Understanding Market potential for biodiesel in Spain: A Pilot study based on consumer preferences

Transcription

1 1 Uderstadig Market potetial for biodiesel i Spai: A Pilot study based o cosumer prefereces A. Gracia, J. Barreiro-Hurlé & L. Pérez y Pérez Agrifood Research ad Techology Cetre of Arago (CITA) Avda Motaaa, 930 E Zaragoza (Spai) agracia@arago.es Cotributed Paper prepared for presetatio at the Iteratioal Associatio of Agricultural Ecoomists Coferece, Beijig, Chia, August 16-22, 2009 Copyright 2009 by Azucea Gracia, Jesús Barreiro-Hurlé & Luis Pérez y Pérez. All rights reserved. Readers may make verbatim copies of this documet for o-commercial purposes by ay meas, provided that this copyright otice appears o all such copies.

2 2 UNDERSTANDING MARKET POTENTIAL FOR BIODIESEL IN SPAIN: A PILOT STUDY BASED ON CONSUMER PREFERENCES A. Gracia #, J. Barreiro-Hurlé & L. Pérez y Pérez ABSTRACT The productio ad use of biodiesel is a importat alterative i Europe to dimiish the emissio of gases i order to fulfill the Kyoto protocol goals. Supported by the Commo Agricultural Policy (CAP), the productio of biodiesel has icreased i Europe but the cosumptio is oly of cosiderable importace i some coutries. I Spai, the cosumptio of biodiesel is still low compared to other EU coutries. The aim of the paper is to aalyze cosumers prefereces for biodiesel i Spai. To do that a choice experimetal approach has bee used to assess cosumers valuatio of differet diesel optios ad calculate the willigess to pay for biodiesel. The data come from a recet survey coducted i Spai (Regio of Aragó). Results idicate that diesel users would pay a extra price of 0.05 to fill biodiesel up istead of fill covetioal diesel, which is the highest premium cosumers are willig to pay for ay diesel attribute cosidered. The paper also explores factors affectig the premium cosumers are willig to pay for this fuel. Keywords: Biodiesel, cosumer, prefereces, Spai, choice experimets # Agrifood Research ad Techology Cetre of Arago (CITA). Avda Motaaa, 930. E Zaragoza (Spai) agracia@arago.es

3 3 UNDERSTANDING MARKET POTENTIAL FOR BIODIESEL IN SPAIN: A PILOT STUDY BASED ON CONSUMER PREFERENCES INTRODUCTION Biofuels has bee put forward as oe of the strategies to allow reductios i greehouse gases emissios. May coutries have set targets for overall trasport related eergy cosumptio to be supplied by biofuels ad i the case of Spai most of the supply of biofuels is cetered i biodiesel. As biodiesel will have to coexist with other fuel optios i the market ad its use is based o volutary idividual decisios, reachig the set targets will be directly related to how cosumers will react to these ew products available i the market. Productio ad distributio costs of biofuels are curretly higher tha fossil oes ad uless subsidized by public authorities they will have to be marketed at higher prices. This is the case i Spai where, eve whe subsidies are i place, biodiesel retail prices are higher tha fossil equivalets. Uderstadig whether cosumers will be willig to pay this higher prices ad why, is a key issue that should be take ito accout whe desigig policies to icrease biodiesel use. I this paper we attempt to aswer both questios, whether there is a willigess to pay ad the factors that explai its variability, by udertakig a demad aalysis for diesel i a midsized Spaish tow. The rest of the paper is structured as follows: the basic of the choice modelig approach used are preseted i the ext sectio. This is followed by a descriptio of how the data has bee obtaied ad its mai characteristics. Results of both the cosumer choice model, mea ad idividual specific willigess to pay for the differet attributes are preseted ad discussed. The paper eds with some tetative coclusios that ca be derived from our results with regards to optimal policy desig for biodiesel itroductio i Spai. METHODOLOGY Choice experimets (CE) have bee selected to assess cosumers prefereces for biodiesel for a umber of reasos. CE is capable of valuig multiple attributes simultaeously, its framework is cosistet with radom utility theory, ad the hypothetical choices preseted are similar to real market decisios (Lusk et al., 2003 ad Adamowicz et al.,1998). Choice modelig is based o Lacastria cosumer theory of utility maximizatio (Lacaster, 1966)

4 4 ad cosumers prefereces over food attributes are modeled i a radom utility framework (McFadde, 1974). I the choice modelig approach cosumers choose betwee alterative products that cotai a umber of attributes with differet levels. Idividuals choose the alterative that provides the greatest utility ad the probability of selectig a alterative icreases as the utility associated with it icreases. The utility fuctio is kow by the idividual but some of its compoets are uobserved by the researcher. Thus, utility is take as a radom variable where utility from the th idividual facig a choice amog j alteratives withi choice set J ca be represeted as, = v + Ujt jt ε jt [1] Where is the umber of respodets; j the umber of alteratives withi choice set J, t the umber of choice occasios, v jt utility determied by the attributes ad their values for 2 alterative j i t choice occasios ad ε a extreme value error term (0, σ ), i.i.d. over jt alteratives ad idepedet of v jt Differet choice models ca be derived cotiget o the specificatio of the desity of uobserved factors f ( ε jt ). The selectio of this fuctio will deped o the assumptios uderlig the cosumer s prefereces. If preferece heterogeeity across cosumers is expected, a geeral specificatio such as the Rado Parameters (RPL) or mixed logit model ca be used. Assumig that v jt is liear i parameters ( v jt ' xjt = β ), each cosumer has his ow vector of parameters β 1 which deviates from the populatio mea β by the deviatio parameters η. The β is radom across idividuals with a desity fuctio f(β). I the RPL model, the coditioal probability that idividual chooses alterative j i a particular choice occasio t, is represeted as: L jt exp( β ' xjt ) ( β ) = [2] exp( β ' x ) i it 1 β does ot carry the subscript t as taste is assumed to vary over respodets but ot over choices.

5 5 For the maximum likelihood estimatio, the coditioal probability of the sequece of choices made by each respodet is obtaied accordig to the followig expressio: S ( β ) = Lj(, t) t ( β ) [3] t where j(,t) represets the alterative chose by perso i choice occasio t. The ucoditioal probability for this sequece is give by: P θ ) = S ( β ) f ( β θ ) dβ. [4] ( Sice the itegral i [4] does ot have a close form, the probabilities have to be simulated by summig over R radom draws of β, which are take from the probability desity fuctio β f ( θ ) (Trai,1998, 1999 ad 2003). For the estimatio of the RPL, Halto draws rather tha radom draws are used sice they provide a more efficiet simulatio for the RPL. Mea radom parameters are derived as the average of the parameters over the R replicatios. The derived stadard deviatio which represets the amout of spread or dispersio aroud the sample populatio is calculated over each of the R draws. I additio to these estimated parameters, the RPL model provides also estimates parameters for each idividual i the sample, reflectig that cosumers preset heterogeeous prefereces. The, to aalyze the reasos behid cosumers heterogeeity i prefereces the coditioal parameters estimates for each idividual will be calculated to obtai idividual-specific WTP. These coditioal parameters are the commo-choice-specific parameter estimates which are coditioed o the choices observed to have bee made (Hesher et al., 2005). These idividual-specific WTP will be regressed i order to idetify the factors determiig cosumers WTP for biodiesel. I derivig WTP estimates based o radom parameters, all the iformatio i the distributio or just the mea or stadard deviatio estimates ca be used. I our study, both WTP have bee calculated by takig the ratio of the parameter estimated for the aalyzed attributes to the price parameter multiplied by mius oe. Mea WTPs are calculated usig the mea estimates parameters ad WTPs for each idividual i the sample are calculated usig the idividual-choice-specific estimates parameters (Campbell, 2007). The reported choice

6 6 model has bee estimated usig NLOGIT 3.0 (Greee, 2002) while regressios fitted to explai WTP heterogeeity have bee estimated usig STATA DATA Data were collected from a survey coducted i Zaragoza, a medium-sized tow located i orthwest Spai, durig April This tow was selected to be represetative of the coutry as it has similar socio-demographics as those of the overall Spaish Cesus of Populatio (see Table A.1. i Appedix). This questioaire had as mai objective the idetificatio of diesel cosumer attitudes, kowledge ad prefereces for differet aspects of biodiesel market developmet. Cosumers were asked questios related to biodiesel fill up habits (where ad why), kowledge about biodiesel, attitudes towards biodiesel, biodiesel cosumptio (actual use of biodiesel, itetio to purchase, place of purchase, etc.), attitudes towards biodiesel purchase, as well as subjective orms ad, perceive behavioral cotrol (Ajze, 1991). The questioaire also cotaied questios o socio-demographic characteristics (i.e. sex, family size ad compositio, age, educatio level, icome).last, the questioaire also icluded questios to implemet the choice experimet. Before the fial questioaire was admiistrated, a pilot survey was udertake to idetify cosumer believes ad kowledge with regards to biodiesel as well as their willigess to pay for this fuel. This pilot, coducted to a small sample of respodets (N=20), allowed us to select the diesel characteristics more importat to buyers ad have a iitial idea of the iterval i which willigess to pay for biodiesel was located. With this iformatio the most relevat attributes to be icluded i the choice experimet desig were selected. The most relevat aspects for cosumers whe decidig which diesel to fill up were the proximity of the petrol statio to buyers every day route ad whether the petrol statio is associated to a atio-wide oil compay (REPSOL, BP, etc.) agaist beig idepedet. Results from the cotiget valuatio ope-eded questios idicate that the maximum premium buyers were willig to pay to fill biodiesel up istead of covetioal oe was 10%. From these results the choice set desig costructed cosidered four distictive diesel characteristics (price, type of diesel, petrol statio locatio ad petrol statio associatio), each of them takig two or more differet levels. Table 1 shows the attributes ad the levels 2 The data set used i the aalysis was gathered by studets from the IAMZ-CIHEAM i the Postgraduate Specializatio Course o Agro-food Marketig

7 7 used. The selectio of price vector was doe as follows. A baselie price sceario was selected correspodig with the mea price of a liter of diesel i differet petrol statios at the time of the survey (1.1 /liter). Based o the results of the CV pre-test a 10% premium was icluded to reflect the highest extra price shoppers were willig to pay to buy biodiesel istead of covetioal oe ad also a 10% discout was icluded to detect whether cosumers ot willig to purchase biodiesel would do it at a discout.. The other attributes etered the choice set with two levels. Table 1. Attributes ad levels used i the experimetal desig. Attribute Price Levels 0.99 per liter 1.10 per liter 1.21 per liter Type of diesel Biodiesel Covetioal diesel Petrol statio i the proximity of the cosumer s every day route Petrol statio associated to a atio-wide oil compay Yes No Yes No Iformatio regardig the meaig of the price, type of diesel ad petrol statio characteristics (proximity ad atio_wide) was preseted to participats right before the choice experimet questio. The choice set desig was created employig a ulabeled octagoal desig created with SPSS 14.0, which resulted i a total of 12 choice sets. I order to avoid a order effect i the resposes, the order of the choice sets was radomized. The 12 selected choice sets were blocked ito three blocks, with each respodet radomly allocated to oe block of four choice sets. Thus, each respodet had to choice four times betwee three diesel optios: two optios which described the hypothetical diesel ad a third optio offerig cosumers the possibility of ot buyig (see o example of choice set i Appedix I). Sample size icluded 121 diesel users. Cosiderig this sample is extracted from a ifiite populatio ad assumig a cofidece level of 95.5% (k=2) ad p=0.5 the error level associated to the reported estimates is ±9%. A stratified radom samplig procedure was used based o age. Target respodets were owers of diesel egie vehicles ad the questioaire was delivered face to face at selected petrol statios located throughout the tow ad its

8 8 suburbs. Iterviewers approached radomly selected idividuals askig them oe screeig questio: whether they have a diesel motor vehicle. I the case of a egative respose, the iterviewer selected radomly aother customer belogig to a give age group, util fidig a participat matchig the requiremet. Sample summary statistics are preseted i table 2. The majority of respodets were male (69.5%). The respodet s average age was about 39, livig i a household with three members. Approximately 60% of respodets state that they have a icome betwee 1,500 ad 3,500 ad more tha half of the sample has complete uiversity studies. Fially, the percetage of households with kids less tha six years old is 20%, ad with people more tha 65 years old, 11%.

9 9 Table 2. Sample characteristics (%, uless stated) ad exogeous variables defiitio. Variable defiitio Name (type) Value Idividual characteristics Geder Male Female FEMALE (dummy: 1=female) Age (Average from total sample) AGE (cotiuous) Educatio of respodet Elemetary School High School Uiversity Average Household Icome a Less tha 1,500 Betwee 1,501 ad 2,500 Betwee 2,501 ad 3,500 Betwee 3,501 ad 4,500 More tha 4,500 UNIVERSITY (dummy: 1=uiversity) HIGH_INCOME (dummy: 1=higher tha 3,500 ) Household Size (Average from total sample) HSIZE (cotiuous) 3.3 Household with kids less tha 6 years old (1=Yes) KIDS6 (dummy:1=yes) 20.3 Household with adults more tha 65 years old (1=Yes) ELDERLY (dummy: 1=yes) 11.0 Cosumers kowledge o biodiesel Cosumer self-assessed kowledge of biodiesel KNOW (dummy: 1=yes) 86.4 Cosumers diesel fill up habits Idividual usually fills diesel up i petrol statios i the proximity of the every day route Idividual usually fills diesel up i petrol statios associated to a atio-wide oil compay Cosumers attitudes towards biodiesel I Lack of iformatio o the effect of biodiesel use o egies FILL_DAILYROUTE (dummy: 1=yes) FILL_NATIONWIDE (dummy: 1=yes) LACK_INFO (5-poit icreasig scale) Biodiesel use cotributes to overexploitatio of atural OVER_EXPLOTATION(5-poit icreasig 3.12 resources scale) Cosumers behaviour model Attitudes towards biodiesel purchase ATT_PURCH (5-poit icreasig scale) 3.92 Subjective orm SUB_NORM (5-poit icreasig scale) 2.97 Perceived behavioral cotrol PERC_CONTROL (5-poit icreasig scale) 3.85 Perceived difficulty to purchase PERC_DIFF (5-poit icreasig scale) 2.48 a 3% of respodets do ot provide iformatio o the icome level MODEL SPECIFICATION AND ESTIMATION RESULTS I the fial specificatio of the utility fuctio, a alterative-specific costat represetig the A ad B choice optio was itroduced (ASC). It is expected that this costat would be positive ad sigificat, idicatig that cosumers will get higher utility from alterative A ad B tha from the o-buy optio C. The type of diesel (BIODIESEL), the proximity of the petrol statio to the buyers every day route (PROXIMITY) ad whether the petrol statio is associated to atio-wide oil compay (NATION_WIDE) variables are effect coded ad the price (PRICE) represets the price levels give to cosumers for each hypothetic product.

10 10 Price is expected to have a egative impact o utility while, the effects of the other variables are the posed questios i the paper. I the RPL, the researcher has to specify the distributio for the radom coefficiets that satisfied his expectatios about cosumer behavior (Trai, 2003). Sice cosumers may either like or dislike the diesel attributes cosidered i the experimetal desig, a ormal distributio is used. The estimatio of the RPL was coducted usig NLOGIT 3.0 treatig price as a fixed coefficiet ad lettig the coefficiets of the other three attributes radom. The results of the RPL estimate are preseted i Table 3 3. The fial specificatio of the utility fuctio relies o statistical tests that support the sigificace of the icluded coefficiets. With respect to the overall fit, the model is statistically sigificat with a chi-square statistics of 379,8 which is higher that the critical value, suggestig that the cosidered diesel characteristics are joitly sigificat, affectig cosumers utility. As expected, the alterative specific costat is positive ad statistically sigificat; idicatig that cosumer utility for purchase alteratives is higher tha the o-purchase optio. Table 3. Mixed Logit Model results for diesel choice. Variable Coefficiet Stad. Error t-statistic Mea Values β Price BIODIESEL PROXIMITY NATION_WIDE Stadard deviatios BIODIESEL PROXIMITY NATION_WIDE Number of observatios 1416 Chi-square Log likelihood Pseudo R The o-radom parameter (PRICE) is egative ad the Wald test idicates that it is statistically sigificatly differet from zero at the 5% of sigificace level. Therefore, price icremets decrease the associated utility level provided by the choice of each diesel products. The mea of the three radom parameters (BIODIESEL, PROXIMITY ad 3 Three idividuals did ot aswer to all choice questios ad have bee subsequetly dropped from the aalysis preseted.

11 11 NATION_WIDE) are statistically sigificat explaiig cosumers utility. The positive value of the mea parameter estimates idicates that utility associated to biodiesel, proximity ad atio-wide associated petrol statio is higher tha for their alteratives (covetioal diesel, petrol statio ot close to usual route ad idepedet petrol statio). Lookig to heterogeeity i preferece, the Wald statistics for the derived stadard deviatio parameters idicates that the dispersio aroud the mea estimate is statistically differet from zero for the three aalyzed characteristics of diesel. I other words, the effect of these attributes o the utility fuctio differs across diesel buyers. However, the iterpretatio of direct estimate parameters is ot eough to fully uderstad cosumers valuatio. Therefore, we calculate the margial values or willigess to pay for the effects of the attributes (BIODIESEL, PROXIMITY ad NATION-WIDE) show i table 4. Mea WTP values are calculated by takig the ratio of the mea parameter estimated for the diesel attributes to the mea price parameter multiplied by mius oe. Table 4. WTP estimates for idividual attributes cosidered ( per liter). Attribute Mea Stadard deviatio BIODIESEL PROXIMITY NATION_WIDE Results idicate that all the cosidered attributes carry a positive premium, with the highest WTP associated with the biodiesel type (0.054 per liter of biodiesel). This implies that, o average, per liter is the premium that makes diesel buyers idifferet betwee the two levels of utility, associated with covetioal diesel ad biodiesel. Slightly behid, the WTP for the proximity of the petrol statios to buyers every day route is Fially, cosumers are willig to pay about half, extra, for a liter of diesel fill up i a petrol statio associated to a atio-wide oil compay. Results preseted idicate that cosumer valuatio for the differet optios to fill up diesel is heterogeeous. However, the model results do ot allow explaiig this heterogeeity. To do so, the determiats of idividual WTP estimates have to be idetified. This ca be achieved by focusig o idividual-specific WTP from the Radom Parameter model istead of o mea WTP. Usig idividual-specific WTP as a depedat variable ad through the

12 12 specificatio ad estimatio of value fuctios for each diesel characteristic statistically sigificat factors explaiig WTP variatio ca be idetified. For the objectives of this paper we would be iterested i kowig oly those factors explaiig WTP for biodiesel. However, we also preset the results idetifyig the factors determiig NATION_WIDE to see whether these are commo to both attributes ad thus, promotio of biodiesel could be made followig the same basics of marketig for well-kow brads. We assume that determiats of WTP heterogeeity are ot oly buyers socio-demographic characteristics but also buyers kowledge o biodiesel, buyers fill up habits ad the factors of Ajze s behavioral model (Azje, 1991). These iclude attitudes towards the product, the attitude that the perso holds toward egagig i the behavior (purchase attitude), the degree of social pressure felt by the perso with regard to the behavior (subjective orm) ad the degree of cotrol that the perso feels he/she has o performig the behavior (perceived behavioral cotrol). Followig Sparks et al., 1997, the perceived behavioral cotrol may be composed of two separate costructs, perceived difficulty ad perceived cotrol. Perceived difficulty meas the skills ad abilities that cosumers ow ad are believed by them to ifluece the degree of persoal cotrol over the behavior. The defiitios of the exogeous variables that explai WTP variability ca be foud i table 2. As far as product kowledge is cocered, diesel buyers were asked whether they kew what biodiesel was, those aswerig yes are cosidered to have a subjective kowledge of the product (KNOW) although the quality of this kowledge has ot bee measured. To measure diesel buyers fill up habits, respodets were asked to idicate whether they usually fill diesel up i petrol statios i the proximity of their every day route (FILL_DAILYROUTE) ad/or whether they usually fill diesel up i petrol statios associated to a atio-wide oil compay (FILL_NATIONWIDE). Over 75% of respodets stated that they usually fill diesel up i petrol statios i the proximity of their every day route ad 21% that they usually fill diesel up i petrol statios associated to a atio-wide oil compay. Cosumers were asked to rate, usig a 5-poit icreasig scale, differet statemets related to agreemet with biodiesel impacts (both positive ad egative) ad importace give to biodiesel characteristics whe decidig to cosume it or ot 4. Oly two of the aspects have 4 As far as impacts are cocered eight statemets were asked for such as biodiesel is less pollutat tha covetioal diesel ; biodiesel ca be produced usig local raw materials or biodiesel use helps fightig

13 13 bee fially icluded: importace give to lack of iformatio o the effect of biodiesel use o egies whe decidig ot to buy biodiesel (LACK_INFO) ad agreemet with the statemet biodiesel use cotributes to the overexploitatio of atural resources (OVER_EXPLOITATION). Fially, the three factors from Ajze s behavioral model have bee measured. Attitudes to biodiesel purchase were measured askig diesel buyers their degree of agreemet i a 5-poit icreasig scale to the setece I believe that buyig biodiesel is good (ATT_PURC). Subjective orm was measured askig diesel buyers their degree of agreemet i a 5-poit icreasig scale to the setece Most people who are importat to me thik that I should buy biodiesel (SUB_NORM). The perceived behavioral cotrol was measured askig diesel buyers their degree of agreemet i a 5-poit icreasig scale to the setece If biodiesel were available i petrol statios, othig would prevet me from buyig it (PERC_CONTROL) ad the perceived difficulty to the setece eve if I should wat to buy biodiesel, I do ot thik I would ever be able to do so (PERC_DIFF). Table 5 presets the value fuctio estimates for the two attributes cosidered. Each equatio has bee estimated by Ordiary Least Square (OLS), as the two edogeous variables are cotiuous. As far as the iterpretatio of the results, first it should be poited out that both models are overall statistically sigificat (F values reject the ull hypothesis that all estimated parameters are equal to zero at the 5% sigificace level) ad they explai a reasoable part of the WTP heterogeeity (adjusted R 2 values are higher tha 20%). Robust t- ratios are reported for idividual parameter sigificace to correct by heteroscedasticity (Greee, 2008). agaist climate chage. As far as cosumptio is cocered statemets were related to reasos for ot buyig biodiesel ad icluded statemet such as Its ot available o my ormal petrol statio ; there is o sufficiet iformatio o biodiesel available or it ca icrease the maiteace costs of my vehicle.

14 14 Table 5. Factors affectig idividual willigess to pay for differet diesel optios. BIODIESEL NATIONWIDE Variables Estimates t-ratios Estimates t-ratios Costat Socio-demographics AGE UNIVERSITY HIGH_INCOME KIDS ELDERLY Cosumers kowledge o biodiesel KNOW Cosumers diesel fill up habits DAILY_ROUTE KNOWN_BRAND Cosumers attitudes towards biodiesel LACK_INFO OVER_EXPLOTATION Cosumers behavior model ATT_PURCH DIFF_PURCH F VALUE Adjusted R Number of observatios=118 / Robust White (1980) t-ratios are reported First, it must be highlighted that factors explaiig the value attached to both, biodiesel ad whether the petrol statio is associated to a atio-wide oil compay are very differet. While the latter depeds maily o buyers socio-demographic characteristics, the former is better explaied by buyers biodiesel kowledge, buyers fill up place, attitudes towards biodiesel ad the Azje s factors. Oly oe buyer socio-demographic characteristics affects WTP for biodiesel, with higher WTP beig associated with idividuals which have uiversity degrees (UNIVER). O the other had, ad as far as kowledge ad habits are cocered, buyers who have heard about biodiesel ad those who state that they usually fill diesel up i petrol statio i the proximity of their every day route are less willig to pay for biodiesel. The role of attitudes ad compoets of Azje s behavioral model are the most sigificat determiats of biodiesel WTP. Buyers who highly believe that there is a lack of iformatio o the effect of biodiesel o egies are less willig to pay for biodiesel. I the same way, buyers who highly believe that biodiesel might cotribute to overexploitatio of atural resources are less willig to pay for biodiesel. This seems to show a clear reluctace to pay for biodiesel associated with lack of product credibility ad/or reliability as well as its

15 15 potetially egative side-effects. This is further reiforced by the fact that o buyers attitudes towards positive biodiesel aspects have bee statistically sigificat determiig WTP. Last, two other costructs sigificatly explai biodiesel WTP; subjective orms ad the perceived difficulty to purchase. The, cosumers with positive subjective orms will be more likely to pay a extra price for biodiesel while cosumers who perceives more difficulty i purchasig biodiesel will be less likely. As metioed, value fuctio for biodiesel differs from that of the more commo attribute of brad. For this cocept, socio-demographics are early the oly variables statistically sigificat explaiig WTP (AGE, UNIVERSITY, HIGH_INCOME, KIDS6 ad ELDERLY). Household icome (INCOME) ad havig attaied a uiversity degree (UNIVERSITY) icrease values attached to the petrol statio be associated to a atio-wide oil compay. However, older people measured by the age of respodet ad the presece of people with more tha 65 years i the household, egative ifluece this WTP. Those household with kids less tha 6 years old are less willig to pay for diesel i petrol statios associated to a atio-wide oil compay. Fially, those buyers who state that they usually fill diesel up i petrol statio associated to atio-wide oil compaies are more willig to pay for diesel i those petrol statios, reflectig the good predictive capacity of our model. CONCLUSIONS I this paper the potetial market for biodiesel has bee explored usig a choice experimet approach. Prelimiary fidigs show that cosumers are willig to pay a premium for biodiesel ad that this product ca have market success eve i the absece of public support policies. However, this must be cofirmed with a supply side aalysis of the extra-cost biodiesel must face whe eterig the market. Biodiesel valuatio does ot seem to resemble that of a commo product attribute such as brad ad is limited by the fact that overall opiio about is ot positive, both from a techical ad evirometal perspective. However, these results might be affected by the fact that fieldwork was udertake precisely durig the period i which higher food prices were a mai media topic ad biofuels were beig blamed for those higher prices. Moreover, our data set is quite limited both i size ad scope ad further research is eeded to cofirm whether these exploratory fidigs ca be extrapolated to society as a whole.

16 16 Refereces Adamowicz, Wiktor, Peter Boxall, Michael Williams ad Jorda Louviere (1998), Stated Preferece Approaches for Measurig Passive Use Values: Choice Experimets ad Cotiget Valuatio, America Joural of Agricultural Ecoomics, 80(1), Ajze I., (1991). The theory of plaed behaviur. Orgaizatioal Behavior ad Huma Decisio Process, 50, Campbell, Day (2007), Willigess to Pay for Rural Ladscape Improvemets: Combiig Mixed Logit ad Radom Effects Models, Joural of Agricultural Ecoomics, 58 (3), Greee, William H. (2002), NLOGIT Versio 3.0 Referece Guide. Plaiview, NY: Ecoometric Software Ic. (2008), Ecoometric Aalysis (6 th Editio). New Jersey: Pretice Hall Hesher, David A., Joh M. Rose ad William H. Greee (2005), Applied Choice Aalysis. A Primer. New York: Cambridge Uiversity Press. Lacaster, Kelvi (1966), A New Approach to Cosumer Theory, Joural of Political Ecoomy 74(2), Lusk, Jayso, Jutta Roose ad Joh A. Fox. (2003), Demad for Beef from Cattle Admiistered Growth Hormoes or Fed Geetically Modified Cor: A Compariso of Cosumers i Frace, Germay, the Uited Kigdom, ad the Uited States, America Joural of Agricultural Ecoomics, 85(1): McFadde, Daiel (1974), Coditioal Logit Aalysis of Qualitative Choice Behavior, i Frotiers i Ecoometrics, Paul Zarembka, ed. New York: Academic Press, STATA/IC StataCorp LP. ( Sparks P., Guthrie C.A., ad Shepherd R., The dimesio structure of the perceived behavioural cotrol costruct. Joural of Applied Social Psychology, 27, Trai, Keeth (1998), Recreatio Demad Models with Taste Differeces over People, Lad Ecoomics, 74 (2),

17 17 (1999), Halto Sequeces for Mixed Logit, Workig Paper No. E00-278, Departmet of Ecoomics, Uiversity of Califoria, Berkeley. (2003), Discrete Choice Methods with Simulatio. Cambridge (UK):Cambridge Uiversity Press. White, Halbert (1980), A Heteroscedasticity-Cosistet Covariace Matrix Estimator ad a Direct Test for Heteroscedasticity, Ecoometrica, 48,