The well-known Solow growth model is the workhorse model of the theory of economic growth, which studies capital accumulation in a model economy as a function of time with capital stock, labour and technology efiiciency as the basic ingredients. The capital is assumed to be in the form of manufacturing equipments and materials. Two important parameters of the model are: the saving fraction $s$ of the output of a production function and the technology efficiency parameter $A$, appearing in the production function. The saved fraction of the output is fully invested in the generation of new capital and the rest is consumed. The capital stock also depreciates as a function of time due to the wearing out of old capital and the increase in the size of the labour population. We propose a stochastic Solow growth model assuming the saving fraction to be a sigmoidal function of the per capita capital $k_p$. We derive analytically the steady state probability distribution $P(k_p)$ and demonstrate the existence of a poverty trap, of central concern in development economics. In a parameter regime, $P(k_p)$ is bimodal with the twin peaks corresponding to states of poverty and well-being respectively. The associated potential landscape has two valleys with fluctuation-driven transitions between them. The mean exit times from the valleys are computed and one finds that the escape from a poverty trap is more favourable at higher values of $A$. We identify a critical value of $A_c$ below (above) which the state of poverty (well-being) dominates and propose two early signatures of the regime shift occurring at $A_c$. The economic model, with conceptual foundation in nonlinear dynamics and statistical mechanics, share universal features with dynamical models from diverse disciplines like ecology and cell biology.
{"title":"Growth, Poverty Trap and Escape","authors":"Indrani Bose","doi":"arxiv-2310.09098","DOIUrl":"https://doi.org/arxiv-2310.09098","url":null,"abstract":"The well-known Solow growth model is the workhorse model of the theory of\u0000economic growth, which studies capital accumulation in a model economy as a\u0000function of time with capital stock, labour and technology efiiciency as the\u0000basic ingredients. The capital is assumed to be in the form of manufacturing\u0000equipments and materials. Two important parameters of the model are: the saving\u0000fraction $s$ of the output of a production function and the technology\u0000efficiency parameter $A$, appearing in the production function. The saved\u0000fraction of the output is fully invested in the generation of new capital and\u0000the rest is consumed. The capital stock also depreciates as a function of time\u0000due to the wearing out of old capital and the increase in the size of the\u0000labour population. We propose a stochastic Solow growth model assuming the\u0000saving fraction to be a sigmoidal function of the per capita capital $k_p$. We\u0000derive analytically the steady state probability distribution $P(k_p)$ and\u0000demonstrate the existence of a poverty trap, of central concern in development\u0000economics. In a parameter regime, $P(k_p)$ is bimodal with the twin peaks\u0000corresponding to states of poverty and well-being respectively. The associated\u0000potential landscape has two valleys with fluctuation-driven transitions between\u0000them. The mean exit times from the valleys are computed and one finds that the\u0000escape from a poverty trap is more favourable at higher values of $A$. We\u0000identify a critical value of $A_c$ below (above) which the state of poverty\u0000(well-being) dominates and propose two early signatures of the regime shift\u0000occurring at $A_c$. The economic model, with conceptual foundation in nonlinear\u0000dynamics and statistical mechanics, share universal features with dynamical\u0000models from diverse disciplines like ecology and cell biology.","PeriodicalId":501372,"journal":{"name":"arXiv - QuantFin - General Finance","volume":"65 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138523136","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Large Language Models (LLMs) have demonstrated remarkable performance on a wide range of Natural Language Processing (NLP) tasks, often matching or even beating state-of-the-art task-specific models. This study aims at assessing the financial reasoning capabilities of LLMs. We leverage mock exam questions of the Chartered Financial Analyst (CFA) Program to conduct a comprehensive evaluation of ChatGPT and GPT-4 in financial analysis, considering Zero-Shot (ZS), Chain-of-Thought (CoT), and Few-Shot (FS) scenarios. We present an in-depth analysis of the models' performance and limitations, and estimate whether they would have a chance at passing the CFA exams. Finally, we outline insights into potential strategies and improvements to enhance the applicability of LLMs in finance. In this perspective, we hope this work paves the way for future studies to continue enhancing LLMs for financial reasoning through rigorous evaluation.
{"title":"Can GPT models be Financial Analysts? An Evaluation of ChatGPT and GPT-4 on mock CFA Exams","authors":"Ethan Callanan, Amarachi Mbakwe, Antony Papadimitriou, Yulong Pei, Mathieu Sibue, Xiaodan Zhu, Zhiqiang Ma, Xiaomo Liu, Sameena Shah","doi":"arxiv-2310.08678","DOIUrl":"https://doi.org/arxiv-2310.08678","url":null,"abstract":"Large Language Models (LLMs) have demonstrated remarkable performance on a\u0000wide range of Natural Language Processing (NLP) tasks, often matching or even\u0000beating state-of-the-art task-specific models. This study aims at assessing the\u0000financial reasoning capabilities of LLMs. We leverage mock exam questions of\u0000the Chartered Financial Analyst (CFA) Program to conduct a comprehensive\u0000evaluation of ChatGPT and GPT-4 in financial analysis, considering Zero-Shot\u0000(ZS), Chain-of-Thought (CoT), and Few-Shot (FS) scenarios. We present an\u0000in-depth analysis of the models' performance and limitations, and estimate\u0000whether they would have a chance at passing the CFA exams. Finally, we outline\u0000insights into potential strategies and improvements to enhance the\u0000applicability of LLMs in finance. In this perspective, we hope this work paves\u0000the way for future studies to continue enhancing LLMs for financial reasoning\u0000through rigorous evaluation.","PeriodicalId":501372,"journal":{"name":"arXiv - QuantFin - General Finance","volume":"202 ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138522836","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
At the peak of the tech bubble, only 0.57% of market valuation comes from dividends in the next year. Taking the ratio of total market value to the value of one-year dividends, we obtain a valuation-based duration of 175 years. In contrast, at the height of the global financial crisis, more than 2.2% of market value is from dividends in the next year, implying a duration of 46 years. What drives valuation duration? We find that market participants have limited information about cash flow beyond one year. Therefore, an increase in valuation duration is due to a decrease in the discount rate rather than good news about long-term growth. Accordingly, valuation duration negatively predicts annual market return with an out-of-sample R2 of 15%, robustly outperforming other predictors in the literature. While the price-dividend ratio reflects the overall valuation level, our valuation-based measure of duration captures the slope of the valuation term structure. We show that valuation duration, as a discount rate proxy, is a critical state variable that augments the price-dividend ratio in spanning the (latent) state space for stock-market dynamics.
{"title":"Valuation Duration of the Stock Market","authors":"Ye Li, Chen Wang","doi":"arxiv-2310.07110","DOIUrl":"https://doi.org/arxiv-2310.07110","url":null,"abstract":"At the peak of the tech bubble, only 0.57% of market valuation comes from\u0000dividends in the next year. Taking the ratio of total market value to the value\u0000of one-year dividends, we obtain a valuation-based duration of 175 years. In\u0000contrast, at the height of the global financial crisis, more than 2.2% of\u0000market value is from dividends in the next year, implying a duration of 46\u0000years. What drives valuation duration? We find that market participants have\u0000limited information about cash flow beyond one year. Therefore, an increase in\u0000valuation duration is due to a decrease in the discount rate rather than good\u0000news about long-term growth. Accordingly, valuation duration negatively\u0000predicts annual market return with an out-of-sample R2 of 15%, robustly\u0000outperforming other predictors in the literature. While the price-dividend\u0000ratio reflects the overall valuation level, our valuation-based measure of\u0000duration captures the slope of the valuation term structure. We show that\u0000valuation duration, as a discount rate proxy, is a critical state variable that\u0000augments the price-dividend ratio in spanning the (latent) state space for\u0000stock-market dynamics.","PeriodicalId":501372,"journal":{"name":"arXiv - QuantFin - General Finance","volume":"196 ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138522838","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
It has been long that literature in financial academics focuses mainly on price and return but much less on trading volume. In the past twenty years, it has already linked both price and trading volume to economic fundamentals, and explored the behavioral implications of trading volume such as investor's attitude toward risks, overconfidence, disagreement, and attention etc. However, what is surprising is how little we really know about trading volume. Here we show that trading volume probability represents the frequency of market crowd's trading action in terms of behavior analysis, and test two adaptive hypotheses relevant to the volume uncertainty associated with price in China stock market. The empirical work reveals that market crowd trade a stock in efficient adaptation except for simple heuristics, gradually tend to achieve agreement on an outcome or an asset price widely on a trading day, and generate such a stationary equilibrium price very often in interaction and competition among themselves no matter whether it is highly overestimated or underestimated. This suggests that asset prices include not only a fundamental value but also private information, speculative, sentiment, attention, gamble, and entertainment values etc. Moreover, market crowd adapt to gain and loss by trading volume increase or decrease significantly in interaction with environment in any two consecutive trading days. Our results demonstrate how interaction between information and news, the trading action, and return outcomes in the three-term feedback loop produces excessive trading volume which includes various internal and external causes.
{"title":"Market Crowds' Trading Behaviors, Agreement Prices, and the Implications of Trading Volume","authors":"Leilei Shi, Bing Han, Yingzi Zhu, Liyan Han, Yiwen Wang, Yan Piao","doi":"arxiv-2310.05322","DOIUrl":"https://doi.org/arxiv-2310.05322","url":null,"abstract":"It has been long that literature in financial academics focuses mainly on\u0000price and return but much less on trading volume. In the past twenty years, it\u0000has already linked both price and trading volume to economic fundamentals, and\u0000explored the behavioral implications of trading volume such as investor's\u0000attitude toward risks, overconfidence, disagreement, and attention etc.\u0000However, what is surprising is how little we really know about trading volume.\u0000Here we show that trading volume probability represents the frequency of market\u0000crowd's trading action in terms of behavior analysis, and test two adaptive\u0000hypotheses relevant to the volume uncertainty associated with price in China\u0000stock market. The empirical work reveals that market crowd trade a stock in\u0000efficient adaptation except for simple heuristics, gradually tend to achieve\u0000agreement on an outcome or an asset price widely on a trading day, and generate\u0000such a stationary equilibrium price very often in interaction and competition\u0000among themselves no matter whether it is highly overestimated or\u0000underestimated. This suggests that asset prices include not only a fundamental\u0000value but also private information, speculative, sentiment, attention, gamble,\u0000and entertainment values etc. Moreover, market crowd adapt to gain and loss by\u0000trading volume increase or decrease significantly in interaction with\u0000environment in any two consecutive trading days. Our results demonstrate how\u0000interaction between information and news, the trading action, and return\u0000outcomes in the three-term feedback loop produces excessive trading volume\u0000which includes various internal and external causes.","PeriodicalId":501372,"journal":{"name":"arXiv - QuantFin - General Finance","volume":"33 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138522784","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jakub Michańków, Łukasz Kwiatkowski, Janusz Morajda
In this paper, we develop a hybrid approach to forecasting the volatility and risk of financial instruments by combining common econometric GARCH time series models with deep learning neural networks. For the latter, we employ Gated Recurrent Unit (GRU) networks, whereas four different specifications are used as the GARCH component: standard GARCH, EGARCH, GJR-GARCH and APARCH. Models are tested using daily logarithmic returns on the S&P 500 index as well as gold price Bitcoin prices, with the three assets representing quite distinct volatility dynamics. As the main volatility estimator, also underlying the target function of our hybrid models, we use the price-range-based Garman-Klass estimator, modified to incorporate the opening and closing prices. Volatility forecasts resulting from the hybrid models are employed to evaluate the assets' risk using the Value-at-Risk (VaR) and Expected Shortfall (ES) at two different tolerance levels of 5% and 1%. Gains from combining the GARCH and GRU approaches are discussed in the contexts of both the volatility and risk forecasts. In general, it can be concluded that the hybrid solutions produce more accurate point volatility forecasts, although it does not necessarily translate into superior VaR and ES forecasts.
{"title":"Combining Deep Learning and GARCH Models for Financial Volatility and Risk Forecasting","authors":"Jakub Michańków, Łukasz Kwiatkowski, Janusz Morajda","doi":"arxiv-2310.01063","DOIUrl":"https://doi.org/arxiv-2310.01063","url":null,"abstract":"In this paper, we develop a hybrid approach to forecasting the volatility and\u0000risk of financial instruments by combining common econometric GARCH time series\u0000models with deep learning neural networks. For the latter, we employ Gated\u0000Recurrent Unit (GRU) networks, whereas four different specifications are used\u0000as the GARCH component: standard GARCH, EGARCH, GJR-GARCH and APARCH. Models\u0000are tested using daily logarithmic returns on the S&P 500 index as well as gold\u0000price Bitcoin prices, with the three assets representing quite distinct\u0000volatility dynamics. As the main volatility estimator, also underlying the\u0000target function of our hybrid models, we use the price-range-based Garman-Klass\u0000estimator, modified to incorporate the opening and closing prices. Volatility\u0000forecasts resulting from the hybrid models are employed to evaluate the assets'\u0000risk using the Value-at-Risk (VaR) and Expected Shortfall (ES) at two different\u0000tolerance levels of 5% and 1%. Gains from combining the GARCH and GRU\u0000approaches are discussed in the contexts of both the volatility and risk\u0000forecasts. In general, it can be concluded that the hybrid solutions produce\u0000more accurate point volatility forecasts, although it does not necessarily\u0000translate into superior VaR and ES forecasts.","PeriodicalId":501372,"journal":{"name":"arXiv - QuantFin - General Finance","volume":"308 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-10-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138522791","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Rédempteur Ntawiratsa, David Niyukuri, Irène Irakoze, Menus Nkurunziza
The term structure of interest rates (yield curve) is a critical facet of financial analytics, impacting various investment and risk management decisions. It is used by the central bank to conduct and monitor its monetary policy. That instrument reflects the anticipation of inflation and the risk by investors. The rates reported on yield curve are the cornerstone of valuation of all assets. To provide such tool for Burundi financial market, we collected the auction reports of treasury securities from the website of the Central Bank of Burundi. Then, we computed the zero-coupon rates, and estimated actuarial rates of return by applying the Nelson-Siegel and Svensson models. This paper conducts a rigorous comparative analysis of these two prominent parametric yield curve models and finds that the Nelson-Siegel model is the optimal choice for modeling the Burundian yield curve. The findings contribute to the body of knowledge on yield curve modeling, enhancing its precision and applicability in financial markets. Furthermore, this research holds implications for investment strategies, risk management, second market pricing, financial decision-making, and the forthcoming establishment of the Burundian stock market.
{"title":"Modeling the yield curve of Burundian bond market by parametric models","authors":"Rédempteur Ntawiratsa, David Niyukuri, Irène Irakoze, Menus Nkurunziza","doi":"arxiv-2310.00321","DOIUrl":"https://doi.org/arxiv-2310.00321","url":null,"abstract":"The term structure of interest rates (yield curve) is a critical facet of\u0000financial analytics, impacting various investment and risk management\u0000decisions. It is used by the central bank to conduct and monitor its monetary\u0000policy. That instrument reflects the anticipation of inflation and the risk by\u0000investors. The rates reported on yield curve are the cornerstone of valuation\u0000of all assets. To provide such tool for Burundi financial market, we collected\u0000the auction reports of treasury securities from the website of the Central Bank\u0000of Burundi. Then, we computed the zero-coupon rates, and estimated actuarial\u0000rates of return by applying the Nelson-Siegel and Svensson models. This paper\u0000conducts a rigorous comparative analysis of these two prominent parametric\u0000yield curve models and finds that the Nelson-Siegel model is the optimal choice\u0000for modeling the Burundian yield curve. The findings contribute to the body of\u0000knowledge on yield curve modeling, enhancing its precision and applicability in\u0000financial markets. Furthermore, this research holds implications for investment\u0000strategies, risk management, second market pricing, financial decision-making,\u0000and the forthcoming establishment of the Burundian stock market.","PeriodicalId":501372,"journal":{"name":"arXiv - QuantFin - General Finance","volume":"205 ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138522777","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Early warning systems (EWSs) are critical for forecasting and preventing economic and financial crises. EWSs are designed to provide early warning signs of financial troubles, allowing policymakers and market participants to intervene before a crisis expands. The 2008 financial crisis highlighted the importance of detecting financial distress early and taking preventive measures to mitigate its effects. In this bibliometric review, we look at the research and literature on EWSs in finance. Our methodology included a comprehensive examination of academic databases and a stringent selection procedure, which resulted in the final selection of 616 articles published between 1976 and 2023. Our findings show that more than 90% of the papers were published after 2006, indicating the growing importance of EWSs in financial research. According to our findings, recent research has shifted toward machine learning techniques, and EWSs are constantly evolving. We discovered that research in this area could be divided into four categories: bankruptcy prediction, banking crisis, currency crisis and emerging markets, and machine learning forecasting. Each cluster offers distinct insights into the approaches and methodologies used for EWSs. To improve predictive accuracy, our review emphasizes the importance of incorporating both macroeconomic and microeconomic data into EWS models. To improve their predictive performance, we recommend more research into incorporating alternative data sources into EWS models, such as social media data, news sentiment analysis, and network analysis.
{"title":"A systematic review of early warning systems in finance","authors":"Ali Namaki, Reza Eyvazloo, Shahin Ramtinnia","doi":"arxiv-2310.00490","DOIUrl":"https://doi.org/arxiv-2310.00490","url":null,"abstract":"Early warning systems (EWSs) are critical for forecasting and preventing\u0000economic and financial crises. EWSs are designed to provide early warning signs\u0000of financial troubles, allowing policymakers and market participants to\u0000intervene before a crisis expands. The 2008 financial crisis highlighted the\u0000importance of detecting financial distress early and taking preventive measures\u0000to mitigate its effects. In this bibliometric review, we look at the research\u0000and literature on EWSs in finance. Our methodology included a comprehensive\u0000examination of academic databases and a stringent selection procedure, which\u0000resulted in the final selection of 616 articles published between 1976 and\u00002023. Our findings show that more than 90% of the papers were published after\u00002006, indicating the growing importance of EWSs in financial research.\u0000According to our findings, recent research has shifted toward machine learning\u0000techniques, and EWSs are constantly evolving. We discovered that research in\u0000this area could be divided into four categories: bankruptcy prediction, banking\u0000crisis, currency crisis and emerging markets, and machine learning forecasting.\u0000Each cluster offers distinct insights into the approaches and methodologies\u0000used for EWSs. To improve predictive accuracy, our review emphasizes the\u0000importance of incorporating both macroeconomic and microeconomic data into EWS\u0000models. To improve their predictive performance, we recommend more research\u0000into incorporating alternative data sources into EWS models, such as social\u0000media data, news sentiment analysis, and network analysis.","PeriodicalId":501372,"journal":{"name":"arXiv - QuantFin - General Finance","volume":"34 ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138522780","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Large language models (LLMs), including ChatGPT, can extract profitable trading signals from the sentiment in news text. However, backtesting such strategies poses a challenge because LLMs are trained on many years of data, and backtesting produces biased results if the training and backtesting periods overlap. This bias can take two forms: a look-ahead bias, in which the LLM may have specific knowledge of the stock returns that followed a news article, and a distraction effect, in which general knowledge of the companies named interferes with the measurement of a text's sentiment. We investigate these sources of bias through trading strategies driven by the sentiment of financial news headlines. We compare trading performance based on the original headlines with de-biased strategies in which we remove the relevant company's identifiers from the text. In-sample (within the LLM training window), we find, surprisingly, that the anonymized headlines outperform, indicating that the distraction effect has a greater impact than look-ahead bias. This tendency is particularly strong for larger companies--companies about which we expect an LLM to have greater general knowledge. Out-of-sample, look-ahead bias is not a concern but distraction remains possible. Our proposed anonymization procedure is therefore potentially useful in out-of-sample implementation, as well as for de-biased backtesting.
{"title":"Assessing Look-Ahead Bias in Stock Return Predictions Generated By GPT Sentiment Analysis","authors":"Paul Glasserman, Caden Lin","doi":"arxiv-2309.17322","DOIUrl":"https://doi.org/arxiv-2309.17322","url":null,"abstract":"Large language models (LLMs), including ChatGPT, can extract profitable\u0000trading signals from the sentiment in news text. However, backtesting such\u0000strategies poses a challenge because LLMs are trained on many years of data,\u0000and backtesting produces biased results if the training and backtesting periods\u0000overlap. This bias can take two forms: a look-ahead bias, in which the LLM may\u0000have specific knowledge of the stock returns that followed a news article, and\u0000a distraction effect, in which general knowledge of the companies named\u0000interferes with the measurement of a text's sentiment. We investigate these\u0000sources of bias through trading strategies driven by the sentiment of financial\u0000news headlines. We compare trading performance based on the original headlines\u0000with de-biased strategies in which we remove the relevant company's identifiers\u0000from the text. In-sample (within the LLM training window), we find,\u0000surprisingly, that the anonymized headlines outperform, indicating that the\u0000distraction effect has a greater impact than look-ahead bias. This tendency is\u0000particularly strong for larger companies--companies about which we expect an\u0000LLM to have greater general knowledge. Out-of-sample, look-ahead bias is not a\u0000concern but distraction remains possible. Our proposed anonymization procedure\u0000is therefore potentially useful in out-of-sample implementation, as well as for\u0000de-biased backtesting.","PeriodicalId":501372,"journal":{"name":"arXiv - QuantFin - General Finance","volume":"6 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138522834","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Irène Irakoze, Rédempteur Ntawiratsa, David Niyukuri
Constructing an accurate yield curve is essential for evaluating financial instruments and analyzing market trends in the bond market. However, in the case of the Burundian sovereign bond market, the presence of missing data poses a significant challenge to accurately constructing the yield curve. In this paper, we explore the limitations and data availability constraints specific to the Burundian sovereign market and propose robust methodologies to effectively handle missing data. The results indicate that the Linear Regression method, and the Previous value method perform consistently well across variables, approximating a normal distribution for the error values. The non parametric Missing Value Imputation using Random Forest (miss-Forest) method performs well for coupon rates but poorly for bond prices, and the Next value method shows mixed results. Ultimately, the Linear Regression (LR) method is recommended for imputing missing data due to its ability to approximate normality and predictive capabilities. However, filling missing values with previous values has high accuracy, thus, it will be the best choice when we have less information to be able to increase accuracy for LR. This research contributes to the development of financial products, trading strategies, and overall market development in Burundi by improving our understanding of the yield curve dynamics.
{"title":"Handling missing data in Burundian sovereign bond market","authors":"Irène Irakoze, Rédempteur Ntawiratsa, David Niyukuri","doi":"arxiv-2309.17379","DOIUrl":"https://doi.org/arxiv-2309.17379","url":null,"abstract":"Constructing an accurate yield curve is essential for evaluating financial\u0000instruments and analyzing market trends in the bond market. However, in the\u0000case of the Burundian sovereign bond market, the presence of missing data poses\u0000a significant challenge to accurately constructing the yield curve. In this\u0000paper, we explore the limitations and data availability constraints specific to\u0000the Burundian sovereign market and propose robust methodologies to effectively\u0000handle missing data. The results indicate that the Linear Regression method,\u0000and the Previous value method perform consistently well across variables,\u0000approximating a normal distribution for the error values. The non parametric\u0000Missing Value Imputation using Random Forest (miss-Forest) method performs well\u0000for coupon rates but poorly for bond prices, and the Next value method shows\u0000mixed results. Ultimately, the Linear Regression (LR) method is recommended for\u0000imputing missing data due to its ability to approximate normality and\u0000predictive capabilities. However, filling missing values with previous values\u0000has high accuracy, thus, it will be the best choice when we have less\u0000information to be able to increase accuracy for LR. This research contributes\u0000to the development of financial products, trading strategies, and overall\u0000market development in Burundi by improving our understanding of the yield curve\u0000dynamics.","PeriodicalId":501372,"journal":{"name":"arXiv - QuantFin - General Finance","volume":"8 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138522778","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pietro Saggese, Esther Segalla, Michael Sigmund, Burkhard Raunig, Felix Zangerl, Bernhard Haslhofer
Entities like centralized cryptocurrency exchanges fall under the business category of virtual asset service providers (VASPs). As any other enterprise, they can become insolvent. VASPs enable the exchange, custody, and transfer of cryptoassets organized in wallets across distributed ledger technologies (DLTs). Despite the public availability of DLT transactions, the cryptoasset holdings of VASPs are not yet subject to systematic auditing procedures. In this paper, we propose an approach to assess the solvency of a VASP by cross-referencing data from three distinct sources: cryptoasset wallets, balance sheets from the commercial register, and data from supervisory entities. We investigate 24 VASPs registered with the Financial Market Authority in Austria and provide regulatory data insights such as who are the customers and where do they come from. Their yearly incoming and outgoing transaction volume amount to 2 billion EUR for around 1.8 million users. We describe what financial services they provide and find that they are most similar to traditional intermediaries such as brokers, money exchanges, and funds, rather than banks. Next, we empirically measure DLT transaction flows of four VASPs and compare their cryptoasset holdings to balance sheet entries. Data are consistent for two VASPs only. This enables us to identify gaps in the data collection and propose strategies to address them. We remark that any entity in charge of auditing requires proof that a VASP actually controls the funds associated with its on-chain wallets. It is also important to report fiat and cryptoasset and liability positions broken down by asset types at a reasonable frequency.
{"title":"Assessing the Solvency of Virtual Asset Service Providers: Are Current Standards Sufficient?","authors":"Pietro Saggese, Esther Segalla, Michael Sigmund, Burkhard Raunig, Felix Zangerl, Bernhard Haslhofer","doi":"arxiv-2309.16408","DOIUrl":"https://doi.org/arxiv-2309.16408","url":null,"abstract":"Entities like centralized cryptocurrency exchanges fall under the business\u0000category of virtual asset service providers (VASPs). As any other enterprise,\u0000they can become insolvent. VASPs enable the exchange, custody, and transfer of\u0000cryptoassets organized in wallets across distributed ledger technologies\u0000(DLTs). Despite the public availability of DLT transactions, the cryptoasset\u0000holdings of VASPs are not yet subject to systematic auditing procedures. In\u0000this paper, we propose an approach to assess the solvency of a VASP by\u0000cross-referencing data from three distinct sources: cryptoasset wallets,\u0000balance sheets from the commercial register, and data from supervisory\u0000entities. We investigate 24 VASPs registered with the Financial Market\u0000Authority in Austria and provide regulatory data insights such as who are the\u0000customers and where do they come from. Their yearly incoming and outgoing\u0000transaction volume amount to 2 billion EUR for around 1.8 million users. We\u0000describe what financial services they provide and find that they are most\u0000similar to traditional intermediaries such as brokers, money exchanges, and\u0000funds, rather than banks. Next, we empirically measure DLT transaction flows of\u0000four VASPs and compare their cryptoasset holdings to balance sheet entries.\u0000Data are consistent for two VASPs only. This enables us to identify gaps in the\u0000data collection and propose strategies to address them. We remark that any\u0000entity in charge of auditing requires proof that a VASP actually controls the\u0000funds associated with its on-chain wallets. It is also important to report fiat\u0000and cryptoasset and liability positions broken down by asset types at a\u0000reasonable frequency.","PeriodicalId":501372,"journal":{"name":"arXiv - QuantFin - General Finance","volume":"26 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2023-09-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138543769","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：