Lending Club Loans - Borrowers' FICO Score and Bad-Loan Experience Index

After reviewing the defaults as a function of the borrowers' FICO score in the previous post, I decide to review FICO range according to bad-loan experience (BLE) index. Please refer to my first post on BLE Index for background information.

The table below shows the BLE Index according to borrowers' FICO score range. The range of BLE Index from 0.23 for FICO score 790-794 to 1.84 for FICO score 660-664 is not considerably wide. It appears that changes in FICO score is not significantly related to credit risk. The borrowers with FICO score below 709 appear to be greater credit risk.

Are you interested in Retirement Planning and Investment Annuities? Learn more from the published works of Moshe Milevsky

The table below shows the BLE Index according to FICO score range for loans issued each year from 2007 to 2012. Except for loans issued in 2012, the pattern of borrowers with FICO score below 709 being greater credit risk is pretty consistent. For loans issued in 2012, even the borrowers with better FICO score have BLE Index greater than 1.10. In my opinion, this observation shows that borrowers with better FICO score are more likely to default during the first year of the loan.

Key Takeaways

• No surprises that borrowers with lower FICO score are greater credit risk.
• The lenders who purchase notes on secondary market may be better of purchasing notes that have aged at least a year.

Lending Club Loans - Borrowers' FICO Score and Defaults

In this post, I will review the borrower's FICO score and its relationship with defaults. At the end of September, Lending Club started reporting the FICO scores in narrow and uniform width bands. For this analysis, I decided to switch historical loan data file to a new one from Oct 14th to be more granular with FICO range. There is general perception that borrowers with lower FICO score are at greater credit risk and more likely to default on their loans. I wanted to find out if this perception is true in peer-to-peer lending segment also.

Loan Volume

As the chart below shows the majority of loans are issued to borrowers that have FICO score below 730. Less than 100 loans are issued to borrowers with FICO score above 825.

Looking for some instant cash back on computer products? Check out Newegg’s Rebate Center.

The chart below shows the loan volume with FICO score for each application year, from 2007 to 2012. The loan applications from borrowers with FICO score 830 and higher were started to be listed on Lending Club in 2012.

The two observations stand out from this chart. First, more and more borrowers with lower FICO scores are applying for loan on Lending Club platform. The strict lending practices by traditional lending channels in the last few years may have resulted in borrowers with lower FICO score to look for alternative lending platforms such as peer-to-peer lending for loans. Second, the increasing smoothness of the FICO curve with time might be due to higher loan volume and Lending Club becoming more granular with borrower's FICO score since about 2010.

Loan Status

The chart below shows the loan status with borrowers' FICO score range for all loans issued since 2007. As expected, there is a clear trend of greater percentage of loans with charged off, default, and late status for borrowers with lower FICO score. Similarly, fewer borrowers with lower FICO score are likely to fully pay their loans.

The chart below shows the loan status of charged off or default with borrowers' FICO score range for application year 2008 through 2011. The pattern of defaults with FICO range each year are very similar to aggregate results shown in the earlier chart. The default rate on loans issued to borrowers with FICO score of 670 and less can be considered exceptionally higher than what normally seen for other borrowers on Lending Club platform.

Key Takeaway

• There were no surprises in expectations of borrowers with lower FICO scores attracted to alternative lending platform such as Lending Club and higher defaults and late loans for borrowers with lower FICO scores.

Lending Club Borrowers' Location and Bad-Loan Experience Index

After reviewing the Lending Club borrowers' state of residence for default rate in my previous post, I was very curious to see how states stack up with respect to bad-loan experience (BLE) index.

As a refresher, the smaller the BLE index the better the parameter is in reducing the credit risk. The green color represents the parameters that have BLE index less than 0.90. The pink color represents the parameters that have BLE index greater than 1.10. For more information on BLE index, please refer to my first post on BLE index, Lending Club Loan Purpose: Default Rate and Bad-Loan Experience Index.

The table below shows the calculated BLE Index according to borrowers' state of residence. There were not many surprises with the states that showed up with highest BLE Index and that also had the highest default rate as discussed in the last post. The BLE index variation from Rhode Island (0.45) to Nevada (1.92) is not considerably wide. It appears that States are not that significantly related to credit risk. From this analysis, it appears Nevada is the only state of some concern.

The table below shows BLE index for each year from 2007 to 2012. The rows are arranged in the same order as the above table for easy comparison. Except Nevada, Florida, and California, none of the other states has BLE index consistently greater than 1.00.

Key Takeaway

• The analysis of borrowers' state based on BLE index doesn't provide any conclusive answers about impact of borrowers' location on credit risk.
• The risk-averse lenders may want to consider excluding borrowers based in Nevada, Florida, and California, the states that consistently show up with BLE index greater than 1.10.

Lending Club Borrower's Location and Defaults

Previously, I analyzed parameters that were specific to loans: Credit Grade, Loan Length, Interest Rate, Loan Amount, and Loan Purpose. Now, I plan to review parameters that are specific to borrowers such as their location of residence, FICO range, income, etc.

In this post, I review the states where borrowers' live and find potential trends and patterns. There is general perception online by Lending Club lenders that California borrowers have higher defaults. I wanted to find out the validity of such views. Personally, in my personal P2P lending, I do not take into consideration the state of residence of borrower. Similarly, the Founders' Filter at PeerCube doesn't exclude loans based on the borrower's state of residence.

Borrower's State of Residence and Loan Volume

As the chart below shows, since 2007, more than 40% of all loans were issued to borrowers residing in California, New York, Florida, and Texas. Less than 100 loans combined have been issued to borrowers residing in Mississippi, Tennessee, Indiana, Idaho, Nebraska, Iowa and Maine.

For further analysis, I decided to combine states in Others  group that accounted for less than 200 loans issued each. In addition to states listed above with less than 100 loans, this group also includes Delaware, Montana, Alaska, Wyoming, South Dakota, and Vermont.

The chart below shows the distribution of states as a function of loan application date. The state composition appears to be more or less same for past several years. Considering California, New York, Florida and Texas being most populous states, the trend of most borrowers from these states is not surprising.

Loan Status

The chart below shows the loan status for borrowers from different states for all loans issued since 2007. One observation right away stands out is that borrowers residing in California don't have exceptionally high default rate. Borrowers residing in Nevada, Utah, Florida, Missouri, and Maryland have higher default rate than borrowers residing in California. Borrowers from Alabama, Arkansas, Washington DC, New Mexico, Rhode Island and Kansas appear to have lowest default rate.

Even after excluding loans from years 2007 through 2009 (charts not shown), there was no evidence that borrowers from California have exceptionally higher default rate.

Key Takeaway

• There is no evidence from above default rate analysis that borrowers from California have greater tendency to default compared to other borrowers.
• The perception of higher defaults for California borrowers may be due to significantly large number of loans issued to borrowers from California (12,607) thus large number of loans in defaults (4.12% * 12,607 = 520) but not necessarily due to higher default rate.

Lending Club Loan Amount and Credit Grade: Bad-Loan Experience Index

This post is continuation of last two posts discussing bad loan experience (BLE) index according to Loan Purpose, Loan Length and Interest Rate. In this post, I will review the two loan parameters that I had previously analyzed using default rate: Loan Amount and Credit Grade.

As a refresher, the green color represents loan parameters that has BLE index less than 0.90 and light red color represents loan parameters that has BLE index greater than 1.10. The smaller the BLE index, the better the loan parameter is in reducing the credit risk.

BLE Index for Loan Amount

The first table below shows the calculated BLE index according to loan amount. The second table shows BLE index according to loan amount for years 2007 through 2012. While reviewing the table, we need to keep in mind that loans for amount greater than $25,000 were only started to be issued in 2011.

It appears from BLE index that loan amount on its own may not be so much relevant in reducing credit risk.  This observation is consistent with the findings of 1940 study in NBER publication. That study also found that the amount of loan is not shown to be significantly related to bad-loan experience. Logically, it makes sense because the borrowers' capacity to make payments on borrowed amount (loan amount in connection with income, debt to income ratio, and revolving credit balance) is more likely to be influential on credit risk instead of loan amount.

---

---

BLE Index for Credit Grade

I consider credit grade to be a composite parameter developed by Lending Club to rate loans based on the methodology described at Interest Rate and How We Set Them. I expect bad-loan experience to show not only the relationship of credit grade with risk but also effectiveness of Lending Club methodology in assigning credit grade. Now, as the BLE index in the first table below shows, Lending Club does one fine job with credit grade. The smaller the BLE index, the better the credit grade. Also, the second table shows that by 2010 the bad-loan experience is rising with the credit grade. It appears Lending Club has done great job adjusting credit grade with credit risk over the years.

BLE Index for Credit Sub Grade

The two tables below show the BLE index according to the credit sub grade. It is clear that while Lending Club has done good job assigning Credit Grade with credit risk, some more work need to be done to do the same at credit sub grade level.

In general, the Credit Grade A and B appear to have lowest BLE index, i.e. better credit risk profile.

Key Takeaways

• The BLE index according to loan amount, on its own, is not a good indicator of credit risk. I expect to get better results when loan amount is analyzed in connection with borrowers' income, debt to income ratio and revolving credit balance (to be analyzed in future posts).
• Lending Club has done good job assigning credit grade according to credit risk; however, more work is required at credit sub grade level.
• Loans with credit grade A and B have lower risk profile according to BLE index. As discussed in previous post, for other credit grades, lender should consider advantage of eliminating X% of bad loans versus disadvantage of eliminating Y% of good loans.

Lending Club Loan Length and Interest Rate: Bad-Loan Experience Index

After reviewing the loan purpose with the Bad-Loan Experience Index (BLE Index) methodology in my last post Lending Club Loan Purpose: Default Rate and Bad-Loan Experience Index, I am very excited about the potential of this Index. I believe it will help me in further improving my understanding of Lending Club loans and what factors influence the credit risk of such loans.

On PeerCube, I started a peer filter for Bad-Loan Experience Index that I will continue to enhance as I review BLE Index for different loan parameters. In this post, I will review the two loan parameters that I had previously analyzed using default rate: Loan Length and Interest Rate.

---

How much is your time worth? Are you spending too much time selecting loans on Lending Club? Try PeerCube: Smart Lending with Peer Insights. PeerCube focuses on quick and collaborative decisions.

---

BLE Index for Loan Length

The table below shows the calculated BLE Index according to loan length. While reviewing the table, we need to keep in mind that 60 month loans were only started to be issued in 2010. The green color represents loan length that has BLE index less than 0.90 and light red color represents loan length that has  BLE index greater than 1.10. As a refresher, the smaller the BLE index, the better it is in reducing the credit risk.

It appears that loans with 36 month loan length are better loans with reduced credit risk than loans with 60 month loan length. This observation is consistent with general tendency of shorter loan length with better credit risk mentioned in 1940 study in NBER publication [PDF]. I also reviewed the BLE index for loan length for year 2010 through 2012 as shown in table below. The trend consistently shows 36 month loans having better credit risks but the gap is closing over past three years.

BLE Index for Interest Rate

The table below shows the BLE Index according to interest rate buckets. Each bucket is 3.00% interest rate wide. The BLE index ranges from 0.35 for loans with interest rate of 3 - 6% to 3.09 for loans with interest rate of 21 - 24%. It appears that interest rate is significantly related to credit risk. The rising BLE index pattern with rising interest rate is also very consistent. Logically, it makes sense: the higher the return, the greater the risk.

The table below shows the BLE index for interest rate for year 2007 through 2012. The BLE index pattern is pretty consistent: the higher the interest rate, the higher the risk.

BLE Index Caveats

No measurement is perfect and BLE Index has its own caveats that we need to consider.

Finding tendencies and not certainties

The following quote from the NBER publication sums it up nicely. It was true in 1940 and, in my opinion, it is true in 2012, too.

"Statistical analysis of the kind we are here attempting are necessarily based on averages and probabilities, and therefore can reveal only tendencies, not certainties."

Weigh the advantage of eliminating X% of bad loans  against the disadvantage of eliminating Y% of good loans

Let's take a hypothetical scenario of  deciding whether to eliminate loans with interest rate between 15 and 18% from consideration. Your decision to review this class of loans is driven by the findings above. The BLE index of 1.91 is much higher than your benchmark of 1.10, so you are deciding not to invest in such loans.

Instead of using a redline approach to eliminate all classes of loan that have BLE index higher than 1.10, weigh the advantage of eliminating, for example in this case, 21.59% of bad loans against the disadvantage of eliminating 11.28% of good loans.

Let's assume there were total 2,000 bad loans and 10,000 good loans and loans with 15 - 18% interest rate have 22% bad loans and 12% good loans. These number are approximate from above and previous post for easier calculations. If you decided to eliminate loans with 15 - 18% interest rate, you will be eliminating 440 bad loans and 1,200 good loans.

You are the only one who can decide whether missing out on 1,200 good loans in order to avoid 440 bad loans is worth or not. Once I finish reviewing all parameters for default rate and BLE index, I will review another measure that may help in this aspect.

Key Takeaways

• To reduce credit risk, lenders may consider avoiding loans with 60 month term.
• Though BLE Index indicates loans with interest rate less than 12% are better, lenders looking for higher yield may need to consider the advantage of eliminating X% of bad loans against disadvantage of eliminating Y% of good loans.

Lending Club Loan Purpose: Default Rate and Bad-Loan Experience Index

Default Rate: Is it a good measure?

I have been using default rate (= Number of loans defaulted or charged off / Total loans issued) since I started analyzing historical loan data from Lending Club. While I continue to use default rate, I am not convinced that default rate is really good measure of risk for loans on P2P platforms. I believe default rate is under-representing the risk of such loans primarily due to high growth in issued loans volume in recent years. Higher number of such recent loans with current and issued status in the denominator of default rate calculation may be artificially lowering the default rate.

Alternate measures for loan analysis

In my last post Lending Club Loans: Loan Purpose and Defaults, I tried to look at data by calculating default rate that excluded loans with current and issued status. This was my attempt to compensate for the influence of recent loan volume.

I don't have much knowledge and experience of consumer lending and finance segment. Recently, I have been reading quite a few books, listed in my Reading List, on Risk Management, Quantitative Finance and Fixed Income Investment. Unfortunately, most of them cover corporate loans and government bonds and ignore consumer financing. I would love to learn from risk management people working in consumer lending segment on how they calculate risk with consumer loans.

Recently, I came across consumer finance and lending publications from National Bureau of Economic Research (NBER) that seem promising and hopefully will provide me guidance on better methodologies for analyzing data for patterns and influences of different parameters. I decided to use one such measure from NBER publication [PDF], "Bad-Loan Experience Index" for analyzing Loan Purpose further.

This analysis is also a good trial for me to consider including such measures on PeerCube for historical loan data analysis. I have database portions for historical loan data taken care of at PeerCube. I just need to figure out how we should be optimally using historical information for making loan selection decisions. Your input is most welcome.

Bad-Loan Experience Index

The Bad-Loan Experience (BLE) Index is the ratio of percentage of bad loans to that of good loans for various options, for example debt consolidation, credit card, car, etc. in each parameter, for example Loan Purpose. This index for all options combined will be 1.00. A ratio greater than 1.00 will indicate worse than average risk. A ratio less than 1.00 will indicate better than average risk. The good loans are defined as the loans that were paid by borrowers without much effort. In Lending Club case, we can consider loan with loan status of Fully Paid as good loans. The bad loans are defined as the loans that were not paid by borrowers. In Lending Club case, we can consider loan with loan status of Default and Charged Off as bad loans.

BLE Index for Loan Purposes

The Table below shows the calculated BLE Index for various Loan Purposes sorted from low to high BLE Index. The green color represent loan purposes that are less than 0.90 and pink color represent loan purposes that are more than 1.10. The variation from loan for car (0.57) to loan for small business (2.34) is considerable. It appears that loan purpose is significantly related to credit risk. Though we need to be aware of the fact that the loan purpose is reported by borrower and there is no secondary method to check intended use with actual use of the loan. Also, borrower may have multiple intended use of the loan.

From this analysis, it appears loans for car, credit card, major purchase, wedding, house and home improvement may be better than loans for small business, renewable energy, medical, other and vacation. These findings are not that different from the takeaways in my last post Lending Club Loans: Loan Purpose and Defaults, only exception being house.

What I found interesting was that how similar the findings from Lending Club data were to the findings from 1940 study in NBER publication for common categories, specifically for car, medical and business.

BLE Index for Loan Purposes by Issued Year

I decided to see if there is any variation in BLE Index by loan issued date. The table below shows the BLE Index for each year from 2007 to 2012. The rows are sorted by BLE Index for 2010. The green color represent loan purposes that are less than 0.90 and pink color represent loan purposes that are more than 1.10.

The loans for credit card, car, and home improvement appears to be consistently better than loans for other, education, vacation, medical, and small business, rest being mixed bag.

Key Takeaways

• Despite the borrower reported nature of loan purpose, it appears to be significantly related to credit risk.
• To reduce credit risk, lenders may consider excluding loans for small business, medical, vacation, education, and other purposes.

Lending Club Loans: Loan Purpose and Defaults

After spending most of my time launching PeerCube, I am happy to get back to analyzing the Lending Club historical loan data for patterns and trends. The analysis below uses the historical data available on September 10, 2012 from Lending Club statistics page.

Previously, I analyzed Loan Issued Date, Interest Rate, Application Date, Credit Grade, Loan Length, and Loan Amount. The common theme among these loan parameters was that none of them are directly reported by borrowers.

Now let's look at some of the loan parameters that are only reported by borrowers. These parameters typically don't require proof from borrowers and have no secondary method of confirmation such as credit report. Personally, I consider such parameters at best subjective and subject to manipulation.

Loan Purpose

The Loan Purpose parameter indicates the purpose for which the loan will be used for.  Lending Club currently has fourteen pre-defined purposes:

1. Car
2. Credit Card
3. Debt Consolidation
4. Educational
5. Home Improvement
6. House
7. Major Purchase
8. Medical
9. Moving
10. Renewable Energy
11. Small Business
12. Vacation
13. Wedding
14. Other

Though loan purpose is reported by a borrower, there is no restriction or check for what purpose borrower may use the loan for. Recently, I have been noticing more and more loans are being reported as debt consolidation and credit card by borrowers even though the title, description, and Q&A indicates that loan should have been classified differently. I hypothesize that some of this reporting is influenced by lenders preference for debt consolidation and credit card loans.

As the chart below shows almost 2/3rd (66.22%) of the all loans issued since 2007 on Lending Club platform are reported with purpose of Debt Consolidation and Credit Card. Individually, less than a percent of loans are issued for Renewable Energy, Educational, Vacation and House Purchase purposes.

The chart below shows the distribution of loan purpose as a function of application date. 2012 is turning out to be a blockbuster year when borrowers are reporting debt consolidation and credit card purpose for almost 3/4th (74.21%) of the loans. Even in 2008 when we went through financial crisis and following year, we don't see such an high level of activity. Will all loans reported to be used for debt consolidation and credit card pay off used for such purpose? I am skeptic.

Loan Status

The chart below shows the loan status as a function of loan purpose for all loans issued since 2007. The loan status of defaults and charged off for loans with purpose of Small Business, Education, and Energy seems to concerning. The chart seems to be validating impression by lenders that debt consolidation and credit card loans are better and renewable energy, educational, and small business loans are bad. Is it really true?

Are we seeing influence of high volume of recent loans issued with debt consolidation and credit card purposes and not enough loans for energy, education, and business? This was the first doubt came to my mind. Let's see if we can figure out ways to take out the influence of volume growth year over year and higher volume of recent loans for debt consolidation and credit card purposes.

I decided to exclude loans with loan status of Current, In Review, and Being Issued. The assumption is that most loans in such status are recent loans. The chart below shows the loan status excluding loans with current, in review, and being issued status as a function of loan purpose. While the education loans don't stand out as a sore thumb when comes to defaults, the defaults of small business and energy loans are still very high and medical loans become a suspect too.

The chart below shows loans with default and charged off status as a function of loan purposes and application year. One observation right away stands out is the higher than normal defaults for loans for small business purpose. Over the years, loans for moving, education, medical, house, and renewable energy purposes have proven to have higher than normal defaults also.

Key Takeaways

• Anecdotal observation of miss-classification of loan purpose to debt consolidation and credit card, the default rate data doesn't support this observation.
• Lenders need to closely watch loans for small business, renewable energy, moving, education, medical, and house purchase purposes as such loans have higher than normal default rate.