Cycles and Seasonais in Inventories: Another Look At Non-Stationarity and Induced Seasonality

The Interest in Understanding Inventory Behavior Is Great; Yet Econometric Results Are Not Impressive, and the Most Widely Used Theoretical Model Has Been Rejected by Several Studies. the Paper Reconsiders Some Simple But Important Features of the Data. the Findings Do Not Correspond to What Is Commonly Assumed About Inventory Behavior. There Are Two Parts to the Paper. the First Considers the Characteristics of Non-Stationarity for Four Industries. the Following Results Emerge From the Analysis: 1) There Is No Consistency Across the Industries, I.E., Different Specifications of Non-Stationarity Occur for the Same Type of Series Across the Industries: 2) There Is No Consistency Across Series Within the Same Industry, I.E., There Is No Uniform Transformation Which Can Be Applied to Each Series Seperately Which Would Yield Stationarity of Shipments, Inventories Nd Industrial Production Series Within the Same Industry and 3) the Characterization of Non-Stationarity with Seasonally Adjusted Data Is Not Consistent with That of Seasonal Adjustment Unadjusted Data (Provided the Seasonal Is Stationary), I.E., Seasonal Adjustment Procedures May Alter the Outcome of Test Statistics for the Characterization of Non-Stationarity. the Second Part of the Paper Looks At the Seasonality in the Series. Ghysels (1986A), Among Others, Observed That the Identification Assumptions Embodied in Univariate Seasonal Adjustment Procedures Are in Principle Incompatible with the Dynamic Economic Theory. a Reduced Form of a Dynamic Model Is Not Necessarily a Frequency-Preserving Mapping From Exogenous Processes Observation, There As to Endogenous Processes. While This Point Has Always Been a Theoretical Observation, There Has Never Been Empirical Support for It. the Second Part of the Paper Reports Empirical Findings of "Leakage" of Power Across Frequencies and Shows That There May Be Considerable Effects of Power Originating From Seasonal Frequencies of Exogenous Series to Non-Seasonal Frequencies of Endogneous Processes. Throughout the Paper There Is Also a Discussion of the Implications the Results Have for 1) the Econometric Studies of the Production Smoothing Model and 2) How One Should Look for Stylized Facts of Inventory Behavior.