Narrow Results

This study attempts to forecast Bitcoin using both traditional and machine learning approaches to determine which methods are more robust. For the traditional method, the GARCH method is used to forecast Bitcoin returns. For the machine learning method, LSTM is used. A hybrid approach of GARCH–LSTM is also applied to the data to compare the results. Hourly data for Bitcoin are obtained from Coin-MarketCap for three years, from 2019 to 2022. The findings of the study reveal that machine learning methods outperform traditional methods. The study has useful implications for researchers.

This paper adopts a new approach to estimating the conditional probability distribution of asset returns. It is evident that the exact conditional mean or variance is inherently unobservable for time series. In practice, the popular way is to derive from different models such as GARCH by assuming distributions such as normal, student t, or skewed t. Thus the accuracy of forecast strongly depends on the assumption of distribution. The new method avoids the need to assume any distribution by using a neural network (NN) to estimate the potentially nonlinear relationship between VaR (Value at Risk) and returns. Our results show that the forecast from neural network outperforms traditional GARCH models.

We estimate an AR(1)/GARCH(1, 1) model that shows the impact of natural-gas prices, hydro conditions, and temperatures on wholesale on-peak electricity prices at the Mid-Columbia (Mid-C) trading hub in the Pacific Northwest of the United States. After controlling for the effects of these three factors, prices are seen to exhibit a weak seasonal pattern, but a strong day-of-week pattern. It is also shown that price spikes can persist for several days. Finally, in support of the GARCH hypothesis, Mid-C prices are seen to have a time-dependent variance that primarily moves with natural-gas prices, and that large price variances tend to persist. Thus, even though buyers might cross hedge using natural-gas futures and temperature-based weather futures, the effectiveness of any hedge is compromised by randomness in hydro conditions. To be sure, a buyer can eliminate the electricity price risk by entering into a forward contract, but only at the expense of what is likely to be a large risk premium embodied in the forward price.

We investigate the impact of scheduled government announcements relating to six different macroeconomic variables on the risk and return of three major U.S. financial markets. Our results suggest that these markets do not respond in any meaningful way, to the act of releasing information by the government. Rather, it is the ‘news’ content of these announcements which cause the market to react. For the three markets tested, unexpected balance of trade news was found to have the greatest impact on the mean return in the foreign exchange market. In the bond market, news related to the internal economy was found to be important. For the U.S. stock market, consumer and producer price information was found to be important. Finally, financial market volatility was found to have increased in response to some classes of announcement and fallen for others. In part, this result can be explained by differential “policy feedback” effects.

This chapter is concerned with the statistical behavior of energy commodity prices. A particularly salient feature of many commodity markets is the unexpectedly rapid changes — or jumps — that result from the arrival of new information. Such a process would contradict the view that energy commodity prices follow a geometric Brownian motion (GBM) process (i.e. log returns are normally distributed). That is, assuming a GBM process for the data-generating mechanism would be insufficient to capture the true dynamics of energy commodity markets. The discontinuous arrival of information necessitates a stochastic process that incorporates this feature, and as such, Jump processes have become an important tool in the analysis of energy markets. While such models allow for multiple jumps in a period, the jump intensity is assumed to be constant over time — a questionable assumption given the dynamics of such energy markets. The autoregressive conditional jump intensity (ARJI) model of Chan and Maheu [2002],which allows for a time-varying jump intensity, is applied to important energy commodity markets. The results indicate the importance of incorporating time-varying jump intensities in energy markets.

Energy-based assets are showing increased susceptibility to volatility arising out of geo-political, economic, climate and technological events. Given the economic importance of energy products, their market participants need to be able to access efficient strategies to effectively manage their exposures and reduce price risk. This chapter will outline the key futures-based hedging approaches that have been developed for managing energy price risk and evaluate their effectiveness. A key element of this analysis will be the breadth of assets considered. These include Crude and Refined Oil products, Natural Gas and wholesale Electricity markets. We find significant differences in the hedging effectiveness of the different energy markets. A key finding is that, Natural Gas and particularly Electricity futures are relatively ineffective as a risk management tool when compared with other energy assets.

The landscape of financial econometrics was forever changed and augmented greatly when Professor Engle introduced the Autoregressive Conditonal Heteroskedasticity model in 1982. It was an ingeniously embedded tool in specifying the dynamics of volatility coupled with the underlying asset price process. This greatly extended the space of stochastic processes including those in the Box Jenkins approach. ARCH was very successfully extended to GARCH by Prof. Bollerslev, and to-date there continues to be a huge number of variations of processes building on the embedded tooling idea. We consider such process modelling to be particularly relevant in estimating risks in market prices and in risk management in today's market of high and persistent volatilities. The topic of maximum likelihood is given more discussion here, as well as ideas of asymptotic efficiency.

The aim of this chapter is to shed new light on hedging discrete volatilities, in particular when using the generalized autoregressive conditional heteroscedasticity (thereafter GARCH) model. Despite its elegance, GARCH does not account for (i) correlation coefficients of debt and equity, (ii) equity parameter, (iii) risk premium, (iv) interest rates, and (v) shocks-stock markets. The unaccounted listed parameters are included into the GARCH(1,1) and the chapter inverts a new model; expanded GARCH; called the capitalized GARCH. The results show that the capitalized GARCH convergences in a similar manner to the GARCH(1,1) in modeling volatility of bonds, commodities, equities, and real estate indices.

It is shown empirically that the likelihood function of the GARCH is multimodal. Hence, the maximum likelihood estimates at local and global maxima will be quantitatively different. Therefore, it is important to start an estimation method with a consistent starting value that converges to global maxima. This study compares two estimation methods, BFGS and DE, on the basis of simulation and the surface is constructed by changing the value of GARCH (1, 1) model. DE is superior and consistent throughout the surface and across distributions. PSX is used as a real-world application and found that the estimates obtained from DE are best and unbiased.