• No se han encontrado resultados

There is also a large evidence that the sovereign bond market has a signicant impact on CDS spreads. For instance, Calice, Vhen and Williams (2011) nd strong liquidity

spill-over between sovereign bond and CDS markets. Thus, similarly to the previous section, we accommodate equation (4.1) in order to take into consideration the bond liquidity spill-over. The main assumption is that there are market-wide liquidity shocks that are transmitted to the sovereign CDS market. In fact, in case of default, CDS sellers of country i can deliver all country-specic eligible bonds even if they may be

dierent from the instruments that investors were initially hedging which imply that country-specic bond liquidity could inuence CDS spreads in addition to the market- wide bond liquidity. Therefore, in this subsection, we analyse liquidity spill-over from (1) the country-specic bond portfolio and (2) the market-wide bond portfolio.

In what follows, we extend the Liquidity-Adjusted CAPM framework dened above and explore all the potential channels through which bond liquidity could inuence CDS spread changes. In other terms, after controlling for default risk and bond liquidity level, we examine the liquidity eect of the country-specic bond portfolio (through β5i)and

of the market-wide bond portfolio (throughβ6i, β7i and β8i). The model is dened as

follows:

E(CDSt)−Def aultRisk=E(cBondt ) +λβ

5i+λβ6i+λβ7i+λβ8i (4.7)

where β5i represents the sensitivity of CDS liquidity to country-specic sovereign

bonds. β5i= Cov(c CDS t −Et−1(cCDSt ), c Bonds L −Et−1(cBondsL )) var(rBonds

L −Et−1(rLBonds)−cBondsL −Et−1(cBondsL ))

cBonds

L represents the average bond bid-ask spread of the country-specic bond port-

folio. rBondsL is the average total return of the country-specic bond portfolio including

the interest payments, as well as the appreciation or depreciation in bond prices. rBondsL

is computed in the following way:

rBondsL t =r Bonds Lt−1 ∗ Pt+At+N Ct+CPt Pt−1+At−1+N Ct−1 (4.9)

wherePt is the clean bond price,At is the accrued interest, N Ct is the next coupon

and CPt is the value of any coupon received on time t or since time t-1.

The aim of the risk factorβ5i of country iis to capture the impact of the Cheapest-

To-Deliver (CTD) option on sovereign CDS liquidity as suggested by the literature13.

If liquidity dries up in the country-specic bond market, it may cause frictions to CDS sellers who, in case of default, have to nd a cheap and liquid bond to deliver. As we stressed above the CTD is more valuable when the number of deliverable bonds is high, this value should be incorporated in the CDS contract, however if the cheapest bond is illiquid then CDS spreads should reect that change. For example for the case of Argentina and Chile we have 20 and 3 deliverable bonds respectively, this dierence have a direct implication on the value of CTD and thus CDS spreads. To estimate the CTD eect, we construct a portfolio of all active and available dollar denominated country-specic sovereign bonds and take the average of the bid-ask spreads to proxy for liquidity.

Moreover, the second goal of β5i is also to capture the potential liquidity spill-over

coming from the bond market through hedging. In fact, if traders use mainly CDSs for hedging purposes then any movement of bond yields either caused by default or liquidity risk can impact CDS spreads14.

β6i, β7i and β8i introduced in equation (4.7) aim to measure the commonality in

liquidity between sovereign CDS and bond markets and are dened as follows:

β6i= Cov(c CDS t −Et−1(cCDSt ), cBondsM −Et−1(cBondsM )) var(rBonds M −Et−1(rMBonds)−c Bonds M −Et−1(cBondsM )) (4.10) β7i= Cov(r CDS t , cBondsM −Et−1(cBondsM )) var(rBonds

M −Et−1(rMBonds)−cBondsM −Et−1(cBondsM ))

(4.11) β8i= Cov(c CDS t −Et−1(cCDSt ), rMBonds−Et−1(rBondsM )) var(rBonds M −Et−1(rMBonds)−c Bonds M −Et−1(cBondsM )) (4.12)

The betas represent the sensitivity of CDS liquidity risk to sovereign bond market liquidity (β6i), the sensitivity of CDS spreads to sovereign bond market liquidity (β7i)

and the sensitivity of CDS liquidity risk to sovereign bond market return (β8i). rBondsM

corresponds to the return of the overall bond portfolio. In order to estimaterMBonds, we

utilize the JP Morgan Emerging Market Bond Index (EMBI) which is the most com- prehensive US dollar denominated emerging market debt benchmark. Included in the index are the US dollar-denominated Brady bonds, Eurobonds and traded loans issued by sovereign and quasi sovereign entities. The countries of the EMBI index are: Ar-

14It is important to note that although the aim of this beta risk factor is to capture systematic liquidity risk coming

from the local bond market, the results could be weakened by the small number of available bonds that some of countries have (i.e. Chile and Malaysia have only two and one traded sovereign bonds respectively).

gentina, Brazil, Mexico, South Korea, Russia, Venezuela, Philippines, Poland, Malaysia, Panama, Bulgaria, Nigeria, China, Ecuador, Peru, Colombia, Morocco, Greece, Turkey, Hungary, Croatia, Lebanon, South Africa, Algeria, Thailand, Chile, Cote D'ivoire15. To

proxy for liquidity in the overall bond market cBondsM , we download for each country

mentioned above the bid-ask spreads of all active sovereign bonds traded in the in- ternational market which make a total of 274 bonds and take the weekly average to measure the overall bond liquidity. The beta estimations are conducted over the entire sample resulting in one set of risk factors for each country (Table 4.2).

Brunnermeier and Pedersen (2009) emphasize that market liquidity is closely linked to the level of funding available. They argue that in crisis time market illiquidity could be reinforced by funding liquidity creating what they call a liquidity spiral. In what follows, we discuss the issues related to the joint eect of market and funding liquidity

Documento similar