Inference for Econometric Modeling in Antidumping, Countervailing Duty and Safeguard Investigations

This paper examines the importance of taking into account standard errors when making inferences from econometric models prepared for antidumping, countervailing duty, and safeguard investigations. This is important since even a properly specified econometric model is only as useful as the inferences that can be made from its results. Analysis of these models typically entail testing whether point estimates are significantly different from zero and then drawing inferences from these point estimates without accounting for their distribution. This paper suggests taking the distribution of the estimates into account by drawing inferences from confidence intervals at various levels of significance and accounting for the impact of sample size and collinearity between imports subject to an investigation and other possible factors that may be affecting the domestic industry. Accounting for the distribution of the estimates provides much more compelling inferences regarding the impact of imports, although these inferences are often more nuanced than those made solely from point estimates