Discuss the assessment of operational, liquidity and insurance risks
Considering liquidity and life and nonlife insurance risk
14.8.5 and 14.9.1 - 14.9.3 from Sweeting is excluded
Liquidity risk:
Risk of not having sufficient short-term or cash type assets to fund its short-term obligations
Funding liquidity risk:
Risk of money markets not being able to supply funding to a business when required
e.g. individual not able to get a mortgage when looking to buy a house due to restricted volumes of mortgage lending by banks
Market liquidity risk:
Lack of capacity in the market to handle asset transactions at the time when the deal is required (without material impact on price)
e.g. difficulty to sell their existing house without discounting the price to make it attractive to potential buyers
Typically quantitative techniques are not possible for liquidity risk
Historical data on liquidity crises is limited
Degrees and nature of every org’s exposure to liquidity risk is different (industry data may not be useful)
Common approach to assessing liquidity risk
Cash inflows and outflows projection under a range of scenarios
Cash outflows
Retail banks
Predicting outflow is problematic
\(\because\) Much of a bank’s liabilities will be in the form of deposits from customers who may withdraw with little or no notice
Large mature pension scheme
May have reasonably predictable cash outflows in respect of size and timing of liability payments
General insurance company
May have very unpredictable cash outflows as both the size and frequency of the claims may be unknown
Cash inflows
Includes:
Revenues/income generated by assets
Generally can be modeled with a reasonable degree of confidence
Potential proceeds from the sales of assets
More difficult (e.g. sale could be forced or made during time of depressed asset prices)
Drawing upon sources of liquidity
Maybe difficult to issue new debt or equity due to poor demand from the capital markets (result of poor credit ratings and/or business results)
When modeling sources of liquidity it is important to allow for factors limiting the extent and speed of liquidity transfers within an organization and between distinct entities
Liquidity analysis
Once we modeled the in/outflows and sources of liquidity
\(\hookrightarrow\) Assess liquidity risk by examining scenarios where the cash outflows > available cash at future points in time
Important to allow for appropriate interactions between risks (between liquidity, market and interest rate risk)
Consideration should be made of both short and long term scenarios
Specific scenarios to consider:
Rising interest rates
(e.g. bank may find depositors transfer funds elsewhere in search of higher returns)
Rating downgrade
(e.g. bank may find depositors transfer funds elsewhere in search of more secure institution)
Large operational loss
Resulting in sudden reduction in cash like asset
Large single insurance claim or a large set of claims from associated events
Resulting in a sudden reduction of cash assets
Loss of control over a key distribution channel
Loss of expected revenues
Impaired capital markets
Equity investors or bondholder won’t be able to provide fresh capital
Sudden termination of large reinsurance contract
Insurer exposed to large cash outflows but without expected inflows from the reinsurance contract
Examine the effect on liquidity of an extreme event or significant change in a key assumption
(e.g. collapse of a major customer, inability to refinance a large debt that is due to mature)
The point at which scenario test becomes stress test is subjective
Demographic risk \(\in\) insurance risk
Arises from population changes (e.g. mortality rates) that impact on both customers and employment
Demographic risk can be broken into:
Level risk (u/w risk):
Risk that the particular underlying population’s claims incidence and intensity is not as expected over the immediate future
(e.g. due to shortcomings in the underwriting process)
Reserve risk:
= Volatility + Cat + Trend
Volatility Risk:
Uncertainty w.r.t. the actual future immediate mortality experience
Arises due to only having a finite pool of policies
\(\therefore\) it is not possible to measure precisely the past underlying rates of the underlying population and going forward, the experience of sub-populations will exhibit statistical variations from that of the underlying population
Cat Risk:
Extreme form of volatility risk (e.g. the occurrence of a natural disaster resulting a large number of deaths)
Trend Risk (cycle risk):
Risk of future (longer term) changes in claims incidence and intensity
Way to determine the current underlying level of mortality
Experience rating
Involves examining the number of deaths in a portfolio of lives to determine the inital mortality rate or central mortality rate
Initial mortality rate:
\(q_x = \dfrac{d_x}{l_x}\)
Applies to the number of lives at the start of the period
Central mortality rate:
\(m_x = \dfrac{d_x}{l_x - (d_x/2)} \approx q_x\)
Applies to the average number of lives over the period at each age
Risk rating
Involes modeling the mortality rate of each homogeneous group as a function of the shared characteristics of their members (Model can take the form of GLM)
Subsequently survivor models might be used to develop other mortality functions (e.g. \(\mathring{e}_x\))
Postcode rating: Relies on being able to identify the shared characteristics of a population from its geographic location and the there are sufficient to model the mortality of that group
Information collected from marketing and other surveys might be used
However, all memebers of the group will not necessarily conform to stereotypical risk factors upon which the model is then based
Combined credibility weightings
Combine the experience rating and risk rating methods by using a subjective credibility weighting factor \(Z\) and combining the two mortality rates in proportion \(Z\) and \(1-Z\)
Both rating methods rely on the data being:
Divided into homogeneous groups (e.g. sex, employee type, etc)
Collected over a period which is sufficiently long to generate adequate data, but not so long that the mortality rates could have varied greatly
Credibility vs relevance
Any portfolio has a finite number of lives so there will be some statistical variation in experience
Volatility risk can be modeled probabilistically or stochastically assuming some underlying statistical process (e.g. Bin or Poisson)
The assessment process should reflect the fact that volatility risk varies by age
Poisson MLE Process
Calculate the expected number of deaths at each age
Using the model to be fitted, and set this equal to the mean of a Poisson distribution
Calculate the probability of the observed number of deaths at each age
Based on the poisson distribution derived in step 1
Fitted parameters are obtained by MLE
i.e. Product of the probabilities (for all ages) that were determined in step 2
Risk of a sudden, temporary increase in mortality (e.g. war, pandemic)
Best modeled using scenario analysis
(e.g. scenario where there is a 20% increase in mortality at all ages)
More complex dependencies can be modeled by copulas
(e.g. consider multiple sources of mortality as separate risk factors each with their own probability distribution and then combine with copula)
Cat risk is one-way only, we can ignore the possibility of a sudden temporary reduction in mortality
Other demographic factors (e.g. proportions married, # of children etc) are less likely to vary unexpectedly
There are some other demographic risk (e.g. lapse rates or pension scheme leavers) may be more volatile
Factors that are less likely to vary unexpectedly can be allowed for when modeling liabilities, by using conservative assumptions for unknown independent variables
Other demographic risks which maybe more significant, are also often dependent on other risks (e.g. lapse increase during economic downtown)
Similarly can be broken into the following
Level risk:
Dealt with the combination of experience and risk rating
Reserving risk:
incl. volatility, CAT risk, trend or cycle risk
Difference with demographic risk
Trend (cycle) risk are more likely to correspond with economic cycle
Best assessed using scenario analysis
Non-life risk have a shorter period of exposure
\(\therefore\) Longer term changes in risk factors are less important than a correct assessment of the risk factors themselves
Non-life insurance risk can be divided based on incidence rates:
High frequency (e.g. motor)
Low frequency (e.g. XoL reinsurance)
There is the added complication of severity to be consider
Key distinction between life and non-life risk is that non-life policies may experience more than one claim and move through different states over the lifetime of the policy