What Should be Taken into Consideration when Forecasting Oil Implied Volatility Index?

• Date: 01 September 2023
• Author: Delis, Panagiotis

1. INTRODUCTION

   In recent years, many studies have focused on oil volatility due to the fact that oil price fluctuations can have large impact, not only on the global economy, but also on the stock market because of its financialization. Moreover, they focus on crude oil since it is considered one of the major inputs in the economy and more specifically because oil volatility could reflect sudden changes in global economic growth. Ferderer (1996) concludes that oil price volatility helps to forecast aggregate output movements in the U.S. Additionally, he provides empirical evidence that there is strong relationship between oil price changes and output growth, which can be explained by the economy's response to oil price volatility. This strong relationship is also found by more recent studies (Elder and Serletis, 2010; Pinno and Serletis, 2013), which have found a significant effect of oil price volatility on aggregate output. Volatility, in general, has triggered the attention of investors, portfolio managers and policy makers. The various stakeholders are interested in implied volatility indices since they help them estimate the cost they are willing to pay in order to proceed to optimal portfolio decisions. So, we have been motivated to study the ability to predict oil implied volatility.

   Recent studies that focus on crude oil volatility forecasting use realized volatility as a volatility measure. There are also studies focusing on other volatility measures, such as stochastic volatility (Ftiti and Jawadi, 2019). Nevertheless, IV indices are considered in many studies' as accurate predictors of future volatility and have been known to provide information about investors' sentiment. Therefore, in this study, we aim to develop a forecasting methodology on the most informative volatility measure, which is also looking-forward by nature, for the crude oil market. We then investigate the dynamics of the oil volatility index (OVX), which measures the market's expectation of the 30-day volatility of crude oil prices extracting information from the options on the United States Oil Fund (USO) for a wide range of strike prices. (2)

   To the best of our knowledge, there are a limited number of studies that implement modeling frameworks in order to directly forecast IV indices, which has attracted the attention of investors, especially during volatile periods. For example, Degiannakis (2008) uses intraday data and conditional volatility estimates to forecast the VIX index. (1) He draws the conclusion that the entire predictive information is provided by VIX itself and that neither interday nor intraday volatility estimates offer incremental forecasting ability on forecasts of VIX. This leads to the conclusion that IV indices are not highly connected to volatility of the underlying asset. Furthermore, Dunis et al. (2013) investigate the predictability of intraday EUR-USD IV by exploiting intraday seasonalities such as overnight effects and they find that IV can be useful in predictions for shorter horizons, within a given day. In this paper, the main objective is to investigate whether more sophisticated modeling frameworks can offer forecasting gains in OVX, since our attention focuses on crude oil, and on which characteristics should be considered. The fact that using such modeling frameworks to predict IV index is one of the main contributions of this paper to the literature. Moreover, focusing on the crude oil market, because of its importance to the global economic outlook, makes this study even more appealing.

   Methodologically, in recent years, various models have been implemented in order to gencrate IV forecasts. A well-known model that has been used for this purpose is the Fractionally Integrated AutoRegressive Moving Average (ARFIMA) model, which captures the property of long memory in the IV series. In addition to the ARFIMA model, the HAR model, proposed by Corsi (2009). has been implemented by Fernandes et al. (2014), who provide evidence that some of the exogenous variables containing mainly information from the S&P500 index seem to have large impact on VIX. However, as per the out-of-sample forecasting results, they conclude that the simple HAR model performs really well and it is difficult even for complicated model specifications to surpass it. The limited literature on forecasting IV indices consists of studies that have implemented not only univariate but also multivariate models. For instance, Konstantinidi et al. (2008) study the impact of certain economic variables on forecasting European and U.S. IV indices under a univariate modeling framework. In the latter study, VAR models are also implemented, which outperform the competing models.

   Regarding the connectedness of oil and other markets, in the case of IV indices, Awartani et al. (2016) provide evidence that volatility transmission from oil to equities is significant and, in general, the transmission from oil to other markets has increased, since the oil price collapse in July 2014. (4) However, the volatility transmission from all the other markets to oil is not so strong. According to Chatziantoniou et al. (2020), there is a transmission of shocks from OVX to VIX but the spillover effects between OVX and VIX do not contain significant predictive information when generating out-of-sample forecasts of OVX. In contrast, Liu et al. (2020) find that there is significant bidirectional IV spillover between the crude oil and stock markets and they draw the conclusion that there is significant positive time-varying correlation between oil and stock IV indices. Moreover, Liu et al. (2013) investigate the short- and long-run cross-market volatility transmission implied by OVX and other important volatility indices in an in-sample analysis. The results indicate that the volatility transmission between the oil market and other markets is short-lived. This paper takes into consideration the above-mentioned results and fills the gap in the literature by investigating the added value of other markets' IV indices (e.g., foreign exchange and commodity markets) in forecasting OVX. This is the first study shedding light on which IV indices enhance the forecasting accuracy of OVX.

   All these studies lead us to the conclusion that forecasting OVX is a topic that needs in depth investigation. One of the contributions of this paper is the extensive investigation of the features, including predictive factors and time series characteristics of OVX, that are important for generating out-of-sample forecasts of OVX. In this regard, we suggest that academics and forecasters should take into account the strong existence of long memory in the time series of IV indices, and more particularly of OVX, which justifies the use of the HAR model. The long memory feature is prerequisite for the use of the HAR model framework. Hence, the investigation of the long memory as well as the incorporation of the DMA estimation technique make the contribution of the paper significant for both academics and investors.

   The second contribution of this paper is to provide a detailed answer concerning which of the IV indices from other asset classes enhance the accuracy of the OVX forecasts. The findings coming out of this research question's investigation can be useful for future studies, since there is no study focusing on the predictive factors of OVX. Relying on the findings of this study, the IV indices from other asset classes offer incrementally higher forecasting accuracy of crude oil volatility, which is a useful outcome for future research. Moreover, the evaluation of the OVX forecasts consists of both statistical loss functions and options straddle trading strategy. The results show that the inclusion of the Dow Jones Industrial Average Volatility Index in the modeling framework is considered of major significance and enhances the predictive ability of the models implemented on short- and mid-run forecasting horizons. This is in line with the literature that finds a strong relationship of oil price volatility and stock markets. More specifically, from this finding, we draw the conclusion that the implied volatility reflecting the future uncertainty of the largest companies' stocks prices is highly significant for generating accurate OVX forecasts. For longer horizons, the Energy Sector ETF Volatility Index appears to have predictive information on OVX, which is explained by the fact that both focus on energy related uncertainty. Moreover, we have concluded that the predictive ability of the HAR models is statistically significant in short- and mid-run forecasting horizons.

   The remainder of the paper consists of the following sections. Section 2 gives a detailed description of the IV indices and provides further information about the dataset used. Moreover, it covers the section in which the long memory of OVX has been tested. In Section 3 the modeling framework has been presented maintaining both the estimation and forecasting frameworks. In Section 4 the evaluation of the generated forecasts has been analytically described, and in Section 5 we provide the results of the statistical and economic evaluation frameworks. Finally, Section 6 concludes the study.

2. IMPLIED VOLATILITY INDICES

   2.1 Data description

   It is vital to start with a description of how IV indices are calculated. This will be useful for differentiating the different volatility measures, namely realized volatility and IV As we have already mentioned, realized volatility reflects the actual volatility of an underlying asset in contrast with IV. which is considered by many studies as a better prediction of future volatility.

   In this study, we focus on the estimate of the expected 30-day volatility of crude oil as priced by USO. which is called OVX. The Chicago Board of Options Exchange (CBOE) methodology, which has been applied to the OVX, uses options on USO, an ETF that is designed to track the price of West...