The Effect of U.S. Electricity Prices and Potential Carbon Pricing on the Purchase of

Energy- Efficient AppliancesPeter SchwarzProfessor of Economics, Belk College of Business and Associate, Energy Production and Infrastructure Center (EPIC) UNC Charlotte, Charlotte NC 28223-00001, USA andVisiting Professorship,China University of Mining and Technology, Xuzhou, China

Craig A. Depken, IIProfessor of Economics, Belk College of BusinessUNC Charlotte, Charlotte, NC 28223-0001, USA Michael HerronData ScientistPremier Healthcare Alliance13034 Ballantyne Corporate Place, Charlotte, NC 28277 Ben CorrellAnalystPricewaterhouseCoopers Advisory Services LLC1333 Main Street #30 Columbia, SC 29201

For presentation at the International Association of Energy Economics North American Meeting, Pittsburgh PA, October 25-28, 2015

Outline

• Introduction• Literature • Data & Empirical Approach• Results• Policy Implications• Conclusions

2 of 24

Introduction• ENERGY STAR • Introduced by USEPA in 1992, USDOE joined in 1996• voluntary labeling program intended to encourage

purchase of energy-efficient products• Provides information on energy savings for four appliances• refrigerators, room air conditioners, clothes washers,

dishwashers

• Stated Justification: • Reduce carbon emissions

3 of 24

Literature:

• Energy efficiency gap: consumers apply too high a discount rate.• Hausman (1979), Dubin and McFadden (1984), most recently

Parry, Evans, and Oates (2014).• Some studies dispute a gap.

• Allcott and Greenstone (2012), Francois Cohen, Matthieu Glachant and Magnus Soderberg (2014), Lance Davis, et al. (2014), Fowlie, et al. (NBER 2015)

• Renters less energy-efficient than owners.• Schwarz (1991), Davis (2008).

• Behavioral explanations: • Defective telescopic faculty, misperceptions, temptation and self-

control.• Pigou (1932), Parry, Evans and Oates (2014), Tsvetanov and Segerson (2014)

4 of 24

Literature

Other variables that affect market share of ES appliances

Attitude towards energy efficiencyACEEE index (Murray and Mills 2011)

RebatesUsed same ES data set (Datta and Gulati 2014)

5 of 24

Data: Sources• U.S. Energy Information Administration, a division of USDOE

• Appliance sales data from national retail chains, representing 70% of market

• Residential electricity prices• Rebates

• U.S. Census Bureau • Percent of housing units that are owner occupied • Percent of adults over age 25 with at least a bachelor’s degree

• U.S. Bureau of Economic Analysis• Per capita income

• American Council for an Energy-Efficient Economy (ACEEE) • State energy-efficiency score

All variables are at the state level for the years 2000-2009

6 of 24

Data: Summary Statistics 

All States 2000-2009 (500 obs.) 

State Means (50 obs.) 

Variable Mean Std Dev Min. Max. Std Dev Min. Max.ENERGY STAR market share

Refrigerator 29.02 8.21 10.54 57.21 3.03 42.28 36.50Dishwasher 59.39 27.01 3.90 99.00 2.76 53.17 65.03

Clothes Washer 28.88 13.84 3.26 60.04 6.01 18.18 42.21Air Conditioner 34.54 14.84 4.09 69.81 6.31 21.11 49.99

Incentives (2009 Dollars)

Refrigerator 3.75 12.67 0 85.18 9.85 0 47.47Dishwasher 2.33 9.11 0 53.21 5.95 0 26.33

Clothes Washer 3.86 14.68 0 113.57 9.83 0 54.06

Air Conditioner -- -- -- -- -- -- --

Residential electricity price (cents/kWh, 2009 dollars) 10.71 3.29 6.39 32.38 3.16 7.09 22.80

Per capita income (2009 dollars) 37,672 5,496 26,866 57,787 5,330 29,919 53,656

Percent of households owner-occupied 70.22 4.88 53.40 81.30 4.73 54.81 78.17

Percent of population with bachelor’s degree or higher 26.25 4.63 15.30 38.20 4.53 16.37 36.26

ACEEE Scores 14.85 10.33 0 50 10.01 0.67 41.17

7 of 24

Empirical Method: Base Model

(Percent ENERGY STAR)jit = β0 + β1 (Electricity Price)it +

β2 (Per Capita Income)it + β3 (Percent Owner Occupied)it +

β4 (Percent Bachelors)it + γXit + εit

• Results using state means from 2000 to 2009, j = appliance, i = state, t = year

• Xit is a matrix of additional control variables

• Regional dummies, incentives, and ACEEE score were added in alternative variants

8 of 24

Estimation Procedure

• Attempted to take advantage of panel aspect of data• Fixed and random effects models performed

poorly• Primarily because of very low

volatility in price of electricity within states.

• Because of the poor performance of the fixed and random effects models, we use• Between estimator, which uses sample

means for each state.

• Regression diagnostics reveal no problems with non-normal, heteroscedastic, or spatially autocorrelated errors.

•9 of 24

Little within-state variation for majority of states

10 of 24

Results  Base Model with ACEEE & Regional Dummy  DW CW RF AC

Electricity Price 0.282* 0.103 0.706*** 0.296

Per Capita Income 0.008 0.060 -0.042 -0.107

Percent Owner Occupied 0.142* 0.333** 0.288*** 0.370**

Percent with Bachelors 0.270*** 0.314 0.172* 0.264

South 0.384 -7.777*** -0.605 -5.521*

West 2.923** 0.759 2.864** -7.300**

Midwest 2.781** -2.259 1.352 -2.054

ACEEE Scores 0.088** 0.159** 0.081** 0.289***

Incentives -- -- -- --

Constant 37.553 -5.641 -4.304 2.081

Adjusted R2 0.649 0.679 0.733 0.538

11 of 24

Results  Base Model with Incentives & ACEEE

  DW CW RF ACElectricity Price 0.004 0.596* 0.599*** --

Per Capita Income -0.003 0.107 -0.020 --

Percent Owner Occupied 0.068 0.365** 0.250*** --

Percent with Bachelors 0.301** 0.383 0.153 --

South -- -- -- --

West -- -- -- --

Midwest -- -- -- --

ACEEE Scores 0.101** 0.155** 0.091** --

Incentives 0.038 0.212 0.079** --

Constant 45.184 -20.332 0.143 --

Adjusted R2 0.500 0.490 0.626 --

12 of 24

Results: Price Effects on RF, AC

Price coefficient is positive in 24/25 specifications/models

Statistically significant at 5% level in 11/25 variants.Refrigerators

Always positive and significant at 1% level.Air conditioners: Positive and significant for two of four specifications

Base model and specification with only ACEEE score included.

13 of 24

Results: Price Effects on DW, CW

DishwashersAt best only weakly statistically related

Statistically significant at 5% level only for the base model with regional dummies.

At the 10% level in only two other specifications.

Clothes WashersOnly significant in base model with incentives included

Insignificant at 5% level in all other models.

14 of 24

Results: Price Effects

Results agree with intuitionDishwashers:

Electricity savings not large enough to justify premium for more efficient appliance

Air conditioners:Quickest payback period and largest elasticity

Though still in inelastic range.

15 of 24

Results: Other Explanatory VariablesPercent owner-occupied housing positive and significant at 10% or better in 16/25 models

Significant at 5% in 9/12 models for RF and ACGenerally insignificant for DW

ACEEE scores positive, statistically significant at 5% or better for all appliances across the three models where it is included.Incentives positive and significant at 5% level for RF and CW when added to the base model

Positive but less significant when added to models including the ACEEE score or regional dummy variables.

16 of 24

Results: Demographic VariablesPercentage of population in state with at least a Bachelor’s degree generally positively related to ES market share.

Significant at 5% or better in 11/25 models across the four appliances.Significant at 5% in 9/12 models for RF and AC

Generally insignificant for DW

Regional dummies added to base modelSouth has lowest share of ES appliances relative to Northeast; West has greater market shares (except for room AC).

Coefficient on income is insignificant in all specifications

17 of 24

Results: Robustness ChecksResults are marginal changes in market shares over the ten year period 2000-2009.

Primary reason for using state sample means is that electricity prices change very slowly within states

Fixed and random effects estimators inappropriate and not well behaved.

How stable are parameter estimates across distribution of market shares?

More sensitive to changes in electricity price at lower or upper end of price distribution?

Quartile regression applied to all the specifications shows all parameter estimates are very stable.

18 of 24

Results: Robustness Checks (cont.)Clustered standard errors by region of country

Very little change in significance.Clustered standard errors based on ACEEE score regardless of where state was located within the country

Again, very little change in parameter estimates.

19 of 24

Policy Implications : Energy Reduction Due to Electricity Price

• Elasticities for base model are 0.34 for AC, 0.21 for RF, 0.05 for CW, and 0.04 for DW.• Given these relatively inelastic responses, even a large

increase in electricity prices might not increase market shares by very much. • Resources for the Future estimates that a carbon price

would increase electricity price by at most 4 cents/kWh. • Based on 2009 data, market share for ES room air

conditioners would increase from 41.4% to 46.2%, for ES refrigerators from 33.4% to 36.0%, for ES clothes washers from 37.0% to 37.6%, and for ES dishwashers from 79.3% to 79.7%.

20 of 24

Policy Implications: Energy and Carbon Reductions Before and After Carbon Price• Based on 2009 data, the decrease in energy use from the four ES

appliances including the 4 cents/kWh carbon price is almost 2,000,000 MWh/year.• Just over 100,000 MWh of the reduction is due to the

4 cent/kWh carbon price. • Using the approximation that each MWh of electricity

generated emits 0.5 metric tons of carbon, ES appliances reduce C emissions by just under 1,000,000 metric tons• Close to 50,000 MWh due to the 4 cent/kWh C tax.

21 of 24

Policy Implications: Carbon Reductions Before and After Carbon PriceAccording to the EIA (2012), the U.S. emitted 5.290 billion metric tons in 2012. • The total percentage reduction is approximately 0.02% per

year, or 0.2% over ten years due to the carbon price, assuming average appliance life. • The annual reduction in carbon emissions is the

equivalent of taking just over 200,000 cars off the road, based on the U.S EPA estimate that the average automobile emits 4.7 metric tons of carbon per year.• Of this total, about 50,000 metric tons are due to the

carbon tax• Equivalent of a little over 10,000 cars per year

22 of 24

Takeaway Points

• In all, the ES program has a modest effect on energy use • And a more modest effect on carbon

reductions• A carbon tax would have an even smaller

marginal indirect contribution through the ES program.

23 of 24

Thank you

24 of 24