A lot of ink, and innumerable pixels, have been spilled on the question of whether Green Bonds outperform, or underperform, comparable vanilla bonds from the same issuer, or the overall market, or what. For all I know, there’s probably some article out there from some financial journalist on Men are from Mars, Women are from Venus, and Green Bonds are from Someplace Special. The laziness and sloppiness with which this issue has been pursued has been disappointing, but not necessarily surprising. There have been several “reports” from banks, and several academic articles, not to mention too many breathless media stories to count, that in the end have really explained less about Green Bonds and how they trade, and, sadly, more about the quality of research being published in the financial sphere.
Thankfully, some of this imbalance is addressed in the new report from The Climate Bonds Initiative (CBI). (Full disclosure–I reviewed this report in preparation.) With this report we finally have some evidence that Green Bonds might outperform in a certain time frame, and against what–in this case the first 28 days of trading, against a relevant index. CBI has indicated that this report will be followed up by others, which is a good thing. While this report provides some useful information, it still leaves a number of questions unanswered (mostly associated with issues outside of the scope of the report).
The CBI report also discusses some Green Bond behavior that should surprise no one–sometimes individual Green Bonds outperform their vanilla counterparts. Of course, sometimes they don’t, and the report has examples of both. Not so much in individual green Bond/Vanilla Bond trading, but when plotted on a traditional yield curve. Sometimes Green Bonds price inside the curve–and sometimes they don’t. Still, this is useful information for some bright empiric to build on. CBI also notes that Green Bonds are nearly always over-subscribed–but acknowledges, which many observers do not, that over the past several years nearly everything has been oversubscribed. Other researchers should absorb something from the caution exercised by CBI on much of their data and profit from it.
It is important to understand some of the methodological difficulties that arise in dealing with this issue, and I outlined these in one of my Green Bond reports when I was still at SocGen (which may or may not still be available on their website for clients.) They are straightforward but considerable, and it’s not clear that many of the commentators opining on the issue really understand these. If you want to see if Green Bonds outperform Vanilla bonds of the same issuer, you can try several approaches. One is look at yield curve behavior, which CBI (and every investment bank in the world, although you probably haven’t seen this published) has done. The more basic one is to simply compare the performance of two bonds of the same issuer with comparable duration, currency, sector classification, and, most importantly, credit rating. It’s this granulation that renders much of this work questionable, because resulting the data set is so small. This is what I wrote in July 2016:
All of which leads to our major concern about any of these claims—it’s not clear to us that the sample size of any analysis is sufficient to warrant any meaningful conclusions. We have attempted this sort of analysis, and have concluded that the data are simply insufficient. We looked at all corporate Green Bonds (excluding supra-nationals, municipals, ABS and REITS) issued between November 2013 and December 2015, and were looking to compare the pricing behaviour of each to a vanilla bond equivalent (in currency, size and duration). This is basically a two year time horizon for the longer bonds in the set, and considerably shorter for more recent issuance. This is not necessarily an insurmountable issue so long as one is prepared to recognize that the two year period in question was a period of significant compression in credit markets in general, during an unusually low interest rate environment that might not be replicable going forward.
More troublesome is the issue of sample size after granulating the universe. The target universe consisted of 55 bond issues across 33 issuers. This universe reflects eliminating issuers whose Green Bonds were their only bonds outstanding, and therefore had no vanilla bond counterpart, and several Green bonds for an issuer where there was no vanilla counterpart in duration, size or currency. The more relevant granulations, however, are those by sector coupled with currency—the largest single sector for our sample period was national development banks in euros—which consisted of seven bond issues from four issuers. US$- denominated Green Bonds from Development banks represented three issues from two issuers during this period. US$-denominated Commercial banks represented seven issues from six issuers. The largest industrial sector, euro-denominated issues from utilities, consisted of seven issues from five issuers. Other sector and rating granulations (we separated High Yield from Investment Grade) produced even smaller sets. The small size of these analytical groups suggests that meaningful judgments would be tenuous at best.
We readily concede that Green Bonds may at times trade differently from vanilla bonds of the same issuer. However, this is true for vanilla bonds of the same issuer as well—sometimes shorter bonds will behave differently from longer bonds, depending on any number of factors. Sometimes a bond won’t actually trade, which makes pricing judgments suspect. Our more general point is that we will likely remain suspicious of broad claims about the behaviour of Green Bonds until the universe is large enough and has a sufficient time period. Any purported analysis that comingles supra-nationals with industrials is likely to be flawed from the outset— there aren’t enough triple-A corporates to do an analysis with anyway, and trying to assess the broad behaviour of the entire Green Bond universe just has too many possible conflations to be meaningful, in our view. This may change as the universe gets larger—right now there do appear to be occasional pockets of demand not being satisfied by potential Green Bond supply.
It comes down to sample size, then. And we’re further along in the Green Bond universe, but it’s not clear to me that the number of suitable bonds for this sort of analysis has increased sufficiently. It may well be several more years of steady issuance before a proper analysis of this sort will generate a meaningful result. Which is why the index data that CBI provides is very relevant–it diminishes the importance of sample size (although it does not eliminate it completely.) The fact that CBI reports a notable difference in tightening between Green Bonds against a relevant index highly suggestive. The fact that Green Bonds tighten in general is a nice point, but unless it’s being used in a comparison with Vanilla bonds or an index, it’s just that–it simply demonstrates the Green Bonds behave like vanilla bonds. And it’s the tightening against the Index that matters here. This is the really significant contribution of the CBI report. The discussion of the performance of the Apple Green Bond is a nice analysis, and is particularly suggestive. What is now needed is some data on whether a broader set of similar bonds issued during the same period demonstrated comparable performance against indexes. Even if that’s the case, it still demonstrates that there are trading opportunities during the first month of issuance. Over the longer term, of course, our earlier comment about the importance of a sufficiently large market for appropriate analysis applies here as well–if this difference is currently material, will it remain so as market size increases any any implied Green Bond scarcity diminishes or disappears entirely?
CBI has, as mentioned, indicated that there will be further reports, and this is to be welcomed. Well-grounded research here is to be encouraged, and good behavior should be rewarded. At some point the corporate Green Bond market will be sufficiently deep to allow for the type of granulated analysis we referred to earlier, and this is to be welcomed. However, even here we would not expect any trading differences to have any longer term impact on trading. It remains the case that there is nothing structurally “special” about Green Bonds–the only difference between Green and Vanilla bonds remains the label. In every other respect the bonds are identical to vanilla bonds in terms of credit measures, and one would expect any differences to be arbitraged away quickly, assuming normal liquidity constraints. So even if it turns out that there is a difference in trading between Green and non-Green bonds in the first 28 days, how long does this persist?
All of this raises a more interesting question about Green Bonds. And it’s not why is there so much interest in the issue of whether they trade differently–personally, it seems to be mainly of interest to people unfamiliar with how bond markets work, rather like the way that intense questions about criteria seem to mainly be of interest to people who don’t actually buy bonds. It’s this–if Green Bonds do trade differently, even within the first 28 days, why is that? Because one of the things we actually don’t know about Green Bonds is why people buy them in the real world. Do people buy the Apple Green Bond because it’s green, or because it’s Apple? The same question can be asked of most of the corporate Green Bond universe–sometimes buyers want a Green Bond, sometimes they’re just looking for triple-B utility or a single-A bank with a seven year maturity. Yes, we know that there are a number of dedicated Green Bond buyers out there. But how large a group is this?
We don’t really have any market data on that–what is the percentage, on average, of purchasers of Green Bond deals that buy them because they’re green? 20%? 80%? We don’t know. Syndicate desks and bank Capital Markets groups know, as do corporate treasurers, but so far they haven’t told us. Another way to think of this is: How many benchmark Green Bonds succeed only from Green Bond buyers? Our own suspicion is not many. This would be useful information as well–it would mean the market sees no difference in the credit status of Green and Vanilla bonds of, say, Iberdrola or KfW. We just don’t know. But what the CBI reports suggests is that, while the issue of Green Bond outperformance remains unresolved, there is clearly data that requires a further look–performance against a relevant index in particular. Equally important, there is clearly no penalty for buying or owning one, so hopefully we can bury that shibboleth once and for all.