Abstract

This thesis consists of three essays. Essay one adds to Chinn and Coibion (2014) by suggesting an alternative measure to investigate the informational content of futures. The paper utilizes the causality measure introduced by Dufour and Renault (1998) and Dufour and Taamouti (2010) to investigate the informational content of the futures’ basis in predicting the price changes. Fourteen commodities and their corresponding prices are examined in three different time horizons and for each, the causality measure and its percentile bootstrap confidence interval are obtained. Results show that in general for energy and agricultural products, the contemporaneous basis of the futures has information to predict the price change in the market, whereas base and precious metals fail in this respect.
Essay two designs a network of realized volatility for the equity market consisted of firms and their customers collected from Compustat customer segment data covering from 1980 to 2017. In the same spirit as Herskovic (2018), two factors are derived from the network. The first is concentration which portrays the node characteristics of the network; whereas the second, sparsity, describes the evolution of the edges. Clustering is used to group the weekly volatility series into nodes and then calculate the concentration factor as the negative entropy of market shares. To complete the network, the causality measure of Dufour and Taamouti (2010) is employed to acquire the spillovers and sparsity. Factors are shown to reduce average pricing errors significantly in the APT setup. To explore further the asset pricing implications, sorted portfolios are created. In tercile portfolios, assets with low-concentration-beta have a 2.15 percent higher return than their counterpart in high-concentration beta annually. For sparsity, the spread between high and low is 0.83 percent. While both high-minus-low investment strategies for two factors show significant returns, results for concentration are more robust to various control variables.
Finally, essay three accommodates the concentration and sparsity factors previously used in analyzing the equity market to address the network of 25 commodity return volatilities for 2000-2017. Similar to previous studies, I document clustering among commodities with a noticeable difference compared to industry categorization. I investigate four commodity-based factors by creating portfolios of futures assets to investigate concentration and sparsity along with the hedging pressure and the basis. The spreads in all long-short strategies are insignificant. Following the law of one price and market integration, I estimate the price of risk of a wide set of equity-based and commodity-based factors. None of the risk premia are significant which appears to be in line with segmented features of the commodity market. Time-series regressions of betas for each commodity and each factor highlights the heterogeneous nature of commodities.