Welfare Effects of Home Automation Technology with Dynamic Pricing Bryan Bollinger Wes Hartmann Duke University Stanford University EI Workshop, May 2016 Welfare Effects of Home Automation Technology with Dynamic Pricing 1 / 20
Motivation How can we better enable consumers to respond to changing market conditions? Electricity Demand Application: ◮ Prices are regulated ◮ Real-time pricing seems ideal ◮ Frequency of effective price variation depends on timeliness of response ◮ Policymaker perspective: Can we reduce reliance on more polluting supply alternatives? ◮ Utility perspective: Can we delay additional capacity investments? Can we combine rich historical data with field experiments to better target treatments? Welfare Effects of Home Automation Technology with Dynamic Pricing 2 / 20
What We Do Non-parametrically estimate average consumer welfare impacts of information and home automation technologies ◮ Use control group demand to condition on unobserved demand shocks Estimate the firm surplus from providing all consumers with each technology Estimate household-level treatment effects for each technology x price treatment combination ◮ Match households based on distributions of pre-treatment consumption as a function of state Estimate the firm surplus with optimal targeting Welfare Effects of Home Automation Technology with Dynamic Pricing 3 / 20
Hardware Options for Dynamic Price & Response In-Home Display (IHD): provides information in convenient location Programmable Communicating Thermostat (PCT): automates response Welfare Effects of Home Automation Technology with Dynamic Pricing 4 / 20
Price Treatments Time of Use with Critical Pricing: ◮ Off-peak price of 4.2c per kWh � Weekends, off-peak times on weekdays and holidays ◮ On-Peak price of 23c per kWh � On-peak times (2p-7p) on weekdays ◮ Critical price of 46c per kWh � Rare price overcall for 2 to 8 hours at any time during the year Variable Peak Pricing: ◮ Off-peak price of 4.5c per kWh � Weekends and off-peak times on weekdays ◮ On-Peak prices average 13.6c per kWH: Price / kWh 4.5 11.3 23 46 � Freq in Days 50 37 23 10 � Communicated by 5pm the day before ◮ Critical price of 46c per kWh � Requires only 2 hours notice Control Pricing: ◮ 8.4c per kWh regardless of day or time ◮ Increases to X after Y cumulative kWh in the month Welfare Effects of Home Automation Technology with Dynamic Pricing 5 / 20
Experiment & Treatment Assignments Treatments CON TOU VPP Total 294 0 0 294 Control 0 310 332 642 Portal 0 224 228 452 IHD 0 214 220 434 PCT All three 0 184 172 356 Total 294 932 952 2,178 Includes some households without AC who were eventually dropped Welfare Effects of Home Automation Technology with Dynamic Pricing 6 / 20
Average Hourly Usage by Treatment Weekdays (w/AC) Weekends, (w/AC) 4 4 3.5 3.5 Usage (KWH) Usage (KWH) 3 3 2.5 2.5 2 2 1.5 1.5 1 1 0:00 14:00 19:00 24:00 0:00 14:00 19:00 24:00 control portal only control portal only ihd and portal pct and portal ihd and portal pct and portal all 3 all 3 Welfare Effects of Home Automation Technology with Dynamic Pricing 7 / 20
Average Effects (w/AC) Table: Average Treatment Effect Regression Results Critical Peak Off-Peak Treatments TOU VPP TOU VPP TOU VPP -0.437 -0.281 -0.324 -0.146 -0.113 0.031 Portal (0.137) (0.136) (0.123) (0.123) (0.082) (0.086) -0.671 -0.414 -0.491 -0.274 -0.138 -0.018 IHD (0.141) (0.137) (0.127) (0.126) (0.090) (0.087) -1.190 -1.219 -0.942 -0.681 0.027 0.118 PCT (0.141) (0.141) (0.127) (0.130) (0.087) (0.094) -1.082 -1.231 -0.830 -0.643 0.080 0.080 All 3 (0.134) (0.149) (0.131) (0.138) (0.097) (0.102) N=2,141 N=2,178 N=2,178 Standard errors in parentheses Greater reduction in TOU likely attributable to higher average price Welfare Effects of Home Automation Technology with Dynamic Pricing 8 / 20
All 3 vs. IHD Demand Curves 0.5 PCT/IHD, 10th perc. IHD, 10th perc. PCT/IHD, 50th perc. 0.4 IHD, 50th perc. PCT/IHD, 90th perc. IHD, 90th perc. 0.3 Price 0.2 0.1 0.0 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 kW Welfare Effects of Home Automation Technology with Dynamic Pricing 9 / 20 Figure: All 3 vs. IHD Demand: 10th, 50th and 90th Percentiles of Control Demand
All 3 vs. IHD Expenditure Differences 0.5 0.5 0.5 0.4 0.4 0.4 0.3 0.3 0.3 Price Price Price 0.2 0.2 0.2 0.1 0.1 0.1 0.0 0.0 0.0 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 kW kW kW Welfare Effects of Home Automation Technology with Dynamic Pricing 10 / 20
All 3 vs. IHD Welfare Differences 0.5 0.5 0.5 0.4 0.4 0.4 0.3 0.3 0.3 Price Price Price 0.2 0.2 0.2 0.1 0.1 0.1 0.0 0.0 0.0 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 kW kW kW Welfare Effects of Home Automation Technology with Dynamic Pricing 11 / 20
All 3 vs. IHD Welfare Differences Aggregating across control states, consumers receive a welfare gain of about $28 over the summer, or $280 NPV. ◮ Breaking the linearity of demand between prices gives bounds of $213 to $486 But , in addition to consumer welfare, the firm benefits from reducing demand during periods of critical peak demand. ◮ The costs of supplying a PCT total $250 ◮ PCT needs to decrease critical period usage by -250/700 = -0.357 kW ◮ PCT (with portal) demand reductions are, on average, nearly a full kWh ◮ But can the utility do better? Welfare Effects of Home Automation Technology with Dynamic Pricing 12 / 20
Household Treatment Effects We match HHs based on their distribution of usage as a function of state variables (time of day, temperature, and recent electricity consumption.). We use a weighted regression for each HH to eatimate that HH’s treatment effect. Welfare Effects of Home Automation Technology with Dynamic Pricing 13 / 20
Effects of Portal by TOU and VPP 1 1 critical critical peak peak 0.9 0.9 off peak off peak 0.8 0.8 0.7 0.7 0.6 0.6 cdf cdf 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 difference in control and portal coef. difference in control and portal coef. Welfare Effects of Home Automation Technology with Dynamic Pricing 14 / 20
Effects of IHD by TOU and VPP 1 1 critical critical peak peak 0.9 0.9 off peak off peak 0.8 0.8 0.7 0.7 0.6 0.6 cdf cdf 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 difference in control and ihd coef. difference in control and ihd coef. Welfare Effects of Home Automation Technology with Dynamic Pricing 15 / 20
Effects of PCT by TOU and VPP 1 1 critical critical peak peak 0.9 0.9 off peak off peak 0.8 0.8 0.7 0.7 0.6 0.6 cdf cdf 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 difference in control and pct coef. difference in control and pct coef. Welfare Effects of Home Automation Technology with Dynamic Pricing 16 / 20
Effects of All Three by TOU and VPP 1 1 critical critical peak peak 0.9 0.9 off peak off peak 0.8 0.8 0.7 0.7 0.6 0.6 cdf cdf 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 difference in control and all 3 coef. difference in control and all three coef. Welfare Effects of Home Automation Technology with Dynamic Pricing 17 / 20
For Which Households Will a PCT Cut 0.357 kWh pct vs portal, tou peak pct vs portal, vpp peak 1.5 1.5 1 1 0.5 0.5 0 0 -0.5 -0.5 pct pct -1 -1 -1.5 -1.5 -2 -2 -2.5 -2.5 -3 -3 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 portal portal Welfare Effects of Home Automation Technology with Dynamic Pricing 18 / 20
CDF of critical treatment effect 1 TOU VPP 0.9 0.8 0.7 0.6 cdf 0.5 0.4 0.3 0.2 0.1 0 -3 -2 -1 0 1 2 3 difference pct and portal coef. Welfare Effects of Home Automation Technology with Dynamic Pricing 19 / 20
In Conclusion Information can improve demand response to variable pricing, ◮ But response technology is much more important (for electricity) Home automation increases consumer welfare but largest gains are to the firm Significant heterogeneity in response can be described by historical usage data ◮ Demographics have little explanatory power Targeted distribution of technologies can avoid installations where the effects of the technology do not justify the cost. ◮ Increases the surplus per installation by 30% Integrating extensive historical behavior into field experiments could yield similarly efficient targeted treatments ◮ Medical treatments condition on limited observable characteristics, but digital daily health data could reveal unobserved differences in treatment effects Welfare Effects of Home Automation Technology with Dynamic Pricing 20 / 20
Recommend
More recommend
