AN ADL MODEL FOR ELECTRICITY AND NATURAL GAS

AN ADL MODEL FOR ELECTRICITY AND NATURAL GAS DEMAND IN COLORADO LEILA DAGHER, PHD AMERICAN UNIVERSITY OF BEIRUT

OUTLINE Introduction ØLiterature Review ØModel ØData and Methodology ØResults ØConclusions Ø

INTRODUCTION • Primary Goal: estimate dynamic price elasticities. • Secondary Goals • • • End-use impact Price elasticity stability Correct price variable

INTRODUCTION • • Elasticities are used by utilities and government agencies for: Forecasting Policy making There is a renewed interest in elasticities as a result of the increased concern in energy pollution the rise in energy prices.

INTRODUCTION • • Capital Intensive → Huge Savings Elasticities are region specific Existing estimates for Colorado are inconsistent with economic theory Stability of estimates

Xcel Energy Service Territory

Electricity Natural Gas

LITERATURE REVIEW PRODUCT US OTHER TOTAL ELECTRICITY 184 221 405 NATURAL GAS 87 95 182 271 316 587 TOTAL

LITERATURE REVIEW • Omission of standard errors especially for the LR elasticities • Wide-ranging estimates o o o Consumer sectors Sample periods Modeling variables Level of analysis Modeling methods and data types

DEMAND MODEL ADL

DEMAND MODEL

DATA AND METHODOLOGY • Unit-root testing • Co-integration testing • Multicollinearity • Data were averaged and logged • Deflator CO CPI • Lagged prices • Frequency conversion • Customers variables were smoothed using an IV

ESTIMATION ISSUES • Spurious Regression • Statistical Inference • Price Endogeneity • Inconsistent Estimates

METHODOLOGY • OLS regression and choose the ARDL model that has uncorrelated errors while optimizing the SIC. • T and F statistics on this model are valid

METHODOLOGY • Lag selection • Residual Diagnostics • Saturation/efficiency indices • Test for model and coefficient stability and • • price asymmetries Monthly bill Dynamic elasticities

RESULTS Electric Small Commercial Variable Coefficient Std. Error t-Statistic Prob. C 7. 982 1. 338 5. 965 0. 000 Qe (-1) 0. 518 0. 082 6. 347 0. 000 Qe (-2) -0. 466 0. 095 -4. 898 0. 000 Qe (-3) 0. 346 0. 097 3. 583 0. 000 Qe (-4) -0. 144 0. 096 -1. 506 0. 135 Qe (-5) 0. 118 0. 079 1. 486 0. 140 -0. 065 0. 028 -2. 280 0. 024 ЄLR=-0. 104 0. 043 Pe(-1)

RESULTS Electric Small Commercial obs 1994 M 07 1994 M 08 1994 M 09 1994 M 10 1994 M 11 1994 M 12 1995 M 01 1995 M 02 1995 M 03 1995 M 04 1995 M 05 1995 M 06 1995 M 07 Qe -0. 0651 -0. 0986 -0. 0859 -0. 0940 -0. 0963 -0. 0995 -0. 1014 -0. 1005 -0. 1011 -0. 1023 -0. 1027 -0. 1030 SE 0. 0284 0. 0429 0. 0377 0. 0374 0. 0406 0. 0407 0. 0415 0. 0423 0. 0420 0. 0421 0. 0426 0. 0427 0. 0428

RESULTS Electric Small Commercial

RESULTS Summary Table Sector Electric Residential Price Elasticity SR LR -0. 015 (0. 028) -0. 028 (0. 075) Electric Small Commercial -0. 065* (0. 028) -0. 103* (0. 043) Electric Large Commercial -0. 015 (0. 010) -0. 016 (0. 011) Natural Gas Residential -0. 028 (0. 019) -0. 070 (0. 049) Natural Gas Small Commercial -0. 003 (0. 021) -0. 016 (0. 072)

SENSITIVITY ANALYSIS • Data aggregation • Seasonal differencing • Different models • Lag selection • Selection criterion • Sample periods

CONCLUSIONS & IMPLICATIONS • Demand is highly inelastic • Surcharges for DSM or RE • Customers do not respond to joint bill • LR range • DE useful tool for end users

THANK YOU!