Back to School – Financing Education With Marketplace Lending

Whether you are a young man or woman heading off to college or an adult whose school days are a distant memory, the month of September carries a familiar association: back to school.  The years we spend in formal education can be among the most transformative and memorable of our lives.  Nevertheless, the rising cost of education has sparked debate in the United States and around the world.   In today’s analysis, we examine the role marketplace lending might play in the financing of education.

In our recent analysis on real estate lending, we included the following graph, showing the major sources of U.S. consumer debt as of 3/31/2014.

Source: U.S. Federal Reserve, 3/31/2014

As of this June, the U.S. Federal Reserve put the balance of outstanding student loan debt at $1.27 Trillion, exceeding the levels for other common sources of spending, including auto loans and credit cards.  Perhaps some students will be turning to marketplace lenders to obtain the funds necessary for their education.   

Education Loan Volumes

While overall educational lending has increased in the United States, leading marketplace lenders LendingClub and Prosper both stopped making education-purpose loans on their platform in 2010.  This is due to certain provisions of federal law surrounding loans for “post-secondary education expenses”.  Loans for this purpose are subject to conditions beyond those required of other loan purposes, making such loans incompatible with the structure of P2P lending platforms.

 Given the lack of data available from U.S. platforms, let’s instead take a look at Europe.  Bondora, the rapidly-growing pan-European marketplace lender, does offer loans for educational purposes, and while the volume for this purpose is still quite small, it does make available some compelling data that should give us a sense of these borrowers and their profile.

Bondora Education Loans – Demographic Data

In the chart below, we show 2014 Bondora loans by country.  Estonia and Spain are the clear leaders, with Spain surpassing even Bondora’s home country. 

As interest rates tend to vary by locality, it is useful to examine the distribution of interest rates by country (excluding those with very low volume).  In the graph below, we can see that rates range from 12.5% to 37.5%, with Estonia having a somewhat wider distribution than Spain.

In the graph below, we compare the interest rates of education loans to those of non-education loans.  There does not appear to be a much of a meaningful difference.

In the way of demographics, Bondora also happens to make available certain data that is typically harder to come by in the U.S., including education, age, gender, and marital status.

 

Credit Quality and Performance

As the increase in education-purpose loan volume on Bondora has been relatively recent, we are not yet in possession of enough performance data to truly compare this versus other loan purposes.  We plan to check back in 6 months or so to conduct a more thorough analysis.

Conclusion

While LendingClub and Prosper stopped making education-purpose loans in 2010, there is certainly a desire in the alternative lending community to fund education.  Companies such as Commonbond and SoFi have been lending to people for the purpose of paying for school or refinancing student loans, with notable success.  There are also some companies introducing loan products for very specific educational uses, such as to fund attendance at various coding schools.  Needless to say, we’d love for some of these companies’ data to be publicly available so that we could understand the borrower dynamics and loan performance.   While we are in the early days of applying marketplace lending to education, we are optimistic about the potential for simpler and more efficient student lending. 

The Economic Health of a Borrower’s City Affects Loan Performance

Harlan Seymour runs a family office in San Mateo, California, with a primary focus on analyzing and investing in fixed income instruments, including marketplace loans.

Introduction

Marketplace lenders like Lending Club and Prosper provide dozens of data points about a borrower’s financial situation.  In an attempt to improve upon average loan performance and generate “Alpha”, sophisticated investors crunch this data to select loans that have been graded more harshly by the lender than the data warrant.

One of the data points provided is the borrower’s locality (city and state).  Analyzing historical loan data shows that, all else being equal, loans issued to borrowers living in economically prosperous cities significantly outperform, while those from troubled localities underperform.

Locality Database

We have created a locality database containing geographic, economic and demographic information for the U.S. as a whole, for all 50 states (plus Washington, D.C.), for all of the over 3,000 U.S. counties, for over 50,000 cities, and for thousands of neighborhoods within larger cities.

Borrowers enter their city and state in their loan applications.  Many borrowers use their neighborhood name as their city name.  For example, Venice Beach, CA is actually a neighborhood of Los Angeles.  Knowing the neighborhood is useful, since economic data (like real estate prices) is available down to the neighborhood level.

We examined the 333,134 loans issued by Lending Club through June 30, 2014.  Using the locality database, we were able to match 99.1% of the cities entered, with many of 0.9% unmatched cities being misspellings (that we decided not to pattern match).  2.6% of the cities matched actually turned out to be neighborhoods:

15,144 individual cities and 161 neighborhoods were matched within the historical loan data.  Using the latitude and longitude of these cities and the loan count for each city, we created a heat map of where borrowers are located (# of loans per city is color coded):

Locality Health Score (LHS)

The locality database contains a rich set of current and historical, economic and demographic values for the U.S. itself, as well as for the neighborhoods, cities, counties and states it contains.  Such values include home prices, unemployment rates, labor force participation, per capita and household incomes, populations, etc.  An optimized weighted average of these values was formulated to create the Locality Health Score (LHS) that runs from 0 (bad) to 1 (good) for each locality. 

As of January 1, 2014, these are the top and bottom 8 U.S. cities with populations over 500,000, as measured by the LHS:

8 top-rated cities

8 bottom-rated cities

LHS vs. Loan Performance

We used Lending Club’s June 30, 2014, historical loan snapshot to analyze how the LHS relates to loan performance (overall default rates) over the lifetime of the loans.  Specifically, we focused on grade A-D loans from 2010-2013.  For the 2013 cohort only “seasoned” loans issued in the first half (H1) of 2013 were included in this study.

Cohorts were broken down by year and grade.  In each cohort, loans with a high LHS (> .8) and loans with a low LHS (< .3) were compared to the overall cohort.  We measured the fraction of bad loans in a cohort as:

(# in grace period + # in default/charged off) / (# loans in cohort)

The results show that high (> .8) LHS loans result in a reduction of the percentage of “bad loans” in every cohort (all years, all grades).  Conversely, low LHS (< .3) loans experience an increase in the percentage of “bad loans” in every cohort (all years, all grades).  Thus, the LHS is highly relevant for marketplace loan selection.   

This chart shows that for every year, 2010-2013, an LHS < .3 (in red) significantly increases the fraction of bad loans vs. the overall fraction (in dark grey), and that an LHS > .8 (in green) significantly decreases the fraction of bad loans:

These charts show that these results are consistent across all years (2010-2013) and all grades (A-D):

How to Use LHS When Investing in Marketplace Loans

My family office invests in marketplace loans using highly tuned filters for selecting “good loans” that should default at a rate significantly lower than other loans of the same grade or interest rate.  The LHS plays a part in these filters.

Since we invest in fractional loans (vs. whole loans), we mainly use the LHS to vary the amount we invest in a given loan: the higher the LHS, the higher the fractional investment.  This biases our marketplace loan portfolios towards borrowers in prosperous localities.  When the next economic downturn comes, borrowers in prosperous localities should be more resilient.

Conclusion

The city or neighborhood in which a borrower resides turns out to have a meaningful impact on the rate at which a borrower may default on his loan(s) versus loans to other borrowers of the same grade.  Furthermore, in an economic downturn, borrowers in more prosperous cities may be more resilient than those in troubled cities.  With the Locality Health Score (LHS), we have developed a single number that coalesces several economic and demographic values for a locality into a simple-to-use and understandable number, ranging from 0 (bad) to 1 (good).

 

August 2014 Marketplace Lending Monthly Recap

Month and Year-to-Date Asset Class Returns 

The Orchard US Consumer Marketplace Lending Index outperformed three of the five comparable U.S. fixed-Income asset classes for the month of August. The Orchard Index outperformed 3-5 year Treasuries, 3-5 year Investment Grade Corporates, and 5-year Municipals and underperformed Intermediate High Yield Corporates and Securitized Products for the month.  

Source: Fixed-income asset class returns based on Barclays Benchmark Indices

Source: Fixed-income asset class returns based on Barclays Benchmark Indices

However, on a year-to-date basis, the Orchard US Consumer Marketplace Lending Index continues to outperform all five comparable U.S. fixed-Income asset classes.

Source: Fixed-income asset class returns based on Barclays Benchmark Indice

Source: Fixed-income asset class returns based on Barclays Benchmark Indice

LendingClub – Pioneering Marketplace Lending & Reshaping Financial Services For Years to Come

In a move promising to transform the landscape of global finance for years to come, LendingClub, the leader in the nascent field of marketplace lending, recently filed for its initial public offering.  Having now made over $5 billion in loans, the success of this booming fintech juggernaut might now seem inevitable.  At this time, however, it is important to understand the state of the world when LendingClub opened its doors, the driving forces behind the need for a new approach to consumer lending, and what this watershed event means for the future of global financial services. 

Rebuilding Real Estate With Marketplace Lending

At $8.2 trillion in the United States alone, mortgage debt dwarfs all other asset classes within consumer credit.  As marketplace lending has grown, its potential to transform the allocation of capital has captured the imagination of an industry, and, increasingly, the public.  While this new way of lending has largely grown around unsecured consumer and small business lending, several exciting companies are working to extend marketplace lending to real estate, an effort that, if successful, promises to have far-reaching implications for our economy.

Source: U.S. Federal Reserve - Data as of 3/31/2014

Many of the circumstances that have necessitated rethinking in the financial world apply particularly to real estate.  If bureaucracy, regulation, outdated technology, and less-than-stellar customer service exist anywhere in banking, they definitely exist in the mortgage process.  As anyone who has been through the process of obtaining a residential mortgage in the past several years can attest, it is not particularly easy or pleasant.  These hurdles apply to commercial real estate as well.

As with any nascent industry, new entrants are offering a diverse set of business models.  Below, we explore some of the most common and interesting examples of how real estate lending is being transformed and democratized through technology. 

Debt Crowdfunding of Real Estate Projects

If you walk around any town or city in the United States, chances are that you will see a residential real estate project.  This might be the construction of a new home or apartment building or the renovation of an investment property.  An example of this is a “flip”, in which an investor purchases a property, renovates it, and then sells it at a (hopefully) higher price a short time later.  A developer will need a sizable chunk of capital to finance such a project, the source of which has traditionally been from personal savings, family and friends, or various “hard money lenders”.

Now, developers have the option of obtaining project funding through an investor marketplace.  On sites including Patch of Land and RealtyMogul, real estate developers can apply for loans to fund their projects through a streamlined, online application process. 

The loans that borrowers receive are typically of 3-12 month duration, with APRs in the 10-20% range.  The investor experience, while maintaining some parallels to unsecured consumer marketplaces such as LendingClub and Prosper, is significantly more specialized to real estate.  Investors can avail themselves of a lengthy project description, property photos, appraisal information, developer biography, construction budgets, financial projections, and risk data.

Equity Crowdfunding of Residential and Commercial Real Estate Projects

It is common practice for any real estate development to be structured as its own corporation, typically an LLC.  In equity crowdfunding of real estate, as practiced on platforms such as RealtyMogul and Fundrise, investors can purchase shares in one of these LLCs.  Shareholders are then able to earn a portion of the cash flow that is generated from a property (e.g. from rental income) as well as the proceeds of the property’s eventual sale.  While equity investing in real estate clearly carries risk, it also affords the investor the opportunity of benefitting from a property’s appreciation in value over time.

Online Mortgage Lending

While the aforementioned categories are very compelling, neither sounds quite like what most people think of as a “mortgage”.  If marketplace lending can transform the way that ordinary people finance their primary residences, it will be a truly great accomplishment.  LendingHome, LendInvest, and Privlo all promise great strides here. Privlo is doing this today, though only in Texas and Idaho at the current time (the others do not yet allow loans on owner-occupied properties). Privlo lets borrowers apply for a mortgage online and also has a “concierge” to walk applicants through the parts of the process that require a human touch.  Borrowers can be approved for a 5 or 7 year adjustable rate mortgage with down payments of at least 20%.  Unlike with traditional banks, the submission and review of documentation is performed entirely through an online portal, a welcome respite from the endless faxing or overnighting of paper experienced by most mortgage-seekers.  Privlo’s value proposition sits with its ability to use data and proprietary algorithms to qualify borrowers who might otherwise have trouble securing a mortgage, albeit at slightly higher than average rates.

Early Days – Challenges & Scale

Changes in a multi-trillion dollar, highly-regulated industry do not come easily.  Indeed, there are quite a few challenges that these new lenders will have to tackle in order to scale their businesses.  The involvement of the U.S. government in mortgage lending casts a long shadow, including a patchwork quilt of federal and state regulations.  By some estimates, Fannie Mae and Freddie Mac guarantee 77% of all mortgages originated in the U.S.  If you include Ginnie Mae, which guarantees loans made by the Federal Housing Administration, that number jumps into the nineties! 

In addition, few of the loan structures available in this market today resemble a traditional mortgage.  One key contributor to that fact is that of loan duration.  While the archetype of an American mortgage is a 30-year, fixed-rate loan, few investors would fund such a loan without a way to create shorter-term liquidity.  In traditional mortgage finance, liquidity is achieved with securitization and secondary loan trading.  As these innovations make their way into marketplace real estate lending, we will likely see things open up for longer duration loans and higher volume. 

Perhaps the most interesting challenge is the fact that we are dealing with an asset that exists in physical space.  As the old saying goes: “location, location, location”.  Real estate markets are highly local, and many investors want to examine a property in person, only wanting to invest in areas where they have a knowledge of market dynamics.  It remains to be seen if investors will be content with online photos and rich web-based content, particularly for properties not close to home.

For many people, a home is both their most expensive purchase and their largest investment.  With mortgage lending as a $8.2 trillion industry, this is perhaps one of the largest and most important sectors ripe for disruption as marketplace lending continues its forward march.

Industry Profile – Zhengyuan Lu of OnDeck

Briefly describe your career and how you came to work at OnDeck?

I have been in the capital markets space for nearly 20 years. Prior to OnDeck, I was a Managing Director of the Asset Finance Group at Gleacher & Company; Managing Director and Head of Structured Products Group at Keefe, Bruyette & Woods; SVP at Fortis Securities and WestLB AG; and Portfolio Manager at PPM America.

You come from a traditional finance background. How has the transition been to a tech company?

The transition has been exciting and OnDeck has been an eye opening experience for me.  While OnDeck has many similar characteristics of traditional lenders that I used to advise, the tech aspect makes OnDeck extremely unique. There is a tremendous amount of electronic data about small businesses that traditional lenders are ignoring or unable to process. Through our technology platform, we are able to digest and analyze a significant amount of data in real time, which allows us to make better credit decisions faster.

What do you like most about working at OnDeck?

It’s inspiring to be a part of a company that is helping small businesses succeed and our economy grow. We are on the forefront of fundamentally transforming an antiquated lending system through technology. 

How does OnDeck differentiate itself from the other online small business lenders?

We’re doing two things to disrupt the industry: we’re making short term working capital more accessible and we’re making the process faster so business owners can spend their time growing their business instead of trying to find financing.  First, our OnDeck Score enables us to make capital more accessible by incorporating business operations and performance data to more accurately assess a business’ health. Second, we’ve deployed a proprietary technology that delivers speed and convenience to small business owners, as well as superior service they deserve. We find that business owners generally want to speak to someone about their loans and so we have loan specialists available Monday through Saturday to talk to customers about their needs and options. 

What is one common misconception people have about OnDeck?

Many of our customers think we are similar to banks. They know we are more tech enabled than banks, but they are blown away with just how fast and easy the process is. We have a 10-minute, online application process (vs. days or months of bank paperwork) and we evaluate small businesses using thousands of data points (vs. relying heavily on business owners’ FICO scores).

What projects are you working on that have you particularly excited?

Our mission at OnDeck is helping small businesses, and small businesses have many financing needs. We are constantly testing new features and products to better serve small businesses, and I am focused on making sure we have flexibility on the financing side to meet all their needs.   We are also growing OnDeck Marketplace, where institutional investors can buy loans we originate directly. Accessing the securitization market is another big step in executing OnDeck’s financing strategies. Through securitization, we are able to continue improving our financing terms while diversifying our investor base, which significantly benefits all key stakeholders in the business.

Education is a big part of growing this industry. What is OnDeck doing to facilitate this?

Educating small businesses is very important to us. When we launched, the direct lending space was in its infancy. Most small business owners were not aware of alternatives to banks. Over the years, we have increasingly added new educational resources for our customers. For example, we have a blog and newsletter that distributes daily tips such as how to apply for a loan and how to better manage your business. We also host OnDeck events and have partnered with organizations such as SCORE to provide small businesses with resources they need to make vital business decisions. 

Are there any other companies in the space that you think are doing very innovative things?

There has been a secular change in the lending space.  Similar to OnDeck innovating the small business lending space, companies like LendingClub, Prosper and SoFi are changing the landscape of the personal lending space.  It is an exciting time for our industry. 

What’s next for OnDeck?

OnDeck’s vision is to transform how money flows to small businesses by leveraging our platform to enable all types of investors to make capital on demand a reality for small business owners. Think of how Priceline transformed travel, or Zillow has transformed home buying – we believe small business financing can be ongoing and friction-free. We’re excited to continue to bring new products to market for our customers and really be a partner to small businesses throughout their entire lifecycle.

What’s next for Zhengyuan Lu?  

My main focus on the capital markets side is to build the most durable and scalable financing platform to ensure that small business can always have access to capital from OnDeck.  We have made significant progress over the last couple of years, and the work is never done. 

What do you do for fun outside of work?

I am an avid traveler.  We started a summer tradition of seeing a new country every year since my youngest turned 4, and he is 11 now.  I am also addicted to running on the weekends.  I think running is therapy for the mind.

Do you have any favorite restaurants?

I live in Chappaqua, and we love Le Jardin Du Roi in town, a small family owned French bistro.  They serve very rustic French food that brings you back to south of France, and they always treat you like a family. 

Do you have any unusual talents?

When my kids were little, I started learning small magic tricks to amuse them.  I kept it going over the year so I have accumulated a lot of magic tricks.

 

 

Discussing the Small Business Lending Evolution

The history of business lending is as old as the earliest societies, and the related innovations are innumerable.  From the interest bearing loans issued in ancient Greece that made lending profitable for investors, to loans secured by collateral in the 1800s that made lending more secure for investors and therefore more affordable for borrowers, to the founding of the Small Business Administration (SBA) in 1953, which established that lending to small businesses is an important feature of our economy.  The landscape has been ever evolving.  The lending products available to a business owner today range from standard loans with a fixed interest rate and term to a merchant cash advance that automatically deducts loan payments from credit card receipts to lenders underwriting loans based on programmatic connections to the borrower’s Etsy account.

Given the size of Prosper and LendingClub, it may seem that the roots of online lending are in consumer lending, but there have also been many small business lenders innovating in similar ways for over 10 years.  There has been a recent storm of publicity and big news from these lenders, including the following:

Why is this an area so ripe for these new entrants?  Banks have traditionally been the go-to place for small businesses to borrow.  However, as businesses change and grow, so do their financing needs.  Making a credit decision on a small business is difficult.  The business’ finances are inherently combined with the personal credit of the owner, leading to questions about what to focus on in an underwriting strategy.  Most traditional lenders will look at some basic metrics on a business to make an underwriting decision (e.g. time in business, industry, revenues, business credit information, and the personal credit of the owner).  However, some of this data is not easily available or verifiable.  Revenues, for instance, can be verified by obtaining audited financial statements, but not all business owners have these, and reviewing them requires the work of a trained analyst.  The availability of alternative data via APIs (and other methods) has enabled lenders with the technical prowess to access and make sense of it in order to lend to businesses that, using traditionally available data, may have been declined.  Most of the companies listed above are pursing such an approach to small business credit underwriting.

Reviewing data from the FDIC, we see the trends over the past 10 years in bank lending to businesses.  Lending by the banks in the <= $1MM loan size segment (generally what is considered to constitute a small business loan) has decreased steadily since the peak in 2008. 

Digging deeper, we see that the $100K - $250K & $250K - $1MM segments follow this trend as well:

However, the < $100K segment displays a very different trend, with loan balances decreasing almost every month since 1995.

Given that the lenders discussed here are generally operating in this segment, it is clear they are positioned to fill an ever growing void.

Loan volume trends for the companies discussed here are not easy to find.  According to the Wall Street Journal, Business Financial Services estimates that online nonbank lenders provided ~$3B in loans in 2013.  Assuming this is mostly in the <$100K segment of lending, it comprises a significant percentage of the borrowing when compared to the FDIC data above.  While this data is hard to verify, given the size and growth of these companies it is clear that their importance will grow and become a serious source of capital for small to medium size businesses.  

In addition to the companies we’ve mentioned, we are also seeing many new offerings from established companies such as Amazon, Square, and Paypal, jumping in to finance their small clients and further validating the size and attractiveness of this asset class.  These companies see the void that needs to be filled and the advantage they have due to the information they collect on the firms using their products.    There are also companies devoted to assisting a business owner in finding the right loan from non-bank lenders, such as Fundera, Lendio, and Biz2Credit.

Many of these online loan originators were funding all or most of the issued loans with bank credit facilities, focusing their technological expertise on the borrower acquisition and underwriting segments of their business model. Orchard is excited about the expansion many are making to the marketplace lending model.  This structure opens up the potential for borrowers to access various types of investors who may have different investment goals than those of a traditional bank.  As long as banks are the only funding source for loans, only those borrowers that meet the criteria necessary for a bank’s business model will be funded.  The ascent of marketplace lending unlocks the possibility of various capital sources, and therefore various business models, to engage in the funding of loans for all purposes – including business growth.  Orchard is excited to be playing a role in this innovative and exciting new time in financial history, enabling the democratization of credit to a wider range of borrowers.

The Price of Credit - Interest Rates Over Time

A core aspect of any lending-related industry is the price of credit.  As marketplace lending has grown and matured since its inception, interest rates for borrowers have moved along with economic conditions and the supply of investor capital. In today’s analysis, we will explore the rates seen across the borrower spectrum on the major online marketplaces and compare them to other major consumer asset classes.

Average Borrower Rates Over Time

In the chart below, we show the mean borrower interest rate over time for LendingClub and Prosper.  Interestingly, these 2 originators have had a significant gap in averages for most of their lifetime, though the lines have recently converged.

Of course, simple averages never tell the whole story.  Each platform offers loans across a broad range of interest rates, and it behooves us to look at the full distribution.  In the chart below, we show a boxplot of interest rates, by originator, for each month of originations.  Since 2009, the lower bounds seem to be relatively similar for both originators, but the upper bounds are quite different.  Prosper offers some loans at interest rates well above those of LendingClub, thus bringing up their average.

To get a closer look at what is happening in the data, let’s take a look at interest rate distributions for distinct periods of time.  The charts below show the distribution of loans made by interest rate band for January 2013 and 2014 respectively.  While Prosper has maintained a presence in the higher interest rates, it has now shifted its distribution toward lower-priced loans.  LendingClub’s distribution has been rather consistent.

Context – National Consumer Credit

The table below is taken directly from the U.S. Federal Reserve website’s August 2014 release.

Interestingly, while these rates have changed over time, they have not fluctuated with the same order of magnitude we have seen on the marketplace lending originators.  This is perhaps an indication that interest rates for marketplace lending have shifted not solely because of the overall economic climate but also due to the rapidly changing nature of a nascent industry and the dynamics of borrowers and investors. As the industry matures, there is a possibility that we will see the same convergence and stability experienced in more traditional forms of lending.

July 2014 Marketplace Lending Index Returns

Month and Year-to-Date Asset Class Returns

As seen in our previous monthly marketplace lending recap, we will continue to benchmark the Orchard US Consumer Marketplace Lending Index against comparable US Fixed Income asset classes.  This will help both old and new investors in understanding their returns.  

The Orchard US Consumer Marketplace Lending Index continues to outperform many of the comparable US Fixed-Income asset classes on both a month-to-date and year-to date basis.  

Source: Fixed-income asset class returns based on Barclays Benchmark Indices

Source: Fixed-income asset class returns based on Barclays Benchmark Indices

Source: Fixed-income asset class returns based on Barclays Benchmark Indices

Source: Fixed-income asset class returns based on Barclays Benchmark Indices

European Platform Analysis: Bondora & Disposable Income

In March of this year, we discovered a European originator with publicly available data called isePankur and wrote a blog post based on that data.  In April, isePankur raised a significant round of capital and renamed itself Bondora.  Since our initial post, we have followed Bondora closely, been impressed with their growth, and admire their openness with data.  Similar to Prosper and Lending Club, Bondora publishes loan data on its website, allowing prospective investors (and the general public) to delve into the borrower attributes and performance of the loans on the platform.  We believe this transparency is great for the investment community.  Marketplace lending works because investors can make informed decisions about the types of loans that make sense for a given investment strategy. This can only be done effectively if there is data available to analyze. 

Given the richness of Bondora’s data, it is possible to run a multitude of analyses on the demographics and performance of loans on the platform.  In fact, Bondora collects certain borrower attributes (e.g. gender & martial status) that are never found on loan applications in the US.  While these are interesting, we felt it would be more relevant for our readership to compare an analysis that we've performed on Prosper and/or Lending Club with Bondora – we will use the disposable income analysis from March. 

First, we’ll take a look at the debt amount provided in the data.  This represents the debt payments the borrower is already committed to making each month prior to applying for this loan.  We see that the payment ranges from ~€200 to ~€650.  The number seems to have increased over time, which (from the lower graph) appears to be from introducing the new geographies (Finland, in particular).

Next, we look at the annual income distribution.  Based on researching the data, the income provided is monthly.  Once annualized, we see that the average income for borrowers on Bondora is around ~€15K.

Diving deeper into the distribution, we look at it by country and credit grade.  This gives us a bit more insight on the income distribution and how Bondora assigns credit groups (analogous to Prosper and Lending Club’s credit grades).  Finland appears to have higher income cutoffs for entry onto the platform.  It is also clear that income is an input into assignment of credit group.

Now that we have an understanding of the monthly debt and income of the borrowers, we can analyze the monthly disposable income.  Bondora calculates this in the same way we did when we analyzed this for Prosper - taking the monthly income and subtracting the monthly debt payments.  Below is the distribution.

The distribution of disposable income by country and credit group shows a more distinct picture than the monthly income distribution.  Disposable income is clearly a large driver to determining credit group.

Next, we look at the default rates by disposable income band.  Unlike Prosper, we do not see the same clear pattern when assessing the default rates (in this case defined as 30+ days past due) versus monthly disposable income.  Monthly disposable income, as calculated here, does not appear to impact the default rate.  This could be because disposable income is such a large factor in underwriting and pricing (as seen in above graphs) that it no longer provides any further insight in analyzing the loans once they are on the platform and funded by investors.

In conclusion, when we did this analysis on Prosper’s data, it was clear that monthly disposable income was a predictor of risk on those loans.  As the disposable income increases, the likelihood of default decreases.  This makes sense, as it would be easier for a borrower with more money every month to make a payment on time.  With Bondora, we don’t see the same pattern.  There are many possible reasons for this, including that Bondora clearly uses this metric in their decision making.  It is possible that the loans that Bondora chooses to make available to investors from borrowers with lower monthly disposable income have passed other credit criteria that make them lower risk in other ways.  Given the richness of Bondora’s data, we will continue to review it and write about it in future posts to answer these questions and others.

Marketplace Lending Meets Cryptographic Currency – Bitcoin Lending on BTCjam

Two of the most talked-about topics in the world of financial technology are marketplace lending and cryptographic currencies.  As we have explored in depth on this blog, marketplace lending is growing rapidly and democratizing access to capital for consumers and businesses around the globe.  Cryptographic currencies such as bitcoin offer the potential for a monetary system not controlled by central banks or governments, as well as the promise of decreased transaction costs and simplified cross-border commerce. 

Recently, some companies have endeavored to combine these two powerful concepts, the largest of which is BTCjam, a bitcoin-based lending marketplace that has processed $4 million in loans since 2012.  Using data available in the BTCjam API, we can explore the characteristics of loans offered for investment and also consider how bitcoin-based lending differs from more familiar approaches.

Data

For this analysis, we used a dataset of available listings downloaded from the BTCjam API on July 27, 2014.  At the time of retrieval, there were 201 listings, representing a total requested amount of $1,109 bitcoin.

Geography

As bitcoin is not dependent on any particular government, cross-border transactions are no more complex than those between 2 immediately adjacent neighbors.  As such, the BTCjam marketplace has attracted borrowers and investors from all over the world.  In the graph below, we see the distribution of active listings by borrower country.  While 38% of listings come from within the United States, 40 distinct nations are represented in the current dataset.

Amount

The average requested amount for listings available on the platform today is 5.5 BTC, an amount roughly equivalent to $3,282 at current exchange rates.  Below is a distribution of requested loan amounts in BTC, followed by the equivalent distribution expressed in dollars.  As we can see, the majority of loans offered on the platform are for less than $2,500, suggesting perhaps that the borrower profile, loan structure, or loan purpose differs from what we’ve seen on the larger marketplace lending platforms or, perhaps, that more performance history needs to be established before investors are comfortable funding larger loans, particularly in a currency that is not as broadly understood.

 

Currency & Exchange-Rate Risk

While we are accustomed to evaluating marketplace lending investments on the basis of credit risk, changes in the value of a currency also have the potential to affect returns.  When loans are denominated in a generally stable currency, such as the Dollar, Euro, or Pound, this risk may be considered relatively minor.  However, given the significant fluctuation in the price of bitcoin, exchange-rate risk may in this case present an even greater potential to impact returns than the underlying credit risk of the borrower!

As we can see from the graph above, based on the Coinbase API, the price of bitcoin in dollars has shifted massively over the past year, from a low of $92.13 one year ago, to a high of $1,126.82, to $595.07 today.  This volatility poses a major risk to borrowers on either side of the transaction and is one of the most common concerns in any discussion of bitcoin-denominated lending.

Fortunately, BTCjam offers 2 distinct types of loans, one of which is directly aimed at mitigating currency risk. 

  • “Bitcoin loans” are made in BTC and repaid in BTC. Both borrower and investor are exposed to changes in the value of bitcoin.  Investors who predict that bitcoin will rise in value would expect to benefit from funding these loans.  Borrowers who predict that bitcoin will decrease in value would theoretically benefit from borrowing under this structure, particularly if they receive their income in a fiat (i.e. government-issued) currency.
  • “Linked loans”, also known as “Bitstamp USD”, are made and repaid in BTC but are indexed to the bitcoin-dollar exchange rate on the date of loan origination.  For example, if a loan is made for a sum of bitcoin equal in value to $5,000, the payments on the loan will be based on the $5,000 initial value, regardless of any change in the value of bitcoin.  In this way, bitcoin serves only as the medium of exchange, and both borrower and investor are insulated from the exchange-rate risk.

In the graph below, we see the breakdown between the 2 types of loans, further segmented by the geographic location of the borrower.  As one might expect, BTCjam customers tend to opt for the dollar-linked loans, particularly those from the United States.


Purpose

A quick look at the data reveals a massive proportion of BTCjam loans to be for the stated purpose of “business”, with a relatively low number for debt consolidation.

 

As business and other are both very broad categories, they merit some more detailed examination.  In reading through the  borrower-provided loan titles and descriptions, a large number of loans appear to be for the purpose of either bitcoin mining or trading/arbitrage.  While these are not broken out as distinct categories, we can perform some text analysis of the loan information to draw our own inference.  As we can see from the graph below, a significant number of borrowers are obtaining loans in order to engage in mining – that is, to invest in the computational hardware needed to run the complex algorithms needed to “discover” new bitcoin.

In addition, many borrowers appear to be obtaining funding for the purpose of trading in the bitcoin exchange markets.

Loan Duration

While we have become accustomed to 3 or 5 year term loans on the mainstream consumer lending marketplaces, BTCjam loans tend to be of shorter and more varied duration, with terms ranging from 7 days to 1 year.    In addition, payment frequency is not necessarily monthly, with available payment cycle intervals of 1, 3, 7 or 30 days.

 

Risk and Interest Rates

As with other marketplace lending originators, BTCjam assigns borrowers an alphabetic score ranging from A+ to E.  This score is based on a variety of factors, including traditional credit data as well as identity verification and social graph information.  For this analysis, we don’t have access to performance data, so we will simply show the credit grade distribution, along with associated interest rates.


Alternative Data and Verification

One unique aspect of BTCjam is its use of alternative data for identity verification and risk-assessment purposes.    The site encourages its members to link their Facebook, LinkedIn, ebay, and PayPal accounts and then makes aspects of this data available to prospective investors.  While its hard to know exactly how these attributes affect credit ratings or interest rates, the  information itself is somewhat interesting.

 

Conclusion

Traditional economic theory teaches that money has 3 main functions: a medium of exchange, a unit of account, and a store of value.  Cryptographic currencies such as bitcoin challenge some of these notions.  For instance, earlier this year, the IRS ruled that bitcoin would be taxed like property, rather than like currency.  Fiat-linked loan options like BTCjam’s BitstampUSD also challenge traditional thinking, utilizing bitcoin as a medium of exchange and unit of account, but preserving the U.S. dollar as a stable store of value.  Clearly, cryptographic currencies still represent a bleeding-edge technology and remain the province of early adopters.  It is not yet clear if, or how, they will influence the evolution of our banking system.  Regardless of one’s feelings surrounding bitcoin as a currency, the promise of a more flexible transaction protocol that lowers costs and facilitates cross-border transactions is extremely compelling, and we are excited to see its impact on the world of marketplace lending.

Industry Profile – Kathryn Ebner of Pave

Given Kathryn’s sunny disposition, one wouldn’t guess that she is a finance industry veteran with many years of traditional banking experience spanning debt capital markets, short term credit trading, and institutional sales.   Trading in her pantsuits for a jeans and a hip tee-shirt, Kathryn seems very at home in her new role as head of business development for Pave as we meet for coffee at the charming Buvette in the West Village.

Revisiting Roll Rates using Markov Chains

In October of 2013, we did a post on Roll Rates for Lending Club.  We defined roll rates and analyzed the patterns on Lending Club data.   For the purposes of both posts, we’ll consider roll rates to be the probability of a loan going from one state of delinquency to another (e.g. 30 days past due to charge off).  Roll rates provide insight into the likelihood of future loss on a loan that has missed just one payment.  This can be used for:

1. Modeling: it is best practice to use more recent data as the current vintages are underwritten and aging during a similar period of time.  To actually charge off, an account needs at minimum 120 days (although – it usually takes at least 10 months or so.)  If you can use an early default predictor like missing 1 payment (i.e. 30 days past due) you can adjust the model with this information. 

2. Portfolio Analysis: If you know the roll rates of a population, you can then extrapolate the current performance of your portfolio earlier because you can use early indications as proxy for later credit loss.  This allows for adjustments in underwriting strategy prior to actually experiencing losses.

3. Loan Loss Reserve Calculations: If an investor keeps a reserve based on expected losses, roll rate calculations can be useful. 

In order to calculate roll rates, it is important to first understand the nature of charge-offs, as the analysis requires some assumptions.  These loans amortize over time, which means if the charge off occurs later in the borrower’s tenure, the principal charged off could become de minimis.  Below is the average dollar loss and percentage of original principal lost based on when the charge-off occurs in the life of the loan.  As expected, the average dollars lost decreases as the loans age, as does the percentage of the original balance that charges off.

As opposed to Lending Club (which only designates past due loans as 31-120 days late), Prosper’s data allows for a more granular assessment of the transitions between the different past due buckets.  Prosper provides the actual days past due of every loan at every month.  In order to calculate loan movement, we need to pick a starting point.  Based on the distribution of 30 days past due (below), 50% of loans that miss one payment do so by month 12, so this will be the starting point.

We analyzed all Prosper loans that have at least 18 months tenure to see the actual change in status from 12 months to 18 months.  We used a statistical method called Markov Chains, also known as transition matrices.  These are mathematical models that calculate the probability of an object moving from one state to another.  In this case, the states are the status of the loan, and the time period we are analyzing is 6 months (from 12 to 18 months).  By applying this model to the historical changes in loan status, we are able to extrapolate the likelihood of a loan in any status moving to another status in 6 month increments from the initial 12 month status.  

Below are the results of the analysis, in which we show the probability of movement between all of the different status buckets.  The likelihood of eventual charge-off on loans that miss just one payment by 12 months is quite high.  When predicting out to 30 months, our analysis shows the likelihood to be about 85%.

The Markov Chain is a useful tool, and was easy to use for this analysis.  In the case of roll rates, this analysis could be accomplished with any starting point and ending point.  In this case, we used 12 months to start and looked out 6 months.  If an analysis required even more recent data, then starting at 3 months and looking to 6 would work.  Making this change to the analysis would bring in more recent data (including all loans with at least 6 months of tenure, instead of 18), but the tradeoff would be a model that may under predict the probability of charge off because it goes out to 9 months instead of 18 months.

Using statistical methods, both basic and complex, can help yield preliminary results for a loan portfolio.  This analysis specifically helps validate the use of early default behavior as an input into a charge-off model.  Based on what we've found, an analysis done on a more recent vintage can yield results that are worthwhile.  Of course, more data is always better, and actual performance is always the optimal input to a model.  In the case where this is not available, statistical methods such as those discussed here are useful.

Repeat Borrowers – Relationship Lending for the Modern Age

Assuming that marketplace lending offers borrowers a good deal and a positive customer experience, we would expect to see repeat business.  In a post last October, we first analyzed repeat borrowers on Prosper.  That analysis found that 10% of loans were issued to those who had previously taken out a Prosper loan.  In Oct. 2013, Prosper originated $50 million in loans. Since then, Prosper has issued an additional $726 million, including $145 million in June 2014 alone.   Given the massive growth in the marketplace, we wanted to see how much of this growth was due to repeat borrowers, how these individuals have performed, and what we can learn from borrowers’ prior experience.

Prior Borrower Market Share

The charts below show monthly loan originations on Prosper as well as the share comprised by repeat borrowers.

As we can see, Prosper’s origination volume has grown significantly over the past 8 months.  Existing borrowers, while growing in absolute number, have comprised a smaller share of monthly issued loans over time.  This pattern is not surprising given the platform’s high rate of growth.  In order for Prosper to expand, it has presumably needed to cast a broader net in acquisition marketing to source borrowers from new channels.

When Borrowers Return

In the first half of 2014, Prosper made 3,455 loans to previous borrowers, totaling $40,185,826 in originations.  Using the data made available by Prosper, we can explore the characteristics of this population.  In the chart below, we group these borrowers by the number of previous loans they had taken prior to the new loan’s date of origination.   While the vast majority had taken only 1 prior loan, 35% had 2 or greater!

For the borrowers with one prior loan, one has to wonder how much time had passed between the issuance of the prior loan and the new one.  In the histogram below, we  show the distribution of time between loans, divided into 3-month intervals and broken out by the term of the previous loan.

Interestingly, a very large proportion of new loans were granted to borrowers who had already received Prosper loans under a year ago.  The vast majority of repeat loans were granted within 2 years of the prior origination, perhaps a surprising finding given that Prosper’s minimum loan term is now 3 years (a small number of 1-year loans were issued in the past, but this practice stopped in early 2013).   If so many borrowers are getting new loans prior to the initial maturity date of an existing loan on the same platform, there are various potential reasons to consider.

  1. The borrower is current on the existing loan and has sufficiently good credit to qualify for an additional loan (potentially to refinance the first one)
  2. The borrower has paid off the existing loan and now qualifies for an additional loan
  3. The borrower defaulted on  the existing loan (note – this scenario would not occur within Prosper’s underwriting policy, as prior defaulters are  restricted from obtaining new credit)

Understanding the frequency and details of the above scenarios is an analysis unto itself and perhaps will be the subject of a future post.

Creditworthiness of Prior Borrowers

Presumably, if Prosper has decided to give an additional loan to a prior borrower, its own credit rating system has favorably evaluated this individual.  Therefore, to get an externally consistent view of the creditworthiness of repeat vs. first-time borrowers, we must use an externally-developed model: in this case, FICO.  In the graph below, we show the distribution of loans by FICO score, split out by the number of prior Prosper loans at the time of origination.  The graph  reveals a relatively stable distribution between new and prior borrowers, with some rightward-skew for those with a previous loan.  Perhaps the higher scores for these individuals reflect good performance on the earlier loan.

Conclusion – Relationship Lending

The concept of “relationship banking” has existed for decades in traditional financial services.  The idea was that customers who were already doing business with a given institution would prefer to come back to that same institution as repeat customers and for additional services.  Embedded in this concept was the assumption that this would allow a bank to make more accurate credit decisions, aided by a wealth of information on existing customers and the ability to consider a customer’s lifetime value.  With this idea in mind, retail banks across the country expanded into a wide variety of financial services, such as checking, savings, credit cards, mortgages, auto loans, student loans, and personal loans.  For a time, this was quite successful.  However, the internet has essentially driven switching costs to zero, and consumers increasingly decide to do business with whichever firm is able to provide the best cost for the best service.  Clearly, this unbundling has been to the benefit of marketplace lenders, who have been able to deliver a positive customer experience without the burden of legacy technologies, and whose operating efficiency has allowed them to offer favorable rates.  As this new wave of loan origination platforms grows to comprise a larger share of overall lending, it will be interesting to observe if they are able to use relationships and data to their advantage to make better decisions and win the loyalty that had been so highly sought by traditional banks.

Vacation Loans on Prosper & LendingClub

As many of us were enjoying a 4th of July weekend vacation, at Orchard we used it as an opportunity to review “vacation” loans on both Prosper and LendingClub.  Both of these lenders offer this as a loan purpose option.  Vacation loans are of interest to us because there has always been some discussion on the general idea of a vacation loan and the performance of these loans on the different discussion boards such as Lend Academy.  Also, if you Google “vacation loans,” LendingClub comes up in the paid adwords section. 

These loans get attention, because they represent an interesting idea: if people need to borrow money to take a vacation, should they take it at all?  However, people have always financed vacations.  Before Prosper and LendingClub existed, the only option was to put the trip on a credit card and pay it off over time.  The interest rates on credit cards are generally higher than those on a marketplace lending term loan.  Given this, is there a difference between consolidating debt (which was potentially incurred on a vacation) and using a loan to fund the vacation in the first place?  For the borrower, a marketplace lending loan is a great new alternative.  For the investor, we’ll have to look at the data to find out if the investment is worthwhile.

Prosper started tracking vacation loans as a separate category in December 2011; LendingClub has been tracking them all along.  For the analysis in this post, we will look at December 2011 data and forward for both Prosper and LendingClub.

Below, we show Prosper’s loan distribution by loans for vacation, and those that are not.  As expected, vacation loans make up a small percentage of overall loans (~1% of listings.)

The actual volume of listings ranges from ~30-120 listings per month.  There are a couple periods of larger volume, but given the relatively small numbers, this is probably due to natural fluctuations and not telling of anything significant.

Some listings become loans, while others do not, as we discussed in this blog post.  Of the vacation loan listings, the percentage that fully fund has been increasing in recent months.  On average, about 55% are funded.  This is lower than the overall average funding rate of 61%.  The difference in funded loans is mainly due to a larger volume of loan requests withdrawn by the borrower.  This could be because the rates are higher than a borrower would like to pay to fund a vacation – or not any better than the rates on a credit card.  Because we are not provided the reason a borrower withdraws a loan, we will not be able to know.

For LendingClub, we see similar patterns.  The percentage of loans is slightly lower at 0.5% on average.  Keep in mind that LendingClub’s data includes only loans that have funded (i.e. listings that do not become loans are not included in the available data), so the percentages below include only funded loans.

The number of loans funded for vacation purposes on LendingClub ranges from ~20 – 140 loans per month.  It appears that there is a bump in the number of loans for vacations during the summertime.  Similar to Prosper, given the relatively low volumes here, it is difficult to know if this represents a real pattern.

Prosper's distribution by credit grade for vacation loans skews to the less risky credit grades in the more recent months.  In fact, there are very few vacation loans in the E or HR credit grades.  It is unclear how much of this is due to a change in applicant mix versus Prosper's underwriting strategy.  What it does mean is that the average interest rate for a vacation loan on Prosper has gone down over time.  

LendingClub vacation loans are skewing in the other direction.  It appears that around December 2012, there may have been a change to the underwriting criteria used by LendingClub, because the distribution changes fairly dramatically.  There is no way to know for certain, but the more recent vacation loans are in riskier credit grades on average.

With the understanding of the relative size and distribution of these loans, we will now assess the default rates (in this case, default is defined as any loan 30+ days past due).  For this analysis, we only look at loans originated from December 2011 – December 2012, excluding more recent vintages.  Prosper vacation loans have a slightly lower default rate than other types of loans; LendingClub vacation loans have a slightly higher default rate than other types of loans.  Overall, the pattern is not clear given the difference between the two originators.  The difference could have to do with either applicant mix or underwriting.

Given that the distribution in credit grades was fairly different for the two originators, we need to look at these default rates by credit grade.  However, it is important to keep in mind that the volumes are small, so these results may not be significant.

For Prosper, it appears that the AA credit grade vacation loans have a higher rate of default than other types of loans (this accounts for only 12 loans total – so this finding could easily be an anomaly.)  Vacation loans outperform the other loan types in all of the other credit grades.

For LendingClub, we see a similarly high default rate on the vacation A grade loans (which accounts for 107 loans).  For the rest of the credit grades, the vacation loans are generally worse.

In conclusion, the number of vacation loans issued by Prosper and LendingClub is small, and therefore it is difficult to say if the patterns and trends we see in the data are due to sample size or the behavior associated with these loans. 

Overall, when compared to the alternatives, it might make sense for a borrower to take out a loan to fund a vacation and, according to the data, the performance may be in line with certain investors' strategies.  According to the numbers in this analysis, these loans could be both profitable to investors and an interesting new way for those who are cash strapped to avoid costly credit card debt in the first place.  Given how new this asset class is, it makes sense to continue assessing the performance and default rates to see how the data plays out before making an assumption about the quality of these loans. 

We hope everyone had a great holiday weekend!

June 2014 Marketplace Lending Index Returns

Month and Year-to-Date Asset Class Returns 

For the month, the Orchard US Consumer Marketplace Lending Index outperformed many of the comparable US Fixed-Income asset classes.  As a reminder, the Orchard Index is designed to measure the performance of direct online lending to US consumers. The Orchard Index tracks the performance of the aggregate amount of loans to consumers originated and funded on eligible US-based online lending platforms. Today those platforms include Lending Club and Prosper with more to be added in the future.

Source: Fixed-income asset class returns based on Barclays Benchmark Indice

Source: Fixed-income asset class returns based on Barclays Benchmark Indice

The Orchard Index serves as a catalyst in promoting marketplace lending as a major investable asset class. Like other asset classes, investors will now have the ability to benchmark their returns against an index. For this reason, we will begin providing monthly recaps of the month-to-date and year-to-date returns of the Orchard Index against five comparable US fixed-income asset classes:

1.     3-5 year Treasuries,

2.     Intermediate High Yield Corporates,

3.     3-5 year Investment Grade Corporates,

4.     5 year Municipals, and

5.     Securitized Products.

US fixed-income returns are sourced from Barclays benchmark indices.        

Source: Fixed-income asset class returns based on Barclays Benchmark Indice

Source: Fixed-income asset class returns based on Barclays Benchmark Indice

Similar to the month-to-date returns, the year-to date returns for the Orchard US Consumer Marketplace Lending Index outperformed several of the comparable US Fixed-Income asset classes.

Orchard Research & Blog

Industry Profile – John Birge of CommonBond

Understanding Loan Statuses

Cash Drag and Marketplace Lending

Borrower Impact on Consumer Marketplace Lending

Understanding Debt-to-Income Ratios