Testing of nonstationary cycles in financial time series data

In this article we propose a method for testing nonstationary cycles in financial time series data. We use a procedure that permits us to test unit root cycles in raw time series. The test has several distinguishing features compared with other procedures. In particular, it has a standard null limit distribution and is the most efficient test when directed against the appropriate (fractional) alternatives. In addition, it allows us to test unit root cycles at each of the frequencies, and, thus, it permits us to approximate the number of periods per cycle. The results, based on the daily structure of Spanish Stock Market prices (IBEX35), show that some intra-year cycles occur, and they take place at approximately 6, 9 or between 24 and 50 periods. The analysis was extended to several other stock market indices of various countries and though the results differ in terms of frequencies, the same conclusions hold, finding evidence of intra-year cyclical effects in all countries. JEL Classification C22; G14.     

A simple non-linear model with fractional integration for financial time series data

This paper provides several examples of simple non-linear time series models with fractionally integrated disturbances. Both types of models (non-linear and fractional integration) have been widely used in recent years when modeling financial data. We use a testing procedure that permits us to test the order of integration in raw time series in the context of non-linear models. The tests are applied to several financial time series, the results showing that when the non-linear sign structure is taken into account, the order of integration of the series is much higher than one, finding thus conclusive evidence against mean reversion in their behavior.     

A new approach for testing the randomness of heteroskedastic time series data

Proposed is a conditional approach for testing the randomness of heteroskedastic time series data as well as for checking the validity of this testing. It is shown that the ordinary serial correlation test works correctly neither for daily sequence of the TOPIX index in Tokyo Stock Exchange nor for heteroskedastic models, while our approach works well for them. It is also shown that our approach is enough powerful for detecting the departure from the randomness. An advantage of this approach is that it allows us to use any quantity for testing. Its application to the TOPIX index detected statistically significant long term correlation which causes both the mean reversion and the outperformance of the Alexander's filter rule over the buy-and-hold strategy.     

Prediction in chaotic time series: methods and comparisons with an application to financial intra-day data

Different prediction methods for chaotic deterministic systems are compared. Two methods of reconstructing the dynamics of the systems are considered with a view to producing a profitable trading model. The methods developed are the ‘nearest neighbours’ method and the ‘radial basis functions’ method. The optimal prediction horizon according to the sampling time step, and a reliable method to measure the prediction error are discussed. These methods are applied to the intra-day series of exchange rates, namely DEM/FRF. Developments concerning the importance of noise when chaotic systems are studied are provided.     

A stochastic frontier approach to modelling financial constraints in firms: An application to India

We propose the use of stochastic frontier approach to modelling financial constraints of firms. The main advantage of the stochastic frontier approach over the stylised approaches that use pooled OLS or fixed effects panel regression models is that we can not only decide whether or not the average firm is financially constrained, but also estimate a measure of the degree of the constraint for each firm and for each time period, and also the marginal impact of firm characteristics on this measure. We then apply the stochastic frontier approach to a panel of Indian manufacturing firms, for the 1997–2006 period. In our application, we highlight and discuss the aforementioned advantages, while also demonstrating that the stochastic frontier approach generates regression estimates that are consistent with the stylised intuition found in the literature on financial constraint and the wider literature on the Indian credit/capital market.     

利用数据挖掘技术开展金融CRM

我国加入WTO后,外资金融机构将纷纷登陆中国,外资银行的进入,不仅会改变中国金融市场竞争的格局,还会进一步加剧金融市场的激烈竞争。目前,银行客户流失率已成为各家银行普遍关注的一个重要问题。商业银行经营有这样两个特点:第一,基础设施投资巨大,生存和发展的途径只能靠吸引和保留足够数量的客户的手段;第二,商品差别小,各个商业银行很难提供差别化的金融产品,长期留住客户比较困难。如果国内银行依然是靠粗放经营、粗略的数据分析和市场判断,将无法与资本雄厚、管理先进、数据集中的外资银行抗衡。如何在激烈的商业竞争中把握市场良机…     

Forecasting financial time-series using data mining models: A simulation study

In this paper, we examine the static and dynamic predictive ability of artificial neural networks and random forests for financial time series within a simulation context. Our simulation design, in which we generate data from an AR(1)-GARCH(1,1) model, allows for several degrees of persistence in the mean equation to mimic the behavior of short and long-horizon asset returns. While the true data generating process beats the data mining techniques in terms of static forecasting, the novelty in this paper is to demonstrate that the data mining techniques outperform the true model under a dynamic forecasting scheme for moderate to highly persistent time series. We provide an empirical application using one-day and long-horizon returns on two exchange rates. Our empirical findings corroborate our simulation results in that the data mining models exhibit superior predictive ability for highly persistent time series. We discuss the importance of our findings for asset allocation and portfolio management.     

A unified framework for robust modelling of financial markets in discrete time

Finance and Stochastics - We unify and establish equivalence between the pathwise and the quasi-sure approaches to robust modelling of financial markets in finite discrete time. In particular, we...     

XBRL and open data for global financial ecosystems: A linked data approach

Information professionals performing business activity related investigative analysis must routinely associate data from a diverse range of Web based general-interest business and financial information sources. XBRL has become an integral part of the financial data landscape. At the same time, Open Data initiatives have contributed relevant financial, economic, and business data to the pool of publicly available information on the Web but the use of XBRL in combination with Open Data remains at an early state of realisation. In this paper we argue that Linked Data technology, created for Web scale information integration, can accommodate XBRL data and make it easier to combine it with open datasets. This can provide the foundations for a global data ecosystem of interlinked and interoperable financial and business information with the potential to leverage XBRL beyond its current regulatory and disclosure role. We outline the uses of Linked Data technologies to facilitate XBRL consumption in conjunction with non-XBRL Open Data, report on current activities and highlight remaining challenges in terms of information consolidation faced by both XBRL and Web technologies.     

Clustering financial time series with variance ratio statistics

This study introduces a new distance measure for clustering financial time series based on variance ratio test statistics. The proposed metric attempts to assess the level of interdependence of time series from the point of view of return predictability. Simulation results show that this metric aggregates time series according to their serial dependence structure better than a metric based on the sample autocorrelations. An empirical application of this approach to international stock market returns is presented. The results suggest that this metric discriminates stock markets reasonably well according to size and the level of development. Furthermore, despite the substantial evolution of individual variance ratio statistics, the clustering pattern remains fairly stable across different time periods.     

Estimation of tail-related risk measures for heteroscedastic financial time series: an extreme value approach

We propose a method for estimating Value at Risk (VaR) and related risk measures describing the tail of the conditional distribution of a heteroscedastic financial return series. Our approach combines pseudo-maximum-likelihood fitting of GARCH models to estimate the current volatility and extreme value theory (EVT) for estimating the tail of the innovation distribution of the GARCH model. We use our method to estimate conditional quantiles (VaR) and conditional expected shortfalls (the expected size of a return exceeding VaR), this being an alternative measure of tail risk with better theoretical properties than the quantile. Using backtesting of historical daily return series we show that our procedure gives better 1-day estimates than methods which ignore the heavy tails of the innovations or the stochastic nature of the volatility. With the help of our fitted models we adopt a Monte Carlo approach to estimating the conditional quantiles of returns over multiple-day horizons and find that this outperforms the simple square-root-of-time scaling method.     

A long-memory integer-valued time series model,INARFIMA, for financial application

A model to account for the long-memory property in a count data framework is proposed and applied to high-frequency stock transactions data. By combining features of the INARMA and ARFIMA models, an Integer-valued Auto Regressive Fractionally Integrated Moving Average (INARFIMA) model is proposed. The unconditional and conditional first- and second-order moments are given. The CLS, FGLS and GMM estimators are discussed. In its empirical application to two stock series for AstraZeneca and Ericsson B, we find that both series have a fractional integration property.     

Combining data and text mining techniques for analysing financial reports

There is a vast amount of financial information on companies' financial performance available to investors in electronic form today. While automatic analysis of financial figures is common, it has been difficult to extract meaning from the textual parts of financial reports automatically. The textual part of an annual report contains richer information than the financial ratios. In this paper, we combine data and text mining methods for analysing quantitative and qualitative data from financial reports, in order to see if the textual part of the report contains some indications about future financial performance. The quantitative analysis has been performed using self‐organizing maps, and the qualitative analysis using prototype‐matching text clustering. The analysis is performed on the quarterly reports of three leading companies in the telecommunications sector. Copyright © 2004 John Wiley & Sons, Ltd.     

Stylised facts of financial time series and hidden Markov models in continuous time

Hidden Markov models are often applied in quantitative finance to capture the stylised facts of financial returns. They are usually discrete-time models and the number of states rarely exceeds two because of the quadratic increase in the number of parameters with the number of states. This paper presents an extension to continuous time where it is possible to increase the number of states with a linear rather than quadratic growth in the number of parameters. The possibility of increasing the number of states leads to a better fit to both the distributional and temporal properties of daily returns.     

A new approach to forecasting exchange rates

Building on purchasing power parity theory, this paper proposes a new approach to forecasting exchange rates using the Big Mac data from The Economist magazine. Our approach is attractive in three aspects. Firstly, it uses easily-available Big Mac prices as input. These prices avoid several potential problems associated with broad price indexes, such as the consumer price index used in conventional PPP studies. Secondly, this approach provides real-time exchange-rate forecasts at any forecast horizon. These high-frequency forecasts could be appealing to those who want up-to-date exchange-rate forecasts. Finally, as our forecasts are obtained through a simulation procedure, estimation uncertainty is made explicit in our framework that provides the entire distribution of exchange rates, not just a single point estimate. Using exchange rates of six major currencies to illustrate the approach, we compare the Big Mac forecasts with those derived from a random walk and the CPI and find some support for our approach, especially at longer term horizons.     

Jump detection with wavelets for high-frequency financial time series

This paper introduces a new nonparametric test to identify jump arrival times in high frequency financial time series data. The asymptotic distribution of the test is derived. We demonstrate that the test is robust for different specifications of price processes and the presence of the microstructure noise. A Monte Carlo simulation is conducted which shows that the test has good size and power. Further, we examine the multi-scale jump dynamics in US equity markets. The main findings are as follows. First, the jump dynamics of equities are sensitive to data sampling frequency with significant underestimation of jump intensities at lower frequencies. Second, although arrival densities of positive jumps and negative jumps are symmetric across different time scales, the magnitude of jumps is distributed asymmetrically at high frequencies. Third, only 20% of jumps occur in the trading session from 9:30 AM to 4:00 PM, suggesting that illiquidity during after-hours trading is a strong determinant of jumps.     

Quant GANs: deep generation of financial time series

Modeling financial time series by stochastic processes is a challenging task and a central area of research in financial mathematics. As an alternative, we introduce Quant GANs, a data-driven model which is inspired by the recent success of generative adversarial networks (GANs). Quant GANs consist of a generator and discriminator function, which utilize temporal convolutional networks (TCNs) and thereby achieve to capture long-range dependencies such as the presence of volatility clusters. The generator function is explicitly constructed such that the induced stochastic process allows a transition to its risk-neutral distribution. Our numerical results highlight that distributional properties for small and large lags are in an excellent agreement and dependence properties such as volatility clusters, leverage effects, and serial autocorrelations can be generated by the generator function of Quant GANs, demonstrably in high fidelity.