Back to Portfolio for the Future™

Advancing the Infrastructure Investment Narrative

August 11, 2014

11-08-09 © zentiliaEDHEC Risk Institute has proposed a valuation and risk-management framework for illiquid infrastructure debt, a framework that it calls “academically robust yet operationally implementable.”

This has been a long time coming. EDHEC’s research program on infrastructure financing goes back to 2012.

Earlier this year, EDHEC’s Frédéric Blanc-Brude put out a position paper explaining that the development of benchmarks that will help investors understand their actual and possible infrastructure positions would offer critical support for “the infrastructure investment narrative.”

The newer publication along these lines, “Unlisted Infrastructure Debt Valuation and Performance Measurement,” on which Blanc-Brude collaborated with Majid Hasab and Omneia R.H. Ismail, gets to work on the problems outlined there, focusing specifically on private project finance (PF) loans, which make up “by far the largest proportion of illiquid infrastructure project debt.”

SPEs and Lenders

PF loans are those made to a special purpose entity (SPE) existing for the duration of the project, where lenders have extensive control rights through covenants and/or embedded options. Blanc-Brude et al call it a “unique form of corporate governance.” Its unique features help render existing valuation and risk-management models unhelpful.

Within the world of PF loans, the paper focuses on two spaces: merchant infrastructure and contracted infrastructure. The former includes projects that generate revenue by selling outputs or services into a market, [like a road that will collect tolls from drivers when complete], while the latter refers to the receipt of contracted-for revenues, with little or no market risk [like a school or hospital to be sold to a municipality].

These two sorts of project have different sorts of risk, and will be structured in different ways. In the case of a merchant infrastructure PF, the debtors will have to get paid faster than the equity owners (or, in technical terms, there will be a rising mean debt-service coverage ratio, DSCR). Also, to protect themselves form revenue-related risks, lenders will demand a longer tail than their counterparts in the contract debt space.

So here are key elements in the new model these three authors propose:

  • Structural credit risk modeling, with “a parsimonious set of empirical inputs,” can value the project debt;
  • The valuation model can derive the relevant risk and return measures;
  • This will entail “arbitrage bounds,” that is, the limits on possible valuations given the characteristics and preferences of investors;
  • The performance of such debt “can be used to populate a centralized database,” which will allow for ongoing monitoring.

DSCR as Proxy for Cash Flow

Intuitively the problem with valuing the project debt is this: the free cash flows of the SPE aren’t easily observed. If it were a matter of public record, then modelers could arithmetically derive their present value. But they aren’t, these are private concerns.

So Blanc-Brude et al. propose the use of the DSCR, “which is typically monitored and recorded by lenders” as a proxy. The DCSR, combined with the base case debt service can allow observers to infer the value, and the volatility, of the SPE’s free cash flow.

Distance to default is also instrumental in this model. That raises the question: what does it mean for a project to default? This is tricky. As I observed in this blog in November 2011, in connection with the causes of the collapse of MFGlobal, default is not necessarily a binary proposition, either A or not-A.

The EDHEC authors observe that in the Basel II framework, the arithmetical way to define default is any DSCR < 1. Now that sounds binary. But … the authors also note that in the Basel II framework PF loan default is defined as “past-due more than 90 days on any material credit obligation to the banking group.” The material credit obligations are a matter of contract, and this allows for the possibility that an SPE will be in default while its DSCR is at 1 or above.

So the distance-to-default is a little tricky. It is the probability that DESCR falls below 1, conditional on the absence of any default until then.

The authors contend that “knowledge of the first two moments of the distribution of DSCRt is sufficient to derive the SPE’s distance to default….”

There is much more to the paper, which we heartily recommend in its entirety.