Duration and Rating Premia in Investment Grade Municipal Bonds
How much difference do ratings make?
In forthcoming work, Albert Pinto and I will propose that Congress create an independent public ratings agency — an FDA — for green finance. As of writing, the business of project assessment and rating classification for green finance is in private hands led by BlackRock. This is a recipe for trouble because private credit ratings are coupled to the financial cycle. For instance, during the mid-2000s, deterioration of rating standards played a crucial role in amplifying the financial boom. We can be virtually certain that, were green ratings to be left to “independent” rating agencies, they will also become decoupled from fundamentals as the green boom gets underway.
Cornell Law Professors Robert Hockett and Saule Omarova have proposed a National Investment Authority (NIA). This is an excellent idea that we support in general. However, we think that there needs to be an “air gap” between the project assessment and rating classification function on the one hand and the project finance function on the other. This is because bureaucracies have a tendency to expand under their own momentum. Ie, they suffer from a version of Parkinson’s law. Because larger mandates increase the power of the bureaucrats responsible, they have a natural incentive to expand their fiefdoms. If the process is left unchecked, the result is a bloated bureaucracy, inefficient use of public resources, and poor capital allocation. This was the process responsible for the otherwise completely unexpected appearance of investment cycles in planned economies and the extraordinarily bloated License Raj in mixed economies like India.
Hockett and Omarova propose surveillance of the NIA by an oversight board. That’s prudent and necessary. But it does not go far enough. A simple solution to the problem is to create an independent agency for project assessment and rating classification that will not be responsible for financing the projects. Technocrats working for the public ratings agency will have little incentive to massage the numbers — it would not increase their revenues (unlike the “independent” credit rating agencies in the mid-2000s) or their power (unlike Gosplan, the Soviet central planner). This amendment is in general consistent with Hockett and Omarova’s design principles.
The basic idea is that well-paid and technically-competent teams of public servants working for the ratings agency will assess the economic viability and contribution to decarbonization of all projects submitted to them by firms, cities, states, and NGOs. They will then stamp these projects with a ratings classification. Only projects that meet predetermined thresholds of economic viability and contribution to decarbonization will enjoy the highest ratings. The ratings will then serve as a public signal for patient capital to invest in the projects and for the NIA to make decisions about derisking and public investment as necessary. High social net present value (NPV) projects that are already attractive to patient capital can be left to them. Projects that are high NPV from a social point of view but are relatively unattractive to patient investors, say, because they are difficult to monetize, will call for derisking and public-private financing as envisioned by the multilateral financial institutions. High value projects that are socially rational but cannot be made attractive to patient capital even through derisking, will have to financed entirely by the NIA, and ultimately the American taxpayer. This is the most sensible way to design the institutions mandated to plan, finance, and steward the energy transition.
As I mentioned on Twitter, cities and states are at the frontlines of the climate crisis. From upgrading the housing stock to mass transit, from electrification to sea walls, most of the work required for decarbonization has to be done at this level. The green finance institutions, however they emerge from the sausage factory in DC, must finance projects undertaken by cities and states. The problem is that cities and states often face high financing costs. In what follows, we’ll try to pin down exactly how high.
The market for municipal bonds is a sleepy corner of the bond market. It accounts for no more than one-tenth of outstanding bonds by value. In wholesale markets for secured funding, muni bonds rarely serve as collateral. They are largely absorbed by patient intermediaries hungry for yield. If we design the green finance institutions well, all this will change. Not just the labor market, but the muni bond market will run hot.
The aggregate size of municipal debt grew rapidly until the financial crisis. For some reason that I have not been able to determine, perhaps due to a legal or regulatory innovation, there was a dramatic jump — from $1.6 trillion to $2.3 trillion — in outstanding municipal debt in the first quarter of 2004. Since the GFC, it has not grown at all. Municipal debt hit $3.2 trillion at the end of 2010, exactly where it stood at the end of 2020.
A better way to think about the scale of the muni market is as a fraction of the financial size of patient intermediaries. Here we look at the ratio of outstanding municipal debt and the aggregate financial assets of pension funds and insurance companies. And we restrict our attention to the period after 2004. From 14-15 percent at the end of the 2000s, the ratio has collapsed to 8 percent. The import is unmistakable — cities and states have been living in austerity. This will have to change if they are to play their part in the energy transition.
It is hard to find data on muni bonds. In order to pin down the financing costs as a function of credit ratings and duration, I have extracted a sample from FINRA. In order to compare apples-to-apples, I chose plain-vanilla muni bonds of the most common type: recently traded, uninsured, fixed-rate, semi-annual coupon bonds, with proceeds going towards “general purpose/public improvement” and paid for by regular tax revenue. Because FINRA only gives API access to institutions, I had to manually copy-paste the data from here into a spreadsheet (available upon request). I tried to obtain data on at least a hundred bonds in each rating class. The dataset contains maturity, rating, coupon, price and yield data on 806 bonds with ratings ranging from AAA to BBB and tenor ranging from 2 weeks to 44 years. Municipal institutions that are not classified as investment grade (BBB- or above) are few and far between — they are basically locked out of the capital markets. So the dataset is more or less representative.
Duration sports a risk premium (ie, yields on higher duration bonds are generally higher) because high duration assets are more exposed to interest rate risk — the longer the asset’s cash flow horizon the more sensitive the price of the asset is to fluctuations in the discount factor. Ratings sport a risk premium because they capture credit risk — not just the risk of outright default, but importantly the risk of rating downgrades. The next figure displays these premia for our sample. I have discretized duration to facilitate comparison — the edges are the quartiles. And I have combined fine-scale rating classifications (eg, A+, A, and A- into A) into four broad buckets (AAA, AA, A, and BBB). The error bars are standard errors of the means. We can see that both duration and ratings have significant risk premia associated with them.
The next table displays the mean yields for each duration-rating combination in our sample. Going from the lowest to the highest duration quartile increases the yield by about 1 percent or 100 basis points (bp); going from AAA to BBB rating increases the yield demanded by investors by about 120bp. Going from low-duration, AAA-rated bonds to high duration, BBB-rated bonds increases the yield from 1.3 percent to 3.5 percent, a premium of 220bp.
These premia are statistically significant as the next figure shows. (We use the Matlab function anovan together with multcompare.) The staggered pattern suggests that these premia are quite stable. Only the lowest duration, highly rated muni bonds do not command a statistically significant premium over the AAA-rated, low-duration baseline.
The analysis above, although it gives us a sense of the relative magnitudes of the premia associated with duration and rating, is not exactly kosher. This is because duration is not independent of ratings. Issuers with lower ratings find it harder to issue bonds of long tenors. But we can easily control for that with linear regression. The next table shows the estimates from a model of yields as a linear function of duration and ratings. The model explains 15 percent of the sample variation in yields. The specification is such that the intercept corresponds to low-duration, AAA-rated bonds. So the estimates in bold can be read as the statistically-significant risk premium commanded by duration and rating after controlling for rating and duration respectively — relative to the low-duration, AAA-rated baseline.
The result is consistent with our earlier estimates. The baseline premium over the risk-free rate, which is at the zero lower-bound, for a AAA-rated, low-duration muni bond is 1.3 percent. One quartile higher duration adds 24bp; one step lower rating adds 39bp. Going from the lowest to the highest duration quartile adds 98bp; going from AAA to BBB adds 121bp. The raw premium on BBB rated, low duration bonds is 2.5 percent.
These estimates give us a straightforward way to pin down the gains from a setup where an independent agency provides reliable ratings and the United States guarantees the principal on muni bonds to fund high NPV green projects. Such a guarantee would imply a de facto AAA rating for said projects and therefore drastically reduce the financing costs for cities and states. But by how much exactly?
Exact calculations require assumptions about duration, coupon rates, and so on. But we can easily pin down a back-of-the-envelope ballpark using the time value of money formula as follows. If cities and states borrow, say, 10 trillion dollars, and promise to pay it back 10 years later, a yield of 1.3 percent (low-duration, AAA baseline) means that they will have to pay investors $11.38 trillion back. If the yield were 2.5 percent (low-duration, BBB ballpark), they would have to pay $12.80 trillion instead. That’s a gain of $1.42 trillion for the American taxpayer. Of course, it scales if you hold the other features constant: if the cities and states borrowed $5 trillion, the gain would be about $717 billion; if they borrowed $20 trillion, the gain would be $2.87 trillion. For reference, Henry Paulson’s market-tranquilizing bazooka was $700 billion, and Biden’s supermassive stimulus package was $1.9 trillion. So the gains are pretty substantial.
Given that a fully publicly-funded energy transition is not viable, derisking is an effective strategy for green finance. A carefully-designed green finance strategy should make the most efficient use of politically-expensive tax dollars. The most efficient way to use the fiscal capacity of the United States is for the taxpayer to bear the risk of capital loss on (genuinely AAA-rated) green investments, thereby allowing patient investors to confidently fund the energy transition by themselves. This strategy mines the risk-bearing capacity of both the United States and patient financial intermediaries. But much depends on the specific design of the institutions now being set up to finance the transition. Albert Pinto and I will convince you that a major pillar of green finance has to be a standalone public ratings agency.
Novice question. Feel free to ignore. How do you control for the variables duration and ratings you are regressing? Thank you.