|
Scenario analysis is a practice that has existed as long
as humans have walked this Earth. A mariner contemplates the possibility
that his ship may be afflicted by storms—or pirates, or scurvy. He thinks
about how he would respond to each, and what the consequences might be. He
toys with different scenarios and different responses. He considers the
likelihood of each scenario. Based on that, and what he perceives would be
the outcome of each, he plans his next voyage.
This sort of "what if" analysis has always been a part of
business decision making. Today, what we call
scenario analysis is a formalization
of the process. Formalized scenario analysis is used in
asset-liability
management and
corporate risk management. It originated in the 1970s and 1980s among
banks and insurance companies when
volatility in interest rates emerged as a threat to their balance
sheets. Its primary use remains the analysis of
interest rate risk, but
corporations apply it for a variety of risks.
Scenario analysis starts with scenarios. A simple analysis
might consider three scenarios, say reflecting assumptions that the economy will experience
strong
growth,
moderate
growth, or
a decline.
Typically, more scenarios are used to permit a variety of eventualities to be assessed. A scenario is specified as a set
of "paths" that will be taken by relevant risk factors. Risk factors are
typically interest rates, but they can also be exchange rates,
equity
prices, commodity prices, implied
volatilities—pretty much any risk factors. Scenarios all have the same
horizon and time step—for example, they might specify values of interest
rates at six-month intervals over the next ten years. Exhibit 1
illustrates one such scenario.
|
|
 |
|
time
(months) |
1-month
Libor |
3-month
Libor |
6-month
Libor |
12-month
Libor |
3-year
Swap |
5-year
Swap |
Prime |
COFI |
|
0 |
3.11 |
3.33 |
3.54 |
3.79 |
4.07 |
4.24 |
6.25 |
2.62 |
|
6 |
3.06 |
3.27 |
3.47 |
3.71 |
3.96 |
4.02 |
6.25 |
2.68 |
|
12 |
3.01 |
3.21 |
3.40 |
3.63 |
3.88 |
3.96 |
6.25 |
2.65 |
|
18 |
2.93 |
3.15 |
3.35 |
3.59 |
3.87 |
3.96 |
6.25 |
2.61 |
|
24 |
2.95 |
3.06 |
3.25 |
3.48 |
3.74 |
3.82 |
6.25 |
2.53 |
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
|
114 |
4.85 |
4.87 |
4.59 |
5.02 |
5.22 |
5.38 |
8.00 |
3.98 |
|
120 |
4.88 |
4.91 |
4.64 |
5.08 |
5.23 |
5.51 |
8.00 |
4.04 |
|
One hypothetical scenario that might be
constructed for a bank as part of its scenario analysis. The horizon
is ten years with a six-month time step. |
Such detailed scenarios can accommodate events such as
narrowing spreads or
flattening of the yield curve. Often, however, scenarios are more
simplistic. Five simple scenarios might have interest rates
rise 200
basis points,
rise 100
basis points,
are flat,
decline 100
basis points, or
decline 200
basis points.
Exhibit 2 illustrates one such scenario.
|
|
 |
|
time
(months) |
1-month
Libor |
3-month
Libor |
6-month
Libor |
12-month
Libor |
3-year
Swap |
5-year
Swap |
Prime |
COFI |
|
0 |
3.11 |
3.33 |
3.54 |
3.79 |
4.07 |
4.24 |
6.25 |
2.62 |
|
6 |
2.11 |
2.33 |
2.54 |
2.79 |
3.07 |
3.24 |
5.25 |
2.62 |
|
12 |
2.11 |
2.33 |
2.54 |
2.79 |
3.07 |
3.24 |
5.25 |
1.62 |
|
18 |
2.11 |
2.33 |
2.54 |
2.79 |
3.07 |
3.24 |
5.25 |
1.62 |
|
24 |
2.11 |
2.33 |
2.54 |
2.79 |
3.07 |
3.24 |
5.25 |
1.62 |
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
|
114 |
2.11 |
2.33 |
2.54 |
2.79 |
3.07 |
3.24 |
5.25 |
1.62 |
|
120 |
2.11 |
2.33 |
2.54 |
2.79 |
3.07 |
3.24 |
5.25 |
1.62 |
|
Scenarios can be quite simple, say
assuming that interest rates immediately decline by 100 basis points
and stay at that level for the duration of the scenario. Note that,
in this scenario, a six-month lag is assumed for the response of
COFI to the sudden decline
in rates. |
Once scenarios have been specified, the next step is to
project what will happen under each one. If a bank is performing the
analysis, this might entail projecting what effect evolving interest rates
will have on the level of demand deposits, commercial loans and
residential mortgage
loans. Performed at each time step, this will generate paths for the
evolution of each. The analysis might project for one scenario, say, declining demand
deposits over time, rising and then declining commercial loans, etc.
Management actions may also be projected under each scenario. Under one
scenario, they might be projected to increase borrowing in the Eurodollar
market and later reallocating resources from one business line to another.
An elaborate analysis might project, under each scenario, a bank's cash-flow
statement and balance sheet at each time step.
Output can take many forms. It may focus on
economic
value, but it is more common to focus on earnings, cash flow or other
accounting results. Exhibit 3 illustrates typical output, detailing the
evolution of net interest income over time for each of several scenarios.
|
|
 |
|
|
Interest Rate Scenario |
|
time
(months) |
decline
200 bp |
decline
100 bp |
flat |
rise
100 bp |
rise
200 bp |
|
0 |
27 |
27 |
27 |
27 |
27 |
|
6 |
38 |
32 |
27 |
23 |
19 |
|
12 |
34 |
30 |
28 |
25 |
22 |
|
18 |
31 |
29 |
28 |
26 |
24 |
|
24 |
30 |
28 |
28 |
27 |
26 |
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
.
.
.
|
|
114 |
36 |
37 |
38 |
40 |
43 |
|
120 |
36 |
37 |
38 |
40 |
43 |
|
Sample output of scenario analysis for a
bank. Five scenarios are considered. Each assumes a specific basis
point change in interest rates, which then remains unchanged for the
duration of the scenario. Projected net interest
income (in millions of USD) is presented over time under each. |
Scenario analysis is an important tool of asset-liability
management. Unlike other tools, such as
gap analysis or
duration, it is very
flexible. Gap analysis cannot address
options risk or anticipate the evolution of future business or
management decisions. Scenario analysis can. Duration can, if implemented
correctly, address options risk, but it only addresses interest rate risk
due to parallel shifts in the
spot curve.
Scenario analysis, by comparison, can consider all sorts of movements in
the spot curve—flattening, steepening, becoming inverted, etc. Gap analysis
and duration cannot address basis
risk. Scenario analysis can.
Scenario analysis has several shortcomings. It only
addresses risk due to the specific scenarios considered. Furthermore,
there is always a risk that scenarios don't consider a long enough
horizon. US thrifts learned this lesson the hard way in the 1980s, when
they performed short-term scenario analysis on the
mortgage-backed
securities (MBSs) they were over-paying for. The analyses captured the
attractive short-term earnings from the MBSs, but the horizon didn't
extend long enough to predict the crushing effect of the subsequent
prepayments. Many of the thrifts
failed.
Scenario analysis is highly dependent on assumptions. Assumptions
must be made about, say, the impact of a flattening term structure on a
bank's business lines or the response of management to a decline in demand
deposits. Output of the analysis is only as good as these
assumptions. The more elaborate the scenario analysis, the more
assumptions that must be made.
Finally, output of scenario analysis tends to be
cumbersome—multiple tables summarizing results as opposed to, say, the
single number that is duration.
In its early days, scenario analysis was largely
a manual process—performed with pencil and paper, and maybe an electronic
calculator. This limited the number of scenarios that could be considered.
Gathering balance sheet information and other inputs from various
departments was a time consuming task. Only simple analyses were
performed.
Advances in computer technology have changed this.
Networked computers and centralized databases makes it easy to gather
inputs. Large numbers of scenarios can be considered, and analyses of
those scenarios can be elaborate.
Sometimes, scenario analysis is
performed as a Monte Carlo simulation. A large number of scenarios
is
randomly generated, and results are calculated under each based upon
formulaic assumptions regarding how business will evolve in given
circumstances. Output, rather than being presented on a
scenario-by-scenario basis, is presented in terms of frequency
distributions. Treating these as probability distributions, summary statistics—such as
earnings-at-risk—are
calculated.
Often, you will hear the terms "scenario analysis" and
"simulation analysis" used interchangeably. If there is a distinction
between them, it is that some practitioners reserve
simulation analysis to refer to the
above form of Monte Carlo analysis. That is how I use the term in this
glossary.
|
|
 |
|
asset-liability
management Techniques for protecting a firm's solvency in the context of accrual accounting.
corporate risk
management Practices that serve to optimize risk taking in a context of
book value accounting.
duration
and convexity Factor sensitivities often used in
asset-liability management.
financial
risk management Practices by which a firm optimizes the
manner in which it takes financial risk.
gap analysis
A technique of asset-liability management used to assess interest rate risk or
liquidity risk.
interest rate risk
Risk due to uncertain future interest rates.
interest rate spreads
An
overview.
liquidity
Used in various senses, all relating to availability of, access to, or
convertibility into cash.
model risk
The risk that models are applied to tasks for which they are
inappropriate or are otherwise implemented incorrectly.
Monte Carlo method
The use of statistical sampling to solve quantitative problems.
option-adjusted spread
Yield spread not attributable to embedded options.
reinvestment
risk Risk from uncertainty in the interest rate at which
future cash flows may be invested.
stress
testing A simple form of scenario analysis typically used to
assess market risk.
valuation
Article about book value and market value accounting. |
|
|
|
 |
 |
Ads by Contingency Analysis
|
|
|
 |
|
Uyemura and Van
Deventer (1993)
and van Deventer, Imai and Mesler (2004)
both offer excellent discussions of scenari analysis for banks.
|
|
|
|
|