Вы находитесь на странице: 1из 4


P&L Forecasting - the new horizon

of stress testing, and beyond
Aleksander Petrov, Yuri Mushkin, Federico Fumagalli


Any use of this material without specific permission of McKinsey & Company is strictly prohibited

Overcoming the limitations of existing P&L forecasting approaches

For banks, models that project profit and loss are emerging as both a regulatory priority and a
vital managerial tool for budgeting and forecasting. Banks have already developed advanced
models for projecting credit losses, prompted by regulatory pressure and internal risk
management over the past decade. But they have not invested much into modelling capabilities
for P&L, particularly in Europe. They need to do so now.

The current state of affairs is inadequate

Profit and loss models, usually known as PPNR (pre-provision net revenue) in the US and PBIL
(profit before interest and liabilities) in the UK, include three main economic elements: net
interest income (NII), non-interest revenue, and operating expenses. For NII calculations the
first port of call is typically ALM (asset liability management) infrastructure which contains the
relevant data, such as loan amortisation schedules, contractual interest rates and pre-payment
assumptions. Banks have started to evolve their ALM calculations under normal macroeconomic
conditions (e.g., for budgeting) and to understand the impact of stressed conditions (e.g., for
regulatory exams), but these systems often lack the capability to predict new originations or
stressed repayments and spreads. For non-interest revenues, such as trading and investment
banking fees, there is no pre-existing framework at all. Nor is there one for modelling operating
expenses such as compensation, marketing and service costs. The tendency is therefore to
compensate through expert judgement in establishing P&L forecasts. Judgemental approaches
have their legitimate place for revenues strongly controlled by managerial actions, yet they lack
precision for income which is sensitive to macro economy.


Robust P&L forecasting requires covering Net Interest Income, NonInterest Revenues and Non-Interest Expenses
P&L components


Asset balances

Current state

Target forecasting approach

For budgeting and planning: managerial

estimate by Lines of Business
For Stress Testing: in case of static B/S,
no estimate performed

Existing balances and new originations projected

with prepayment assumptions estimated with
statistical macro-economic models

Existing balances are projected based

on simplified (judgement based)
approach based on regulatory rules
communicated with macro scenario

Existing balances and new originations rates

estimated with earning spread assumptions
based on statistical macro-economic models

For budgeting and planning: managerial

estimate by Lines of Business
For Stress Testing: in case of static B/S,
no estimate needed

Separate models for retail and wholesale

segments, and type of deposit (term vs. nonmaturity)
Wholesale funding forecasted with rules to close
balance sheet

Simplified projections based on

regulatory rules communicated with
macro scenario

Term of deposits and spreads to scenario riskfree rate are projected based on statistical
macro-economic models.

Simplified projection at aggregated P&L

item level, often based on expert
judgement based or simple regression of
the aggregated item vs. macro variables

Separate models for each fee or revenue type

(e.g., late fees, transaction fee income, deal
volumes), split between: number of events and
average fee

Often rules-based (according to

regulatory limitations), following
centrally-defined principles (e.g.,
percentage of total assets)

Different models for each P&L item, separating

(i) compensation-based costs estimated on FTE
plus production-based volume incentives, (ii)
fixed line items (e.g., occupancy, external
services) estimated top-down, and (iii) other
expenses based on estimated levels of activities
(e.g., numbers of transactions, customers)

Earning asset yields

Net interest
Deposit / funding

Deposit / funding costs

Number of fee generating



Average fee per event


Compensation expenses
Non-comp expenses
External services

McKinsey & Company

In response to these inadequacies, banks have attempted to run P&L regressions on macroeconomic factors, yet often such ad-hoc models are not statistically robust, because they use
short data series and have limited granularity. For example, they do not distinguish between
volumes and margins and show insufficient segmentation between product classes.

External and internal forces are pushing banks to bolster their P&L
modelling capabilities
Profit and loss modelling is coming under growing scrutiny from various external stakeholders,
including regulators. System-wide stress tests in the UK and the rest of Europe are expanding
their focus and paying closer attention to income forecasts, therefore banks have to improve
their methodologies. Yet the benefits of strong P&L projection capabilities go beyond a
satisfactory score on exams. Well designed and justified P&L projections compensate high
proportion of credit losses (especially in the case of a dynamic balance sheet assumption) and
therefore can lead to significant capital savings.
From an internal budgeting perspective, such methodologies help reduce the role of expert
judgement and bring a healthy challenge to a budgeting process that has historically relied
heavily on market-share assumptions. A strong P&L model suite that has been properly
documented and validated can lead to better capital allocation decisions and overall steering.
That in turn can increase the confidence of stakeholders (i.e., the Board, regulators, and

investors) in the capital management process. Streamlining their modelling capabilities would
also allow banks to produce estimates for internal budget, EBA, or PRA stress tests using a
single execution machine.
For the lines of business (LOB), a new P&L analytical suite can add rigour to the evaluation of
key business decisions. If properly rolled out and integrated in the core businesses, these tools
could bring a common methodology to the analysis of new markets and product opportunities.
Augmented by business intuition, they can serve as a bridge to engage the LOB in the stresstesting and budgeting process.

Building better P&L forecasting models will require a significant

A robust P&L framework requires infrastructure and analytical capabilities that often do not
exist in European banks. Leading practices involve disaggregation of revenue drivers between
prices and volumes and require choosing the right level of granularity to model the P&L. In
addition, banks need to ensure that assumptions in P&L forecasts are consistent between
products and across other forecasting models in the bank, such as credit loss and RWA
Banks therefore need to design their modelling frameworks carefully, by focusing on
consistency, data quality and controls, estimation methods, segmentation of the product lines,
and business engagement. For meaningful results, institutions need data covering a full interest
rate cycle and multiple business cycles, typically going back up to 10 years. A reasonable
ambition for banks would be to cover more than 80 per cent of their balance sheet through
statistical models, with the rest addressed by non-statistical models or expert judgement. To
develop sound econometric projections, modellers typically need to consider differencing timeseries methods and rigorous diagnostic tests like stationary, autocorrelation and out-of-sample
testing. It is also crucial for banks to define where non-statistical models are appropriate, such
as in businesses with low macroeconomic dependence, few internal or external data sources, or
small scale. In these cases, non-modelled approaches that draw on industry-level data or employ
historical stress analysis may be appropriate.



Higher granularity of income and cost modelling will be required to

produce robust estimates for internal and external stakeholders
Projection approach


2014 Stress Tests1

2016 Stress Tests1





First lien fixed mortgages


First lien floating mortgages


Buy to let mortgages

Second lien mortgages


Auto loans

Student loans


Credit cards

Other consumer loans

C&I term loans

Commercial Real Estate loans


Other loans and leases






Trading assets




Intangibles (incl. goodwill)

Other assets

Other assets



Demand deposits


Other deposits

Other liabilities (non-deposits)



Total assets


Other liabilities

Total liabilities

Historical average

Primarily judgement

Primarily statistical model

McKinsey & Company |

1 Existing ALM platform maybe used to calculate stressed NII by adjusting through judgement ( or statistical models ) the build-in assumptions such as prepayments, new origination, spreads etc.

Banks will also need to invest in expertise and processes. P&L modelling is a new discipline that
requires staff with different skill-sets compared with existing staff focused on risk and credit loss
estimates; developers need training in advanced econometric techniques and validation teams
require similar expertise for effective challenge. In addition, business divisions, risk, and finance
must coordinate closely with each other to validate statistical outcomes and check that results
incorporate independent variables that make plausible business sense.
Given these challenges, statistical P&L forecasting will not be a sprint. Banks should invest
significantly to develop a centralised plan for the entire model landscape, ensuring sufficiently
rich data sources and improved methodology and expert capabilities. The result will be tools and
processes that can produce estimates useful for both internal budgeting and regulatory stress
tests and that rely on a single execution machine.
If banks want to anticipate (and possibly shape) the regulatory agenda as well as incorporate
enhanced forecasting capabilities into their budgeting and planning processes they will
probably be well advised to act now. Indeed, several large European institutions have already
started. The way ahead towards robust P&L forecasting methodologies and tools is arduous, but
the substantial investment required will be time and money well spent.
Aleksander Petrov is a Partner and Yuri Mushkin is a Senior Expert, both in the London office,
Federico Fumagalli is a Consultant in the Milan office.