Introducing the Bondora Rating

Stats & Data

As announced in some earlier newsletters, we are going to replace the current credit groups and payment history criteria with a credit scoring model that will assess each loan application separately based on specific criteria and assign them a certain Bondora Rating accordingly.

With the introduction of the new credit scoring system, every loan application will be assigned a Bondora credit Grade. Credit Grades will allow investors to easily consider a loan application’s level of risk because each rating represents an estimated average annualized expected loss rate range to the investor.

In addition to facilitating credit evaluation process, the new credit scoring system enables risk-based pricing of loans, meaning that lower risk borrowers will get better rates than the riskier ones.

As shown in the figure below, the rating scale will consist of seven grades from AA (the safest grade) to F (the riskiest grade accepted for the marketplace).

In addition, there will be a grade (HR) for high risk loans with the annualized expected loss of 25% or higher. The HR grade will not be included in pre-defined Portfolio Managers, but will be available for investors on the Primary Market and custom Portfolio Managers.

The expected loss is expressed as a percentage of the original loan amount.

Grade Min Expected Loss Max Expected Loss
AA 0% 2.0%
A 2.0% 3.0%
B 3.0% 5.5%
C 5.5% 9.0%
D 9.0% 13.0%
E 13.0% 18.0%
F 18.0% 25.0%
HR 25.0%

In short, the credit model accounts for risks for each loan application separately on both the micro-level (PD, LGD, EAD%, M), which in turn account for the country of residence, credit history and other criteria, and the macro-level (unexpected risk, country risk factor).

The calculations are based on 3 different data sources:

A) Internal data that is available to all investors through the public datasets.

And from proprietary sources that cannot be shared with the public:

B) External data that we cross-check with other databases.

C) Behavioral data based on actions taken on our website.

This will allow pricing each loan application with a fair interest rate for both the investors and borrowers, while still allowing for a buffer for unexpected risks.

The following paragraphs will give a more thorough explanation of how the credit scoring model works and what it accounts for in the calculations.

Expected Losses (EL)

Technically, each loan’s expected loss is calculated as the product of the three key credit risk parameters:

EL% = PD*LGD*EAD%

PD, Probability of Default, refers to a loan’s probability of default within one year horizon. PD-s are estimated based on the internally developed scorecards that are enhanced with the external credit bureau data and scores (Krediidiinfo data for Estonia, Equifax scores for Spain, Asiakastieto Consumer Risk Indicator for Finland).

LGD, Loss Given Default, gives the percentage of outstanding exposure at the time of default that an investor is likely to lose if a loan actually defaults. This means the proportion of funds lost for the investor after all expected recovery and accounting for the time value of the money recovered. In general, LGD parameter is intended to be estimated based on the historical recoveries. However, in new markets where limited experience does not allow us more precise loss given default estimates (currently: Finland, Spain and Slovakia), a LGD of 90% is assumed.

EAD%, Exposure at Default (expressed as a percentage of the original loan amount), indicates outstanding investor exposure at the time of default, including outstanding principal amount plus accrued but unpaid interests.

Further, the calculated expected loss is adjusted for loan term as in various situations, longer term loans are considered riskier. Currently, a Maturity Factor M of 1.3 is assumed for loans with duration exceeding one year.

Pricing algorithm

Put simply, with the introduction of the risk-based pricing, a loan’s interest rate (I) will be calculated as the sum of a loan’s expected loss rate (EL%) and the expected return of that loan, the E(R):

I = EL% + E(R)

From an investor’s perspective this means that his or her expected return is derived as:

E(R) = I – EL%

Expected return should be sufficient to cover the cost of capital at a given risk level. The minimum acceptable return or the cost of capital is most commonly determined by the capital asset pricing model (CAPM). In addition, an active investor might assume certain alpha return.

Accordingly, Bondora pricing algorithm for estimating the E(R) component is based on the CAPM; CAPM parameters are “translated” into the context of credit business.

For those interested in details: as market index, we have selected Markit iBoxx Eurozone Retail ABS Index EUR as the most representative currently available index for Bondora loans.

A loan’s beta estimate consists of the two building blocks:

1) Unexpected loss corresponding to a loan’s individual risk characteristics

Unexpected losses are losses above expected levels that an investor may incur in future, but which cannot be predicted neither in terms of timing or severity. Our estimate of a loan’s unexpected loss is derived from the fairly standard Basel II IRB Risk Weight Function that is generally used by banks for calculating their regulatory capital requirements.

2) Country risk factor

Country risk accounts for the fact that different eurozone countries are likely to react differently to systemic risk events. Indeed, while efforts are being made towards the single market, considerable financial fragmentation was observed during the past crisis and can be assumed in future crises should this be the case. The country risk factor accounts for that fact. It should be interpreted as country beta, given the eurozone average beta of 1.