*Article* **The Eucalyptus Firewood: Understanding Consumers' Behaviour and Motivations**

### **Nadia Palmieri \*, Alessandro Suardi, Francesco Latteriniand Luigi Pari**

CREA Research Centre for Engineering and Agro-Food Processing, Via della Pascolare, 16, Monterotondo, 00015 Rome, Italy; alessandro.suardi@crea.gov.it (A.S.); francesco.latterini@crea.gov.it (F.L.); luigi.pari@crea.gov.it (L.P.)

**\*** Correspondence: nadia.palmieri@crea.gov.it; Tel.: +39-06-9067-5219

Received: 1 October 2020; Accepted: 28 October 2020; Published: 30 October 2020

**Abstract:** Italy is one of the world's major importers of firewood, despite the large amount of Italian eucalyptus plantations that could satisfy part of the country's internal demand. The demand is critical for farmers to understand developing market dynamics and people's willingness to buy a product is related to several parameters, including different supply methods. This study aimed to analyse the willingness to consume domestic eucalyptus firewood, and the related motivations of consumers considering the preferred supply method. Data was collected through a web-survey and analysed applying a multilevel regression. In general, the sample showed that attention is paid to both the type of wood and its origin, and that there is a preference for loose firewood as a supply method. Our findings suggest that factors such as age, experience, and familiarity with a product, the supply method, attitude towards novelty, provenience, and energetic density of firewood have an important role in shaping individual inclination towards consuming domestic eucalyptus firewood. This implies that the owners of eucalyptus plantations should target mostly young and detail-oriented consumers, and should also try to clearly give information regarding the origin of the product and its technical characteristics.

**Keywords:** consumer choices; eucalyptus; firewood; Italy; multilevel logistic regression model; willingness to consume

#### **1. Introduction**

Eucalyptus (*Eucalyptus* spp.) forests and agro-forestry plants cover about 20 million hectares in the world. These species show faster biomass growth if compared to other species [1] and can be used to obtain pulp, paper, and firewood. In regards to the environment, the management of eucalyptus plantations can be considered more sustainable than other energy crops [2]. Due to its fast growth and modesty, eucalyptus can contribute to biodiversity conservation [1,3]. Moreover, it shows an important role in climate change mitigation [1] due to its high capacity of carbon sequestration during growth [4,5].

In Italy, agro-forestry plants mainly aim towards the bioenergy production of eucalyptus (*Eucalyptus* spp.), poplar (*Populus* spp.), and black locust (*Robinia pseudoacacia* L.), which cover more than 100,000 ha [6]. Considering this, eucalyptus could provide a biomass that can fulfill about 72% of Italian demand [7–9]. Notwithstanding this large availability of wooden biomass, Italy is one of the major global importers of wood for energy purpose [10], used particularly for domestic heating [11,12]. The major part of imported fuelwood originates from the Balkans' area, and this is mainly due to the lower cost of labour and of raw materials in these countries [11,13].

Considering the growing importance of Eucalyptus plantations throughout the world, many researches have focused on both environmental impacts related to eucalyptus management [1,4,14–16] and cultivation methodology [10]. However, lower attention has been put on the economic aspects of

cultivation, i.e., investigation of the supply chain analysis [2,17,18] as well as of the demand of such species' wood.

Consumer choice is a key variable for farmers to understand developing market dynamics [19], moreover people's willingness to buy a given product is related to several parameters [20]. In fact, consumer attitude is an important aspect to analyse because it allows us to study the acceptance of a particular good by consumers [21]. According to this, it could be important to investigate the willingness to consume domestic eucalyptus fuelwood, given that eucalyptus firewood shows similar heating value to other species, such as oak [22], and thus it could be an interesting alternative.

In this framework, the study aimed to understand people's willingness to consume domestic eucalyptus firewood and their motivations, considering also people's preferred supply method. This last parameter is indeed important in consumer behaviour [23], even if, to the best of our knowledge, current literature studies that consider this wood supply chain are lacking.

This paper is structured as follows: Section 2 describes the materials and methods used. The results and discussion are presented in Section 3 and the conclusions are handled in Section 4.

#### **2. Materials and Methods**

#### *2.1. Data Collection, Sample and Questionnaire*

Data was collected through a web-based survey performed during the period January–April 2020, from an initial sample of 300 Italian people. In particular, consumers were recruited through invitations to participate in the online survey (performed by Google drive) via social networks (Instagram, Twitter, and Facebook). Moreover, snowball sampling recruitment was also adopted, using the interpersonal relations of the authors (via email) to reach a larger number of participants [24]. For these reasons, the sample was not representative of the Italian population, which happens in many studies about consumer behaviour where a convenience sample was used [21,25–30]. Subsequently, 18 respondents were excluded from the survey because they stated not to be domestic firewood consumers. The final sample was consequently made up of 282 consumers. Before starting the survey, an interview with 80 consumers was carried out in order to understand if the investigated topic was understandable through our questionnaire.

The questionnaire was made up of three sections: (1) Consumers' behaviour towards firewood; (2) consumers' willingness to consume domestic eucalyptus firewood and related motivations; and (3) socio-demographic features of respondents.

The first two sections applied a five-point Likert scale (1 = totally disagree; 2 = disagree; 3 = indifferent; 4 = agree, and 5 = totally agree), with the exception of some questions in Section 2 in which categorical variables (i.e., *Forn*, *Will*, *Cons*, and *Familiarity*) have been used [23]. Furthermore, the Cronbach Alpha coefficient for each item group was calculated to assess the reliability of the scale, which showed a good level (from 0.70 to 0.90).

In the first section of the questionnaire we analysed the consumers' attitudes about firewood species, ethical aspects of the choice, geographic provenience (*Prov*) (i.e., if firewood comes from tropical countries or Mediterranean ones), and the origin of firewood (*Origin*) (if firewood comes from an agro-forestry plant or natural stand) [29].

The second part of the questionnaire investigated respondents' willingness to consume domestic eucalyptus fuelwood (*Will*) with a binary choice (Yes vs. No), their willingness to pay per 100 kg of biomass (*Price*), and the amount of eucalyptus firewood which the consumer would be willing to consume yearly (*Will\_q*). It is important to underline that, if respondents were not willing to consume it, the quantity of eucalyptus firewood was considered zero.

The questionnaire also requested to indicate the consumer's familiarity (*Familiarity*) with eucalyptus fuelwood and if they have consumed it in the past (*Cons*) [30], and in both cases the question was asked as a binary choice (Yes vs. No). It is important to underline that familiarity was investigated using the question: "Have you ever heard about eucalyptus as firewood alternative?"

Moreover, another important aspect considered in the survey is consumers' motivation to use eucalyptus firewood. These aspects were investigated by asking questions related to curiosity (*Curiosity*), to technical characteristics (*Energetic*), as well as to environmental aspects [29]. In particular, the question about curiosity was "You are willing to consume domestic eucalyptus firewood for curiosity (How much do you agree with the following statements? Express your judgment by putting a tick from 1 to 5. 1 = totally disagree. 5 = totally agree)", while questions on technical characteristics were "You are willing to consume domestic eucalyptus firewood if it had better combustion behaviour (wood burning duration) than other firewood species (How much do you agree with the following statements? Express your judgment by putting a tick from 1 to 5. 1 = totally disagree. 5 = totally agree)" [23]. Moreover, it is important to highlight that the questions concerning the environmental aspects of firewood were asked to respondents but these questions were subsequently excluded because, using the stepwise procedure, they did not fit in the applied model. Finally, respondents had to select their preferred supply method (*Forn*), i.e., loose firewood, firewood arranged in pallets, or firewood in 10–15 kg bags. Other questions about where consumers usually buy firewood (i.e., *Woodman*, *Retail*, *Internet*, etc.) were asked to respondents but these questions were also excluded because they did not fit in the applied model.

In the last section of the questionnaire, data about socio-demographic features of the sample, such as age (*Age*), gender, area of residence, and education level were collected [31–34].

#### *2.2. Statistical Analysis*

A logistic regression model describes the relationships between the willingness to consume a particular product and the motivations of consumers, without including the variability among predictors of different levels (in case of data with nested structure) [35]. Given the nested structure of the studied sample, this study utilized multilevel logistic regression model to understand the consumers' willingness to consume domestic eucalyptus firewood and their motivations, considering the people's preferred supply method.

The multilevel logistic regression model consists of an extension of the regression model in which data are arranged in groups and coefficients can differ among the various groups [19].

In particular, several steps were followed to lead the analysis [19,35]. In fact, in the first step, it is necessary to understand if the dataset show a nested structure calculating the intra-class correlation (ICC) coefficient [36]. After such a check, analysis can be carried out, calculating, and comparing: The simplest two level model, an intermediate model, and the full multilevel logistic regression model [35]. This procedure allowed us to have a final model accounting both for the effects of the lower-level predictor variables and for higher level ones [35].

In particular, following some authors [19,35] and as above mentioned, to examine the existence of a nested structure of the data, an intra-class correlation (ICC) coefficient was calculated [19,35,37]. The ICC coefficient estimates the heterogeneity of the dependent variables among groups i.e., people's preferred supply method. Values of ICC range from 0 to 1, in which 0 indicates that probability does not vary among groups while 1 means the result probability only differentiate between groups. In our case, the ICC coefficient was 0.1432, therefore the calculated ICC of the total dataset means 14% of the difference in the probability of willingness to consume domestic eucalyptus fuelwood was related to the difference in people's preferred supply method. Therefore, to build up the single-level logistic regression model would not be appropriate to describe the relationship between the probability of willingness to consume domestic eucalyptus fuelwood and the motivations of consumers without considering the preferred supply method [38].

Successively, the following steps were applied:

• *The first step: Building up of the simplest two level model.*

The simplest two level model represents the model in which intercepts casually vary among groups [19]. In particular, the simplest two level model was described as:

$$Pr(\mathbb{W}ill = \text{Yes}|\mathbf{x}) = \gamma O + \mathbb{u}O\text{j} + r\text{ij} \tag{1}$$

where *Will* refers to the probability that people are willing to consume domestic eucalyptus firewood, γ*0* is the fixed intercept, *u0j* represents the random intercept, and *rij* represent the error. In other words, γ*0* represents the overall average probability that people are willing to consume domestic eucalyptus firewood of the total dataset, while *u0j* represents the variety in the average probability that people are willing to consume domestic eucalyptus firewood. Equation (1) shows two sources of errors in its random part (*u0j* + *rij*) i.e., the between-groups variance (σ*1*) and the within-group variance (σ*2*). The parametric estimation results for the empty model are given in Table 1.

**AIC BIC LogLik** 300.3 307.2 −148.1 Random effects σ<sup>1</sup> = 0.38 σ<sup>2</sup> = 0.62 Fixed effects Value Standard Error z value *p*-value Intercept 0.51 0.1739 2.965 <0.001 Note: The AIC (Akaike information criterion) and the BIC (Bayesian information criterion) are the well-known model fit indices.

**Table 1.** Parametric estimation results for the simplest two level model Equation (1).

Source: Our elaboration.
