A Dynamic, Turbulence Based Approach to Active Management and Portfolio Rebalancing.

?

The problem.

Confronted with the perennial challenge of portfolio construction, the timing of new investments and the periodic asset rotation, investment managers’ set of options can span anywhere on the spectrum between purely passive to highly active approaches. Should one adhere to a passive approach and a formulaic rebalancing rule to a pre-determined strategic asset allocation (SAA) on a periodic basis, or should one exercise discretion, and if so -- what should guide such discretion?[i]

Evolution in the financial landscape.

Adding to the difficulty of the above question, has been the evolution in landscape of financial markets in the last 2 decades or so. In particular, the elevated degree in co-movement of risk assets and by deduction the diminishing importance of idiosyncratic factors. During periods of market distress, we often observe co-movement among asset classes on the downside that we characterize as ‘contagion’, and in non-distressed times we observe co-movement on the upside, cunningly referred to as a 'junk rally' or market ‘meld-up’. In both circumstances, asset returns are largely independent of idiosyncratic factors and driven primarily the so-called ‘global risk factor’ – a catch-all proxy for the elusive concept of risk aversion. [ii] [iii]

To demonstrate this phenomenon, we estimate the so-called Absorption Ratio, a measure that captures the variability in asset returns explained by the first two principal components (on a rolling 12-month basis) for a representative opportunity set of liquid assets. During the last 50 years this ratio varied between 95% and 60%, however, in the last 2 decades, the frequency of this ratio in elevated, 4th quartile, territory is considerably higher. In the post-2000 era, the Absorption Ratio is in the 4th quartile 38% of the time in our dataset (107 out of 276 months) versus only 13% of the time (43 out of 324 months) in the pre-2000 era. ??

Much has been written on the likely underlying drivers behind this phenomenon. [iv] [v] Notwithstanding some distinctions between developed and emerging markets, these drivers generally fall under two broad categories: market structural factors and behavioral factors.

For the first cluster, we cite the heightened market integration along with the risk-taking capacity of global financial intermediaries and the eruption in market access – including access by retail investors. Along the global expansion of financial intermediaries, regulators guided by the Basel Committee on Banking Supervision, implemented the Basel Capital Accord[i] which introduced regulatory capital charges, initially for credit risk in the late 90’s (Basel I), 8% of risk weighted assets, followed by a more complex approach (Basel II) that introduced charges for market risk and operational risk [xvii], coincidentally in the aftermath of the 1997 Asian Financial Crisis and the 1988 Russia default and collapse of LTCM. ?

At the same time the turn of the millennium was also characterized by a rapid expansion in globalization, marked by China’s entry in the WTO in 2021. Amongst the plethora of potential culprits, recent academic research puts the onus on the interaction of constraints by global financial intermediaries in explaining the rise of global systemic factors hence the structural break in the absorption ratio around the turn of the millennium [xviii],[xix].

On the behavioral cluster, we cite the proliferation of likeminded risk management methods, to an extent underpinned by the regulatory efforts mentioned above, may have contributed to exacerbating stresses on financial markets at least opportune instances. A second contributor has been the asymmetric risk-taking incentive structure manifested by systemic backstops and mega bailouts carried out by monetary authorities of major economies. Cementing it all together, straddling both the pre-2000 and post-2000 timeframe, has been the great moderation of interest rates. Characteristically this period was underpinned by the decline in the yield of the 10-year US Treasury bond from a high of 16% to a mere 0.5% -- that inter-alia, added further pressure on market participants to reach out the risk spectrum for additional yield.

Typical asset rotation and rebalancing approaches.

In our pursuit for optimal asset rotation and rebalancing strategies we swiftly dismiss the active but heuristic approach, observed at one extreme of the spectrum of investment practices. Such approach, fooled by randomness, typically leads to noisy asset rotation, momentum chasing and the churning of transactions for no apparent benefit.[vi] On the other end of the spectrum, we encounter a rigidly passive approach, agnostic on the timing of introducing new assets to a portfolio and having a formulaic rebalancing method, which, while it could be considered an investment philosophy, suffers from numerous caveats. A passive approach is essentially based on the belief that humans cannot fully synthesize current public information, or conversely said all public information is perfectly reflected in market prices – a strong form of the Efficient Markets Hypothesis (EMH). Such a belief may have led many investors for instance, at the end of 2Q 2008, amid the US sub-prime loan and housing crisis just weeks after Bear Stearns collapsed, to rebalance and add risk to their portfolios. Passive rebalancing can also be prone to exploitation by market participants with a good grasp on both the quarterly reconstitution of major indices and long-term investor flows.[vii] ?It is not entirely accidental that there has been a proliferation in so called early-roll strategies.

?

Dynamic Asset Allocation.

In the light of the structural phenomenon in financial markets described above and confronted with the pitfalls of both the active but heuristic and passive strategies, we gravitate in the direction of a macro-based systematic approach. We advocate for a dynamic asset-allocation (DAA) approach that hinges on the persistence of market-based risk-on and risk-off indicators in order to scale exposure to risk.

Our journey for a robust risk-on and risk-off metric has, over recent years, guided us beyond the commonly used Financial Conditions Indicators (FCIs) [viii], swap spreads, TED spreads, technical indicators and heuristics, towards the statistical concept of Mahalanobis Distance. [ix] [x] This concept lends itself very well to finance applications given the multi-variate nature of market returns. We follow the process of classifying Mahalanobis Distances beyond a specific threshold, as financial turbulence in line with earlier academic work in the field.[xi] [xii] [xiii] Mahalanobis Distances allow us to detect atypical behavior in assets based on their historical patterns, which is typically manifested by extreme price movements and a breakdown in the cross-asset correlation structure. Such conditions are associated with market illiquidity and heightened risk aversion. Empirically, episodes of ‘turbulence’ are characterized by significant drawdowns in risk-assets along with rallies in safe-haven assets.

?

Risk regime and empirical returns.

We calibrate our own Mahalanobis Distance metric for a representative opportunity set of liquid assets and compute returns during turbulent versus non-turbulent episodes over a 50-year period, encompassing multiple economic cycles, a plethora of market crashes and monetary policy cycles. Our approach captures significant drawdowns in turbulent episodes for risk-assets such as US and International Equities, Commodities and REITs while perceived ‘safe-haven’ assets such as Gold and US Treasuries perform extremely well in those episodes.

?The empirical analysis, performed at the 90th percentile of metric, captures material drawdowns in risky assets at times of turbulence as compared to times of non-turbulence: US equities -9.9% annualized return during turbulent episodes vs 12.8% during non-turbulent episodes, International Equities -9.1% vs. 10.5%, US REITs -32% vs +9% and so on. Safe-haven assets on the other hand, rally significantly at times of turbulence and do moderately at non-turbulent times (US Treasuries +10.7% vs. 5.5% and Gold +37% vs. +4%).

?

Robustness checks using out-of-sample portfolio simulation. ?

We use simulation to test the robustness of our signals, absent the perfect hindsight of the empirical findings. Our benchmark for the exercise is the typical 60:40 portfolio of US equities and US Treasury bonds, rebalanced quarterly, not allowing for leverage or other frictions. On the other hand, our DAA strategy is characterized also by the same starting point and the same opportunity set, with the crucial distinction that we allow a maximum deviation of 40% from the 60:40 portfolio. This deviation is dynamically driven off the financial turbulence signal. This portfolio reallocation process is smoothed by an algorithm that at one extreme dynamically re-allocates the full 40% permissible deviation to Treasuries when Financial Turbulence is at extreme levels, and at the other side of the spectrum dynamically re-allocates the full permissible deviation to the risk asset i.e. to US equities when turbulence converges to low levels. To maintain the robustness of the back-test (avoid overfitting), we calibrate the model with subset of our dataset and use it to simulate the performance of our strategy in an out-of-sample fashion for the 13-year period from 2010 to date.

We observe (left hand panel) that our dynamic strategy’s monthly returns, depicted in green, exhibit on average smaller drawdowns against those of the Buy and Hold (“BH”) 60:40 strategy, depicted in blue. Inevitably, DAA gives up some of the upside on occasions such as the post-covid melt-up rally, but given the mathematics of compounding the downside protection at episodes of market drawdowns, such as the European debt crisis, more than offset the give-up of some of the upside at non-turbulent episodes. The DAA strategy offers downside protection for instance against the 2018 market correction, that ensued from the intensifying of US-China trade tensions and the inflationary woes post pandemic. For the full out-of-sample period 2011 to end of 2023 the DAA strategy (right hand side panel) accumulates to 3.6x against 2.8x by the BH strategy – a non-negligible differential. The table below provides the results in greater detail.

?

Risk & Performance Characteristics for the Dynamic Asset Allocation Strategy (DAA) vs. Buy and Hold (BH)

Outperformance of the dynamic strategy (DAA) over the buy and hold strategy (BH) is the equivalent of +1.61% per annum (10.21% vs. 8.60% p.a) and it is delivered with considerable reduction in portfolio risk. The DAA for instance exhibits lower annualized portfolio volatility (8.2% vs. 8.9%), lower Conditional Value at Risk (CVaR: 4.4% vs. 5.3%), material reduction in the negative skewness of portfolio returns (-0.07 from -0.27) and more importantly a significant reduction in the fatness of portfolio returns (kurtosis 0.36 vs. 0.66). The drawdown of the DAA is also smaller (-16.5% vs. -18.5%) experienced in the early 2022 correction that ensued when the US Fed rose rates sharply from near zero. The deviations in the DAA from the BH portfolio result for the full period in a Tracking Error (TE)of 3%, implying a risk adjusted excess return ratio aka. Information Ratio of 0.53 computed as the annualized excess return divided by the annualized TE.[xiv] We also find low and statistically insignificant correlation coefficients between Financial Turbulence and the Momentum factor, and against all the Fama-French 5-factors (Market Risk Premium, Value, Size, Conservativeness and Profitability).[xv]

?

Some parting thoughts.

Our findings, utilizing an array of model assumptions and estimation methods, demonstrate that Mahalanobis Distances, capture effectively returns to risk i.e. drawdowns in risk-assets at times of turbulence along with gains in safe-haven assets, and the inverse during non-turbulent times. Turbulence signals offer strong downside protection and perform robustly algorithmically in dynamic asset allocation (DAA) context due to persistence. The framework allows us to construct turbulence resilient portfolios across the liquid asset opportunity set that outperform their passive counterparts. By extension the framework can add value by informing investment decisions on asset rotation and periodic portfolio rebalancing in a discretionary blend [xvi] type investment approach.

Turbulence derived risk-on and risk-off signals capture the market pulse and can be potentially deployed to the full spectrum of risk-asset irrespective of the opportunity set used in the calibration. The efficacy of these signals is manifested by the rise of global systemic component as the primary driver of risk-asset returns both in normal but more so at time of distress. ?

?Last, the framework lends itself well to variety of risk management applications such as: (i) the simulation of new assets in a portfolio; and (ii) the development of robust stress scenarios using ‘high-turbulence’ episodes. These scenarios are unique given the diversity in this approach and complement well the typically high volatility episodes employed by most risk solution vendors and risk managers.

?Significant advances in Artificial Intelligence and Machine-Learning methods underpin our renewed interest in this topic. Risk-on / risk-off signals are produced daily and available via web and API. Please reach out for more details at: tk@intellicore.ai

?

End Notes:

[i] In practice, even in relatively passive mandates, a degree of flexibility around the neutral strategic asset allocation (SAA) is often embedded, typically quantified in terms of tracking error (TE) relative to the SAA. This margin allows for departures from the neutral SAA position, which can significantly impact value—either positively or negatively. The measurement of TE often does not typically encompass structural deviations that are frequently made in relation to the SAA, such as the incorporation of systematic betas (like value, small capitalization, momentum, etc.) versus the policy benchmark that underestimates risks. Additionally, potential drawdowns resulting from deviations during market distress can far exceed those in more stable periods, as the typical assumption of normally distributed asset returns rarely holds in such conditions.

[ii] See “FX co-movements: disentangling the role of global market factors, carry-trades and idiosyncratic components” by José Gonzalo Rangel, 2011. ?

[iii] See “Portfolio Flows, Global Risk Aversion and Asset Prices in Emerging Markets” by Hui Tong and Shang-Jin Wei?2, IMF Working paper #14/156, August 2014.

[iv] ?See “Global production linkages and stock market co-movement” by Raphael Auer, Bruce Muneaki Iwadate, Andreas Schrimpf and Alexander Wagner, BIS Working paper 1003, February 2022.

[v] See “Asset managers in emerging market economies”, Ken Miyajima and Ilhyock Shim, BIS Quarterly Review, September 2014.?

[vi] Active management approaches that consistently erode capital are often characterized by opaque investment processes, deficiencies in transparency and accountability, and are susceptibility to behavioral biases. 'Fooled by randomness', as coined by Nassim Taleb, selective memory, and the extrapolation of recent trends into the future are just a few examples of behavioral biases they fall prey to. It is a travesty how frequently asset owners, trustees, and retirement plan beneficiaries are oblivious to such issues, including the opportunity cost and the cumulative impact of their plans consistently underperforming their strategic asset allocation (SAA) benchmarks. During years with strong absolute returns, investment teams may tout these returns while downplaying any underperformance, even if it's as minimal as 1%. Conversely, in years with negative returns, the same individuals are quick to highlight even marginal outperformance relative to the SAA. Often oversight bodies and asset owners are blind sighted to the degree market exposure embedded in the portfolio through leverage, major implementation deviations and tactics to boost portfolio performance, under the umbrella of “active management” -- that tends to work well until a catastrophic, a black swan event, strikes.

[vii] See Jiaguo Wang, Yaqiong Yao, Adina Yelekenova, ETF Rebalancing, Hedge Fund Trades, and Capital Market”, SSRN 4324054, October 2023.?

[viii] Financial Conditions Indicators (FCI’s) are in general lagged indices whose composition and definition of commonly used FCI’s is amended mid-journey without communication of the revised composition, that renders them unusable. Often, vintage readings are (quietly) revised also without communication to users.

[ix] P.C. Mahalanobis, “On the generalized distance in statistics.”, Proceedings of the National Institute of Sciences (Calcutta), 1936, 2, pp.49–55.

[x] The concept of Mahalanobis distances originates from the work of Professor P.C Mahalanobis (1927) a mathematician from India that developed the early mathematical process to identify skulls that belonged to different casts and tribes. Using multiple characteristics of human skulls, the method allowed to identify resemblance and consequently dissimilarities. By extension dissimilarity in markets, given an opportunity set of N asset classes, is estimated using Mahalanobis distances by observing Euclidean distances in a multi-variate context. ????

[xi] See Sebastian Stoekl and Michael Hanke “Financial Applications of Mahalanobis Distance”, Applied Economics and Finance, Vol. 1, No. 2, Nov 2014.

[xii] This line of research has been pursued by financial stability watchdogs, academics and leading monetary authorities, in assessing systemic risk and developing early warning indicators from a financial fragility standpoint, particularly in the aftermath of the Asian Currency Crisis of the late 1990’s.

[xiii] See Mark Kritzman and Yuanzhen Li, “Skulls, Financial Turbulence, and Risk Management", FAJ v66, 2010.?

[xiv] Of note is that we obtain similar results from several variations to the model shown here with out-of-sample information ratios ranging between 0.4 and 0.6, using alternative model estimation and data refinement methods. ?

?[xv] Rolling regression coefficients against the Fama-French factors:

?[xvi] A blend approach entails the synthesis of a number of quantitative and qualitative dimensions, in order to facilitate active portfolio views. These dimensions typically entail the assessment of asset valuations, review of technical factors, investor flows and positioning, analysis of the macroeconomic and geopolitical environment --and provide a platform from which well-grounded investment decisions could take place. As a minimum, the identification of the current investment regime from the proposed framework can augment such a blend of considerations, provide an alternative view and the basis for well-grounded active investment decisions.

[XVII] The Basel I Capital Accord, introduced in 1988, and Basel II, which began implementation in most major countries by 2006, represented two significant steps in the evolution of global banking regulation, focusing on the management of credit risk and the adequacy of capital banks must hold against their risk exposures. The transition from Basel I to Basel II marked a significant shift towards a more risk-sensitive, comprehensive, and nuanced framework for banking regulation. It aimed to encourage better risk management practices among banks and to align capital requirements more closely with the actual risks banks face, while also enhancing transparency and market discipline. The evolution brought about unintended consequences and the two accords differed in a number of dimensions:? (a) Risk Sensitivity and Complexity: Basel I provided a simpler framework with broad, standardized risk weights for assets. It primarily focused on credit risk and set a uniform capital requirement of 8% of risk-weighted assets, without much differentiation based on the actual riskiness of individual asset classes. Basel II introduced a more risk-sensitive approach, offering banks three different methods for calculating capital requirements for credit risk: the Standardized Approach, the Foundation Internal Ratings-Based (IRB) Approach, and the Advanced IRB Approach. Basel II also incorporated operational risk into the capital framework for the first time and refined the treatment of market risk. (b) Internal Ratings-Based (IRB) Approaches: Basel I did not allow banks to use their internal assessments of risk. Basel II allowed banks to use their internal ratings systems for determining the riskiness of their assets, under the IRB approaches. This enabled banks with advanced risk management systems to tailor their capital requirements more closely to their actual risk profile. (c) Operational Risk: Basel I focused almost exclusively on credit risk and to a lesser extent on market risk, without addressing operational risk. Basel II introduced a formal requirement for banks to hold capital against operational risk, acknowledging the potential financial impact of operational failures, such as systems failures, fraud, or other disruptions. (d) Supervisory Review and Market Discipline: Basel I was mainly focused on setting minimum capital requirements without a structured process for supervisory review or enhancing market discipline.? Basel II established the "Three Pillars" framework, expanding beyond minimum capital requirements (Pillar 1) to include a supervisory review process (Pillar 2) and enhanced market discipline through disclosure requirements (Pillar 3). This framework aimed to provide a more comprehensive approach to risk management and bank supervision. (e) Scope of Application:? Basel I applied a one-size-fits-all approach, which was relatively straightforward but lacked sophistication in addressing the complexities of modern banking activities.? Basel II was designed to be more flexible and applicable to banks with varying complexity and international exposure. Its differentiated approaches allowed more sophisticated banks to use advanced risk management techniques to calculate their capital requirements.

[XVIII] See “A Theory of the Global Financial Cycle”, NBER 29217, September 2021.

[XIX] See “Uncertainty Shocks, Capital Flows, and International Risk Spillovers”, NBER 30026, May 2022.

?

?

?

?

?

?

?

?

?