How do Experts Forecast Sovereign Spreads?*
  • 1 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund
  • | 2 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Contributor Notes

Author’s E-Mail Address: mpoplawskiribeiro@imf.org

This paper assesses how forecasting experts form their expectations about future government bond spreads. Using monthly survey forecasts for France, Italy and the United Kingdom between January 1993 and October 2014, we test whether respondents consider the expected evolution of the fiscal balance—and other economic fundamentals—to be significant drivers of the expected bond yield differential over a benchmark German 10-year bond. Our main result is that a projected improvement of the fiscal outlook significantly reduces expected sovereign spreads. This suggests that credible fiscal plans affect market experts’ expectations and reduce the pressure on sovereign bond markets. In addition, we show that expected fundamentals generally play a more important role in explaining forecasted spreads compared to realized spreads.

Abstract

This paper assesses how forecasting experts form their expectations about future government bond spreads. Using monthly survey forecasts for France, Italy and the United Kingdom between January 1993 and October 2014, we test whether respondents consider the expected evolution of the fiscal balance—and other economic fundamentals—to be significant drivers of the expected bond yield differential over a benchmark German 10-year bond. Our main result is that a projected improvement of the fiscal outlook significantly reduces expected sovereign spreads. This suggests that credible fiscal plans affect market experts’ expectations and reduce the pressure on sovereign bond markets. In addition, we show that expected fundamentals generally play a more important role in explaining forecasted spreads compared to realized spreads.

I. Introduction

The global financial crisis fallout significantly unsettled European sovereign bond markets. Bond spreads jumped during the crisis mainly on the back of a deteriorating outlook for public finances, worsening macro-economic conditions, and rising international risk aversion. Views diverge over the importance of these drivers of sovereign bond market turbulence. One interpretation is that markets paid little attention to fiscal and economic fundamentals before the crisis in the European Monetary Union (EMU), and thus spreads were excessively low. With the deterioration of the fiscal and economic outlook during the crisis, bond investors rediscovered the role of those fundamentals (Ghosh et al., 2013). Another interpretation is that risk premia are mostly driven by market sentiment and as such independent of the underlying expected fundamentals (Blommestein et al., 2012, De Grauwe and Ji, 2013). Thus, uncertainty about policy outcomes could make investors sell sovereign bonds out of fear and lead to a mispricing in bond markets (see also Akerlof and Shiller, 2009). Each of these views carries a different policy prescription. Under the former view, fiscal consolidation and improved economic conditions are necessary steps in curbing bond spreads. Under the latter view, bond market reactions may be difficult to predict as they are fueled by sentiment and instincts, therefore pursuing virtuous macroeconomic policies may be insufficient to reduce spreads.

Some papers attribute a role to the sovereign’s fiscal position in determining realized bond spreads, i.e., the actual spreads observed in bond markets (e.g., Codogno et al., 2003; Akitoby and Stratmann, 2008, Schuknecht et al., 2010; Aizenman et al., 2013). However, other studies fail to find fiscal variables among the main determinants of bond spreads for advanced economies. Conversely, they highlight that bond yield spreads are increasingly driven by international factors that reflect global investor risk aversion (Bellas et al., 2010; Favero et al., 2010; Manganelli and Wolswijk, 2009). As they look into the reaction of realized spreads to realized fundamentals, however, most of those studies do not treat expectations explicitly.

More recently, Laubach (2009) suggested directly relating realized sovereign spreads to proxies of expectations on fiscal and other macro developments. The main idea behind this approach is that market prices incorporate the expectations of the future path of fiscal and economic fundamentals, rather than their current or past values. In this context, future implicit or explicit liabilities, such as the size of bank rescue packages and the position of the domestic banking sector, are often found to explain developments in government bond spreads (Ejsing and Lemke, 2009; Attinasi et al., 2009; Gerlach et al., 2010). D’Agostino and Ehrmann (2014) use survey-based forecasts from the Consensus Economics (hereafter generally referred to as “CE”) dataset to investigate the time-varying relationship between realized sovereign bond spreads of the G7 countries and expectations about fiscal and other macroeconomic fundamentals. Their results show that fundamentals and general risk aversion were downplayed in the years preceding the global financial crisis, but an over-pricing of the same factors occurred during the European sovereign debt crisis. Yet, this strand of the literature does not investigate how experts’ forecasts about future bond yield differentials react to changes in their perception of future macroeconomic and fiscal developments.

Indeed, which variables underline the individual experts’ forecast of sovereign spreads is still unknown in the literature. In particular, it is not clear whether forecasting experts value predominantly expected fundamental or non-fundamental factors in their projections. In addition, the relative weight of each specific expected fundamental factor in predicting spreads is also not known. This paper addresses both unknowns. First, we test the importance of expected macroeconomic fundamentals as well as a global risk factor in the forecast of sovereign spreads. Second, we estimate the contribution of three key expected fundamentals—i.e., the overall fiscal balance, defined as general government primary balance plus interest payments (hereafter simply referred to as ‘fiscal balance’), GDP growth, and CPI inflation—in shaping the forecasted spreads.

Addressing these two unknowns is important because they shed light on how expectations about future sovereign spreads are formed, which is directly related to the need to anchor market expectations on fiscal policy and sovereign risk premia. Indeed, according to a recent literature, credibly anchoring the market’s fiscal expectations can significantly improve the effectiveness of macroeconomic policies (Leeper, 2009 and 2010) by alleviating the risk premium and ensuring fiscal solvency (Bi, 2012; and Bi et al., 2013).

Understanding how spread forecasts are shaped is further important because realized observations for financial variables—including bond spreads—do not necessarily reflect expectations about future fiscal and macroeconomic conditions in the same way as future (expected) financial variables. In fact, fundamentals are often found to play an important role in explaining the long-run predictability of financial variables and exchange rates, but generally have little power in accounting for current market developments.3 In addition, a vast finance literature has shown that other factors—including high frequency news and “irregular” trading behaviors (e.g., bandwagon effect, excess speculation, and major trading manipulation)—are important in explaining the short-term dynamics of financial variables and exchange rates;4 whereas their role is less clear in explaining the medium or long-term forecasts of the same variables (e.g., Odean, 1999; Shleifer, 2000; Hirshleifer, 2001; Claessens and Kose, 2013; and Hommes, 2014).

Overall, this literature suggests that fundamentals should affect the expectations and realizations of financial variables (including sovereign spreads) in substantially different ways. However, instead of directly inquiring about such differences via a survey to sovereign bond traders (Cheung and Chinn, 2001) or by designing a laboratory experiment (Hommes et al., 2005), we look into the forecasters’ predictions themselves and investigate how these projections relate to forecasted and realized sovereign bond spreads.

This paper, therefore, uses a regression-based methodology to analyze the role of expected fundamentals in determining spread forecasts. This econometric approach appears to be the most appropriate when the underlying forecast model is not observed. Moreover, the other alternative of directly inquiring about such differences via a survey to sovereign bond traders would be rather difficult to implement and would not deliver the same number of observations, high frequency data and degrees of freedom that our approach does.

Our empirical analysis employs the CE survey-based monthly dataset of individual forecasters The CE survey primarily includes forecasting experts from financial institutions and (private and public) research centers. We focus on France, Italy, and the U.K., and use a period sample from January 1993 until October 2014, which includes a substantial part of the recent financial crisis. Using GMM methods to address possible reverse causality issues, we estimate whether market experts’ projections for the fiscal balance, GDP growth and CPI inflation—as well as a global risk factor—play a significant role in explaining their forecast of the government bond spreads.

Our main finding is that the projections of fiscal and other macroeconomic fundamentals significantly explain the one-year-ahead expectations of the French, Italian and U.K. sovereign spreads over Germany. In particular, an improvement in the one-year-ahead projected fiscal outlook reduces expected spreads. This is especially the case for the Italian spread: a 1 percent rise in the expected surplus ratio to GDP reduces forecasted spreads by around 38 basis points; whereas the effect is smaller—but still significant—for the French and British spreads. The projected fiscal balance and real GDP growth have a substantially larger effect on the expectations of future government spreads compared to regressions based on realized spreads.

The results also show that, until the recent financial crisis, a weaker growth outlook was associated with a reduction in forecasted spreads. This might reflect a flattening of the yield curve in bad economic times due to the expectation of easier monetary policy. Nevertheless, this relationship is reversed for Italian spreads during the financial crisis. In this period, expectations of higher growth considerably reduced the expected Italian 10-year bond spread, suggesting that markets perceive future growth as crucial to the future sustainability of public finance.

We further check whether the distance of an economy from its fiscal limit matters for the determination of the expected spreads in Italy and France, as opposed to the United Kingdom. The fiscal limit can be described as the point at which the government no longer has the ability and willingness to increase its borrowing capacity through changes in tax policy (Leeper, 2013 and 2015). Its importance also depends on whether a country controls its own monetary policy. For example, as part of the EMU, France and Italy delegate the control of their own monetary policy to the European Central Bank (ECB). Hence, for those two countries, the domestic nominal government debt becomes effectively equivalent to real debt and must be backed by real surpluses (Leeper, 2015). For this reason, as those countries’ debt levels approach their fiscal limits, their sovereign yields and spreads should react more strongly. On the other hand, the U.K. controls its own monetary policy and issues nominal bonds, which are a claim to pounds in the future. Therefore, in the event of a very severe crisis that puts the sustainability of public finances at risk, it is more likely that the British authorities would tolerate a deviation from the Bank of England’s inflation target rather than a default on government debt. For these reasons, the fiscal limit is expected to be less binding than in France and Italy.

We test this hypothesis based on the distance of an economy from its fiscal limit as estimated by Polito and Wickens (2015). Our results indeed corroborate the discussion above and suggest that for France and Italy — and in contrast to the U.K.—the sensitivity of the expected spread to the fiscal space becomes stronger when the fiscal limit is approached.

The rest of the paper is structured as follows. Section II describes the dataset, Section III presents the methodology and Section IV the results. Conclusions and policy implications follow in Section V.

II. Data: Expert Forecasts

We use Consensus Economics data to investigate the relationship between the projections of macroeconomic fundamentals and of sovereign bond spreads. CE conducts a survey—mainly based on OECD countries—among professional economists working for commercial or investment banks, government agencies, research centers and university departments. Most of the surveyed experts provide forecasts for their own country only. However, there are also a few experts working for international financial institutions or research institutes that provide forecasts for several countries simultaneously. The survey queries respondents every first week of each month about current and future developments for a number of macroeconomic and financial variables, including the yields on 10-year benchmark government bonds. The forecasts are then published early in the second week of the same month.5

Unlike other surveys, individual forecasts in CE should not suffer a bias owing to the release of strategic forecasts, as often happens for official projections released by governmental agencies (Ottaviani and Sorensen, 2006; D’Agostino and Ehrmann, 2014). In addition, evidence shows that CE forecasts are less biased and more accurate than forecasts of some international institutions.6 CE data is public, which help to prevent a participant from reproducing others’ forecasts and also limits the possibility of herding (Trueman, 1994). Moreover, forecasters are bound in their survey answers by their recommendations to their clients, and discrepancies between the survey and their private recommendation would be hard to justify (Keane and Runkle, 1990). Overall, we can reasonably argue that the CE survey data broadly reflects the spectrum of expectations of market experts.

We focus on Italy, France and the U.K., with data covering the period from January 1993 to October 2014. Apart from Germany, these are the only European Union countries for which fiscal forecasts are reported in the CE survey for a long time span. Including the U.K. has two main advantages. First, it allows us to study a non-euro area country, observing how the results for this country may differ from countries belonging to the EMU.7 Second, the CE survey includes many observations for the U.K., which allows us to significantly expand our dataset. The last CE survey in our sample (October 2014) comprises 52 forecasters in France, 42 in Italy, and 86 in the United Kingdom. However, we use a subset of these respondents. In fact, despite the gradual expansion of the dataset, fiscal forecasts have not always received the same attention from forecasters over time. Some forecasters stopped producing projections for the fiscal balance, while others that were initially included left the sample owing to closures, mergers or other reasons. Moreover, new forecasters joined the CE survey only at a later stage. Therefore, we apply a double criterion to select our sample. First, we do not consider those forecasters that have participated for fewer than 12 consecutive months in the CE survey. Second, among those forecasters, we select only those with no gaps between two consecutive forecasts that are larger than 36 months. This reduces the panel to 19 forecasters in France, 25 forecasters in Italy, and 43 in the U.K. (all forecasters are listed in the Appendix).8 Overall, our dataset is characterized by a large number of observations, i.e., around 1,500 for France, 1,200 for Italy and 2,500 for the United Kingdom. In the following, we describe the CE variables used in the paper.

A. Expected Bond Yield Spreads

In each month m of year t, the CE survey provides the forecast of the 10-year government bond yield for month m+3 and m+12. We calculate the expected spread at the single forecaster level as follows. First, we collect the individual forecast (12-month-ahead) of the 10-year government bond yield for the respective country (France, Italy, or U.K.). Second, we construct the expected spread vis-à-vis Germany—i.e., the dependent variable in our regressions—based on the average of all fixed horizon 12-month-ahead forecasts for Germany included in the CE dataset. We use the average forecast for the German yield because most forecasters do not report both the forecast for the domestic bond yield and the German benchmark one.9

This bond spread includes an exchange rate premium for the U.K. over the full sample and, for France and Italy, in the pre-EMU period. To filter out this effect, we subtract the forward swap spread from the forecast spread. The forward swap spread is the difference in the 10-year fixed interest rate from forward swap contracts denominated in the two currencies. Given that swap contracts are free from default risk, the difference purely reflects exchange rate risk.

Figure 1 summarizes the information of all spread forecasts by plotting the inter-quartile range of the forecasters’ distribution. It then reports the realized spread for the three countries. The figure highlights a rather large degree of disagreement among forecasters on the 12-month-ahead spread. The range between the upper and the lower quartile in forecasted spreads is generally no less than 50 basis points. The 1999-2007 period has been characterized as having a strong consensus among forecasters, and the forecast distribution tends to track quite closely the realized spread. However, for France and Italy, the range in forecasts increased to about 100 basis points during the recent crisis. For the U.K., despite the less pronounced increase in realized spreads after 2008, there is also a rise in the dispersion of forecasts. During the global financial crisis, realized spreads appear to be generally under-predicted, revealing that experts tended to be ex-ante overly optimistic about developments in sovereign debt markets.

B. Expected Macroeconomic Fundamentals

In addition to government bond yields, the CE survey provides projections for the fiscal balance, GDP growth and CPI inflation (among other macroeconomic variables) for the current year t and the year ahead t+1. For these variables, the monthly update implies a shrinking horizon of the forecast, e.g., as we move from January to December of each projection year t, the horizon until the target year t+1 is progressively reduced. Hence, following Dovern et al. (2012), we compute the forecast at a constant horizon of 12 months. For the fiscal balance, GDP growth and CPI inflation and in each projection month m of year t, we therefore calculate the constant-horizon (12-month-ahead) forecast as a weighted average of the same-year and the year-ahead bond rate with arithmetic weights (1m12) and m12 respectively.

Regarding the fiscal balance, CE asks respondents for a forecast in nominal terms (i.e. amounts of millions).10 In order to transform this forecast into one of the fiscal balance as a ratio to GDP, we divide the fixed-horizon one-year-ahead forecast of the nominal overall balance (surplus) in a certain month m by the GDP forecast for the same year. As the CE dataset only provides forecasts of GDP growth rates, we compute the year-ahead nominal GDP level forecast by applying the CE growth rate to the latest available estimate for the same year GDP level. The latter is taken from IMF WEO (see Appendix A for more details).11

Figure 2 displays the inter-quartile range of the 12-month-ahead fiscal balance ratio forecasts. This range is generally small and contained in an interval of just half a percentage point around the median for all three countries under examination. The range increases to 1 percentage point when there are major turning points in the realized fiscal balance. This strong consensus among forecasters continues even during the global financial crisis. This agreement does not imply accurate forecasting. On the contrary, the realized fiscal balance—i.e., the annualized quarterly fiscal balance to GDP ratio—deviates from the forecasted balance quite considerably. Figure 2 shows that, during the recent crisis, there has been a tendency to overestimate the fiscal balance for all three countries in our sample, and especially for the United Kingdom.

III. Methodology

Our empirical model follows the standard specification adopted by the literature on sovereign spreads in the EU, in which the dependent variable (the sovereign spread) is assumed to depend on a set of domestic variables, foreign variables and other controls, often including a global risk factor (Schuknecht et al., 2010; Poghosyan, 2012, Favero and Missale, 2012; and Favero, 2013). Different from the above-mentioned papers, given that our aim is to relate financial market experts’ predictions of the long-term 10-year bond spread to the balance ratio to GDP forecast while controlling for other variables, we use the 12-month ahead forecast of the sovereign spreads for France, Italy and the U.K. as dependent variables in our regressions.

In particular, we test whether expectations of the 12-month-ahead fiscal balance (surplus) for a certain forecaster i in some month m of a year t, which will be denoted by Ei,t,m [fiscal_balancet+1,m, influence the expected 12-month ahead bond spread Ei,t, m [spreadt+1,m]. In addition to the fiscal balance, other economic developments may also matter for the forecasted government bond yield. Therefore, we control the relationship between the forecasted spread and fiscal balance forecast for the expected 12-month ahead economic growth Ei,t,m [growtht+1,m], and the inflation rate Ei,t,m [inflationt+1,m].

We expect the fiscal balance to have a negative effect on the projected spread. This is justified by the fact that the perception of better future fiscal positions should ease the pressure on sovereign securities, thus resulting in lower spreads. Regarding the expected GDP growth, it can be argued that low growth tends to be associated with a flattening of the yield curve due to an expected future easing in monetary policy. Therefore, this channel would suggest a positive coefficient of growth on spreads. However, the expectation of lower future growth may, in some circumstances, imply additional fiscal adjustments in order to keep public finances under control (Alesina et al., 1992). If a permanently lower level of growth casts doubt on the long-term sustainability of public finances, higher risk premia on sovereign bonds may be the consequence. This risk factor is likely to be more important for high debt countries (Bi, 2012). This channel would therefore suggest a negative coefficient of growth on sovereign spreads. Overall, the effect of growth on spreads will depend on which of the two channels prevail: in “normal times” or for low-debt countries, we expect the first channel to prevail, while in “crisis times” or for high-debt countries, the second channel should be stronger.

In turn, if market participants expect higher inflation in the future, they may foresee a higher interest rate on government bonds. Inflation expectations push up the entire term structure, as monetary policy is expected to react by tightening. Hence, the response of sovereign interest rates and spreads to higher expected inflation should be positive.12

Besides the macroeconomic variables, global risk factors can be important in determining the expectations on spreads as most studies analyzing realized spreads suggest. The inclusion of these global risk factors is important to avoid biased estimators (e.g., Dell’Erba and Sola, 2011). In the literature, global risk has been generally proxied with a common “global” factor, corporate risk premia in the U.S., or indicators of market volatility like the VIX Index. Other papers have used the bond rate of a reference country (Blommestein et al., 2012). In line with this approach, we include the forecast of the 10-year U.S. government bond yield as a proxy of international developments and global risk.13 Descriptive statistics of all variables are shown in Table 1.

Most studies that analyze realized bond yield differentials define all explanatory variables in relative terms to a benchmark country (see, e.g., Favero and Missale, 2012; and Favero, 2013). Similarly, we take the difference between the forecast of the fiscal balance, GDP growth, and inflation of each forecaster, and the mean forecast of those variables among surveyed experts in Germany, as explanatory variables.

In sum, the baseline specification to test the relationship between projections of the overall fiscal balance and the expected sovereign spread for our panel of forecasters can be described as follows:

Ei,t,m[spreadt+1,m]=αi+β1Ei,t,m[fiscalbalancet+1,m]+β2Ei,t,m[growtht+1,m]++β3Ei,t,m[inflationt+1,m]++β4Ei,t,m[Zt+1,m]+Σn=212θnmonthn+ɛi,t,m.(1)

Ei,t,m [xt+1,m] is the 12-month-ahead forecast for fundamental x released at month m of year t by respondent i. In model (1), x is therefore the fiscal balance, inflation, and the GDP growth rate while Z represents the global factor, i.e., the interest rate on the U.S. 10-year government bond.

We also include a forecaster specific fixed-effect αi to account for unobserved heterogeneity across forecasters. Accordingly, we estimate (1) with panel fixed effects.14 Moreover, monthly dummies are used to account for regular seasonal effects in the budget forecasts (see, for example, Onorante et al., 2009). In fact, in France and Italy the regular budget cycle comprises the following steps. First, an initial draft budget is prepared (usually in late Spring). Second, the budget is submitted to the Parliament in September or October. Third, the final budget is voted on in December. The U.K. also follows a similar cycle, albeit lagged by three months given that the budget year covers the period from 1 April until 31 March. In these various stages, the government (generally, the Ministry of Finance) releases new information on the new fiscal measures planned for the next years and of its own budget forecasts. This implies that all forecasters receive the same budget news every year in the same months. Hence, we add to (1) time dummies (Σn=212θnmonthn) for each month, but January, represented by the index n (February, n = 2 ; March, n = 3 ;…; December, n = 12) to control for these effects.15 Each of these dummies equals to one whenever n = m and equals zero otherwise.

The estimation of (1) by fixed effects is not straightforward as the forecast of surplus, inflation, and real GDP growth are all potentially endogenous. The reason is that, for each forecaster, all three variables are likely to be part of the same forecasting model for the spread. To address this possible endogeneity issue, we estimate (1) using the Generalized Method of Moments (GMM). To construct strong and valid instruments, we exploit the fact that respondents update their forecast each month. Such updates reflect the ‘news’ that each forecaster receives over time and that is used for the forecast revision. We define the news as the month-to-month change (Ei,t,m[xt+1,m]–Ei,t,m–1[xt+1,m–1]), and the year-on-year change (Ei,t,m[xt+1,m]–Ei,t–1,m[xt,m]) of the first three regressors in (1), i.e., the fiscal balance, economic growth and inflation.

Using this instrument is, in our view, appropriate for the following reasons. First, the news is likely to be highly correlated with the year-ahead forecasts of the explanatory variables. In fact, when forecasters construct their macroeconomic projections in any month m, developments over the last month (the ‘news’) are taken into account in updating the forecasts. Second, the news is fully exogenous to the explanatory variables, that is, there is no feedback of the forecast series to the news during that month. Third, there is no direct impact from the forecast spread (i.e., our dependent variable) on the news.

We perform a battery of tests to check the validity and strength of the chosen instruments. The validity of the instruments is tested with the Hansen J-test. A p-value of less than 0.05 implies a rejection of the validity of instruments (at the 95 percent significance level). In addition, we compute the (HAC robust first-stage) LM test for weak instruments of Kleibergen and Paap (2006). A low statistic on this KP-test indicates weak instruments, which causes a bias in the coefficient estimates and size distortion in hypothesis tests (Stock et al., 2002). The null of weak instruments is rejected if the statistic is larger than the Stock-Yogo critical values for a single endogenous regressor.16

IV. Results

A. Baseline Results

Table 2 reports the estimates for the baseline model for the three countries’ spreads over the entire January 1993–October 2014 sample. The main result is that expectations of a higher fiscal balance significantly reduce forecasted spreads for all the three countries. This effect is particularly strong for Italian spreads: a 1 percent rise in the expected surplus to GDP ratio reduces forecasted spreads by about 38 basis points. The effect is smaller for French and British spreads (around 4 and 6 basis points, respectively), but the coefficient is still highly statistically significant.17

The growth outlooks also matter to forecasters. A higher expected economic growth leads to larger forecast spreads in all three countries. As previously discussed, in normal times, an expected improvement in the economic outlook is likely to be associated with expectations of tighter monetary policy in the medium-term and an upward shift in the term structure. For the full period sample, this effect seems to outweigh the fall in spreads triggered by fading concerns over the sustainability of public finances on the back of better growth prospects.

The effect of expected higher inflation is positive and significant, but for the U.K. only. This might be explained by a higher premium on U.K. bonds—and possibly tighter future monetary policy—when investors expect higher future inflation. For Italy and France, the coefficient on expected inflation is not significant. This may be related to the fact that during the EMU period, realized inflation and inflation expectations were relatively low in these two countries. As Dick et al. (2013) argue, in such a specific environment, it may be that investors cared relatively less about inflation compared to real growth or uncertainty.

The coefficient on the external global risk factor indicates that more risk in the U.S., as reflected in a higher expected interest rate on the 10-year U.S. government bond, raises forecasted spreads in all three countries in our sample. This global risk effect is particularly strong for Italian spreads, more moderate for French and rather weak for the British spreads. The regression diagnostics also show that the model is well identified: the KP test for weak instruments validates the use of the chosen instruments.

How do these results compare to those arising from the use of realized spreads? We substitute the forecast with the realized spread at time t as the dependent variable. This implies that the left hand side variable is identical for all forecasters, instead of being specific to each forecaster. As shown in Table 3, it emerges that the use of realized spreads leads to smaller coefficients—in absolute values—for the fiscal balance and GDP growth, compared to the regression based on spread forecasts. In particular, it turns out that the expected surplus reduces the realized French and Italian sovereign bond spread. However, the size of the effect is substantially smaller than in the case of spread forecasts. Expectations by experts of a 1 percent cut in the forecast deficit implies, on average, a fall of about 28 basis points in the Italian realized spread, which is around 10 basis points less than what we find using the spread forecasts. The effect on the French spread appears to be more than halved. For the U.K. spread, the effect falls by around 8 basis points and even turns positive as compared to the reaction of the forecasted spread.

Regarding GDP growth, in all three cases the effect is smaller compared to results in Table 2. Interestingly, for France and the U.K., the sign of the growth coefficient is now reversed: a positive growth outlook tends to reduce realized spreads. For Italy, the coefficient on growth remains positive but it is now not significant. With respect to inflation, the coefficient becomes not significant for U.K. and positive and significant, at the 5 percent level, for Italy. For France, the effect on inflation is now reversed and becomes negative. Regarding the external global-risk factor, the coefficient in Table 3 is broadly similar compared with the results in Table 2 for the projected spreads.

All in all, these results suggest that the expected evolution of fundamentals—in particular the fiscal surplus and GDP growth—is quantitatively more relevant in explaining spread projections than spread realizations. This indicates that, ex-ante, market participants seem to assign a bigger role to these factors than what is observed ex-post. This also suggests that realized spreads are likely to reflect other factors (e.g. bandwagon effect, excess speculation) beyond the expected evolution of fundamentals.18

B. Debt limit and Fiscal Space

One goal of our analysis is to deepen our understanding of how experts perceive the evolution of sovereign spreads in reaction to concerns on debt sustainability. In our baseline specification, we use the fiscal balance to investigate this reaction. Now, we treat the expectations of debt sustainability more explicitly by looking at the distance of an economy from its “fiscal limit”. The latter can be defined as the point where the government no longer has the ability and the willingness to increase its borrowing capacity (Leeper, 2013 and 2015). Exceeding the fiscal limit would result in a debt-GDP ratio that cannot be sustained without appreciable risk of default or higher inflation. We therefore incorporate a measure of “fiscal space”, defined as the distance between the time-varying fiscal limit estimated by Polito and Wickens (2015) for France, Italy and the U.K. and the debt-GDP ratio for the same countries and time periods.19 Polito and Wickens (2015) develop a measure of fiscal limit—which is an adaptation to an open economy with distortionary taxation on income from labor, capital, and consumption of the fiscal limit derived by Davig et al. (2011) and Bi (2012)—and derive time series for this limit for a set of advanced economies. A positive fiscal space would indicate that countries would have “room for maneuver” before approaching an unsustainable level of public finance. Therefore, in general, we would expect to observe a negative relationship between the fiscal space and the projected spread. In fact, experts are likely to assess a diminishing fiscal space as a risk to sustainability and thus to raise their forecast of spreads.

One important reason why experts evaluate public finances to be sustainable or not depends on monetary policy. By being part of the euro area, France and Italy do not control their own monetary policy. Consequently, for both countries, the nominal government debt that they issue becomes effectively equivalent to real debt and must be backed by real surpluses (Leeper, 2015). For this reason, as those countries’ debt levels approach their fiscal limit, their sovereign yields and spreads should tend to rise. On the other hand, the U.K. issues nominal debt and controls its own monetary policy. Nominal bonds are thus a claim to its currency in the future. In this context, an expansion of British nominal debt can be unbacked by future surpluses, as the “fiscal theory of the price level” would predict (e.g., Leeper, 1991; Sims, 1994). Thus, in the event of a very severe crisis that puts the sustainability of public finances at risk, it is likely that —faced with a choice between defaulting on government debt or violating the inflation target— the British government would try to avoid default at the cost of deviating from its inflation target. This is a critical difference between EMU countries and the U.K., which retains control of its monetary policy. All in all, we expect that—for EMU members—the sensitivity of the spread should be larger as the fiscal space gets smaller because they cannot resort to monetary policy to address perceived risks to sustainability. Therefore, we expect the reaction of the spread to the fiscal space to be non-linear for these countries: it should become stronger the closer debt gets to the limit.

We test this hypothesis by including the fiscal space variable in our baseline regression, for the three countries under examination. We first test the reaction on the full EMU sample. Second, we divide the sample in two periods, labelled as “high” and “low” fiscal space regime, that coincide with values of the fiscal space above or below the average fiscal space for each country. These two regimes allow us to test whether there are non-linear effects of the fiscal space variable on the expected spread, depending on whether a country is inside the EMU or not.

Results, shown in Table 4, indicate that the coefficient associated with the fiscal space variable is—as expected—negative for the full EMU sample, for all three countries. When we look at the two separate regimes, we find that the coefficient is broadly stable for the U.K. over the two regimes. Conversely, for France and Italy we observe a very strong decline of the coefficient in the low fiscal space regime as compared to the high fiscal space regime. This is particularly evident for Italy, for which the coefficient is not significant in the high fiscal space regime, but becomes highly significant and negative in the low fiscal space regime. These findings indicate that, for countries in the EMU, approaching the fiscal limit leads to a higher perceived risk for the sustainability of public finances, which is translated in a stronger reaction of the (projected) sovereign spread.20

C. Additional Results

Alternative indicators for global risk

Financial integration has made investors’ international portfolios co-move more strongly, such that portfolio adjustments tend to uniformly shift in one direction. In times of uncertainty, this has generally implied a flow towards safe havens. These developments have made bond yields increasingly sensitive to global conditions, rather than to country-specific risk factors (Alper and Forni, 2011; Kumar and Okimoto, 2011). So far, we have used the U.S. yield as forecasted by CE experts as an indicator of international market movements and found this variable to significantly influence spreads forecasts. The reason for adopting this measure is that alternative indicators of global risk, which have been commonly used in the literature, are realized series and not projections. At the same time, several studies argue that other financial factors may also play a role in explaining the dynamic of sovereign spreads, especially in periods of market turbulence. We thus replace the forecasted U.S. 10-year yield with these alternative indicators for global risk.

First, we include the AAA-BAA U.S. corporate bond spread, which is frequently used in the related literature (Codogno et al., 2003). This spread reflects international liquidity and credit risks. The first block in Table 5 shows that the main findings of Table 2 remain broadly unchanged. In particular, the AAA-BAA corporate spread appears to raise forecasted spreads. The impact is quantitatively stronger in Italy and the United Kingdom. In Italy, such a strong impact may be due to the high public debt and the fiscal situation in the country. In the U.K., results may reflect the strong interlinkages between U.S. and British financial markets.

The global financial crisis tied fiscal budgets to banking sector bailouts, so financial stability is likely an important indicator of risk to fiscal sustainability. Therefore, we also use the IMF’s Financial Stress Index, FSI (see Cardarelli et al., 2011). We take the difference between the FSI for the U.S., and the French, Italian and British’s FSI, as an indicator of potential financial sector problems in the three countries. The second block in Table 5 reveals that higher financial stress in the U.S.—or a lower financial stress in Italy and the U.K.—has been associated with lower expected spreads, with a particularly strong effect for Italy.

We further test the interaction between the FSI with the projected fiscal balance. In the third block in Table 5, we observe a negative and, indeed, highly significant effect for Italy and the U.K.: in those countries, higher surpluses lead to an additional reduction in spreads when financial stress is high in the United States. However, this effect appears to be quantitatively small. Finally, the interaction coefficient between the fiscal balance and the BAA-AAA U.S. corporate spreads remains negative for the U.K. and France, but turns positive for Italy. However, the coefficients on the country-specific fundamentals in this regression remain broadly consistent with the baseline findings.

All in all, we find that the impact of global-risk factors tend to be statistically significant but small. Moreover, the overall fit of the regression does not significantly improve when these factors are included. This might be due to the fact that forecasters not paying a great deal of attention to other financial elements than those in the baseline model when making their bond spread projections.

Primary balance

Next, we estimate a panel model in which we use the expected primary balance—instead of the fiscal balance—on the right-hand-side of equation (1). This variable may help to address the possible issue of reverse causality, given that interest payments depend on interest rates on sovereign securities. Therefore, a change in interest rate conditions could have an effect on the expectations of the fiscal balance for the next year. We correct the overall fiscal balance ratio by subtracting the realized interest payments to GDP ratio.

Results reported in Table 6 confirm the main findings from the baseline regression: the sign and statistical significance of the coefficients on fiscal and macro-economic fundamentals is always preserved. In particular, the future primary balance still has the expected negative sign while the coefficient for growth is positive for all three countries. The coefficient for inflation is further positive in the specific case for the United Kingdom.

Realized spreads

The realized level of spreads may also be an important determinant of the spread forecasts. Therefore, we include the lagged realized spreads (for month m-1) in the regression to test whether this variable is important in explaining spread forecasts for the m+12 horizon. We use the end-of-month value of the realized spread for month m-1, and not for the current month m because the latter is not fully known by forecasters when producing their forecast in month m.21 It is important to note that, contrary to the forecasted spread, the realized spread is the same for every single forecaster.

Table 7 shows that the lagged market spreads have significant coefficients in the regressions and increase their fit. This result shows there is a partial adaptation of the spread forecasts to the realized spread on the market. This is likely to be the result of news being incorporated gradually over time. This also shows that forecasters weigh both the role of macroeconomic fundamentals (expected fiscal balance, inflation and output growth) and past financial conditions in preparing their forecasts.

Data in levels

So far, we have expressed all variables as a difference of the CE forecast from surveyed experts for a particular country relative to the mean forecast of this variable from surveyed experts in Germany. Using all variables in difference vis-à-vis a benchmark country is common practice in the literature on realized spreads (e.g., Favero and Missale, 2012). However, we have also estimated model (1) directly on the data in levels. In other words, while we still use the expected spread against the German 10-year bond as the dependent variable, following D’Agostino and Ehrmann (2014) we now also include separately the country specific variables and the German variables—all of them in levels—on the right-hand-side of the regression equation.22

As highlighted in Table 8, the results for the country-specific fundamentals are similar to those in the baseline model. In particular, it is shown that a higher surplus significantly lowers expected spreads in the three countries of our sample. Nevertheless, it also appears that German economic developments matter for the forecast of bond spread of the other countries, as reflected in the coefficients of the German variables. This is in line with what one would expect: spreads should reflect events both from the country under analysis but also from the benchmark country. In particular, a higher surplus in Germany tends to decrease the yield for that country, thus increasing the spreads. Additionally, a higher growth in Germany is accompanied by a higher yield for that country relative to other countries’ yields, thus lowering the spread. Finally, the effect of German inflation is moderate for the U.K. and strong for Italy.23

D. Sub-Sample Analysis

The recent global financial crisis has been reflected in increased tension and turbulence in European sovereign bond markets. Bond spreads for fiscally vulnerable countries jumped to very high levels, and increased volatility of bond markets stood in sharp contrast with market developments in the first phase of the EMU (Figure 1). The higher volatility in bond spreads during the financial crisis suggests that there may have been significant changes in market participants’ expectations, possibly related to a reassessment of the sustainability of public finances as the financial crisis developed. For the fiscal balance, Figure 2 also shows that the distribution of forecasts has deviated (even over prolonged spells) from the realized balance.

Therefore, we analyze how market experts changed their spread predictions over time. As in D’Agostino and Ehrmann (2014), we look at three different subsamples for the relationship between expected spreads and expected fundamentals, covering different periods of the EMU institutional architecture. The first sample is the one ahead of the currency union, or pre-EMU period: from the start of our sample (i.e., January 1993) until December 1998. We then focus on the subperiod from the start of the EMU and circulation of the euro (January 1999) until the pre-Lehmann crisis (August 2008). Over this period, sovereign bond markets were particularly calm, as reflected in relatively flat spread over this period (Figure 1). The third subsample covers the financial crisis period, i.e., a subsample starting from the Lehman Brothers collapse (September 2008), covering the ensuing eruption of the euro area debt crisis (2011-2012), and ending in the last available observation of our sample (October 2014)..24

Table 9 presents the results for these three periods. In the first (pre-EMU) subsample, forecasters consider the projected fiscal balance, other macroeconomic fundamentals, and the global factor as relevant for the projected spread. These results are broadly in line with the full-sample estimates (Table 2). For Italy and U.K., higher expected deficits are significantly associated with higher projected sovereign spreads. For France, however, the coefficient has a positive and significant coefficient. Expectations of a higher US yield also reduce expected spreads in the three countries in the sample.

In the Italian case, the large and negative coefficient for the projected fiscal outlook appears to be associated with the convergence process to meet the Maastricht criteria. During that period, Italy undertook considerable fiscal efforts to meet the Maastricht criteria. It reduced its budget deficit from about 10 percent of GDP in 1993 to below 3 percent in 1998. This involved the introduction in 1997 of a one-off tax (so called “Euro tax”) worth around 0.6 percent of GDP. For comparison, France had a budget deficit below 6 percent of GDP already in 1993.

Regarding the initial (pre-crisis) EMU period, the role of the macroeconomic fundamentals is significantly weaker compared to the full-sample estimates. This is probably due to the fact that during this period investors did not fully price-in country-specific fundamental factors in sovereign yields, which were substantially more flat and similar across European countries and thus spreads were excessively low. In line with Ghosh et al. (2013), our results for this subsample period may suggest that, although forecasters continued to attach some relevance to macroeconomic fundamentals in their assessment of bond prices, they may have also expected that countries running into times of economic or financial distress could receive some sort of assistance (financing or transfers) from the rest of the monetary union. Thus, spreads were relatively unreactive to country-specific developments.

The period covering the financial crisis provides some insights on how expectations about future spreads have been formed during a period of severe market turbulence. The expected fiscal balance regains strong relevance during the crisis. It is significant and negative for the three countries and especially sizable for Italy. This suggests that, indeed, bond investors rediscovered a role for those fundamentals (Ghosh et al., 2013). In particular, concerns about fiscal sustainability are likely to have pushed expected spreads up during the crisis, especially in the most fiscally vulnerable countries.

In turn, expectations on the growth outlook become again important during this period. In particular, the coefficient for this variable substantially increases for the United Kingdom. Yet, it becomes negative, large and highly significant for Italy. This last result suggests that better economic prospects for Italy become relevant for the evolution of expected spreads. In the view of the experts, higher growth would support fiscal sustainability and, therefore, is accompanied by lower expected spreads in this critical phase of the business cycle (see also Alesina et al., 1992, and Cottarelli and Jaramillo, 2012). This difference, with Italy on the one side, and the U.K. on the other, might be due to a different perception about the sustainability of public finances for these countries. Markets might have perceived that sustainability was not an issue for the U.K. (and, partially, for France) during the crisis. Therefore, the link between growth and interest rates was positive in the U.K. during the crisis, as in “normal times”. Expected inflation in this more recent period also shows a high and positive coefficient in the regression for Italy. This may indicate that the prospect of a prolonged period of low inflation, which have in part come about since mid-2013, was reflected in a drop in expected sovereign spreads in this country.25

V. Conclusion

Expectations about macroeconomic and fiscal developments have triggered reactions in sovereign bond markets, especially since the start of the global financial crisis. Some argue that bond markets are rediscovering fundamentals. Others claim that, in a very uncertain economic environment, sentiments spurred by political events and contagion from vulnerable countries might have pushed financial markets into an equilibrium in which spreads are unrelated to fiscal or other macroeconomic fundamentals.

In this paper, instead of analyzing how realized spreads reacted to macroeconomic fundamentals, we test how expectations about future government bond spreads are formed by market experts. This allows us to check whether expectations about macro and fiscal fundamentals matter to market experts’ perception of future sovereign risk premiums, helping to ensure the sustainability of public finance (Bi, 2012). We employ a survey-based monthly dataset of individual forecasters (from Consensus Economics) composed of financial institutions, but also of private and public research centers. Our sample covers the period from January 1993 to October 2014, and therefore includes a substantial part of the recent financial crisis. We employ the expert-specific spread between the 12-month-ahead forecast of the 10-year government French, Italian and British bond yields relative to the average expected 10-year government bond yield for Germany. We then test whether the 12-month-ahead projections for the fiscal balance, GDP growth and CPI inflation played a significant role in explaining the projected government spreads.

Our findings suggest that forecasters do consider expected fiscal and other macroeconomic fundamentals in forming their expectations about the 12-month ahead sovereign bond spreads. A better expected fiscal outlook significantly reduces expected spreads. While true for all countries, this effect is particularly strong and robust for Italian forecast spreads. For this country, a 1 percent rise in the expected surplus-to-GDP ratio reduces the forecast spread by 38 basis points. Higher expected GDP growth tends to be positively and significantly associated with forecast spreads. Moreover, international developments, as reflected in global risk factors, are also relevant to spread forecasts. Our findings further indicate that fundamentals tend to play a significantly more important role in explaining expectations about future government spreads, compared to regressions based on realized spreads. In fact, realized market spreads are likely influenced by noisy trading and other contingent market conditions.

Additional evidence on the importance of stable fiscal policies and expected economic growth also emerges from the changing relationship between the forecast of fundamentals and of spreads. The fiscal policy outlook is more important in the period prior to the EMU and during the financial crisis. In the latter period, experts also find that, for Italy, lower expected growth tends to lead to higher forecast spreads.

Our analysis also sheds light on the role of fiscal limits and their different effects on expected sovereign spreads for countries participating in a monetary union versus a country that controls its own monetary policy. As expected, improvements in fiscal positions as reflected in an increased distance of the debt level from a country-specific “fiscal limit”, are significantly associated with a decline in expected spreads. This is particularly true for the countries participating in the EMU (France and Italy), where the fiscal limit is more binding as debt must be backed by real surpluses. For the U.K., which controls its own monetary policy, the fiscal limit does not affect differently the expected spreads depending on whether the distance between the debt level and the limit is higher or lower.

Overall, our results support the call to increase fiscal transparency and anchor fiscal expectations (Leeper, 2010). Besides significantly improving the effectiveness of fiscal policy (Leeper, 2009), anchoring expectations about the future path of fiscal policy (and other macro-fundamentals) can ease financial markets’ concerns and expectations about the long-term sustainability of public finances (Bi, 2012). A sound, stable and credible macro-fiscal framework implemented by fiscal authorities can lead to stable expectations of the evolution of sovereign bond markets, reducing the country’s expected risk premium.

Tables and Figures

Table 1.

Descriptive Statistics, January 1993–October 2014

article image
Source: Cardarelli et al. (2011), Consensus Economics, Datastream, ECB, and authors’ calculations. Notes:

Percent of forecasted GDP.

Scalar.

Table 2.

Panel Baseline Regressions, Jan. 1993–Oct. 2014a

article image
Notes:

All variables are forecasts for the year ahead and (apart from US yields) represent differences from the values for Germany; significance at *** p<0.01, ** p<0.05, * p<0.1; monthly dummies not reported.

KP LM test is the Kleibergen Paap LM test for weak instruments.

Table 3.

Panel with Realized Spreads, Jan. 1993–Oct. 2014a

article image
Notes:

All variables (apart from the spreads) are forecasts for the year ahead and (apart from US yields) represent differences from the values for Germany; significance at *** p<0.01, ** p<0.05, * p<0.1; monthly dummies not reported.

KP LM test is the Kleibergen Paap LM test for weak instruments.

Table 4.

Panel with high and low values for the Fiscal Space, EMU period (Jan. 1999–Oct. 2014)a

article image
Notes:

All variables are forecasts for the year ahead and (apart from US yields) represent differences from the values for Germany; significance at *** p<0.01, ** p<0.05, * p<0.1; monthly dummies not reported. High- and low fiscal space split the sample for those observations in which fiscal space is above or below the average sample.

KP LM test is the Kleibergen Paap LM test for weak instruments.

Table 5.

Panel including alternative Global Risk variables, Jan. 1993–Oct. 2014a

article image
Notes:

All variables are year-ahead forecasts and (apart from the financial variables) represent differences from Germany; significance at *** p<0.01, ** p<0.05, * p<0.1; monthly dummies not reported.

KP LM test is the Kleibergen Paap LM test for weak instruments.

Table 6.

Panel Using Primary Surplus, Jan. 1993–Oct. 2014a

article image
Notes:

All variables are forecasts for the year ahead and (apart from US yields) represent differences from the values for Germany; significance at *** p<0.01, ** p<0.05, * p<0.1; monthly dummies not reported.

KP LM test is the Kleibergen Paap LM test for weak instruments.

Table 7.

Panel Using Realized Spreads as Explanatory Variables, Jan. 1993–Oct. 2014a

article image
Notes:

All variables are year-ahead forecasts and (apart from the financial variables) represent differences from Germany; significance at *** p<0.01, ** p<0.05, * p<0.1; monthly dummies not reported.

KP LM test is the Kleibergen Paap LM test for weak instruments.

Table 8.

Panel Using Data in Levels, Jan. 1993–Oct. 2014a

article image
Notes:

Significance at *** p<0.01, ** p<0.05, * p<0.1; monthly dummies not reported.

KP LM test is the Kleibergen Paap LM test for weak instruments.

Table 9.

Panel with Different Period Subsamples, Jan. 1993–Oct. 2014a

article image
Notes:

All variables are forecasts for the year ahead and (apart from US yields) represent differences from the values for Germany; significance at *** p<0.01, ** p<0.05, * p<0.1; monthly dummies not reported.

KP LM test is the Kleibergen Paap LM test for weak instruments.

Figure 1.
Figure 1.
Figure 1.

Realized Values and Market Forecast (12-month ahead) of the 10-year Government Bond Spreads vs. Germany

Citation: IMF Working Papers 2016, 100; 10.5089/9781484362068.001.A001

Source: CE dataset (forecast spreads) and Datastream (realized spreads).Note: the figure shows the forecast for month m of year t+1, as expressed in month m of year t (black lines), together with realized series for month m of year t+1 (red line).
Figure 2.
Figure 2.
Figure 2.

Realized Values and Market Forecast of Fiscal Balance (as percent of GDP)

Citation: IMF Working Papers 2016, 100; 10.5089/9781484362068.001.A001

Source: CE dataset (government balance forecast) and Datastream (realized government balance).Note: the figure shows the annualized government balance forecast for month m of year t+1, as expressed in month m and year t (black lines), together with realized quarterly values for quarter q (that includes month m) of year t+1 (blue line).

Appendix A. Description of variables

article image

B. Forecasters in Italy, France and the U.K.

article image
article image

C. Calculation of the forecasted overall balance (as a ratio of GDP)

CE provides forecasts for the total deficit only in nominal values (local currency). Hence, we follow Heppke-Falk and Hüfner (2004) and Poplawski-Ribeiro and Rülke (2011) to construct a forecast measure of deficit ratio to GDP (percentage of GDP). For that, we cannot simply scale the nominal value deficit forecast by the GDP forecast, since the CE surveys for growth rates only, and not for the GDP in nominal value.

We construct a measure of the expected nominal year-ahead GDP forecast of forecaster i at month m and year t as follows. In the first step, we take a real-time measure of real GDP in levels for a particular year t. We use the real-time forecast of the same-year real GDP (in levels) coming from the most recent IMF World Economic Outlook (WEO) vintage available at any particular month m of year t. The IMF WEOs are published either in April or October, hence from May to October we use the April issue, and the October issue in the other months.

The second step is to compute the year-ahead GDP forecast in nominal value. We multiply the real-time (WEO) measure of same-year real GDP (in levels), EWEO,t [yt,m ]f, by the year-ahead market (CE) forecasts for GDP growth, Ei,t,m[growtht+1,m], and inflation, Ei,t, m[inflationt+1,m], for each forecaster i at a particular month m of year t. The expected year-ahead nominal GDP value for each country is then

Ei,t,m[yt+1,m]=EWEO,t[yt,m]×(1+Ei,t,m[growtht+1,m]+Ei,t,m[inflationt+1,m]).

The year-ahead expected overall balance for each country is then:

Ei,t,m[fiscal_balancet+1,m]=Ei,t,m[fiscalbalancet+1,mnom]Ei,t,m[yt+1,m],

where Ei,t,m[fiscalbalancet+1,mnom] is the CE forecast of the nominal overall balance by forecaster i in month m of year t for one year-ahead t+1.

References

  • Aizenman, J., M. Hutchison, and Y. Jinjarak, 2013, “What is the risk of European sovereign debt defaults? Fiscal space, CDS spreads and market pricing of risk,” Journal of International Money and Finance, Vol. 34, pp. 3759.

    • Search Google Scholar
    • Export Citation
  • Akerlof, G.A., and R.J. Shiller, 2009, “Animal Spirits: How Human Psychology Drives the Economy, and Why It Matters for Global Capitalism” (Princeton: Princeton University Press).

    • Search Google Scholar
    • Export Citation
  • Akitoby, B., and T. Stratmann, 2008, “Fiscal Policy and Financial Markets,” Economic Journal, Vol. 118, pp. 19711985.

  • Al-Eyd, A., and S.P. Berkmen, 2013, “Fragmentation and Monetary Policy in the Euro Area,” IMF Working Paper No. 13/208, October (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Alesina, A., M. de Broeck, A. Prati, and G. Tabellini, 1992, “Default risk on government debt in OECD countries,” Economic Policy, Vol. 7, No. 15, pp. 428463.

    • Search Google Scholar
    • Export Citation
  • Alper, E., and L. Forni, 2011, “Public Debt in Advanced Economies and its Spillover Effects on Long-term Yields,” IMF Working Paper No. 11/210, August (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Aristei, D., and D. Martelli, 2014, “Sovereign bond yield spreads and market sentiment and expectations: Empirical evidence from Euro area countries,” Journal of Economics and Business, Vol. 76, pp. 5584.

    • Search Google Scholar
    • Export Citation
  • Attinasi, M.G., C. Checherita, and C. Nickel, 2009, “What Explains the Surge in Euro Area Sovereign Spreads During the Financial Crisis 2007-2009?” ECB Working paper No. 1131 (Frankfurt: European Central Bank).

    • Search Google Scholar
    • Export Citation
  • Bao, T., J. Duffy, and C. Hommes, 2013, “Learning, forecasting and optimizing: An experimental study,” European Economic Review, Vol. 61(C), pp. 186204.

    • Search Google Scholar
    • Export Citation
  • Bao, T., J. Duffy, and C. Hommes, C. Hommes, J. Sonnemans, and J. Tuinstra, 2012, “Individual expectations, limited rationality and aggregate outcomes,” Journal of Economic Dynamics and Control, Vol. 36(8), pp. 11011120.

    • Search Google Scholar
    • Export Citation
  • Barberis, N., A. Shleifer, and R. Vishny, 1998, “A model of investor sentiment,” Journal of Financial Economics, Vol. 49, pp. 307343.

    • Search Google Scholar
    • Export Citation
  • Batchelor, R., 2001, “How Useful Are the Forecasts of Intergovernmental Agencies? The IMF and OECD versus the Consensus.” Applied Economics, Vol. 33, pp. 225235.

    • Search Google Scholar
    • Export Citation
  • Bellas, D., M.G. Papaioannou, and I. Petrova, 2010, “Determinants of Emerging Market Sovereign Bond Spreads: Fundamentals vs Financial Stress” IMF Working Paper No. 10/281, December (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Bi, H., 2012, “Sovereign Default Risk Premia, Fiscal Limits, and Fiscal Policy, European Economic Review, Vol. 56, No. 3, pp. 389410.

    • Search Google Scholar
    • Export Citation
  • Bi, H., E. M. Leeper, and C. Leith, 2013, “Uncertain Fiscal Consolidations,” Economic Journal, Royal Economic Society, vol. 0, pages F31F63, 02.

    • Search Google Scholar
    • Export Citation
  • Blommestein, H., S. Eijffinger, and Q. Zongxin, 2012 “Animal Spirits in the Euro Area Sovereign CDS Market,” CEPR Discussion Papers No. 9092 (London: Centre for Economic Policy Research).

    • Search Google Scholar
    • Export Citation
  • De Bondt, W.F., and R. Thaler, 1985, “Does the Stock Market Overreact?” Journal of Finance, Vol. 40, No. 3, pp. 793805.

  • Cardarelli, R., S. Elekdag, and S. Lall, 2011, “Financial stress and economic contractions,” Journal of Financial Stability, Vol. 7, No. 2, pp. 7897.

    • Search Google Scholar
    • Export Citation
  • Cheung, Y.W., and M.D. Chinn, 2001, “Currency traders and exchange rate dynamics: a survey of the US market,” Journal of International Money and Finance, Vol. 20, pp. 439471.

    • Search Google Scholar
    • Export Citation
  • Cheung, Y.W., and M.D. Chinn, and I.W. Marsh, 2000, “How do U.K.-Based Foreign Exchange Dealers Think Their Market Operates,” NBER Working Paper No. 7524 (Cambridge, MA: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation
  • Citibank, 2012, “Sovereign CDS ban already having an impact on CEEMEA,” Emerging Markets Strategy, October 18th.

  • Claessens, S., and M. Kose, 2013, “Financial Crises Explanations, Types, and Implications,” IMF Working Paper No. 13/28 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Codogno, L., C. Favero, and A. Missale, 2003, “Yield spreads on EMU government bonds,” Economic Policy, pp. 505532.

  • Cottarelli, C. and L. Jaramillo, 2012, “Walking Hand in Hand: Fiscal Policy and Growth in Advanced Economies.” IMF Working Paper No. 12/137, May (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • D’Agostino, A., and M. Ehrmann, 2014, “The Pricing of G7 Sovereign Bond Spreads–The Times, They are a-changing”, Journal of Banking & Finance, Vol. 47(C), pp. 155176.

    • Search Google Scholar
    • Export Citation
  • Davig, T., E. Leeper, and T.B. Walker, 2011, “Inflation and the fiscal limit,” European Economic Review, Vol. 55(1), pp. 3147.

  • De Grauwe, P., and J. Ji, 2013, “Self-Fulfilling Crises in the Eurozone: An Emprical Test”, Journal of International Money and Finance, Vol. 34, pp. 1536.

    • Search Google Scholar
    • Export Citation
  • De Grauwe, P., 2015, “Design failures of the Eurozone”, VOX Column, 5 September 2015.

  • Dell’Erba, S., and S. Sola, 2011, “Expected Fiscal Policy and Interest Rates in Open Economy”, IHEID Working Papers No. 07–2011 (Geneva: Graduate Institute of International and Development Studies).

    • Search Google Scholar
    • Export Citation
  • Dick, C.D., M. Schmeling, and A. Schrimpf, 2013, “Macro-Expectations, Aggregate Uncertainty, and Expected Term Premia,” European Economic Review, Vol. 58, Issue C, pp. 5880.

    • Search Google Scholar
    • Export Citation
  • Dovern, J., U. Fritsche, and J. Slacalek, 2012, “Disagreement Among Forecasters in G7 Countries,” The Review of Economics and Statistics, Vol. 94, No.4, pp. 10811096.

    • Search Google Scholar
    • Export Citation
  • Ejsing, J. and W. Lemke, 2009, “The Janus-Headed Salvation: Sovereign and Bank Credit Risk Premia During 2008-09, ECB working paper No. 1127 (Frankfurt: European Central Bank).

    • Search Google Scholar
    • Export Citation
  • Favero, C. A., 2013, “Modelling and forecasting government bond spreads in the euro area: A GVAR model,” Journal of Econometrics, Elsevier, vol. 177(2), pages 343356.

    • Search Google Scholar
    • Export Citation
  • Favero, C.A., and A. Missale, 2012, “EU Public Debt Management and Eurobonds.” EU Parliament Economic Policy Note (Brussels: European Parliament).

    • Search Google Scholar
    • Export Citation
  • Favero, C.A., and A. Missale, M. Pagano, and E. von Thadden, 2010, “How Does Liquidity Affect Government Bond Yields?” Journal of Financial and Quantitative Analysis, Vol. 45, No. 1, pp. 107134.

    • Search Google Scholar
    • Export Citation
  • Furceri, D. and A. Zdzienicka, 2013, “The Euro Area Crisis: Need for a Supranational Fiscal Risk Sharing Mechanism?” IMF Working Paper No. 13/198 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Gerlach, S., S. Alexander, and G.B. Guntram, 2010, “Banking and Sovereign Risk in the Euro area.” Deutsche Bundesbank Discussion Paper No. 09/2010 (Frankfurt: Deutsche Bundesbank).

    • Search Google Scholar
    • Export Citation
  • Ghosh, A., J. Ostry, and M.S. Qureshi, 2013, “Fiscal Space and Sovereign Risk Pricing in a Currency Union,” Journal of International Money and Finance, Vol. 34, pp. 131163.

    • Search Google Scholar
    • Export Citation
  • Heppke-Falk, K., and F. Hüfner, 2004, “Expected Budget Deficits and Interest Rate Swap Spreads: Evidence for France, Germany and Italy.” Deutsche Bundesbank Discussion Paper No. 40/2004 (Frankfurt: Deutsche Bundesbank).

    • Search Google Scholar
    • Export Citation
  • Hirshleifer, D., 2001, “Investor Psychology and Asset Pricing,” Journal of Finance, Vol. 56, No. 4, pp. 15331597

  • Hommes, C.H., 2014, “Behaviorally Rational Expectations and Almost Self-Fulfilling Equilibria,” Review of Behavioral Economics, Vol. 1(1-2), pp. 7597.

    • Search Google Scholar
    • Export Citation
  • Hommes, C., J. Sonnemans, J. Tuinstra, and H. van de Velden, 2005. “A strategy experiment in dynamic asset pricing,” Journal of Economic Dynamics and Control, Vol. 29(4), pp. 823843.

    • Search Google Scholar
    • Export Citation
  • Keane, M.P., and D.E. Runkle, 1990, “Testing the Rationality of Price Forecasts: New Evidence from Panel Data,” American Economic Review, Vol. 80, pp. 714735.

    • Search Google Scholar
    • Export Citation
  • Kleibergen, F., and R. Paap, 2006, “Generalized Reduced Rank Tests Using the Singular Value Decomposition,” Journal of Econometrics, Vol. 133, No. 1, pp. 97126.

    • Search Google Scholar
    • Export Citation
  • Kumar, M. and T. Okimoto, 2011, “Dynamics of International Integration of Government Securities’ Markets,” Journal of Banking and Finance, Vol. 35, No. 1, pp. 142154.

    • Search Google Scholar
    • Export Citation
  • Laubach, T., 2009, “New Evidence on the Interest Rate Effects of Budget Deficits and Debt,” Journal of the European Economic Association, Vol. 7, No. 4, pp. 858885.

    • Search Google Scholar
    • Export Citation
  • Leeper, E.M., 1991, “Equilibria Under ‘Active’ and ‘Passive’ Monetary and Fiscal Policies,” Journal of Monetary Economics, Vol. 85, pp. 473492.

    • Search Google Scholar
    • Export Citation
  • Leeper, E.M., 2009, “Anchoring Fiscal Expectations,” NBER Working Paper No. 15269, August (Cambridge, MA: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation
  • Leeper, E.M., 2010, “Monetary Science, Fiscal Alchemy,” Jackson Hole Economic Symposium Paper, October (Kansas City: Federal Reserve Bank of Kansas City).

    • Search Google Scholar
    • Export Citation
  • Leeper, E.M., 2013, “Fiscal Limits and Monetary Policy,” NBER Working Paper No. 18877, March (Cambridge, MA: National Bureau of Economic Research).