Four Filters to Increase Your P2P Lending ROI

When you look at the spread of late payments from the post I published earlier this week it is obvious that some people get a loan from Lending Club or Prosper with no intention of ever paying it off. Wouldn’t it be nice if we could recognize these people before committing our investment dollars to their loan. Well I believe with some simple filtering you can eliminate most of them.

While p2p lending is still new, there are thousands of loans that have reached or are close to maturity and we can do some statistical analysis on these loans. Ken at Lendstats.com has been busy this month upgrading his filtering options for his Prosper and Lending Club loan statistics. I have spent several hours this week poring over his changes to work out which criteria have the most impact on p2p investors ROI.

After looking at all the filters available on Lendstats I decided on four filters that I think will make the most impact while still leaving the investor with plenty of loans to choose from. I chose filters that take data from the borrowers credit report because I didn’t want to rely on unverified information. People who are scammers, with no intention of repaying the loan, will likely provide misleading or incorrect information on their loan application. So I didn’t want to rely on any inputted data from the borrower. These filters obviously don’t catch all defaults but I was able to reduce the number of defaults by well over 50%.

Prosper’s Numbers Using the Filters

Below is the table for Prosper loans. Because Prosper has been around since 2005 I was able to select a large pool of loans that have reached maturity.

Prosper 1/05 - 12/07No. of loansROIDefaultsDefault Rate
All Loans17,426-6.3%672438.6%
Inquires: 03,5411.0%87424.7%
Loan Size < $20,00033832.0%81724.2%
Open Credit Lines >= 513582.8%27220.0%
Utilization <= 70%8183.2%12715.5%

You can see that the ROI for all loans originated from the very beginning of Prosper until December 2007 was -6.3%. But after applying the first filter the ROI jumps to positive 1.0%. Defaults go down dramatically as well. Each successful filter improves ROI while at the same time reducing defaults. When the last filter is applied you end with an ROI of 3.2%, not stellar by any means but far better than the -6.3% that you started out with. Here is the link for the Prosper loan filters on Lendstats.

Explanation of the Four Filters

As I said these filters are taken from information from the borrowers credit report (apart from the loan size obviously). Here is a brief explanation of the filters:

Inquiries: 0

These are what are known as “hard inquiries” that appear on a person’s credit report whenever they request credit or a loan. So, if a borrower has recently applied for a credit card or an auto loan this will appear on their credit report. In our filter we look for no hard inquiries at all over the past six months. This may seem a little harsh but even when you expand this number to one the ROI goes down. This filter is the easiest way to improve your ROI and so it is the first and most important filter.

Loan Size < $20,000

This one is self-explanatory. We filter out all loans of $20,000 or more. This makes logical sense because the higher the loan, the higher the payment and presumably the more difficult it is to pay back.

Open Credit Lines >= 5

This refers to the number of open lines of credit, namely credit and charge cards. I am not sure why five is the magic number but if you go much higher or lower than five then you start to negatively impact ROI.

Utilization <= 70%

Credit Utilization refers to the amount of credit used in ratio to the total available credit. You might think that the lower the number the better the return for the p2p investor but the sweet spot seems to be a maximum of 70%.

Lending Club’s Numbers Using the Filters

Lending Club started originating loans in June 2007 so if I chose to analyze only loans that had reached maturity then it would have been a very small sample size (589 loans total). So, I took all the data through December 2008; these loans are all at least two years old and should provide enough data for an analysis. Here are Lending Club’s numbers:

LC 1/07 - 12/08No. of loansROIDefaultsDefault Rate
All Loans2,963-1.0%58719.8%
Inquires: 01,0013.4%12812.8%
Loan Size < $20,0009384.2%11211.9%
Open Credit Lines >=57334.4%8611.7%
Utilization <= 70%5255.3%608.8%

You see the same trends as with Prosper’s data. One quick note about Lending Club’s stats. At the time of this writing the Lending Club loan filter was not yet available to the public on Lendstats. Ken indicated that its release was imminent (and he let me play around with the beta – thanks Ken), but if you don’t see a link to it on his site you can try this link here.

How to Implement These Filters on Prosper and Lending Club

Prosper makes it very easy to use filters when choosing loans on their site. They have an advanced search page where you can add all these filters very easily. At the time of this writing there were 32 loans available on Prosper and after applying these four filters there were eight loans to choose from.

On Lending Club it is a little more difficult. Their filtering isn’t as flexible as Prosper’s, so you have to download the “In Funding Loan Data” from their downloads page. You can save the CSV file and then bring it into Excel to do the filtering there. I prefer this method because you get all the loan data in the spreadsheet and you can add many more filters to the data to find the loans worth considering. At the time of this writing applying these four filters left you with 140 loans to chose from out of a total of 426 loans on the site.

A Word of Caution About Using Filters

These four filters that I have mentioned here are only a starting point. I like them because they will tend to catch most of the scammers – those people who have no intention of repaying their loan. Luckily this is a small minority of borrowers but they likely cause a significant percentage of defaults.

I chose these four filters because they provide substantial improvements with both Prosper and Lending Club loans. But now some words of caution. With additional filters you can slice and dice the numbers so narrowly that you can back into a fantastic return. For example, here is a filter that boasts a 20.86% ROI but if you look closely it contains only 73 loans (out of 31,000) so it is not a significant enough sample. I tried to provide filters that left a decent number of loans to choose from, eliminating the probable worst performing loans. The other thing to be careful about is the possibility that future loans may perform differently than those in the past. As more loans mature I will be re-testing these filters to make sure they still perform well.

Having said all that I believe if you use these filters you will get above average returns on your p2p lending investment. But for the serious investor these filters should only be a starting point. You should add your own filters and other criteria when choosing loans. I would like to hear from others – what factors do you look for when deciding which loans to invest in on Lending Club or Prosper?

Peter Renton is the chairman and co-founder of LendIt Fintech, the world’s first and largest digital media and events company focused on fintech.

LendIt Fintech conducts three conferences a year for the leading fintech markets of the USA, Europe, and Latin America. LendIt also provides cutting-edge content all year long via audio, video, and written channels.

Peter has been writing about fintech since 2010 and he is the author and creator of the Fintech One-on-One Podcast, the first and longest-running fintech interview series.

Peter has been interviewed by the Wall Street Journal, Bloomberg, The New York Times, CNBC, CNN, Fortune, NPR, Fox Business News, the Financial Times, and dozens of other publications.

Subscribe
Notify of
21 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Mike
Mike
Jan. 21, 2011 1:37 pm

Some good food for thought in this exercise. The unfortunate bottom line is that 1 in 5 LC loans default over the life of the loan, and even after using all the screens your return is a measly 5%. I eliminate any loan from contention where the borrower commits a spelling or grammatical error in answering a question. I look for a stable employment history, no public records or delinquencies, high monthly income (greater than 8k), and credit consolidation loans where banks have jacked up interest rates on the customer.

Mike
Mike
Jan. 21, 2011 2:14 pm

Gotcha. When do you think you’ll have enough of a sample size to do a similar analysis for Prosper and LC version 2.0?

David
David
Jan. 21, 2011 3:17 pm

Hey Peter,

Ken just added a custom filter ROI calculator to his site for both Prosper and Lending Club at

https://www.lendstats.com/loansearch/lc/lcloanfilter.php

and

https://www.lendstats.com/loansearch/loanfilter.php

They are pretty fun to play around with.

David
David
Jan. 21, 2011 3:24 pm

Whoops, jumped the gun a bit there. Just noticed that you have both of those links in your post. Feel free to delete that comment.

KenL
Jan. 21, 2011 7:34 pm

Hi everyone, I’m glad to see so many of you using my site. I do my best to present the data as accurately and practically as possible. I just wanted to chime in here and tell everyone that active listings that fit the selected filter criteria are now displayed with the LC filter, but I see Peter has already mentioned that. I’d also like to point out that your browser will remember which listings you looked at by changing the color of the listing link. It’s a real time-saver. This however will work best if you use the same browser and the same computer for selecting listings.

Dan B
Dan B
Jan. 21, 2011 9:09 pm

I think we will need a few more years of data to reach the types of conclusions that are being suggested here with any confidence. I apply probably 10 filters when I pick loans & I have no doubt that all together they are effective in helping me pick better……..but I wouldn’t be surprised if any one of those 10 make no difference or are even detrimental if you were to run the numbers right now. And I wouldn’t stop using a particular filter even if the numbers don’t support it right now.
Simply put, the sample size is way to small right now & results can be off by 5-7% on any given filter & still be within the margin of error or 1 standard deviation. Please understand that I’m not trying to diminish any ones effort & work in compiling this information. And I think that the use of “filters” is a necessity if one is to have decent returns. I just think that we should sit down & think through each filter option instead of relying on conclusions that are based on an insufficient sample size. Once we get to the point where there are a couple of thousand notes within every filtering option then we can be accurate in saying this will improve or decrease your performance by this or that percentage.

Dan B
Dan B
Jan. 23, 2011 2:59 am

I use………VERIFIED INCOME
3 yr term only
2 or less inquiries
no public record
60 or more months last delinquency
DTI under 20%
rev balance util under 60%
2 yrs job
B notes or lower
No relisted loans

Mike
Mike
Jan. 24, 2011 6:23 pm

Dan B, just curious why you avoid the 60 month notes. Is it just because you don’t want to tie up your $$ for that length of time?

Dan B
Dan B
Jan. 25, 2011 3:35 pm

@Mike….I don’t feel comfortable with the extra time nor do I feel that the extra % involved is adequate compensation. I have absolutely no proof or evidence that the 60 month terms will negatively or positively impact returns.
It’s just a personal preference as I have real difficulty in grasping a 5 year time period going forward in my own life, to say nothing of others.

KenL
Jan. 31, 2011 2:55 pm

Hi everyone, I thought I’d chime in again. I personally like the 60 month loans. I think they are a great alternative for riskier borrowers. So far they are showing about the same rate of delinquency as the 36 month term over the same time frame, and with a higher interest rate. How they perform later in the term is still an unknown, but so far all indications are good.

I guess I’ll say something in defense of my filters. Criticisms of a too small sample set for some of the filters are correct. Because of this I view some of my filters as an experiment in motion. Nevertheless there are a few filters with a significant data set and they should be quite reliable.

I have purposefully presented the data in a way so everyone can make their own decision on the validity of each filter. Even if there isn’t enough loans to validate the significance of a filter, a person can still take a look at the calculation sheet and learn some valuable lessons. In the calculation sheet returns are not only calculated for variables that fall within the filtered range, returns are calculated for each variable outside of the filter range too. This data is represented by the shaded lines and maybe just as valuable.

Take the business filter https://lc.lendstats.com/index100.php?i=8 for example. If one studies the calculation sheet one can see that it is best for borrowers to have a lower DTI a lower utilization % and a longer credit history. Of course this is true for most potential borrowers but with borrowers looking for a business loan, holding these criteria to stricter standards is even more important and actually this is just common sense. A great thing about these calculation sheets is that they put numbers behind such common sense.

Kathy F
Kathy F
Feb. 1, 2011 2:03 pm

I too use filters based on on doing a lot of reading on the forums at http://www.prospers.org when I first started. I have both Prosper and LC loans but I am not reinvesting anything in Prosper now and continue to reinvest only at LC. I was worried about Prosper going out of business because I read that they had to keep borrowing money to stay in business. I only invested about $1400 total as an experiment. Of my 16 prosper loans, 3 are paid off and 1 is charged off but is making reduced payments in collections. I have 53 loans with LC, 5 are already paid off and 1 is charged off. The one Lending Club charge off was C- rated, I went out on a limb that time. The person only made a few months payments and then declared bankruptcy. My LC ROI is 8.39%. At Prosper, my return has been 10.37% based on actual gains divided by initial investment.

Anyway, I also use filters of no delinquencies, no public records, DTI< 20%, no big loans <$20K, they must be employed in same job for at least a few years, and I only invest in A or B loans, except for the one time a took a chance. I usually do 36 mos loan but have done a few 60 month. I usually only do those people who are trying to refinance their credit card debt to a lower interest rate. That is smart of them and it means they are not taking on new debt with LC loan. I don't loan for new business ventures.

I like the idea of verified income as a filter but it seems that many of the loans are not verified. I also don't understand why so many of the loans don't have any descriptions or explanations. That usually turns me off. I don't usually ask questions because it is too time consuming waiting and trying to find the answers later. But I always read the loan request carefully and if anything feels wrong to me, I pass and see if I can find something better. But I don't have much money in this, it is just an experiment for me. I appreciate anyone else sharing their advice on reducing loan defaults in their portfolio. I will definitely use the new filter of zero inquiries when I am looking for new loans to fund.

By the way, I found your site from you comment at Get Rich Slowly.

David J
David J
Feb. 8, 2011 2:35 pm

I just recently found LC and put $1k into it, only investing $25 per loan and doing mostly B loans with a few C’s. I review each loan request (don’t use the portfolio tool) and use my gut as a judge, as well as many of the same filters you all are using.

One thing I would love to see, and don’t know why they don’t, is verifying everyone’s income. I mean, why wouldn’t they? A bank would. I see some loan requests where they state $15k per month gross income, not verified of course, and they’re looking for an $8k loan. Doesn’t add up. Reading answers to questions, looking at spelling, grammar, ALL CAPS, etc. allows me to weed a lot of it out too.

I have just seen the first part of my loans issued, still have about 15 that are still funding, so we’ll see how this experiment goes.

ab
ab
Feb. 1, 2015 2:42 am

Right now I invest in As and Bs at only $25/loan. But I will check the filters more. 52% of the portfolio is prime so I am pushing towards 75%