Regime-based dynamic asset allocation using a diverse set of features
(2023) In Master's Theses in Mathematical Sciences FMSM01 20231Mathematical Statistics
- Abstract
- This paper presents a novel approach to dynamic asset allocation using regime-based modeling and a diverse set of features. The objective of this study is to select the most important features that reflect the state of the market, apply an effective method for detecting market regimes using the selected features, and design an effective trading strategy that leverages the identified sequence of market states. The strategy will take into account realistic constraints, such as transaction costs, to ensure its practical applicability. Finally, the strategy will be evaluated using various performance metrics and by comparing it with relevant benchmarks and traditional, static asset allocation strategies.
Statistical methods and common... (More) - This paper presents a novel approach to dynamic asset allocation using regime-based modeling and a diverse set of features. The objective of this study is to select the most important features that reflect the state of the market, apply an effective method for detecting market regimes using the selected features, and design an effective trading strategy that leverages the identified sequence of market states. The strategy will take into account realistic constraints, such as transaction costs, to ensure its practical applicability. Finally, the strategy will be evaluated using various performance metrics and by comparing it with relevant benchmarks and traditional, static asset allocation strategies.
Statistical methods and common knowledge of the financial markets are used to construct and identify which features are most relevant for describing the state of the market. The detection of market regimes is then accomplished using variations of a hidden Markov and a Gaussian mixture model with the selected features as input. Considering the prevailing market regimes, an asset portfolio is constructed by solving an optimization problem, aiming to maximize returns while taking into account transaction costs and variance of the portfolio.
We show that our method may outperform traditional, static asset allocation strategies by achieving higher risk-adjusted returns than relevant benchmarks, while also providing evidence that our approach is robust to changes in market conditions, effectively reducing the occurrence of large drawdowns. Overall, this paper hopes to make a valuable contribution to the field of dynamic asset allocation by using a broad set of features as the basis for regime-based modeling of the financial markets. It may provide important insights for investors who seek to maximize returns while managing risk in an uncertain financial environment. (Less) - Popular Abstract
- Using a diverse set of financial data to model hidden market regimes as a means to beat the market
By using features reflecting the state of the market and applying an effective method for detecting market regimes, we show that it is possible to outperform the S&P 500 while reducing downside risk.
Investors are constantly looking for ways to beat the market, while also being cautious of not losing money in the process. With many fund managers failing to outperform passive index investing, proponents of the efficient market hypothesis are getting their convictions validated. However, there are numerous examples of investors who have been able to achieve extraordinary returns in a consistent manner over long periods of time. One of the... (More) - Using a diverse set of financial data to model hidden market regimes as a means to beat the market
By using features reflecting the state of the market and applying an effective method for detecting market regimes, we show that it is possible to outperform the S&P 500 while reducing downside risk.
Investors are constantly looking for ways to beat the market, while also being cautious of not losing money in the process. With many fund managers failing to outperform passive index investing, proponents of the efficient market hypothesis are getting their convictions validated. However, there are numerous examples of investors who have been able to achieve extraordinary returns in a consistent manner over long periods of time. One of the most distinguished examples of such is perhaps Renaissance Technologies, a quantitative investment management firm that employs advanced mathematical and statistical methods, which from 1988 to 2018 accomplished an astonishing annualized return of 66 %.
While such a track record is hard to match, we propose a method for dynamic asset allocation using regime-based modeling and a diverse set of features that is shown to outperform passive investment strategies at a lower risk, net of fees.
As a first step, statistical methods and common knowledge of the financial markets are used to construct and identify which features are most relevant for describing the state of the market, which include features related to e.g., equities, bonds, commodities, currencies, sentiment indicators and economic variables. Next, we assume there are underlying unobservable (‘hidden’) market regimes that cannot be observed directly, but rather indirectly by observing the features selected in the previous step. The unobservable market regimes are then detected by using variations of a so-called hidden Markov model, with the selected features as input. Depending on the prevailing market regimes, a portfolio of assets is constructed with the aim of maximizing returns while considering transaction costs and the variance of the portfolio (a measure of risk).
In its simplest implementation, one may assume the state of the market may be represented by either of two market regimes, which may be labeled as “bull” and “bear”, characterized by an upward- and downward trend in asset prices respectively. It is possible to extend this assumption to an arbitrary number of market regimes which, however, comes at the cost of a loss of interpretability of each regime.
We found that assuming a four-state regime-model and allocating in a dynamic portfolio consisting of ETFs tracking the S&P 500, gold and bonds, our method achieved an annualized return of almost 9 % (assuming transaction costs of 0.02%) from the start of 2008 until the beginning of 2023, compared to just under 7.5 % for a passive ‘buy-and-hold’ strategy in the S&P 500. Moreover, using our method, we managed to achieve a remarkable reduction in downside risk in comparison to the passive ‘buy-and-hold’ strategy, demonstrated by a maximum drawdown (largest percentage drop of the portfolio from a peak to a trough) of just half the size. The low market-beta (a measure of volatility relative to the overall stock market) of roughly 0.25 also displays the attractiveness of the method for investors who seek means for diversification and risk-hedging, both desirable traits in unstable market conditions. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9126109
- author
- Hellberg, Markus LU and Rosén, Jacob LU
- supervisor
- organization
- alternative title
- Regimbaserad dynamisk tillgångsallokering utgående från ett brett spektrum av finansiella data
- course
- FMSM01 20231
- year
- 2023
- type
- H2 - Master's Degree (Two Years)
- subject
- publication/series
- Master's Theses in Mathematical Sciences
- report number
- LUTFMS-3484-2023
- ISSN
- 1404-6342
- other publication id
- 2023:E57
- language
- English
- id
- 9126109
- date added to LUP
- 2023-06-21 11:13:51
- date last changed
- 2023-07-03 14:00:00
@misc{9126109, abstract = {{This paper presents a novel approach to dynamic asset allocation using regime-based modeling and a diverse set of features. The objective of this study is to select the most important features that reflect the state of the market, apply an effective method for detecting market regimes using the selected features, and design an effective trading strategy that leverages the identified sequence of market states. The strategy will take into account realistic constraints, such as transaction costs, to ensure its practical applicability. Finally, the strategy will be evaluated using various performance metrics and by comparing it with relevant benchmarks and traditional, static asset allocation strategies. Statistical methods and common knowledge of the financial markets are used to construct and identify which features are most relevant for describing the state of the market. The detection of market regimes is then accomplished using variations of a hidden Markov and a Gaussian mixture model with the selected features as input. Considering the prevailing market regimes, an asset portfolio is constructed by solving an optimization problem, aiming to maximize returns while taking into account transaction costs and variance of the portfolio. We show that our method may outperform traditional, static asset allocation strategies by achieving higher risk-adjusted returns than relevant benchmarks, while also providing evidence that our approach is robust to changes in market conditions, effectively reducing the occurrence of large drawdowns. Overall, this paper hopes to make a valuable contribution to the field of dynamic asset allocation by using a broad set of features as the basis for regime-based modeling of the financial markets. It may provide important insights for investors who seek to maximize returns while managing risk in an uncertain financial environment.}}, author = {{Hellberg, Markus and Rosén, Jacob}}, issn = {{1404-6342}}, language = {{eng}}, note = {{Student Paper}}, series = {{Master's Theses in Mathematical Sciences}}, title = {{Regime-based dynamic asset allocation using a diverse set of features}}, year = {{2023}}, }