A. Introduction

1. The debt restructuring in Greece in February-March 2012 effectively extended senior creditor status to Eurosystem bond purchases. By mid-February 2012, the ECB and Eurosystem central banks (henceforth ECB) swapped their Greek bonds acquired under the Securities Market Program (SMP) for new bonds of identical structure and nominal value, with different serial numbers. This provided the ECB an exemption from the subsequent private sector debt swap under so-called private sector involvement (PSI), effected through retroactive collective action clauses.² By swapping its debt purchases in advance of the PSI announcement, the ECB effectively received preferential (i.e., senior) creditor status on its Greek bond holdings.³

2. According to the ECB, this exemption from PSI was “special” because its bond market interventions were undertaken solely for monetary policy purposes. Treating Greece as a special case may imply that the subordination of private debt holders will not be repeated in other contexts. At the same time, the ECB has cautioned against incurring losses on the SMP, which could–in an extreme case–require recapitalization of the central bank and result in reduced financial independence.

3. Currently, a key question is whether the SMP has become less effective after the Greek PSI exemption–or even earlier–due to subordination risk. This treatment of the ECB may effectively have reshaped the seniority structure of all official holdings of sovereign bonds. From a market risk-return perspective, this implied a possible mispricing of many euro area bond markets leaving future SMP beneficiaries subject to rating downgrades as ECB interventions reduce the private investor base and increase losses in the event of restructuring. Anecdotal market evidence indeed confirms that the impact of SMP purchases has become controversial, although this may already have been priced in prior to the Greek debt exchange in February 2012. Indeed, after the euro area summit of heads of state and government on July 21, 2011, when PSI for Greece was first announced, subordination effects were already under consideration.⁴

4. At the same time, while the SMP helped to temporarily reduce stress in government bond markets, due to the limited scale/time horizon of its effective use, it did not appear fully effective (see Figure 1). Especially after its launch in May-June 2010 and after reactivation in August 2011, interventions were sizeable and helped stem the rise in periphery yields and the escalation of bond market volatility. Purchases amounted to €36bn and €21bn in the first two months of the program but dropped to less than a few billion afterwards. However, as purchases were scaled back, volatility and broad euro area financial market risks increased again and SMP interventions were scaled up again in November-December 2010. Similarly, as sovereign market stresses increased again over the summer of 2011, the reactivation of the SMP in August and September 2011 with purchases of €51bn and €37bn led to a reduction in broad market stresses. However, as stress in periphery debt markets re-emerged, SMP purchases increased again to €40bn in November 2011 before the ECB adopted other measures to help (periphery) banks in need of funding.

Sources: Bloomberg; and IMF staff calculations.Notes: Bond yields refer to 10-year benchmark yields. Weighted series take government debt as weights. Bond volatility follows a GARCH(1,1). Scaled EUR risk is the principal component score of 13 euro area interbank market spreads, corporate CDS spreads, euro area equity risk premium, SovX and EMBIG CDS, and euro exchange rate implied volatility.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Sources: Bloomberg; and IMF staff calculations.Notes: Bond yields refer to 10-year benchmark yields. Weighted series take government debt as weights. Bond volatility follows a GARCH(1,1). Scaled EUR risk is the principal component score of 13 euro area interbank market spreads, corporate CDS spreads, euro area equity risk premium, SovX and EMBIG CDS, and euro exchange rate implied volatility.

• Download Figure
• Download figure as PowerPoint slide

Sources: Bloomberg; and IMF staff calculations.Notes: Bond yields refer to 10-year benchmark yields. Weighted series take government debt as weights. Bond volatility follows a GARCH(1,1). Scaled EUR risk is the principal component score of 13 euro area interbank market spreads, corporate CDS spreads, euro area equity risk premium, SovX and EMBIG CDS, and euro exchange rate implied volatility.

• Download Figure
• Download figure as PowerPoint slide

5. This paper looks at various ways to quantify the extent of subordination arising from ECB debt purchases. It first looks at illustrative empirical evidence, aimed at documenting developments in government bond prices and CDS risk premia around/after the ECB debt swap. Next, it looks at theoretical models to quantify and illustrate the potential effect of subordination on bond prices and CDSs. Finally, the paper offers some tentative policy conclusions. The main finding is that the impact of ECB seniority is primarily related to perceived probability of default and the proportion of outstanding debt already in the hands of the central bank. Moreover, credibility of SMP interventions matters.

B. Empirical Evidence on ECB Subordination Risk

6. Insufficient data make it difficult to quantify the subordination risk from ECB debt purchases directly. ECB debt purchases under the SMP were largely put on hold after the PSI exemption. Only one SMP intervention took place afterwards, in the week of March 5–9, 2012 for an amount of €27 million, the fifth-smallest intervention since the start of SMP purchases in May 2010, based on weekly Eurosystem financial statements. Hence, there is insufficient quantitative evidence to establish empirically the relation between subordination risk from ECB debt purchases and sovereign yields.

7. However, an event study analysis of ECB news on its senior status can provide some gauge of subordination risk.⁵ In this context, innovations in bond yields and CDS premia in the days surrounding the July 22, 2011 ECB statement (see footnote 4) and the ECB debt swap on February 16, 2012 are examined. Following MacKinlay (1997), standardized cumulative abnormal yields and CDS spreads are estimated (in first differences). The underlying model is ΔY_i = a_i+b_i ΔY_LCH + c_iΔSovX and ΔCDS = d_i+e_iΔSovX, where Y_i denotes the government bond yield in country i, CDS is the sovereign credit default swap spread, Y_LCH is the LCH Clearnet benchmark 10-year yield on AAA countries and SovX is the SovX Western Europe CDS index. Standardization makes it easier to compare cumulative abnormal yields/spreads. The models are estimated using 70 daily observations with data up until one month ahead of the event, in order to avoid coefficient bias due to the events.

8. The results show little impact of the ECB debt swap on periphery yields, but the initial PSI announcement and the ECB’s non-participation did have substantial short-term negative effects. The market effect of the ECB, as a large creditor, shifting to a preferred debtor status, did not seem significant when evaluated around the time of the debt swap announcement (Figure 2). However, this appears to reflect the fact that this may have been anticipated and priced in by the market already. Indeed, longer-dated periphery bond yields and CDS default risk premia seem to have risen unexpectedly following President

Event Study Analysis of ECB Subordination on Bond Yields and CDS Spreads

Citation: IMF Staff Country Reports 2012, 182; 10.5089/9781475505788.002.A004

Sources: Bloomberg; and IMF staff calculations.Note: Periphery is an aggregation of Spain, Italy, and Portugal. Scales are standard normal, i.e., values exceeding +/-1.96 are statistically significant at the 5 percent level.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Event Study Analysis of ECB Subordination on Bond Yields and CDS Spreads

Citation: IMF Staff Country Reports 2012, 182; 10.5089/9781475505788.002.A004

Sources: Bloomberg; and IMF staff calculations.Note: Periphery is an aggregation of Spain, Italy, and Portugal. Scales are standard normal, i.e., values exceeding +/-1.96 are statistically significant at the 5 percent level.

• Download Figure
• Download figure as PowerPoint slide

Event Study Analysis of ECB Subordination on Bond Yields and CDS Spreads

Citation: IMF Staff Country Reports 2012, 182; 10.5089/9781475505788.002.A004

Sources: Bloomberg; and IMF staff calculations.Note: Periphery is an aggregation of Spain, Italy, and Portugal. Scales are standard normal, i.e., values exceeding +/-1.96 are statistically significant at the 5 percent level.

• Download Figure
• Download figure as PowerPoint slide

Trichet’s public statement that the ECB would not be participating in the voluntary Greek PSI on July 22, 2011 – one day after an important euro area summit had agreed on modalities of additional support for Greece (including PSI) and on greater flexibility of EFSF loans to the other program countries. In the days ahead of the summit, bond yields and CDS default risk premia had come down substantially, but this was reversed after the ECB’s statement – although it cannot be excluded that markets may have been disappointed by some other summit-related news (for instance, the realization that size of the EFSF was not increased and large implementation risks).

C. Theoretical Approaches to Quantifying Subordination Risks

9. It is possible to incorporate the subordination effect in a closed-form bond price model or in a reduced form CDS model. In both models, the subordination effect depends on three factors: probability of default, loss given default (one minus recovery rate), and the share of ECB bond holdings.

Sovereign bonds–closed form model

10. ECB seniority matters when the recovery rate is not close to zero or to 100 percent. For example, suppose a country’s debt trades at 60 cents to par before establishment of the ECB’s senior creditor status (Figure 3). This could reflect market estimates of 40 percent default probability with 100 percent loss given default (i.e., zero recovery value) and 60 percent (non-default) probability of full repayment. In this case (case I), seniority does not matter because upon default, no single bond holder is repaid. However, a pre-announcement price of 60 cents is also consistent with another case: 80 percent default probability with 50 percent recovery (case II). In this case, seniority matters. For example, if the ECB’s share is 50 percent, the ECB can recover all its claims at face value because the country will repay first 50 percent of the total debt to senior creditors. For the ECB, the ex post (shadow) price of the bond becomes the full value. For private bond holders, the defaulted country will not have anything left to repay after the repayment to the ECB. The ex post market price of bond reflect this ex post zero recovery and will be traded at 20 cents since there is still a 20 percent probability of non-default with full repayment of 100. The annex develops this model more formally.

Two Cases for Original Price q=60 (ECB Share=50%)

Citation: IMF Staff Country Reports 2012, 182; 10.5089/9781475505788.002.A004

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Two Cases for Original Price q=60 (ECB Share=50%)

Citation: IMF Staff Country Reports 2012, 182; 10.5089/9781475505788.002.A004

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Two Cases for Original Price q=60 (ECB Share=50%)

Citation: IMF Staff Country Reports 2012, 182; 10.5089/9781475505788.002.A004

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

11. The net impact of subordination thus depends on three major factors:

• Probability of default (PD). This increases linearly with the subordination effect, i.e., the difference between the original price and ex post price (after subordination). Figure 4 (top panel) plots this effect over various default probabilities for a range of ECB debt market shares. Evidently, the larger the ECB market share, b, the lower the market price ex post for any given PD.

• Loss given default (LGD). If large, then having the senior status does not translate into a large advantage. Also, for a small loss given default, senior status is not valued much since even junior creditors could recover a large portion of the face value. The overall effect—the difference between the original price and ex post price—is thus not monotonic. Figure 4 (middle panel) illustrates this effect for an 80 percent default probability and various ECB market shares over various LGD values.

• The ECB share.⁶ As long as the SMP’s share is small relative to the original recovery rate, even with the ECB expecting full repayment, the loss on private sector holdings will be limited. Therefore, the difference between the original price and the ex post price will be small. In contrast, when the ECB holds a large share relative to the recovery rate, the ECB’s senior status will lower the ex post recovery rate for the private sector dramatically. Figure 4 (lower panel) shows that this effect is monotonic but not linear. In particular, when the ECB’s share is larger than the original recovery rate, there is a kink, above which private bond holders will not receive anything ex post in bad states of nature. Note that what is important in pricing the bonds is the expectation of the ECB’s share including future SMP. Increases in the expected ECB share (or related uncertainty) can create further negative effects.

Bond Prices and SMP Determinants

Citation: IMF Staff Country Reports 2012, 182; 10.5089/9781475505788.002.A004

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Bond Prices and SMP Determinants

Citation: IMF Staff Country Reports 2012, 182; 10.5089/9781475505788.002.A004

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Bond Prices and SMP Determinants

Citation: IMF Staff Country Reports 2012, 182; 10.5089/9781475505788.002.A004

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

12. SMP purchases thus have two main effects on bond prices and yields:

• Raising subordination risk. The rise in yields for distressed sovereigns at the time of announcement of a debt swap (in favor of the ECB) reflects a net expected transfer of value from private bond holders to the ECB. This subordination will also undermine the effects of any future SMP purchases because sovereigns will face higher issuance cost on any bonds not purchased by the ECB.

• Improving liquidity. To the extent that the SMP improves liquidity conditions, it would reduce the probability of default and hence increase the value of residual bonds. This happens when the ECB provides vital liquidity to the market, and ultimately to peripheral governments. Investors would be reassured that the ECB is willing to smooth out temporary liquidity shocks. Also, a lower interest rate could decrease the default probability by improving debt sustainability. This beneficial effect would be enhanced by a credible and clearly communicated SMP intervention strategy.

Therefore, the net impact of the subordination versus the liquidity support effect on bond prices is ex ante ambiguous.

CDS–reduced form model

13. Along the same lines, CDS pricing implicitly reflects liquidity and seniority effects. CDS and bonds should be perfectly cointegrated, as they are assets with exactly the same cash flow and thus the same price.⁷ At the same time, the CDS model allows illustrating the role of SMP interventions somewhat differently and illustrates the role of credibility in SMP interventions (although this is also implicit in bond prices). More specifically:

• Liquidity effect. Starting from a standard CDS pricing formula, SMP interventions – by lowering sovereign yields, assuming the intervention is credible and sustained (see below) – reduce the cumulative probability of default (PD), which has a non-linear (but less than proportional) impact on CDS spreads:

  $CDS=LGDx(1-{(1-PD)}^{1/n})$

  where n denotes the number of periods (years). In the subsequent analysis, we abstract from real world versus risk-neutral PDs, which is analyzed elsewhere.⁸

• Subordination risk. Due to perceived senior creditor status, SMP purchases may increase private sector loss given default (LGD) and possibly offset the lower probability of default.

14. LGD for private bondholders increases with debt restructuring needs and the size of ECB holdings. Intuitively, in case of a debt restructuring, the LGD will depend on how much debt/GDP reduction is needed (ΔD/GDP) and on the participation rate. Using this fact, official sector holdings have a non-linear impact on expected private sector loss given default/haircut.⁹ If the ECB is expected to be exempt from PSI, as in Greece, a higher ECB share of debt holdings, b, will increase the private sector haircut or LGD.

Combining these elements gives:

$LGD=(\mathrm{\Delta }D/GDP)/(1-b)$

This makes it clear that the negative effect of SMP purchases rises progressively with an increase in debt as it entails larger haircuts (see top panel of Figure 5).

Role of SMP Purchases in a CDS Model

Citation: IMF Staff Country Reports 2012, 182; 10.5089/9781475505788.002.A004

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Role of SMP Purchases in a CDS Model

Citation: IMF Staff Country Reports 2012, 182; 10.5089/9781475505788.002.A004

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Role of SMP Purchases in a CDS Model

Citation: IMF Staff Country Reports 2012, 182; 10.5089/9781475505788.002.A004

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Substitution in the previous equation yields

$CDS=(\mathrm{\Delta }D/GDP)/(1-b)x(1-{(1-PD)}^{1/n})$

This shows that an increase in the share of SMP purchases b increases CDS spreads but more strongly so when CDS prices are already high (i.e, when PD is more elevated) and to a limited extent when CDS premia are low, i.e., only when default or restructuring fears come into play (see lower panel of Figure 5). When putting this in a dynamic context (although not explicitly done in this model), the beneficial effect of SMP interventions can be shown to hinge on the ECB’s credibility of its SMP intervention strategy: subordination may lead to self-fulfilling default dynamics (upward shift of the PD line), if, similar to models of currency crises, the central bank’s purchases are not able to offset the increased private sector loss given default owing to the ECB subordination effect. This may occur when the ECB is not able to lower spreads (or yields) which would otherwise help improve debt sustainability and contribute to lowering the PD (or at least keep the PD unchanged).

D. Conclusions

15. This SIP illustrates that the SMP may have a subordination effect, but this effect will be important only at the margin-as was the case in Greece- and depends on the program’s credibility. The rise in yields for the most distressed sovereigns at the time of announcement of the ECB’s exemption from the Greek PSI reflected a net expected transfer of value from private sovereign bond holders to the ECB. This de facto subordination may undermine future SMP interventions because sovereigns may face higher issuance cost on any bonds not purchased by the ECB. As shown theoretically, this subordination effect depends on three factors: probability of default, loss given default and the share of ECB bond holdings. At low default probability levels, when LGD is relatively low and the debt market share of SMP is not too high, subordination risk plays a limited role. This is also borne out by the analysis in a CDS model, where further SMP interventions have a negative effect only when adjustment needs are very high or CDS spreads are already extremely high. The latter also helps to illustrate the importance of the ECB’s credibility on the SMP: if it is low, SMP interventions may be unable to stop self-fulfilling debt default dynamics. This may occur when the ECB is not able to reduce sovereign spreads or yields, which would otherwise support debt sustainability and be conducive to lowering the PD.

16. Should anything be done to accommodate market fears about subordination by SMP purchases? SMP subordination currently does not seem to play a large role in pricing and markets. However, in principle it is possible to attenuate market fears about SMP subordination (beyond what is captured in our stylized models) by transferring some of the benefits for the ECB back to private sector bond holders.