Backtesting Analysis Research Articles

Quantitative Strategies Based on FinGPTs Analysis of the Federal Reserves Attitude

This paper explores the impact of the Federal Reserve (Fed) on financial markets, specifically through the sentiment scoring of Fed meeting minutes by the AI analysis model FinGPT and its correlation with market performance. The article first introduces the method of sentiment analysis using FinGPT and conducts an in-depth analysis of the meeting minutes from 1990 to 2020 using this method. Furthermore, key market data, such as the S&P 500 Index, real estate, and financial sector indices, are used to assess the actual impact of the Feds attitude on market dynamics. Through quantitative trading and backtesting analysis, this study verifies the performance and effectiveness of a single-factor strategy in different markets. The results show that while the Feds sentiment scores have high predictive validity and market adaptability in the financial sector, their performance in the real estate market and the broader stock market index is relatively weaker. The findings indicate that the market applicability of the strategy is significantly influenced by market type and economic environment. Future strategy optimization needs to consider these factors to enhance adaptability and efficiency across different markets.

Stock Selection Analysis of Innovative Multifactor Model in New Energy Sector

As of 2024, China’s new energy sector has made significant progress, ranking first in the world in terms of installed photovoltaic (PV) and wind power capacity. The proportion of new energy in the energy mix has increased significantly, driving China’s energy transition and sustainable development. Meanwhile, with the rapid development of Internet technology, quantitative investment strategies based on artificial intelligence technology are expected to show great growth potential in the future financial sector. In this context, this paper will construct a multi-factor model with the help of machine learning, combine value analysis and technical analysis, select candidate factors, develop exclusive factors for the characteristics of the new energy industry, test the validity of the factors, construct a multi-factor stock picking strategy and carry out backtesting analysis. The results demonstrate that the strategy can effectively manage risks and achieve excess returns, which is theoretically and practically instructive for the construction of investment portfolios. The results provide an effective investment strategy reference for financial institutions such as banks, brokerage firms and fund companies.

Addressing the financial impact of natural disasters in the era of climate change

The objective of our study is to predict the financial losses that may result from natural disasters, along with their level of volatility, over a period of 1 to 15 years. Volatility can lead to significant fluctuations in Profit and Loss (P&L) for companies that are affected by unexpected events. To achieve this goal, we created a novel two-factor square-root model that allows us to establish a correlation between the frequency of occurrences and volatility, using correlated Brownian motions. Moreover, we utilized a Generalized Pareto Distribution (GPD) to estimate the maximum potential loss in terms of Value at Risk (VaR) for each specific type of natural disaster. To ensure the reliability of our predictions, we compared our results to those of four reference models and conducted a backtesting analysis. This approach is particularly suitable for insurance companies seeking to maintain stable reserves, but it can also be adapted for any other type of business that is vulnerable to extreme events and aims to safeguard a consistent cash flow for their stakeholders.

Interlinkages across US sectoral returns: time-varying interconnectedness and hedging effectiveness

This study examines the time-varying asymmetric interlinkages between nine US sectoral returns from January 2020 to January 2023. To this end, we used the time-varying parameter vector autoregression (TVP-VAR) asymmetric connectedness approach of Adekoya et al. (Resour Policy 77:102728, 2022a, Resour Policy 78:102877, 2022b) and analyzed the time-varying transmitting/receiving roles of sectors, considering the positive and negative impacts of the spillovers. We further estimate negative spillovers networks at two burst times (the declaration of the COVID-19 pandemic by the World Health Organization on 11 March 2020 and the start of Russian-Ukrainian war on 24 February 2022, respectively). Moreover, we performed a portfolio back-testing analysis to determine the time-varying portfolio allocations and hedging the effectiveness of different portfolio construction techniques. Our results reveal that (i) the sectoral return series are strongly interconnected, and negative spillovers dominate the study period; (ii) US sectoral returns are more sensitive to negative shocks, particularly during the burst times; (iii) the overall, positive, and negative connectedness indices reached their maximums on March 16, 2020; (iv) the industry sector is the largest transmitter/recipient of return shocks on average; and (v) the minimum correlation and connectedness portfolio approaches robustly capture asymmetries. Our findings provide suggestions for investors, portfolio managers, and policymakers regarding optimal portfolio strategies and risk supervision.

Using Crypto-Asset Pricing Methods to Build Technical Oscillators for Short-Term Bitcoin Trading

This paper examines the trading performances of several technical oscillators created using crypto-asset pricing methods for short-term bitcoin trading. Seven pricing models proposed in the professional and academic literature were transformed into oscillators, and two thresholds were introduced to create buy and sell signals. The empirical back-testing analysis showed that some of these methods proved to be profitable with good Sharpe ratios and limited max drawdowns. However, the trading performances of almost all methods significantly worsened after 2017, thus indirectly confirming an increasing financial literature that showed that the introduction of bitcoin futures in 2017 improved the efficiency of bitcoin markets.

Estimating Value-at-Risk and Expected Shortfall: Do Polynomial Expansions Outperform Parametric Densities?

We assess Value-at-Risk (VaR) and Expected Shortfall (ES) estimates assuming different models for the standardized returns: distributions based on polynomial expansions such as Cornish-Fisher and Gram-Charlier, and well-known parametric densities such as normal, skewed-t and Johnson. This paper aims to analyze whether models based on polynomial expansions outperform the parametric ones. We carry out the model performance comparison in two stages: first, with a backtesting analysis of VaR and ES; and second, using loss functions. Our backtesting results show that all distributions, except for normal ones, perform quite well in VaR and ES estimations. Regarding the loss function analysis, we conclude that polynomial expansions (specifically, the Cornish-Fisher one) usually outperform parametric densities in VaR estimation, but the latter (specifically, the Johnson density) slightly outperform the former in ES estimation; however, the gains of using one approach or the other are modest.

Hedging Option Books Using Neural-SDE Market Models

We study the capability of arbitrage-free neural-SDE market models to yield effective strategies for hedging options. In particular, we derive sensitivity-based and minimum-variance-based hedging strategies using these models and examine their performance when applied to various option portfolios using real-world data. Through backtesting analysis over typical and stressed market periods, we show that neural-SDE market models achieve lower hedging errors than Black–Scholes delta and delta-vega hedging consistently over time, and are less sensitive to the tenor choice of hedging instruments. In addition, hedging using market models leads to similar performance to hedging using Heston models, while the former tends to be more robust during stressed market periods.

Estimating risks of option books using neural-SDE market models

In this paper, we examine the capacity of an arbitrage-free neural-SDE market model to produce realistic scenarios for the joint dynamics of multiple European options on a single underlying. We subsequently demonstrate its use as a risk simulation engine for option portfolios. Through backtesting analysis, we show that our models are more computationally efficient and accurate for evaluating the Value-at-Risk (VaR) of option portfolios, with better coverage performance and less procyclicality than standard filtered historical simulation approaches.

Hedging Option Books Using Neural-SDE Market Models

We study the capability of arbitrage-free neural-SDE market models to yield effective strategies for hedging options. In particular, we derive sensitivity-based and minimum-variance-based hedging strategies using these models and examine their performance when applied to various option portfolios using real-world data. Through backtesting analysis over typical and stressed market periods, we show that neural-SDE market models achieve lower hedging errors than Black--Scholes delta and delta-vega hedging consistently over time, and are less sensitive to the tenor choice of hedging instruments. In addition, hedging using market models leads to similar performance to hedging using Heston models, while the former tends to be more robust during stressed market periods.

Polynomial adjusted Student-t densities for modeling asset returns

We present a polynomial expansion of the standardized Student-t distribution. Our density, obtained through the polynomial adjusted method in Bagnato, Potí, and Zoia (2015. “The Role of Orthogonal Polynomials in Adjusting Hyperbolic Secant and Logistic Distributions to Analyse Financial Asset Returns.” Statistical Papers 56 (4): 1205–12340), is an extension of the Gram–Charlier density in Jondeau and Rockinger (2001. “Gram-Charlier Densities.” Journal of Economic Dynamics and Control 25 (10): 1457–1483). We derive the closed-form expressions of the moments, the distribution function and the skewness–kurtosis frontier for a well-defined density. An empirical application is also implemented for modeling heavy-tailed and skewed distributions for daily asset returns. Both in-sample and backtesting analysis show that this new density can be a good candidate for risk management.

Tax-Rate Arbitrage: Realization of Long-Term Gains to Enable Short-Term Loss Harvesting

We look at an enhanced loss-harvesting strategy, tax-rate arbitrage, which exploits the differential between short- and long-term tax rates. Our study relies on ATBAT, an After-Tax Back-Testing Analysis Tool that lets us examine tax-managed strategies over numerous historical periods. For the ideal tax-rate arbitrage investor, one who is subject to federal only tax rates, who has a long horizon and a planned liquidation date, and who launches the strategy from all cash, tax-rate arbitrage generated an average of 0.78% in excess after-tax active return at a 10-year horizon relative to a standard loss-harvesting strategy. Other investors with different profiles may benefit from tax-rate arbitrage but typically to a lesser extent.

Risk Analysis of China Stock Market During Economic Downturns–Based on GARCH-VaR and Wavelet Transformation Approaches

Accurately measuring the risk of stock market is important for improving the risk management level of stock market and reducing losses for the investors, especially during economic downturns. In this paper, to study the stock market risk in economic downturns, we first adopted wavelet transformation analysis to detect and define those phases. This paper selected the Shanghai Composite Index and eleven representative sector indexes from January 1, 2008 to January 31, 2020, established GARCH models based on student-t and GED distributions, and found the volatility aggregation and the leverage effect of the stock market. This paper measured the risk of stock market by the parameter method, using conditional volatility estimated by the GARCH models, and carried out back-testing analysis and the Kupiec failure rate test on measurement accuracy of VaR. To study the stock market risk during economic downturns, this paper calculated and compared the risks of different sectors during the corresponding phases, and the results showed that the risks in real estate, industrial, telecom service sectors are much higher during financial downturns. Finally, this paper summarized the conclusion and put forward suggestions for investors from the perspective of risk management.

Forecasting robust value-at-risk estimates: evidence from UK banks

In this paper, we present a novel approach for forecasting Value-at-Risk (VaR) by combining a Bayesian GARCH(1,1) model with Student's-t distribution for the underlying volatility models, vine copula functions to model dependence, and the peaks-over-threshold (POT) method of extreme value theory (EVT) to model the tail behaviour of asset returns. We further propose a new approach for threshold selection in extreme value analysis, which we call a hybrid method. The empirical results and back-testing analysis show that the model captures VaR quite well through periods of calmness and crisis; therefore, it is suitable for use as a measure of risk. Our results also suggest that with a correct implementation of the VaR model, Basel III is not needed.

Adaboost-SVM Multi-Factor Stock Selection Model Based on Adaboost Enhancement

In recent years, the applications of machine learning techniques to perfect traditional financial investment models has gained a widespread attention from the academic circle and the financial industry. This paper takes CSI300 stocks as the object of the research, uses Adaboost to enhance the classification ability of original linear support vector machine, and combines all major factors to build Adaboost-SVM multi-factor stock selection model based on Adaboost enhancement. In the backtesting analysis, the stock selection strategy of original linear support vector machine was compared with the Adaboost-SVM multi-factor stock selection strategy based on Adaboost enhancement. The result shows that the Adaboost-SVM multi-factor stock selection strategy based on Adaboost enhancement possesses stronger profitability and smaller income fluctuation than the original algorithm model.

Research on Risk Measurement in Financial Market Based on GARCH-VaR and FHS——An Example of Chinese Bond Market

Accurately measuring the risk of bond market is very important for improving the risk management level of bond market and maintaining the stability of the financial system. Taking ChinaBond New Composite Wealth (gross) Index as the research object, this paper selects the closing price from January 1, 2002 to March 30, 2018, establishes the GARCH, EGARCH and GJR-GARCH model based on normal distribution and t distribution, and finds out the volatility aggregation and the leverage effect of the bond market. Then, this paper use two methods to measure the risk of the bond market: first, we estimate the value at risk (VaR) of the bond market by the parameter method, using conditional variance estimated by the GARCH models, and we carry out backtesting analysis and the Kupiec failure rate test on measurement accuracy of VaR. The results show that t distribution hypothesis and elimination of autocorrelation of the yield rate can improve the accuracy and robustness of the estimation of the VaR; second, we simulate the future revenue path of the bond market and compare it with the actual loss, using Filtered Historical Simulation (FHS) based on Bootstrap method. The results show that the bond market has leverage effect. The maximum possible loss under extreme conditions can be far greater than the maximum possible revenue. But the estimated VaR under 95% confidence level can predict future risks very well. Finally, according to the conclusion, this paper puts forward some suggestions for regulators and investors from the perspective of risk management.

Measuring long-term tail risk: Evaluating the performance of the square-root-of-time rule

Abstract This paper focuses on risk over long time horizons and within extreme percentiles, which have attracted considerable recent interest in numerous subfields of finance. Value at risk (VaR) aggregates several components of asset risk into a single quantitative measurement and is commonly used in tail risk management. Due to realistic data limits, many practitioners might use the square-root-of-time rule (SRTR) to compute long-term VaR. However, serial dependence and heavy-tailedness can bias the SRTR. This paper addresses two deficiencies of the study by Wang et al. (2011), who propose the modified-SRTR (MSRTR) to partially correct the serial dependence and use subsampling estimation as the benchmark to verify the performance of MSRTR. First, we investigate the validity of the subsampling approach through numerical simulations. Second, to reduce the heavy-tailedness bias, we propose a new MSRTR approach (MSRTR ∗ ) in light of the Central Limit Theorem (CLT). In the empirical study, 28 country-level exchange-traded funds (ETFs) from 2010 to 2015 are considered to estimate the 30-day VaR. After modifying both serial dependence and heavy-tailedness, our approach reduces the bias from 26.46% to 5.97%, on average, compared to the SRTR. We also provide a backtesting analysis to verify the robustness of the MSRTR ∗ . This new approach should be considered when estimating long-term VaR using short-term VaR.

Measuring Long-Term Tail Risk: Evaluating the Performance of the Square-Root-of-Time Rule

This paper focuses on risk over long time horizons and within extreme percentiles, which have attracted considerable recent interest in numerous subfields of finance. Value at risk (VaR) aggregates several components of asset risk into a single quantitative measurement and is commonly used in tail risk management. Due to realistic data limits, many practitioners might use the square-root-of-time rule (SRTR) to compute long-term VaR. However, serial dependence and heavy-tailedness can bias the SRTR. This paper addresses two deficiencies of the study by Wang et al. (2011), who propose the modified-SRTR (MSRTR) to partially correct the serial dependence and use subsampling estimation as the benchmark to verify the performance of MSRTR. First, we investigate the validity of the subsampling approach through numerical simulations. Second, to reduce the heavy-tailedness bias, we propose a new MSRTR approach (MSRTR) in light of the Central Limit Theorem (CLT). In the empirical study, 28 country-level exchange-traded funds (ETFs) from 2010 to 2015 are considered to estimate the 30-day VaR. After modifying both serial dependence and heavy-tailedness, our approach reduces the bias from 26.46% to 5.97%, on average, compared to the SRTR. We also provide a backtesting analysis to verify the robustness of the MSRTR. This new approach should be considered when estimating long-term VaR using short-term VaR.

Tail-Related Risk Measurement and Forecasting in Equity Markets

Parametric, simulation-based and hybrid methods are utilized to estimate various risk measures such as Value-at-Risk (VaR), Conditional VaR and coherent Expected Shortfall. An exhaustive backtesting analysis is performed for London’s FTSE 100 index and a comparative evaluation of the predictability of the investigated models is performed with the use of various statistical tests. We show that optimal tail risk forecasting necessitates that many factors be considered such as asset structure and capitalization and specific market conditions i.e., normal or crisis periods. Specifically, for large capitalization stocks and long investment horizons parametric modeling accounted for relatively better risk estimation in normal quantiles, whilst for short-term trading strategies, the non-parametric methods are more suitable for measuring extreme tail risk of small-cap stocks.

A Monte-Carlo-based Latent Factor Modeling Approach with Time-Varying Volatility for Value-at-Risk Estimation

The normal probability distribution assumption, to model price changes in Finance, belongs to the largest imperfections in the Value-at-Risk (VaR) estimation. In fact, the financial returns are rather distributed leptokurtic than normally and the empirical distributions are often skewed. In these cases, the normal distribution assumption results in over or underestimation of VaR especially when the quantiles are very high/low. Therefore, it is necessary to put emphasis on respecting the leptokurtic and skewed return distribution. In this paper, we propose a new approach for portfolio VaR estimation, which combines the standard latent factor model with the generalized quadratic autoregressive conditionally heteroskedastic model (GQARCH). This new “hybrid” specification provides an alternative, compact, model to handle co-movements, heteroskedasticity and intra-frame correlations in financial data. For maximum likelihood estimation we have used an iterative approach based on an extended version of the Kalman filter algorithm combined with the Expectation Maximization (EM) algorithm. Using a set of historical data, from the Tunisian foreign exchange market, the model parameters are estimated. Then, the fitted model combined with a modified Monte-Carlo simulation algorithm was used to predict the VaR of the Tunisian public debt portfolio. Through a Backtesting analysis, we found that this new specification produces far more accurate forecasts for the VaR compared to the mixture of factor analyzers and other competing approaches.

Backtesting the Lee–Carter and the Cairns–Blake–Dowd Stochastic Mortality Models on Italian Death Rates

This work proposes a backtesting analysis that compares the Lee–Carter and the Cairns–Blake–Dowd mortality models, employing Italian data. The mortality data come from the Italian National Statistics Institute (ISTAT) database and span the period 1975–2014, over which we computed back-projections evaluating the performances of the models compared with real data. We propose three different backtest approaches, evaluating the goodness of short-run forecast versus medium-length ones. We find that neither model was able to capture the improving shock on mortality observed for the male population on the analysed period. Moreover, the results suggest that CBD forecasts are reliable prevalently for ages above 75, and that LC forecasts are basically more accurate for this data.