ORIGINAL ARTICLE

Self-financed efficiency incentives: case study of Mexico

Anand R. Gopal & Gregory Leventis & Amol Phadke &

Stephane de la Rue du Can

Received: 22 June 2013 /Accepted: 26 March 2014# The Author(s) 2014. This article is published with open access at Springerlink.com

Abstract Numerous countries use public funds to sub-sidize residential electricity for a variety of socioeconom-ic objectives. These subsidies lower the value of energyefficiency to the consumer while raising it for the gov-ernment. Further, while it would be especially helpful tohave stringent Minimum Energy Performance Standards(MEPS) for end uses in this environment, they are hardto strengthen without imposing a cost on ratepayers. Inthis second-best world, where the presence of subsidieslimits the government’s ability to strengthen standards,we find that efficiency-induced savings in subsidy pay-ments can be a significant source of financing for energyefficiency incentive programs. Here, we introduce theLawrence Berkeley National Laboratory (LBNL)Energy Efficiency Revenue Analysis (LEERA) modelto estimate the greatest appliance efficiency improve-ments that can be achieved in Mexico by the revenueneutral financing of incentive programs from savings insubsidy payments yielded by the same efficiency im-provements. We analyze Mexico’s tariff structures andthe long-run marginal cost of supply to calculate themarginal savings for the government from applianceefficiency. We find that these avoided subsidy payments

alone can provide enough revenue to cover the fullincremental manufacturing cost of refrigerators that are29 % more efficient and televisions that are 36 % moreefficient than baseline models. For room air conditioners(ACs), the same source of financing can contribute up toone third of the incremental manufacturing cost of amodel that is 10 % more efficient than baseline. Weanalyze the sensitivity of our results to the most impor-tant parameters and find our main conclusion thatefficiency-induced avoided subsidy payments will con-tribute significantly to financing efficiency incentiveprograms in Mexico to be significant and robust.

Keywords Financial incentives . Energy efficiency.

Developing countries . Energy subsidies . Appliancemarket transformation .Mexico

Introduction

Electricity consumption subsidies are common in coun-tries around the world. While subsidies are found inOECD countries, the majority of subsidy programs arein developing countries, including the major emergingeconomies (Morgan 2008). In most of these countries,electricity and fuel subsidies were introduced as socialprograms to reduce the cost of energy for the poor(Komives et al. 2006). Hence, reducing or eliminatingsubsidies involves substantial political risk and is usu-ally not part of the energy policy dialogue (2010).Further, subsidies make it harder to introduce orstrengthen Minimum Energy Performance Standards

Energy EfficiencyDOI 10.1007/s12053-014-9263-9

Electronic supplementary material The online version of thisarticle (doi:10.1007/s12053-014-9263-9) contains supplementarymaterial, which is available to authorized users.

A. R. Gopal (*) :G. Leventis :A. Phadke :S. de la Rue du CanEnvironmental Energy Technologies Division, LawrenceBerkeley National Laboratory,1 Cyclotron Rd, Berkeley, CA 94720, USAe-mail: argopal@lbl.gov

(MEPS) for end uses, as greater stringency is frequentlynot cost-effective from the consumer perspective(Letschert et al. 2011).

Lawrence Berkeley National Laboratory (LBNL) isdeveloping the LBNL Energy Efficiency RevenueAnalysis (LEERA) model to help design incentive pro-grams that meaningfully improve appliance efficiencywith self-financing from efficiency-induced savings insubsidy payments. LEERA calculates the financial sav-ings that will accrue to the government from the deploy-ment of more efficient models for each type of appliance.It then draws on the product-specific technoeconomicanalyses of the Super-efficient Equipment and ApplianceDeployment (SEAD) Initiative,1 which calculate country-specific incremental manufacturing costs of higher effi-ciency appliance models. These incremental costs andgovernment financial savings are then compared to helpincentive program designers understand the amount ofself-financing that will be available at each level of appli-ance efficiency improvement. Where the amount of self-financing available is greater than the program costs, anappliance incentive program could be entirely financed byavoided subsidy payments. Themodel can support severaltypes of incentive program design.

The version of LEERA we present here can mosteffectively support policy in cases where revenue flowsfor electricity subsidies and end-use efficiency incentiveprograms are fungible, and decisions regarding fundingallocations for either option rest with the same authority(i.e., the Government ofMexico). These conditions holdtrue in a large number of developing countries becausethere are no established funding mechanisms for effi-ciency incentive programs nor too many restrictions onsources of financing for them (Sarkar and Singh 2010).Therefore, this version of LEERA cannot directly sup-port incentive program design in the USA and in manyEuropean countries where stricter policies govern thesources of financing for energy efficiency (Geller et al.2006), although it can still identify incentive opportuni-ties. In future improvements to the model, we plan toextend its applicability to cases common in the USA andEurope where decisions to invest in efficiency are madeby privately owned utilities and their government regu-lators (Geller et al. 2006).

In this paper, we analyze refrigerators, televisions(TVs), and room air conditioners (ACs) for residential

use in Mexico, a sector that receives generous nettaxpayer-funded electricity subsidies (Komives et al.2009). Our goal is to help Mexican regulators under-stand, precisely, the extent to which they can transformthe markets for these major end-use appliances if reve-nue from avoided subsidy payments was to be used tofinance incentive programs. The paper is structured asfollows. We first present an overview of energy subsi-dies and the theory of their impact on demand for energyefficiency. Next, we introduce and explain the LEERAmodel. This is followed by a presentation and discussionof results for Mexico and their implication for appliancemarket transformation and financing for incentive pro-grams. Finally, we discuss broader applications ofLEERA.

Energy subsidies and energy efficiency

Studies of global energy subsidies find that they aresubstantial and most are in developing countries(Morgan 2008). Globally, approximately $420 billionis spent annually on energy subsidies, making it one ofthe most subsidized sectors (Badcock and Lenzen 2010;Lewis 2012). Although most of these subsidies are forpetroleum, substantial support is directed toward elec-tricity consumption (Foster and Yepes 2006). In 2008,the UNEP estimated that the economic value of subsi-dies going to the electricity sectors in Russia, China,India, Saudi Arabia, and South Africa approached orexceeded $5 billion per year in each country (Morgan2008). Importantly, even though the stated goals of mostsubsidy programs are to reduce poverty, there is consid-erable evidence that they are not well targeted (Komiveset al. 2006).

Despite the massive amounts spent on subsidies,there is a paucity of data on energy subsidy programsat the country level. Studies have lamented the lack of aglobal or even OECD-wide inventory of programs(Badcock and Lenzen 2010; Gadgil and Anjali Sastry1994). Badcock and Lenzen undertook a comprehensivereview of subsidies for energy generation, but they didnot find a consistent definition of electricity subsidies, aconsistent method of accounting for them, or a consis-tent method for estimating them (Badcock and Lenzen2010). Even the European Union does not use a uniformevaluation method for each member country (Baconet al. 2010). Part of the difficulty in evaluating andanalyzing subsidies is the numerous forms that subsidies

1 To learn more about the SEAD Initiative, please visit http://superefficient.org

Energy Efficiency

can take including direct cash transfers, tax credits,rebates, accelerated depreciation, cross subsidies, pricecaps, subsidized loans, waived dividends, risk assump-tion, or delayed system maintenance (Komives et al.2005). Further, many countries, like India, have un-planned subsidies where government-owned utilitiesfrequently recoup their losses from the general fund onan ad hoc basis (Abhyankar and Phadke 2012).

Improving energy efficiency in subsidized regimes

From an energy policy perspective, subsidies causeoverconsumption of energy and lead to inefficient allo-cation of societal resources (2010). From an energyefficiency perspective, end-use electricity subsidies typ-ically make efficiency programs more challenging toimplement (Bouton et al. 2010). Even in the absenceof subsidies, society underinvests in energy efficiencydue to market failures like first-cost barriers, consumerinformation asymmetry, and environmental externalitiescaused by energy production and use (Jaffe and Stavins1994; Meier and Jollands 2007). Figure 1 shows thedeadweight loss resulting from these market failures ifelectricity is priced at the privately optimal marginalcost (PPRIV) instead of the socially optimal marginalcost (PSOC).

Electricity subsidies further increase this deadweightloss. Figure 2 shows a market in which the price toconsumers (PSUB) for electricity is reduced belowPPRIV due to subsidies. Electricity becomes even

cheaper compared to its socially optimal cost, resultingin even greater demand (QSUB). However, subsidiesmake energy efficiency more valuable to the govern-ment, which can decrease its subsidy burden by reduc-ing end-use energy consumption. From a theoreticaleconomic perspective, a rollback of subsidies wouldbe a first-choice energy policy (Komives et al. 2009).However, as we discuss earlier, such policies have prov-en to be politically challenging (Bacon et al. 2010). Thisis because the government is not just concerned byeconomic efficiency but also by political reality andthe need to maintain broad popularity. Financial incen-tives, on the other hand, are a politically feasible effi-ciency policy that can transform the market without anychanges to existing subsidy program design. In the nextsection, we describe the methodology, assumptions,inputs, and sources of data for the LEERA modeland explain how the model supports the design ofincentive programs self-financed by appliance effi-ciency improvements.

The LEERA model

The objective of the LEERA model is to calculate thesavings from avoided subsidy payments achieved byappliance energy efficiency and to treat these as reve-nues to finance incentives for the same efficient appli-ances. It does this by calculating the subsidy on themarginal unit of electricity consumed by a representa-tive appliance-owning household, multiplying that bythe annual energy savings from the deployment of a

Fig. 1 Deadweight loss due to social externalities from electricityconsumption without subsidized tariffs. MXN Mexican Pesos,PSOC socially optimal price, PPRIV privately optimal price, QSOC

socially demanded quantity, QPRIV privately demanded quantity,kWh kilowatt hours

Fig. 2 Additional deadweight loss due to electricity consumptionsubsidies. PSUB subsidized electricity price, MCSUB marginal pri-vate cost under subsidized pricing and demand, QSUB quantitydemanded under subsidized pricing

Energy Efficiency

more efficient appliance, and calculating the presentvalue of the associated subsidy payment savings overthe life of the appliance. Using this approach, we gen-erate a curve of government financial savings for eachpercentage improvement in appliance efficiency relativeto the baseline efficiency. We can compare this curve ofrevenues from avoided subsidy payments to varioustypes of incentive program costs. In this paper, wecompare the avoided subsidy revenue curve to the in-cremental manufacturing cost curve.

In its avoided subsidy payment calculations, LEERAonly includes subsidies that cover the difference be-tween retail price and long-run marginal cost of gener-ation because only this revenue is likely to be availablefor financing incentive programs. LEERA does notinclude efficiency-induced abatement of externalitiesin the avoided subsidy calculation because these arenot real streams of revenue unless policies that regulatefor externalities already exist. Hence, it can be arguedthat LEERA underestimates the overall subsidy burdenof the government, because the government will mostlikely have to bear the long-run costs of social andenvironmental externalities. LEERA also ignores otherforms of subsidies that are unplanned but frequent, likegrants or interest-free loans to the utility for systemupgrades, preferential tax rates to utilities, etc. Due totheir ad hoc, unplanned nature, these are hard to quantifyas efficiency-induced revenue streams available for in-centives. However, all of these forms of subsidy pay-ments will be reduced by improvements in applianceefficiency, which implies that we are underestimatingthe self-financing potential in our results.

First, the amount of money that the governmentavoids spending for each unit of electricity saved iscalculated. This is done by calculating the differencebetween the tariff at which electricity savings are real-ized and the supply cost. LEERA assumes that appli-ance efficiency savings occur at the consumer margin,and hence, the model uses the following equation tocalculate avoided subsidy payment per unit ofelectricity:

Avoided subsidy payment ($/kWh)=long-run mar-ginal cost of supply (LRMC) ($/kWh)−marginal tariffat which electricity savings occur ($/kWh).

The difference between MCSUB and PSUB in Fig. 2 isthe avoided subsidy payment calculated in the equationabove.

Next, LEERA multiplies this avoided subsidy pay-ment per unit by the annual electricity savings from

deploying appliance models that are more efficient thanthe baseline model. Baseline models are determined inthe SEAD technoeconomic analyses using market sur-veys and forecasts for each appliance in each SEADcountry (Park et al. 2011; Shah et al. 2013). Please referto the Electronic supplementary material, the SEADtechnoeconomic reports, and the LBNL report in sup-port of refrigerator MEPS revision for Mexico(Letschert et al. 2011; Park et al. 2011; Shah et al.2013) for more details on the baseline models we usein our analysis. LEERA then takes the present value ofthese annual financial savings over the life of the appli-ance to get the full value, to the government, of avoidedsubsidy payments at each level of improved efficiency.These subsidy savings are then compared to the incre-mental manufacturing costs of more efficient appliancemodels. It is important to note that incrementalmanufacturing costs of higher efficiency appliances tendto drop over time as demand grows (Dale et al. 2009).Hence, the self-financed efficiency improvement poten-tial could be greater than we estimated in this paper.

We also correct for rebound using estimates fromliterature (Davis et al. 2012; Gavankar and Geyer 2010;Maxwell et al. 2011; Nadel 2012). We apply an 11 %(0 % direct+11 % indirect) rebound for refrigerators andTVs and a 24% (13% direct+11% indirect) rebound forroom ACs. Note that by including indirect rebound, weare choosing an approach that is more conservative incounting energy savings than many other studies(Gavankar and Geyer 2010; Maxwell et al. 2011). Thisis another factor that reduces the self-financing potentialfor efficiency incentives we report in our results.

For example, a baseline refrigerator model inMexicouses 480 kWh per year (Letschert et al. 2011). Thus,switching to a 25 % more efficient model would yieldenergy savings of 107 kWh per year.2 We calculate thesubsidy for refrigerator use by a representative house-hold to be $0.14 per kWh, which translates to savedsubsidy payments of $13 per year. The net present valueof this revenue stream, at a real discount rate of 4 %,over the course of the refrigerator’s 15-year lifetime is$142. The incremental cost to produce a model that is25 % more efficient than the baseline model is $107(Letschert et al. 2011). Therefore, an upstream govern-ment incentive could be set at a level that covers 100 %

2 Twenty-five percent corrected for an 11 % rebound effect resultsin a 22.25 % actual savings. 480 kWh×22.25 %=106.8 kWhsaved per year.

Energy Efficiency

of the incremental cost of making a more efficientrefrigerator and still leave $35 in savings from avoidedsubsidy payments as a result of deploying the moreefficient model in place of a baseline model.

In this paper, we present results for refrigerators, roomACs (split style), and TVs.We plan to extend the analysisto other appliances and countries, as cost curves for eachare completed by the SEAD technoeconomic analyses.Baseline unit energy consumption (UEC) and incremen-tal manufacturing costs for room ACs and TVs are fromthe SEAD technoeconomic analysis (Park et al. 2011;Shah et al. 2013). For refrigerators, we use data fromLBNL’s analysis in support of harmonization ofMexicanand US refrigerator standards (Letschert et al. 2011). Thediscount rate we use is from the SEAD technoeconomicanalyses for each country (Shah et al. 2013).

Applying LEERA to Mexico

For this paper, we apply LEERA to the Mexican resi-dential electricity sector. The state-owned utility,Comisión Federal de Electricidad (CFE), provides allresidential electricity in Mexico. Rates are set by acomplex increasing block tariff (IBT) system in whichtariff zones are defined by average regional temperature(CFE 2012a). For electricity generation, fuel oil makesup 18 % of the electricity generation mix and usuallyoperates on the margin because it is the most expensiveform of generation (Sicilia Salvadores and HorstKeppler 2010; 2012a). In this section, we describehow LEERA calculates the long-run marginal cost(LRMC) of generation and the marginal tariff at whichsavings occur for each appliance in Mexico.

TVs and refrigerators have almost 100 % residentialpenetration rates (Davis et al. 2012; Komives et al. 2009).Hence, the LEERA model calculates marginal tariffs ofhouseholds that own these two appliances by taking theaverage, seasonally adjusted customer electricity con-sumption for each residential tariff zone and applying thetariff rate at that consumption level (CFE 2012a). Thesemarginal tariffs for each zone are then weighted by thezone’s proportion of all customers (CFE 2012b) to get anationally representative marginal tariff at which savingsfrom more efficient refrigerators and TVs will occur.

LEERA calculates the marginal tariff for room ACsdifferently because they are only present in 39 % ofhouseholds and their use is greater in hotter regionswhere consumers are subsidized more heavily. FromSEAD UEC data (Shah et al. 2013) and IEA data on

Mexican household share of AC energy consumption(Ellis 2009), LEERA calculates the minimum energyconsumption of an AC-owning household. The modelthen uses this consumption level to determine whichhouseholds in each tariff zone have ACs and their cor-responding marginal tariff. A nationally representativeACmarginal tariff is determined by taking an average ofthe marginal AC tariffs in each zone weighted by thenumber of AC-owning households in each zone. Formore details on the calculation of AC marginal tariffs,please refer to the Electronic supplementary material.

Given fuel oil’s significant share of the generation mix(18 %), a 64 % capacity factor (IEA 2012), and becausefuel oil generators almost always operate on the margindue to their high costs, it is sound to assume that theseplants will be the marginal generators for efficiencysavings. Even if the incentive programs we propose inthis paper are extremely successful, they are unlikely toreduce total Mexican electricity demand by 18 %, whichis the level necessary to completely avoid the need forfuel oil generators. Hence, it would be sufficient to justinclude the cost of fuel oil generation when calculatingthe LRMC in LEERA. However, because the share offuel oil generation is expected to gradually drop over thecoming two decades (Komives et al. 2009; 2012b), wemake the conservative assumption that 10 % of the long-run margin will be from substantially cheaper natural gasgeneration. This assumption reduces the revenues fromavoided subsidy payments calculated by LEERA andtherefore reduces the self-financing efficiency potentialin our results.

To calculate the cost of fuel oil generation, LEERAuses its opportunity cost: the international market price.This is necessary because the prices charged for fuel oilby the state-owned oil company Pemex to the state-owned utility CFE are not public. In any case, it is verylikely that Pemex would sell any fuel oil not needed forpower generation at market price, probably to interna-tional shipping companies that use Mexican ports. SincePemex is state-owned, this fuel oil sale revenue thatresults directly from improved appliance efficiency willaccrue to the government. We use the 2012–2022 aver-age of the reference forecast fuel oil price from theAnnual Energy Outlook 2013 to calculate the long-runvariable cost of generation (AEO 2013). Based on powerplant efficiencies, this translates to a variable generationcost of approximately US$0.22 per kWh (Badcock andLenzen 2010; Honorio 2003; Lewis 2012). We correctthis generation cost for transmission and distribution

Energy Efficiency

losses, approximately 17 % in Mexico (2012b), to cal-culate the LRMC of end-use delivered fuel oil electricity.The same procedure is used to calculate the LRMC ofnatural gas generation where we again use the AnnualEnergy Outlook 2013 reference forecast of natural gasprices for electricity generation (2013). Assuming a gen-erous power plant efficiency of 49 % (Honorio 2003;Sicilia Salvadores and Horst Keppler 2010), this trans-lates to a variable natural gas generation cost of US$0.04per kWh after adjusting for transmission losses. Sincefossil fuel prices tend to be volatile, we analyze thesensitivity of our final results to oil and natural gas pricesby including calculations for low and high EIA pricescenarios. Finally, we do not include any fixed costs inour LRMC calculation. If we were to include the fixedcosts, which are typically more than double the variablecosts for marginal generators, we would find greaterincentive self-financing potential in our final results.

For readers interested in more detail, please refer tothe Electronic supplementary material where we haveprovided the entire LEERAMexico analysis and data inspreadsheet form.

Results

We find that efficiency-induced savings in subsidy pay-ments could finance incentives that cover the entire in-cremental manufacturing cost of refrigerators that are29 % more efficient than baseline models. In the case ofLED-LCD TVs, the full incremental cost of models thatare 36 % more efficient than baseline models could befinanced with just half of the savings from avoided sub-sidy payments. For room ACs, revenue from avoidedsubsidies could finance an incentive that would coverabout one third of the incremental manufacturing cost ofa 10 % efficiency improvement (see Figs. 3, 4, and 5). Ofall the principal input parameters in this analysis, we findthat forecasts of oil and natural gas prices are the mainparameters with uncertainty levels high enough to signif-icantly change our results. Hence, we assess the sensitiv-ity of our results for each appliance and find that our mainconclusions are qualitatively robust (see Figs. 6, 7, and 8).We discuss our findings for each appliance below.

Refrigerators

The substantial market transformation potential we findfor Mexican refrigerators is due to three main reasons:

the large subsidies on each unit of refrigerator energyconsumption, relatively high annual UEC, and the longlife of the appliance. First, because refrigerators areowned by most households, rich or poor in all tariffzones, energy savings frommore efficient models occur,on average, at highly subsidized tariff rates, therebyyielding more monetary savings to the government.Second, the UEC for refrigerators is high with no directrebound. This is because refrigerators are alwaysplugged in and are in continuous operation. Thus, anyincrease in efficiency translates to substantial annualenergy savings. Finally, refrigerators have long lifetimes(15 years), so the large annual monetary and energysavings continue over a long time period. For thesereasons, we find that efficiency-induced subsidy savingscan yield revenue equal to the incremental manufactur-ing cost of refrigerator models that are 29 % moreefficient than baseline models (see Fig. 3). Figure 3 alsoshows that if an upstream incentive program was imple-mented that covered the full incremental cost differencefor every model up to 29%more efficient than baseline,the total income from subsidy savings would actually begreater than the expenditures on incentives. This netpositive revenue could be used to cover other costs(administrative, management, etc.) of implementingthe program.

LED-LCD televisions

Currently, LED-LCD TVs have low market penetrationin Mexico but are expected to constitute nearly 95 % ofthe national stock within a decade (Park et al. 2011).Almost all new purchases today are LED-LCD models(Park et al. 2011). Therefore, in our analysis, we choosean already efficient LED-LCD TVas our baseline mod-el. Further, TV UEC is substantially lower than refrig-erators yielding less annual revenue potential from sub-sidy savings. We still find the market transformationpotential for TVs to be greater than refrigerators fortwo reasons. First, two technologies, Dual BrightnessEnhancement Film (DBEF) and Local Dimming, sub-stantially increase LED-LCD TV efficiency at lowcosts3 (Park et al. 2011). Second, TV energy savingsalso occur at the same highly subsidized tariff rates as

3 For more information on DBEF and Local Dimming technolo-gies, please refer to the SEAD TVAnalysis that can be found here:http://www.superefficient.org/Activities/Technical%20Analysis/SEAD%20TV%20Analysis.aspx

Energy Efficiency

refrigerators since both appliances are similarly distrib-uted among households. Figure 4 shows that TVefficiency-induced subsidy savings can yield revenuethat is substantially greater than the incrementalmanufacturing cost of all super-efficient LED-LCDTV models. Note that we would see even greater poten-tial if we set the baseline TV model to reflect today’smarket average efficiency in Mexico. Such a baselinewould be appropriate to support the design of an earlyreplacement incentive program. In this case, however,we choose to show a result that can support the design ofan upstream program targeting only new purchases.Finally, it is important to keep in mind that furtherincreases in TV efficiency involve a fundamental tech-nological shift (i.e., OLED TVs) (Park et al. 2011), andhence, we should not assume that the two curves we seein Fig. 4 will continue to diverge at higher efficiency

improvements, as the costs of OLED TVs are veryuncertain at this time (Park et al. 2011).

Room ACs

Room ACs differ from TVs and refrigerators in twoways that reduce their market transformation potentialin this analysis. First, improved AC efficiency yieldsless than half the savings from avoided subsidy pay-ments when compared to TVs and refrigerators. This isbecause AC savings occur, on average, at less subsi-dized tariff rates. Second, studies have shown that ACefficiency improvements result in larger rebound, espe-cially for households in hot regions, although the effectis weaker in wealthier households (Davis et al. 2012).So, we use a rebound value for ACs of 24 % (13 %direct+11% indirect) based on estimates in the literature

$0

$50

$100

$150

$200

$250

$300

0% 5% 10% 15% 20% 25% 30% 35%

Su

bsi

dy

Sav

ing

s, M

anu

fact

uri

ng

Co

sts

(US

D)

Appliance Model Energy Efficiency Improvement Over Baseline

Revenues from avoided subsidy payments Incremental Manufacturing Cost

Fig. 3 Revenue potential fromavoided subsidy payments andincremental manufacturing costsfor refrigerators

$0

$5

$10

$15

$20

$25

$30

0% 5% 10% 15% 20% 25% 30% 35%

Su

bsi

dy

Sav

ing

s, M

anu

fact

uri

ng

Co

sts

(US

D)

Appliance Model Energy Efficiency Improvement Over Baseline

Revenues from avoided subsidy payments Incremental Manufacturing Cost

Fig. 4 Revenue potential fromavoided subsidy payments andincremental manufacturing costsfor TVs

Energy Efficiency

that account for these effects (Maxwell et al. 2011;Nadel 2012). Therefore, even with high baseline con-sumption and a relatively long life (582 kWh per year,12 years), Fig. 5 shows that AC efficiency-inducedsavings in subsidy payments do not yield enough reve-nue to cover the entire incremental manufacturing costof more efficient roomACmodels. However, Fig. 5 alsoshows that avoided subsidy revenue would be sufficientto contribute one third of the incremental manufacturing

cost of a model that is 10 % more efficient than thebaseline. Note that AC use is a much greater contributorto daily peak demand than refrigerators or TVs. Hence,inefficient ACs cost CFE, and therefore, the governmenta lot more than just electricity consumption subsidies inthe long run. Therefore, the long-run savings that resultfrom better power system planning due to improved ACefficiency could be a second source of financing for anAC incentive program.

$0

$20

$40

$60

$80

$100

$120

$140

$160

$180

$200

0% 5% 10% 15% 20% 25% 30% 35%

Su

bsi

dy

Sav

ing

s, M

anu

fact

uri

ng

Co

sts

(US

D)

Appliance Model Energy Efficiency Improvement Over Baseline

Revenues from avoided subsidy payments Incremental Manufacturing Cost

Fig. 5 Revenue potential fromavoided subsidy payments andincremental manufacturing costsfor room air conditioners

Fig. 6 Revenue potential sensitivity analysis for refrigerators

Energy Efficiency

Sensitivity analysis: model input parametersand uncertainty

As with any model, LEERA includes assumptions andchoices, both parametric and epistemic, the most

important of which we explain and justify in the earliersection describing the model. However, it is important toassess the sensitivity of our main findings to uncer-tainties in key input variables to see if our conclusionscould be fundamentally different. In order to decide

Fig. 7 Revenue potential sensitivity analysis for TVs

Fig. 8 Revenue potential sensitivity analysis for room air conditioners

Energy Efficiency

where to focus our sensitivity analysis, we compilea summary of key model input parameters, theirsources, our understanding of uncertainty in each,and the qualitative effect of changes in each param-eter on the market transformation potential of in-centives financed by efficiency-induced subsidysavings (shown in Table 1).

LEERA’s input parameters can be separated into oneof three categories based on the parts of the analysis thateach parameter influences. The three categories are thefollowing:

1. Electricity generation and supply cost calculation2. Calculation of tariffs at which efficiency savings

occur for each appliance3. Costs, use, and behavior related to each appliance

In the first category, the variables with the highestuncertainty are the price forecasts for fuel oil and naturalgas. There are low levels of uncertainty regarding thelong-run marginal fuel mix for reasons that we explainin the earlier section describing the model. Still, wechoose a conservative value for the fuel mix that dimin-ishes our revenue estimates so any sensitivity analysison this parameter will only further enhance our conclu-sions. In the second category, there is medium uncer-tainty in the ratio of summer to winter consumption for arepresentative household because this ratio is estimatedbased on the difference in IBT consumption thresholdsfor each season and not from empirical data (for moreinformation refer to the Electronic supplementary mate-rial). In the third category, there is a medium degree ofuncertainty in the magnitude of the rebound effect, butbecause we use a value close to the highest estimates ofrebound from the literature, sensitivity analyses on thisparameter will only show greater efficiency improve-ment potential. All the other parameters have low tonegligible uncertainty. Hence, from Table 1, we con-clude that uncertainties in fuel oil and natural gas priceforecasts are the most likely to qualitatively change ourconclusions.When we apply LEERA to other countries,we expect fossil fuel price forecasts to be the mostuncertain parameters in those cases as well.

We run our analysis for the low and high EIA pricescenarios for fuel oil and natural gas to assess the sen-sitivity of our results to these price forecasts (2013).Figures 6, 7, and 8 show that our main findings for eachappliance are qualitatively robust. Recall from the sec-tion describing the model that our choices on what to

exclude when calculating subsidy levels and the long-run marginal supply cost, the high values we use forrebound and the share of natural gas on the margin allreduce self-financing potential for efficiency incentivesfurther underlining the significance and robustness ofour findings. Even if prices of fuel oil and natural gasremain low over the next decade, we still see substantialrevenue-neutral market transformation potential for TVsand refrigerators. Conversely, if the high price projec-tions come to pass, we see 15 % revenue neutralefficiency improvement potential for room ACsand greater than 36 % improvement potential forrefrigerators and TVs.

Figure 6 shows that efficiency-induced subsidy sav-ings can still yield revenue equal to the incrementalmanufacturing cost of refrigerator models that are20 % more efficient than baseline models even if fueloil and natural gas follow the EIA low price scenario. Ifthese commodities follow the high price scenario,efficiency-induced subsidy savings will yield enoughrevenue to finance incentives for models that are 36 %more efficient than baseline. Hence, we find that thecase for a self-financed refrigerator incentive program inMexico is strong and robust.

Figure 7 shows that our results for TVs are evenmorerobust than for refrigerators. In all fuel oil and naturalgas price scenarios, the efficiency-induced subsidy sav-ings are greater than the full incremental cost of an LED-LCD TV that is 36 % more efficient than baseline.Hence, our analysis makes a strong, robust case that aTV incentive program can be largely self-financed bythe efficiency-induced subsidy savings in Mexico

Figure 8 shows that if the EIA’s low price scenario forfuel oil and natural gas materializes, it could completelyeliminate the Mexican government’s subsidy burden atthe tariff rate where savings fromACs occur. Hence, ACenergy savings will not yield revenue from savings insubsidy payments. However, if the high price scenariooccurs, we find that subsidy savings revenue wouldalmost equal the incremental cost of a model 15 % moreefficient than baseline. When this revenue is coupledwith the other benefits from AC efficiency improve-ments that we discuss in the AC result subsection,the Mexican government could finance an ACincentive program that can lead to significant mar-ket transformation. In summary, our conclusionthat the self-financing potential for AC incentivesin Mexico is small and significantly lower than forTVs and refrigerators is robust.

Energy Efficiency

Tab

le1

LEERAanalysisforMexico:

parameters,sources,anduncertainty

Parameter

Value

used

inanalysis

Source(s)

Uncertainty

(negligible,

low,m

edium,high)

Effecto

nLEERA’sestim

ateof

appliance

efficiency

improvem

entp

otential

Electricity

generatio

nandsupply

Long-runmarginalg

eneration

fuelmix

90%

Fueloil–10

%naturalg

asIEAMexicodata(2012a)

Low

Greater

shareof

fueloilincreases

efficiency

improvem

entp

otential

Fueloilp

rices

2012–2022Average

oftheAEO

2013

Reference

forecast

EIA

AnnualE

nergyOutlook

2013

(2013)

High

Higherfueloilp

ricesincrease

efficiency

improvem

entp

otential

Naturalgasprices

2012–2022Average

oftheAEO

2013

Reference

forecast

EIA

AnnualE

nergyOutlook

2013

(2013)

High

Highernaturalg

asprices

increase

efficiency

improvem

entp

otential

Power

plantE

fficiency

Fueloil4

4%,naturalgas49

%Honorio

(2003);S

iciliaSalvadores

andHorstKeppler

(2010)

Neglig

ible

Higherpower

plantefficiencydecreases

efficiency

improvem

entp

otential

Value-added

tax(VAT)rate

forresidentialelectricity

16%

Mexican

FederalL

awNegligible

HigherVATratedecreasesefficiency

improvem

entp

otential

Transmission

andDistribution

(T&D)losses

17.5

%EIA

Mexicopower

sector

data

(2012b)

Low

HigherT&Dlosses

increasesefficiency

improvem

entp

otential

Marginaltariff

Tariffrates

Increasing

blocktariffrateslisted

byCFE

CFE

tariffs(CFE

2012a)

Neglig

ible

Highertariffsdecrease

efficiency

improvem

entp

otential

Household

electricity

consum

ption

Seasonally

adjusted

CFE

Salesdata

CFE

Salesdata(CFE

2012b)

Low

Higherconsum

ptiondecreasesefficiency

improvem

entp

otential

Ratio

ofwinterto

summer

consum

ptionin

each

tariffzone

Rangesfrom

50to

17%

winter–83

%summer

(inhottertariffzones)

Calculatedfrom

CFEtariffdata

(CFE

2012a)

Medium

Highersummer

consum

ptionincreases

efficiency

improvem

entp

otential

Discountrate

3.81

%SE

ADtechnoeconom

icanalysis

reports

Low

Higherdiscount

ratedecreasesefficiency

improvem

entp

otential

Baselineunitenergy

consum

ption

(UEC)(kW

h)Refrigerator480/year;airconditioner

582/year;television(TV)102/year

SEADtechnoeconom

icanalysis

reports

Low

Higherbaselin

eUECincreasesefficiency

improvem

entp

otential

Rebound

Indirect–11%

forallappliances;

direct–refrigeratorandTV0%;air

conditioner

13%

GavankarandGeyer

(2010);

Maxwelletal.(2011);N

adel

(2012)

Medium

Higherrebounddecreasesefficiency

improvem

entp

otential

Appliancecosts,use,

andbehavior

Appliancelifetim

eRefrigerator15

years;airconditioner

12years;TVs8years

SEADtechnoeconom

icanalysis

reports

Low

Longerlifetim

eincreasesefficiency

improvem

entp

otential

Applianceefficiency

improvem

ent

costs

Increm

entalm

anufacturing

costsof

higher

efficiency

modelsof

each

appliance

SEADtechnoeconom

icanalysis

reportsandLBNLTSD

for

refrigerator

MEPS

Low

Higherincrem

entalm

anufacturing

costs

decrease

efficiency

improvem

ent

potential

Energy Efficiency

Discussion

The LEERA model can support financial incentive pro-gram implementation in a number of ways. It can showhow much, if any, appliance efficiency improvementscan be achieved through financing incentives withavoided subsidy payments at a zero or positive net cashflow impact to the government. In turn, this informationcan help inform incentive levels and incentive programdesign. For example, LEERA can calculate the appli-ance efficiency improvement potential from self-financing for upstream, midstream, or downstreamprograms. If program administrative costs areknown, it can estimate the efficiency improvementpotential and the share of self-financing that willgo toward these costs. We also plan to extendLEERA to quantify the efficiency-induced incen-tive financing potential from avoided additions togeneration capacity and reduced pollution from theenergy system. Importantly, we can apply themodel to quantify such self-financing potentialfor energy efficiency within the business modelsof privately owned utilities that are common in theUSA and Europe.

LEERA could also be used to support standards andlabeling programs. For example, LEERA’s calculationof efficiency-induced subsidy savings can be added toother consumer cost effectiveness metrics to calculatethe national cost effectiveness of strengthening MEPS.Where standards are in place, LEERA can be used tocompare existing MEPS with higher efficiency levelsthat could be obtained at zero net cash flow impact to thegovernment.

Finally, this model allows policymakers to com-pare and contrast the savings, both energy andfinancial, and the drivers of those savings, fordifferent end uses. In countries that subsidize res-idential electricity—those contemplating implemen-tation of financial incentive programs as well asthose with programs in place—LEERA can beused to help policymakers implement and improvefinancial incentive programs. We plan several im-provements to LEERA: developing the ability toanalyze the impacts and implications of peak con-sumption and cross subsidization and linkingLEERA with LBNL’s Bot tom Up EnergyAnalysis System (BUENAS) (McNeil et al. 2013)to estimate macro impacts of using avoided subsi-dies to finance incentives.

Conclusion

Many countries around the world, including a numberof emerging economies, subsidize electricity consump-tion, which promotes increased and inefficient energyconsumption. Countries that subsidize electricity oftenfind it politically difficult to lower or eliminate subsidiesand are frequently unable to strengthen MEPS for eco-nomic and political reasons. In this environment, gov-ernments have an opportunity to use efficiency-inducedsavings in subsidy payments to self-finance applianceincentive programs that improve end-use energy effi-ciency. The LEERA model supports the design of suchincentive programs down to the level of specific appli-ance models.

In the case of Mexico, we find that savings fromavoided subsidy payments can finance incentives thatcover the entire incremental manufacturing cost of re-frigerators that are 29 % more efficient than baselinemodels. In the case of LED-LCD TVs, the full incre-mental cost of models that are 36 % more efficient thanbaseline LED-LCDTVs could be financedwith just halfof the savings from avoided subsidies. For room ACs,revenue from avoided subsidies could finance an incen-tive that would cover about one third of the incrementalmanufacturing cost of a 10 % efficiency improvement.We assess the sensitivity of our results to key parametersand find the results for all three appliances to be robust.

Acknowledgments This work was funded by the Bureau ofOceans and International Environmental and Scientific Affairs,US Department of State, and administered by the US Departmentof Energy in support of the Super-efficient Equipment and Appli-ance Deployment (SEAD) initiative through the USDepartment ofEnergy under contract no. DE-AC02-05CH11231. We thank theentire US SEAD team. In particular, we thank Nihar Shah, WonYoung Park, Michael McNeil, and Virginie Letschert for theirassistance in this work. FromMexico, we thank Rodrigo Gallegosfor advising the development of the model.

Open Access This article is distributed under the terms of theCreative Commons Attribution License which permits any use,distribution, and reproduction in any medium, provided the orig-inal author(s) and the source are credited.

References

Abhyankar, N., & Phadke, A. (2012). Impact of large-scale energyefficiency programs on utility finances and consumer tariffsin India. Energy Policy, 43, 308–326.

Energy Efficiency

AEO. (2013). Annual Energy Outlook 2013. Washington, DC:U.S. Energy Information Administration.

Bacon R, Ley E, Kojima M (2010) Subsidies in the energy sector:an overview. World Bank

Badcock, J., & Lenzen, M. (2010). Subsidies for electricity-generating technologies: a review. Energy Policy, 38, 5038–5047. doi:10.1016/j.enpol.2010.04.031.

Bouton S, Creyts J, Kiely T, et al. (2010) Energy efficiency: acompelling global resource. McKinsey&Company

CFE (2012a) Tarifas-CFE. http://app.cfe.gob.mx/Aplicaciones/CCFE/Tarifas/Tarifas/tarifas_casa.asp. Dec 2012a

CFE (2012b) Estadística de Ventas-CFE. http://app.cfe.gob.mx/Aplicaciones/QCFE/EstVtas/. Accessed 2 Apr 2013b

Dale, L., Antinori, C., McNeil, M., McMahon, J. E., & SydnyFujita, K. (2009). Retrospective evaluation of appliance pricetrends. Energy Policy, 37(2), 597–605. doi:10.1016/j.enpol.2008.09.087.

Davis L, Fuchs A, Gertler P (2012) Cash for Coolers: how appli-ance replacement programs affect energy use. EnergyInstitute at Haas and National Bureau of EconomicResearch and the Haas School of Business at the Universityof California, Berkeley, Berkeley, CA

Ellis M (2009) Gadgets and gigawatts: policies for energy efficientelectronics. Paris, France

Foster V, Yepes T (2006) Is cost recovery a feasible objective forwater and electricity? The Latin American experience. TheWorld Bank

Gadgil, A. J., & Anjali Sastry, M. (1994). Stalled on the road to themarket: lessons from a project promoting lighting efficiencyin India. Energy Policy, 22, 151–162. doi:10.1016/0301-4215(94)90132-5.

Gavankar S, Geyer R (2010) The rebound effect: state of thedebate and implications for energy efficiency research. BrenSchool of Environmental Science and Management, SantaBarbara, CA

Geller, H., Harrington, P., Rosenfeld, A. H., et al. (2006). Policesfor increasing energy efficiency: thirty years of experience inOECD countries. Energy Policy, 34, 556–573. doi:10.1016/j.enpol.2005.11.010.

Honorio L (2003) Efficiency in electricity generation. Eurelectric:Union of the Electric Industry, VGB Powertech, Brussels,Belgium

IEA (2012) IEA Statistics: electricity information (2012 editionwith 2011 data). II.7,–IV.489, IV.499.

Jaffe, A. B., & Stavins, R. N. (1994). The energy-efficiency gap:what does it mean?Energy Policy, 22, 804–810. doi:10.1016/0301-4215(94)90138-4.

Komives K, Foster V, Halpern J, Wodon Q (2005) Water, electric-ity, and the poor : who benefits from utility subsidies? WorldBank, International Bank for Reconstruction andDevelopment, Washington, DC

Komives K, Halpern J, Foster V,Wodon Q (2006) The distribution-al incidence of residential water and electricity subsidies. 1–26.

Komives K, Johnson TM, Halpern JD, et al. (2009) Residentialelectricity subsidies in Mexico: exploring options for reformand for enhancing the impact on the poor. World Bank,International Bank for Reconstruction and Development,Washington, DC

Letschert V, McNeil MA, Lutz W, et al. (2011) Normas deDesempeño Energético Mínimo para Refrigeradores enMéxico: Análisis de Impactos Financieros y Energéticos.Lawrence Berkeley National Laboratory (LBNL), Institutode Investigaciones Eléctricas Prepared by order of theCollaborative Labeling and Appliance Standards Program(CLASP) for the Comisión Nacional para el Uso Eficientede la Energía (CONUEE)

Lewis M (2012) The missing link in the oil-price debate—declin-ing crude exports. 1–8.

Maxwell D, Owen P,McAndrew L (2011)Addressing the reboundeffect: final report. European Commission ENV DG, GlobalView Sustainability Services, France

McNeil M, Letschert V, Rue du Can S, Ke J (2013) Bottom–UpEnergy Analysis System (BUENAS)—an international ap-pliance efficiency policy tool. Energy Efficiency 6:191–217–217. doi: 10.1007/s12053-012-9182-6

Meier, A., & Jollands, N. (2007). Mind the gap: quantifyingprincipal-agent problems in the energy efficiency. (C. May,Ed.). Paris: OECD/IEA

Morgan T (2008) Reforming energy subsidies: opportunities tocontribute to the climate change agenda. United NationsEnvironment Programme, Division of Technology, Industryand Economics

Nadel S (2012) The rebound effect: large or small? ACEEESummer Study

Park, W. Y., Phadke, A., Shah, N., & Letschert, V. (2011).TV energy consumption trends and energy-efficiencyimprovement options. Berkeley: Lawrence BerkeleyNational Laboratory.

Sarkar, A., & Singh, J. (2010). Financing energy efficiency indeveloping countries—lessons learned and remaining chal-lenges. Energy Policy, 38, 5560–5571. doi:10.1016/j.enpol.2010.05.001.

Shah, N., Waide, P., & Phadke, A. (2013). Cooling the planet:Opportunities for deployment of superefficient room air con-ditioners. Berkeley: Lawrence Berkeley National Laboratory.

Sicilia Salvadores M, Horst Keppler J (2010) Projected costs ofgenerating electricity–2010 edition. IEA, Nuclear EnergyAgency, OECD, Paris, France

(2010) IEA, OPEC, OECD, World Bank joint report: analysis ofthe scope of energy subsidies and suggestions for the G-20initiative. 1–81

(2012a) IEA Energy Statistics—electricity for Mexico. http://www.iea.org/files/503/electricitydata.html. Oct 2012a

(2012b) Mexico power sector data, Energy InformationAdministration. http://www.eia.gov/files/484/country-data.html. Nov 2012b

Energy Efficiency