Multi asset: the road to normalisation

Aurèle Storno Chief Investment Officer, Multi Asset

How often do we read quotes such as “this time is different”, we live in a “new normal” or we have embarked on a “new paradigm”? The world is always changing and the ever-increasing news flow can often blur our understanding of what is driving recent evolutions of macro data and related market ups and downs. As a team of systematic portfolio managers, our primary objective is to maximise the "signal-to-noise" ratio in all aspects of our work. Any anomalies detected by our models always spark our curiosity regarding what constitutes noise versus signal. Furthermore, anytime we spot what may be significant abnormalities, we aim to distinguish whether they are likely to become the new normal or will eventually normalise. This quarter’s Simply Put focuses on exploring the concept of normalisation and defining what it takes for us to consider the investment environment as "normal" following the influx of novelties over the past five years.
 

Data evolution or data revolution?

The exponential increase of data and information certainly supports trendy techniques such as artificial intelligence (AI), but is all of this new information relevant? We are creating a world with more and more data, which we ultimately fail to navigate as it becomes overwhelming. So techniques such as those offered by AI are becoming increasingly popular. But is this a revolution or rather a natural evolution of an increase in available data, coupled with much higher computing power while using broadly the same old techniques (e.g. correlation/regression analysis, hypothesis testing, multivariate analysis, dimensionality reduction, etc.)? This convergence of factors will surely open doors to future innovation and research breakthroughs across all industries, which would eventually have happened anyway but may now occur at a faster rate. In other words, productivity has increased.

But what about investment decision-making in an ever-changing and data-overloaded world? Being quantitative investors, we have naturally investigated several AI techniques but, for now, have not implemented anything meaningful, as signal “explicability” remains a challenge in our view. We still believe there are long-standing patterns that we may want to identify systematically, using quantitative techniques and models: we aim to objectively trust these models on a daily basis while challenging them with our intuition and eventually risk-control everything – we never forget that things can go wrong or happen differently in a changing world!
 

Of old and new volatility regimes

In titling this section, we suggest that we are shifting back from an abnormal state to a normal one; from a statistical point of view, this means we were wandering in a dense fog and have now emerged into a clearer, more familiar world or one that at least inspires more comfort. We have come from an abnormal regime, as discussed in last quarter’s Simply put, which in simple terms was characterised by high inflation, high rates, an inverted yield curve and a positive correlation between equities and bonds. Maybe not a dense fog, but certainly not the norm. We argued this mix of ingredients was not supportive of future high risk-adjusted excess returns across the multi-asset space. But once again, markets have surprised us, this time on the positive side and we are glad we happened to be running a full risk exposure in our strategy for the last several months. Section 2 provides more detail on this.

As you know, we are always interested in understanding what the current risk regime is about, as we believe in the close link between risk and returns: we invest in “risk premia” to deliver an excess return above the risk-free rate. We therefore care about Sharpe and Calmar ratios and we naturally prefer more diversification. So let’s look again at the bond and equity duo, which drive most of the risk in our diversified portfolios, especially their respective volatility (we discuss their correlation in section 4,). Figure 1 shows the long-term evolution of commonly used bond and equity volatility proxies, the MOVE and VIX indices, since 1990. At first glance, we note long-term patterns: the two tend to move together much of the time, but they also have periods of disagreement. We have seen such a period recently and it reminds us a little of the 1990s.

FIG 1. MOVE versus VIX evolution

Source: Bloomberg, LOIM as at March 2024.
 

From rates vol and asset vol

However, we are less interested in figuring out whether the 1990s is a guide to investing today – we want to approach the topic without context or any timing dependence. We have argued in the past that instability in the bond market reflects discrepancies in investors’ expectations of future growth and inflation. We have also argued that when rate volatility remains high(er) for long(er) it is possible that other assets eventually become more volatile too. Now that’s not necessarily in sympathy, but rather because rates are the discounting tool for all future cash flows and when they remain unstable it eventually makes the whole valuation process more fragile and trickier.

The very recent past has seen bond volatility normalise a bit, which is good news, while it remains relatively high from a historical point of view. In figure 2, we look at the relationship between the MOVE and VIX indices over that same period. We split the MOVE index into quartiles, based on its three-month rolling average percentile. We then looked at how the VIX index behaves in the four sub-states of the MOVE, noting these initial intuitive facts:

• The VIX is higher on average when the MOVE is higher, and conversely
• The range of the VIX is narrower for the bottom quartile of the MOVE …
• … and broader for the top quartile of the MOVE.

These facts suggest that there is a link between volatilities, maybe reflecting our intuition that uncertainty in rates goes hand in hand with fragility elsewhere.

But let’s come back to our abnormal-to-normal topic. To illustrate this trend, we have highlighted in figure 2 where the MOVE and VIX stood in mid-2022, at the end of 2022, at the end of 2023 and most recently (21 March, at the time of writing). Clearly, the trend is for normalisation from a peak in stress during late September 2022. At that time, and in subsequent quarters, we wrote extensively on the topic: bond volatility stood in the 5th percentile and not just temporarily. Meanwhile, equity volatility was less extreme and starting to disconnect downwards on a solitary journey, while bonds were still struggling and stuck at very high levels for some time. In that environment, our long volatility strategies in the rates markets managed to deliver solid returns, while our allocation to sovereign bonds was on the low side. From September 2022 (when both assets were stressed) onward we have seen a regular normalisation of equity risk towards its median (conditional to MOVE’s level) in late December 2022. But in late 2023 and early this year, the disconnect between the two risk metrics was abnormally high, with equities reaching the low range while bonds remained in the top quartile. It is only very recently in March 2024 that the MOVE dropped a notch into the third quartile. Another step in the right direction, from this perspective.

FIG 2. VIX versus MOVE scatterplot

Source: Bloomberg, LOIM as at March 2024.
 

The path ahead

Things are obviously changing, but to what extent? Measuring this should help assess whether portfolios are still on track to deliver in line with expectations. At this stage, our view can be summarised as follows:

1. Disinflation continues but caution is warranted for Q2
2. A soft-landing scenario remains but the macro environment is decelerating, and
3. Correlations have just started to normalise and rate volatility only slightly

So this is a step in the right direction, but not the end of the road. This quarter, we are glad to share our thoughts with you once more, covering the usual breadth of topics. Section 2 highlights our portfolio activity and the rather stable allocation we have had throughout the quarter, which remains on the high side of our risk budget. Section 3 digs into the macro world and wonders what normalisation means for the economy and its relation to markets. Correlation is still a hot topic and section 4 hints about how we approach a topic that is potentially tricky for portfolios and certainly for a trading signal. Finally, we continue to research and are currently reviewing existing investments, such as long volatility and commodities, while preparing to deploy a new valuation-based strategy into our overlays, see section 5 for more on this topic.

Alain Forclaz Deputy CIO, Multi Asset

Following the November CPI report and the change in tone from central bankers at the beginning of the Q4 2023, our market exposure has swiftly normalised. Since then, we have maintained, and even slightly increased, a significant capital exposure to the markets. This is particularly noteworthy when compared to the past two years, as our notional exposure to markets has approached its historical average from historical low levels experienced previously, while our asset class allocation still struggles to achieve such normalisation. In fairness our All Roads allocation now includes a higher proportion of bonds compared to the end of Q4 last year, driven by sentiment, macroeconomic factors, and risk considerations. However, we are still relatively distant from the average observed since the launch of our strategy. The diversification benefits of bonds do not currently present a compelling case. Our preference continues to lean towards cyclical assets, with a slight appetite for credit, equities and commodities at the moment. This section provides a comprehensive 360° perspective from our investment cockpit.
 

Duration risk gradually normalising

It has become somewhat of a tradition: in the beginning of each of the past seven editions of this quarterly publication, the portfolio management section highlights the high volatility of bonds compared to their historical levels. Figure 3 illustrates our risk dashboard, showcasing the volatility estimates of the risk premia that make up our investment universe. It highlights that developed and emerging equities, credit, and commodities are below their median levels of risk, falling within the 1st quartile of their history. This situation has further improved over Q1, with a continued decline in volatility, aligning with the positive market sentiment that has driven investor activity since the beginning of the year.

This positive sentiment is evident in the performance of equities and the strong interest investors have shown in large credit bond issuance programmes during Q1 in the primary market. However, the area where disproportionally high volatilities can still be observed is in duration, and to a lesser extent, inflation. Although duration volatility has experienced a modest decline, it remains in the higher quartile as we analyse these data. We have discussed various aspects of this situation and its underlying factors in different pieces, ranging from the persistently volatile inflation environment to the uncertain prospects of monetary policy and the apparent divergence of views among investors¹.

Currently, we view this prominent characteristic of our All Roads mix as a cyclical element that is gradually evolving into a more structural component. Rates volatility remains high, although it has recently shown a decline, and investors must learn to navigate and adapt to it.

FIG 3. Risk premia volatilities

This chart shows the timeseries evolution of our proprietary volatility models per risk premia. Dotted line shows historical median and red zone shows the 4th quartile of volatilities.

Source: Bloomberg, LOIM as at March 2024.
 

Trend is your very happy friend

Market sentiment has provided incredible support throughout Q1, particularly evident towards the end of the quarter. In the previous quarter, we highlighted the favorable performance of developed equities, credit, and commodities, while bonds, emerging equities, and small cap stocks lagged behind. Over the course of Q1, the broadening of the rally is clearly depicted on our trend dashboard, shown in figure 4. As the quarter draws to a close, not only do equities exhibit a robust uptrend, but previously overlooked segments of the market have progressed enough for their signals to shift from red to green. Emerging equities and emerging assets, in particular, appear to be attracting investment flows once again with the exception of Chinese equities. Bonds also showed some positive trend this quarter although signals are weakening at the time of writing. As a strong undercurrent for these positive price signals, our risk appetite indicators reveal a similar picture.

FIG 4. Trend-following signals

Source: Bloomberg, LOIM as at March 2024.
 

A strong bull consensus

Our risk appetite signal continues to reflect strongly bullish market sentiment. The various signals that make up our indicator convey a consistent message. Firstly, risk appetite is currently high, without a doubt. It surpassed recent historical levels, hovering around 90% on our risk appetite gauge, which ranges from 0% (indicating extreme risk aversion) to 100% (representing extreme risk appetite). Secondly, when examining the individual underlying models, their level of agreement is unusual compared to the period following the recovery from 2020. The black line representing the average across the signals is closely aligned with each of the three individual signals. Regardless of how we assess risk appetite, whether through forward-looking, market-based metrics, or historical models (risk/reward analysis, or the behavior of risk-on assets over time), the conclusion remains consistent across the board.

The current strength of this signal complements the cross-asset trends mentioned earlier. Markets have experienced a robust Q1, and the concern now is that this consensus may collide with previously disregarded elements of the backdrop, such as the economic situation and escalating political risks. In the following paragraph, let us delve into the macro backdrop.

FIG 5. LOIM Global Risk Appetite index and components

Source: Bloomberg, LOIM, as at March 2024.
 

Soft-landing across the board

As our regular readers will be familiar with, our market-based investment signals are complemented by our macro indicators, the dashboard of which is presented in figure 6. Throughout Q1, the growth situation has been showing significant improvement, consistent with the trend we have been identifying since last summer. However, as the quarter drew to a close, a series of data points marked the end of this "recovery without a recession," potentially dampening market enthusiasm. Inflation, on the other hand, continues to exhibit an upward trend from low levels (i.e. consistent with slowing disinflation rather than outright resumption of inflation). Surprising inflation numbers, particularly in the US, could introduce a second potential headwind for the markets.

Meanwhile, our monetary policy signal continues to reflect the transition of central banks from a hawkish stance to a more dovish one. As long as this shift is not accompanied by a genuine recession, all three indicators remain consistent with the narrative of a disinflationary soft landing that has supported the markets since November. Taking these factors into account, our macro signals have been favoring emerging assets throughout the quarter, and given the deteriorating growth trend, they have also shown a preference for bonds.

FIG 6. Macro nowcasting signals

Source: LOIM at March 2024. For illustrative purposes only.
 

Foot still firmly on the accelerator

All told, the primary evolution in our asset allocation over this quarter is a slight increase in our bond and commodity allocation, at the expense of equities. Figure 7 illustrates the breakdown of our All Roads portfolio, separating market exposure from an allocation rebased to 100%. Of note, our market exposure has continued to normalise since the end of Q4, reaching ~140% at time of writing.

This results in no small part from running at maximum risk budget across our portfolios. Mechanically, the only way for risk budget from here on in is down, so our strategies stand ready to lower our market exposure if markets were to enter a meaningful turn for the worse.

As far as allocations, bonds and commodities benefited from a slightly reduced volatility, the positive trend signals, and the shift in our growth signal.

Nevertheless, we are still far from our average historical allocation for bonds, which is around 45%. Achieving this target would require a more favorable macroeconomic backdrop, stronger and more consistent trends, and most importantly, a substantial reduction in bond volatility to rebuild their diversification potential. The correlation analysis conducted in section 4 does not currently indicate a strong case in favor of such diversification².

FIG 7. All Roads allocation, decomposed

Source: Bloomberg, LOIM, as at March 2024. Holdings and/or allocations are subject to change. For illustrative purposes only.

Florian Ielpo Head of Macro, Multi Asset

Post-COVID-19 disruptions have caught many seasoned market observers off-guard, creating the temptation for them (and us) to rethink some of our long-term convictions. Across the macro and market spheres, three of the areas continue to catch our eye. First, the reliability of leading indicators, which traditionally hold a strong relationship with GDP growth figures, has come into question due to an intriguing anomaly observed over the past year that growth has been much stronger than the levels these leading indicators steer economists to forecast. Second, the structure of earnings across different indices has revealed disparities between corporations and their ability to capitalise on economic growth. Third, concerns arise over the market pricing of earnings, with the potential existence of a bubble-like state in equity indices. As we navigate through these complexities, it becomes crucial to examine the need for normalisation and its potential implications for the broader economic and market landscape from all these angles to help better prepare us to face Q2 from a fundamental perspective.

Have nowcasting indicators become useless?

In the realm of economic analysis, there is typically a reliable relationship between the leading indicators of the US economy and the corresponding GDP growth figures. However, the past year has presented an intriguing anomaly in this regard. Despite the low levels indicated by the Institute for Supply Management (ISM) index, the official hard numbers revealed unexpectedly strong growth. When a regression analysis is conducted to examine the connection between GDP growth and leading indicators, it has consistently shown a significant disparity throughout the year, as illustrated in figure 8. This discrepancy between actual growth and the customary monthly indicators surprised economists (and us) as it deviated from the expected pattern. A similar situation occurred in the late 1990s when a slowdown in Asia influenced leading indicators in the US without significantly affecting actual growth. During that time, growth surpassed expectations by approximately 4.5%. However, it is important to note that such deviations are historically temporary. We can now anticipate that the relationship between leading indicators and actual growth will normalise. This normalisation process will restore some of the predictive power to the nowcasting indicators that guide the economic analysis of most economists. The macro signals described in the previous section are, at the moment, showing how the process of a disinflationary soft-landing is continuing:

• Growth pressures remain low. After a phase of acceleration, economic indicators have been turning south once more, negatively surprising economists at the end of Q1
• Inflation pressures are showing signs of potential growth in the US, while continuing to fade in Europe
• Monetary policy signposts so far remain on the dovish side

The situation of slower disinflationary growth experienced in the previous year is expected to have a broader impact on expectations for the current year. The Federal Reserve anticipates a 2.1% real GDP growth rate in the United States, while analysts forecast a moderate increase of approximately 5 to 7% in US stocks' earnings. These projections align well with our indicators and do not appear to be contradictory. The key factor necessary for a comprehensive understanding is the normalisation of the relationship between these indicators and the eventual reports on GDP. Once this relationship is restored to its usual state, a clearer picture will emerge.

FIG 8. US GDP growth vs estimation from leading indicators (left) and sorted spread between the two (right)

Source: Bloomberg, LOIM as at March 2024.

Something’s off with earnings

Another important aspect that warrants careful consideration and normalisation is the structure of earnings across different indices. One particular area of focus is the so-called "magnificent seven" companies, which have garnered significant attention in recent times. These companies not only command attention but also generate substantial revenues. However, when evaluating the broader landscape of US corporate earnings in relation to GDP, a notable disparity becomes apparent. In general, corporations – especially small and mid-sized enterprises – have struggled to keep pace with GDP growth and harness its full potential for their own benefit. To illustrate this disparity, figure 9 presents a compelling chart that compares US corporate earnings with US GDP. The chart highlights the divergence between these two phenomena, emphasising the challenges faced by corporations in capitalising on economic growth, resulting in their lag behind GDP expansion. When assessing the overall earnings of US corporates, they find themselves approximately 7% behind their typical connection to US GDP. Shifting the focus to the earnings of companies in the S&P 500, a different narrative is unfolding. S&P 500 companies have exhibited a remarkable ability to extract substantial revenues from both domestic and international economies. Consequently, their earnings have surpassed the pace of GDP growth. This phenomenon is evident when conducting comparative calculations. Rough estimates suggest that S&P 500 earnings stand approximately 25% higher than their normal relationship with global GDP. Furthermore, these earnings are approximately 13% higher than their historical relationship to US GDP. The discrepancy between this situation and the larger set of US corporates sheds light on their challenges in general, as they strive to align their earnings’ trajectory with the pace of economic growth. 

The abnormal relationship between economic growth and earnings growth raises the imperative for normalisation as the economy navigates a soft landing. The expectation is that this normalisation process will pave the way for GDP gains that benefit local and smaller-sized companies. By promoting a more balanced economic landscape, diversified equities and credit exposures can help investors profit from this normalisation. Should this soft-landing process be a success, the growth-earnings normalisation process could foster greater stability and resilience within the market.

FIG 9. Earnings and profits compared to their nominal GDP-implied levels
 

Source: Bloomberg, LOIM. As at 22 March 2024. For illustrative purposes only.

Market valuations to normalise too?

The process of macro and earnings normalisation raises a pertinent question regarding the market pricing of these earnings. The complex and disrupted relationship between earnings and economic indicators may have significant implications for the pricing of major stock indices. In fact, it leads to a concerning observation: equity indices could currently be in a bubble-like state. This realisation could serve as a compelling reason to exercise caution and consider stepping away from equity and credit markets, as there is a potential risk that could emerge in the near future – but we think that situation deserves a second look.

However, it is important to note that this risk does not stem from fundamental valuation arguments. The charts presented in figure 10 provide valuable insights into the relationship between equity indices and their prospective earnings. These charts indicate that when regression analysis is applied to assess fundamental valuations, the current pricing of equity indices does not exhibit significant exaggeration. In the United States, the S&P 500 appears to be approximately 10% away from fundamental valuations, while a similar pattern is observed for the Euro Stoxx index. Smaller cap stocks worldwide, on the other hand, look fairly priced.

Overall, it becomes evident that there is more work to be done in terms of macro and earnings normalisation than in how markets are currently discounting this information. This observation is crucial to keep in mind as we enter into the second quarter. While the macroeconomic and earnings landscape requires attention and adjustment, the market pricing of these factors does not demonstrate extreme deviations from fundamental valuation. 

FIG 10. Equity indices vs earnings-based regression (top) and difference between the two in % (bottom)

Source: Bloomberg, LOIM, as at 22 March 2024.  

Sui Kai Wong Senior Portfolio Manager, Multi Asset

An important ingredient of our investment approach is the control of risk and we have provided a fair amount of content on this matter, with a key focus on its premier element: volatility. Since the emergence of the inflation wave, we have been vocal about the increase in fixed income volatility and, notably, its persistence. That element alone has played an important role in adapting our asset allocations to this new environment. However, not all risks are captured by volatility, and without even touching on extreme risks, we want to look at the topic of correlation. In order to deliver a “quality” stream of multi-asset returns, correlation structures are an essential input to portfolio construction, as they are the number one ingredient to cross-asset diversification. Given this quarter’s Simply Put focuses on the normalisation of markets and the economy, our special focus will explore the essential topic of multi-asset correlation fluctuations, considering whether the Federal Reserve’s pivot and incumbent rally have been enough to restore the historical function of correlation and therefore given diversification back some of its past appeal? This is an important consideration for multi-asset portfolio managers such as ourselves. Here is what our models and estimates have to say. 

Some have, some have not

A risk model is at the basis of all that we do. That risk model encompasses a mixture of different models aimed at capturing the fluctuations of volatility and trying to differentiate between signals and noise within asset returns. The second key feature of that risk model is our correlation model. It is well-known that portfolio construction is particularly sensitive to misspecifications of correlation (due to the inversion of the variance-covariance matrix), which can lead to erratic portfolio allocations if too unstable or too reactive. Therefore, it is crucial that we also discern between the signals and the noise when it comes to correlation. A stylised fact about the correlation between asset classes is that it can evolve over time: many investors use static long-term correlation assumptions which can be misleading, while short-term assumptions can prove unreliable. To tackle this, we use a Dynamic Conditional Correlation (DCC) model as the basis for our correlation model. Notably, DCC was developed by Robert Engle, as a natural extension of the GARCH (Generalised AutoRegressive Conditional Heteroskedasticity) model in the multivariate space.

Figure 11 highlights the recent trends we have detected in the different asset classes that make up our All-Roads portfolios, using estimates from our DCC model. When comparing risk premia correlations over the 2019-2024 period, with the one observed in mid-November 2023 (probably the most extreme correlation regime we have seen so far) and in mid-March 2024, clear trends appear. First, the commodity-bond correlation has remained unaffected by the unusual period we have just been through and has remained marginally negative. The true oddity is in bonds’ relation to risky assets such as equities and credit spreads. The unusually positive equity-bond and credit-bond correlation seen in the price action of markets in November last year has recently seen a reversal. It is now slightly negative in the equity-bond case and slightly positive in the case of credit. This is not enough to be considered a normalisation, but this improvement raises the question: are we witnessing a burgeoning trend back to the 2019-2024 period or not? It takes further modelling steps to answer that question.
 
FIG 11. Recent trends in the correlation between risk premia
   

Source: Bloomberg, LOIM as at March 2024.

A tale of long-, mid- and short-term trends

While the correlation between assets may be dynamic, we are mindful that some of the short-term changes come from exogenous shocks and are therefore unpredictable. We find that risks from these shocks can be better managed through signals elsewhere in the portfolio construction process, such as volatility and momentum signals. For correlations, we want to pay particular attention to the structural elements that are driving changes in relationships between asset classes. To that end, we add an important macro element that helps link these dynamic correlation estimates to possible fundamental drivers. 

Our risk model extracts a medium- and a long-term component from these correlation estimates (via the DCC model using monthly returns). This decomposition is obtained by crossing these correlation measures with our growth and inflation regime measures. The medium-term focuses on the regime-based patterns of correlations over the past five years, while the long-term component reflects the same regime-based analysis performed over the entirety of our dataset. When we contrast these estimates with the DCC correlation measure based on daily returns, we can further decompose correlation into a short-term, noisier, component i.e. the residual. This represents the dynamics of the recent co-movements in price actions. The medium-term component highlights the typical correlation between risk premia during the current economic regime based on their recent evolutions, while the longer-term analysis does the same for past decades – providing a secular perspective. 

Figure 12 shows the outcome of these calculations. The conclusion from our previous analysis remains: the oddity of the period and the element requiring normalisation is not the commodity-bond correlation for it has been stable during the changes in recent macro regimes. It is the equity-bond and credit-bond correlations that are again at the heart of the current correlation disruption. Our daily estimates have increased across the board and are now showing a clear uptrend for the entire period – not just since 2021. This uptrend fails to be apparent in the long-term analysis – it is too recent to be seen as a structural shift of markets. The most interesting element here is probably the red line: for equity-bond correlation, the mid-term component has advanced beyond daily estimates since the summer of 2022. Now, in the current environment of declining growth and inflation, this result shows a pervasive weakening of the diversification power of bonds within the context of multi-asset investing. A trend that is doing the opposite of normalisation as it has been rising almost continuously over the past 18 months. The effect is smaller in the case of the credit-bond correlation, but the direction of travel is the same. For now, daily estimates may be pointing to normalising correlation, but our decomposition based on macro regimes seems to send another message: the “last mile” of correlation normalisation could take longer than widely expected – that’s the message from our regime-based approach to measuring correlations at least.

FIG 12. Daily correlation fluctuations versus trend components in our correlation measures
 
Source: Bloomberg, LOIM as at March 2024.

Julien Royer Quant Analyst, Multi Asset

Multi-asset portfolios have the advantage of benefiting from cross-asset diversification. Diversification may however be a double-edged sword. While volatility diversification entails the good behaviour of risk-based portfolios when asset returns are near the centre of the distribution – the most frequent case – such portfolios face two hurdles. On the one hand, volatility diversification may hide risk concentration in tail events, for example as assets re-correlate in times of market stress. On the other hand, diversification may underweight assets with strong performance, penalising the portfolio in good times. The addition of Alternative Risk Premia as overlays to risk-based portfolios mitigates both effects and leads to a more balanced portfolio in any market scenario.

Concave or convex: harvesting diversification in the tails

Alternative Risk Premia may spuriously appear as any standard investment strategies. As they are transparent, rule-based, systematic strategies, they are sometimes perceived as passive investments that should be treated like any asset in the construction of a portfolio. They are in fact more complex objects, presenting asymmetric payoffs that should be carefully handled in portfolio construction to enhance tail risk diversification. On the one hand, concave strategies present positive asymmetry but often come with a fat left tail: returns are on average positive but can exhibit sharp drawdowns. On the other hand, convex strategies present the opposite payoff – a fat right tail but negative asymmetry – as they can present large positive returns that come with the cost of small negative returns on average. 

Carry and trend-following strategies are probably the most well-known examples of concave and convex strategies. The former can enhance performance, mitigating the potential over-diversification of risk-based portfolios in good times³, while the latter acts as a tail-hedge strategy in bad times, acting as a tail diversifier. Figure 13 provides an intuitive illustration of this asymmetry that we borrowed from Lemperiere et al. (2017)⁴. The chart presents the cumulative performance of standard cross-asset trend and carry strategies from 2009 to 2024 in a non-standard way: the monthly returns are ordered by their amplitude rather than by date. The right part of the chart thus captures the returns of the strategies with the largest amplitude whereas the left part shows returns closer to 0. This allows us to see that in the case of carry, the most extreme returns are negative and erase nearly 80% of the strategy’s cumulative performance (which shows positive asymmetry with fat left tail). In the case of trend, the most extreme returns are positive and largely compensate for the more frequent small, negative returns (which shows negative asymmetry with fat right tail). Over the period, both strategies delivered a comparable return (3.5% for carry and 4% for trend) and are relatively uncorrelated (-12%). Moreover, both strategies are decorrelated with a standard volatility parity portfolio constructed using Global Aggregate bonds and Developed Market equities (32% for carry, -18% for trend) emphasising the diversification benefits of combining both concave and convex strategies.

FIG 13. Illustration of the non-normal distribution of Alternative Risk Premia

This chart shows the cumulative performance of cross-asset trend and carry strategies when the monthly returns are ordered by their relative amplitude.
Source: Bloomberg, LOIM as at 29 February 2024. Past performance is not a guarantee of future results. For illustrative purposes only.

Since January 2022, two features of the carry/trend pair have been noteworthy. First, the correlation between the two ARPs is at its lowest point for more than 10 years. In particular, carry struggled in 2022 when market conditions deteriorated while trend returned a positive performance before inverting its path in 2023 when carry rebounded. Interestingly, year-to-date, both strategies are positive and correlation has started to inflect, another sign of normalisation.

FIG 14. Three-year rolling correlation between cross-asset trend and carry

Source: Bloomberg, LOIM as at 29 February 2024. For illustrative purposes only.

Diversifying convexity 

While long-only multi-asset portfolios – however diversified – present a concave payoff, convex strategies are paramount to obtaining a good risk-adjusted return. Trend-following strategies are often paired with long volatility strategies that mimic a long put position to hedge extreme downside risk. Such strategies may be implemented in different ways depending on the asset class (rates volatility, equity volatility) and the convexity profile. Figure 15 presents the convexity profile of the rates and equity strategies that we use in our All Roads portfolios. It is noteworthy that over a long horizon⁵, all implementations yield a remarkably stable convex profile.  

Figure 15. Long-volatility strategies convex profiles

This chart shows the dispersion of monthly long volatility strategies returns as a function of the monthly returns of S&P500 and 10-year US Treasuries. Dotted lines represent the long-term realised convexity profile of the strategies while the dots represent points from the January 2022 to February 2024 period.
Source: Bloomberg, LOIM, as at 29 February 2024.

The period from January 2022 to February 2024, however, represents an anomaly when compared to the historical behaviour of these strategies. Indeed, over the period most equity long-volatility strategies have failed to benefit from the negative returns of their asset class. However, rates volatility strategies have helped mitigate the effect of higher interest rates, emphasising the necessity of having multiple convex strategies to efficiently hedge multi-asset portfolios.  

As markets normalise, concave payoffs – including our core portfolio – should benefit from favourable market conditions. But the last two years have been a stark reminder that drawdowns can take many forms and tail risk diversification is key. Recently, we have researched a new long-volatility rates strategy and developed a novel Valuation strategy⁶ that will help us deliver performance independently of the roads paths markets may take.