New Measures of Australian Corporate Credit Spreads

Transcription

1 New Measures of Austraian Corporate Credit Spreads Ivaio Arsov, Matthew Brooks and Mitch Kosev* Austraian corporations access bond markets both domesticay and offshore. Despite this, there is a ack of pubicy avaiabe data on bond market conditions faced by non-financia corporations (NFCs). This gap in the data is particuary apparent at onger maturities where the ow eve of bond issuance, especiay in the domestic market, makes it difficut to gauge the ong-term credit spreads faced by resident issuers. To address this ack of data, the artice presents a method for estimating aggregate credit spreads of Austraian NFCs across maturities ranging from 1 to 1 years. The estimation method is simpe, transparent and reativey robust in sma sampes. The Bank wi commence pubishing the estimated credit spreads monthy from December 213. Introduction A number of Austraian NFCs are we-estabished issuers in bond markets, both domesticay and offshore. Despite this, historica data on Austraian credit spreads are imited, especiay at onger maturities. 1 To address this, the artice presents newy constructed measures of secondary market credit spreads for bonds issued by Austraian NFCs. Aggregate measures of spreads are estimated as a weighted average of observed spreads of outstanding bonds issued by Austraian NFCs, with the weights determined by the distance between the bonds residua maturities and the target tenor of the estimated spread. 2 The Bank wi pubish these measures of credit spreads and yieds in a new Statistica Tabe F3 Aggregate Measures of Austraian Corporate Bond Spreads and Yieds. 3 Whie * The authors are from Domestic Markets Department. Thomas Wiiams aso contributed to the preiminary anaysis whie he was in Domestic Markets Department. 1 Besides the difficuty in constructing comprehensive sampes of bond data from a range of sources, another factor expaining the ack of such measures is the cost of sourcing, storing and anaysing the data for individua bonds necessary to produce measures of aggregate credit spreads. 2 For a given point in time, the bond s residua maturity measures the time eft unti its maturity date. 3 The new tabe wi repace the existing Statistica Tabe F3 from December 213. the pubication of spreads across tenors (3, 5, 7 and 1 years) is subject to the issuance of bonds with such maturities, the method used is reativey robust to changes in issuance trends. The artice proceeds by examining the issuance patterns of Austraian NFCs since 2. This informs the construction of the sampe used to estimate the aggregate credit spreads. The estimation method is then discussed. The artice concudes by presenting the resuts obtained for aggregate monthy credit spreads of and Austraian NFCs since 25. The Austraian n-financia Corporate Bond Market Bond issuance by Austraian NFCs has grown markedy since the eary 2s. 4 Over this period, around three-quarters of Austraian NFCs bond issuance has been in offshore markets. Most of the offshore issuance has been denominated in US doars, refecting the depth and size of the US bond market (Graph 1). Corporate bond issuers, incuding Austraian NFCs, source much of their onger-term bond funding from the US market where investor demand for onger-dated paper 4 See Back et a (212) for a history of Austraian corporate bonds. Buetin DECEMBER Quarter

2 Graph 1 Gross Bond Issuance by Austraian NFCs* Gross issuance 3 15 Yrs 1 5 Offshore Domestic Weighted average tenor** 3 15 Yrs 1 5 Graph 2 Gross Bond Issuance by Austraian NFCs* Credit rating at issuance AA and A 8 4 AAA AA AA+ AA- A+ A 3 A * For 213, issuance is year to end vember ** Weighted by face vaue; domestic weighted average tenor omitted in 28 due to the sma number of bond issues Source: RBA is strong. As a resut, the average tenor of offshore issuance has been around 8 years, which is much onger than the typica tenor of sighty above 5 years in the domestic market. Bond issuance by Austraian corporations increased foowing the onset of the goba financia crisis, aongside a decine in equity funding and a goba retrenchment in bank oan funding. In recent years, bond funding, and debt funding more generay, have become more attractive for companies because the extended period of ow nomina interest rates has seen the cost of debt funding decine to muti-year ows. The majority of issuance by Austraian resident NFCs has historicay come from companies with a broad A credit rating (A+, A or A-), athough issuance by NFCs with a broad BBB credit rating (BBB+, BBB or BBB-) has increased significanty since 29 (Graph 2). Very few resident Austraian NFCs have been rated AA- or higher, resuting in very itte issuance at these ratings. Issuance of AA bonds by Austraian NFCs occurred with greater frequency in the years preceding the goba financia crisis. These were mainy credit-wrapped bonds bonds issued by ower-rated entities that achieved significanty higher ratings through insurance provided by speciaist bond insurers. These were issued by airports, infrastructure financing vehices and utiity companies. The disappearance of bond 15 BBB+ BBB BBB * For 213, issuance is year to end vember Source: RBA insurers foowing the onset of the goba financia crisis has contributed to increased bond issuance at the ower ratings. As a resut of the historica prominence of offshore issuance, the majority of outstanding Austraian NFC bonds are denominated in foreign currencies, particuary in US doars. By face vaue, around two-thirds of the bonds currenty outstanding are rated A, and this share has increased over time (Graph 3). Cose to haf of the bonds are denominated in US doars, distributed eveny across tenors, whie euro-denominated bonds account for around 2 per cent of bonds outstanding. Most of the remaining outstanding bonds are denominated in Austraian doars, though these tend to be concentrated at the shorter residua maturities (i.e. of ess than 5 years). US doardenominated securities account for an even arger share of the outstanding bonds. Amost a of the bonds outstanding with residua maturities above 7 years are denominated in 16 Reserve bank of Austraia

3 US doars. Austraian doar-denominated bonds are sighty ess than 2 per cent of the tota outstanding at this rating, and are skewed heaviy towards shorter residua maturities. Over time, the vaue and number of outstanding Austraian NFC bonds with onger residua maturities has increased significanty, especiay in the 7 to 1 year range. Graph 3 Austraian NFC Bonds Outstanding * 213* 4 1 Face vaue and credit rating at issuance Number of bonds (RHS) * As at end vember Source: RBA Sampe Construction Residua maturity (years) LHS: n USD n AUD n EUR n Other The paucity of Austraian doar-denominated issuance by NFCs, particuary at onger tenors, makes it impractica to estimate credit curves across a range of tenors soey from domesticay issued bonds. Therefore, the sampe incudes bonds denominated both in Austraian doars and foreign currencies. An important feature of the onger-term bond issuance by Austraian NFCs, and corporate issuers more generay, is the issuance of bonds with embedded options at onger maturities. Refecting this, the sampe incudes buet bonds and bonds with embedded options, such as caabe bonds. 5 The data in the sampe are month-end from January 2 to vember 213, covering bonds with residua maturities over one year. The sampe incudes a bonds identified by Boomberg that were outstanding after 1 January 199 and were issued by non-financia entities incorporated in Austraia. 6 The sampe is restricted to fixed-rate NFC-issued bonds raising the equivaent of at east A$1 miion. The sampe covers bonds denominated in Austraian doars, US doars and euros. A tota of 555 securities met these criteria, comprising 455 buet bonds and 1 bonds with embedded options. 7 Where a US doar-denominated bond ine had both 144A and Reguation S series, the atter were omitted to avoid dupication, as these are effectivey the same bond but issued under different regimes, reducing the sampe by 77 securities. 8 A further seven securities were excuded because of other forms of dupication. 9 The sampe aso excudes a number of securities that were downgraded mutipe notches by credit ratings agencies during, or shorty after, the onset of the goba financia crisis. This meant excuding seven credit-wrapped securities, 5 Buet bonds are redeemabe ony at maturity, whie bonds with embedded options may have significanty different maturities or cashfows, and as a resut different prices and yieds, if the embedded options are exercised. Therefore, the atter require the use of optionadjusted spreads to account for the vaue of their optionaity. 6 n-financia corporations are identified based on their cassification by Boomberg in a group other than banking, commercia finance, consumer finance, financia services, ife insurance, property and casuaty insurance, rea estate, government agencies, government deveopment banks, governments regiona or oca, sovereigns, supranationas and winding-up agencies. 7 The bonds with embedded options incude caabe, convertibe and puttabe bonds. 8 Issuers raising bond funding in US doars can issue two types of securities for the same bond ine that are intended for different investors and cassified as either 144A or Reguation S (Reg S). Securities issued under the US Securities and Exchange Commission s Rue 144A are privatey paced into the US market and are sod to Quaified Institutiona Buyers. Reg S securities are issued in the Eurobond market for internationa investors and are exempt from registration under the US Securities Act Each security type is typicay assigned its own Internationa Securities Identification Number. 9 These incude dupicate securities avaiabe to accredited investors, bonds with warrants and a second series of a bond ine. Buetin DECEMBER Quarter

4 mosty bonds that were originay AA and were issued by airports and utiity companies prior to 28, and one bond which was downgraded to sub-investment grade status in eary 29. The anaysis primariy uses corporate bond price data from Boomberg s BVAL pricing source, which combines information from a number of sources to generate a best-avaiabe data point. 1 These data are suppemented, in order of priority, by Boomberg generic prices and UBS data. 11 This approach is guided by an overarching objective to produce transparent credit spread measures derived from price observations provided by mutipe contributors. The ack of historica bond price data is the main impediment to producing historica credit spread measures. However, the number of vaid observations in the sampe bonds with prices, reported face vaues and avaiabe credit ratings improves consideraby over time as a share of the tota number of bonds outstanding in the sampe (Graph 4). Prior to 25, around 25 per cent of the bond sampe had adequate pricing data (mainy prices avaiabe from UBS). Data avaiabiity increases to around 5 per cent of the sampe between 26 and 29, which argey refects the avaiabiity of Boomberg s BVAL pricing after 28, before improving thereafter to over 8 per cent by 213. The anaysis is conducted directy on corporate bond spreads over corresponding swap rates, which has two advantages: (1) credit spreads to swap can be sourced directy and consistenty from Boomberg; (2) hedging the credit spreads of foreign currency 1 BVAL pricing is Boomberg s best-avaiabe data point for a given instrument at a point in time. The BVAL price methodoogy uses three ayers of price information, appied in the foowing order: (1) directy observed data based on traded market prices is used where avaiabe; (2) where insufficient data are avaiabe, BVAL appies its proprietary correation mode to derive a price from comparabe bonds; (3) a reative vaue yied curve or pricing matrix may be used to derive a price where the correation mode is unabe to generate an estimate. A sma number of negative bond spread observations were excuded from the sampe. 11 UBS data have previousy been used by the Bank to generate the corporate bond spreads and yieds series in Statistica Tabe F3, but are ony avaiabe for Austraian doar-denominated buet bonds. For the new dataset presented in this artice, UBS data are used ony to generate historica estimates before Graph 4 Average Number of Vaid Securities by Credit Rating S&P ratings, average annua number of bonds outstanding AA+ AA AA- Tota number of bonds A+ A A- BBB+ BBB BBB Sources: Boomberg; RBA; Standard & Poor s; UBS AG, Austraia Branch bonds into Austraian doar equivaent spreads, which as expained ater is required for comparabiity within the sampe, requires the cacuation of swap spreads. Traditionay, most bonds issued by Austraian NFCs have been buet securities, where the face vaue of the bond is redeemed at maturity. For these securities, obtaining the spread to swap is straightforward and is cacuated as the security s yied to maturity over the corresponding interpoated swap rate. For Austraian doar bonds, this is the spread over the Austraian doar swap curve, which is an interpoated quartery swap rate for tenors between 1 and 3 years, and a semiannua swap rate for tenors of 4 years and above. 12 The anaysis is compicated by the incusion of bonds with embedded options, where the optionaity affects the underying vaue of the bond and, in turn, its yied and swap spread. This requires the use of an option adjusted spread (OAS), which measures the spread that is not attributabe to the vaue of the 12 Whie not used in this artice, the new measures of credit spreads reported in Statistica Tabe F3 Aggregate Measures of Austraian Corporate Bond Spreads and Yieds, incude the credit spread to Commonweath Government securities (CGS) rates. These are cacuated by adding to the estimated credit spread to swap at each tenor the corresponding swap to CGS spread Reserve bank of Austraia

5 option. 13 The incusion of securities with embedded options increases the sampe of bonds with vaid pricing (Graph 5). Because these securities tend to be over-represented at onger maturities, their incusion assists in producing credit spread measures for onger tenors. Graph 5 Number of Vaid Securities by Optionaity Monthy n Bonds without options n Bonds with make-whoe ca options ony n Bonds with other options* 1 1 The fina sampe captures around 9 securities and around 6 securities on average in 213, with a significant increase in the sampe size over time (Graph 6). The number of unique issuers incuded in the sampe has aso increased over time Graph 6 Number of Vaid Securities by Tenor Monthy Tota unique issuers * These bonds can incude make-whoe ca provisions aong with other optionaity Sources: Boomberg; RBA; Standard & Poor s; UBS AG, Austraia Branch For comparabiity within the sampe, the credit spreads on foreign currency denominated bonds are converted to their Austraian doar equivaent spreads, which corresponds to the foreign exchange risk on the foreign currency bonds being competey hedged (see Appendix A for a discussion of the hedging method). Given the sma number of bonds with a credit rating above AA-, ony bonds with broad A and BBB credit ratings are incuded in the anaysis. Individua bond ratings issued by Standard & Poor s (S&P) are used where avaiabe, and S&P s issuer rating otherwise. 13 Conceptuay, the OAS is the constant spread that has to be added to the spot yied curve in an interest rate option pricing mode to equate the present vaue of a bond s cash fows with its market price. The option pricing mode needs to make assumptions, incuding about interest rate voatiity. Many bonds in the sampe incude optionaity in the form of make-whoe cas. However, the theoretica vaue of make-whoe ca options is sma (Powers and Tsypakov 28) and is not currenty incorporated into Boomberg s OAS spread cacuation. As a resut, the spreads on bonds with ony a make-whoe ca option are the simpe spreads rather than the OAS. Boomberg s estimates of the OAS are used for the bonds with a other forms of optionaity n 1 2 years n 2 4 years n 4 6 years n 6 8 years n 8 1 years n 1+ years Sources: Boomberg; RBA; Standard & Poor s; UBS AG, Austraia Branch Based on the avaiabe data, it is possibe to produce aggregate credit spreads from 25 onwards. The choice of starting point attempts to baance the representativeness of the estimates with a desire to produce a historica time series of reasonabe ength. However, because of the smaer sampe size, estimates generated for the bonds prior to 28, and before ate 29 for the bonds, may be ess representative of the underying market than the estimates obtained for more recent years. Estimating the Austraian NFC Credit Spread Curve Axis abe Pane abe Sub-pane abe Curve abe (Curve sub-abe) Dates n Legend text Sources/footnotes : Robust estimates of credit spreads for a given rating require the avaiabiity of a sufficient number of bonds, distributed widey across tenors. Despite the increase in the number of bonds in the broad A and BBB ratings since the mid 2s, the number of bonds within the fine credit ratings remains reativey ow. As a resut, aggregate credit spreads are estimated separatey at ony the broad A and BBB credit ratings. Buetin DECEMBER Quarter

6 A number of methods are avaiabe to estimate credit spreads. In this artice, aggregate credit spreads of and Austraian NFCs are estimated for a given (target) tenor as the weighted average of the Austraian doar equivaent credit spreads over the swap rate. The method is appied to the cross-section of bonds in the sampe that have the desired credit rating. The weights are determined by a Gaussian kerne that assigns a weight to every observation in the cross-section depending on the distance of the observation s residua maturity and the target tenor according to a Gaussian (norma) distribution centred at the target tenor. 14 This method recognises the fact that the observed spreads on bonds with residua maturities cose to the target tenor contain more information about the underying spread at that tenor than spreads on bonds with residua maturities further away. The advantage of the Gaussian kerne over other more simpistic weighting methods, such as an equay weighted average, is that it uses the entire cross-section of bonds, abeit with weights approaching zero as the distance of the bonds residua maturity from the target tenor increases. This provides a robust method capabe of producing estimates even when the number of avaiabe observations is reativey sma. The advantage of the Gaussian kerne over parametric methods, that have been popuarised in the iterature on the estimation of government yied curves, is its simpicity. Aso, it does not impose a particuar functiona form on the credit spread curve but aows the observed data to determine its shape For a discussion of the Gaussian kerne and kerne methods more generay, see Li and Racine (27). 15 A number of estimation methods were investigated, athough the detais are not reported here. These methods produced very simiar estimates of credit spreads across tenors and broad credit ratings. These methods incuded a range of parametric modes estimated by east squares regressions appied to the cross-section in each period. In particuar, the Neson and Siege (1987) method was examined in detai owing to its wide use in practice for estimating government yied curves (BIS 25); this method has aso been adapted for the estimation of corporate bond yied and spread curves (Xiao 21). However, in our sampe these modes dispayed spurious statistica properties, producing very high mode fit but argey statisticay insignificant coefficients. Other studies have aso found evidence of possibe over-fitting of the data using parametric methods, particuary in the case of the Neson and Siege mode (Annaert et a 213). Gaussian kerne weighting Formay, the Gaussian kerne average credit spread estimator S(T) at target tenor T for a given broad rating and date is: S(T) = Σ N i=1 w i (T; σ)s i where w i (T; σ) is the weight for the target tenor T of the i th bond in the sub-sampe of bonds with the given broad rating, and S i is the observed spread on the i th bond in the sub-sampe of N bonds with the given broad rating. The parameter σ (sigma), which is measured in years, contros the weight assigned to the spread of each observation based on the distance between that bond s residua maturity and the target tenor (sigma is the standard deviation of the norma distribution used to assign the weights). It determines the effective width of the window of residua maturities used in the estimator, with a arger effective window producing smoother estimates. The genera form of the weighting function is: K (T w i (T; σ) = i T; σ)f i Σ N K (T j T; σ)f j j = 1 where K(τ; σ) is the Gaussian kerne function 16 giving weight to the i th bond based on the distance of its residua maturity from the target tenor ( T i T ). 17 F i is the face vaue of the i th bond, which recognises that 16 A kerne function is a symmetric, continuous and bounded rea-vaued function that integrates to 1. When the function is constrained to be non-negative it corresponds to a continuous distribution function. There are a arge number of candidate kernes, with the Gaussian being the most widey used. Linton et a (21) examines the appication of kerne-based methods to the estimation of yied curves and estabishes statistica properties of these estimators. Investigation of a number of other kerne specifications showed that the particuar choice of kerne had itte materia impact on the credit spread estimates. 17 At the end points of the tenor range (1 and 1 years, but particuary at the 1-year tenor), the Gaussian kerne, and other simiar methods, may be somewhat biased because there are no observations beow and above the target tenor. Effectivey, the weighted average is cacuated from bonds on ony one side of the estimation window. At the 1-year tenor, this is aso an issue due to the sparse issuance above 1 years, but is ess probematic for the bonds, for which some observations are avaiabe. However, the degree of bias depends on the true shape of the credit spread curve, with steeper curves resuting in more biased estimates. (1) (2) 2 Reserve bank of Austraia

7 Spread (bps) arger bond issues are typicay more activey traded and are therefore ikey to more accuratey refect market conditions. 18 Finay, the Gaussian kerne is: K (T i T; σ) = 1 (T exp i T ) 2 2π σ 2σ 2 The mechanics of the Gaussian weighting method are iustrated in Graph 7 which shows its appication to estimating the 5-year credit spread for Austraian NFCs. (For carity, the iustration abstracts from the impact of the bonds face vaues on the weights.) The Gaussian kerne assigns positive weights to a bond spread observations with a BBB rating in the sampe s cross-section on the estimation date, but assigns greater weights to the bonds around the 5-year maturity point. This contrasts, for exampe, with the equay weighted average where observations in the sampe are assigned the same weight within some pre-specified range of the residua maturity around the 5-year tenor, but zero weight otherwise. Graph 7 5-year BBB Kerne Weights Credit spread to swap, end August year tenor (3) Weight Optima smoothing of the Gaussian kerne The Gaussian kerne method provides a degree of fexibiity in weighting the observations around the target tenor through the choice of the vaue of the smoothing parameter, σ. There is a natura trade-off between the goodness-of-fit of the estimates, measured as the sum of squared residuas between the observed spreads and the estimated spreads, and the smoothness of the resuting credit curve. 19 Sma vaues of the smoothing parameter produce estimates with smaer residuas by assigning higher weights to bonds cosest to the target tenor. However, when the vaue of the smoothing parameter is too ow the estimates are unikey to be representative of the true credit spread for that tenor, as they refect more of the noise in the observations. Moreover, the estimates are highy variabe (for sma changes in the tenor), and can produce credit spread curves that are quite irreguar and for which there is itte economic justification. 2 Conversey, higher vaues of the smoothing parameter produce ess variabe estimates but may have arger residuas within the sampe. The optima choice of the smoothing parameter can be made objectivey by casting the choice in the same framework as the one for smoothing spines. This invoves expicity trading off the goodness-of-fit of the estimates and the smoothness of the credit spread curve (Anderson and Seath 21). 21 The fina choice of the optima smoothing is aso guided by the economic pausibiity of the credit spread estimates Term to maturity x Credit spreads (LHS) Equa weighting (RHS) Sources: Boomberg; RBA; Standard & Poor s Gaussian weighting (RHS).1 18 The resuts produced by the Gaussian kerne are very simiar when the face vaues of the bonds are not used in the weighting function; that is, when the weights are based ony on the distance between the residua maturities in the sampe and the target tenor. 19 This probem is not unique to credit spreads estimation. Indeed, it is an important consideration in the estimation of government yied curves, where the smoothness of the curve has a direct impact on the quaity of the estimated forward rates, which are often used to provide an indication of market expectations of future monetary poicy. 2 The sope at each point of the credit curve for a singe issuer can be interpreted as an indicator of the instantaneous probabiity of defaut at that point. There are no intuitive or theoretica reasons to expect that the sope of the credit curve, and hence the probabiity of defaut, shoud change significanty for a sma change in the tenor. Athough this argument ony appies to the credit spread curve of a singe entity, it is nonetheess a desirabe feature of an aggregate measure of credit spreads for a given credit rating. In other words, a priori it woud seem reasonabe for the sope of the credit curve to change graduay. 21 A number of risk-free yied curve estimation modes use smoothing spines. For exampes and detais on this approach, see Anderson and Seath (21) and Yaup (212). Buetin DECEMBER Quarter

8 Using the smoothing spine framework, the optima vaue of the smoothing parameter on a singe date is chosen as the one that minimises the sum of squared residuas of the credit spread estimates, whie penaising excessive irreguarity (or curvature ) of the estimated credit spread curve: 2 2 d S(u; σ) min σ (1 λ) Σ N [S(T i ; σ) S i ] 2 1 i=1 +λ 1 du du 2 where S(T i ; σ) is the Gaussian kerne estimate of the spread at the tenor of the i th bond T i, and S i is the observed spread of the i th bond. The first term of Equation (4) measures the goodness-of-fit of the Gaussian kerne estimate, whie the second term measures the curvature of the estimated spread curve. 22 The parameter λ( λ 1) contros the trade-off between the fit and the curvature terms in the objective function, with higher vaues putting more weight on smoothness. There is itte consensus in choosing the trade-off parameter λ. 23 Considering a range of pausibe vaues for λ showed that for vaues above.9 there is itte difference between the optima sigmas. Reativey ow vaues of λ tend to produce optima vaues for the smoothing parameter that appear too sma because they resut in a arge increase in curvature without a significant improvement in the fit. Consequenty, the optima sigma was chosen from the candidates generated by the higher vaues of λ. For the bonds in the sampe, a smoothing parameter of 1½ years is optima and is aso reativey stabe from 28 (i.e. the point after which the sampe size of the bonds increases notaby; Graph 8). The choice of the optima vaue for the smoothing parameter is ess cear for the BBB sampe, with somewhat higher vaues for sigma before In practice, the second derivative in the curvature term is measured by the second difference cacuated over a fine grid of tenors. 23 A popuar choice in the risk-free yied curve iterature is to set λ by minimising the generaised cross-vaidation (see Yaup (212) for an overview). However, even in this setting a choice sti needs to be made about the penaty that is appied to the number of mode parameters and there is itte consensus on this choice. Therefore, the generaised cross-vaidation approach was not pursued in the anaysis. (4) but coser to 1½ years thereafter. In the interests of simpicity, credit spreads are estimated with a fixed vaue of 1½ years for the smoothing parameter throughout the whoe estimation period; that is, the weights around each target tenor are determined from a norma distribution with a standard deviation of 1½ years which is centred at the target tenor. This assigns around 5 per cent of the weight to observations with residua maturities within one year of the target tenor, around 8 per cent within two years and around 95 per cent within three years. The choice of 1½ years for the smoothing parameter is further supported by an examination of the credit spreads produced from a range of vaues for the smoothing parameter (½, 1, 1½ and 2). This reveaed that ower vaues of the smoothing parameter (1 year or ess) produce, at times, counterintuitive crossing of the credit spread estimates of different tenors, whie higher vaues of sigma produce what appear to be excessivey smooth resuts. Sigma Graph 8 Optima Smoothing Parameter (sigma) Tweve-month moving average Lamda.95 Lamda.5 Lamda Sources: Boomberg; RBA; Standard & Poor s; UBS AG, Austraia Branch Austraian n-financia Credit Spread Curves The credit spreads estimated with the Gaussian kerne are ow and quite stabe across tenors prior to 27 (Graph 9). Broady speaking, the estimated spreads since 27 have exhibited the expected movements during episodes of financia stress (e.g. during the 28 9 goba financia crisis) Sigma Reserve bank of Austraia

9 Graph 9 Austraian NFC Spreads to Swap* Austraian doar spreads Graph 1 Austraian NFC Spread Curve Sope 1-year ess 3-year spread to swap BBB curve sope year 5-year 7-year 1-year * Shaded areas indicate episodes of financia distress, which are identified by Boom (29) as the consecutive months during which the deviation of the VIX from its trend is above the 95th percentie; these episodes are: the US subprime crisis (coinciding with the coapse of Lehman Brothers in 28); intensification of concerns over Greek sovereign debt in 21; and a broadening of concerns over euro area sovereign debt in 211 Sources: Boomberg; RBA; Standard & Poor s; UBS AG, Austraia Branch and changes in market expectations (e.g. in mid 213 when financia markets brought forward their expectations for the unwinding of unconventiona monetary poicies by the US Federa Reserve). 24 For the most part, since 25, the credit spread curves of the and Austraian NFCs have tended to shift in parae across tenors. However, at the height of the goba financia crisis in ate 28, when spreads experienced their most significant increase, the credit spread curve steepened sharpy, especiay for the bonds (Graph 1). Despite the decine in credit spreads since mid 212, the BBB credit spread curve has steepened by around 5 basis points, suggesting that investors in bonds have, in recent years, required greater compensation for taking credit exposures at onger tenors. Interestingy, this has been accompanied by an increase in issuance of bonds at onger tenors and the downgrade of a number of previousy issuers to the BBB rating. 24 The episodes of financia market stress are identified from the deviation of the option-impied voatiity of the S&P 5 index (VIX) from its traiing average (Boom 29). It is not possibe to determine with certainty whether the dramatic increase in BBB spreads in ate 28, particuary at the shorter tenors, is overstated owing to the sma sampe size around this time. During this period, the sampe is heaviy infuenced by the sharp widening of the bond spreads for a major diversified mining company, having considerabe infuence on the spread estimates around the 5-year tenor A curve sope Sources: Boomberg; RBA; Standard & Poor s; UBS AG, Austraia Branch Overa, the Gaussian kerne method produces effective weighted average tenors that are very cose to each of the target tenors (Graph 11). The exception is the 1-year tenor where the effective tenor is coser to 9 years. This refects the dearth of issuance of bonds with tenors of 1 years or more. twithstanding the sighty shorter effective tenor for the 1-year point, the estimates of the 1-year spread from the Gaussian kerne are distinct from the estimates of the 9-year spread as the two are estimated by appying different weights to the bonds in the sampe. There are very few aternative measures of Austraian credit spreads against which the Gaussian kerne estimates can be compared (Graph 12 and Graph 13). 25 The Gaussian kerne estimates for bonds have been consistenty beow the credit spread series in the Bank s previous Statistica Tabe F3 since mid 27 because the atter incude (non-bank) financia corporations, such as rea estate investment trusts (REITs). These bonds have tended to have higher credit spreads for the same rating than non-financia entities since 27. The new measures improve the previous series in the Bank s Statistica Tabe F3 25 The avaiabe aternative measures are the 1 5 year credit spreads previousy pubished by the RBA in Statistica Tabe F3 and Boomberg s proprietary fair vaue curves. Currenty, Boomberg s fair vaue curve indices for Austraian and corporate bonds are avaiabe up to a maximum tenor of 7 years, with historica data starting in the eary 2s; the Boomberg indices are produced using a method which is not discosed pubicy in detai. Buetin DECEMBER Quarter

10 Yrs Graph 11 Effective Tenor 1-year 7-year 5-year 3-year Sources: Boomberg; RBA; Standard & Poor s; UBS AG, Austraia Branch Graph 12 Austraian NFC Spread to Swap Comparison RBA Statistica Tabe F3 (1 5 year) 29 5-year 213 Gaussian Boomberg AUD Yrs Boomberg USD Sources: Boomberg; RBA; Standard & Poor s; UBS AG, Austraia Branch Graph 13 Austraian NFC Spread to Swap Comparison Gaussian 1-year by significanty expanding the sampe, separating financia and non-financia corporations, estimating spreads across different maturities, and by using a more robust method that gives greater weight to more representative observations. The new measures of credit spreads for the and bonds are simiar to corresponding measures produced by Boomberg prior to ate 28. However, they have diverged from the Boomberg series since then, which refects in part the counterintuitive behaviour of the Boomberg spreads between 29 and The new credit spread measures presented herein have a number of advantages. First, the method of the construction is more transparent. Second, the sampe is arger due to the incusion of bonds issued in foreign currencies. Third, the method used is reativey robust, aowing for the estimation of spreads at onger maturities than are avaiabe esewhere. The Gaussian kerne estimates generay track the movement of spreads in goba credit markets, in terms of both the timing and severity of their reaction to episodes of financia stress. The eves of the Gaussian kerne estimates of Austraian NFC spreads have diverged from their US equivaents since mid 211, with US credit spreads decining even further than Austraian NFC spreads since then. Despite this divergence in eves, the co-movement between Austraian NFC spreads estimated with the Gaussian kerne and the corresponding US spreads has remained high. Concusion This artice presents a method for estimating aggregate credit spreads across tenors ranging up to 1 years for Austraian NFCs. The estimation method is simpe, transparent and robust in sma sampes. The Bank wi commence pubishing monthy Boomberg USD Sources: Boomberg; RBA; Standard & Poor s; UBS AG, Austraia Branch The Boomberg Austraian doar fair vaue curve appears to be overy smooth between eary 29 and ate 21. These measures did not increase as much as coud be expected in eary 29, given that the goba financia crisis was at its most severe at that time, and as was observed in other measures of Austraian and foreign corporate bond spreads. Moreover, the Boomberg spread measures remained eevated for an extended period of time between eary 29 and 21, whie credit spreads gobay decined sharpy foowing the introduction of extraordinary poicy measures; this was especiay true of bond spreads. 24 Reserve bank of Austraia

11 credit spreads estimates from December 213. The newy constructed credit spread measures wi provide richer information than is currenty avaiabe pubicy, aowing the pubic researchers, investors, reguators and others to examine deveopments in the Austraian credit market in more detai. R Appendix A: Hedging Foreign Currency Bond Spreads To provide comparabiity within the sampe between bonds denominated in different currencies, the anaysis converts spreads of foreign currency bonds into their Austraian doar equivaent. This is consistent with corporate bond investors taking a view on credit and interest rate risk ony and fuy hedging the foreign exchange risk associated with hoding foreign currency bonds using derivatives. Moreover, constructing estimates that assume a borrowers hedge in a consistent way aows the anaysis to abstract from differences in the methods of hedging (incuding the use of natura hedges) and the extent of the coverage of hedging. The conversion to Austraian doar equivaent spreads reies on the existence of a we-functioning cross-currency swap market, which is the case for Austraia. Estimating the cost of hedging Estimating the cost of hedging foreign-issued bonds into their Austraian doar equivaent spread invoves a number of stages. This process is ony intended as an approximation of the materia costs invoved. Briefy, these are: Cross-currency basis swap: used to convert foreign currency payments into Austraian doars. This is generay the most significant hedging cost. Interest rate swap: a basis swap is used to hedge between semiannua coupon payments and the 3-month foreign currency interbank rate, which is typicay used as the benchmark for cross-currency basis swap contracts. A basis swap is aso used to convert the resuting 3-month Austraian doar equivaent spread to a 6-month equivaent spread for comparabiity with the semiannua coupons on Austraian doar bonds. Conversion factor: adjusts for interest rate differentias when cacuating the spread between benchmarks denominated in different currencies. The conversion factor is the ratio of price sensitivities, which transates the reative vaue of a one basis point change in the interest rate of one currency into the change in another. 27 The impact of hedging foreign currency bonds The Austraian doar cross-currency basis swap accounts for most of the foreign currency hedging cost. However, since 28 the cost of receiving a 3-month foreign currency interbank rate in exchange for a 6-month interbank rate has become a arger component of tota hedging costs. To iustrate the evoution of hedging costs, a hypothetica 1-year constant maturity foreign currency bond trading at a foreign currency swap spread of 1 basis points, is hedged from both US doars and euros. The tota hedging cost has been reativey stabe since 21 at around 5 basis points for a theoretica US doardenominated bond (Graph A1) and 7 basis points for a euro-denominated bond (not shown). 28 Comparing a seection of bonds aso indicates that hedging foreign currency bonds into their Austraian doar equivaent tends to aign the spreads more cosey with those of comparabe Austraian doar bonds. There are ony a few instances of comparabe pairs of (matched) bonds from the same borrower: bonds with simiar residua maturities and issued in a foreign currency and Austraian doars. Despite the sma number of bonds in the sampe avaiabe for comparison, the avaiabe observations suggest that the difference between the Austraian doar 27 The conversion factor is approximated by the ratio of changes in present vaue from a one basis point shift in the swap curve at a given tenor for each currency. This approximation, aso known as the PV1 (the present vaue of a one basis point shift in the swap curve), is commony used by market participants in practice. 28 The anaysis assumes that hedging a euro-denominated bond into Austraian doars first requires euro-denominated cash fows to be converted into US doar cash fows (incorporating the three factors outined above), from which Austraian doar equivaent spreads can be estimated. The additiona stage refects the fact that cross-currency basis swaps not invoving a US doar eg tend to be iiquid, and market practice is to hedge first into US doar cash fows. Buetin DECEMBER Quarter

12 bond spreads and the hedged foreign currency spreads on comparabe bonds tends to be smaer than the difference if the foreign currency bond was unhedged (Graph A2). Graph A1 Decomposition of Hedging Adjustment* Theoretica US doar-denominated bond References Anderson N and J Seath (21), New Estimates of the UK Rea and mina Yied Curves, Bank of Engand Working Paper 126. Annaert J, AGP Caes, MJK De Ceuster and H Zhang (213), Estimating the Spot Rate Curve Using the Neson- Siege Mode: A Ridge Regression Approach, Internationa Review of Economics and Finance, 27, pp USD to AUD conversion factor Tota hedging cost BIS (Bank for Internationa Settements) (25), Zerocoupon Yied Curves: Technica Documentation, BIS Papers 25. Back S, J Kirkwood, A Rai and T Wiiams (212), A History of Austraian Corporate Bonds, RBA Research Discussion Paper Testra Wesfarmers AUD cross-currency basis swap 29 Spread to USD LIBOR * Theoretica 1 year constant maturity bond; assumes constant foreign currency spread to swap of 1 basis points Sources: Boomberg; RBA Graph A2 Comparison of Bond Pairs Matched bonds A$ spread to swap Differences EUR* Ju 215 A$ Apr 215 EUR* Ju 215 A$ Sep 214 Hedged spread** Unhedged spread*** Boom N (29), The Impact of Uncertainty Shocks, Econometrica, 77(3), pp Li Q and JS Racine (27), nparametric Econometrics: Theory and Practice, Princeton University Press, Princeton. Linton O, E Mammen, J Niesen and C Tanggard (21), Yied Curve Estimation by Kerne Smoothing Methods, Journa of Econometrics, 15, pp Neson CR and AF Siege (1987), Parsimonious Modeing of Yied Curves, The Journa of Business, 6(4), pp Powers E and S Tsypakov (28), What is the Cost of Financia Fexibiity? Theory and Evidence for Make-whoe Ca Provisions, Financia Management, 37(3), pp Xiao J (21), Term Structure Estimation for U.S. Corporate Bond Yieds, RiskMetrics Journa, 2(1), pp Yaup PJ (212), Modes of the Yied Curve and the Curvature of the Impied Forward Rate Function, Journa of Banking and Finance, 36(1), pp Wooworths15 5 US$* Apr 216 A$ Mar * Austraian doar equivaent spread of the foreign currency bond ** Difference between the spread of the Austraian doar bond and the hedged spread on the foreign currency bond *** Difference between the spread of the Austraian doar bond and the unhedged spread on the foreign currency bond Sources: Boomberg; RBA Reserve bank of Austraia