CFA Institute

Despite posting a 9.6% return in fiscal 2024, large US college and university endowments once again fell short of market benchmarks — by a staggering 9.1 percentage points. The culprit? A combination of return smoothing and persistent structural underperformance. As the data shows, over the long term, endowments heavily invested in alternatives are falling well behind low-cost indexed portfolios. This post breaks down why and what the numbers really reveal about endowment strategy since the global financial crisis (GFC). The Data Is in The National Association of College and University Business Officers (NACUBO) recently released its annual survey of endowment performance. Funds with greater than $1 billion in assets had a return of 9.6% for the fiscal year ended June 30, 2024. A Market Index, the construction of which is based on US endowment funds’ typical market exposures and risk (standard deviation of return), returned 18.7%. That endowments underperformed their market index by a whopping 9.1 percentage points is a result that needs interpretation. Vexing Valuations Fiscal 2024 was the third consecutive year in which endowment returns were visibly distorted by return smoothing. Return smoothing occurs when the accounting value of assets is out of sync with the market. Exhibit 1 illustrates the effect. The endowment returns for fiscal years 2022, 2023, and 2024 were greatly attenuated relative to the Market Index. The US stock and bond markets declined sharply in the final quarter of fiscal year 2022. Private asset net asset values (NAVs) used in valuing institutional funds at year-end 2022 did not reflect the decline in equity values. This was caused by the practice of using NAVs that lag by one or more quarters in portfolio valuations. The equity market rose sharply the following year, and once again marks for private assets lagged as NAVs began to reflect the earlier downturn. This pattern repeated itself in 2024. The overall effect was to dampen the reported loss for 2022 and tamp down gains in 2023 and 2024. (See shaded area of Exhibit 1.) The pattern of distortion appears to have largely run its course in 2024. Exhibit 1: Performance of Endowments with Greater than $1 Billion in Assets. Dismal Long-Term Results Notably, the long-term performance of large endowments is unaffected by recent valuation issues. The annualized excess return of the endowment composite is -2.4% per year, in line with past reporting by yours truly. Exhibit 2 shows the cumulative effect of underperforming by that margin over the 16 years since the GFC. It compares the cumulative value of the composite to that of the Market Index. The typical endowment is now worth 70% of what it would have been worth had it been invested in a comparable index fund. At this rate of underperformance, in 12 to 15 years the endowments will be worth half what they would have been worth had they indexed. Exhibit 2 also illustrates the impact that return smoothing had on results for the final three years — an apparent sharp performance gain in 2022 resulting from return smoothing, followed by two years of reckoning. Exhibit 2: Cumulative Endowment Wealth Relative to Market Index. Parsing Returns I examine the performance of five NACUBO endowment-asset-size cohorts (Figure 3). These are fund groupings that range from less than $50 million in assets to more than $1 billion. Stock-bond mix explains a lot. Exhibit 3 shows that large funds invest more heavily in equities and earn higher total returns, accordingly. Ninety to 99% of the variation in total return is associated with the effective stock-bond allocation. There is nothing new here. (See, for example, Brinson et al., 1986). Excess return is the difference between total return and a market index based on the respective stock-bond allocations, as illustrated in Exhibit 1. All the excess returns are negative. Exhibit 3: Parsing Returns (fiscal years 2009 to 2024).         Cohort     Effective Stock-Bond Allocation     Annualized Total Return   Percent of Total Return Variance Explained by Asset Allocation (R2)       Excess Return 1  <$50 million 68-32% 6.0% 99% -1.2% 2  $51 – 100 71-29 5.8 99 -1.4 3  $101 – 500 76-24 6.0 97 -1.9 4  $501 – 1000 80-20 6.5 94 -2.3 5  >$1000 million 83-17 6.9 90 -2.4 Alts Explain the Rest Exhibit 4 shows the relationship of excess returns and the average (over time) allocation to alts for the five NACUBO endowment-asset-size cohorts. The relationship between them is inverse. For each percentage point increase in alts exposure, there is a corresponding decrease of 28 basis points in excess return. The intercept is -0.9%. Ninety-two percent of the variation in excess return (R2) is associated with the alts exposure. This tells us that, of the small percentage of return variation that goes unexplained by traditional asset allocation, 92% is explained by exposure to alts. Exhibit 4: Relationship of Excess Returns and Exposure to Alts. Why have alts had such a perverse influence on performance? The answer is high cost. I estimate the annual cost incurred by Cohort #5 funds has averaged 2.0% to 2.5% of asset value, the vast majority of which is attributable to alts. A Simple Story If you can tolerate the risk, allocating to equities pays off over time. Allocating to alts, however, has been a losing proposition since the GFC. And the more you own, the worse you do. It is a pretty simple story, really. source

For more on the crypto and blockchain phenomena, read Valuation of Cryptoassets: A Guide for Investment Professionals by Urav Soni and Rhodri Preece, CFA, from CFA Institute Research and Policy Center. A few foundational microeconomic assumptions and a discounted cash flow (DCF) framework can help inform crypto buy and sell decisions. “Crypto’s Unanswered Question: At What Price?” by Franklin J. Parker, CFA, highlights a conversation I often have with other charterholders, investors, and clients. These discussions have led to both valuable thought exercises and rousing debates. I am not a crypto expert and certainly not a crypto “bro.” I have no strong opinion on whether cryptoassets are undervalued or overvalued, the future of money and commerce or a fad that we’ll all look back on amusingly. Nevertheless, I believe crypto investors can employ a logical valuation framework by which they can make reasonable and informed crypto investment decisions. By applying a discounted cash flow (DCF) model, relying on microeconomic principles as inputs, and using gold and other commodities as guides, we can define a range of prices at which we could expect a reasonable, risk-adjusted rate of return over a given time horizon for a particular cryptoasset. Because cryptoasset prices are directly observable, using a DCF valuation framework, we only need to estimate a future price or range of future prices for a particular cryptoasset, which we can discount back to the present at a required cost of capital. The net present value of our expected future price would equal our estimated intrinsic value today. By comparing that to spot prices, we can make our buy and sell decisions. Admittedly, some elements of this future price estimation process involve a high degree of uncertainty, but others can be reasonably estimated with a modest amount of effort. For example, we know that, over the long run, profit-maximizing businesses will only produce if the marginal revenue exceeds the marginal cost to produce. As such, the marginal cost of mining a crypto coin sets a floor price around which supply will fluctuate. In the case of cryptoassets, the variable costs are reasonably simple to assess — computing costs / energy consumption, taxes, and transaction fees — and because computers can be turned on and off quickly, mining activities can be adjusted quickly depending on price fluctuations. In fact, we can observe this quick response function at work when we juxtapose hash rates over spot prices or estimated mining profitability. Accounting for pre-ordained “halvings” in the mining algorithm, estimating future variable costs associated with cryptoassets, is relatively simple and straightforward. Moreover, crypto miners presumably require a reasonable return on their physical capital investment over time, so we must also include an estimate for the future cost of hardware as well as other capital and fixed costs. With estimates for variable costs, fixed costs, and an assumed required cost of capital for the miners, we can calculate the range of prices at which a cryptoasset will be mined, thus setting the price floor at which we’d expect it to trade. Estimating a cryptoasset’s price ceiling, or the degree to which the actual price could exceed the price floor, is more challenging because it depends on demand, which entails a large degree of uncertainty. But all investments involve uncertainty and investors employ various logical approaches to work through it. For example, we can assess the various demand drivers that influence cryptoasset owners by evaluating it as money. Like gold, cryptoassets are generally divisible into smaller units, countable and fungible (unit of account), used by some to hedge against inflation (store of value), and used to buy and sell goods (medium of exchange). As such, cryptoassets generally meet the criteria for the definition of money, which allows us to measure a cryptocurrency’s demand based on its value as money and more specifically, its utility in these use cases. As a store of value, a cryptoasset may increase in price as confidence in fiat currency collapses or fears of inflation or hyperinflation spike. As a medium of exchange, a cryptoasset may rise in value the more it is used in domestic and international commerce as a method of buying and selling goods and services. We could incorporate a demand component based on the attractiveness of its anonymity — which has utility for both legal and illicit purposes — and we could even incorporate our expectations about how central banks might use cryptoassets to diversify their holdings in the future. A cryptoasset’s value across these various use cases would influence demand, and with it, the price of the cryptoasset itself. Presumably, the sum of a cryptoasset’s utility exceeds its cost and cryptoassets would continue to exist. The point is that, as with all investments, some assumptions must be made about future conditions, and as with gold, some of the key assumptions involve potential demand. Unlike gold, which has a long history, and, therefore, offers some sense for what demand will reasonably look like from various users, cryptoassets lack a long history of use and demand; its story as money is still being written. Nevertheless, this is where the individual assumptions of the investor come into play: their own personal risk tolerance, their investment goals, objectives, and required rate of return, and, ultimately, their own personal determination about the potential risk and potential return, and whether, given their expectations for risk and return, a cryptoasset is an attractive investment. We may all argue about the inputs and assumptions that go into the framework, but that is, after all, exactly what makes financial markets work; the interaction of millions of investors applying their own assumptions and expectations to various investment opportunities using a logical framework in order to avoid speculation. Which brings me to my answer to Parker’s unanswered question: “At What Price?” I don’t know at what price, but I know how someone who wants to answer that question could answer it for themselves. For more on this topic, check out Valuation of Cryptoassets: A Guide for Investment Professionals by Urav Soni and Rhodri

How is risk defined in portfolio optimization objective functions? Usually with a volatility metric, and often one that places a particular emphasis on downside risk, or losing money. But that only describes one aspect of risk. It doesn’t capture the entire distribution of outcomes investors could experience. For example, not owning an asset or investment that subsequently outperforms could trigger an emotional response in an investor — regret, say — that resembles their reaction to more traditional definitions of risk. That’s why to understand risk for portfolio optimization purposes, we need to consider regret. Among different investors, the performance of speculative assets such as cryptocurrencies could potentially evoke different emotional responses. Since I don’t have very favorable return expectations around cryptocurrencies and consider myself relatively rational, if the price of bitcoin increases to $1 million, I wouldn’t sweat it. But another investor with similarly unfavorable bitcoin return expectations could have a much more adverse response. Out of fear of missing out on future bitcoin price increases, they might even abandon a diversified portfolio in whole or in part to avoid such pain. Such divergent reactions to bitcoin price movements suggest that allocations should vary based on the investor. Yet if we apply more traditional portfolio optimization functions, the bitcoin allocation would be identical — and likely zero — for the other investor and me, assuming relatively unfavorable return expectations. Considering regret means moving beyond the pure math of variance and other metrics. It means attempting to incorporate the potential emotional response to a given outcome. From tech to real estate to tulips, investors have succumbed to greed and regret in countless bubbles throughout the years. That’s why a small allocation to a “bad asset” could be worthwhile if it reduces the probability that an investor might abandon a prudent portfolio to invest in that bad asset should it start doing well. I introduce an objective function that explicitly incorporates regret into a portfolio optimization routine in new research for the Journal of Portfolio Management. More specifically, the function treats regret as a parameter distinct from risk aversion, or downside risk — such as returns below 0% or some other target return — by comparing the portfolio’s return against the performance of one or more regret benchmarks, each with a potentially different regret aversion level. The model requires no assumptions around return distributions for assets, or normality, so it can incorporate lotteries and other assets with very non-normal payoffs. By running a series of portfolio optimizations using a portfolio of individual securities, I find that considering regret can materially influence allocation decisions. Risk levels — defined as downside risk — are likely to increase when regret is taken into account, especially for more risk-averse investors. Why? Because the assets that inspire the most regret tend to be more speculative in nature. Investors who are more risk tolerant will likely achieve lower returns, with higher downside risk, assuming the risk asset is less efficient. More risk-averse investors, however, could generate higher returns, albeit with significantly more downside risk. Additionally, allocations to the regret asset could increase in tandem with its assumed volatility, which is contrary to traditional portfolio theory. What are the implications of this research for different investors? For one thing, assets that are only mildly less efficient within a larger portfolio but potentially more likely to cause regret could receive higher allocations depending on expected returns and covariances. These findings may also influence how multi-asset funds are structured, particularly around the potential benefits from explicitly providing investors with information around a multi-asset portfolio’s distinct exposures versus a single fund, say a target-date fund. Of course, because some clients may experience regret does not mean that financial advisers and asset managers should start allocating to inefficient assets. Rather, we should provide an approach that helps build portfolios that can explicitly consider regret within the context of a total portfolio, given each investor’s preferences. People are not utility maximizing robots, or “homo economicus.” We need to construct portfolios and solutions that reflect this. That way we can help investors achieve better outcomes across a variety of potential risk definitions. For more from David Blanchett, PhD, CFA, CPA, don’t miss “Redefining the Optimal Retirement Income Strategy,” from the Financial Analysts Journal. If you liked this post, don’t forget to subscribe to the Enterprising Investor. All posts are the opinion of the author. As such, they should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute or the author’s employer. Image credit: ©Getty Images / jacoblund Professional Learning for CFA Institute Members CFA Institute members are empowered to self-determine and self-report professional learning (PL) credits earned, including content on Enterprising Investor. Members can record credits easily using their online PL tracker. source

Editor’s Note: This is the first in a series of articles that challenge the conventional wisdom that stocks always outperform bonds over the long term and that a negative correlation between bonds and stocks leads to effective diversification. In it, Edward McQuarrie draws from his research analyzing US stock and bond records dating back to 1792. CFA Institute Research and Policy Center recently hosted a panel discussion comprising McQuarrie, Rob Arnott, Elroy Dimson, Roger Ibbotson, and Jeremy Siegel. Laurence B. Siegel moderated. The webinar unveils divergent views on the equity risk premium and McQuarrie’s thesis. Subscribe to Research and Policy Center, and you will be notified when the video airs. Edward McQuarrie: When I tell acquaintances that I’ve put together the historical record of stock and bond performance back to 1792, the first reaction is generally, “I didn’t know there were stocks and bonds 200 years ago!” They aren’t familiar with Jeremy Siegel’s book, Stocks for the Long Run, which is now in its 6th edition, where he presents a 200-year series of stock and bond returns that he first compiled 30 years ago.  The book conveys a simple message with compelling support from that history. That is, stocks have always made buy-and-hold investors wealthy, and the wealth accumulation possible from stocks far exceeds that of any alternative, such as bonds. My new research, “Stocks for the Long Run? Sometimes Yes, Sometimes No,” which was published in the Financial Analysts Journal, suggests otherwise. I will begin to explain those findings in this article. But first, a reminder of the theoretical support that undergirds Siegel’s “Stocks for the Long Run” thesis. Risk and Return Bonds, especially government bonds, are a “fixed income” asset. Investors get the coupon and the return of principal at maturity. Nothing less, but also nothing more. The risk is minimal, and the promised return is accordingly small, because it is largely assured. Stocks are risk assets. No guarantees. Your investment could go to zero. Notably, investors are risk averse. No utility maximizer would put a penny into stocks without a promise of upside, i.e., the potential for a strong return far enough in excess of fixed income returns to compensate for the much greater risk of investing in stocks. Therefore, over any lengthy interval, after short-term fluctuations shake out, stocks can be expected to outperform bonds, exactly as Siegel’s history shows. The Conundrum If stocks will assuredly outperform bonds over intervals of, say, 20 years or more, where is the risk? And if stocks aren’t risky over long intervals, why should their returns exceed the returns on bonds that are not risky? The logic behind “Stocks for the Long Run” blows up. Theory says, “Expected return is a positive function of risk.” But the Siegel history shows no risk for holding stocks over longer intervals. The stock investor always wins. The Resolution I call it a conundrum, not a paradox, because it is easily resolved. All that is needed is a demonstration that sometimes, regardless of the holding period, stocks do blow up, leading to underperformance in either absolute terms or relative to bonds. Stocks can win most of the time, over intervals of any length, if they lose some of the time, over intervals of any length. Those occasional shortfalls are sufficient to restore risk. And risk is the key to any reasonable expectation of earning an excess return over a government bond benchmark. I found those shortfalls in the historical record I compiled. The Updated Historical Record My Financial Analysts Journal article contains a summary of how I compiled the historical data. The online appendix goes into more detail and includes the raw data if you’d like to play around with it. Here is a chart depicting the updated historical record: What do you see? For almost 150 years, stocks and bonds produced about the same wealth. It was a horse race, with the lead swinging back and forth. Stocks would occasionally leap ahead, as in the 1920s, but would also occasionally fall behind, as in the decades before the Civil War. Net, the picture is one of parity performance until World War II. Then, during and after the war, the wealth lines dramatically diverge. Over the four decades from 1942 through 1981, stocks piled up an enormous lead over bonds. The stock investor would have turned $10,000 into $136,900 real dollars. The bond investor would have lost money, turning $10,000 into a real $4,060. Think about that for a moment: You would have lost money in “safe” government bonds. After the war, bonds proved to be a risk asset. Again, what happened after World War II was not that stocks performed extraordinarily well. If you mentally draw a straight line from the beginning of the stock line in 1792 to its end in 2019 (this chart stops before the pandemic), there is not much deviation in the second panel. There was a slight upward displacement through about 1966, but the inflation of the 1970s and the bear market of 1973 to 1974 brought stocks back on trend. Rather, bonds performed extraordinarily poorly during this period. Nowhere else in the chart do you see a multi-decade period of ever-declining bond wealth. The decades through WW I come closest, but the decline was abrupt and abbreviated — nothing like the multi-decade swoon that followed the second world war. The third panel represents my innovation in chart design. In a conventional multi-century chart, once wealth lines have diverged, as in the middle panel, the human eye cannot detect if parity performance has resumed. In a Siegel-style chart, (see p. 28 in the 6th edition of “Stocks for the Long Run” or p. 82 in the 5th edition), what you see is a gap in stock and bond performance that appears to continue to the present day. To see the return to parity performance that did occur after 1981, it is necessary to reset the bond wealth line equal to the stock line as of 1981. Once that is

Asset owners have dramatically increased their allocations to private markets over the past two decades, driven largely by a mistaken belief that private debt and equity deliver returns that are orders of magnitude above those of public markets. What makes most investors believe that private capital funds are such clear outperformers? The use of since-inception internal rate of return (IRR) as the industry’s preferred performance metric and the media’s coverage of the sector’s performance are to blame. The myth of the Yale model — a belief of superior returns stemming from a heavy allocation to private equity funds — is based solely on a since-inception IRR. While there is no ideal substitute for since-inception IRR, investors — especially retail investors — should understand that IRR is not equivalent to a rate of return on investment (ROI). This is the first in a three-part series in which I frame the problem, offer techniques for critical evaluation of fund performance reports, and propose alternative approaches to metrics and benchmarks. The call to action is for regulators or the industry, through self-regulation, to ban the use of since-inception IRR in favor of horizon IRRs. This simple action would eliminate many of the most misleading figures that are presented to investors and would facilitate comparisons. Figure 1 illustrates the migration of institutional assets to private capital over the past two decades. Recently, high-net-worth individuals and more broadly retail investors have joined the trend. The resulting growth in assets under management (AUM) might be unprecedented in the history of financial markets. Private capital fund AUM grew fifteen-fold — 14% per annum over the last 25 years.[1] Figure 1: Evolution of AUM of all private capital funds. Why did capital fly out of traditional asset classes and into private capital funds? The main cause seems to be a strong belief in superior returns. But here’s a reality check on performance. Below are performance metrics, using one of the largest databases available — the MSCI (private-i) — and including all 12,306 private capital funds with a total of $10.5 trillion in AUM, over the entire history of the database. Median IRR of 9.1% Pooled IRR of 12.4% 1.52 total value to paid-in capital (TVPI): TPVI is the sum of distributed and current valuation, divided by the sum invested. 1.05 Kaplan-Schoar Public Market Equivalent (KS-PME): KS-PME is the ratio of present value of capital distributed and current valuation, by present value of capital invested. A score of 1.05 indicates a slight outperformance over the benchmark S&P 500 Index and 1.4% per annum of direct alpha (annualized outperformance over that benchmark). The Source of the Belief: Evidence from News Coverage and Practitioner Publications These performance figures are good, but not spectacular when compared to long-term US stock market returns. According to data on Ken French’s data library, the US stock-market has averaged 12% per annum over nearly 100 years from 1927 and 2023.[2] Most importantly, the returns do not seem commensurate with the spectacular growth in private market AUM. Thus, the puzzle: What makes most investors believe that private capital funds are such clear outperformers? It would be interesting to conduct a survey among both retail and institutional investors to ask for the source of their belief. However, it is difficult to obtain many responses to a survey of this type and to extract what really drives a given belief. An alternative route is to collect information online, mostly from the media. This is the approach I take. While it has its own limitations and is necessarily imprecise, it can nonetheless give a sense of how people convey their beliefs. Exhibits 1 to 9 show some potentially influential articles and statistics. They are spread over time, starting in 2002 (Exhibit 1) and ending in 2024 (Exhibit 9). Exhibit 1 is an extract from a newspaper article covering the fact that a first-time fund was going to be the largest fund ever raised in Europe at the time. Such a situation is rather unusual as funds tend to start small and grow over time. There is, however, no such thing as a pure first-time fund, and the person raising the money had executed nine deals before raising that first-time fund. The article mentions two performance metrics, one is spectacular (62% per annum), the other one not so spectacular (£2.1 per £1 invested gross of fees). Given that this track record led to the largest fund ever raised at the time (2002), it is possible that investors reacted to the 62% annual figure. Sixty-two percent feels extraordinary indeed. In Exhibit 2, Bloomberg shares the Figure 1 from a widely distributed article, “Public Value, a Primer in Private Equity,” first published in 2005 by the Private Equity Industry Association. This figure compares an investment in the S&P 500 to one in top quartile private equity funds from 1980 to 2005. The S&P 500 delivered 12.3% per annum but the top quartile of private equity firms delivered 39% per annum. A 39% return for one quarter of all private equity funds is extraordinary indeed. Exhibit 3 is an extract from an article by The Economist, which wanted to explain the sharp increase in AUM of private equity in 2011. The Economist points to the poster child for private equity investing: the Yale Endowment track record. The article says that the university’s private-equity assets have produced an annualized return of 30.4% since inception. That investment program was launched in 1987; hence Yale Endowment obtained a 30.4% annual return over a 25-year period. This is certainly extraordinary. Exhibit 4 shows the investment memo of a large public pension fund, Pennsylvania’s Public School Employees’ Retirement System (PSERS). The investment committee recommends investing in Apax VII, and the main argument appears to be a gross return of 51% and a 32% net return. The memo states that this performance places Apax in the top decile of private equity firms. No other performance metrics are mentioned. Once again, these numbers appear extraordinary. This fund (Apax VII) closed

Over the past two decades, investors have poured capital into private assets, drawn by the promise of higher returns than public markets. But as Ludovic Phalippou highlights in “The Tyranny of IRR,” many investors are beginning to question whether private equity (PE) returns truly live up to their internal rate of return (IRR) figures. A key reason for the mismatch lies in partial investment. Unlike public assets, PE funds call capital gradually and return it in stages, meaning that a large portion of the committed capital may sit idle for years. This reduces the investor’s gain, even as IRR remains high. IRR compounds the problem by only considering capital deployed by the fund manager, not the full amount contributed by the investor. As a result, it overstates performance and hides the drag of unused capital. To understand what investors truly earn, we need a metric that captures this dilution. Enter the capital deployment factor (CDF) — a simple yet powerful tool that measures how much of the paid-in capital was put to work. It reveals not just how much was used, but also how much gain was lost due to partial investment. The CDF quantifies the impact of partial investment by showing what portion of paid-in capital was actually used to generate returns. Because gain is proportional to the CDF, it also indicates how much potential return was forfeited due to idle capital. What does the CDF reveal about the impact of partial investment on real PE funds? It shows that it is very significant, as the CDF of PE funds rarely exceeds 60% over their lifetime and typically falls to between 15% and 30% at the time of liquidation. A side effect of partial investment is that IRR becomes unreliable for comparing performance: Funds with the same IRR but different capital deployment levels can produce very different gains from the same capital paid in. By contrast, the CDF allows investors to calculate the IRR a fund would need to match the gain of another fund or a liquid asset for the same capital outlay. Capital Deployment Factor The CDF shows the fraction of the amount paid in by the investor that was deployed by the PE fund manager. It can be calculated at any time knowing the fund’s IRR, TVPI and duration. The TVPI is the total value to paid-in indicator at time t, IRR is the internal rate of return since inception expressed on an annualized basis, and DUR the number of years elapsed from inception to time t. For example, a PE fund with an IRR = 9,1% per annum and a TVPI = 1,52X, after 12 years: What does this CDF figure mean? It means that over the 12-year period, only 28.2% of the capital paid in by the investor was used by the fund manager to generate the gain. In other words, just over one dollar in four was put to use to produce wealth. The IRR and TVPI figures above were compiled by Phalippou from a vast and reputable PE fund database. IRR = 9.1% per annum representing the median IRR for PE funds in the database, and TVPI = 1.52x, their average TVPI. The duration reflects the average 12-year life of a PE fund. The CDF = 28.2% is thus broadly representative of the median PE fund at its date of liquidation. How does the CDF affect the investor? The impact of partial investment is considerable, since the gain is reduced in proportion to the CDF, as shown by the gain equation: PAIDINt is the total amount the investor paid in up to time t and Gaint, the gain at time t. Thus, the median PE fund sees its gain reduced by a factor of 0.282 owing to partial investment. What is the CDF’s typical range for PE funds?  It varies throughout the fund’s life. We found it rarely exceeds 60% during its lifetime and falls somewhere between 15% and 30% at liquidation. Venture capital funds and primary funds of funds tend to have higher CDFs than buyout funds, as illustrated in Figure 1. Figure 1. Who controls the CDF? The CDF is dictated by the PE fund manager, since the manager alone decides on the timing of flows. The CDF increases if the manager calls the capital earlier. The CDF also increases if payments are deferred. If the full amount is called in at the beginning and both capital and gain are repaid at the end of the measurement period, the CDF is equal to 100%. Comparing Returns Two funds are equivalent in terms of performance when they have generated the same gain from the same amount paid in. This formula expresses this equivalence criterion by giving the IRR that fund A must have if it is to generate the same gain as fund B out of the same amount paid in. Let’s look at an example: Fund(A): DUR = 12 years; CDF = 20.0%; IRR = ?. Fund(B): DUR = 12 years; CDF = 28,2%; IRR = 9,1% per year. What IRR should fund A have for its performance to be equivalent to that of fund B? Thus, fund A must have an IRR = 11.26% per annum for its performance to be equivalent to that of fund B, which has an IRR = 9.1%. The reason is fund A’s manager has used fewer of the resources at his disposal than fund B’s manager, which is reflected in their respective CDFs. If fund A has an IRR greater than 11.26%, it is considered to have outperformed fund B. Let’s now assume that fund C has a CDF = 100% and the same duration as fund B. For fund C to have equivalent performance to fund B, its IRR could be much lower at: A CDF = 100% implies that the amount paid in remained fully invested throughout the 12-year period, with no interim cash flows, the capital and gain being recovered by the investor at the end of the period. This would

Is there a trade-off between diversity and investment performance? It’s a common question with a definitive answer: No That’s my conclusion after conducting an extensive review of the literature on the relationship between diversity and investment risk and performance. An Overview of the Studies In total, the research I analyzed comprises 56 studies published over the past 28 years that combined examine almost 50 years of data. They largely focus on gender diversity. In fact, 45 of the 56 examine gender diversity only. Only 11 considered racial, ethnic, and other types of diversity, and most of this cohort also took gender diversity into account. This emphasis is largely a function of the available data. Information on first names and pronoun use is easily accessible to researchers who can use it to make assumptions about gender. To examine other forms of diversity, however, researchers need self-identification data, which is harder to find, though some clever studies do leverage public information about portfolio managers’ birthplaces to explore cultural and socioeconomic diversity. Still, whatever the distinctions among the studies’ methods and focus, the results are consistent across the board regardless of the form of diversity under analysis. Fifty-two of the 56 studies focus on portfolio management. Roughly one third of these examine diversity at the team level and the rest at the individual level. The four remaining studies consider the ownership of the firm hired to manage the investment team. Of course, ownership and portfolio management at many firms may have considerable overlap. Diversity and Investment Performance: The Results With that background, the findings on investment performance are as follows: No Difference or Mixed: There were 15 findings of either no variation in performance or outperformance only in some circumstances, whatever the characteristics of the manager. Most of these were academic studies of mutual funds. Outperformance: 26 findings noted an association between diversity ad outperformance. More than half of these were based on studies of hedge funds, private equity funds, or venture capital funds, and were produced by industry firms. Underperformance: Seven findings associated diversity with underperformance. (These 48 findings do not add up to the 56 total studies because some studies have multiple findings on performance, while others focus on risk or other portfolio characteristics and draw no conclusions on performance.) In my assessment, the evidence for the “No Difference or Mixed” is strongest. Why? Because such findings are heavily tilted toward academic studies that are more likely to be risk-adjusted, peer-reviewed, and based on standardized and heavily scrutinized mutual fund data. Nevertheless, the impressive showing of the “Outperformance” category implies that diversity may have a more positive affect on investment performance. On the whole, the weight of the evidence indicates that diversity is associated with performance that is at least as good as the mean. Investment Performance and Diversity: Research Findings Focus and Conclusion Diversity and Risk More than half of the studies address portfolio risk. The results appear straightforward at first glance, with almost two-thirds associating diversity with lower risk. However, when it comes to risk-raking, we need to distinguish between personal accounts and professional investors. The findings on personal accounts are quite consistent. There is no indication that women take more risk than men. These studies draw from large data sets, such as all accounts at a major brokerage firm. Their findings are among the oldest in the literature and have been replicated periodically over the past 28 years. They have almost become accepted wisdom. However, while there may be a strong association between gender and risk-taking in personal accounts, factors other than gender may be driving the results. While most studies control for income and marital status, other factors can affect risk taking, such as risk tolerance and financial knowledge. According to one cross-border study, gender differences in risk-taking are not present in countries with more gender equality, which supports the hypothesis that gender is not determinative when it comes to risk. Studies of risk-taking by professional investors further support this hypothesis: 11 such studies find women professional investors take less risk, four find no difference in risk-taking, and four find that women take more risk. Overall, this literature suggests that something other than gender may be driving the results. Future studies will hopefully zero in on what that driver might be. Diversity and Risk in Investing: Research Findings Conclusion The evidence shows that diversity and investment performance co-exist. Investors don’t have to make a choice between the two. Is there a study I missed? Let me know at www.versanture.com/contact. For more on the relationship between diversity and investment results, don’t miss “Diversity and Investment Performance: A Summary of the Research.” If you liked this post, don’t forget to subscribe to Enterprising Investor. All posts are the opinion of the author. As such, they should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute or the author’s employer. Image credit: ©Getty Images/ matdesign24 Professional Learning for CFA Institute Members CFA Institute members are empowered to self-determine and self-report professional learning (PL) credits earned, including content on Enterprising Investor. Members can record credits easily using their online PL tracker. source

In an era where data privacy and efficiency are paramount, investment analysts and institutional researchers may increasingly be asking: Can we harness the power of generative AI without compromising sensitive data? The answer is a resounding yes. This post describes a customizable, open-source framework that analysts can adapt for secure, local deployment. It showcases a hands-on implementation of a privately hosted large language model (LLM) application, customized to assist with reviewing and querying investment research documents. The result is a secure, cost-effective AI research assistant, one that can parse thousands of pages in seconds and never sends your data to the cloud or the internet. I use AI to augment the process of investment analysis through partial automation, also discussed in an Enterprising Investor post on using AI to augment investment analysis. This chatbot-style tool allows analysts to query complex research materials in plain language without ever exposing sensitive data to the cloud. The Case for “Private GPT” For professionals working in buy-side investment research — whether in equities, fixed income, or multi-asset strategies — the use of ChatGPT and similar tools raises a major concern: confidentiality. Uploading research reports, investment memos, or draft offering documents to a cloud-based AI tool is usually not an option. That’s where “Private GPT” comes in: a framework built entirely on open-source components, running locally on your own machine. There’s no reliance on application programming interface (API) keys, no need for an internet connection, and no risk of data leakage. This toolkit leverages: Python scripts for ingestion and embedding of text documents Ollama, an open-source platform for hosting local LLMs on the computer Streamlit for building a user-friendly interface Mistral, DeepSeek, and other open-source models for answering questions in natural language The underlying Python code for this example is publicly housed in the Github repository here. Additional guidance on step-by-step implementation of the technical aspects in this project is provided in this supporting document. Querying Research Like a Chatbot Without the Cloud The first step in this implementation is launching a Python-based virtual environment on a personal computer. This helps to maintain a unique version of packages and utilities that feed into this application alone. As a result, settings and configuration of packages used in Python for other applications and programs remain undisturbed. Once installed, a script reads and embeds investment documents using an embedding model. These embeddings allow LLMs to understand the document’s content at a granular level, aiming to capture semantic meaning. Because the model is hosted via Ollama on a local machine, the documents remain secure and do not leave the analyst’s computer. This is particularly important when dealing with proprietary research, non-public financials like in private equity transactions or internal investment notes. A Practical Demonstration: Analyzing Investment Documents The prototype focuses on digesting long-form investment documents such as earnings call transcripts, analyst reports, and offering statements. Once the TXT document is loaded into the designated folder of the personal computer, the model processes it and becomes ready to interact. This implementation supports a wide variety of document types ranging from Microsoft Word (.docx), website pages (.html) to PowerPoint presentations (.pptx). The analyst can begin querying the document through the chosen model in a simple chatbot-style interface rendered in a local web browser. Using a web browser-based interface powered by Streamlit, the analyst can begin querying the document through the chosen model. Even though this launches a web-browser, the application does not interact with the internet. The browser-based rendering is used in this example to demonstrate a convenient user interface. This could be modified to a command-line interface or other downstream manifestations. For example, after ingesting an earnings call transcript of AAPL, one may simply ask: “What does Tim Cook do at AAPL?” Within seconds, the LLM parses the content from the transcript and returns: “…Timothy Donald Cook is the Chief Executive Officer (CEO) of Apple Inc…” This result is cross-verified within the tool, which also shows exactly which pages the information was pulled from. Using a mouse click, the user can expand the “Source” items listed below each response in the browser-based interface. Different sources feeding into that answer are rank-ordered based on relevance/importance. The program can be modified to list a different number of source references. This feature enhances transparency and trust in the model’s outputs. Model Switching and Configuration for Enhanced Performance One standout feature is the ability to switch between different LLMs with a single click. The demonstration exhibits the capability to cycle among open-source LLMs like Mistral, Mixtral, Llama, and DeepSeek. This shows that different models can be plugged into the same architecture to compare performance or improve results. Ollama is an open-source software package that can be installed locally and facilitates this flexibility. As more open-source models become available (or existing ones get updated), Ollama enables downloading/updating them accordingly. This flexibility is crucial. It allows analysts to test which models best suit the nuances of a particular task at hand, i.e., legal language, financial disclosures, or research summaries, all without needing access to paid APIs or enterprise-wide licenses. There are other dimensions of the model that can be modified to target better performance for a given task/purpose. These configurations are typically controlled by a standalone file, typically named as “config.py,” as in this project. For example, the similarity threshold among chunks of text in a document may be modulated to identify very close matches by using high value (say, greater than 0.9). This helps to reduce noise but may miss semantically related results if the threshold is too tight for a chosen context. Likewise, the minimum chunk length can be used to identify and weed out very short chunks of text that are unhelpful or misleading. Important considerations also arise from the choices of the size of chunk and overlap among chunks of text. Together, these determine how the document is split into pieces for analysis. Larger chunk sizes allow for more context per answer, but may also dilute the focus of the topic

The relationship between capitalization rates (cap rates) and interest rates is more nuanced than first meets the eye. Understanding their interplay is a cornerstone of real estate investment analysis. In this blog post, we dissect historical data, discuss current opportunities, and forecast real estate valuations for the second half of 2024. Cap rates measure the ratio of a property’s net operating income (NOI) to its purchase price. Because interest rates influence the cost of borrowing, they affect property prices and investor returns. It is often assumed that cap rates move in tandem with interest rates because, in theory, rising interest rates lead to higher cap rates, which in turn lower property values. However, empirical data suggests that this relationship is not as straightforward as the theory. Historical Context and Theoretical Foundations While nominal interest rates (i.e., the interest rate that doesn’t take inflation into account) have an impact on real estate values, they do not have the same effect on cap rates as do real interest rates (i.e., the interest rate that has been adjusted for inflation). First, we can distill the relationship between cap rates and interest rates through the lens of inflation expectations. In a previous blog post, we noted that there is ample research supporting real estate’s ability to function as an inflation hedge. As such, real estate values may suffer if the increase in cap rates is driven by higher real rates, rather than inflation expectations​​. Prior cycles of rising rates provide time-series data that are not influenced by current conditions. A 2016 white paper by TIAA Global Real Assets concluded that cap rates, as proxied by the NCREIF Property Index (NPI), do not necessarily move in lockstep with interest rates. Using the 10-Year Treasury yield as a proxy, it cited a positive correlation of 0.7 between cap rates and interest rates from Q4 1992 to Q3 2015.  And cap rates are not in real time. Because they are periodically set via appraisals or sparse transaction data, cap rates lag private market values for other real assets. Cap Rates: Range-Bound or Fixed in the Moment? Several other factors influence the dynamic between cap rates and interest rates, including other real estate fundamentals, broader macroeconomic performance, capital flows, and investor risk appetites. One of the most prevalent views on the cap rate-to-interest rate relationship is that cap rates move within a range as measured by their relationship to a risk-free rate such as the 10-year Treasury yield.  This basis point spread can be viewed as a protective buffer from any expected rises in interest rates, and it compresses or expands over time. This preferred gage has not shown a consistent behavioral pattern over time, however, and there are several instances in history when cap rates and US Treasuries did not move in unison, with lagged or minimal movement.  The correlation (five-year rolling basis) between US Treasury yields and cap rates fluctuated between -0.82 and 0.79 from 1983 to 2013, according to an analysis by Morgan Stanley. The firm identified eight key periods within that timeframe when corporate bond rates and/or the 10-year US Treasury yield moved upward. Notably, cap rates moved in the opposite direction during five of those periods. The key question here is whether the analysis was based on concurrent cap rates — fixed in the moment — or considered potential lags in cap rate data. Given the periodic appraisal-based valuations associated with private real estate, there is a lag in valuation adjustments, which also smooths volatility. In a different scenario to the analysis, Morgan Stanley adjusted its cap rates by a one-year period and arrived at a similar place. Other Factors Influence Cap Rates Morgan Stanley identified several other drivers to the cap rate-to-interest rate relationship, including credit availability, supply and demand, and increases in real rates. The effects of credit availability are intuitive: increased availability of debt capital at more compelling rates is beneficial to overall transaction volumes. This intensifies competition for assets, which further benefits seller pricing within this generally illiquid asset class and compresses cap rates. A countering effect to increased competition can be the available supply of real estate within a certain sector or market. Simply put, the availability of alternative investment options can drive cap rate expansion by lowering underlying prices. The opposite is true in markets with few investment alternatives: in these markets, underlying real estate prices rise and cap rates compress. Peter Linneman’s Fall 2020 newsletter reported a correlation between 10-year Treasury yields and cap rates over a ~20-year window. When he dissected cap rates over distinct time periods, however, the data pointed to other factors that influenced them more directly. Linneman and his co-authors intuited that capital flows should play a significant role in driving cap rates, given that availability of capital and increased competition for assets will significantly impact real estate asset values and compress cap rates.  Linneman’s research findings demonstrate the benefit of examining the components driving capital availability at any given time. His multivariate model utilizes the flow of mortgage funds relative to gross domestic product (GDP) as a proxy for liquidity and historic cap rates as well as the unemployment rate as proxies for market dynamics and risk, respectively. Ultimately, this model is nearly as accurate in predicting forward cap rates as the regression model of cap rates to real rates is descriptive. Most notably, a key finding is that when mortgage debt grows by 100 bps faster or slower than GDP, cap rates expand by 22 bps for multifamily properties and 65 bps for office properties, suggesting that an increase in mortgage debt as a percentage of GDP drives down value. The model also finds that an increase in unemployment slightly expands cap rates.  When investors are withdrawing capital at the same time lending becomes more restrictive, transaction volume and pricing will fall. This is consistent with commercial real estate (CRE) capital markets over the past one to two years, predominantly driven by the higher rate environment, a volatile stock market,

Younger generations in the United States and other developed nations increasingly worry about economic trends that threaten to jeopardize their financial futures. The United States struggles with rising national debt, sparking fears that the dollar might lose its dominant status as the world’s reserve currency. This issue is further complicated by policies that favor easy money and significant budget deficits, potentially leading to skewed approaches to political economy like Modern Monetary Theory (MMT). Issuing central bank digital currencies (CBDCs) can help developed countries including the United States counteract these trends. CBDCs are digital forms of money issued by central banks, unlike decentralized cryptocurrencies such as Bitcoin. They come in public and institutional forms, serving as a digital replacement for physical cash and facilitating smoother interbank transactions. Major economies around the world continue to rack up large budget deficits each year, primarily to sustain existing entitlement programs. They use fiscal stimulus as a short-term fix for sluggish gross domestic product (GDP) growth without a matching increases in tax revenue. The Congressional Budget Office (CBO) predicts that if current trends continue, over the next 30 years the average US federal budget deficit will swell to 8.5% of GDP and national debt will balloon to 166% of GDP. Given this trajectory, achieving a budget surplus seems increasingly slim. To make matters worse, interest rates will climb to support the growing debt burden, limiting the government’s ability to finance its entitlement programs. This scenario hints at a potential slowdown in economic growth and escalating fiscal challenges. The rising cost of servicing its burgeoning debt will force the government to raise taxes or cut spending, further complicating economic management. Elevated public debt could also divert savings from productive investments, hindering long-term economic growth. If economic growth falls behind debt accumulation, the United States may encounter financial instability, including potential crises or the need for debt restructuring. Hence, managing fiscal policy, inflation, GDP growth, and debt remains a crucial, albeit complex task. Effective debt management is essential to avoid significant interest rate hikes and ensure economic stability. However, relying on “easy money” policies and excessive debt could lead to political and economic strategies influenced by MMT, which promotes government spending funded by money creation to achieve full employment and focuses less on deficits unless they spur inflation. Such a shift could deeply impact economic stability and the United States’ international financial stance. The Dangers of MMT If MMT becomes more influential in economic policy, it will radically alter how fiscal strategies are framed. According to MMT, taxation’s primary role is to manage inflation and allocate resources more effectively. This theory also promotes a shift in the Federal Reserve’s priority from conventional monetary policies toward more direct fiscal involvement. MMT in effect empowers substantial government spending on social and environmental programs, because it asserts that monetary creation can support such spending without negative consequences until inflation becomes a pressing issue. Thankfully, MMT has not garnered widespread acceptance among leading economists in the developed world. Even so, the political environment often favors short-term, straightforward solutions suggested by policymakers rather than more nuanced and logical alternatives. This tendency poses a significant risk to developed nations’ long-term global economic leadership. The Dollar’s Pivotal Role in World Finance The structural economic challenges in the United States date back to the inflationary pressures of the 1970s. These issues were managed by the neoliberal policies of the 1980s, which, despite their flaws, leveraged the US dollar’s pivotal role in world finance. This role was cemented by globalization and financial market advancements in the 1990s. However, the era of economic growth came to a screeching halt with the 2008 global financial crisis. Interestingly, this crisis reinforced the dollar’s reputation as a safe-haven asset, channeling global capital into US Treasury bonds amidst widespread economic turbulence.  Today, the US dollar remains dominant in global reserves, bolstered by significant foreign investments in Treasuries. Countries like Japan and China hold substantial assets to manage their currencies and support trade surpluses. While the dollar accounts for more than 60% of global forex reserves, the rising national debt threatens its stability, emphasizing the need for careful management. The interest costs on federal debt have overtaken US military spending, potentially leading to reductions that could weaken the military’s ability to secure trade routes and ensure geopolitical stability — key factors that have traditionally enhanced the United States’ investment allure and economic confidence. Maintaining this stability is vital for the dollar’s continued role in international trade and as the primary reserve currency. This, in turn, allows the United States to influence global economic policies, impose economic sanctions, and shape economic trends globally. The dollar’s dominance makes international borrowing easier and cheaper, creating a strong demand for dollar-denominated assets and helping to keep interest rates low. However, if the dollar’s dominance begins to fade, the United States might struggle to afford its deficits, leading to a diminished global economic influence. Nearly 90% of international transactions are conducted in US dollars or euros. Any major shift toward “de-dollarization” would be painful not just for the United States but for the world, potentially reducing the financial quality of life for the average individual. To counteract this trend, two primary actions are required: First, the United States must adopt stricter fiscal discipline, stabilize entitlement programs, and increase tax revenues. Second, there must be a focus on modernizing and digitalizing the US Dollar. If the dollar’s dominance wanes, the United States’ ability to manage its deficits and maintain its economic power will take a serious hit. The notion that decentralized cryptocurrencies could dominate global financial flows remains unfeasible. The idea that Bitcoin or other cryptocurrencies could replace traditional fiat currencies is a topic of debate. The reason? It would have major implications for credit markets. If cryptocurrencies take over, banks might lose their position as intermediaries, potentially reducing their influence over credit creation. The high volatility of cryptocurrencies like Bitcoin could increase credit risk, making risk management tougher and discourage market participation.  Moreover, without central bank tools