VAR versus expected shortfall: why Priips has got it wrong

Hardwiring of older risk measure into Priips means risk ratings could mislead investors

Missing the point: the Priips 2.5% VAR methodology effectively disregards any differences in the products' KIP barrier levels

David Stuff is co-founder and chief executive of Cube Investing, a distributor focused on structured investments; Arthur Noble is director of research at Cube Investing

Despite its troubled birth, there is much to admire about the aims of Europe's forthcoming regulatory overhaul of the retail structured products market. First and foremost, the Packaged retail and insurance-based investment products (Priips) regulation will, for the first time, put structured products on a level footing with other retail investments in terms of ease of comparability of their perceived riskiness, by applying a system of risk ratings.

That, at least, is the intention. Unfortunately for the ordinary punter, inherent flaws in the way the risk rating system has been designed mean it is likely to be far less useful than it might have been. Specifically, we argue that regulators have erred in basing the calculation required to generate a product's risk rating on the value-at-risk methodology rather than expected shortfall, which has better statistical credentials and will be hardwired into dealers' modelling practices by the Fundamental review of the trading book (FRTB).

Our analysis shows that a VAR methodology puts products into one of two buckets: those with capital protection and those with capital at risk. For products where capital is at risk through a down-and-in put, the VAR methodology does not help distinguish between products with lower or higher barriers. The volatility of the product is instead determined by the underlying asset.

Using the 2.5% VAR percentile is tantamount to calculating volatility using the worst case for the underlying, and disregards barrier levels and kick-out features

We anticipate the use of a VAR methodology may have a number of ramifications. For instance, a product's risk number as shown on its Key information document (Kid) may not be widely used for products that feature knock-in puts. Wealth managers and investors may instead look for risk measures that offer more granularity, so they can better identify the risk of different products.

In addition, we may see the development of more products that use put spreads or binary puts, as these can be structured to look less risky under the Priips methodology. As others have pointed out, using the 2.5% VAR percentile is tantamount to calculating volatility using the worst case for the underlying, and disregards barrier levels and kick-out features.

Our analysis shows that Priips' VAR-based volatility calculation does not separate products that have different barrier levels or different knock-out schedules. Instead, the volatility simply reflects the worst case of the underlying assets. This is a fundamental drawback to the VAR methodology; as John Hull puts it in his well-known Risk Masterclass: "Where VAR asks the question 'how bad can things get?', expected shortfall asks 'if things do get bad, what is our expected loss?'."

To illustrate how the Priips VAR methodology fails to reflect a product's important risk-mitigating features, we take those public single-index autocall products currently available pre-strike in the UK market that have a kick-in put – there are 19 of them – and compare volatilities using two differing methodologies: the Priips' 2.5% VAR approach, and expected shortfall.

The graph plots the products' volatility under the Priips methodology against their expected shortfall volatilities (blue triangles), and their expected knock-in put (KIP) payoffs against their expected shortfall volatilities (red dots).

The distribution of results suggests volatilities under Priips methodology are very similar across all products, making it hard to tell them apart – despite the products having KIP barriers that range from 50% to 70%. This is because all these products would, using the Priips simulations (including the use of a zero equity risk premium) have at least a 2.5% chance of the KIP being triggered.

The Priips 2.5% VAR methodology effectively disregards any differences in the products' KIP barrier levels and the differences that any intervening autocall events might have – for example their frequency or timing, and their defensiveness – and simply looks at the worst outcome that can occur.

Expected shortfall volatilities, by contrast, show a significant spread of volatilities, reflecting the products' different riskiness. To illustrate this, the round dots (right-hand axis) show the expected shortfall from the KIP, showing that the higher expected shortfall volatilities are associated with higher expected shortfalls.

VAR is a point on a returns distribution, whereas expected shortfall measures the average of everything that happens to the left of that point, and so represents its contents; VAR does not care how ugly things become further to the left.

To further illustrate the problem, we now compare all public single-index autocall products currently available in the UK market that have already struck, and have published secondary market prices (158 of them):

The graph shows two distinct clumps of products. Their clump-iness and the discontinuity between them makes it difficult to tell products apart: products with a low volatility under Priips correspond to products with low probability of triggering the KIP, so that the 2.5% VAR lands in the typical return-of-capital-only scenario.

High Priips volatility products correspond to products where the 2.5% VAR lands firmly in the KIP scenario, so that the volatility is determined by a near-worst case return of the underlying asset.

For the expected shortfall approach, we use the region to the left of the worst 10% of returns, rather than the Priips 2.5%. This makes derived volatility responsive to the entire topography of the left tail, rather than just one point way out on the horizon. This has two advantages: volatilities are representative of all these left tail returns, not just one VAR point, and there is no discontinuity: volatility progressively increases to reflect increasingly poor average returns in the left tail.

The expected shortfall approach allows investors to distinguish between products that have higher or lower barriers. The kick-out schedule will also impact the volatility calculation. The conclusion we draw from our analysis is that for products with a knock-in put, a volatility calculated using expected shortfall provides investors with a much more accurate and useful measure of the risk of the product than a VAR method.