Author Topic: Automeated investing questions  (Read 2052 times)

OleBill

  • Newbie
  • *
  • Posts: 22
    • View Profile
    • Email
Automeated investing questions
« on: July 13, 2017, 06:17:52 PM »
I have a dilemma. If you have been following my analysis blog, billlanke.blogspot.com, you know that I’m very concerned about the performance of my Lending Club account. To simplify, I opened an IRA account in early 2015. I purchase over 3,000 notes in the next 2 years using Automated Investing. The first year my account balance grew by 8%. This annual rate fell regularly finally going slightly negative by mid-2017. My analysis concluded the two biggest factors were “bad luck in the notes selected for me” and a note distribution skewed toward the riskier notes.
 
I did develop a technique that identified my bad luck impact. When I added this back my growth rate improved to 2.92%. Lending Club estimates that my “Adjusted Net Annualized Return” for my current notes is 2.86%. When I look at Lending Club’s automated investing estimated returns, none are anywhere near as bad as mine. This leads me to several questions and I would be interested in your opinions.

Do other large investors rely primarily on automated investing?

Is this approach a good idea?

Is there something inherent in the automated investing process that selects notes that are worse than randomly selected notes in the same grade?

Are Lending Club’s estimated returns consistently overblown?

I see a lot of people bailing out of Lending Club. Is that a small minority or a growing trend?



AnilG

  • Hero Member
  • *****
  • Posts: 1104
    • View Profile
    • PeerCube
Re: Automeated investing questions
« Reply #1 on: July 13, 2017, 07:28:06 PM »
Quote
The first year my account balance grew by 8%.

This is very common in first year of new accounts. Most of your loans are new and haven't aged much, i.e. delinquencies and defaults haven't hit your account. Also, loans are amortized so during earlier phases of loan lifecycle, majority of repayment is interest and reduction in principal is minimal that positively skews the performance at account basis.

Quote
This annual rate fell regularly finally going slightly negative by mid-2017.

This is not out of ordinary. As you lent lump sum amount at the start, the defaults are coming in lump sum too. Also every default has outsize impact on your portfolio performance. On average, each default makes you lose ~70% of original lent principal. To make up for this principal loss due to default, your portfolio has to generate too much interest income to break-even, basically you need anywhere from 5 to 20 loans to pay off on time with full interest for each defaulted loan.

Quote
a note distribution skewed toward the riskier notes

IME, performance chasing by lending to riskier notes is one of the biggest reason for portfolio underperformance. Most people get so fascinated with high average projected/expected returns that they forget that those projections come with much higher variance/standard deviation/volatility. Risk, Expected Reward, and Volatility go hand-in-hand.

Quote
I did develop a technique that identified my bad luck impact.

I would like to know what technique and how you developed it to identify impact of "luck". Typically luck is a random error.

Quote
Do other large investors rely primarily on automated investing? Is this approach a good idea?

If you are lending $25 a pop in lot of loans, automated is the way to go if you take into the account the value of your time spent on reviewing the loans. Even if a review of loan takes 30 seconds and you lend to 50% of the loans reviewed, you can only lend to 60 loans maximum in an hour. Valuing you time at $30 an hour, you are spending 50 cents on each $25 note, a 2% hit on return even before the note has been issued.

Quote
Is there something inherent in the automated investing process that selects notes that are worse than randomly selected notes in the same grade?

Even performance of portfolios of randomly selected loans has a distribution (mean and variance), any one portfolio can fall anywhere on that distribution.

Quote
Are Lending Club’s estimated returns consistently overblown?

Any estimated return claimed by anyone is just that the "average" estimated/projected/expected return. Without knowing the variance/standard deviation, you don't know what the expected range of returns might be achieved. I don't give any weightage to estimated returns claims unless they are accompanies by range. It is basic statistics.

I have a dilemma. If you have been following my analysis blog, billlanke.blogspot.com, you know that I’m very concerned about the performance of my Lending Club account. To simplify, I opened an IRA account in early 2015. I purchase over 3,000 notes in the next 2 years using Automated Investing. The first year my account balance grew by 8%. This annual rate fell regularly finally going slightly negative by mid-2017. My analysis concluded the two biggest factors were “bad luck in the notes selected for me” and a note distribution skewed toward the riskier notes.
 
I did develop a technique that identified my bad luck impact. When I added this back my growth rate improved to 2.92%. Lending Club estimates that my “Adjusted Net Annualized Return” for my current notes is 2.86%. When I look at Lending Club’s automated investing estimated returns, none are anywhere near as bad as mine. This leads me to several questions and I would be interested in your opinions.

Do other large investors rely primarily on automated investing?

Is this approach a good idea?

Is there something inherent in the automated investing process that selects notes that are worse than randomly selected notes in the same grade?

Are Lending Club’s estimated returns consistently overblown?

I see a lot of people bailing out of Lending Club. Is that a small minority or a growing trend?
---
Anil Gupta
PeerCube Thoughts blog https://www.peercube.com/blog
PeerCube https://www.peercube.com

OleBill

  • Newbie
  • *
  • Posts: 22
    • View Profile
    • Email
Re: Automeated investing questions
« Reply #2 on: July 14, 2017, 11:01:30 AM »
Thanks for your insightful response. I will be considering your points as I move forward. You asked "I would like to know what technique and how you developed it to identify impact of "luck"". Following is a summary lifted from my blog posts.

1. I used the total amount paid on a note divided by the amount loaned as a measure of return for each note. For those still active this would be the values as of 03/31/2017.

2. I developed a match code for each note "mm_yyg##". The first 5 characters are the month and year of issue, "g" is the grade, and "##" is the term of the loan.

3. I built a table of these match codes with the loan amount, amount paid, and computed return. An example entry would be "15_11B36,9415,114455500,82012305,72".  So in the month of Nov 2015, there were 9,415 grade B, 36 month notes issued. As of 03/31/2017 the average amount paid on these notes was 72%.

4. Automated investing selected 8 of these notes for me in that month for a total investment of $150. With a return rate of 72% for randomly selected notes, I would expect to have received $108 in payments thus far. I have received $103.85, thus far, a shortfall of $4.15. This is was I would consider the luck of the draw.

5. Some of my notes are behind the payment curve and some ahead. In fact, 75% of the notes selected for me by automated investing are behind for a total of negative $1764 .

One can look at these results as "unlucky" or some inherent problem with automated investing's selection process. That's why I raised the question.

 

fliphusker

  • Sr. Member
  • ****
  • Posts: 463
    • View Profile
    • Email
Re: Automeated investing questions
« Reply #3 on: July 14, 2017, 11:35:55 AM »
Am I missing something here or did you not do any backtesting?
Did you just set it up to invest in certain large parameters such as grade and term and not filter further?
Thanks for your insightful response. I will be considering your points as I move forward. You asked "I would like to know what technique and how you developed it to identify impact of "luck"". Following is a summary lifted from my blog posts.

1. I used the total amount paid on a note divided by the amount loaned as a measure of return for each note. For those still active this would be the values as of 03/31/2017.

2. I developed a match code for each note "mm_yyg##". The first 5 characters are the month and year of issue, "g" is the grade, and "##" is the term of the loan.

3. I built a table of these match codes with the loan amount, amount paid, and computed return. An example entry would be "15_11B36,9415,114455500,82012305,72".  So in the month of Nov 2015, there were 9,415 grade B, 36 month notes issued. As of 03/31/2017 the average amount paid on these notes was 72%.

4. Automated investing selected 8 of these notes for me in that month for a total investment of $150. With a return rate of 72% for randomly selected notes, I would expect to have received $108 in payments thus far. I have received $103.85, thus far, a shortfall of $4.15. This is was I would consider the luck of the draw.

5. Some of my notes are behind the payment curve and some ahead. In fact, 75% of the notes selected for me by automated investing are behind for a total of negative $1764 .

One can look at these results as "unlucky" or some inherent problem with automated investing's selection process. That's why I raised the question.

AnilG

  • Hero Member
  • *****
  • Posts: 1104
    • View Profile
    • PeerCube
Re: Automeated investing questions
« Reply #4 on: July 14, 2017, 01:47:52 PM »
He used Lending Club's automated lending. He got suckered into marketing pitches that prominently  highlights  "high" expected  average returns and skip or bury the expected range in fine prints or put performance disclaimers in fine prints.

Am I missing something here or did you not do any backtesting?
Did you just set it up to invest in certain large parameters such as grade and term and not filter further?
---
Anil Gupta
PeerCube Thoughts blog https://www.peercube.com/blog
PeerCube https://www.peercube.com

SLCPaladin

  • Full Member
  • ***
  • Posts: 202
    • View Profile
Re: Automeated investing questions
« Reply #5 on: July 14, 2017, 02:54:17 PM »
Quote
If you are lending $25 a pop in lot of loans, automated is the way to go if you take into the account the value of your time spent on reviewing the loans. Even if a review of loan takes 30 seconds and you lend to 50% of the loans reviewed, you can only lend to 60 loans maximum in an hour. Valuing you time at $30 an hour, you are spending 50 cents on each $25 note, a 2% hit on return even before the note has been issued.

This is quite possibly the best way to quantify opportunity cost with respect to Lending Club and note selection that I have ever read. I cringe to calculate the negative return of all the hours I've spent on this forum trying to attain a modicum of understanding of the industry! In Anil's case and many others, the gift of insight they provide is truly a service!

OleBill

  • Newbie
  • *
  • Posts: 22
    • View Profile
    • Email
Re: Automeated investing questions
« Reply #6 on: July 14, 2017, 04:56:30 PM »
I did not back test. I can't see how anyone could back test automated investing. As SLCPaladin points out, AnilG's description of the effort involved in individually picking notes is impractical when you are dealing with 1000's of notes.

The only way you could succeed at that  is to try to build a model that is better than Lending Club's when it comes to evaluating notes. I decided that was unlikely, hence opted for the automated investing approach.

If you look at the automated investing edit function, there is a sample of mixtures of notes, all with impressive returns. And there is the small print "Historical Returns: 3.22% - 8.6%" with a footnote. The footnote explains that 10% to 90% of these selections with at least 100 notes should experience results in this returns range. Since I was going to deal with 1000's of notes I falsely felt comfortable that the assumption that my reuturns would regress to the mean. That turned out not to be true.

rawraw

  • Hero Member
  • *****
  • Posts: 2786
    • View Profile
Re: Automeated investing questions
« Reply #7 on: July 14, 2017, 05:15:33 PM »
I did not back test. I can't see how anyone could back test automated investing. As SLCPaladin points out, AnilG's description of the effort involved in individually picking notes is impractical when you are dealing with 1000's of notes.

The only way you could succeed at that  is to try to build a model that is better than Lending Club's when it comes to evaluating notes. I decided that was unlikely, hence opted for the automated investing approach.

If you look at the automated investing edit function, there is a sample of mixtures of notes, all with impressive returns. And there is the small print "Historical Returns: 3.22% - 8.6%" with a footnote. The footnote explains that 10% to 90% of these selections with at least 100 notes should experience results in this returns range. Since I was going to deal with 1000's of notes I falsely felt comfortable that the assumption that my reuturns would regress to the mean. That turned out not to be true.
There is a guy on this forum whose LC account is measured in the millions (IIRC) and he does it all manually.  It can be done, but just not common.

Back testing is pretty easy.  There are several sites that aggregated the data and make it available.

And you did regress to the mean.  It just so happened that you invested in the risky assets right before they started performing poorly for the industry.  I doubt your returns are significantly worse than anyone else here who was chasing yield

AnilG

  • Hero Member
  • *****
  • Posts: 1104
    • View Profile
    • PeerCube
Re: Automeated investing questions
« Reply #8 on: July 14, 2017, 07:09:55 PM »
Let me give you counterpoint to my own argument. You may realize that the hours you spent doing manual lending and understanding the industry may not be considered wasted. OP would have been less dis-satisfied if he had started lending manually first. There are several benefits of manual lending:

You learn about lending. Most of us have been borrowers and never lenders so we don't know what a lender should look into borrowers. Lending has a totally different mindset than borrowers. By lending manually, you learn to identify whom you want to lend to. You build your intuition about good and bad borrowers and the conflict/tension/incentive misalignment that exist between lenders, borrowers, and platform. I encourage most new lenders to consider lending manually for first 4-6 months to really understand how things work and why they work that way.

There is another benefit of manual lending at the opening of account or anytime when a large lump sum cash deposit is made. When you use automated lending, the service will try to deploy your cash as quickly as possible. This creates a loan bulge in your portfolio timeline where loans from this short period dominate your portfolio. Your fortune will rise and die with that loan bulge. When default wave comes for a particular vintage, if you have loan bulge from that vintage in your portfolio, you are screwed. This initial bulge in your portfolio takes several cycle of reinvesting to dissipate. Also, it is very hard for your portfolio to recover if you encounter high number of defaults in a short period of time. For every 1% loss in your portfolio, your portfolio has to gain >2% just to break-even.

The impact is similar to baby boomer (population bulge). They suck up all resources available at whatever point of lifecycle that bulge is at. When you lend manually, you are forced to slow down and distribute your loan selection over time and in the process alleviating bulge issues and hopefully learning as you go along with lending.

Humans have tendency to take credit when things go well but transfer the blame to someone/something else when things go bad. When you lend manually, you don't get the opportunity to transfer blame to someone else as you are the freaking reason for whatever happened. When same thing happens with automated lending, all kind of conspiracy theories are formed and put forward about automated lending/platform being bad.

I know several PeerCube subscribers who lend manually and maintain large loan portfolios. My wife also lends manually in her accounts while I use automated lending for my accounts.

Quote
If you are lending $25 a pop in lot of loans, automated is the way to go if you take into the account the value of your time spent on reviewing the loans. Even if a review of loan takes 30 seconds and you lend to 50% of the loans reviewed, you can only lend to 60 loans maximum in an hour. Valuing you time at $30 an hour, you are spending 50 cents on each $25 note, a 2% hit on return even before the note has been issued.

This is quite possibly the best way to quantify opportunity cost with respect to Lending Club and note selection that I have ever read. I cringe to calculate the negative return of all the hours I've spent on this forum trying to attain a modicum of understanding of the industry! In Anil's case and many others, the gift of insight they provide is truly a service!
---
Anil Gupta
PeerCube Thoughts blog https://www.peercube.com/blog
PeerCube https://www.peercube.com

Fred93

  • Hero Member
  • *****
  • Posts: 2203
    • View Profile
Re: Automeated investing questions
« Reply #9 on: July 14, 2017, 07:34:23 PM »
I did not back test. I can't see how anyone could back test automated investing.

Its pretty simple.  You take the criteria you specified, and apply it to a historical database of LC loans, and see how it performed.  You would have learned things that would have been helpful to you.  You would have learned, for example, that the high risk loans performed quite differently in different years.

You looked only at the average over all time that LC gave you.  Because LC has been growing exponentially, that all-time average is very heavily weighted toward recent years.  It is much more helpful to look to see how performance varied over different years, because all years won't be the same as some recent good years.


Quote
the effort involved in individually picking notes is impractical when you are dealing with 1000's of notes.

Irrelevant.  Back testing and manual investing are entirely different things.