Funds try to predict behaviour of mystery investors

New EU rules on liquidity stress-testing force fund managers to hunt out clues on investors

  • From September, European fund managers must use redemption scenarios specific to individual funds as part of a new stress-testing regime but managers often do not know who their end-investors are.
  • “We fully understand that we have published guidelines for an existing legal requirement for which the underlying data is not perfect all the time… Firms must do the best they can with the information available,” says Patrik Karlsson at Esma.
  • Firms can use widely available historical flow data to estimate future redemptions. Many question its predictive power, though.
  • Managers can make assumptions on a fund’s investor profile by looking at the fund’s type, strategy and distribution channel. Access to more reliable or granular data can be negotiated in talks with distributors.
  • HSBC Global Asset Management uses a “waterfall approach”, increasing the number of factors in its model as more data becomes available.

The fund manager holds her head in her hands. A shiver in the markets has caused the unit price of her fund to slide. Some investors may want their money back.

But the fund manager has a problem. She can’t second-guess the size of future redemption requests because she knows so little about the end-investors. Are they retail clients prone to knee-jerk reactions? Or institutional investors with single large holdings capable of blasting through the fund’s liquidity reserves?

Fund managers across the European Union are in a similar bind as they prepare for new rules on liquidity stress-testing for alternative investment funds and popular retail funds known as Ucits, due to come into force on September 30.

The rules aim to help fund managers model liquidity risk to insulate themselves during future stress scenarios – a global pandemic, for example.

But as a large majority of funds in Europe are sold through intermediaries, such as banks, insurance companies and fund distribution platforms, the identity of end-investors is, in many cases, a mystery to fund managers. Without knowing what kind of institutions or individuals hold shares in a given fund, it is difficult to predict how those investors may behave in different scenarios. 

The European Securities and Markets Authority, architect of the new rules, acknowledges the problem.

“We are aware that this is not an ideal world that we live in right now,” says Patrik Karlsson, policy officer in the investment management team at Esma. “We fully understand that we have published guidelines for an existing legal requirement for which the underlying data is not perfect all the time, but this is an important first step. Let’s see how the availability of data evolves from now on.”

For now, he adds, “firms must do the best they can with the information available”. 

Many fund management firms think distributors should be compelled to disclose investor details to managers, but Karlsson says this would require extra legislation by the European Commission.

Comprehensive liquidity stress-testing should involve three main steps. First, funds are run through a variety of redemption shocks. Examples include historical events, such as the Lehman Brothers collapse; specific redemption levels – say, 5, 10 or 20%; a certain worst percentage of historical net flows; and a hypothetical adverse scenario – rising interest rates or credit spread widening, for example.

Second, a fund’s liquidity buffers are measured. And lastly, potential net outflows and liquid assets are compared to work out a fund’s ability to meet redemptions in a stressed scenario.

With the compliance deadline only seven months away, asset managers are scrambling to develop models and methods to satisfy the guidelines, even for cases of sparse information. HSBC Global Asset Management, for example, uses what it calls “a waterfall approach” to deal with the data limitations.

Meanwhile, software vendors are expanding their offerings to help funds stress-test liabilities (see box: Is there an app for that?).

But the variability of investor information between funds and between management companies is hindering the development of reliable and consistent liability stress-testing across the industry. 

Are you concentrating?

A string of fund suspensions in recent years has made liquidity risk a priority for funds and regulators. In the most high-profile case, Neil Woodford’s Equity Income fund was gated on June 3 last year as its substantial holdings of illiquid assets meant it was unable to immediately meet a large redemption request. On that day, Kent County Council signalled its intention to pull out its entire £250 million ($325 million) investment – almost 7% of the fund’s net asset value – piling further pressure on Woodford after months of outflows.   

Esma’s emphasis on redemption scenarios specific to each fund addresses such investor concentration risks, forcing managers to plan for various combinations of redemption levels and asset sales.

Although the fund sector is not entirely new to redemption modelling, the practice has so far been confined mostly to the biggest firms.

Chiara Sandon, senior policy adviser at the European Fund and Asset Management Association, notes that big asset managers were doing it even before the Esma guidelines came out. Amish Doshi, from the finance and risk practice at Accenture, based in London, says smaller firms have little experience of the required modelling techniques.

“Historically, I’m not sure I’ve seen much focus on redemption modelling, albeit the events of the last few years may have prompted institutions [to start]. But especially the small houses will have to build the models from scratch,” he says. In contrast, stress-testing of fund assets “has always been strong”.

Likewise, Scott Bauguess, a former senior official at the Securities and Exchange Commission, says that, outside of a handful of the largest US asset managers, most firms rely on finger-in-the-air methods and will struggle to develop the sophisticated approach necessary to model redemption risk.

“Most of the asset managers probably operate from their gut in terms of thinking this is going to be problematic or it’s not. I’ve spoken to many large asset managers in the past and asked them about liquidity management and some of them, although very large, were not quantitatively oriented.”  

He adds: “Very few [US] asset managers really have strong modelling skills and can do this sort of quantitative analysis that’s implied by the Esma guidelines.”

But even managers with robust modelling skills will run into a more fundamental problem: the shortage of data on end-investors.

We will often see the largest investor in our fund is a distribution platform. That’s the biggest data challenge

David Regan, Janus Henderson Investors

Often the only type of information available to fund managers is data on past inflows and outflows, which limits redemption scenarios to those based on historical patterns.

In such cases, HSBC Global Asset Management uses the flow information to calculate various data points such as the median, the maximum and the ninety-fifth percentile historical outflows, says Michael Hall, global head of risk transformation and innovation at the firm.

If fund managers also know what share of the fund is held by each of the biggest investors, they can sharpen their redemption scenarios. For example, if one investor owns 30% of the fund, then the chances of a 30% redemption in one day might be higher than if the biggest investor holds only 5%.

Information about the types of investors in the fund – whether they are insurance companies, retail investors and so on – allows managers to estimate the “relative stickiness” of their investment, as Hall puts it. Institutional and retail investors react to market and business events at different speeds and sometimes in different ways.

“The more data that is available, the more factors can be incorporated into the modelling process to give, hopefully, more insightful metrics with a greater degree of confidence,” says Hall. “Where we have fewer data points, we then switch to a different estimation model.”

The equation becomes more complicated where retail investors buy a fund through a platform. Managers may need to clump these investors together as a single combined entity.

“If the platform is taking you off its recommended list, it could have the same effect as being a single investor,” says Derbhil O’Riordan, Dublin-based partner at law firm Dillon Eustace advising investment funds.

Predictive power

Data on inflows and outflows is available to all fund managers – even those selling funds through intermediaries, as Woodford did – since distributors pass on buy and sell orders for managers to fulfil. But available historical data sometimes doesn’t go back far enough. What’s more, “the predictive power of historical redemption data is quite limited”, says David Regan, global head of financial risk at Janus Henderson Investors.

“History doesn’t necessarily predict anything,” Regan says. “There are no trends. Redemptions can be just simple asset allocation within another firm and you’re not going to be able to predict that.”

Esma says in its guidelines that liquidity stress-testing “should not overly rely on historical data, particularly as future stresses may differ from previous ones”. In addition to historical scenarios, fund managers should employ hypothetical ones, it adds.

What is more useful is looking at past redemptions and purchases in relation to market events, such as general market downturns, sector-specific issues, changes in market volatility or events in peer funds such as suspensions, Regan suggests. 

“That’s probably the best indicative power we can have. Certain market events in certain areas of the market can predict flows,” he says. Managers may be looking at how the current high volatility affects inflows and outflows.

 

Germany’s financial regulator made a similar observation in a 2018 article. Investors’ redemption behaviour can vary “hugely” depending on their type, investment horizon and portfolio or tax situation but in certain market phases, in particular when there are extreme fluctuations in prices, they can behave in “unexpectedly similar” ways, Bafin wrote.

The availability of investor data other than buy and sell orders is, at best, ad hoc as distributors are under no legal obligation to pass on any information to fund managers. 

Risk.net contacted four fund distribution platforms. One platform, which targets only retail investors, says it does not supply any information to fund managers apart from flow data. The other three – Allfunds, Credit Suisse Fund Lab and Aegon UK’s Cofunds – did not respond to repeated requests for comment.

As for asset managers, what they know about their end-investors varies from firm to firm.

“We will often see the largest investor in our fund is a distribution platform,” says Regan at Janus Henderson. “That’s the biggest data challenge.”

A large US-based asset manager says it receives some information from its fund distributors, mostly broker-dealers. They disclose the identity of any individual or institutional shareholder whose holding exceeds 5% of the fund – but only if the shareholder has a segregated account at the distributor, says a senior risk professional at the asset manager. Often, end-investors’ holdings across the industry are pooled in omnibus accounts and in such cases the US firm does not know who the large investors in its funds are.

Omnibus, or nominee, accounts pose a challenge for HSBC, too. “For some of our funds the availability of underlying investor information is limited – for example, we don’t know the type and number of investors – because it’s often hidden away in nominee structures. This is a challenge the whole industry faces,” says Hall.  

Out of the shadows

It is not clear why intermediaries do not routinely provide data on end-investors, especially since what fund managers need is not clients’ personal details or company names but their broad characteristics such as the size of their holding.

O’Riordan at Dillon Eustace speculates that the information may go out of date very quickly as flows change, and intermediaries do not want to be responsible for keeping it up to date.

Two influential industry bodies are pushing to force distributors out of their secrecy. In a January report, the European Fund and Asset Management Association and the International Capital Market Association wrote: “For the purpose of improved risk management, we believe that the communication of basic information to fund managers including at least investor profiles and shares/units held by these categories of underlying investors should be mandatory and free of charge.”

As things stand, fund managers will have to get by with the data they have. Esma’s guidelines do not explicitly mandate specific details to be used but instead provide examples of the kind of information on investors that could be incorporated into stress tests – and Karlsson emphasises the word ‘could’.  

A stress test based on assumptions is better than no stress test at all

Patrik Karlsson, Esma

However, firms still must meet the overarching requirement. Esma states: “Managers should understand the potential risks associated with the fund’s investor base and be able to demonstrate that those risks play a material factor in the ongoing liquidity risk management of a fund.”

The difficulty for many funds will be to develop meaningful stress tests without basic facts on the investor base. Karlsson suggests the regulator is prepared to offer some latitude on this front.

“It may not be possible for collective investment schemes to know fully who the underlying clients are. And in such cases, some assumptions should be made in order to conduct stress tests,” he says. “A stress test based on assumptions is better than no stress test at all.”

Assumptions on a fund’s investor profile can be made by looking at the fund’s type, strategy and distribution channel, says Mark McKeon, global head of investment analytics at State Street’s Global Exchange, which provides data and analytics services for clients of the bank.

To obtain more reliable or granular information, fund managers may have to sharpen their deal-making skills. What investor data is passed on should be the subject of the initial commercial negotiation between the fund management company and the distributor, says O’Riordan. In other words, funds shouldn’t shy away from pressuring distributors to pass on investor info as part of the deal.  

As well as a push factor for distributors, there is a pull factor for investors to disclose information. Large investors who purchase funds through intermediaries have an incentive to keep fund managers informed about their redemption plans because they too may be disadvantaged by high transaction costs in the event of sudden large redemption requests.

Testing ground

Fund managers can use additional data, such as investor type, to improve the modelling of stress scenarios. HSBC is going one step further. The asset manager is developing complex reverse stress tests, which take a hypothetical outcome – say, a fund’s failure to meet a set level of redemption requests – and work backwards to see what combination of factors would trigger that outcome. Factors could include redemption rates, liquidation volumes and market impact costs.

Payden & Rygel, a US-headquartered investment manager that manages $5.6 billion in Ucits funds, takes a different approach. The firm runs regular stress tests to ensure that each of its Ucits funds can be redeemed in full within approximately 10 days in a number of scenarios. The scenarios assume normal market conditions or moderate stress, says Robin Creswell who heads up Payden’s London-based business.

“Ucits rules imply that in a normalised market it should be possible to redeem a fund in full within a 10-day framework without materially impacting the net asset value of the fund beyond the underlying market value of the securities being sold,” he says. 

This means the manager must aim to avoid any market impact from hurriedly selling the fund’s holdings. The existing rules provide the option to limit redemptions to no more than 10% of a Ucits fund’s net asset value on any one dealing day.

At times of extreme stress, the Ucits directive states that a fund may temporarily suspend redemptions “in exceptional cases where circumstances so require and where suspension is justified having regard to the interests of the unitholders”. One such scenario would be drying up of liquidity in certain derivatives where it is expected to return after a few days, according to an example given in a 2019 legal guide European Financial Services Law.    

Given the complexity of liquidity stress-testing, the timeline for implementing Esma’s guidelines is “quite short”, says McKeon at State Street.

Many asset managers plan to complete the work by the end of June so they can show the output of the stress-testing models to their boards and other relevant parties, he says. 

Although the requirements come into force on September 30, in some countries they may be implemented after September if they require changes to national legislation that cannot be completed in time, says Karlsson at Esma. 

The Central Bank of Ireland, which oversees funds in Europe’s second-biggest hub for Ucits, tells Risk.net it is currently considering how to incorporate Esma’s provisions into Ireland’s regulatory framework and expects that fund management companies will comply with the guidelines from September 30. 

The CSSF in Luxembourg, Europe’s Ucits capital, did not respond to questions from Risk.net, while the UK’s Financial Conduct Authority declined to comment.

Ready, steady, model

Apart from the data gaps, there is little regulatory guidance on how to perform liquidity stress simulations for funds, as Esma acknowledged a few days after publishing its new rules. The regulator did, though, offer a 37-page paper containing a stress simulation framework dubbed STRESI. The paper also outlined a range of modelling options and guiding principles that can be used to customise simulations both for assets and liabilities.

Esma’s advice for scenarios of pure redemption shock included using individual fund data rather than aggregated data by fund styles, using individual fund net flows and opting for expected shortfall rather than value-at-risk to calibrate the shock. 

The Esma paper also discussed possible second-round effects. Asset sales by the fund after the original wave of redemptions can reduce the remaining assets’ price, triggering further outflows and so further liquidations.

The simulation can be run again to include possible third-round effects, Esma wrote. But it noted: “The predictive power of third-round effects declines dramatically and therefore does not add robust and reliable outputs.”

Is there an app for that?

State Street’s truView software includes a tool that satisfies Esma’s new liquidity stress-testing guidelines, both on the asset and liability side, says Mark McKeon.

The tool calculates a fund’s redemption coverage ratio by comparing the asset and liability profiles, and also allows clients to conduct fund-level reverse stress tests.

The truView team is experimenting with gathering more granular investor information, including investor concentration and type, says McKeon. The idea is to source the data on a particular fund from third parties, such as transfer agents, and pass it back to the client that manages the fund. The data then may be used to adjust the fund manager’s redemption scenarios.

State Street will decide whether to progress with this initiative towards the end of the second quarter, depending on client feedback.

The StatPro Revolution software already allows asset managers to run simulations on their portfolios with customised levels of redemptions, but the vendor is working on enhancements to help clients better understand redemption risk. 

New features on the liability side, which StatPro plans to release in the second quarter, will enable asset managers to input data such as fund flows or investor type, and answer questions such as: what level of redemptions by the largest clients would not allow the fund manager to meet the request.

Earlier this year, MSCI built modelling tools that can incorporate a fund’s redemption history, investor concentration and investor type, and is already using this tool to set stressed redemption scenarios for existing and prospective clients.

The firm sources the information either from the fund managers themselves or third-party data vendors. 

Bloomberg doesn’t offer tools for redemption modelling. Instead its liquidity tool focuses on modelling the market liquidity of each portfolio’s assets. 

The vendor notes a recent survey it conducted at an event for managers of Ucits funds. Bloomberg asked them how they plan to model redemptions to comply with Esma’s liquidity stress-testing guidelines. A majority planned to use a preset list of redemption percentages – for example, 1%, 2%, 5% and 10% – and/or amounts based on the largest outflow observed for the fund, while only a few intended to build a model for investor behaviour.

Correction, March 12, 2020: Due to an error in the editing process, an earlier version of this article incorrectly stated that Janus Henderson Investors is actively plotting fund flow data against wider market events for predictive purposes. The company only suggests that such a technique could be effective.

Editing by Alex Krohn

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.

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