New expected loss methodology

By Luke OMahony

Today we have updated our expected loss calculations and also expanded the information we provide on our statistics page.

On 3 April 2017, we announced that we would revise the guidelines and methodology for assessing expected future losses and update our expected future loss estimates each quarter. An overview of the revised methodology is in the table below.

This means that four times a year, we bring in the most recent data for each type of lending and re-examine our forecasts based on this data, adjusting them as necessary. The most recent update has gone live today, and has led to a decrease in expected future losses and a consequential increase in the Coverage Ratio.

We also said that we would expand the data we publish on the current and past performance of our lending, and this can be accessed here.

A guide to the methodology for expected future losses

Old methodology

New methodology

Segmentation of loans

One-dimensional – loans segmented by risk grade. No differentiation by term, channel or origination date.

Multi-dimensional – loans segmented by risk grade, term, channel and origination date.

Loan performance reference data

Losses on all loans mechanically refer to data taken from a 12 month sample of loans written between 18 months and 30 months ago.

Losses on loans > 9 months old refer to actual data for that specific segment of loans, which is extrapolated out. Losses on loans < 9 months old are calibrated based on cohorts with similar characteristics (e.g. average score, loan size, APR, early delinquency rates) and are back-tested and adjusted on a quarterly basis.