Total assets and what is supported by each component (2)
Kreps
total assets = reserves + surplus
reserves support mean of assets & liabilities
surplus supports variability of assets & liabilities
Total capital (C)
Kreps
total capital = mean outcome + risk load
Desirable qualities for an allocatable risk load (3)
Kreps
General form of riskiness leverage models
Kreps
R = integral of f(x) * (x - mu) * L(x) dx
f(x)dx can be called dF-bar for joint probability distributions
where f(x) = joint probability distribution and L(x) = riskiness leverage function for total losses
Risk load (R) and capital (C) across multiple LOB
Kreps
R = sum of R(k)'s C = sum of C(k)'s
where k = individual LOB
Advantage of co-measures
Kreps
they are automatically additive
Disadvantage of co-measures
Kreps
can be challenging to find appropriate forms of the riskiness leverage function L(x)
Conditions for negative risk loads (2) and when it is desirable
(Kreps)
desirable for hedges, occurs when there is a low correlation with total losses
Properties of riskiness leverage models (4)
Kreps
Coherent risk measures
Kreps
satisfy sub-additivity requirement
R(x + y) <= R(x) + R(y)
Super-additivity
Kreps
R(x + y) > R(x) + R(y)
not coherent
Types of riskiness leverage functions, L(x) (7)
Kreps
Risk-neutral form of the riskiness leverage function, L(x)
Kreps
L(x) = c
Situations when a risk-neutral form of L(x) might be appropriate (2)
(Kreps)
Variance form of the riskiness leverage function, L(x)
Kreps
L(x) = (beta / surplus) * (x - mu)
Relevant part of the distribution when using the variance form of L(x)
(Kreps)
entire distribution (just as much risk associated with good & bad outcomes)
Surplus (S) when using the variance form of L(x)
Kreps
S = sqrt(beta * var(x))
Forms of the riskiness leverage function, L(x) where the risk load increases quadratically (2)
(Kreps)
2. semi-variance, SVaR
TVaR form of the riskiness leverage function, L(x)
Kreps
L(x) = theta(x - x(q)) / (1 - q)
where theta is a step function with:
theta(x) = 0 for x <= 0 and
theta(x) = 1 for x > 0
x(q) = value of x so F(x(q)) = q
Relevant part of the distribution when using the TVaR form of L(x)
(Kreps)
only the high end of the distribution is relevant
VaR form of the riskiness leverage function, L(x)
Kreps
L(x) = delta(x - x(q)) / f(x(q))
where delta(x) = 0 everywhere except 0 and integrates to 1
Coherent riskiness leverage function, L(x)
Kreps
TVaR
Capital (C) when using the VaR form of L(x)
Kreps
C = x(q) = VaR
x(q) = value of x so F(x(q)) = q
Relevant part of the distribution when using the VaR form of L(x)
(Kreps)
only the single VaR point is relevant