The development of a predictive system that correctly forecasts trading signals is crucial for algorithmic trading and investment management. Technical analysis has been used by many researchers for financial market prediction. Numerous technical indicators (TIs) are computed by setting a time-frame parameter called the input window length. This paper therefore investigates how the input window length and forecast horizon together affect the predictive performance of the model. Market-specific TIs are extracted through a random forest technique. These TIs are used as inputs for an artificial neural network and a support vector machine to forecast the future direction of trading signals. The data set consists of 22 years of daily prices for the Pakistan Stock Exchange. This research finds the 15 most relevant features for the Pakistan Stock Exchange from a list of 34 TIs. The prediction system performs best when the forecast horizon is more than 15 days, which shows the dependency of the input variable parameter selection and the forecast horizon. This unique pattern is studied using multiple confusion metrics. The findings of this study may improve the prediction accuracy of a trading strategy based on technical analysis.
The vector autoregressive model has long been used for portfolio analysis, while a recent extension (VARX) incorporates exogenous factors. Despite its increased forecasting precision, the applicability of the VARX model is seriously hampered in the absence of sparsity, as the number of coefficients to be estimated grows quadratically with the number of series. We introduce a novel regularization method for the VARX model in the context of portfolios, where weighted links between portfolios are used to construct a penalty function for the autoregressive parameter matrixes. To test the prediction performance of the new method, we first cluster the time series into several groups using wavelet decomposition and hierarchical clustering, after which we construct data sets of different homogeneity. Our method is advantageous in two ways: the computation time for the model is significantly reduced, and the forecasting precision of the model is enhanced by 50% compared with existing regularization methods.
Banking systems are at the center of the financial infrastructure of any country. It has become apparent after the subprime crisis that such systems cannot be studied by looking at their components individually (that is, in isolation). Thus, an integrated approach is needed.In this paper we introduce a numerically friendly, yet general, algorithm that allows us to represent a banking system in a realistic way. We start with a detailed description of the banks' balance sheets and we incorporate two different features to account for connectivity effects: interbank loans and correlated-exposures to a common universe of credit-risk loans. The driving force behind the model is the progressive deterioration of the loan portfolios to which the banks are exposed. This method is useful to identify the weaknesses of a given banking system, and it is also helpful to assess the merits of different potential responses from the regulator's viewpoint. An example based on the banking system of a Latin American country is used to demonstrate the merits of this approach.Some of our findings are in agreement with those of previous authors, namely, that the probability of developing cascades is proportional to the degree of connectivity of the banking networks, and that the maximum resilience for a given set on banks is achieved for some "intermediate" (optimal?) level of connectivity. However, we also find -- in contrast with a few earlier researchers -- that investigating the features of a banking system by collapsing banks individually (and keeping other features unaltered) can give a misleading view of the system resilience.
A block-structured model for the reconstruction of directed and weighted financial networks spanning multiple countries is developed. In a first step, link probability matrixes are derived via a fitness model that is calibrated to reproduce a desired density and reciprocity for each block (ie, country and cross-border submatrix). The resulting probability matrix allows for fast simulation through bivariate Bernoulli trials. In a second step, weights are allocated to a sampled adjacency matrix via an exponential random graph model (ERGM) that fulfills the desired row, column and block weights. This model is analytically tractable, calibrated only on scarce publicly available data and closely reconstructs known network characteristics of financial markets. In addition, an algorithm for the parameter estimation of the ERGM is presented. Further, calibrating our model to the European Union interbank market, we are able to assess the systemic risk within the European banking network by applying various contagion models.
Computation of spectral structure and risk measures from networks of multivariate financial time series data has been at the forefront of the statistical finance literature for a long time. A standard mode of analysis is to consider log returns from the equity price data, which is akin to taking first difference ($d = 1$) of the log of the price data. Sometimes authors have considered simple growth rates as well. Either way, the idea is to get rid of the nonstationarity induced by the {\it unit root} of the data generating process. However, it has also been noted in the literature that often the individual time series might have a root which is more or less than unity in magnitude. Thus first differencing leads to under-differencing in many cases and over differencing in others. In this paper, we study how correcting for the order of differencing leads to altered filtering and risk computation on inferred networks. In summary, our results are: (a) the filtering method with extreme information loss like minimum spanning tree as well as filtering with moderate information loss like triangulated maximally filtered graph are very susceptible to such d-corrections, (b) the spectral structure of the correlation matrix is quite stable although the d-corrected market mode almost always dominates the uncorrected (d = 1) market mode indicating under-estimation in the standard analysis, and (c) the PageRank-based risk measure constructed from Granger-causal networks shows an inverted U-shape evolution in the relationship between d-corrected and uncorrected return data over the period of analysis 1972-2018 for historical data of NASDAQ.
This paper describes asset price and return disturbances as result of relations between transactions and multiple kinds of expectations. We show that disturbances of expectations can cause fluctuations of trade volume, price and return. We model price disturbances for transactions made under all types of expectations as weighted sum of partial price and trade volume disturbances for transactions made under separate kinds of expectations. Relations on price allow present return as weighted sum of partial return and trade volume "return" for transactions made under separate expectations. Dependence of price disturbances on trade volume disturbances as well as dependence of return on trade volume "return" cause dependence of volatility and statistical distributions of price and return on statistical properties of trade volume disturbances and trade volume "return" respectively.
A growing body of studies on systemic risk in financial markets has emphasized the key importance of taking into consideration the complex interconnections among financial institutions. Much effort has been put in modeling the contagion dynamics of financial shocks, and to assess the resilience of specific financial markets - either using real network data, reconstruction techniques or simple toy networks. Here we address the more general problem of how shock propagation dynamics depends on the topological details of the underlying network. To this end we consider different realistic network topologies, all consistent with balance sheets information obtained from real data on financial institutions. In particular, we consider networks of varying density and with different block structures, and diversify as well in the details of the shock propagation dynamics. We confirm that the systemic risk properties of a financial network are extremely sensitive to its network features. Our results can aid in the design of regulatory policies to improve the robustness of financial markets.
This work studies contagion risk through the portfolio investment channel using network analysis and simulation on bilateral cross-country data. The importance of the portfolio channel in the transmission of financial shocks reflects the high interconnectedness of the global financial system, which diminished in the aftermath of the global financial crisis but has resumed in recent years. The network representing cross-country portfolio investments turns out to be highly concentrated around the main financial centers, which act as global hubs connecting with nodes that are not linked directly. Using a network simulation model based on the assumption that international investors rebalance their portfolios after an idiosyncratic shock, reducing investments in countries to which they are overexposed, we find that contagion effects may be significant even when the shock originates in a peripheral country. In addition, the model suggests contagion risk has risen since the global financial crisis, owing to the greater financial integration of emerging economies.
This paper analyzes how systemic risk structurally evolved between 2007 and 2017. The main contributions of the paper to the literature include the methodology, analysis and potential use for macroprudential policies. The methodology, known as network analysis, comprises direct (credit and liquidity risk) and indirect (concentration risk) contagion channels as well as other specificities that improve the methodologies exploited so far in the literature. Using a consolidated sample, which varies between 14 and 17 banks over the period 2007–17, we show that the structural systemic risk of the Portuguese banking system reduced between 2007 and 2017. Further, in line with most of the literature, this paper highlights that direct contagion is not significant compared with contagion that stems from banks’ common exposures to asset classes. Finally, this paper supports the role played by capital in mitigating structural systemic risk, and the model behind the analysis can be used to perform stress tests with a macroprudential dimension as well as to calibrate structural capital buffers such as the other systemically important institutions and systemic risk buffers.
Cryptocurrencies represent an asset class featuring two unique properties: they are not backed by sovereigns, and their supply is fixed exogenously. This combination becomes apparent in their volatility, which is driven only by demand-side factors. In particular, cryptocurrencies represent an extreme case of the excess volatility puzzle, with asset prices moving more than the fundamentals. We explore the effects of market capitalization on the dynamics of cryptocurrencies within both returns and volatility networks and show that these cryptocurrencies exhibit scaling properties in volatility with respect to market capitalization. The dependency network suggests that currencies with a larger market share have a larger presence in the dominant eigenspectrum, and they exert more influence in the comovement network. In these regards, we find parallels between the dynamics of cryptocurrencies and those of more traditional asset classes. Our findings have implications for both researchers and practitioners in terms of modeling and analyzing the collective behavior of financial assets.
Recent regulatory reforms like the mandatory clearing of standardized swap contracts and mandatory trading on centralized execution platforms have significantly changed the derivatives landscape. These reforms have, in certain cases, led the market to increasingly trade on multilateral platforms, potentially affecting the average cost of execution. Prior research has examined the effects of centralized trading on execution costs and has generally found reduced costs, especially for entities with higher transaction volumes or greater execution flexibility. We use detailed information on the trading of credit index swaps, the most actively traded credit derivatives instrument, between May 2014 and Sep 2016. We find that the customers who trade with a higher number of dealers (high network degree) and those who trade with the most active dealers (high network centrality) incur lower trading costs; for at least the less liquid indices, this cost improvement increases as trade size increases. We also identify a few liquidity trends: measures like average daily volumes, average price impact, and price dispersion have remained steady or have improved. However, during the same period, trade sizes for certain indices may have declined slightly.
Automated digital consultancy platforms ("robot advisors") reduce costs and improve the perceived service quality, speeding it up and making user involvement more transparent. These improvements are often offset by risk classification models that are simpler than those employed in traditional consultancy. We show how to exploit the available data to build portfolios that better fit the risk profiles of investors. This is made possible, on the one hand, by constructing groups of homogeneous risk profiles based on user responses to the markets in financial instruments directive (MIFID) questionnaire, and, on the other hand, by constructing homogeneous clusters of financial assets based on their risk and return performance. We also show that machine learning methods and, specifically, neural network models can be used to "automatize" the previous classifications and, eventually, to assess whether an investor's portfolio matches their risk profile.
We present new evidence on the structure of euro area securities markets using a multilayer network approach. Layers are broken down by key instruments and maturities as well as the secured nature of the transaction. This paper utilizes a unique dataset of banking sector crossholdings of securities to map these exposures among banks and economic and financial sectors. We can compare and contrast funding and exposure networks among banks themselves and of banks, non-banks and the wider economy. The analytical approach presented here is highly relevant for the design of appropriate prudential measures, since it supports the identification of counterparty risk, concentration risk and funding risk within the interbank network and the wider macro-financial network.
The global nature of derivatives markets, and the presence of large key financial institutions trading in several markets across the globe, calls for taking a "macro" view of the interconnections arising in the clearing network. Based on the analysis of derivatives transactions data reported under the European Market Infrastructure Regulation (EMIR), we reconstruct the network of relationships in the centrally cleared derivatives market and analyze its topology, providing insight into its structural features. The centrally cleared derivatives network is modeled in the form of a multiplex network, where each layer is represented by a derivatives asset class market. In turn, each node represents a single counterparty in that market. On the basis of different centrality measures applied to the collapsed aggregate network and to the multiplex network, the critical participants of the euro area centrally cleared derivatives market are identified and their level of interconnectedness is analyzed. This paper provides insight into how the collected data pursuant to the EMIR can be used to shed light on the complex network of interrelations underlying the financial markets. It provides indications of the structural features of the euro area centrally cleared derivatives market and discusses policy-relevant implications as well as future applications.
In this paper, we investigate a credit rating problem based on the network of trading information (NoTI). First, several popular tools, such as assortativity analysis, community detection and centrality measurement, are introduced for analyzing the topology structures and properties of the NoTI. Then, the correlation between the characteristics of the network and the credit ratings is investigated to illustrate the feasibility of credit risk analysis based on the NoTI. Sovereign rating based on the world trade network is analyzed as a case study. The correlation between the centrality metrics and the sovereign ratings conducted by Standard & Poor's clearly shows that highly ranked economies with vigorous economic trading links usually have higher credit ratings. Finally, a simulation is conducted to illustrate the degree of improvement in credit rating prediction accuracy if the NoTI is considered as an additional attribute.
This paper examines the relationship between the topology of interbank networks and their ability to propagate localized, idiosyncratic shocks across the banking sector via banks’ interbank claims on one another. We begin by creating a wide variety of networks and heterogeneous balance sheet structures using a generative algorithm capable of replicating key characteristics of real-world interbank networks. On each network, we run a standard financial contagion model with cascading defaults. Our modeling framework differentiates between random and targeted shocks of varying magnitude. Interbank contagion comprises a direct channel via banks’ cross-exposures and an indirect channel due to liquidity effects and external asset fire sales. Last, we develop an empirical model to test which global features of the network, aggregate banking sector balance sheet and shock properties drive contagion dynamics. Our results show a strong stabilizing role played by system leverage across all specifications. Among the global network measures, average path length, network density and assortativity are shown to drive the number of failures in a manner consistent with their definition. Similarly, the centralities of the shocked banks (across all definitions) play a significant role in the default cascade. By contrast, allowing for liquidity effects diminishes the explanatory power of the network on both global and local scales. However, taking the change in asset price (primarily driven by liquidity effects) as the dependent variable reestablishes the link between the network structure and the extent of financial contagion.
To capture the systemic complexity of international financial systems, network data is an important prerequisite. However, dyadic data is often not available, raising the need for methods that allow for reconstructing networks based on limited information. In this paper, we are reviewing different methods that are designed for the estimation of matrices from their marginals and potentially exogenous information. This includes a general discussion of the available methodology that provides edge probabilities as well as models that are focussed on the reconstruction of edge values. Besides summarizing the advantages, shortfalls and computational issues of the approaches, we put them into a competitive comparison using the SWIFT (Society for Worldwide Interbank Financial Telecommunication) MT 103 payment messages network (MT 103: Single Customer Credit Transfer). This network is not only economically meaningful but also fully observed which allows for an extensive competitive horse race of methods. The comparison concerning the binary reconstruction is divided into an evaluation of the edge probabilities and the quality of the reconstructed degree structures. Furthermore, the accuracy of the predicted edge values is investigated. To test the methods on different topologies, the application is split into two parts. The first part considers the full MT 103 network, being an illustration for the reconstruction of large, sparse financial networks. The second part is concerned with reconstructing a subset of the full network, representing a dense medium-sized network. Regarding substantial outcomes, it can be found that no method is superior in every respect and that the preferred model choice highly depends on the goal of the analysis, the presumed network structure and the availability of exogenous information.
Network science is being increasingly utilized to assist in the search for causes of irregular behavior in financial markets. The search gained greater impetus after traditional finance theories were unable to predict the extent of the most recent global financial crisis. The increased abilities of researchers to access and manipulate data has also opened new avenues of investigation, including the discovery of key networks and the agents that interact within them. In this paper, an analysis of the temporal networks formed between US institutional investors and Standard & Poor's 500 stocks between 2007 and 2010 is presented, with the results identifying key relationships between the density of these networks and the movement of the market. The analysis also identified the changing behavior of investors, as their risk aversion varied ahead of the market's price movements. To a lesser degree, relationships between the return of individual stocks and their investor networks are reported.
We provide evidence that the presence of bankers in the board of directors reduce information asymmetry between credit markets and firms. We show that the impact of the presence of bankers on leverage is driven by firms with low level of debt. This effect is amplified the more connected the bankers are to the corporate world. Additionally the results are more pronounced for less transparent firms. Our findings suggest that the connectedness of bankers play a key role in reducing information asymmetry.
