Bill Draving Company, Inc.

Mortgage Banking Consultant


BDCI

About

Resume

Services

Contact

EMAIL

Portfolio Stratification - The Mechanics


Stratification Coding

Creating a stratification is a little like taking the family car apart. There are many pieces and you don't recognize most of them. Naturally, you can never get it back together correctly. Leftover pieces are the plague of anyone foolish enough to take things apart.

My personal role in life has been that of the mad scientist; forever exploring why things work and how to make then work better. My home and office is littered with things that will never work again.

But you can’t do that to a portfolio worth 30 million dollars. You must take it apart and reassemble it so the final result produces a greater value than the individual pieces.

When you break apart something that has a hundred thousand pieces you better have a plan or schema that is logical and consistent. Yet too often, analysis schemes are completely random. Creating stratification ranges of .28% interest rate spreads makes no sense at all.

In Part 1 of this series, you learned the importance of stratification to achieve your secondary marketing goals. This article delves into the how-to mechanics of it all.

Our basic task in a stratification is as follows:
1. Decode the portfolio.
2. Create the stratification ranges.
3. Run reports for each stratification.
4. Analyze the results
5. Use the stratification results to improve the value of the portfolio.

Decode the Portfolio

Creating elaborate reports provides as much value as sorting all the parts when you take your car apart, It is interesting and helps your personal understanding of the task but it does not improve the value of the car.

Stratification and decoding is certainly an analytical task. It is usually impossible to get anyone’s attention when you start talking about things like codes, fields and ranges. These are words used by the technical people. To help you on this short journey into computer world let’s quickly define a few basic words:

FIELD. Fields are the actual data values in the computer record. A computer record is composed of hundreds of "fields". The physical format of each field is exactly the same for every loan

TYPE. Every portfolio is composed of thousands of different types of loans. A single loan may have as many as 50 different types.

CODES. Codes define the qualitative characteristics; investor, insurer, collection agent. A portfolio has many types of loans. Every loan type has many different codes. The codes distinguish one loan from another.

Also, there are several coding formats
1. User defined codes.
2. Standard code.
3. Industry defined codes.
4. Geographic codes.
5. Due Diligence codes.

Coding Schemes

Stratifications can be quantitative as well as qualitative. They are coding schemes that you define. You can use numbers, letters or loan groups. Stratifications fall into the following categories:

a. Balance. Every $10,000. There are hundreds of different balances on a typical portfolio. The most important are the large balances!

b. Rate spread . A typical portfolio has dozens of different rate fields. Rates never have codes.

c. Date ranges. Dates are stratified by one, three or five year spreads. Date fields include first payment date, due date, acquisition date and sale date.

d. Type code. A few type codes are : insurer, amortization, loan purpose, property type, lien type.

e. Categories. Categories are usually defined by each company. They include investor, acquisition code and branch code.

Grouping loans

A group is a bunch of loans that have similar characteristics. One loan could be in many groups. We usually stratify the portfolio into groups that have similar characteristics. You may have your own special group structure already defined.

One of the most critical groupings is your acquisition groups. A good stratification will break out each acquisition to show the original balance, the current balance, the prepayments and a valuation comparison.


Standard Codes

A complicated stratification task is the comparison of two different portfolios. The worst case is when the servicing systems are different. To handle separate systems we create an intermediate coding system. We translate both systems' codes to a common format. This format is the same for every system. We "decode" each portfolio. The second step is to create the stratification categories. Since each portfolio has a common set of codes, it is easy to compare the stratifications side by side. Finally, based on these common codes we can create new aggregations.

We use the following common coding fields:

CDTYPX
CDARMX
CDPAYX
CDINVX
CDDELX
CDPREX
CDGRPX
CDSTAX
CDACQX
CDSALX
 Loan Type
Amortization types
Payment type code
Investor code
Delinquency code
Prepayment code
Valuation code
Geographic code
Acquisition code
Sale/disposition code

 Each field, or key, contains all critical information about that set of loan characteristics. This decode-encode system is also helpful when applying information to the portfolio data.

Stratification fields

This is a simple checklist of fields that everyone else uses for stratification purposes. There is a very definite purpose for each field.

Loan balance
Original loan balance
Loan interest Rate

Loan insurance type
Amortization type
Geographic state
Delinquency category
Due date
Foreclosure category
Bankruptcy category

 Loan amortization term
Loan maturity term
Loan age
Remaining loan term
Maturity date
First payment date
Origination date
Call or balloon date
Acquisition date
Sale date
Prepayment rate
Loan to value ratio
Service fees
Guarantee fees


System Overview

Now that you have mastered the details, let me preset a brief overview of the system. Every system has two basic design considerations:

1. What do I want from the system and
2. Can I justify the cost of my wants.

 When you construct a stratification system, look at these "wants":

  • How big a system should I have.

  • How many ranges do I want to create for each field.

  • Is there any benefit from this information. Does it tell me something I want to know.

  • Can I get consistent results each month.

  • Do the results provide month to month comparisons

How Costly

The factors which most affect the cost of the system are:

Computer Time - The time needed to process your data is going to determine the amount of information you receive. Every piece of hardware and every piece of software will produce different results. Purchase only the fastest computer system available. Stratification takes a long time. It is best done on a personal computer where there is a single fixed cost. There is not enough money in the world to run portfolio analysis software on a mainframe.

Disk Space - Today, disk space is the only critical variable. What was a huge disk space three years ago is now inoperable for most computer systems. Two gigabytes is a minimum disk requirement for all new systems. I know it sounds ridiculous.

Reports Needed - How many reports can I actually read and analyze? Don't be a company that spends 90% of the time producing information and 10% of the time analyzing it. Remember, computers can help analyze as well as produce information.


Why do we need stratifications?

1. Comparisons - Period to period comparisons are easier with a stratified portfolio. - The greatest question ever asked - "I bought a portfolio 14 months ago. How much is left, what is the value today and how good were my valuation assumptions?" If you can answer this question for every acquisition, you're doing a good job.

2. Prepay Assumptions - Stratifications help determine prepay assumptions. - You can not apply prepayment assumptions using interest rate alone. It simply is not accurate. You are saying that every loan has the exact same risk exposure. A loan in Alaska is not the same as a loan in New England. You must go deeper than interest rate to apply prepayment assumptions.

3. Evaluation assumptions - The only way you can avoid endless keypunching is to use a stratification database. Once you decode the portfolio you can automatically apply all your assumptions to thousands of segments in a matter of seconds.

4. Discount rate risk analysis - Everyone wants to apply a discount rate to each loan in the portfolio. This seems reasonable since each loan may have a value of several thousand dollars. Actually, a typical loan is worth more than the family car.

5. Servicing productivity - Servicing productivity is measured primarily through delinquency ratios and loans serviced per employee. Decoding the portfolio will tell you the productivity for each type of loan.

6. Merge industry databases thru stratification ranges - Industry databases are readily available. This is information that has been accumulated from many companies and put into a computer database. The only problem is how to apply this information to your portfolio. Although it is difficult, you can use stratification matrices to put this information into your own analysis.

7. Loan performance - We measure loan performance by assigning a risk factor to each stratification range. We use these factors to create a giant matrix that calculates a single factor for each loan. You can use this performance factor to buy loans, sell loans or reward your managers.

8. Measure dispersion - Dispersion is one of those technical terms that you either understand or you don't. Statistically correct stratification ranges are the only means to determine dispersion.

9. Evaluate portfolio performance - The only way you can measure performance is thru stratification.. It is impossible to measure the performance of one loan. The results don't mean anything. One loan is either current or delinquent, it has a high balance or low balance, it is FHA or VA. Only thru proper aggregation can we build a picture of performance that we can apply to tomorrow’s loan.