on the accuracy of the records kept, both system based and physical. The scale and depth of any due diligence audit at the start of the investment in the pool is therefore critical in gaining confidence in the repayment forecasts.
But that's just the customer, what about the asset itself - the house? How was the valuation carried out, where is it, what does the geographical distribution of the properties look like? When the customer pays as expected, nobody is really worried about the asset value, in the event of a default and a potential loss the value of the asset comes into play and the valuation done at the outset and the current valuation is now critical. So, we're back to understanding the controls and the accuracy of data that relates to the asset valuations.
There's a third aspect that will drive our pool performance, and that is the product that the customer was sold. The proportion of the mortgages that reside on Fixed Rate products compared to those on a Standard Variable Rate has implications for both margin (fixed rates being squeezed when rates rise) and on prepayment performance (when perhaps a very market driven fixed rate expires into a punitive variable rate). The various sub-pools of products and how they will change over time is another key area the wary investor should understand and stress.
Finally, there is some degree of help in this initial assessment of the pool , and that is the historical performance of both the wider mortgage business (it gives an indication of overall lending and management capability) and of the performance of similar pools of mortgages to the ones in the securitised pool. Helpfully, this historical analysis can be used both at the initial assessment of the pool and the ongoing monitoring of performance. It is this quantitative assessment we will look at in more detail here and which is used alongside the qualitative assessment detailed above.
Quantitative Analysis Of Mortgage Backed Securities
We've already touched upon two of the most significant events that a mortgage will experience and that is the early settlement of the mortgage and the complete non-payment (default) of the mortgage. We'll use these two events as the basis of the modelling we are about to undertake. The most important element of both these events is the point at which they occur in the life of the mortgage. There is a subtle but important point to make here and that is we are for the time being ignoring the traditional 'point in time' (dynamic) analysis "this mortgage defaulted in September 2012" and referring to the age of the mortgage itself (static analysis) "this mortgage defaulted in month 7". This gives you the ability to align many mortgages with different starting dates on the same basis and build a picture of a 'typical mortgage' profile. The capture of this event information is one of the most difficult aspects of the development of static pool analysis and great care should be taken to ensure that the definitions of events (what counts as a mortgage prepayment?) both matches the current business process and is consistently maintained over time.
Dynamic Analysis
The event date itself forms the basis of traditional 'dynamic pool' modelling. That is to say we can plot these events in a traditional 'point in time' way, Prepayments in January, February, March and look at those patterns of behaviour over time. It informs medium to long term trends (prepayments rising or falling over time), seasonality (the drop in prepayments before Christmas and the rise in mortgage arrears after Christmas !) and helps to identify short term or 'intra month' patterns (for example a short spike in arrears caused by a failure in the payment system)
Close scrutiny of the dynamic events of the pool and the wider business will therefore yield valuable information about the assets and capability you are buying into. Take the example of a mortgage pool that has seen two spikes in its arrears performance over the last twelve months. On investigation it seems that these two spikes were caused by a failure in the direct debit (AP for the US) system. The impact was short lived but none the less had a impact on the business, and that leads you to ask further questions (what happened to customer complaints over the same period?) and ultimately to the real question you are trying to answer : "is this company well managed?".
Dynamic pool analysis informs your view over time, it can be plotted against real events such as macro-economic changes, changes to management and adjustments to lending policy. However, it has a limitation in that is does not tell you how a typical mortgage behaves over time. All the events that happen 'in January' are the sum of all the events on mortgages that are 1 month old to mortgages that are 150 months old. This is where static pool analysis comes to the rescue.
Static Pool Analysis
For Static Analysis we need to convert the 'event date' (in this example the prepayment date) against the advance date of the mortgage and derive a 'static event date' (the months since the loan advance). This instantly gives you the ability to treat every mortgage as is of it were advanced on the same day and view how those mortgages perform in the context of the mortgage itself. A good example of this would be the rise in prepayments in month 25 of a pool of 2 year fixed rate mortgages.
This bring us to a place where we can form a prepayment curve for the portfolio in question and begin to understand how they perform over time. We can imagine a pool of perhaps tens of thousands of mortgages, some of which have had partial prepayments in the past, some of which had settled entirely and some of which have performed as expected with no additional or late payments. Taking that imaginary pool we can again calculate the 'static event date' (the number of months after the loan origination the event took place) and plot those outcomes as a time series based on the Static Event Date.
This gives you a useful quantum of the prepayments over time but to make it really useful we need to understand what proportion of the pool is being eroded by prepayments and this is very simply the sum of the total prepayments for each month divided into the total original advance amount of the mortgage.
Very quickly then we have calculated the performance of the pool at a macro level and can use that base prepayment curve in a whole variety of ways to model future performance. But we rarely model and manage a pool at the macro level... We like to understand how different parts of the pool behave and we can do exactly the same process but for subsets of the overall portfolio. For example, how do fixed rate assets perform differently to floating rate assets? How about sub prime lending compared to super prime lending, what about customer demographics or regional variations? Using mortgage level attributes you can easily calculate the same prepayment curves for any variety and combination of factors. The trick here is not to over analyse and begin creating hundreds of small pools which you monitor. Rather you need to arrive at a suitable level of granularity where you can understand the customer, product and mortgage attribute dynamics, particularly where there are strong correlations of good or bad performance. Using stratification tables is a useful way of highlighting hotspots in your portfolio.
Stratification Tables
Stratification analysis is exactly as you would imagine - it is primarily focused on the different 'layers' within a portfolio and how behaviour manifests itself as those layers are examined. Imagine taking a core through layers of rock, moving deeper and deeper through the sample you'd see changes in colour and texture. From a portfolio perspective, you can choose where to drill to see that core sample. Lets use 'house value' as an example, where we have decided to examine the portfolio pool by breaking it down into bands of house value. Perhaps grouping the portfolio into mortgages that have a house value of £75k - £100k; £101k - £150k etc. Now, in terms of what we want to analyse we need a single point in time measure that we can apply to all those individual pools to compare behaviour. In this example we might use "36 months prepayments" (where we are measuring the total erosion of the original pool by the 36th month). However, unlike comparing the prepayment curves for these different pools on the same chart (which is useful in itself for comparing the life cycles of the mortgages) we can compare lots of metrics alongside each other for the 'stratification' in question. This allows you to see the relationship between metrics and come to deeper understanding of the data leading you to examine hotspots more closely.
Static Pool Analysis is a powerful tool and used with Stratification tables like the ones above you can begin to get some real insight into the nature of your portfolio. However, Static Pool analysis is a very slowly changing measure and since it is concerned with the performance of your portfolio over a long time period, a single 'Prepayment Curve' for example changes only subtly each month. The movement of the curve once established is almost glacial and changes in performance can take months to become apparent. This is where we need to combine the powerful analysis of Static Pool with the more immediate measure of Dynamic Analysis and produce some Cohort Analysis.
Cohort Analysis
The basis of Cohort Analysis is the Static Pool data we discussed above, but we add a Dynamic Analysis perspective which shows more readily and immediately changes in performance.
Lets take a key performance indicator of "12 Month Prepayments" (where we are monitoring the amount of prepayments or balance erosion by the 12th month of the mortgage). If we use some sample Prepayment Data we can see the overall Prepayment Curve and within it the single 12 month data point (in Red). Cohort Analysis takes this single data point and measures its change over time (on a Dynamic month by month basis). By calculating the 'Period To Date' averages of that data point over time we can quickly see whether 12 Month Prepayments are stable, rising or falling.
In the sample above then, we can see 12 month prepayments are falling over time (something that is already quite obvious in the numbers in this small sample) but also that our final data point (0.18%) ties in with the current Static Pool position for point 12 (the red dot). What you have then is the Dynamic month by month performance, tied together with the that Static Analysis and a very easy approach to track the long run and short run metrics at the same time.
In this short article we started to discuss the importance of Securitisation as a financial tool and the factors that make up the main considerations for the investor. Even using some fairly basic approaches we can begin to understand the make up and history of a portfolio which leads us to ask the right questions about the way in which the portfolio is managed and the key factors which underpin its assets and performance.