Over a year ago, I did a post about big data and peer to peer lending. Big data was all the rage in terms of creating filters for loan selection, and still is for many small investors. Everyone was slicing and dicing the data including stupid things like if FL has more defaults than other states, then don't lend to FL residents. The 'logic' in this method means we are clearly excluding many outstanding FL borrowers. My 3 factors in this year old post of why not to be misled by big data still apply.
Now don't get me wrong, the fact that we have access to all this data and we can analyze and manipulate it however we want is awesome. This decentralization and democratization of lending and of lending information is a big win/win for all involved.
Credit analysis like I espouse, where we look at numerous factors in relation to each other (e.g a lower FICO score can be made up for by many years of steady employment in their field), is taking hold as a more common sense form of filtering.
While analyzing credits for over $100 million in equipment finance transactions, it's become clear that strength in some areas can make up for deficiencies in other areas. Tax liens and judgments are still worth filtering since if a borrower isn't going to pay Uncle Sam or pay when the Court orders them to, they likely are not going to pay us either. Most filters cannot be looked at in a vacuum this way though and I'm happy to report that many people are learning this valuable lesson. Here are 2 reasons to continue to not be misled by big data. If you aren't a big data analytics person (I'm not) then hopefully you will find encouragement from the following.
Reason 1 Not to be Misled: Loans are too new
Lending Club's own data from its Statistics page helps reinforce some of the issues of analyzing big data too finely.
As of September 30 2014, LC has issued a whopping $6.2 billion in loans. That's pretty awesome. Just looking at it visually, you can see that $4 billion of these loans were originated in 2013 or 2014. This means that at least 2/3 of these loans have not gone full term and the only defaults that can be measured by loans over these 2 years are early payment defaults. Some of the earliest 2013 loans are getting close to being able to be realistically measured since many defaults historically occur around month 18 but we are still a few months short of that.
The infographic on the Stats page tells us that total loan volume at year end 2012 is $1.178 billion. This means that just a hair over $5 billion of the total $6.2 billion in volume is within the last 2 calendar years. This does not include the huge 4th quarter that LC had which included the historic act of going public. The 4th quarter numbers will only skew this data even more.
Reason 2 Not to be Misled: Unproven in full economic cycles
Due to the newness of the loans, and the quiet period that both LC and Prosper went through where Prosper was sold to its current owners, peer loans have not gone through a full economic cycle of a boom and bust period. The most recent recession, which I am going to be liberal and count through Q4 2010, has LC at only $202 million in loan volume per the Infographic. This is 3.25% of the total loan volume listed in this graphic and definitely not meaningful for data analysis purposes.
Until billions of dollars of LC loans term out and loan origination and performance are proven through full economic cycles (at least 1 of them), it is difficult, and in my view a mistake, to completely rely on big data for your loan filtering. Common sense credit analysis of evaluating why one 720 FICO score is better than another will always be here to help us enhance our returns and reduce risk.