Comparisons of Calibration Methods for the RECS Engineering End-Use Estimates

• Activity Number: 253 - SPEED: Government Statistics, Health Policy, and Marketing
• Type: Contributed
• Date/Time: Monday, July 31, 2017 : 3:05 PM to 3:50 PM
• Sponsor: Government Statistics Section
• Author(s): Shaofen Grace Deng* and Greg Lawson and Chrishelle Lawrence
• Companies: U.S. Energy Information Administration and U.S. Energy Information Administration and U.S. Energy Information Administration
• Keywords: RECS ; engineering end-use models ; calibration ; residential end uses

Abstract

Since the 1970s, the U.S. Energy Information Administration (EIA) has conducted the Residential Energy Consumption Survey (RECS), which collects residential household characteristics with energy use patterns and the corresponding household consumption totals and expenditures from energy suppliers. One primary purpose use of the data collected in RECS is to produce household end-use estimates. In the previous rounds of the RECS, EIA used a nonlinear regression model approach to disaggregate the total annualized consumption data into different end uses based on the housing characteristics data and weather data. For the forthcoming RECS 2015 survey, EIA is planning a new approach to end-use estimation in making use of engineering models. However, calibration will be needed to ensure the end uses sum to the supplier reported consumption totals while ensuring the engineering model outputs remain representative. Here we report on our research of the calibration step: we compare a simple normalization calibration with an "informed" calibration approach, which utilizes the relative uncertainties and correlations of the end uses.