Evaluation of Waste in Food Services: A Structural Equation Analysis Using Behavioral and Operational Factors

Abstract

To make the activities of food service companies more sustainable, it is essential to continuously improve their production processes. Understanding which factors are associated with the phenomenon of waste, as well as how they are causally related, is essential for proposing more effective actions to mitigate it. However, the vast majority of studies consider only the analysis of behavioral factors in food waste. To fill this gap, this work evaluates the behavioral and operational dimensions of the phenomenon studied, assessing the implications of the interdependence and causality relations for the respective factors of these dimensions, bringing a broader perspective to the waste problem. The behavioral dimension is developed from the Theory of Planned Behavior (TPB) (Motivation (MO), Intention (IN), and Waste Behavior (CD)) associated with the operational aspects, which are treated in this study as Opportunity (OP), Ability (HB), and Waste Control and Management (CGD), through the use of structural equation modeling (SEM) using the partial least squares in a public food service establishment. The sample size was calculated using the GPower calculator. The composition of the sample was defined considering (i) public profile; (ii) definition of the number of indicators; and (iii) definition of the power of the statistical test and the effect of exogenous variables (f²). Contact was made by sending an email. A return of 102 respondents was obtained. A minimum value of 86 observations was necessary to achieve a statistical power of 80% to identify R². Highlighting the participation of some indicators, Situational Factors 36%, Environmental Beliefs 29%, Health Risks 40%, Training 35%, and Technical Skills 42%, have a strong influence on the average variance extracted (AVE) in their constructs. The proposed model showed the importance of alignment between the factors of the behavioral and operational dimensions in view of improvements in production processes and reduction of waste in food service units.

1. Introduction

The food insecurity index has been increasing worldwide in recent years. The United Nations, through the 17 Sustainable Development Goals (SDGs) established by the 2030 Agenda, included food insecurity as one of the topics to be discussed by UN members [1]. The Food and Agriculture Organization of the United Nations (FAO) data points to an alarming number of about 821 million people in a state of malnutrition and struggling to access basic foods in the world [2]. On the other hand, these figures are opposed to the volume of waste produced by a part of the population worldwide [3,4,5].

When it comes to food waste, developed countries tend to have higher rates of waste [6]. Food waste per capita in developed regions, such as Europe and North America, reaches 95–115 kg/year. In developing regions, such as sub-Saharan Africa and South and Southeast Asia, this value is about 6–11 kg/year [7].

Brazilian families waste considerable amounts of food when compared to other families located in Latin America [8,9]. Usually, this waste is associated with several factors, such as cultural, sociodemographic, and motivational factors, the control and management of waste, and skills, among others [7,8,9,10,11,12,13].

Even though avoiding food waste is suggested as the most promising initiative for decreasing the environmental impact of food waste, little is known about consumers’ behavior towards food waste and the determinants of consumer food waste. Compared to the body of literature aiming to estimate the amount of food waste and its consequences, studies on consumer behavior toward food waste are scarce. Most of the existing studies evaluate the subjective aspects involved in the factors that promote waste through theories such as the Theory of Planned Behavior (TPB) and Perceived Behavioral Control (PBC). The Theory of Planned Behavior (TPB) [14] is one of the most widely applied and accepted models of the belief–attitude behavior relationship within the health literature. It is a cognitive theory, established on the assumption that most conscious behavior is rational and goal-oriented [15]. Essentially, the model implies a causal link between attitudes and behavior that is mediated by intentions.

According to the TPB, behavior is directly influenced by behavioral intentions, which are, in turn, shaped by three sets of considerations [14]. First, beliefs about the outcome of the behavior, as well as evaluations of these outcomes, are implied in the formation of the attitude towards the behavior, which could be a positive or negative attitude. Second, the beliefs an individual holds regarding the expectations of others about her behavior as well as the individual’s motivation to comply with these expectations form a subjective norm. The third consideration refers to beliefs about any factors that may either prevent or facilitate the adoption of the behavior (ability, resources, opportunities, etc.).

On the other hand, the waste that occurs in food service units may be due to operational aspects, such as poor production planning (serving or preparing too-large servings); inadequacy of handling and management of purchase orders; inability to consume food before the expiration date; difficulty in correctly interpreting the instructions on packaging labels; and the behavioral aspects mentioned above [16,17,18].

Although some authors [19,20] have found that food waste has been investigated more frequently in the consumption phase, the literature shows that many of the studies are focused on the assessment of losses, including the stages of harvesting, transportation, and distribution. However, the authors in [19,20] mention that the development of studies on the consumption phase is growing. In this context, public food service companies, because they have high levels of waste, need studies that indicate the underlying causal factors associated with food waste, thus providing subsidies for the proposition of more sustainable measures.

The objective of this study is to evaluate the implications of interdependence and causality for the behavioral and operational dimensions present in food waste. Understanding these relationships makes it possible to define the most critical predictive factors for the output of the model, while also indicating which indicators need more attention on the part of managers. This article analytically addresses the relationships of interdependence and causality between latent variables present in the phenomenon of food waste, by the application of structural equation modeling (SEM). Latent variables will be called constructs in this study. The proposed model considers the Theory of Planned Behavior (TPB) associated with an operational aspect, treated in this study as Opportunity (OP), Ability (HB), and Waste Control and Management (CGD). It is important to highlight that models were not found in the current literature considering the behavioral and operational dimensions concurrently for the phenomenon studied.

2. Literature Review

2.1. The Theory of Planned Behavior (TPB) in Food Waste

The Theory of Planned Behavior (TPB), has been used as a basis to analyze the relationships between factors that induce the behavior of consumers [21]. Some authors highlight the application of TPB to quantitative research on consumer behavior in relation to food waste [12,13,22]. According to [13], TPB is considered a flexible theory that allows for additional concepts. Thus, factors such as moral norms and environmental beliefs, for example, could be included in TPB.

TPB is based on factors such as beliefs, intentions, and consumer attitudes to evaluate consumer IN [14]. Studies aimed at understanding consumer behavior were explained with attitudes [23]. However, it was verified that attitudes alone were difficult to use as a predictive factor of behavior [23]. This problem was addressed by factoring in intention as a mediator among the underlying factors that cause the behavior, such as attitudes. Another additional development of TPB was Perceived Behavior Control (PBC).

PBC includes external factors that can affect behavior, regardless of intention. When TPB is used to evaluate the behavior associated with environmental preservation—for example, how to reduce food waste—factors such as moral norms, environmental and sociodemographic beliefs, among others, can be inserted [13,24,25]. Socioeconomic factors have also been shown to affect the behavior of food waste [13,26,27,28].

The results of the studies by [12,13,22] using TPB to analyze the behavior of food waste differed on which underlying factors have a greater impact on the behavior associated with waste.

Studies [12] indicate that factors such as attitudes and intentions not to waste food have a greater impact on food waste behavior. According to the same authors, moral norms and perceived behavior control (PBC) had no significant impact on food waste. The authors [13] highlight in their results that the intent factor of avoiding food waste would not be a good predictor of food waste behavior. On the contrary, ref. [22] mention that the PBC and moral norms are good predictors of food waste behavior, as well as factors of intention and attitudes.

The TPB theory emphasizes that attitudes represent positive or negative evaluations regarding self-performance for a specific behavior. Subjective norms are associated with the perception of social pressure or important beliefs for others, influencing the behavior of those being observed, such as in a restaurant. Perceived behavioral control, on the other hand, indicates the perceived ease or difficulty of behaving in a specific way; intentions, on the other hand, represent the will to behave in a specific way. Finally, behavior can be defined as any observable action that has a personal or social meaning [29].

In the context of food waste, TPB has been widely used as a theoretical framework to explain consumer behavior. For example, ref. [30] explained the behavior of separating food waste at home using TPB associated with some situational factors. Results indicated that 13.7% of the variation in separation intention was affected by other factors not included in the study.

Similarly, ref. [31] investigated the reduction of household food waste using TPB as well, and the results indicate that the intention to reduce fruit and vegetable waste was predicted by attitude, subjective norms, and perceived behavioral control. Furthermore, the remaining evidence indicates that the final proposed model explains 8% of the variation in food waste reduction behavior; [32] used TPB to explain food waste behavior in a similar study. The authors defined the intention construct as the “intention to reduce food waste” and the behavior construct as the “food waste behavior”. Because the two had a negative relationship, the proposed model included other constructs, such as emotions and habits, as explanatory variables. The model was able to explain 46% of the variation in food waste behavior.

Other authors evaluated different factors associated with food waste. The author in [33,34] highlighted the influence of sociodemographic aspects on the behavior of food waste, such as income. The same authors cite low-cost foods, such as starch, as being more likely to be left on the plate than high-cost foods such as proteins. Greater purchasing power enables greater demand for more diversified foods, which leads to greater food waste [12,28].

2.2. Preliminary Evaluation Model and Hypotheses for Food Waste

The preliminary model elaborated is based on the proposed MOA (Motivation (MO), Opportunities (OP), and Ability (HB) approach) by [31,32] to investigate how the setting of goals interferes in the reduction of food waste as well as which factors are most related. This approach takes into account the motivational background motivation (MO) necessary to establish specific goals, as well as barriers that may hinder its implementation with regard to the infrastructure of the production unit in the food service case opportunity (OP), as well as the absence of skills and knowledge ability (HB). More precisely, motivation can be considered as the driving force for the establishment of intention, and it can be positive or negative, including subjective values, attitudes, and norms [33].

It is important to note that no studies have been found so far that consider the implications of behavioral and operational perspectives for the proposition of structural models aimed at reducing food waste in the food service and making it more sustainable. The proposed preliminary model consists of six constructs: Motivation (MO); Intention (IN); Waste Behavior (CD); Skill (HB); Opportunity (OP); Waste Control and Management (CGD), and Food Waste (DS).

The construct Opportunity (OP) refers to the influences that infrastructure, such as manufacturing space, production sets, and utensils, has on food service consumers and workers, and how this influences their behavior in the face of waste. The construct Ability (HB) refers to the set of skills and knowledge necessary to implement the necessary actions to mitigate waste [33].

In the preliminary model elaborated in this study, the constructs Motivation (MO), Intention (IN) and Waste Behavior (CD) are the input constructs of the model, that is, of the first order. These constructs will establish a structural relationship with the other constructs of the model, where the construct Motivation (MO) establishes a relationship of formation and mediation with the Intention (IN) and Waste Behavior (CD) constructs, bringing the basis of the TPB theory to the model. These constructs were adopted because they present a set of indicators that make it possible to evaluate causal relationships with regard to the behavioral and operational aspects in a broader way (besides being adopted in other consolidated studies) [32,33]. In the context of food waste, TPB has been widely used as a theoretical basis to explain the behavior of clients [34].

More precisely, motivation includes drivers of intention setting, including values, attitudes, and subjective norms. In this sense, when dealing with the problem of food waste, the motivation to act upon a positive goal should be present, and barriers that hinder its implementation should be absent. Such potential barriers are factors that drive consumers to believe that they are unable to reduce food waste [32].

The construct Motivation (MO) demonstrates a person’s willingness to perform actions that reduce the likelihood or amount of waste that may occur. Low motivation can make consumers and food service workers resistant to incorporating new behaviors into their daily routines, which reinforces their commitment to reducing waste [33]. The construct Motivation (MO) presents a set: MO1–Educational level; MO2–Awareness, MO3–Attitude, MO4–Situational factors, and MO5–Social norms. In relation to these, hypothesis H1 indicates that there is a direct relationship between the Motivation (MO) and Waste Behavior (CD) construct. In this relationship, aspects such as purchasing power, level of education, and financial disengagement are cited as the ones that most impact this relationship [27,34].

The construct Intention (IN), which is closely related to Motivation (MO), expresses the intention or willingness to act. The latent variable intention, which is the motivation and willingness to act, is what drives behavior [14]. The intentional process is determined by subjective norms, attitudes, and perceived behavior control (PBC) of the consumer. The authors in [22] indicate that attitudes, PBC, and moral norms indirectly affect the food waste behavior of customers, through their intention to perform the behavior. The specific behavior of an individual increases when they hold a positive attitude toward their behavior if they think that other people expect them to engage in a particular behavior [32].

The results from [12,22] explain that the intention to avoid food waste is a primary factor determining food waste behavior. The authors in [12,13,22] point out that it is more natural to measure intention to avoid food waste than to expect consumers to form an intention to waste food. This is because consumers tend to perceive themselves as averse to the practice of waste. This construct plays a fundamental role in mediating the Motivation (MO) construct in relation to CD. In this relationship, Intention (IN) is responsible for modulating the behavior of CD waste through MO. That is, Waste Behavior (CD) is preceded by the causal influence of Motivation (MO). However, the participation of construct IN will be more relevant in the expression of waste behavior by mediating the relationship. Construct Intention (IN) presents a set of indicators: IN1–Healthy food; IN2–Environmental attitudes; IN3–Environmental beliefs; IN4–Local gastronomy; and IN5–Scarcity of time. Regarding hypothesis H2, it indicates that there is a direct relationship between the Motivation (MO) and Intention (IN) construct [14,22,30]. The H3 hypothesis indicates that the Waste Behavior (CD) construct is directly influenced by IN. The authors [13,22,35] mention that external factors, such as administrative aspects, can impose barriers on this causal relationship, reducing engagement in the adoption of preventive behavior.

The latent variable Waste Behavior (CD) concerns the consumer’s food waste behavior, which is a complex phenomenon resulting from the interaction of several behavioral aspects. In the case of these studies, we have the example of the latent variables Intention (IN) and Motivation (MO) being present. The Waste Behavior (CD) construct is shaped by social, economic, and personal factors, and is the outcome of the interaction of decisions, values, and engagements and directly implies the decision-making process that ends with the behavior of wasting food. The authors in [36] argue that food waste behavior is not unrelated to the pro-environmental objective, and that food waste behavior has a marked habitual and pronounced emotional component. In general, the latent variable Waste Behavior (CD) is influenced by several factors, such as habits, emotions, subjective norms, perceived behavioral control, intentions, etc., all of which have a subjective role to play in determining food waste behavior [37].

It is possible to highlight that the construct Waste Behavior (CD) is directly related to the Food Waste (DS) and Waste Control and Management (CGD) constructs, and carries with it the effect of the causal relationships of its predecessor constructs, Motivation (MO) and Intention (IN). However, it also brings the influence of aspects that are inherent to its indicators, such as aspects related to environmental awareness, habits, and healthy diets [13,14,22]. The H4 hypothesis indicates that the Waste Behavior (CD) construct directly influences the Food Waste (DS) construct.

The domain of the latent variable Ability (HB) represents the individual’s skills or knowledge base related to his capacity for action [38], and in the case of this study, the skills and knowledge needed to analyze, control, and act to reduce waste. In this sense, the Ability construct reflects the set of competencies, skills, and knowledge that employees have to develop their activities, whether in the stages of handling and preparation, sanitization, cooling and freezing, or even inventory management [39].

The construct Ability (HB) concerns a person’s expertise in performing activities as well as problem solving [33]. This study refers to the activities of handling, preparation, and distribution of meals. In this sense, the construct Waste Control and Management (CGD) is directly influenced by the Ability (HB) construct, because improving the skills of employees enables better efficiency of activities in general, thereby reducing waste rates [37,38]. The construct Ability (HB) presents the following indicators: HB1–Technical Knowledge, and HB2–Technical skills. The H5 hypothesis indicates that the construct HB directly influences the Waste Control and Management (CGD) construct.

The domain of the latent variable Opportunity (OP) represents the environmental or contextual mechanisms that enable action [39], as has also often been described under the label of situational or operational constraints [40,41]. It is worth noting that the reflection of this latent variable indirectly influences the Motivation (MO), Intention (IN), and Waste Behavior (CD) constructs. On the other hand, the Opportunity (OP) construct directly influences the Waste Control and Management (CGD) construct, implying the efficient development of its related activities [39].

The construct OP indicates the availability and accessibility of materials and resources needed to avoid food waste. Some examples of aspects relevant to this construct are materials and technologies, infrastructure, human resources training, etc. [39,40]. The authors [41] mention that operational or plant aspects such as lack of adequate equipment for the correct packaging, heating, and refrigeration contribute directly to waste. Construct OP presents the following indicators: OP1–Infrastructure; OP2–Utensils and technological innovation; OP3–Operational scheduling; and OP4–Training and qualification. The H6 hypothesis indicates that the Opportunity (OP) construct directly influences the Waste Control and Management (CGD) construct.

In relation to the construct Waste Control and Management (CGD), it can be highlighted that this concerns the use and improvement of techniques, tools, and methodologies that enable the identification and control of factors responsible for the generation of waste in the food service [42]. The authors [43] mention some potential factors contributing to the generation of food waste, such as absence of demand predictions, inaccuracy in the definition of schedules of activities and menus, and the adoption of inadequate methods and operating procedures. In addition, the hiring of professionals duly trained to perform technical functions in the preparation and handling of food is vital to make food service activities more sustainable [41]. Hypothesis H7 indicates that the Waste Control and Management (CGD) construct directly influences the Food Waste construct.

2.2.1. Preliminary Evaluation Model for Food Waste

From the literature review, a preliminary structural theoretical model was proposed. This model took into account the main indicators that make up the constructs, which portray some different aspects involved in the phenomenon of food waste.

As a contribution, it is important to highlight that, so far, no studies have been found in the literature with models considering these two perspectives. This fact can be considered positive if we take into account the innovation of the approach proposed by the tested model. This condition, as a starting point, restricts discussions of the results obtained from the latest research. The result of the structural model can serve as an instrument for evaluating the dynamics of waste because it makes it possible to evaluate the hypothesis and the participation of its respective indicators in the structural relationship, facilitating the visualization of which constructs and indicators present greater criticality for the output of the model. Thus, the result of this model can assist managers and production supervisors in making more assertive freight decisions, to minimize waste.

Figure 1 shows the preliminary theoretical model highlighting the relationships between constructs and indicators. These relationships are formed by hypotheses that will be tested in the development of the predictive structural model of waste dynamics. Because the data collected are not normally distributed, due to their qualitative nature, i.e., since they are categorical data that must be measured using a scale of values such as the Likert scale, their distribution is not normal, compromising the consistency of the measurement model and necessitating normalization. In this study, the non-parametric bootstrap procedure [44,45] was performed, using the PLS-SEM algorithm of the SmartPLS3.2 software to test the coefficients for their significance.

2.2.2. Justification of the Hypotheses of the Preliminary Model

The preliminary structural theoretical model is composed of seven hypotheses.

Table 1. Structural theoretical model hypotheses.

Hypothesis	Description	References
Hypothesis 1	MO directly impacts CD	[22,43,46]
Hypothesis 2	MO directly impacts IN	[12,13,14,22]
Hypothesis 3	IN directly impacts CD	[12,13,14,22]
Hypothesis 4	CD directly impacts DS	[11,31]
Hypothesis 5	HB directly impacts CGD	[11,47]
Hypothesis 6	OP directly impacts CGD	[48]
Hypothesis 7	CGD directly impacts DS	[10,11,49]

lists the hypotheses present in the structural theoretical model, indicating the causative reflexive relationships between the constructs. The first of them, H1, concerns the direct relationship between the Motivation (MO) construct and the CD construct in such a way that the indicators that compose it, for example, sociodemographic factors (income) are associated with the intention of waste [33,34]. Studies conducted by [26] mention that people with greater purchasing power tend to feel less guilty about wasting food when they eat meals outside the home; that is, they adopt a negative attitude from the intentional point of view before their practice. Financial disengagement is also seen by [42] as a motivational factor that directly influences the intention to avoid waste.

The H2 hypothesis indicates that the Motivation (MO) construct relates directly to the construct Intention (IN). The intentional process, which is the motivation and willingness to act, is what drives behavior. This, in turn, is determined by subjective norms, attitudes, and PBC [14]. These findings suggest that consumers who are in control are not only more motivated to reduce food waste, but are also more likely to implement behaviors that lead to its reduction.

It has been shown that PBC influences intention while having a strong direct impact on food waste levels [12,13,22].

Studies conducted by [22] confirmed that attitudes, PBC, and moral norms indirectly affect the Waste Behavior (CD) through the intention to maintain a certain behavior. The results of [12,22] explain that the intention to avoid food waste is one of the main factors that determine Waste Behavior (CD). The authors [12,13,22] point out that consumers tend to perceive themselves as aversive to waste, so it is more natural to measure the intention of avoiding food waste rather than waiting for consumers to form an intention to waste food.

The H3 hypothesis indicates that the Intention (IN) construct is directly related to Waste Behavior (CD). According to [14], Waste Behavior (CD), that is, the behavior associated with the practice of Food Waste (DS), is influenced both by Motivation (MO) and Intention (IN), as well as by external factors that may impose barriers to them in the engagement of preventive behavior, for example [13,22,50].

In Hypothesis H4, it is observed that the Waste Behavior (CD) construct is directly related to the construct Food Waste (DS). For [51], internal situational factors such as appetite, smell, taste, menu, food temperature, etc., as well as those considered external, that is, related to the restaurant, such as infrastructure, operational, and administrative factors directly influence the behavior of waste.

Hypothesis H5 indicates a direct relationship between Ability (HB) and the Waste Control and Management (CGD) constructs. The prevention of food waste through the use of technical skills by employees requires technical knowledge, which can be acquired through professional training and continuing training in the companies where they work [12]. In other words, even when individuals are motivated to reduce food waste, they need technical knowledge to integrate this behavior into their work routine. Thus, the Ability (HB) construct provides basic subsidies to the Waste Control and Management (CGD) construct through trained human resources with good levels of technical knowledge and specific skills [52].

In Hypothesis H6, it is observed that the Opportunity (OP) construct is directly related to Waste Control and Management (CGD). The Opportunity (OP) construct is based on the availability and accessibility of materials and resources needed to mitigate waste and improve processes, providing subsidies to the Waste Control and Management (CGD) construct [53]. In the case of food waste, some aspects are described in the literature as relevant for its mitigation from the perspective of the Opportunity (OP), such as time, schedule, technology, and infrastructure [54].

The author in [55] mentions that a large percentage, or almost 100%, of food arrives from distribution channels/suppliers in conditions and patterns for consumption. However, in the storage stage, foods begin to lose quality “because they are stored in inadequate places, either because there is not enough refrigeration equipment, lack of space, or also a lack of care or knowledge to store the products not controlling the validity in the correct way” [55].

Hypothesis H7 indicates that Waste Control and Management (CGD) is directly related to the variable Food Waste (DS). The use and improvement of techniques, tools, and methodologies that enable the identification and control of factors responsible for waste generation enable Waste Control and Management (CGD) to reduce waste levels in the food service [55]. The absence of demand predictions and inaccuracy in the definition of activity schedules and menus, associated with inadequate operating methods and procedures, are considered potential factors in the generation of waste [56].

The authors in [57] mention that operational or facility aspects such as lack of adequate equipment for the correct packaging, heating, and cooling contribute directly to waste. In addition, other factors should be taken into account in the process of waste control and management, such as the hiring of professionals properly qualified for the performance of technical functions in the preparation and handling of food.

3. Methods

The proposed method is divided into two major steps. The first is structured Section 3.1, Section 3.2, Section 3.3 and Section 3.4, and concerns the preparation of the survey data collection instrument, the respondents’ demographic data, the sample size, and the validity and reliability of the questionnaire. The second concerns to implementation of structural equation modeling (SEM) by the SmartPLS (student v. 3.3.7) software, supported by the parameters for validation of the measurement and structural mode.

3.1. Preparation of the Survey Data Collection Instrument

3.1.1. Item Development

The preliminary definition of the domains of latent variables that refer to the concept, attribute, or unobserved behavior that is the target of the study was an initial condition for the development of the items. In the case of this study, the domains were defined a priori, based on the literature review [12,13,44,45,58], where the relationship with their respective items is of a reflexive nature. The authors in [59] cite that a well-defined domain will provide a working knowledge of the phenomenon under study, specify the boundaries of the domain, and ease the process of item generation and content validation.

Although a considerable number of studies were used as a theoretical basis for the development of this study, the measurement scales used in the data collection instrument were more specifically based on the studies developed by the following authors: [60,61,62,63].

In this study, the inductive method was used to generate the items adopted in the preliminary model through individual interviews and confirmatory factor analysis (CFA). The inductive method can be developed from “qualitative data obtained through direct observations and exploratory research methods, such as focus groups and individual interviews, Exploratory Factor Analysis (EFA), and Confirmatory Factor Analysis (CFA), which can be used to inductively identify domain items” [64,65].

In the content validity stage, 10 specialists were selected from the department of nutrition. This step was developed to assess if the items adequately measure the domain of interest. The specialists observed the following criteria: (i) adequacy of the latent variables defined for the proposed model; (ii) sufficiency and adequacy of the indicators defined to evaluate each latent variable; and (iii) the semantic quality of the sentences with regard to their purpose and adequacy for the defined public. In Supplementary Materials File S1 and S2.

3.1.2. Scale Development

The pre-testing questions were carried out to ensure the questions and answers are meaningful. Initially, 32 items were generated; however, only 27 were actually selected in this phase, after analysis based on predefined criteria (see File S2). For data collection, an online self-administered cross-sectional survey was adopted, in which the questionnaire was used with the application of the Likert scale. The questionnaire is composed of 27 questions. According to [66,67] the elaboration of a questionnaire basically consists of developing well-written questions related to the proposed problem, translating the specific objectives of the research, and selecting the class and type of research to be used.

The survey was developed in defined [67] phases, according to [66]: (a) specification of the objectives; (b) operationalization of concepts and variables; (c) preparation of the data collection instrument; (d) pre-test of the instrument; (e) sample selection; (f) data collection and verification; (g) analysis and interpretation of data; and (h) presentation of the results.

Contact was made by sending an e-mail to the target audience explaining the survey and asking for collaboration. E-mails were sent to undergraduate and graduate courses such as those in Gastronomy, Nutrition and Production Engineering, as well as to the nutrition department of the food service unit. A return of 102 respondents was obtained. It is worth mentioning that a minimum value of 86 observations was necessary to achieve a statistical power of 80% to identify R² values of at least 0.1, with a probability of error of 5% [68].

The target audience selected to participate in the survey was defined based on some criteria: being food service customers; being collaborators and knowing the facilities and work routine of the food service; being undergraduate and graduate students of any major; and having already had at least three meals at the food service establishments. The target audience was defined, taking into account their knowledge related to the provision of service. In this case, undergraduate students present academic knowledge about the problem of waste and, in the case of graduate students, they have technical knowledge about the activities of handling, preparation, and distribution of meals. Figure 2 and Figure 3 show the percentage of student and employee participants in the survey.

For this research, the ordinal scale was used, in view of the need to evaluate the phenomenon in terms of its situation within a set of ordered levels, which vary from a minimum to a maximum degree. The other classes were not used, considering that the nominal scale has essentially qualitative records in which the expressed variables can be compared, based on a relation of equality or difference.

The attitude scale aims to measure human attitudes through the respondents’ verbal responses, opinions, and assessments, through which individuals respond to a specific situation, with examples of methodologies including Thurstone, Guttman, Osgood, Stapel, and Likert [67].

In this study, the five-point Likert attitude scale was adopted in view of the possibility of measuring both the interviewee’s opinion about a given context and the degree of intensity of the responses (File S1).

3.1.3. Demographic Profile

Table 2. Demographic profile of the participants surveyed.

Demographic Profile of the Participants
Age (Years)	Frequency	%	Education Level	Frequency	%
≤18	33	32	Undergraduate student	61	60
24 to 34	40	39	Postgraduate student	28	27
35 to 44	23	23
45 to 54 <55	4	4	Food Service Employee	13	13
	2	2
	102
Sex
Male	39
Female	54
Other	9
Income
Minimum Wage (MW)	18
1 MW to 2 MW	29
<2 MW	55

presents the demographic profile of the data collected (18 to 57 years old) considered in the study. About 18% receive one minimum wage, 28% up to two salaries, and 54% more than two salaries. About 60% of respondents are undergraduate students, 27% are graduate students, and 13% are employees.

3.1.4. Sample Size

The sample size was calculated using the GPower calculator (3.1.9.7). This software allows for estimating the statistical power and sample size through some statistical tests.

The power of the test is the probability of the test rejecting H0 when H0 is false, which means that the power of the test is equal to 1 − β. In this analysis, the standardized value of 0.80 was considered. The significance level is an assumption of the probability of type I error, which represents the rejection of the null hypothesis when it is true.

The composition of the sample was defined considering the following items:

(i)

Public profile.

(ii)

Definition of the number of indicators.

(iii)

Definition of the power of the statistical test and the effect of exogenous variables (f²).

The authors in [68,69] recommend the use of the test power of 0.80 and the average effect size (f²) equal to 0.15.

3.1.5. Validity and Reliability of the Questionnaire

Although this work makes use of an already validated scale (the Likert scale proposed by Rensis Likert in 1932), in the questionnaire elaboration phase it is necessary to evaluate the validity and reliability of this data collection instrument.

For [67,68], reliability is related to the constancy of the results when applied repeatedly, while validity indicates the instrument’s ability to actually measure what should be measured. According to [70,71], to validate a questionnaire for reliability, three criteria can be addressed: stability, internal consistency, and equivalence.

The stability of a measure is an estimate of the consistency of the repetitions of the measures, and its evaluation can be performed through test–retest (intraclass correlation coefficient (ICC)) [71]. Considering that there was no repetition of the application of the questionnaire, this criterion was not adopted.

Internal consistency is a way of measuring the reliability of an instrument—in this case, the questionnaire, its items, and dimensions—based on a single administration, removing the temporal influence on its application. Considering a scale with multiple items, all of them must be consistent and express part of the explained variance of the construct [70].

The authors in [68] point out that the idea of internal consistency is that the scale items should measure the same construct, and thus be highly correlated. Equivalence, on the other hand, is the degree of agreement between two or more evaluators regarding the scores of an instrument, with the Kappa coefficient being a measure used for its evaluation [71]. Considering that there was only one evaluator of the questionnaire, this criterion was not used.

The verification of the internal consistency can be performed using the Cronbach’s alpha coefficient and composite reliability performed in this work, with the help of the SmartPLS^® (student v. 3.3.7) software [72], to ensure the quality of the data collection instrument for this research.

3.2. Structural Equation Modeling (SEM)

The evaluation by SEM allows simultaneous analysis between multiple variables and a better understanding of the relationships of causality and interdependence in their relationships. The PLS-SEM has been widely used, due to the possibility of being able to work with non-normal data, a small sample, and the use of both formative and reflective indicators [31,34], used in this study.

The constructs are classified as exogenous and endogenous. Exogenous predictors are those that usually assume an entry position in the structural model; that is, they are of the first order, and thus they do not suffer from influence or effect of other variables present in the model. On the other hand, the endogenous constructs are dependent, may be of second or greater order within the model, and thus are influenced by other variables that are present predictors.

The equations represent the measurement model as shown [36]:

X = λx ξ + δ

(1)

Y = λy η + ε

(2)

where X and Y are the variables observed; λx is the matrix of coefficients that associates the observed variables and the exogenous latent variables; ξ is the vector of exogenous latent variables; δ is the vector of measurement errors in exogenous variables; λy is the matrix of coefficients that associates the observed variables and the endogenous latent variables; η is the vector of endogenous latent variables; and ε is the vector of measurement errors in endogenous variables. It is noteworthy that errors external to measurement do not interfere with the indirect measurement of the construct [37,38].

The structural model can be expressed by [36]:

η = βη +Γξ + ζ

(3)

where η is a vector of endogenous latent variables; β is the coefficient matrix that binds endogenous variables; Γ is the matrix of coefficients relating exogenous variables to endogenous variables, ξ is the vector of exogenous variables; and ζ is the vector of latent errors in equations.

3.3. Parameters for Validation of the Measurement and Structural Mode

In order to obtain the most appropriate model regarding reliability and predictive validity, the evaluation of the proposed model was performed by means of an exploratory approach of structural equations using the partial least square (PLS) method with SmartPLS (student v. 3.3.7). The literature highlights that PLS-SEM can be applied even when the sample size is too small or even smaller than the number of observed variables [73]. For this reason, we chose to use this evaluation method in this study.

The first stage was the evaluation of the measurement or external model, considering its reliability and convergent and discriminant validity for the measures that represent the set of indicators of each construct.

Table 3. Evaluation measures for reflective indicators.

Method/Tool	Type of Evaluation	References
Convergent validity	Average variance extracted (AVE)	[70]
Internal consistency	Cronbach’s alpha Composite reliability rho_A
Discriminating validity	Cross-loading Fornell and Larcker’s criteria	[74,75]

Note: indicators with factor loads above 0.95 indicate that the items are redundant, reducing the construct validity [76]. Cronbach’s ideal alpha values should be between 0.70 and 0.95. Cronbach’s alpha tends to underestimate reliability when the sample size is small (<100). The ideal is to adopt the composite reliability measure. The coefficient rho_A [74] returns an average value between Cronbach’s alpha and Composite reliability (Hair et al., 2019) [77].

shows the parameters for evaluating the measurement or external model.

The AVE was used for the convergent and divergent validity test. AVE indicates the mean commonality for each latent variable for a reflective model. For a model to be considered appropriate, the value of the AVE must be greater than 0.5 [75,77].

According to [68], Cronbach’s alpha also signals whether the indicators ensure convergent validity for constructs and reliability. “By convention, the same cutoff points apply: greater than or equal to 0.80 for a good scale, 0.70 for an acceptable scale, and 0.60 for an exploratory scale”. However, Cronbach’s alpha is a conservative measure that tends to underestimate reliability. Thus, due to the limitations of Cronbach’s alpha, it is technically more appropriate to use composite reliability [68].

For composite reliability, the acceptable cut-off point ranges from 0 to 1, with 1 being the perfect estimate [68]. In a model suitable for exploratory purposes, composite reliability must be equal to or greater than 0.6 [75,77]; equal to or greater than 0.70 for a model suitable for confirmatory purposes [78]; and equal to or greater than 0.80 is considered good for confirmatory purposes [79].

In the evaluation of cross loads, the highest factor loads of the observed variables should be exclusively related to the constructs, explaining the discriminant’s variation. The criterion of Fornell and Larcker is met when a construct shares more variation with its associated indicators than with any other construct [68]. In the evaluation of the structural model, the hypotheses are tested for aspects of collinearity, influence effect, and predictive validity, among others.

Table 4. Structural model evaluation measures.

Purpose	Indicator/Procedure	References
Collinearities	Collinearities	[80]
Evaluate the variance of endogenous variables that is explained by the structural model	Pearson coefficient of determination (R2)	[69]
Evaluate how much each construct is useful for model tuning	Cohen effect size or indicator (f2)
Assess the accuracy of the adjusted model	Predictive validation or Stone–Geisser indicator or cross-validity redundancy (Q2)	[68]
Evaluate causal relationships	Path coefficient

Source: adapted from [78,79].

shows the measures for evaluating hypothetical relationships within the structural model.

To evaluate the collinearity, each set of predictor constructs was examined separately for each subpart of the model.

For this purpose, tolerance values below 0.20 (variance inflation factor (VIF) above 5) in predictor constructs are classified as a critical level of collinearity [68]. For the hypothetical relationships between the constructs represented by the path coefficients, the values usually fall between the −1 and +1 limits.

Then, we have the evaluation of Pearson’s coefficients of determination (R²). Secondly [68,81], the R² evaluates the amount of explained variance of the endogenous variables formed by its predecessor constructs. Thus, it represents a predictive measure of endogenous constructs.

In the area of social and behavioral sciences, ref. [69] suggests the classification of small effect for R² = 2%, an average effect for R² = 13%, and a large effect for R² = 26%. This convention will be adopted in this study.

Following the evaluation, Cohen’s effect size or indicator (f²) aims to evaluate the contribution of an exogenous construct to an R² value in the endogenous latent variable. The guidelines are that values of 0.02, 0.15, and 0.35, respectively, represent small, medium, and large effects of the exogenous latent variable [68,69].

Subsequently, the Stone–Geisser (Q²) value is analyzed, in which the way the model approaches what was expected of it is analyzed, that is, the predictive relevance of the path model for a given dependent construct for values greater than zero [68,81]. For this, the “blindfolding” technique of reusing samples is used, which systematically omits data points and provides a prognosis of their original values [77].

3.4. Limitations of the Study

As a limitation of this study, we can highlight the use of only reflective constructs in the proposed model. It should also be noted as a limitation of this study, the reduced number of studies considering the evaluation by SEM in the food service. This fact limits discussions of the findings in light of the latest research.

4. Results and Discussion

The results are divided into three parts: Section 4.1. Analysis of the measurement or external model, highlighting the external model validation results; Section 4.2. Structural model analysis, presenting SEM modeling results; and Section 4.3. Participation of indicators in their respective constructs, evidencing the participation of indicators in their respective constructs, thus allowing an assessment to identify the most critical indicators in each construct.

4.1. Analysis of the Measurement or External Model

The analyses of the preliminary model initially started from the evaluation of the measurement model, taking into account parameters associated with the reliability of the collected data as well as its convergent and discriminant validity, presented in Table 3. Evaluation measures for reflective indicators are shown below in Figure 4, presenting the first result obtained from SmartPLS3 for the measurement model.

The results of the first evaluation by the execution of the PLS-SEM algorithm made it possible to analyze the factor loading of the set of indicators for the respective constructs through the confirmatory factor analysis (CFA).

The authors in [82,83] cite factor loads below 0.609–0.708 as acceptable in exploratory research, which uses categorical data and evaluation of consumption behavior. Following iterations and analyses, indicators with the factor loads MO1-0.496, IN4-0.177 were removed from the measurement model because they jeopardized the constructs’ convergent validity and model reliability.

According to [83], in the CFA it is important to carefully examine the effects of the exclusion and inclusion of an indicator present in a construct, because often the exclusion of an indicator will decrease composite reliability as well as its extracted average variance (AVE). In addition, the aspects of theoretical relevance need to be considered.

Table 4 presents the measures of internal consistency (Cronbach’s alpha, CR, and rho_A > 0.7) and convergent validity (>0.5). The results indicate that the reliability of the measurement model is within the tolerance limit, making it possible to proceed to the evaluation and validation of the structural model [68].

Table 5. Test results of the measurement model with validated indicators.

Construct	Indicator	Description	Load	Cronbach’s Alpha	rho_A	Composite Reliability	AVE
Motivation (MO)	(MO2)	Awareness	0.765
	(MO3)	Attitudes	0.713	0.868	0.942	0.903	0.701
	(MO4)	Situational factors	0.935
	(MO5)	Social norms	0.911
Intention (IN)	(IN1)	Healthy food	0.620
	(IN2)	Environmental attitudes	0.666
	(IN3)	Environmental beliefs	0.760	0.668	0.603	0.771	0.458
	(IN5)	Scarcity of time	0.653
Waste behaviour (CD)	(CD1)	Abundance habits	0.796
	(CD2)	Health Risks	0.834	0.727	0.728	0.846	0.647
	(CD3)	Healthy habits	0.782
Ability (HB)	(HB1)	Technical Knowledge	0.741
	(HB2)	Technical skills	0.838	0.740	0.768	0.850	0.654
	(OP4)	Training	0.843
Opportunity (OP)	(OP1)	Infrastructure	0.841
	(OP2)	Utensils and technological innovation	0.803	0.682	0.688	0.826	0.614
	(OP3)	Operational scheduling	0.700
Waste control and management (CGD)	CGD1	Refrigeration and heating equipment	0.636
	CGD2	Methods and procedures	0.698	0.506	0.521	0.749	0.501
	CGD3	People management	0.781
Food waste (DS)	DS1	Extensive land use	0.892
	DS2	CH4 emissions	0.837
	DS3	Food insecurity	0.611	0.843	0.935	0.884	0.608
	DS4	Extensive use of water	0.794
	DS5	Extensive use of pesticides and fertilizer	0.735

presents the results of the measurement model test using the PLS-SEM algorithm, with values referring to factor loads and reliability parameters.

In relation to the path model, the hypotheses defined as preliminary may or may not converge. This is related to the collected data’s normality distribution; that is, if it approaches a normal Gaussian distribution. As the collected data are categorical, the nonparametric bootstrap test will be performed considering the parameters t_value 1.96 and p_value (two-tailed) 5%, indicating that the path coefficient is significant and different from zero at a significance level of 5% (α = 0.05; two-tailed) [68]. Critical t values with significance levels of 1% (α = 0.01; two-tailed test) and 10% (α = 0.10; two-tailed test) indicate a probability of error of 2.57 and 1.65, respectively [68]. The non-parametric bootstrap procedure adopted in this study considers the following parameters: t_value 1.96 and p_value (two-tailed) 5%.

According to [68], the most conservative reliability measure Cronbach’s Alpha also signals that the indicators ensure convergent validity for the constructs and reliability of the proposed model, besides indicating the reliability of the collected data, indicating possible biases present in the observed variables, or even if the set of answers is reliable. It is possible to notice in Table 5 that the latent variable Intention (IN), and OP and CGD present some values below those recommended by the literature. However, it is worth mentioning that this difference is minimal for the values used in exploratory research.

Cronbach’s alfa values ranged from 0.506 to 0.868, and it is possible to consider them acceptable from the point of view of the internal consistency of the model. By convention, the same cutoff points apply: greater than or equal to 0.80 for a good scale, 0.70 for an acceptable scale, and 0.60 for a scale for exploratory purposes [68]. However, the same authors point out that, as a conservative measure, Cronbach’s alpha tends to underestimate reliability for samples smaller than 100. Using this, it is technically more appropriate to work together with other measures, such as composite reliability and AVE [68].

The composite reliability measure indicates the degree of association between the constructs and their respective indicators. For composite reliability, the acceptable cut-off point ranges from 0 to 1, with 1 being the perfect estimated reliability [68]. For [75,77], composite reliability must be equal to or greater than 0.6 in an exploratory model, equal to or greater than 0.70 in models with confirmatory purposes [78], or greater than 0.80 [79]. The results for the proposed model ranged from 0.749 to 0,903, indicating a satisfactory level of reliability.

The constructs should explain at least 50% of the variance of their respective indicators, thus presenting an average commonality for each latent variable in a reflective model. For a model to be considered adequate, the value of the AVE must be greater than 0.5, as well as higher than the cross loads [75,77]. In situations where the AVE has values below 0.5, this means that the variance of the error exceeds the explained variance.

The results found ranged from 0.458 to 0.701, indicating that the explained variance of the constructs contributes to the internal consistency of the model and carries its convergence. In this case, the values found for the latent variable IN also present a value of less than 0.5, but the difference is very small. The results found for the measures adopted in the evaluation of constructs make it possible to affirm that there is reliability and validity of convergence to the measurement model, taking into account the fulfillment of the set of parameters.

The discriminant validity was evaluated by the criterion of cross-loading and that of Fornell–Larcker. For the first criterion, it is possible to observe in Table S1 (File S5) that the factor loads of the indicators and their respective constructs are higher than those of their crossing with the other constructs present in the model. Thus, it can be inferred that the model has validity for the discriminant. Table S2 (File S5) also presents the result of the discriminant’s validity using the Fornell–Larcker criterion. The criterion of Fornell and Larcker evaluates whether a given construct shares more variance explained by its respective indicators than with any other construct present in the model [68].

According to Table S2 (File S5), the values of the main diagonal of the matrix are greater than the correlations presented in the secondary diagonals. This implies that constructs relate without collinearity, because the square root of the AVE is greater than the correlations of the measured construct.

The results found for the Fornell–Larcker criterion indicate that the constructs present in the model behave differently in the structural model, implying that the variance captured by each construct in reaction to its set of indicators concerns a specific dimension of the observed phenomenon, in this case, food waste. The authors [68] highlight that the indirect measures of each construct are independent; that is, they are effectively related as their respective constructs, thus avoiding the occurrence of cross-loadings that would imply the existence of collinearity between constructs.

4.2. Structural Model Analysis

It is important to highlight that the main objective of PLS-SEM is to maximize the explained variance, that is, the R² value of the endogenous latent variables in the path [68]. For this reason, the evaluation of the quality of the measurement of structural models focuses on metrics of the predictive capabilities of the model, which is different from the evaluation used in the measurement model, where the quality criteria are used to focus on the reliability and validity of the convergent and discriminant. For the structural model, the most important evaluation metrics are R² (explained variance), f² (effect size), Q² (predictive relevance), and the size and statistical significance of structural path coefficients [68].

Table 6. Structural model result.

Hypothesis		R2	B	Std error	t_Value	p_Value	Decision	f2	q2
H1	MO → CD	0.148	0.152	0.141	1.057	0.282	Not supported	0.00	0.02
H2	MO → IN	0.082	0.288	0.122	2.363	0.018	Supported	0.09	0.03
H3	IN → CD	0.148	0.307	0.161	1.875	0.056	Not supported	0.10	0.05
H4	CD → DS	0.154	0.042	0.149	0.289	0.776	Not supported	0.00	0.01
H5	HB → CGD	0.415	0.245	0.104	2.388	0.018	Supported	0.08	0.02
H6	OP → CGD	0.415	0.316	0.119	2.671	0.008	Supported	0.14	0.04
H7	CGD → DS	0.154	0.368	0.164	2.251	0.025	Supported	0.11	0.07
H8	CD → CGD	0.415	0.295	0.077	3.845	0.000	Supported	0.05	0.01

shows the results of the structural model tests (explained variance (R²), predictive relevance (Q²), and effect size (f²)) after the bootstrap procedure. Here, it is important to highlight as a contribution the insertion of a new H8 hypothesis, which relates to the Food Waste (CD) construct and the Waste Control and Management (CGD) construct, discussing the direct relationship between behavioral theory and waste control management, until then little discussed in the literature.

The evaluation of the structural model makes it possible to observe how the results of the measurement model are close to the theory [68]. Figure 5 presents the result in the theoretical structural model after validation of the measurement model and application of the bootstrap procedure. After the procedure with the insertion of hypothesis H8, the path coefficients of the structural relationship showed a slight change in their values, not interfering with the consistency of the model.

The results of the structural theoretical model after tests and adjustments are discussed in more detail below.

The results found for the proposed model indicate that H1 MO → CD is not supported, that is, the null hypothesis H0 is not true, as already expected, since the variables (MO, IN, and CD) establish a complete mediation relationship, as can be seen in Figure 5. The value of R² (0.148) shows a moderate explained variance of 15% for the CD construct in relation to its MO predictor construct. It is worth mentioning that in exploratory research and evaluating consumption behavior [82,84], values close to 0.20 are considered high.

The value of β (0.152) indicates that there is a positive and weak relationship between constructs. In relation to the t_value (1.057), this attests that there is no statistical significance among constructs at the level of 5%. Regarding f², which indicates the contribution or effect of the variance explained for an endogenous variable, it has no effect on the endogenous latent variable (0.00). Authors such as [69] indicate that values for f² lower than (0.02) indicate that there is no influence of the exogenous variable on the endogenous variable in the case of MO → CD. Finally, the values relative to predictive relevance, q², (0.02) indicate that there is small predictive relevance for the endogenous variable.

The MO construct has a direct relationship with the CD construct. In the case of this model, the results indicate a complete mediation relationship, hence the relationship is not supported after the bootstrap procedure, does not present statistical significance nor f² effect, and has small predictive relevance (q²). Studies indicate that sociodemographic factors, especially income, have a greater impact on this relationship [33,34]. The financial disengagement on the part of students is seen by [42] as a motivational factor that directly influences the intention to avoid waste. The authors in [26] mention that people with greater purchasing power have a negative attitude from the point of view of intentionality towards the practice of waste when they eat outside the home.

It is important to highlight that in this causal relationship between MO and CD, the employees involved in the production process must be properly trained to ensure that indicators such as situational factors are within the defined standards. Thus, it is necessary to be attentive to the monitoring of the heating and cooling temperatures of the equipment to ensure that the food is served at the ideal temperature, a fact that prolongs the perishability of the prepared food. The proper preparation of menus and trying to vary the type of meal served is vital for better acceptance by customers

The results state that H2 MO → IN is supported, that is, the null hypothesis H0 is true. The value of R² (0.082) indicates a small variance explained by the Intention (IN) construct in relation to its predictor construct MO, taking into account the exploratory nature of the research. The value of β (0.288) indicates that there is a positive and moderate relationship between constructs. Regarding the t_value (2.363), this attests that there is statistical significance among constructs, with a significance of 5%. In relation to f², this showed small effect between the exogenous variables and endogenous variables (0.09). Predictive relevance with a value of q² (0.03) indicates a low relevance for the endogenous variable Intention (IN).

The relationship between the Motivation (MO) and Intention (IN) constructs is directly influenced by subjective norms, attitudes, and PBC [14]. The authors in [12,13,22] point out that consumers who are in control of the situation of the act of wasting are not only more motivated to reduce it, but also effectively implement behaviors that lead to its reduction.

When discussing food waste in an everyday context, taking into account the volume of waste produced every day in food services, studies show [71] that the absence of periodic educational campaigns aimed at employees and customers has an impact on the motivation and intention of these individuals when the subject is food waste. This aspect is aggravating, considering that a significant portion of those who are producing waste consider food waste inevitable when it occurs from time to time, and trivialize its consequences. In this sense, food service unit managers must provide their customers and employees with access to information to build new values and beliefs in line with sustainability.

The results state that H3 IN → CD is not supported, that is, the null hypothesis H0 is not true. The value of R² (0.148) indicates a moderate explained variance of 15% for the CD construct in relation to its predictor construct. The value of β (0.307) indicates that there is a positive and average relationship between constructs. In relation to the t_value (1.875), this attests that there is no statistical significance among constructs, with a significance of 5%. In relation to f², which indicates the contribution or effect of the variance explained for an endogenous variable, it presents an average ratio of (0.10). The values relative to predictive relevance q² (0.05) indicate that there is a small predictive relevance for the endogenous variable CD. Although this hypothesis is not supported after the bootstrap procedure, it was not excluded from the structural model, considering its theoretical relevance. Furthermore, from a statistical point of view, its results for the t_value (1.875) and p_value (0.061) are very close to the minimum values considered significant.

The environmental impacts resulting from food choices depend on the nature of the food. Meat and other animal products tend to have greater environmental impacts, due to their higher trophic level and related inefficiency in production. The lack of clear information for customers about the environmental impacts that each food produces does not help in the consumption of more sustainable items, limiting individuals to only consider aspects such as price and calorific value [85,86].

The results indicate that H4 CD → DS is not supported, that is, the null hypothesis H0 is not true. The value of R² (0.154) indicates a mean explained variance of 16% for the Food Waste (DS) construct in relation to its CD predictor construct. The value of β (0.042) indicates that there is a small positive relationship between constructs. The t_value (0.289) indicates that there is no statistical significance among constructs, with a significance of 5%, and this is because there is a complete mediation relationship between the two latent variables CD → DS, as can be seen in Figure 5. The results for f² (0.00) and q² (0.01) indicates that there is no effect of the exogenous variable on the endogenous variable, as well as no predictive relevance.

According to [14], it is the waste behavior associated with the practice of producing waste that is modulated by the Motivation (MO) and Intention (IN) constructs, as well as by internal and external situational factors [13,22,50]. For [51], internal situational factors such as appetite, smell, taste, menu, food temperature, etc., as well as those considered external, that is, related to the restaurant, such as infrastructure, operational, and administrative factors, directly influence the behavior of producing waste. Studies by [87] show that consumers find it boring to eat the same meal multiple times in a row, and prefer to have a large variety of foods always [88]. For [89], the individuals may throw away edible leftovers because they want something new and fresh or the food does not meet their expectations of flavor or texture, for example.

The Waste Behavior (CD) construct is preceded by structural relationships with Intention (IN) and Motivation (MO) and correlated with their respective indicators. It is important to consider the main indicators present in these relationships, as they will impact the structural relationships of constructs of an operational nature. Thus, as mentioned in the previous paragraphs, aspects associated with situational factors, healthy habits, environmental beliefs and attitudes, etc., must be perceived by managers and managers as critical indicators for the behavioral dimension, taking into account the subjectivity related to its nature.

Studies conducted by [27] cite that it is necessary to adopt policies aimed at consumer awareness, modifying the behavior regarding food waste, in addition to arousing consumer responsibility and commitment to this issue. However, other studies have assessed the fact that certain interventions aimed at raising awareness do not sufficiently reduce food waste [90], because there are other more complex causes associated with behavioral aspects that make the process of awareness ineffective in practice. The author [91] proposes that efforts can be directed at material aspects related to controlling the portions served, for example.

The results indicate that H5 HB → CGD is also supported; that is, the null hypothesis H0 is true. The value of R² (0.415) indicates a high explained variance β (0.245) for the Waste Control and Management (CGD) construct in relation to its HB predictor construct. The value of β (0.245) indicates that there is a positive and moderate relationship between the constructs. The t_value (2.388) indicates that there is statistical significance among the constructs, of 5%. The result for f² indicated an average effect of the exogenous construct for the endogenous (0.08). There is a small predictive relevance among the variables’ q² (0.02).

Despite the simple fact that individuals are motivated in an organization to implement actions that promote the reduction of food waste, ref. [52] mention that the absence of certain technical skills and specific knowledge ends up limiting the field of action of individuals in the face of waste mitigation. The realization of continuous training by food service companies is fundamental to the implementation of waste control and management actions. [12].

In relation to H6 OP → CGD, the results indicate that this relationship is also supported, that is, the null hypothesis H0 is true. The value of R² (0.415) indicates a large explained variance of 42% for the Waste Control and Management (CGD) construct in relation to its OP predictor construct. The value of β (0.316) indicates that there is an average positive relationship between constructs. The t_value (2.671) indicates that there is a statistical significance among the constructs, of 5%. Predictive relevance as a value of q² (0.04) indicates that there is a small predictive relevance among constructs. The result for f² indicated an average effect of the exogenous construct on the endogenous construct (0.14). The literature highlights the participation of technological and infrastructure factors as preponderant factors for the generation of waste when not adequate for the demands of production processes [19,92,93,94,95].

Some aspects related to OP constructs can be cited as having a large influence on the Waste Control and Management (CGD) construct. Storage in inadequate locations, refrigeration equipment, insufficient heating, lack of space, as well as lack of care or technical knowledge regarding storage, are cited [54,55,57].

The results indicate that H7 CGD → DS is also supported, that is, the null hypothesis H0 is true. The value of R² (0.154) indicates an average explained variance of 15% for the construct Food Waste (DS) in relation to its Waste Control and Management (CGD) predictor construct. The value of β (0.368) indicates that there is an average positive relationship between constructs. The t_value (2.251) attests that there is a statistical significance among the constructs, of 5%. The result for f² indicated a medium effect of the exogenous construct on the endogenous construct (0.11). Predictive relevance as a value of q² (0.07) indicates a small predictive relevance for the endogenous latent variable DS. The authors [26,95,96] cite the difficulties of the proper use of methods for the quantification of food waste, as well as highlighting the reluctance of some organizations to work with a focus on reducing waste, or even the fact that the reduction of waste is not part of their list of priorities in their operational actions.

The absence of demand predictions, in addition to the adoption of inadequate operational methods, are some of the factors mentioned by [56] as relevant to the relationship between the Waste Control and Management (CGD) and DS constructs. The author in [55] also highlights the importance of using and improving techniques and tools that enable the identification and control of factors responsible for waste generation.

The non-use of planning instruments such as a technical data sheet brings implications to the steps of manipulation and preparation, because the non-use of the technical data sheet will not allow the employee responsible for this stage to perform the controls related to the menu planning. The authors in [97,98] affirm that the technical data sheet is a useful tool to assist the planning, and serves as a form of control for some variables such as correction factor, number of portions, cooking factor and yield. The previous knowledge of this information is fundamental for the identification of the quantity of foods to be produced, and avoiding waste, besides allowing for a better financial control relating to the items used [99].

Regarding H8, the results state that the hypothesis is supported, that is, the null hypothesis H0 is true. The value of R² (0.415) indicates a high explained variance of 42% for the Waste Control and Management (CGD) construct in relation to its OP predictor construct. The value of β (0.295) indicates that there is a moderate positive relationship between the constructs. The t_value (3.845) attests that there is a statistical significance among the constructs, of 5%. The result for f² indicated a small effect of the exogenous construct on the endogenous construct (0.05). Predictive relevance as a value of q² (0.01) indicates that there is no predictive relevance.

The H8 hypothesis (β₈ = 0.295) has a moderately positive and direct relationship with Waste Control and Management (CGD). The behavioral variable plays a predominant role in the aspects of control and management of food waste. This relationship makes clear the importance of the behavioral dimension for the implementation of actions of an operational nature, because, in practice, the emotional, psychological, and motivational state of the employees will affect how they use machines, equipment, and utensils, as well as their compliance with rules and procedures of tasks in the stages of handling and preparation of food.

When evaluating the proposed model considering the set of constructs, it is possible to clearly observe the presence of two perspectives: one formed by the set of constructs that relate to the behavioral part of the model, formed by constructs (MO, IN, CD) and supported by the TPB theory, and another related to the operational part, focusing on the control of waste management and the constructs (HB, OP, CGD).

Considering the results of the AVE and f², it is clear that there is a strong relationship between the Waste Control and Management (CGD) construct and its respective Ability (HB) and Opportunity (OP) predictor constructs, in relation to the behavioral construct of the model, represented by the Waste Behavior (CD) construct, with an R² 0.415, that is, about 42% of the explained variance. In view of this finding, the results found converged on the insertion of a new H8 hypothesis, which concerns the causal relationship between the independent construct Waste Behavior (CD) and the Waste Control and Management (CGD) dependent construct. This new hypothesis, besides being supported and presenting internal consistency for the structural model, confirms the need for alignment between behavioral and operational factors.

4.3. Participation of Indicators in Their Respective Constructs

When we observe in Figure 6. the participation of the indicators in their respective constructs, it is possible to identify the contribution of each indicator in the variance of the construct; that is, how much each indicator is explained by this, since the relationship between them is reflective in nature.

Observing the percentages of each indicator for the latent variable MO, the indicator MO4, Situational factors, with 36%, stands out among the other indicators belonging to this construct, indicating that aspects such as flavor, texture, temperature, menu, emotional state, and others have expressive importance in the act of waste. When compared to situational factors, the MO5 Social Norms indicator comes in second place, with 33% participation in the latent variable. This indicator concerns how we behave in front of other people, such as the waste carried out in an attempt to impress, by eating little or, in more extreme cases, to appear economical in the face of the situation [46].

The indicator MO2 Awareness with 16% and MO3 Attitudes with 15% show a low participation in the mean variance of the construct; however, these two indicators, when used strategically, can contribute significantly to reducing waste. Previously cited studies indicate that it is possible to change diners’ attitudes through periodic awareness campaigns to combat waste.

Within a management approach, it would be possible to use this information to promote actions aimed at reducing waste. The indicator MO4 Situational factors, because it is of internal control character and presents participation of about one-third of the latent variable, should be monitored with greater attention by management, since the nature of the indicator allows fast and efficient action in the control of waste.

When we evaluate the participation of the indicators for the IN construct, it is observed that there is a balance between the set of its indicators, with regard to their respective participation, with a variation of up to 6% between them. The participation of the indicator IN3 29% Environmental Beliefs, emphasizes that aspects related to environmental impacts arising from food waste should be widely used as an instrument, combined with the reduction of waste, through periodic awareness campaigns showing the damage caused to the environment by waste production. Although the IN2 23% Environmental attitudes indicator showed 6 percentage points less than the IN3 indicator, in practice, environmental attitudes have a more effective impact on reducing waste, that is, it is necessary to go beyond the borders of environmental beliefs and take proactive attitudes aimed at combating food waste.

The other indicators show a similar participation, in the construct Healthy food, 22%, and Scarcity of time, 26%. These indicators have an important participation in the AVE of the IN construct; however, their management is limited only to the guidance and awareness of the diners.

Observing the results of the participation of indicators for the Waste Behavior construct, the Health Risks CD2 40% indicator indicates the predominance of behavior associated with the risk of contamination by spoiled food or the presence of viruses or bacteria, such as salmonella. The CD2 indicator points to an issue wherein it should be made clear to diners that the control patterns of sanitary factors are ensured, as well as the stages of handling and preparation adequately taking care of the sanitization and hygiene aspects of the foods to be processed. This care guarantees the diners’ safety in relation to the aspects of contamination and intoxication because of spoiled food served, thus avoiding the waste associated with these issues.

The CD1 indicator highlights that negative habits lead to negative attitudes, thus increasing actions that are uncommitted to the reduction of waste, whether at the time customers and diners put their dishes together, or during the production planning and purchasing, or even stock management.

The indicators that relate to the OP construct can be highlighted by the participation of the Infrastructure OP1 with 46%, Utensils and technological innovation OP2 with 35% and Operational scheduling OP3 with 20%. These results indicate the importance of investments in infrastructure and technological innovation, as well as the standardization of production processes. Proper management of these indicators contributes to improving the efficiency of the activities carried out within the food service units, whether in the stages of receiving raw materials, or with more accurate stock control, and in the food handling and preparation, considering the monitoring of correction factors, as well as the food distribution [44].

In the participation of Ability (HB) construct indicators, the Technical skills HB2 indicator presented a 42% share, standing out more than the Technical construct.

The Knowledge HB1 indicator presented 24%. The result of the OP4 indicator was 35%, present in the latent variable HB through the exchange with the OP construct, a signal that the development of employees’ technical skills by training reflects the improvement in control and waste management, emphasizing the importance of allocating financial resources for these purposes [44,100].

With about 51% participation in the People’s Management CGD3 indicator, which makes up the CGD construct, the importance of human resources management in hiring qualified professionals is highlighted. Next, we have the indicator for CGD2 methods and procedures, with 29%. Although this indicator has lower participation in the Waste Control and Management (CGD) construct’s explained variance, it has great relevance in the control and reduction of food waste in practice, and in many cases it is used as one of the key indicators of the production unit, because much of the waste generated within production occurs due to a lack of implementation of methods and appropriate management techniques.

Food waste in the food service negatively impacts the environment throughout its production chain. Evaluating the indicators of the Food Waste (DS) construct, it is possible to note that the indicators Extensive land uses DS1 with 28%, Emissions CH4 DS2 with 27%, and Water consumption DS4 with 18% are more critical, considering their impacts on the environmental and social spheres.

As a contribution, it is important to note that no studies with models considering these two perspectives have been found in the literature thus far. This fact can be considered positive if we take into account the innovation of the approach proposed by the tested model. The result of the structural model can serve as an instrument for evaluating the dynamics of waste, because it makes it possible to evaluate hypothesis by hypothesis and the participation of their respective indicators in the structural relationship as presented in Table 6 and Figure 6 facilitating the visualization of which constructs and indicators present a greater criticality for the output of the model.

The proposed model can help in the adoption of good practices aimed at reducing the environmental impacts promoting their sustainability. More sustainable practices, besides contributing to environmental improvement, are also perceived as instruments that assist the competitive advantage of companies [89,101,102]; for that reason, this model simultaneously addressed the behavioral and operational dimensions focused on the sustainability of their actions, which meets the needs and attributes of its customers and other stakeholders.

The development of this research tries to propose a tool that helps the mitigation of waste at its origin—that is, an understanding of the hidden complexities in the behavioral and operational relationships present in the activity of producing meals in the food service, enabling a better understanding of the root causes and underlying factors of food waste generation.

5. Conclusions

This study proposed the development of a structural model to analyze the causal relationships between behavioral and operational factors responsible for reducing efficiency and generating waste in food service. Although the recent literature presents some studies on the behavior of food waste in the food service, through the modeling of structural equations, these usually do not consider the aspects of the operational nature, leaving a gap that can help the decision-making by managers, improving the productive and environmental indicators.

The preliminary model was developed based on a systematic literature review, defining seven hypotheses and twenty-seven indicators. Of the eight hypotheses present in the preliminary structural theoretical model, five were confirmed, representing about 63%.

However, the three unsupported hypotheses were maintained in the model because they established a complete mediation relationship for H1 and H4, as well as having theoretical relevance.

When we evaluate the set of hypotheses, the results of the structural model indicate that the hypotheses H2, H5, H6, H7, and H8 are more relevant for the output of the model, represented by construct DS.

Hypothesis H2 presents a moderate effect, considering the nature of the research. This relationship implies that the CD construct is influenced by the IN and MO constructs. Observing the result of the participation of the MO construct indicators, the Situational Factors Indicator MO4 has 36% and the Social Norms MO5 33%, together presenting half of the explained variance of the construct. This shows the need for management alignment with these aspects and more frequent monitoring of these indicators.

Observing the hypotheses H5 and H6 and the results of the participation of the indicators in these constructs, the HB construct indicator, Technical skills HB2, presented a participation of 42%, signaling that the development of the technical skills of employees by training improves efficiency in the control and management of waste. The HB1 Technical Knowledge indicator, with 24% participation, emphasizes the importance of access to information as an instrument for knowledge training. In the OP construct, the indicator Training OP4, with 35%, highlights the need for continuous training by food service units for their employees, similar to the HB2 indicator. The indicators Operational scheduling (OP3) with 20% and (OP1) Infrastructure with 46% should also be seen as timely, considering that the improvement of FC, Correction Factors, depends directly on these indicators.

The H8 hypothesis highlights the participation of the indicators Abundance habits (CD1) with 32%, Health Risks (CD2) with 40% and Healthy habits (CD3) with 28% of the CD construct, in the formation of the CGD construct. The CD1 indicator serves to draw the attention of customers and employees to the losses resulting from this type of habit, either at the time of putting their dishes together or even during production planning. The CD2 indicator shows the importance of making it clear to customers and diners that the food service unit ensures the maintenance of POPs and sanitary factors, as well as that the stages of handling and preparation adequately take care of the aspects of sanitization and hygiene of the foods to be processed. The CD3 indicator shows it is essential to guide diners about the propensity to waste foods that are considered healthy, usually low in sugar and fat.

Analyzing hypothesis H7, an important influence of the CGD construct on the construct DS was observed. In addition to its indicators, the CGD3 People’s Management indicator, with 51% participation, highlights the importance of human resources management in hiring professionals with technical skills and skills to develop specific activities such as food handling and preparation. The CGD2 Methods and Procedures indicator with 29% indicates the importance of implementing appropriate management methods and techniques for consolidating continuous improvement actions in each production step. The CGD1 20% indicator highlights the importance of investing in the acquisition and periodic maintenance of freezing, refrigerating, and heating equipment for proper food preservation.

Considering the indicator s of the DS construct, the indicators Extensive land use (DS1) with28%, Emissions CH₄ (DS2) with 27%, and Water consumption (DS4) with 18% are more critical, considering their impacts on the environmental and social spheres. These are followed by 11% for Global food supply (DS3) and 17% for Exacerbated use of pesticides and fertilizers (DS5).

Analyzing the structural relationships of the proposed model, it is possible to observe which aspects are more critical from a managerial point of view. The results indicate that environmental aspects should be widely used throughout employee and customer awareness campaigns. The hiring of professionals with technical skills and abilities to develop specific activities, such as handling and preparing food, also indicates that the development of the technical skills of employees, through training and access to information, is perceived as important when we observe the operational constructs.

Future research could incorporate other formative constructs for the operational part of the model, focusing on the measurement of food waste in steps such as food handling and preparation. The insertion of new formative constructs into the model will improve its prediction of the generation of waste. In addition, further studies may focus on other states in the region that have different climatic conditions, food dependence, and sociocultural aspects.