scholarly journals Non-linearities and GARCH Effects in the Emerging Stock Markets of South Asia

2007 ◽  
Vol 32 (3) ◽  
pp. 23-38 ◽  
Author(s):  
Raj S Dhankar ◽  
Madhumita Chakraborty
Keyword(s):  
Time Series ◽  
South Asia ◽  
Linear Dependence ◽  
Stock Markets ◽  
Theory And Practice ◽  
Future Price ◽  
Conditional Heteroscedasticity ◽  
Identical Distribution ◽  
Non Linear ◽  
Autoregressive Conditional Heteroscedasticity

Up to the beginning of the last decade, financial economics was dominated by linear paradigm, which assumed that economic time series conformed to linear models or could be wellapproximated by a linear model. However, there is increasing evidence that asset returns may be better characterized by a model which allows for non-linear behaviour. Though more efforts are now being directed towards the Asian stock markets in the light of their increasing importance to the investment world and the world economy, there is an extremely sparse literature, which utilizes recent advances in non-linear dynamics to examine the data generating process of the South-Asian stock markets. This study investigates the presence of non-linear dependence in three major markets of South Asia: India, Sri Lanka, and Pakistan. It was, however, realized that merely identifying non-linear dependence was not enough. Previous research has shown that the presence of nonlinear characteristics usually takes the form of ARCH/GARCH (Autoregressive Conditional Heteroscedasticity or Generalized Autoregressive Conditional Heteroscedasticity) type conditional heteroscedasticity. Keeping this in view, this study investigates whether the non-linear dependence is caused by predictable conditional volatility. It has been found that the simple GARCH (1, 1) model has fitted all the market return series adequately and accounted for the non-linearity found in the series. The findings reveal the following: The application of the BDS test developed by Brock, et al., (1996) strongly rejects the null hypothesis of independent and identical distribution of the return series as well as the linearly filtered return series for all the markets under study. With the possibility of linear dependence causing the rejection of independent and identical distribution (IID) being eliminated by linear filtering, the study also shows that non-stationarity of return series is also not a cause for non-IID behaviour by applying Augmented Dickey Fuller test and Phillips-Perron test. This implies the presence of non-linear dependence in the return series. For researchers in the developing countries, it is time to embrace the shift to non-linearity as it would provide a better understanding of the underlying dynamics of financial time series. However, the results are not necessarily inconsistent with efficient market hypothesis, simply because non-linearity does not essentially imply predictability as the future price changes can be predictable but only with a time horizon too short to allow for excess profits. The implications of non-linear dependence and presence of GARCH effects go beyond the issue of market efficiency. The common assumption of constant variance underlying the theory and practice of option pricing, portfolio optimization, and value-at-risk (VaR) calculations needs to be revised. If the assumed stochastic processes do not adequately depict the full complexity of the true generating processes, then any derivatives in question may be mis-priced.

Long memory volatility in Asian stock markets

2017 ◽  
Vol 29 (3) ◽  
pp. 423-442 ◽  
Author(s):  
Geeta Duppati ◽  
Anoop S. Kumar ◽  
Frank Scrimgeour ◽  
Leon Li
Keyword(s):  
Long Memory ◽  
Moving Average ◽  
Realized Volatility ◽  
Theory And Practice ◽  
Long Term Memory ◽  
Conditional Heteroscedasticity ◽  
Term Memory ◽  
Content Type ◽  
Autoregressive Conditional Heteroscedasticity

Purpose The purpose of this paper is to assess to what extent intraday data can explain and predict long-term memory. Design/methodology/approach This article analysed the presence of long-memory volatility in five Asian equity indices, namely, SENSEX, CNIA, NIKKEI225, KO11 and FTSTI, using five-min intraday return series from 05 January 2015 to 06 August 2015 using two approaches, i.e. conditional volatility and realized volatility, for forecasting long-term memory. It employs conditional-generalized autoregressive conditional heteroscedasticity (GARCH), i.e. autoregressive fractionally integrated moving average (ARFIMA)-FIGARCH model and ARFIMA-asymmetric power autoregressive conditional heteroscedasticity (APARCH) models, and unconditional volatility realized volatility using autoregressive integrated moving average (ARIMA) and ARFIMA in-sample forecasting models to estimate the persistence of the long-term memory. Findings Given the GARCH framework, the ARFIMA-APARCH long-memory model gave the better forecast results signifying the importance of accounting for asymmetric information when modelling volatility in a financial market. Using the unconditional realized volatility results from the Singapore and Indian markets, the ARIMA model outperforms the ARFIMA model in terms of forecast performance and provides reasonable forecasts. Practical implications The issue of long memory has important implications for the theory and practice of finance. It is well-known that accurate volatility forecasts are important in a variety of settings including option and other derivatives pricing, portfolio and risk management. Social implications It could be said that using long-memory augmented models would give better results to investors so that they could analyse the market trends in returns and volatility in a more accurate manner and reach at an informed decision. This is useful to minimize the risks. Originality/value This research enhances the literature by estimating the influence of intraday variables on daily volatility. This is one of very few studies that uses conditional GARCH framework models and unconditional realized volatility estimates for forecasting long-term memory. The authors find that the methods complement each other.

scholarly journals Time Series Modelling of Inflation in Botswana Using Monthly Consumer Price Indices

2016 ◽  
Vol 8 (3) ◽  
pp. 15
Author(s):  
Kesaobaka Molebatsi ◽  
Mpho Raboloko
Keyword(s):  
Time Series ◽  
Confidence Intervals ◽  
Price Index ◽  
Moving Average ◽  
Consumer Price Index ◽  
Conditional Heteroscedasticity ◽  
Autoregressive Integrated Moving Average ◽  
Time Series Modelling ◽  
Error Terms ◽  
Autoregressive Conditional Heteroscedasticity

<p>This paper identifies an autoregressive integrated moving average (ARIMA (1,1,1)) model that can be used to model inflation measured by the consumer price index (CPI) for Botswana. The paper proceeds to improve the model by incorporating the generalized autoregressive conditional heteroscedasticity (ARCH/GARCH) model that takes into consideration volatility in the series. Ultimately, CPI is forecast using the two models, ARIMA (1, 1, 1) and ARIMA (1, 1, 1) + GARCH (1, 2) and compared with the actual CPI. Both models perform well in terms of forecasting as their 95 percent confidence intervals cover the actual CPI. Marginal differences that favour the inclusion of the ARCH/GARCH components were observed when testing for normality among error terms. The paper also reveals that volatility for Botswana’s CPI is low as shown by small values of ARCH/GARCH components.</p>

scholarly journals PEMODELAN RETURN HARGA SAHAM MENGGUNAKAN MODEL INTERVENSI–ARCH/GARCH (Studi Kasus : Return Harga Saham PT Bayan Resources Tbk)

2018 ◽  
Vol 7 (2) ◽  
pp. 110-118
Author(s):  
Dea Manuella Widodo ◽  
Sudarno Sudarno ◽  
Abdul Hoyyi
Keyword(s):  
Time Series ◽  
Stock Price ◽  
Time Series Data ◽  
Step Function ◽  
Series Data ◽  
Model Data ◽  
Conditional Heteroscedasticity ◽  
Constant Variance ◽  
High Volatility ◽  
Autoregressive Conditional Heteroscedasticity

The intervention method is a time series model which could be used to model data with extreme fluctuation whether up or down. Stock price return tend to have extreme fluctuation which is caused by internal or external factors. There are two kinds of intervention function; a step function and a pulse function. A step function is used for a long-term intervention, while a pulse function is used for a short-term intervention. Modelling a time series data needs to satisfy the homoscedasticity assumptions (variance of residual is homogeneous).  In reality, stock price return has a high volatility, in other words it has a non-constant variance of residuals (heteroscedasticity). ARCH (Autoregressive Conditional Heteroscedasticity) or GARCH (Generalized Autoregressive Conditional Heteroscedasticity) can be used to model data with heteroscedasticity. The data used is stock price return from August 2008 until September 2018. From the stock price return data plot is found an extreme fluctuation in September 2017 (T=110) that is suspected as a pulse function. The best model uses the intervention pulse function is ARMA([1,4],0) (b=0, s=1, r=1). The intervention model has a non-constant variance or there is an ARCH effect. The best variance model obtained is ARMA([1,4],0)(b=0, s=1, r=1)–GARCH(1,1) with the AIC value is -205,75088. Keywords: Stock Return, Intervention, Heteroscedasticity, ARCH/GARCH 

scholarly journals African Stock Markets and Return Predictability

2016 ◽  
Vol 8 (5(J)) ◽  
pp. 91-99
Author(s):  
Gyamfi NE ◽  
Kyei KA ◽  
Gill R
Keyword(s):  
Time Series ◽  
Stock Markets ◽  
Weak Form ◽  
Return Predictability ◽  
Stationary Time Series ◽  
Conditional Heteroscedasticity ◽  
Rolling Window ◽  
Linear And Nonlinear ◽  
Window Approach ◽  
Over Time

This article re-examines the return predictability of eight African stock markets. When returns of stocks are predictable, arbitrageurs make abnormal gains from analyzing prices. The study uses a non-parametric Generalised Spectral (GS) test in a rolling window approach. The rolling window approach tracts the periods of efficiency over time. The GS test is robust to conditional heteroscedasticity and it detects the presence of linear and nonlinear dependencies in a stationary time series. Our results support the Adaptive Market Hypothesis (AMH). This is because, indices whose returns were observed to be predictable by analyzing them in absolute form and therefore weak - form inefficient showed trends of unpredictability in a rolling window.

Comparison of linear and non-linear GARCH models for forecasting volatility of select emerging countries

2020 ◽  
Vol ahead-of-print (ahead-of-print) ◽  
Author(s):  
Sudhi Sharma ◽  
Vaibhav Aggarwal ◽  
Miklesh Prasad Yadav
Keyword(s):  
Global Financial Crisis ◽  
Emerging Countries ◽  
Expected Returns ◽  
Garch Models ◽  
Conditional Heteroscedasticity ◽  
Content Type ◽  
Volatility Models ◽  
Non Linear ◽  
Forecasting Volatility ◽  
Autoregressive Conditional Heteroscedasticity

PurposeSeveral empirical studies have proven that emerging countries are attractive destinations for Foreign Institutional Investors (FIIs) because of high expected returns, weak market efficiency and high growth that make them attractive destination for diversification of funds. But higher expected returns come coupled with high risk arising from political and economic instability. This study aims to compare the linear (symmetric) and non-linear (asymmetric) Generalized Autoregressive Conditional Heteroscedasticity (GARCH) models in forecasting the volatility of top five major emerging countries among E7, that is, China, India, Indonesia, Brazil and Mexico.Design/methodology/approachThe volatility of financial markets of five major emerging countries has been empirically investigated for a period of two decades from January 2000 to December 2019 using univariate volatility models including GARCH 1, 1, Exponential Generalized Autoregressive Conditional Heteroscedasticity (E-GARCH 1, 1) and Threshold Generalized Autoregressive Conditional Heteroscedasticity (T-GARCH-1, 1) models. Further, to examine time-varying volatility, the distinctions of structural break have been captured in view of the global financial crisis of 2008. Thus, the period under the study has been segregated into pre- and post-crisis, that is, January 2001–December 2008 and January 2009–December 2019, respectively.FindingsThe findings indicate that GARCH (1, 1) model is superior to non-linear GARCH models for forecasting volatility because the effect of leverage is insignificant. China has been considered as most volatile, whereas India is volatile but positively skewed and Indonesia is the least volatile country. The results can help investors in better international diversification of their portfolio and identifying best suitable hedging opportunities.Practical implicationsThis study can help investors to construct a more risk-adjusted returns international portfolio. Further, it adds to the scant literature available on the inconclusive debate on the choice of linear versus non-linear models to forecast market volatility.Originality/valueEarlier studies related to univariate volatility models are mostly applications of the models. Only few studies have considered the robustness while applying the models. However, none of the studies to the best of the authors’ searches have considered these models for identifying the diversification opportunity among the emerging countries. Hence, this study is able to derive diversification and hedging opportunities by applying wide ranges of the statistical applications and models, that is, descriptive, correlations and univariate volatility models. It makes the study more rigorous and unique compared to the previous literature.

scholarly journals Modeling the Behavior of Inflation Rate in Albania Using Time Series

2017 ◽  
Vol 13 (3) ◽  
pp. 7257-7263
Author(s):  
Rozana Liko
Keyword(s):  
Time Series ◽  
Serial Correlation ◽  
Inflation Rate ◽  
Garch Models ◽  
Conditional Heteroscedasticity ◽  
One Year ◽  
Autoregressive Conditional Heteroscedasticity

In this paper, time series theory is used to modelling monthly inflation data in Albania during the period from January 2000 to December 2016. The autoregressive conditional heteroscedastic (ARCH) and their extensions, generalized autoregressive conditional heteroscedasticity (GARCH)) models are used to better fit the data. The study reveals that the inflation series is stationary, non-normality and has serial correlation.   Based on minimum AIC and SIC values the best model turn to be GARCH (1, 1) model with mean equation ARMA (2, 1)x(2, 0)12. Based on the selected model one year of inflation is forecasted (from January 2016 to December 2016).

scholarly journals Volatility of cryptocurrencies

Notitia ◽  
2020 ◽  
Vol 6 (1) ◽  
pp. 13-23
Author(s):  
Branimir Cvitko Cicvarić
Keyword(s):  
Time Series ◽  
Financial Time Series ◽  
Information Criterion ◽  
Garch Models ◽  
Arch Model ◽  
Model Estimate ◽  
Conditional Heteroscedasticity ◽  
Financial Time ◽  
Egarch Model ◽  
Autoregressive Conditional Heteroscedasticity

Many models have been developed to model, estimate and forecast financial time series volatility, amongst which are the most popular autoregressive conditional heteroscedasticity (ARCH) model introduced by Engle (1982) and generalized autoregressive conditional heteroscedasticity (GARCH) model introduced by Bollerslev (1986). The aim of this paper is to determine which type of ARCH/GARCH models can fit the best following cryptocurrencies: Ethereum, Neo, Ripple, Litecoin, Dash, Zcash and Dogecoin. It is found that the EGARCH model is the best fitted model for Ethereum, Zcash and Neo, PARCH model is the best fitted model for Ripple, while for Litecoin, Dash and Dogecoin it depends on the selected distribution and information criterion.

scholarly journals MODEL GENERALIZED AUTOREGRESSIVE CONDITIONAL HETEROSCEDASTICITY IN MEAN UNTUK MERAMALKAN VOLATILITAS RETURN SAHAM

2020 ◽  
Vol 9 (1) ◽  
Author(s):  
Syarifah Zela Hafizah, Dadan Kusnandar, Shantika Martha
Keyword(s):  
Time Series ◽  
Garch Model ◽  
Conditional Heteroscedasticity ◽  
Volatility Clustering ◽  
Model Time Series ◽  
Autoregressive Conditional Heteroscedasticity

Volatilitas menunjukkan fluktuasi pergerakan harga saham. Semakin tinggi volatilitas maka semakin tinggi pula kemungkinan mengalami keuntungan dan kerugian. Data time series yang sering memiliki volatilitas yang tinggi adalah data keuangan. Data time series di bidang keuangan sering memiliki sifat volatility clustering atau sering disebut sebagai kasus heteroskedastisitas. Pada umumnya, pemodelan data time series harus memenuhi asumsi varian konstan (homoskedastisitas). Untuk mengatasi masalah heteroskedastisitas, model time series yang dapat digunakan adalah ARCH/GARCH. Model GARCH merupakan pengembangan dari model ARCH yang dapat digunakan untuk menggambarkan sifat dinamik volatilitas dari data. Salah satu bentuk pengembangan dari model GARCH adalah Generalized Autoregressive Conditional Heteroscedasticity in Mean (GARCH-M). Tujuan dari penelitian ini adalah untuk mengimplementasikan model GARCH-M pada peramalan volatilitas return saham. Data yang digunakan dalam penelitian ini adalah return penutupan harga saham mingguan S&P 500 dari September 2013 sampai Juni 2019. Model terbaik yang digunakan untuk peramalan volatilitas pada return harga saham S&P 500 adalah MA (1) GARCH (1,1)-M.Kata Kunci: saham, volatilitas, GARCH-M

scholarly journals RNNLBL : A Recurrent Neural Network and Log Bilinear based Efficient Stock Forecasting Model

2020 ◽  
Vol 9 (4) ◽  
pp. 1676-1682 ◽  
Keyword(s):  
Neural Network ◽  
Time Series ◽  
Recurrent Neural Network ◽  
Stock Price ◽  
Future Price ◽  
Forecasting Model ◽  
Price Forecasting ◽  
Short Term ◽  
Stock Price Forecasting ◽  
Non Linear

Recent years have seen the wide use of Time series forecasting (TSF) for predicting the future price stock, modeling and analyzing of finance time series helps in guiding the trades and investors decision. Moreover considering the stock as the dynamic environment, it is pronounced as the non-linearity of time series which affects the stock price forecast immediately. Hence, in this research work we propose intelligent TSF model, which helps in forecasting the early prediction of stock prices. The proposed stock price forecasting model employed both short-term (i.e. recent behavior fluctuation) using log bilinear (LBL) model and long-term (i.e., historical) behavior using recurrent neural network (RNN) based LSTM (long short term memory )model. Subsequently, this model is mainly helpful for the home brokers since they do not possess enough knowledge about the stock market. Proposed RNNLBL hybrid model shows the satisfying forecasting performance, these results in overall profit for the investors and trades. Furthermore, proposed model possesses a promising forecasting in case of the non-linear time series since the pattern of non-linear pattern are highly improbable to capture through these state-of-art stock price forecasting models.

Sign in / Sign up

Export Citation Format

Share Document