I have some questions on section 3.3 (The Process): Just to check, are we modelling both assets and liabilities (separately) using stochastic models? And, if so: In the liability projection model, our stochastic variables may be investment returns (to calculate bonuses in the case of WP products; to calculate the cost of any guarantees), mortality improvements (in the case of annuities) and possibly lapses (?). Is that right? What else am I missing? In the asset projection model, our stochastic variables would be the returns on the assets (risk free rates, equity returns etc)? The results of the asset projection model would then be fed into the liability projection model If the products modelled aren't investment related (i.e. not UL or WP), how would the asset model results be used in the liability model? In terms of the ESG, am I right that in saying that this model broadly provides a range of economic scenarios that will then serve as the distributions of the stochastic variables in the asset and liability projection models above? Also, page 8 specifically says: "This model produces....future economic outcomes, which are then used as an input to the stochastic projection model." What's confusing me here is the singular form of "stochastic projection model". Aren't we using the output of the ESG as inout in both A and L models?
As described in that section, most companies have a model which focuses on projection of liabilities, and this will normally include stochastic inputs for investment returns on assets, interest rates and inflation rates. There may also be a stochastic model used for mortality improvements, particularly for annuities as you say, but these assumptions could be done deterministically. Lapse assumptions would tend to be deterministic rather than stochastic/random, but they could be linked dynamically to other assumptions, e.g. investment returns. There may also be a separate model which projects individual actual assets explicitly, and the relevant stochastic variables will be as above i.e. investment returns on various asset types, interest rates (to influence bond yields) and inflation (if index-linked bonds are held), possibly also credit defaults (if corporate bonds are held). [In terms of your reference to "risk-free rates", note that the projections do not have to be for a market-consistent risk-neutral valuation - but these may be used to represent yields on government bonds, say.]
1. Mortality rates can be modelled stochastically as well 2. What is fed into the liability projection model is the ESG file which contains input with economic data. The asset projections results from the asset model are not fed into the liability projections but work in parallel. 3. I think the singular "stochastic projection model" refers to the ALM type model where we have two modules for liabilities and assets combined into 1 single model.
A company won't necessarily have a separate model which generates the value of assets at each time period, it may all be done within the overall liability model. In which case investment earnings on the assets backing the liabilities need to be calculated within the liability model, and this is the case irrespective of the type of business written. The rates of interest used to value liabilities in future projection periods will also depend on projected investment conditions at those future dates.
Yes, that's right, bearing in mind that most companies will project assets and liabilities together within the same model, rather than having a separate asset model.
Hi - as mentioned later in Section 3.3, ESGs are not all calibrated on a market-consistent risk-neutral basis. Stochastic projections could instead be "real world". See also the following thread for some examples of when the latter would be used: https://www.acted.co.uk/forums/index.php?threads/economic-scenario-generators.14329/#post-53552
Hi Lindsay My question is more around why the projection is assets and liabilities shouldn't follow an approach that is more in line with SII (market consistent), given that capital requirements is one of the main drivers of a suitable investment strategy. [Sorry, I should've been more specific]
When calculating the liabilities and associated capital requirements in future projection periods the company would likely want to use a market consistent approach as this is what is required by the regulator. But when projecting forwards the assets, the company may prefer to use a "real world" approach. This will give them useful information on the actual (expected future) amounts of assets and actual "real world" probabilities, bearing in mind that market consistent risk-neutral calibrations are somewhat artificially constructed in that respect. Does that help? Market-consistent calibrations are most useful when determining the present value of something (such as the best estimate liability). Real-world calibrations are most useful when wanting to consider what the future might actually look like.
