What are your favorite and least favorite criteria?

[Watermen]

Just wondering? I seen so many discussuon, so many different analysis, some of them are even contradicting, some of them does not even make sense.

For example, is loan amount correlated to default rate and ROI?

Some people said that lower amount lead to higher default rate, but some will never invest in loan more than certain amount, say $5000 or $10,000.

What is your taught?

So far I go by my intuition, I stated 1 month ago, therefore aside from reading blogs and analysing data from Prosper stat or NRS, I can only go by what is common sense to most people. However, I understand that this can sometime be counter-productive, as it is not evidence-based. But again, is evidence-base analysis really better than some common sense criteria?

So far my favorites are:
1. No delinquencies or public records.
2. Have a job.
3. Make more than $5000 per month.
4. Credit line no earlier than 1980, reason because person may be too old and not able to pay back. (I am not sure if such a thought is reasonable.)
5. Work in the military, seems like a good one.
6. Work for at least 3 years.
7. Own a house or at least have a mortgage.
8. Loan less than $35,000. (I am worry some people just want to borrow the max amount and run away.)
9. Have a good description. Have a goal in the description, my favorites are trying to improve credit score, be debt free, explain how you can and want to pay the loan.

My least favorites:
1. No job. It makes me wonder how you going to pay back.
2. No house of your own.
3. Work as an accountant, police or lawyer.
4. Low amount loan despite having a reasonably high income, for example borrowing $5000 when you make $7000 per month, why the hell do these people want to borrow? Can't they try to save?
5. From CA or NV. That is what everyone is saying, not sure if it is valid.
6. Too old, because that person can die suddenly and the loan can be default.
7. Note graded A, B, C on LC. Or Interest less than 17%.
8. Empty description or meaningless description. (If you are trying to persuade someone for money, don't you need to work hard to persuade the person who is trying to lend you?)
9. Loan money for business or vacation. I am ambiguous for wedding or engagement ring, not sure if these two are high or low risk?

* By the way, not too sure if anyone realized about verified income? What is the effect on ROI and default? If you have your income verified right from the beginning, does it make the loan less risky? Please share your experience.

[rev]

I've written a blog post when I first introduced Interest Radar, and I think it might interest you:
http://interestradar.wordpress.com/2012/09/02/quick-start-guide/

I think 3 out of your 9 favorite filters are reversed to what the numbers tell. You can play with the filters and see for yourself.

[sociallender]

Watermen,

I agree with you.  There are so many available loan filters...  I was confused too about which ones to use and when.  I think interest radar is a great site to develop successful strategies / filters. On the surface, I think your filters look good to me.   IMO, I don't think there is a golden rule that consistently provides optimum ROI.  However, it does help and  provide peace of mind which is a good thing regarding investing psychology.

The following doesn't answer your question directly but may be of help to you:

I run statistical analysis on all loans to create the types of rules you are describing but with a high degree of accuracy albeit low occurrences.  For example, a rule may be: 

total_credit_lines > 62, amount_requested <= 18000, revolving_line_utilization <= 45.1, amount_requested > 3825, employment_length > 3

In this case, there were 36 out of 36 loans of which all were fully paid.  The rub is loans that meet this criteria are few and far between.  However, statistical software creates 10s of thousands of rules for each class (fully paid and charged off).  I filter my loans based on these types of rules to compliment a classification my machine learning algorithm.  The system does provide statistical probabilities which provides the supporting evidence you were looking for.  If you are interested, I post the statistical spreadsheets to my blog daily at sociallender.blogspot.com.

One feature you may find useful is once you have a filter that you like (such as the one you posted), you can apply this filter to all the currently available loans on my site.  What is nice is it also provides you with the statistical probability of the loans being fully paid and whether it matched any of the rules i described above.  To do this:

Go to my blog home page
Click the "filter" link next to the "Evaluate" spreadsheet.
Enter the criteria of your filter and click submit (click link to generated spreadsheet)

You can also save your filter for future use as well, just copy the filter text provided at the bottom of the page.

Hope that helps,
John

[Grant]

You would do well to use analytic software of some sort and move away from your intuition.  I can tell you that some of your "good" factors from your intuition are incorrect for the types of loans I invest I and seem of your negative factors are also incorrect.  The key phrase being "for the loans I invest in."

When I first started last summer, I was overwhelmed by some of the conflicting opinions presented as fact as well as some of the statements some make ("I've never had a late note!") which aren't true on face value (I.e. either they were incredibly lucky, conservative, or they aggressively sell their notes on processing day 4). 

Before I learned about Interest Radar and LendStats, I ran my own regression analysis using SPSS software. Some of my own assumptions were absolutely incorrect.... For example, I thought vacation spending was "bad". Turns out FOR THE LOANS I INVEST IN, vacation is a lower-yielding category but are OK.

One key for any analytics is that you have to have enough loans in the analysis to make your result statistically significant.  You can whittle down your criteria very tightly like SocialLender does above.  I doubt the criteria he proposed will hold up over time.  Their analysis is probably statistically insignificant given the small number of loans meeting that characteristic.  My main criteria is based on 324 loans and I do have statistical significance to a high degree.

I currently use Interest Radar and am impressed by the filters and analytical available. I've also been impressed with the responsiveness REV has shown both the community and individual investors.

[American in Busan]

But again, is evidence-base analysis really better than some common sense criteria?

Can you give me any example in real life where evidence based analysis is worse than common sense criteria? Just one.

[brycemason]

Can you give me any example in real life where evidence based analysis is worse than common sense criteria? Just one.

Disagreements with my wife. Keep the peace.

[Watermen]

I've written a blog post when I first introduced Interest Radar, and I think it might interest you:
http://interestradar.wordpress.com/2012/09/02/quick-start-guide/

I think 3 out of your 9 favorite filters are reversed to what the numbers tell. You can play with the filters and see for yourself.

Thanks for your recommendation, your blog is very interesting.

Are those data really from 2008 onward and you have all the data from all the loans since then?


After spending sometime playing with your site. I was surprise with a few findings.

People who own a house seems to have a higher default rate than those who have mortgage or rent.

People who borrow for vacation seems to have pretty good return rate, worst is business, medical...

I wonder if you have problem logging into to this website after registration, it keep saying that my password is wrong, despite the fact that I reset it.

[rev]

Are those data really from 2008 onward and you have all the data from all the loans since then?

Yes, the data is from 2008 onward (since LC restructured).
You can download that data yourself from this page: https://www.lendingclub.com/info/download-data.action

[Watermen]

Are those data really from 2008 onward and you have all the data from all the loans since then?

Yes, the data is from 2008 onward (since LC restructured).
You can download that data yourself from this page: https://www.lendingclub.com/info/download-data.action

I can't log in anymore, it keep saying that my password is wrong...

[rev]

You can download that data yourself from this page: https://www.lendingclub.com/info/download-data.action

I can't log in anymore, it keep saying that my password is wrong...

You can't log in to Lending Club or Interest Radar?
If IR, send me an e-mail (webmaster@interestradar.com) and I can try to help, but right now I see you logged in...

[sociallender]

@Grant,

324 is still statistically insignificant as well considering over 20,000 loans.  more if you use loss factoring on non mature loans.  I have since raised my threshold to 500 for all minority rules but this is still statistically insignificant.  I think you may be confused that I am trying to select loans based on just these minority filters.  I am using them as a complement to classification as an additional layer of confidence.  I show very good statistical evidence across the entire range of loans (not just a small sample that people use as filters) with traditional classification techniques. 

[rawraw]

But again, is evidence-base analysis really better than some common sense criteria?

Can you give me any example in real life where evidence based analysis is worse than common sense criteria? Just one.

The subprime market and related securitization.  I didn't even have to go back a few years for that one.

[New Jersey Guy]

So many debatable variables!

Personally, I like Interest Radar to weed out the bulk of the loans that should be in the scrap heap to begin with.  Rev's insight into this is far beyond mine, so I trust his filters and so far, they've served me well.  It's a real time saver.

I always hope that what does pop up in IR are basically good notes.  Now I'm comfortable.  But, I do scrutinize each note that I am considering.

*  I don't invest in vacations.
*  I quit investing in home improvements where a flippin' hot tub is involved!
*  I, too, won't buy low dollar loans where the borrower makes big money (for the same reason you won't.)
*  I agree with your thoughts concerning a description. However, poor spelling and grammar are point losers with me.
*  I don't agree totally about your thoughts on old people (I'm 54).  I know of whipper snappers younger than me currently pushing daisys.
*  I don't like borrowers running over 70% of their credit limits, especially high wage earners.  Sorry, but to me it's irresponsible.

[veggivet]

But again, is evidence-base analysis really better than some common sense criteria?

Can you give me any example in real life where evidence based analysis is worse than common sense criteria? Just one.

Loans with verified income have a higher default rate than loans without income verification.

[American in Busan]

Loans with verified income have a higher default rate than loans without income verification.

I don't believe your analysis. Is it based on evidence?