One of the recurring questions ever since the ECB's foray into sterilized monetization in May 2010, when it started buying up Greek bonds, is what is the total loss that the Eurosystem (with an €81.5 billion in capital) has accumulated over the life of the Securities Market Programme (SMP). And while the answer "lots" is often times sufficient, and in reality would be "many, many lots" if it weren't for the fact that for the ECB cost basis does not matter courtesy of several money printers, thus allowing it become a marginal buyer just when everyone else is a seller, here is a more granular analysis by Barclays' Laurent Fransolet who calculates total paper losses of about €30 billion to date for the €211 billion of securities purchases since program inception. In other words, over the past year and a half the ECB has lost 14% on its monetization portfolio: which while not horrible, is not very good for an entity that can simply bid up any security to a price of its choosing - remember: there is no cost basis, and no opportunity cost to investment decisions, when you happen to print the instrument of purchase. As such, in theory, the ECB could easily have a 100 limit bid on all PIIGS bonds and it would experience zero losses. Furthermore, with BTP yields once again north of 7.1%, we are quite confident that the €90 or so billion in Italian bonds held by the ECB will i) increase substantially ii) will record many more losses, especially once 2012 unrolls completely and the ECB has to deal with not only portfolio liquidations of Italian bonds but primary issuance.
First, here is Barcalys' estimate of how many various bonds the ECB owns by country of origin:
The ECB only gives a total weekly cash purchase number (and weekly redemptions), but nothing else. In particular, it does not publish any country or maturity split of what it has bought. Although the amounts bought in each bond markets have varied every week, on the basis of the pressure in these markets, we make the assumption that the ECB has been buying bonds broadly in proportion to the size of the underlying bond markets. Such a proportionality rule is likely since: 1) anecdotal evidence suggests that it has also been prevalent in the Covered Bonds Purchase Programme (CBPP); and 2) a proportionality rule also applies to the share of buying by the various national central banks and the ECB itself. So in the first phase of the SMP (ie, SMP1), based on our proportionality assumption, the ECB would have bought roughly 50%, 25% and 25% respectively of Greek, Irish, and Portuguese bonds. On the same bases, in the second phase of the SMP (ie, SMP2), the ECB would have bought roughly 66% of Italian bonds and 33% of Spanish bonds (and a minimal amount of Portuguese and Irish bonds). While we see the advantages of proportionality (it is clear and explainable), in our view, more flexibility, and in particular, greater focus on the problem areas, is something the ECB should consider for both SMP2 and CBPP2.
The ECB bought a total (in cash) of €218bn between May 2010 and end of December 2011, although about €7bn of bonds/bills have matured since the start of the SMP (€5bn in Greece and €2bn in Portugal, based on the ECB weekly announcements), putting the current total ECB holdings at about €211bn (and probably close to €220-225bn in nominal terms). We would estimate, based on the weekly ECB purchases officially reported, the proportionality rule above and anecdotal evidence (the week-to-week pressure on the various markets) that of the current €211bn SMP holdings (in cash terms at time of purchases), Greek bonds account for around €36bn, Portuguese bonds €20bn, Irish bonds €19bn, Spanish bonds €46bn, and Italian bonds €90bn. Obviously, these numbers should be taken as rough guides rather than more definitive estimates.
The Greek holdings are likely substantial in the context of the upcoming PSI: €36bn at purchase price and probably around €40bn in nominal terms. Still, the target amount of €100bn of nominal debt reduction in the Greek PSI (assuming a 50% nominal haircut on €200bn of private debt holdings) implies that the ECB would not participate. Indeed, the ECB was very clear that it would not participate in a PSI when the idea was initially floated. However, we note that it has been fairly quiet on the topic since then, and other (presumably large) investors have been calling for the ECB to participate in the PSI. In our view, the ECB holdings could eventually be included in the PSI, while the bonds are still on the Eurosystem balance sheet or after they have been transferred to another entity (eg EFSF/ESM).
So where are the losses concentrated? For now Greece. Soon - everywhere else.
Now that we have estimated a country split for each weekly purchase released by the ECB, we can estimate how much the current mark-to-market on these purchases is, using total return bond indices for each country (thus taking into account price moves as well as coupons, see Figure 2). We have assumed that the average maturity of the bonds bought was around 5 years in most markets (maybe a bit longer in Ireland, due to the structure of the market), as the ECB does not seem to have bought bonds longer than 10 years, or very short-dated bonds.
Figure 3 shows our estimate of the mark-to-market (MTM) for the Eurosystem’s purchases between May 2010 and end 2011 – again, we would stress that these are only broad estimates, dependent on a number of assumptions. Roughly, we estimate the ECB has a ‘paper loss’ of around €30bn, of which around 80% would likely be due to losses on the purchases made in May and June 2010, at the inception of the programme. We would estimate that the MTM losses on Greek bonds are the biggest, at around €20-25bn (more than a 60% MTM loss, after coupon payments), with Portuguese losses at probably around €5bn (about a 23% MTM loss), Irish ones at around €1bn, Italian ones at around €2-3bn (- 2%, or on average around a 50bp yield loss), and a small gain on Spanish bonds. Thus, our analysis suggests most of the MTM losses have been on the bonds bought in the initial stages of SMP1, and in particular on Greek bonds. In contrast, we estimate that the MTM losses since the start of SMP2 could have been relatively small: probably less than€2bn. This is remarkable, since the total cash amount bought under the SMP2 is so much larger than under SMP1 (€140bn vs €77bn). In our view, at the margin, such a relatively benign mark-to-market on Italian and Spanish SMP purchases facilitates ECB SMP activity to support the market going forward.
Of course the ECB is also incurring the cost of sterilizing these purchases on a weekly basis: according to our calculations (on the basis of the weekly drain auction results), the total cost up until now, over the life of SMP1 and SMP2, has amounted to about €1.1bn.
A quick reminder that the ECB is joint and several: i.e., losses are socialized. Germany will be delighted.
Note that, according to the Eurosystem accounting rules, the gains and losses on the SMP purchases are mutualised across the ECB and the national central banks (NCBs), in proportion to their share in the capital of the ECB (which itself is defined on the basis of relative GDP and population) – this is in contrast with the CBPP, the P/L of which resides in each national central bank and the ECB. Note as well that the SMP purchases are not marked to market as such on the Eurosystem balance sheet, but held at cost. The capital and reserves of the Eurosystem amounted to €81.5bn and the revaluation accounts to €394bn as at the end of 2011 (up +€3.3bn and +€63bn, respectively, versus end 2010, mostly due to gold holdings).
Lastly, on the question of who was selling, we find that the bulk of Italian purchases are likely due to portfolio liquidation:
The large buying of Italian and Spanish bonds in the SMP2 raises the question: who has been selling to the ECB? There are two potential sellers: existing investors (typically non domestic ones) and the Italian and Spanish treasuries, as they increase the stock of bonds in the course of their regular auctions. To estimate the relative proportion between the two, one has to compare the ECB SMP buying with the net issuance by the two treasuries (ie, the changes in the stock of debt). This is shown in Figure 4: if the SMP buying is higher than the net issuance, then the ECB is absorbing the net amount liquidated by investors. Of course, there might also be some switches between types of investors (eg, international and domestic investors), but only the residual would have to be taken by the ECB (ie, the liquidations). Conversely, if ECB SMP buying is lower than the net issuance, investors are taking down the new supply in some way.
Figure 4 illustrates that in both August and September (calendar months) there was limited new net issuance (close to zero in fact) and most of the ECB buying (€55bn and €30bn respectively in August and September) was on the back of large portfolio liquidations (€45bn and €39bn respectively, for Italy and Spain). In October and November, new net supply was much heavier (€29bn combined, mostly coming from Italy), while SMP buying was lower (€17bn and €29bn respectively). Hence, during that period, the bulk of the ECB buying was actually on the back of new issuance, while portfolio liquidations were (surprisingly) more modest (29+17-29=€17bn for October and November, almost equally split in time). In December, net issuance was remarkably heavy (€11bn in Italy, €12bn in Spain), while the SMP was barely active (less than €5bn), hence investors absorbed the supply (at the price of still very high yields), likely helped to some extent by some short covering ahead of year-end and the early December summit, as well as buying related to the ECB 3yr LTRO facility conducted at the very end of December. but announced on December 8. In summary, looking at Figure 4 shows the heavy liquidations of August and September have faded and that more recently, the ECB SMP buying has been relatively limited compared with the issuance.
Interestingly, the analysis also shows that in Italy, investor liquidations dominated (roughly 2/3 vs 1/3 of new issuance). In contrast, in Spain, the split between liquidations and net issuance was closer to 50/50. In our view, this could be because a lot of investors were large outright and relative holders of Italy and liquidated positions once the spotlight turned to Italy, while a lot of international investors have likely gradually been reducing their Spanish holdings over the past two years.
Barclays conclusion is that the losses will hardly be an impediment to conducting ongoing sterilized bond monetizaton, and will allow it to continue expanding its balance sheet outright ("our base case scenario would be that it will probably remain less active than it has been in Q3 and the beginning of Q4, and given its track record, the ECB will likely be willing still to intervene, with some confidence."), even as it proceeds with more bank guarantee programs such as the LTRO, which are an indirect monetization, whereby the ECB keeps banks on life support merely so these can go ahead and purchase bonds themselves, which for all intents and purposes is even more monetization only one degree removed. In the meantime, look for the ECB balance sheet to keep expanding, pushing the EURUSD ever lower, until it becomes a real impediment to doing business. As such, while the ECB may or may not do outright QE, and we have been qiote skeptical the ECB would proceed with outright monetization (see: Ze Germans), what it is doing is, semantics aside, monetization as is. So the ball is now really in the Fed's court. And strange days like today (NFP data) which serve the purposes of briefly boosting Obama's rating, but delay QE only force the market to do precisely what it has done: i.e., turn red. Because as pointed out in December, the only thing that matters for the market is the amount of gross liquidity available and injected at the margin. Everything else is legacy noise.