Pricing No-Negative-Equity-Guarantee for Equity Release Products under a Jump ARMA-GARCH Model Presenter: Sharon Yang Co-authors: Chuang-Chang Chang Jr-Wei Huang National Central University, Taiwan 2015/4/9 1 Outline Introduction. Investigation of House Price Return Dynamics With Jumps. Valuation Framework for No-Negative-Equity-Guarantee. Numerical Analysis. Conclusion. 2015/4/9 2 Introduction 2015/4/9 3 What are Equity Releasing Products? A kind of home equity conversion that allows the elder persons to borrow money with their home as the collateral . The loans accrue interest are only repaid once the people is died or leave the house. Such products are needed for “equity rich and cash poor” persons. For example: a rolled-up mortgage Loan Value: K --- Kt Kevt at time t Property Value: H 0 ---> H t Age x Die(x+s) Loan Period 4 2015/4/9 4 The Risk from Lender Prospective The loan value may exceed the value of the property. v Ke H How to deal with such risk? Using Insurance. Ex: HECM program in the united states. Securitization Writing a no-negative-equity-guarantee(NNEG) Payoffs: v Max[(Ke H ),0] an European put option on the mortgaged property 2015/4/9 5 Purpose of this study Can Black & Sholes option pricing formula apply to value NNEG? No! We built up a general framework which considers the dynamics of the house price return with jumps. 2015/4/9 6 Purpose of this study-Con’t Li et al . (2010) conclude that the Nationwide House Price Index has the following statistical properties: there is a strong positive autocorrelation effect among the log-returns the volatility of the log-returns varies with time; a leverage effect is present in the log-return series ARMA-EGARCH Model Chen et al.(2010) use the ARMA-GARCH model to price reverse mortgage for the HECM program in the U.S.. 2015/4/9 7 Purpose of this study-Con’t We consider a jump model that incorporate both autocorrelation effect and volatility cluster. a Jump ARMA-GARCH Model 2015/4/9 8 An Investigation of House Price Return Dynamics with Jumps 2015/4/9 9 Jumps in House Price Returns? According to the quarterly data from 1952 to 2008, it can show that the quarterly housing price changed more than three standard deviations. 2015/4/9 10 Jumps in HousePrice or Equity Returns ? Chen et al. (2009) study U.S. mortgage insurance premium using Merton jump diffusion process for house price returns. Merton (1976) build a jump diffusion model with a continuous-time basis. dH t dt dWt dJ t Ht NT J T (V j 1) j 1 2015/4/9 11 Jumps in House Price or Equity Returns ? Kou (2002) also considers the leptokurtic feature and proposes a double exponential jump-diffusion model. The return distribution of assets may have a higher peak and two (asymmetric) heavier tails than those of the normal distribution. f ( y) p 1e1 y1{ y0} q 2e2 y1{ y0} , 2015/4/9 1 1,2 0, 12 Jumps in House Price or Equity Returns ? Chan and Maheu (2002) and Duan et al. (2006, 2007) both examine the jump effect with equity returns under a GARCH model Dynamic jumps in return v.s. Constant jumps in both returns and volatility. 2015/4/9 13 Jumps in House Price or Equity Returns ? Chan and Maheu (2002) Dynamic jumps in return s m Nt i 1 j 1 k 1 Yt c iYt 1 j t j t Vt , k q ht w i 1 2 i t i p j ht j , j 1 exp(t )t j P( N t j | t 1 ) , j 0,1, 2... j! t 0 t 1 t 1 2015/4/9 14 Jumps in House Price or Equity Returns ? Duan et al. (2006, 2007) Constant jumps in both returns and volatility. rt t ht J t Jt z (0) t Nt zt( j ) j 1 J t 1 ht 0 1ht 1 2 ht 1 c 2 2 1 ( ) where 2 zt(0) ~ N (0,1), zt( j ) ~ N ( , 2 ) N t ~ Poisson( ) 2015/4/9 15 Jumps in House Price or Equity Returns ? We extend Chan and Maheu (2002) to consider the dynamic jump effect with house price returns under an ARMA-GARCH model and develop a framework for pricing the NNEG. 2015/4/9 16 ARMA-GARCH Model Ht eYt H t 1 Yt c Y i 1 ht w i 1 i t i j j 1 q p i 1 j 1 t j t 2 i t i j ht j p q m s i j 1 j 1 Y t follows an ARMA process. ht follows a GARCH process. 2015/4/9 17 Dynamic Jump ARMA-GARCH Model s m Yt c iYt 1 j i 1 j 1 q ht w i i 1 J t t j t J t p 2 t i j ht j j 1 N (t ) V j 1 t, j Return jump size : Vt N (t , t2 ) Number of jumps between t-1 and t: N (t ) The case for a dynamic jump: 2015/4/9 Poisson( t ) t 0 t 1 t 1 18 A Comparison of Model Fitting Model Selection, 1953Q4~2008Q4 Model Log-Likelihood AIC BIC Geometric Brownian Motion 499.1072 -4.5192 -4.4883 ARMA-GARCH 567.8156 -5.4861 -5.2542 ARMA-EGARCH 586.4799 -5.4871 -5.2303 Merton Jump 516.2469 -4.6477 -4.5706 Double Exponential Jump 506.3450 -4.5481 -4.4555 Constant Jump ARMAGARCH 592.8361 -5.6193 -5.3149 Dynamic Jump ARMAGARCH 607.5076 -5.6512 -5.3014 Diffusion 2015/4/9 19 A Comparison of Model Fitting Model Selection, 1958Q4~2008Q4 Model Log-Likelihood AIC BIC Geometric Brownian Motion 448.9902 -4.4699 -4.4369 ARMA-GARCH 498.4404 -5.3309 -5.0791 ARMA-EGARCH 505.4725 -5.2110 -4.9395 Merton Jump 465.1300 -4.6013 -4.5188 Double Exponential Jump 452.3087 -4.4907 -4.4061 Constant Jump ARMAGARCH 519.8619 -5.3594 -5.0330 Dynamic Jump ARMAGARCH 522.0326 -5.3817 -5.0016 Diffusion 2015/4/9 20 A Comparison of Model Fitting Model Selection, 1968Q4~2008Q4 Model Log-Likelihood AIC BIC Geometric Brownian Motion 343.6102 -4.2701 -4.2317 ARMA-GARCH 405.9601 -5.2722 -4.9021 ARMA-EGARCH 397.5060 -5.0637 -4.7417 Merton Jump 345.2055 -4.2526 -4.1565 Double Exponential Jump 345.4001 -4.2425 -4.1272 Constant Jump ARMAGARCH 415.0801 -5.3314 -4.9211 Dynamic Jump ARMAGARCH 416.0165 -5.3679 -4.9124 Diffusion 2015/4/9 21 The Valuation Framework for NoNegative-Equity-Guarantee 2015/4/9 22 Pricing No Negative Equity Guarantee Let us define the following notation: K : the amount of loan advanced at time zero; H t : the value of the mortgaged property at time t; r : the constant risk-free interest rate; v: the roll-up interest rate; g : the rental yield; : the average delay in time from the point of home exit until the actual sale of the property. 2015/4/9 23 Pricing No Negative Equity Guarantee Assuming the person dies in the middle of the year Considering the delaying time Payoff Max[( Ks 1/ 2 H k 1/ 2 ),0] Valuation EQ [er ( s 1/ 2 ) Max[( Ks 1/ 2 Hk 1/ 2 ), 0] 2015/4/9 24 Pricing No Negative Equity Guarantee w x 1 VNNEG (0) t 0 w x 1 t 0 Q r ( s 1/ 2 ) p q E [ e Max[( K s 1/ 2 H k 1/ 2 ), 0] | 0 ] s x xs 1 , H 0 , K , v, r , g ) s px qx sV (0, s 2 1 where V (k , H 0 , K , v, r , g ) is calculated using simulations. 2 The value of P under measure Q can be obtained using conditional Esscher transform. 2015/4/9 25 Pricing No Negative Equity Guarantee Under the risk-neutral measure Q, the return processes of and to characterize the jump ARMA(s,m)-GARCH(p,q) model become Q Nt h YtQ r g t VtQ,k 2 k 1 q p i 1 j 1 htQ w i (t i htQi ) 2 j htQ j Special Case: Constant Jump htQ Yt r g 2 Q q p i 1 j 1 htQ w i (t i htQi )2 j htQ j , 2015/4/9 26 Pricing No Negative Equity Guarantee Black and Sholes V (0, s 1/ 2 , H 0 , K , v, r , g ) BSM =Ke( v-r )( s 1/2 ) N (-d 2 ) - H 0e (- g ( s 1/2 )) N (-d1 ), Merton Jump V (0, s 1/ 2 , H 0 , K , v, r , g ) MJ exp *( s 1/2 ) j 0 2015/4/9 ( *( s 1/2 ) ) j ( v-r MJ )( s 1/2 ) Ke N (-d 2MJ ) - H 0e (- g ( s 1/2 )) N (-d1MJ ), j! 27 Making Numerical Analysis 2015/4/9 28 Numerical Analysis 2015/4/9 29 Numerical Analysis 2015/4/9 30 Numerical Analysis 2015/4/9 31 Conclusion This article contributes to the literature in the following ways. Dynamic Jump ARMA-GARCH model can better capture the dynamics of house price return. The estimation of the proposed jump ARMAGARCH model is carried out and presents a better fitting result compared with various house price return models proposed in the literature. 2015/4/9 32 Conclusion This article contributes to the literature in the following ways. The risk neutral pricing framework for the jump ARMA-GARCH model is derived using the conditional Esscher transform technique. Numerical result shows that incorporating the jump effect in house price returns is important for pricing NNEG. 2015/4/9 33 The End. Thanks! 2015/4/9 34