Comparing different economic linkages among commodity futures

Gasoline (GA) and kerosene (KO) are extracted from crude oil (CO), such that the three fuel commodities share a chemical link. On the other hand, GA also shares an industrial link with natural rubber (NR) and palladium (PA) as complementary commodities that are heavily consumed by the automobile industry. We contrast the information content embedded in the two economic linkages. Focusing on TOCOM futures contracts written on the five commodities and centering on GA, we confirm that incremental information provided by either CO, KO or NR, PA over a buy-and-hold strategy and a naive forecast, are both statistically and economically significant. While the chemical link forecast is more profitable, a double-link forecast generated from a VECM with two cointegrating vectors (KO-GA and GANR prices) outperforms both single-link forecasts based on risk-adjusted profit net of transaction costs. Further comparisons against the profitability of commodity-based momentum strategies documented in Erb and Harvey (2006) and Miffre and Rallis (2007) show that the double-link forecast holds its own against the most profitable of the 75 momentum strategies considered. This strongly suggests that not only are there incremental profits to be gained from harnessing and combining economic links among commodity futures, the resultant incremental profits are economically significant against other proven commodity-based trading strategies in the existing literature.

Economic Bias of Weather Forecasting: A Spatial Modeling Approach

The value of accurate weather forecast information is substantial. In this paper we examine competition among forecast providers and its implications for the quality of forecasts. A simple economic model shows that an economic bias geographical inequality in forecast accuracy arises due to the extent of the market. Using the unique data on daily high temperature forecasts for 704 U.S. cities, we find that forecast accuracy increases with population and income. Furthermore, the economic bias gets larger when the day of forecasting is closer to the target day; i.e. when people are more concerned about the quality of forecasts. The results hold even after we control for location-specific heterogeneity and difficulty of forecasting.