Interpreting LendingClub Net Annualized Return from Loan Payment Data
Summary
The document examines how to calculate LendingClub’s net annualized return from historical loan payment data. It lays out a monthly formula that combines interest, late charges, service charges, charge-offs, and recoveries, weighted against remaining principal. The author asks whether principal cancels in the numerator and whether loan lifetime totals can substitute for installment-level monthly figures when aggregating observed payments.
As an attempted cross-check, the author classifies charged-off, defaulted, late, and grace-period loans as defaults, yet obtains a portfolio return higher than LendingClub’s advertised historical figure. This discrepancy motivates questions about the interpretation of the formula and treatment of incomplete loan histories. The document presents no resolution or validated recalculation, so its return estimate and aggregation assumptions remain uncertain. In particular, collapsing cash flows across months may affect annualization and weighting, and the supplied data may not support all required monthly inputs.
Key ideas
- The formula aggregates monthly interest, fees, losses, recoveries, and remaining principal to estimate return.
- The author questions whether principal factors cancel in the weighted numerator.
- A portfolio calculation produces a return above the advertised historical level, prompting an audit of assumptions.
- Replacing monthly installment data with lifetime totals is proposed but not verified.
Tags
Full text
# Calculating Net Annualized Return on LendingClub historical data
# Calculating Net Annualized Return on LendingClub historical data
I am interested in the formula LendingClub provides as their measure of "Net Annualized Return":
$\big(1 + \frac{\sum_{i=1}^N{((I_i + L_i - S_i - C_i + R_i) / P_i) P_i}}{\sum_{i=1}^N{P_i}}\big)^{12}-1$
where the variables $I,L,S,C,R,$ and $P$ denote interest, late charges, service charges, charge offs, collections recoveries, and principal remaining, and $i=1...N$ parametrizes months in the portfolio. For example, if we have one loan that matured over Jan 2008 - Jan 2011 and another that matured over Feb 2009 - Feb 2014, $i = 1...N$ would index over all the months from Jan 2008 to Feb 2014.
First, I would like to confirm that in the numerator, the $P_i$ are indeed supposed to cancel, and LendingClub is merely trying to illustrate that they are taking a "fraction of principal" explicitly. The variables are all available in the full historical payment data (see All payments at the bottom of the link) LendingClub provides. However, I am having some trouble applying the definition.
Taking the conservative estimate and assigning all loans that are Charged Off, Default, Late (31 - 120 days), Late (16 - 30 days), and In Grace Period as defaulted and incorporated into the $C_i$ charge-off term, I still get a 10.9% portfolio-wide NAR. This seems high, since LendingClub advertises historical returns of ~6%.
It should be noted that LendingClub does not provide this information on an installment-by-installment basis except principal remaining, $P_i$, so I simply took total interest received, late fees received, etc. over the lifetime of the loan. I believe this substitution is acceptable because the sum ranges over all non-censored installments, so after collapsing sums it should produce the same result regardless of whether the loan yet reached maturity.Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.