Author Topic: Best Lending Club Analysis Site?  (Read 12984 times)

Zach

  • Administrator
  • Hero Member
  • *****
  • Posts: 622
    • View Profile
    • Email
Re: Best Lending Club Analysis Site?
« Reply #15 on: February 20, 2013, 03:41:52 AM »
Zpbsfg,

Rawraw hit the nail with his comment about circular logic. If you do an analysis of a certain set of data and come to find a correlation, that correlation will always the best fit for the same set of data. But it doesn't prove validity of correlation until the correlation successfully predicts outcome for new/different set of data.

As for BLE, I shared the original research paper of 1928 and showed BLE calculation for several loan attributes using lending club historical data on my blog last year. Later on i found out that similar BLE variations are underpinnings of loan underwriting models for consumer loans. My own BLE finding was that Lending Club did a great job assigning credit grade A through D for their loans.

The composite BLE Index that I use on PeerCube is straight derivation of original research. Actually, writing a detailed description of BLE for PeerCube is on my list of things to do. I am also working on calculating composite BLE for historical loans. Unfortunately, it is arduous task as BLE includes most, if not all, loan attributes and a multidimensional array manipulation. And every month of new historical data requires new calculations for base numbers to use for new loans.

Only time will tell how good BLE predicts the risk of new loans.

BTW, there is nothing different about lending club loans, except the term p2p, than your conventional amortized loans such as mortgage and auto loans or any other equal repayment loans. The basic models stay the same.

Btw, I am currently working on background research for returns of lending club loans as there seems to be lot of misinformation out there. It has been another of my pet peeves that seems to be becoming generally accepted way of figuring returns.

Thanks for the info, I look forward to seeing your future posts about this topic.

nonattender

  • Hero Member
  • *****
  • Posts: 713
  • I am not here.
    • View Profile
Re: Best Lending Club Analysis Site?
« Reply #16 on: February 20, 2013, 10:09:01 AM »
A simple aphorism may be instructive, here, for the un-initiated:

"Make your models fit your data --- not your data fit your models."

("Circular logic" being a higher level form of "confirmation bias".)
A little nonsense now and then is relished by the wisest men.

rev

  • Sr. Member
  • ****
  • Posts: 308
    • View Profile
    • Interest Radar
    • Email
Re: Best Lending Club Analysis Site?
« Reply #17 on: February 20, 2013, 10:29:37 AM »
I normally avoid this type of thread because some people just let lose empty criticism with no basis. Of course being a "statistician" is more complicated than just looking for past correlations and applying them to future data, but since when did someone here become the authority to tell who does that and who doesn't? I missed the coronation, I hope someone posts it on youtube.

I'm not an amateur statistician, because I'm not a statistician at all, and I have no aspirations in the field. My background is completely different, but I don't think I should be advertising credentials for the benefit of my service. Thinking now, I actually find pretty uncongenial to brag about one's credential as a source of reliability, but then repeat over and over that you can't look at past performance to predict future behavior. Shouldn't we be looking at actual performance instead? Where are the numbers? Why criticize before even looking at the evidence? "Do what I say, don't do what I do"?

But I respect everyone in the community that is making an effort to maximize returns or minimize loss. I naively expect others to have the same attitude.

Anyway, my IR01 score is around for over 7 months now, so if you think this is enough to peek a preview of its performance, serve yourself:
http://www.interestradar.com/analysis?VD=2012-08&VD=9999-99&S1=1&S1=2&S1=3&S1=4&ShowBreakdown=1&SortColumn=1
(signup required; subscription required if you've used the site for more than 30 days)

nonattender

  • Hero Member
  • *****
  • Posts: 713
  • I am not here.
    • View Profile
Re: Best Lending Club Analysis Site?
« Reply #18 on: February 20, 2013, 10:47:59 AM »
BTW, there is nothing different about lending club loans, except the term p2p, than your conventional amortized loans such as mortgage and auto loans or any other equal repayment loans. The basic models stay the same.

Out of pure curiosity, what are your thoughts on adverse selection wrt "p2p" loans versus "conventional amortized" loans?
(I understand that the mechanics of the models remain essentially the same, though am curious if you attempt to account.)
A little nonsense now and then is relished by the wisest men.

AnilG

  • Hero Member
  • *****
  • Posts: 1088
    • View Profile
    • PeerCube
Re: Best Lending Club Analysis Site?
« Reply #19 on: February 20, 2013, 02:44:44 PM »
Can you elaborate on what do you mean by adverse selection wrt p2p?

Thanks



Out of pure curiosity, what are your thoughts on adverse selection wrt "p2p" loans versus "conventional amortized" loans?
(I understand that the mechanics of the models remain essentially the same, though am curious if you attempt to account.)
---
Anil Gupta
PeerCube Thoughts blog https://www.peercube.com/blog
PeerCube https://www.peercube.com

nonattender

  • Hero Member
  • *****
  • Posts: 713
  • I am not here.
    • View Profile
Re: Best Lending Club Analysis Site?
« Reply #20 on: February 20, 2013, 04:57:30 PM »
Can you elaborate on what do you mean by adverse selection wrt p2p?

"wrt" = "with respect to".

If that's not the confusion, then, wrt the confusion, this is what I meant:

Your comment that there were no differences in "p2p" loans versus "conventional amortized" loans struck me as odd, given that
essentially every P2P platform in the world has had to combat serious adverse selection, due to a variety of p2p-intrinsic factors:

- Nature of borrower acquisition (online / targeted v. untargeted / etc)
- Nature of borrower information (single bureau, proprietary scoring products, limited datapoints)
- Unsecured nature of the notes (versus, say, auto or mortgage loans as you mention)
- Etc, etc...

The platforms seem to have made great progress in this area, so, I'm curious whether you redouble their efforts or whether you
discount the activity completely or whether you feel you simply can't access symmetric information (say, acquisition method) such
that you could even begin to factor any ways in which "pure p2p" data models may naturally diverge from "conventional" one(s)?

If you merely meant that the mechanics of the models were indifferent to the origin of the data then I understand your comment.

If you meant, however, that there are no - and that we should not expect - qualitative divergence(s) from conventionality, then -
essentially, you're denying that "p2p loans" exist - in essence! - as you will have treated p2p data without ANY contradistinction.

That make sense?  It's a fine point - and I am rather rusty - but I well remember the days when P2P platforms relied upon model
predictions where the training data was culled from "conventional" products (including revolving, auto, mortgage, secured, etc)...

We've, thankfully, gotten through those days - but, not without some serious effort to build proprietary models around selection,
though largely on the part of the platforms and never on the part of any individual modeler, at least so far as I have yet to see...

(Anyone with me?) :)
A little nonsense now and then is relished by the wisest men.

AnilG

  • Hero Member
  • *****
  • Posts: 1088
    • View Profile
    • PeerCube
Re: Best Lending Club Analysis Site?
« Reply #21 on: February 20, 2013, 08:55:23 PM »
Personally, I haven't looked in the finer distinctions between p2p and conventional lending. I assume higher information asymmetry and the lending process and terms driven by a loan broker (Lending Club) instead by a lender (ex: Banks) will result in lower profitability for p2p lenders. My comment about p2p loans versus conventional loans are more of high-level perspective on the lending process.

I don't see anything that is very different w.r.t. acquisition method (both use similar mechanics for borrower acquisition). The difference between unsecured p2p loans and collateral based amortized loans is same as the difference between unsecured loans such as credit cards and unsecured line of credits from bank and collateral based amortized loans.

The risk for p2p lender is higher due to higher information asymmetry between p2p lender and borrower which will be adjusted for by lenders by using higher discount rate and demanding higher returns. In the end, p2p lending transfers the power from lender to broker over conventional lending. With more power, comes more responsibility and more rewards. So it will be a balancing act for Lending Club but LC profitability will be much more than conventional loan brokers and even p2p lenders.

Do you think anything else should be different?


---
Anil Gupta
PeerCube Thoughts blog https://www.peercube.com/blog
PeerCube https://www.peercube.com

Keltset

  • Full Member
  • ***
  • Posts: 176
    • View Profile
Re: Best Lending Club Analysis Site?
« Reply #22 on: February 20, 2013, 10:04:32 PM »
wow... I just read this thread in entirety and felt like I should take a shower. AnilG, Rev, Bryce, and others out there that are providing models have been generally transparent. There is to a degree of secrecy (specifically with Bryce, or even Rev) as to what those models are doing but the reality is that you either trust the model and make informed decisions or your don't.

I don't have a huge amount of experience with AnilG's BLE or with rev's IR numbers. Either way the request from zpbsfg to prove performance is kind of a problem. It's easy to fit models to historical data... Forward projection and expectation meeting is a different story. If you want to criticize any form of model you need to use the model to date to determine efficiency... There is data available for Rev's, I'm not sure if AnilG has posted such a thing but you are looking for a holy grail without proving it to yourself first. If anyone's metrics out their don't meet your needs, then by all means please ask some productive or specific questions regarding the model that can help -all- of us learn. I don't doubt for one second that those heavily involved with stats here are not willing to look at and respond to valid questions and even admit error should they find such. I have found Rev, Bryce, and AnilG to all be fairly forthcoming with stats specific and valid questions and responding to them appropriately and not in a sheer defensive manor.

ehhh... I thought about just deleting this and not responding at all but I was rubbed in a negative way (even if un-intended) after reading through this... So here you are take my words for what they are worth... If they are of worth let me know what they are worth and I'll start charging...

rawraw

  • Hero Member
  • *****
  • Posts: 2756
    • View Profile
Re: Best Lending Club Analysis Site?
« Reply #23 on: February 21, 2013, 06:27:02 AM »
Quote
Personally, I haven't looked in the finer distinctions between p2p and conventional lending. I assume higher information asymmetry and the lending process and terms driven by a loan broker (Lending Club) instead by a lender (ex: Banks) will result in lower profitability for p2p lenders. My comment about p2p loans versus conventional loans are more of high-level perspective on the lending process.
I don't mean to be a nag, but I'm very familiar with the banking sector and how banks are run.  No bank makes a 9% ROA (I'm sure there are some, but just for the sake of ignoring the tails).  A P2P lender has no overhead and just the 1% servicing fee.  Banks have interest costs on the money they use to fund loans, over head costs, regulatory requirements and associated premiums, etc.  Now, if you are talking about the ROE -- the return a bank gets just on it's shareholder supplied capital, I think the market average is something like 9% in the USA.  I personally am beating that over 3 years with P2P.

People think we get less information than banks.  If you look at loans files of many banks for unsecured credits, normally you have income and credit reports.  The detail is required on big corporate deals that have assignment of leases, multiple indebtedness mortgages, appraisals, etc.  The risk we run is really hoping LC's method of ensuring its a real person on the other side works well.  But in terms of information, I don't think it's as asymmetric as I thought it was before entering the world of banking.  There is a degree of it (We can't cut up the person's credit card, loan proceeds don't go straight to the credit card company like a bank would do).  But I don't think it's enough to warrant alarm.
« Last Edit: February 21, 2013, 06:29:52 AM by rawraw »

American in Busan

  • Newbie
  • *
  • Posts: 44
    • View Profile
    • Email
Re: Best Lending Club Analysis Site?
« Reply #24 on: February 21, 2013, 07:35:29 AM »
I like Nickel Steam Roller the best overall. Great interface.

AnilG

  • Hero Member
  • *****
  • Posts: 1088
    • View Profile
    • PeerCube
Re: Best Lending Club Analysis Site?
« Reply #25 on: February 21, 2013, 11:06:08 AM »
Rawraw

I don't have experience in banking sector so I will take your words for it. Fair enough. My knowledge mostly has come from talking with actuaries and model developers for financial services firms. Based on my discussions about models development and type of information used in models and your comments about light loan files, I can only assume that the most information is crunched by models in the backend and minimal information required for final decision is provided to frontline staff.

IMO, a few important attributes about borrowers not available to p2p lenders are age, gender, marital status, profession, education, and types of different debt. Msa code will also be useful.

Sorry, I can't comment on p2p lender performances. Good for you if you are consistently beating the markets. Hopefully trend will continue long term. Inmy experience, There tends to be positive confirmation bias in such claims. I rarely hear someone saying they are underperforming the market.

Thanks for participating in discussion. I always learn something new from such conversation. If you ever in Seattle area, let me know. I will be happy to pick your brain over a beer.


I don't mean to be a nag, but I'm very familiar with the banking sector and how banks are run.  No bank makes a 9% ROA (I'm sure there are some, but just for the sake of ignoring the tails).  A P2P lender has no overhead and just the 1% servicing fee.  Banks have interest costs on the money they use to fund loans, over head costs, regulatory requirements and associated premiums, etc.  Now, if you are talking about the ROE -- the return a bank gets just on it's shareholder supplied capital, I think the market average is something like 9% in the USA.  I personally am beating that over 3 years with P2P.

People think we get less information than banks.  If you look at loans files of many banks for unsecured credits, normally you have income and credit reports.  The detail is required on big corporate deals that have assignment of leases, multiple indebtedness mortgages, appraisals, etc.  The risk we run is really hoping LC's method of ensuring its a real person on the other side works well.  But in terms of information, I don't think it's as asymmetric as I thought it was before entering the world of banking.  There is a degree of it (We can't cut up the person's credit card, loan proceeds don't go straight to the credit card company like a bank would do).  But I don't think it's enough to warrant alarm.
---
Anil Gupta
PeerCube Thoughts blog https://www.peercube.com/blog
PeerCube https://www.peercube.com

rawraw

  • Hero Member
  • *****
  • Posts: 2756
    • View Profile
Re: Best Lending Club Analysis Site?
« Reply #26 on: February 21, 2013, 06:58:26 PM »
Rawraw

I don't have experience in banking sector so I will take your words for it. Fair enough. My knowledge mostly has come from talking with actuaries and model developers for financial services firms. Based on my discussions about models development and type of information used in models and your comments about light loan files, I can only assume that the most information is crunched by models in the backend and minimal information required for final decision is provided to frontline staff.

IMO, a few important attributes about borrowers not available to p2p lenders are age, gender, marital status, profession, education, and types of different debt. Msa code will also be useful.

Sorry, I can't comment on p2p lender performances. Good for you if you are consistently beating the markets. Hopefully trend will continue long term. Inmy experience, There tends to be positive confirmation bias in such claims. I rarely hear someone saying they are underperforming the market.

Thanks for participating in discussion. I always learn something new from such conversation. If you ever in Seattle area, let me know. I will be happy to pick your brain over a beer.


I don't mean to be a nag, but I'm very familiar with the banking sector and how banks are run.  No bank makes a 9% ROA (I'm sure there are some, but just for the sake of ignoring the tails).  A P2P lender has no overhead and just the 1% servicing fee.  Banks have interest costs on the money they use to fund loans, over head costs, regulatory requirements and associated premiums, etc.  Now, if you are talking about the ROE -- the return a bank gets just on it's shareholder supplied capital, I think the market average is something like 9% in the USA.  I personally am beating that over 3 years with P2P.

People think we get less information than banks.  If you look at loans files of many banks for unsecured credits, normally you have income and credit reports.  The detail is required on big corporate deals that have assignment of leases, multiple indebtedness mortgages, appraisals, etc.  The risk we run is really hoping LC's method of ensuring its a real person on the other side works well.  But in terms of information, I don't think it's as asymmetric as I thought it was before entering the world of banking.  There is a degree of it (We can't cut up the person's credit card, loan proceeds don't go straight to the credit card company like a bank would do).  But I don't think it's enough to warrant alarm.
It's no problem man.  Just remember a few things.  The first is only the largest banks and financial services companies are going to be using actuaries (There are what, 4000 banks in USA?).  You are clearly a smart guy and love math.  But the banks that use guys like you aren't the most common, they are just the largest.  So remember that there are plenty of banks out there that don't do those things and fare pretty well compared to the big boys, with much less risk (arguably).

Secondly, I plan on making a post comparing P2P to the big banks and how I think these filter strategies may go wrong if blindly trusted.  I see a lot of similarities and I want to make sure people understand the risks when getting involved in such activities.

And I use sites like NSR and Lendstats to see what the 'market' return on LC is.  That may not be a correct metric, but it's the best I'm aware of. 

I hate correcting people on forums, especially when I realize I'm probably much younger than the rest of you.  But hopefully I can add a little insight from my limited experience thus far in finance and banking.

AnilG

  • Hero Member
  • *****
  • Posts: 1088
    • View Profile
    • PeerCube
Re: Best Lending Club Analysis Site?
« Reply #27 on: February 22, 2013, 01:05:44 AM »
Peter tweeted this blog post that I thought you may like. The blogger compared P2P loans with Asset based Securities (ABS).

http://blog.samuelhupc.com/2013/02/21/p2p-lending----risk--reward-analysis-.aspx


Secondly, I plan on making a post comparing P2P to the big banks and how I think these filter strategies may go wrong if blindly trusted.  I see a lot of similarities and I want to make sure people understand the risks when getting involved in such activities.
---
Anil Gupta
PeerCube Thoughts blog https://www.peercube.com/blog
PeerCube https://www.peercube.com

rawraw

  • Hero Member
  • *****
  • Posts: 2756
    • View Profile
Re: Best Lending Club Analysis Site?
« Reply #28 on: February 22, 2013, 05:10:38 AM »
Peter tweeted this blog post that I thought you may like. The blogger compared P2P loans with Asset based Securities (ABS).

http://blog.samuelhupc.com/2013/02/21/p2p-lending----risk--reward-analysis-.aspx


Secondly, I plan on making a post comparing P2P to the big banks and how I think these filter strategies may go wrong if blindly trusted.  I see a lot of similarities and I want to make sure people understand the risks when getting involved in such activities.
That is interesting, thanks for sharing

RomanLegend

  • Newbie
  • *
  • Posts: 49
    • View Profile
Re: Best Lending Club Analysis Site?
« Reply #29 on: July 16, 2017, 06:01:28 PM »
Hello,

Please check out my YouTube video and please give my LC Investment strategy a try.  It is free for now in Beta!  :) 

YouTube Intro:  https://www.youtube.com/watch?v=dIE5bVgj8eE&t=3s

Register Here:  LCPicks.com

Let me know if you have any questions along the way just send me a message and I'll help out.

Thanks,

-Chris