arXiv:1706.07864v4 [math.PR] 30 Nov 2018 oe;-al [email protected] Korea;e-mail: [email protected] mail: rcs ycnieigteitniyo oso rcs sarealiza a as process gene Poisson natural of a intensity and the (1955) considering Cox by by process introduced first is process Cox A Introduction 1 Keywords MSC2010: rcs eitosfrCxpoesswt htnoise shot a with processes Cox for deviations Precise ∗ † eateto ahmtc,Dn- nvriy ua,Saha-gu Busan, University, Dong-A Mathematics, of Department Tallahassee University, State Florida Mathematics, of Department eitosfrso os o rcs sn h eetmod- recent the using th process Cox study we noise shot paper, for this deviations In number a established. models, be such easily of can structure results the to Due considere the be procedure. can on tion it so depends and only process stochastic not a intensity as considered its because flexible is and cess statistic, theory, queue finance, insurance, in applied ecnie o rcs ihPisnso os nest w intensity noise shot Poisson with process Cox a consider We o rcse,Pisnso os,Mod noise, shot Poisson processes, Cox : rmr 05;scnay6F5 60F10. 60F05, secondary 60G55; primary aliCheng Zailei eebr3 2018 3, December intensity ∗ Abstract 1 onsoSeol Youngsoo − φ ovrec,Peiedeviations. Precise convergence, L336 ntdSaeo America;e- of State United 32306, FL , adn-ar 5,3,Rpbi of Republic 37, 550, Nakdong-daero , ayohrfils o pro- Cox fields. other many satose randomiza- step two a as d φ utain n precise and fluctuations e ovrec method. convergence iebtas a be can also but time fueu n general and useful of † ihhsbe widely been has hich ino admmea- random a of tion aiaino Poisson a of ralization sure (Møller, 2003). The Cox process provides the flexibility of letting the intensity not only depend on time but also allowing it to be a . Hence, it can be viewed as a two-step randomization procedure which can deal with the stochastic nature of catas- trophic loss occurrences in the real world. Moreover, shot noise processes (Cox, 1980) are particularly useful to model claim arrivals; they provide measures for frequency, magnitude and the time period needed to determine the effect of catastrophic events within the same framework; as time passes, the shot noise process decreases as more and more losses are settled, and this decrease continues until another event occurs which will result in a positive jump. Therefore, the shot noise process can be used as the intensity of a Cox process to measure the number of catastrophic losses. The shot noise Cox processes were introduced in Møller (2003) and further generalized in Hellmund, Prokeˇsov´a, and Jensen (2008) with- out discussing the statistical inference for the model. Note that the class of shot noise Cox processes also includes the very popular Poisson Neyman-Scott processes like the Thomas process((Thomas, 1949),[(Illian et al., 2008), Section 6.3.2]). The Cox model has been used widely in many aspects: insurance, finance, queue the- ory, statistic etc.(cf.(Albrecher and Asmussen, 2006),(Cox, 1955),(Dassios and Jang, 2003)). Dassios and Jang (2003) applied the Cox process with Poisson shot noise intensity to pricing stop-loss catastrophe reinsurance contract and catastrophe insurance derivatives. Albrecher and Asmussen (2006) studied the asymptotic estimates for infinite time and finite time ruin probabilities of the risk model. There are many applications of large deviations such as insurance, portfolio manage- ment, risk management, queue system, ((Macci and Stabile, 2006), (Gao and Yan, 2008),(Shen, Lin, and Zhang, 2009),(Ganesh, Macci, and Torrisi, 2007)). Macci and Stabile

(2006) studied the large deviation principles for the Markov modulated risk process with rein- surance. Gao and Yan (2008) extended the result considered in Macci and Stabile (2006) to the sample path large and moderate deviation principles. Shen, Lin, and Zhang (2009) ob- tained the precise large deviation results for the customer arrival based risk model which can

2 be treated as a generalized Poisson shot noise process. Recently, Macci and Torrisi (2011) took into account the large deviation estimations for the ruin probability of the risk pro- cesses with shot noise Cox claim number process and reserve dependent premium rate. Many researchers have made great efforts of the precise large deviations for the loss process of a classical insurance risk model and have obtained a lot of inspiring results. Precise large devi- ations for the loss process have been widely investigated. For the classic results, we refer the reader to Cline and Hsing (1991), Kl¨uppelberg and Mikosch (1997), Mikosch and Nagaev (1998), and Ng et al. (2004), among others. It is worth mentioning that the results of pre- cise large deviations for random sums are particulary useful for evaluation of some risk measures such as conditional tail expectation and value at risk of aggregate claims of a large insurance portfolio; see McNeil et al. (2005) for a review of risk measures. Wang and Wang (2007) investigated precise large deviations of multi-risk models. Previous works on insurance applications using a shot noise process or a Cox pro- cess with shot noise intensity can be found in Br´emaud (2000), Dassios and Jang (2003), Jang and Krvavych (2004), Torrisi (2004), Albrecher and Asmussen (2006), Macci and Torrisi

(2011), Zhu (2013) and Schmidt (2014). However, previous study on the large deviations and moderate deviations for Cox process only gives the leading order term (Gao and Yan, 2012), but not the higher order expansion which make more accurate computational tractability for an financial application. Distributional properties for such a model generate computational tractability for an financial application. Transform, distribution, and moment formulae com- putational tractability for a range of applications in portfolio credit risk. The transform for- mulae facilitate the valuation, hedging, and calibration of a portfolio credit derivative, which is a security whose payoff is a specified function of the portfolio loss, and which provides insurance against default losses in the portfolio. In this paper, we assume that N , t 0 is a Cox process with Poisson shot noise { t ≥ }

3 intensity, that is, the intensity λt of Nt is stochastic and at time t, given by

t λ = ν + g(t s)dN¯ , (1) t − s Z0 where N¯ is a Poisson process with intensity ρ,ν > 0 and g : R R being locally bounded. t + → + We study precise deviations for shot noise Cox process using the recent mod-φ convergence method developed in Feray, Meliot, and Nikeghbali (2013). In many applications in finance, insurance, and other fields, more precise deviations are desired, which motivates us to study the precise deviations for Cox process with a shot noise due to higher order expansion given by the precise deviation for Cox process. Taylor expansions have numerous applications to simplify complex functions in finance such as the price of a portfolio of options or bonds. The more terms used in the expansion the more accurate the approximation. Higher or- der Taylor expansions are obtained using higher partial derivatives. Higher derivatives are derivatives of derivatives. Thus higher order expansion is one of the most important con- cepts in mathematical finance literature. In particular, using the re-normalization theory called mod-φ convergence method, we construct an extremely flexible framework to study the precise deviations and limit theorems. This type of convergence is a relatively new concept with many deep ramifications. It is worth mentioning that the results of precise large deviations for Cox process with a shot noise are particulary useful for calculations and evaluations in financial applications using the higher order expansion given by the precise deviations which is more applicable than the large or moderate deviations principle. Very recently, precise deviations for Hawkes processes was studied in Gao and Zhu (2017) using the mod-φ convergence method. In particular, Gao and Zhu (2017) used Hawkes processes for large time asymptotics, that strictly extends and improves the existing results in the literature. In order to apply the results of mod-φ convergence theory, we manipulate the the moment generating function for shot noise Cox process, and then the precise deviations principle and precise moderate deviation and fluctuation results are obtained by the mod-φ

4 convergence method after careful analysis and series of lemmas. The structure of this paper is organized as follows. Some auxiliary results and the main results are stated in Section 2. The proofs for the main theorems are contained in Section 3.

2 Statement of the main results

This section states the main results of this paper. It consists of two key lemmas and the precise deviations principle and furthermore precise deviation principle and fluctuation re- sults. The main strategy of proving the precise deviations principle is by showing the mod-φ convergence as defined in Feray, Meliot, and Nikeghbali (2013) and apply their Theorem 3.4 and 3.9 to get the main Theorem. We start with the definition of mod-φ convergence and then the assumptions which we will use throughout the paper. Let us first recall the definition of mod-φ convergence, see e.g. Definition 1.1. (Feray, Meliot, and Nikeghbali,

zXn 2013). Let (Xn)n N be a sequence of real-valued random variables and E[e ] exist in a strip ∈ := z C : c < (z) < d , with c

tnη(z) zXn e− E[e ] ψ(z), (2) → where t + as n . Then we say that X converges mod-φ on with parameters n → ∞ →∞ n S(c,d)

(tn)n N and limiting function ψ. ∈ 1 Assume that φ is a lattice distribution and the convergence is at speed ν . Then O tn Theorem 3.4. Feray, Meliot, and Nikeghbali (2013) says that for any x R in the interval ∈

5 (η′(c), η′(d)) and θ∗ defined as η′(θ∗)= x, assume that t x N, then, n ∈

tnF (x) e− a1 a2 aν 1 1 P(X = t x)= ψ(θ∗)+ + + + − + , (3) n n t t2 ··· tν 1 O tν 2πtnη′′(θ∗) n n n−  n ! p as n , where F (x) := supθ R θx η(θ) is the Legendre transform of η( ), and similarly, →∞ ∈ { − } · if x R is in the range of (η′(0), η′(d)), then, ∈

tnF (x) e− 1 b1 b2 bν 1 1 P − (Xn tnx)= θ∗ ψ(θ∗)+ + + + ν 1 + ν , (4) ≥ 2πt η (θ ) 1 e− tn tn ··· t − O t n ′′ ∗ − n  n ! p as n , where (a )∞ , (b )∞ are rational fractions in the derivatives of η and ψ at θ∗, →∞ k k=1 k k=1 that can be computed as described in Remark 3.7. (Feray, Meliot, and Nikeghbali, 2013).

Then let’s recall Theorem 3.9. (Feray, Meliot, and Nikeghbali, 2013), which states the precise moderate deviation.

Consider a sequence (Xn)n N that converges mod-φ, with a reference infinitely divisible ∈ 1/6 law φ that is a lattice distribution. Assume y = o((tn) ). Then,

P X t η′(0) + t η (0)y = P ( R(0, 1) y)(1+ o(1)). (5) n ≥ n n ′′ N ≥  p  1/2 On the other hand, assuming y 1 and y = o((t ) ), if x = η′(0) + η (0)/t y and h ≫ n ′′ n is the solution of η′(h)= x, then p

tnF (x) e− P Xn tnη′(0) + tnη′′(0)y = (1 + o(1)) (6) ≥ h 2πtnη′′(h)  p  And Corollary 3.13 (Feray, Meliot, and Nikeghbali,p 2013) gives a more explicit form of

1/4 Theorem 3.9. (Feray, Meliot, and Nikeghbali, 2013). If y = o((tn) ), then one has

2 3 (1 + o(1)) y η′′′(0) y P ′ − 2 Xn tnη (0) + tnη′′(0)y = e exp 3/2 (7) ≥ y√ π 6(η′′(0)) √tn 2    p  1/2 1/m More generally, if y = o((tn) − ), then one has

6 m 1 (i) i/2 i (1 + o(1)) − F (η′(0)) (η′′(0)) y P X t η′(0) + t η (0)y = exp (8) n ≥ n n ′′ √ − i! (i 2)/2 y 2π i=2 tn− !  p  X

Here we consider Nt as a Cox process with Poisson shot noise intensity defined in (1), we assume throughout this paper that

g 1 = ∞ g(t)dt < • k kL 0 ∞ R 1 g(t)= +2 as t • O tν →∞  By the definition of mod-φ convergence, the moment generating function of Nt is finite, which implies the first assumption. We also need the second assumption for the convergence speed in mod-φ convergence. Our main results for the precise large deviations and precise moderate deviations for the Cox process with Poisson shot noise intensity are stated as follows.

2.1 Precise Large Deviations

Definition 2.1. We say that the sequence of random variables (Xn)n N converges mod-φ at ∈ 1 1 speed ν if the difference of the two sides of Equation 2 can be bounded by CK ν for O tn tn any z in a given compact subset K of .   S(c,d)

Theorem 2.2. (i) For any x> 0, and tx N, as t , ∈ →∞

tI(x) I′′(x) a1 a2 aν 1 1 P(N = tx)= e− ψ(θ∗)+ + + + − + , (9) t 2πt t t2 ··· tν 1 O tν r −  ! where for any θ C ∈

ψ(θ) := eρϕ(θ) (10)

7 and

u ∞ ∞ (eθ 1) R g(s)ds (eθ 1) R g(s)ds ϕ(θ)= e − 0 e − 0 du, (11) 0 − Z h i which is analytic in θ for any θ C and I(x) is defined as ∈

I(x) = sup θx η(θ) (12) θ R{ − ∈

and

θ θ (e 1) g 1 η(θ)=(e 1)ν + ρ e − k kL 1 (13) − −  

θ∗ is the solution of η′(θ∗)= x and (ak)k∞=1 are rational fractions in the derivatives of η and

ψ at θ∗.

Note that θ∗ is unique for each x because it is easy to show that η′′(θ) > 0. (ii) For any x > ν + ρ g , as t , k kL1 →∞

tI(x) I′′(x) 1 b1 b2 bν 1 1 P(N tx)= e− ∗ ψ(θ∗)+ + + + − + , (14) t ≥ 2πt 1 e θ t t2 ··· tν 1 O tν r − − −  ! where (bk)k∞=1 are rational fractions in the derivatives of η and ψ at θ∗.

The key to prove main Theorem is to verify the mod-φ convergence. More precisely, we need to show the following three lemmas.

Lemma 2.3. Y has an infinitely divisible distribution. Y is some random variable defined as

θ η(θ) θ (e 1) g 1 θY e = exp (e 1)ν + ρ e − k kL 1 = E[e ], (15) − − h  i

Note that infinitely divisible is a limitation of the method of mod-φ convergence. Fortunately, the limiting distribution in the case of the Cox process with shot noise is indeed infinitely

8 divisible.

Lemma 2.4. For any θ C, ∈

u ∞ ∞ (eθ 1) R g(s)ds (eθ 1) R g(s)ds ϕ(θ)= e − 0 e − 0 du, (16) 0 − Z h i is well-defined and analytic in θ, and

tη(θ) θNt ρϕ(θ) e− E[e ] ψ(θ) := e , (17) → as t , locally uniformly in θ. →∞ Lemma 2.5. N , t 0 converges mod-φ at speed 1 as t . { t ≥ } O tν →∞  2.2 Precise Moderate Deviations

By Theorem 3.9. (Feray, Meliot, and Nikeghbali, 2013), we have the following theorem:

Theorem 2.6. (i) For any y = o(t1/6), as t , →∞

P N (ν + ρ g 1 )t + √t ν + ρ g 1 ( g 1 + 1)y = Ψ(y)(1 + o(1)), (18) t ≥ k kL k kL k kL  p 

2 where Ψ(y) := ∞ 1 e x /2dx. y √2π − (ii) For anyRy 1 and y = o(t1/2), as t , ≫ →∞

tI(x∗) e− P Nt (ν + ρ g L1 )t + √t ν + ρ g L1 ( g L1 + 1)y = (1 + o(1)), (19) ≥ k k k k k k θ∗ 2πtη′′(θ∗)  p  p 1 where x∗ =(ν + ρ g 1 )+ ν + ρ g 1 ( g 1 + 1)y and θ∗ is the solution of η′(θ∗)= x∗ k kL √t k kL k kL and the notation a b meansp b is much less than a. ≫ Remark 2.7. Note that Theorem 2.6 (i) is an improvement of the classical because here we allow y = o(t1/6) for t . Theorem 2.6 (ii) is the moderate → ∞ deviations.

9 By using Corollary 3.13 (Feray, Meliot, and Nikeghbali, 2013), we can get a more explicit form of Theorem 2.6.

Theorem 2.8. (i) For any y = o(t1/4), as t , →∞

P N (ν + ρ g 1 )t + √t ν + ρ g 1 ( g 1 + 1)y t ≥ k kL k kL k kL 2 3 (20)  (1 + o(1))p y η′′′(0) y  − 2 = e exp 3/2 y√ π 6(η′′(0)) √t 2  

1/2 1/m (ii) For any y = o(t − ), where m 3, as t , ≥ →∞

P N (ν + ρ g 1 )t + √t ν + ρ g 1 ( g 1 + 1)y t ≥ k kL k kL k kL  m 1  (21) (1 + o(1)) p− I(i)(η (0)) (η (0))i/2yi = exp ′ ′′ , √ − i! t(i 2)/2 y 2π i=2 − ! X where I(x) is defined in (12) and η(θ) is defined in (13).

3 Proofs

In this section, we give a proof of the main theorems and lemmas. The key idea to prove the main theorem is to apply the recently developed mod-φ convergence method.

3.1 Proofs of the results in Section 2.1

Proposition 3.1. Assume that Nt is Cox process with Poisson shot noise intensity defined in (1), then the moment generating function of Nt is

t θ u θNt θ (e 1) R g(s)ds E e = exp (e 1)tν + ρ e − 0 1 du (22) − − " Z0 #     Proof. By the definition of a Cox process with Poisson shot noise intensity and Fubini’s

10 theorem, we have

θ t θNt (e 1) R λsds E e = E e − 0

h θ t is ¯   (e 1)[νt+R0 R0 g(s u)dNuds] = E e − −

h θ t t ¯ i (e 1)[νt+R0 [R g(s u)ds]dNu] = E e − u −

hθ θ t t i (e 1)tν (e 1) R [R g(s u)ds]dN¯u = e − E e − 0 u −

t (23) h t i (eθ 1)tν (eθ 1) R g(s u)ds = e − exp ρ e − u − 1 du " 0 − # Z   t t−u (eθ 1)tν (eθ 1) R g(s)ds = e − exp ρ e − 0 1 du " 0 − # Z   t u θ (eθ 1) R g(s)ds = exp (e 1)tν + ρ e − 0 1 du . " − 0 − # Z  

Remark 3.2. From (22), we have

θ 1 θNt θ (e 1) g 1 lim log E e =(e 1)ν + ρ e − k kL 1 . (24) t →∞ t − −     Then, by G¨artner-Ellis theorem (see for details (Bordenave and Torrisi, 2007)), we can say that ( Nt ) satisfies the large deviation principle with the good rate function t ∈·

θ θ (e 1) g 1 I(x) = sup θx (e 1)ν + ρ e − k kL 1 (25) θ R( − − − ) ∈ h  i

Proof of Lemma 2.3. It suffice to show that

θ θ (e 1) g 1 θY exp (e 1)ν + ρ e − k kL 1 = E e − − h  i   where Y is an infinitely divisible random variable. To this end, Y can be interpreted as

R Y = X + Z. Here X is a Poisson random variable with parameter ν, Z = i=1 Zi is a P 11 compound Poisson random variable, R is a Poisson random variable with parameter ρ, Zi are i.i.d. Poisson random variables with parameter g 1 , also, X and Z are independent k kL thus

E eθY = E eθX E eθZ

θ    θ   (e 1) g 1 = exp (e 1)ν exp ρ e − k kL 1 (26) − − h θ i  θ  (e 1) g 1 = exp (e 1)ν + ρ e − k kL 1 − − h  i Therefore, Y has an infinitely divisible distribution.

Proof of Lemma 2.4. We first consider

E eθNt tη(θ) θNt e− E[e ]= . θ (eθ 1) g 1 exp (e 1)νt +ρt e − k kL 1 − − h  i Using (22), we have

E eθNt

θ θ (e 1) g 1 exp (e 1)νt +ρt e − k kL 1 − − h t  i θ u θ (e 1) R g(s)ds (e 1) g 1 = exp ρ e − 0 1 du ρt e − k kL 1 " 0 − − − # (27) Z     t u ∞ (eθ 1) R g(s)ds (eθ 1) R g(s)ds = exp ρ e − 0 e − 0 du " 0 − # Z h i ψ(θ) := eρϕ(θ). →

To prove this, it suffices to show that

t u ∞ (eθ 1) R g(s)ds (eθ 1) R g(s)ds e − 0 e − 0 du − Z0 h u ∞ i ∞ (eθ 1) R g(s)ds (eθ 1) R g(s)ds e − 0 e − 0 du (28) → 0 − Z h i as t , locally uniformly in θ C. It is equivalent to show that for any compact set → ∞ ∈

12 K C, ǫ> 0, N s.t. when t > N, ⊂ ∀ ∃

u ∞ ∞ (eθ 1) R g(s)ds (eθ 1) R g(s)ds e − 0 e − 0 du < ǫ, (29) − Zt h i

for θ K, and we have ∈

u ∞ ∞ (eθ 1) R g(s)ds (eθ 1) R g(s)ds e − 0 e − 0 du − Zt h i (30) ∞ θ (eθ 1)ξ ∞ = (e 1)e − g(s)ds du − Zt Zu h i

u by the mean value theorem, ξ(u) g(s)ds, ∞ g(s)ds . ∈ 0 0 R R  ξ ∞ θ (eθ 1) R g(s)ds ∞ (e 1)e − 0 g(s)ds du − Zt Zu h i (31) ˆ ∞ ∞ < CK g(s)dsdu Zt Zu

θ ∞ ˆ θ (e 1) R0 g(s)ds 1 where CK = supθ K (e 1)e − . According to our assumption, g(t)= tν+2 , ∈ − O so ∞ ∞ g(s)dsdun= 1 as t . o  t u O tν →∞ R R ˆ  So we have for ǫ′ = ǫ/CK there exists N such that if t > N,

∞ ∞ g(s)dsdu < ǫ/CˆK (32) Zt Zu

then for t > N,

ξ ∞ θ (eθ 1) R g(s)ds ∞ (e 1)e − 0 g(s)ds du < ǫ (33) − Zt Zu h i so we have proved (29). Thus, we conclude that

t u ∞ (eθ 1) R g(s)ds (eθ 1) R g(s)ds ρϕ(θ) exp ρ e − 0 e − 0 du ψ(θ) := e , (34) " 0 − # → Z h i

13 as t , locally uniformly in θ C, where →∞ ∈

u ∞ ∞ (eθ 1) R g(s)ds (eθ 1) R g(s)ds ϕ(θ)= e − 0 e − 0 du. 0 − Z h i

Hence, ϕ(θ) is well-defined and is analytic in θ.

Proof of lemma 2.5. By definition 2.1, it suffices to show that

t ∞ 1 exp ρ G(u)du exp ρ G(u)du

u ∞ (eθ 1) R g(s)ds (eθ 1) R g(s)ds as t . Here G(u)= e − 0 e − 0 . According to (30) and (31), →∞ −

t ∞ exp ρ G(u)du exp ρ G(u)du − " Z0 # " Z0 # ξ′ t ∞ = ρ2 exp ρ G(u)du G(u)du G(u)du (36) − " Z0 # Z0 Z0  ∞ ∞ 1 < C¯ g(s)dsdu = C K K tν Zt Zu  

as t . Here ξ′ (t, ). Thus we have proved Lemma 2.5. →∞ ∈ ∞

Completion of the Proof of Theorem 2.2. From Lemma 2.3 and Lemma 2.4, we have es- tablished the mod-φ convergence. Hence, by Theorem 3.4. (Feray, Meliot, and Nikeghbali, 2013), for any x> 0, and tx N, ∈

tI(x) I′′(x) a1 a2 aν 1 1 P(N = tx)= e− ψ(θ∗)+ + + + − + , (37) t 2πt t t2 ··· tν 1 O tν r −  ! and for any x > ν + ρ g , k kL1

tI(x) I′′(x) 1 b1 b2 bν 1 1 P(N tx)= e− ∗ ψ(θ∗)+ + + + − + , (38) t ≥ 2πt 1 e θ t t2 ··· tν 1 O tν r − − −  !

14 where I(x) is defined in (25). This completes the proof of Theorem 2.2.

3.2 Proofs of the results in Section 2.2

Since we have established the mod-φ convergence and Nt is a lattice distribution, the proof of Theorem 2.6 and Theorem 2.8 follows from Theorem 3.9 and Corollary 3.13 in (Feray, Meliot, and Nikeghbali, 2013), respectively. Acknowledgements

Youngsoo Seol is grateful to the support from the Dong-A University research grant.

References

Albrecher, H. and Asmussen, S. 2006. Ruin probabilities and aggregate claims distributions for shot noise Cox processes, Scand. Actuar. J. 2:86–110.

Bordenave, C. and Torrisi, G.L. 2007. Large deviations of Poisson cluster processes, Stoch.Models 24:593-625.

Br´emaud, P. 2000.An insensitivity property of Lundberg’s estimate for delayed claims, J. Appl. Prob 37:914–917.

Cline, D.B.H. and Hsing, T. 1991.Large deviation probabilities for sums and maxima of

random variables with heavy or subexponential tails, Preprint, Texas A&M University.

Cox, D. R. 1955. Some statistical methods connected with series of events, J. R. Stat. Soc. Ser. B Stat.Methodol 17(2):129–164.

Cox, D. R. and Isham, V. 1980. , Chapman and Hall, London.

Dassios, A. and Jang, J. 2003. Pricing of catastrophe reinsurance and derivatives using the Cox process with shot noise intensity, Finance Stoch. 7:73–95.

15 Feray, V., Meliot, P.-L, and A. Nikeghbali. 2013. Mod φ convergence I: Normality zones and − precise deviations, arXiv:1304.2934.

Ganesh, A, Macci, C, and Torrisi, G.L. 2007. A class of risk processes with reserve-dependent premium rate: sample path large deviations and importance sampling, Queuing System. 55: 83–94.

Gao, F. Q. and Yan, J. 2008. Sample path large deviations and moderate deviations for risk

processes with reinsurance, J. Appl. Prob 45: 800–817.

Gao, F. Q. and Yan, J. 2012. Exponential martingale and large deviations for a Cox risk process with Poisson shot noise intensity, J. Math. Anal. Appl. 394:74–83.

Gao, F. and Zhu, L. 2017. Precise deviations for Hawkes processes, arXiv:1702.02962.

Hellmund, G., Prokeˇsov´a, M. and Jensen, E. B. V. 2008. L´evy-based Cox Point processes, Adv. in Appl. Probab. 40:603–629.

Illian, J., Penttinen, A., Stoyan, H. and Stoyan, D. 2008. Statistical Analysis and Modeling of Spatial Point Patterns, Chichester:Wiley.

Jang, J. and Krvavych, Y. 2004. Arbitrage-free premium calculation for extreme losses using the shot noise process and the Esscher transform, Insurance: math. Econom. 35:97–111.

Kl¨uppelberg, C. and Mikosch, T. 1997. Large deviations of heavy tailed random summs with applications in insurance and finance, J. Appl. Prob. 34(2):293-308.

Macci, C. and Torrisi, G.L. 2011. Risk processes with shot noise Cox claim number process and reserve dependent premium rate, Insurance: math. Econom. 48:134–145.

Macci, C. and Stabile, G. 2006. Large deviation for risk processes with reinsurance, J. Appl. Prob. 43:713–728.

16 McNeil, A.J., Frey, R. and Embrechts, P. 2005. Quantitative Risk Mangagment. Concepts, Techniques and Tools, Princeton, NJ:Princeton University Press.

Mikosch, T. and Nagaev, A.V. 1998. Large deviations of heavy-tailed sums with applications in insurance, Extremes, 1(1):81-110.

Møller, J. 2003. Shot noise Cox processes, Adv. in Appl. Probab. 35(3):614–640.

Ng, K.W., Tang, Q.H., Yan, J.A. and Yang,H.L. 2004. Precise large deviations for sums of

random variables with consistently varying tails, J. Appl. Prob. 41(1):93-107.

Schmidt, T. 2014. Catastrophe insurance modeled by shot noise processes, Risks, 2:3-24.

Shen, M.X., Lin, Z.Y. and Zhang, Y. 2009. Precise large deviations for the actual aggregate loss process, Stoch. Anal. Appl. 27:1000–1013.

Thomas, M. 1949. A generalization of Poisson’s binomial limit for use in ecology, Biometrika, 36:18-25.

Torrisi, G. L. 2004. Simulating the ruin probability of risk processes with delay in claim settlement, Stochastic Process. Appl. 112:225-244.

Wang, S.J. and Wang, W.S. 2007. Precise large deviations for random variables with consis- tently varying tails in multi risk models, J. Appl. Prob. 44 (4):889-900.

Zhu, L. 2013. Ruin probabilities for risk processes with non-stationary arrivals and sub- exponential claims, Insurance Math. Econom. 53:544–550.

17