Online Program Home
My Program

Abstract Details

Activity Number: 437
Type: Contributed
Date/Time: Tuesday, August 2, 2016 : 2:00 PM to 3:50 PM
Sponsor: Business and Economic Statistics Section
Abstract #319868 View Presentation
Title: Predicting Industry Output with Statistical Learning Methods
Author(s): Peter Meyer* and Wendy Martinez
Companies: Bureau of Labor Statistics and Bureau of Labor Statistics
Keywords: Modeling ; Productivity ; Industry output ; Cross-validation ; Colinearity

The Bureau of Labor Statistics uses estimates of industry output to produce industry-level productivity statistics, but full output measures are not available until more than a year after the end of the reference year. Our objective is to generate preliminary estimates of the value of output produced by each of 86 U.S. manufacturing industries within four months of the end of the year. Many predictors are available by industry for the reference year, including output proxies from the Federal Research and the Census Bureau, price indexes of industry output, industry imports and exports, and employment counts and wages paid. Previous work used multiple regression, and found that the quality of early predictions using this methodology has not been high for all industries, perhaps because the predictors are collinear. In this analysis, we test more sophisticated statistical prediction methodologies, including ridge regression, LASSO, ElasticNet, multivariate adaptive regression splines, and more. Performance of these methods are compared using 2007-2014 data and tested with cross-validation methods.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

Copyright © American Statistical Association