The impact of sovereign split ratings between global credit rating agencies on foreign exchange market

Nội dung text: The impact of sovereign split ratings between global credit rating agencies on foreign exchange market

1. THE IMPACT OF SOVEREIGN SPLIT RATINGS BETWEEN GLOBAL CREDIT RATING AGENCIES ON FOREIGN EXCHANGE MARKET Do Thi Thu Ha - Do Cam Nhung*1 ABSTRACT: The behaviour of foreign exchange market in reaction to split sovereign ratings, which is hardly mentioned on recent studies, is a topic of concern, especially in the post-crisis period. We analyse difference of opinion in sovereign credit ratings and their influence on currency market. The first question relates to the response of exchange rates to sovereign credit rating changes. Both negative and positive rating announcements have impacts on the market though the influence of the former events are much more significant than the later ones. The second question investigates how split sovereign ratings affect the information content of subsequent sovereign rating events for global foreign exchange market. The behaviours of spot exchange rate are concerned. The findings reveal that exchange rate spreads are significantly responsive to S&P negative events with lower pre-event ratings than either Moody’s and Fitch, while only Moody’s positive events with prior split ratings versus S&P significantly influence the foreign exchange rates. Also, Fitch actions play as a “tie-breaker’’ in currency market, hence split ratings involved Fitch significantly reduce the market uncertainty and volatility. Additionally, exchange rate volatility is only responsive to S&P rating announcements with pre-event split ratings. Also, Fitch actions play as a “tie- breaker’’ in currency market, hence split ratings involved Fitch significantly reduce the market uncertainty and volatility. Keywords: Credit rating, foreign exchange market, sovereign rating 1. INTRODUCTION The 2007-2009 global financial crisis and the following European sovereign debt crisis has put credit rating agencies (CRAs) on spotlight of public attention. As the economic uncertainty has raised a huge demand for sovereign credit ratings, the crisis and post-crisis period witnessed more actions of sovereign credit rating compared with pre-crisis period (Bửninghausen and Zabel, 2015). Also, during the Euro debt crisis, downgrade actions from CRAs are criticized to exacerbate the severity of the crisis (Alsakka and ap Gwilym, 2013). These facts raise a question about the market impact of credit rating announcements, particularly the market response following sovereign rating changes, given their decisive role in credit rating market and the stability of global financial market (ESMA, 2013). The influences of sovereign ratings on bank and corporate ratings are analysed by Williams et al. (2013), and Borensztein et al. (2013) respectively. Prior studies also reveal empirical evidences of sovereign credit signals exerting significant impact on bond market, equity market and CDS market (e.g. Gande and Parsley, 2005; Ferreira and Gama, 2007; Hill and Faff, 2010; Alsakka et al., 2017). * Banking Academy of Vietnam, 12 Chua Boc, Dong Da, Hanoi 10000, Vietnam.
2. 218 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA During the two crises, the behaviour of currency market in reaction to news is also a topic of debate. The foreign exchange rate is considered to be highly sensitive to news, therefore experiences unexpected volatility during crisis time (Evans and Lyons, 2015). Accordingly, the value of the US dollar during the global financial crisis - which is driven by negative news related to the US economy - has strong impact on global currency market (Fratzscher, 2009). Most governments and corporates consult multiple rating assigned by at least two leading CRAs, in order to enhance their ranking in the market. This creates the phenomenon of split rating which is disagreement on rating opinions of two CRAs for the same issuers at the same time. However, prior empirical studies remain silent on the influence of split ratings on spot foreign exchange market – the largest and most liquid financial market in the world (BIS, 2013). These issues are addressed in this paper. The study analyses the sovereign credit actions including rating changes, outlook and watch signals from the big three global CRAs: Moody’s, S&P, and Fitch. The study reveals the impact of rating actions on global foreign exchange market varies across CRAs. By using event study and fixed effect modelling approach, the investigation focuses on the effect of pre- event split rating on foreign exchange rates and volatility. The first research question considers the response of foreign exchange market to sovereign credit rating announcements. Accordingly, negative rating events are found to exert stronger impact on exchange rates than positive events. The findings support prior literature that negative actions are more informative to financial markets, therefore generate stronger market reaction than positive ones. It is also revealed that the most significant dynamic of the foreign exchange rate is recorded after negative events from S&P, and positive events by Moody. This is in line with Alsakka and ap Gwilym’s (2010) suggestion that S&P tends to be the “first mover” in downgrade actions, whereas Moody’s positive actions likely lead the reaction of the market. Arising from the first baseline question in the perspective of split ratings, the second question addresses the issue of whether split ratings contain valuable information, whereby they influence the impact of subsequent rating events on foreign exchange market. The pre-event inferior rating of S&P than Moody’s and Fitch is found to substantially affect the information content of S&P negative actions. Also, the market impact of Moody’s positive events is significantly affected by Moody’s prior rating which is superior than S&P action. The study reveals indicative evidences that split rating across CRAs is informative and affects the response of exchange rate volatility to following rating events. The magnitude of the effect depends on the level of market’s uncertainty after rating events. The more the event is predicted, the stronger the influence of pre-event split rating is. 2. LITERATURE REVIEW 2.1. Sovereign credit ratings and market impacts Sovereign ratings are the credit ratings for sovereign issuers. It reflects the long-term default risk of a sovereign. In other words, it measures the ability and the willingness to make payments of a government. According to SEC (2016), credit ratings for government securities accounted for 80.95%, 81.13% and 72.06% of total ratings of Moody’s, S&P and Fitch, respectively. Prior studies for emerging countries show that sovereign credit signals can affect the international bank flow into stock and bond markets (Kim and Wu, 2011). Also, sovereign credit ratings are able to attract attention of foreign investors and help the emerging governments to present financial transparency, whereby sovereign ratings make ways for private sectors to enter global capital market (Alsakka and ap Gwilym, 2009).
3. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 219 In addition, sovereign credit ratings impose ‘ceiling effect’ in the credit rating industry. It act as an upper bound and a determinant of credit rating for corporates and banks (Borensztein et al., 2013, Almeida et al., 2017). Tran et al. (2014) show that CRAs have downgraded numbers of European banks during years from 2011 to 2013, which is explained directly from the descent in the Europe governments’ financial capacities. In terms of empirical studies, the research attention is brought to the linkage between sovereign rating events and the pricing of debt instrument such as derivatives (Afonso et al., 2012), spot foreign exchange rate (Alsakka and ap Gwilym, 2012b) and stock index option (Tran et al., 2014. Prior literatures distinguish between negative and positive rating signals. Following that, the former signals have significant impact on stock and bond markets, while the market impact of the latter one is negligible. For instance, Brooks et al. (2004) show that the stock market indices are only responsive to sovereign rating downgrades while there is no evidence about abnormal returns triggered by upgrade events. The asymmetric effect of these signals is thought to be due to stronger negative reputational impacts for a CRA which gives a later downgrade announcement compared with other CRAs (Alsakka and ap Gwilym, 2010). In fact, debt issuers are more likely to publish positive news prior to an upgrade rather than showing a potential downgrade by leaking negative information to the market, therefore the negative credit announcements become more informative than the positive news (Ganda and Parcel, 2005). Furthermore, the cross-border transmission effect, that is, a sovereign rating change in one country significantly influences sovereign credit spreads of other countries, is found by Gande and Parsley (2005), and Ferreira and Gama (2007). The cross-country spill-over effect whch is only triggered by negative rating event (but not positive new) is confirmed to be asymmetric (Gande and Parsley, 2005) 2.2. Split ratings and market impacts Split rating denotes the situation when a specific sovereign is assigned different ratings by CRAs. In case of split rating, the rating of a CRA can be either higher or lower than others. Livingston et al. (2010) use ‘superior’ and ‘inferior’ to refer this comparison. As far as the market impact of corporate split ratings is concerned, Livingston et al. (2010) indicates that, to price the corporate credit risk, the CRA’s rating are separately considered but the opinion of the more conservative CRA (Moody) is more heavily weighted. Bongaerts et al. (2012) show that, in case where corporate issuers have divergence in credit opinions between S&P and Moody’s, the third rating from Fitch is found as a ‘tie-breaker’ of split ratings. In addition, split rated bonds with higher level of risk are more expensive than the non-split rated bonds. Santos (2006) shows that split ratings significantly affect bond yields and the effect is dependent with general economy’s condition and the credit quality of issuers. Particularly in adverse economic environment, issuers with average credit quality may incur additional cost which increase the bond’s price. Also, in case of split ratings caused by information opacity in the corporate bond market, Fitch ratings add information which is valued and priced by both investors and issuing firms (Livingston et al., 2016). In terms of sovereign split ratings, Vu et al. (2015) find empirical evidences about the strong impact of split sovereign rating on bond spreads. The study separates split rating event into 2 groups which is (i) positive (negative) events on the superior ratings (inferior ratings) and (ii) positive (negative) events on the inferior ratings (superior rating). It is suggested that the former types of events convey information which is more valuable to the market participants than the latter types of event, therefore have stronger impact on
4. 220 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA bond market. Generally explaining for the effect of split sovereign ratings on bond spread, Vu et al. (2015) suggests that split ratings which widen following a downgrade event on inferior ratings are associated with an increase in uncertainty about sovereign default risk, then the market reacts to inconsistent information as a result. This suggestion is consistent with a viewpoint of investors’ behaviour under uncertain- quality news and ambiguity aversion in Epstein and Schneider (2008). It is shown that when market information is put in uncertainty, investors acts as if they take a worst-case earnings forecast returns, hence bad news often induce stronger investors’ reactions than good news. 2.3. Empirical evidences for an existence of credit rating event’s impacts on spot foreign exchange market Alsakka and ap Gwilym (2012b) analyse the response of foreign exchange spot market to sovereign credit signals during 1994- 2010. Focusing on emerging countries, they indicate the different reaction to sovereign credit signal of foreign exchange market compared with stock and bond markets. Significant responses of exchange rate to upgrade announcements and positive watch signals is found in emerging exchange markets, while all other empirical studies find little evidence about the response of bond and stock markets to upgrades (e.g. Sy, 2002; Ferreira and Gama, 2007). Furthermore, the strongest reactions of the foreign exchange market are in response to Fitch ratings. This opposes with Brooks et al. (2004) suggestion that S&P actions give greatest influences over stock market, while the effect of Fitch ratings is minimal. The asymmetric influence of CRAs to spot foreign exchange market (particularly to developed markets) are affirmed by Alsakka and ap Gwilym (2013). It is also indicated that the magnitudes of effects during crisis period are far stronger compared to pre-crisis time. Furthermore, sovereign creditworthiness is a strong explanatory variable for foreign exchange volatility (Bisoondoyal-Bheenick et al., 2011). Tran et al. (2014) offers new insights by estimating the impact of multiple sovereign ratings on market volatility including the movement of spot foreign exchange market and forward exchange market. Accordingly, sovereign rating news have a statistically significant impact on the market volatility, and additional rating signals are able to help reduce the market uncertainty. About the economic rationale of sovereign rating impact on foreign exchange market, Tran et al. (2014) suggest that, the casual linkages between sovereign credit ratings, the fiscal indices of a country and its exchange rate is the main rationale for sovereign rating to have impact on a country’s foreign exchange market. Despite using different method to rate the creditworthiness of a government, CRAs generally determine their sovereign rating based on a country’s fiscal condition (Moody’s, 2016; S&P, 2016). Taking S&P for example, there are five main criteria considered to determine a sovereign credit rating, which are political, external, economic, monetary and fiscal factors. In another words, changes in a country’s economic condition are associated directly with its rating level assigned by CRAs (Bennell et al., 2006; Afonso et al., 2011). On the other hands, the sovereign rating level is considered to influence back to fiscal indices of a country. For example, empirical studies emphasize that sovereign ratings are an important factor explaining the change of sovereign bond yields which are benchmarks to a country’s market risk (Kaminsky and Schmukler, 2002). In addition, theoretical models predict a positive relationship between a fiscal expansion and the real exchange rate (Erceg et al., 2005). Particularly, a government budget in balance is expected to appreciate the real exchange rate. Empirical studies also take the economic linkage between exchange rate and fiscal health of a country into consideration. As opposed to other theoretical models, Kim and Roubini (2008)
5. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 221 empirically evidence that a fiscal policy shock, or a national budget deficit shock in the USA can appreciate the US dollar, hence depreciating the real exchange rate. In general, based on the given economic interaction between sovereign credit rating, fiscal conditions and exchange rate, the study aims to find out a strong impact of sovereign rating signals on the foreign exchange market. 3. DATA SAMPLE The chosen dataset is an unbalanced data panel of 52 countries with a period ranging from January 2008 to July 2017. The disagreement in sovereign rating opinion of three world leading CRAs (Moody’s, Standard and Poor’s and Fitch) is examined. There are 2 numerical rating scales which are (i) 18-notch rating scale which only considers changes in rating levels (ii) 52-point comprehensive credit rating scale (CRR) that counts outlook and watch in addition to actual rating. Following CCR scale, rating symbols are transformed into number as from 1 to 52. Then, 1 and 2 are added as adjustment for positive outlook and watch status respectively. In contrast, for negative outlook and watch announcement, the actual rating point is substracted 1 and 2 respectively. Accordingly, for every solo downgrade or upgrade, the CCR is adjusted by 3 points, whereas, combined events of rating changes and outlook (watch) signals might cause a change of 4 to 5 units in CCR point. Generally, negative events and positives events are respectively attributed to the decreases and increases in the CCR measured using the 52-point rating scale. There are 4 different types of events. The first type including solo downgrades and upgrades reveal the changes of ratings expressed by number of notches only. The next two types are the solo events of outlook and watch without any changes in rating level. Negative (positive) outlook signals covers the cases of (i) a sovereign’s outlook changes from stable to negative (positive) and (ii) a sovereign’s outlook changes from positive (negative) to stable. Negative (positive) watch signals arise from (i) cases when a sovereign is under a review for possible downgrade (upgrade), and (ii) the action of confirming rating of a sovereign which is on watch for potential downgrade (upgrade). The forth type is combined events of downgrades (upgrades) with outlooks or watch signals. Table 1. Summary of the sovereign rating data sample January 2008 - July 2017 Moody’s Fitch S&P Total 1 Solo rating downgrades 21 29 32 82 2 Solo negative watch signals 27 16 37 80 3 Solo negative outlook signals 40 52 57 149 4 Combined events of rating downgrades and watch 8 5 12 25 5 Combined events of rating downgrades and outlook 45 45 54 144 6 Negative events 141 147 192 480 7 Solo rating upgrades 39 45 54 138 8 Solo positive watch signals 7 4 1 12 9 Solo positive outlook signals 50 47 69 166 10 Combined events of rating upgrades and outlook 10 5 9 24 11 Combined events of rating upgrades and watch 1 0 0 1 12 Negative watch to negative outlook signal 5 3 11 19 13 Positive events 112 104 144 360 14 Total credit events (rows 6 + 13) 253 251 336 840 15 All rating downgrades (rows 1 + 4 + 5) 74 79 98 251 16 – of which by >1-notch (% row 15) 37.84% 27.85% 20.41%
6. 222 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA 17 All rating upgrades (rows 7 + 10 + 11) 50 50 63 163 18 – of which by >1-notch (% row 17) 6% 14% 9.52% 19 1-point negative action 55 64 82 201 20 2-point negative action 35 28 49 112 21 3-point negative action 14 27 37 78 22 >3-point negative action 37 28 24 89 23 Negative actions using 52-point scale 141 147 192 480 24 1-point positive action 61 51 81 193 25 2-point positive action 31 19 28 78 26 >2-point positive action 20 34 35 89 27 Positive actions using 52-point scale 112 104 144 360 Table 2. Agreement/ disagreement on the 52 sovereign ratings. January 2008 - July 2017 S&P and S&P and Moody’s and Moody’s Fitch Fitch Panel I –18-notch rating scale Daily observation 130000 130000 130000 Split % of whole sample 49.59% 46.85% 44.86% Higher rating from first CRA; % of split 34.98% 44.74% 59.56% 1-notch higher rating from first CRA 18130 25456 27680 > 1-notch higher rating from first CRA 4424 1796 7052 1-notch lower rating from first CRA 32354 28377 18304 > 1-notch lower rating from first CRA 9565 5281 5277 Pane lI – 52-point rating scale Daily observation 130000 130000 130000 Split % of whole sample 64% 59.61% 59.75% Higher rating from first CRA; % of split 33.83% 43.81% 59.27% 1-point higher rating from first CRA 6246 6898 8994 2-point higher rating from first CRA 7983 5542 8184 3-point higher rating from first CRA 8688 17227 17389 4-point higher rating from first CRA 811 2542 4009 5-point higher rating from first CRA 680 733 1603 > 5-point higher rating from first CRA 3745 1007 5848 1-point lower rating from first CRA 13289 10399 8259 2-point lower rating from first CRA 7171 5185 6589 3-point lower rating from first CRA 21207 19587 9525 4-point lower rating from first CRA 4175 3584 1953 5-point lower rating from first CRA 1232 1218 1174 > 5-point lower rating from first CRA 7976 3576 4144
7. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 223 As can be seen in Table 1, there are more negative events (480 events) than positive events (360 events) issued by three CRAs throughout the sample period. S&P is also the most active CRAs with 336 events. Besides, Moody takes stronger actions than other CRAs in terms of downgrades, when 37.84% of rating downgrades assigned by Moody’s is multiple-notch adjustments. Besides, multiple-notch downgrades occur more frequently than multiple-notch upgrades by all three of CRAs. For example, in Fitch, 27.85% of all rating downgrades is multiple-notch change, whereas the figure for multiple-notch upgrade is 14%. Credit actions based on the 52-point scale are summarized in Row 19-27 of Table 1. As can be observed that, the majority of the action are 1-point change with 201 out of 480 negative actions (41.87%) and 193 out of 360 actions (53.61%). Table 2 reports split ratings between CRAs. There are 130,000 daily observation for each CRAs. It can be seen from Row 2 of the Table 3.2 that split ratings occur frequently across CRAs. According to 18-notch rating scale, 49.59% (46.85%) of daily observation show the divergence in credit opinions between S&P and Moody’s (Fitch), and the percentage of split ratings between Moody’s and Fitch is 44.86%. As outlook and watch status are taken into consideration, using the 52-point rating scale reveals more rating disagreements across CRAs. The split rating between S&P and Moody’s accounts for 64% of the whole sample. The average rating difference between two CRAs in each group is approximately 3 CCR units, which is equivalent to one notch difference on the 18-notch rating scale. The rating differences vary from one CCR unit to 18 CCR units (6 notches). Table 2 shows that S&P appears to assign more inferior ratings than superior ratings compared with either Moody’s or Fitch, therefore the S&P negative signals are expected to be more informative to the foreign exchange rate than the other two CRAs. Moody’s is likely to assign higher ratings than the other CRAs and the number of positive ratings assigned by Moody’s is highest of the three CRAs, therefore positive signals from Moody’s are expected to trigger stronger impact on the foreign exchange rate than positive rating from S&P and Fitch. The dataset of daily exchange rates comprises 52 currencies which are named in BIS (2016). Due to the dominance of US dollars which accounts for 88% of foreign exchange market’s turnover (BIS, 2016), all the exchange rates is based on the US dollars. Daily data of exchange rate and exchange rate volatility are obtained from Data Stream (the primary source is Thomson Reuters). The exchange rates are denoted with direct quotation which refers to domestic currency units per one US dollar. Also, the study use the natural logarithms of the exchange rates, following prior studies. The exchange rate per US dollar is measured over short window to control for the effect of event clustering (Gande and Parsley, 2005), and the exchange rate return is calculated in [-1,+1] window as follows: ∆EXt = 1000 x (Ln (Exchange rate)t+1 – Ln (Exchange rate)t-1) The study also uses 1-month maturity volatility to capture the volatility of foreign exchange market. The return of exchange rate volatility is estimated as follows: ∆EVt = 1000 x (Ln (Exchange rate volatility)t+1 – Ln (Exchange rate volatility)t-1) 4. METHODOLOGY To examine the behaviour of spot foreign exchange market in response to sovereign credit events with pre-event rating differences, the study employ a multivariate regression model framework, and a baseline regression model is constructed as follows: ∆ ∆EXi,s= α + β it + γ2* it-1 + γ3* PRIOREVENTSit + γ4* VIXit + θ*yt + à*countryi + εit (1) 퐂퐂퐑 퐂퐂퐑
8. 224 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA ∆EVi,s= α + it + γ2* it-1 + γ3* PRIOREVENTSit + γ4* VIXit + θ*yt + à*countryi + εit (2) The baselineβ ∆퐂퐂퐑Eq. (1) with 퐂퐂퐑the dependent variable is ∆EXi,s measures the impact of sovereign credit rating change on the movement of exchange rate (∆EXi,s), while the Eq.(2) caculates the effect on exchange rate volatility ( t). ∆EXi,s is 휎퐑the daily return of exchange rate of sovereign i around the event date t, in the [-1,+1] time windows where date -1 is the previous working day of the event date, and date +1 is the next working day of the event date. The exchange rate are denoted with direct quotation, hence the increase in ∆EXi,s represents a depreciation of the domestic currency against US dollar. i,s is the log-change in the interest rate volatility of sovereign i in [-1,1] time window ∆푬푽 it is the change of the sovereign ratings of country i at time t based on 52-point rating scale. Slope∆퐂퐂퐑 β illustrates the movement of the exchange rate in response to one point change in CCR. it-1 is the comprehensive credit rating assigned to sovereign i before the day of credit event (day t). It is used as a control variable which refers to economic, political and market conditions of sovereign i. 퐂퐂퐑 PRIOREVENTSit is the cumulative change of CCR in one month prior to day t (30 days before a certain credit rating event). This is used as a variable controlling the event clustering. A high probability of upgrade is indicated through net positive changes of CCR in 30 days before even date, whereas net negative changes of CCR refer a downgrade trend. PRIOREVENTSit is equal to zero in case where events is not classified as downgrade or upgrade within one month horizon. VIXit: is a proxy of global risk aversion. It is defined as the contemporary logarithmicchanges within the window of [0, +1] where date 0 is the event date and date +1 is the next trading day. Finally, the model use year dummy variable yt for the fixed-effect of time trends, and country dummy variable countryi,t for the fixed-effect of countries. The baseline models are applied in turn to negative events and positive events by each CRA. The models generally estimate whether the foreign exchange market is responsive to sovereign credit rating events, and whether the effect is heterogeneous across CRAs. This model does not account for the pre-event differences between ratings of CRAs. As been suggested in prior literatures, a pre-event split rating between CRAs is influential in the magnitude of how a rating change impact markets on the event date. The pre- event rating which is different from the rating of another CRA (second rating) can be inferior (lower) rating or superior (higher) rating than the second rating. To examine the effect of pre-event ratings on the market impact of rating events by three CRAs, the study estimates a ‘differential model’ as follows: EXi,s 1* 2* 2* 3* PRIOREVENTS 4* VIX (3) ∆ =t α + β ∆퐂퐂퐑i it *it SUPit + β ∆퐂퐂퐑it * INFit + γ 퐂퐂퐑it-1 + γ it + γ EV * * * * PRIOREVENTS * it + θ*yi,s + à*country1 + ε 2 2 3 4 VIX (4) ∆ =t α + β ∆퐂퐂퐑i it *it SUPit + β ∆퐂퐂퐑it * INFit + γ 퐂퐂퐑it-1 + γ it + γ itEquation + θ*y + (3) à*country measures+ the ε movement of spot exchange rate following split rating across CRAs, while the Equation (4) estimates the behaviour of exchange rate volatility. The rating change variable (∆CCRit) interacted with two dummy variables namely SUPit and INFit which respectively reflect superior rating (higher rating) and inferior rating (lower rating) compared with another CRA. SUPit is equal to one (INFit is equal to zero) if the CRA assigned a higher pre-event rating than another CRA, and if the CRA assigned an inferior rating before event date, the value of SUPit is zero (INFit is one). The variable CCRit is defined as the average of ratings of two CRAs whose pre-event ratings are different.
9. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 225 Following prior literatures (Ferreira and Gama, 2007; Vu et al., 2015), non-event data is added in the sample. These data point is randomly collected during the horizon sample to match with event date. The 61-day window is employed to choose non-event days which are not preceded (followed) by a rating event within 30 days before (after) that non-event day. The number of non-event dates is equal to the event dates, whereby the number of observations is double. 5. EMPIRICAL RESULTS 5.1. The baseline model The empirical results for Eq. (1) are summarized in Table 3. At 1% significance level, exchange rates are responsive to S&P negative signals, while Moody’s negative signals explain the exchange rate movement at the significance level of 5%. As been shown in the data description, S&P tends to rate lower than Moody’s and Fitch, therefore the result is consistent with Livingston et al. (2010) suggestion that the market perception of sovereign risk is driven by the most conservative CRA. Also, Fitch signals have insignificant impact on foreign exchange, which consistent with prior researches (e.g.V u et al., 2015). In terms of positive signals, only positive news assigned by Moody significantly induces market reactions. This is consistent with the result of Alsakka and ap Gwilym (2012b) that foreign exchange market significantly react to positive credit signals, while prior studies find little evidence about the response of bonds and stock market to upgrade signals (Gande and Parsley, 2005; Ferreira and Gama, 2007). Table 4 represents the empirical result of Eq (2). It is noteworthy that exchange rate volatility is highly sensitive to negative credit events from CRAs. The consistent signs of coefficient estimates of S&P’s and Moody’s negative events is different from the results of Tran et al. (2014) that Moody’s rating news tend to reduce the foreign exchange, as opposed to S&P. Tran et al. (2014) suggest that Moody’s actions might serve as the confirmation of the market expect, therefore these signals are likely to reduce the market uncertainty. However, during crisis period when credit signals might not be predicted by the market, the “additional information’’ role of Moody’s downgrade is mitigated, hence the market’s volatility increase in reaction to either S&P and Moody’s negative signals. Also, there is no evidence that positive announcements could significantly explain the dynamics of foreign exchange market. The result supports the prediction of the asymmetric market reaction to negative and positive events. Negative events associated with high level of sovereign risk might affect investor’s sentiment and raise the market’s uncertainty, as a result. Meanwhile, positive events is likely to convey less information than negative events hence insignificantly influence the market’s dynamics. 5.2. Split rating effect The results of Eq. (3) and Eq. (4) which compare the market effect of superior ratings and inferior rating on split rated sovereigns are presented from Table 5 to Table 10. On the one hand, the model of S&P negative events assigned for sovereigns which are unequally rated by Moody’s is in line with my prediction that exchange rate movement is more responsive to negative events on inferior ratings than those on superior rating. Table 5 illustrates that at 5% significance level, the market impact of S&P negative events are driven by pre-event disagreements between S&P and Moody (Fitch). Accordingly, the exchange rate immediately increases by 0.08% after 1- point CCR downgrade on inferior S&P rating (versus Moody’s). Also, the pre-event split ratings between S&P and Fitch significantly explain the movement of exchange rate in response to S&P negative signals. However, the impact of pre-
10. 226 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA event split rating between S&P and Fitch is stronger compared with the split pair of S&P and Moody’s. To some extent, the market impact of split rating which have Fitch rating involved is unexpected. One justification might be that Fitch ratings play as additional information in case where market is disturbed by different opinions between S&P and Moody’s (Tran et al., 2014). On the other hand, in terms of positive events, only the positive events by Moody’s on the superior ratings (versus S&P) impose a significant impact. In contrast, positive rating signals assigned by S&P and Fitch - under the effect of pre-event ratings which are split with another CRA - insignificantly affect the foreign exchange market. The result also supports the view that S&P tends to lead in sovereign downgrades while Moody’s tends to be the ‘first mover’ in upgrades (e.g. Alsakka and ap Gwilym, 2010). Table 8 to Table 10 report the results of Eq. (4) in the dynamics of foreign exchange market in reaction to credit rating announcements with pre-event split ratings. It should be noted that 1-unit changes in the CCR cause varying effect on the CCR depending on the investors’ behaviours in response to market’s uncertainty. Only the market impact of S&P ratings are affected by pre-event ratings which are unequally rated by Moody’s and Fitch whereas there is no significant evidence that pre-event split ratings induce market volatility on and after the days of Moody’s and Fitch news. As been observed from Table 10, currency market’s volatility also strongly reacts to S&P positive signals on the prior superior ratings than Fitch. The magnitude of reduction in volatility is 0.77% following one point CCR upgrade by S&P. This is in line with the result of Vu et al. (2015) that positive events on superior ratings and negative events on inferior ratings can strengthen the believe of investors in current movement of the market. These kinds of event break the inherent ambiguity in preceding split ratings between CRAs, hence trigger a strong reaction in foreign exchange rate market. Also, negative actions on inferior rating conveying a growing level of sovereign risk and uncertainty can instigate the market’s volatility. In contrast, if a superior rating is confirmed by a positive event, the market tends to be stabilized due to a positive sign of sovereign creditworthiness. Interestingly, S&P downgrade on preceding superior rating than Fitch also significantly influence the market volatility. As been discussed in the market impact of Fitch ratings, especially during crisis period, Fitch actions play a confirmation role of S&P and Moody’s announcements and split ratings between them. If Fitch rates a lower rating than S&P, investors are likely to expect a subsequent negative sign action. This is the reason why S&P downgrades following a superior rating (inferior rating of Fitch) can reduce the market uncertainty. It is consistent with the suggestion that “additional information’’ are likely to reduce the financial market volatility (Tran et al., 2014).
11. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 227 Table 3. OLS regression of the exchange rates’ reactions to negative sovereign rating events on split rated sovereigns (Eq.1) Panel A - Negative events S&P Moody’s Fitch Coeff t-val Coeff t-val Coeff t-val Constant 0.4965 1.62 -0.3084 -0.6 0.4743 0.800 (0.3061) (0.5107) (0.5923) ∆CCR 0.0854 2.66 0.0887 2.38 -0.0061 -0.150 (0.0321) (0.0373) (0.0405) CCR -0.0030 -0.33 0.0301 2.53 0.0161 1.410 (0.0090) (0.0119) (0.0114) PRIOEVENTS -0.0702 -1.46 -0.0176 -0.15 -0.0034 -0.030 (0.0482) (0.1210) (0.1224) VIX Index 8.5865 3.95 10.4378 2.7 6.6103 2.120 (2.1739) (3.8670) (3.1122) No of observations 384 282 294 Adjusted R-squared (%) 29.66 11.6 9.57 Panel B – positive events S&P Moody’s Fitch Coeff t-val Coeff t-val Coeff t-val Constant -0.1233 -0.3 -0.0529 -0.16 -0.0008 0 (0.4088) (0.3409) (0.5499) ∆CCR -0.1075 -1.81 -0.1471 -2.88 -0.0072 -0.13 (0.0595) (0.0510) (0.0543) CCR -0.0008 -0.07 0.0160 1.61 0.0127 1.04 (0.0122) (0.0099) (0.0122) PRIOEVENTS -0.1144 -1.62 -0.1029 -1.19 -0.2216 -1.62 (0.0705) (0.0868) (0.1368) VIX Index 13.8764 2.78 2.5600 0.69 4.7847 1.85 (4.9995) (3.6873) (2.5818) No of observations 288 224 208 Adjusted R-squared (%) 0.87 13.52 3.25 The table presents the coefficient estimated of Eq. (1) using the sample of 52 sovereigns rated by each agency. Year dummies and country dummies are included. The ‘bold’ coefficients are significant at the 10% level or less. : significant at 1% level : significant at 5% level *: significant at 10% level.
12. 228 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA Table 4. OLS regression of the exchange rate volatility’s reactions to negative sovereign rating events on split rated sovereigns (Eq.2) Panel A - Negative events S&P Moody’s Fitch Coeff t-val Coeff t-val Coeff t-val Constant -6.0408 -2.73 -1.2768 -0.43 0.7247 0.18 (2.2161) (2.9703) (3.9182) ∆CCR 0.3924 1.69 0.3623* 1.67 0.3382 1.26 (0.2327) (0.2169) (0.2676) CCR 0.0656 1 0.0545 0.79 0.0793 1.05 (0.0654) (0.0690) (0.0756) PRIOEVENTS -0.5045 -1.45 -0.4329 -0.62 -0.8381 -1.04 (0.3491) (0.7036) (0.8097) VIX Index 75.2177 4.78 61.4129 2.73 101.6488 4.94 (15.7383) (22.4904) (20.5863) No of observations 384 282 294 Adjusted R-squared (%) 7.25 8.24 13.5 Panel B – positive events S&P Moody’s Fitch Coeff t-val Coeff t-val Coeff t-val Constant 0.9874 0.43 -1.1481 -0.38 -2.9611 -1.05 (2.3006) (3.0203) (2.8264) ∆CCR -0.3195 -0.93 0.0425 0.1 0.4011 1.44 (0.3442) (0.4393) (0.2791) CCR -0.0194 -0.29 0.0947 1.05 0.0219 0.35 (0.0668) (0.0900) (0.0625) PRIOEVENTS -0.0363 -0.06 -0.1106 -0.21 0.1706 0.24 (0.5857) (0.5206) (0.7034) VIX Index 39.8893 1.6 38.2896 1.04 34.7684 2.62 (24.8837) (36.9348) (13.2703) No of observations 288 224 208 Adjusted R-squared (%) 0.19 -1.05 11.83 The table presents the coefficient estimated of Eq. (2) using the sample of 52 sovereigns rated by each agency. Year dummies and country dummies are included. The ‘bold’ coefficients are significant at the 10% level or less. : significant at 1% level : significant at 5% level *: significant at 10% level.
13. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 229 Table 5. OLS panel regressions of exchange rate’s reactions to S&P’s sovereign rating events on split rated sovereigns (Eq.3) Explanatory Variables S&P’s negative events S&P’s positive events S&P vs. Moody’s S&P vs. Fitch S&P vs. Moody’s S&P vs. Fitch Coeff t-val Coeff t-val Coeff t-val Coeff t-val Constant 0.2596* 0.81 0.0440 0.14 -0.2006 -0.710 -0.0558 -0.1 (0.3194) (0.3067) (0.2818) (-0.0558) ∆CCR*SUP 0.0616 1 -0.0579 -0.52 0.0183 0.320 -0.0551 -0.75 (0.0616) (0.1111) (0.0579) (0.0732) ∆CCR*INF 0.0802 2.5 0.1246 3.73 0.0553 0.630 -0.0597 -0.48 (0.0320) (0.0334) (0.0873) (0.1253) CCR 0.0192 2.19 0.0189 1.86 0.0055 0.710 -0.0010 -0.1 (0.0088) (0.0102) (0.0078) (0.0104) PRIOEVENTS -0.0463* -1.78 -0.0030 -2.03 -0.0210 -0.500 -0.0013 -0.49 (0.0261) (0.0015) (0.0419) (0.0026) VIX Index 5.1054 2.62 13.1845 5.37 2.5636 1.470 7.4411 2.44 (1.9458) (2.4530) (1.7441) (3.0535) No of observations 348 350 204 204 Adjusted R-squared (%) 38.9 33.77 11.46 3.46 The model is controlled for year and country dummies. The ‘bold’ coefficients are significant at the 10% level or less. : significant at 1% level : significant at 5% level *: significant at 10% level. Table 6. OLS panel regressions of exchange rate’s reactions to Moody’s sovereign rating events on split rated sovereigns (Eq.3) Explanatory Variables Moody’s negative events Moody’s positive events Moody’s vs. S&P Moody’s vs. Fitch Moody’s vs. S&P Moody’s vs. Fitch Coeff t-val Coeff t-val Coeff t-val Coeff t-val Constant 0.0723 0.14 -0.0107 -0.02 0.0821 0.250 -0.1085 -0.3 (0.5308) (0.4709) (0.3266) (0.3579) ∆CCR*SUP 0.1535* 1.73 0.1434 1.36 -0.1651 -2.130 -0.0484 -0.56 (0.0886) (0.1057) (0.0775) (0.0869) ∆CCR*INF 0.0061 0.13 -0.0173 -0.45 0.0495 0.700 0.0250 0.38 (0.0479) (0.0388) (0.0710) (0.0664) CCR -0.0008 -0.05 0.0283 1.9 0.0030 0.230 0.0138 0.93 (0.0170) (0.0149) (0.0130) (0.0149) PRIOEVENTS 0.0209 0.48 0.0389 1.06 0.0799 1.280 0.0388 0.69 (0.0435) (0.0368) (0.0626) (0.0566) VIX Index 9.3786 3 17.1890 4.44 7.3175 2.200 13.8478 2.54 (3.1215) (3.8673) (3.3217) (5.4626) No of observations 244 240 174 158 Adjusted R-squared (%) 15.05 10.92 -3.01 -3.78 The model is controlled for year and country dummies. The ‘bold’ coefficients are significant at the 10% level or less. : significant at 1% level : significant at 5% level *: significant at 10% level.
14. 230 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA Table 7. OLS panel regressions of exchange rate’s reactions to Fitch’s sovereign rating events on split rated sovereigns (Eq.3) Explanatory Variables Fitch’s negative events Fitch’s positive events Fitch vs. S&P Fitch vs. Moody’s Fitch vs. S&P Fitch vs. Moody’s Coeff t-val Coeff t-val Coeff t-val Coeff t-val Constant -0.1417 -0.27 -0.1009 -0.19 -0.2463 -0.610 -0.0059 -0.15 (0.5193) (0.5347) (0.4046) (0.0401) ∆CCR*SUP -0.0745 -1.19 0.0517 0.73 0.0199 0.400 0.0002 0.04 (0.0626) (0.0703) (0.0502) (0.0049) ∆CCR*INF 0.0143 0.29 -0.0169 -0.43 -0.1060 -1.060 -0.0145 -1.4 (0.0490) (0.0390) (0.1000) (0.0103) CCR 0.0198 1.42 0.0004 0.04 -0.0148 -1.270 -0.0015 -1.52 (0.0139) (0.0119) (0.0116) (0.0010) PRIOEVENTS -0.0189 -0.44 0.0422 1.15 0.0173 0.310 0.0004 0.1 (0.0432) (0.0369) (0.0564) (0.0043) VIX Index 8.6009 2.65 5.0082 1.89 6.0566* 1.850 0.3045 1.16 (3.2418) (2.6555) (3.2801) (0.2635) No of observations 204 294 164 176 Adjusted R-squared (%) 18.94 5.48 6.51 1.26 The model is controlled for year and country dummies. The ‘bold’ coefficients are significant at the 10% level or less. : significant at 1% level : significant at 5% level *: significant at 10% level. Table 8. OLS panel regressions of exchange rate volatility reactions to S&P’s sovereign rating events on split rated sovereigns (Eq.4) Explanatory Variables S&P’s negative events S&P’s positive events S&P vs. Moody’s S&P vs. Fitch S&P vs. Moody’s S&P vs. Fitch Coeff t-val Coeff t-val Coeff t-val Coeff t-val Constant 1.8734 0.81 3.2494 1.53 -4.935* -1.810 8.8806 2.71 (2.3083) (2.1215) (2.7282) (3.2829) ∆CCR*SUP 0.3978 0.89 -1.8048 -2.35 0.4206 0.750 -0.7707* -1.73 (0.4453) (0.7687) (0.5601) (0.4453) ∆CCR*INF 0.4328* 1.87 -0.1209 -0.52 -0.2393 -0.280 -1.1535 -1.51 (0.2312) (0.2310) (0.8451) (0.7620) CCR 0.0263 0.41 0.0076 0.11 -0.0338 -0.450 -0.0113 -0.18 (0.0634) (0.0703) (0.0758) (0.0632) PRIOEVENTS -0.0773 -0.41 -0.0090 -0.87 0.3841 0.950 -0.0183 -1.14 (0.1883) (0.0104) (0.4054) (0.0161) VIX Index 70.3740 5 92.906 5.48 83.5818 4.950 41.1053 2.21 (14.0636) (16.9689) (16.8841) (18.5674) No of observations 348 350 204 204 Adjusted R-squared (%) 12.39 10.87 16.16 13.73 The model is controlled for year and country dummies. The ‘bold’ coefficients are significant at the 10% level or less. : significant at 1% level : significant at 5% level *: significant at 10% level.
15. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 231 Table 9. OLS panel regressions of exchange rate volatility reactions to Moody’s sovereign rating events on split rated sovereigns (Eq.4) Explanatory Variables Moody’s negative events Moody’s positive events Moody’s vs. S&P Moody’s vs. Fitch Moody’s vs. S&P Moody’s vs. Fitch Coeff t-val Coeff t-val Coeff t-val Coeff t-val Constant -3.4436 -1.14 -4.3023* -1.76 1.0473 0.290 0.3842 0.15 (3.0310) (2.4418) (3.6573) (2.5553) ∆CCR*SUP -0.1229 -0.24 -0.8607 -1.57 -0.2934 -0.340 -0.5629 -0.91 (0.5057) (0.5482) (0.8682) (0.6206) ∆CCR*INF -0.1070 -0.39 -0.0927 -0.46 0.1524 0.190 0.6216 1.31 (0.2735) (0.2014) 0.7950 0.4737 CCR 0.1709 1.76 0.1458 1.88 -0.0080 -0.050 0.0139 0.13 (0.0973) (0.0775) 0.1458 0.1062 PRIOEVENTS -0.1050 -0.42 0.1602 0.84 0.6595 0.940 0.3881 0.96 (0.2484) (0.1910) 0.7011 0.4038 VIX Index 61.3976 3.44 124.7874 6.22 -141.4684 -3.800 76.8198* 1.97 (17.8247) (20.0515) 37.1948 38.9968 No of observations 244 240 174 158 Adjusted R-squared (%) 11.72 20.53 0.17 3.98 The model is controlled for year and country dummies. The ‘bold’ coefficients are significant at the 10% level or less. : significant at 1% level : significant at 5% level *: significant at 10% level. Table 10. OLS panel regressions of exchange rate volatility reactions to Fitch’s sovereign rating events on split rated sovereigns (Eq.4) Fitch’s negative events Fitch’s positive events Explanatory Variables Fitch vs. S&P Fitch vs. Moody’s Fitch vs. S&P Fitch vs. Moody’s Coeff t-val Coeff t-val Coeff t-val Coeff t-val Constant -3.1104 -0.72 -2.4005 -0.54 -4.4975 -1.200 -0.0123 -0.33 4.3090 4.4205 3.7391 0.0374 ∆CCR*SUP 0.0487 0.09 0.4387 0.75 0.7083 1.530 0.0044 0.96 0.5195 0.5816 0.4640 0.0046 ∆CCR*INF 0.3352 0.82 0.4663 1.45 0.1673 -1.060 0.0056 0.58 0.4070 0.3227 0.1000 0.0096 CCR 0.0930 0.8 0.1154 1.18 -0.1373 -1.280 -0.0005 -0.52 0.1157 0.0982 0.1072 0.0009 PRIOEVENTS -0.4109 -1.15 -0.0236 -0.08 0.2320 0.450 -0.0014 -0.35 0.3582 0.3047 0.5208 0.0040 VIX Index 72.7932 2.71 108.5484 4.94 57.73348* 1.900 0.5298 2.15 26.8991 21.9548 30.3151 0.2460 No of observations 204 294 164 176 Adjusted R-squared (%) 12.97 5.46 -1.86 -6.24 The model is controlled for year and country dummies. The ‘bold’ coefficients are significant at the 10% level or less. : significant at 1% level : significant at 5% level *: significant at 10% level.
16. 232 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA 6. CONCLUSION There is wide interest in sovereign credit ratings and split ratings and their impact on financial mar- kets, but the relation between sovereign rating actions and foreign exchange rates attracts little attention. However, the economic rationale of sovereign rating impact on the currency market can be inferred from the casual linkages between sovereign credit ratings and fiscal indices of a country, as well as the direct influence of a sovereign fiscal health over its currency and exchange rate. Prior papers exhibit evidences about these relations, but empirical studies remain silent on the information content of split ratings to for- eign exchange market, and particular to exchange rate volatility. This question is empirically addressed in this paper. The sample reveals a high probability of split ratings occurred during the study period. The results provide empirical insight on the information content of sovereign credit rating changes and split ratings for the foreign exchange market. There are empirical evidences about the market’s asymmetric responses not only between negative and positive rating signals but also across CRAs. Particularly, S&P negative rating announcements appear to demonstrate a ‘first mover’ feature in global currency markets, whereas Moody’s is likely to lead the market reaction to positive credit events. This offers a considerable contribution to prior literatures on market impacts of sovereign credit ratings. Furthermore, the study analyses the currency market impact of certain types of rating announcements from the perspective of split ratings. Particularly, the distinct significant effect of inferior (superior) ratings on negative (positive) events, which was found in bond and stock markets by Vu et al. (2015), is confirmed in this study when it comes to foreign exchange rate. The effect of split rating on reaction of foreign exchange rate to rating events is in line with the view- point of Epstein and Schneider (2008) about trading attitude under ambiguity aversion. Specifically, the influence of split ratings which raise uncertainty about sovereign default risk is driven by how ambiguous the market is after the events concerning prior split ratings. Ambiguity increases when inferior ratings are downgraded and superior ratings are upgraded, hence risk – averse market participants react strongly to negative rating changes on prior inferior ratings and positive announcement on pre-event superior ratings. Further, the study shows a significant relation between exchange rate volatility and split ratings across CRAs, which has not been considered in prior papers regarding the market impacts of split ratings. Only the information content of S&P events to exchange rate volatility are affected by prior split ratings (with Moody’s and Fitch). Particularly, the action of downgrading S&P superior ratings versus Fitch’s, which is unexpected, strongly give rise to currency market’s ambiguity. Although the study contributes to literature in a number of respects and also have practical implica- tions for financial institutions, there remains some limitations. The sample is based on the biggest currency markets with 17 out of 52 countries using Euro, therefore the findings seem inclined to the European market while the reactions of developing foreign exchange markets is not the focus. Moreover, other econometric models which account for both the behaviours of exchange rate and exchange rate volatility (e.g. GARCH family) could be employed as robustness tests. In addition, investigations on split ratings assigned for cor- porate and banking sectors are potential suggestion for future studies. Also, the question of whether split ratings have spill-over effects on global foreign exchange markets should be considered in future studies. REFFERENCE [1] Afonso, A., Furceri, D., & Gomes, P. 2012. Sovereign credit ratings and financial markets linkages: application to European data. Journal of International Money and Finance, 31(3), 606- 638.
17. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 233 [2] Afonso, A., Gomes, P. & Rother, P. 2011. Short-and long-run determinants of sovereign debt credit ratings. International Journal of Finance & Economics, 16 (1), pp. 1-15. [3] Almeida, H., Cunha, I., Ferreira, M.A. & Restrepo, F. 2017. The real effects of credit ratings: The sovereign ceiling channel. The Journal of Finance, 72 (1), pp. 249-290. [4] Alsakka, R. & Ap Gwilym, O. 2013. Rating agencies’ signals during the European sovereign debt crisis: Market impact and spillovers. Journal of Economic Behavior & Organization, 85 pp. 144- 162. [5] Alsakka, R. & ap Gwilym, O. 2012b. Foreign exchange market reactions to sovereign credit news. Journal of International Money and Finance, 31 (4), pp. 845-864. [6] Alsakka, R. & ap Gwilym, O. 2010. Leads and lags in sovereign credit ratings. Journal of Banking & Finance, 34 (11), pp. 2614-2626. [7] Alsakka, R. & ap Gwilym, O. 2009. Heterogeneity of sovereign rating migrations in emerging countries. Emerging Markets Review, 10 (2), pp. 151-165. [8] Alsakka, R., ap Gwilym, O. & Vu, H. 2017. Differences of opinion in sovereign credit signals during the European crisis. The European Journal of Finance, 23 (10), pp. 859-884. [9] Bửninghausen, B. & Zabel, M. 2015. Credit ratings and cross-border bond market spillovers. Journal of International Money and Finance, 53 pp. 115-136. [10] Bennell, J.A., Crabbe, D., Thomas, S. & Ap Gwilym, O. 2006. Modelling sovereign credit ratings: Neural networks versus ordered probit. Expert Systems with Applications, 30 (3), pp. 415-425. [11] BIS. 2013. Triennial Central Bank survey of Foreign exchange turnover in April 2013. BIS. 2016. Triennial Central Bank Survey Foreign exchange turnover in April 2016. [12] Bongaerts, D., Cremers, K.J. & Goetzmann, W.N. 2012. Tiebreaker: Certification and multiple credit ratings. The Journal of Finance, 67 (1), pp. 113-152. [13] Borensztein, E., Cowan, K. & Valenzuela, P. 2013. Sovereign ceilings “lite”? The impact of sovereign ratings on corporate ratings. Journal of Banking & Finance, 37 (11), pp. 4014-4024. [14] Brooks, R., Faff, R.W., Hillier, D. & Hillier, J. 2004. The national market impact of sovereign rating changes. Journal of banking & finance, 28 (1), pp. 233-250. [15] Epstein, L.G. & Schneider, M. 2008. Ambiguity, information quality, and asset pricing. The Journal of Finance, 63 (1), pp. 197-228. [16] Erceg, C. J., Guerrieri, L., & Gust, C. 2005. Expansionary fiscal shocks and the US trade deficit. International Finance, 8(3), 363-397. [17] European Securities and Market Authority (ESMA), 2013. CRAs’ market share calculation according to Article 8d of the CRA regulation. Available via: calculation.pdf. [18] Evans, M.D. & Lyons, R.K. 2005. Do currency markets absorb news quickly? Journal of International Money and Finance, 24 (2), pp. 197-217. [19] Ferreira, M.A. & Gama, P.M. 2007. Does sovereign debt ratings news spill over to international stock markets? Journal of Banking & Finance, 31 (10), pp. 3162-3182. [20] Gande, A. & Parsley, D.C. 2005. News spillovers in the sovereign debt market. Journal of Financial Economics, 75 (3), pp. 691-734. [21] Hill, P., Brooks, R. & Faff, R. 2010. Variations in sovereign credit quality assessments across rating agencies. Journal of Banking & Finance, 34 (6), pp. 1327-1343. [22] Kaminsky, G. & Schmukler, S.L. 2002. Emerging market instability: do sovereign ratings affect country risk and stock returns? The World Bank Economic Review, 16 (2), pp. 171-195.
18. 234 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA [23] Kim, S. & Roubini, N. 2008. Twin deficit or twin divergence? Fiscal policy, current account, and real exchange rate in the US. Journal of International Economics, 74 (2), pp. 362-383. [24] Kim, S. & Wu, E. 2011. International bank flows to emerging markets: Influence of sovereign credit ratings and their regional spillover effects. Journal of Financial Research, 34 (2), pp. 331- 364. [25] Livingston, M. & Zhou, L. 2016. Information Opacity and Fitch Bond Ratings. Journal of Financial Research, 39 (4), pp. 329-358. [26] Moody’s Investor Service. 2016. Retrieved from Rating Methodology: Sovereign Bond Ratings.(online): Available: http:/www.moodys.com [27] Santos, J.A. 2006. Why firm access to the bond market differs over the business cycle: A theory andsome evidence. Journal of Banking & Finance, 30 (10), pp. 2715-2736. [28] SEC. 2016. Securities and Exchange Commission, 2016. Annual Report on Nationally Recognized Statistical Rating Organizations. Available via: [29] Tran, V., Alsakka, R. & ap Gwilym, O. 2014. Sovereign rating actions and the implied volatility of stock index options. International Review of Financial Analysis, 34 pp. 101-113. [30] Vu, H., Alsakka, R. & ap Gwilym, O. 2015. The credit signals that matter most for sovereign bond spreads with split rating. Journal of International Money and Finance, 53 pp. 174-191. [31] Williams, G., Alsakka, R. & Ap Gwilym, O. 2013. The impact of sovereign rating actions on bank ratings in emerging markets. Journal of Banking & Finance, 37 (2), pp. 563-577.
19. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 235 APPENDICES APPENDIX A. COUNTRIES SAMPLE Argentina Estonia Korea Russia Australia Finland Latvia Saudi Arabia Austria France Lithuania Singapore Bahrain Germany Malaysia Slovakia Belgium Greece Malta Slovenia Brazil Hong Kong SAR Mexico South Africa Bulgaria Hungary Netherlands Spain Chile India New Zealand Sweden China Indonesia Peru Switzerland Colombia Ireland Philippines Thailand Croatia Israel Poland Turkey Cyprus Italy Portugal United Kingdom Czech Republic Japan Romania Appendix B. How the ratings are mapped to an 18 notch rating scale and the 52 point CCR. S&P Rating Moody’s Rating Fitch Rating 18-notch scale 52 point CCR scale AAA Aaa AAA 18 52 AA+ Aa1 AA+ 17 49 AA Aa2 AA 16 46 AA- Aa3 AA- 15 43 A+ A1 A+ 14 40 A A2 A 13 37 A- A3 A- 12 34 BBB+ Baa1 BBB+ 11 31 BBB Baa2 BBB 10 28 BBB- Baa3 BBB- 9 25 BB+ Ba1 BB+ 8 22 BB Ba2 BB 7 19 BB- Ba3 BB- 6 16 B+ B1 B+ 5 13 B B2 B 4 10 B- B3 B- 3 7 CCC+/CCC/CCC- Caa1/Caa2/Caa3 CCC 2 4 CC/C/SD Ca/C/D CC/C/D 1 1