36 Read more

The no-change forecast tends to be relatively accurate for horizons up to twelve months. This is consistent with existing literature and the view among agents in the oil **market**. As stated in the introduction, Hamilton (2008) suggests that oil prices are completely unpredictable. Although presenting different results, the no-change model being favoured by Alquist and Kilian (2010), Cher- nenko et al. (2004) favouring the **futures** model, and the spread regression being favoured by Mc- Callum and Wu (2005), these studies do not contradict each other nor the findings above. The divergence in forecast accuracy can mainly be explained by differences in the sample period. The estimates presented in this section are based on a larger sample size compared to the mentioned studies. Oil price predictability is also affected by a range of factors and the tie to the spot price of oil is, to put it mildly, complicated (Zagaglia, 2010). Thus, the connection between the oil price and macroeconomic determinants is important to bear in mind.

Show more
38 Read more

CBOE introduced its current VIX methodology on September 22, 2003 (The CBOE **volatility** index - VIX, 2003) to fulfill the above requirements and was based on (Demeterfi, Derman, Kamal, & Zou, 1999b, 1999a), where a closed-form formula for the expected value of RV (Barndorff-Nielsen & Shephard, 2002) was derived using call and put prices. Notably, it utilized the S&P 500 index, which is far more representative of the total **market**, both near-term and next-term options and a broader range of strike prices. CBOE publishes historic data using both methodologies, VIX (new) and VXO (old) dating back to 1990 (VIX Options and **Futures** Historical Data, n.d.), (historic stock prices used in calculation of RV can be found at (S&P500, n.d.)). Here we call 1990 through September 19, 2003 VIX Archive and VXO Archive and from September 22, 2003 through December 30, 2016 VIX Current and VXO Current.

Show more
27 Read more

Beine et al. (2007) studied the relation between central bank intervention and the **volatility** of two major exchange rates. They applied the bipower variation to decompose the **realized** **volatility** into its continuous and jump components, deducing that interventions trigger considerable jumps. Fuentes et al. (2009), meanwhile, compared four estimators— namely **realized** **volatility**, **realized** range, **realized** power variation, and **realized** bipower variation—by investigating their in-sample distributional pattern and an out-of-sample forecast. Their analysis used a seven-year sample of prices for 14 stocks listed on the NYSE. The forecast was then generated with a GARCH framework. The authors then concluded that the combination of all four intraday measures gave the lowest forecast errors in about half the sampled stocks. Bollerslev et al. (2009) applied nonparametric **realized** variation and bipower variation measures constructed from high- frequency data with the aim of developing a discrete-time daily stochastic **volatility** model that could distinguish between the jump and continuous components of return movements. They suggested that the model allows for the consideration of structural inter-dependencies between shocks to returns and **volatility** components. Andersen et al. (2011) also applied a **volatility** decomposition method based on long samples of high-frequency data for equity and bond **futures** returns. Their results suggested that dynamic dependencies and variability in the continuous element can be well described by an approximate long-memory HAR-GARCH model. In addition, the dynamic dependencies in the identified significant jumps seemed to be well expressed by the ACH model with a simple log-linear structure for the jump sizes. The authors highlighted the superior forecasting performance of the model that considered both components of **volatility** when compared to other commonly used models. In order to take into account the impact of jumps, Barunik et al. (2016) applied a GARCH forecasting model with decomposed **realized** **volatility** measurements. They therefore decomposed **volatility** into several timescales, thus approximating the behavior of traders at corresponding investment horizons. They then compared forecasts by employing some current **realized** **volatility** measures for FOREX **futures** data for the recent financial crisis. Their results indicated that separating jump variation from the integrated variation improved forecasting performance.

Show more
15 Read more

and efficient forecast of future **realized** **volatility**, and it outperformed historical **volatility** in forecasting. They claimed that the results were different because they used longer time series and non-overlapping data with lower sampling frequency. Their method of data collection was different from that of Canina and Figlewski’s study in 1993, which used daily option trading data in a shorter time period. Christensen and Prabhala also suggested that the option trading data before the October 1987 **market** crash were different from the data after. It might help explain why implied **volatility** was biased in previous work.

Show more
32 Read more

created new financial risks for emitting firms. To deal with these risks, options are traded since October 2006. Because the EU ETS is a new **market**, the relevant underlying model for option pricing is still a controversial issue. This article improves our understanding of this issue by characterizing the conditional and unconditional distributions of the **realized** **volatility** for the 2008 **futures** contract in the European Climate Exchange (ECX), which is valid during Phase II (2008-2012) of the EU ETS. The **realized** **volatility** measures from naive, kernel-based and subsampling estimators are used to obtain inferences about the distributional and dynamic properties of the ECX emissions **futures** **volatility**. The distribution of the daily **realized** **volatility** in logarithmic form is shown to be close to normal. The mixture-of-distributions hypothesis is strongly rejected, as the returns standardized using daily measures of **volatility** clearly departs from normality. A simplified HAR-RV model (Corsi, 2009) with only a weekly component, which reproduces long memory properties of the series, is then used to model the **volatility** dynamics. Finally, the predictive accuracy of the HAR-RV model is tested against GARCH specifications using one-step-ahead forecasts, which confirms the HAR-RV superior ability. Our conclusions indicate that (i) the standard Brownian motion is not an adequate tool for option pricing in the EU ETS, and (ii) a jump component should be included in the stochastic process to price options, thus providing more efficient tools for risk-management activities.

Show more
34 Read more

The contributions of this article are several. First, this work is pioneer in using ANN methods to model **agricultural** **futures** prices **volatility**. Corn **futures** **volatility** is characterized by strong nonlinearities that are related to: U.S. Department of Agriculture (USDA) report release days (Adjemian and Irwin 2018); strong seasonality (Karali and Power 2013; Egelkraut, Garcia and Sherrick 2007); time-to-delivery effects (Goodwin and Schnepf 2000); and changes in macroeconomic conditions (Karali and Power 2013). Combined, these effects create nonlinearities of unknown forms that require highly flexible and adaptive methods such as ANN models. Second, by using a HAR framework, we allow for the heterogeneous effects of different **market** participants on the **volatility** which reflects shorter- and longer-term trading activity. Third, we also conduct multi-step horizons forecasting. While forecasting corn **volatility** using option prices, previous research has traditionally used a 2-month horizon due to options expirations (e.g. Wu et al. 2015), other **market** participants may rely on medium- or short-term forecasting to adjust their risk strategies. For instance, traders relying on USDA weekly export sales reports 24 (such as shippers) or weekly corn options traders 25 will be more interested in a shorter-term forecasting of the corn **futures** prices **volatility**. **Futures** markets **volatility** forecasts are also crucial in the pricing of the options, a higher **volatility** will result in a higher priced options (Kroner, Kneafsey and Claessens 1995). On the other hand, high-speed liquidity providers might rely on short-run **volatility** forecasts to decide how to adjust their strategies for the next day. As a result, our findings should be useful to a wider range of **market** participants.

Show more
149 Read more

Shanker (2017) estimated the indices of adequate and excess speculation for 21 different **futures** contracts for the period 1986-2015. She used the indices of adequate and excess speculation to investigate the relationship between speculation and **volatility** of the NYMEX’s West Texas Intermediate (WTI) crude oil **futures** contract over the period 31 January 1986 till 29 December 2015, while controlling for **market** fundamental risk. In order to estimate the unobservable variables -- balancing hedging and balancing speculative contracts, which are necessary to estimate the indices, she applied a Kalman (1960) filter approach with inequality constraints imposed on the state variables, which are the time-varying intercept and slope of the true linear relationship between the speculative and hedging ratio for each contract. Wang (2001) also developed a sentiment index based on COT positions in six actively traded **agricultural** **futures** markets. His sentiment index, 𝑆𝐼 𝑖,𝑡 is also calculated for each trader type, 𝑖 (non-commercial, commercial and non-reporting), based on current aggregate positions and historical extreme values over the previous three years. Wang defined the sentiment index of trader type 𝑖 in **market** 𝑗 at week 𝑡 as:

Show more
74 Read more

The aims of the paper are to analyse **agricultural** commodity **futures** returns using several conditional **volatility** models, namely GARCH, EGARCH and APARCH, and fractionally integrated conditional **volatility** models, namely FIGARCH, FIEGARCH and FIAPARCH, as an extension of existing results. The paper differs from existing studies in three respects. First, due to changes in the financial and economic environment, such as the 2008-09 global financial crisis, an increasing number of **market** participants, product yield uncertainty, changes in the demand and supply position of **agricultural** commodities and growing international competition, **agricultural** commodity **futures** markets have matured considerably over the last decade. An extension in the sample period from 2000 to 2009, giving an additional 2,500 observations, is intended to allow a suitable analysis of these issues.

Show more
34 Read more

More specifically, Tian et al. (2017a) use a two regime-switching Markov models to forecast **realized** **volatility** for five **agricultural** commodities traded in the Chinese mar- ket, namely, Soybean, Soybean oil, White Sugar, Gluten Wheat and Cotton. They find evidence that regime switching dynamics offer predictive gains compared to both a sim- ple AR(1) and a Markov-Switching AR(1) model. Yang et al. (2017) also use intra-day data from the Chinese commodity **futures** markets (Zhenzhou Commodity Exchange and Dalian Commodity Exchange) of Soybean, Cotton, Gluten Wheat and Corn **futures** prices and employ a similar strategy with Tian et al. (2017b), where the HAR model is extended with potential predictors (such as day-of-the-week dummies, past cumulative returns and the jump component) and forecasts are generated based on bagging and combination methods. Their conclusions suggest the forecasts based on the HAR models with bagging and principal component combination methods are able to outperform the AR model. Finally, Tian et al. (2017b) use Soybean, Cotton, Gluten Wheat, Corn, Early Indica Rice and Palm **futures** prices, traded in the Chinese **market**, to construct and forecast their **realized** **volatility** measure. Furthermore, the authors use several other **realized** **volatility** measures (such as daily log-range **volatility**, **realized** threshold multi-power variation and the **realized** threshold bi-power variation) and the jump component, as potential predictors of the **realized** **volatility**. Their predictive models allow both predictors and coefficients to vary over time. Their findings show that the Dynamic Model Average and Bayesian Model Average models are able to exhibit superior predictive ability, relatively to the simple HAR model. More importantly, they show that the HAR model with time-varying sparsity produces the most accurate forecasts for all the chosen commodities.

Show more
49 Read more

The influence of derivatives on the commodity markets was also examined through the such- called ‘herding phenomenon’. In 1990, Pindick and Rotenberg propose to define herding by a situation where traders are alternatively bullish or bearish on all commodities for no plausible economic reasons. According to these authors, herding is a possible explanation for the excess co-movement that they observe on the prices of seven different commodities. Pindick and Rotenberg show first of all that the prices of raw commodities have a persistent tendency to move together. They try to explain this co-movement by macro-economic variables: the expected inflation, the growth in industrial production, the consumer price index, several exchange rates, interest rates, money supply, and the S&P common stock index. However, they find that the co-movement is well in excess of anything that could be explained by these common variables, and they conclude that herding could explain this excess. Booth and Cinner (2001) extend this work. They analyse the linkage among **agricultural** commodity **futures** prices on the Tokyo Grain Exchange, and they suggest two explanations of long- term co-movements among the prices of **agricultural** **futures** contracts: common economic fundamentals or herd behaviour by **market** participants. Contrary to Pindick and Rotenberg, their results support the common economic fundamentals assumption.

Show more
39 Read more

To keep our discussion concise we have left out some important issues that can be explored in future work. We select two examples. First, in practice advances in **realized** **volatility** model- ing may not be translated so neatly into improvements in modeling the conditional distribution of returns. Two aspects of the link between **realized** **volatility** and returns should be studied more carefully. The assumption that returns standardized by **realized** **volatility** are approximately normal and independent seems to be inadequate for some series. Is there a role for jumps in ad- justing the distribution? Do the problems in measuring **realized** **volatility** make this relation less straightforward? We have also only considered a simple model for the dependence between return and **volatility** innovations. Second, we have mostly analyzed the performance of different models in one day ahead applications. Because financial quantities are so persistent many incongruent models are misleadingly competitive at very short horizons. More emphasis should be placed in investigating whether different models are consistent with a realistic longer horizon dynamics. Our analysis suggest that to do so we may need a more solid understanding of asymmetric effects.

Show more
33 Read more

Recently, Andersen et al. (2005) suggested modeling and forecasting the **realized** **volatility** of exchange rate, stock and bond returns by extracting the component due to jumps and including it as an explanatory variable in a HAR-RV regression model of Müller et al. (1997) and Corsi (2003). In some cases, the jump component turned out to be highly signi…cant and considerable increases in the coe¢ cient of determination were observed. This suggests that gains in forecasting the **realized** **volatility** could be made by separately modeling and forecasting the jump and continuous sample path components and obtaining forecasts of the **realized** **volatility** as their sum instead of considering the aggregate **realized** **volatility**, as conjectured by Andersen et al. (2005). The purpose of this paper is to study whether such an approach really would be bene…cial and whether the potential gains in forecast accuracy depend on the way the decomposition is made. To this end, we examine the returns of the Euro against the U.S. Dollar. To model the **realized** **volatility** and the continuous components, the mixture-MEM model previously shown to …t well to comparable exchange rate data by Lanne (2006) is employed. The jump components are modeled by means of standard Markov-switching models.

Show more
26 Read more

Our results reveal that Canadian stock **market** experiences signi�cant price discontinues (jumps) and average jump arrival rate is about 0.17 jumps per day. We �nd that about 55% of jumps are due to the overnight returns and about 90% of jumps occur within 30 minutes of the **market** opening for trading – providing a strong evidence of jump clustering. While looking at the jump intensi- ties, our results show an asymmetric distribution of positive versus negative jumps for intraday returns but such asymmetry disappears when we include overnight returns in our analysis. Berk- man et al. (2012) and Lou et al. (2018) suggest that institutional investors tend to trade relatively more during the day and individual investors trade relatively more overnight. Such di�erences in jump charecteristics in intraday versus overnight returns potentially re�ecting the corresponding clientele e�ects. Therefore, it is important to incorporate overnight returns in jump risk analysis. 1 In our paper, we further show that although the e�ect of jump component in **volatility** fore- casting is statistically signi�cant, its economic signi�cance is very nominal - large portion of re- alized **volatility** is coming from the continuous component. When we examine e�ect of **market**

Show more
36 Read more

To keep our discussion concise we have left out some important issues that can be explored in future work. We select two examples. First, in practice advances in **realized** **volatility** model- ing may not be translated so neatly into improvements in modeling the conditional distribution of returns. Two aspects of the link between **realized** **volatility** and returns should be studied more carefully. The assumption that returns standardized by **realized** **volatility** are approximately normal and independent seems to be inadequate for some series. Is there a role for jumps in ad- justing the distribution? Do the problems in measuring **realized** **volatility** make this relation less straightforward? We have also only considered a simple model for the dependence between return and **volatility** innovations. Second, we have mostly analyzed the performance of different models in one day ahead applications. Because financial quantities are so persistent many incongruent models are misleadingly competitive at very short horizons. More emphasis should be placed in investigating whether different models are consistent with a realistic longer horizon dynamics. Our analysis suggest that to do so we may need a more solid understanding of asymmetric effects.

Show more
34 Read more

This paper estimates the long memory **volatility** model for 16 **agricultural** commodity **futures** returns from different **futures** markets, namely corn, oats, soybeans, soybean meal, soybean oil, wheat, live cattle, cattle feeder, pork, cocoa, coffee, cotton, orange juice, Kansas City wheat, rubber, and palm oil. The class of fractional GARCH models, namely the FIGARCH model of Baillie et al. (1996), FIEGACH model of Bollerslev and Mikkelsen (1996), and FIAPARCH model of Tse (1998), are modelled and compared with the GARCH model of Bollerslev (1986), EGARCH model of Nelson (1991), and APARCH model of Ding et al. (1993). The estimated d parameters, indicating long-term dependence, suggest that fractional integration is found in most of **agricultural** commodity **futures** returns series. In addition, the FIGARCH (1,d,1) and FIEGARCH(1,d,1) models are found to outperform their GARCH(1,1) and EGARCH(1,1) counterparts.

Show more
34 Read more

This section investigates a relevant framework of the conditional variance model through comparison among different specifications. The univariate conditional **volatility** model, namely the GARCH model of Bollerslev (1986), EGARCH model of Nelson (1991) and APARCH model of Ding et al. (1993), and fractionally integrated class of models, namely the FIGARCH model of Baillie et al. (1996), FIGARCH model of Chung (1999), FIEGARCH model of Bollerslev and Mikkelsen (1996), FIAPARCH model of Tse (1998), and FIAPARCH model of Chung (1996) with Gaussian errors, are estimated by QMLE, which allows for asymptotically valid inferences when the standardized innovations are not normally distributed. Corresponding estimates are obtained using the BFGS algorithm. The computations are performed using the Ox/G@RCH 4.2 econometrics software package of Laurent and Peters (2006).

Show more
35 Read more

Several previous papers have observed and provided application of fractional integrated models in many fields, namely stock returns (Bollerslev and Mikkelsen (1996), Degiannakis (2004) and Niguez (2007), Lux and Kaizoji (2007), Kang and Yoon (2007), Jefferis and Thupayagale (2008), Ruiz and Veiga (2008)); exchange rate (Baillie et al. (1996), Davidson (2004), and Conrad and Lamla (2007)) and inflation rate (Baillie et al. (2002)). However, in the literature to date, there have been few applications of the fractionally integrated GARCH class models to commodity **futures** markets. Barkoulas et al. (1997) examined the fractional structure of commodities spot prices, namely aluminum, cocoa, coffee, copper, rice and rubber. They found that some commodity spot price time series display a fractional structure, and the fractional orders vary among these commodities because the processes involved in the price movements of each commodity varies.

Show more
34 Read more

With the recent availability of high-frequency financial data the long range dependence of **volatility** regained researchers’ interest and has lead to the consideration of long memory models for **realized** **volatility**. The long range diagnosis of **volatility**, however, is usually stated for long sample periods, while for small sample sizes, such as e.g. one year, the **volatility** dynamics appears to be better described by short-memory processes. The ensemble of these seemingly contradictory phenomena point towards short memory models of **volatility** with nonstationarities, such as structural breaks or regime switches, that spuriously generate a long memory pattern (see e.g. Diebold and Inoue, 2001; Mikosch and St˘ aric˘ a, 2004b). In this paper we adopt this view on the dependence structure of **volatility** and propose a localized procedure for modeling **realized** **volatility**. That is at each point in time we determine a past interval over which **volatility** is approximated by a local linear process. Using S&P500 data we find that our local approach outperforms long memory type models in terms of predictability.

Show more
36 Read more

probabilistic model) rather than theory of statistical inference (where we utilize observed data to infer on the unknown data generating process). We note that Gonçalves and Meddahi (2009) adopt a similar conditioning strategy to derive the second-order properties of the **realized** **volatility** statistic. Section 3.1 discusses some technical issues to derive the unconditional version of our large deviation result. Section 3.2 extends the baseline model to allow some specific form of leverage effects and derive analogous large and moderate deviation results. Our large deviation analysis can be considered as a starting point to derive more general properties (e.g., the unconditional large deviation theorems) or to compare the **realized** **volatility** statistic with other estimators for the integrated **volatility**, such as the ones by Barndorff-Nielsen et al. (2008), Christensen, Oomen and Podokskij (2010), and Zhang, Mykland and Aït-Sahalia (2005a).

Show more
38 Read more