Author Topic: New Feature: Secondary Market Recommendations  (Read 14831 times)

turing

  • Full Member
  • ***
  • Posts: 222
    • View Profile
    • PeerToPeerQuant.com
    • Email
New Feature: Secondary Market Recommendations
« on: February 04, 2015, 01:20:55 PM »
A number of people asked us to add recommendations for secondary market notes because they can't buy on the primary market.  We just launched the service today.

Basics:

All notes recommended must pass these criteria:
  • Pass our model's qualifications (based upon info at note approval)
  • Current on payments
  • FICO score hasn't dropped more than 15 points from original FICO score

You can click a 'history' link to see full details on the note.  This is always free.

Clicking 'invest' button has same pricing structure as primary market notes.  It takes you to Foliofn with that note ready to add to your cart.

Full info and screenshots:
http://blog.peertopeerquant.com/2015/02/note-recommendations-for-lending-club.html

This is a new feature so please give us your feedback.

rawraw

  • Hero Member
  • *****
  • Posts: 2793
    • View Profile
Re: New Feature: Secondary Market Recommendations
« Reply #1 on: February 04, 2015, 01:25:47 PM »
The model takes into account the premium/discount and fee?

turing

  • Full Member
  • ***
  • Posts: 222
    • View Profile
    • PeerToPeerQuant.com
    • Email
Re: New Feature: Secondary Market Recommendations
« Reply #2 on: February 04, 2015, 01:43:23 PM »
The model takes into account the premium/discount and fee?

Our model does not decide if that premium/discount is a "good" investment or not.  If that is what you are asking.

The model just looks at borrower's info at loan approval to make sure it meets our qualifications (just like primary notes), but also makes sure the note is current now and hasn't had a long-term FICO decline.

We basically cull the list down for people, but they still need to do some due diligence before investing.  For instance near term FICO drops are not accounted for at this time.

If I misunderstood the question, please let me know.

rawraw

  • Hero Member
  • *****
  • Posts: 2793
    • View Profile
Re: New Feature: Secondary Market Recommendations
« Reply #3 on: February 04, 2015, 01:47:36 PM »
The model takes into account the premium/discount and fee?

Our model does not decide if that premium/discount is a "good" investment or not.  If that is what you are asking.

The model just looks at borrower's info at loan approval to make sure it meets our qualifications (just like primary notes), but also makes sure the note is current now and hasn't had a long-term FICO decline.

We basically cull the list down for people, but they still need to do some due diligence before investing.  For instance near term FICO drops are not accounted for at this time.

If I misunderstood the question, please let me know.
No you got it. I guess I'm confused what a note being selected actually means and worry your users may have different understanding of what you are claiming to pick. A note is being selected based on credit data and original interest rate being ok but ignoring the current yield? Is that correct

turing

  • Full Member
  • ***
  • Posts: 222
    • View Profile
    • PeerToPeerQuant.com
    • Email
Re: New Feature: Secondary Market Recommendations
« Reply #4 on: February 04, 2015, 02:11:24 PM »
The note is selected based upon the original credit data, status = "current", and FICO score not dropping too much since approved.

We do not take into account original interest rate or YTM in our recommendations.  We show that info to users when the see our recommendations.  Below is a screenshot.

Does that make sense now?

Fee

  • Full Member
  • ***
  • Posts: 109
    • View Profile
Re: New Feature: Secondary Market Recommendations
« Reply #5 on: February 04, 2015, 03:08:23 PM »
The note is selected based upon the original credit data, status = "current", and FICO score not dropping too much since approved.

Does it look at all FICO scores along the way, or just the current vs the original?

turing

  • Full Member
  • ***
  • Posts: 222
    • View Profile
    • PeerToPeerQuant.com
    • Email
Re: New Feature: Secondary Market Recommendations
« Reply #6 on: February 04, 2015, 03:36:30 PM »
The note is selected based upon the original credit data, status = "current", and FICO score not dropping too much since approved.

Does it look at all FICO scores along the way, or just the current vs the original?

At this point, only the current vs original.  Lending Club doesn't provide the historic FICO scores in any download files so we don't have a record of all historic scores.

If users think it would be helpful, we might be able to collect FICO scores on notes and show recent changes.  This would take more development work so we didn't include it in the initial service until we heard feedback from users.

rawraw

  • Hero Member
  • *****
  • Posts: 2793
    • View Profile
Re: New Feature: Secondary Market Recommendations
« Reply #7 on: February 05, 2015, 04:56:45 AM »
The note is selected based upon the original credit data, status = "current", and FICO score not dropping too much since approved.

We do not take into account original interest rate or YTM in our recommendations.  We show that info to users when the see our recommendations.  Below is a screenshot.

Does that make sense now?
But if it's not based on yield, what's the difference between a pick and non pick? Probability of default?

turing

  • Full Member
  • ***
  • Posts: 222
    • View Profile
    • PeerToPeerQuant.com
    • Email
Re: New Feature: Secondary Market Recommendations
« Reply #8 on: February 05, 2015, 10:31:24 AM »
But if it's not based on yield, what's the difference between a pick and non pick? Probability of default?

Ah, I see what you are asking now.  Yes, it is probability of default.  The picks are notes where we think the borrowers have a lower chance of default than the interest rate they are paying.

I did misspeak previously when I said interest rate isn't factored in.  We do consider that relative to the expected default rate. 

For notes that are NeverLate, this is a good way to think about the picks:  These are notes we would have invested in when they were being approved and nothing has changed since then to tell us other wise (current on loan, made all payments, and FICO is about the same).

We also include loans that have been late because some people like to bargain shop and find those diamonds in the rough that have been late once like 6 months ago and they feel are worth the discount.

rawraw

  • Hero Member
  • *****
  • Posts: 2793
    • View Profile
Re: New Feature: Secondary Market Recommendations
« Reply #9 on: February 05, 2015, 10:34:22 AM »
But if it's not based on yield, what's the difference between a pick and non pick? Probability of default?

Ah, I see what you are asking now.  Yes, it is probability of default.  The picks are notes where we think the borrowers have a lower chance of default than the interest rate they are paying.

I did misspeak previously when I said interest rate isn't factored in.  We do consider that relative to the expected default rate. 

For notes that are NeverLate, this is a good way to think about the picks:  These are notes we would have invested in when they were being approved and nothing has changed since then to tell us other wise (current on loan, made all payments, and FICO is about the same).

We also include loans that have been late because some people like to bargain shop and find those diamonds in the rough that have been late once like 6 months ago and they feel are worth the discount.
So the way I should interpret this is that at par  this note is good. At a discount, this note is great. And at a premium, we do not know if this is good or bad. Is this correct?

turing

  • Full Member
  • ***
  • Posts: 222
    • View Profile
    • PeerToPeerQuant.com
    • Email
Re: New Feature: Secondary Market Recommendations
« Reply #10 on: February 05, 2015, 12:21:39 PM »
Yes.  For NeverLate notes.  That is a great way to look at it.

DEFINITELY CHECK RECENT FICO CHANGES THOUGH.  We only compare most recent vs original.  Sometimes FICO can decline a lot in short term but still stay above original.

Would recent FICO change be useful to you as a criteria?

rawraw

  • Hero Member
  • *****
  • Posts: 2793
    • View Profile
Re: New Feature: Secondary Market Recommendations
« Reply #11 on: February 05, 2015, 02:11:31 PM »
Yes.  For NeverLate notes.  That is a great way to look at it.

DEFINITELY CHECK RECENT FICO CHANGES THOUGH.  We only compare most recent vs original.  Sometimes FICO can decline a lot in short term but still stay above original.

Would recent FICO change be useful to you as a criteria?
I think so, as preliminary research done by Anil suggests its predictive (and you'd expect it to be).  I've often thought you can probably quantify the impacts of FICO trends to determine if a note is a good/bad pick with the original credit data.  But it seems like it'd be hard to collect the data.  People like Fred and Core may already do this as well.  I know for debt consolidation loans, there seems to be a trend of a FICO pop after origination if they actually paid off credit cards.  And some loans maintain their new scores and others start dropping again.  When my account was smaller, I paid very close attention to this and sold all the latter.  My account is too big to pay attention to the pop, so I just pay attention to the drop.

rawraw

  • Hero Member
  • *****
  • Posts: 2793
    • View Profile
Re: New Feature: Secondary Market Recommendations
« Reply #12 on: February 05, 2015, 02:19:49 PM »
Now that I understand, I still don't get why the model doesn't change to use current YTM instead of interest rate.  Seems like the same logic would hold -- comparing how it is priced to its default probability to find notes that meet whatever hurdle rate you have set.  But I may be missing something

Also, a bit of unsolicited advice.  I somehow think this current offering may come back to haunt you.  While it's noble you are making an effort to meet customer demands, if they don't acutely understand what these suggestions are and are not they may fault the entire model for being broken just because of user error.  May be something to consider

turing

  • Full Member
  • ***
  • Posts: 222
    • View Profile
    • PeerToPeerQuant.com
    • Email
Re: New Feature: Secondary Market Recommendations
« Reply #13 on: February 06, 2015, 10:35:52 AM »
Also, a bit of unsolicited advice.  I somehow think this current offering may come back to haunt you.  While it's noble you are making an effort to meet customer demands, if they don't acutely understand what these suggestions are and are not they may fault the entire model for being broken just because of user error.  May be something to consider

Good feedback.  I can definitely understand what you mean.  That is something we will need to discuss and consider the pros/cons.

Now that I understand, I still don't get why the model doesn't change to use current YTM instead of interest rate.  Seems like the same logic would hold -- comparing how it is priced to its default probability to find notes that meet whatever hurdle rate you have set.  But I may be missing something

Our genetic model doesn't rank notes as better or worse than the others.  They either pass or they fail.  Just like other criteria based selection models.  That being the case, and all else being equal, a note with a bigger discount would be a better investment.  That is why we rank them by markup instead of YTM.