Friday, May 13, 2016

The Empiricals' New Clothes...



The rule of law….?
One of the credit unions with which I have a chance to work received a diktat from Duke Street this Spring requiring the use of risk-based lending.  Some of NCUA's "capital market specialists" had validated through "rigorous and robust" analyses that the use of credit scoring in lending and rate setting was "empirically and statistically accurate".


Snap, Crackle and Pop Analysts !!
(All Puffed Up!)
Evidently using the same modeling skills and analytical techniques developed in overseeing the Corporate system, these NCUA  experts even mentioned sophisticated institutions like Fannie and Freddie ( both currently in conservatorship i.e. - bankrupt!) as examples to follow when formulating "best  practices" lending and pricing .

No I'm not making this up; wish I were....

But, why don't you decide for yourself about the "empirical truth" of risk-based lending theory…..  




Q #1: A credit bureau score is based on the analysis of millions of individual credit profiles and can forecast the likelihood of default within the next 24 months of a particular risk segment (generally CUs use ranges from A-E, with A being good and E being not so!).  For example, the A group may have a 720+ score, an implied default rate of 1/2 % and carry an interest rate of 4.00% for a car loan; the E group may have a score of less than 620, an implied default rate of 10%, and carry an interest rate of 15% for a car loan.  

Is the credit score determination by the credit bureau "empirically and statistically accurate"?

A #1: Yes, believe we can all agree that at the 
"macro" level the analysis regime is correct and in this example that 10% of those in the E segment will default.

The analysis is absolutely empirically sound!

Q #2:  Using that E group again as an example, if it's agreed as an empirical absolute that 10% will default;

What does it empirically say about the other 90% in the E group?

A #2:  Empirically it says that

... 90% in the E group will pay and not default!


Q #3: Can the credit bureau or the credit score tell you if a specific, individual E member is in the 10% "default group" or the 90% "will pay" group?

Can the model tell the CU in advance which member will default?

A #3:  According to the credit bureau the empirical and statistically accurate answer is 


"NO".

Q #4:  When did it become alright to over-charge in a big, big way so many innocent, "will pay" consumers?  Is this NCUA's definition of " helping members of modest means"?

A #4:  Well, yes!  You can bet your sweet FICO these Agency endorsed loan pricing techniques will certainly help keep folks means modest!


 Q #5:  But those NCUA capital market specialists and 

… econometric model magicians have a deeper, clearer understanding of all this, don't they?

A #5:  Most folks outside the Beltway, believe these specialists' track record is empirically embarrassing and statistically extremely expensive... as with this case, the model appears to be wrong about 90% of the time at the individual level!


Well, a "90% false positive"is about par for the course these days it seems…. 


More mis-"guidance", cut from entirely shabby, empirical cloth.



Why don't you raise your voice on behalf of the consumer?

(… don't wait for never!)

3 comments:

Mike Higgins said...

It should be called "perceived" risk based pricing. For 90 percent, there was no risk, in your example. The 90 percent need to see some form of economic renumeration for holding up their end of the promissory agreement. Sadly, most do not. They were mis-classified and paid the price for it.

tony costanzo said...

Jim - We need the 90% to pay the premium for the 10% that default. It is like the Corporate Federal Credit Union failures / conservatorship / liquidation. The 90% that survived paid the premium for the 10% that went into default. So the 90% did not create the failure of U.S Central, WesCorp, Members United FCU, etc. - they must pay. And soon credit unions will pay for the Taxi Medallion Credit Unions. It is the NCUA Business Model. Take from the pool of those still standing to pay the costs of those that went belly up and failed.

Anonymous said...

I don't believe NCUA has any business telling a credit union to use risk based pricing but it is in their purview to share best practices with credit unions who might be struggling with profitability due to high risk lending that does not charge for the risk taken. There are only a few out there who can divine if any member will actually pay their loans back (Jim's folks are the exception), therefore most of us have to use risk based pricing to offset the risk we take in that portion of our portfolios. What is often overlooked in these discussions is the fact that the 90% who do pay us back despite the higher interest rate actually saved 10% or more compared to where they formerly borrowed (at the pay here dealership).
Georgia Birddog