The impact of sovereign rating actions on bank ratings in emerging markets

Journal of Banking & Finance 37 (2013) 563–577

Contents lists available at SciVerse ScienceDirect

Journal of Banking & Finance journal homepage: www.elsevier.com/locate/jbf

The impact of sovereign rating actions on bank ratings in emerging markets Gwion Williams, Rasha Alsakka, Owain ap Gwilym ⇑ Bangor Business School, Bangor University, Bangor, LL57 2DG, UK

a r t i c l e

i n f o

Article history: Received 11 August 2011 Accepted 24 September 2012 Available online 8 October 2012 JEL classification: G15 G24 Keywords: Sovereign ceiling Sovereign upgrades/downgrades Sovereign watch Bank ratings Emerging markets

a b s t r a c t This paper analyses the effects of sovereign rating actions on the credit ratings of banks in emerging markets, using a sample from three global rating agencies across 54 countries for 1999–2009. Despite widespread attention to sovereign ratings and bank ratings, no previous study has investigated the link in this manner. We find that sovereign rating upgrades (downgrades) have strong effects on bank rating upgrades (downgrades). The impact of sovereign watch status on bank rating actions is much weaker and often insignificant. The sensitivity of banks’ ratings to sovereign rating actions is affected by the countries’ economic and financial freedom and by macroeconomic conditions. Ratings of banks with different ownership structures are all influenced strongly by the sovereign rating, with some variation depending on the countries’ characteristics. Emerging market bank ratings are less likely to follow sovereign rating downgrades during the recent financial crisis period. Ó 2012 Elsevier B.V. All rights reserved.

1. Introduction Sovereign ratings have been a focus of widespread attention during 2010–2012, most obviously in the case of eurozone sovereigns including Greece and Spain and in the downgrading of the USA by Standard & Poor’s (S&P) in August 2011. The IMF (2010) highlights that sovereign credit risk is one of the main current threats to global economic stability. Related to this, Duggar et al. (2009) identify that 71% of defaults by rated corporates and subsovereigns in emerging markets have occurred during sovereign crises. They also suggest that sovereign credit risk is a key factor in corporate defaults outside sovereign credit events. The aim of this paper is to investigate to what extent sovereign rating actions affect the credit ratings of banks in the same country. The paper models: (i) the effects of sovereign credit rating upgrades, downgrades and watch status on bank credit ratings; and (ii) how country characteristics and bank ownership influence the sensitivity of bank ratings to recent sovereign rating changes. The paper aims to provide insights into the rating policies applied by the world’s largest credit rating agencies (CRAs). A crucial factor motivating the analysis is the notion of the sovereign rating ‘ceiling’. This means that generally the sovereign rating represents the highest achievable rating for non-sovereigns

⇑ Corresponding author. Tel.: +44 1248 382176. E-mail addresses: [email protected] (G. Williams), r.alsakka@ bangor.ac.uk (R. Alsakka), [email protected] (O. ap Gwilym). 0378-4266/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jbankfin.2012.09.021

within that country. Although the largest CRAs no longer apply this ceiling as an absolute rule, it is still the prevailing situation in the vast majority of cases. For example, many non-sovereigns were downgraded in August 2011 following the USA downgrade. The sovereign ceiling inevitably has a greater impact on non-sovereign ratings in countries with lower sovereign ratings. For example, if the sovereign has a speculative grade rating, the potential rating scale for a non-sovereign issuer in that country is compressed. This paper focuses on emerging markets, where the effect of the sovereign ceiling is much more apparent. Sovereign rating changes and outlook/watch signals affect bond and stock markets in emerging markets. The literature also shows that these effects are not only significant at the domestic level, since sovereign rating news is found to affect markets in other countries. In particular, negative sovereign rating news causes significant spillovers into other countries’ stock and bond markets, while positive news has an insignificant effect (e.g. Brooks et al., 2004; Gande and Parsley, 2005; Ferreira and Gama, 2007). The economic and market impact of sovereign rating actions are discussed further in Section 2. The large growth in debt issuers has increased the demand for ratings and the influence of CRAs in capital markets. The credibility of CRAs has been questioned over the past few years, in particular during the 2007 US subprime mortgage crisis. Yet, CRAs still control the gateway into capital markets for bond issuers, as well as providing debt market participants with valuable signals due to their access to private information. In general, the vast majority

564

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577

of studies on credit ratings have used data from a single agency (usually Moody’s or S&P). More recently, a few studies have highlighted important inter-agency differences (e.g. Hill et al., 2010). This paper contributes to the literature by examining the effects of sovereign rating actions on bank ratings. We consider the economic significance of detected relationships by calculating the effects of changes in the independent variables (sovereign rating change and sovereign watch status) on the probability of bank rating upgrades and downgrades of one and two or more notches, and the probability of no rating change, i.e., the marginal effects (see Livingston et al., 2008). We utilise a large dataset of emerging market bank ratings from the three global CRAs, namely Moody’s, S&P and Fitch. Several robustness checks are performed using sub-samples by agency, bank ownership and time periods. We also investigate whether the characteristics of the countries, in terms of economic/financial freedom and macroeconomic conditions, influence the sensitivity of bank ratings to sovereign rating changes, and whether this impact varies across different bank ownership status. The main results are as follows. Emerging market banks have very high probabilities of being upgraded (downgraded) soon after an upgrade (downgrade) to their corresponding sovereign rating. These effects are fairly consistent for all three CRAs, although some results imply that Moody’s is the least likely agency to migrate bank ratings simultaneously with the sovereign rating. We find that the sensitivity of bank ratings to sovereign rating actions does vary depending on the country’s overall economic and financial freedom and macroeconomic factors. The results are not driven by bank ownership, because state-owned, foreign-owned, and local privately-owned bank ratings are all affected very strongly. However, local privately-owned banks’ ratings are the most sensitive to sovereign upgrades, and foreign-owned banks’ ratings are the most sensitive to sovereign downgrades. This can be explained by varying levels of influence of country characteristics on bank ratings’ sensitivity to sovereign rating actions across different ownership structures of banks. The rest of the paper is organised as follows. Section 2 reviews the relevant literature, Sections 3 and 4 discuss the data and methodology, Section 5 presents the empirical results and Section 6 concludes the paper.

2. Literature review 2.1. Bank ratings The literature linked to bank ratings is limited, and there is no prior research which documents how bank ratings are affected by sovereign rating signals. Ferri et al. (2001) examine the effect of linking banks’ capital requirements with external credit ratings in non-high income countries, under the proposed Basel II regime. They find that the capital requirements of banks in these countries would become more volatile since the bank ratings seem to be strongly correlated to sovereign ratings. Caporale et al. (2012) show that bank ratings reflect banks’ financial position and country of origin, whereby a bank in a less stable/developed/rich economy appears to have a lower rating. Using a sample of S&P credit ratings for 86 countries during 2002–2008, Shen et al. (2012) find that banks with higher ratios of profitability, liquidity and capital adequacy and better ratios of efficiency (cost-to-income) and asset quality (loan loss provisions to net interest revenues) tend to be assigned higher ratings. The influence of financial ratios on bank ratings is greater in low information asymmetry countries (such as industrial or high-income countries) but reduced in countries with high information asymmetry (such as middle-income countries and emerging market countries). Shen et al. (2012) also show that

larger bank assets and higher sovereign credit ratings boost bank credit ratings. Poon et al. (2009) and Bannier et al. (2010) investigate bank ratings, but their focus is on whether unsolicited ratings are biased downward. These studies indicate that solicited bank ratings tend to be significantly higher than unsolicited ratings. Using S&P ratings for 460 commercial banks in 72 countries, excluding the United States, for the period 1998–2003, Poon et al. (2009) point out that banks with solicited ratings tend to be larger, have relatively less nonperforming loans to gross loans, and have higher returns on equity than banks with unsolicited ratings. Bannier et al. (2010) show that observed downward bias in unsolicited bank ratings (by S&P for 1996–2006) is driven by strategic factors within the rating process, and seems to increase along with bank’s opaqueness. On a related issue, Morgan (2002) analyses ratings assigned by Moody’s and S&P across different US industries to determine whether there are more split ratings in the banking sector than in others, with split ratings used as a proxy for opaqueness. He finds that the proportion of split ratings is much higher in the banking and insurance sectors. He argues that banks are more opaque than other corporates, thus making it more challenging to quantify the risks arising from the nature of banks’ assets and capital structure. Cash, loans and trading assets increase the uncertainties involved with quantifying banks’ risks, whilst there is less uncertainty for banks with more fixed assets and capital. Similarly, Iannotta (2006) uses split ratings to test whether banks are relatively more opaque than other industries. For European data on firms rated by Moody’s and S&P, he finds that the probability of a split rating increases by more than 20% when the issuer is a bank, compared to other industries. 2.2. Rating heterogeneity Assessing the factors causing credit rating migration is a topical theme in recent credit ratings literature. The CRAs use outlook and watch as indicators of possible future rating changes, in order to retain rating stability whilst providing more information for market participants.1 These instruments have been found to provide an important economic function (e.g. Bannier and Hirsch, 2010). Vazza et al. (2005) analyse the behaviour of S&P corporate issuer ratings placed on outlook and watch. They find that such issuers have a higher probability of a rating change in the direction specified by the watch or outlook. 70% (64%) of issuers placed on positive (negative) watch were subsequently upgraded (downgraded), normally within 90 days of having been placed on watch. 44% (35%) of issuers placed on positive (negative) outlook were subsequently upgraded (downgraded), within a 6-month to 2-year period after being placed on positive or negative outlook. Using data from six CRAs, Alsakka and ap Gwilym (2009) analyse the dynamics of sovereign ratings in emerging economies, while including the watch status. They find that sovereigns placed on watch have a higher probability of a rating change in the direction specified by the watch status, within 12 months of being placed on the watch list. An important element of credit rating migrations is rating momentum, where the prediction of a future rating change for an issuer is dependent on its previous change. The implication is that an issuer that has experienced a previous upgrade (downgrade) is more likely to be upgraded (downgraded) in its next rating update. For S&P rated corporate issuers, Lando and Sködeberg (2002) find strong evidence of downgrade momentum except for issuers in the BB, CCC+ and CCC rating categories, while little 1 Outlook reflects a CRA’s medium-term (1–2 years) view on the development of a credit rating, while watch is a stronger indication focused on a typical ex-ante target horizon of 3 months.

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577

evidence of upgrade momentum. Hamilton and Cantor (2004) consider Moody’s rated corporate issuers, and find evidence for both upgrade and downgrade momentum. The evidence for downgrade momentum is significantly stronger than for upgrade momentum, in line with previous literature. Hamilton and Cantor (2004) conduct an additional bivariate analysis, where they consider rating momentum and outlooks simultaneously. They find that rating momentum almost disappears when the outlook is controlled for. Further, Alsakka and ap Gwilym (2009) find evidence for downgrade momentum but not for upgrade momentum from their sample of sovereigns rated by six CRAs. They show that the larger the previous downgrade of an issuer, then the more likely it is that the issuer will suffer another downgrade, compared to an issuer which only received a one notch previous downgrade. They also find that rating momentum is much less significant after accounting for watch status. 2.3. The economic and market impact of sovereign rating actions Sovereign ratings represent assessments of the ability and willingness of governments to meet their financial obligations. They affect the dynamics of capital markets and influence the cost of capital. Brooks et al. (2004) show that sovereign rating downgrades have a strong negative impact on stock markets but there is limited evidence of abnormal returns linked to upgrades. Gande and Parsley (2005) and Ferreira and Gama (2007) reveal that sovereign downgrades incorporate valuable information for sovereign bond spreads and aggregate stock market returns of other countries, particularly in emerging economies, neighbouring countries, and during crisis periods, while upgrades have an insignificant impact. Ismailescu and Kazemi (2010) analyse whether emerging market CDS spreads respond to sovereign rating changes. They find that positive signals add new information to the markets, while negative news is anticipated and hence reflected in the CDS spreads. These results are quite contradictory to earlier studies that find negative rating signals to have the greatest effect on CDS spreads (e.g. Norden and Weber, 2004). However, they find that negative signals significantly widen CDS spreads for investment grade issuers, and positive announcements significantly narrow CDS spreads for speculative grade issuers. Kim and Wu (2008) examine whether S&P sovereign ratings help attract international capital (international banking, foreign direct investment and portfolio flows) and thus induce domestic financial sector development in 51 emerging markets. They find that sovereign rating news is an important stimulus for all three kinds of international capital flows. They find significant domestic bond market developments after improvements in sovereign ratings. Kim and Wu (2011) examine whether sovereign rating actions in emerging markets impact international bank flows from the G7 countries, and also examine regional spillover effects. They find that positive sovereign rating changes have a strong positive effect on international bank flows from G7 countries into the emerging countries. Positive rating actions in one region tend to draw the international bank flows away from the other three regions. Durbin and Ng (2005) analyse whether investors apply the sovereign ceiling rule by comparing firms’ bond spreads with those of the corresponding sovereign. They find several cases where the sovereign spread is greater than or equal to the corporate spread, which suggests a violation of the sovereign ceiling rule. They explain that some of the firms have lower spreads than the sovereign because they have a very high proportion of foreign currency earnings, a substantial amount of overseas assets, an affiliation with foreign companies, and/or a strong tie with the home government. Borensztein et al. (2007) analyse a sample of S&P rated sovereign and non-sovereign issuers from emerging economies to consider

565

the impact of sovereign ratings on the corporate ratings. They find that a one notch upgrade to a sovereign rating increases the nonsovereign ratings by one-sixth of a notch, on average. A one notch downgrade in the sovereign ratings causes a half-notch downgrade for the non-sovereigns, on average.

3. Data 3.1. Credit ratings The dataset consists of end-of-month long-term (LT) foreigncurrency (FC) ratings for sovereigns and banks in 54 emerging countries. The source of the ratings data is the InteractiveData Credit Ratings International database. The sample is based on selecting emerging market banks which are rated by at least one of the three largest CRAs (Moody’s, S&P and Fitch) during the period 30th November 1999–31st December 2009. We define an ‘emerging market’ by using the World Bank’s country classification, according to the countries’ GNI per capita. The World Bank classifies countries into four different categories: low-income (LI), lower middle income (LMI), upper middle income (UMI) and high-income (HI). All LI, LMI and UMI economies are categorized as emerging markets. A strict duration of 3 months is used throughout the sample, whereby any bank rating action which is more than 3 months later than the most recent relevant sovereign action is omitted. There are two reasons behind this choice. First, due to the research questions, we need to place a restriction on the time elapsed between the sovereign and the bank rating actions. Bank rating actions which are more than 3 months later than a sovereign action are very likely to be driven by other factors. Second, the CRAs express an ex-ante target of 90 days to take action once an issuer is placed on watch (e.g. Hamilton and Cantor, 2004; Vazza et al., 2005).2 Following Alsakka and ap Gwilym (2010), actual rating changes are identified according to a 20-point numerical rating scale (Aaa/ AAA = 1, Aa1/AA+ = 2, Aa2/AA = 3 . . . Caa3/CCC = 19, Ca/CC, C/SDD = 20) by notches on the basis of monthly intervals. Panel I of Table 1 summarises the dataset. There are 514 observations for 178 banks from 36 countries rated by Moody’s. There are 440 observations for 151 banks from 40 countries rated by S&P, and 796 observations for 278 banks from 41 countries rated by Fitch. This gives a total of 1750 end of month observations. We observe that S&P is the agency most likely to rate the bank the same as the sovereign, with almost 80% of banks’ observations rated at the sovereign ceiling. Moody’s is the agency least likely to assign the same rating to the bank and sovereign with just over 55% observations. Also, Moody’s is the most likely to assign a lower rating to the bank than the sovereign, whilst Fitch is the most likely to rate a bank higher than the sovereign. There are 189 (105) bank upgrades (downgrades) by Moody’s, 234 (154) by S&P, and 423 (247) by Fitch. There are also 103 (23) sovereign upgrades (downgrades) by Moody’s, 116 (72) by S&P, and 103 (54) by Fitch. These statistics reflect the strong upgrade trend in emerging markets during this time period (in particular pre-2007), which can be explained by higher commodity prices, higher oil and natural gas prices and larger pools of inexpensive skilled labour which fuelled economic growth. Panel I of Table 1 also summarises the sovereign watch actions, and identifies 42 (15) positive (negative) cases of watch status by Moody’s, 0 (19) by S&P, and 9 (18) by Fitch. These figures highlight differences in the policies of the three CRAs. Moody’s assigns posi2 The second reason is actually more important in sample construction. In reality, the bank and sovereign rating actions in the sample are normally within less than one month of each other.

566

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577

Table 1 Descriptive statistics of the data sample. Moody’s

S&P

Fitch

Total

Panel I – Credit rating data Banks Countries Rated banks State-owned Foreign-owned Local privately-owned Observations Upgrades Downgrades No change B=S BS

36 178 41 50 50 514 189 105 220 285 179 50

36.8% 20.4% 42.8% 55.5% 34.8% 9.7%

40 151 40 37 50 440 234 154 52 343 75 22

53.2% 35.0% 11.8% 77.9% 17.1% 5.0%

41 278 39 103 74 796 423 247 126 519 160 117

53.2% 31.0% 15.8% 65.2% 20.1% 14.7%

54 425 74 136 125 1750 846 506 398 1147 414 189

48.3% 28.9% 22.8% 65.5% 23.7% 10.8%

Banks pre-crisis Observations Upgrades Downgrades No change

406 165 76 165

40.6% 18.7% 40.7%

325 174 107 44

53.5% 32.9% 13.6%

601 352 160 89

58.6% 26.6% 14.8%

1332 691 343 298

51.9% 25.7% 22.4%

Banks crisis Observations Upgrades Downgrades No change

108 24 29 55

22.2% 26.9% 50.9%

115 60 47 8

52.2% 40.9% 6.9%

195 71 87 37

36.4% 44.6% 19.0%

418 155 163 100

37.1% 39.0% 23.9%

Sovereign actions Upgrade Down grades Up watch Down watch Total

103 23 42 15 183

116 72 0 19 207

103 54 9 18 184

322 149 51 52 574

Panel II – Other explanatory variables Variable

Description

Economic freedom Financial freedom Government spending Fiscal freedom Trade freedom Investment freedom Monetary freedom Business freedom Property rights Freedom from corruption GDP per capita GDP growth Inflation Current acc bal. Fiscal balance External debt

A composite measure of 10 economic factors determining the degree of economic freedom in the country A measure of banking efficiency and independence from government control and interference in the financial sector A measure of the level of government expenditures as a percentage of GDP A measure of the tax burden imposed by government A measure of the absence of tariffs that affect imports and exports A measure of the absence of constraints on the flow of investment capital A measure of price stability with an assessment of price controls A measure of the ability to start, operate, and close a business and government efficiency in the regulatory process A measure of the extent to which a country’s laws protect private property rights A measure of the extent to which corruption prevails in a country GDP per capita for the previous year (Thousands US$) Average annual real GDP growth on a year-over-year basis for the previous three years (%) Average annual consumer price inflation growth on a year-over-year basis for the previous three years (%) Average annual current account balance relative to GDP for the previous three years (%) Average annual central government deficit or surplus relative to GDP for the previous three years (%) Total external debt relative to exports for the previous year (%)

Panel I presents summary statistics for the credit rating dataset, that consists of end of month bank and sovereign ratings and watch (only for the sovereigns) from emerging countries for November 1999–December 2009. Emerging countries are defined using the World Bank classification of GNI per capita. B = S, B < S, and B > S are defined as follows: Banks rated the same as the sovereign, banks rated worse than the sovereign, and banks rated better than the sovereign, respectively. Each banks’ ownership status was identified using BankScope. The pre-crisis period is November 1999–December 2006, whilst the crisis period is January 2007–December 2009. Panel II presents a description of other utilised variables, source of the data: DataStream (WDI, IMF, and Oxford Economics) and Heritage Foundation. All freedom measures are measured by a scale from 0 to 100, where 100 represents the maximum freedom.

tive watch most frequently, S&P tends not to put any sovereign on positive watch status, yet S&P assigns negative watch most frequently. This is in line with Alsakka and ap Gwilym (2010), who find Moody’s to be the leader among the three CRAs in upgrading sovereigns, while S&P tends to be the first mover in downgrading sovereigns. We also find that sovereigns that have been on positive watch are subsequently upgraded more often than those on negative watch are subsequently downgraded. For Moody’s, the 42 positive watch cases led to 38 rating upgrades (33 within 3 months), and for Fitch the 9 positive watch cases led to 8 rating upgrades (all within 3 months). In the case of downgrades, 15 negative watch sovereign actions led to 6 rating downgrades (5 within 3 months) by Moody’s, 19 led to 11 (9 within 3 months) by S&P, and 18 led to 9 (5 within 3 months) by Fitch.

Of the 189 (105) bank upgrades (downgrades) by Moody’s, 180 (73) of them are linked to sovereign upgrades (downgrades). For S&P, of 234 (154) bank upgrades (downgrades), 231 (154) are linked to sovereign upgrades (downgrades). For Fitch, of 423 (247) bank upgrades (downgrades), 422 (222) are linked to sovereign upgrades (downgrades). These statistics give an indication of the strength of the link between sovereign and bank rating changes. From the total of 846 (506) bank upgrades (downgrades) from the three CRAs, only 4 (12) are linked to sovereign downgrades (upgrades), suggesting that banks are highly likely to have a rating change in the same direction as the sovereign. Using BankScope, the ownership status of the rated banks is identified, to investigate whether ratings of state-owned banks behave differently to those of foreign or local privately-owned banks.

567

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577

There are 74 state-owned banks in the sample, 41 rated by Moody’s, 40 by S&P, and 39 by Fitch. There are 136 foreign-owned banks, 50 rated by Moody’s, 37 by S&P and 103 by Fitch, and 125 local privately-owned banks, 50 rated by Moody’s, 50 by S&P, and 74 by Fitch.3 We also split the sample into pre-crisis and crisis periods. The pre-crisis period is 30th November 1999–31st December 2006 and the crisis period is 1st January 2007–31st December 2009. This captures any differences between rating events before and during the recent financial crisis. There are far more observations in the pre-crisis period, and the bank rating upgrades clearly dominate over bank downgrades in this period, with 165 (76) upgrades (downgrades) by Moody’s, 174 (107) by S&P, and 352 (160) by Fitch. During the crisis period, bank upgrades and downgrades are more evenly matched, with 24 (29) upgrades (downgrades) by Moody’s, 60 (47) by S&P, and 71 (87) by Fitch. 3.2. Other explanatory variables We investigate the characteristics of the countries where the sovereign ceiling effect on bank ratings is the most or least pronounced. We first use the Heritage Foundation’s Economic freedom index, which indicates a ranking of a country’s policies and performance in terms of providing economic freedoms. It comprises ten elements of economic freedom, and each of the freedoms is individually scored on a scale of 0 to 100, where 100 represents the maximum freedom. The country’s overall economic freedom index is a simple average of its scores on the 10 individual freedoms. The 10 freedoms are not mutually exclusive, and therefore we also examine their effects separately. They are grouped into four categories: i. ‘Open markets’: which includes – Financial freedom. A measure of banking efficiency and independence from government control. It indicates the openness of the banking system to foreign competition, the level of government regulation of financial services, the degree of government intervention in the financial sector, and the level of financial and capital market development. – Investment freedom. A measure of the absence of constraints on the investment capital flow. It reflects restrictions on foreign exchange, payments, transfers, capital transactions, and real estate purchases, and whether any industry is closed to foreign investment. – Trade freedom. A measure of the absence of tariffs, reflecting an economy’s openness to the import of goods and services. ii. ‘Limited government’: which includes – Fiscal Freedom. A measure of the tax burden imposed by government. – Government spending. A measure of the level of government expenditures relative to GDP. Excessive government spending may lead to inefficiency, bureaucracy, lower productivity, and waste. A higher score indicates lower government spending. iii. ‘Regulatory efficiency’: which includes – Business freedom. A measure of the ability to start, operate, and close a business, reflecting overall regulation and the government’s efficiency in the process. – Monetary freedom. A measure of price stability with an assessment of price controls via microeconomic intervention.

3 The ownership of the remaining 90 banks could not be identified in BankScope as clearly belonging to one of our three categories.

– Labour freedom. A measure of the legal and regulatory framework of a country’s labour market. iv. ‘Rule of law’: which includes – Property rights. The extent to which a country’s laws protect private property rights and the extent to which its government enforces those laws. – Freedom from corruption. Indicates the extent to which corruption prevails in a country. Corruption leads to insecurity and uncertainty in economic relationships. The above data are available at annual frequency from Heritage Foundation.4 We expect bank ratings in countries with higher freedom scores to be less affected by sovereign rating actions, and vice versa. Also, the extent of the influence may vary depending on the bank’s ownership status. Some of these freedom indices are expected to have stronger effects in our context, such as Economic freedom, Financial freedom, Investment freedom, Government spending and Business freedom. Beck et al. (2006) use both Economic and Banking freedom indices, and find that countries with greater freedoms are less likely to experience a banking crisis. Macroeconomic factors are likely to affect the sensitivity of bank ratings to sovereign rating changes. To investigate this issue, we use GDP per capita, GDP growth, inflation rate, current account balance, fiscal balance, and external debt. These variables are selected to be in line with factors that are emphasised in the theoretical and empirical literature as determinants of sovereign ratings (e.g. Hill et al., 2010) and are also likely to affect the quality of banks’ assets (e.g. Beck et al., 2006). The data is obtained from DataStream, with the underlying sources being IMF, World Bank and Oxford Economics. Panel II of Table 1 provides a summary description of the Freedom indices and the macroeconomic variables. 4. Methodology The impact of sovereign rating actions on bank ratings is examined by employing the ordered probit modelling approach. This is a widely accepted approach in credit ratings literature, in order to account for the discrete, ordinal nature of credit ratings and rating changes. The model estimates the upgrade, downgrade and no rating change probabilities for the bank credit ratings.5 The rating changes are identified by notches (0, 1, and 2 or more) using the 20-point rating scale. The specification of the model is defined as follows:

Dyi;a;t ¼

2 X bSch ni;a þ cpwi;a þ kwi;a þ #rating i;a;t þ ei ;

ei

n¼0

 Nð0; 1Þ

ð1Þ

Dyi;a;t is an unobserved latent variable linked to the observed ordinal response categories yi,a,t by the measurement model: 2

3 0 if yi;a;t 6 l1 ðno rating changeÞ 6 1 if l < y 6 l ðrating upgrade=downgrade of 1 notchÞ 7 yi;a;t ¼ 4 5 1 2 i;a;t  2 if l2 < yi;a;t ðrating upgrade=downgrade of 2 or more notchesÞ

where lm represent thresholds to be estimated using the maximum likelihood estimation (MLE), along with parameters b, c, k and #, subject to the constraint that l1 < l2. i = 1, . . ., 54 countries, a = Moody’s, S&P or Fitch, t = 1, . . ., 109 months, and n = 0, 1, 2 or more notch rating change. yi,a,t is an ordinal variable; BUPi,a,t or BDNi,a,t. BUPi,a,t 4 Due to absence of data for some countries/years, the Labour freedom index is not included in our analysis. 5 A ‘no rating change’ occurs only if an issuer (in this case a bank) is either put on watch status, or taken off watch status, in which case no actual rating change has occurred.

568

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577

(BDNi,a,t) = 1, 2 if a bank from country i is upgraded (downgraded) by 1, 2 or more notches, respectively, by agency a in month t; 0 otherwise. Sch_ni,a  SUP_ni,a or SDN_ni,a. SUP_ni,a (SDN_ni,a) is a dummy variable taking the value of 1 if an emerging sovereign i is upgraded (downgraded) by n notches by agency a, up to 3 months prior to month t, with n = 1 for a 1 notch upgrade (downgrade), and n = 2 for a 2 or more notch upgrade (downgrade); 0 otherwise. pwi,a  PWi,a or NWi,a. PWi,a (NWi,a) is a dummy variable taking the value of 1 if a sovereign i was (in its previous rating action, which has a maximum 3 month ‘lookback’ period due to the CRAs target of 90 days to take action once an issuer is placed on watch) placed on positive (negative) watch by agency a, up to 3 months prior to month t; 0 otherwise. wi,a  pwi,a or nwi,a. pwi,a (nwi,a) is a dummy variable taking the value of 1 if a sovereign i is on positive (negative) watch, by agency a, up to 3 months prior to month t; 0 otherwise. ratingi,a,t = 1, 2, . . ., 19, 20. This is the transformed (numerical) rating of sovereign i by agency a, in month t. This is a control variable to account for the economic situation in the country at the time of the bank rating action. We follow recent literature in examining the upgrade and downgrade models separately, as they have been found to be driven by different factors (e.g. Livingston et al., 2008; Alsakka and ap Gwilym, 2009). The dependent variables (bank upgrade/bank downgrade) are always related to the independent variables (sovereign rating information) through the same CRA and country. Each dependent variable observation is no more than 3 months later than the independent variables. We expect sovereign rating actions to significantly affect bank ratings, and we expect positive coefficients for sovereign upgrades (downgrades), since we expect them to induce bank upgrades (downgrades). We also expect positive coefficients for the positive (negative) watch variables, because if the sovereign has recently been on positive (negative) watch status, we expect the sovereign to subsequently be upgraded (downgraded), which in turn will induce bank upgrades (downgrades). Further, we calculate the marginal effects (MEs) to estimate the economic significance of each independent variable (Livingston et al., 2008). The marginal effects show the impact of a sovereign rating action (rating change, watch) by agency a on country i on the probability of bank rating changes of 0, 1, or 2 or more notches by agency a on banks from country i. To examine whether country characteristics influence the sensitivity of bank ratings to sovereign rating changes, we estimate the following logit regression model:

CDb;i;a;t ¼ a þ w Freedom Indexi;yt þ ei

ð2Þ

where CDb,i,a,t takes the value of 1 if bank b in country i is upgraded/ downgraded at time t following a sovereign i rating upgrade/downgrade by agency a up to 3 months prior to month t, 0 otherwise. Freedom Indexi,yt is the freedom index score of country i at time yt (the year of the bank rating change). We estimate separate versions of Eq. (2) for Economic freedom and for each of its components (See Panel II of Table 1). To further examine the effect of the countries’ freedom characteristics on the sensitivity of bank ratings to sovereign rating changes, we estimate the following ordered probit model:

COb;i;a;t ¼ h Freedom Indexi;yt þ ei

ð3Þ

COb,i,a,t is an ordinal variable equal to 3 if there is a rating change for bank b from country i at time t following a sovereign i rating change (in the same direction) by agency a up to 3 months prior to month t, and the outcome is sovereign rating > bank rating, equal to 2 when the outcome is sovereign rating = bank rating, equal to 1 when the outcome is sovereign rating < bank rating, and 0 otherwise. To examine the effect of macroeconomic factors on bank ratings’ sensitivity to sovereign rating changes, the following models are estimated:

CDb;i;a;t ¼ a þ f1 GDP per capi þ f2 GDP growthi þ f3 Inflationi þ f4 Current acc bali þ f5 Fiscal bali þ f6 External debt i þ ei

ð4Þ

COb;i;a;t ¼ d1 GDP per capi þ d2 GDP growthi þ d3 Inflationi þ d4 Current acc bali þ d5 Fiscal bali þ d6 External debt i þ ei

ð5Þ

where CDb,i,a,t and COb,i,a,t are defined as in Eqs. (2) and (3) respectively; GDP per cap is GDP per capita for the previous year for country i (thousands US$); GDP growth is the average annual real GDP growth on a year-over-year basis for the previous 3 years for country i (per cent); Inflation is the average annual consumer price inflation growth on a year-over-year basis for the previous 3 years for country i (%); Current acc bal is the average annual current account balance relative to GDP for the previous 3 years for country i (%); Fiscal bal is the average annual central government deficit or surplus relative to GDP for the previous 3 years for country i (%); External debt is the total external debt relative to exports for the previous year for country i (%). The macroeconomic variables are defined in line with Hill et al. (2010).6 5. Empirical results 5.1. Model (1) – whole sample and agency comparisons Table 2 presents the estimation results of Eq. (1) for the whole sample, and also for the sample split by CRAs. For the whole sample, we find that a bank is 63.2% (19.3%) more likely to be upgraded by 1 (2 or more) notches if the sovereign had a 1 notch upgrade. If the sovereign has been upgraded by 2 or more notches, a bank is 87.1% more likely to be upgraded by 2 or more notches. We also find that banks from countries with poorer sovereign ratings are more likely to be upgraded, than banks from countries with better sovereign ratings. A bank is 26.0% (26.7%) more likely to be downgraded by 1 (2 or more) notches following a 1 notch sovereign downgrade, and 62.2% more likely to be downgraded by 2 or more notches following a 2 or more notch sovereign downgrade. If the sovereign has previously been on positive watch then a bank is 33.1% (7.0%) less likely to be downgraded by 1 (2 or more) notches, and if the sovereign is currently on positive watch then a bank is 20.2% (6.7%) less likely to be downgraded by 1 (2 or more) notches. In general, the marginal effects (MEs) are economically smaller for Moody’s than for S&P and Fitch. For sovereigns receiving a 1 notch upgrade (downgrade), a bank has increased probabilities of a rating upgrade (downgrade) of 1, and 2 or more notches as follows: Moody’s: 36.9% (14.7%), and 42.7% (13.6%); S&P: 70.7% (26.6%), and 7.7% (26.4%); and Fitch: 78.5% (20.1%), and 17.3% (33.7%), respectively. The results are stronger for sovereign upgrades (downgrades) of 2 or more notches, where a bank has subsequent increased probabilities of a rating upgrade (downgrade) of 2 or more notches as follows: Moody’s: 61.3% (17.5%); S&P: 98.0% (66.7%); and Fitch: 99.7% (87.8%), respectively. These results suggest that Fitch is the agency most likely to upgrade (downgrade) an emerging market bank following an upgrade (downgrade) of the corresponding sovereign, due to the marginal effects being larger. If a sovereign issuer was on negative watch on its previous rating action then a bank has increased probabilities of being downgraded by 1 (2 or more) notches of 9.9% (7.9%) by Moody’s, and decreased probabilities of 9.9% (8.2%) by S&P. If a sovereign 6 The GDP growth, Inflation, Current acc bal, and Fiscal bal use the average values of the previous three years to minimize the business cycle effect, reflecting the ‘rating through the cycle’ approach used by CRAs.

569

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577 Table 2 Estimation results of Eq. (1) for the whole sample and sub-samples split by agency. Explanatory variables

BUP Coef.

Whole sample SUP_1 SUP_26 SDN_1 SDN_26 PW NW pw nw Rating Moody’s SUP_1 SUP_26 SDN_1 SDN_26 PW NW pw nw Rating S&P SUP_1 SUP_26 SDN_1 SDN_26 PW NW pw nw Rating Fitch SUP_1 SUP_26 SDN_1 SDN_26 PW NW pw nw Rating

BDN t-Value

2.79 3.53 0.71 0.33 0.02 0.57 0.13 0.91 0.05 Pseudo R2

8.83** 9.98** 1.42 0.56 0.17 1.09 0.37 1.93 2.83**

2.56 2.71 0.50

8.93** 8.91** 0.81

0.05

0.31

0.32 0.07 0.02 Pseudo R2

1.03 0.12 0.71

2.66 4.90

5.57** 4.78**

0.00 Pseudo R2

0.05

4.17 6.40

10.43** 13.27**

0.00

0.01

1.65

1.63

0.01 Pseudo R2

0.17

Marginal effects

Coef.

0

1

26

82.5% 44.3%

63.2% 42.8%

19.3% 87.1%

1.6% 43.3%

1.3% #Obs.

0.3% 1244

79.6% 76.7%

36.9% 15.4%

42.7% 61.3%

33.1%

#Obs.

409

78.4% 15.1%

70.7% 82.9%

7.7% 98.0%

43.5%

#Obs.

286

95.8% 28.7%

78.5% 70.9%

17.3% 99.7%

60.4%

#Obs.

549

t-Value

Marginal effects 0

1

26

52.7% 55.8% 40.1%

26.0% 6.4% 33.1%

26.7% 62.2% 7.0%

26.9%

20.2%

6.7%

29.5%

#Obs.

904

28.3% 34.1%

14.7% 16.7%

13.6% 17.5%

17.8% 20.9%

9.9% 13.8%

7.9% 7.2%

2.2% 21.9%

1.4% #Obs.

0.8% 325

0.20 0.04 1.50 2.22 1.13 0.05 0.69 0.04 0.02 Pseudo R2

0.66 0.11 11.24** 10.34** 2.22* 0.46 2.81** 0.33 1.14

0.00 0.14 0.77 0.92 0.89 0.49 0.69 0.04 0.07 Pseudo R2

0.00 0.31 3.81** 3.5** 1.63 2.22* 2.52* 0.17 2.11*

1.76 2.16

6.75** 4.09**

53.0% 28.4%

26.6% 38.3%

26.4% 66.7%

0.58

3.16**

18.1%

9.9%

8.2%

0.27 0.14 Pseudo R2

1.65 3.79**

4.0% 27.6%

1.5% #Obs.

2.5% 206

1.75 3.14

8.18** 8.68**

53.8% 46.0%

20.1% 41.8%

33.7% 87.8%

0.11

0.54

0.19 0.04 Pseudo R2

0.94 1.34 29.3%

#Obs.

373

This table reports the results of ordered probit estimation (Eq. (1)) with robust standard errors using data from Moody’s, S&P, and Fitch. The first section reports the results for the whole sample, whilst the next three sections split the sample by agency. The dependent variable is BUP (BDN) (which equals 0, 1 or 2 if an emerging bank from country i is upgraded (downgraded) by agency a by 0, 1, 2 or more notches, respectively, in month t; 0 otherwise). SUP_ni,a (SDN_ni,a) is the sovereign rating change. PWi,a (NWi,a) is the previous watch status of the sovereign rating. pwi,a (nwi,a) is the current watch status of the sovereign rating. ratingi,t is the sovereign rating. We also estimate and report the impact of each variable on the probability of a rating change (marginal effect), but only for variables with significant (at 5% or better) coefficients. Where no coefficients are reported, there were insufficient observations for that independent variable. The estimates of the two threshold parameters are significant at the 1% level in all estimations, and are not shown here. ** Significant at 1% level. * Significant at 5% level.

is currently on positive watch, then a bank is 13.8% (7.2%) less likely to be downgraded by 1 (2 or more) notches by Moody’s. For both Moody’s and S&P, banks from countries with poorer sovereign ratings are more likely to be downgraded than banks from countries with better sovereign ratings. 5.2. Bank ownership comparisons Table 3 presents the estimation results of Eq. (1) for three subsamples according to bank ownership. We estimate sub-samples by bank ownership as robustness checks for whether the results in Table 2 are driven by banks of a particular ownership category e.g. state-owned. The results in Table 3 show that state, foreign

and local privately-owned banks are all affected very strongly by sovereign rating actions. However, ratings of local privately-owned banks are the most sensitive to sovereign upgrades, and ratings of foreign-owned banks are the most sensitive to downgrades. If a sovereign receives a 1 notch upgrade (downgrade) then a bank has increased probabilities of a rating upgrade (downgrade) of 1, and 2 or more notches as follows: state-owned: 67.3% (32.0%), and 17.7% (19.0%); foreign-owned: 57.2% (37.6%), and 25.1% (33.0%); and local privately-owned: 68.0% (17.1%), and 22.5% (17.7%), respectively. The MEs are again very strong for sovereign upgrades (downgrades) of 2 or more notches, where a bank has subsequent increased probabilities of a rating upgrade (downgrade) of 2 or more notches as follows: state-owned: 87.8%

570

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577

Table 3 Estimation results of Eq. (1) for sub-samples split by bank ownership. Explanatory variables

BUP Coefficient

State-owned SUP_1 SUP_26 SDN_1 SDN_26 PW NW pw nw Rating Foreign-owned SUP_1 SUP_26 SDN_1 SDN_26 PW NW pw nw Rating Local privately-owned SUP_1 SUP_26 SDN_1 SDN_26 PW NW pw nw Rating

2.92 3.41

BDN t-Value

5.11** 5.26**

0.25

1.06

0.30

0.50

0.05 Pseudo R2

1.33

2.79 3.35

6.53** 6.48**

*

Marginal effects

Coefficient

0

1

26

85.0% 34.5%

67.3% 53.3%

17.7% 87.8%

34.9%

#Obs.

297

82.3% 48.8%

57.2% 35.7%

25.1% 84.5%

0.55 0.03 1.20

2.54 0.04 1.42

17.1%

0.06 Pseudo R2

2.51*

2.2% 40.8%

1.8% #Obs.

0.5% 359

3.38 4.20

9.30** 8.08**

90.5% 44.3%

68.0% 50.9%

22.5% 95.4%

0.47

1.60

0.30

0.51

0.00 Pseudo R2

0.05 45.1%

10.9%

#Obs.

t-Value

0.64

1.05

1.38 1.59

4.06** 3.05**

0.55 1.14 0.85 0.05 Pseudo R2

1.63 1.92 2.44* 1.52

0.95 0.05 2.35 3.48

1.67 0.08 8.35** 6.93**

0.14 0.34 0.10 0.04 Pseudo R2

1.05 0.82 0.40 1.36

0.63 0.34 0.93 1.74 0.85 0.18 1.16 0.27 0.04 Pseudo R2

1.19 0.43 4.07** 4.51** 1.17 0.97 2.51* 1.21 1.14

Marginal effects 0

1

26

51.0% 53.4%

32.0% 19.2%

19.0% 34.2%

32.5%

19.6%

12.9%

33.2%

#Obs.

159

70.6% 57.4%

37.6% 29.8%

33.0% 87.2%

41.6%

#Obs.

277

34.8% 48.8%

17.1% 2.3%

17.7% 51.1%

42.3%

31.0%

11.3%

22.3%

#Obs.

233

6.2%

329

This table reports the results of ordered probit estimation (Eq. (1)) with robust standard errors using data from Moody’s, S&P, and Fitch, for three sub-samples according to bank ownership. For variable definitions see Table 2. We also estimate and report the impact of each variable on the probability of a rating change (marginal effect), but only for variables with significant (at 5% or better) coefficients. Where no coefficients are reported, there were insufficient observations for that independent variable. The estimates of the two threshold parameters are significant at the 1% level in all estimations, and are not shown here. * Significant at 5% level. ** Significant at 1% level.

(34.2%); foreign-owned: 84.5% (87.2%); and local privately-owned: 95.4% (51.1%), respectively. If a sovereign issuer is currently on negative watch then a stateowned bank is 19.6% (12.9%) more likely to be downgraded by 1 (2 or more) notches. If a sovereign was previously on positive watch, then foreign-owned banks are 10.9% (6.2%) more likely to be upgraded by 1 (2 or more) notches. Also, foreign-owned banks in countries with poorer sovereign ratings are more likely to be upgraded, than foreign-owned banks in countries with better sovereign ratings. If a sovereign is currently on positive watch then local privately-owned banks are 31.0% (11.3%) less likely to be downgraded by 1 (2 or more) notches. 5.3. Bank rating to sovereign rating comparison Table 4 presents the estimation results of Eq. (1) for three subsamples according to a bank’s rating relative to the sovereign rating (prior to the latest rating action). There are three categories: banks with the same rating as the sovereign (at the sovereign ceiling), banks with a poorer rating than the sovereign, and banks with a better rating than the sovereign. As a general overview, banks rated at the sovereign ceiling are the most sensitive to sovereign rating actions, whilst banks with a poorer rating than the sovereign are least sensitive to sovereign rating actions.

If a sovereign receives a 1 notch upgrade (downgrade) then a bank has increased probabilities of a rating upgrade (downgrade) of 1, and 2 or more notches as follows: banks = sovereign: 82.9% (36.4%), and 12.2% (36.5%); banks < sovereign: 36.4% (13.1%), and 39.2% (26.3%); and banks > sovereign: 47.4% (22.8%), and 30.7% (24.4%), respectively. The marginal effects are again very strong for sovereign upgrades (downgrades) of 2 or more notches, where a bank has subsequent increased probabilities of a rating upgrade (downgrade) of 2 or more notches as follows: banks = sovereign: 98.8% (85.7%); banks < sovereign: 72.8% (30.4%); and banks > sovereign: 77.9% (87.1%), respectively. Also, banks rated at the sovereign ceiling are more likely to be upgraded in countries with a poorer sovereign rating than they are in countries with better sovereign ratings. If a sovereign was previously on positive watch, banks with a better rating than the sovereign are 8.4% (13.9%) more likely to be upgraded by 1 (2 or more) notches. 5.4. Pre-crisis and crisis comparison Table 5 presents estimation results of Eq. (1) for two sub-samples of the pre-crisis and crisis periods, to capture potential policy changes by CRAs during or after the 2007–2009 financial crisis. The results suggest that bank ratings are strongly affected by sovereign upgrades in both time periods, however, the banks are less likely to

571

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577 Table 4 Estimation results of Eq. (1) for sub-samples split by bank rating relative to sovereign rating. Explanatory variables

BUP Coefficient

Banks = sovereign SUP_1 SUP_26 SDN_1 SDN_26 PW NW pw nw Rating Banks < sovereign SUP_1 SUP_26 SDN_1 SDN_26 PW NW pw nw Rating Banks > sovereign SUP_1 SUP_26 SDN_1 SDN_26 PW NW pw nw Rating

3.94 5.30

BDN t-Value

11.06** 11.93**

0.02

0.09

0.77

0.99

0.10 Pseudo R2

3.97**

2.35 2.36 0.40

6.76** 5.45** 0.75

0.19 0.49 0.47 0.50 0.00 Pseudo R2

0.71 0.75 1.04 0.95 0.16

3.54 4.32

8.16** 10.47**

1.48

2.22*

0.34

0.81

0.05 Pseudo R2

0.97

Marginal effects

Coefficient

0

1

26

95.1% 43.4%

82.9% 55.4%

12.2% 98.8%

3.5% 59.4%

3.3% #Obs.

0.2% 793

75.6% 49.0%

36.4% 23.8%

39.2% 72.8%

24.5%

#Obs.

316

78.1% 61.4%

47.4% 16.5%

30.7% 77.9%

22.3%

59.4%

8.4%

#Obs.

t-Value

0.55 0.14 2.45 3.48

1.00 0.26 13.35** 11.32**

0.07 0.10 0.06 0.01 Pseudo R2

0.75 0.29 0.53 0.41

0.40 0.08 1.03 1.05 1.02 0.41

1.17 0.15 5.38** 3.41** 1.84 1.50

0.38 0.04 Pseudo R2

1.68 1.24

1.42 3.04

3.59** 3.9**

0.62 0.24 0.10 0.07 Pseudo R2

1.15 0.45 0.19 1.70

Marginal effects 0

1

26

72.9% 63.3%

36.4% 22.4%

36.5% 85.7%

42.6%

#Obs.

587

39.4% 39.6%

13.1% 9.2%

26.3% 30.4%

13.1%

#Obs.

217

47.2% 38.7%

22.8% 48.4%

24.4% 87.1%

26.3%

#Obs.

100

13.9%

135

This table reports the results of ordered probit estimation (Eq. (1)) with robust standard errors using data from Moody’s, S&P, and Fitch, for three sub-sample according to the banks rating compared to the sovereign, where banks = sovereign, banks < sovereign, and banks > sovereign represent banks rated the same as, poorer, and better than the sovereign, respectively, prior to the latest rating action. For variable definitions see Table 2. We also estimate and report the impact of each variable on the probability of a rating change (marginal effect), but only for variables with significant (at 5% or better) coefficients. Where no coefficients are reported, there were insufficient observations for that independent variable. The estimates of the two threshold parameters are significant at the 1% level in all estimations, and are not shown here. * Significant at 5% level. ** Significant at 1% level.

be downgraded following sovereign downgrades in the crisis period compared to the pre-crisis period. If a sovereign receives a 1 notch upgrade (downgrade) then a bank has increased probabilities of a rating upgrade (downgrade) of 1, and 2 or more notches as follows: pre-crisis: 61.7% (31.8%), and 24.8% (34.3%); and crisis: 72.9% (19.0%), and 10.7% (19.3%), respectively. For sovereign upgrades (downgrades) of 2 or more notches, banks have subsequent increased probabilities of a rating upgrade (downgrade) of 2 or more notches as follows: pre-crisis: 89.3% (71.0%); and crisis: 99.6% (45.3%), respectively.

5.5. Characteristics of countries 5.5.1. Economic freedom indices Table 6 presents estimation results of Eq. (2) for the whole sample, and for the sub-samples with different bank ownerships. For the whole sample, we find, in line with our expectations, that the higher the score of Economic freedom, Financial freedom, Government spending, Investment freedom and Business freedom, the less likely banks are to be upgraded following recent sovereign rating upgrades. This implies that the strong impact of sovereign rating upgrades on bank ratings (see Section 5.1) does vary depending

on the country’s policies and performance in terms of providing economic, financial, investment and business freedoms. Comparing different banks’ ownership, the results show that local-owned banks in countries with lower (higher) scores of Economic, Financial, Investment and Business freedoms are more (less) likely than state and foreign-owned banks to be upgraded following recent sovereign rating upgrades. The ME analysis suggests that if the Economic freedom, Financial freedom, Investment freedom and Business freedom decrease by 1%, the probabilities that localowned banks will be upgraded following recent sovereign rating upgrades increase by 1.80%, 0.91%, 0.54% and 0.84% respectively, compared with 0.74%, 0.27% 0% (i.e. insignificant coefficient), and 0.77% for foreign-owned banks, and 0.69%, 0.40%, 0% and 0% for state-owned banks. This implies that local banks are the most sensitive to sovereign upgrades, which is in line with our results in Section 5.2. It is not surprising that reactions of state-owned banks’ ratings to sovereign upgrades are the least influenced by the countries’ scores of ‘freedom’. In the case of downgrades, bank ratings’ sensitivity to sovereign rating downgrades are not affected by the degree of overall Economic freedom of the countries. In other words, the strong impact of sovereign rating downgrades on bank ratings (shown in Section 5.1) is not influenced by the degree of overall economic

572

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577

Table 5 Estimation results of Eq. (1) for sub-samples split into pre-crisis and crisis periods. Explanatory variables

BUP Coefficient

Pre-crisis SUP_1 SUP_26 SDN_1 SDN_26 PW NW pw nw Rating Crisis SUP_1 SUP_26 SDN_1 SDN_26 PW NW pw nw Rating

3.07 3.58

BDN t-Value

11.32** 11.37**

0.10

0.66

0.14 0.51 0.04 Pseudo R2

0.41 1.09 2.23*

2.89 5.80

5.51** 6.4**

0.40 0.66 0.10

1.01 0.86 0.15

0.03 Pseudo R2

0.74

Marginal effects

Coefficient

0

1

26

86.5% 43.2%

61.7% 46.1%

24.8% 89.3%

1.4% 40.6%

1.1% #Obs.

0.3% 989

83.6% 48.1%

72.9% 51.5%

10.7% 99.6%

58.4%

#Obs.

255

t-Value

2.01 2.80 0.41 0.06 0.54 0.22 0.02 Pseudo R2

10.58** 9.75** 0.83 0.57 1.63 1.94 0.67

1.06 1.48

4.96** 3.17**

0.28 0.72 0.55 0.03 Pseudo R2

0.87 1.89 1.88 1.27

Marginal effects 0

1

26

66.1% 68.0%

31.8% 3.00%

34.3% 71.0%

38.9%

#Obs.

641

38.3% 36.8%

19.0% 8.5%

19.3% 45.3%

15.9%

#Obs.

263

This table reports the results of ordered probit estimation (Eq. (1)) with robust standard errors using data from Moody’s, S&P, and Fitch. We split the sample into the pre-crisis (30/11/1999–31/12/2006) and crisis (1/1/2007–31/12/2009) periods. For variable definitions see Table 2. We also estimate and report the impact of each variable on the probability of a rating change (marginal effect), but only for variables with significant (at 5% or better) coefficients. Where no coefficients are reported, there were insufficient observations for that independent variable. The estimates of the two threshold parameters are significant at the 1% level in all estimations, and are not shown here. * Significant at 5% level. ** Significant at 1% level.

freedom of the countries. However, the results reveal that bank ratings in countries with lower scores of Government spending and Trade freedom are more likely to follow the sovereign downgrades, while banks in countries with higher scores of Financial and Business freedoms, are more likely to be downgraded following sovereign rating downgrades. For different bank ownership, we find that Financial freedom is only important for foreign-owned banks, and it is plausible for foreign-owned banks in countries with lower levels of Financial freedom to be less affected by sovereign rating downgrades (ME is +0.35%). This is consistent with the findings in Section 5.2 that foreign-owned banks are the most sensitive to sovereign downgrades. The levels of Government spending and Business freedom are only important for the sensitivity of stateowned banks’ ratings, and it is not surprising that state-owned banks in countries with a higher score of Government spending are less likely to follow sovereign rating downgrades (ME is 0.77%). The insignificant coefficients for all measures of freedom in the case of local-owned banks’ downgrades suggest that their strong reactions to sovereign rating downgrades (see Section 5.2) do not vary according to the countries’ levels of freedom.7 Using the ordered probit model (Eq. (3)) to further investigate the effect of ‘freedom’ characteristics on the reactions of bank ratings to sovereign rating changes provides very similar findings to the results of the logit model (Eq. (2)). In Table 7, we only present the results of Eq. (3) for Economic freedom, Financial freedom and Government Spending in the interests of brevity, but the results for other freedom indices are available on request. Our focus in this discussion is on the marginal effects. If the Economic freedom,

7

The results of Property and Corruption freedoms are not presented here in the interests of brevity, but are available on request. They are insignificant in most cases, which is not surprising given their more distant relevance to the implications of the sovereign ceiling.

Financial freedom and Government spending scores increase by 1%, the probabilities that a bank rating will not follow a recent sovereign rating upgrade increase by 1.68%, 0.66% and 0.33% respectively. In more detail, the probabilities that a bank is upgraded and consequently has a higher rating than the sovereign increase by 0.14%, 0.05% and 0.03%, the probabilities that a bank is upgraded and consequently has a rating equal to the sovereign rating decrease by 0.88%, 0.34% and 0.17%, and the probabilities that a bank is upgraded and consequently has a lower rating than the sovereign rating decrease by 0.94%, 0.37% and 0.19%. Similar to the logit model, the results of the ordered probit model show that bank ratings’ sensitivity to sovereign rating downgrades is not influenced by the countries’ overall Economic freedom, and that the scores of Financial freedom and Government spending components show weaker effects compared to their role in the case of banks’ reactions to sovereign upgrades. Looking at different banks’ ownership, the ME analysis here also confirms that local privately-owned banks are the most (least) sensitive to sovereign upgrades (downgrades), and foreign-owned banks are the most sensitive to sovereign downgrades. 5.5.2. Macroeconomic variables Table 8 presents estimation results of Eq. (5) for three sub-samples according to bank ownership. The aim is to examine the effect of a country’s macroeconomic conditions on bank ratings’ sensitivity to sovereign rating changes, using the ordered probit model.8 The results show that the higher the GDP growth of a country, the more (less) likely bank ratings are to follow recent sovereign rating 8 The related results of Eq. (4) using a logit model to examine the effects of macroeconomic variables produce qualitatively similar results to Eq. (5). We also estimate Eq. (5) for the whole sample; we find a similar picture to that obtained from the results of the three sub-samples. The results are not presented here in the interests of brevity, but are available on request.

Table 6 Estimation results of Eq. (2) for the whole sample and sub-samples split by bank ownership. All Banks Coef. Panel I – Upgrades (1) Economic freedom (2) (3) (4)

Financial freedom Government spending Fiscal freedom Trade freedom

(6)

Investment freedom

(7)

Monetary freedom

(8)

Business freedom

#Obs. Panel II – Downgrades (1) Economic freedom (2)

Financial freedom

(3)

Government spending

(4)

Fiscal freedom

(5)

Trade freedom

(6)

Investment freedom

(7)

Monetary freedom

(8)

Business freedom

#Obs.

0.01 Pseudo 0.01 Pseudo 0.01 Pseudo 0.01 Pseudo 0.02 Pseudo 0.01 Pseudo 0.01 Pseudo 0.02 Pseudo 904

ME%

Coef.

6.43**

1.45 (51.79) 2.92% 0.59 (36.57) 2.81% 0.48 (32.02) 1.79%

0.03 Pseudo 0.02 Pseudo 0.03 Pseudo 0.004 Pseudo 0.03 Pseudo 0.01 Pseudo 0.004 Pseudo 0.01 Pseudo 297

R2 6.44** R2 4.80** R2 0.87 R

2

0.04% 1.91

R2 3.88** R2

0.27% 0.38 (29.78) 1.09%

0.95 R2 **

6.93 R2

0.07% 1.06 (46.65) 3.72%

0.59 R2 **

2.79 R

2

2.88** R2

0.03% 0.35 (27.10) 0.63% 0.35 (23.63) 0.71%

0.81 R2 3.55** R

2

R

2

0.05% 0.60 (36.44) 1.30%

1.07

ME%

Coef.

2.32*

0.69 (22.74) 0.69% 0.40 (16.56) 1.29% 0.66 (34.44) 3.06%

0.03 Pseudo 0.01 Pseudo 0.01 Pseudo 0.02 Pseudo 0.68 Pseudo 0.01 Pseudo 0.01 Pseudo 0.04 Pseudo 359

R2 2.16* R2 2.51** R2 0.30 R

2

0.02% 1.78

R2

1.63% 0.55

R2

0.09% 0.34

R2

0.03% 0.76

R2

0.20%

0.67 R2

0.23% 1.38

R

2

2.37* R2

R2 3.42**

0.13% 0.44 (21.77) 0.96%

0.95% 0.77 (49.66) 1.79%

1.38 R2

0.92% 1.78

R

2

2.23% 1.35

0.30% 1.27

R2

0.02 Pseudo 0.02 Pseudo 0.03 Pseudo 0.04 Pseudo 0.03 Pseudo 0.01

Foreign-owned t-Value

0.90% 0.01 Pseudo R2 0.04 Pseudo R2 159

0.39 2.68**

0.07% 0.95 (40.50) 3.63%

0.01 Pseudo 0.01 Pseudo 0.003 Pseudo 0.03 Pseudo 0.02* Pseudo 0.001 Pseudo 0.01 Pseudo 0.01 Pseudo 277

Local privately-owned

t-Value

ME%

Coef.

2.03*

0.74 (23.78) 0.91% 0.27 (16.36) 0.68%

0.08 Pseudo 0.04 Pseudo 0.02 Pseudo 0.01 Pseudo 0.01 Pseudo 0.03 Pseudo 0.01 Pseudo 0.04 Pseudo 329

R2 1.97* R2 1.12 R2

0.27% 1.84

R

2

0.71% 0.55

R2

0.07% 0.76

R2

0.14% 0.92

R2 **

2.71 R2

0.22% 0.77 (34.19) 2.07%

0.57 R2 **

3.37 R

2

0.08% 0.35 (26.97) 0.92%

0.46 R2

0.06% 1.77

R2 2.05* R

2

R

2

0.85% 0.60 (28.40) 1.21%

0.10 0.00% 1.56 R2

0.65% 1.43

R2

0.53%

0.004 Pseudo 0.010 Pseudo 0.01 Pseudo 0.01 Pseudo 0.02 Pseudo 0.02 Pseudo 0.01 Pseudo 0.02 Pseudo 233

t-Value

ME%

3.54**

1.80 (56.13) 3.39% 0.91 (56.07) 5.30%

R2 4.22** R2 1.46 R2

0.62% 0.93

2

R

0.20% 0.64

R2 2.52** R2

0.11% 0.54 (29.85) 1.71%

0.63 R2 2.74** R2

0.12% 0.84 (37.11) 2.09%

0.20 R2

0.01% 1.06

2

R

0.35% 1.13

R2

0.40% 0.27

R2

0.02% 1.46

2

R

0.73% 1.89

2

R

1.12%

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577

(5)

0.07 Pseudo 0.03 Pseudo 0.02 Pseudo 0.01 Pseudo 0.01 Pseudo 0.02 Pseudo 0.01 Pseudo 0.05 Pseudo 1244

State-owned t-Value

0.85 R2

0.24% 1.81

R2

1.05%

This table reports the results of logit estimation (Eq. (2)) with robust standard errors using data from Moody’s, S&P, and Fitch, for the whole sample and three sub-samples according to bank ownership. The dependent variable is CD (a dummy taking the value of 1 if bank b in country i is upgraded/downgraded at time t following sovereign i rating upgrade/downgrade by agency a up to 3 months prior to month t, 0 otherwise). The independent variable is the Freedom Index score of country i at time yt (see Panel II of Table 1 for definitions). We run separate models for Economic freedom and for each of its components. We also estimate the marginal effects (ME), i.e. the percentage changes in the dependent variables when the explanatory variable changes value by 1%, and from minimum to maximum (the latter is reported in parentheses), but only for variables with significant (at 5% or better) coefficients. * Significant at 5% level. ** Significant at 1% level.

573

Explanatory variables

Upgrades Coefficient

Downgrades t-Value

Whole sample (1) Economic freedom

0.05

7.68**

(2)

Pseudo R2 0.02

8.20**

(3)

Financial freedom

Gov spending

Pseudo R 0.01

2

4.56**

Pseudo R2 2.17*

0.03 2

(2)

Financial freedom

Pseudo R 0.01

(3)

Gov spending

Pseudo R2 0.01

2.67**

2.30

Pseudo R2 Foreign-owned (1) Economic freedom

0.02

2.41*

(2)

Financial freedom

Pseudo R2 0.01

2.83**

(3)

Gov spending

Pseudo R2 0.01 Pseudo R2

1.52

Local privately-owned (1) Economic freedom

0.07

4.52**

(2)

Financial freedom

Pseudo R2 0.03

5.34**

(3)

Gov spending

Pseudo R2 0.01 Pseudo R2

Marginal effects (%)

Coefficient

Avr |Chg|

0

1

2

3

0.91 (31.22) 2.79% 0.36 (21.01) 2.57% 0.18 (13.45) 0.68%

1.68 (58.94)

0.14 (3.51)

0.88 (25.17) 1244 0.34 (24.11) 1244 0.17 (8.32) 1244

0.94 (37.28)

0.52 (16.55) 1.03% 0.25 (10.09) 1.32% 0.20 (12.59) 0.94%

0.97 (31.64)

0.60 (21.34) 297 0.28 (12.71) 297 0.23 (9.66) 297

0.44 (11.76)

0.46 (14.44) 0.83% 0.20 (11.94) 0.94%

0.84 (26.93)

0.50 (16.08) 359 0.22 (13.51) 359

0.42 (12.81)

0.66 (40.86) 0.33 (24.15)

#Obs. 0.03 (2.75) #Obs.

0.46 (19.12) 0.38 (23.05)

0.07 (1.46) #Obs. 0.03 (1.06) #Obs. 0.02 (2.12) #Obs.

0.37 (22.25)

0.08 (1.96) #Obs. 0.03 (1.63) #Obs.

0.25% 1.28 (34.78) 3.64% 0.56 (30.69) 4.29%

#Obs. 0.05 (1.15)

2.32 (69.10) 1.02 (61.39)

#Obs.

359

0.24 (0.46)

0.68 (25.61) 329 0.31 (27.84) 329

#Obs. 0.11 (1.13) #Obs.

1.24 0.20%

#Obs.

329

0.37 (17.91) 0.19 (18.59)

t-Value

0.003

0.49

Pseudo R2 0.01

2.25*

Pseudo R 0.01

2

2.99**

Pseudo R2

0.21 (07.47) 0.17 (15.51)

0.01 Pseudo R 0.01

1.88 (43.95) 0.82 (32.42)

1

2

3

#Obs. 0.00 (0.22)

904 0.17 (13.25)

0.09 (6.81)

#Obs. 0.01 (0.14) #Obs.

904 0.21 (13.37) 904

#Obs.

159

#Obs. 0.01 (0.26) #Obs.

159 0.60 (30.76) 159

#Obs. 0.00 (0.03)

277 0.32 (19.10)

#Obs.

277

0.07%

#Obs.

277

0.45%

#Obs.

233

0.02%

#Obs.

233

0.71%

#Obs.

233

0.01% 0.13 (10.14) 0.27% 0.16 (11.37) 0.50%

0

20.28 (0.26) 0.33 (22.60)

0.11 (9.37)

0.34 2

Pseudo R2 0.02

0.05% 1.42

2.58**

Pseudo R2

0.18 (10.37)

Marginal effects (%) Avr |Chg|

0.003

0.26

Pseudo R2 0.01

2.13*

Pseudo R2 0.003 Pseudo R2

0.46

0.02

1.35

Pseudo R2 0.002

0.30

Pseudo R2 0.01 Pseudo R2

0.70% 0.37 (24.66) 2.99%

0.01% 0.19 (11.36) 0.54%

0.75 (49.06)

0.38 (22.72)

0.14 (18.56)

0.06 (3.59)

1.67

This table reports the results of ordered probit estimation (Eq. (3)) with robust standard errors using data from Moody’s, S&P, and Fitch, for the whole sample and three sub-samples according to bank ownership. The independent variable is CO (an ordinal variable equal to 3 if there is a rating change for bank b from country i at time t following a sovereign i rating change (in the same direction) by agency a up to 3 months prior to month t, and the outcome is sovereign rating > bank rating, equal to 2 when the outcome is sovereign rating = bank rating, equal to 1 when the outcome is sovereign rating < bank rating, and 0 otherwise). The independent variable is Economic Freedom, Financial Freedom, or Government spending score of country i at time yt (see Panel II of Table 1 for definitions). We also estimate the marginal effects (ME), i.e. the percentage changes in the dependent variables when the explanatory variable changes value by 1%, and from minimum to maximum (the latter is reported in parentheses), but only for variables with significant (at 5% or better) coefficients. The estimates of the three threshold parameters are significant at the 1% level in all estimations, and are not shown here. * Significant at 5% level. ** Significant at 1% level.

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577

State-owned (1) Economic freedom

574

Table 7 Estimation results of Eq. (3) for the whole sample and sub-samples split by bank ownership.

Table 8 Estimation results of Eq. (5) for three sub-samples according to bank ownership. Explanatory variables

State-owned GDP per capita

Upgrades

Downgrades

Coefficient

t-Value

2.12*

0.05

**

GDP growth

0.12

4.68

Inflation Current acc bal. Fiscal balance

0.01 0.04 0.09

1.26 1.43 2.54**

External debt

0.00

0.40

Pseudo R

2

0

1

2

3

0.95 (15.01) 2.13 (38.84)

1.77 (27.67) 3.96 (70.79)

0.14 (2.36) 0.30 (4.50)

1.17 (15.32)

0.74 (14.71)

1.57 (22.28)

2.92 (41.53)

0.77 2.56**

Inflation Current acc bal.

0.003 0.09

0.72 5.66**

Fiscal balance External debt

0.03 0.00

1.64 0.13

Pseudo R2 Local privately-owned GDP per capita 0.01

0.22

GDP growth

0.08

2.59**

Inflation Current acc bal.

0.003 0.08

0.64 5.14**

Fiscal balance External debt

0.02 0.00

0.85 0.04

0.22 (3.03)

#Obs.

2.61 (2.38)

1.92 (22.97)

1.65 (77.68)

1.22 (21.59)

297

t-Value

2.58**

0.07

**

0.13

3.46

0.004 0.05 0.05

0.54 1.66 1.09

0.0002

2.96**

Pseudo R

2

1.32 (22.60)

2.34 (42.09)

0.24 (3.12)

1.55 (25.79)

1.08 (19.42)

0.03 0.09

1.40 3.74**

1.73 (42.75)

3.15 (82.51)

0.31 (2.99)

2.04 (33.28)

1.42 (52.22)

0.001 0.02

0.28 -1.20

0.01 0.0002

0.95 7.69**

4.84%

#Obs.

359

Pseudo R2 0.06

2.57**

1.55 (26.21)

2.80 (48.40)

0.32 (4.03)

0.90 (14.43)

2.22 (37.99)

0.06

1.86

1.58 (45.47)

2.85 (90.94)

0.32 (2.12)

0.91 (30.33)

2.25 (58.49)

0.002 0.001

0.40 0.04

0.02 0.0002

0.88 2.89**

7.19%

#Obs.

329

Pseudo R2

Marginal effects (%) Avr |Chg|

0

1

2

3

1.41 (25.10) 2.50 (34.29)

2.81 (50.19) 5.00 (68.58)

0.04 (0.12)

2.35 (36.71)

0.42 (13.36)

0.07 (0.55)

4.18 (49.41)

0.75 (18.63)

0.003 (20.25) 8.68%

0.01 (40.51)

0.00 (1.97)

0.01 (35.89) #Obs.

0.00 (2.65) 159

1.87 (29.08)

3.74 (58.15)

0.05 (1.38)

3.26 (47.89)

0.44 (8.89)

0.004 (27.60) 3.98%

0.01 (55.21)

0.00 (9.30)

0.01 (43.63) #Obs.

0.00 (2.28) 277

1.21 (22.29)

2.42 (44.58)

0.03 (0.03)

1.40 (20.65)

0.99 (23.90)

0.003 (21.43) 3.04%

0.01 (42.86)

0.00 (2.29)

0.01 (31.34) #Obs.

0.00 (9.24) 233

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577

0.02 0.07

Coefficient

Avr |Chg|

4.87%

Foreign-owned GDP per capita GDP growth

Pseudo R2

Marginal effects (%)

This table reports the results of ordered probit estimation (Eq. (5)) with robust standard errors using data from Moody’s, S&P, and Fitch, for three sub-samples according to bank ownership. The independent variable is CO (see Table 7 for its definition). The independent variables are: GDP per capita, GDP growth, Inflation rate, Current account balance relative to GDP, Fiscal balance relative to GDP and External debt relative to exports of country i (see Panel II of Table 1 for definitions). We also estimate the marginal effects (ME), i.e. the percentage changes in the dependent variables when the explanatory variable changes value by 1%, and from minimum to maximum (the latter is reported in parentheses), but only for variables with significant (at 5% or better) coefficients. The estimates of the three threshold parameters are significant at the 1% level in all estimations, and are not shown here. ** Significant at 1% level. * Significant at 5% level.

575

576

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577

upgrades (downgrades). The average MEs for the GDP growth variable in the upgrade (downgrade) cases are 2.13% (2.50%) for stateowned banks, 1.32% (1.87%) for foreign-owned banks and 1.55% (0%, i.e. insignificant) for local-owned banks. This implies that state-owned bank ratings’ sensitivities to sovereign rating changes are those most affected by GDP growth. Also, local bank ratings’ reactions to sovereign upgrades are more affected by GDP growth than foreign-owned banks, while foreign-owned bank ratings’ reactions to sovereign rating downgrades are more affected by GDP growth than local banks. This is consistent with the findings in Section 5.2 that local banks are more sensitive to sovereign upgrades than foreign-owned banks, and foreign-owned banks are more sensitive to sovereign downgrades than local banks. Unsurprisingly, External debt is only important for downgrades; the higher the level of External debt, the lower the probability that bank ratings follow recent sovereign rating downgrades. If External debt changes value from minimum to maximum (or by 1%), the probabilities of state-owned, foreign-owned and local bank ratings to follow recent sovereign downgrades alter by 20.3% (0.003%), 27.6% (0.004%) and 21.4% (0.003%) on average, confirming that foreign-owned banks are the most sensitive to sovereign rating downgrades. GDP per capita is relevant for state-owned bank ratings’ reactions to sovereign rating changes and for local bank ratings’ sensitivity to sovereign rating downgrades. The higher a country’s GDP per capita, the more likely for state-owned (local) bank ratings to follow recent sovereign rating changes (downgrades). Current account balance is only influential for foreignowned and local bank ratings’ reactions to sovereign rating upgrades. The higher the Current account balance, the more likely are foreign-owned and local bank ratings to follow recent sovereign rating upgrades. Fiscal balance is only important for stateowned bank ratings’ sensitivity to sovereign rating upgrades. Bank ratings’ sensitivity to sovereign rating changes is not affected by Inflation.

6. Conclusions For emerging market countries, we analyse the effects of sovereign rating actions on the rating change probabilities of banks in the same country. The sample consists of the three global CRAs, rating 425 banks in 54 emerging countries. The results show that sovereign upgrades and downgrades have a strong impact on bank ratings, very much as expected. The previous and current sovereign watch status is most often insignificant and generally has much weaker marginal effects than the sovereign upgrades and downgrades. The signs of the coefficients of the sovereign upgrade and downgrade are always as expected. Banks are more likely to be upgraded following sovereign upgrades, than they are of being downgraded following sovereign downgrades. The strong impact of sovereign rating upgrades on bank ratings does vary depending on the country’s policies and performance in terms of providing Economic, Financial, Investment and Business freedoms. The higher the country’s freedom score, the less likely are bank ratings to follow recent sovereign rating upgrades. In contrast, bank ratings’ sensitivity to sovereign rating downgrades is not affected by the degree of overall Economic freedom of the countries. We also find that countries’ macroeconomic conditions affect bank ratings’ sensitivity to sovereign rating changes. In particular, the higher the country’s GDP growth, the more (less) likely bank ratings are to follow recent sovereign rating upgrades (downgrades), while the higher the country’s External debt, the lower the probability that bank ratings follow recent sovereign rating downgrades. Moody’s is the agency which is least likely to migrate bank ratings in tandem with the sovereign rating, since the marginal effects are economically smaller than they are for S&P and Fitch, in

particular for downgrades. Fitch is the agency which is most likely to migrate bank ratings in tandem with the sovereign rating. We also show that banks with poorer ratings than their sovereign are those which are least likely to follow sovereign rating changes, compared to banks that are rated at or above the sovereign rating. The above effects are not driven by state-owned banks. We find that foreign-owned banks are more likely to follow sovereign downgrades, while local privately-owned banks are most likely to follow sovereign upgrades. This can be explained by country characteristics, which influence the sensitivity of bank ratings to sovereign rating changes. Local privately-owned banks in countries with lower (higher) scores of Economic, Financial, Business and Investment freedoms are more (less) likely than state-owned and foreign-owned banks to be upgraded following recent sovereign rating upgrades. We find that state-owned bank ratings’ reactions to sovereign upgrades are those which are least influenced by the countries’ freedom scores. The reactions of local banks’ ratings to sovereign rating downgrades do not vary according to the countries’ ‘freedoms’, while the degree of Financial freedom is only important for foreign-owned bank ratings’ sensitivity to sovereign rating downgrades. We also find that local privately-owned bank ratings’ reactions to sovereign rating upgrades are more affected by GDP growth than are those of foreign-owned banks. Foreignowned bank ratings’ reactions to sovereign downgrades are more affected by GDP growth and External debt than are those of local banks. Also, GDP per capita mainly influences state-owned bank ratings’ reactions to sovereign rating changes, while Current account balance is mainly relevant for foreign-owned and local bank ratings’ reactions to sovereign rating upgrades. This study makes a unique contribution to the literature by examining the effects of sovereign rating actions on bank ratings. We find that emerging market bank ratings are frequently constrained by the sovereign ceiling. This highlights the importance of sovereign ratings for banks in emerging markets, and how strongly CRAs apply the sovereign ceiling in the majority of cases. An obvious question for future research is whether emerging market bank ratings should be so closely tied to the sovereign rating. These features have increasing relevance to some developed countries, as the heavy government debt burdens (e.g. in some eurozone countries) continue to make sovereign rating actions more frequent and influential.

Acknowledgements We are very grateful to an anonymous referee for motivating much of the analysis in Sections 3.2 and 5.5. We are also grateful for comments from the editor and from participants at the 2011 British Accounting Association Annual Conference.

References Alsakka, R., ap Gwilym, O., 2009. Heterogeneity of sovereign rating migrations in emerging countries. Emerging Markets Review 10, 151–165. Alsakka, R., ap Gwilym, O., 2010. Leads and lags in sovereign credit ratings. Journal of Banking and Finance 34, 2614–2626. Bannier, C., Hirsch, C., 2010. The economic function of credit rating agencies – what does the watchlist tell us? Journal of Banking and Finance 34, 3037–3049. Bannier, C., Behr, P., Güttler, A., 2010. Rating opaque borrowers: why are unsolicited ratings lower? Review of Finance 14, 263–294. Beck, T., Demirgüç-Kunt, A., Levine, R., 2006. Bank concentration, competition, and crisis: first results. Journal of Banking and Finance 30, 1581–1603. Borensztein, E., Cowan, K., Valenzuela, P., 2007. Sovereign Ceiling Lite? The Impact of Sovereign Ratings on Corporate Ratings in Emerging Market Economies. Working Paper. IMF. Brooks, R., Faff, R., Hillier, D., Hillier, J., 2004. The national market impact of sovereign rating changes. Journal of Banking and Finance 28, 233–250. Caporale, G., Matousek, R., Stewart, C., 2012. Ratings assignments: lessons from international banks. Journal of International Money and Finance 31, 1593–1606.

G. Williams et al. / Journal of Banking & Finance 37 (2013) 563–577 Duggar, E., Emery, K., Gates, D., Paulo, S., Lemay, Y., Cailleteau, P., 2009. Emerging market corporate and sub-sovereign defaults and sovereign crises: perspectives on country risk. Moody’s Investors Service (February). Durbin, E., Ng, D., 2005. The sovereign ceiling and emerging market corporate bond spreads. Journal of International Money and Finance 24, 631–649. Ferreira, M., Gama, P., 2007. Does sovereign debt ratings news spill over to international stock markets? Journal of Banking and Finance 31, 3162–3182. Ferri, G., Liu, L.G., Majnoni, G., 2001. The role of rating agency assessments in less developed countries: impact of the proposed Basel guidelines. Journal of Banking and Finance 25, 115–148. Gande, A., Parsley, D., 2005. News spillovers in the sovereign debt market. Journal of Financial Economics 75, 691–734. Hamilton, D., Cantor, R., 2004. Rating transitions and default rates conditional on outlook. Journal of Fixed Income 14, 54–71. Hill, P., Brooks, R., Faff, R., 2010. Variations in sovereign credit quality assessments across rating agencies. Journal of Banking and Finance 34, 1327–1343. Iannotta, G., 2006. Testing for opaqueness in the European banking industry: evidence from bond credit ratings. Journal of Financial Services Research 30, 287–309. IMF, 2010. Resolving the Crisis Legacy and Meeting New Challenges to Financial Stability. IMF Global Financial Stability Report: Meeting new Challenges to Stability and Building a Safer System (April). Ismailescu, I., Kazemi, H., 2010. The reaction of emerging market credit default swap spreads to sovereign credit rating changes. Journal of Banking and Finance 34, 2861–2873.

577

Kim, J., Wu, E., 2008. Sovereign credit ratings, capital flows and financial sector development in emerging markets. Emerging Markets Review 9, 17–39. Kim, J., Wu, E., 2011. International bank flows to emerging markets: influence of sovereign credit ratings and their regional spillover effects. Journal of Financial Research 34, 331–364. Lando, D., Sködeberg, T., 2002. Analyzing rating transitions and rating drift with continuous observations. Journal of Banking and Finance 26, 423–444. Livingston, M., Naranjo, A., Zhou, L., 2008. Split bond ratings and rating migration. Journal of Banking and Finance 32, 1613–1624. Morgan, P., 2002. Rating banks: risk and uncertainty in an opaque industry. American Economic Review 92, 874–888. Norden, L., Weber, M., 2004. Informational efficiency of credit default swap and stock markets: the impact of credit rating announcements. Journal of Banking and Finance 28, 2813–2843. Poon, W., Lee, J., Gup, B.E., 2009. Do solicitations matter in bank credit ratings? Results from a study of 72 countries. Journal of Money, Credit and Banking 41, 285–314. Shen, C.-H., Huang, Y.-L., Hasan, I., 2012. Asymmetric benchmarking in bank credit rating. Journal of International Financial Markets, Institutions & Money 22, 171– 193. Vazza, D., Leung, E., Alsati, M., Katz, M., 2005. Creditwatch and ratings outlooks: valuable predictors ratings behaviour. Standard and Poor’s (May).