Marshall Flashcards
1
Q
what are risk margins intended to do?
Marshall et al.
A
provide an extra bit of cushion above the central estimate (i.e. mean)
2
Q
what are less sophisticated approaches to determining CoVs?
Marshall et al.
A
- ignore individual characteristics of valuation portfolio
- based on “model” portfolios
3
Q
what are more sophisticated methods to determining CoVs?
Marshall et al.
A
-combine quantitative analysis (i.e. stochastic modeling) with qualitative analysis of sources of uncertainty not captured quantitatively
4
Q
what is the most common approach to populating the correlation matrix?
(Marshall et al.)
A
- relies heavily on actuarial judgment
- key risks believed to cause correlation between valuation portfolios are categorized as high, medium, or low
- each category is assigned a correlation coefficient value
5
Q
why do quantitative approaches to populating the correlation matrix tend to be the exception?
(Marshall et al.)
A
-most techniques require a significant amount of data, time, and cost to produce credible and intuitive results
6
Q
what two distributions are most commonly selected for risk margin analysis?
(Marshall et al.)
A
- lognormal
- normal (especially at lower probabilities of adequacy where it can generate a higher risk margin than logN)
7
Q
what does bolt-on approach refer to?
Marshall et al.
A
- any approach that does not involve a single unified distribution of the entire distribution of possible future claim costs
- ie: separate analyses to develop a central estimate of insurance liabilities and estimate risk margins
8
Q
why is it impossible to develop a purely quantitative model for representing the range of possible future claim cost outcomes?
(Marshall et al.)
A
- judgment is essential to assessing insurance liabilities and risk margins
- well-fitting models only reflect past sources of uncertainty
9
Q
what is a claims portfolio?
Marshall et al.
A
- aggregate portfolio for which risk margins must be estimated
ex: XYZ Insurance
10
Q
what are valuation classes?
Marshall et al.
A
- portfolios that are considered individually as part of the risk margin analysis
ex: Property vs. Auto
11
Q
what is a claim group?
Marshall et al.
A
a group of claims with common risk characteristics
12
Q
at the highest level, what are two sources of uncertainty?
Marshall et al.
A
systemic risk
independent risk
13
Q
what does systemic risk represent?
Marshall et al.
A
risks that are common across valuation classes or claim groups
14
Q
what are the two sources of systemic risk?
Marshall et al.
A
- internal systemic risk
- external systemic risk
15
Q
what is internal systemic risk?
Marshall et al.
A
risks internal to the insurance liability valuation/modeling process
16
Q
what are 3 examples of internal systemic risk?
Marshall et al.
A
- model structure
- model parameterization
- data accuracy
17
Q
what is internal systemic risk also referred to as?
Marshall et al.
A
model specification risk
18
Q
what is external systemic risk?
Marshall et al.
A
risks external to the insurance liability valuation/modeling process
19
Q
what is an example of external systemic risk?
Marshall et al.
A
-future trends in claim cost outcomes (material costs, labor costs) that may cause actual experience to differ from what is expected based on the current environment and trends
20
Q
what does independent risk represent?
Marshall et al.
A
-risks that occur due to the randomness inherent in the insurance process
21
Q
what are two sources of independent risk?
Marshall et al.
A
- random component of parameter risk
- random component of process risk
22
Q
what does the random component of parameter risk represent?
Marshall et al.
A
-extent to which the randomness associated with the insurance process affects the ability to select appropriate parameters in the valuation models
23
Q
what does the random component of process risk represent?
Marshall et al.
A
pure effect of the randomness associated with the insurance process
24
Q
what types of risk are traditional quantitative modeling techniques best suited for analyzing?
(Marshall et al.)
A
- independent risk
- historical external systemic risk
25
what types of risk do traditional methods fail to fully capture?
(Marshall et al.)
- internal systemic risk
| - future external systemic risk
26
what considerations should go into splitting the claims portfolio into valuation classes?
(Marshall et al.)
- need to balance practical benefits of higher level allocation with insights from granular allocation
- retain as much consistency as possible between central estimate analysis and risk margin analysis
27
why might it be preferable to align valuation classes with the valuation portfolios analyzed for central estimate purposes?
(Marshall et al.)
- allows risk margin analysis to be conducted in the context of the central estimate analysis
- allows quantitative and qualitative analysis to be aligned with central estimate valuation
28
what are two reasons why it might not be possible to conduct quantitative analysis at the same granularity used for central estimate purposes?
(Marshall et al.)
- credibility concerns
| - implementation costs
29
when it's not possible to conduct quantitative analysis at the same granularity as central estimate purposes, what can we do?
(Marshall et al.)
-conduct quantitative analysis on aggregated valuation classes and allocate down to more granular classes
30
what is an example of a valuation class that typically requires further allocation into claim groups, and why?
(Marshall et al.)
home portfolio:
- dev patterns differ between cat and non-cat claims
- liability claims tend to behave differently from other home claims
31
what are three reasons why stochastic modeling techniques do NOT capture all sources of uncertainty?
(Marshall et al.)
- good models fit past data well -> tends to remove past episodes of external risk, leaving only random sources of uncertainty behind
- if past external risk is not removed completely - must consider if we expect these to continue into the future
- models normally don't catch uncertainty arising from internal systemic risk
32
what are five modeling techniques?
| Marshall et al.
- Mack
- Bootstrapping
- Stochastic CL
- GLM
- Bayesian
33
what is a useful supplement for any analysis of independent risk?
(Marshall et al.)
internal and external benchmarking
34
what are three sources of internal systemic risk?
| Marshall et al.
- specification error
- parameter selection error
- data error
35
what is specification error?
| Marshall et al.
error that arises because the model cannot perfectly model the insurance process
36
what is parameter selection error?
| Marshall et al.
model cannot adequately measure all predictors of future claim costs or trends in these predictors
37
what is the difference between the parameter risk component of independent risk and the parameter selection error component of internal systemic risk?
(Marshall et al.)
- parameter selection error relates to actual selection of model parameters
- parameter risk relates to the proper parameterization of the selected predictors
38
what is data error?
| Marshall et al.
- error that arises due to lack of credible data
- can also refer to an inadequate knowledge of the portfolio being analyzed, including pricing, underwriting, and claims management processes
39
describe an approach to analyzing internal systemic risk
| Marshall et al.
- balanced scorecard is developed to objectively assess the model specification against a set of criteria
- for each source of internal systemic risk, risk indicators are developed and scored against the criteria
- scores are aggregated for each valuation class and mapped to a quantitative measure (CoV) of the variation arising from internal systemic risk
40
what subjective decisions must be made within the balanced scorecard approach to analyzing internal systemic risk?
(Marshall et al.)
- risk indicators
- measurement and scoring criteria
- weight given to each risk indicator
- CoVs that map to each score
41
what are risk categories within external systemic risk? (7)
| Marshall et al.
- economic and social risks
- legislative, political, and claim inflation risks
- claim management process change risk
- expense risk
- event risk
- latent claim risk
- recovery risk
42
what are economic and social risks?
| Marshall et al.
uncertainty associated with inflation, social trends, etc.
43
what are legislative, political, and claim inflation risks?
| Marshall et al.
uncertainty associated with changes in the political landscape, shifts/trends in the level of claim settlements, etc.
44
what is claim management process change risk?
| Marshall et al.
uncertainty associated with changes in claim reporting, payment, estimation, etc.
45
what is expense risk?
| Marshall et al.
uncertainty associated with the cost of managing the run-off of the insurance liabilities or the cost of maintaining the unexpired risk until the date of loss
46
what is event risk?
| Marshall et al.
uncertainty associated with claim costs arising from events (e.g. cats), either natural or man-made
47
what is latent claim risk?
| Marshall et al.
uncertainty associated with claims that arise from sources that are not currently covered (e.g. asbestos)
48
what is recovery risk?
| Marshall et al.
uncertainty associated with recoveries, either reinsurance or non-reinsurance
49
how do the valuation actuary and business unit management interact in the risk margin analysis?
(Marshall et al.)
- normally used to discuss all aspects of the portfolio management process
- discuss potential external systemic risks that may impact portfolio to inform both central estimate valuation AND identification and quantification of external systemic risk
50
what risk category accounts for most of the uncertainty in property classes?
(Marshall et al.)
event risk dominates volatility of premium liabilities
51
what risk category accounts for most of the uncertainty in long-tailed classes?
(Marshall et al.)
legislative, political and claim inflation dominate the volatility of both outstanding claim liabilities and premium liabilities
52
what are disadvantages of quantitative methods used to assess correlation effects?
(Marshall et al.)
- very complex
- require large amounts of data
- heavily influenced by past correlations (problem for external risk that future may differ from past)
- difficult to separate past correlation effects between independent risk and systemic risk OR to identify pure effect of each past systemic risk
- internal systemic risk cannot be modeled using standard correlation modeling techniques
53
how can we consider correlation effects for independent risk?
(Marshall et al.)
assume ind't risk to be uncorrelated with any other source of uncertainty, either within a particular valuation class or between classes
54
how can we consider correlation effects for internal systemic risk?
(Marshall et al.)
- assumed to be uncorrelated with ind't risk and with each source of external systemic risk
- tends to be correlated between valuation classes AND between OCLs and PLs
55
how can we consider correlation effects for external systemic risk?
(Marshall et al.)
- reasonable to assume each risk category is uncorrelated with ind't risk and internal systemic risk
- correlation may exist between risks that belong to similar external risk categories -> must aggregate to broader, uncorrelated categories
56
what are the five correlation bands names and thresholds?
| Marshall et al.
- Nil = 0%
- Low = 25%
- Medium = 50%
- High = 75%
- Full = 100%
57
how does the normal distribution represent risk margins for insurance liabilities?
(Marshall et al.)
-produces higher risk margin at lower probabilities of adequacy, irrespective of the size of the consolidated CoV
58
how does the lognormal distribution represent risk margins for insurance liabilities?
(Marshall et al.)
- produces higher risk margin at higher probabilities of adequacy as long as consolidated CoV is not too high
- extremely high CoVs can cause risk margin to reduce as a percentage of the CoV
59
how can actuaries sensitivity-test the final risk margins?
| Marshall et al.
vary the key assumptions and measure the impacts
60
what key assumptions should be sensitivity-tested?
| Marshall et al.
-CoVs and correlation coefficients assumed for independent risk, internal systemic risk, and each external systemic risk category
61
what happens to our central estimate after including the risk margin?
(Marshall et al.)
our provision for a particular insurance liability will be larger than its central estimate
62
how would we scenario test changing key assumptions to produce a central estimate at the level of our risk-margin inclusive provision for insurance liabilities?
(Marshall et al.)
- valuation outcomes (e.g. claim frequencies, claim severities, loss ratios, etc.) can be compared against the central estimate for various assumption scenarios
- can consider the risk margin inclusive valuation outcomes as well as changes required to produce these outcomes in the context of emerging experience
63
what internal checks should proposed CoVs be subjected to?
| Marshall et al.
- for each source of uncertainty, CoVs should be compared between valuation classes for OCLs, PLs and total insurance liabilities
- comparison should be made between OCLs and PLs within classes as well
64
for indepedent risk, what two main dimensions should internal benchmarking be done for?
(Marshall et al.)
- portfolio size (larger portfolio -> lower volatility arising from random effects)
- length of claim run-off (longer run-off, more time for random effects to impact)
65
how should OCL CoVs vary by short vs. long-tail portfolios?
| Marshall et al.
- OCL CoVs for short-tail portfolios should be lower than similarly-sized long-tailed portfolios
- OCL CoVs for short-tailed portfolios should be substantially lower than smaller long-tailed portfolios
66
how should CoVs vary by PL vs. OCL for long-tailed lines?
| Marshall et al.
- PL CoVs for long-tailed portfolios should be higher than OCL CoVs for same portfolios due to law of large numbers
- PLs for long-tailed lines will be smaller, resulting in more volatility
- larger difference for small portfolios (with higher ind't risk components) than for large portfolios (smaller ind't risk components)
67
how should CoVs vary for PL vs. OCL for short-tail portfolios?
(Marshall et al.)
- PL CoVs for short-tailed should be lower than OCL CoVs for same portfolios due to law of large numbers
- PLs for short-tailed lined will be larger since most of the OCLs will have already closed -> less volatility for PLs
68
for internal systemic risk, how can the CoVs be compared?
| Marshall et al.
compare in the context of each valuation class:
- classes with homogeneous claim groups should have similar CoVs
- long-tailed portfolios should have higher CoVs than short-tailed portfolios (more complicated)
69
for external systemic risk, how should CoVs vary by long vs. short-tail, and what are two exceptions?
(Marshall et al.)
- long-tailed portfolios should have higher CoVs than short-tailed portfolios
- exceptions: event risk and liability risk for home classes
70
when is external benchmarking most beneficial?
| Marshall et al.
when little information is available for analytics
71
how should (and shouldn't) external benchmarking be used?
| Marshall et al.
- primarily used as a reasonability check
| - should NOT form basis of risk margin analysis
72
what must be considered when using external benchmarks?
| Marshall et al.
-differences between benchmark portfolios and claims portfolios being analyzed
73
what does a hindsight analysis do?
| Marshall et al.
compares past estimates of OCL's and PL's against the latest view of the equivalent liabilities
74
how does a hindsight analysis treat any movement/variation in past estimates of OCL/PL's vs. latest view of liabilities?
(Marshall et al.)
-convert to a coefficient of variation reflecting actual past volatility -> this volatility is combination of ind't risk, internal systemic risk, and external systemic risk
75
when is hindsight analysis more useful?
| Marshall et al.
for short-tailed portfolios where serial correlation between consecutive valuations is less significant
76
why should hindsight analyses be interpreted carefully?
| Marshall et al.
- models may have improved since previous valuations
- future external sources of systemic risk may be significantly different from past episodes (which could make it appear as though estimates were off)
77
what does mechanical hindsight analysis do?
| Marshall et al.
applies a mechanical approach to estimating the OCL's and PL's by systematically removing the most recent claims experience
78
what types of risk can mechanical hindsight analysis be used to analyze?
(Marshall et al.)
- independent risk
- internal systemic risk
- all past sources of uncertainty
79
by focusing on what, can mechanical hindsight analysis be used to analyze independent risk?
(Marshall et al.)
by focusing on periods with stable experience
80
by focusing on what, can we use mechanical hindsight analysis to analyze internal systemic risk?
(Marshall et al.)
by applying the technique using a number of actuarial methods and observing the differences in volatility
81
by focusing on what, can we use mechanical hindsight analysis to analyze all past sources of uncertainty?
(Marshall et al.)
by applying the approach across all past periods
82
what do current documentation practices of risk margin analysis range from and to?
(Marshall et al.)
from a thorough discussion of all assumptions underlying the risk margin analysis to no discussion at all
83
at least how often should a full application of the risk margin analysis framework be applied?
(Marshall et al.)
every three years
84
when central estimate valuations are conducted, what key assumptions of risk margin analysis should be reviewed?
(Marshall et al.)
- any emerging trends
- emerging systemic risks
- changes to valuation methodologies
85
when a model is fit to past systemic episodes and trends, what should residual volatility exclusively reflect?
(Marshall et al.)
random effects ONLY (i.e. the independent risk)
86
what factors does the choice of statistical method depend on?
(Marshall et al.)
- valuation class being assessed
- materiality of ind't risk for that class relative to overall claims portfolio risk margin
- cost and effort associated with applying the method
87
why is a GLM a good choice for analyzing ind't risk for OCLs?
(Marshall et al.)
-GLM is used for reserving purposes to identify the key factors that have contributed to past claim costs
88
what statistical methods can be used to support the analysis of premium liabilities' independent risk?
(Marshall et al.)
- bootstrapping
- GLM
- Bayesian
89
what is a simpler approach than statistical methods for assessing independent risk for premium liabilities?
(Marshall et al.)
- assume: valuation class where central estimate of claim cost component of PL is calculated by combining projected claim F and average claim size
- for both claim F and avg. claim size, past systemic episodes (ex: seasonality, inflation) can be removed
- CoV can be derived for residual volatility
90
why is an objective assessment of the model infrastructure and its ability to reflect/predict the insurance process important?
(Marshall et al.)
models represent a simplified view of the insurance process
91
what approach is used to assess internal systemic risk?
| Marshall et al.
balanced scorecard approach
92
how does the balanced scorecard approach work for internal systemic risk?
(Marshall et al.)
- for specification, parameter, data risk components -> qualitative assessment of modeling infrastructure is completed
- risk indicators are weighted together to develop an overall score for each valuation class
- scores are mapped to a CoV
93
what is the qualitative assessment of modeling infrastructure used in the balanced scorecard approach to assessing internal systemic risk?
(Marshall et al.)
- identifying risk indicators for each risk component
| - score on scale of 1 to 5 (5 represents best practice)
94
what does the risk indicator weight reflect in the balanced scorecard approach?
(Marshall et al.)
-reflects each risk indicator's importance to the modeling infrastructure
95
what additional risk indicators might be required for PL's?
| Marshall et al.
-whether OCL's are used as an input
-whether credible portfolio level pricing analysis is used as an input
(both would drive risk score up and CoV down)
96
what is an example of a risk indicator that may be less relevant for short-tail portfolios in PL assessment?
(Marshall et al.)
-large variance in OCLs may only affect the most recent accident periods and have a small impact on projected claim frequencies and avg. claim sizes
97
how can actuaries get an idea for the amount of potential variability (range of CoV) associated with a particular modeling approach?
(Marshall et al.)
analyzing both the scoring approach AND past model performance, combined with judgment
98
what is the minimum CoV suggested, and why?
| Marshall et al.
5% because no model can perfectly replicate the insurance process
99
why aren't CoV scales linear?
| Marshall et al.
marginal improvement in outcomes between fair and good modeling approaches is less than the marginal improvement between poor and fair modeling approaches
100
for internal systemic risk, why are CoVs for long-tailed portfolios higher than those for short-tail portfolios?
(Marshall et al.)
it's more difficult to model insurance process underlying long-tailed portfolios
-key predictors are less stable for long-tailed portfolios
101
how can communication with business experts help with evaluating external systemic risk?
(Marshall et al.)
- portfolio and claim management can provide insights into emerging trends and future sources of external systemic risk
- tailored to important topics for both central estimate and risk margin purposes
102
what six categories should discussions with business experts about external systemic risk consider?
(Marshall et al.)
- underwriting and risk selection
- pricing
- claims management
- expense management
- emerging portfolio trends
- environment in which the portfolio operates
103
what is an example of a risk category that is more relevant for higher probabilities of adequacy?
(Marshall et al.)
-probability of an emerging latent risk is extremely low -> would not have a material impact on a 75th percentile risk margin
104
how can one identify key risks?
| Marshall et al.
- scoring system can be used to rank individual risk categories for each valuation class in order of importance
- risk categories that show up as important for a number of classes are key risks
105
what are four examples of economic and social risks?
| Marshall et al.
- levels of standard inflation (AWE and CPI)
- general economic conditions
- fuel prices
- driving patterns
106
what general economic conditions constitute as economic and social risks? (3)
(Marshall et al.)
- unemployment rates
- GDP growth
- interest rates
107
what is an example of an economic/social risk that is less relevant for auto OCLs than auto PLs?
(Marshall et al.)
uncertainty around driving conditions
108
how does uncertainty around AWE (average weekly earnings) impact some classes more than others?
(Marshall et al.)
tends to be greater for long-tailed classes
109
for what type of claims are legislative, political, and claims inflation risks more material, and how are the risks treated?
(Marshall et al.)
- long-tailed classes
| - risks are aggregated and referred to as superimposed inflation
110
for long-tailed classes, what are four examples of risks within the legislative, political, and claims inflation risk category?
(Marshall et al.)
- precedent setting in courts
- changes in medical costs
- changes in legal costs
- systemic shifts in large claim frequency or severity
111
for short-tailed risk, what is an example of a risk within the legislative, political, and claims inflation risk category?
(Marshall et al.)
risk that claim inflation will increase at a level different from that used for central estimate purposes
112
how can discussions with claim managers help to analyze claim management process change risk?
(Marshall et al.)
- help identify current or future changes to process
- identify past systemic episodes impacted by claim management process & process changes contributing to those episodes
- consider reporting patterns, payment patterns, reopening rates, etc.
113
is claim management process risk typically more relevant for OCLs or PLs, and why?
(Marshall et al.)
- OCL
- for example, OCLs for short-tailed classes are materially impacted because this risk impacts the emergence pattern for credible claim estimates
114
why should an actuary spend time with product and claim management to consider expense risk?
(Marshall et al.)
- to understand key drivers of policy maintenance and claim handling expenses
- allows actuary to identify key sources of variation relative to central estimate assumptions
- help identify past systemic episodes
115
what does event risk relate to?
| Marshall et al.
single events that cause a large number of claims
116
for event risk for OCLs, what should the focus of discussions with claim management be on?
(Marshall et al.)
- focus on material outstanding events
- set expectations on final claim cost outcomes
- range of dev. patterns on past events may influence view on uncertainty
117
for event risk for PLs, what can be used to quantify event risk?
(Marshall et al.)
- past experience for event claims
- output from proprietary cat models
- output from models for perils not covered by cat models
118
when analyzing past experience to help quantify PL event risk, what are four examples of parameters we should allow for changes in?
(Marshall et al.)
- portfolio size
- geographical spread
- inflation
- reinsurance arrangements
119
for what classes can latent claim risk be material?
| Marshall et al.
workers compensation and liability classes
120
why is latent claim risk difficult to quantify?
| Marshall et al.
it has extremely low frequency and extremely high severity
121
what is an example of where recovery risk may be significant?
(Marshall et al.)
subrogation for auto insurance
122
when assessing recovery risk, what is the goal?
| Marshall et al.
identify systemic events that may lead to different recovery outcomes than those used for central estimate purposes
123
for non-reinsurance recoveries, how can a range of reasonable outcomes about recovery risk be developed?
(Marshall et al.)
by analyzing past experience and talking to claim management about current/future trends in recovery levels
124
for reinsurance recoveries, how can a range of reasonable outcomes about recovery risk be formed?
(Marshall et al.)
- determining how shifts in reinsurance market conditions (such as catastrophes) impact the ability of the reinsurer to pay
- discussions with reinsurance management to help identify possible scenarios and quantify impact
