Economic woes and regulation add to VA hedging challenge

Uncertain global economic conditions mean that consumer demand for variable annuity products remains high. Yet these same economic conditions make the hedging of the embedded guarantees more challenging, while the steps taken by regulators to constrain the risks of derivatives have brought additional layers of complexity to the hedging process. Insurers, as a result, are becoming more sophisticated in how they manage their exposures.

maze

Variable annuities (VAs) have always posed challenges, both to those that issue them and to the regulators that oversee the market. Insurers must design and hedge the products in such a way that they can meet their long-term liabilities. Regulators must assess whether the products – and the promises they make in terms of their in-built guarantees – present a risk to the overall financial system.

While the debate around the systemic risk of VAs continues, the current difficult and uncertain global economic conditions mean that consumer demand for these products remains high. In the US, VA assets-under-management topped $1.6 trillion in 2011.

Yet, these same economic conditions make the hedging of VA guarantees more challenging, and the steps taken by regulators to constrain the risks of derivatives – the primary mechanism used by insurers to reduce their VA exposures – have brought additional layers of complexity to the hedging process.

Although research has shown that, during the financial crisis, most VA hedging programmes were more than 90% effective in achieving their goals, in the subsequent years insurers have had to up their game in how they manage their exposures.

“Historically, if companies were hedging their VAs, they were often making simple assumptions about market movements,” says Pouyan Djahani, a member of the annuity solutions group at Aon Benfield Securities, based in Toronto. Even where they were using stochastic approaches, it was often via simplistic models such as Black-Scholes. “But the companies are hedging with long-dated options going out 30 or 40 years, so these were not realistic models,” he says.

The guarantees in VAs, which can come in many forms but basically guarantee the performance of the underlying funds or their eventual payout, are in effect written put options. As such VA hedging programmes are essentially about hedging the sensitivities (greeks) of these options to market and economic parameters, most notably movements in the price of the underlying assets (delta), the volatility of the price of the assets (vega), the rate of change in the delta (gamma) and change in interest rates (rho).

Companies hedge these sensitivities in the capital markets using exchange-traded derivatives, such as futures and options, or over-the-counter derivatives, such as swaps and swaptions. Companies can also choose not to hedge – an approach usually only taken for small blocks of business where the cost of hedging might outweigh the benefits.

Since the financial crisis, hedging programmes have become more sophisticated, with more realistic models and more dynamic approaches in terms of the frequency of the risk analysis and hedge trades. However, the overall approach to VA hedging depends to some degree on a company's attitude to the VA market. A number of insurers, particularly in the US, have decided that VAs are more trouble than they are worth and have either cut back or pulled out of the business. However, unless they sell off their books, they – like the companies still issuing such products – must still hedge their exposures.

“Where companies are running off legacy books of VAs, there is a strong effort towards controlling costs, defining efficient processes for managing and monitoring risk and carrying out hedging,” according to Sam Nandi, actuarial group leader in the financial risk management group at Milliman, who is based in Chicago.

Companies that have remained in the VA business are no doubt cost conscious as well, but many have also sought to design risk out of their products where possible. “Compared with traditional VA products, most of the products offered in the market today have been de-risked to some degree,” says Nandi.

The most common approach to de-risking is through the control and design of the funds in which the VA customer invests. Whereas previously customers tended to be given the choice of a wide range of conventional mutual funds, often managed by a brand-name asset manager, now they are likely to be offered a more restricted and proprietary set of funds that will have a dynamic hedging element built into them. This benefits both the customer and the insurer, says Nandi.

“The customer is no longer just buying mutual funds but also risk management, and the company has a fund with a more favourable risk profile on which to write the guarantees that the customers value in VAs,” he says.

One result of the de-risking of underlying funds is to reduce the amount of VA-related hedging that has to be done on the company's balance sheet. On the other hand, there is greater complexity in actual calculations and processes required to carry out the overall hedging strategy. Furthermore, since the funds are managed in response to changing market conditions, perhaps even intra-day, the calculation of the risk characteristics of the liabilities is also more complex.

“Unless you rigorously model the dynamics of the funds in the liability analytics, there will be the potential to misstate the risk, or not take the appropriate credit for the reduced risk profile,” says Nandi.

Most VA hedging programmes today cover delta and rho, according to industry observers. The degree to which delta is hedged in the fund or on the balance sheet can depend on market conditions, with companies generally keeping more delta within the fund during times of unstable markets. The degree to which rho is hedged in the fund or on the balance sheet tends to come down to a company's view as to where it should best manage interest rate risk and differs widely from company to company. De-risked funds largely hedge vega internally.

One particular problem for dynamic hedging programmes can arise from volatility spikes because of their impact on gamma, says Djahani of Aon Benfield. VA books can be characterised as negative gamma portfolios. This means that when the market goes up, delta goes down. “To compensate in a dynamic hedging programme and stay delta neutral, you have to buy the underlying – but the market [price] has gone up. When the market goes down, since the portfolio is gamma negative, the delta goes up, so you have to sell to reduce delta. In other words, you are buying high and selling low, which can affect profit and loss,” he says.

A solution to this so-called 'gamma bleed' is to hedge gamma. “But hedging gamma requires more complex instruments such as equity options, which only a few companies are prepared to use, so many remain exposed to gamma bleed,” says Djahani.

Another focus of VA hedging programmes is credit risk. Here too, the issues are becoming more complex than was previously the case because the market has moved from using Libor to an overnight index swap (OIS) curve to discount the collateral attached to derivatives deals. Also, as regulators push OTC derivatives trading towards centralised clearing, collateral costs and complexity become more significant. The challenges of credit risk hedging in today's environment add to the requirement for more sophisticated modelling capabilities.

But developing accurate, stable and reliable models that can cope with all the risk factors in an integrated way is by no means straightforward. “Even in benign market conditions it is challenging to get the Monte Carlo-based modelling environment right,” says Edward Perry, senior quantitative analyst at Phoenix Life Insurance in Connecticut. “We have found it seems best to strike a balance between good calibrations to market data and good control over the correlative structures involved – in other words, getting the marginal distributions right, while also getting the joint distribution right.”

For Phoenix, it means using a Heston model for stock indexes volatility and a reduced-factor Libor market model for interest rates. “There are probably better ways to handle each marginal risk, but this approach of making things as simple as possible – but not simpler, to use Einstein's phrase – seems optimal for us,” says Perry.

One consequence of the heightened complexity of the modelling allied with the large books of VA policies is the need for high-performance computing to perform the required calculations in a timescale that supports a dynamic hedging approach. Insurers are turning to the latest parallel processing technologies for help. US insurance and investment group Transamerica Capital Management (TCM), uses a computing grid – a network of 64-bit processors managed by software from DataSynapse. Others, such as Aon Benfield, Milliman and Phoenix, are using graphical processing units (GPUs) – microprocessors originally developed for computer graphics and optimised for parallel processing.

“Very large data sets – thousands or even millions of client accounts, each of which features complex embedded optionality – makes the raw power of one's computing environment crucial for effective VA hedging. GPU technology in the Monte Carlo context is becoming de rigueur,” says Perry of Phoenix. Another high-performance option, offered for example by Aon Benfield, is cloud computing – renting the processing time from a third-party provider of massive online computing resources, such as Amazon, Microsoft or Google. 

Even more important than the processing power, say insurers and hedging services providers, is intellectual power. There are relatively few people with the capabilities and experience to manage hedging programmes for large and complex books of VAs. Turnover is high as qualified staff are aggressively fought over, while those companies simply running off books find it particularly difficult to retain skilled, ambitious people. Access to intellectual resources is now one of the key selling points for VA hedging services providers, such as Aon Benfield and Milliman, which they bundle in with their sophisticated models and high-performance computing facilities.

Some companies hesitate handing over hedging responsibilities to consultants for what can be an impactful area of business. Some are reluctant because they believe they understand their liabilities better than a service provider ever could, so they will make better informed hedging decisions. Others that have developed de-risked funds do not want to reveal their methods, especially where the service provider also offers de-risked funds of its own. For TCM it was also the sheer size of its VA business, worth around $50 billion, and issues around control and responsibility that made it reject putting its hedging programme into the hands of consultants.

“We chose not to outsource our VA hedging to consultants because we did not have confidence in our ability to manage our profitability by outsourcing this aspect of our business,” says Hunt Blatz, California-based vice-president and head of model development at TCM. “It is one thing to ask a consultant to take a look at one particular problem and vet their solution, and another to give the consultant ownership of a daily production operation without the consultant assuming financial responsibility for that work.”

TCM uses analytics and hedging automation software from New York-based Numerix for its hedging programme, as does Phoenix. Elsewhere, a range of in-house and third-party software is deployed. Aon Benfield has its own hedging platform, although it incorporates pricing and risk analytics from Vancouver-based Fincad. Milliman used to mainly sell its MG-Alfa and MG-Hedge software for companies to install and use on their own, but now the company focuses on hosting the software bundled with advisory and hedging services. Other third-party vendors include Towers Watson, SunGard and IBM Risk Analytics.

A significant challenge for tool developers is the trade-off between sophisticated dynamic hedging capabilities and ease of use, says Curt Burmeister, head of buy side products at IBM Risk Analytics, who is based in Boston. “We have found is that it is hard to have a system that is really good at dynamic hedging yet still easy for actuaries who are not programmers to customise and keep updated as products evolve,” he says.

One of the key issues for TCM as it moved from an in-house hedging platform to a third-party system was to try to reduce the requirement for expert programming skills for actuaries. “Aegon preferred to give the hard-core coding of the system’s kernel to a professional software development company rather than have it as part of the actuarial role,” says Blatz.

The company uses a combination of Numerix's CrossAsset analytics and its Leading Hedge hedging platform. CrossAsset provides a pay-off scripting language for describing complex products such as VAs, and then automatically translates this into modelling code. The models are then transferred to the automated production environment of Leading Hedge.

The availability of the relatively easy to use scripting language has advantages for TCM, by easing the problem of finding experienced staff. “[We] can now leverage graduate students from actuarial sciences with modest programming experience who are able to start adding value in weeks instead of several months,” says Blatz.

With benign market and economic conditions still appearing to be a long way off, it is likely that consumer demand for VAs will remain high. Turning this demand into a profit is very much a challenge of running effective hedging programmes.

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact info@risk.net or view our subscription options here: http://subscriptions.risk.net/subscribe

You are currently unable to copy this content. Please contact info@risk.net to find out more.

The future of life insurance

As the world constantly evolves and changes, so too does the life insurance industry, which is preparing for a multitude of challenges, particularly in three areas: interest rates, regulatory mandates and technology (software, underwriting tools and…

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here