| dc.description.abstract | This thesis uses monthly time series data on producer’s price index of almonds, pecans, walnuts and peanuts in the United States. Main questions this thesis addresses are: Are almonds, peanuts, pecans, and walnuts markets integrated? What is a causal structure between observed products? To answer these questions, we use two methods. First, develops a vector autoregressive model to discover relationship between observed products. Second, uses directed acyclic graphs to understand how innovations from each market are conveyed to other markets in contemporaneous time and to find current market pattern from raw data.
Results provided in this thesis may benefit producers by explaining price fluctuations and their possible future values. The following information is a summary of our study. Our result of causal graph from difference of producer price index unfolds the following way. Peanut is a leader and walnut is a follower. Almond and pecan are in between where peanut causes pecan and from pecan causal arrow goes to almond with a further route to walnut. Causal model from vector autoregressive model indicated that almond does not interact with others while new information from walnut will affect peanut negatively and new information from pecan will positively affect peanut. Our findings from forecast variance error decomposition from vector autoregressive model showed us that observed products explain a small percentage of errors in a range from one to six percent meaning that influence between them is insignificant. We also compared forecast values of vector autoregressive model with other simpler model such as Naïve, Autoregressive Integrated Moving Average, Seasonal Naïve and Exponential Smoothing. | |
