The day after #10 - Rethinking the macro and cross-asset research: what we have learned from the Covid-19 crisis

Covid-19 accelerated markets’ entropy and risk takers will evolve their investment approaches

Crises create disruption and Covid-19 is no exception, bringing new complexities, new opportunities and new risks to the investment landscape.

From our research angle, we realised that the pandemic triggered:
1. A fast-evolving macro financial environment.
It has become much more difficult to anticipate and to predict performance, in both the short and medium term, because of the lack of hard data (even the calculation of consumer prices has become difficult because of lockdowns!) and the unknown consequences of the unconventional policies put in place.

2. A common global ‘health’ shock affecting economies and spill-overs to the full spectrum of asset classes.
Financial market volatility initially spiked on increased uncertainty regarding the short-term effects of the virus on lives and due to the potential long-term changes to livelihoods. Later, a sequencing of policy announcements, measures implemented and pandemic-driven news resulted in volatility increasing.

3. A deep disconnect between macro fundamentals and markets.
We have seen a crisis of global confidence and a sudden contraction of global growth. Asset classes saw very significant declines based on the common ‘unknown’, the virus. Then, they bounced back based on a liquidity-driven rally fuelled by central banks’ implementation of aggressive accommodative policies. In contrast to the past, economic policy interventions have occurred very quickly, and CBs have been creative and unconventional in their responses to the emergency. Analysts had to provide forecasts at a time when companies stopped providing guidance on earnings per share and top lines. Policy accelerators caused risk assets to drift higher, overshadowing the pain being experienced by the real economy.

4. A larger role for central authorities in market functioning. Monetary policies in particular have been overwhelming asset classes’ specific drivers, blurring market correlations and inducing (dangerous) mispricing.
Central banks increased their stimulus over and over again. They have been crossing new frontiers¹, elongating the cycle once again via more debt creation. They are
managing yield curves while increasing the risk of asset bubbles, should fiscal policies not deliver as needed.
Governments addressed the health emergency first (international travel restrictions, domestic measures, improvement of medical treatment) and then implemented fiscal policies, not only reacting (state aid/grants), but also trying to be pre-emptive in order to address the profound damage Covid-19 will inflict on labour markets and growth.

The net effect of these interventions remains difficult to ascertain, but it is clear we need to evolve our thinking in order to analyse this new macro financial ecosystem.

How we have analysed this evolving situation and our findings

We soon realised that the global health crisis was having a huge impact on economies via an unusual shock potentially triggering some structural changes. However, we remain convinced that macro factors anchored financial markets. Growth, inflation, rates, policy efforts and liquidity are likely to remain pivotal, so, the nature of this crisis requires an evolution in the way we look at these factors.

In particular:
1. We moved away from the traditional macro/markets/micro causalities towards using non-linear big data nowcasting indicators. They will help trace the new cross-sectional risk factors commonalities that we believe will represent the new market movers. In this way, we track asymmetry and velocity of the current macro environment to fine-tune forecasts.
During the Covid-19 outbreak, our Bayesian approach to macro forecasts has been providing highly volatile outcomes which are subject to continuous re-assessment. 
The overall fluidity of the situation, the lack of hard data, and uncertainty on the size and timing of fiscal policies are distorting the contribution to the fundamental identity of national income accounting. Consumption, investment, public expenditure and net trade are running on evolving statistical relationships. They are constrained by a large number of assumptions and they will likely face a structural break. 
Are consumers’ preferences structurally going to change? How fast and synchronised will deglobalisation be? Which sectors will see local on-shoring? Will their weights in the economy change? When will labour-intense sectors, such as hospitality, recover? The answers to these questions will likely affect productivity and address growth in the medium to long term. However, as risk-takers, we needed to get a better focus on the short-term reality to understand the appropriate investment positioning. Therefore, we complemented our structural approach to macro forecasts with the support of fast data and micro evidence.
– We utilized big data to process a vast quantity of panel data, being conscious that n-dimensional problem solving required managing large databases. This shift needed higher analytical capacity.
– We expanded our coverage to almost 300 asset classes to catch the deeper co-movements that had been taking place across and within asset classes.
We thought about valuable regularities but we needed to detect them automatically: technology with the Python coding has been of great help. The major advantage of learning system is the ability to adapt to a changing environment. At the same time, the identification of specific financial regimes², rather than random sampling, improves the application of machine learning to investments, automating the calculations at a time of growing data management complexity.

2. We built GREAT (Global Risks Exposure Attribution Tool) to systematise our analyses.
Once we verified that the macro factors we had identified are the key drivers of financial markets, we realised that the factors’ asymmetric profile plays a prominent role. For example, DM equities’ sensitivity to prices (CPI, PPI) shifts from positive when in a period of disinflation/reflation to negative in hyperinflation. These sensitivities shifts are asset class-specific, even within the same cluster (equity, fixed income and commodity). For example, Japanese equities’ sensitivities to monetary aggregates is three times the sensitivity of US equities. As a consequence, we identified for each factor a ‘low’, ‘medium’ and ‘high’ level and distinguished the sensitivity of each asset class in the three cases (“importance of the macro factor’s first derivative”). Lastly, we ensured that in addition to the statistical significance, results were economically sound. All this information will be entered into GREAT and used for top-down assessment of risk exposure, as we will discuss in the next paragraph.

3. If macro factors influence markets dynamics in a more complex and unusual manner, then a macroeconomic (risk) factor-based approach³ is suited to capturing the broad common risks across an investment universe. This is why we calculate asset classes’ sensitivities to these factors on some proxies based on our internal forecasts and elaborations.

We revisited and consolidated our macro factor-based approach to asset allocation to enforce portfolio diversification, reduce volatility, and possibly improve portfolios outcome.
The investment universe continues and will continue to fast co-move with diverse sensitivity to common factors that are now driving the markets: namely, central banks’ support for markets and pandemic- related news. As a result, correlations are biased and unusual: for example, the mid-March capitulation, when all asset classes were drifting lower, followed by the most recent rally, which has included fallen angels.

Moreover, the risk is that such an environment exacerbates the weakness of traditional asset allocation which operate on silos of asset-class bucket holdings (equity, fixed income, commodities, FX). Covid-19 has been the opportunity for our Regime-Based Approach to move one step forward towards operating in a more explicit factor-based fashion, using non-linear thinking and artificial intelligence to empower the search for cross-asset opportunities amid unconventional market behaviours. In our opinion, a macro (risk) factor⁴ asset allocation that allows for the budgeting of risk assigned to each macro factor while positioning cross asset according to the investor’s high convictions⁵ is the right approach in an environment in which macro factors are not only influencing markets, but are the “minimum common factor” on the a global scale.⁶

GREAT allows for the calculation of portfolio exposure to risk factors (beta) while tracking beta changes and risk budget consistency when different factor tilts are input.

The entire process has been fascinating and as well as challenging.

Remaining in control of the full process was paramount. Our regime-based framework articulated in the base and alternative scenarios with some clear references in terms of macro-economic assumptions and financial consequences had been quite helpful to fix reference values to initialise with some a priori the clustering procedure when flooded with tons of results. Furthermore, we also had to move beyond the comfort zone of time series analysis and parametric estimation. Consistency, reliability and significance of results become more challenging when directed at large volumes of information and evolving structural relationships. Machines deliver faster than humans, but it is the human brain that has to define the problem, to apply the most suitable pragmatic technique, and, eventually, to use common sense and judgement to interpret the significance of the final solution.We haven’t changed our thinking in that regard.

Factor-based asset allocation cuts across traditional asset classes, diversifying returns streams on a ‘truly’ cross-asset dimension.

We identified these macro (risk) factors as the key top-down discriminants cross assets amid the five financial regimes we defined ex ante. We derived state-dependent sensitivities for our investment universe.

As the path to reach the new equilibrium in the day after will not be straight, we have been adopting non-linear thinking to elaborate big data aimed at improving the comprehension of asset class sensitivities around risk factors.

Connecting dots

The pandemic made a virtue of necessity.

1. We penciled in base and alternative scenarios and framed them into our financial regimes’ mapping to spot some asset class behaviour references.
Our base scenario sees a global deep, short-lived recession in H120, followed by a short rebound in Q320. In a sequencing progression, regions will slowly revert to pre-crisis levels (EM in Q221, AEs Q422). The game changers for the alternative scenarios are the discovery of a vaccine and the effectiveness of policy boosters on the real economy⁸. We then populate this frame with the information resulting from high-frequency macro and micro evidence to add more accurate fluctuations around the medium-term trends detected by the scenarios. 
The pandemic altered our perception of financial regimes’ cycles compared with the beginning of the year. We entered a contraction (risk-off phase) possibly bottoming out in Q320. Thanks to policy boosters we will progress into a sequencing revovery with significant divergences both across and within regions.
In this environment, growth, rates, inflation, monetary and fiscal policies are strongly interconnected. The potential mismatch of one regarding others could affect the results overall. If we change perspective, though, this situation is building common factors that are now driving the markets. As a result, markets will move with diverse sensitivity to these common factors, paving the way for a macro factor-based asset allocation.

2. We calculate expected returns on a broader asset class spectrum and frame them into base and alternative scenarios. We accelerated and sharpened our use of artificial intelligence to capture non-linear connections between macro factors and markets and dynamically assess the exposure to the macro (risk). To control and ensure the consistency of the outcome in terms of market sequencing and investment consequences, the algorithms run within the base and alternative scenarios information window⁹.

3. We use GREAT to nest our high convictions (including those on risk factors specifically) and derive the asset allocation preference.

High convictions and asset class preferences

Recovery is our central case. Hence, growth factors remain paramount. The recession should be confined to H120, followed by a short rebound in Q320, and then slow and bumpy convergence to pre-crisis levels (2022). Price dynamics should evolve from disinflation to below CB targets, with potential spikes down the road on components base effects (oil). Policy accelerators support risk assets. However, the decoupling from their fundamentals increases downside risks. In the medium term, we expect EPS to bounce back in a V-shaped fashion.
– GROWTH factor to increase from low to medium.
– INFLATION factor to stay low.

We expect ultra-accommodative monetary policies to persist, leading to stable and low interest rates worldwide. Potential upside for rates will be tempered by the strong demand coming from central banks while the recalibration vs the short end of curves will ease tensions in the long end. In fact, governments are financing the emergency with short-dated issuance. Central banks’ purchasing programmes and state guarantees safeguard default rates, at least in the short term. Turbocharged monetary policy should support easy financial conditions. Carry appeal in a low-yield environment overcomes the compressed expected returns.
– DEBT& MONETARY AGGREGATES from medium to high
– NOMINAL RATES FACTOR to stay low.
– LIQUIDITY FACTOR to stay high.

In the table, we present the asset class preference according to our outlook convictions. Sensitivity to (geo)political risk is not explicitly considered. This is when we intervene with the qualitative layer to exclude PXJ and Latam equities from the recommendations, for example. Govies, gold and IG tick almost all our boxes, while European and US equity represent a more tactical play.

Learning never exhausts the mind

As a result of Covid-19, the financial world has become increasingly complex, irrational, non-linear and does not fit into a rigid and static analysis architecture. Macro factor expectations are more than ever driving financial markets.

The crisis triggered a transition to a new economic and financial equilibrium dynamically evolving into waves.

The current situation is a perfect storm: the world’s entropy has increased exponentially.

In such an environment, we believe that assessing macro risk factors is key to generating returns for investors and a cross-asset approach is required. In our opinion, tilting to individual macro-economic risk factors and combining asset classes under this new framework should facilitate diversification, improve risk budgeting and eventually improve portfolio returns.

Very often in the past, many investment ideas perceived as alpha opportunities have been involuntary beta transfer on factors not fully understood by risk takers. For this reason, a disciplined top-down approach that evaluates a factor in its articulation and interdependence across asset classes allows for the transformation of a specific risk into clear investment opportunity.

Research methodology has to evolve to fit the need to find answers to external shocks quickly in a flexible manner.

Over the last quarter, we moved forward on our implementation of machine learning and big data to economics, taking advantage of the learning systems’ ability to adapt to a changing macro financial environment.

Our approach to asset allocation evolved as well: we shifted from asset class to risk factor diversification, as Covid-19 is a shock shared across all economies and asset classes. Risk exposure is global, but the impacts on asset classes are local.

We believe that – in order to be competitive – asset mixes will need to be defined on factor tilts rather than traditional risk/return combinations. In addition, looking at return streams, portfolio objectives will need to shift from being benchmark-based to being goal-oriented. This will decide any change to asset allocation⁷.

You cannot use up creativity. The more you use, the more you have.

Dr Maya Angelou