Investing Curiouscatblog

Tag: peer to peer lending

  • LendingClub Filters, Selecting Loans and Automated Investing

    This post continues our series on peer-to-peer lending (and LendingClub): Peer to Peer Portfolio Returns and The Decline in Returns as Loans Age, Investing in Peer to Peer Loans. LendingClub, and other peer-to-peer lenders let you use filters to find loans that meet your criteria. So if you chose to take more, or less, risk you can use filters to find loans fitting your preferences. Those filters can also be applied to automate your lending.

    There are resources online to help you understand the past results of various investing strategies (returns based on various filters). Some filter are just a trade-off of risk for return. You can invest in grade A (a LendingClub defined category) loans that have the lowest risk, and the lowest interest rates and historical returns. Or you can increase your risk and get loans with higher interest rates and also higher historical returns (after factoring in defaults).

    historical chart of returns by grade at Lending club
    Description of chart: This chart shows the historical performance by grade for all issued loans.

    This chart includes all loans that were issued 18 months or more before the last day of the most recently completed quarter. The historical returns data in the chart is updated monthly.

    Adjusted Net Annualized Return (“Adjusted NAR”) is a cumulative, annualized measure of the return on all of the money invested in loans over the life of those loans, with an adjustment for estimated future losses.

    LendingClub lets you set filters to use to automatically invest in new loans as funds are available to invest (either you adding in new money or receiving payments on existing loans). This is a nice feature, there are items you can’t filter on however, such as job title. And also you can’t make trade-offs, say given x, y and z strong points and a nice interest rate in this loan I will accept a bit lower value on another factor.

    So I find I have to be a bit less forgiving on the filter criteria and then manually make some judgements on other loans. For me I add a bit higher risk on my manual selections. I would imagine most people don’t bother with this, just using filters to do all the investing for them. And I think that is fine.

    Practically what I do so that I can make some selections manually is to set the criteria to only be 98% invested. This will cause it to automatically invest any amount over 2% that is not invested. You can set this to whatever level you want and also is how you can make payments to yourself. I will say I think one of the lamest “features” of LendingClub is that is has no ability to send you regular monthly checks. So you have to manually deal with it.

    It should be simple for them to let you set a value like send me $200 on the 15th of each month. And then it manages the re-investments knowing that and your outstanding loans. But they still don’t offer that feature.

    As I said one of the factors in setting filters is managing risk v. reward but the other is really about weaknesses in the algorithm setting rates. You can just see it as risk-reward trade-off but I think it is more sensible to see 2 different things. The algorithm weaknesses are factors that will fluctuate over time as the algorithm and underwriting standards are improved. For example, loans in California had worse returns (according to every site I found accessing past results). There is no reason for this to be true. If a person with the exactly same profile is riskier in California that should be reflected in higher rates and thus bring the return into balance. My guess is this type of factor will be eliminated over time. But if not, or until it is, fixed filtering out loans to California makes sense.

    Once you set your filter criteria then you select what balance you want between A, B, C, D, E and FG loans. I set mine to

    A 2%
    B 16%
    C 50%
    D 20%
    E 10%

    I actually have a bit over 1% in FG (but I select those myself). In 2015 the makeup of the loans given by LendingClub was A 17%, B 26%, C 28%, D 15%, E 10%, F and G 4%.

    Related: Where to Invest for Yield Today (2010) – Default Rates on Loans by Credit Score – Investing in Stocks That Have Raised Dividends Consistently – Investment Risk Matters Most as Part of a Portfolio, Rather than in Isolation

    Sadly Lending Club uses fragile coding practices that result in sections of the site not working sometimes. Using existing filters often fails for me – the code just does nothing (it doesn’t even bother to provide feedback to the user on what it is failing to do). Using fragile coding practices sadly is common for web sites with large budgets. Instead of using reliable code they seems to get infatuated with cute design ideas and don’t bother much making the code reliable. You can code the cute design ideas reliably but often they obviously are not concerned with the robustness of the code.

  • Default Rates on Loans by Credit Score

    Credit scores are far from a great measure of whether a person is a great credit risk for a specific loan, in my opinion. However, they are very widely used and therefor, very important. They also are somewhat useful. And lenders don’t base judgement solely on credit scores, they consider many other factors, if they have any sense at all.

    Credit scores range from 300 to 850. They are calculated by various credit reporting organizations, including FICO. They factor in payment history, percent of outstanding credit available that is used, credit report checks, length of outstanding credit accounts, etc..

    Metlife report on consumers and credit scores provides some interesting data.

    Credit score range Default rate*
    740-850 .4%
    680-739 2.8%
    620-679 7.5%
    550-619 17%
    300-459 33.8%

    * Default rate in this case means, 90 days past due. MetLife got this data from the Consumer Financial Health Study dataset**.

    Peer to peer lending platform, Lending Club, limits loans to those with a minimum credit score of 660 (remember there are multiple organizations that provide credit scores, this minimum is based on Lending Club’s score). In general I see scores above 700 in A and B loans, scores from 650-700 in C and D loans. Remember the credit score is not the only factor setting the rate (you will see scores above 700 in the C loans sometimes, etc.). Credit scores provide some insight but are just 1 factor in approving loans or setting rates (an important one but not a completely dominant one).

    About 38% of people have credit scores from 750-850. Another 37% from 600-749 and about 25% from 350-599.

    chart of Default rate by credit score (731-750) from 2003 to 2010
    via online Vantage Score presentation

    Vantage Score decided to make their score range go up to 1000, not the standard 850. Maybe a 750 score for them is comparable to 680? They say super-prime is 900+ (750-850 on more common scale), prime is 701-900 (680-739), near-prime 641-700 (620-679), subprime 501-640 (550-619). Anyway that chart shows the changing default rates from 2003 to 2010 by type of loan.

    This Federal Reserve report on meeting between Federal Reserve Board staff and Fair Isaac Corporation (FICO) 20 June 2013 has some interesting material.

    For guidance, the following table generally matches a borrower’s odds-of-default with the corresponding FICO 8 score (calculated on performance from Oct 2008 – Oct 2010). Of course, the range of scores and odds-of-default [the data is related to mortgages] will vary with each model as creditors develop and validate their own credit scoring models.

    Odds-of Default
    		FICO 8 Score
    		percent of population**
    5:1 610 9%
    10:1 645 9%
    20:1 685 6%
    30:1 705 6%
    40:1 720 6%
    50:1 735 9%
    100:1 770 30%

    As you can see at a 610 level, 20 loans out of 100 defaulted. At 685 just 5 in 100 defaulted and at 770 just 1 in 100 did.

    ** I had to adjust this, because the report didn’t report it in this form, so it a very approximate measure (I made estimates for something like scores from 735 to 769 etc.). Again this is data from the Oct 2008 – Oct 2010 period. The rest of the population (about 25%) would have scores below 610.

    Related: The Impact of Credit Scores and Jumbo Size on Mortgage Rates (2009) – Your FICO credit score explained – $2,540,000,000,000 in USA Consumer Debt

    This page references a Fed report (that I can’t find) that found the following default rates on new loans for the two years after origination, 2000-2002:

    Credit score range Default rate*
    under 520 41%
    520-559 28%
    560-599 23%
    600-639 16%
    640-679 9%
    680-719 4.4%
    over 720 <1%

    ***

    The Consumer Financial Health Study respondents were asked to self-assess their credit quality and for permission to pull their actual credit scores.8 Forty-five percent of survey participants granted permission, yielding an “opt-in” sample size of 3,215. We appended two objective measures of creditworthiness to the dataset: Experian provided VantageScore 3.0 credit scores, and LexisNexis Risk Solutions provided RiskView scores. VantageScore is a generic credit scoring model that was created by the three major credit bureaus (Equifax, Experian and TransUnion) and, in addition to tradeline data, includes rent, utility and cell phone payment data when it is available in consumer credit files.

  • Peer to Peer Portfolio Returns and The Decline in Returns as Loans Age

    This is a continuation of my previous post: Investing in Peer to Peer Loans

    LendingClub suggest a minimum of 100 loans (of equal size) to escape the risk of your luck with individual loans causing very bad results. Based on this diversity the odds of avoiding a loss have been very good (though that obviously isn’t a guarantee of future performance), quote from their website (Nov 2015):

    With just $2,500 you can spread your investment across 100 Notes. 99.9% of investors that own 100+ Notes of relatively equal size have seen positive returns.

    chart of expected returns

    This chart, from LendingClub, shows a theoretical (not based on past performance) result. The basic idea is that as the portfolio ages, more loans will default and thus the portfolio return will decline. This contrasts with other investments (such as stocks) that will show fluctuating returns going up and down (over somewhat dramatically) over time.

    For portfolios of personal loans diversity is very important to avoid the risk of getting a few loans that default destroying your portfolio return. For portfolios with fewer than 100 notes the negative returns are expected in 12.8% of the cases (obviously this is a factor of the total loans – with 99 loans it would be much less likely to be negative, with 5 it would be much more likely). I would say targeting at least 250 loans with none over .5% would be better than aiming at just 100 loans with none over 1% of portfolio.

    There are several very useful sites that examine the past results of Lending Club loans and provide some suggestions for good filters to use in selecting loans. Good filters really amount to finding cases where Lending Club doesn’t do the greatest job of underwriting. So for example many say exclude loans from California to increase your portfolio return. While this may well be due to California loans being riskier really underwriting should take care of that by balancing out the risk v. return (so charging higher rates and/or being more stringent about taking such loans.

    So I would expect Lending Club to adjust underwriting to take these results into account and thus make the filters go out of date. Of course this over simplifies things quite a bit. But the basic idea is that much of the value of filters is to take advantage of underwriting weaknesses.

    chart of LendingClub returns as portfolio ages historically

    This chart (for 36 month loans) is an extremely important one for investors in peer to peer loans. It shows the returns over the life of portfolios as the portfolio ages. And this chart (for LendingClub) shows the results for portfolios of loans issued each year. This is a critical tool to help keep track to see if underwriting quality is slipping.

    (more…)

  • Investing in Peer to Peer Loans

    Peer to peer lending has grown dramatically the last few years in the USA. The largest platforms are Lending Club (you get a $25 bonus if you sign up with this link – I don’t think I get anything?) and Prosper. I finally tried out Lending Club starting about 6 months ago. The idea is very simple, you buy fractional portions of personal loans. The loans are largely to consolidate debts and also for things such as a home improvement, major purchase, health care, etc.).

    With each loan you may lend as little as $25. Lending Club (and Prosper) deal with all the underwriting, collecting payments etc.. Lending Club takes 1% of payments as a fee charged to the lenders (they also take fees from the borrowers).

    Borrowers can make prepayments without penalty. Lending Club waives the 1% fee on prepayments made in the first year. This may seem a minor point, and it is really, but a bit less minor than I would have guessed. I have had 2% of loans prepaid with only an average of 3 months holding time so far – much higher than I would have guessed.

    On each loan you receive the payments (less a 1% fee to Lending Club) as they are made each month. Those payments include principle and interest.

    historical chart of returns by grade at Lending club
    This chart shows the historical performance by grade for all issued loans that were issued 18 months or more before the last day of the most recently completed quarter. Adjusted Net Annualized Return (“Adjusted NAR”) is a cumulative, annualized measure of the return on all of the money invested in loans over the life of those loans, with an adjustment for estimated future losses. From LendingClub web site Nov 2015, see their site for updated data.

    Lending Club provides you a calculated interest rate based on your actual portfolio. This is nice but it is a bit overstated in that they calculate the rate based only on invested funds. So funds that are not allocated to a loan (while they earn no interest) are not factored in to your return (though they actually reduce your return). And even once funds are allocated the actual loan can take quite some time to be issued. Some are issued within a day but also I have had many take weeks to issue (and some will fail to issue after weeks of sitting idle). I wouldn’t be surprised if Lending Club doesn’t start considering funds invested until the loan is issued (which again would inflate your reported return compared to a real return), but I am not sure how Lending Club factors it in.

    (more…)