Factors Influencing the Gulf and Pacific Northwest (PNW) Soybean Export Basis: An Exploratory Statistical Analysis

Growth in the export marketing of soybeans has drawn attention to the basis volatility in these market channels. Indeed, there has been greater growth in soybean exports compared to other commodities and this is due in part to the growth of exports to China. Concurrently, there has been substantial volatility in the basis at the primary U.S. export locations: the U.S. Gulf and the Pacific Northwest (PNW). This variability is caused by traditional variables affecting the basis but is also influenced by shipping costs, international competition, and inter-port relationships. Further, there seems to be distinct seasonal patterns that vary across marketing years. The purpose of this study is to examine the impact of supply/demand, export competition and logistical variables on both the average level and seasonality of U.S. export basis values for the 2004/05 through 2015/16 marketing years (September through August for U.S. soybeans). This study examines the impact of a wide range of supply, demand, transportation, and other market variables upon both the average level and seasonality (by marketing year) of the basis at the two major U.S. export locations, Gulf and Pacific Northwest (PNW). The explanatory dataset contains more variables (27) than observations (12 marketing years from 1994/95 through 2015/16); therefore, it presents challenges from both a sparsity and a multicollinearity perspective. To address these issues, a statistical regression technique, called partial least squares (PLS) is utilized. This technique has advantages over using principal components regression (PCR) since derivation of the components is directed towards maximizing the covariance between the dependent (Y) and explanatory (X) variable sets rather than just explaining the variance of X. Seasonality is investigated in this study utilizing agglomerative hierarchal clustering (AHC) to group similar marketing years by seasonal pattern called seasonal analogs. These seasonal analogs were then related to the explanatory variable set using a two-sample statistical test (Lebart, Morineau and Piron 2000) that compares the means of a subset and its parent set to explain the impact of the explanatory variables. The results indicate that the average market year level of the basis is primarily influenced by export competition from Brazil and export demand – particularly from China; however, domestic demand (soybean crush) also has some influence. Rail transportation costs to both the Gulf and PNW have an influence on the basis level; however, barge and ocean freight rates appear to not have a significant influence on the level of the basis. Application of AHC resulted in the identification of 5 and 4 distinct analogs (over the 12 marketing years in the dataset) for the Gulf and PNW respectively. Application of the two-sample mean difference tests to the analogs indicate that the seasonal pattern of the export basis is more heavily influenced by internal logistical conditions (late railcar placement and secondary railcar values), pace of farmer marketings, transportation cost differentials (between ports), and individual port export activity (ships in port and export inspections) rather than international and domestic demand.