Loan Supply and Demand Information for Lending Club Retail Platform
Reading List: Securitization, Forecasting, and Food
Predicting Demand on Lending Club Retail Platform from Loan Availability Data
In our last blog post Loan Supply and Demand Information for Lending Club Retail Platform, we summarized the available PeerCube resources for loan supply and demand information for Lending Club retail platform. We have multiple information resources for loan listings. However, there is only one information source for loan funding, SEC filings of Lending Club sales report. Lending Club files this reports weekly. Can we use a more current listing information source for estimating the loan funded on Lending Club platform? In this blog, we will take a stab at it using the archived data for Loan Availability on Lending Club Primary Platform.
Dataset
We archive the new loan and total loan count for both Lending Club and Prosper platforms with timestamp in a database table. At the time of retrieval for this blog post, the dataset contained 244,308 observations for Lending Club retail platform since June 2013. We removed 123 observations where Total Loan and New Loan values were zeros as these were most likely errors.
Using the new loan and total loan count, we calculated the loans that exited the platform since the last time system checked the available loans on Lending Club platform.
exit_loant=2 = total_loant=1 + new_loant=2 - total_loant=2
Why did we use the term Exit Loan instead of Funded Loan? The loans can be removed and no longer detected on the platform due to one of the following reasons:
- Loans are being fully funded by lenders (Funded Loans),
- Loans are removed by Lending Club for variety of reasons such as application withdrawn by borrowers, or
- Loans are expired due to not being fully funded.
Therefore, any prediction, if any, about loan demand only based on Exit Loan is going to over-estimate actual demand.
Data Exploration
The two charts below show the time-series for Total Loans, New Loans and Exit Loans. First Chart displays the data since July 2013 while Second Chart shows a narrow window since January 2016. A few observations:
- The growth in availability of loans is quite apparent from the charts. While in 2013 the total loans were averaging about 500, by 2016 the total available loans were averaging over 1,000.
- The high count of exit loans seem to appear near the peak of total loans. This relationship should be obvious from our equation for calculating exit loans. Such large count for exit loan may also indicate that the removal of loans from platform at these times may not be the result of loans being fully funded by retail lenders only. If Exit loans only consisted of loans being funded by retail lenders, the peaks in Exit Loans will be shorter, distributed over time and appear often.
- There seems to be a sudden increase in the volume of new loans released to retail platform between March 21 and May 9, 2016. Though this increase in loan supply was easily absorbed by retail lenders, this pattern is very unusual considering the events transpired at Lending Club in that time frame. What was the motivation behind releasing unusually high number of new loans to retail platform?


Analysis
The biggest challenge in any attempt to estimate the volume of loans funded was to identify, in the absence of actual funded data, the instances when loans were more likely funded and when loans were more likely to be removed from the platform for other reasons.
As shown below in R code output, we didn't notice any pattern using exit loan count and exit loan as percentage of total loan that could be used for estimating portions of exit loans that may be funded by lenders on retail platform.
> library(dplyr) > lc0[,c(3:4,6:7)] %>% arrange(desc(exit_loan)) %>% head(10) total_loan new_loan datetime_ct exit_loan 1 2460 0 2015-07-01 04:01:10 1598 2 651 0 2016-06-07 15:51:03 1367 3 1363 0 2016-03-29 07:14:09 463 4 874 7 2014-06-01 04:00:09 445 5 431 0 2016-03-31 10:24:04 437 6 469 0 2015-10-02 19:38:06 389 7 480 195 2014-01-20 23:35:21 384 8 4058 0 2015-07-01 02:07:11 355 9 631 0 2015-10-20 15:51:07 348 10 1393 0 2016-03-28 12:14:10 334 > lc0[,c(3:4,6:7)] %>% mutate(exit_tot_pct = exit_loan * 100 / total_loan) %>% arrange(desc(exit_tot_pct)) %>% head(10) total_loan new_loan datetime_ct exit_loan exit_tot_pct 1 17 0 2015-10-01 20:38:01 139 817.6471 2 28 0 2014-02-28 10:35:03 213 760.7143 3 17 0 2015-10-01 19:38:02 99 582.3529 4 60 0 2015-12-31 16:07:02 137 228.3333 5 651 0 2016-06-07 15:51:03 1367 209.9846 6 48 1 2013-09-18 14:30:02 82 170.8333 7 58 3 2013-09-19 10:03:02 97 167.2414 8 49 0 2015-12-31 19:38:02 77 157.1429 9 116 0 2015-10-01 19:24:02 176 151.7241 10 29 19 2013-08-23 18:20:12 38 131.0345
We also tried to make use of the assumption that loans exiting from the platform at the time of new loans release were most likely to be funded by retail lenders. However, while new loans are released in an instance, the loans take a while to be fully funded and removed from the platform. The loan availability data didn't provide any reasonable method for distinguishing the loans exiting from platform due to being fully funded by retail lenders from loans exiting the platform for other reasons.
By aggregating the loan availability data over different time periods, we noticed that the pattern in Exit Loan almost mirrors that in New Loan; moreover, the count of loans exiting the platform is almost same as the count of new loans added to the platform.
Conclusion
While we can review the loan listing trends with time from the dataset, any attempts to predict the loan funded from limited information in the dataset didn't produce meaningful results. The only timely available loan funding information source appears to be the SEC filings of sales report by Lending Club. We thank analysts at Morgan Stanley for pointing us to this source. Using these SEC filings, we display the Historical Weekly Loan Funding Trend on Retail Platform at Loan Availability on LC Primary Platform.
Comments: (3)
- Rick Meechan | Friday July 8, 2016, 11:05 am
I'm new. would like to use Ms. T's formula. how do I automate it?
- Anil Gupta | Friday July 8, 2016, 1:39 pm
Rick,
What is "Ms.T's formula"? Once you log in to PeerCube, all available automation tools are listed at upper-right corner of the home page.
- Cchateau | Saturday July 16, 2016, 4:45 pm
You seem really knowledeable. Are you still investing in prosper loans? I'm posting my listing because I have an emergency home repair to the roof. Insurance won't cover it because they said it was normal wear and tear!! Every bit helps. I can send you my contact info if you haven any questions at all. Thank you.
https://www.prosper.com/listings#/detail/5249389