Upload
thierry-moudiki
View
922
Download
2
Embed Size (px)
Citation preview
‘Sensibilité d’une rente viagère à l’extrapolationde la courbe des taux dans un contexte LTGA’,
JIRF 2015
Thierry Moudiki (ISFA, Université Lyon 1)
May 21st, 2015
Solvency II economic balance sheet, technical provisions
I Solvency II? QIS5? LTGA?I Simplified balance sheet
Assets Liabilities
Assets - Available capitalat - Solvency Capital Requirement (SCR)market - Risk Marginvalue - Best Estimate Liabilities
I ‘Fair’ valuation ∼ Assets marked to market + Liabilitiesmarked to model
I Technical provisions = Best Estimate Liabilities + RiskMargin
Best Estimate Liabilities (BEL) explained
BELt =∑T
E[D(t,T )CFT |Ft ⊗ Tt ]
I D(t,T ): stochastic discount factorI CFT : future cash-flowsI Ft ⊗ Tt : financial and technical information at tI Simple case: CFT deterministic/highly predictableI Difficult case: CFT depending on financial and technical
information
In the article:
I CFT : deterministic, mortality risk mutualizedI d(t,T ) := E[D(t,T )|Ft ] = ? = critical input
Solvency II term structure of discount factorsHow to derive d(t,T ) within Solvency II?
Liquid part
I Par swap ratesI Parallel shift for credit risk deduction (10 bps)I + Volatility adjustment (optional)I + Matching adjustment (optional)
Extrapolated part
I Last Liquid Point (LLP)I Ultimate Forward Rate (UFR)I Convergence speed (α) ∼ convergence period (years)I Bootstrap, interpolation, extrapolation: Smith-Wilson
Solvency II term structure of discount factors (cont’d)
Solvency II interpolation/extrapolation explained
Solvency II term structure of discount factors (cont’d)
UFR?
I Endogeneous UFR = f∞ = f∞(t)I Vasicek-Fong (1982): French Institute of Actuaries
d(t, s) =∑
mβm,se−mf∞(s−t)
I Nelson & Siegel (1987) or Svensson (1994): Central Banks
d(t, s) = exp (−(s − t) (f∞ + β1K1(t, s) + β2K2(t, s)))
I Exogeneous UFR = f∞ = constantI Smith-Wilson (2001): Solvency II
d(t, s) = e−f∞(s−t) +∑
mβmKm(t, s)
Solvency II term structure of discount factors (cont’d)I Problem with an endogeneous UFR: potential volatility
induced in liabilitiesI ECB data; calibration with Svensson method (Blue = AAA
Bonds, Red = All Bonds)
2004 2006 2008 2010 2012 2014
01
23
45
67
Calibration date
UF
R
Solvency II term structure of discount factors (cont’d)
The case for a fixed exogeneous UFR for insurance
I High demand on long-term swaps from pension funds,artificially driving long rates down?
I More stable liabilities valuationI UFR = Expected inflation + Expected real interest ratesI UFR = fixed since 5 years = 4.2% = 2% + 2.2%
Results from the article
Based on a fixed annuity valuation
I Fixed cash-flows depending on a mortality table (mutualizedmortality risk)
I Additional benefits cash-flows depending of implied forwardI Aggregated cash-flows discounted, using Smith-Wilson for,
interpolation and extrapolation (swap curve at 12/31/2011)
Results on the valuation
1. On interpolation with a ‘complicated’ (realistic, but not socommon) benchmark curve
2. On the dependence with UFR and convergence speed3. On the dependence with LLP and convergence speed
Results from the article (cont’d)On interpolation
Implied forwards on “A Curve where all cubic splines producenegative rates” (Hagan & West (2006))
Results from the article
A ‘Shiny App’ showing the impact of LLP, UFR and α on aannuity
I Just run the following commands in R (internet connexionrequired, tested on Safari and Chrome)
# Loading Shiny packagelibrary(shiny)# Run the app in your browser; execs code from my GistrunGist("ee4e7b9506a09e5d7cb8")
“Problem” with an exogeneous UFRI Hedging current UFR (increasing basis with low rates)?I Reasonable macroeconomic assumptions (2%+2.2%) ?
UFR implied vs market implied forwards rates on EUR swap curves
impl
ied
forw
ard
rate
s
0 5 10 15 20 25 30
0.00
0.01
0.02
0.03
0.04
20132014