EDHECinfra Senior Relationship Manager Grace Chen answers five key questions about what the project of creating benchmarks for infrastructure investors entails, and how our research team is addressing the challenges involved.
She discusses the reasons behind creating such benchmarks and the data collection strategy being implemented to create representative samples of “investable” infrastructure assets on both equity and debt sides. She also addresses existing methodologies and their shortcomings and why innovation is necessary in the private asset space is institutional investors are going to make real forays in infrastructure investments. Grace also addresses the type of benchmarks that EDHECinfra is going to build.
Senior Relationship Manager
For several years much has been said about these issues but not much done. EDHECinfra is a 15-strong team with a multi-million dollar budget created to address this gap.
We do this firstly, by defining what infrastructure investing is. Next, we have developed asset pricing and risk models that can take into account the specifics of infrastructure assets. On that basis, we can determine which data needs to be reported so this technology can be applied using information that already exists and can be collected, even standardised.
Indeed, a very important part of our work is data collection. We have been collecting data from both infrastructure creditors and investors since 2015. As of March 2016, we already built the world’s largest database for infrastructure cash flows and our database is still expanding.
By the end of 2016, we will publish results about the risk-adjusted performance of portfolios of private infrastructure debt or equity designed to be used as benchmarks.
This will allow better allocation of assets, better attribution of performance and better calibration of risks models.
These benchmarks will be published on an annual basis and cover different types of equity and debt infrastructure investments around the world.
We have relationships with an expanding base of long-term infrastructure investors. From them, we collect realised and forecast cash flow data of their individual investments. We speak to creditors and bond holders, from whom we get realised and forecast debt service and ratios, and to equity investors and their managers to collect realised and forecast dividend data. We also obtain financial models from public sector regulators or through freedom-of information requests. We also collect historic data from the statements of accounts and annual reports from individual infrastructure investments.
In each country or market, we begin by surveying the entire population of “investible infrastructure” firms, whether they are individual projects, utilities or regular corporations. In numerous countries there are hundreds such firms, that is, infrastructure firms that one could invest in, even though the immense majority is not for sale. We categorise this population of investable infrastructure by what we call “physical “ and “business model” attributes, in other words, by the major characteristics that we expect can explain the variance of asset prices.
We then create a representative sample of this population taking into account their attributes in this particular market. So, for example, if investable infrastructure a given country consists of 13% of merchant projects in the $100-$300m range, we aim to build a sample which is representative of this fact.
Once we know which firms we want to collect data about, we obtain data from our contributors and from audited accounts. We read each statement of accounts – year by year, company by company. We clean the data, put them in our own system. The end result is a high quality database of infrastructure investment data.
By 2017, we expect our database to cover in excess of 1,500 firms in both OECD and emerging markets dating back 15 to 20 years.
NAVs for example are based on appraisals and can be affected by the “stale price” issue that leads to smoothing and underestimating the volatility of returns. The use of constant IRRs also precludes computing the correct risk measures when risk profiles are expected to change over time.
The challenge is that infrastructure investments are extremely illiquid: individual firms are seldom traded and that makes it very difficult to apply methods used in other private investment spaces such as real estate or edge funds. Hence, we need to rely on more advanced statistical techniques.
Using such models, we are now able to derive a robust estimate of expected cash flows (in the statistical sense) and their volatility controlling for the attributes of different types of infrastructure firms. [here is a recent paper on the dynamics of debt service cover ratios]
We can also derive valuations and return measures by estimating the term structure of discount factors that best explains observed transaction prices given what we know about cash flow volatility in such investments. [see for example]
The combination of advanced cash flow models with adequate pricing models allows us to derive all the necessary risk and performance metrics that can put private infrastructure investment on an equal footing with other asset classes, from Sharpe ratios, to duration to conditional VaR.
Recent debates around the calibration of Solvency-II have shown that there is little usable information.
On the private debt side, rating agencies have documented default frequencies in certain classes of private project debt but this is not enough to arrive at valuation and portfolio construction results and build investment benchmarks.
On the equity side, investors have been using combinations of public equity indices – so-called “listed infrastructure” – or indices created by averaging the internal rates of return or IRRs of private assets. But listed infrastructure indices do not capture the characteristics of private infrastructure projects well at all.
They are extremely concentrated in a few stocks and, in a forthcoming paper, we show that they tend to be more volatile than the market as a whole.
Likewise, as any finance textbooks could tell you, you cannot just average IRRs… and with smoothed returns, these approaches fail to inform us about risk-adjusted performance or return correlations.
In due course, such benchmarks can also be designed to help address broader investment solutions to achieve a combination of duration and risk-adjusted performance or estimate the factor loadings of infrastructure investments.
In the coming months, we will launch a widespread industry consultation about what kind of private infrastructure benchmarks would be the most useful and relevant for investors.