Enterprise risk analyses, such as stress testing and scenario analyses, are increasingly popular with institutional investors. Alternative investments, such as hedge funds, private equity and real estate, present challenges to enterprise risk analysis.
In this article, BNY Mellon’s Global Risk Solutions and HedgeMark Risk Analytics, a BNY Mellon company, explore these issues and offer examples of possible solutions.
Key observations and insights to best practices include:
To view the full paper, please click here.
Enterprise risk analysis defined
Risk management can mean different things to different people. Enterprise Risk can include all the many aspects of risk that affect an organisation including but not limited to market risk, reputational risk, regulatory risk, compliance risk, operational risk, and legal risk.
Enterprise Risk Analysis can also mean forward-looking (“ex-ante”) risk calculations that estimate investment risks across multiple asset classes. This paper focuses on the definition of Enterprise Risk Analysis that estimates ex-ante risks of an investment programme with multiple asset classes owned by a single organisation, such as a pension plan or a charitable foundation. Ideally, this risk management discipline should be a component of a larger “enterprise-wide” risk management framework that also considers risks other than investment risk.
Over the past decade, institutional investors have become more aware of the importance of risk management across asset classes. Since the advent of modern portfolio theory, portfolio managers have understood risk to be important in selecting and monitoring investments within a single portfolio strategy. Evaluating risk across investment strategies has become common practice more recently. It used to be that reviewing a few risk measures based on the volatility of monthly return streams such as a Sharpe Ratio or Information Ratio would be considered an effective risk management process.
With the continuing increased focus on risk by regulators and other stakeholders, many institutional investors require a more comprehensive understanding of how risk operates across investments within an entire investment programme. Some regulators require reporting pertaining to stress testing and scenario analysis, such as through Form PF for US investment advisers to hedge funds, and pursuant to Solvency II for insurance companies and UCITS for European investment funds. These new risk reporting requirements have generated the need for sophisticated risk calculation tools and services to help address these changing requirements. Many firms are establishing a separate chief risk officer function to supplement the risk management responsibilities inherent within the investment management functions. These new risk management functions require a significant amount of data to enable evaluation of investment risks across asset classes.
Exposure and volatility
In considering investment risk, two common approaches to risk management are based on the definition of risk as return volatility or the definition of risk as exposures.
Exposure-based risk analysis identifies the relevant risk factors for each investment, enabling the preparation of summaries of portfolio allocations to various categories of exposures or factors, or investment characteristics. Investment strategies or broad asset classes have characteristics that may be more relevant to one strategy than another. For example, duration is a key risk characteristic for fixed income investments but less relevant for common stock. Some characteristics can be relevant across asset classes. Credit ratings are often associated with bonds but may also be associated with the issuers of common stock. Economic sectors are most often associated with common stock, but corporate debt and private equity investments can also be associated with economic sectors.
Currency or country exposure is relevant for most investments, and liquidity is also a consideration for all investments. Each of these measures provides insights into different aspects of investment risks, but the estimation and aggregation of these risks requires transparency into portfolios.
A covariance matrix enables an understanding of the volatility of the total fund relative to the correlations of each of the component asset classes and sub-strategies or investments. This approach enables an understanding of which investments contribute to the total risk of the fund and which investments help off-set some of the risks of other investments.
The covariance matrix is generally based on historical returns to estimate how investments move in relation to each other, so ex-ante risk is also based on ex-post returns. Often, risk systems generate covariance matrices based on several months or several years of daily returns of each of the specific securities. Some risk systems also calculate covariance matrices based on monthly returns of the investments.
Covariance matrices based on daily or monthly returns make sense in evaluating the relative risk profile of publicly traded liquid investments that are priced frequently. Many institutional investors have been increasing their allocations to alternative investments such as private equity, real estate and hedge funds, which often do not provide sufficient liquidity or transparency to facilitate ex-ante analysis. Real estate and private equity investments are often priced quarterly, so the apparent risk of these investments based on daily volatility alone could underestimate the actual risk of losing money in the event an investor wanted to sell. The typical return profile of private equity includes an initial period of negative returns followed hopefully by a prolonged period of growth, resulting in an initial public offering and pay-outs to investors. This return profile is appropriately measured by long-term internal rates of returns rather than independent monthly or daily time-weighted rates of return. Private equity investments will often provide transparency into the underlying companies, but these companies are not valued frequently and detailed cash flow information generally is not provided for specific underlying companies. Real estate investments often provide transparency about underlying investments as well on a quarterly frequency.
Hedge funds may be constructed with liquid or illiquid investments, but hedge fund managers often share limited information periodically with investors. Hedge fund managers that provide descriptions of exposures and characteristics in manager letters often define summary categories differently. For example, one manager’s definition of Europe might include the United Kingdom and all countries in the European continent, while another manager might define Europe to mean only those countries in the European Economic and Monetary Union. Investors with large allocations to alternatives often spend a significant amount of resources and time finding ways to normalise summary data so exposures can be aggregated across managers and asset classes.
Approaches to incorporating alternative investments into enterprise risk analysis
Institutional investors that want to incorporate alternative investments into ex-ante risk analysis have to make some assumptions. These assumptions come with different levels of data management requirements that could lead to different levels of confidence about the reasonableness of the potential conclusions.
Third-party risk aggregators
Where detailed holdings are available, taking the time to incorporate the most detailed information enables the most flexibility in viewing different slices of data, and the most confidence in the accuracy of the calculations. This is particularly important with hedge funds or any portfolio containing derivatives, because derivatives may move in non-linear ways that can only be captured if the detailed terms and conditions of the derivatives have been modeled. Where hedge funds may not be willing to provide full transparency to investors, they may be willing to provide transparency to third-party risk aggregators using non-disclosure agreements. These risk aggregators can then provide calculated risk exposures using the underlying holdings to the investors without revealing the underlying holdings in detail.
Where detailed holdings are not available even to third-party risk aggregators, managers will often provide summary exposures. Where detailed holdings are available but the investments are illiquid, or where summary exposures only are available, it is necessary to proxy these summary exposures with relevant indexes that represent relevant factors of interest to the investor.
Some investors with smaller allocations to alternative investments may choose to assume that an entire portfolio or even asset class is not worth the trouble of estimating detailed factor exposures. In this case, it is possible to proxy an investment or asset class to an index with various factor adjustments that is intended to mimic the performance of the asset.
Private investments present a problem in calculating ex-ante risk. Private investments are rarely valued on a daily basis, and funds generally report their underlying holdings on a lagged (often quarterly) basis. To understand the risk of the entire fund we need to incorporate the private investments in some manner. The typical approach is to proxy the funds to an equity index – although lagged valuations, the difficulty of valuing private holdings, and the infrequency of the valuations generally lead to the conclusion that public and private equity are not highly correlated given traditional correlation methods.
Public market proxy for private equity
Intuitively many investors consider private equity to be more like public equity than anything else to which it can be compared. In absence of greater knowledge about the private investments, utilising a public market proxy makes sense. Many investors do get additional information such as sector and country on private equity holdings. With that additional information more granular analysis can be done. From an exposure standpoint, being able to incorporate the private equity sectors with the public sector does provide a more complete exposure picture. With the additional information we can also proxy individual sectors to a public market sector benchmark. The assumption is that a private equity information technology stock will be more like a public information technology stock than the broad index.
So does it make a difference if you proxy against a single index or go deeper and proxy against country/sector indices? To explore this question, we analysed an actual private equity investment programme for an institutional investor that uses a detailed proxy process. The private equity portfolio consisted of over 70 limited partnerships with hundreds of underlying positions. Utilising the look-through capabilities of the Burgiss Group’s Private Informant service we receive each limited partnership’s prorated shares of each private (as well as public) company that is held by the partnership. Any cash held in the portfolio is also captured. Each company is assigned a sector and a country. We then proxy each company to a similar sector/country or region public market index. We compare that detailed sector/country proxy analysis to an alternative model using a high-level proxy for the entire private equity portfolio. Fig.1 clearly indicates a different VaR and conditional VaR for the two different approaches.
Sample real estate portfolio
We also went through the same process with our sample real estate portfolio. In this case we only went down to the country/region level for the more detailed process. Once again we compared the same portfolio to a single real estate public market index. As Fig.2 indicates, even taking the minor step of getting the proxy in the correct region produces a different risk number.
This sample real estate portfolio is 90% US; thus, the impact to expanding the proxies to countries or regions is not a significant change. Expanding the detailed proxy process to real estate types such as office, residential, etc. presumably would highlight the same diversification benefits we saw with private equity.
The results of this exercise have shown a more detailed proxy methodology for private investments does produce different risk statistics. In this analysis the calculated risk is less for the more detailed proxy methodology than the single equity market index. The diversification of the different benchmarks as well as the inclusion of cash in these funds lowers the risk statistics.
TOTAL FUND ANALYSIS
In order to evaluate the impact that different levels of transparency can have when looking at the total plan (e.g., including all of the plan assets including traditional stock and bond funds with alternative assets), we created two synthetic plan allocations.
A sample Conservative Portfolio has allocations as follows:
A sample Aggressive Portfolio includes the same underlying funds but with the following weights:
We then created three different versions of each one of these portfolios, so there were six in total. The first used all position-level information for all investments including all of the alternative funds. Position-level information is used for hedge funds while private equity and real estate use the underlying company-level information which is then modeled to a factor based on the region and sector and a random beta to that factor, usually in the 0.9 to 1.2 range, but averaging around 1. The second iteration used exposure information for the hedge funds, public market information for the long positions, and a single factor (e.g., the S&P 500 for the PE/VC and REITs Exposure information being defined as investor letters, Open Protocol Enabling Risk Aggregation (OPERA) reports and the like) for other alternative positions. The third version used the same position-based funds for the long funds, but modeled all the alternatives to a single factor based on the pertinent asset class. In most cases, the incorporation of alternative investments into a total plan analysis will usually be undertaken through a combination of the above methodologies, but for the purposes of the analysis we look at each approach separately to demonstrate some of the inherent differences.
Different levels of transparency
The ability to calculate and decompose the risk and review exposures is greatest when position-based portfolios are used. Different levels of transparency impact the fundamental exposure-based reporting, as well as more sophisticated measures such as VaR-based statistics, historical stresses, and shocks. The accuracy of the underlying data becomes even more important and difficult to manage the more opaque the assets that are held in the portfolio. The collection, aggregation, and synthesis of alternative data with the traditional holdings of the portfolio have traditionally been a manual process supported by institutional investors. To do it well takes a significant amount of operational alpha and the process is often fraught with issues due to the completely manual nature (e.g., input errors, interpretive differences of staff over time, spreadsheet errors, etc.).
Evaluating our findings
One of the most fundamental ways plan sponsors continue to evaluate risk is by looking at their exposures to regions, countries, and sectors. Due to the rapid expansion into alternatives this provides some challenges. An example of this is shown in Fig.3 and Fig.4 by looking at the apparent differences in regional exposures using different levels of transparency. One common question may be what is our overall exposure to a country such as China, and what type of instruments (stocks, bonds, etc.) are included in that exposure. If full position-level transparency is available on all of the funds, this can be easily determined, but if the only information available is exposure information, country-level exposure may be limited to an estimated exposure based on the collective conversations with the relevant managers with Asian exposure. Furthermore, Fig.3 and Fig.4 show 4% greater exposure to Asia Pacific in the proxy portfolio compared to the position-based portfolio, and this difference holds true for the market value allocations and the notional value allocations. It is interesting that the exposure-based portfolios show higher allocations to Asia Pacific for both notional and market values than the position-level analysis would suggest.
Also when loading exposures or using proxies, some data is classified in a general exposure to Global or Other, because of the variety of categories and levels of detail reported by different managers. For example, Fig.3 and Fig.4 (see previous page) show the different apparent weights of the Global bucket using the proxy compared to using the position-level detail. Almost 9% of the book is in Global with the proxy analysis, but on a position basis this drops to almost zero as those instruments are classified into their respective regions. This type of impact is apparent in all topologies (e.g., sector, industry, country, instrument type, etc.) the more position-level data available.
Impact on Value at Risk
The Value at Risk statistics shown are calculated using a one-month Monte Carlo analysis at a 95% confidence level, unless otherwise specified.
In both model portfolios (Aggressive and Conservative) we see the VaR as a percentage of the market value is lowest for the position-based version and highest for the proxy versions (see Fig.5), and have a minimal impact on the Conditional VaR as a percentage of market value. This may indicate a benefit from diversification impact when detailed positions enable the analysis to capture idiosyncratic risk at its maximum level. This result appears generally consistent across all of the VaR-based statistics for the six model portfolios analysed.
Shocking the portfolio
In addition to VaR, there are other ways to evaluate the risk of the portfolio. Two of the more frequently used ex-ante approaches are looking at historical scenarios to see how the current portfolio would react if that same scenario were to replay itself and shocking the portfolio based on the user’s view of what is going to happen in the market. The treatment of the alternative assets will materially impact the results and it is helpful to consider the impact and/or limitations that can occur when using exposure and/or single factor proxies.
CLICK IMAGE TO ENLARGE
In Fig.6 are some results of a variety of popular historical scenarios. Fig.6 uses a green, yellow, orange scheme to denote loss/gain levels within the respective Aggressive and Conservative portfolios.
Further decomposing results
Historical shocks are an interesting illustration of how a current portfolio would react if that exact scenario repeated, but almost everyone will agree that history is highly unlikely to exactly repeat. There are simply too many variables in the market for them to converge again, so investors want the ability to stress the portfolio on what they think will happen next or come up with a nightmare scenario and see how the portfolio would react. This is often referred to as shocking the portfolio. The treatment of alternatives has a material impact on the results.
Using the same model portfolios we have throughout this section we stressed the portfolios assuming the follow scenario:
The position-based portfolios show less of a loss than the proxy versions in both the Aggressive and Conservative portfolios (see Fig.7). This is again exacerbated in the Aggressive portfolio compared to the difference in the Conservative portfolio. The loss in Aggressive position is 14.96% while the loss in the same Aggressive Proxy version is 18.21%; however, in the Conservative position you see an expected loss of 13.35% compared to a loss of 14.61% in Conservative proxy. This is due to the treatment of the alternative assets.
As mentioned above, the ability to drill into the results in multiple dimensions is enhanced by greater transparency.One example of evaluating this at a lower level would be by asset class or instrument type. In Fig.8 you can see that the proxy portfolios have five basic asset classes (alternative investments, cash and equivalents, derivatives, equity and fixed income). This is a typical way things may be aggregated when a lack of better transparency exists. As we go up the transparency continuum we start adding asset classes which help to better explain the risk. In the exposure buckets we add commodities as an asset class and finally when we get to position-based we add FX and other. The position-based methodology is a bottom-up approach that looks at every asset and categorises accordingly versus a top-down approach that assumes portfolios match the investor’s defined investment hierarchy. More often than not assets in the alternative investments get a broad market index proxy and sometimes these get a beta adjustment to that proxy. When assuming a broad market index or indices you lose the idiosyncratic risk that you capture when you do this at the position level and it explains much of the difference between the position and exposure portfolios. The exposure-based results fall in the middle due to the diversification impact (using multiple factors with respective weights per fund) achieved versus using a single proxy for investments without any level of transparency.
CLICK IMAGE TO ENLARGE
Further analysis could consider sectors, duration bands, or other categories relevant to the investments. One common way to look at this is in historical stresses, sensitivities or VaR-based reports to compare a bottom-up approach to a top-down approach. Often this will highlight an exposure to an asset class that was not intended. A great example of this occurred in 2008 when well-diversified investors were unaware of their exposure to the equity market. Hedge funds and other alternatives were often included in an absolute return bucket and investors were surprised by their exposure and the corresponding losses. In fairness, when correlations go to one and there is a liquidity crisis there may not be a safe place to invest, but better information might have helped them avoid some of the surprises with the losses incurred.
CONCLUSIONS AND BEST PRACTICES
As the examples above illustrate, the assumptions that are used in incorporating alternative investments into enterprise risk analysis matter. Different approaches to data management can lead to different potential conclusions about the apparent risks within an investment programme.
The analyses discussed earlier, and discussions with institutional investors about their experiences in using enterprise risk analyses, suggest the following best practices:
Risk management as a function is necessary whenever there is the desire to take and understand risks. The practice of risk management has changed significantly over time, but it continues to be a discipline without definitive standards. The best practices discussed herein are intended to help others address the challenges presented by opaque or illiquid alternative investments.