CAPM may be ‘one of the founding frameworks of modern finance’, but for determining the net asset value of unlisted infrastructure it is terribly inadequate. Frédéric Blanc-Brude, the director of the EDHEC Infrastructure Institute explains why and offers an alternative.

Originally published in Infrastructure Investor.

When reporting the NAV of unlisted assets like infrastructure using discounted cashflows, best practice consists of deriving a discount rate from an asset pricing model, which can be calibrated and adjusted to reflect the assets’ exposure to the risk of the relevant market – for example, the market for private infrastructure equity, as well as asset-specific or deal-specific components, such as control, distress, etc.

Alas, after many hours of interviews, surveys and desk-based research, we found that the real-life application of this approach often only pays lip service to best practice. There are three reasons for this.

People use the wrong model

The asset pricing model used is typically the Capital Asset Pricing Model (CAPM), a long-discredited approach to empirical asset pricing. While it is one of the founding frameworks of modern finance, multiple researchers – including economist Eugene Fama and his colleague Kenneth French – have demonstrated that CAPM is unable to explain historic equity returns. Research has shown instead that multiple factors simultaneously explain the returns of financial assets, but only if you use data from the right market. This is where things get worse.

The wrong data

Unlisted infrastructure equity investments have no straightforward listed proxies. Even if a listed proxy could be found, the resulting beta and market return still do not come from the right market and do not match infra investors’ expected returns. So, when using CAPM for private infra, people use smoothed inputs: a fixed beta (like 0.5) and a very averaged equity risk premia obtained from major stock indices (like 5 percent). The risk-free rate used is also a very average short-term rate (like 1 percent) instead of a yield curve representing the horizon of the investment. The result is a completely ad hoc and smooth discount rate that captures neither the risks (betas) nor the price of risk (risk premia) of an infrastructure asset. By now it should be clear that this is not really using CAPM, only plugging numbers into a formula. From there it gets uglier.

Ad hoc adjustments

Since the resulting CAPM cost of equity is unlikely to be the expected returns of equity investors in infrastructure, it needs to be plugged with an “illiquidity premium” (like 100 basis points). In the end, the calibration of and adjustments made to the discount are completely ad hoc and mix market and asset-level considerations, obfuscating what make this asset different from the average company in the same peer group and what is a reflection of market factors.

Even though these discount rates now have little to do with the CAPM, let’s call them “CAPM+” discount rates for convenience.

Does it matter if the discount rate is wrong?

In practice, at the time of investing, the discount rate is not a major concern. What matters to investors is the price and the expected yield given the cashflows. At that time, the determination of the CAPM+ discount rate consists of finding the right plug (the “illiquidity premium”) to match the price, given some (wrong) CAPM inputs.

But once set, CAPM+ discount rates tend to become stale and very hard to change because:

• they do not refer to a genuine and representative market benchmark;
• nor do they refer to the company’s relative exposure to this market;
• nor do they describe the difference between this company and the “average company” represented by the benchmark.

As they become stale, CAPM+ discount rates diverge from their true market value and become less and less representative of the market cost of equity. If the illiquidity premium was 100bps last year, what is it today? Still 100? 150? 50? No one knows, because there is no way to know. But we do know that in real life, market prices do change.

This is where one may start to worry about CAPM+ because the long-term nature of infrastructure investments means that such errors in the estimation of the cost of equity can lead to very large errors in NAV estimation.

For example, with a standard payout profile for a 30-year infrastructure investment project, a 100bps-error in the cost of equity is equivalent to a 20 percent error in future dividends. If an analyst gets the project cashflows wrong by 20 percent, do they keep their job? What about a 100bps error in the discount rate?

This is what happened to the shareholders of the Thames Water HoldCo. A collective case of myopia about the level of the applicable discount rate (not helped by Ofwat, which also uses CAPM) in the face of an increasingly risky business. By not changing the discount rate to reflect the level of risk and the market price of risk, investors in Thames Water failed to see that they had been losing money for years.

In the end, if your discount rate is stale, with significant changes in supply and demand (for infra assets) and interest rates changing all along the yield curve, there is no telling where the actual fair market NAV of an asset might be today. The larger the discount rate error, the more enormous the unrealised gains or losses.

Love asset pricing

Today, reporting the fair value of private infrastructure investments is more of an issue than it used to be. With infrastructure found in packaged retail and insurance-based investment products (PRIIPs), DC plans, evergreen structures, and even 401ks, it may really be time to start worrying about CAPM+.

Luckily, infrastructure is for geeks. There is no need to stay wedded to a pricing framework that is broken when modern finance and quantitative research offer a wealth of options to conduct scientific and robust asset pricing. Leveraging the power of better models than the CAPM and machine learning, it is possible to capture the factors (the multiple betas) that drive infrastructure market prices and to document the market dynamics. Separating the signal from the noise in recent transactions to calibrate a multi-factor model is a well-understood approach and, with the right data, it achieves very robust results.

Of course, a model of market prices does not predict every detail of the value of every asset. It just does a much better job than the CAPM at estimating where the infrastructure market is at. From this starting point – we call it a valuation anchor – investors can integrate asset-specific pricing components but without the illiquidity premium black box (since the model is calibrated with illiquid transaction prices).

Instead, idiosyncratic components can be clearly defined, and updated if needed, while the systematic part of the discount rate is updated each time new transactions take place. This allows estimating the NAV of infrastructure investments in practically real time – also known as nowcasting: NAV 2.0.

To find out more, see our asset valuation solutions