Journal of International Money and Finance: Atish R. Ghosh, Jonathan D. Ostry, Mahvash S. Qureshi

Fiscal space and sovereign risk pricing in
a currency union
Atish R. Ghosh
*
, Jonathan D. Ostry, Mahvash S. Qureshi
Research Department, International Monetary Fund, HQ1-09-612, 700 19th Street, NW,
Washington DC 20431, USA
JEL classications:
E62
H62
H63
Keywords:
Fiscal space
Sovereign risk pricing
Eurozone
a b s t r a c t
We examine how membership in a currency union affects public
debt sustainability and market assessments of default risk in
eurozone countries. We argue that there exist offsetting effects:
expectations of bailouts tend to make a given level of debt more
sustainable, lowering bond yields and CDS rates, but constraints on
the use of monetary policy would tend to have the opposite effect,
pushing rates up especially as room for scal maneuver gets
exhausted. We develop a formal concept of scal space (which
takes account of the notion of scal fatigue under which there are
limits to the governments ability to raise the primary surplus in
response to higher debt), and apply it to the eurozone countries,
investigating in particular how currency union membership affects
CDS and bond rates during both quiet and turbulent times for
a given amount of scal space. We nd that in quiet times, CDS and
bond rates for eurozone members were below what would be
expected given their scal space (a bonus from currency union
membership). But when the crisis erupted, CDS and bond yields
rose more sharply for eurozone members than would be predicted
based on their available scal space. Our interpretation is that
sovereign bailouts did not occur with the hoped-for alacrity in
euro-crisis countries, generating sharper penalties for sovereigns
that belong to a currency union.
2012 International Monetary Fund. Published by Elsevier Ltd. All
rights reserved.
* Corresponding author. Tel.: 1 202 623 6288; fax: 1 202 589 6288.
E-mail addresses: aghosh@imf.org (A.R. Ghosh), jostry@imf.org (J.D. Ostry), mqureshi@imf.org (M.S. Qureshi).
Contents lists available at SciVerse ScienceDirect
Journal of International Money
and Finance
j ournal homepage: www. el sevi er. com/ l ocat e/ j i mf
0261-5606/$ see front matter 2012 International Monetary Fund. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jimonn.2012.11.008
Journal of International Money and Finance 34 (2013) 131163
1. Introduction
Events in Europe over the past couple of years have rekindled interest in the public nances of
monetary unions. Although eurozone members with debt sustainability problems are not the only
advanced economies to be facing such concerns (among others are Japan, the United Kingdom, and the
United States), their membership in the monetary union may bring particular challenges (as well as
advantages) in dealing with their unsustainable public nances. In this paper, we examine how
membership in a currency union (CU) affects debt sustainability, and what this implies for market
pricing of default risk of eurozone members, given their available scal space.
There are three key implications of CU membership for the conduct of scal policy and the sustain-
ability of the public nances. First, andmost simply, because CUs typically reect more thanjust economic
considerations, and often represent important historical and political ties as well, members of the union
may expect to receive assistance (nancing or transfers) fromthe rest of the union in times of economic
or nancial distress. In other words, when the budget constraint binds, there is some likelihood that it
will be shifted outwards. Second, because CU members do not have their own independent monetary
policy, macroeconomic stabilization will need to rely more heavily on scal policy. The corollary is not
just that CU members should normally try to keep more scal space than countries with independent
monetary policy, it is also that their risk premiumwill rise more steeply as their scal space is exhausted
because they lack policy instruments to help offset the impact of scal consolidation on economic
activity (and hence on revenues). Third, the scal theory of the price level means that CUmembers must
have money dominant policy regimes since theycannot rely onhigher inationor engineer a change in
the price level in order to erode the real value of their debt. As such, CUmembers must be willing to raise
taxes or reduce spending to ensure that the intertemporal budget constraint is satised.
These three implications go in opposite directions for market assessments of debt sustainability.
The possibility that a CU member will be bailed out by other members of the union pushes out its
budget constraint and makes a given level of debt more likely to be sustainable.
1
Conversely, the
greater need to rely on scal policy for macroeconomic stabilization implies higher government bond
yields or credit default swap (CDS) rates especially when the governments room for maneuver (its
scal space) is low. Likewise, the limited scope for raising the ination tax or eroding the real value of
government debt means that CU members do not have the same safety valve as non-CU members,
and should therefore face correspondingly higher spreads on their public debt.
2
To what extent are these implications reected in markets assessment of eurozone sovereign risk?
To answer this, we compare the behavior of government bond yields and CDS rates for eurozone
countries with those of other advanced economies, controlling for scal space.
3
Intuitively, market
assessment of risk should depend on the governments ability to service its debt that is, its available
scal space and any systematic divergence fromthe scal-space implied rates for eurozone countries
should reect peculiarities associated with membership in the CU.
While several recent studies analyze the market pricing of sovereign risk for the eurozone countries
in the context of the global nancial crisis examining implicitly or explicitly the role of macroeco-
nomic fundamentals, notably, scal space (e.g., Attinasi et al., 2009; Fontana and Scheicher, 2010;
1
More generally, the provision of liquidity by the common central bank could also contribute to lower spreads. In principle, if
the prospect of bailouts reduces spreads for countries expected to receive such bailouts, then it should raise them for countries
expected to provide the bailouts. But one implication of the model discussed below is that risk spreads only rise as debt
approaches the countrys debt limit. Therefore, unless the prospective bailouts threatened the debt sustainability of the
stronger members of the union, it is unlikely that there would be any appreciable increase in their spreads.
2
While the risk of being exposed to higher ination in non-CU member countries could lead investors to require higher
nominal yields than otherwise, the ability to inate away debt lowers the likelihood of a discrete default (which is likely to be
more disruptive and value-destroying than modestly higher ination), thereby implying lower risk premia than otherwise. (For
CDS rates, there is no reason for inationary expectations to lead to higher spreads for non-CU members because ination is not
a credit event.) Moreover, especially during crisis, the absence of a monetary policy tool to stimulate the economy to
counteract the contractionary effects of scal consolidation could weaken growth, making adjustment more difcult, and
increasing the likelihood of default, which would also imply higher spreads.
3
The earlier literature on this topic focused on emerging market economies, which were more susceptible to debt problems;
for references to that literature, see Attinasi et al. (2009).
A.R. Ghosh et al. / Journal of International Money and Finance 34 (2013) 131163 132
Aizenman et al., 2011; Bernoth and Erdogan, 2012; Maltritz, 2012) our analysis differs from them in
two key respects. First, we explicitly take into account the implications of CUmembership for eurozone
countries, and how they have varied before and during the crisis, arguing that any deviations of
sovereign risk pricing from macroeconomic fundamentals may partly reect rational behavior by
market participants (e.g., anticipation of bailouts before the crisis, and realization of lack of safety
valves and existence of policy constraints during the crisis). Second, most commonly in the existing
literature, public debt and/or the scal decit are used as proxies for scal space. While these variables
are certainly reasonable to use, and are likely to be correlated with scal space, they are inherently
backward looking. Yet solvency (and hence the likelihood of default) depends on the governments
ability or willingness to raise revenues or cut expenditures over the (possibly innite) future.
4
For our concept of scal space, therefore, we go beyond the usual debt or decit proxies and follow
Ghosh et al. (2011) to model how primary adjustments are undertaken. Consistent with empirical
evidence (Bohn, 1998, 2008; Mendoza and Ostry, 2008), we posit that governments typically act
responsibly, raising the primary balance in response to rising public debt. But there is a limit to this
process: as the requisite primary balance rises, it becomes increasingly difcult to keep cutting primary
expenditures or raising taxes. As long as (eventually) the increase in the primary balance does not keep
pace with the rising interest burden as debt increases, there will be a point the debt limit at which,
barring an extraordinary scal effort, debt dynamics become explosive and the government becomes
unable to fully meet its obligations. The distance between current (or projected) debt levels and this
debt limit constitutes the countrys scal space.
The advantage of our scal space measure over other proxies such as the debt ratio is evident from
the model itself: two governments with the same public debt might face very different spreads
depending on the distance to their debt limit. Also, the model predicts that the response of spreads to
scal space will be highly non-linear: essentially zero when there is ample scal space but rising to
innity (complete shut out from the markets) as scal space is exhausted. And, relatedly, markets will
tend to react very late and governments will be able to borrow at close to the risk-free rate until they
have almost run out of scal space.
Applying our concept of scal space to a sample of 23 advanced economies over 19702007, we
estimate scal space in the eurozone both individually and for the union in aggregate. We nd a wide
range of available scal space across advanced economies in general, but also among the eurozone
countries. Our estimates suggest that Greece, Italy, Japan and Portugal have the least scal space, while
Belgium, France, Ireland, Spain, and the United States are also constrained in their degree of scal
maneuvre. Australia, Korea, New Zealand, Norway, and Sweden generally have the most scal space to
deal with unexpected shocks.
Equippedwiththis concept of scal space, we turnnext to market assessments of debt sustainability in
eurozone countries, and ndthree main results. First, consistent with the model, we nd that our concept
of scal space is a better predictor of government bond yields or of CDS rates than just the level of public
debt, and that the relationship is nonlinear (government bond yields and CDS spreads rise more sharply
as scal space is exhausted). Second, from the time of the introduction of the euro till the 2008 global
nancial crisis, rates for eurozone members were lower than they should have been, given their scal
space. This suggests that the rst effect mentioned above dominated, and markets expected bailouts in
the event that a eurozone member ran into debt sustainability problems.
5
Third, when the crisis erupted,
the opposite happened: rates for eurozone members became higher than they otherwise should have
been on the basis of their available scal space. This implies that as the expected bailouts of countries
running into debt difculties did not occur (or at least did not happen with the hoped-for alacrity),
markets recognized that the constraints of CU membership made it more difcult to maintain debt
sustainability and this is what is (at least partly) reected in the wider rates for eurozone members.
4
Recognizing this, some of the studies (e.g., Aizenman et al., 2011) use debt divided by the average tax revenue over the past
ve years (as an indicator of the governments ability to collect taxes in the future), while others (e.g., Heppke-Falk and Hfner,
2004; and Attinasi et al., 2009) use projected scal decits.
5
Of course, another possibility is that markets were irrationally exuberant at the introduction of the euro, underpricing
default risk for members of the union.
The rest of this paper is organizedas follows. Section2 provides a brief snapshot of howthe eurozone
got into its current scal situation. Section 3 discusses debt sustainability dynamics in currency unions,
and their implications for primary balance adjustment and scal space. Section 4 presents our scal
space estimates for the eurozone countries, and for several other advanced economies for comparison.
Section 5 analyzes the markets pricing of risk in the eurozone countries both before and after the global
nancial crisis, controlling for their available scal space. Section 6 concludes.
2. How did it all happen?
One of the major accomplishments of the European single currency, much heralded at the time, was
the convergence of interest rates across the monetary union. During the 1980s and much of the 1990s,
government bond yields had diverged widely across Europe, with sovereign issues by traditionally-
high ination countries such as Greece, Italy, Portugal, Spain priced at 400 to 1000 basis points
(bps) above corresponding German government bunds. With the advent of the single currency in 1999,
however, these spreads narrowed dramatically, never exceeding 50 bps until the onset of the global
nancial crisis in 2008 (Fig. 1). This compression of spreads reected the removal of any currency risk,
since countries in the eurozone would no longer be able to erode the real value of their debt by
devaluing their currency. Implicitly, however, markets were also assuming that outright default was
unthinkable. Indeed, the ip side to the removal of any currency risk was the loss of the safety valve of
partial default through ination and currency depreciation that had hitherto been available.
Henceforth, a government unable to fully honor its obligations would either need a bailout or would
have to outright default on the face value of its debt.
Eurozone convergence of key economic indicators ination, interest rates, output growth, decits
and public debt ratios continued through much of the 2000s (albeit at a slowing pace) until, on the
eve of global nancial crisis (end-2007), public debt in proportion of GDP ranged from 25 percent in
Ireland, 35 percent in Finland and Spain, 45 percent in the Netherlands, 65 percent in France and
Germany, 70 percent in Portugal, to just over 100 percent in Italy and Greece (Table 1).
In the aftermath of the global nancial crisis, public debt rose across the board, reecting various
combinations of nancial sector restructuring costs, discretionary stimulus spending, automatic
stabilizers, and the collapse of asset-price related revenues (Table 2). Germany saw its public debt
increase by some 18 percent of GDP, roughly the same as France. But the most dramatic increase was in
0
2
4
6
8
10
12
14
16
18
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
Italy
Greece
Portugal
Ireland
Belgium
Spain
France
Netherlands
Finland
Austria
Introduction of Euro
in non-physical form
Beginning of
financial crisis
IMF
approves
Greece
program
European Council
announces Euro
arrives in 2002
Introduction of
Euro notes and
coins
IMF
approves
Ireland
program
IMF approves
Portugal
program
Note: Long-run government bond yield spreads over German bonds.
(In percentage points)
Fig. 1. Long-term Govt. Bond yields in eurozone countries, 19952011. Source: IFS database.
Ireland, whose public debt ratio rose more than four-fold to about 105 percent, mostly corresponding
to the costs of bailing out its over-sized banking sector and the loss of revenue when the property
bubble punctured. The second-largest increase was experienced by Greece, where the ratio of public
debt to GDP rose to 163 percent (the highest in the eurozone), on the back of higher expenditure, lower
revenues, and the discovery of scal skeletons that implied discrete revisions to the public debt
Table 1
Gross public debt in eurozone countries, 20002011 (In percent of GDP).
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
Austria 66.2 66.8 66.2 65.3 64.7 64.2 62.3 60.2 63.8 69.5 71.8 72.2
Belgium 107.8 106.5 103.4 98.4 94.0 92.0 88.0 84.1 89.3 95.9 96.2 98.5
Finland 43.8 42.5 41.5 44.5 44.4 41.7 39.6 35.2 33.9 43.5 48.4 48.6
France 57.3 56.9 59.0 63.2 65.1 66.7 63.9 64.2 68.3 79.0 82.4 86.3
Germany 60.2 59.1 60.7 64.4 66.2 68.5 67.9 65.2 66.7 74.4 83.2 81.5
Greece 103.4 103.7 101.5 97.3 98.8 100.3 106.1 105.4 110.7 127.1 142.8 163.3
Ireland 37.5 35.2 31.9 30.7 29.1 27.1 24.7 24.8 44.2 65.2 92.5 105.0
Italy 108.5 108.2 105.1 103.9 103.4 105.4 106.1 103.1 105.8 116.1 118.7 120.1
Netherlands 53.8 50.7 50.5 52.0 52.4 51.8 47.4 45.3 58.5 60.8 62.9 66.2
Portugal 48.4 51.1 53.7 55.7 57.5 62.5 63.7 68.3 71.6 83.1 93.4 106.8
Spain 59.3 55.6 52.6 48.8 46.3 43.2 39.7 36.3 40.2 53.9 61.2 68.5
Source: IMFs WEO database.
Table 2
Public debt and decit in selected eurozone countries, 20072011 (In percent of GDP).
2007 2008 2009 2010 2011
Greece Gross debt 105.4 110.7 127.1 142.8 163.3
Revenue 40.0 40.0 37.5 39.0 40.5
Expenditure 41.9 44.2 48.1 45.5 45.7
of which: Interest expenditure 4.7 5.0 5.1 5.7 6.8
Residual 1.1 5.7 9.1 15.4
Ireland Gross debt 24.8 44.2 65.2 92.5 105.0
Revenue 36.3 35.0 33.7 34.3 34.3
Expenditure 31.6 36.9 43.9 61.9 41.5
Residual 17.4 10.8 0.2 5.3
Italy Gross debt 103.1 105.8 116.1 118.7 120.1
Revenue 46.1 45.9 46.5 46.0 46.0
Expenditure 47.1 48.2 51.3 50.4 49.9
Residual 0.5 5.4 1.8 2.4
Portugal Gross debt 68.3 71.6 83.1 93.4 106.8
Revenue 41.1 41.1 39.7 41.6 44.7
Expenditure 44.0 45.5 48.9 49.8 48.2
Residual 1.1 2.3 2.2 9.9
Spain Gross debt 36.3 40.2 53.9 61.2 68.5
Revenue 41.1 37.1 34.9 36.1 35.1
Expenditure 36.8 39.0 43.3 43.4 42.6
Residual 2.1 5.3 0.0 0.2
France Gross debt 64.2 68.3 79.0 82.4 86.3
Revenue 49.8 50.0 49.2 49.6 51.0
Expenditure 51.7 52.5 55.9 55.9 55.6
Residual 1.5 4.0 2.9 0.7
Germany Gross debt 65.2 66.7 74.4 83.2 81.5
Revenue 43.7 44.0 44.9 43.6 44.6
Expenditure 43.7 44.2 48.2 47.9 45.6
Residual 1.3 4.4 4.5 2.8
Source: IMFs WEO database.
series. Both Portugal and Spain sawtheir debt ratios rise by some 3040 percent of GDP, but given their
very different starting points, Portugals debt, at about 107 percent of GDP, now stands considerably
higher than that of Spain (68.5 percent of GDP). Moreover, while Spains decits reect a roughly equal
split of lost revenue and greater expenditure (each about 6 percent of GDP), for Portugal it was mainly
higher expenditure (4 percent of GDP).
The overall result has been an increase in the public debt ratios of the eurozone countries from an
average of 63 percent at end-2007 to 92 percent as of end-2011, with the minimum ratio in the region
rising from 25 to 50 percent of GDP, and the maximum rising from 105 to 163 percent of GDP. The
difference between the highest and the lowest debt ratios in the union has thus widened from 80 to
about 113 percent of GDP.
Even more dramatic has been the widening dispersion of government bond yields across members
of the union. With German bund yields themselves falling because of accommodative monetary policy
and safe haven effects, spreads over bunds have reached 1300 bps for Greece, 700 bps for Ireland,
283 bps for Spain, and 763 bps for Portugal (Fig. 1). Curiously, however, the rise in these countries
yields occurred with a delay through 2009 and even early 2010, spreads barely reacted (even though
actual and prospective public debt ratios were clearly higher by end-2009). Moreover, even though the
countries that need to pay higher yields are generally those with higher debt ratios, the correlation is
far fromperfect: for instance, Spains public debt is only 68 percent of GDP almost exactly the same as
Germanys debt ratio prior to the onset of the global nancial crisis. As discussed below, these
observations are no coincidence, but rather follow from the rational behavior of markets in a world
where governments may face limited scal space.
3. Public nances in a currency union
Regardless of its exchange rate regime or monetary arrangement, the government will (at least
eventually) be forced to satisfy its intertemporal budget constraint (IBC).
6
This constraint requires that
the present value of primary (i.e., non-interest) expenditure be no greater than the present value of
revenue (including seignorage) in real terms, net of initial obligations (the real value of base money and
the outstanding stock of government debt):
X
N
j 0
s
tj
s
tj
T
tj
g
tj
1 r
j

M
t1
D
t1
P
t
(1)
where g
t
is primary expenditure, s
t
is tax and non-tax revenues, s
t
is seignorage, T
t
is the real value of any
unrequited transfers received by the government, M
t
is the stock of high-powered money, and D
t
is
nominal stock of outstanding government debt. In addition to this solvency constraint, the government
might face rollover or liquidity risk i.e., the possibility that, even though the government could satisfy
its IBC, the market is unwilling to lend (perhaps because it fears that the government will choose not to
satisfy its IBC due to political or other costs of adjustment). Liquidity risk could then force
the government to default because it can neither meet all of its maturing obligations immediately nor
roll themover. Inwhat follows, however, we will largely abstract fromliquidity risk, andfocus onthe IBC.
3.1. Implications of currency union membership
What are the implications of membership in a currency union? These are threefold: rst, given the
enormous economic cost and nancial disruption that a disorderly break up of a monetary unionwould
imply, there is signicant likelihood that members encountering debt servicing difculties would
receive some form of assistance fromother members of the union. Indeed, in the case of Europe, break
6
Bohn (2008) argues that debt will be sustainable provided that the primary balance always reacts positively to the lagged
level of debt (regardless of the size of the response as long as it is strictly positive), which ensures that debt grows slower than
the rate at which it is discounted. This concept of sustainability, however, does not rule out an innite debt-to-GDP ratio, which
we consider to be implausible.
up of the eurozone would also entail signicant political costs, with the post-war dream of a united
Europe set back by decades, possibly forever. All the more reason, therefore, why markets would expect
some form of bailout either outright transfers or ofcial nancing in the event that a member runs
into problems.
Transfers of course make insolvency less likely and, in practice, given that illiquidity problems can
readily turn to solvency problems, so does the availability of nancing especially at below-market
interest rates. Current eurozone crisis countries have received both forms of support; European
Union (EU)-level transfers (EU federal spending net of the national contribution to the budget) and
exceptional nancing.
7
Moreover, regardless of whether such transfers are in fact received, market
perceptions about the possibility of such transfers may delay the point at which the market cuts off
funding the government when there are doubts about its ability to otherwise satisfy the IBC.
Second, unless business cycles happen to be synchronized across members of the monetary union,
national scal policy will have to shoulder a greater burden for macroeconomic stabilization.
8
In the
run up to euro adoption, there was a large body of literature comparing the eurozone countries to the
United States with the main conclusion that the correlation of shocks and de facto mobility of labor are
lower in Europe (Bertola and Ichino, 1995; Clark and van Wincoop, 2001). While the average corre-
lation of GDP growth in eurozone countries did increase (from about 0.50 over 19901999 to 0.65 over
20002008), it is still considerably lower than that in the United States, including because the federal
budget is much smaller in Europe. The upshot is that members of a CU will want to keep more scal
space in order to be able to use scal policy for macroeconomic stabilization as monetary policy will
necessarily be geared to the average cycle of the union. But the corollary is that, as scal space is
exhausted, countries in the CU will see steeper increases in spreads because they do not have scope for
expansionary monetary policy to offset the effects of the scal contraction on output, and thus on
revenues.
Third, while in general the IBC can be satised by adjusting taxes and expenditures or by increasing
seignorage and eroding the real value of outstanding obligations, members of a monetary union do not
have the latter option.
9
In a money dominant regime, the government restores solvency by reducing
primary expenditures or raising revenues. In a scal dominant regime, the government restores
solvency by increasing seignorage or by engineering a discrete change in the price level through a burst
of ination that erodes the real value of its nominal liabilities (base money and government debt). The
union as a whole could of course decide to tolerate higher ination, but barring that, individual
countries must conduct policy within a money dominant rather than a scal dominant regime.
Accordingly, CU members must be willing to undertake the primary balance adjustments necessary to
ensure the sustainability of public debt if membership in the union is to be viable over the longer run.
3.2. Primary balance adjustment and scal space
SinceCUmembers must relyonprimarybalanceadjustment toensuredebt sustainability, it is natural
to gauge the sovereigns solvency according to the behavior of its primary balance. Here we assume that
governments typically act responsibly, raising the primary balance in response to rising debt in order to
meet the mounting interest payments, and ensure long-term solvency. But this process of raising the
primary balance to meet the higher interest payments cannot continue indenitely (because eventually
the balance would need to exceed GDP), and in practice it becomes increasingly difcult to keep cutting
7
In addition, budget nancing has also been provided indirectly through ECB Emergency Liquidity Assistance (ELA), Long-
term Renancing Operation (LTRO), and Securities Market Program (SMP). While these programs are intended to provide
liquidity to the nancial system, banks in turn may increase their purchase of government bonds, and thus channel the liquidity
provision into budgetary nance.
8
In practice, governments typically respond asymmetrically to cycles, with expansions during downturns not being fully
offset by commensurate contractions during booms, leading to a positive debt bias and correspondingly less scal space.
9
The constraint on the ination tax could lead to lower nominal and possibly real interest rates as markets do not have to
worry about ination eroding the real value of their claims (CDS rates should be unaffected by this as ination is not a credit
event). But to the extent that this constraint implies a higher probability of outright default, it could imply higher real interest
rates and higher CDS rates in CU member countries.
primary expenditures or raising taxes. In other words, beyond a certain point, scal fatigue sets in
whereby the primary balance maystill be anincreasing functionof laggeddebt but at a decreasingrate.
As noted in Ostry et al. (2010) and Ghosh et al. (2011), a reasonable characterization of the primary
balance function, pb
t
f(d
t1
), therefore, may be a cubic function (which indeed turns out to be the case
empirically): at lowlevels of debt, there may be littleor evena negative relationshipbetweentheprimary
balance and debt; as debt increases, the relationship turns positive, but at sufciently high levels of debt,
thecurve attens out. Once theincrease inthe primary balance (inresponsetohigher debt) does not keep
pace with the mounting interest payment (given by the interest rate-growth rate differential multiplied
by the stock of outstanding debt), there will be a point the debt limit, d at whichthe primarysurplus is
insufcient to meet the interest payment; thus, debt increases further, and in the absence of an
extraordinary scal effort, thedebt dynamics becomeexplosive andthegovernment necessarily defaults.
If the primary balance response to debt is purely deterministic, then the government will be able to
borrow from the market at the risk-free rate up to d, but will be completely shut out of the market
thereafter. If theprimarybalanceis subject toshocks, thenthemarket interest ratewill begintorisebefore
d, again becoming innite at the debt limit.
The deterministic casecanbereadilyillustrated. Fig. 2shows thetwocomponents of the standarddebt
dynamics equation, d
t
d
t1
r gd
t1
pb
t
: the primary balance function (with its posited cubic
shape), andtheinterest payment schedule, r gd
t1
. As shown, therearetwointersections of thesetwo
schedules. The lower intersection, labeledd
, is thelong-runequilibriumdebt ratiotowhichtheeconomy
will conditionally converge. This equilibrium is dynamically stable: below d
, the interest payment

exceeds the primary balance and debt increases; above d
, the primary balance exceeds the interest

payment, anddebt decreases. Thereis asecondintersection, d, however, whichis dynamicallyunstable. In
particular, once debt exceeds d, the primary balance is never sufcient to meet the interest payment, and
debt increases further along an explosive trajectory. It is natural to think of d as the debt limit, and the
distancebetweenthe current level of debt andd as the available scal space. The sovereignborrows at the
risk-free rate up to d but faces an innite interest rate (i.e., is shut out of the market) thereafter.
The models comparative statics are intuitive: an increase in the interest rate or a decline in the
output growth rate rotates the r gd
t1
counter clockwise in Fig. 2, raising the long-run debt level,
d
, while lowering the debt limit, d, and the available scal space. Structural reforms (or other
extraordinary efforts) that improve the primary balance shift the f(d
t1
) schedule upwards, lower the
long-run debt level, d
, while raising the debt limit and corresponding scal space.

A stochastic version of the model where the primary balance is subject to explicit stochastic shocks
pb
t
f d
t1

t
is similar, though complicated by the endogeneity of the default risk premium and
Fig. 2. Determination of debt limit in the deterministic case. Source: Authors illustration based on Ghosh et al. (2011).
the interest rate charged by the market (Ghosh et al., 2011). Specically, as the debt limit is approached
(beyond
b
d in Fig. 3), the market starts charging a risk premium to reect the higher probability of
default. But the higher interest rate raises the likelihood that the primary balance will be insufcient to
meet interest payments, and thus the probability of default, which in turn must be reected in the risk
premium. The debt limit is reached when this xed point problem between the rising risk premium
and the probability of default has no solution (at a nite interest rate).
The stochastic model provides some useful additional insights. First, the model emphasizes that the
interest rate charged by the market should not depend on the debt level per se, but rather on the
available scal space (i.e., the difference between the debt level and the debt limit). While scal space
and the debt level will typically be negatively correlated (countries with lower debt typically have
more scal space), this need not be the case: if one country has a better track record of scal adjust-
ment, it may have more scal space and be charged a correspondingly lower risk premium by the
market than another country with a poorer track record but lower debt.
Second, less obviously, the market interest rate will react very late i.e., only when the country is
close to its debt limit. In fact, it can be shown that if shocks to the primary balance function are
uncorrelated over time, and if all debt was one-period debt, then the point at which the market would
start charging a risk premium (
b
d in Fig. 3) is within one support of the shock to the primary balance.
10
This is because the market will lend at the risk free rate as long as it is assured that the debt can be
rolled over (even if the rollover itself is not at the risk-free rate). With correlated shocks and multi-
period debt, the implication will not be as stark (i.e., the interest rate will start increasing before
debt is within one support of the shock belowthe debt limit), the model does suggest that interest rates
will tend to increase surprisingly late, and then rise dramatically as the debt limit is approached an
observation that is very much consistent with the experience of eurozone countries that are now in
debt distress (and a salutary tale for governments that take comfort from successful bond auctions).
Third, again resonating with the experience of some eurozone countries, data revisions that suggest
that the support of the shock to the primary balance is larger than previously believed will lead to an
immediate lowering of the debt limit, potentially turning a sustainable level of debt unsustainable
even though the larger shock has not necessarily been realized. A mean preserving (symmetric)
increase in uncertainty lowers the debt limit in this model because the nature of debt means that
creditors lose when there are negative realizations of the shock to the primary balance that lead to
default but do not gain when there are positive realizations (as the return to lending is xed). Hence,
even mean-preserving increases in uncertainty are usually bad news that lower the debt limit.
4. Estimating scal space
An important advantage of the concept of scal space outlined above is that it can be readily
estimated. In what follows, we do so for eurozone countries (and other advanced economies for
comparison), proceeding in three steps. We begin by estimating the scal reaction function for
a sample of 23 advanced economies allowing for a cubic function to capture the possible nonlinear
response of primary balance to lagged gross debt, and controlling for various economic, structural and
institutional variables, as well as country-specic xed effects.
11
10
To put this in perspective, if the debt limit is 150 percent of GDP, and if the maximum shock to the primary balance is 5
percent of GDP, then the sovereign would be able to borrow at the risk-free rate until debt reached 145 percent of GDP, but the
interest rate would rise sharply over the next 5 percent of GDP of debt until the country is completely shut out of the market by
the time debt reaches 150 percent of GDP.
11
We use gross rather than net debt in the estimations for two reasons. First, net debt data is not always comparable across
countries due to different denitions and treatment of asset components, particularly, those of the social security system.
Second, while it makes sense to exclude intragovernment debt holdings, at least some of this debt tends to be part of a social
security fund that covers pension liabilities. If this debt is to be netted out (thus yielding the consolidated balance sheet of the
general government), then the unfunded pension liabilities need to be added. However, this requires complex actuarial
computations and assumptions (including whether to compute such liabilities on a closed- or open-group basis). Among the
high- (net or gross) debt countries, Japan stands out in the difference between gross and net debt, which is substantial (around
100 percent of GDP), but the main conclusions about Japans (lack of) scal space do not change depending on whether net or
gross debt is used.
Estimates of the scal reaction function for two sample periods (19702007 and 19852007) are
reportedinthe Appendix (Table A1), where data availability for the shorter periodallows for inclusionof
a richer set of structural variables as determinants of the primary balance.
12
In both sample periods, the
coefcients of the cubic functional form, capturing the increasing but slowing response of the primary
balance to debt, are statistically signicant. The estimated coefcients of other determinants are also
plausible and broadly in line with earlier studies (e.g., Roubini and Sachs, 1988; Abiad and Ostry, 2005;
Bohn, 2008; and Mendoza and Ostry, 2008): for example, primary balances respond positively to the
output gap (output above potential implies a large balance, and vice versa); temporary increases in
government outlays (for example, military spending), as captured by the government expenditure gap
variable, affect the primary balance negatively; and more open economies exhibit better scal perfor-
mance, as do countries with stronger institutions (as proxied by the political stability index).
Next, we use the estimated coefcients for the 19852007 sample (Table A1, col. [4]), and two
different interest-rate growth differentials: (i) the historical average (over 20002007) of the real long-
term government bond yield relative to the projected real GDP growth rate (ve-year average); and (ii)
the IMFs current projections of the real long-term government bond yield relative to the real GDP
growth rate (for 2016), to compute the debt limit for all countries in the sample.
13
Table 3 shows that
there is signicant variation across countries in the resulting estimates (reecting the country-specic
intercepts of the primary balance reaction function given by the xed effects and the values of the
other non-debt independent variables in the regression and the interest rate-growth rate differen-
tial).
14
Specically, the long-run debt ratio to which countries will conditionally converge ranges from
a zero or positive asset position(indicated by d
0) to debt of about 105 and 120 percent of GDP under

the historical and projected interest rates, respectively, with an average of about 35 percent of GDP. The
estimated debt limit, d, ranges from about 167 to 245 percent of GDP, with a mean and median of about
200 percent of GDP, under the historical interest rate-growth rate differential. Since, the interest rate-
growth rate differentials are generally projected to be less favorable in the post-crisis years than the
historical experience, the corresponding average debt limit is about 195 percent of GDP for the projected
rates.
Once the debt limit, d, has been computed, estimates of available scal space are given simply by the
difference between current (or projected) debt levels and the debt limit. In three cases (Greece, Italy
and Japan), no estimate of d
(or d) is reported under the projected interest rates. This is because, given
these countries estimated primary balance reaction function and assumed interest rate-growth rate
differential, public debt would not be expected to converge to a nite steady-state debt ratio (it follows
that there is no maximum debt level below which convergence occurs, so it is not meaningful to
calculate d either).
15
For Greece and Italy, however, d
exists using the historical interest rates.

Based on the projected interest rates and debt level for 2016, among the eurozone countries, Greece
and Italy do not have any available scal space, while Portugal, Ireland, and Spain are also highly
constrained. By contrast, Finland, Germany, and Netherlands have the most additional scal space.
Among the comparator group of advanced countries, Japan has no available scal space, while
12
See the Appendix for the description of variables and data sources.
13
For the calculation of d
and d, output and government expenditure gaps are assumed to be closed.

14
The estimates of scal space include xed effects for individual countries to capture differences in average primary
balances, but assume slope homogeneity. While the slope homogeneity assumption may be questioned, the issue is more
complex here because the cubic scal fatigue behavior is postulated to hold over the entire range of debt (from 0 to about 180
percent of GDP), whereas individual countries debt ratios may be observed over only a portion of that range. Our analysis for
countries that, in the sample, we do not observe at high debt ratios must therefore rest on the assumption that these countries
if their debt were to increase would behave similarly (in terms of scal fatigue) as countries that we do observe at high debt
ratios. As discussed in Ghosh et al. (2011), this is a reasonable conjecture for three reasons. First, our sample consists only of
advanced economies, which share similar characteristics, and is therefore relatively homogeneous. Second, the contrary
assumption that some countries would never exhibit scal fatigue, and would always increase the primary balance in line
with rising debt seems implausible, if only because the primary surplus cannot exceed GDP. Third, slope homogeneity across
countries cannot be rejected (in almost every case) over the debt ranges (low-to moderate; moderate-to-high) in which they
are observed.
15
In terms of Fig. 2, the primary balance reaction function is always below the interest rate schedule, so there is no
intersection.
Australia, Korea, Norway and Sweden have ample scal space. At about 7780 percent of GDP, the point
estimates of scal space for the US and Ireland are similar, while UKs scal space stands at about 111
percent of GDP.
16
These scal space estimates are just point estimates, which are a complicated function of the
underlying parameter estimates of the primary balance reaction function. Since the debt limit is
essentially the annuity value of the primary balance, even modest uncertainty in the estimates of the
primary balance translates into signicant differences in the calculation of the debt limit. To take
account of uncertainty about the coefcient estimates in the primary balance regression, Table 4
reports the estimates of scal space in terms of the probability (computed using Monte Carlo tech-
niques) that a country has a given amount (0, 50, or 100 percent of GDP) of remaining scal space at the
projected interest rate-growth rate differential.
Consistent withthediscussionabove, theprobabilitythat GreeceandItaly have additional scal space
at the projected interest rate-growth rate differential is very low, while for Portugal, the probability of
some positive scal space is moderate (about 56 percent).
17
For the other eurozone member countries,
however, the probability that these countries have at least some additional scal space is about 80
percent or higher. Among the other advanced countries, Japan has negligible probability of having
positive scal space, but for the remaining countries, this probability is at least about 80 percent. Intu-
itively, these estimates of scal space depend on the debt level projected for 2016 and the debt limit,
whichdepends onthecountrys historical trackrecordof primarybalances, as well as ontheinterest rate-
growth rate differential. Thus, even a country whose current debt is relatively high (e.g., Belgium) may
enjoyadditional scal space if, inthe past, it has beenscally responsible adjusting the primary balance
to rising debt once the shock had passed and if its projected interest rate-growth rate differential is
small. By the same token, countries can expand their scal space through credible adjustment plans.
In assessing these results, several points should be borne in mind. First, the reported estimates of
scal space are against projected debt levels in 2016, and do not take account of future age-related
spending and possible contingent liabilities (though if estimates of such liabilities are available, it is
Fig. 3. Determination of debt limit with stochastic shocks. Source: Authors illustration based on Ghosh et al. (2011).
16
Using net debt ratios for end-2016 (instead of gross debt ratios) to compute the scal space leads to a similar ranking of
countries in terms of the available scal space. Specically, Australia, Korea, Norway, Sweden along with New Zealand have the
lowest net debt ratios in 2016 and ample scal space, while Greece, Italy, Japan and Portugal have the highest net debt ratios,
and no or limited available scal space. These results are available upon request.
17
Theseestimates arebasedonthebehavior of primarybalances over theperiod19802007. For Italy, whichundertooksignicant
adjustment in the 1990s, estimates based on the 19902007 data suggest somewhat more scal space than reported here.
straightforward to subtract them from the reported scal space).
18
Second, the debt limit on which
the scal space estimates are predicated is by no means a desirable or optimal level of debt: for
a variety of reasons, including rollover risk, governments will want to ensure that they do not exhaust
their scal space and that debt remains well below its calculated limit. Thus, even countries that are
currently estimated to have substantial scal space may need to undertake medium-term scal
adjustment. Third, as discussed above, a key feature of the model is that a revision in market estimates
of the support of a possible shock to the primary balance even if the shock does not occur can trigger
a rise in market interest rates, potentially undermining debt sustainability. Finally, since estimates of
Table 3
Long-run debt limit and scal space in the eurozone and other advanced countries (In percent of GDP).
Country Debt to GDP
(end-2016)
a
Interest rate-growth rate
differential
d* d
Historical
b
Projected
c
Historical
b
Projected
c
Historical
b
Projected
c
Australia 17.0 0.9 0.2 0.0 0.0 211.7 204.9
Austria 70.6 0.7 0.4 0.0 0.0 191.1 193.5
Belgium 90.9 0.4 1.8 40.0 67.8 193.0 174.2
Canada 76.3 0.3 0.6 71.9 84.7 191.2 178.0
Denmark 48.6 0.5 0.5 0.0 0.0 203.6 203.6
Finland 54.2 0.5 0.1 0.0 0.0 195.3 199.1
France 87.5 0.4 0.2 81.6 74.5 181.7 189.0
Germany 72.7 1.1 0.1 48.0 0.0 185.7 197.1
Greece 143.7 1.9 3.1 75.4 . 201.5 .
Iceland 83.0 1.5 0.5 0.0 0.0 198.3 207.9
Ireland 112.2 3.5 1.0 0.0 0.0 230.3 189.3
Israel 67.4 1.3 0.6 77.0 49.3 175.2 200.4
Italy 120.7 0.6 3.9 105.2 . 166.8 .
Japan 253.6 0.1 0.4 . . . .
Korea 24.9 1.9 2.5 0.0 0.0 227.3 231.4
Netherlands 78.7 0.1 0.1 0.0 0.0 201.6 201.6
New Zealand 32.7 0.6 1.8 0.0 0.0 210.6 200.1
Norway 49.6 1.3 1.4 0.0 0.0 245.3 244.9
Portugal 110.7 0.3 1.7 77.7 120.1 188.6 145.1
Spain 90.7 1.7 2.2 0.0 79.3 213.8 166.2
Sweden 24.1 0.5 0.2 75.8 71.6 183.0 187.2
United Kingdom 90.1 0.7 1.0 78.1 57.8 180.7 201.0
United States 112.8 1.1 0.2 52.2 63.0 203.6 193.3
Mean 83.2 35.6 33.4 199.1 195.4
Median 78.7 20.0 0.0 196.8 198.1
Source: IMFs WEO database and authors estimates.
Notes: d is the debt limit, above which debt grows without bound given the countrys past primary balance behavior; d* is the
long-run average debt ratio to which the economy converges conditional on not exceeding d; (.) indicates that given the scal
reaction function and the interest rate-growth rate differential, the public debt dynamics are not on a sustainable path to
converge to a nite stable steady state debt ratio; 0 indicates that convergence achieved at a negative d*, implying a positive
government asset position. Fiscal space is the difference between and projected debt for 2016. All results are based on the
estimated scal reaction function reported in Table A1 (col. 4).
a
IMFs WEO projection for 2016.
b
Average of 2000-07 based on the difference between real long-term government bond yield and projected (5-year average)
real GDP growth rate (in percentage points).
c
Based on the (IMFs WEOprojected) real long-termgovernment bond yield and real GDP growth rate for 2016 (in percentage
points).
18
While our estimates of scal space do not account for implicit future social security liabilities, a simple (rank) correlation of our
scal space estimates with the European Commissions (EC) scal sustainability indicator namely, the required primary balance
adjustment over 20112030 taking into account age-related spending (and assuming a debt target of 60 percent of GDP by 2030, or
debt stabilization at end-2012 level if debt ratio is less than 60 percent of GDP; source: IMF (2012)) is about 0.6 (and the null
hypothesis that the two series are independent is rejected at 1 percent signicance level). Moreover, if we compare ECs
sustainability indicator withthe maximumprimary decit (net of age-related spending) that a countrycould run at the projected
interest rate-growth rate differential over the next 30 (or 50) years before hitting its debt limit, the correlation is about 0.9.
scal space are based on the assumption that future policy reactions to rising debt will be similar to
those in the past, a nding that a country has little or no scal space is not a prediction that public debt
will necessarily explode and the government will default history is not destiny but rather that
something must change and scal policy cannot proceed on a business as usual basis. Already many
of the countries reported here as having little scal space, including in the eurozone, have embarked
upon adjustment programs intended to represent a clean break from the past.
4.1. Eurozone wide scal space
Given talk about greater scal union being the solution to the eurozones current debt sustainability
problems, it is interesting to ask how much scal space the zone as a whole would have were it
a genuine scal union (with countries behaving as they have done historically).
19
A simple way to
estimate the eurozone wide scal space is to take the average of the scal space estimates for the euro
member countries reported in Table 3, weighted by some measure of economic scale (such as the share
of their nominal GDPs). A more formal way, however, is to estimate the scal reaction function
including the eurozone countries as a single entity post-1999 (taking the nominal GDP weighted
average of the relevant variables), and to compute the debt limit and available scal space using the
average interest rate and growth rate differential for the region.
Table 4
Estimated probability of scal space in eurozone and other advanced countries (In percent).
Country Debt to GDP
(end-2016)
a
Interest rate-Growth
rate differential
b
FS > 0 FS > 50 FS > 100
Australia 17.0 0.2 100.0 100.0 100.0
Austria 70.6 0.4 98.7 98.7 91.9
Belgium 90.9 1.8 97.7 97.1 11.5
Canada 76.3 0.6 90.8 90.7 54.0
Denmark 48.6 0.5 100.0 100.0 100.0
Finland 54.2 0.1 99.9 99.9 99.7
France 87.5 0.2 98.5 98.3 53.4
Germany 72.7 0.1 99.4 99.4 94.7
Greece 143.7 3.1 2.7 0.0 0.0
Iceland 83.0 0.5 91.2 91.2 79.5
Ireland 112.2 1.0 98.7 94.9 9.4
Israel 67.4 0.6 100.0 100.0 99.9
Italy 120.7 3.9 0.8 0.0 0.0
Japan 253.6 0.4 0.2 0.1 0.1
Korea 24.9 2.5 100.0 100.0 100.0
Netherlands 78.7 0.1 100.0 100.0 97.4
New Zealand 32.7 1.8 89.1 89.1 89.0
Norway 49.6 1.4 100.0 100.0 100.0
Portugal 110.7 1.7 55.9 36.6 0.8
Spain 90.7 2.2 82.7 78.8 7.2
Sweden 24.1 0.2 93.8 93.8 93.8
United Kingdom 90.1 1.0 99.9 99.9 80.4
United States 112.8 0.2 99.6 98.1 12.3
Mean 83.2 82.6 81.2 59.8
Median 78.7 98.7 98.1 80.4
Source: Authors estimates.
a
IMFs WEO projection for 2016 (in percent of GDP).
b
Interest rate-growth rate differential (in percentage points) is based on the (IMFs WEO projected) real long-term
government bond yield and real GDP growth rate for 2016.
19
Going forward, of course, debt sustainability may not become an issue for eurozone countries if they abide by the estab-
lished scal rules, but past experience allows room for only cautious optimism. Despite the assurances at the time of euro
adoption that the EMU would be based on strict budgetary discipline, and no member country would be permitted to accu-
mulate debts in excess of 60 percent of GDP or to run scal decits larger than 3 percent of GDP, several euro-member countries
(including Germany) have violated these rules in several years without facing any serious penalties.
The estimation results for this exercise are presented in the Appendix (Table A2) as before, the
relationship between primary balance and (lagged) debt to GDP ratio is nonlinear, and the (negative)
cubic termis statistically signicant in all specications. Combining the estimated coefcients reported
in Table A2 (col. [4]) with the average (projected) interest rate-growth rate differential for the region in
2016, we obtain estimates for the regions debt limit and available scal space. Fig. 4(a) shows that as
a scal union, the fragility of the eurozone would be considerably reduced (even if member countries
behaved as they did in the past). Specically, under the average historical interest rates, the region
would have a combined scal space of about 123 percent of GDP. The scal space drops at the projected
interest rates, but nevertheless, at about 100 percent of GDP, remains considerable. Simulations taking
into account the uncertainty associated with parameter estimates further show that the probability of
positive scal space or that in excess of 50 percent of GDP for the region is about 95 percent, while that
in excess of 100 of GDP is about 53 percent (Fig. 4(b)).
5. Fiscal space and market responses to the currency union
How has the market assessed debt sustainability of eurozone countries given their scal space?
Anecdotal evidence suggests that CU membership may have affected the risk premium for eurozone
countries both before and during the crisis. Fig. 5, for example, compares the debt ratios and government
bondyields for theUnitedKingdom(anoneurozonecountry) andSpain(aeurozonemember).
20
Through
the rst half of the 2000s, Spains public debt inproportion to GDP was higher than that of the UK, but its
government bond yield was persistently lower than that of the UK. This suggests some under-pricing of
risk though it might have been justied by market expectations of bailouts for members experiencing
debt problems. But if so, it also means that the market must have downplayed the constraints on stabi-
lization policy and scal dominance that membership in a monetary union would imply.
21
(a) Debt limit and fiscal space (In percent of GDP)
1
Probability of given fiscal space (In percent)
2
Notes: Estimates based on the estimated fiscal reaction function reported in Table A3 (col. 4). Debt limit is the maximum debt level above which
debt grows without bound given the country's past primary balance behavior. Fiscal space is the difference between debt limit and WEO
projected debt to GDP ratio for end-2016. Projected interest rate-growth rate differential (in percentage points) is the average of the difference
between real long-term government bond yield and real GDP growth rate for 2016 for the 11 eurozone countries in the sample. Historical interest
rate-growth rate differential (in percentage points) is the difference between real long-term government bond yield and projected (5-year average)
real GDP growth rate averaged over 2000-07 for the eurozone countries.
0
50
100
150
200
Historical interest rate-growth
rate differential (-0.35)
Proj. interest rate-growth rate
differential (1.31)
Debt limit
Fiscal space
0
20
40
60
80
100
FS>0 FS>50 FS>100
(b)
Fig. 4. Fiscal space in the eurozone as a scal union. Source: Authors estimates.
20
See De Grauwe (2011) for a further comparison of debt and bond rate dynamics in Spain and the UK. In the case of the UK,
quantitative easing by the Bank of England may also have contributed to lower bond yields.
21
Aizenman et al. (2011) estimate a formal model of sovereign risk pricing in 60 advanced and emerging countries over
20002006, and nd that in the South-west peripheral eurozone countries, sovereign risk was considerably underpriced
relative to international norms in the pre-global nancial crisis period.
The pattern, however, reversed quickly with the onset of the global nancial crisis. Although UKs
debt ratio rose faster than that of Spain, the yield on Spanish long-term government bonds reacted
more sharply, rising some 200 bps higher than the yield on British government bonds. The CDS rates
(not shown here) mimic the same pattern while the CDS rate for the UK was higher than for Spain in
2008, the trend has reversed since 2009, and the difference between Spanish and the UKCDS rates rose
to over 300 bps at the end of 2011. Financial markets thus seemto be placing a much higher default risk
on Spanish debt, although in percent of GDP it is lower than that of the UK.
To examine this more formally, we estimate regressions of the determinants of CDS rates which is
a direct measureof howmarkets perceive(andprice) sovereigncredit riskandof long-termgovernment
bondrates for our sampleof 23advancedeconomies over theperiod20002011.
22
Webeginwithasimple
specication that includes only scal space (in log percent of GDP to capture the expected non-linear
effect on default probability as scal space is exhausted), a dummy variable for eurozone membership,
eurozone, and real per capita income as a control variable (Table 5, col. [1]).
23
While higher per capita
income is associated with signicantly lower CDS rates, we nd that there is indeed a premium for CU
membership: eurozone countries experience some 50 percent lower CDS rates than other advanced
economies, and this difference is statistically signicant. Fiscal space is also a highly statistically signi-
cant determinant of CDS rates: decreasingscal spacefrom100percent of GDPto50percent of GDPraises
CDS rates byabout 23percent, while further decreasingscal space byabout 50percentage points of GDP
(to 1 percent of GDP of scal space) implies a 215 percent increase in CDS rates.
24
0
2
4
6
0
20
40
60
80
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
UK: Gross debt (in % of GDP) Spain: Gross debt (in % of GDP)
UK: Long-term govt. bond yield (in %)-right-axis Spain: Long-term govt. bond yield (in %)-right-axis
Global financial
crisis
Fig. 5. Gross debt to GDP and long-term Govt. Bond yields in UK and Spain, 20002011. Source: IFS and IMFs WEO databases.
22
For this purpose, we compute scal space annually for each country by using the estimated coefcients of the scal reaction
function reported in Table A1 (col. [4]), and the yearly real interest rate-growth rate (average of the 5-year projected real growth
rates) differential as discussed above. CDS spreads are end-of-year observations, while government bond yields are annual
averages. To address potential reverse causality concerns between CDS rates/bond yields and scal space, we also estimate all
specications using lagged scal space in the robustness analysis, but nd the results to be very similar.
23
Wedonot includecountryxedeffects inthespecications toestimatetheimpact of eurozonemembership(whichis acountry-
specic time invariant variable in our sample). Instead, as discussed below, we control for other relevant country-specic charac-
teristics suchas (logof) real GDP per capita, trade openness, scal balance toGDP, current account balance to GDP, andinationas in
other studies (e.g., IMF, 2010; Aizenman et al., 2011). Country xed effects are, however, included in the robustness analysis.
24
Computed in basis points, a decrease in scal space from 100 percent of GDP to 50 percent of GDP raises CDS rates by about
5 bps for eurozone member countries (and 9 bps for non-members); and a further decrease in scal space by about 50
percentage points of GDP (to 1 percent of GDP of scal space) implies an increase in CDS rates of 52 bps for eurozone countries
(and 105 bps for others).
The impact of scal space and eurozone membership remains statistically signicant, but is reduced,
whenwe include a dummy variable, crisis, to capture the effect of the global nancial crisis (20082011)
on CDS spreads (col. [2]). The effect of crisis itself on CDS spreads is signicantly positive implying, on
average, about a 1500percent (i.e., a 16-fold) increaseinCDSspreads inthecrisis years.
25
However, tosee
if the effect of crisis onspreads has beenlarger for eurozone members or inother words, if the premium
associated with being in the eurozone turned negative during the crisis we add to our specication an
interactionterm, eurozone crisis, whichis equal tounity for eurozone countries during the crisis period
20082011 (col. [3]). If CDS rates widened more for eurozone countries than other advanced economies
in the crisis, then the estimated coefcient on the interaction term should be positive.
The results presented in Table 5, col. [3] show that indeed eurozone countries experienced a larger
increase in CDS rates during the crisis specically, rates rose on average by about 2400 percent (i.e.,
25-fold) for the eurozone countries and by about 830 percent for other advanced economies.
26
Again,
greater scal space is signicantly associated with lower CDS rates, while membership in the eurozone
is itself associated with lower CDS rates, controlling for scal space. The result holds when we control
for other possible determinants of CDS rates such as overall scal balance to GDP, trade openness, the
Table 5
Sovereign risk pricing, scal space and Euro membership, 20002011.
Dependent variable CDS spread Real long-term government bond yield
(1) (2) (3) (4) (5) (6) (7) (8)
Fiscal space (log) 0.293*
(0.166)
0.146**
(0.061)
0.103*
(0.060)
0.132*
(0.070)
0.416*
(0.225)
0.462*
(0.250)
0.379
(0.224)
0.428**
(0.166)
Eurozone 0.710**
(0.326)
0.460*
(0.247)
1.001***
(0.335)
1.111***
(0.381)
0.395
(0.262)
0.385
(0.273)
1.003***
(0.301)
1.321***
(0.426)
Crisis 2.776***
(0.186)
2.232***
(0.326)
2.019***
(0.394)
0.748*
(0.379)
1.622***
(0.358)
2.024***
(0.383)
Eurozone Crisis 0.994**
(0.375)
0.926**
(0.411)
1.943***
(0.642)
1.877***
(0.580)
Real GDP per
capita (log)
0.801*
(0.394)
1.284***
(0.399)
1.261***
(0.356)
0.967***
(0.278)
0.705
(0.454)
0.614
(0.491)
0.584
(0.448)
0.706
(0.413)
Trade openness 2.279**
(0.839)
3.788**
(1.678)
Ination 0.130
(8.163)
0.606
(16.348)
Current account
bal./GDP
0.030***
(0.010)
0.012
(0.028)
Fiscal balance/GDP 0.026*
(0.014)
0.023
(0.052)
VIX index (log) 0.134
(0.394)
0.711*
(0.350)
Constant 13.148***
(3.510)
15.708***
(4.081)
15.624***
(3.637)
11.231***
(3.113)
11.458***
(3.847)
10.982**
(4.047)
10.594**
(3.760)
8.411*
(4.326)
Observations 151 151 151 151 276 276 276 275
R-squared 0.122 0.747 0.765 0.810 0.171 0.214 0.286 0.347
Notes: CDS spread is log transformed. Real government bond yield is in percent. Fiscal space (log) is the difference between the
debt limit (determined using the scal reaction in Table A1 (col. 4) and the real long-run bond yield and (average 5-yr. projected)
real GDP growth rate differential) and the actual debt ratio. Eurozone is a binary variable equal to 1 if the country is a eurozone
member and 0 otherwise. Crisis is a binary variable equal to 1 for 20082011. Trade openness is the ratio of sum of exports and
imports to GDP (in percent). Ination is 5-year lagged average. Clustered standard errors (by country) reported in parentheses.
*,**,*** indicate signicance at 10, 5 and 1 percent levels, respectively.
25
Evaluated at mean values (for scal space and per capita GDP), this translates into 72 bps higher spreads for eurozone
countries (and 114 bps higher spreads for others advanced economies in our sample) in the crisis.
26
Evaluated at meanvalues (for scal space andper capita GDP), this translates into 95 bps higher spreads for eurozone countries
(and89bps higher spreads for others advancedeconomies inour sample) inthecrisis. Theincreasefor eurozonecountries thus fully
offset their pre-crisis advantage resulting in almost identical predicted spreads for eurozone and non-eurozone countries (about
a 100 bps). These averages of course mask important differences across eurozone members (e.g., for Greece, the predicted spreadin
2011 is 355 bps, whereas for Germany, it is 65 bps; see Fig. A1).
current account balance to GDP, ination, and a measure of global risk aversion (the VIX index) (col.
[4]). Consistent with other studies (e.g., Barrios et al., 2009; IMF, 2010; Aizenman et al., 2011), we nd
that an improved current account balance is associated with lower CDS spreads, indicating that, for
a given scal space, a higher level of external indebtedness implies higher CDS rates.
27
A higher scal
balance included to capture rollover/liquidity risk is also associated with lower CDS spreads. By
contrast, higher trade openness is related with higher CDS spreads likely capturing the vulnerabilities
arising fromgreater reliance on external trade while the impact of ination and the VIX index on CDS
rates is statistically insignicant.
The right hand panel of Table 5 (cols. [5][8]) repeats the analysis using real long-term government
bond yields. The results support our earlier ndings using CDS rates and show that both the available
scal space and CU membership matter signicantly for bond yields a decrease of scal space from
100 percent of GDP to 50 percent of GDP is associated with about 13 percent higher real yields, while
a decrease from 50 percent of GDP to 1 percent of GDP is associated with about 65 percent higher real
yields [col. 5]. Controlling for scal space, eurozone countries on average paid 17 percent lower bond
yields than other advanced economies. The crisis itself was associated with a decline in government
bond yields as central banks ooded the nancial system with liquidity, and on average advanced
countries paid some 30 percent lower yields (col. [6]). The impact of the crisis is, however, not uniform
Table 6
Sovereign risk pricing, debt ratio and Euro membership, 20002011.
(1) (2) (3) (4) (5) (6) (7) (8)
Debt ratio 0.007
(0.006)
0.007*
(0.003)
0.006*
(0.003)
0.010***
(0.003)
0.002
(0.005)
0.003
(0.005)
0.003
(0.004)
0.001
(0.004)
Eurozone 0.637*
(0.333)
0.464
(0.271)
1.111***
(0.325)
1.264***
(0.302)
0.278
(0.229)
0.277
(0.233)
1.141***
(0.331)
1.449***
(0.441)
Crisis 2.871***
(0.191)
2.211***
(0.302)
1.922***
(0.381)
0.477
(0.462)
1.724***
(0.380)
2.385***
(0.426)
Eurozone Crisis 1.144***
(0.340)
1.088***
(0.348)
2.618***
(0.835)
2.586***
(0.742)
Real GDP per
capita (log)
0.828**
(0.331)
1.416***
(0.420)
1.398***
(0.370)
1.204***
(0.248)
0.607*
(0.329)
0.561*
(0.321)
0.542
(0.329)
0.565
(0.413)
(1.223)
2.446
(1.610)
Ination 0.185***
(6.141)
0.404
(15.257)
Current account
bal./GDP
0.029**
(0.012)
0.028
(0.035)
Fiscal balance/GDP 0.012
(0.019)
0.081
(0.068)
(0.381)
0.904**
(0.363)
Constant 11.626***
(3.338)
15.889***
(4.227)
16.165***
(3.684)
11.352***
(2.514)
8.434**
(3.413)
8.072**
(3.345)
8.312**
(3.348)
5.073
(4.368)
Observations 151 151 151 151 276 276 276 275
R-squared 0.054 0.748 0.775 0.833 0.015 0.033 0.174 0.242
Notes: CDS spread is log transformed. Real government bond yield is in percent. Debt ratio is gross general government debt to
GDP (in percent). Eurozone is a binary variable equal to 1 if the country is a eurozone member and 0 otherwise. Crisis is a binary
variable equal to 1 for 20082011. Trade openness is the ratio of sum of exports and imports to GDP (in percent). Ination is 5-
year lagged average. Clustered standard errors (by country) reported in parentheses. *,**,*** indicate signicance at 10, 5 and 1
percent levels, respectively.
27
Gros (2011) argues that external debt is a key determinant of sovereign debt risk in the eurozone countries, for even if these
countries do not have the option of issuing their own currency, they still retain full taxing powers and can always service their
domestic debt by taxing the residents. Taxing foreigners however is more complicated as governments are generally bound by
existing treaties and international norms, and do not have a free hand in taxing non-citizens.
across countries as is illustrated by the positive interaction termbetween eurozone membership and
crisis variables in col. [7] and eurozone members on average paid some 17 percent higher real yields
in the crisis, while other advanced economies paid about 57 percent lower real yields in the crisis.
These results hold when controlling for overall scal balance, trade openness, ination, the current
account balance, and the VIX index (col. [8]).
How do these results compare with using debt ratio as a proxy for scal space? Table 6, col. [1]
shows, perhaps surprisingly, that the estimated effect of public debt ratio on CDS spreads is statistically
insignicant. Membership in the eurozone remains statistically signicant, with the estimated
premium equal to some 50 percent. The estimated coefcient of debt turns statistically signicant
when other control variables are included in the regression (col. [2]col. [4]), but is always statistically
insignicant for real government bond yield (col. [5]col. [8]). Nevertheless, eurozone membership is
always associated with lower government bond yields, except for during the crisis when eurozone
countries on average experienced higher yields than other advanced economies. The weaker results
using public debt conrm the intuition above that it is scal space not just the debt ratio that
matters for default probability and therefore for CDS rates.
28
Overall, the results conform to the hypotheses laid out in Section 3. Membership in the eurozone
was associated with lower interest rates and CDS spreads before the crisis, either because of expec-
tations of bailouts in the event of a crisis or because of irrational exuberance at the introduction of the
euro. Once the crisis erupted, the additional constraints imposed by CU membership became apparent
(as did the reluctance of stronger members of the union to bailout their weaker partners), resulting in
sharper increases in interest and CDS rates for eurozone members than for other advanced economies.
This does not imply that the observed spreads in the post-crisis period are necessarily rational; even
allowing for eurozone spreads to be different since 2008 onward (i.e., once markets recognized that
bailouts would not be automatic and that CU membership imposed constraints on adjustment), there
are sizable residuals for some eurozone countries (Figs. A1 and A2). While these are presumably
justiable in the case of Greece (which defaulted), for other countries such as Ireland, Portugal, Spain,
and Italy, it is hard to rule out the possibility that they reect multiple equilibria whereby pessimistic
expectations lead to excessively wide spreads that are then justied by the greater default risk given
the high spreads.
5.1. Quarterly data
While the analysis conducted above using annual data is informative, it may mask important
variations in sovereign risk pricing during the course of the year, particularly given rapidly evolving
crisis dynamics. For example, the CDS rates for Greece increased by about 800 bps over 2010, and
by 2500 bps during the course of 2011. Similarly, CDS rates for Portugal increased by about 770
basis points and for Italy by some 325 bps over 2011. To take into account these sharp variations in
the underlying data series, especially during the crisis, we also estimate the above regressions
using quarterly data, where both CDS rates and government bond yields are averaged over the
quarter, and crisis is dened over 2008:Q32011:Q4, but scal space is measured on an annual
basis.
The results obtained from this exercise are very similar to those reported above. First, on average,
there is a premium for eurozone membership, with CDS rates about 53 percent lower for eurozone
countries than for other advanced economies (Table 7, col. [1]). Second, the crisis itself had a large
impact on sovereign risk pricing across the board, with CDS rates on average rising by 955 percent
during the crisis period (col. [2]). Third, during the crisis, CDS rates rose more sharply for eurozone
countries: on average, increasing by 1500 percent, while the corresponding increase for other
advanced economies was some 540 percent (col. [3]). That the premium associated with CU
membership declined in the crisis for eurozone countries holds even after controlling for other vari-
ables in col. [4], of which scal balance, the VIX index, and trade openness are estimated to be
28
Similar results are obtainedusing net debt insteadof gross debt ratio. Specically, we do not nd a statistically signicant effect
of net debt ratios onreal government bondyields whiletheimpact onCDSspreads is stronger (all results areavailableuponrequest).
signicantly associated with CDS rates. The signicance of the crisis variable despite controlling for the
VIX index is interesting, and suggests that factors in addition to a rise in global risk aversion perhaps
contagion fears from a global economic slowdown contributed to the across-the-board rise in CDS
spreads.
Muchthe same picture is obtainedusing real government bond yields (Table 7, cols. [5][8)]). Prior to
the crisis, government bond yields were very nearly 30 percent higher than during the crisis (col. [6]).
The ight to quality and liquidity provision during the crisis resulted in about 62 percent lower real
government bondyields for other advancedeconomies, but for eurozonecountries, real yields roseby26
percent in the crisis (col. [7]).
5.2. Sovereign risk pricing and credit ratings
In addition to macroeconomic fundamentals, several recent studies have emphasized the role
played by credit ratings in inuencing CDS rates and bond yields (e.g., Gande and Parsley, 2005;
Ismailescu and Kazemi, 2010; Afonso et al., 2012).
29
Since ratings can be issued or changed at will, and
are supposed to capture the likelihood of default which is what the CDS (and, to a somewhat lesser
extent, real bond yields) are supposed to price their inclusion in the regression should render the
other variables statistically insignicant. But, interestingly, we do not always nd this to be the case.
30
The estimated effect of the scal space variable on CDS spreads, for example, remains statistically
Table 7
Sovereign risk pricing, scal space and Euro membership, 2000Q12011Q4.
(1) (2) (3) (4) (5) (6) (7) (8)
Fiscal space (log) 0.343**
(0.128)
0.187***
(0.046)
0.145***
(0.048)
0.135*
(0.072)
0.422**
(0.178)
0.465**
(0.203)
0.379**
(0.173)
0.385**
(0.147)
Eurozone 0.753**
(0.306)
0.550**
(0.236)
0.994***
(0.329)
1.148***
(0.305)
0.239
(0.243)
0.235
(0.255)
0.761**
(0.308)
0.919**
(0.362)
Crisis 2.356***
(0.160)
1.859***
(0.274)
1.532***
(0.307)
0.634*
(0.368)
1.482***
(0.361)
1.700***
(0.356)
Eurozone Crisis 0.929**
(0.337)
0.778**
(0.331)
1.912*** 1.997***
(0.618) (0.578)
Real GDP per
capita (log)
1.496***
(0.431)
1.758***
(0.374)
1.759***
(0.336)
2.013***
(0.369)
1.087**
(0.435)
1.022**
(0.457)
0.989**
(0.414)
1.536***
(0.515)
(0.906)
0.630
(1.682)
Ination 0.120
(0.170)
0.190
(0.143)
Current account
bal./GDP
0.006
(0.023)
0.095***
(0.031)
(0.018)
0.041
(0.032)
VIX index (log) 0.574***
(0.159)
0.054
(0.203)
Constant 19.978***
(4.052)
20.708***
(3.876)
20.796***
(3.445)
21.098***
(3.999)
15.019***
(3.891)
14.737***
(3.981)
14.259***
(3.681)
19.025***
(5.416)
Observations 595 595 595 595 1101 1101 1101 1097
R-squared 0.224 0.730 0.749 0.784 0.138 0.158 0.202 0.259
Notes: See note in Table 5 for variable denitions. Clustered standard errors (by country) reported in parentheses. *,**,***
indicate signicance at 10, 5 and 1 percent levels, respectively.
29
Of course, one needs to be wary of endogeneity in examining this relationship, though Afonso et al. (2012) nd reverse
causality from CDS spreads and government bond yields to credit ratings to be more of a concern when using very high
frequency (e.g., daily) data.
30
To estimate the effect of ratings on CDS spreads and bond yields, we transform the credit rating information of Moodys and
Standard & Poor into a discrete variable using a linear scale as reported in Table A4.
Table 8
Sovereign risk pricing and credit rating, 2000Q12011Q4.
Dependent
variable
CDS spread Real long-term government bond yield
Specication Moody S & P Moody S & P
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16)
Fiscal space
(log)
0.243*
(0.118)
0.083***
(0.025)
0.024
(0.047)
0.018
(0.045)
0.213*
(0.118)
0.022
(0.033)
0.008
(0.045)
0.016
(0.043)
0.379**
(0.173)
0.418**
(0.198)
0.340**
(0.130)
0.325**
(0.124)
0.439**
(0.182)
0.473**
(0.206)
0.408***
(0.115)
0.393***
(0.110)
Eurozone 0.390
(0.341)
0.174
(0.145)
0.682**
(0.270)
0.701**
(0.281)
0.373
(0.339)
0.075
(0.181)
0.590**
(0.281)
0.613**
(0.284)
0.127
(0.286)
0.103
(0.309)
0.813**
(0.373)
0.839**
(0.359)
0.277
(0.257)
0.255
(0.268)
0.965**
(0.351)
0.982***
(0.344)
Crisis 2.368***
(0.138)
1.671***
(0.261)
1.645***
(0.273)
2.420***
(0.149)
1.662***
(0.280)
1.635***
(0.282)
0.687*
(0.331)
1.709***
(0.359)
1.754***
(0.361)
0.628*
(0.357)
1.692***
(0.352)
1.732***
(0.356)
Eurozone
Crisis
0.602*
(0.301)
0.596*
(0.300)
0.594*
(0.321)
0.601*
(0.317)
2.000***
(0.573)
1.906***
(0.547)
1.993***
(0.584)
1.967***
(0.565)
Credit rating 0.239***
(0.058)
0.247***
(0.034)
0.226***
(0.025)
0.212***
(0.027)
0.209***
(0.052)
0.258***
(0.039)
0.238***
(0.031)
0.231***
(0.033)
0.100
(0.123)
0.118
(0.117)
0.087
(0.134)
0.045
(0.131)
0.028
(0.120)
0.015
(0.117)
0.033
(0.131)
0.043
(0.128)
Rating
downgrade
0.436**
(0.194)
0.329**
(0.139)
1.739***
(0.579)
0.760*
(0.400)
Rating upgrade 0.032
(0.253)
0.187
(0.146)
0.162
(0.295)
0.525**
(0.245)
Real per
capita (log)
0.407
(0.454)
0.634**
(0.237)
0.979**
(0.382)
1.026**
(0.381)
0.446
(0.528)
0.469
(0.280)
0.728*
(0.378)
0.765*
(0.384)
0.732
(0.790)
0.599
(0.786)
1.253
(0.772)
1.364*
(0.738)
1.188
(0.722)
1.077
(0.728)
1.631**
(0.640)
1.667**
(0.624)
Trade
openness
1.554
(0.907)
1.494
(0.888)
1.859**
(0.876)
1.788**
(0.849)
0.400
(1.603)
0.217
(1.490)
0.734
(1.574)
0.652
(1.512)
Ination 0.183
(0.135)
0.178
(0.135)
0.086
(0.130)
0.096
(0.131)
0.119
(0.199)
0.160
(0.196)
0.217
(0.189)
0.232
(0.186)
CA./GDP 0.004
(0.018)
0.002
(0.018)
0.002
(0.018)
0.004
(0.018)
0.090**
(0.035)
0.097***
(0.034)
0.095***
(0.029)
0.099***
(0.030)
Fiscal./GDP 0.027
(0.016)
0.026
(0.016)
0.028*
(0.015)
0.027*
(0.015)
0.037
(0.034)
0.038
(0.031)
0.042
(0.033)
0.042
(0.031)
VIX index
(log)
0.611***
(0.149)
0.628***
(0.150)
0.643***
(0.154)
0.662***
(0.154)
0.060
(0.205)
0.084
(0.208)
0.045
(0.207)
0.070
(0.207)
Constant 12.845***
(3.980)
13.341***
(2.173)
14.452***
(4.018)
14.627***
(4.008)
12.892***
(4.398)
11.980***
(2.520)
11.945***
(3.919)
12.153***
(3.960)
13.074**
(5.979)
12.422*
(6.041)
17.797**
(6.738)
17.984**
(6.475)
15.551***
(5.196)
15.024***
(5.261)
19.354***
(5.247)
19.387***
(5.162)
Observations 595 595 595 595 595 595 595 595 1101 1101 1097 1097 1101 1101 1097 1097
R-squared 0.297 0.808 0.844 0.846 0.273 0.804 0.844 0.846 0.147 0.170 0.264 0.276 0.139 0.159 0.259 0.264
Notes: Real long-term government bond yield and (log) CDS are averaged over the quarter. Credit rating is the average (numeric) rating over the quarter. Credit rating downgrade (upgrade) is
a binary variable equal to 1 if the sovereign credit rating was lowered (raised) in the quarter, and zero otherwise. All other variables are dened as in Tables 3 and 4. Clustered standard errors (by
country) reported in parentheses. *,**,*** indicate signicance at 10, 5 and 1 percent levels, respectively.
signicant when we include Moodys rating with scal space and the eurozone membership dummy
variable (while controlling for real per capita income), and also when we control for the crisis (Table 8,
cols. [1] and [2]).
31
The estimated effect of the scal space variable however turns insignicant when
we control for other variables (cols. [3] and [4]), and is also mostly insignicant when we include the
Standard & Poor (S&P) rating in the specication (cols. [5][8]). The effect of both Moodys and S&Ps
ratings on CDS spreads is strongly signicant with an increase in ratings implying lower CDS spreads.
By contrast, both Moodys and S&Ps ratings have a statistically negligible impact on real govern-
ment bond yields, while scal space remains signicant in all specications (Table 8, col. [9][16]). For
a given credit rating, however, changes in ratings have an impact on both CDS spreads and bond yields,
though the effects are asymmetric: sovereign downgrades by both Moodys and S&P are of greater
magnitude and statistical signicance than corresponding upgrades (Table 8, cols. [4], [8], [12], [16]).
32
Despite controlling for credit ratings or changes in the ratings, the estimated pre-crisis positive and
post-crisis negative premium for eurozone countries (as indicated by the estimated coefcients for the
dummy variable eurozone, and the interaction termeurozone crisis) remains. This suggests that either
credit ratings are not fully efcient in the sense of incorporating all available information, or the market
overreacts, with wider spreads than justied by the underlying probability of default (as captured by
ratings).
5.3. Robustness
While the estimation results presented in Tables 58 provide strong empirical evidence that CU
membership matters for both CDS rates and bond yields, controlling for scal space and other
macroeconomic variables as well as credit ratings, we further check the robustness of our key results by
estimating various alternative specications for both annual and quarterly data. The obtained results
are however similar for both types of data; hence, in what follows, we present the results for quarterly
data (those for annual data are available upon request).
A pertinent concern in the estimation of sovereign risk pricing is the endogeneity between the scal
space variable and CDS spreads/government bond yields. To take account of this concern, Table 9 (cols.
[1] and [2]) reports the estimation results using (four-quarter) lagged scal space instead of the
contemporaneous scal space variable in both the parsimonious and general specications. While
the estimated coefcient of scal space increases slightly, this has no appreciable impact on the other
variables of interest: eurozone countries continue to enjoy a pre-crisis premium, but saw sharper rises
in CDS spreads than other advanced economies during the crisis.
In cols. [3] and [4], we add country xed effects (FE) to our specications, while both country-xed
and time effects (FE/TE) are included in cols. [5] and [6]. This raises the R-squared of the regressions
marginally, and somewhat weakens the statistical signicance of the scal space variable. The euro-
zone and crisis dummy variables drop out from the regression due to perfect collinearity with the
country xed effects for eurozone countries, and the year effects for 20082011, respectively, but the
estimated coefcient of the interaction term between the eurozone and crisis dummy variables
remains strongly signicant. In cols. [7] and [8], we split the crisis period and introduce dummy
variables for individual years to analyze the impact of each year separately on CDS spreads. As ex-
pected, the largest impact of the crisis on CDS spreads in general, as well as in eurozone countries, was
in 2011 when political uncertainty over potential bailouts in the eurozone intensied.
Next, in cols. [9] and [10], we replace our measure of global risk aversion, the VIX index, with the
LIBOR-US T-bill (TED) spread, but again doing so has little impact on the results. Finally, we consider
31
That our measure remains signicant even after controlling for the rating suggests that perhaps the rating is not fully
capturing the available scal space (at least) in our sample of countries. Moodys recently adopted the approach of Ghosh et al.
(2011) followed in this paper to assess the available scal space in advanced economies (source: http://www.moodysanalytics.
com/w/media/Insight/Economic-Analysis/Special-Studies/2011-12-13-Fiscal-Space.ashx).
32
Several studies (e.g., Norden and Weber, 2004; Hull et al., 2004; Afonso et al., 2012) nd a similar result, hypothesizing that
agents care more strongly about utility losses than they do about gains of equal magnitudes. The asymmetric impact may also
reect that governments have an incentive to leak information that would lead to an upgrade (and therefore upgrades are
priced in by markets even before they are announced), whereas they have no such incentive when it comes to downgrades.
Table 9
Robustness: CDS spreads, scal space and Euro membership, 2000Q12011Q4.
Specication Lagged FS FE FE/TE Crisis years Ted spread Financial contagion
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Fiscal space (log) 0.194***
(0.048)
0.261***
(0.069)
0.133*
(0.077)
0.087
(0.076)
0.112*
(0.064)
0.075
(0.054)
0.164***
(0.044)
0.188***
(0.054)
0.145***
(0.048)
0.137*
(0.070)
0.171***
(0.051)
0.119
(0.071)
Eurozone 0.985***
(0.325)
1.077***
(0.304)
1.075***
(0.323)
1.162***
(0.312)
0.994***
(0.330)
1.149***
(0.306)
0.303
(0.461)
0.756
(0.528)
Crisis 1.889***
(0.277)
1.426***
(0.286)
1.793***
(0.287)
1.415***
(0.393)
1.861***
(0.272)
1.839***
(0.284)
2.257***
(0.394)
1.778***
(0.391)
(0.318)
0.702**
(0.290)
0.848**
(0.359)
0.860**
(0.380)
0.954***
(0.244)
0.754***
(0.260)
0.927**
(0.334)
0.772**
(0.332)
0.422
(0.402)
0.445
(0.367)
y2008 1.162***
(0.216)
0.384*
(0.211)
y2009 1.905***
(0.296)
1.122***
(0.267)
y2010 1.936***
(0.353)
1.474***
(0.355)
y2011 2.187***
(0.327)
1.691***
(0.335)
Eurozone y2008 0.621**
(0.251)
0.623**
(0.267)
Eurozone y2009 0.333
(0.378)
0.154
(0.378)
Eurozone y2010 1.191***
(0.405)
1.009**
(0.373)
Eurozone y2011 1.548***
(0.413)
1.402***
(0.387)
Real GDP per
capita (log)
1.846***
(0.341)
1.884***
(0.376)
1.226
(2.070)
1.556
(2.501)
1.762
(2.037)
3.648*
(2.096)
1.842***
(0.340)
1.963***
(0.405)
1.760***
(0.335)
2.032***
(0.366)
1.280***
(0.345)
1.734***
(0.532)
(0.898)
0.615
(3.733)
0.006
(4.635) 1.873*
(0.981)
2.274**
(0.984)
2.827**
(1.292)
Ination 0.108
(0.195)
0.500**
(0.222)
0.618***
(0.170)
0.003
(0.208)
0.133
(0.173)
0.264
(0.197)
CAB/GDP 0.018
(0.023)
0.021
(0.030)
0.013
(0.026)
0.015
(0.031)
0.003
(0.026)
0.030
(0.020)
Fiscal
balance/GDP
0.768***
(0.164)
0.585***
(0.158)
0.949***
(0.117)
1.051***
(0.112)
0.397**
(0.166)
VIX index (log) 0.028*
(0.015)
0.032
(0.030)
0.008
(0.024) 0.023
(0.020)
0.029
(0.019)
0.035
(0.021)
A
.
R
.
G
h
o
s
h
e
t
a
l
.
/
J
o
u
r
n
a
l
o
f
I
n
t
e
r
n
a
t
i
o
n
a
l
M
o
n
e
y
a
n
d
F
i
n
a
n
c
e
3
4
(
2
0
1
3
)
1
3
1
1
6
3
1
5
2
Table 9 (continued)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Ted spread (log) 0.009
(0.061)
0.015
(0.066)
Bank lending
to EMU
0.426
(0.676)
1.175
(0.958)
Bank to
EMU*Crisis 0.210
(0.613)
0.223
(0.668)
Observations 595 595 595 595 595 595
595 595 595 595 482 482
R-squared 0.763 0.806 0.828 0.851 0.888 0.916 0.816 0.863 0.749 0.772 0.767 0.798
Country xed
effects
No No Yes Yes Yes Yes No No No No No No
Year effects No No No No Yes Yes
Yes Yes No No No No
Notes: Dependent variable is (log) CDS spread averaged over the quarter. Bank lending to EMU is the share of foreign claims on EMU countries in the sample in total foreign claims
(immediate borrower basis; quarterly observations as reported by BIS). Ted spread is average of daily observations over the quarter. All other variables are dened as in Table 7. Constant is
included in all specications. Clustered standard errors (by country) reported in parentheses. *,**,*** indicate signicance at 10, 5 and 1 percent levels, respectively.
A
.
R
.
G
h
o
s
h
e
t
a
l
.
/
J
o
u
r
n
a
l
o
f
I
n
t
e
r
n
a
t
i
o
n
a
l
M
o
n
e
y
a
n
d
F
i
n
a
n
c
e
3
4
(
2
0
1
3
)
1
3
1
1
6
3
1
5
3
Table 10
Robustness: Real bond yield, scal space and Euro membership, 2000Q12011Q4.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Fiscal space (log) 0.239*
(0.125)
0.173
(0.124)
0.829***
(0.160)
0.781***
(0.167)
0.825***
(0.170)
0.755***
(0.182)
0.344*
(0.176)
0.356**
(0.151)
0.361*
(0.177)
0.372**
(0.158)
0.315
(0.195)
0.338**
(0.140)
Eurozone 0.743**
(0.306)
0.953**
(0.356)
0.771**
(0.320)
0.947**
(0.363)
0.691**
(0.315)
0.877**
(0.357)
0.572*
(0.325)
1.140***
(0.372)
Crisis 1.379***
(0.347)
1.719***
(0.356)
1.140***
(0.308)
1.364***
(0.354)
1.322***
(0.351)
1.532***
(0.349)
0.492
(0.543)
0.959
(0.581)
(0.672)
2.459***
(0.617)
0.962*
(0.516)
1.116**
(0.517)
0.825*
(0.462)
1.067**
(0.464)
1.925***
(0.641)
1.991***
(0.587)
1.742**
(0.725)
1.694**
(0.646)
y2008 1.447***
(0.394)
1.803***
(0.486)
1.899***
(0.262)
1.899***
(0.262)
y2009 1.425***
(0.383)
2.027***
(0.478)
0.955
(0.569)
1.299**
(0.588)
y2010 1.025**
(0.468)
2.009***
(0.570)
1.183**
(0.443)
1.442***
(0.439)
y2011 1.257
(0.762)
2.161**
(0.886)
2.136***
(0.558)
2.457***
(0.452)
(0.377)
1.120***
(0.360)
(0.718)
1.732**
(0.730)
(0.638)
1.494**
(0.547)
(1.222)
3.156**
(1.133)
Real GDP per capita
(log)
0.823**
(0.370)
1.319**
(0.505)
5.045***
(1.292)
4.987**
(2.066)
5.651**
(2.313)
3.392
(3.060)
0.940**
(0.402)
1.472**
(0.533)
0.790*
(0.416)
1.326**
(0.538)
0.553
(0.533)
1.596***
(0.455)
Trade openness 0.144
(1.753)
2.590
(4.456)
2.293
(4.960)
0.765
(1.663)
0.680
(1.814)
2.528*
(1.310)
Ination 0.140
(0.130)
0.122
(0.176)
0.126
(0.170)
0.164
(0.133)
0.140
(0.137)
0.145
(0.254)
CAB/GDP 0.095***
(0.030)
0.066*
(0.035)
0.069**
(0.030)
0.089**
(0.036)
0.082**
(0.034)
0.069**
(0.030)
(0.037)
0.055
(0.042)
0.072
(0.046)
0.040
(0.036)
0.040
(0.033)
0.066
(0.047)
(0.198)
0.265
(0.185)
0.520***
(0.166)
0.191
(0.214)
0.239
(0.204)
Ted spread (log) 0.350***
(0.060)
0.235***
(0.067)
A
.
R
.
G
h
o
s
h
e
t
a
l
.
/
J
o
u
r
n
a
l
o
f
I
n
t
e
r
n
a
t
i
o
n
a
l
M
o
n
e
y
a
n
d
F
i
n
a
n
c
e
3
4
(
2
0
1
3
)
1
3
1
1
6
3
1
5
4
Table 10 (continued)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Bank lending to EMU 0.158
(1.249)
0.521
(1.326)
Bank lending to EMU*Crisis 2.481
(2.227)
2.030
(2.238)
Observations 1009 1009 1101 1097 1101 1097 1101 1097 1009 1009 816 812
R-squared 0.150 0.204 0.373 0.390 0.402 0.420 0.235 0.286 0.208 0.249 0.220 0.284
Country xed effects No No Yes Yes Yes Yes No No No No No No
Year effects No No No No Yes Yes Yes Yes No No No No
Notes: Real long-term government bond yield is averaged over the quarter. Bank lending to EMU is the share of foreign claims on EMU countries in the sample in total foreign claims
(immediate borrower basis; quarterly observations as reported by BIS). Ted spread is average of daily observations over the quarter. All other variables are dened as in Table 7. Constant is
included in all specications. Clustered standard errors (by country) reported in parentheses. *,**,*** indicate signicance at 10, 5 and 1 percent levels, respectively.
A
.
R
.
G
h
o
s
h
e
t
a
l
.
/
J
o
u
r
n
a
l
o
f
I
n
t
e
r
n
a
t
i
o
n
a
l
M
o
n
e
y
a
n
d
F
i
n
a
n
c
e
3
4
(
2
0
1
3
)
1
3
1
1
6
3
1
5
5
possible nancial contagion from the eurozone countries during the crisis by including the countrys
exposure through cross-border bank loans to (other) eurozone countries, and its interaction with the
crisis variable. Although the contagion variable (pre- or during the crisis) is insignicant, its inclusion
statistically weakens the signicance of eurozone membership (most likely because of high correlation
between the eurozone membership and bank exposure to EMU variables in this smaller sample).
33
In Table 10, we repeat the above robustness analysis for real bond yields, and nd the results to be
very similar to those reported in Table 7. Specically, in all estimations (Table 10, cols. [1][12]), we nd
that eurozone countries had lower government bond yields relative to other advanced economies
before the crisis. But this pattern changed dramatically during the crisis when eurozone countries on
average experienced higher yields than their other advanced counterparts.
6. Conclusion
This paper examines howmembership in a currency union affects debt sustainability, and what this
implies for market pricing of default risk of eurozone members given their available scal space. We
argue that CU membership carries three key implications: rst, when the budget constraint of a CU
member binds, it may be expected to receive assistance fromother members of the union, thus making
a given level of debt more sustainable and lowering the likelihood of default. Conversely, however, the
greater need to rely on scal policy for macroeconomic stabilization implies higher government bond
yields (or credit default swaps (CDS) rates) especially when the government has almost exhausted its
scal space. Likewise, the limited scope for raising the ination tax or eroding the real value of
government debt means that CU members do not have the same safety valve as non-CU members,
and should therefore face correspondingly higher spreads on their public debt.
In examining how the market assesses default risk of CU members, it is essential to control for
countries scal space. Applying a forward-looking concept of scal space, as developed in Ghosh et al.
(2011), to a sample of 23 advanced economies, our estimates suggest that Greece, Italy, Japan and
Portugal have the least scal space, while Belgium, France, Ireland, Spain, and the United States are also
constrained in their degree of scal maneuvre. Australia, Korea, New Zealand, Norway, and Sweden
generally have the most scal space to deal with unexpected shocks.
We nd that our concept of scal space is a better predictor of government bond yields or of CDS rates
than just the level of public debt, and the relationship is nonlinear, rising more sharply as scal space is
exhausted. Controlling for scal space, we nd that, during the pre-crisis period (19992008), CDS and
bond rates for eurozone members were systematically and signicantly lower than would be predicted
given their scal space perhaps because markets anticipated possible bailouts (or suffered from irra-
tional exuberance over the introductionof the euro). But whenthe crisis erupted, the opposite happened:
rates for eurozone countries rose more sharply than for other advanced economies. This suggests that as
the expected bailouts of countries running into debt difculties did not occur with the hoped-for alacrity,
markets began to recognize that the policy constraints of CU membership made it more difcult to
maintain debt sustainability and this is what is reected in the wider rates for eurozone members.
It does not follow, however, that observed spreads are necessarily rational; even allowing for post-
2008 eurozone spreads to be different (i.e., once markets recognized that bailouts would not be
automatic and that currency union membership imposed constraints on adjustment), there are sizable
residuals for some eurozone countries. While these are presumably justiable in some cases, in others
it is hard to rule out the possibility that they reect multiple equilibria whereby pessimistic expec-
tations lead to excessively wide spreads that are then justied by the greater default risk given the high
spreads. If that is the case, then there is a clear role for ofcial nancing to provide liquidity to help
avoid an illiquidity problem from becoming an insolvency problem. But if that is not the case, then
provision of ofcial nancing might simply delay necessary adjustment. While this paper has gone
some way toward understanding the determinants of spreads for currency union countries, future
research will need to try to pin down more precisely how much of the observed spreads is justied on
33
In addition to a countrys nancial exposure to all eurozone countries, we also analyze the effect of nancial exposure to
Greece only, but nd the results to be very similar.
the basis of fundamentals, including the countrys scal space and its monetary and exchange rate
arrangements, and how much must be attributed to multiple equilibria or market skittishness.
Acknowledgments
We are grateful to Peter Srensen, IMF colleagues, and participants at the Danmarks Nationalbanks
conference on the European Sovereign Debt Crisis and the European Central Banks Sovereign Debt
Workshop for helpful comments and suggestions, and to Hyeon Ji Lee for excellent research assistance.
The views expressed herein are those of the authors and should not be attributed to the IMF, its
Executive Board, or its management.
Appendix A
Fig. A1. Actual (solid line) and tted (dashed line) CDS spreads (in logs) for selected eurozone countries: annual data. Note: Based on
estimates reported in Table 5, col. [4].
0
2
4
6
8
2003q2 2005q2 2007q2 2009q2 2011q2
Austria
0
2
4
6
8
2003q12005q12007q12009q12011q1
Belgium
0
2
4
6
8
2003q12005q12007q12009q12011q1
France
0
2
4
6
8
2003q12005q12007q12009q12011q1
Germany
0
2
4
6
8
2003q12005q12007q12009q12011q1
Greece
0
2
4
6
8
2007q4 2008q4 2009q4 2010q4 2011q4
Ireland
0
2
4
6
8
2003q12005q12007q12009q12011q1
Portugal
0
2
4
6
8
2004q2 2006q12007q42009q32011q2
Spain
0
2
4
6
8
2003q12005q12007q12009q12011q1
Italy
Fig. A2. Actual (solid line) and tted (dashed line) CDS spreads (in logs) for selected eurozone countries: quarterly data. Note: Based on
estimates reported in Table 7, col. [4].
Table A1
Estimation results for the scal reaction function.
Sample specication 19702007 19852007
(1) (2) (3) (4)
Lagged debt 0.208*** 0.225*** 0.081 0.086
(0.059) (0.061) (0.076) (0.070)
Lagged debt_square 0.003*** 0.003*** 0.002* 0.002*
(0.001) (0.001) (0.001) (0.001)
Lagged debt_cubic 0.00001*** 0.00001*** 0.00001* 0.00001**
(3.0e-06) (3.0e-06) (3.0e-06) (3.0e-06)
Output gap 0.497*** 0.491*** 0.485*** 0.441***
(0.047) (0.046) (0.053) (0.053)
Government expenditure gap 0.185*** 0.184*** 0.183*** 0.183***
(0.047) (0.045) (0.052) (0.047)
Trade openness 0.091* 0.146***
(0.050) (0.054)
Ination 3.400 4.620**
(2.519) (2.008)
Oil price
a
8.775*** 9.529***
(3.216) (3.244)
Age dependency 0.072
(0.101)
Future dependency 0.015
(0.067)
Nonfuel price
a
3.005
(8.362)
Political stability 0.068**
(0.030)
IMF arrangement 1.142
(0.999)
Fiscal rules 0.300
(0.347)
Observations 642 642 496 491
Number of countries 23 23 23 23
R-squared 0.282 0.316 0.304 0.405
AR (1) coefcient 0.791 0.760 0.819 0.749
Notes: Dependent variable is general government primary balance to GDP (in percent); in all specications, country-specic
xed effects included, and error term assumed to follow an AR (1) process; robust standard errors reported in parentheses;
***, **, and * denote signicance at 1, 5, and 10 percent levels, respectively.
a
Applies to oil and nonoil commodities exporters only.
Table A2
Estimation results for the scal reaction function with eurozone.
Sample specication 19702007 19852007
(1) (2) (3) (4)
Lagged debt 0.226*** 0.253*** 0.102 0.111
(0.060) (0.061) (0.077) (0.071)
Lagged debt_square 0.003*** 0.003*** 0.002* 0.002**
(0.001) (0.001) (0.001) (0.001)
Lagged debt_cubic 1.2E-05*** 1.2E-05*** 6.0E-06* 6.2E-06**
(0.000) (0.000) (0.000) (0.000)
Output gap 0.499*** 0.487*** 0.491*** 0.453***
(0.050) (0.049) (0.058) (0.056)
Government expenditure gap 0.176*** 0.161*** 0.169*** 0.136**
(0.056) (0.056) (0.061) (0.056)
Trade openness 0.093* 0.119**
(0.054) (0.057)
Ination 7.293*** 0.67
(2.683) (5.728)
Oil price
a
8.912*** 9.676***
(3.244) (3.359)
Age dependency 0.012 0.072
(0.094) (0.113)
Nonfuel commodity price
a
0.403
(7.689)
Political stability 0.082**
(0.035)
Fiscal rules 0.215
(0.352)
Observations 551 551 406 401
R-squared 0.285 0.329 0.311 0.407
AR (1) coefcient 0.796 0.751 0.832 0.755
Notes: Dependent variable is general government primary balance to GDP. Eurozone countries are treated as a single entity over
19992007, and (nominal GDP) weighted average is taken for primary balance, lagged debt, output gap and government gap
(simple average is taken for trade openness, ination rate, age dependency ratio, political stability index, and scal rules index).
In all specications, country-specic xed effects are included, and error term is assumed to follow an AR (1) process. Robust
standard errors in parentheses. *, **, *** indicate signicance at 10, 5 and 1 percent levels, respectively.
a
Applies to oil and nonoil commodities exporters only.
Table A3
Variable denitions and data sources.
Variable Description Source
Age dependency ratio Ratio of the dependent to working
age (1564 years) population.
UNs database. Available online at: http://data.un.
org/Data.aspx?qdependencyratio(percent)
&dPopDiv&fvariableID%3a42
CDS spread In logs Bloomberg
Crisis Binary variable equal to 1 for 200811
(2008Q3:2011Q4 in quarterly data)
Currency union Binary variable equal to 1 for eurozone
countries in the sample
Current account
balance to GDP
In percent Authors calculations based on IMFs World
Economic Outlook (WEO) database.
Fiscal rules Binary variable equal to 1 if a country
has any type (expenditure, revenue,
balance budget, and debt) of scal
rule in a given year, and 0 otherwise.
IMFs Fiscal Rules database, 19852009
Future age dependency
ratio
Projected age dependency ratio 20
years ahead.
UNs database. Available online at: http://data.un.
org/Data.aspx?qdependencyratio(percent)
&dPopDiv&fvariableID%3a42
Government bond yield
(long-term)
In percent IMFs WEO database
Government
expenditure gap
Difference between actual and potential
(calculated using the Hodrick-Prescott
lter) real government consumption
spending
Authors calculations based on WEO database
IMF arrangement Binary variable equal to one if a
country has an IMF support program
in a given year, and zero otherwise.
IMFs History of Lending Arrangements database
Available online at http://www.imf.org/external/
np/n/tad/extarr1.aspx.
Ination Three-year lagged moving average of
CPI ination
Lagged debt to GDP ratio In percent IMFs WEO database
Nonfuel commodity price Log of (trend) nonfuel commodity price
index applied to nonfuel commodity
exporters only.
Oil price Log of (trend) oil price applied to oil
exporters only.
Output gap Difference between actual and potential
(calculated using the HodrickPrescott
lter) real GDP.
Political stability index Smaller (larger) values indicating
higher (lower) risk.
International Country Risk Guide (ICRG) dataset
Primary balance to
GDP ratio
In percent IMFs WEO and OECD databases
Real GDP growth rate In percent IMFs WEO database
Real GDP per capita In logs World Banks World Development Indicators
Trade openness Sum of exports and imports to
GDP (in percent)
VIX index In logs Bloomberg
References
Abiad, A., Ostry, J., 2005. Primary Surpluses and Sustainable Debt Levels in Emerging Market Countries. IMF Policy Discussion
Paper 05/6.
Afonso, A., Furceri, A., Gomes, P., 2012. Sovereign credit ratings and nancial market linkages: application to European data.
Journal of International Money and Finance 31 (3), 606638.
Aizenman, J., Hutchison, M., Jinjarak, Y., 2011. What Is the Risk of European Sovereign Debt Defaults?. In: Fiscal Space, CDS
Spreads and Market Pricing of Risk. NBER Working Paper No. 17407.
Attinasi, M., Checherita, C., Nickel, C., 2009. What Explains the Surge in Eurozone Sovereign Spreads during the Financial Crisis
of 200709?. ECB Working Paper no. 1131/2009.
Barrios, S., Iversen, P., Lewandowska, M., Setzer, R., 2009. Determinants of Intra-euro-area Government Bond Spreads during the
Financial Crisis. European Economy Economic Papers No. 388.
Bernoth, K., Erdogan, B., 2012. Sovereign bond yield spreads: a time-varying coefcient approach. Journal of International
Money and Finance 31 (3), 639656.
Bertola, G., Ichino, A., 1995. Wage Inequality and Unemployment: United States vs. Europe. In: NBER Macroeconomics Annual,
vol. 10. University of Chicago Press, Illinois.
Bohn, H., 1998. The behavior of U.S. public debt and decits. Quarterly Journal of Economics 113 (3), 949963.
Bohn, H., 2008. The sustainability of scal policy in the United States. In: Neck, R., Sturm, J. (Eds.), Sustainability of Public Debt.
MIT Press, Cambridge, pp. 1549.
Clark, T., van Wincoop, E., 2001. Borders and business cycles. Journal of International Economics 55 (1), 5985.
De Grauwe, P., 2011. The Governance of a Fragile Eurozone. CEPS Working Paper No. 346.
Fontana, A., Scheicher, M., 2010. An Analysis of Euro Area Sovereign CDS and Their Relation with Government Bonds. ECB
Working Paper no. 1271/2010.
Gande, A., Parsley, D., 2005. News spillovers in the sovereign debt market. Journal of Financial Economics 75 (3), 691734.
Ghosh, A., Kim, J., Mendoza, E., Ostry, J., Qureshi, M., 2011. Fiscal Fatigue, Fiscal Space and Debt Sustainability in Advanced
Economies. NBER Working Paper No. 16782. Also forthcoming in the Economic Journal.
Gros, D., 2011. External versus Domestic Debt in the Euro Crisis. Available online. http://www.voxeu.org/index.php?qnode/
6550.
Heppke-Falk, K., Hfner, F., 2004. Expected Budget Decits and Interest Rate Swap Spreads Evidence for France, Germany and
Italy. Discussion Paper Series 1: Studies of the Economic Research Center No. 40/2004.
Hull, J., Predescu, M., White, A., 2004. The relationship between credit default swap spreads, bond yields, and credit rating
announcements. Journal of Banking and Finance 28 (11), 27892811.
IMF, 2010. Global Financial Stability Report: April 2010. International Monetary Fund, Washington DC.
IMF, 2012. Fiscal Monitor: April 2012. International Monetary Fund, Washington DC.
Islamilescu, J., Kazemi, H., 2010. The reaction of emerging market credit default swap spreads to sovereign credit rating changes.
Journal of Banking and Finance 34 (12), 28612873.
Maltritz, D., 2012. Determinants of sovereign yield spreads in the Eurozone: a Bayesian approach. Journal of International
Money and Finance 31 (3), 639656.
Table A4
Coding of Moody and S&P ratings.
Credit rating Moody S&P Transformation
Investment grade Aaa AAA 20
Aa1 AA 19
Aa2 AA 18
Aa3 AA 17
A1 A 16
A2 A 15
A3 A 14
Baa1 BBB 13
Baa2 BBB 12
Baa3 BBB 11
Speculative grade Ba1 BB 10
Ba2 BB 9
Ba3 BB 8
B1 B 7
B2 B 6
B3 B 5
Caa1 CCC 4
Caa2 CCC 3
Caa3 CCC 2
Ca CC 1
C C
SD
D
Mendoza, E., Ostry, J., 2008. International evidence on scal solvency: is scal policy responsible? Journal of Monetary
Economics 55 (6), 10811093.
Norden, L., Weber, M., 2004. Informational efciency of credit default swap and stock markets: the impact of credit rating
announcements. Journal of Banking and Finance 28 (11), 28132843.
Ostry, J., Ghosh, A., Kim, J., Qureshi, M., 2010. Fiscal Space. IMF Staff Position Note SPN/10/11.
Roubini, N., Sachs, J., 1988. Political and Economic Determinants of Budget Decits in the Industrial Democracies. NBER Working
Paper No. 2682.

Journal of International Money and Finance: Atish R. Ghosh, Jonathan D. Ostry, Mahvash S. Qureshi

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Journal of International Money and Finance: Atish R. Ghosh, Jonathan D. Ostry, Mahvash S. Qureshi

Uploaded by

Copyright:

Available Formats

Fiscal space and sovereign risk pricing in

, the interest payment

, the primary balance exceeds the interest

, while raising the debt limit and corresponding scal space.

0) to debt of about 105 and 120 percent of GDP under

exists using the historical interest rates.

and d, output and government expenditure gaps are assumed to be closed.

You might also like