Comparison of Pooled and Household-Level Usage Impact Analysis Jackie Berger Ferit Ucar IEPEC Conference – August 14, 2013.

Slides:



Advertisements
Similar presentations
Agency for Healthcare Research and Quality (AHRQ)
Advertisements

New Paradigms for Measuring Savings
Experimental and Ex Post Facto Designs
1-Way Analysis of Variance
A Two-Level Electricity Demand Model Hausman, Kinnucan, and Mcfadden.
Automated Demand Response Pilot 2005/2004 Load Impact Results and Recommendations Final Report © 2005 Rocky Mountain Institute (RMI) Research & Consulting.
1 Home Gas Consumption Interaction? Should there be a different slope for the relationship between Gas and Temp after insulation than before insulation?
LECTURE 3 Introduction to Linear Regression and Correlation Analysis
2005 LIEE Impact Evaluation Final Report January 23, 2007 Presentation to the Low Income Oversight Board West Hill Energy and Computing, Inc. with Ridge.
Rebate programs for water efficient appliances: Are municipalities just flushing money down the drain? Jonathan Lee Center for Environmental & Resource.
Evaluating Hypotheses Chapter 9. Descriptive vs. Inferential Statistics n Descriptive l quantitative descriptions of characteristics.
© 2000 Prentice-Hall, Inc. Chap Multiple Regression Models.
Multiple Regression Models. The Multiple Regression Model The relationship between one dependent & two or more independent variables is a linear function.
Evaluating Hypotheses Chapter 9 Homework: 1-9. Descriptive vs. Inferential Statistics n Descriptive l quantitative descriptions of characteristics ~
Are Building Codes Effective at Saving Energy? Evidence from Residential Billing Data in Florida Grant D. Jacobsen UC Santa Barbara Matthew J. Kotchen.
FINAL REPORT: OUTLINE & OVERVIEW OF SURVEY ERRORS
NJ Comfort Partners Evaluation Jackie Berger August 21, 2014.
17 June, 2003Sampling TWO-STAGE CLUSTER SAMPLING (WITH QUOTA SAMPLING AT SECOND STAGE)
BPA Pre-Pilot, Monmouth  14 homes with installed DHP, single zone, single compressor.  11 Monmouth, 2 Moses Lake, 1 Tacoma  Savings.
Overview of the 2009 LIEE Impact Evaluation Workshop 1: “Overview of Lessons Learned” October 17, 2011.
The new HBS Chisinau, 26 October Outline 1.How the HBS changed 2.Assessment of data quality 3.Data comparability 4.Conclusions.
Chapter 8 Introduction to Hypothesis Testing
1Managed by UT-Battelle for the Department of Energy Michael Blasnik M Blasnik & Associates Greg Dalhoff Dalhoff Associates, LLC David Carroll APPRISE.
Performance Metrics for Weatherization UGI LIURP Evaluation Yvette Belfort Jackie Berger ACI Home Performance Conference April 30, 2014.
Manufactured Housing Duct Sealing Pilot - Independent Evaluation Results Tom Eckhart, Howard Reichmuth, Jill Steiner Regional Technical Forum February.
T tests comparing two means t tests comparing two means.
Lecture 12 Statistical Inference (Estimation) Point and Interval estimation By Aziza Munir.
Residential Behavior Programs RTF Subcommittee Ryan Firestone September 17, 2015.
1 NORTHWEST ENERGY EFFICIENCY ALLIANCE Northwest Ductless Heat Pump Pilot Project Impact & Process Evaluation: Billing Analysis Ecotope, Inc. February.
Agresti/Franklin Statistics, 1 of 106  Section 9.4 How Can We Analyze Dependent Samples?
Sampling Class 7. Goals of Sampling Representation of a population Representation of a population Representation of a specific phenomenon or behavior.
Measures that Save The Most Energy Jackie Berger David Carroll ACI New Jersey Home Performance Conference March 5, 2010.
EvergreenEcon.com ESA 2011 Impact Evaluation Research Plan Public Workshop #1 February 20, 2013 Presented By: Steve Grover, President.
EvergreenEcon.com ESA 2011 Impact Evaluation Draft Report Public Workshop #2 August 7, 2013 Presented By: Steve Grover, President.
Experience you can trust. 1 RECAP Technology Tour Date: 2004.
Demand Side Management Programs National Energy and Utility Affordability Conference Denver, Colorado David Carroll June 18, 2008.
Demand Response and the California Information Display Pilot 2005 AEIC Load Research Conference Myrtle Beach, South Carolina July 11, 2005 Mark S. Martinez,
© 2013 Cengage Learning. All Rights Reserved. May not be scanned, copied or duplicated, or posted to a publicly accessible website, in whole or in part.
Measures that Save The Most Energy Jackie Berger David Carroll ACI New Jersey Home Performance Conference January 25, 2007.
New Evidence on Energy Education Effectiveness Jackie Berger 2008 ACI Home Performance Conference April 8, 2008.
Achieving Higher Savings in Low-Income Weatherization Jacqueline Berger 2015 IEPEC Conference ― Long Beach, California.
BGE Limited Income Pilot Programs - Evaluation ACI Home Performance Conference March 2012.
EMV Results for online Energy Education Study conducted by Lei Wang, PhD October 2011.
Energy Education in the Home Jackie Berger 2014 BECC December 9, 2014.
Evaluating Impacts of MSP Grants Hilary Rhodes, PhD Ellen Bobronnikov February 22, 2010 Common Issues and Recommendations.
Review of the New England “Mini-Pilot” DHP Evaluation Why we ignore this study.
© Copyright McGraw-Hill 2000
2009 Impact Evaluation Concerns ESAP Workshop #1 October 17, 2011.
Evaluating Impacts of MSP Grants Ellen Bobronnikov Hilary Rhodes January 11, 2010 Common Issues and Recommendations.
2016 Long-Term Load Forecast
Using Feedback as a Tool for Household Energy Conservation: An Experimental Approach Kannika Thampanishvong Policy Dialogue “Transition to Green Economy.
Why Data Matters Building and Sustaining a Business Case NEAUC Conference June 18, 2014.
Impact of Energy Efficiency Services on Energy Assistance NEUAC Conference June 18, 2014.
Utilities’ Update on Energy Savings Assistance Program Studies Ordered in D LIOB Meeting August 21, 2013 Sacramento, California.
Heat Pump Research Project Sponsored by the Heat Pump Working Group April 5, 2005.
T tests comparing two means t tests comparing two means.
Learning Objectives After this section, you should be able to: The Practice of Statistics, 5 th Edition1 DESCRIBE the shape, center, and spread of the.
© 2007, Itron Inc. Statistically Adjusted End-Use Model Overview & Thoughts about Incorporating DSM into a Forecast May 4, 2009 Frank A. Monforte, Ph.D.
Best Practices in Residential Energy Efficiency
Evaluating Impact Do it Right or Not At All
12 Inferential Analysis.
More on Specification and Data Issues
Health and Safety Investments to Increase Energy-Saving Opportunities
South Jersey Gas Home Performance Program & Evaluation
WAP Warm Climate Weatherization: Opportunities for Energy Savings
Health and Safety Investments to Increase Energy-Saving Opportunities
12 Inferential Analysis.
Behavior Modification Report with Peak Reduction Component
Evaluating Low-Income Programs Why and How
Jackie Berger Home Performance Conference April 3, 2019
Presentation transcript:

Comparison of Pooled and Household-Level Usage Impact Analysis Jackie Berger Ferit Ucar IEPEC Conference – August 14, 2013

Presentation Outline Motivation Billing Analysis Usage Impact Models Model Results Summary and Next Steps 2

MOTIVATION 3

Usage Impacts Were expected energy savings results obtained? Are the treatments cost-effective? Should measure selection procedures be revised? Should installation procedures be reviewed? Should contractors be re-trained? 4

Analysis Method Goal: develop most accurate estimate of program savings. Weigh costs and benefits of various approaches to measurement. Consider possible causes of mis-measurement or bias. 5

What Are You Measuring? 6 Approach MeasuresIssues Engineering Estimate Expected usage change based on measures alone Assumptions Installation quality Other usage changes Usage 2 - Usage 1 Actual change in usage Weather Other factors Weather Norm Usage 2 - Usage 1 Change in usage if both periods had average weather Other factors Weather Norm Usage 2 - Usage 1 w/Comp. Group Other factors held constant (prices, economy, market information, etc.) Best estimate of program impact

Analysis Approaches 7 Approach CostAccuracyAttrition Reasons for Exclusion Engineering Estimate $****None Engineering Estimate with Retrofit Data $$**** Retrofit Data Missing Billing Analysis $$$***** Usage Data Missing or Inadequate Metering$$$$*****Cost

Billing Analysis 8

Data Requirements 9 Core Data RequiredSupplemental Data Energy billing data Read date, real or estimated, usage, units Energy efficiency measures Measure-specific impacts Service delivery date Divides period into pre- and post- treatment Service delivery providers Provider-specific impacts Weather data Local weather station, daily temp for pre and post period and longer normalization period Housing unit characteristics Relation between housing /household characteristics and savings Household characteristics

Challenges 10 Concern Data Attrition Savings do not represent treated population Sample Size Low precision for savings estimates Cannot estimate for sub-groups Comparison Group Need to control for exogenous factors Not able to do random assignment Difficult to find comparable population Later program participants Earlier program participants Comparable households

Treatment and Comparison Group Example 11 Post-Treatment Period PreYr 2 – Quasi Post 2010 Post Yr 2 – Quasi Post SERVICE DELIVERY DATE SERVICE DELIVERY DATE Pre-Treatment Period Post Yr 1 – Quasi Pre Pre Yr 1 – Quasi Pre Treatment Group Comparison Treated Year Before Comparison Treated Year After

Usage Impact Models 12

House-by-House 13 NAC = 365α + βHo(τ) H o (τ) = long term average heating degree days Fi = average daily consumption in time interval i H i (τ) = heating degree days to reference temperature τ in interval i i = random error term Regression Analysis on Each Individual Home: Fi = α + βHi(τ) + i

Pooled Analysis 14 F it = average daily usage during the pre- and post-treatment periods H it = average daily base 60 HDDs POSTt = a dummy variable that is 0 in the pre-period and 1 in the post-period ε it = estimation error term PRE USAGE α i = average daily baseload usage in pre-treatment period. β1 = average daily usage per HDD in the pre-treatment period. POST USAGE α i + β2= average daily baseload usage in the post-treatment period. β1 + β3= average daily usage per HDD in the post-treatment period. SAVINGS β2 = average daily baseload savings β3 = heating usage savings per HDD. F it = α i + β1* H it + β2*POST t + β3*POST t *H it + ε it

Advantages 15 House-by-HousePooled Detailed attrition analysis Utilizes all billing data Post-analysis of usage and savings Direct estimate of savings is furnished Analysis of high- and low-saving homes Regression models can be used to estimate savings by measure Exogenous factors can be included in the model Relationship between usage and housing/ household characteristics can be explored

Disadvantages 16 House-by-HousePooled Less robust when energy use response to degree days varies within household Inclusion of many parameters can make final results difficult to interpret Requires close to a full year of pre- and post-treatment data Alternative functional form may be required Substantial attrition can bias the analysis Limited ability to furnish information on savings distribution and conduct exploratory analysis

When to Use 17 House-by-HousePooled Availability of close to one year of pre-and post-treatment usage data for significant % of treatment and comparison Limited data availability and concern for attrition bias associated with excluding homes Data available on treatment, home, or households, that can be used to assess factors related to higher or lower savings Supplemental data not available Study sponsors not interested in supplemental analysis

Model Results 18

Program 1 Results Gas Heating Jobs 19 ModelObs. Pre- Use Post- Use Savings ccf% Not Normalized 1,1661, (±11)6.6% House-by-House 1,1661, (±10)5.8% Pooled Regression 1,1661, (±10)6.4% Pooled-Month Dummy 1,1661,0841,02064 (±10) 5.9% Pooled-all obs.1,439 1, (±9)6.3%

Program 2 Results Gas Heating Jobs 20 ModelObs. Pre- Use Post- Use Savings ccf% Not Normalized 1,2111, (±12) 20.2% House-by-House 1,2111, (±10) 6.5% Pooled Regression 1, (±9) 6.3% Pooled-Month Dummy 1,2111, (±10) 6.5% Pooled-all obs. 1,6651, (±8)6.9%

Program 1 Results Electric Baseload Jobs 21 ModelObs. Pre- Use Post- Use Savings kWh% Not Normalized 4,05511,15310, (±73)3.2% House-by-House 4,05511,37010,1471,223 (±78)10.8% Pooled Regression 4,05510,6249, (±55) 8.4% Pooled-Month Dummy 4,05510,7989, (±56) 7.8% Pooled-All Obs. 5,375 10,7289, (±53) 7.8% Pooled-Month-All 5,375 11,19010, (±55) 6.8%

Program 2 Results Electric Baseload Jobs 22 ModelObs. Pre- Use Post- Use Savings kWh% Not Normalized 2,44011,0229,7651,257 (±93)11.4% House-by-House 2,44010,75810, (±99)5.7% Pooled Regression 2,44010,1399, (±69)6.3% Pooled-Month Dummy 2,4409,7799, (±82)6.7% Pooled-All Obs. 4,65410,2879, (±56) 5.5% Pooled-Month-All 4,6549,8539, (±66)5.8%

Program 1 Results Electric Heating Jobs 23 ModelObs. Pre- Use Post- Use Savings kWh% Not Normalized 14417,84617,77967 (±541)0.4% House-by-House 14419,66218,5341,128 (±503)5.7% Pooled Regression 14417,94017, (±559) 4.8% Pooled-Month Dummy ,73818, (±586) 4.6% Pooled-All Obs ,83016, (±491) 4.7% Pooled-Month-All ,84620, (±515) 3.9 %

Program 2 Results Electric Heating Jobs 24 ModelObs. Pre- Use Post- Use Savings kWh% Not Normalized 13418,10314,2983,805 (±646)21.0% House-by-House 13419,40217,8991,503 (±665)7.7% Pooled Regression 13417,02015,5051,515 (±543)8.9% Pooled-Month Dummy ,17715,6141,562 (±647) 9.1% Pooled-All Obs ,88415,2631,621 (±391)9.6% Pooled-Month-All ,19316,3741,819 (±457) 10.1%

Electric Baseload Pre-Treatment UsageObs. Pre- Use Post- Use Savings kWh% < 8,000 kWh7036,8626, % 8,000 – 12,000 kWh1,9729,7208, % > 12,000 kWh1,38016,02413,8522, % Program 1 Results Household Characteristics 25 OwnerObs. Pre- Use Post- Use Savings kWh% Owner2,26311,43310,1551, % Renter1,79211,29110,1371, %

Electric Baseload Supplemental HeatObs. Pre- Use Post- Use Savings kWh% Supplemental Heat2,05412,29310,8121, % No Supp Heat2,00110,4239, % Program 1 Results Household Characteristics 26 Level of ServiceObs. Pre- Use Post- Use Savings kWh% Basic3,53311,29110,1611, % Major52211,90610,0511, % Major measures include refrigerators, air conditioner, and water heater replacements.

Electric Baseload MeasuresObs. Pre- Use Post- Use Savings kWh% Air Conditioner7811,3639,6261, % No Air Conditioner3,97711,37010,1571, % Refrigerator42011,2119,3761, % No Refrigerator3,63511,38810,2351, % AC/Refrigerator2711,4329,2322, % AC/ No Refrigerator5111,3279,8341, % No AC/ Refrigerator39311,1969,3861, % No AC/ No Refrigerator3,58411,38910,2411, % Program 1 Results Household Characteristics 27

Summary and Next Steps 28

Summary Overall savings results fairly consistent Differences between models rarely statistically significant Gas usage results were more consistent Electric baseload varied most 29

Conclusions Sources and potential biases caused by large data attrition should be explored. When additional analysis is desired for many subgroups and data attrition is low, house-by- house may be favored. When data attrition is high and only overall usage results are desired, the pooled regression may be preferred. 30

Next Steps Additional exploration of differences. Explore deletion of various types and numbers of observations from house by house. Compare results with different levels of attrition. Test different functional forms for the pooled model. 31