CONSISTENT RESIDENTIAL EFFICIENCY IMPROVEMENTS ACROSS END-USES: THEORETICAL AND EMPIRICAL INSIGHTS Mike Blackhurst Assistant Professor The University Of Texas At Austin Civil, Architectural, & Environmental Engineering mike.blackhurst@austin.utexas.edu Multiple Perspectives on Technical Efficiency What happens if you double the efficiency of your air conditioner? The technologist says, “You use half the energy.” The economist says, “You turn down the thermostat.” The social scientist says, “Who made the decision?” The “Rebound Effect” o aka “Jevon’s paradox” or “the energy efficiency paradox” o Efficiency decreases resources needed for service o Efficiency also decreases the cost of service, which… o Induces income and substitution effects and… o Likely other behavioral responses and drivers Rebound Terminology Category Description Example Direct rebound Homeowners use more of the more efficient service Consumer drives more with a more fuel efficient car Indirect rebound Homeowners re-spending on other goods and services Savings from efficient lighting spend on 2nd refrigerator Economywide rebound More efficient production and shifts in demand alter economic structure and growth A more efficient steam engine increases production changes structural relationships and leads to economic growth Magnitude of Rebound Debated Technically Feasible After Direct Rebound After direct + indirect rebound Net Energy Elasticity (% Change in Energy / % Change in Efficiency) 0% -20% -40% -60% -80% -100% Technically feasible energy savings Single-Service Rebound Model o Start with technical definition of efficiency: E = S / e ¶E e ¶e E = he ( E) = he (S )-1 he (E) = -1 when he (S ) = 0 o Direct rebound usually estimated as own-price elasticity of demand he (E) ~ -h p( E)-1 o Indirect rebound (re-spending) is estimated by modeling by income and substitution effects in response to a discrete efficiency change Avg. Technology Choices per Household (STech/SHHs) Challenge to Single Service Model 4 OTHERS HEAT PUMP AC WEATHERIZED 3 CFLs WELL INSULATED 2 PROGRAM. THERMOSTAT ES CLOTHES WASHER 1 ES DISHWASHER ES REFRIGERATOR 0 < 5k $1 5k $1 .9 19 $ - k $ k 20 29 -$ k .9 0k $3 - 9k 9. 3 $ 0k $4 - 9k 4. 5 $ 5k $5 - 9k 9. 7 $ 0k $8 - 9. $9 9k 2009 Annual Household Before-Tax Income 0 $1 0k - Modified from Blackhurst and Ghosh (under review) 99 $1 k .9 > k 20 1 $ Two Service Model E = Ei + E j = f ( pi , p j ,ei ,e j , M ,s ) 0 0 0 0 ¶E ¶f ¶f ¶pi ¶f ¶p j ¶f ¶e j ¶f ¶M ¶f ¶s + + + + + = ¶ei ¶ei ¶pi ¶ei ¶p j ¶ei ¶e j ¶ei ¶M ¶ei ¶s ¶ei Two Service Model ¶E ¶f ¶f ¶e j = + ¶ei ¶ei ¶e j ¶ei he ,cc(Ei ) = he ( Ei )+ he (E j )he (ei ) i i i j he ,cc(E j ) = he (E j )+ he ( E j )he (e j ) i i j i Ej Ei he ,cc ( Ei + E j ) = he ,cc( Ei ) + he ,cc( E j ) i i i E E Two Service Model: Re-Arranged he C = ,cc (EC +ET ) æE ET ö C - ç + he (eT ) ÷ C E ø è E E + [he ( EC )+1] C C E + he ( ET ) C + C direct rebound for C (1st order) ET E he (eT )he ( EC ) C technical response (1st and 2nd order) indirect rebound from C to T ind. of e correlation (1st order) EC E E + he (eT )éëhe ( ET )+1ùû T C T E indirect rebound from j to i from e correlation (2nd order) indirect rebound from i to j from e correlation (2nd order) Application of Two-Service Model Would homeowners in more efficient homes drive more? o Include electricity (C) and transportation (T) services o Used constant elasticity of substitution (CES) production function V(M , pC , pT ,eC ,eT ) = max {U(Y )} M ³ X + pC EC + pT ET EC = C / eC ; ET = T / eT s ( s -1) s ( s -1) s ( s -1) s ù é Y = ë(1- aC - aT )X + aC C + aT T û o Can provide draft manuscript for more details ( s -1) Empirical Assumptions Parameter Base Case Income category ($1,000) $25-30 Min. $40-45 Max. $70$75 Ref BLS 2011 Short-run elasticity of sub., sSR 0.15 0.1 0.2 Long-run elasticity of sub., sLR 0.8 0.7 0.9 BLS 2011, Dahl 1993, Brons 2008, Graham 2002 Electricity Nominal Shares, aC 1.3% 0.4% 2.2% BLS 2011 Gasoline Nominal Shares, aT 2.9% 0.8% 5.1% BLS 2011 Electricity Real Shares 27% 26% 31% BLS 2011 Gasoline Real Shares 73% 74% 69% Efficiency correlation, heT(eC) 2.1 0.5 6 heC(eT) 0.48 2.00 0.17 BLS 2011 Replacements assuming different code- and above-code performance Rebound Across Resid and Trans Sectors: Driven by Changes in Electricity Efficiency Technically Technically feasible elasticity Direct Direct rebound Rebound feasible elasticity i/E + hei(ej)Ej/E) -1(E-1(E i/E + hei(ej)Ej/E) Cross-sector (indirect), Cross-sector, independent from C to T of c.c. of c.c. independent + e (Ei)+1]EE/E /E [he[h i(Eii)+1] ii + ei(Ej) Ej/E hhei(E j) Ej/E Cross-sector, Cross-sector, Fromfrom trans toCresid T to with with c.c. c.c. Cross-sector, Cross-sector, From resid to trans from C to T withwith c.c.c.c. ei(ej) hej (Ei )Ei/E +heh(e + hei(ej) [hej(Ej)+1] Ej/E = i j) hej (Ei )Ei/E hei(ej) [hej(Ej)+1] Ej/E -0.5 (-1) Energy Elasticity, DE E De e -1.5 -2.5 -3.5 Short-run response Long-run response -4.5 Results shown for median income range ($40-$45k) Net Energ Elastici Rebound Across Resid and Trans Sectors: Driven by Changes in Electricity Efficiency Technically Technically feasible elasticity Direct Direct rebound Rebound feasible elasticity i/E + hei(ej)Ej/E) -1(E-1(E i/E + hei(ej)Ej/E) Cross-sector (indirect), Cross-sector, independent from C to T of c.c. of c.c. independent + e (Ei)+1]EE/E /E [he[h i(Eii)+1] ii + ei(Ej) Ej/E hhei(E j) Ej/E Cross-sector, Cross-sector, Fromfrom trans toCresid T to with with c.c. c.c. Cross-sector, Cross-sector, From resid to trans from C to T withwith c.c.c.c. ei(ej) hej (Ei )Ei/E +heh(e + hei(ej) [hej(Ej)+1] Ej/E = i j) hej (Ei )Ei/E hei(ej) [hej(Ej)+1] Ej/E -0.5 Energy Elasticity, DE E De e -1.5 -2.5 -3.5 Short-run response Long-run response -4.5 Results shown for median income range ($40-$45k) Net Energ Elastici Rebound Across Resid and Trans Sectors: Driven by Changes in Vehicle Efficiency Technically feasible elasticity -1(Ei/E + hei(ej)Ej/E) 1.5 0.5 Energy Elasticity, DE E De e Technically feasible elasticity Direct rebound + [hei(Ei)+1] Ei/E Direct Rebound, hei(Ei) + Cross-sector, from resid. to trans. independent of c.c. hei(Ej) Ej/E Cross-sector (indirect), independent of c.c. + Cross-sector, from trans. to resid. with c.c. hei(ej) hej (Ei )Ei/E + Cross-sector, From resid to trans with c.c. Cross-sector, from resid. to trans. with c.c. hei(ej) [hej(Ej)+1] Ej/E = Cross-sector, From trans to resid with c.c. Technical response -0.5 -1.5 Short-run response Long-run response -2.5 Results shown for median income range ($40-$45k) E Other Behavioral Drivers Behavior or Driver Effect on technology… Choice Cost minimization, income constraint Reference(s) Use High implicit discount rate observed Hausman 1979; Sanstad et al. 1995 Demographic Education levels, ownership, & tenure increased technology adoption ? Hartman 1998; Michelson & Madner 2011 Physical household characteristics Increased home age and size promote technology adoption ? Michelson and Madner 2011 Environmental awareness and valuation Increased awareness & valuation increased adoption ? Cummings and Taylor 1999; Hanley et al. 1990; Bateman et al. 2011 Technological awareness Homeowners misperceive technology performance at extremes; Self-reported awareness increased adoption ? Attari 2010; Nair et al 2010 Other Behavioral Drivers Do homeowners correlate or compensate drivers of energy technology choice and use? o Limited qualitative insights • Correlation and compensation observed across a variety of “green” behaviors [Thøgersen & Ölander 2003] • Self-reported behavior changes with PV adoption [Keirstead 2007; McAndrews; Schweizer-Reis et al. 2000 ] o Implications for rebound? Empirical Research o Estimate the impact of marginal technical change within and across end uses on electricity use and rebound • If choose technology A versus • If choose both technology A and technology B Pecan Street Research Institute Static data High resolution consumption data Representative Sample Data Sample includes one year of monthly electricity consumption for 79 homes Variable Range Climate Monthly CDD Mean= 292, SD= 257 Structural Floorspace (square feet) Windows area (square feet) Age of the house Occupancy Tenure HH income Thermostat set point – summer TV hours per month Dishwasher loads per month Clothes washer loads per month Education (interval) Attic insulation R-value Air conditioning Energy Efficiency Ratio (EER) No. of devices Dummy variables, Programmable thermostat, Double pane windows, Energy star appliances, Solar PV (count = 37), EV (count = 14), Electric heater Electricity consumption (KWh/month) Mean= 2,019, SD= 719 Mean= 245, SD= 106 Mean= 21.4, SD= 23.6 Mean= 2.7, SD= 1.2 Mean= 6.6, SD= 7.6 Mean= $128k SD= $62k Mean= 76.9, SD= 2.2 Mean= 107, SD= 71.9 Mean= 14.3, SD= 8.1 Mean= 17.1, SD= 9.2 Demographic Selfreported behaviors Technology choices Electricity Mean= 28.6, SD= 8.4 Mean= 10.5, SD= 1.7 Mean= 3.34, SD= 1.8 Mean= 963, SD= 938 Model Specification o Where • Yitλ represents monthly electricity consumption • βj are the predictor coefficient fixed effects • βi are the coefficient estimates for random effects • Sijλ represents a series of household structural factors • Dijλ represents a series of household demographic factors • Bijλ represents household behaviors and cognitive factors • Xij interaction terms for different technology choice combinations • Ri represents the household identification codes Results with No Interaction Terms Explanatory variable ProgTherm ES Refrig 1/sqrt(Sq Ft) Devices CWloads Home R value Cooling Degree Days EV ES DW 2-P window ES Clothes washer PV AC EER 1/occupancy Dishwasher loads 1/Window Sq Ft Income Constant (b0) Coefficient -0.236 -0.164 -75.7 0.067 -1.958 -0.009 0.001 0.087 -0.085 -0.081 -0.069 0.054 0.006 0.141 0.001 -2.086 2.00E-07 8.348 p-value 0.026 0.025 < .0001 0.027 0.053 0.087 < .0001 0.339 0.325 0.346 0.378 0.487 0.738 0.465 0.879 0.924 0.78 < .0001 % change in Y for: 1 unit (or *10%) increase in X -21.0% -15.1% 8.35%* 7.00% 1.20%* -0.91% 0.14% 9.14% -8.14% -7.78% -6.68% 5.61% 0.65% -0.21%* 0.09% 0.09%* 0.00% - Rebound from Marginal Efficiency Gains: Demonstrative Empirical Results Single Pane Window Multi-pane window 7.1 Fitted Value (kWh) 7 6.9 6.8 6.7 6.6 6.5 6.4 6.3 6 7 8 9 10 11 12 Air Conditioning Efficiency (EER) 13 14 Rebound with Marginal Efficiency Gains Individual Coeff.--> -0.24 * Prog. Therm. Prog. Therm. Multi-Pane Windows <19 Home 19-29 Insulation >30 AC 7 Energy 11 Efficiency 14 ES Refrigerator ES Dishwasher * ES Clothes Washer * 1 Devices 4 (count) 9 -0.081 Multi Pane -0.009 * Home R value <19 19-29 >30 0.006 AC EER 7 11 14 * * * * * * -0.16 * -0.085 -0.069 0.068 * ES ES Devices (count) ES CW Refrig. Dishw. 1 4 9 * * * * * * * Multi-pane windows* installed, * * increased * AC efficiency * * * * * * * * * * Multi-pane windows installed * at indicated AC efficiency * * * * * * Legend Increase in consumption (rebound) No effect Net energy savings AC = Air conditioning; EER = energy efficient ratio; ES = Energy Start; CW = Clothes washer * Significant to 10% level Rebound with Marginal Efficiency Gains Individual Coeff.--> -0.24 * Prog. Therm. Prog. Therm. Multi-Pane Windows <19 Home 19-29 Insulation >30 AC 7 Energy 11 Efficiency 14 ES Refrigerator ES Dishwasher * ES Clothes Washer * 1 Devices 4 (count) 9 -0.081 Multi Pane -0.009 * Home R value <19 19-29 >30 0.006 AC EER 7 11 14 * * * * * * -0.16 * -0.085 -0.069 ES ES ES CW Refrig. Dishw. * * * * * * * * * * * * * 0.068 * Devices (count) 1 4 9 * * * * * * * * * * * * * * * Legend Increase in consumption (rebound) No effect Net energy savings AC = Air conditioning; EER = energy efficient ratio; ES = Energy Start; CW = Clothes washer * Significant to 10% level Preliminary PV Results o Order of technical change matters Order of technical change Increase AC Efficiency Increase Insul. Install multipane Windows Purchase EnergyStar Appliances Have PV before efficiency - - + - Install PV after efficiency change + low EER - high EER + low R-values - high R-values - - + Statistically significant increase in electricity consumption - Statistically significant decrease in electricity consumption Implications o Literature is mixed as to whether consumers correlate or compensate valuations across energy technology choice/use o Empirical work suggests consumers MAY leverage efficiency gains for services ACROSS end uses; our results are also mixed o Rebound is relative to the current efficient technical state of the home and order of technical change o These findings suggest the dominant singleservice rebound paradigm is misleading Implications o Consistent efficiency change across end uses can mitigate consumer responses; however… o Consumers can and do expend energy services; thus… o Models of rebound need to recognize service expansion Implications o The literature assumes PV exclusively replaces conventional grid energy sources; however… o Behavioral implications of PV are entirely unclear o Consumers will treat long-run operating cost of PV as zero o Results are mixed with respect to consumers responses to both efficiency change and installation of PV Related Ongoing/Future Work o Rebound across resources (water/electricity/natural gas/gasoline) o Comparing Empirical and Estimated Energy Consumption (RECS/BeOpt) o Does Weather Influence the Use of PV for Discretionary Electricity End Uses? o Estimating Total and End-Use Residential Water (Energy) Demands Using Energy (Water) Demands o Comparing the Observed and Estimated Performance of Residential Water Efficient Fixtures and Appliances Acknowledgements o This work was funded by • The University of Texas at Austin • Bill and Melinda Gates Foundation Fellowship o PhD students • Nour El-Imane Bouhou • Pamela Torres • Alison Wood o MS Students • Bruk Berhanu • Neftali Torres o Post doc • Sarah Taylor-Lange CONSISTENT RESIDENTIAL EFFICIENCY IMPROVEMENTS ACROSS END-USES: THEORETICAL AND EMPIRICAL INSIGHTS Mike Blackhurst Assistant Professor The University Of Texas At Austin Civil, Architectural, & Environmental Engineering mike.blackhurst@austin.utexas.edu References o Blackhurst, MF, and NK Ghosh. “The Rebound Effect with Consistent Efficiency Improvements and Implications for Cross-Sector Rebound.” Ecological Economics (submitted for review). o Attari, S. Z., M. L. DeKay, C. I. Davidson, and W. B. de Bruin. 2010. “Public Perceptions of Energy Consumption and Savings.” Proceedings of the National Academy of Sciences 107 (37): 16054–16059. o Thøgersen, J., and F. Ölander. 2003. “Spillover of Environment-Friendly Consumer Behaviour.” Journal of Environmental Psychology 23 (3): 225–236. o Keirstead, J. 2007. “Behavioural Responses to Photovoltaic Systems in the UK Domestic Sector.” Energy Policy 35 (8): 4128–4141. o McAndrews, K. “To Conserve or Consume: Behavior Change in Residential Solar PV Owners.” The University of Texas at Austin, 2012. o Hausman, Jerry A. “Individual Discount Rates and the Purchase and Utilization of Energy-Using Durables.” The Bell Journal of Economics 10, no. 1 (April 1, 1979): 33–54. doi:10.2307/3003318. o Sanstad, Alan H., Carl Blumstein, and Steven E. Stoft. “How High Are Option Values in Energy-Efficiency Investments?” Energy Policy 23, no. 9 (1995): 739–743. o Hartman, R. S. “Self-Selection Bias in the Evolution of Voluntary Energy Conservation Programs.” The Review of Economics and Statistics (1988): 448–458. o Michelsen, C., and R. Madlener. “Homeowners’ Preferences for Adopting Residential Heating Systems: A Discrete Choice Analysis for Germany.” FCN Working Papers (2011). o Cummings, Ronald G., and Laura O. Taylor. “Unbiased Value Estimates for Environmental Goods: A Cheap Talk Design for the Contingent Valuation Method.” The American Economic Review 89, no. 3 (June 1, 1999): 649–665. o Nair, Gireesh, Leif Gustavsson, and Krushna Mahapatra. “Factors Influencing Energy Efficiency Investments in Existing Swedish Residential Buildings.” Energy Policy 38, no. 6 (June 2010): 2956–2963. doi:10.1016/j.enpol.2010.01.033. o Bateman, Ian J., Georgina M. Mace, Carlo Fezzi, Giles Atkinson, and Kerry Turner. “Economic Analysis for Ecosystem Service Assessments.” Environmental and Resource Economics 48, no. 2 (2011): 177–218. o Dahl, C. A. “A Survey of Energy Demand Elasticities in Support of the Development of the NEMS” (1993). http://mpra.ub.unimuenchen.de/13962/. o Brons, Martijn, Peter Nijkamp, Eric Pels, and Piet Rietveld. “A Meta-Analysis of the Price Elasticity of Gasoline Demand. A SUR Approach.” Energy Economics 30, no. 5 (September 2008): 2105–2122. doi:10.1016/j.eneco.2007.08.004. o Graham, Daniel J., and Stephen Glaister. “The Demand for Automobile Fuel: A Survey of Elasticities.” Journal of Transport Economics and Policy (2002): 1–25. o BLS (U.S. Bureau of Labor Statistics). “Consumer Expenditure Survey,” 2011. http://www.bls.gov/cex/. Single service rebound model o Using technical definition of efficiency: E = S / e ¶E e = he ( E) = he (S )-1 ¶e E he (E) ~ -h p( E)-1 o Using CES production function V(M , p,e ) = max {U(Y )} M ³ X + pE E = S /e s ( s -1) s ( s -1) s ù é Y = ë(1- a S )X + aS S û he (E) ~ -h p( E)-1 ( s -1) Rebound with Marginal Efficiency Gains Individual Coeff.--> -0.24 * Prog. Therm. Prog. Therm. Multi-Pane Windows <19 Home 19-29 Insulation >30 AC 7 Energy 11 Efficiency 14 ES Refrigerator ES Dishwasher * ES Clothes Washer * 1 Devices 4 (count) 9 -0.081 Multi Pane -0.009 * Home R value <19 19-29 >30 0.006 AC EER 7 11 14 * * * * * * -0.16 * -0.085 -0.069 0.068 * ES ES Devices (count) ES CW Refrig. Dishw. 1 4 9 * * * * * * * Multi-pane windows* installed, * * increased * AC efficiency * * * * * * * * * * Multi-pane windows installed * at indicated AC efficiency * * * * * * Legend Increase in consumption (rebound) No effect Net energy savings AC = Air conditioning; EER = energy efficient ratio; ES = Energy Start; CW = Clothes washer * Significant to 10% level Rebound Across Resid and Trans Sectors: Driven by Changes in Electricity Efficiency Technically Technically feasible elasticity Direct Direct rebound Rebound feasible elasticity i/E + hei(ej)Ej/E) -1(E-1(E i/E + hei(ej)Ej/E) Cross-sector (indirect), Cross-sector, independent from C to T of c.c. of c.c. independent + e (Ei)+1]EE/E /E [he[h i(Eii)+1] ii + ei(Ej) Ej/E hhei(E j) Ej/E Cross-sector, Cross-sector, Fromfrom trans toCresid T to with with c.c. c.c. Cross-sector, Cross-sector, From resid to trans from C to T withwith c.c.c.c. ei(ej) hej (Ei )Ei/E +heh(e + hei(ej) [hej(Ej)+1] Ej/E = i j) hej (Ei )Ei/E hei(ej) [hej(Ej)+1] Ej/E -0.5 (-1) Energy Elasticity, DE E De e -1.5 -2.5 -3.5 Short-run response Long-run response -4.5 Results shown for median income range ($40-$45k) Net Energ Elastici