28.05.2013 Views

Etude des marchés d'assurance non-vie à l'aide d'équilibres de ...

Etude des marchés d'assurance non-vie à l'aide d'équilibres de ...

Etude des marchés d'assurance non-vie à l'aide d'équilibres de ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

tel-00703797, version 2 - 7 Jun 2012<br />

ceiling function. In<strong>de</strong>ed, we have<br />

4.4. Focus on the <strong>de</strong>pen<strong>de</strong>nce structure<br />

P (W = k) = qP (0 = k) + (1 − q)P (⌈Y ⌉ = k) = qδk,0 + (1 − q)(1 − δk,0)e −θ(k−1) (1 − e −θ ).<br />

Furthermore, P (W ≤ x) = q + (1 − q)P (⌊Y ⌋ < ⌊x⌋) = q + (1 − q)(1 − e −ν(⌊x⌋) ) for all x ∈ R+.<br />

In particular for x = k ∈ N, we have FW (k) = q + (1 − q)(1 − e −θk ).<br />

Let Wi = Ii⌈Yi⌉ with Ii’s and Yi be conditionnaly in<strong>de</strong>pen<strong>de</strong>nt exponential random variables,<br />

i.e. Yi|Θ = θ ∼ E(θ). We have<br />

and<br />

P (Wi ≤ x) = 1 − (1 − q)LΘ(⌊x⌋),<br />

P (W1 ≤ x, W2 ≤ y) = 1 − (1 − q)(LΘ(⌊x⌋) + LΘ(⌊y⌋)) + (1 − q) 2 LΘ(⌊x⌋ + ⌊y⌋).<br />

Let DW1,W2 be the distribution function of the pair (FW1 (W1), FW2 (W2)). Similarly to the<br />

previous subsection, we get<br />

<br />

0<br />

DW1,W2 (u, v) =<br />

P (W1 ≤ lu, W2 ≤ lv)<br />

if u < q or v < q<br />

otherwise<br />

.<br />

And we get for u, v ≥ q,<br />

DW1,W2 (u, v) = 1 − (1 − q)(LΘ(⌊lu⌋) + LΘ(⌊lv⌋)) + (1 − q) 2 LΘ(⌊lu⌋ + ⌊lv⌋).<br />

There is also a jump when u or v equal q. In<strong>de</strong>ed, we have DW1,W2 (q, v) = q −q(1−q)LΘ(⌊lv⌋)<br />

and DW1,W2 (u, v) ≥ q2 for u, v ≥ q.<br />

Differences between DZ1,Z2<br />

and DW1,W2<br />

Let u ∈ Im(FW1 ) and v ∈ Im(FW2 ), i.e. there exist n, m ∈ N, such that u = FW1 (n) and<br />

v = FW2 (m). Firstly, we have<br />

u = FW1 (n) = FZ1 (n + 1) and v = FW2 (m) = FZ2 (m + 1).<br />

Since we have lu = n and lv = m, we have<br />

DZ1,Z2 (u, v) = 2 − (1 − q)(LΘ(n) + LΘ(m)) − 1 + (1 − q) 2 LΘ(n + m) = DW1,W2 (u, v).<br />

Now, we focus on the maximal difference between these two functions, after <strong>de</strong>riving an<br />

elementary lemma.<br />

Lemma 4.4.5. Let f be a completely monotone function, c > 0 and p ∈ [0, 1]. The function<br />

fc,p : x ↦→ f(x) − pf(x + c) is also completely monotone.<br />

Proof. As f being completely monotone, (−1) n f (n) is <strong>de</strong>creasing. Thus, we have<br />

(−1) n f (n) (x) ≥ (−1) n f (n) (x + c) ≥ p(−1) n f (n) (x + c),<br />

as x < x + c. Hence, (−1) n f (n)<br />

c,p (x) ≥ 0, i.e. fc,p is completely monotone.<br />

Proposition 4.4.6. Let ∆i,j be the difference between functions DW1,W2 and DZ1,Z2 at the<br />

point u = FW1 (i) and v = FW1 (i). The maximal stairstep is ∆∞,1 = ∆1,∞.<br />

203

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!