So I noticed I had made some embarrassing mistakes on my post about recent developments in the Spanish Sovereign debt market. However, while looking for an alternative illustration of Spanish woes, and blatantly engaging in confirmation bias, I came across a disturbing observation, which if it is true, has some very sad implications for the state of financial analysis. This post is written in a particularly familiar tone. It seemed the most appropriate tone to take the reader around the steps it took me to arrive where I now stand.
But lets start with my mistakes before considering those of others.
Unfortunately, I have found some errors in the aforementioned post.
1. First, the comment I had about the failure of bill issuance by Spain since March 20, was wrong. I presented the two tables below, as evidence of the issuance difficulties encountered by Spain, arguing that the difference between “solicited” and “adjudicated” was evidence of failure and financing difficulties. (On Wednesday, March 20 2012: € 3,547.70 Mn and € 1,410.50 Mn “adjudicated”, compared to € 7,575.08 Mn and € 4,123.08 Mn “solicited”, in 12 month and 18 month bills respectively).
Similarly, the table below made the same argument for the bill auction, which took place on March 27, 2012.
My error was to misunderstand the meaning of “solicited”. In the framework of analysis I was using, I construed the market as one for liquidity, rather than a market for debt. As a result, demand was misconstrued as being constituted by the public sector. Incorrectly assuming that any “solicitations” would be carried out by the liquidity demanding public sector, it was easy to misunderstand the tables above.
However the market described in the tables above is characterised as one for bonds, and so the public sector provides the supply by conducting debt auctions. Any “solicitations” are carried out by the financial agents demanding the debt securities. As a result, in and of themselves, the table above do not provide any evidence of financing difficulties experienced by Spain. Actually, given that all bid-to-cover ratios (“ratio de cobertura”) are above 2, the auctions could be characterised as having considerable demand.
The problem is that these tables offer a rather limited description of the result of the debt auctions. Instead of describing the financing needs of the auctioning authority (the government in this case), they begin by describing the demand for the auctions given their coupon and their remaining maturity. Because the government gets to set the allotted amount, the bid-to-cover ratios are generally biased to making the government looking good.
Instead, a better measure of the success of a sovereign debt auction would be one of either:
- For a newly issued security, the best is to compare coupons over maturities, with previously issued lines of bonds.
- For reissues of the same security, at a previously determined coupon the best is to compare public sector financing needs with allotted amounts.
2. In point three (“3.”) of the post I present a figure charting the movements of deposits of and with the Bank of Spain. In the original post I referred to these as describing “recourse to the Marginal lending and to the Deposit facilities “. This was inaccurate. The figures represent total borrowings of and from the Bank of Spain, rather than only those with overnight maturities.
These mistakes do not necessarily impact on the overall conclusion. However the first correction does rise questions regarding the influence of primary market developments in determining the yield movements. This takes me to the next section.
About those CDS figures…
While blatantly engaging in confirmation bias and looking for further evidence of Spanish trouble, I found this interesting chart, kindly provided by FT Alphaville, accompanies by the comment “Looking back over the last few months, credit default swap markets seemed of the opinion that Italy was the greater risk. Not so anymore:”
Now, finding this image made me happy. It confirmed that Spain was considered riskier that Italy and that a change had occurred recently. However, after looking around I could not find any reason for why the crossings had occurred in August 2011 and February 2012. I thought I could argue that Italian CDSs rose in August in the run up to the negotiation of the second bailout for Greece and the fiscal compact. But then, why did they not fall following the approval of both aggreement?
This nagging question itched at me enough that I decided to check the actual data. So I went on Bloomberg and decided to compare 5 year Italian CDSs with their Spanish counterpart. Guess what: I got a pretty similar picture (the first of the two below)! But this is when it gets tricky. I’m geeky enough that I decided to check the situation for the last 5 years (second picture immediately below) and it all breaks apart. There’s little apparent connection between the two figures, although they represent the same variables, simply over different periods of time (the first one covers movements in the last 6 months and the second one covers movements over the last 5 years (arguably actually only 3 and a half)).
To understand this mess, it is important to understand what is being measured and how it is being measured. Unless stated otherwise (say in the case of official policy rates) Bloomberg’s interactive charts, for reasons unknown to me, show the evolution of variables over a certain period of time, with relation to the first point. So at t=0, the first observation, the variable is equal to “0” (zero). So the charts don’t show the actual values of the variable but only changes from the pre-selected starting period. This means that in principle one can take the variables and collate them accurately. However, because Bloomberg only provides charts (which is already extremely generous), rather than the full dataset, it becomes impossible to do this, so the best we can do is look at changes in relation to different period, which can be confusing.
I presume that the Markit figure shown above provided by FTAlphaville is probably a slighlty longer version of the screenshot of the Bloomberg interactive chart, immediately above, for the evolution of 5 Year CDS over the last 6 months. Considering how the FTAlphaville figure begins in April 2011 and how the (+/-) 5 year Bloomberg variable behaves, this is likely to be true. Indeed, by April 2011, Spanish CDSs were considerably below their October (Bloomberg chart) level. This makes them likely to cross the Italian ones “faster” (in the sense of shortly after the beginning of the series/chart) if such a chart begins in April , rather than in October. This is what happens, as the first crossing occurs in 1 month after the beginning of the Bloomberg series, rather than a full four months as in the chart provided by FTAlphaville. Of course, I could be missing something. If so please let me know. I’m happy to apologise again. I seem to be getting some practice. 😉
This fact is very important. It didn’t take me long to find a time framework that fits the less popular view that Spain is doing just fine (which according to the long term view over the last 5 years, is clearly not true).
In this post I have attempted to correct some errors I had made in a previous post. I hope these are taken cared of. I have now posted a notice regarding this fact on the original post. I left the original there in order to remind myself and other of this. April be it is slightly masochistic of me…
However in finding these errors, I also stumbled across some other interestingly confusing facts, that I attempted to promptly discuss in the lines above. I am not accusing the writer of that post of having made a mistake. I am merely pointing to the inconsistencies that arise when we compare the charts provided in that article and the ones found via the Bloomberg charts over the last 6 month with the 5 years Bloomberg figure. Clearly, if this post is evidence of anything else, it is that I make mistakes. May be the same is true of the author of that FTAlphaville post. If not, please let me know. I’ll be sure to retract these comments as soon as possible. I’m not being cheeky…