Core Path · L3.3.10

Expected loss

12 minCredit riskExpected lossPricingRAROC

Know these first

Math Atoms

By the end you will be able to

  1. 1Compute expected loss for a loan and for the bank's corporate and retail books.
  2. 2Explain expected loss as an expectation over outcomes, and why it adds up across a portfolio.
  3. 3Build a risk-based loan price from funding, expected loss, costs and a return on capital.
  4. 4Use RAROC to judge a proposed price, and explain why one price for all borrowers fails.

Why it matters

Graybeck Plastics owes RiskAtoms Bank £12m: about 2% of the corporate book. Yet it accounts for 36% of the book's expected loss. Expected loss is the cost of lending that the bank knows about in advance. It has to be priced into every loan and set aside in provisions, and it shows at once where the risk in a book sits.

Core explanation

The average loss, and what it costs

Illustrative dataRiskAtoms Bank plc's credit book

One made-up lending book on one made-up date (31 December, year 0), in pounds: a ten-grade master scale, a one-year rating transition matrix, twelve invented corporate borrowers (£587m of exposure, expected loss £2.81m), retail portfolios and twenty past workout cases. The firms are invented and the numbers are round illustrative values, not anyone's data. The same book runs through every lesson in Module 3.3.

The three parameters of the last three lessons meet here. Expected loss (EL) is PD×LGD×EAD\text{PD} \times \text{LGD} \times \text{EAD}, a one-year figure unless stated (L2.2.03: Credit and counterparty risk). For the twelve corporate borrowers, with PDs from the master scale (L3.3.04: Internal ratings and master scales), LGDs after collateral (L3.3.08: Loss given default and recovery) and EADs after CCFs (L3.3.09: Exposure at default):

BorrowerGradePDLGDEADELShare of EL
Graybeck Plastics Ltd911.31%75.0%£12.00m£1,017.9k36.2%
Hastemoor Construction Ltd84.53%55.0%£24.25m£604.2k21.5%
Bramwick Stores plc61.13%45.0%£76.00m£386.5k13.7%
Larkhollow Freight Ltd72.26%35.1%£29.00m£230.1k8.2%
Vellacourt Hôtels SAS72.26%21.7%£39.00m£191.2k6.8%
Mirefield Care Homes Ltd61.13%19.3%£44.25m£96.5k3.4%
Quorrel Systems BV50.57%45.0%£30.00m£77.0k2.7%
Pellcourt Properties Ltd50.57%10.9%£120.00m£74.4k2.6%
Thornvale Engineering Ltd40.28%31.0%£52.50m£45.6k1.6%
Ostreth Energie GmbH30.14%40.0%£70.00m£39.2k1.4%
Sallowmere Logistics GmbH40.28%30.8%£45.00m£38.9k1.4%
Cindral Pharma Inc20.07%40.0%£45.00m£12.6k0.4%
Corporate book£587m£2.81m100%

The book loses 0.48% of its exposure in an average year: its EL rate. Two weak borrowers, Graybeck and Hastemoor, hold 6% of the EAD and 58% of the EL. The large, well-rated, secured loans barely register. Retail is a much bigger book with a similar EL rate:

Retail segmentEADPDLGDEL
Mortgages: LTV up to 60%£1,470m0.3%5%£0.22m
Mortgages: LTV 60–80%£1,640m0.6%12%£1.18m
Mortgages: LTV 80–95%£710m1.5%20%£2.13m
Personal loans: Score band A£190m1.2%65%£1.48m
Personal loans: Score band B£130m3.5%70%£3.19m
Personal loans: Score band C£50m9.0%75%£3.38m
Credit cards: Transactors£91m0.8%80%£0.58m
Credit cards: Revolvers£204m4.5%85%£7.80m
Retail£4,485m£19.96m

Retail loses 0.45% a year on average, and the bank as a whole expects to lose £22.8m a year. Mortgages are most of the retail exposure, but cards and personal loans, with high PDs and LGDs, carry most of its expected loss.

An expectation, not a forecast

Expected loss is an expectation in the sense of M-C02: Expectation: a probability-weighted average over everything that can happen. Take Bramwick Stores plc, and suppose that if it defaults, the bank loses either 30% (a quick sale of the business) or 60% (a long administration), equally likely:

OutcomeProbabilityLossProbability × loss
No default98.870%£0.0m£0
Default, quick sale (LGD 30%)0.565%£22.8m£128,820
Default, long administration (LGD 60%)0.565%£45.6m£257,640
Expected loss£386,460

The average LGD is 45%, so EL is the same £386,460 as before. But in 98.9% of years the loss is zero, and otherwise it is £22.8m or more. The standard deviation of the loss is £3.81m, about 10 times the EL. Expected loss is the cost the bank charges for and provides for; the spread around it is unexpected loss, which needs capital (L3.3.11: Unexpected loss and default correlation).

Expected loss adds up

Because the expectation of a sum is the sum of the expectations, the book's EL is just the sum of the loans' ELs, whatever the correlation between them. Diversification does not reduce expected loss. It only narrows the spread of losses around it.

Risk-based pricing

A loan's interest must pay for four things. Sort each cost first:

Sort itWhich part of the price pays for it?

  • Interest the bank pays depositors and bondholders on the money it lends on

  • The average yearly loss from borrowers of this grade defaulting

  • Relationship managers, credit analysts, systems and monitoring

  • A bad year with losses far above average

With funding at 4.50%, operating cost 0.40% a year and a pre-tax hurdle rate of 15% on capital, the required rate is r=f+PD×LGD+c+K(h−f)r = f + \text{PD}\times\text{LGD} + c + K(h - f). The capital per £, KK, comes from the one-factor model of the next lesson: the extra loss, above EL, in a 1-in-1,000 year, with an asset correlation of 0.20. For now, take it as given. It is a simplified version of the regulatory IRB formula, which also adds a maturity adjustment and lets the asset correlation fall as PD rises (M4.2: Credit risk capital). By grade, for an unsecured loan with LGD 45%:

GradePDEL rateCapital KCapital chargeRequired rate
10.03%0.01%0.48%0.05%4.96%
20.07%0.03%0.93%0.10%5.03%
30.14%0.06%1.57%0.16%5.13%
40.28%0.13%2.59%0.27%5.30%
50.57%0.26%4.23%0.44%5.60%
61.13%0.51%6.59%0.69%6.10%
72.26%1.02%9.95%1.05%6.96%
84.53%2.04%14.36%1.51%8.45%
911.31%5.09%20.84%2.19%12.18%

The price rises slowly through the strong grades and steeply through the weak ones, because both EL and capital grow with PD. This is risk-based pricing. Suppose instead the bank charged every corporate borrower the same rate, the book's EAD-weighted average required rate of 5.82%. The 7 strong borrowers would be overcharged and could refinance elsewhere; the 5 weak ones, Graybeck at a required 17.03% among them, would be glad to stay. That is adverse selection.

The formal version

Expected loss. Let 1D\mathbf 1_D be 1 if the borrower defaults within the year (probability PD) and 0 otherwise. The loss is L=EAD×LGD×1DL = \text{EAD} \times \text{LGD} \times \mathbf 1_D. If EAD and LGD are taken as their averages given default:

E[L]=E[EAD×LGD∣D] Pr⁡(D)+0×Pr⁡(no D)=PD×E[LGD∣D]×EAD(EAD fixed given default)E[∑iLi]=∑iE[Li]=∑iPDi LGDi EADi\begin{aligned} \mathbb E[L] &= \mathbb E[\text{EAD} \times \text{LGD} \mid D]\,\Pr(D) + 0 \times \Pr(\text{no } D) \\ &= \text{PD} \times \mathbb E[\text{LGD} \mid D] \times \text{EAD} \quad \text{(EAD fixed given default)} \\ \mathbb E\Bigl[\sum_i L_i\Bigr] &= \sum_i \mathbb E[L_i] = \sum_i \text{PD}_i\,\text{LGD}_i\,\text{EAD}_i \end{aligned}

The LGD and EAD that belong in the formula are those given default. If recoveries are worse and lines fuller in the years when defaults are high, those averages are higher than in a normal year, which is why banks use downturn LGDs and CCFs measured on defaulted facilities.

Price and return on capital. Per £1 of EAD for a year, with capital KK and the rest funded at ff, the profit after expected loss and costs must give the hurdle return on KK:

r−f(1−K)−EL−c=hKr=f+EL+c+K(h−f)RAROC(r)=r−f(1−K)−EL−cK\begin{aligned} r - f(1-K) - \text{EL} - c &= h K \\ r &= f + \text{EL} + c + K(h - f) \\ \text{RAROC}(r) &= \frac{r - f(1-K) - \text{EL} - c}{K} \end{aligned}

A loan priced at rr earns exactly hh; any lower rate earns less. (Tax, fees and the loan's term are left out to keep the structure clear.)

Try it

Move the grade from 1 to 9 and watch the capital charge overtake the expected loss. Then propose a rate and see whether it meets the hurdle.

Try itPrice a loan

  • Funding4.50%
  • Expected loss (PD × LGD)0.51%
  • Operating cost0.40%
  • Capital charge K × (hurdle − funding)0.69%
  • Required loan rate6.10%
Required rate6.10%
Expected loss a year£101.7k
Capital per £ (K, 99.9%, ρ = 0.20)6.59%
Capital for this loan£1,317k
Return on capital at that rate13.5%
Against the hurdlefalls short

Worked example

Pricing Bramwick Stores plc's facility

From expected loss to a price

Step 1 of 5
  1. 1

    Expected loss

    EL=0.0113×0.45×£76,000,000=£386,460,EL rate=0.509%\text{EL} = 0.0113 \times 0.45 \times \text{£}76{,}000{,}000 = \text{£}386{,}460, \quad \text{EL rate} = 0.509\%

Regulatory lens · looking ahead

Regulatory capital for credit risk is set for unexpected loss; expected loss is meant to be covered by pricing and provisions. Banks on the IRB approach compare their expected loss with their provisions, and a shortfall reduces capital (M4.2: Credit risk capital). Accounting provisions use expected credit losses too, over twelve months or the loan's lifetime under IFRS 9 (L3.3.18: Provisioning basics, M4.7: Accounting for risk). Risk-adjusted performance measures and capital allocation come back in M6.5: Portfolio credit models.

Spark Atoms

Spark AtomWhy banks do not just charge more
Joseph Stiglitz and Andrew Weiss showed in 1981 that raising the interest rate can make a lender worse off: safer borrowers drop out and riskier ones, who may never repay, stay. So banks may ration credit rather than raise its price, an idea that later formed part of the work for which Stiglitz shared the 2001 Nobel prize in economics.
Spark AtomThe loss you almost never have
Bramwick Stores plc's expected loss is £386,460 a year. In 98.9% of years the bank loses nothing on it, and in the rest it loses at least £22.8m. No single year ever looks like the expected loss.

Key takeaways

  • EL = PD × LGD × EAD; the corporate book's is £2.81m on £587m (0.48%), the bank's £22.8m.
  • Expected loss is an average over outcomes; a single loan almost never loses exactly its EL.
  • Portfolio EL is the sum of the loans' ELs: diversification does not reduce it.
  • A risk-based price is funding + EL + costs + capital × (hurdle − funding).
  • RAROC tests a proposed rate; one price for all borrowers invites adverse selection.

Reaction

Check what stuck

Question 1 · objective 1

A loan has PD 2%, LGD 40% and EAD £5m. What is its one-year expected loss, in £?

£

Question 2 · objective 3

Funding costs 4%, operating cost is 0.5%, the EL rate is 0.5%, capital is 6% of EAD and the hurdle rate is 14%. What loan rate is needed, in %?

%

Question 3 · objective 2

The corporate book's expected loss is £2.81m. In most years, what is the actual credit loss on the book?

Question 4 · objective 4

If the bank charged every corporate borrower the same 5.82%, what would happen over time?

Question 5 · objective 2

Why does the bank's expected loss not shrink when it adds more borrowers, while its unexpected loss per £ does?

Glossary

Expected loss (EL)
Expected loss: the average credit loss over a horizon (usually one year), PD × LGD × EAD for one exposure, summed over a portfolio.
EL rate
Expected loss divided by EAD: the share of the exposure lost on average each year, PD × LGD.
Risk-based pricing
Setting each loan's rate from its own risk: funding, expected loss, costs and a return on the capital it needs.
Economic capital
Capital the bank holds against a loan so that losses above the expected level, up to a high quantile, do not make it insolvent.
Hurdle rate
The minimum return shareholders require on the capital they put up.
RAROC
Risk-adjusted return on capital: income after funding, expected loss and costs, divided by the capital the loan needs.
Adverse selection
When one price is charged to borrowers of different risk, the safer ones leave for cheaper lenders and the riskier ones stay.

Deep Dives & Origins

Deep Dives

  • A. Saunders, M. M. Cornett and O. Erhemjamts, Financial Institutions Management: A Risk Management Approach, 11th ed., McGraw-Hill, 2023, the chapters on credit risk (individual loan risk, which covers the return on a loan and RAROC, and loan portfolio risk). Intermediate.
  • C. Bluhm, L. Overbeck and C. Wagner, Introduction to Credit Risk Modeling, 2nd ed., Chapman & Hall/CRC, 2010, chapter 1 on expected and unexpected loss. Advanced.

Origins

  • J. E. Stiglitz and A. Weiss, “Credit Rationing in Markets with Imperfect Information”, American Economic Review 71(3), 1981.
  • Basel Committee on Banking Supervision, An Explanatory Note on the Basel II IRB Risk Weight Functions, BIS, 2005.

Watch this lesson

Video version

Prefer to watch? The same lesson as a 12-minute video will sit here, made from this page's script.