Bio-based and Applied Economics 9(3): 225-240, 2020 ISSN 2280-6180 (print) © Firenze University Press ISSN 2280-6172 (online) www.fupress.com/bae Full Research Article DOI: 10.13128/bae-7768 Investigating determinants of choice and predicting market shares of renewable-based heating systems under alternative policy scenarios CRISTIANO FRANCESCHINIS*, MARA THIENE University of Padova, Italy Abstract. Fostering the uptake of heating technologies based on renewable resources is an important part of the EU energy policy. Yet, despite e!orts to promote their dif- fusion, heating systems based on fossil fuels are still predominant. In order to better tailor energy policies to citizens preferences, it is crucial to collect accurate informa- tion on their determinants of heating choices. At this purpose, we adopted a choice experiment and a latent class model to analyze preferences of householders in the Veneto region (North-East Italy) for di!erent heating systems and their key features. We focused on three devices based on biomass and three on fossil fuels, and account- ed for technical, economic and environmental characteristics of such systems. Model estimates highlight the presence of substantial preference heterogeneity among the population, which can be partially explained by citizens socio-demographics. We also use model outputs to simulate market shares for heating systems under alternative policy scenarios. Results provide interesting suggestions to inform the design of poli- cies aimed at fostering the adoption of biomass-based heating systems. Keywords. Ambient heating systems choice, Latent class Model, Market shares, Willingness to pay. JEL Codes. C01, Q42, Q47. 1. Introduction Developing a strategy to increase the sustainability of the heating sector is a priority for the European Union, in order to reduce energy imports and dependency and meet the greenhouse gas emission target established under the Paris Agreement. Currently, heating and cooling account for half of the EU energy consumption and 75% of the fuel used in this sector comes from non-renewable resources (European Commission, 2016). To tackle such issues, the 2030 Climate and Energy Policy Framework adopted by the European Council in 2014 includes three key targets for 2030: i) a 40% reduction of green- *Corresponding author. E-mail: cristiano.franceschinis@unipd.it Editor: Meri Raggi. 226 Cristiano Franceschinis, Mara Thiene house gas (GHG) emissions compared to 1990 levels, ii) a 27% share of renewable energy in gross "nal energy demand and iii) a 27% increase in energy e#ciency (European Council, 2014). $e targets for renewables and energy e#ciency were revised upwards in 2018 at 32% and 32.5% respectively. Member States are obliged to adopt integrated National Climate and Energy Plans (NECPs) for the period 2021-2030 to de"ne how they plan to achieve such goals. Member States submitted their dra% plans in 2018 and "nal plans must be submitted by the end of 2019. Italy, in its dra% plan, set the targets of a 33% reduction of GHG emis- sions compared to 2005 levels and a 30% share of renewable energy on "nal consumption to be achieved by 2030 (Italian Government, 2018). In 2017, the values for the two targets were 18% and 17% respectively. $us, as emphasized in the EU Country Report 2019 (European Commission, 2019), further e!orts are needed to ensure the achievement of 2030 objectives. Among the speci"c targets set by the plan, there is an annual increase of 1.3% of renewables share in the sector of residential heating and cooling. To achieve such target – among other measures – the plan aims to promote an active role by citizens on the energy demand mar- ket and the uptake of micro-generation technologies based on renewables. As such, instal- lation of renewable based residential heating systems in new buildings and replacement of fossil fuel technologies in existing ones plays a crucial role in the energy system transition. To entice the active participation of citizens, it is important to collect information on their heating preferences, in order to retrieve determinants of heating choices. Information about heating preferences can be collected via choice experiment, an increasingly popular method for stated preferences analysis. For example, Rommel and Sagebiel (2017) investigated preferences of German homeowners for micro-cogeneration units for residential use. $eir results suggest how householders have a strong interest in adopting such technologies, with willingness to pay (WTP) values ranging from 11.000 to 23.000 Euros. Features of micro-cogeneration products, as well as socio-demographics characteristics of houseowners, were found to substantially a!ect their WTP. Scarpa and Willis (2010) investigated WTP for the adoption of di!erent renewable micro-generation technologies in the UK. Speci"cally, they focused on solar photovoltaic, micro-wind, solar thermal, heat pumps, biomass boilers and pellet stoves. $eir results suggest that householders are willing to adopt such technologies, but for most of them WTP values do not cover capital cost. A similar study was carried out by Su et al. (2018) in Lithu- ania. Authors found householders to prefer solar energy-based technologies over the other renewable based ones. Claudy et al. (2010) also estimated consumers’ WTP for di!erent microgeneration technologies, namely micro wind turbines, wood pellet boilers, solar panels and solar water heaters. $e study showed how WTPs vary substantially among di!erent technologies and how consumers attitudes and beliefs about the technologies signi"cantly in&uence their WTPs. Rouvinen and Matero (2013) focused on preferences towards di!erent types of heating systems (based on fuel used) and examined the role of system features on householders’ choices in Finland. Investment cost was identi"ed as the most impactful attribute on householders’ decisions, but non-monetary attributes played a signi"cant role as well. Results also provided evidence of preference heterogeneity, par- tially linked to individuals’ characteristics. Similarly, Michelsen and Madlener (2012) ana- lyzed the in&uence of sensitivity to di!erent heating systems’ attributes on homeowners’ adoption decision. $eir "ndings suggest that importance attached to di!erent attributes a!ects technological features choice: for example, people focused on energy saving are 227Market shares of renewable-based heating systems more likely to adopt condensing boilers with thermal support, while consumers attach- ing a strong value to use of renewables prefer pellet-"red boilers. Furthermore, they found socio-demographics and spatial factors to a!ect preferences. Ruokamo (2016) explored homeowners’ attitudes towards innovative hybrid home heating systems, described in terms of fuel used and key features, such as costs, comfort of use and environmental impact. $e author found that such technologies are generally well accepted by house- owners and that their preferences are strongly a!ected by socio-demographic character- istics. Yoon et al. (2015) compared householders’ WTP for district heating and individual heating. While they found citizens to be generally willing to pay more for district heating, substantial di!erences emerged when accounting for preference heterogeneity: consumers with higher income and education were found to prefer district heating, while those more concerned about costs were willing to pay more for the individual one. In this paper we present the results of a choice experiment aimed at eliciting house- holders’ preferences towards di!erent heating systems in the Veneto region (Italy). Spe- ci"cally, we focus on six di!erent heating systems, three based on renewables (chip wood, "rewood and wood pellet) and three on fossil fuels (methane, oil and LPG). Veneto is a fairly populated region (almost "ve million residents) characterized by air pollution mainly related to high road tra#c in all main cities and by the presence of large industrial districts in several sectors, such as tanning, cement production and furniture manufacturing. When accounting speci"cally for carbon dioxide emission, the residential impact is substantial as well, around the 20% of total emissions (ARPAV, 2015). To decrease the negative impact of residential sector on the production of greenhouse gases, since 2014 the regional authority supports the purchase of biomass-based heating systems by annually allocating "nancial subsidies (up to '1,600 for stoves and '5,000 for boilers). Such policy, however, seems to only marginally meet the expectations of the popu- lation, in terms of fostering the adoption of such technologies. For example, in 2018 only 29 citizens applied for the funding and 25 requests were approved, for a total of around '55.000, out of '500,000 allocated policy budget. In 2019 the number of requests was high- er (76, of which 66 approved) but again most of the policy budget was not used (around '120,000 allocated out of a budget of '500,000)1. $us, a better understanding of underlying factors motivating householders to stick with a fossil fuel system or to switch to a renewable one, is of crucial to reach the goals of the energy transformation process in the region. Building on the evidence provided by the literature on how preferences towards dif- ferent heating systems are highly heterogeneous and on how socio-demographics play an important role in such variability, we adopt a Latent Class approach and use socio-demo- graphic characteristics of respondents to predict probability to belong to di!erent classes. $is approach allows to: i) identify di!erent segments of the population according to socio- demographic characteristics; ii) explore how preferences towards di!erent heating systems and their features vary across segments. We then use the estimates of such model to predict market shares for alternative heating systems within two policy scenarios, considering/based on a reduction of: i) investment costs for biomass fueled heating systems; ii) operating costs for biomass-based technologies. Both simulated scenarios are in line with the policies imple- mented by the Veneto region, the idea being that our empirical results may become useful 1 Data retrieved from https://www.regione.veneto.it/web/ambiente-e-territorio/rottamazione-stufe-bando-2019. 228 Cristiano Franceschinis, Mara Thiene to better tailor the features of such policies to the population of the region. Reduction of investment cost is a commonly adopted policy to foster use of renewable based technologies, such as the subsidies provided by the Veneto region. Reduction of operating costs is a pos- sible e!ect of targeted policies as well (e.g. subsidies on fuel purchase). $e objective of our study is twofold: on one hand to investigate how socio-demo- graphic characteristics in&uence citizens choice of heating systems, in order to gain insights on the determinants of adoption of renewable based technologies; on the other, to identify which, among a set of possible policy interventions, can be more e!ective in terms of fostering the di!usion of such technologies among the population. $e remainder of the paper is organized as follows: section 2 describes data collection (sampling procedure, survey design and administration); section 3 formally describes the econometric approach; section 4 reports the results of our study and section 5 draws its conclusions. 2. Data collection and survey $is section reports a succinct description of sampling procedure and survey. Further details can be found in Franceschinis et al. 2016, 2017. Data were collected with the support of a market research "rm via a web-based sur- vey addressed to a sample of householders of the Veneto region. We used a random sam- ple of householders, strati"ed on the main socio-demographics (age, education, gender, place of residence). A total of 1,557 questionnaires were collected, out of which 1,451 were complete and used for the analysis. $e questionnaire was structured in "ve sections. $e "rst focused on heating system and energy resources currently used by respondents. $e following section included the choice experiment, which is described in detail below. $e third section included follow-up questions about the choices made in the previous section. $e fourth section presented attitudinal questions related to respondents’ psychological traits. $e last collected socio-demographic information. $e choice experiment involved a hypothetical scenario in which respondents were asked to select the heating system they would adopt if they had to renovate their cur- rent one among a set of alternative options. $e heating systems presented to respondents were six, three based on biomass ("rewood, chip wood and wood pellet) and three on fossil fuels (methane, LPG and oil). Each alternative system was described in terms of six attributes: i) investment cost, ii) investment duration, iii) annual operating cost, iv) CO2 emissions, v) "ne particle emissions and vi) required own work. $e respective levels were system-speci"c and are reported in Table 1. Investment cost is the cost for purchasing and installing the heating system. Possible public incentives were not accounted for in de"ning the levels of the attribute. Investment duration refers to the lifespan of the heating system, from purchase to dismantling. Operating costs include fuel price, maintenance costs and electricity costs for those systems that need it to work. CO2 emissions and "ne particles emissions refer to the quantity of CO2 and "ne particles released by the fuel combustion processes. To facilitate the evaluation of CO2 emissions levels, respondents were informed that 1,000 kg of CO2 corresponds to the emissions from driving 6,000 km in a new gen- eration car. To illustrate "ne particles health impacts, respondents were informed that “it has been estimated that if annual "ne particle emissions for one house are 2,000 g, then 229Market shares of renewable-based heating systems the total emissions of 10,000 similar houses cause one premature death per year”. Final- ly, required own work refers to the time required to ensure the faultless operation of the heating system (e.g., cleaning and handling fuel loads). $e choice of attributes and their levels was based on earlier studies and on feedback from experts. $e annual operating cost and CO2 and "ne particle emissions were computed using as reference the energy consumption of an average detached house with a living area of 120 m2. $e experimental design adopted in the study was an e#cient availability design (Rose et al., 2013), according to which only three alternatives were shown in each choice task. $e combination of levels that appeared in each scenario was de"ned with three dif- ferent sub designs, namely near orthogonal, D-e#cient (Scarpa and Rose, 2008; Rose and Bliemer, 2009) and serial designs. For the latter, an orthogonal design was used for the "rst respondent. A%er the choice sequence was completed, a multinomial logit model was estimated in the background and statistically signi"cant parameters were used as priors to generate an e#cient design. $is process continued a%er each respondent and priors were continuously updated to generate a gradually more e#cient design. Overall, the design generated 60 choice scenarios blocked in six groups, so that each respondent faced 10 of them. $e sample was split so to have the same number of respondents assigned to the three di!erent sub designs. An example of choice scenario is reported in Table 2. 3. Econometric approach In our study we estimated a latent class model to investigate variation of tastes towards heating systems types and their features among the householders of the Veneto Table 1. Choice Experiment attributes and levels. Attributes Firewood Chip wood Wood Pellet Methane Oil LP Gas Investment cost (') 9,500; 11,000; 12,500 11,500; 13,000; 14,500 13,000; 15,000; 17,000 4,000 4,800; 5,600 4,500; 5,500; 6,500 4,000; 5,000; 6,000 Investment duration (y) 15; 17; 19 17; 20; 23 16; 19; 22 16; 18; 20 16; 18; 20 14; 17; 20 Operating cost ('/y) 1,200; 2,000; 2,800 2,000; 2,800; 3,600 2,500; 3,750; 5,000 4,000; 5,500; 7,000 6,000; 8,000; 10,000 9,000; 12,500; 16,000 CO2 Emissions (kg/y) 150; 225; 300 300; 375; 450 375; 450; 525 3,000; 3,750; 4,500 3,900; 4,575; 5,250 3,525; 4,125; 4,725 Fine particle emissions (g/y) 4,500; 6,000; 7,500 2,250; 3,750; 5,250 750; 1,500; 2,250 15; 30; 45 150; 450; 750 15; 30; 45 Required own work (h/m) 5; 10; 15 1; 2; 3 1; 2; 3 - 0.5; 1; 1.5 0.5; 1; 1.5 230 Cristiano Franceschinis, Mara Thiene Region. $e model is based on the Random Utility $eory (Luce, 1959; McFadden, 1974), according to which a respondent n facing a set of J mutually exclusive alternatives has utility Ui for alternative i as a function of attributes Xk, so that: Uni=!xni+"ni (1) where "ni is the unobserved error assumed to be i.i.d. extreme value type I. To account for heterogeneity in sensitivity to attributes Xk, we adopted a latent class model. Such model assumes the existence of C classes of respondents, where C is exog- enously de"ned by the analyst, based on information criteria indexes. Preference vary across classes but are homogeneous within them. As the classes are latent, an equation explaining the probabilistic assignment of individual n into class c needs to be de"ned. Using a logit formulation for the class allocation model, with Zn being a vector of socio- economic variables and (c a vector of estimated coe#cients, the probability that individual n belongs to segment c is given by (Bhat, 1997): (2) Speci"cally, the variables we used in Z vector are: i) age, ii) education, iii) income, iv) currently owning a biomass-based heating system. $en, the probability that individual n chooses alternative i, conditional on belonging to class c, takes the logit form (Hensher and Greene, 2003): (3) Where Xi represents the vector of attributes associated with each alternative and )nc the vector of estimated coe#cients for class c. $e estimated parameters of the latent class model were used to simulate the market shares of di!erent heating systems under di!erent policy scenarios. Speci"cally, the sce- narios involve reductions of investment cost (ranging from none to 50% reduction) and operational costs (same levels as previous case) for biomass-based heating systems. We Table 2. Example of choice scenario. Attributes Wood Pellet LP Gas Firewood Investment Duration (years) 19 20 19 Fine particles emissions (g/year) 2,250 15 7,500 CO2 emission (kg/year) 375 3,525 150 Required own work (hours/month) 1 1 15 Investment cost (') 17,000 5,000 12,500 Operative cost (') 3,750 9,000 1,200 Your choice ○ ○ ○ 231Market shares of renewable-based heating systems computed choice probabilities in each scenario with the logit formula described in Equa- tion 3, by including in it estimated coe#cients )nc and by varying the levels of investment and operational costs according to the reduction scenarios. 4. Results $is section reports the results of our study. In the "rst part of the section the estimates of the latent class model are presented, while the second part focuses on the policy scenarios. 4.1 LC model estimates $e "rst step of our modelling approach involves the identi"cation of the optimal number of classes. As suggested by the literature (Hurvich and Tsai, 1989; Nylund et al., 2007), we referred to the AIC and BIC information criteria, which both favour a speci"ca- tion with 4 classes (Table 3). Class membership probabilities are 23% for class 1, 36% for class 2, 16% for class 3 and 25% for class 4 (Table 4). Results of latent class model with four classes are reported in Table 22. $e table also reports WTP values for heating sys- tems features, which were computed with respect to the investment cost. To class 1 are more likely to belong older individuals with low income and education, who currently do not own a biomass-based heating system. Such class exhibits a strong preference towards methane-fuelled technologies with seemingly no interest in biomass- based ones. As it concerns the attributes, it can be noticed how members of this class are very sensitive to installation and operational costs, which is consistent with their feature of individuals with low income. $is class seems also sensitive to technical features of heating systems, and it shows a preference for systems with a long duration (WTP value of '0.38 for each additional year of duration) and which require low amount of time for maintenance ('0.27 to avoid an hour of work per month). Emissions, instead, do not seem to a!ect choices of members of this class. Moving to class 2, to this class are more likely to belong younger individuals with high education and income who currently do not possess a biomass-based technology. 2 A part of these results was included in the report “Veneto 100% Rinnovabile: fotogra"a e prospettive” by the Interdepartmental Centre Giorgio Levi Cases for Energy Economics and Technology (University of Padova), available at http://levicases.unipd.it/wp-content/uploads/2019/11/Relazione-"nale.pdf Table 3. Information criteria for alternative model speci!cations. Number of classes Number of parameters LL AIC BIC 1 (MNL) 11 -15,713 31,448 31,506 2 28 -15,081 30,218 30,367 3 44 -15,017 30,122 30,356 4 60 -14,894 29,908 30,227 5 76 -14,886 29,924 30,328 232 Cristiano Franceschinis, Mara Thiene As it concerns preferences towards di!erent types of heating system, it can be noticed how members of such class show a strong interest in biomass fuelled system, especially those based on wood pellet. $is suggests that such class has a strong potential in terms of switching from a fossil fuel-based system to a biomass one. $is seems corroborated by the high sensitivity to carbon dioxide and "ne particles emissions of members of this class, which are those willing to pay the most to avoid them among all classes ('0.73/kg/ year for CO2 and '0.28/g/year for "ne particles). At the opposite, in this class there seems to be no concern about technical features of heating systems. To class 3, instead, are more likely to belong older individuals who currently own bio- mass-based heating system. As in the previous class, there seems to be a strong interest in biomass technologies, but in this case, "rewood is the preferred fuel. As it concerns heat- ing systems features, members of this class seem to strongly appreciate technologies with long investment duration ('1.58 for each additional year, largest value among all classes) and low emissions of carbon dioxide ('0.36 to avoid a kilogram per year). Finally, members of class 4 (the baseline class) seem to be interested mainly in meth- ane-based systems and in those fuelled by wood pellet. $ey also have a strong aversion to oil-based technologies. As it concerns systems’ features, they seem interested in low main- Table 4. LC model results. Class size Class 1 23% Class 2 36% Class 3 16% Class 4 25% Estimate WTP Estimate WTP Estimate WTP Estimate WTP Class membership function Intercept 0.16 0.24 -0.11 -0.07 Age 0.31 -0.31 0.12 Degree -0.22 0.43 -0.23 Income -0.15 0.32 0.02 Owning a biomass fuelled system -0.22 -0.11 0.19 Heating system features Investment cost -0.41 -0.89 -0.18 -0.49 Operational cost -0.38 -0.95 -0.31 -0.46 Investment duration 0.16 0.38 0.34 0.38 0.28 1.58 0.17 0.35 Required own work -0.11 -0.27 -0.24 -0.27 0.09 0.50 -1.41 -2.87 CO2 emissions -0.01 -0.03 -0.73 -0.82 -0.06 -0.36 -0.02 -0.04 Fine particle emissions 0.04 0.10 -0.28 -0.31 -0.05 -0.25 0.01 -0.01 Heating system type Firewood 2.00 4.99 1.89 1.55 Chip wood -3.96 1.94 0.99 1.21 Wood pellet 0.42 10.88 0.46 4.20 Methane 4.81 7.65 0.19 6.29 Oil -0.39 2.14 -0.04 -1.26 Note: Coe"cients statistically signi!cant at 95% in bold. WTP values were computed with respect to the investment cost. 233Market shares of renewable-based heating systems tenance requirements and to a lesser degree to avoiding emissions, with low WTP values ('0.04/kg/year to avoid CO2 and '0.01/g/year to avoid "ne particles). 4.2 Market shares for di!erent heating systems in alternative policy scenarios In this sub-section we report results of market shares simulations for di!erent heat- ing systems in two sets of policy scenarios: i) reduction of investment cost for biomass fuelled heating systems, speci"cally none, 10%, 20%, 30%, 40% and 50%; ii) reduction of operating costs for biomass based technologies (same range as above). In the "rst part of the sub-section (4.2.1) we present the average market shares weighted by class size for dif- ferent heating systems; in the second (4.2.2) we report the shares for biomass technologies within each class. 4.2.1 Average market shares in the investment cost reduction scenarios Table 5 and Figure 1 illustrate weighted average market shares for di!erent heating systems under the investment cost reduction scenario. In the baseline scenario, i.e. no investment cost reduction, most of the population would choose a methane heating sys- tem to replace the current one (64.40%), followed by wood pellet (12.61%), LPG (8.82%), "rewood (7.35%), oil (5.39%) and chip wood (1.23%). Moving to the 10% reduction scenario, it can be noticed how shares for biomass fuelled technologies slightly increase (around 1% for each system). A 20% reduction seems to trigger a stronger response, with an increase of around 3% for wood pellet and 1.5% for the other biomass-based systems compared to the 10% reduction scenario. In the 30% reduction scenario the share for bio- mass-based systems further increases, in particular for wood pellet technologies, with a share of around 21% compared to the 12.61% of the baseline scenario. $e fourth scenario (40% reduction), instead, does not show substantial di!erences compared to the previous one. Finally, in the last scenario (50% reduction) around 30% of citizens would choose a wood pellet "red system, around the 17% a "rewood one and around the 5% a chip wood one. Overall, in such scenario, nearly half of the population would choose to adopt a bio- Table 5. Average market shares under the investment cost reduction scenarios. Investment cost reduction Heating system None (baseline) 10% 20% 30% 40% 50% Chip wood 1.23 1.83 2.69 3.21 4.55 5.49 Firewood 7.35 7.90 10.32 13.86 14.93 16.59 Wood pellet 12.61 13.94 15.21 21.36 23.79 26.91 Biomass total 21.19 23.67 28.22 38.43 43.27 49.00 Methane 64.60 63.66 62.49 55.91 52.32 47.33 Oil 5.39 4.84 3.53 2.31 1.70 1.21 LPG 8.82 7.84 5.76 3.36 2.71 2.46 Total 100.00 100.00 100.00 100.00 100.00 100.00 234 Cristiano Franceschinis, Mara Thiene mass-based system. $e overall increases of the share of biomass-based systems compared to the baseline scenario are: 2.5%, 7%, 17.2%, 22% and 28% for the alternative investment cost reductions. 4.2.2 Average market shares in the operational costs reduction scenarios Table 6 and Figure 2 report the estimated average shares under the alternative opera- tional cost reduction scenarios. Firstly, it is of interest to notice how reducing operational costs seems to have a stronger e!ect in terms of increasing biomass systems market shares compared to the reduction of investment cost. $is seems true in each scenario (i.e. for each magnitude of the reduction) and is particularly evident in the 50% reduction sce- nario, under which the overall biomass technologies share is around 54% for operational costs reduction and 49% for investment cost reduction. In terms of fostering di!usion of biomass "red systems, this seems to suggest how policies aimed at decreasing operational costs for citizens may be more e!ective than those providing a reduction of the invest- ment cost. By looking more closely at the operational costs reduction scenarios, it can be noticed that, similarly to the previous scenario, a 10% reduction does not lead to a substantial increase in the shares of biomass technologies (between 1% and 2% increase for each system). A 20% reduction has an only slightly stronger e!ect, with an increase of about the 3% for biomass systems shares compared to the previous scenario. A similar relative increase is also shown for the 30% and 40% reduction scenarios. Finally, in the last sce- nario there is a substantial increase in the shares for biomass technologies, especially as it concerns wood pellet, which would be chosen by around the 30% of the population. Overall, it seems that a reduction of the operational costs would strongly favour the dif- 0 10 20 30 40 50 60 70 80 90 100 None 10% 20% 30% 40% 50% M ar ke t s ha re (% ) Investment cost reduction Chip wood Fire wood Wood pellet Methane Oil LPG Figure 1. Average market shares under the investment cost reduction scenarios. 235Market shares of renewable-based heating systems fusion of wood pellet systems and only to a lesser degree the di!usion of other biomass- based systems. $is may be due to higher operational costs of pellet "red heating systems, compared to other biomass ones. 4.2.3 Market shares in the investment cost reduction scenarios within each class In this section we move from the population-level picture to a class-speci"c analysis, to explore how preference heterogeneity in&uences the di!usion of biomass heating sys- tems. Speci"cally, we report and discuss probabilities to choose a biomass-based technol- ogy as replacement of the current one within each class in di!erent policy scenarios. Table 6. Average market shares under the operational costs reduction scenarios. Operational costs reduction Heating system None (baseline) 10% 20% 30% 40% 50% Chip wood 1.23 2.11 3.37 3.88 4.91 6.11 Firewood 7.35 8.65 11.21 14.65 15.32 17.71 Wood pellet 12.61 14.99 18.07 23.16 25.89 29.94 Biomass total 21.19 25.75 32.65 41.69 46.12 53.76 Methane 64.60 63.19 60.66 54.08 51.80 45.19 Oil 5.39 4.65 2.60 1.71 0.98 0.46 LPG 8.82 6.41 4.09 2.52 1.10 0.59 Total 100.00 100.00 100.00 100.00 100.00 100.00 Figure 2. Average market shares for under the operational costs reduction scenarios. 0 10 20 30 40 50 60 70 80 90 100 None 10% 20% 30% 40% 50% M ar ke t s ha re (% ) Operational costs reduction Chip wood Fire wood Wood pellet Methane Oil LPG 236 Cristiano Franceschinis, Mara Thiene Starting from class 1, Table 7 and Figure 3 show how in class 1 the market share for biomass devices in the baseline scenario is extremely low (2.60%). Such value is consist- ent with the pro"le illustrated in Section 4.1, which highlighted how members of this class are characterized by little interest in biomass technologies and absence of sensitivity to car- bon emissions. Moving to the cost reduction scenarios, their e!ect on adoption probability seems limited. Only in the 50% reduction scenario there seems to be a substantial increase in the biomass share (8% increase compared to the 40% reduction scenario). It seems that a very strong incentive is needed to foster di!usion of renewable based systems in this class. Moving to class 2, the share for biomass devices in the baseline scenario equals 32.11%. Such results – together with the LC estimates – suggest that this class includes individuals who currently own a fossil fuel "red heating system and around one third of them would switch to a biomass fuelled one, even with no cost reduction. $is seems to corroborate the potential of this class in terms of increasing the di!usion of renew- Table 7. Class-speci!c biomass systems market shares under the investment cost reduction scenarios. Class Investment cost reduction None (baseline) 10% 20% 30% 40% 50% Class 1 2.60 3.39 5.91 8.66 12.11 19.99 Class 2 32.11 36.11 40.18 44.18 48.12 50.08 Class 3 61.16 64.12 73.18 78.81 82.18 88.88 Class 4 19.98 23.11 27.61 32.81 38.11 45.18 Figure 3. Class-speci!c biomass systems market shares under the investment cost reduction scenarios. 0 10 20 30 40 50 60 70 80 90 100 None 10% 20% 30% 40% 50% B io m as s m ar ke t sh ar e (% ) Investment cost reduction Class 1 Class 2 Class 3 Class 4 237Market shares of renewable-based heating systems able based technologies across the population. As for the previous class, the e!ect of the investment cost reduction seems limited. In this case, however, the low e!ect may be linked to the high income of its members, that could make them less sensitive to costs. Class 3 exhibits the highest biomass system adoption probability in the baseline scenario (61.6%). Overall, this class seems characterized by individuals that currently use a biomass system and show a high probability of choosing one of the same kind as replacement. Importantly, this class seems to be strongly a!ected by cost reduction, with an around 28% increase of the biomass devices share between the baseline and the 50% reduction scenario. $is might be due to the low income of members of this class. Finally, biomass systems share in class 4 equals 19.89% in the baseline scenario and 45.18% in the 50% reduction one, thereby suggesting a high sensitivity to investment cost reduction. 4.2.4 Market shares in the operational costs reduction scenarios within each class Table 8 and Figure 4 report market shares within each class in the operational costs reduction scenarios. By comparing results with those reported in the previous section, it is interesting to notice how class 2 and 3 are a!ected more strongly by operational costs reduction, while classes 1 and 4 are a!ected more by investment cost reduction. $is seems to be related to di!erent sensitivity to the two costs highlighted in Section 4.1: classes 2 and 3 are more sensitive to operational costs, and as such reducing it has a stronger e!ect in such classes, while the opposite is true for classes 1 and 4. For example, in class 1, at a 50% reduction the share is around 15% for operational costs and 20% for investment cost. At the opposite, in class 3 the share in 5% higher in the case of operational cost reduction. 5. Discussion and conclusions In the light of the importance of increasing the sustainability of the residential heating sector, it is crucial to inform energy policies with an accurate knowledge of the determi- nants of citizens heating choices. To this purpose, we designed a choice experiment aimed at investigating preferences towards heating systems and their features among the citizens of the Veneto region. We analysed choice data by means of a latent class model and we used the estimates to forecast market shares for di!erent heating systems under alternative policies scenarios. Table 8. Class-speci!c biomass systems market shares under the operational costs reduction scenarios. Class Operational costs reduction None (baseline) 10% 20% 30% 40% 50% Class 1 2.60 3.11 4.88 7.91 10.12 15.18 Class 2 32.11 37.14 42.24 46.11 49.88 53.11 Class 3 61.16 66.89 76.11 81.56 86.88 92.21 Class 4 19.98 22.41 26.22 30.33 34.16 40.11 238 Cristiano Franceschinis, Mara Thiene $e results of our study suggest how householders’ preferences towards di!erent heating systems and their features are strongly heterogeneous and how such variability can be par- tially explained by householders socio-demographic characteristics. Such "ndings support those of previous studies of the energy literature (e.g. Yoon et al., 2015, Ruokamo, 2016). Importantly, our estimates highlight the presence of population segments which seem to have a strong potential in terms of switching from a fossil fuel system to a renewable-based one. To this segment are more likely to belong individuals who already own a biomass- based heating systems and young individuals with high income and education level. At the opposite, our results highlight the existence of segments of the population with low interest towards the adoption of renewable-based technologies. Such segments are characterized by individuals with low income and education who currently own a fossil fuel system. $e simulations of policy scenarios allowed us to retrieve some important information about the e#ciency of di!erent policy measures in terms of fostering the di!usion on bio- mass technologies. Overall, we found that measures aimed at reducing operational costs for householders may induce a broader uptake of biomass appliances compared to those which target investment cost, even if the opposite is true in some segments of the population. $is is particularly important in the context of the Veneto region, where subsidies for invest- ment cost are currently in place and they seem to be only partially successful in nudging citizens towards the adoption of biomass-based appliances. We also found that only a large reduction of costs (i.e. 40% or 50% reduction) has a substantial e!ect on the increase of biomass systems shares, in classes with low interest in such technologies. $is suggests that current incentives provided by authorities may not be enough to entice such segments of the population to switch from a fossil fuel to a biomass-based technology. Figure 4. Class-speci!c biomass systems market shares under the operational costs reduction scenar- ios. 0 10 20 30 40 50 60 70 80 90 100 None 10% 20% 30% 40% 50% Bi om as s m ar ke t s ha re (% ) Operational costs reduction Class 1 Class 2 Class 3 Class 4 239Market shares of renewable-based heating systems Acknowledgments $e research was funded by Interdepartmental Centre Giorgio Levi Cases for Energy Economics and Technology (University of Padova), “Sustainability of introduction of pel- let based heating systems in a mountain area”. References ARPAV, 2015. Inventario regionale delle emissioni in atmosfera. Bhat, C.R. 1997. An endogenous segmentation mode choice model with an application to intercity travel. Transportation Science, 31(1):34–48. Claudy, M.C., Michelsen, C., O’Driscoll, A., Mullen, M.R. 2010. Consumer awareness in the adoption of microgeneration technologies: an empirical investigation in the Republic of Ireland. Renewable & Sustainable Energy Reviews, 14(7):2154-2160. European Commission. 2016. An EU Strategy on Heating and Cooling. https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:52016DC0051&from= EN European Commission. 2019. Country Report Italy 2019. https://ec.europa.eu/info/sites/info/files/file_import/2019-european-semester-country- report-italy_en.pdf European Council. 2014.: Conclusions. European Council, General Secretariat of $e Council. https://www.consilium.europa.eu/uedocs/cms_data/docs/pressdata/en/ ec/145397.pdf Franceschinis, C., Scarpa, R., $iene, M., Rose, J., Moretto, M., Cavalli, R. 2016. Exploring the Spatial Heterogeneity of Individual Preferences for Ambient Heating Systems. Energies, 9, 407. Franceschinis, C., $iene, M., Scarpa, R., Rose, J., Moretto, M., Cavalli, R. 2017. Adop- tion of Renewable heating systems: an empirical test of the di!usion of innovation theory. Energy, 125:313-326. Hensher, D. and Greene, W. (2003). $e mixed logit model: the state of practice. Transpor- tation, 30(2):133–176. Hurvich, M., and Tsai, C. (1989). Regression and time series model selection in small samples. Biometrika, 76: 297–307. Italian Government. 2018. Proposta Di Piano Nazionale Integrato Per L’energia E Il Clima. https://www.mise.gov.it/images/stories/documenti/Proposta_di_Piano_Nazionale_Integra- to_per_Energia_e_il_Clima_Italiano.pdf Luce, D. 1959. Individual choice behavior: A theoretical analysis. John Wiley and Sons. McFadden, D. 1974. Conditional logit analysis of qualitative choice behavior. In Zarembka P. (ed), Frontiers in econometrics. Academic Press, New York, pp 105–142. Michelsen, C.C., Madlener, R. 2012. Homeowners’ preferences for adopting innovative residential heating systems: A discrete choice analysis for Germany. Energy Econom- ics, 34(5):1271-1283. Nylund, K. L., Asparouhov, T. and Muthén, B.O. (2007). Deciding on the number of class- es in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural Equation Modeling: A Multidisciplinary Journal, 14(4): 535-569. 240 Cristiano Franceschinis, Mara Thiene Rommel, K., Sagebiel, J. 2017. Preferences for micro-cogeneration in Germany: Policy implications for grid expansion from a discrete choice experiment. Applied Energy, 206:612-622. Rose, J.M., Bliemer, M.C.J. 2009. Constructing efficient stated choice experimental designs. Transport Reviews, 29(5):587-617. Rose, J.M., Louviere, J.J., Bliemer, M.C.J. 2013. E#cient stated choice designs allowing for variable choice set sizes. In: International choice modelling conference, Sydney, Aus- tralia, 3rd-5th July 2013. Rouvinen, S., Matero, J. 2013. Stated preferences of Finnish private homeowners for resi- dential heating systems: A discrete choice experiment. Biomass and Bioenergy, 57:22–32. Ruokamo, E. 2016. Household preferences of hybrid home heating systems – A choice experiment application. Energy Policy, 95:224-237. Scarpa, R., Rose, J.M. 2008. Design e#ciency for non-market valuation with choice mod- elling: how to measure it, what to report and why. Australian Journal of Agricultural and Resource Economics, 52:253-282. Scarpa, R., Willis, K. 2010. Willingness-to-pay for renewable energy: Primary and discre- tionary choice of British households’ for micro-generation technologies. Energy Eco- nomics, 32(1):129-136. Su, W., Mengling, L., Zeng, S., *treimikien+, D., Bale,entis, T., Ali-auskait+-*e-kien+c, I. 2018. Valuating renewable microgeneration technologies in Lithuanian households: A study on willingness to pay. Journal of Cleaner Production, 191:318-329. Yoon, T., Ma, Y., Rhodes, C. 2015. Individual Heating systems vs. District Heating sys- tems: What will consumers pay for convenience? Energy Policy, 86:7. Investigating determinants of choice and predicting market shares of renewable-based heating systems under alternative policy scenarios Cristiano Franceschinis, Mara Thiene Multi-country stated preferences choice analysis for fresh tomatoes Maria De Salvo1,*, Riccardo Scarpa2,3,4, Roberta Capitello2, Diego Begalli2 “Not my cup of coffee”. Farmers’ preferences for coffee variety traits. Lessons for crop breeding in the age of climate change Abrha Megos Meressa, Ståle Navrud* Does the place of residence affect land use preferences? Evidence from a choice experiment in Germany Julian Sagebiel1,*, Klaus Glenk2, Jürgen Meyerhoff3 The use of latent variable models in policy: A road fraught with peril? Danny Campbell*, Erlend Dancke Sandorf