Unit 18
Unit 18
Unit 18
Structure
18.0 Objectives
1 8.1 Introduction
18.2 What is DFA
18.2.1 ALM in Life insurance.
18.2.2 Objective of DFA
18.2.3 Fixing the Time Period
18.2.4 DFA and Efficient Frontiers
18.2.5 Solvency Testing
18.2.6 Structure of a DFA Model
1 8.3 Stochastically Modelled Variables
18.3.1 Evaluation o f Variables
18.3.2 Short-Term Interest Rate
18.3.3 Term Structure
18.3.4 General Inflation
18.3.5 Change by Line of Business
18.3.6 Stock Returns
18.3.7 Non-Catastrophe Losses
18.3.8 Catastrophes
18.3.9 Underwriting Cycles
18.3.10 Payment Patterns
1 8.4 Corporate Model
18.5 Let Us Sum Up
18.6 Key Words
18.7 Some Usehl Books
18.8 Answer or Hints to Check Your Progress
- -
18.0 OBJECTIVES
After going through this unit, you will be able to
understand the scope of dynamic financial analysis (DFA) in insurance
industry; and
apply the DFA technique to life and non-life insu&nce problems.
18.1 INTRODUCTION
Insurance companies, especially those dealing with non-life insurance, often
witness pricing cycles accompanied by volatile insurance profits and
increasing catastrophe losses. Under such circumstances, shareholders .
is, therefore, necessary to evaluate the economic factors behind such cycles Dynamic Financial Analysis
and to identify their nature and interrelationships.
I i)
ii)
What is DFA,
Stochastically modelled variables and
iii) Corporate model.
The following discussion is based on the work entitled, Introduction to
Dynamic Financial Analysis (see Kaufmann, et a1 (2001)). To facilitate
learner's exposition to formulation of the technical ideas, we have followed
the sequence and notation of the above work borrowing freely. Keeping in
view the introductory requirement of the theme for our course we propose to
use the important portions of the work to help understand the basic
formulations of DFA. However, once we understand the basic concepts, if
will be useful to go through the entire work.
i
: return 1
risk
I
measures that mix together systematic risk(non-diversifiable by shareholders)
and non-systematic risk, which blurs the shareholder value perspective.
Moreover, efficient frontiers might give misleading insight if they are used to
address investment decisions once the concept of systematic risk has been
plugged into the equation.
It can be seen from Figure 18.2 that (i) the stochastic scenario generator
produces realisations of random variables representing the most important
drivers of business results. A realisation of a random variable in the course of
simulation corresponds to fixing a scenario; (ii) data source consists of
company specific inputs (e.g., mean severity of losses per line of business and
per accident year), assumptions regarding model parameters (e.g., long-term
mean rate in a mean reverting interest rate model), and strategic &sumptions
(e.g., investment strategy); and (iii) the output provided by the DFA model,
can then be analysed to improve the strategy, i.e., to make new strategic
assumptions.
Check Your Progress 1
1) What is ALM approach to life insurance? Why this approach cannot be
relied upo? for nonlife insurance?
...................................................................................
....................................................................................
..................................................................................
....................................................................................
....................................................................................
2) What is DFA?
....................................................................................
....................................................................................
......................................................................................
....................................................................................
Non-Life Reserving and 3) What do mean by efficient frontiers in DFA?
Accounts
....................................................................................
4) Explain the concept of solvency testing.
-I - strategic assumptions I
analyze output.
revise strategy
Looking at the list of variables given above it will become apparent that these
belong to asset and liability categories of the financial statement of a
company. While evaluating the variables we may follow such a categorisation.
It is necessary to be able to tackle the problem of evaluating interest rate risk.
Since nonlife insurance companies are often exposed to interest rate behaviour
due to large investments in fixed income assets, it will be useful to assume
that interest rates are strongly correlated with inflation. Consequently, the
inflationary influences of the future changes in claim size and claim frequency
must be included. As there is a correlation between interest rates and stock
returns, investment returns are affected. The model should explicitly recognise
such a feature.
Let us turn to the liability side. Four sources of randomness, viz., non-
catastrophe losses, catastrophe losses, underwriting cycles, and payment
patterns may have to be considered from it. We may simulate catastrophes
separately due to their different statistical behaviour from non-catastrophe
losses. Such a separation helps in having more homogeneous data for non-
catastrophe losses, which makes fitting the data by well-known (right skewed)
distributions easier; in case of reinsurance programme evaluation of external
events with severity of claims have also to be treated separately.
Underwriting Cycles: In case of non-life insurance, underwritings reflect
market and macroeconomic conditions, which affect the business results.
Therefore, it would be useful to include these in a DFA model.
Losses influence insurance firms by their size and piecewise payment over
tiqe. Such features increase the uncertainties of the claims process by
C
introducing the time value of money and future inflation considerations.
Consequently, it is necessary to model claim frequency and severity keeping
L
in view the uncertainties involved in the settlement process. In order fa alIow
for reserving risk we may use stochastic payment patterns as a means of
estimating loss reserves on a gross and on a net basis.
Note that for each of the DFA components considered, there are numerous
alternatives evaluation methods, which may be more appropriate in particular
situations. We have provided a model framework here that serves as a
suggested reference point that can be adjusted or improved on case-by-case
basis.
By setting y = 0.5, we arrive at CIR also known as the squaie root process
b = long-term mean,
a = constant that determines the speed of reversion of the interest rate toward
its long-run mean b,
s = volatility of the interest rate process,
CIR is a mean-reverting process where the short rate stays positive almost
surely.
For simulating the short rate dynamics over the projection period we need to
discretise the mean reverting model (2.2). This is done in (2.3) as
where
where
Non-Life Reserving and
Accounts
The continuously compounded spot rates R,,,. at time t derived from
where
To model the change in loss frequency 6; i.e., the ratio of number of losses
divided by number of written exposure units, the change in loss severity 6:,
and the combination of both of these, 6:, the following formulae can be
used:
where
Remember that a geometric Brownian motion for the stock price can be
projected as a lognormal distribution for 1+ r",
where
Again, it may be noted that stock returns can be modelled differently than we
have done above.
Dynamic Financial Analysis
18.3.7 Non-Catastrophe Losses
The loss numbers N', and mean loss severities x,! = -c:
1
x,!( I ) for
N,!
deviations a l ' * ' , a x . ' of historical data can be used for loss frequencies and
mean loss severities. We also have to take into account inflation and written
exposure units. Because loss frequencies behave more stable than loss
numbers, estimation of loss frequencies instead of relying on estimates of loss
numbers can be resorted to.
We consider the negative binomial distribution with mean m:" and variance
v,'",' where variables m and v represent mean and variance of different factors
N
!, - ~ ~ ( a , pj =) O,l,2,
, (2.17)
.
Next, we take up claim size distribution for high frequency, low severity
losses.
X {. X ;.... independent,
where
x,i =p X . / ~ ; x . ~
m, J
When the number of losses is multiplied with the mean severity, the total
(non-catastrophic) loss amount in respect of a certain line of business:
x:=, N/X/ is obtained.
18.3.8 Catastrophes
~ons'ider losses triggered by catastrophic events like windstorm, flood,
hurricane and earthquake. Such events, are modelled through negative
binomial, Poisson, or binomiai distribution with mean mM and variance vM
We assume that there are no trends in the number of catastrophes:
-
M f NB, Pois, Bin, ...(mean mM, variance vM),
4
where
he total (economic) loss (i.e., not only the part the insurance company in
consideration has to pay) for each catastrophic event i E { I , ..., M,) can be
simulated by taking probability distribution, GPD (generalised Pareto
We have seen GPD in Unit 15 where G5,Pwas considered
distrib~tion)G{,~.
q%,,q,2,
,... i.i.d., (2.2 1)
where
When q' is generated, we split it into pieces reflecting the loss portions of
different lines of business:
Consider the transition probabilities i,i,j e (1, 2, 31, which denote the
probability of changing from state i to state j from one year to the next and
are assumed to be equal for each projection year. That means, the Markov
chain is homogeneous. p, form a matrix T such that
Dynamic Financial Analysis
The set of transition probabilities p,, i, j E (1, 2, 3) above depicts more than
remain constant over time. It is not unreasonable to that some of these are
zero, because there may exist some states that cannot be attained directly from
certain other states. Thus, taking the attainable states, L , the matrix Tcan be
assigned dimension L x L:
.To fix the transition probabilities p, in any of the above mentioned cases,
The model deals with uncertainties of the claim settlement process, i.e., to
pay at the random time. It is not difficult to see that a whole loss portfolio
belonging to a specific line of business exists. We need to consider its
aggregate yearly loss payments in different calendar years (or development
periods). The incremental payment of aggregate losses stemming from one
and the same accident year forms a payment pattern. To form a broad idea on
technique related these payments you may see the portion dealing with loss
triangle of Unit 17. If we consider yearly loss payments pertaining to a
specific accident year t then the ifhdevelopment year refers to calendar year
Non-Life Reserving and
Accounts t + i . In the following we will denote accident years by t, and development
years by t, .
Here we need to distinguish two cases: First, we have to explain the modelling
of outstanding loss payments pertaining to previous accident years followed
by a description to model loss payments in respect of future accident years.
For that purpose we use a chain-ladder procedure.
Since a lognormal distribution usually provides a good fit to historical loss
development factors we may use it for outstanding loss payment.
Reserve Estimation
For each accident year tl we have to estimate the ultimate claim amount in
each development year t through:
where
2) What are the practical issues you may consider for selecting the interest
rate while building a DFA model?.
....................................................................................
where ;,
I .
.I, .
PI= earned premiums,
I, = market value of assets (including realised capital gains in year t),
C', = equity capital,
Z, = losses paid in cqlendar year t,
each line of business, written premiums 4' for renewal class j change in loss
where
Studies have concluded that the written premiums given by equation (3.2)
would come close to be adequate if the realisations of all random variables
referring to projection year t ,w
( 6 ,c m m ) were known in advance and
'
For these variable see (2.1 I), (2.1 0), (2.9), (2.8) and (2.4).
where
r
a'$, oJ = parameters that can be estimated based on historical data. Thus,
where
=t written premiums charged for the last year and still valid just before
the start of the first projection year.
Taking written premiums 4' (k) from (3.3) where k denotes line of business,
the total earned premiums of all lines and renewal classes will be:
where