A TOTAL ENTERPRISE APPROACH TO ENDOWMENT MANAGEMENT

By Mark A. Schmid, VP & CIO and Que Nguyen, MD of Strategy – The University of Chicago

Introduction

The crisis of 2008 and the ensuing losses in risky portfolios, including endowment portfolios, continue to affect the financials of universities, despite the rebound in asset prices over the past two years. Although the smoothing inherent in most payout formulas dampened the immediate impact of investment losses, it also spread out the pain over a longer time period.

Endowments have always been and continue to be institutions with a long term outlook for investments, but universities do have immediate financial obligations in their day to day operations. Hence, the singular focus on long term investing had the unintentional impact of shifting much of the burden of bearing short term volatility onto the operational side of the university.

In order to balance these risks more equally between operations and investments, the University of Chicago set out to examine our long term investment strategy from a total enterprise perspective.

The first step of a total enterprise approach was a deeper integration and communication between the Investment Office of the University and Finance and Administration. This included Investment Office involvement in budget planning and review, liquidity and capital resource planning, and coordination with the administration on operating plans.

As the University was embarking on an ambitious long term growth plan, including the development of new programs, involving increasing faculty and capital expenditure, the coordination efforts underscored the need for the different parts of the University to have realistic, long term plans in place.

In support of these efforts, the Investment Office launched a project to develop a well structured framework for our investment strategy, which takes into account the particulars of our University, such as growth, debt, and wealth.

While many have likened the Total Enterprise approach to the asset-liability framework often used by pension plans, we found that the problem was far more complex. Pension plans have a well defined liability imposed by regulatory and accounting practices. Universities are largely free from such regulation, and as such our liabilities are much less defined.

A university’s risk is both much different as well as more varied than that of a mature pension fund, where the liability risk is dominated by interest rates. Furthermore, the variety of investments in endowment portfolios creates challenges in assessing high level economic risk, as well as translating that high level risk budget into an implementable portfolio strategy. We now turn to a discussion of how we approached this project.

TEAM Discussion

In the beginning of 2010 we launched an initiative to evaluate the overarching investment strategy and risk taking of the endowment in the context of the overall University. This project, called Total Enterprise Asset Management (TEAM), sought to frame the investment strategy of the endowment in the context of the long term operating goals and risks of the University, rather than as a stand-alone, total return fund.

In marrying the asset (the endowment) and the liability (the University’s operating goals) sides, we found that the problem was large and complex, and needed to be reduced to a set of well understood, common economic drivers that could be evaluated. For example, if we believe that GDP growth influences both investment returns and growth in Grants, how can we measure that?

In this context, the TEAM approach had several sub-components to achieve project goals. These included developing a fundamental economic model of risk expected returns, economic analysis of the University’s operating exposures, and then marrying both in an internally consistent Monte Carlo simulation to determine the trade-off between risk taking and wealth accumulation for the University.

Risk and Liquidity Modeling

The “endowment model” of investing emphasized capturing illiquidity premia with the belief that illiquid categories added diversification benefits to a portfolio. However, in tail events such as 2008, much of these diversification benefits evaporated at a time when liquidity was challenged. To monitor and develop a deeper understanding of risks in the endowment portfolio, we use a Factor Risk Model.

We extensively map detailed portfolio holdings to 90 public market proxies, and the model then calculates factor exposures and a volatility estimate for each fund and the portfolio in aggregate.

Currently, Staff monitors exposures to U.S. equities, emerging markets, credit, real estate, commodities, interest rates, and inflation factors, in addition to the global equity factor. The Global Equity Factor (GEF) has become our primary risk governance target, as equity factor risk accounts for over 90% of the volatility the portfolio experiences. The table on this page summarizes our factor exposures as of October 28, 2011.  

(See Figure 1 below) Fig. 1 Factor Risk Model

Factor	BETA	Total BETA
–	–	–
Global Equity Factor	–	0.75
U.S. Equities	0.27	0.73
Emerging Markets	0.18	0.47
Credit	0.15	0.89
Real Estate	0.10	0.43
Commodities	0.11	0.47
Inflation	0.04	0.65
Interest Rates	0.09	-0.22

Footnotes 1. Multiple-regression beta. Answers the question: “If there is a 1% move in a factor and that move is uncorrelated with the moves in the other factors then what will the percent move be in TRIP?” 2. Single-regression beta. Answers the question: “If there is a 1% move in a factor and all the other factors move by their usual amount in conjunction with the event that caused the factor move by 1%, then what will be the percent move be in TRIP?”

We update this information weekly and distribute to the Investment Committee and senior members of the University Administration. This weekly report provides valuable look-through information that more precisely describes our risk exposures based on true economic risk drivers, as compared to a more traditional assessment of risk based on asset class classifications. We continuously review this model and enhance it as the market and portfolio evolves.

In the past year, we have built several improvements into the model, including: integration of the GEF beta; expansion of the mapped index population; improvements in the mapping process; incorporation of fund-level debt into levered betas; movement to more robust data & regression methodologies; introduction of a new data warehouse; development of a new robust risk architecture to support complete rewrite of the model/ risk engine as an automated modular solution with error handling and a suite of risk diagnostic tools, to replace the initial brittle spreadsheet model; ability to produce new or ‘alternate-view’ risk measures in production; new suite of derivatives’ analytics and risk metrics; and created a full-revaluation ‘fat-tailed’ VaR engine.

In addition to market risk, we have also developed an integrated liquidity model for the total endowment pool. Using this model, we project ten years of monthly flows by incorporating the most up-to-date information available for cash flows, market valuations, and redemption terms.

The asset-level liquidity model consists of an illiquid drawdown model for private investments and a redemption model for hedge fund and liquid investments. Return assumptions are examined via scenario analysis, providing to a deep understanding of the potential range of future asset allocations and liquidity, as well as expected returns.

Expected Returns Framework

In the same way that we sought to understand the economic drivers our risk position, we sought to understand the fundamental drivers of investment returns. Long term expected returns had always been based on looking at long histories of asset class returns and projecting them forward. While this approach is probably sound for an economy and capital market experiencing little dislocation, it can lead to very misleading results if we are starting from a point of extreme dislocation in the economy and markets.

For example, over the past 30 years, bonds have returned 11.5% per year. However, with yields on 10-year Treasury’s now hovering near 2%, such an expectation would seem ridiculous. Comparatively, stocks have returned 10.8% over the past 30 years. However, 30 years ago, the trailing P/E multiple on the S&P 500 was 8x, compared to the 15x it currently represents, and contributed an annualized gain of more than 2% per year to returns. Unless we expect P/E multiples to close to double in the future, this 2% component of past returns is not repeatable.

Our proprietary expected returns model is designed to evaluate the underlying economic drivers of long term returns for a variety of asset types. Here, we show the important drivers for equities and bonds.  (Figure 2) Fig. 2

Expected Returns Framework

As can be seen from the graphics above, fundamental drivers of returns include GDP growth, inflation, and risk premia. At this stage of the project, we chose to focus on two major asset categories, Equities and Bonds in order to simplify the problem of integration with the University financials. Our risk analysis of the endowment portfolio had found that the dominant risk factor in endowment was an equity factor, with a secondary factor being a “safe assets” or bonds exposure. This had become more pronounced in the financial crisis, as the correlations of a variety of risk assets moved closer to 1, and even post crisis, remained elevated.

A realistic assessment of the benefit to equity risk taking is crucial in properly evaluating how much risk taking is warranted. For example, if equities are expected to outperform bonds by 4% per year, much more risk taking is warranted than if equities are only expected to outperform bonds by 2% per year. As our strategy of investing in private structures and hedge funds had added significant alpha to returns over time, we also modeled in the returns and “alpha” that was aimed to capture the benefits of manager selection, liquidity premia, and other endowment-style advantages.

Economic Modeling of University Financials

The next step entailed modeling University financials with respect to economic drivers consistent with our expected returns. In using the University financials, we focused on examining figures from FY 1992 to 2009, as we believe the accounting practices since then as well as the operating of the University since that time is closer to current practices than pre-1992 periods. In some cases, such as evaluating compensation, we extended the dataset further back, to the 1970s to better capture the effects of inflation. In modeling the University financials, we simplified the problem by focusing on the broad categories of income and expenditure sources as follows:  

(Figures 3 & 4)

Examples of relationships we had found include a positive relationship between Gifts and Equity market returns, Grants and GDP Growth and Deficit Growth.  We also found strong absolute growth trends in Net Tuition and Compensation, without much inflation impact.

Integration of Investments and University Fundamental Drivers in Monte Carlo Framework

While the main economic drivers of Investments and University financials can be identified, they are by no means deterministic. For example, we know that there is a strong relationship between GDP growth and Grant growth, any given year can underperform or outperform the central relationship if the University has interesting project ideas or successes.

Thus, in marrying the asset and liability side, a significant amount of uncertainty must also be incorporated in the evaluation process in order to properly trade off risk taking and wealth accumulation at the University. In doing so, we turned to Monte Carlo methods.

Our approach to Monte Carlo for TEAM incorporated two important features. The first feature is that the Monte Carlo incorporates some mean reverting features not always considered in financial modeling, but is more reflective of economic reality. Secondly we sought to create internally consistent scenarios for the economy, the University and the investment markets, rather than model them separately. The graphic on the following page illustrates this integrated approach within our Monte Carlo: (Figure 5) Fig. 5

Drivers in Monte Carlo Framework

To this end, we started with central scenarios for the economy and incorporated the long term capital and growth plans for the University. Some of the key underlying variables include the following:

Key Economic Variables

◆ Inflation

◆ Real GDP Growth

◆ Profitability relative to GDP

◆ P/E Multiples

◆ Real Bond Yields

Key University Variables

◆ Net Tuition Growth

◆ Gifts

◆ Grants

◆ Auxiliary Income

◆ Compensation

◆ Supplies & Other

In generating economic scenarios, we used a bootstrap of historical experience since the 1950s to create realistic scenarios of the evolution of the economy. We also added random noise to simulate the uncertainty of the relationship between the University, investment markets, and the economy.

Our simulation involved 1000 scenarios over 20 years. Depending on the amount of equity risk in the endowment portfolio, the University wealth outcome at the end of 20 years varies widely. At the end of the 20 years, we evaluated several financial metrics and related them to the amount of equity risk taking in the endowment portfolio. Any number of financial metrics can be examined, either on the low or high end. We chose to examine the metrics characterizing the risk of an undesirable financial outcome to the University. The higher the probability of such outcomes, the more likely our operating goals would need significant adjustment (called “off-ramps”).

These included the probability that:

◆ Endowment falls below Restricted Endowment adjusted by inflation

◆ Endowment falls below Restricted Endowment adjusted by GDP growth

◆ Ratio of Real Endowment (i.e. inflation adjusted) to Faculty falls more than an acceptable level

◆ Ratio of Expendable Endowment to Debt falls below an acceptable level

Additionally, we examined the expected accumulated wealth of the University vs. the one year drawdown of the endowment portfolio. A greater likelihood of a significant one-year drawdown represents a greater level of operational risk to the University.

The graphs below illustrate this exercise.  (Figures 6 & 7) Fig 6

Probability of Significant Off-Ramps vs. Equity Exposure

 Range of Endowment Balance and Probability of Drawdown: Year 20 Fig. 7

The top graph clearly shows that, given the University’s growth liabilities, there is a benefit to owning growth oriented assets (i.e. equities). However, beyond a beta of 0.6-0.7, the downside risks no longer decrease as the volatility of equities begin to offset their growth benefit to the University. Read differently, this graph seems to say that the University does not “need” more equities than a 0.7 beta would imply.

The red line in the second graph shows that the University can benefit from owning more equities because of the long term wealth accumulation benefits. The blue line shows the risk of a significant wealth drawdown over the course of one year, and includes the impact of the payout in addition to the market returns. The blue line is higher for an all fixed income portfolio, because bonds currently yield so much less than a typical payout.

Adding equities decreases this risk through growth benefit up to a point. At higher and higher levels of equities, drawdown risk increases more quickly, while incremental wealth accumulation shrinks. Beyond a 0.7 to 0.8 beta, the incremental wealth accumulation is small relative to the increase in short term drawdown risk for the endowment.

Qualitative Risk Assessment

To complement our quantitative framework, we also engaged in a qualitative assessment of University risk profile. While the quantitative approach of the Enterprise Model provides the foundation for strategy, a qualitative approach supplements the model by allowing the incorporation of non-quantified considerations.

During a strategy status update in the November 2010 Investment Committee meeting, it was suggested that the Investment Office consider the risk profile of another well-defined institutional investor to allow for a robust assessment of risk. It was concluded that a large pension fund could be a relevant comparison since endowments used to be managed at a 60/40 risk profile. Many mature large pension plans today target 55-60% global equity risk (GEF); whereas, large endowments have migrated to an 85-90% global equity level over the past 10-20 years.

An example of a reason universities can take more equity risk is that we have the ability to reduce costs and capex, whereas defined benefit plans have little leeway in reducing promised benefits. An example of a reason universities should take less risk than a pension plan is that Universities can’t issue equity to fund shortfalls, similar to corporations.

Selection of Investment Risk Profile and Illiquidity Budget

After a thorough discussion of the quantitative analysis, the qualitative assessment, and a review of peers, the Investment Committee decided to have a long run, central tendency global equity beta of 0.75, with authority to vary between 0.7 and 0.8. Additionally, we chose a long-term illiquidity target of 35%, including private investments and sidepockets. We viewed this as being a sensible position which balances the desire of wealth accumulation with appropriate levels of institutional risk taking.

Strategic Asset Allocation

Having chosen our high level risk posture of global equity exposure and illiquidity budget, we then set out to form a sensible portfolio. For example, how much of our global equity exposure should come from private equity as opposed to real estate?

In this task, we first expanded our expected return model to cover more asset classes, including Global Equity, Private Equity, Real Estate, Distressed Debt, Absolute Return, Natural Resources, Fixed Income, and Credit.

We also added a new investment category, called Portfolio Protection, encompassing tail hedging strategies that seek to benefit disproportionately when markets are volatile to the downside, and budget for small, contained losses when markets are more stable and rising. While many institutions embed such strategies within a hedge fund portfolio, we chose to make the category an explicit capital allocation commitment for the sake of transparency and clarity from a governance standpoint.

Given a set of expected returns and a risk model, it is tempting to simply construct a mean-variance efficient frontier. However, efficient frontiers are most meaningful when the curvature is steep enough to distinguish between the risks and returns of portfolios on the frontier. As constraints are imposed, the available efficient frontier shortens and flattens. The graphic (figure 8) illustrates this principle. Fig. 8

Efficient Frontier: Normal Economic Scenario

After imposing all our constraints, the available efficient frontier in green has a minimum risk point of 13.5%, and maximum risk point of 15.5%, and the difference in expected returns is 0.7% per year. The differences in portfolio composition can be significant, but the expected outcomes are not meaningfully different. Hence, in choosing a portfolio, we should consider alternative economic scenarios which may occur over next decade.

For example, our risk model, liquidity model, and expected returns model were all based on assumptions that the economy return to a “normal” state of growth and inflation over the next decade. However, in studying history, the last 10 years was a period of below normal growth accompanied by high volatility, while the 1990s were a period of above normal growth accompanied by low volatility. We simplified potential economic scenarios into four categories:

◆ Normal growth a period of moderate inflation and real growth exceeding debt growth

◆ Stagflation a period of high inflation and real growth less than debt growth

◆ Deep Recession/Debt Deflation a period of zero to negative inflation with debt exceeding real growth

◆ Innovation a period of low to zero inflation with real growth far exceeding debt growth

In studying the history of financial markets, we can understand the impact each scenario will have on asset price fundamentals, and hence investment returns. For example, in an Innovation environment, we would likely witness strong GDP and EPS growth along with elevated P/E multiples for equities. We also looked at the risks and challenges each environment might pose to the University: (Figure 9) Fig. 9

Economic Scenarios

We developed return and risk expectations for each economic scenario. Within each economic scenario, we formed maximum return portfolios with 0.75 beta to GEF and 35% illiquidity constraint. We found the following to be true of the portfolio in the four scenarios:

◆ Inflation Real assets will be preferred

◆ Growth Private Equity will be preferred

◆ Deflation/Recession Long bonds and Portfolio Protection will be preferred

◆ Balanced/Normal Absolute Return will be preferred

Each portfolio will have similar outcomes in a normal environment, but they have more differentiated outcomes in other scenarios. No single portfolio will be the “best” for all environments. Thus, the appropriate portfolio should take into account both the likelihood of economic regimes as well as the vulnerability of the University in each regime.

In our analysis of the types and magnitudes of risk to the University in each scenario, we were interested to find that the University is much more vulnerable in a lingering deflation environment than in a stagflation environment. In a deflationary environment, higher economic and political uncertainty creates challenges in maintaining gifts and grants growth.

The need for financial aid grows as tuition growth becomes more constrained. Although the nominal cost of debt decreases, the real cost of debt rises when growth in all sources of income become more difficult. By contrast, in a stagflation environment, universities have been able to raise tuition above the rate of inflation while constraining compensation growth below the rate of inflation due to a poor labor market environment. At the same time, the real cost of debt declines particularly for universities with long dated fixed rate debt.

A stagflation environment may not be enjoyable, but by no means will it do as much harm to a university as a deflationary bout. Thus, in assigning probabilities to each economic scenario, we overemphasized the deflationary risks, as we felt the pain in such an outcome would be more severe than in a stagflation outcome. Based on this analysis, we were able to form a portfolio that integrated total enterprise and factor risk budgeting within a traditional asset class framework.

Summary and Implications

The TEAM project was completed over the better part of 18 months and has changed the governance of our investment process. Here we discuss three important implications of our chosen path:

◆ Focus on factor based risk management Endowments have traditionally de-emphasized risk management in the belief that risk matters less in the long term. The TEAM approach puts risk monitoring and budgeting at the center of our governance structure. While we continue to pursue strong returns, we must do so without taking on excessive risk to the University.

◆ Dynamic Adaptive Expected Returns Our expected returns model is fundamentally based, and thus evolves with market fundamentals. For example, if interest rates were to suddenly rise to 7%, this would lead to higher expected return for fixed income. Likewise if P/E multiples increase significantly, this would lead to lower expected returns for equities. This dynamic, adaptive framework works to center long term financial planning on a more sustainable growth path. As the past 2 years have seen very strong returns to risk assets in our portfolio, our expected returns model has led us to decrease expected returns going forward. In this way, the University does not make plans based on unsustainable market dislocations.

◆ De-emphasis of Peer Comparisons Our decision to have a lower equity and illiquidity risk posture than many large endowments creates an incomparability between our returns and those of our peers. A more appropriate comparison would require a risk adjustment of returns to similar risk levels. In adopting the TEAM approach the Board of Trustees and the University have accepted that peer comparisons are less important to the long run health of the University than tailored risk management.

The TEAM project has been a time consuming exercise for the Investment Office, and has involved the University Administration and the Investment Committee of the Board of Trustees. However, now that the framework has been built, we have the ability to revisit the analysis as conditions at the University change. While our goal remains to generate strong investment returns, we now explicitly recognize that this has to be done in a risk posture appropriate to the goals and needs of the University.