Why do we need new asset allocation models?

The limitations of existing models

The primary role of asset allocation is to design a portfolio that optimally balances risk and return, tailored to an investor's specific objectives and risk constraints. A key aspect of this process is achieving diversification while avoiding unintended exposures. Since asset allocation largely determines a portfolio's overall risk and return profile, it is the most critical factor influencing investment outcomes. Therefore, asset allocation models should be evaluated based on two main criteria: (i) how accurately they represent the individual investor's objectives and risk constraints, and (ii) how effectively they capture the risk-return trade-off faced by that investor. To add value, asset allocation models must succeed in both aspects of the asset allocation process.

Today, investors and asset allocators have access to a wide array of asset allocation models and portfolio optimization tools, each with different theoretical foundations and features.^[1] However, despite their diversity, many of these models share several common limitations, which we discuss below.

Limited modelling of investors’ objectives and constraints

Most of the current asset allocation frameworks provide limited scope for customization to meet the unique objectives and constraints of individual investors.^[2] This is a crucial yet often overlooked aspect of the asset allocation process.

Asset allocation approaches have evolved significantly since 1950s, when foundational research on portfolio diversification and asset pricing first emerged. Early approaches to asset allocation were largely centered around broad diversification and mean-variance optimization, often motivated by use of the Capital Asset Pricing Model (CAPM). A typical asset allocation was a “60/40” equity/bond portfolio with the customisation largely limited to adjusting the equity share depending on the degree of investor’s risk aversion.

To address the limitations of the CAPM model, academics developed the Intertemporal CAPM (ICAPM), a multi-period model that accounts for changing expected returns and risks. It thus provides a theoretical framework for investors to design dynamic investment strategies that hedge investor-specific risks. While theoretically appealing, its implementation is not straightforward, which is one of the reasons why it has not been widely adopted by investors.

Historically, the scope for customisation was constrained, not only by the lack of appealing and implementable frameworks, but also by limited availability of investable assets.

Modelling of private assets

Since the 1990s, new ways of deploying capital such as private equity, private debt, real estate, and hedge funds introduced new dimensions for asset allocators. These additions presented an opportunity for investors to refine their exposures and better align their portfolios with specific investment objectives and constraints. Having the tools available does not automatically lead to optimal outcomes, however.

Many models supporting asset allocation are not able to handle private assets in a manner consistent with the liquid part of the portfolio. While including private assets and alternatives into the asset mix broadens the scope for customisation and has the potential to improve the risk-return trade-off, it also significantly increases the complexity of the portfolio.

A key challenge is that public and private equity, government bonds and hedge funds have overlapping exposures to underlying macro drivers, which are difficult to identify—particularly in the case of private assets. Achieving optimal diversification and avoiding unintended risks requires a deep understanding of these overlapping exposures.^[3] This has fuelled a long-standing debate over whether to focus on traditional "asset classes" or adopt a "factor-based" approach.^[4] Our approach sidesteps this debate, as we recognize that all asset classes inherently relate to macro drivers. In our modelling framework, macro drivers are interpreted as "risk factors," to which all asset classes are exposed in different ways. These risk factors differ from the tradable factors used in risk models or those derived from sorting securities based on attributes like value, growth, or quality.

Underestimation of long-term portfolio risks

For decades, investors benefited from a favorable environment characterized by steadily declining interest rates, globalization, and a rules-based international order—conditions that created a supportive backdrop for traditional 60/40-style allocations. In other words, we have witnessed a favourable outcome from a long-term risk perspective.

The existing allocation models tend to underestimate and/or misrepresent long-term portfolio risks, which are largely driven by secular shifts in key macroeconomic variables. This issue is becoming increasingly relevant as favorable global conditions may not persist unchanged. In a less accommodating macroeconomic environment—marked by heightened volatility and a greater likelihood of tail-risk events across asset classes—the need for better modeling of long-term portfolio risks will become more critical.

Complex generation of optimal portfolios

Many existing models fail to offer clear, intuitive insights into the optimal asset mix they propose. In addition, the optimal portfolios tend to be sensitive to small changes in estimates in expected returns, which compounds the previous issue. This is primarily because they often rely on dynamic statistical models, such as Vector Autoregressions (VARs), to represent complex relationships between asset class returns based on historical data. A lack of intuitive understanding can undermine the influence these models have on asset allocation decisions, particularly at board and committee levels where such decisions are often made.^[5]

Looking forward

The investment landscape continues to evolve, with increasingly innovative ways to deploy capital and growing opportunities in private markets across asset classes. These developments raise the bar for asset allocation models, requiring more sophisticated approaches.

Our asset allocation approach empowers investors and asset managers to address the challenges posed by an ever-evolving global economy and investment landscape. Our simulation-based approach means it’s possible to directly model investors, enabling us to tailor asset allocations to their specific needs and constraints. The simulation-based approach also allows us to model the risk-return trade-off in a forward-looking way and create forward-looking scenarios.

Our flexible modelling approach allows us to model not only the returns on asset classes but also their prices. This flexibility in representing asset classes is instrumental for incorporating private assets into the asset allocation process in a consistent manner, avoiding ad-hoc assumptions.^[6]