1  

Energyconsumptionandurbansprawl:evidencefortheSpanishcase

Abstract

Electricity accounted for 41,2% of the total final energy consumption ofSpanishhouseholdsin2014.InSpain,electricityisstillgeneratedfromfossilfuels,whichaffectsthelevelofgreenhousegasemissionsandSpanishenergydependency. Therefore, policies oriented towards reducing householdelectricityconsumptioncouldhelptoachieveamoreenergy‐sustainableandlow‐carbon economy. Some of those policies seek energy building savingsfrom energy rehabilitation of the thermal envelope, the renewal ofappliances,ortheconstructionofnearly‐zeroenergybuildings.However,therelevance of urban structure andurbanphenomena, such as urban sprawl,hasnotnormallybeenconsideredforhouseholdelectricitysavings.

This paper focuses on the impact of urban design and urban sprawl onhouseholdelectricityconsumption.Thus,anelectricityconsumptionmodelisestimated by using urban characteristics as its determinants, such us thelevelofurbanagglomerationor if the families live inadetached, sprawlinghouse,andalsohouseholdsocioeconomicvariables.Themodel isestimatedby using 2014 Household Budget Survey (HBS) microdata and applyingOrdinary Least Squares as well as quantile regressions as econometricprocedures.

Results confirm that living in a detached house significantly increases theelectricityconsumptionwhileurbanagglomerationshavetheoppositeeffect.SprawlisoccurringrapidlyinSpanishcitiesand,accordingtoourresults,itcould constitute a main source of increase in electricity demand in thefollowingyears.Thispapercommentsontherelevanceofurbanpoliciesandurbanplanningfromtheperspectiveofhouseholdenergyefficiency.

Keywords: household energy consumption, house characteristics, urbansprawlandurbanenergyefficiency.

2  

1.Introduction

Currently,55%oftheworld’spopulationlivesinurbanareas,aproportionthatis

expected to increase to66%by2050and85%by2100. In a rapidlyurbanizing

world, the way in which cities are planned, built, operated and redefined has a

hugesocialandeconomicimpact.Urbanpoverty,inequalities,affordablehousing,

mobilityandcongestionareattheheartoftraditionalurbanchallenges.However,

the environmental problems of pollution, increased global energy consumption

and climate change as well as the energy dependency of many countries draw

attention to sustainable urban growth and, particularly, to the understanding of

energy consumption patterns in cities, and drive innovation in urban energy

savingsandgreenbuilding.

Urban energy efficiency engages a complex and nuanced spectrum of issues. In

recent years, one of themain focuses of innovation in cities was placed on the

capacity of new technologies to develop “smart cities”, those that aremonitored

withnetworksofdevices that giveprecise informationaboutpollutionand local

mobility,allowingabetterunderstandingofcitylifeandproblems.Manyscientists

are also investigating the use of new materials and building technologies to

improve housing energy performance. By contrast, the relevance of energy

consumption of urban phenomena, such as urban sprawl, is still not widely

studied.

Thefirstcitiestobegintoexperienceurbansprawlweretheemergentcitiesofthe

centralandwesternUnitedStates.Thismodelofasprawlingcityrapidlyextended

firsttoLatinAmerica(seeGilbert1996andPolèseandChampain2003)andlater

to Asian cities (see Bunnel et al., 2002), finally becoming a global phenomenon.

Traditionally, most old European cities grew differently from the new cities of

AmericaorAsia:citiesontheoldcontinentwerestronglyconcentratedarounda

denselypackedhistoricalcentreand itscommercialandbusinessextensionsand

usually adhered to a monocentric growth model with a strong centre and

hierarchicalstructureofsub‐centres.However,urbansprawlinEuropehasgrown

inmanycasesoverthelastfourdecades(seeCouchetal.,2007).Accordingtothe

EuropeanCommission (2006), countries in the east and southofEuropeare the

onesmostatriskofanexplosiveprocessofurbansprawl.

3  

The caseof Spain is oneof themost interesting inEurope. In someareasof the

Iberian Peninsula there is very high building pressure due to tourism and the

demandforasecondresidence.TheSpanisheconomyhasbeendrasticallyaffected

bytheconstructionsectorsufferingoneofthebiggestrealestatebubblesofallof

Europe(Romero,2012).Spainhadveryfasteconomicgrowthduringthelastfour

decades of the past century, presenting a very strong, concentrated process of

urbanization.CitiessuchasMadridandBarcelonadoubledtheirpopulationinless

thantwentyyears.Otherbigmetropolitanareasof thecountryexperiencedsuch

growththatdifferentcitiesor townsgrewintoone.Ruralareasalso lostmostof

their population in just two decades. The strong changes in income per capita,

social customs and land use pressures make Spain a victim of sprawled urban

growth(Rubieraetal.,2016).

Inthispaper,weareparticularlyinterestedinmeasuringandevaluatingtheeffect

of increasing urban sprawl in Spanish cities on residential energy consumption.

Residentialenergyconsumptionconstitutesoneof the largestsourcesofSpanish

finalenergydemand,19%in2014accordingtotheEuropeanCommission(2016).

Thus,wearegoingtofocusonelectricityconsumptionasitisthelargestsourceof

householdenergydemand.Electricityaccountedfor41,2%ofthetotalfinalenergy

consumptionin2014withanincreasingtrend(householdelectricityconsumption

increased21,8%from2004to2014,seeEuropeanCommission,2016). Itshould

benotedthatnearly72%ofSpanishelectricity isstillgeneratedfromfossil fuels

(Spanish Government, 2016), which affects the level of greenhouse gas (GHG)

emissions and Spanish energy dependency (European Commission, 2016).

Therefore, household electricity saving achieves amore energy‐sustainable, low‐

carbonandclimatefriendlyeconomy.Toreducehouseholdenergyconsumption,it

isimportanttoanalyseitsdeterminants.Mostoftheexistingempiricalworkshave

used localclimate,homeappliances,householdsizeorhouseholdcharacteristics,

amongother aspects, as determinants of energy consumption forhouseholds. In

contrast,thispaperfocusesthispaperfocusesontheimpactofurbandesignand

urbansprawlonhouseholdelectricityconsumption.

The phenomenon of urban sprawl has been studied within different disciplines

(geography, urban planning, the environment, economics, sociology and even

4  

public health) and from very different standpoints, which has led to numerous

definitions. Glaeser and Kahn (2004) and, more recently, Jaeger and Schwick

(2014) have compiled one of the most complete reviews of leading papers on

urbansprawlanditsconsequences.Althoughallthestudiesconsiderurbansprawl

asacomplexphenomenonhavingmanydimensions, theyallcoincide inthe idea

that sprawled cities alwaysdenote the extent of the area that is built up and its

dispersioninthelandscape,themoreareabuiltoverandthemoredispersedthe

buildings,thehigherthedegreeofurbansprawl.Thus,fromtheindividualpointof

view(household)sprawlmeansthepredominanceofdetachedhouses insteadof

the building of apartments and low building density. Thus, from the energy

consumptionperspective, theeffectof sprawlcouldbe identifiedby theeffectof

livingindetachedhousesinlessintenselyagglomeratedareas.

According to statistical data from the Household Budget Survey of the National

StatisticalInstitute,in2014,approximately35%ofthepopulationofSpainlivedin

houses,with11% living indetachedhousesand24.2% in semi‐detachedhouses

(INE, 2014). The remaining percentage of the population is distributed among

other types of houses, such as flats.What we propose in this paper is to study

energy consumption factors of households, including in the analysis the urban

environment (urbansizeandurbanversus ruralareas)and the typeofhouse to

identify how they affect energy consumption. The obtained results could orient

state and local governments in energy efficiency strategies of urban planning. A

more electricity efficient urban design can not only decrease the energy

dependencyoftheSpanisheconomyandachieveenergysavingsbutalsoimprove

theenvironmentbyreducingGHGemissionsandhelpto fulfil theEU's2020and

2030climateandenergygoals(CounciloftheEuropeanUnion,2010,2014).

The paper is structured as follows. In the next section,we review the empirical

modelsofenergyconsumptionand theirpreviousconclusions, focusingon those

papers thatparticularly include in someway theurban/rural effect andhousing

characteristics. From this literature review we obtain the household electricity

model specification to be estimated. Section three revises the dataset, the

HouseholdBudget Survey (INE,2014) and specific characteristicsof theSpanish

case.Theestimationstrategyandmainresultsarepresentedinsectionfour.The

5  

main conclusions are summarized in the final section, along with a number of

recommendationsregardingpolicy.

2. Household energy consumption in urban areas: literature review and

empiricalmodelproposal

Theeffectofseveralvariablesonhouseholdenergyconsumptionhasbeenwidely

studied.Previousstudieshavefoundthathouseholdenergyconsumptioncouldbe

explained by various factors, including household economic characteristics

(Pachauri 2004, Druckman and Jackson 2008, Cayla et al. 2011, Brounen et al.

2012, O'Neill and Chen 2002, Santin 2011, Rahut et al. 2016), household

sociodemographicfactors(Kaza2010,Miahetal.2011,Rahutetal.2014,Joneset

al.2015,Kelly2011,Brounenetal.2012,Kaza2010,O'NeillandChen2002,Özcan

etal.2013),physicalcharacteristicsofthehouseorconstructionquality(Daioglou

etal.2012,Kavousianetal.2013,Valenzuelaetal.2014),homeappliances(Stiri

2014,Huang2015),residential location(Fengetal.2011,DruckmanandJackson

2008,Daioglouetal.2012),localclimatefactors(Kavousianetal.2013,Valenzuela

etal.2014),andenergycosts(LarsenandNesbakken2004,Kasparian2009,Niuet

al.2016).However, thenumberofstudiesconsideringhousing typeand/or local

degreeofurbanization(rural‐urban)asadditionaldriversofenergyconsumption

has not been extensive (Wiesmann et al. 2011,Heinonen and Junnila 2014, Stiri

2014,Huang2015).

For example, Wiesmann et al (2011) estimated electricity consumption at the

municipality level for 2001 and at household level by using the Portuguese

consumer expenditure survey that was collected in 2005 and 2006. They used

socioeconomic and demographic household characteristics, but also housing

characteristicsasexplanatoriesvariables.Regarding thesecondsones, theyused

dwellingtype(detachedhouse,semidetachedhouse,smallapartmentbuildingand

large apartment building) and the level of urbanization (rural, half‐urban and

urban).TheOrdinaryLeastSquaresestimationsatbothlevelswereconsistentand

indicatethatanapartment inabuildingconsumed less thandetachedhouse,but

alsourbanhouseholdsconsumesmoreelectricitythanruralhouseholds.

6  

Heinonen and Junnila (2014) studied Finland residential energy consumption

patternsindifferenthousingtypes(apartmentbuildings,row‐terracedhousesand

detachedhouses)anddegreesofurbanization(cities,semi‐urbanareasandrural

areas). They found significant behavioural differencesbetweendifferenthousing

types and rural‐urbanmodes. In fact, they found lowerenergyuse in apartment

buildingscomparedtodetachedhousesandlessintensiveenergyconsumptionfor

all types of houses in rural areas. In this study, Heinonen and Junnila (2014)

holisticallyanalysedtheresidentialenergyconsumptionindifferentbuildingtypes

byusingtheHouseholdBudgetSurveyfor2006.Thus,theyexplicitlyobservedthe

residential energy consumption patterns of 3984 households and compared

explicitlybybuilding type,butnoeconometricmodelwasestimated, thus itwas

notpossibletogeneralizetheresults.

Stiri(2014)analysedtheimpactofhouseholdcharacteristicsandbuildingphysical

attributes(housingtypeandsize)onresidentialenergyconsumption intheUSA.

He used information provided by the Residential Energy Consumption Survey

(RECS)for11,590householdsin2009.Thecategoriesofhousingtypeconsidered

weresinglefamilydetached,singlefamilyattachedapartmentswith2–4unitsand

apartments with more than 5 units. By using a structural equation model he

demonstrated that the impact of housing type on energy consumption is higher

thanthecorrespondingdirectimpactfromhouseholdcharacteristics.

Huang (2015) employed a quantile regression to analyse the determinants of

household electricity consumption in Taiwan over the period 1981‐2011. The

information was provided by Taiwan's Family Income and Expenditure Survey,

where 15,000 households were sampled each year. In regard to housing type,

Huang(2015)foundthatlargerhousingareasandmulti‐floorhousescontributed

to higher household electricity consumption and that urban households showed

higherelectricityusethandidruralhouseholds.

In general, existing empirical studies seem to agree that energy consumption is

lower for apartment buildings than for detached houses; however, the results

relatedtoruralversusurbanareambiguous:someof themconcludethatenergy

use seems to be lower in rural areas than in urban areas (Poumanyvong and

Kaneko2010,Niuetal.2012,HeinonenandJunnila2014,Huang2015)andothers

7  

find theopposite (Normanet al.2006,Rickwoodet al. 2008).However,manyof

theconcretefindingsarecountryspecific,andtheresultsalsorelyonthemodel,

thestudiedenergytype,thestudiedperiod,theconsideredexplainedvariablesor

theusedeconometrictool.

Thispaperanalyseshouseholdelectricityconsumptionfor2014inSpain.Forthe

Spanishcase,toourbestknowledge,onlyLabandeiraetal.(2006,2012),Blázquez

et al. (2013) and Romero‐Jordán et al. (2014, 2016) have analysed household

electricitydemandbyusingasitsdeterminantsnotonlyhouseholdsocioeconomic

anddemographiccharacteristicsandclimatefactorsbutalsosomevariablerelated

tothelocationofthehouse.

Labandeiraetal.(2006)estimated,forthefirsttimeinSpain,aresidentialenergy

demand system using 51,691 household annualmicrodata for the period 1975–

1995, by applying an extension of the Almost Ideal DemandModel (Deaton and

Muellbauer, 1980). They also estimated the model by considering types of

municipalities(rural,village,city).Labandeiraetal.(2012)estimatedthedemand

model using monthly data for the period 2005‐2007 with information from

422,696households, by applyingapanel randomeffectsmodel.Theyperformed

estimations for different regions of the country. Blázquez etal. (2012), by using

aggregatepanel data at theprovince level for47 Spanishprovinces, estimated a

log–log demand equation for electricity consumption using a dynamic partial

adjustment for the period from 2000 to 2008. Romero‐Jordán et al. (2014)

analysed the determinants of household electricity demand with a panel data

partial adjustment model of the Spanish regions between 1998 and 2009.

Furthermore, Romero‐Jordán et al. (2016) estimated these determinants by

applyingaquantileregressionmethodusingdatafrom2006–2012.Asexplanatory

variables related to housing location, they used the region and the type of

municipality(ruralorurban)wherehousingissited.

Asshown,theabovementionedliteraturehasconsideredinsomewaythelocation

of households in Spain as electricity consumption determinants, but the urban

sprawldeterminanthasnotyetbeenspecificallyconsidered.Thispaperfocuseson

theimpactofurbansprawlonSpanishhouseholdelectricityconsumptionin2014.

Apart from the impact of household socioeconomic characteristics and physical

8  

characteristics of the dwelling, we also studied the urban environment and

housingtype(detachedindependenthouseorpartofabuilding).Finally,asprior

studies (Kaza 2010, Valenzuela et al. 2014, Huang 2015, Niu et al. 2016, and

Romero‐Jordán et al. 2016 for the Spanish case) suggested that explanatory

variables may exhibit variation in magnitude and sign depending on household

energy use quantile, the empirical study is extended by estimating a quantile

regression model. The obtained information could be very useful for designing

specificenergyefficiencymeasuresongroupswithhigherelectricityconsumption.

3.TheSpanishcase:datasetandselectedvariables

The data used in this analysis are obtained from the Household Budget Survey

(HBS)oftheNationalStatisticalInstitute(INE),asurveythatprovidesinformation

about the patterns of consumption, income and other socioeconomic and

demographic characteristics of Spanish households. The dataset is formed from

21,790observations that aredisaggregatedacross the17 regionsat theNUTS‐II

level.This survey is obtainedyearly since2006. In thispaper, themost recently

availableinformationof2014isused.

The HBS is composed of three data files: (i) the Household file, which provides

informationabouttheAutonomousCommunitythatthehouseholdbelongsto,the

citysize,thepopulationdensity,theincomelevel,numberofmembers,numberof

employees, some characteristics about the household head such as age, sex,

education level, marital status, among others, and some information about the

characteristicsofthehouse,suchasthesizeinm2,theconstructionyearor,among

others, if it is a detached house, a semidetached house or an apartment in a

residentialbuilding;(ii)theExpenditurefile,whichrepresentsallhouseholdswith

any expenditure; and, (iii) the Household members file, which contains all the

informationabouteachmemberinthehousehold.

As an explicative variable of our analysis,we use the electricity consumption in

terms of kW (LELECTRICITYQ, in logarithm). In accordance with the literature

quoted in theprevioussection,weuseastandardsetofvariables toexplain this

electricity consumption assuming the limitations of the database. The set of

selectedvariablesissummarizedinTable1.

9  

A particularly relevant variable in any consumption analysis is the price of the

product.TheHBSgiveusanespeciallyprecisewayofobtainingtherealfinalprice

thatthefamiliesarefacingfortheirelectricityconsumption.Householdsnotonly

reporttheirexpendituresforeachgoodbutalsothephysicalamounttheybought.

Therefore, individual prices atwhich households purchase the electricity can be

recovered by dividing monetary expenditures on it by physical quantities (kW)

consumed. In accordancewith thedemand system literature (Deaton, 1988),we

refertothese“prices”asunitvalues.Theempiricalexperiencedemonstratedthat

unitvaluesareveryusefulasanindicatoroftimeandspatialpricevariations.The

LELECTRICITYP variable is the logarithm of this unit value obtained for each

household.

Although all households in Spain need to use electricity for certain purposes, in

somebuildings,itispossibletouseotherenergysourcessuchasgas,petrolorcoal

forheatingorhotwater.Thedatabasegivesusinformationtoproposeproxiesthat

inform us about the consumption of these other energy sources. Using this

information,wedefinethedummiesHEATINGandHOTWATERtotakeavalue1if

thehouseonlyusedelectricity forall theenergyconsumptionand0 if thehouse

used other energy sources for heating, in the first case, and hot water, in the

secondone.

To complete this basic information with socioeconomic or housing observable

characteristicsweadd to themodel variables such asHOUSEAGE, a dummy that

takes a value 1 if the house is older than 25 years and 0 otherwise, INCOME, a

categoricalvariablethatgivesusinformationabouttheincomeleveloftheheadof

the family, FAMILYMEMBERS, a variable that informs us about the number of

membersof the family,andHOUSE,adummyvariable, that takesavalue1 if the

family lives inadetachedisolatedorsemidetachedhouseand0 if they live inan

apartmentorflatinaresidentialbuildingorsimilarconstruction.

For thepropose of this researchwe focus special attention on this last variable,

HOUSE,because it informsusabouttherelevanceof living inadetachedisolated

house typical in sprawled cities in comparisonwith the usual type of houses in

compacted cities, flats in buildings. We complete this variable with others that

provide complementary information about the urban environment. URBAN is

10  

anotherdummythattakesvalue1whenthehouseislocatedinamunicipalitywith

more than 100,000 inhabitants and 0 otherwise. Although the limit of 100,000

inhabitants is not themost proper to delimit an agglomeration in the particular

caseoftheSpanishurbansystem,thisboundaryisabletodelimitquiteprecisely

thelargermunicipalitiesandmaincitiesofthecountry(seePolèseetal.,2009).We

alsoincludespatialvariablesthatallowustomeasuretheregionalandlocaleffect.

In particular, we add a dummy variable of the regional zone (NUTS2 level of

desegregation)inwhichthehouseislocated(variableREGION).

>>>INSERTARROUNDHERETABLE1<<<

4. Estimation strategy: a two‐stage estimation method and quantile

regression

Basedontheaboveinformation,thebasicempiricalmodelthatweproposecanbe

writtenas

[1]

where LELECTRICITYQ is our dependent variable, HOUSE and URBAN are the

variablesthatwefocusouranalysisontoobservetherelevanceofurbansizeand

typeofhouse,andXisavectorofallcontrolvariablesnormallyconsideredinthe

literature: LELECTRICITYP, HOTWATER, HEATING, INCOME, OWNERSHIP,

HOUSEAGEandREGION.Finallyuistherandomerrortermoftheestimation.

Itshouldbenotedthathousingsocioeconomiccharacteristicscouldinfluencenot

onlyhouseholdelectricityconsumption(Rahutetal.2016)butalsohousingtype

choice.Forexample,havingalargefamilysizeorahighincomecouldincreasethe

likelihoodof living inadetachedhouse(Beguin,1982;Boehm,1982).Therefore,

anendogeneityproblemcanemergeifthedecisiontoliveinadetachedhouseis

partiallyaffectedby theanticipatedenergyconsumption,aswell asby theother

repressors included in the model, making the OLS estimator inconsistent. To

address this problem, the dummy of detached is instrumented by defining a

variable,observableintheHBS,associatedwithdetachedbutnotcorrelatedwith

theenergyconsumption.

11  

Using this instrumental variable, we propose a two‐stage estimation method

(2SLS).Thedecision to live inadetachedhouseornot is instrumentedbyusing

twoadditionalexogenousvariables:

‐ RURAL:whichindicatesifthehouseissituatedinaruralarea.Itshouldbe

remarked that this “rural”classification is related to theenvironment,not

the population. The HBS classifies the households situated in industrial

rural settings, agricultural rural settings and fishing rural settings;

conversely,therearehouseholdslocatedindifferenturbansettingssuchas

luxury urban, semi‐luxury and poor urban settings. With all this

information adummyvariable is constructed and takes the value1 if the

householdsarelocatedinaruralsettingand0otherwise.

‐ LFUELEXP: a continuous variable defined as the logarithm of the

expenditureinEurosoffuelusedfortransportmotorssuchascars.Thisis

clearly related to the type of house, a detached isolated house normally

implies higher expenditure in fuel but is independent of electricity

consumption.

Aftercontrollingtheendogeneityproblembythisprocedureanadditionalissueis

thatweareinterestedinobservingiftheconsumptionpatternschangedepending

onthedistributionofconsumption.Thisispossibleusingthequantileregressions

(QR)approach(KoenkerandBasset,1978),whichfitsquantilesofconsumptionof

electricitytoalinearfunctionofcovariates.Initssimplestform,theleastabsolute

deviationestimatorfitsmedianstoalinearfunctionofcovariates.Themethodof

quantile regression is more attractive because medians and quantiles are less

sensitive to outliers than means, and therefore ordinary least squares (OLS).

Indeed, the likelihood estimator is more efficient than the OLS one. Quantile

regressionsallowthatdifferentsolutionsatdifferentquantilesmaybeinterpreted

as differences in the response of the dependent variable to changes in the

regressors; thus,quantileregressionsdetectasymmetries in thedatathatcannot

be detected by OLS. However, the most important feature is that quantile

regression analyses the similarity or dissimilarity of regression coefficients at

different points of the dependent variable, which in this case is the electricity

12  

consumption; it allows one to consider the possible heterogeneity across the

intensityinelectricitydemand.

Thefollowingexpressionpresentstheadaptationofequation[1]intermsofQR:

[2]

wherecoefficients representthereturnstocovariatesattheτthquantileofthe

logarithmofelectricityconsumption.

QRestimatescanbeaffectedbythesameendogeneityproblemsdiscussedforthe

caseofOLS.Thisissueisaddressedbyapplyingtheinstrumentalvariablequantile

regression (IVQR) estimator developed by Chernozhukov and Hansen (2005,

2006).TheIVQRestimationisbasedonthesameinstrument,thevariableRURAL,

asthe2SLSestimation.

Themodel isestimatedbyusingthe least‐absolutevalueminimizationtechnique

andbootstrapestimatesoftheasymptoticvariancesofthequantilecoefficientsare

calculatedwith20repetitions.

5.Mainresults

ThefinalresultsofalltheseestimationstrategiesaresummarizedinTable2.The

firstcolumnofthetablepresentstheresultsofthe2SLSprocedure.Thefollowing

columnsshowthecoefficientsandtheirsignificanceunderthedifferentquantiles

10,25,50,75and90,usingIVQR.Basictests,R2andχ scoreforoveridentifying

restrictionsarepresentedattheendofthetable.

First, we are going to discuss briefly the control variable results to focus our

attention.Second,wewillfocusontheurbancontextconclusions.

>>>INSERTARROUNDHERETABLE2<<<

5.1.Socioeconomiccharacteristics

In general, all the socioeconomic and household and housing characteristics are

significantandsimilartowhatwecanexpectaccordingtopreviousliterature.

Thepriceofelectricityhastheexpectednegativeandsignificanteffect.Thevalue

of the coefficient is close to1 and is stable across thequantiledistribution.This

13  

indicatesthattheelectricitydemandinSpain,accordingtoourresults,isunitary.

This result is slightly higher than that obtained by Labandeira et al. (2006) and

Labandeiraet al. (2012)andclearlyhigher than thatobtainedbyBlazquezetal.

(2013).Nevertheless,theresultisinlinewithwhatwasobtainedininternational

studies of electricity demand (see the summary of international resultsmade in

Labandeiraetal.2012).

Asexpected,thehousesthatcombinegas,petrolorcoalforheatingandhotwater,

with electricity for the remainder, reduce in a very significant way the final

consumptionofelectricity.Thiseffectishigherforthehigherquantiles,indicating

thatthehigherthedemand,thelargertheeffectofusingotherenergysources.Our

result is consistent with most of the previous literature for the Spanish case.

Blázquezetal.(2013)usedagaspenetrationrate(percentageofhouseholdsthat

haveaccesstogas)asanexplanatoryvariableandfoundasignificantandnegative

sign;Moreover,Romero‐Jordánetal(2014,2016)foundthatagreaterpenetration

of electric heating and electric water heating had a clear positive impact on

Spanish electricity consumption. Those studies indicated a clear competition

betweenelectricityandgas in theprovisionofheatingandhotwater systems in

Spanishhouseholds.Moreover,theimpactofthisvariableincreaseswhenquantile

increasesasRomero‐Jordánetal(2016)alsofound.

Regardingtheageofthehouse,wefinditissignificantandpositive,indicatingthat

olderhouses consumemoreelectricity.However, this effect isnot significant for

thehigherquantilesofthedistribution.

Thenumberofmembersinthefamilyisalsoclearlysignificantandpositivewith

the expected result that the larger the family is the higher the electricity

consumption. This resultwas also found byBlázquez et al. (2013) andRomero‐

Jordán et al (2016) in the quantile regression‐ they also found this result was

stableamongthequantilesaswefind.ContraryRomero‐Jordánetal(2014)found

householdsizenotsignificant.

Our results indicate that to be an owner of the house significantly increases

electricityconsumption,andsimilarresultswerefoundbyLabandeiraetal.(2006)

forSpain.

14  

The demand for electricity is responsive to the level of income with elasticity

clearly below 1. This result is stable for the different income levels and for the

different quantiles in the quantile estimation. This means that income growth

apparently results in a less than proportional increase in electricity demand,

implyingthatapossibleconvergenceinthepercapitaincomeofSpainwiththatof

the most advanced countries in the EU‐15 will not translate into proportional

increases in electrical equipment, and in turn into proportional increases in

electricityconsumption.

Regionaldummiesarenotstableacrossthedistributionandshowdifferenteffects.

TheclearesteffectisfortheCanaryIslandsdummy,whichreflectsthehigherfinal

realpricesoftheelectricityduetotheisolatedsituationofthisarchipelago.Inthe

remaining cases, the effects change between quantiles, preventing clear

conclusions. Previous papers obtained similar results indicating that households

located in areas with Mediterranean climate (Regions South and East) tend to

consumemoreelectricityonaveragethantheircounterpartsinotherpartsofthe

country(Jordánetal2016).

5.2.Urbancontext:sprawlandagglomerationeffects

Ourinterestismainlyfocusedonanalysisofthetypeofhouseandthesizeofthe

urbanarea,whatwecalltheurbancontext.

Theestimations indicate thatahousehold living inanurbanarea(municipalities

with more than 100,000 inhabitants) significantly decreases electricity

consumptioncomparedtothoselivinginasmallcityorinaruralarea.Ourresults

for thequantileregressionare thesame forallquantilesof thedistributionwith

theexceptionofthehigherelectricityconsumers,thosesitedinquantile90,asfor

themthevariableisnotsignificant

Previous literature about variables similar to this one is inconclusive. Romero‐

Jordánetal. (2016) foundthathouseholds in lowerquantiles(1‐30quantiles)of

the electricity consumption distribution consumedmore in urban areas than in

ruralareas.However,fortheremaininghouseholds,theconsumptionincreasesed

iftheyweresituatedinruralareas.Incontrast,Labandeiraetal.(2006)concluded

that rural and village households (those living inmunicipalitieswith fewer than

10,001inhabitantsandwithmorethan10,000inhabitantsbutfewerthan50,001,

15  

respectively) spent more on electricity that households living in an urban

municipality (municipalities with more than 50,000 inhabitants) for the period

1973 to1995.However,once theyaccounted forbothdirectand indirecteffects

through an interaction term between total household expenditure and type of

municipality,theyfoundthatruralhouseholdsspentmoreonelectricity.

There are several possible explanations for our result. First, it could be due to

building construction practices in high density urban areas resulting in a larger

proportion of apartment buildings with comparatively less heating and cooling

requirements than detached or semidetached houses (which consume more

electricity as we have shown). Second, urban environments change the life

patternssuchthatmoretime isspentawayfromhome,moreoftenhaving lunch

anddinneroutside,consumingmore intensivelyexternalservices for takingcare

of children or dependentmembers of the family, andmore frequently pursuing

leisureactivities.

Finally, the variable in which we focus special attention is the type of house.

Basically,thisvariableinformsusabouttheimpactof livinginanisolatedhouse,

insteadofaflatsbuilding,ontheenergyconsumption.Whatwecanobserveisthat

detachedorsemidetachedhousesconsumesignificantlymoreelectricitythanflats,

apartmentsandotherboundingsolutions.Theeffectissignificantlyrelevant,even

after controlling for all the socioeconomic variables and after solving a possible

endogeneityproblem.Ourresultsaresimilartothosestudiesanalysingtheimpact

of the typeofhousingonhouseholdelectricityconsumptionas itwasshowed in

the above literature review‐section 2 (Heinonen and Junnila 2014, Stiri 2014,

Huang2015,Wiesmannetal2011).

It isalsoveryinterestingobservingwhatoccursacrossthequantilesdistribution

because thiseffect isnotsignificant in thosehouseswith lowerenergy‐intensity,

butitishigherforthehigherquantilesofelectricityconsumption.Supposedlythe

householdswith lower incomeand less energypurchase capacity are in the low

partofthedistributionandthosethatcanaffordhigherdemandareinthehigher

part of the distribution. Thus, this result implies that the effect of sprawl is

especiallyrelevantforhigh‐incomefamilies.

16  

5.Conclusionsandpolicyimplications

ThispaperestimatedtheimpactofurbansprawlandmunicipalitytypeonSpanish

household electricity consumption. Moreover, household socioeconomic and

demographic characteristics are also analysed as determinants of electricity

consumption. By using 2014 Household Budged Survey microdata with OLS, as

wellasaquantileregressioneconometricapproach,ourresultsconfirmbehaviour

ofSpanishfamiliessimilartothestandardsinotheradvancedeconomiesanalysed

inpreviousliterature.

Oneofthefactorsthatcontributemoretohigherelectricitydemandisthenumber

ofmembersinthefamily,butthesizeofSpanishfamiliesissupposedtobestable

in the future. A unitary electricity‐price indicates that households seem to be

reactive to energy prices; thus, we can be optimistic about the effectiveness of

pricing policies to reduce electricity consumption in Spanish households. At the

sametime,thelowincome‐elasticityobservedindicatethatwecannotexpectlarge

increases inelectricitydemandeven in thepossiblecontextof incomegrowth in

thenearfuture.

Ontheotherhand,olderhousesconsumemoreelectricity.Newconstructionsand

materialsaremuchmoreefficient intermsofenergyconsumption.Theyarealso

alwayspreparedtousesolarenergy.Theuseofotherenergysourcesforheating

andhotwater, such as gas, can clearly help in reducing electricity consumption.

Theproblemisthattheaccessibilityandpenetrationofthegassupplyisrestricted

to some areas connectedwith the national gas network. These areas aremainly

cities with higher density, so the accessibility to gas is going to be reduced for

sprawledhouseslocatedfarfromthecentreofthecity.

We include in our electricity consumption model two variables to evaluate the

effectofurbanagglomerationsandsprawlbehaviour.Weidentifythatlivinginan

urban areawithmore than100,000 inhabitants,which corresponds to themain

cities in Spain, has a significant effect on reducing the electricity demand.

Simultaneously, we found that detached and semidetached houses, typical in

sprawledlandscapes,clearlyconsumemoreelectricity.

The new trends and housing demands of Spanish families show an increase in

sprawl.Fordifferentreasons,thenumberofcitiesthatsufferhighlevelsofsprawl

17  

isgrowingandtheproportionoffamiliesthatchoosetoliveindetachedhousesis

increasing significantly.Detachedand sprawledhousesmeanmore consumption

of petrol for mobility and reduce the possibility of gas penetration because the

infrastructure in a low‐density context ismuchmoreexpensive.Whatwewould

like to test in thispaper is if italso increases theelectricitydemand.We include

severalurban/ruralvariables,aswellasaspecificdummyvariable,totestif,after

controllingforallthesocioeconomicandhousingcharacteristics,wecanobservea

significantpositiveeffectoflivinginadetachedhouse.Ourresultsclearlyconfirm

thishypothesis.Thus,increasingenergyefficiencyindetachedhousesisessential

formakingprogresstowardsasustainablecity.

In that sense, the Spanish Energy Saving and Efficiency Action Plan 2011‐2020

elaboratedby theMinistryof Industry,TourismandCommerce,and Institute for

Energy Diversification and Saving‐ IDEA (Spanish Government 2011), includes

recommendationsonenergysavingmeasuresintheBuildingsector,inaccordance

withthecontentsofDirective2010/31/EUrelatedtoenergyefficiencyinbuildings

(European Commission, 2010) and the Resource Efficient Europe strategy

(European Commission, 2011). Themeasures included in this Action Plan 2011‐

2020willinvolvesavingsoffinalenergyfor2020worth2,867ktoeintheBuilding

sector. In the residential building sector, savings are derived from energy

rehabilitationofthethermalenvelopeinbuildingstock,renewalofboilersandair

conditioning equipment, improvements in the energy efficiency of thermal

installations and indoor lighting installations in existingbuildings, andalso from

constructionofnearly‐zeroenergybuildingsandtherehabilitationofexistingones

with high energy qualification (of 8.2 million m2/year). A summary of other

policiesadoptedbydifferentcountriestoreduceenergyconsumptioninbuildings

canbefoundinAllouhietal.(2015).

Another strategy to achieve amore sustainable city is the introduction of small

installations of renewable thermal energy and cogeneration in household

electricity consumption. In that sense, theSpanishGovernment (2014)approved

theLaw413/2014,whichestablishes,forthefirsttimeinSpain,theconditionsof

auto‐consumptionofelectricalenergypowerfromrenewableenergy.

18  

As shown, the quantile regression indicated that electricity consumption

determinantsvarieddependingonthehouseholdenergyusequantile.Thus,smart

meteringandvariousconsumption‐feedbacksystems(Podgorniketal.2016,Salo

et al. 2016), in combination with personal consultations (Stieß and Dunkleberg

2013, Laakso and Lettenmeier, 2016) are proposed instruments for obtaining

more efficient and sustainable household energy consumption according to

specificcharacteristicsofeachhousehold.

Inlinewithcurrentresults,furtherresearchonthistopicwouldbedesirable.For

example, toexamine the impactof thedegreeofurbanizationand/or the typeof

energy used by the household on household carbon dioxide emissions could be

veryinterestinginfutureresearch,asotherauthorshavenoted(Hanetal.,2015,

Liuetal.,2011,WangandYang,2014,Lietal.,2015,Yeetal.,2017).

19  

References

BeguinH(1982):Theeffectofurbanspatialstructureonresidential‐mobility.TheAnnalsofRegionalScience16,16–35. 

BlázquezL,BoogenN,FilippiniM(2013):ResidentialelectricitydemandinSpain:Newempiricalevidenceusingaggregatedata.EnergyEconomics36,648–57. 

BoehmTP(1982):Ahierarchicalmodelofhousingchoice.UrbanStudies19,17–31. 

BrounenD, KokN, Quigley JM (2012): Residential energy use and conservation:economicsanddemographics.EuropeanEconomicReview36,931–45. 

Bunnel, T., Drummond, L.B.W., Ho, K.C. (2002): Critical reflections on cities inSouthernAsia.Singapore:BrillAcademicPress.

CaylaJM,MaiziN,MarchandC(2011):Theroleofincomeinenergyconsumptionbehavior:evidencefromFrenchhouseholdsdata.EnergyPolicy39,7874‐83. 

ChernozhukovV,HansenC(2005):AnIVmodelofquantiletreatmenteffects.Econometrica73,245–261. 

ChernozhukovV,HansenC(2006):Instrumentalquantileregressioninferenceforstructuralandtreatmenteffectmodels.JournalofEconometrics132,491‐525. 

Couch, C., Ldontidou, L., Petschel‐Held, G. (2007): Urban Sprawl in Europe.Landscapes,Land‐UseChange&Policy.Oxford:BlackwellPublishing.

Council of the European Union (2010: Europe 2020 — A strategy for smart,sustainableandinclusivegrowth.COM(2010)2020final,Brussels. 

Councilof theEuropeanUnion(2014):2030FrameworkforClimateandEnergy.Availableatwww.ec.europa.eu/energy/2030. 

Daioglou V, Van Ruijven BJ, Van Vuuren DP (2012). Model projections forhouseholdenergyuseindevelopingcountries.Energy37,601‐15. 

Deaton AS, Muellbauer JN (1980): An Almost Ideal Demand System. AmericanEconomicReview83,570‐97. 

DeatonA(1988):Quality,quantityandspatialvariationofprice.AmericanEconomicReview70,418‐430. 

DruckmanA,JacksonT(2008):HouseholdenergyconsumptionintheUK:ahighlygeographically and social‐economically disaggregatedmodel. Energy Policy36,3177‐92. 

EuropeanCommission(2010):Directive2010/31/UEoftheEuropeanParliamentandoftheCouncilof19May2010ontheenergyperformanceofbuilding.EuropeanCommission,Brussels 

European Commission (2011). Energy Roadmap 2050. European Commission,Brussels. 

European Commission (2016): Eurostat database. Retrieved 26 September 2016http://ec.europa.eu/eurostat/data/database. 

20  

Feng Z‐H, Zou L‐L, Wei Y‐M (2011): The impact of household consumption onenergyuseandCO2emissionsinChina.Energy36,656‐70 

Gilbert, A. (1996): The Mega‐city in Latin America. Tokyo: United NationsUniversityPress.

Glaeser, E., Kahn, M. (2004): Sprawl and urban growth. In Henderson, V. andThisse, J.F. (eds.) Handbook of Regional and Urban Economics, vol. 4.Amsterdam:North‐Holland,2481‐2527.

Heinonen J, Junnila S (2014): Residential energy consumption patterns and theoverall housing energy requirements of urban and rural households inFinland.EnergyandBuildings76,295–303. 

Huang W‐H (2015): The determinants of household electricity consumption inTaiwan:Evidencefromquantileregression.Energy87,120‐133 

Jaeger,J.A.G., Schwick, C. (2014):Improving the measurement of urban sprawl:WeightedUrbanProliferation (WUP) and its application to Switzerland.EcologicalIndicators,38,294‐308.

JonesRV,FuertesA,LomasKJ(2015):Thesocio‐economic,dwellingandappliancerelated factors affecting electricity consumption in domestic buildings.Renewable&SustainableEnergyReviews43,901‐17. 

Kasparian J (2009): Contribution of crude oil price to households' budget: theweightofindirectenergyuse.EnergyPolicy37,111‐4. 

KavousianA,RajagopalR,FischerM(2013):Determinantsofresidentialelectricityconsumption: using smart meter data to examine the effect of climate,buildingcharacteristics,appliancestock,andoccupants'behavior.Energy55,184‐94. 

KazaN(2010):Understandingthespectrumofresidentialenergyconsumption:aquantileregressionapproach.EnergyPolicy38,6574‐85. 

KellyS (2011): Dohomes thataremoreenergyefficientconsume lessenergy?AstructuralequationmodeloftheEnglishresidentialsector.Energy36,5610–20. 

KoenkerR,&BassetJrG(1978).Regressionquantiles.Econometrica,46(1),33‐50. 

KyröR,HeinonenJ,SäynäjokiA, JunnilaS(2011):Occupantshavelittle influenceon theoverall energyconsumption indistrictheatedapartmentbuildings.EnergyandBuildings43,3484–90. 

LabandeiraX,LabeagaJM,López‐OteroX(2012):Estimationofelasticitypriceofelectricitywithincompleteinformation,EnergyEconomics34,627–33 

Labandeira X, Labeaga JM, Rodríguez M (2006): A residential energy demandsystemforSpain.EnergyJournal27,87–112. 

Larsen BM, Nesbakken R (2004): Household electricity end‐use consumption:results from econometric and engineeringmodels, Energy Economics 26,179–200. 

21  

MiahMD,FoysalMA,KoikeM,KobayashiH(2011):Domesticenergy‐usepatternby the households: a comparison between rural and semi‐urban areas ofNoakhaliinBangladesh.EnergyPolicy39,3757‐65. 

Nesbakken R (1999): Price sensitivity of residential energy consumption inNorway.EnergyEconomics21,493‐515. 

Niu S, Jia Y, Ye L, Dai R, Li N (2016): Does electricity consumption improveresidential living status in less developed regions? An empirical analysisusingthequantileregressionapproach.Energy95,550‐60. 

Niu S, Zhang X, Zhao C, Niu Y (2012): Variations in energy consumption andsurvival status between rural and urban households: a case study of theWesternLoessPlateau,China.EnergyPolicy49,515‐27. 

O'NeillBC,ChenBS(2002):DemographicdeterminantsofhouseholdenergyuseintheUnitedStatesPopulationanddevelopmentreview28,53–88. 

ÖzcanKV,GülayE,Üçdoğru(2013):EconomicanddemographicdeterminantsofhouseholdenergyuseinTurkey.EnergyPolicy60,550‐57. 

Pachauri S (2004): An analysis of cross‐sectional variations in total householdenergy requirements in India using micro survey data. Energy Policy 32,1723‐35. 

Polèse,M.,Champain,C.(2003):Laevolucióndeloscentrosurbanos:LaexperienciadeAméricadelNorte.Washington:WorldBank.

PoumanyvongP,KanekoS(2010):Doesurbanizationleadtolessenergyuseandlower CO2 emissions? Across‐country analysis. Ecological Economics 70,434‐44. 

RahutDB,DasS,DeGrooteH,BeheraB(2014):DeterminantsofhouseholdenergyuseinBhutan.Energy69,661‐72. 

Rahut DH, Behera B, Ali A (2016): Household energy choice and consumptionintensity: Empirical evidence from Bhutan. Renewable and SustainableEnergyReviews53,993–1009. 

Romero,J.(2012):ConstrucciónresidencialygobiernodelterritoriodeEspaña.Dela burbuja especulativa a la recesión, causas y consecuencias. CuadernosGeográficos,47,17‐46.

Romero‐Jordán D, del Río P, Peñasco C (2014): Analysing the determinants ofhouseholdelectricitydemandinSpain.Aneconometricstudy.InternationalJournalofElectricityandPowerEnergySystem63,950–61. 

Romero‐Jordán D, del Río P, Peñasco C (2016): An analysis of the welfare anddistributive implications of factors influencing household electricityconsumption.EnergyPolicy88,361‐70. 

Rubiera, F., González. V., Rivero, J.L. (2016): Urban sprawl in Spain: differencesamongcitiesandcauses.EuropeanPlanningStudies,24(1),204‐226.

Santin OG (2011): Behavioural patterns and user profiles related to energyconsumptionforheating.EnergyBuilding43,2662–72. 

22  

SpanishGoverment(2016):LibrodelaEnergíaenEspaña2014,SpanishMinistryofIndustry,EnergyandTourism:Madrid(Spain). 

Spanish Government (2011): Spanish Energy Saving and Efficiency Action Plan2011‐2020‐ Ministry of Industry, Tourism and Commerce and IDEA.Madrid. 

SpanishGovernment(2014):RealDecreto413/2014,de6de junioporelqueseregulalaactividaddeproduccióndeenergíaeléctricaapartirdefuentesdeenergíarenovables,cogeneraciónyresiduos‐BOE‐A‐2014‐6123,Madrid. 

StiriH (2014):Building andhouseholdX‐factors and energy consumption at theresidentialsector.Astructuralequationanalysisoftheeffectsofhouseholdand building characteristics on the annual energy consumption of USresidentialbuildings.EnergyEconomics43,178–184. 

ValenzuelaC,ValenciaA,WhiteS, Jordan JA,CanoS,Keating J, et al. (2014): Ananalysis of monthly household energy consumption among single‐familyresidencesinTexas2010.EnergyPolicy69,263‐72. 

PodgornikA,SucicB,BlazicB(2016):Effectsofcustomizedconsumptionfeedbackonenergyefficientbehaviour in low‐incomehouseholds. JournalofCleanerProduction130,25‐34 

SaloM,NissinenA,LiljaR,OlkanenE,O'NeillM,UotinenM(2016):Tailoredadviceand services to enhance sustainable household consumption in Finland.JournalofCleanerProduction121,200‐7. 

AllouhiA, El FouihY,KousksouT, JamilA, Zeraouli Y,MouradY (2015): Energyconsumption and efficiency in buildings: current status and future trends.JournalofCleanerProduction109,118‐130. 

Stieß I, Dunkleberg E (2013): Objectives, barriers and occasions for energyefficient refurbishment by private homeowners. Journal of CleanerProduction48,250–259. 

LaaksoS,LettenmeierM(2016):Household‐leveltransitionmethodologytowardssustainablematerialfootprints.JournalofCleanerProduction132,184–191. 

Ye H, Ren Q, Hu X, Lin T, Xu L, Li X, Zhang G, Shi L, Pan B (2017): Low‐carbonbehavior approaches for reducing direct carbon emissions: Householdenergyuseinacoastalcity.JournalofCleanerProduction141,128‐136. 

Han L, Xu X, Han L (2015): Applying quantile regression and Shapleydecompositiontoanalyzingthedeterminantsofhouseholdembeddedcarbonemissions: evidence from urban China. Journal of Cleaner Production 103,219‐230. 

LiuLC,WuG,WangJN,WeiYM(2011):China'scarbonemissionsfromurbanandruralhouseholdsduring1992–2007.JournalofCleanerProduction19,1754–1762. 

Wang Z, Yang L (2014): Indirect carbon emissions in household consumption:evidence from the urban and rural area in China. Journal of CleanerProduction78,94–103. 

23  

Li Y, Zhao R, Liu T, Zhao J (2015): Does urbanization lead to more direct andindirect household carbon dioxide emissions? Evidence from China during1996–2012.JournalofCleanerProduction102,103‐114. 

Norman J,MacLeanHL,KennedyCA(2006):Comparinghighand lowresidentialdensity: life‐cycle analysis of energy use and greenhouse gas emissions.JournalofUrbanPlanningDevelopment132,10‐21. 

RickwoodP,GlazebrookG,SearleG(2008):Urbanstructureandenergy–areview.UrbanPolicyandResearch26,57–81. 

Wiesmann D, Lima Azevedo I, Ferrão P, Fernández JE (2011): Residentialelectricityconsumption inPortugal: findingsfromtop‐downandbottom‐upmodelsEnergyPolicy39,2772–9.