Lecture 1 of 4-part series. Spring School on Risk Management, Insurance and Finance European University at St. Petersburg, Russia.

Transcription

1 Principles and Lecture 1 of 4-part series Spring School on Risk, Insurance and Finance European University at St. Petersburg, Russia 2-4 April 2012 s University of Connecticut, USA page 1

2 s Outline s 5 page 2

3 s The allocation of capital is the term typically referring to the subdivision of a company s aggregate capital across its various constituents: lines of business its subsidiaries product types within lines of business territories, e.g. distribution channels types of risks: e.g. market, credit, pricing/underwriting, operational Company is typically involved in the financial services industry e.g. banks, insurance companies. A very important component of : identifying, measuring, pricing and controlling risks page 3

4 s The purpose of capital Knowing how much capital you need for your overall business is a key aspect of ERM. is the amount set aside, usually in excess of assets backing all liabilities, so that the firm: could withstand and absorb unexpected losses from all risks it is facing; would remain solvent with high probability; and is able to cover obligations to its customers as promised. Economic capital vs regulatory capital: Economic capital is usually calculated based on true market value (or economic) terms. Regulatory capital is usually calculated on the basis of prescribed guidelines by regulatory authorities. page 4

5 for capital computations We will assume that how we measure capital is known and given. s Requires understanding all aspects of risks (or losses) the company is facing. modeling the distribution of losses understanding expectation and variation of these losses understanding possible inter-dependencies of these losses Some well known risk measures may be used: Value-at-Risk or percentile or VaR Conditional tail expectation or Tail-VaR If X is the random loss, then ρ[x] is some risk measure. page 5

6 s - quick review A risk measure is a mapping ρ from a set Γ of real-valued random variables defined on (Ω, F, P) to R: ρ : Γ R : X Γ ρ[x]. Let X, X 1, X 2 Γ. Some well known properties that risk measures may or may not satisfy: Law invariance: If P[X 1 x] = P[X 2 x] for all x R, ρ[x 1 ] = ρ[x 2 ]. Monotonicity: X 1 X 2 implies ρ[x 1 ] ρ[x 2 ]. Positive homogeneity: For any a > 0, ρ[ax] = aρ[x]. Translation invariance: For b R, ρ[x + b] = ρ[x] + b. Subadditivity: ρ[x 1 + X 2 ] ρ[x 1 ] + ρ[x 2 ]. page 6

7 s Some important concepts Conditional Tail Expectation (CTE): (sometimes called TailVaR) CTE p [X] = E [ X X > F 1 X (p)], p (0, 1). In general, not subadditive, but it is so for continuous random variables. Comonotonic sum: S c = n (0, 1). The Fréchet bounds: where i=1 F 1 X i (U) where U is uniform on L F (u 1,..., u n ) C(u 1,..., u n ) U F (u 1,..., u n ), Fréchet lower bound: L F = max ( n i=1 u i (n 1), 0 ), and Fréchet upper bound: U F = min(u 1,..., u n ). page 7

8 Some special distributions Distribution density f X (x) Quantile Q p [X] CTE p [X] s Normal 1 2πσ e 1 2 Gamma ( x µ σ β α Γ(α) x α 1 e βx ) 2 µ + Φ 1 (p)σ µ + φ(φ 1 (p)) p σ no explicit form ) F X (x p;α+1,β F X (x p;α,β) α β ( 1 Lognormal 2πσx e 1 log(x) µ ) 2 ( ) 2 σ e µ+φ 1 (p)σ e µ+σ2 /2 Φ σ Φ 1 (p) 1 p Pareto ab a (x+b) a+1 b [ (1 p) 1/a 1 ] a a 1 Q p[x] + b a 1 page 8

9 s Illustrative case study For purposes of showing illustrations, we will consider an insurance company with five lines of business: auto insurance - property damage auto insurance - liability household or homeowners insurance professional liability other lines of business We will measure loss on a per premium basis and denote the random variable by S for the entire company and X i for the i-th line of business, i = 1, 2, 3, 4, 5. are described in the subsequent slides. page 9

10 s Line of Loss Premium business distribution share Parameters Mean Variance Auto (PD) Gamma 30% α = 360, β = Auto (liab) Lognormal 20% µ = 0.362, σ = Household Gamma 15% α = 56.25, β = Prof liab Pareto 15% a = 6.92, b = Other Lognormal 20% µ = 0.784, σ = auto (PD) auto (liab) household prof liab other auto (PD) 1.00 auto (liab) household prof liab other correlation between lines of business page 10

11 s Graph of densities - by lines of business density fxi(x) auto (PD) auto (liab) household prof liab other x page 11

12 Distribution of the aggregate loss Distribution of aggregate loss s Density loss per premium mean SD median min max VaR 0.95 [S] CTE 0.95 [S] page 12

13 Stand-alone capitals s Line of business VaR 0.95 [X i ] CTE 0.95 [X i ] Auto (PD) Auto (liab) Household Prof liab Other page 13

14 Insurance company with multiple lines of business auto property damage K 1 s insurance company K auto liability K 2 household K 3 professional liability K 4 other lines K 5 page 14

15 Proportional capital allocation s Many well-known allocation formulas fall into a class of proportional allocations. Members of this class are obtained by first choosing a risk measure ρ and then attributing the capital K i = γ i ρ [X i ] to each business unit i, i = 1,..., n. The factor γ i is chosen such that the full allocation requirement is satisfied. This gives rise to the proportional allocation principle: K i = K n j=1 ρ[x j] ρ[x i], i = 1,..., n. page 15

16 s Covariance capital allocation Because of its popularity, we also consider here for purposes of early illustrations this allocation using covariance. The covariance is based on the fact that when we have an aggregate loss that is a weighted sum such as then it is easy to see that S = n c j X j, j=1 [ n ] Var[S] = Cov c j X j, S = j=1 n c j Cov[X j, S] In some sense, this is a special case of the proportional allocation formula with the factor γ i chosen that gives rise to the covariance allocation principle: j=1 K i = c icov[x i, S] K, i = 1,..., n. Var[S] page 16

17 Proportional and covariance allocation results s proportional allocation covariance allocation based on based on Line of business VaR CTE VaR CTE Auto (PD) Auto (liab) Household Prof liab Other Total page 17

18 Results of covariance vs proportional allocations other s prof liab household auto (liab) auto (PD) covariance proportional page 18

19 s Dhaene, J., Tsanakas, A., Valdez, E.A., and S. Vanduffel (2012). Optimal capital allocation principles. Journal of Risk and Insurance, 79(1), Sweeting, P. (2011). Financial, International Series on Actuarial Science, Cambridge University Press. Tang, A. and E.A. Valdez (2006). Economic capital and the aggregation of risks using copulas. Proceedings of the 28th International Congress of Actuaries, Paris, France. page 19