Estimating Returns on Consumer Loan Portfolios with Outstanding Balances
Summary
The document presents a proposed way to compare bundles of peer-to-peer consumer loans that have not matured. The available data include borrower characteristics recorded at origination and payment histories observed to date. The questioner seeks a valuation method, ideally a structural credit model, for loans with a common 36-month maturity, and describes a current return-on-investment calculation.
That calculation aggregates interest received, servicing fees, and expected losses on each loan's remaining principal, then scales the net amount by an estimate of invested principal based on loan rates. Expected loss is assigned using fixed percentages for statuses such as late, charged off, current, or fully paid. The document gives these inputs and the formula but provides no answer evaluating their accuracy or recommending a replacement. The status-based loss assumptions are therefore a rough proposal, not validated estimates; incomplete loan histories and outstanding balances limit conclusions about final performance.
Key ideas
- The goal is to compare performance across groups of consumer loans that have not matured.
- The proposed ROI combines interest received, servicing fees, and expected losses on remaining principal.
- Expected losses are assigned using fixed assumptions tied to each loan's payment status.
- The document does not validate those assumptions or supply a structural credit model.
Tags
Full text
# How to value a portfolio of non-mature consumer loans?
# How to value a portfolio of non-mature consumer loans?
I'm looking for the best way to value a portfolio of consumer loans that have NOT reached maturity and for which I do observe the payment/default history to date?
I'm working with a large database of consumer (peer-to-peer) loan data. The majority of these loans have yet to reach maturity. Still, I observe the payment history as of, say, today. For each loan, I observe a vector of borrower characteristics (accurate to the time of loan origination) including credit score, city/state, occupation, monthly income, debt/income, homeowner indicator, loan purpose, previous loan outcomes, etc. All the loans mature after 36 months.
I'd like to compare the performance of various bundles of these loans. I'm not sure what's the approach. Ideally, I'd like to fit the data to some structural credit model.
Until now, I've been using the following approach for computing ROI (found here http://www.nickelsteamroller.com/#!/stats/roi):
$ROI \triangleq \frac{\sum_{i} \left(I_i - S_{i} - \textsf{E}[L_{i}] \right)}{\sum_{i} \left(\frac{I_{i}}{r_{i}}\right)}$
where
$I_{i} \triangleq $ cumulative interest payments received by creditors on loan $i$,
$r_{i} \triangleq $ interest rate on loan $i$,
$S_{i} \triangleq $ cumulative loan servicing fees paid by creditors on loan $i$, and
$\textsf{E}[L_{i}] \triangleq $ expected loss on remaining principle for loan $i$. To compute expected loss, I use the following loss estimates conditional on loan status:
1 Month Late: 70.00%
2 Months Late: 80.00%
3+ Months Late: 95.00%
Charged Off: 100.00%
Current: 0.00%
Default: 96.00%
Fully Paid: 0.00%
In Grace Period: 30.00%
Late: 50.00%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.