Rethinking Asset Allocation Part 1

As we emerge from the swim-or-drown situation caused by the global pandemic, the world we used to live in has changed dramatically - financial markets included. Indeed, if on the one hand the recent crisis has sparked unprecedented levels of volatility across all asset classes, on the other hand, we have all experienced how keeping pace with technology has become vital to deal, survive and seize the opportunities of this new complex world. Yet, we would be wrong to think that all is gone.

In this two-part paper, we will share 10 asset allocation insights to help asset and investment managers face the new investing challenges. From understanding why the scientific method has become essential to decode the inner dynamics of financial markets to untapping the full potential of factor-based analysis, these insights will help us frame where opportunities lie, showing not only how, but also where the latest breakthroughs in the field of machine learning and quantitative finance are leading the world of investing. How will asset allocation look like in the 2020s?


Evolution is definitely a ubiquitous concept of our lives. Whether we talk about the current economic landscape, socio-political dynamics or the impact brought by the new technological breakthroughs, we must find a way to deal with it, both as humans and as investors. Yet, evolution rarely occurs instantaneously. Rather, it is more often the result of several incremental changes that compound over time and lead us to a new understanding of reality. 

This is why, when we look at it from day to day, we perceive it as noisy, while it appears more clearly if we consider the big picture.

In this sense, investing makes an interesting example to consider for two reasons: its forward-looking nature and its increasing complexity. Indeed, as financial markets continuously gather investors’ expectations about the future, investing turns out to be a two-sided job: it requires having the right tools to deal with and interpret complexity, but also to position oneself accordingly to anticipate the gradual unfolding of market dynamics.

From this standpoint, over the past few years the increasing adoption of Artificial Intelligence (AI) and Machine Learning in investing has marked a significant acceleration in our ability to make more informed investment decisions. As they grow into mature technologies, their ability to constantly adapt, and gain a deeper understanding of the inner dynamics of financial markets has marked the crucial shift from human and statistical reasoning to a new era of assisted decision-making.

In this sense, the asset management industry is facing today the challenge of keeping pace with the evolution – rather than revolution – of how technology is being utilized to design and improve more efficient investment strategies. As we are going to see, during this paper we will depict 10 key research insights useful to frame how asset allocation is going to look like in the near future.

In this first part, we will start by retracing the footsteps of the academical and empirical research surrounding asset allocation, pointing at the direction towards which this industry is moving. This will clear the path to examine how investment and correlation analysis are benefiting from the application of AI and machine learning models. 

Eventually, we will discuss how having a tight grip on investment factors, portfolio rebalancing and ESG investing is quickly becoming essential for asset and investment managers to better understand the risks their portfolios are exposed to.

Start integrating AI into your investment process with the most advanced no-code solution.

Fill out the form to book a demo of Sphere®, and learn how it can bring value to your investment decisions.

Book A Custom Demo

By submitting you are agreeing to the Terms of Use and Privacy Policy
Thank you! Your submission has been received.
Oops! Something went wrong while submitting the form.

Asset Allocation in the 2020s: Revolution or Rediscovery?

Tens of thousands of scientific papers have been written on asset allocation during the last 70 years. Yet, if we look at the big picture, we see that the fundamental questions that this field of research addresses have not changed at all.

Indeed, the fil rouge across all of them has been finding a reliable strategy to allocate wealth efficiently across securities. In other words, a mechanism to build portfolios that meet a set of investment objectives and constraints.

In this sense, what goes under the name of “modern approach to investing” is a body of knowledge that sets its foundations in the need of applying the scientific method to both understand the dynamics of financial markets and build models that can assist us in making better investment decisions.

However, to develop insights about the direction currently taken by the research in the field of asset allocation, what appears to be useful is to consider the degree by which the recent breakthroughs in the field of AI have dramatically improved our ability to go beyond the limitations (in terms of restrictive assumptions and computing power) posed by traditional quantitative models.

Indeed, since Harry Markowitz pioneered the first comprehensive theory on portfolio decisions in the early 1950s, concepts like volatility, correlation, efficiency and diversification have remained at the heart of portfolio management. Most notably, they have helped the next generations of researches (e.g. Sharpe, Fama, French, Black, Litterman, Michaud) to further investigate the complex dynamics of financial markets, for example, introducing the notion of investment factors – sources of excess returns that can be exploited systematically – or the idea to average a large number of future probabilistic scenarios to refine and improve the portfolio positioning.

Yet, for a long time, being able to successfully apply in the real world what they had discovered has encountered many obstacles, due in part to the restrictive assumptions of their models (e.g. backward-looking approach, uniperiodal setting) and to the absence of reliable statistical tools that could cut through the complexity of financial markets.

From this standpoint, AI and Machine Learning are finally bridging the gap between theory and practice. As they can learn the hidden structure of data by processing millions of data points each day, they are enabling asset and portfolio managers to make better investment decisions, and build unbiased investment strategies that quickly adapt when markets change.

As we will see, especially for what concerns asset allocation, these sophisticated tools are also marking a significant transition towards a new approach to portfolio management, taking portfolio construction one step further: a forward-looking approach to diversification to successfully navigate the complexity of financial markets.

Forecasting Correlations is the Right Way To Go

Asset and investment managers have good reasons to say that diversification – reducing risk without affecting returns – is their top priority. To build diversified portfolios investors rely on correlation, that is, the degree and the direction of how a set of securities move with respect to each other. 

Indeed, by selecting assets that tend to move in opposite or independent directions, the resulting portfolio will likely be less risky, as the losses recorded by one can be smoothed out by the gains of the others. 

From this perspective, it appears clearly why correlation analysis plays a key role in building efficient portfolios. Yet, analyzing and forecasting correlations has revealed to be anything but an easy task. This is due in part to the fact that financial markets are driven by complex (or hidden) dynamics, and also because financial data exhibits a low signal-to-noise ratio, meaning that it does not lend itself to be easily interpreted. In this sense, reliably forecasting correlations has turned into a strategic edge to improve investment decision-making.

The first attempts to forecast correlations were based on their historical values, as Markowitz initially suggested. However, as the future rarely turns out to be similar to the past, these forecasts revealed to be inaccurate because the correlation between securities tends to vary sharply over time. As we will see, relying exclusively on a backward-looking approach can easily mislead investors.

The nature of this instability is twofold: not only correlations among securities often vary over time, but they can also change according to the sampling frequency (e.g. weekly or monthly data) and time horizon (e.g. a month, a year) that are chosen to calculate them. 

This curtain of noise makes it difficult for investors to grasp immediately the information that correlations provide, as there are thousands of combinations to consider before being confident with correlation estimates. 

In this sense, not having a rational and forward-looking approach to correlation analysis can be dangerous for investors because obtaining reliable estimates will be far more difficult.

From a practical point of view, the fact that correlations vary over time has a profound effect on the diversification benefit – the contribution to diversification brought by each asset in the portfolio.

For instance, if two securities move from a negative to a positive correlation, the benefit will decrease and the portfolio will likely become riskier, as assets that were first thought to be uncorrelated are now exposed to the same source risk.

Exhibit 1 can help us understand the degree of these fluctuations. The plot shows the rolling 5-years correlations between different asset classes and the US stock market represented by the S&P 500 Index – in the last 20 years. 

For each asset class, three elements show us how much they vary over time. First, the horizontal line includes the entire interval between the lowest and the highest correlation values obtained during the period. Then, the coloured rectangle highlights the second and third quartiles respectively, separated by the median. Finally, the current correlation value is indicated by the white dot. For instance, we can see that the correlation of EU Corporate Bonds with the S&P 500 Index has changed a lot historically, moving from a minimum of about -0.24 to a maximum (and current) value of about 0.5, with a median around 0.11.

Distribution of the rolling 5-year correlations with the S&P 500 Index for different asset classes from 1999 to 2020.
Exhibit 1Distribution of the rolling 5-year correlations with the S&P 500 Index for different asset classes from 1999 to 2020.

From the considerations made above, it naturally emerges the significant role that Artificial Intelligence can play in forecasting correlations. Indeed, AI’s capacity to adapt to the unfolding of market dynamics enables asset managers to extract more efficiently the information contained in the traditional variance-covariance matrix. 

By having a probabilistic – rather than a static – forecasting approach based on thousands of simulations, investors can have a 360-degree understanding of the potential future shifts in correlations. Consequently, they can discount different future evolutions of the market and build portfolios better positioned to sustain unexpected shocks.

Clearly, the adoption of AI in finance opens the door to a holistic vision of correlations. Today, asset and investment managers can forgo traditional backward-looking models and leverage these new tools to reach a thorough understanding of the complex dynamics of correlations and, most importantly, of their evolutions in the future. 

By doing so, they will be able to build more stable and efficient portfolios, dynamically adapting to ever-changing market scenarios.

Seeing Risk Through the Factor Lenses

As investing involves risk, managing its exposure and picking the right securities are any investment manager’s first concerns when it comes to building a portfolio. Indeed, when we talk about asset allocation, the idea that there is a reward in bearing uncertainty – that gains are proportional to risks – is not only intuitive but a deeply-rooted assumption in the investment culture. Explaining it, however, has required a deeper understanding of which are the hidden drivers of investment returns. 

What has been unveiled during the last decades is that, instead of pure skill, a great portion of the excess return over a benchmark results from the exposure to a set of industry or firm-specific characteristics of financial securities, the so-called investment factors. In this sense, starting from the famous equity risk premium, many other risk premia have emerged during the years, paving the way to the modern factor-based approach to portfolio management. 

Exhibit 2 - Key facts about the main equity investment factors and drivers of excess returns.

The giant leap caused by this discovery has allowed asset managers to gain a deeper look-through of investment risk and to use factors as the building blocks of their investment strategies. 

In this sense, by breaking down returns into risk-based components, it has become possible to explore which are the drivers that influence investment returns, and in turn, this has allowed to have better control over its performance.

However, successfully using investment factors to harvest risk premia does not come off easily. Indeed, it requires great care and expertise to not fall prey to the potential sources of bias that may lead to the wrong investment decisions.

As we have discussed extensively in our paper “Fifty Shades of Alpha”, the issues with investment factors generally belong to one of these two groups. On the one hand, academic factors (i.e. investment factors as described in the financial literature) are not directly investable due to several reasons that include, for instance, excessive turnover, large drawdowns or a low efficiency ratio due to trading costs. This requires to further refine and improve the original factors in order to identify which is the best way to exploit them in an investment strategy. 

On the other hand, as investment factors reflect the underlying dynamics of financial markets, their behaviour (in terms of performance) tends to vary over time, meaning that although they represent a persistent source of excess return, they are not meant to always earn a positive risk premium. 

Yet this feature – being pro or countercyclical – offers an important insight for what concerns diversification.

Since standalone factors alternate periods of sustained overperformance with moments of underperformance, switching between them – the ability to move between Betas – is becoming increasingly strategic for asset and investment managers as a way to allocate capital more efficiently.

Eventually, as we learn more about the true nature of risk and investment returns, it is inevitable to see in the rise of factor investing and alternative risk premia the new face of portfolio management – transparent, rule-based and factor-driven.

Portfolio Rebalancing is More Than You Know

Rebalancing a portfolio asset allocation is a quite common practice in the asset management world. Yet, although a wide literature has examined the pros and cons of following different rebalancing strategies, rethinking why – rather than how – it adds value to portfolio management is receiving increasing interest as an additional dimension to unlock precious insights.

In this sense, as asset allocation is one of the primary determinants of risk, rebalancing means readjusting the portfolio weights to reflect its original risk targets. In this way, sticking to the plan allows investors to better control the risks their portfolios are exposed to, meeting the original risk-return profile and avoiding being overly exposed to undesired bets.

Additionally, rebalancing represents the basis on which a disciplined and scientific investment process can meet its investment objectives. Amid today’s chaotic markets, a rational, scientific approach to investing is incredibly beneficial to unlock value in the long run. In this sense, rebalancing allows asset and investment managers to forecast risk more confidently, rather than being severely influenced by unexpected market movements.

Indeed, not rebalancing a portfolio can be extremely damaging for investors. In fact, as assets perform differently over time, their market values and their weightings in the portfolio will change accordingly. Even a portfolio that is initially well-diversified will become less diversified over time if no rebalancing occurs, as winners earn themselves higher weights and losers decrease to smaller ones. Consequently, if a portfolio is not rebalanced, its assets can become under- or overweight compared to the initial target. 

For instance, Exhibit 3 shows the evolution of the weights of two hypothetical 50/50 portfolios. The one on the left is rebalanced periodically and thus its weight resembles closely the original asset allocation, despite small periodic fluctuations. On the other hand, the one on the right (i.e. the drifting portfolio) increasingly gains more exposure to equity than originally planned, making it more vulnerable to a severe drawdown in case of a market downturn. 

Exhibit 3 - Rebalanced vs. Drifting portfolios: Evolution over time of the asset allocation of two hypothetical 50/50 portfolios.

In this sense, Exhibit 4 gives us further proof about the drifting risk that investors face when the portfolio is not rebalanced. The plot shows the evolution of the asset allocation of two 60/40 portfolios from January 1960 to December 2019. 

One portfolio is never rebalanced, while the other is rebalanced twice a year, at the end of June and December. We immediately notice how the drifting portfolio – the grey line – over time becomes increasingly more exposed to global equity, from the original 60% up to more than 80%. On the contrary, the red line shows the behaviour of the rebalanced portfolio, which is much closer to the original asset allocation. 


Exhibit 4 - Changes stock exposure for a rebalanced and a drifting portfolio, from January 1960 to December 2019. Source: Vanguard

Eventually, two important insights emerge, highlighting the need to efficiently factor rebalancing into asset allocation. On the one hand, sticking to the scientific method is the best tool investors have to maintain a consistent risk profile and consequently meet their own objectives. On the other hand, rebalancing protects investors from being unawarely exposed to hidden market dynamics. 

In this sense, rebalancing allows asset and investment managers not only to achieve actual diversification but also to build portfolios capable of adapting to the gradual unfolding of financial markets. Indeed, combining a dynamic rebalancing process with a forward-looking approach to diversification and to risk-premia allocation represents the key for investors to mitigate volatility and generate substantial value over the long-term.

ESG Investing: You Cannot Ruin an Ice Cream with Chocolate Sprinkles

Recently, environmental, social and governance (ESG) investing has grown exponentially, becoming one of the most important trends in finance. According to McKinsey, ESG-oriented investing now tops $30 trillion, up tenfold since 2004. This extraordinary growth has made many economists wonder about the existence of a link between ESG and performance. 

However, although there is little evidence that supports the idea that ESG can represent an investment strategy by itself, it is clear the role it plays in mitigating additional risks typically not accounted for (e.g. environmental, operational risks). 

As we discussed in our recent paper “ESG Is Not Enough”, there are several insights that emerge from a 360-degree analysis of ESG data and their integration into investment strategies. ESG-screened portfolios do not harm historical returns but can lead to very similar results to the broader market in terms of risk, performance and correlation. As a consequence, today ESG can represent a valuable starting point for more sophisticated investment strategies that can finally put performance and sustainability together.

In this sense, although being coined less than 15 years ago, the term ESG is connected to a wider branch of finance that has been around for centuries: responsible investing. Its recent exponential growth has occurred for several reasons. 

First, there is now greater attention to climate change and other environmental issues because of rising awareness about their potential dramatic consequences on human life. 

Secondly, it has become clear that social and environmental risks represent actual investment risks and can thus be an additional source of value creation or destruction. 

Lastly, investors have learned that ESG screenings can meet two goals at the same time: reducing portfolios’ risk exposures and increasing their efficiency. 

As a result, investors are reassessing their drivers of risk and value, expanding their traditional frameworks to evaluate the long-run prospects of a business. Among these, ESG indicators have become an increasingly adopted solution to account for companies’ externalities. Specifically, investors have been using ESG ratings as synthetic proxies to identify sustainable investments.

In general, ESG refers to a set of variables regarding a firm’s environmental, social and governance levels that can give valuable information about the long-term potential and attractiveness of a company. For instance, Exhibit 5 lists a set of commonly used – yet not exhaustive – variables to look for in an ESG analysis to monitor a company’s sustainability level.

Exhibit 5 - List of variables commonly included in ESG Analysis

As ESG is gaining more relevance in finance, asset and investment managers have been looking for the best way to concretely turn sustainability into real value. In our paper, we highlighted that ESG-screening may represent a Pareto-efficient choice. Indeed, an ESG-oriented approach does not destroy value and yields investment outcomes which are mostly in line with broad market performance. Consequently, they can be combined with Artificial Intelligence to create more sophisticated investment strategies.

Indeed, today Artificial Intelligence and Machine Learning can support investors to integrate ESG data into profitable investment solutions. More specifically, the impact of AI and ESG investing on portfolio management is threefold. First, AI can help investors conduct an in-depth and comprehensive analysis of the unstructured and raw data surrounding the environmental, social and governance profile of a company. Secondly, it can process this data efficiently, finding an effective and scalable way to identify the investment signal hidden in all the noise. Lastly, it translates the processed and cleaned signal into an actual strategy capable of supporting investors in their decision-making process.

While ESG investing continues its tremendous growth, asset and investment managers are paying increasing attention to how they embed ESG data into their analyses. To complete the puzzle, they need one last piece: to put and rethink ESG at the centre of the investment process, as a solid building block to develop more sophisticated investment strategies aimed at finally combining sustainability and efficiency.

The Bottom Line

As the world keeps getting more and more interconnected, the demand for a data-driven and adaptive approach to financial markets will keep getting stronger. All investors entering the new decade should consider whether they are using the right tools to navigate this complexity efficiently.

In this first part, we explored what are the key drivers of performance for asset managers in the 2020s and why correctly forecasting correlations will still play a major role for them. 

Moreover, we have discussed the significance of factor-based analysis and portfolio rebalancing, examining their impact on asset allocation. Finally, we have seen how AI can bring ESG investing to the next level.

In the second part, we will further investigate why modern portfolio theory still plays a relevant role in asset allocation and what math can teach us about investing. We will look at the mechanisms behind portfolio optimization and at the errors we must avoid to do it properly. Finally, we will show how AI and Machine Learning are allowing investors to build more efficient and well-positioned portfolios that follow the gradual unfolding of financial markets.

Start Integrating AI Into Your Investment Process