BIS Working Papers
Does sovereign risk in local and foreign currency differ?
by Marlene Amstad, Frank Packer and Jimmy Shek
Monetary and Economic Department
JEL classification: G15, F30, G21, G28
Keywords: Global banks, dollar funding, money market funds, relationship frictions, US Money Market Fund reform
This publication is available on the BIS website (www.bis.org).
© Bank for International Settlements 2017. All rights reserved. Brief excerpts may be reproduced or translated provided the source is stated.
ISSN 1020-0959 (print)
ISSN 1682-7678 (online)
Does sovereign risk in local and foreign currency differ?
Marlene Amstad, Frank Packer and Jimmy Shek
Historically, sovereign debt in local currency has been considered safer than debt in foreign currency. Yet the literature offers scant theoretical or empirical guidance as to why such a gap exists, or why it appears to have slowly and steadily diminished for all regions over the past two decades, as expressed in the ratings widely used by global investors and regulators to assess credit risk. We suggest and empirically test five hypotheses. We find that differences in inflation do not explain the assessed gaps between local and foreign currency credit risk. The banking sector’s vulnerability to sovereign debt problems is a significant determinant of the spread, but does not account for its decline over time. Instead, the surge in global reserves, and to lesser extent the reduced reliance on overseas foreign currency borrowing (ie the decline of original sin), as well as lower global volatility, appear to have lessened the gap. But if these variables were to go into reverse, the gap could again widen.
JEL classification: F31 F33 F34 F41 H63.
Keywords: sovereign risk, local currency debt, foreign currency debt, credit ratings.
The authors appreciate comments from Claudio Borio, William Coen, Torsten Ehlers, Leonardo Gambacorta, Ko Nakayama, Ugo Panizza, Andrew Rose, Toshitaka Sekine, Hyun Shin, Shigenori Shiratsuka, Christian Upper, Charles Wyplosz and other participants in seminars at the Bank of Japan, the Reserve Bank of Australia and the Bank for International Settlements, as well as from an anonymous referee. This research was supported with a research fellowship by the BIS. The paper was also prepared in part while Marlene Amstad was a visiting scholar at the Institute for Monetary and Economic Studies (IMES), Bank of Japan, and has also been published in the IMES Discussion Paper Series, No. 2018-E-1. The views expressed in this paper are those of the authors and do not necessarily reflect the official views of the Bank for International Settlements or the Bank of Japan.
This paper asks whether the sovereign risk in domestic currency differs in level and underlying determinants from its counterpart in foreign currency. The answers have wide-ranging consequences for regulators and investors. With financial markets that are more and more global, and local and foreign currency instruments at times trading on the same platforms in any market, the distinction between local and foreign currency credit risk can be seen as increasingly important.
Sovereign debt in local currency has historically been considered safer than sovereign debt in foreign currency. Bonds issued by the same emerging market sovereign were almost always rated higher when denominated in the sovereign's own domestic currency rather than a foreign currency, reflecting the view that sovereigns are much less likely to default on local currency obligations as they can always tax their subjects or print their own currency to service the debt.
And yet, the view that local currency debt is safer has diminished over the past few decades. Graph 1 illustrates for a broad sample of 73 emerging economies (EMEs) that the average difference in the average local currency rating minus its foreign counterpart used to be as large as three notches in 1996 but had almost fully closed by 2015. In part this may be due to the increased recognition that sovereigns do default in local currency: the seminal work of Reinhart and Rogoff (2009) documents that de jure defaults of sovereigns in their own currency have been rather common since 1800.
Second, we document a slow and steady convergence of sovereign risk in local and foreign currency over the past 20 years. The proportion of sovereigns without any gap has risen steadily, while gaps larger than one rating notch is now extremely rare. Third, we analyse the determinants of the difference in local versus foreign sovereign risk, and their statistical and economic significance. We do so first by estimating separate panel regressions for foreign and local currency sovereign risk between 1996 and 2015. Then we estimate the probability and extent of a gap in local and foreign currency risk using a panel trinominal ordered logistic regression. We also approximate the empirical relevance of different factors in explaining the observed closing of the gap over the past twenty years.
For measures of sovereign risk in both foreign and local currency, we take the average of credit ratings provided by the three major global agencies, Standard and Poor's, Moody's and Fitch. This is motivated by several considerations. First, global investors routinely reference credit ratings when measuring credit risk. Second, while there have been international efforts to reduce the reliance on ratings (FSB, 2014), sovereign ratings remain a tool in many regulatory frameworks, e.g. to weigh the bank capital requirements for sovereign exposure. Third, rating agencies provide assessments of both local and foreign credit risk across a very large number of countries; by contrast market-based measures of sovereign risk such as sovereign CDS are for the most part available only for foreign currency obligations. Further, market-based measures are impacted by various capital market frictions and nearterm market conditions (including the pricing of liquidity and counterparty risk), while ratings are by construction aiming to abstract from these and focus on credit risk over the medium to longer term. In addition, compared to ex-post realisations of credit risk, i.e. defaults, ratings are available over a much wider array of countries, and at regular frequencies.
We find limited evidence of sovereigns' willingness to inflate away their local debt in our measures of credit risk. Thus, declining trend inflation has little to do with a diminishing distinction between foreign and local currency sovereign risk. Instead, higher FX reserves, lesser dependence on foreign currency borrowing (decline of original sin), and lower global volatility better account for the observed narrowing of the gap. In particular, our results point towards the role of an unprecedented two- decade long surge in FX reserves in lifting the relative creditworthiness of foreign currency debt. If FX reserves should drain, ability to borrow in local currency decline and global volatility further increase, more significant gaps in local and foreign currency sovereign risk may re-emerge.
The paper is organised as follows. Section 2 reviews the literature and develops five hypotheses about what might justify a gap between local and foreign currency sovereign risk. Section 3 introduces the data used to test the hypotheses as well as the empirical methodology. Section 4 reports the results, and Section 5 summarises and draws some policy conclusions.2. What might justify a gap?
The literature on the factors determining sovereign credit risk is large. However, sovereign risk has almost always been discussed in the context of external debt. As exceptions, Reinhart and Rogoff (2009, 2011) turned their attention to the so-called "forgotten history" of domestic debt and related defaults. Many models assume that domestic debt is always honoured, not least because governments could simply print money and inflate the problem away. In fact, Reinhart and Rogoff document that, while domestic default has been somewhat less frequent than for external debt, it has not been rare; there have been nearly 70 cases since 1800 vs. around 250 defaults on external debt.
Instead of focusing only on inflation, we propose five hypotheses why the credit risk entailed in local currency debt might differ from that of foreign currency sovereign debt. All of these hypotheses presume that local currency is more creditworthy, but that the degree of the difference depends on several factors. Section 2.1 presents the five hypotheses and related explanatory variables. Section 2.2 reviews commonly used control variables that are commonly related to sovereign creditworthiness in the literature, but for which we have no strong prior of a differential effect across local vs. foreign currency obligations.2.1. Five hypotheses
The creditworthiness of domestic currency sovereign obligations is often viewed as higher than that of foreign currency debt, reflecting the presumed greater ability and willingness of sovereigns to service debt in their own currency. The empirical literature to date has generally focused exclusively on the determinants of sovereign risk in foreign currency, rather than on whether and how the determinants might differ between foreign and local currency sovereign risk. Below, we suggest five hypotheses for how the gap between foreign and local currency risk might be related to observable factors. Our paper empirically estimates the influence and significance of the factors. To the best of our knowledge, we are the first to propose and test these hypotheses.
The first factor we focus on is inflation. Most models of sovereign risk posit a negative relationship between a sovereign's inflation and the creditworthiness of its external debt. This is because high inflation is a good proxy for problems in the government's finances as well as political and other forms of institutional instability (Cantor and Packer, 1996). Such problems should negatively affect the creditworthiness of domestic debt as well.
However, the relationship between inflation and domestic currency debt has another dimension: namely, to the extent that printing currency is inflationary, the sovereign has the ability to generate inflation to reduce the real burden of local currency debt. This consideration suggests that inflation has a positive aspect in the case of local currency debt, since the ability of sovereigns to repay may be positively affected by inflation in a fashion that is not the case with foreign currency debt. By this logic the gap between domestic and foreign currency risk should widen for high inflation countries.
A competing hypothesis with opposite implications should be kept in mind. Reinhart and Rogoff (2011) note high inflation episodes tend to go "hand in hand" with domestic default. This suggests that the sovereign's willingness to monetise rather than default on domestic debt may in fact be diminished in times of high inflation, perhaps because of increased awareness of the costs of hyperinflation and political resistance to further inflation at high levels.
The importance of reserves to creditworthiness was taken to heart by many nations in the wake of the Asian financial crisis of the late 1990s, as illustrated by the massive build-up of reserves by central banks in the region over the past few decades. The Asian financial crisis was triggered by the inability of many countries in the region to service their foreign currency debts.
Here we use the term "original sin" as short-hand for the lack of development of the local currency international debt markets relative to those in foreign currency. The ability to raise funds internationally in local currency has been shown to be a general positive for the creditworthiness of external sovereign debt in many empirical models, though it may have lost some explanatory power since the global financial crisis (Amstad and Packer, 2015). There are many possible reasons for this importance, including the fact that the ability of a nation's borrowers to raise local currency funds in overseas markets may proxy for other aspects of financial, social and economic development. However, the reason we focus on here is that having large local currency borrowing in international markets is likely to be highly correlated for the degree to which the sovereign as well as large corporate borrowers in the country have debt burdens that are hedged against movements in the FX value of the currency, and thus less risky.
To the extent that the lesser dependence on foreign currency debt improves sovereign creditworthiness through the currency risk hedging channel, we would expect the positive impact to be much more pronounced in the case of the creditworthiness of foreign currency liabilities. Local currency debt is already likely to be supported by local currency revenue, whereas foreign currency debt can be influenced by the risks of currency mismatch.
Another factor we concentrate on is the exposure of the banking sector to sovereign bonds. While it has been known for some time that the fiscal costs of banking crises are immense (Bordo et al, 2001), and thus banking system problems can threaten the creditworthiness of the sovereign, more recently there has been recognition of the damage that declines in sovereign creditworthiness can themselves impose on the banking sector. This suggests a reinforcing negative feedback loop. In the well-known model of Acharya et al (2014), the channel for this is the sovereign debt owned by the bank, and the potential for collateral damage on the bank balance sheet. Gennaioli et al (2014) document how the decline in lending following a sovereign default is larger in those countries where financial institutions are more developed and banks hold more sovereign bonds, which in turn implies an even greater vulnerability for the sovereign.
Given that the domestic banking sector tends to hold domestic rather than foreign currency government debt, the domestic banking sector will be more damaged by defaults on domestic currency debt, and the "doom loop" will be more likely to operate through that channel. Thus, ceteris paribus, we would expect any advantage to local currency debt to be reduced in countries where banks have greater exposure to the sovereign, and for the gap between domestic and foreign currency sovereign obligations to be smaller.
Similar to the effects of inflation earlier, an alternative hypotheses with opposite implications should be kept in mind. Due to knowledge of the damage that banking sector crises can cause, the sovereign's willingness to pay can possibly be improved by banks' holdings of government bonds. Once again, this effect can be viewed to act more strongly on the local currency obligations more likely to reside on banks' balance sheets. The ultimate sign of the influence of bank holdings of sovereign debt on overall creditworthiness as well as the gap thus depends on the relative importance of its impact on the sovereign's ability versus willingness to pay.
It is likely that the impact from any shift in global market volatility would have more impact on foreign than local currency credit risk. For one, VIX measures implied volatility for a US dollar-based index, and thus should correlate more with the risks of foreign currency sovereign debt obligations, usually also denominated in US dollars. In addition, global credit markets respond to swings in the risk appetite of global investors, who trade more in foreign than local currency securities. Local currency debt should be more naturally sheltered from the spill-overs of global volatility. Indeed, Du and Schreger (2016) report the pass-through of VIX into their measure of the local currency credit spread as small and not statistically different from zero for most of the sample countries. In stark contrast they find the pass-through of VIX into the foreign currency credit spread as statistically significant and positive in most countries.2.2. Control variables
With the above five hypotheses, we have proposed five variables that might influence sovereign risk in domestic debt differently from that of foreign debt. Needless to say, the literature suggests that many other variables that might influence sovereign credit risk more generally; but these do not necessarily have a differential effect. As control variables, we take measures of economic, fiscal and institutional strength as well as the default history. These variables include log per capita GDP, public debt/GDP, interest/revenue, short-term external debt over GDP, the exchange rate regime, corruption perceptions (See Table 2a and 2b for a description of all variables used in the analysis).
2.3. Related Literature
Compared to research on foreign currency sovereign risk, the literature on the determinants of local currency sovereign risk measured by ratings, or more specifically, the distinction between local and foreign currency sovereign ratings, is relatively sparse. Packer (2003) provides an early overview of the sovereign ratings "gap", noting the discrepancy between the higher sovereign ratings for local versus foreign currency, and the lack of a uniform relationship between the denomination of debt and the likelihood of default. As for the determinants of local versus foreign currency ratings, Kisselev and Packer (2006) undertake an analysis of the gap between local and foreign currency ratings for 101 countries. In contrast to our first hypothesis above, they found that higher inflation diminished local currency ratings more than foreign ratings and thus led to a smaller gap. The authors did not explore dependence on foreign currency borrowing, banks' holding of government debt and global volatility as possible determinants, however. They also estimated regressions only over the period 1995-2003, when domestic currency ratings were still a relatively recent product of the ratings agencies.
Another related strand of literature uses foreign versus domestic currency defaults as the dependent variable. As noted earlier, Reinhart and Rogoff document the frequency of the two types of sovereign default, but in terms of assessing the determinants of the two types, the principal reference is now the recently published work of Jeanneret and Souissi (2016). They find evidence consistent with some of the above-mentioned hypotheses: ie, a larger banking sector would make a sovereign less likely to default on domestic debt. But they find that high levels of inflation make a sovereign more likely to default on local currency obligations not less. Further, they do not find evidence that global factors - as proxied by VIX and other metrics - make a difference for either local or foreign currency defaults.
3. Data and Methodology
3.1. Foreign and Local Currency Credit Ratings
We follow the practice of many global investors and regulators and measure sovereign creditworthiness in foreign and local currency using the sovereign credit ratings of the three major global rating agencies, Fitch, Moody's and Standard and Poor's. Our sample comprises 73 emerging market economies for which we are able to obtain both foreign and local currency sovereign ratings from at least one of the three major rating agencies at some point between 1996 and 2015 (Table 1). We take year-end ratings, as these are most likely to reflect economic developments of the year as captured in the annual data.
For some of the quantitative analysis, we transform the ratings data, drawing upon the method for estimating the determinants of sovereign ratings used by Cantor and Packer (1996). Namely, we convert ratings, averaged across the major agencies, into a linear variable, with AA+/Aa1=19, AA/AA2=18 and so forth. For the most part, we rely on multinomial ordered logistic regressions where the average rating is treated as a qualitative dependent variable. While non-linearity may well characterise certain relationships between indicators and ratings, the alternative linear regression approximation has been fruitful in a wide array of empirical investigations, for use of ratings as both explanatory and dependent variables. Therefore we also report OLS specifications, where ratings are measured numerically, as a sensitivity test.
Similarly, for our base case analysis of the ratings gap, we rely on trinomial ordered logistic regressions that only distinguish between the existence of a gap and no gap, as well as the existence of a large gap of two notches or more. But as a sensitivity test, we run linear regressions where the "gap" is calculated as the numeric difference between the numeric values of the local and foreign currency ratings (Annex Table 2).
Local currency sovereign ratings are a relatively recent product of the major rating agencies. The sovereign ratings given by the major rating agencies when they first entered the business in the early 2oth century were ratings on foreign currency
obligations, as sovereigns needed these to access international bond markets (Flandreau et al, 2011). However, starting in the 1990s, in part reflecting efforts to increase investor participation in domestic currency bonds, an increasing percentage of sovereigns obtained local currency ratings as well. From 5% of our EME sample in 1995, more than one-half had a local currency rating by 2000, and the share had steadily increased to include all EMEs in our sample from 2013 onwards (Graph 2) This rapid increase in the provision of local currency ratings has been broadly shared among the regions of Emerging Asia, Latin America, Emerging Europe, and the Middle East and North Africa.
As mentioned in the introduction, the difference between local and foreign currency ratings for the same sovereign - or the "ratings gap" - has steadily declined over the sample period. The average ratings gap started at 3 notches in 1995, had fallen to 1 notch in 2005 and has been close to zero from 2013. Encompassing all regions, the gap has narrowed as a result of the foreign currency rating increasing towards a more stable local currency rating. In terms of frequencies, while a two or more notch gap existed for nearly four-fifth the sample in 1996, it had declined to 5% by 2015 (Graph 3). By contrast, the percent of sovereigns with no gap had risen from less than 20% in 1996 to 75% in 2015. In advanced economies, the decline in the ratings gaps occurred more rapidly and from a lower starting level.
3.2. Explanatory variables
The explanatory variables we use to test our main hypothesis described in the previous chapter are inflation, foreign reserves, international debt securities and loans issued in domestic versus foreign currencies ("original sin"), banking sector exposure to government debt, and VIX. The measurement for all variables except "original sin" is straightforward. As a robustness test, we include a narrow and a broad measure for original sin, whereby the narrow version includes only international debt securities, while the broad version also includes bank debt. The definitions and sources for the explanatory and control variables are provided in Tables 2a and 2b, and some summary statistics in Table 3. We also include country and/or time fixed effects in all specifications.
We proceed to test the five main hypotheses using two complimentary methods. First, we estimate two regressions, of foreign and local currency ratings, respectively, on the same set of explanatory variables and compare the sign, size and significance of the corresponding coefficients. Specifically, we run panel multinomial ordered logistic regressions separately for foreign and local currency ratings, where we estimate the probability of the ratings being in one of 19 categories between Ca/CC and AA+/Aa1. y,Li,t represents the rating of each sovereign i at time t, indexed to be either foreign or local:
We employ variables related to the main hypotheses (Xitthypothesis) as well as the control variables (Xi,tcontrolJ introduced above. We run statistical significance tests
comparing the sign of the coefficient on the local and foreign currency rating regressions. As another robustness check, we also report in the annex the result of the OLS regression, in which the same explanatory variables are used to estimate the average of the numerical value of the ratings, as calculated according the linear transformation described earlier: AA+/Aa1=19, AA/Aa2=18, and so forth.
Secondly, we focus on the gap itself, and estimate trinomial ordered logit models where we estimate the probability of the ratings gap being in one of three categories - less than or equal to zero; between zero and 2 notches, and 2 notches or greater - to be a function of the above explanatory variables:
Although this second exercise does take into account the degree of the ratings gap, it is much less granular than the earlier multinomial logits. Rather, the question asked is simply whether the existence or non-existence of a ratings gap, as well as the existence of a very large ratings gap, is consistent with the main hypotheses. Once again, we run robustness checks on the statistical significance of the estimates, reporting different definitions of the original sin variable, and reporting in the annex the result of OLS regressions of the same explanatory variables on the numerical value of the average gap in terms of notches.4. Results
We now turn to regression models of foreign and local currency ratings that include variables that are directly related to our hypotheses introduced in section 2, and yet at the same time control for a number of key factors. In 4.1., referring to different specifications of an ordered logistic regression and checking against an OLS regression, we report whether the sign and relative size of the coefficients in local and foreign currency risk regressions are consistent with our five hypotheses. In 4.2, focusing on regressions where the likelihood and extent of the gap is the dependent variable, we test which drivers of the gap are statistically significant, and in 4.3 we explore their quantitative contribution to the decline of the gap over the past two decades.4.1. The drivers of foreign versus local currency sovereign risk
The separate ordered logistic regression samples for foreign currency ratings in columns (i) and (ii) comprise 56 EM sovereigns; the sample for local currency ratings in columns (iii) and (iv) comprise 47. Results from the base specification of the multinomial logit with country fixed effects are shown in Table 4 (results from alternative specifications using OLS are shown in Annex Table 1; results from logit models using different combinations of country and time fixed effects are shown in Annex Tables 3a and 3b). As in other work, we find that ratings both foreign and local are well explained by a relatively parsimonious set of variables, with adjusted R- squared of more than 44% for logit and around 90% in all reported OLS specifications. The control variables related to government debt, international short-term debt burden, economic development (per capita GDP), institutional development (corruption), default history and exchange rate regimes generally have the expected sign and the coefficients are usually at levels of statistical significance. In those few cases where the sign is not as expected, the coefficient is with only one exception insignificant.
Inflation is a highly significant (negative) driver of credit risk in both regressions, however there is little evidence for a lesser impact on local currency credit risk as predicted by H1. Indeed, comparing Table 4's specification (i) with (iii), as well as (ii) with (iv), we see that the estimated negative impact of inflation is actually greater for local currency ratings than foreign, which goes against hypothesis H1. This latter outcome is consistent with the view that at high levels of inflation, the negative impact on local currency creditworthiness can be greater due to the increased costs of monetization and increased incentives to default on local currency debt. In fact, test statistics suggest that in the case of at least one specification, the impact of local currency creditworthiness is significantly more negative, in a statistical sense, than that on foreign creditworthiness. Similar to the findings of Jeanneret and Souissi (2016) with regard to default, we find that despite the sovereign's often cited ability to turn on the printing press and inflate out of local currency obligations, the relative benefits to local currency obligations from a high inflation environment are limited.
The alternative specifications using OLS (Annex table 1) or time fixed effects (Annex tables 3a and 3b) shows also an estimated impact of inflation on local currency creditworthiness that is more negative than that on foreign currency creditworthiness.
One possible reason that inflation does equal or more damage to the ratings of local currency obligations may be the fact that some sovereigns issue inflation- indexed bonds. In those cases, any incentive to tolerate higher inflation for the sake of relieving the burden of local currency obligations would presumably be reduced. We report in the Annex estimations where a variable has been added which interacts inflation with a dummy variable which is 1 in a given year when the share of inflation- indexed bonds of overall sovereign debt is higher than 5% and zero otherwise. In two of four specifications, the negative impact of inflation on both local and foreign ratings is reduced when the sovereign issues indexed bonds (Annex Table 4). However, for those sovereigns without indexing, the greater negative impact on local currency obligations relative to foreign currency obligations is maintained in all specifications.
As anticipated by H2, FX reserves have a positive impact of much higher economic and statistical significance for the assessed creditworthiness of foreign currency debt. The difference in coefficients is statistically significant regardless of which definition of the original sin variable is used in Table 4. For a rough estimate of their economic impact, we refer to the coefficients of the OLS specification of Annex Table 1, where an increase in the ratio of FX reserves to GDP by 10 percentage points will raise the foreign currency rating between 0.3-0.5 notches and local currency rating around 0.2 notches. This is consistent with the well-established view that reserves should help to protect against a shortage of foreign currency, and thus should improve foreign currency creditworthiness. The time fixed effect models in Annex Tables 3a and 3b also show FX reserves as expected to be positively related to creditworthiness but does not show as clear a difference between the coefficients on FX reserves in the foreign versus local currency equations.
So-called original sin, which reflects the degree to which a country's borrowers are dependent on foreign currency denominated debt to raise overseas funds, is also a significant determinant of sovereign creditworthiness, and consistent with H3, it has a disproportionate impact on foreign currency obligations. While less original sin improves the credit ratings of both local and foreign currency obligations, it does more so for foreign currency ratings (Table 4). When using the broad definition of original sin in the OLS regressions (i) and (iii) of Annex Table 1, a 10 percentage point decline in original sin corresponds to around a 0.4 notch increase in the foreign currency rating, as opposed to a 0.1 notch increase in the local rating. Similarly, when using the specifications with the narrower measure for original sin ((ii) and (iv) of Annex Table 1), the impact of a 10 percentage point decline in original sin corresponds to a 0.15 notch increase in the foreign currency rating, as opposed to an increase of 0.03 notch in the local rating, which is also of much lower statistical significance. In fact, test statistics suggest that in the case of both pairs of specifications reported in Table 4, the impact of original sin on foreign currency creditworthiness is significantly more negative, in a statistical sense, than that on local creditworthiness. Similarly, the time fixed effects model documents a greater negative impact of original sin on foreign currency obligations.
The results also support H4, as banking sector exposure to government debt is significantly negatively related to sovereign creditworthiness in all specifications in Table 4. The negative relation is also stronger for local currency debt, indicating that the bank holdings of sovereign debt are associated with a narrower gap. In equations (i) and (ii) of Annex Table 1, a 10 percentage point increase in bank holdings of sovereign debt leads to a drop in foreign currency ratings of 0.08 notches and 0.11 notches, respectively, and 0.16 and 0.19 in local currency. That bank holdings of government debt are associated with a decline in the sovereign's creditworthiness strongly suggests that the increased risks of a feedback loop between financial institutions and sovereign debt, resulting from greater bank exposure to the sovereign, dominate any potential positives from increased willingness to pay on the part of the sovereign.
Consistent with H5, global volatility does reduce creditworthiness, and much more so for foreign currency obligations in both an economic and a statistical sense (Table 4). This is consistent with the view that the risk on foreign - mostly USD denominated - currency obligations is more vulnerable to swings in USD referenced global factors such as VIX. Estimating the economic impact from the OLS regression result (i) reported in Annex Table 1, an increase of VIX by 10 percentage points lowers the foreign currency rating by more than 0.2 notches, and statistically significantly so, versus a much smaller and insignificant effect on the domestic currency rating.4.2. Predicting the gap
This section explores the determinants of the difference in credit risk from a different perspective, asking whether the same explanatory variables examined earlier affect the probability and extent of a gap in local and foreign currency sovereign risk for any single sovereign. To some extent, we have already done this when comparing the size of coefficients in the local and foreign currency regressions, though now we will estimate regressions where the gap itself is the dependent variable. We run a trinomial logistic regression on the same variables as before, distinguishing between the three categories of no gap, a small gap above zero up to two notches, and a large gap that rounds to two notches or more. We include country fixed effects.
The explanatory power of Table 5's specifications is greater than that reported in the earlier regressions of Table 4. The pseudo R-squared measure suggests that between 60-70% of the variation in outcomes can be explained by the variables.
Despite the marked difference in empirical approach, there remains a lack of support for H1; rather, the contrary hypothesis that high levels of inflation might hurt local currency ratings more than foreign currency ratings, and thus reduce the gap is supported in one of the two specifications in Table 5. When the variable that interacts inflation with the existence of inflation-indexed bonds is added to that specification, the negative association of inflation with the gap is eliminated when there in indexing (Annex Table 5). However, in no case is a significant positive relationship between the gap and inflation uncovered. Conceivably, at high levels of inflation the risks of hyperinflation are too great to avoid default via the printing press.
On the other hand, the support in Table 5 for hypotheses (2), (3), (4) and (5) is quite solid. FX reserves are negatively associated with the difference between local and foreign currency creditworthiness, and the variable is highly significant independent of the specification. The hypothesised original sin effect, or the tendency of greater dependence on foreign currency borrowing to have a disproportionately negative impact on foreign currency risk, is supported by a positive sign for the coefficient on the gap in both specifications of Table 5, and is highly statistically significant for the broad definition (p=.0001) and only slightly less so for the narrow definition of original sin (p=.0561). Bank ownership of sovereign debt is negatively associated with the gaps between local and foreign currency creditworthiness, consistent with a stronger "doom loop" disproportionately reducing the local currency rating. The p-values for the statistical significance of the coefficients in the two specifications are 0.0891 and 0.0001, respectively.
The coefficients on global volatility, as represented by VIX, also support hypothesis (5). Global volatility tends to be associated with an increase in the gap between foreign and local currency credit risk. The earlier regressions of Table 4 suggest this is because foreign currency creditworthiness is more negatively impacted than local currency creditworthiness by global volatility.
Finally, all the main findings are robust to a transformation to a linear model. The OLS regression estimations in Annex Table 2, where the dependent variable is defined simply as the average notch difference in local and foreign currency ratings, reports similar results, supportive of the hypotheses on reserves (H2), original sin (H3), banking sector exposure (H4) and volatility (H5).4.3. Accounting for the gap’s decline
At the outset of this paper, we noted that the gap between local and foreign currency ratings had declined over the last two decades. What light do the above results shed on the reasons for this? We focus on FX reserves, original sin, and global market volatility: three variables that the regression analysis suggests drive the gap. FX reserves has increased over the sample, while both reliance on foreign currency debt and VIX has declined. Combined with the signs on the coefficients in the ratings regressions, this implies that each of these three variables could have accounted for some of the declining gap over the sample period. (The measure of banking sector exposure, although a significant determinant in the panel regression and volatile over the period, was little changed at the end of the sample period from the beginning, which suggests it would not account for the declining gap.)
As an indicative exercise, we take a three-year rolling average of these three explanatory variables and multiply the change from the first to the last observations (2015 value-1997 value) by the different coefficients for the variables in the ratings gap OLS regressions, and reported in column (i) of Annex Table 2. The results indicate the movement of the gap that is consistent with the change of each explanatory variable over the sample period.
FX reserves, which on average more than doubled as percent of GDP over the sample period, stand out in importance among the country variables in terms of explaining the decline in the gap over the entire sample period. The increase in foreign reserves alone in our sample of emerging market economies over the period corresponds with 0.4 notch decline in the gap. Smaller is the impact of diminishing original sin, which corresponds with a declining gap of around 0.3, slightly less than one-third of a notch. The increase in VIX over the sample period corresponds with a relatively small 0.1 notch decline in the gap between foreign and domestic currency ratings. This exercise, while rough, suggests that among our explanatory variables, it was the increase in FX reserves that most accounts for the decline in the gap between the perceived risk of foreign and local currency debt over the sample period.5. Conclusions
Differences in sovereign risk denominated in local versus foreign currency matter for investors and regulators alike. Despite frequent reference to the greater creditworthiness of domestic currency obligations, a consistent theoretical framework has been so far missing in the literature. Often a remark that sovereigns can simply "inflate" their way out of domestic obligations is all that is provided.
This paper contributes to the literature in several ways. First, we document that the difference in the level of local versus foreign currency risk has steadily decreased over time for all regions. Second, we develop five hypotheses why the currency of denomination might matter for credit risk, and why there might be a gap between domestic and foreign currency creditworthiness. Four hypotheses result in implications for the relation of the gap to observable country-level variables - inflation, FX reserves, reliance on foreign currency debt markets, and banking sector ownership of government debt. Another hypothesis relates the gap to global risk and volatility, as proxied by VIX.
Based on our analysis of a panel of 73 EME rated sovereigns between 1995 and 2015, we find no support for the often cited hypothesis that high inflation might increase the relative creditworthiness of local currency obligations. Consistent with the default evidence found in other studies, sovereigns' willingness to inflate away their local debt appears to be limited, and higher inflation can even diminish the relative standing of local currency obligations.
At the same time, we find support for the hypotheses that higher FX reserves, less reliance on foreign currency borrowing (declining original sin), greater banking sector exposure to government debt, and lower global risk and volatility can narrow the gap. Among the explanatory variables, it is higher FX reserves that appear to have the most power to explain the trend decline of the gap among EMEs.
The analysis of this paper has focused on EMEs, but it would be of interest to extend the analysis to advanced economies. In many such jurisdictions, government debt burdens have been growing inexorably, and the value of the monetization option is more frequently entering public discourse.
Our results suggest there is little reason to think that the convergence of foreign and local currency risks is necessarily a permanent phenomenon. Particularly if FX reserves start to drain, foreign currency borrowing rebounds, and global volatility rises further, the distinction between local and foreign currency credits risks might once again grow in importance.
Acharya, V, I Drechsler and P Schnabl (2014): "A Pyrrhic Victory? Bank Bailouts and Sovereign Credit Risk", Journal of Finance, vol 69, pp 2689-2739.
Amstad, M and P Frank (2015): "Sovereign ratings of Advanced and Emerging Economies after the crisis", BIS Quarterly Review, December.
Amstad, M, E Remolona and J Shek (2016): "How do Global Investors Differentiate Between Sovereign Risks? The New Normal versus the Old", Journal of International Money and Finance, vol 66 (September).
Amstad, M, S Kong, F Packer and E Remolona (2016): "A spare tire for capital markets: Fostering corporate bond markets in Asia", BIS Papers, no 85.
Augustin, P, M Chernov and D Song (2017): "Sovereign Credit Risk and Exchange Rates: Evidence from CDS Quanto Spreads", Technical Report, McGill University.
Bank for International Settlements (2013): "Treatment of sovereign risk in the Basel capital framework", BIS Quarterly Review, December.
Bank for International Settlements (2016) 86th Annual Report.
Bar-Isaac, He and J Shapiro (2013): "Ratings quality over the business cycle", Journal of Financial Economics, vol 108, pp 62-78.
Basel Committee on Banking Supervision (BCBS) (2017): "The regulatory treatment of sovereign exposures", BCBS Discussion Paper, December.
Becker, B and T Milbourn (2011): "How Did Increased Competition Affect Credit Ratings?" Journal of Financial Economics, vol 101 (3), pp 493-514.
Bordo, M, B Eichengreen, D Klingebiel, M Martinez-Peria and A Rose (2001): "Is the Crisis Problem Growing More Severe?", Economic Policy, April, pp 52-82.
Borio, C and F Packer (2004): "Assessing new perspectives on country risk," BIS Quarterly Review, December.
Cantor, R and F Packer (1996): "Determinants and impact of sovereign credit ratings", FRBNY Economic Policy Review, October.
Cochrane, J (2011): "Presidential Address: Discount Rates", Journal of Finance, vol 66 (August), pp 1047-1108.
Cosset, J-C and A Jeanneret (2014): "Sovereign credit risk and government effectiveness", Working Paper, HEC, Montreal.
Dilly, M and T Mahlmann (2016): "Is there a 'Boom Bias' in Agency Ratings?", Review of Finance, vol 20 (May), pp 979-1011.
Du, W and J Schreger (2016): "Local Currency Sovereign Risk", Journal of Finance, vol 71 (June), pp 1027-70.
Eichengreen, B and R Hausmann (1999), "Exchange Rates and Financial Fragility", New Challenges for Monetary Policy", Federal Reserve Bank of Kansas City.
Eichengreen, B, R Hausmann and U Panizza (2007): "Currency mismatches, debt intolerance and original sin: why they are not the same and why it matters", in Capital Controls and Capital Flows in Emerging Economies: Policies, Practices and Consequences, NBER Volume, University of Chicago Press.
Eichengreen, Barry and Ashoka Mody, 2000,"What Explains Changing Spreads on Emerging Market Debt?", in Capital Flows and the Emerging Economies: Theory, Evidence, and Controversies, National Bureau of Economic Research, pp 107-34.
Eichengreen, B and A Mody (2004): "Do Collective Action Clauses Raise Borrowing Costs?", The Economic Journal, 114 (April), pp 247-64.
Engel, C and J Park (2017): "Debauchery and Original Sin: The Currency Composition of Sovereign Debt", mimeo.
Ferri, G, LG Liu and G Majnoni (2001): "The role of rating agency assessments in less developed countries: Impact of the proposed Basel guidelines", Journal of Banking and Finance, vol 25 (1), pp 115-48.
Financial Stability Board (FSB), 2014, "Thematic Review on FSB principles for reducing reliance on CRA ratings," May.
Fitch Ratings (1998): "The rating impact of the euro", Sovereign comment, June.
Flandreau, M, N Gaillard and F Packer (2011): "To err is human: credit rating agencies and the interwar foreign government debt crisis", European Review of Economic History, vol 15 (3), pp 495-538.
Forbes, K and F Warnock (2012): "Capital flow waves: Surges, stops, flight, and retrenchment", Journal of International Economics, vol 88 (2), pp 235-51.
Gennaioli, N, A Martin and S Rossi (2014): "Sovereign default, domestic banks, and financial institutions", Journal of Finance vol 69, pp 819-66.
Goldstein, M and P Turner (2004): "Controlling currency mismatches in emerging markets", Institute for International Economics, Washington.
Jeanneret, A and S Souissi (2016): "Sovereign defaults by currency denomination", Journal of International Money and Finance, vol 60, pp 197-222.
Kisselev, K and F Packer (2006): "Mind the gap in Asia," BIS Papers, no 30, pp 174-99.
Kumhof, M and E Tanner (2008) "Government debt: A key role in financial intermediation", in Carmen M. Reinhart, Carlos Vegh, and Andres Velasco, eds: Money, Crises and Transition, Essays in Honor of Guillermo A. Calvo.MIT press, Cambridge, MA.
Packer, F (2003): "Mind the gap: domestic versus foreign currency sovereign ratings", BIS Quarterly Review, September, pp 55-63.
Pan, J and K Singleton (2008): "Default and Recovery Implicit in the Term Structure of Sovereign CDS Spreads", Journal of Finance, vol 63, pp 2345-84.
Ratha, D, K Prabal and S Mohapatra (2010): "Shadow sovereign ratings for unrated developing countries", World Development, vol 39 (3), pp 295-307.
Reinhart, C and K Rogoff (2009): "This Time is Different: Eight Centuries of Financial Folly", Princeton University Press, Princeton.Reinhart, C and K Rogoff (2011): "The forgotten history of domestic debt", Economic Journal, vol 121.