A Logistic Regression Model for the Analysis of Attitudes and Behaviours Towards Functional Foods Among Senior Consumers Aged 60+ Years

Abstract

The aim of this study was to establish models of attitudes and behaviours of senior consumers towards functional foods. Due to the ageing societies in modern Europe, it is crucial to identify opportunities to ensure the well-being of seniors. This is all the more important because, in every branch of the economy, this social group still plays a significant role in its operation. One method that seniors can use to promote their health is to include functional foods in their daily diets. Therefore, it is important to skilfully model this social group’s attitudes and behaviours. For this purpose, this article proposes models based on logarithmic regression. Due to its properties, on the one hand, this method is a scientist-friendly tool, and on the other hand, it allows for the accurate modelling of a real problem. The four analytical and forecasting models proposed in this article were based on survey research conducted in a distinct social group. The models characterise seniors’ food neophobia and attitudes towards functional foods according to independent descriptive variables that influence the dependent variable. Marketers should use the results obtained to prepare sales strategies for products and functional foods among seniors.

1. Introduction

Europeans are living longer than ever before, and the age profile of society is rapidly advancing. Demographic ageing in the EU means that the proportion of people of working age is shrinking, while the number of older people is growing; this pattern will continue over the next couple of decades, as the post-war Baby Boom Generation completes its move into retirement. In 2018, almost one-fifth of the population in the European Union was over the age of 65, and this proportion will increase to 28.5% by 2050 [1]. Chronic non-communicable diseases (NCDs) accounted for 71% of deaths worldwide in 2016 and were responsible for 90% of deaths in Poland [2]. The main non-communicable diseases include cardiovascular diseases and complications, respiratory diseases, and related diseases that affect the life expectancy of those suffering from them. These health problems are strongly influenced by diet. The connection between healthy ageing and nutrition is an emerging scientific problem [2,3,4,5,6,7,8]. Ensuring the well-being of people and preventing non-communicable diseases at the end of life are also top priorities for policymakers [9,10,11]. Functional foods can play a key role in achieving these goals; therefore, their development is one of the most important fields of innovation in the food sector [12,13,14,15,16].

The ageing process correlates with a decline in many physiological functions affecting nutritional status. Among the older population, the risk of inadequate diet and malnutrition is significant [17]. The dietary preferences of older people differ from those of other age groups in several aspects, which justifies a research approach tailored to their needs [4,18,19,20]. Relevant studies have shown that older people are generally interested in healthy eating [21,22]. It has also been proven that they are willing to spend money on products that meet their needs [23,24,25,26]. However, some specific research findings regarding older consumers’ preferences should be considered.

For example, food neophobia (i.e., avoiding trying new products) has emerged as a critical problem among older consumers [27]. Compared to younger customers, older consumers pay more attention to information on the label, including nutritional properties, which highlights the marketing potential of functional foods in this age group [19,28,29].

Meeting the nutritional needs of older consumers through the product types they desire benefits the public health community and the food industry. However, the heterogeneity of the elderly market is a challenge and requires market segmentation. Although many researchers have proposed ways to segment the elderly consumer population, not enough attention is paid to the food market for this group of customers. Therefore, marketing activities that focus on older people’s nutritional needs can benefit both public health and the food industry.

However, from a marketing perspective, meeting these needs is a challenge. First, older adults who have decreased appetites are unlikely to be able to meet their nutritional needs through increased intake. Therefore, a more promising approach may be the commercialisation of nutrient-enriched foods [30], a type of functional food [31] that is relatively rich in nutrients, considering its volume. Commercialising these products may pose a second challenge. All subgroups of the older population may not readily accept functional foods specifically targeted at older consumers due to age-related stigmatisation [32,33]. In contrast, when functional foods are advertised as healthy alternatives to conventional products, older people are generally willing to try them [24,25,34]. Therefore, the general concept of functional foods may appeal to older people. However, specific functional foods may not accomplish this, which poses a third challenge to meeting the nutritional needs of older adults. Although the senior population of older adults can be classified based on age range (defined here as age 55 and over), it is highly heterogeneous in composition [35]. During the decade from age 50 to 60, consumers undergo many life changes and thus become less similar [36]. Therefore, older adults may have similar nutritional needs (i.e., nutrient requirements), but their food desires (i.e., product preferences) differ significantly.

Thus, the main aim of this article is to propose models of the attitudes and behaviours of senior consumers towards functional foods. To this end, this paper presents logit regression models. Them giving the answers to the behaviours and attitudes of seniors towards functional food is consistent with implementing Sustainable Development Goals No. 2 and 3 (https://sdgs.un.org/2030agenda (accessed on 7 November 2024)). Furthermore, the proposed models also allow one to verify the following hypotheses:

• neophobia is an attitude not only a food dysfunction;
• the neophobia as the attitude impacts both consumer food and eating habits and consumer awareness about the composition;
• the attitudes towards functional food relate to everyday eating habits and consumer awareness regarding health-promoting food ingredients described by frequency of consumption;
• the level of neophobia relates to negative attitudes towards functional food;
• a relationship exists between attitudes towards functional foods and behaviours towards healthy foods.

This article consists of the following sections. First is the Introduction, where we briefly place the study about senior consumer behaviour and attitudes in a broad context and highlight why it is important. Next is Section 2, which describes the theoretical basis and analysis used to obtain the results, which are presented in Section 3. Finally, we provide some general conclusions.

2. Materials and Methods

2.1. Functional Food

The concept of functional foods arose in Japan. The Ministry of Education, Science and Culture (MESC) introduced the national functional food programme. Its main focus was the relation between food science and medicine. In 1991, the Japanese Ministry of Health and Welfare (MHW) defined this new type of food as the “Food for Special Health Use (FOSHU)”. Later, in 2001, a broader definition was introduced: Food with Health Claims (FHC). Commonly, there were two main groups [37]:

• Food with Nutrient Function Claims—food that contains additives (ingredients), currently 12 vitamins and 5 mineral products. These foods can be controlled and distributed without the required registration or notification of compliance with legal requirements, provided that they meet the requirements of the established standards.
• Foods for Specified Health Use—officially approved food that has a positive physiological effect on the human body. Such food contains ingredients with health-promoting properties that are formally recognised as having a physiological effect on the human body and should be consumed to maintain health. To sell food as FOSHU, it is necessary to assess its safety and compelling impact on health, and the declaration must be approved by the Minister of Health and Care each time.

Japan’s approach to, research on, and legislation of functional foods have set a benchmark for other countries and provided a basis for their regulations. For instance, the Food and Nutrition Board of the United States defines functional food as food, or its components, modified to provide more significant health benefits [37]. This definition draws inspiration from Japan’s pioneering work [38].

Canadian law defines functional foods as conventional foods or foods like those in a traditional, regular diet that have a proven beneficial effect on health beyond their basic nutritional functions. These foods have the potential to not only improve health but also significantly reduce the risk of chronic diseases, offering a promising outlook for the future [39].

In Australia and New Zealand, novel foods are functional foods if they are traditional foods intended for consumption as part of a regular diet but have been modified so that their functions go beyond simple nutritional requirements. They are not the same as nutraceuticals [40].

In China, functional foods are foods with effects intended for consumption by specific groups of people. They affect the function of the human body but do not cure diseases [41].

In India, functional foods include alternative products, including modified foods or their ingredients, that provide health effects beyond traditional functional functions [42].

In 1996, the Functional Food Science in Europe (FUFOSE) research programme, funded by the European Commission, was launched. According to the results of this project, in 1999, in the final FUFOSE document, functional food was defined as food that has been proven to have a beneficial effect on one or more body functions beyond its nutritional impact. This effect consists of improving health and well-being and reducing the risk of diseases. Functional foods must resemble conventional foods in form and demonstrate beneficial effects in amounts that are expected to be generally consumed in the diet: these are not tablets or capsules but part of a healthy diet ([43,44,45,46]).

2.2. Elderly Consumers and Functional Food

The number of adults in the world aged 60 and over is expected to triple by 2100, increasing from 841 million in 2013 to 2 billion in 2050 and almost 3 billion in 2100 [47,48]. Statistical data clearly illustrate the ongoing change in the demographic structure in highly developed countries. An increasing share of their population comprises a group of seniors discreetly taking over a key position on the market. Additionally, let’s take into account the fact that the financial status of seniors is determined not only by income but also by the amount of resources they have accumulated so far. They can spend more money on their needs. Thus, they constitute a group with great potential to create demand, provided these goods and services fit into their new lifestyle [49]. Older people do not want to be perceived as having difficulties associated with ageing; they want to occupy an essential place in society and be treated subjectively. It is consistent with the term “active ageing” adopted by the World Health Organization in the late 1990s. It is optimising opportunities for health, participation and safety to improve quality of life with advancing age [50]. It refers to the continued involvement in social, economic, cultural, spiritual, and civic relationships, as well as the ability to be physically active or actively participate in the world of work. Furthermore, “older people” have become a segment of interest because of their social importance and spending, as many studies have shown their high purchasing power [51].

The large diversity of the senior consumer group makes it difficult to characterise this group precisely. Nevertheless, certain standard features that distinguish seniors from other consumer groups on the market can be identified. Senior consumers are demanding, rational, consciously approach shopping, need more time to decide, prefer specific information about the product, strive to obtain the most complete information that will facilitate making a rational choice. They pay attention to the product’s practicality, are individualists (they are less susceptible to peer pressure), are emotionally involved, are loyal to specific brands, and are also interested in novelties. They tend to trust companies with a good reputation and social involvement.

In the literature, there are various divisions of older consumers. The proposed typology related to age is the division of Salomon, Bamossi and Askegard [52]:

• older consumers (55–64 years old),
• relatively old consumers (65–74 years old),
• old consumers (75–84 years old),
• very old consumers (over 84 years old).

Despite the perceived attractiveness of seniors as a new and growing consumer group, the knowledge of marketers and other media broadcasters (including public ones) about this target group is relatively poor. First, they perceive these consumers as a monolithic group, not constantly aware of the diverse segments within it. This is not about simple age, gender or income differentiation but a combination of psychographic variables, based on which we can distinguish subgroups characterised by specific features in terms of, among others, shopping attitudes, values, and lifestyle. It is worth noting that considering the diversity of the older people in the media message not only increases the chance of reaching a selected group of recipients with a specific message but can also affect the perception of old age as such.

Although there is no single, universally accepted definition of old age, there is agreement in sociological discourse that old age is a static concept. Ageing is, however, considered a dynamic process that should be analyzed about the biological, social, and psychological dimensions of human life. Most often, three phases of old age are distinguished (according to WHO [53]): the so-called young old, or “young old” (age III: 60–74), old old—“old old”, “mature old age” (age IV: 75–89), and oldest old, long life—“long-lived” (over 90 years).

Maintaining good health in old age is of great importance. WHO has defined healthy ageing as “the process of developing and maintaining functional abilities that enable well-being in old age” [53]. Adequate, healthy nutrition is a critical factor in achieving this goal. Ageing has a profound impact on the physiology and functions of the body, affecting the nutritional requirements of older people.

As mentioned above, older people are willing to spend money on products that meet their needs, such as functional foods. Most age-related disorders can be prevented by using appropriate nutritional interventions and consuming foods rich in nutrients and antioxidants.

2.3. Food Neophobia

The definition of food neophobia (FN) shows that it is the reluctance to eat or the avoidance of new foods [54]. Previous studies on food neophobia tend to treat it as a disease entity or a nutritional dysfunction [54,55]. They focus more on its negative effects on human health and the causes of its occurrence [54,55,56,57]. However, when analyzing the literature on the subject, it should be noted that FN primarily affects children up to a certain age, then sometimes disappears partly or entirely, only to begin to occur in the senior group [56,57,58]. This approach shows that neophobia can be treated as a disease. However, some researchers show that neophobia is the attitude to eat novel, not-known food, sometimes also healthy food [58,59]. More and more often, functional food products are becoming this new unknown food for consumers. According to this aspect, some research results show the relationship between neophobia and eating behaviour, dietary patterns, and food choice motives [56,59,60,61,62,63].

Thus, we assume that neophobia is an attitude, not a disease or behaviour. Furthermore, it impacts behaviours in eating novel food, i.e., functional food. Therefore, understanding the relationship between food neophobia, attitudes to healthy eating habits and behaviours towards functional food in more depth appears to be a vital issue.

2.4. Logistic Regression

The main results presented in this article are based on logistic regression [64,65]. In general, a logistic regression model (like multiple linear regression) is used to examine the impact of many independent variables $\left(X_1, X_2,\dots , X_k\right)$ on one dependent variable $\left(Y\right)$ [64,65,66,67,68,69,70,71]. However, the dependent variable only takes two values, positive/negative or high/low, in the presented model. They are coded as {1}, indicating a distinguished value—having a given feature—or {0}, meaning the absence of the given feature [56,57,58,59,60,61,62].

Function (1), on which the logistic regression model is based, calculates the probability of variable Y taking a distinguished value. It is not used to determine the value of a two-level Y variable. The equation is as follows [64,65,66,67,72]:

$$P\left(Y=1|X_1, X_2,\dots ,X_k\right)={\displaystyle \frac{e^Z}{1+e^Z}},$$

(1)

where $P\left(Y=1|X_1, X_2,\dots ,X_k\right)$ is the probability of adopting the distinguished value {1}, provided that specific values of the independent variables are obtained, i.e., the so-called predicted probability for 1;$X_1, X_2,\dots ,X_k$(k = 3 or k = 4 in Section 3.2) are the independent, explanatory variables; and Z is most often a linear relationship given by

$$Z={\beta }_0+{\displaystyle \sum }_{i=1}^k{\beta }_i{X}_i,$$

(2)

with parameters ${\beta }_1,{\beta }_2,\dots ,{\beta }_k$, and k = 3 or k = 4 in Section 3.2.

Now, we introduce the logit function as the following Formula (3) [64,65,66,67,72]:

$$ln\left({\displaystyle \frac{P}{1-P}}\right)=Z,$$

(3)

where P indicates the probability of an event (e.g., positive attitude).

Based on the coefficients, for each independent variable in the model, an easy-to-interpret measure is calculated—the individual odds ratio (OR) [64,65,66,67,72]:

$$O{R}_i=e^{{\beta }_i}.$$

(4)

The odds ratio obtained using (4) expresses the change in the chance of the highlighted value {1} occurring when the independent variable increases by 1 unit. This result is adjusted for the remaining independent variables in the model to assume that they remain constant while the independent variable under study increases by one unit.

We interpret the OR value as follows [64,65,66,67,72]:

• When OR_i > 1, the examined independent variable has a positive influence on the probability of obtaining the distinguished value {1}; i.e., the OR value expresses how much the chance of occurrence of the distinguished value {1} increases when the independent variable increases by one level.
• When OR_i < 1, the examined independent variable has a negative influence on the probability of obtaining the distinguished value {1}; i.e., the OR value expresses how much the chance of occurrence of the distinguished value {1} decreases when the independent variable increases by one level.
• When OR_i ≈ 1, the independent variable under study does not influence the probability of obtaining the distinguished value {1}.

Additionally, we should consider the covariance matrix when building the logit regression model. The following are well-known facts [68,69,70,71]:

• A positive covariance value means that if one variable increases, the other tends to increase;
• A negative covariance value means that if one variable increases, the other tends to decrease.

This, in turn, affects the meaning of the OR sign. According to this, we should interpret the OR value as follows:

• When OR_i > 1, the examined independent variable has a positive influence on the probability of obtaining the distinguished value {1}; i.e., the OR value expresses how much the chance of occurrence of the distinguished value {1} increases when the independent variable decreases by one level.
• When OR_i < 1, the examined independent variable has a negative influence on the probability of obtaining the distinguished value {1}; i.e., the OR value expresses how much the chance of occurrence of the distinguished value {1} decreases when the independent variable decreases by one level.

In this way, we obtain a dual approach to the model’s interpretation.

Logit models transform probabilities into log odds in Formula (4) for individuals. Our research uses only two states (with and without a given feature). Thus, we must consider only two chances for values {1} and {0} in the logit model. The general formula for the chance function (FC) for both values is given by Formula (5) (based on (2) and (3)) [64,65,66,67,72]:

$$F{C}_{X_i}\left(j\right)=\left({\displaystyle \frac{P\left(j\right)}{1-P\left(j\right)}}\right)=e^Z=e^{{\beta }_0+{\beta }_1{X}_1+{\beta }_2{X}_2+\dots +{\beta }_i{X}_i\left(j\right)+\dots +{\beta }_k{X}_k},$$

(5)

where j = 1, 2 is the number of categories for the selected variable X_i, k = 3 or k = 4 in Section 3.2.

Therefore, the formula for the general OR takes the following form (exemplary for variable X₁) [64,65,66,67,72]:

$$OR\left[F{C}_{X_1}\left(2\right)/F{C}_{X_1}\left(1\right)\right]=\left({\displaystyle \frac{e^{{\beta }_0+{\beta }_1{X}_1\left(2\right)+{\beta }_2{X}_2+\dots +{\beta }_k{X}_k}}{e^{{\beta }_0+{\beta }_1{X}_1\left(1\right)+{\beta }_2{X}_2+\dots +{\beta }_k{X}_k}}}\right)={\left(e^{{\beta }_1}\right)}^{\left[X_1\left(2\right)-X_1\left(1\right)\right]}.$$

(6)

This formula tells us how much the chance of occurrence of the highlighted value will change when the multiplicity of the quotient changes by $\left[X_1\left(2\right)-X_1\left(1\right)\right]$. This means the odds ratio calculated for non-adjacent categories of the distinguished variable: i.e., it is an $\left[X_1\left(2\right)-X_1\left(1\right)\right]$-fold odds ratio.

Verifying the obtained model is very important in logistic regression [65,66,67,68,69,70,71,72]. In this case, we must take into account two things. The first is the statistical significance of individual variables in the model or, equivalently, the significance of the odds ratio [64,65,66,67,72]. The second is the quality of the built model. Below, the most useful methods for both assessments of the model are described.

Based on the coefficient and its estimation error, we can conclude whether the independent variable for which this coefficient was estimated significantly impacts the dependent variable. For this purpose, we use the Wald test.

A good model should meet two primary conditions: suitable and simple. We can assess the quality of a logistic regression model using several measures, which are based on the following values:

• L_FM—the maximum of the likelihood function of the full model (with all variables);
• L₀—the maximum of the likelihood function of the model containing only the intercept;
• n—the sample size.

Information criteria are based on the entropy of information carried by the model (model uncertainty); i.e., they estimate the information lost when a given model is used to describe the phenomenon under study. We should therefore choose a model with the minimum value of a given information criterion. The following criteria can be used [64,65,66,67,72]:

• Akaike’s information criterion is given by the following formula:

  $$AIC=-2\ln L_{FM}+2k.$$

  (7)
• The improved Akaike’s information criterion, based on (7), is defined as follows [64,65,66,67,72]:

  $$AICc=-2\ln L_{FM}+{\displaystyle \frac{2k\left(k+1\right)}{n-k-1}}.$$

  (8)
• The Bayesian Information Schwarz’s Criterion—BIC—considers the sample size used, similarly to AICc [64,65,66,67,72]:

  $$BIC=-2\ln L_{FM}+k\ln n.$$

  (9)

AIC, AICc, and BIC are compromises between goodness of fit and complexity. The second sum element in the information criteria formulas (loss or penalty function) measures the model’s simplicity. It depends on the number of variables in the model (k) and the sample size (n). In both cases, this element increases with the number of variables; this increase is faster the smaller the number of observations.

However, the information criterion is not an absolute measure: i.e., if all compared models misdescribe reality in the information criterion, there is no point in looking for a warning. Thus, as with the other methods, we introduce the Pseudo-R², also called the McFadden R², which is a measure of model fit (equivalent to the coefficient of multiple determination R² calculated for linear multiple regression). The following formula is used [64,65,66,67,72]:

$$R_{pseudo}^2=1-{\displaystyle \frac{\ln L_{FM}}{\ln F_0}}.$$

(10)

The value of this coefficient is in the range [0; 1), where values close to 1 mean a perfect fit of the model, and 0 means no fit at all. Since the coefficient does not take the value 1 and is sensitive to the number of variables in the model, its corrected value is determined as follows [64,65,66,67,72]:

$$R_{Naglekerke}^2={\displaystyle \frac{1-e^{-\left(\frac{2}{n}\right)\left(\ln L_{FM}-\ln F_0\right)}}{e^{-\left(\frac{2}{n}\right)\ln F_0}}},$$

(11)

or

$$R_{Cox-Snell}^2=1-e^{{\displaystyle \frac{2\ln L_{FM}-2\ln F_0}{n}}}.$$

(12)

The likelihood ratio test is a basic tool for estimating the significance of all variables in the model. It is based on a chi-square distribution with k degrees of freedom.

Another possibility to assess the significance of variables is the Hosmer–Lemeshow test [64,65,66,67,68,69,70,71,72]. It compares the observed frequencies of occurrence of the distinguished value O_g and the predicted probability E_g for different subgroups of data. If O_g and E_g are close enough, then it can be assumed that a well-fitting model has been built. For calculations, the observations are first divided into G subgroups—usually into deciles (G = 10). It is based on a chi-square distribution with G-2 degrees of freedom. In the case of the Hosmer–Lemeshow test, we note that the lack of significance is desirable because it indicates the similarity of the observed numbers and the predicted probabilities.

The third method applied for testing the significance of the variables is the area under the receiver operating characteristic curve (ROC curve). This curve, built based on the value of the dependent variable and the predicted probability of the dependent variable P, allows for the assessment of the ability of the built logistic regression model to classify cases into two groups: {1} and {0}. The resulting curve, particularly the area under it, illustrates the classification quality of the model. When the ROC curve coincides with the diagonal y = x, assigning a case to the selected class ({1} or {0}) based on the model is as good as randomly dividing the studied cases into these groups. The classification quality of the model is good when the curve is well above the diagonal y = x, i.e., when the area under the ROC curve is much larger than the area under the line y = x, and therefore, its classification ability is better than 0.5 [64,65,66,67,72].

3. Results

This section presents the empirical survey results about attitudes towards functional food. It consists of three subsections: assumptions and research methodology (to describe the general assumptions for the research), numerical results (to provide a concise and precise description of the experimental results), and experimental conclusions and remarks.

3.1. Assumptions and Research Methodology

The research’s whole concept tries to help realise the Sustainable Development Goals Nos. 2 and 3. In the context of ageing societies, the models proposed in this part are intended to support the functional food segment of senior consumers. It is essential to know their attitudes and be able to link them to behaviours related to functional food. This makes it possible to help this social group adapt to the changes associated with ageing.

A survey was conducted to identify factors influencing food neophobia and attitudes towards functional food in 510 seniors. A Likert scale [73] was used to assess eating attitudes and behaviours in the survey. This is an ordinal scale in which a set of short, most often declarative sentences are defined. Their task is to characterise attitudes towards the examined object. Each item of this scale is assigned an ordinal, bipolar, usually 5-point intensity scale [73,74,75].

In the case of the prepared survey, areas containing statements subject to analysis were assessed so that each of them was given a point value ranging from 1 for a highly incorrect statement to 5 for an extremely correct statement. Then, the points obtained by each person were added up, and on this basis, the entire analysed area was assessed. When assessing areas, the point values were calculated so that the lowest possible value to be achieved was 0 (evaluation logic). The characteristics of the assessments of these areas are included later in the descriptions of statistical variables regarding eating attitudes and behaviours. The point values of individual statements and those of entire regions were then used for further comparative analyses in this work.

For each respondent, the sum of ratings describing opinions on individual issues was calculated (ranging from 9 to 45 points). Then, based on all of the sums of the ratings, two indicators were calculated, namely, the average value of the sums (X = 27.4 points) and the standard deviation (SD = 6.1 points). Missing data were removed case by case, and only the attitudes of people who rated each of the 11 statements were measured. These indicators were used to distinguish three numerical ranges: from 9 points to X−SD (21.3 points), from X−SD (21.3 points) to X+SD (33.5 points), and from X+SD (33.5) to 45 points, which corresponded to a low level of neophobia (16.7% of respondents), an average level (66.4%), and a high level (16.9%) of neophobia, respectively. Only two values were considered for the analysis following logistic regression assumptions: low {0} and high {1} levels of neophobia. The intermediate level was omitted.

For each respondent, the sum of ratings describing opinions on individual issues was calculated (ranging from 6 to 30 points). Then, based on all of the sums of ratings, two indicators were calculated, namely, the average value of the sums (X = 13.4 points) and the standard deviation (SD = 4.0 points). Missing data were removed case by case, and only the attitudes of people who rated each of the six statements were measured. These indicators were used to distinguish three numerical ranges: from 6 points to X−SD (9.3 points), from X−SD (9.3 points) to X+SD (17.4 points), and from X+SD (17.4) to 30 points, which corresponded to positive (16.7% of respondents), neutral (67.3%), and negative (15.9%) attitudes towards health and healthy nutrition, respectively. Only two values were considered for the analysis under logistic regression assumptions: positive attitudes {0} and negative attitudes {1}. Indifferent attitudes were omitted.

3.2. Numerical Results

The survey research allowed us to propose four models of attitudes and behaviours (two for neophobia and two for attitudes towards functional food). This impact study allowed us to determine, in detail, the following relationships:

• The relationship between neophobia and consumer eating and food habits—Model 1;
• The relationship between neophobia and consumer awareness about the composition—Model 2;
• The relationship between attitudes towards functional food and everyday eating habits—Model 3;
• The relationship between attitudes towards functional food and consumer awareness regarding health-promoting food ingredients described by frequency of consumption—Model 4.

The selection of dependent and independent variables for the constructed models was not accidental. Models 1 and 3 examine the strength of the relationships of the level of neophobia and attitudes towards functional food with the daily eating habits of seniors. In turn, Models 2 and 4 measure the strength of the relationships of the same dependent variables with the approach to the ingredients of food products. With Model 2, we examine a more general relationship than with Model 4. Models constructed in this way also allow for studying the existence of a relationship, and possibly its strength, between neophobia and attitudes towards functional food. That is, the method we propose serves to link attitudes and consumer behaviours. This is of great importance for such heterogeneous groups as seniors.

The general statistics for the particular models obtained under these assumptions are in

Table 1. Basic data about the surveyed sample and survey results.

Values	Model 1	Model 2	Model 3	Model 4
Number of surveys	510	510	510	510
Cardinality—missing data	355	357	354	357
Severity level	0.05	0.05	0.05	0.05
Cardinality	155	153	156	153
Number of variables in the model	3	4	3	3
Group size {0}	77	76	78	77
Group size {1}	78	77	78	76

Source: [own studies].

.

For further research, we define the explanatory variables for Models 1–4.

• Explanatory variables for Model 1:

Y¹—neophobia, dependent variable (values: high = 1—means ‘lack of trust in new, unfamiliar foods and not eating them’; low = 0—means ‘high level of trust in new, unfamiliar foods and eating them’);

$X_1^1$—independent variable “I pay attention to the health benefits of the food I eat” (values: 5—definitely yes; 4—probably yes; 3—neither yes nor no; 2—probably not; 1—definitely not);

$X_2^1$—independent variable “I make sure that my diet contains a lot of vitamins and minerals” (values: 5—definitely yes; 4—probably yes; 3—neither yes nor no; 2—probably not; 1—definitely not);

$X_3^1$—independent variable “I only choose fresh, unprocessed products” (values: 5—definitely yes; 4—probably yes; 3—neither yes nor no; 2—probably not; 1—definitely not).

2.

Explanatory variables for Model 2:

Y²—neophobia, dependent variable (values: high = 1—means ‘lack of trust in new, unfamiliar foods and not eating them’; low = 0—means ‘high level of trust in new, unfamiliar foods and eating them’);

$X_1^2$—independent variable “enriched with omega-6 and omega-3 fatty acids” (values: 5—definitely yes; 4—probably yes; 3—neither yes nor no; 2—probably not; 1—definitely not);

$X_2^2$—independent variable “containing probiotic bacteria” (values: 5—definitely yes; 4—probably yes; 3—neither yes nor no; 2—probably not; 1—definitely not);

$X_3^2$—independent variable “with reduced sugar content” (values: 5—definitely yes; 4—probably yes; 3—neither yes nor no; 2—probably not; 1—definitely not);

$X_4^2$—independent variable “with reduced sodium content” (values: 5—definitely yes; 4—probably yes; 3—neither yes nor no; 2—probably not; 1—definitely not).

3.

Explanatory variables for Model 3:

Y³—attitudes towards functional food, dependent variable (values: negative = 1—means ‘I only eat traditional foods.’; positive = 0—means ‘I eat functional foods.’);

$X_1^3$—independent variable “I am constantly trying new and different foods” (values: 5—I strongly agree; 4—I agree; 3—hard to say; 2—I disagree; 1—I strongly disagree);

$X_2^3$—independent variable “I eat almost everything” (values: 5—I strongly agree; 4—I agree; 3—hard to say; 2—I disagree; 1—I strongly disagree);

$X_3^3$—independent variable “When preparing dishes, I like to try new recipes” (values: 5—I strongly agree; 4—I agree; 3—hard to say; 2—I disagree; 1—I strongly disagree).

4.

Explanatory variables for Model 4:

Y⁴—attitudes towards functional food, dependent variable (values: negative = 1—means ‘I only eat traditional foods.’; positive = 0—means ‘I eat functional foods.’);

$X_1^4$—independent variable “lowering cholesterol levels” (values: 5—several times a week; 4—once a week; 3—several times a month; 2—once a month; 1—less than once a month or not at all);

$X_2^4$—independent variable “with reduced sugar content” (values: 5—several times a week; 4—once a week; 3—several times a month; 2—once a month; 1—less than once a month or not at all);

$X_3^4$—independent variable “anti-ageing (improving the appearance of the skin, enriched with vitamins and minerals; most often, these products contain green tea and herbal extracts)” (values: 5—several times a week; 4—once a week; 3—several times a month; 2—once a month; 1—less than once a month or not at all).

Based on the assumptions mentioned above, the parameters of logistic regression models were estimated using specialised PQstat software ver. 1.8.6. [76]. These data are included in Table 2, Table 3, Table 4, Table 5, Table 6, Table 7, Table 8 and Table 9.

The parameters for Model 1 are introduced in Table 2.

Table 2. Numerical results for Model 1 parameters.

Term	Est. Value βi	Std. Dev. βi	−95% CI	+95% CI	Wald Stat.	p-Value	OR	−95% CI	+95% CI
Free	−1.441237	0.519616	−2.45967	−0.4228	7.693178	0.005543	0.236635	0.085464	0.655204
$X_1^1$	0.847827	0.308298	0.243574	1.45208	7.562634	0.005959	2.334569	1.275801	4.271992
$X_2^1$	0.578859	0.287697	0.014983	1.14274	4.048324	0.044215	1.784002	1.015096	3.135332
$X_3^1$	−0.483777	0.229919	−0.93441	−0.0331	4.427336	0.035368	0.616451	0.392818	0.967398

Source: [own study].

Table 2. Numerical results for Model 1 parameters.

Term	Est. Value βi	Std. Dev. βi	−95% CI	+95% CI	Wald Stat.	p-Value	OR	−95% CI	+95% CI
Free	−1.441237	0.519616	−2.45967	−0.4228	7.693178	0.005543	0.236635	0.085464	0.655204
$X_1^1$	0.847827	0.308298	0.243574	1.45208	7.562634	0.005959	2.334569	1.275801	4.271992
$X_2^1$	0.578859	0.287697	0.014983	1.14274	4.048324	0.044215	1.784002	1.015096	3.135332
$X_3^1$	−0.483777	0.229919	−0.93441	−0.0331	4.427336	0.035368	0.616451	0.392818	0.967398

Source: [own study].

The p-value determined in the Wald test for every variable (shown in column 7 in Table 2) is less than 0.05. Thus, all explanatory variables are significant for this model.

Next, a covariance matrix was calculated to fully consider the odds ratio in the model. The matrix shows that the variances for individual variables have negative signs (see Table 3).

Table 3. Covariance matrix for Model 1.

	Y1	${\mathit{X}}_1^1$	${\mathit{X}}_2^1$	${\mathit{X}}_3^1$
Y1	0.270001	−0.041391	−0.04681	−0.0426
$X_1^1$	−0.041391	0.095048	−0.03917	−0.0192
$X_2^1$	−0.046809	−0.039166	0.08277	−0.0198
$X_3^1$	−0.042575	−0.019214	−0.01976	0.05286

Source: [own study].

Table 3. Covariance matrix for Model 1.

	Y1	${\mathit{X}}_1^1$	${\mathit{X}}_2^1$	${\mathit{X}}_3^1$
Y1	0.270001	−0.041391	−0.04681	−0.0426
$X_1^1$	−0.041391	0.095048	−0.03917	−0.0192
$X_2^1$	−0.046809	−0.039166	0.08277	−0.0198
$X_3^1$	−0.042575	−0.019214	−0.01976	0.05286

Source: [own study].

Hence, the chance of occurrence of a high level of neophobia depends on the variables mentioned above, as described by the Odds Quotient:

• Variable $X_1^1$—paying attention to the health benefits of the food one eats: OR [95%CI] = 2.33 [1.28; 4.27]. Thus, the less we pay attention to the health benefits of food, the greater the chance of a high level of neophobia (2.33 times higher).
• Variable $X_2^1$—ensuring that the diet is rich in vitamins and minerals: OR [95%CI] = 1.78 [1.02; 3.14]. Thus, the less we care about whether a diet is rich in vitamins and minerals, the greater the chance of a high level of neophobia (1.78 times higher).
• Variable $X_3^1$—choosing only fresh and unprocessed products: OR [95%CI] = 0.62 [0.39; 0.97]. Thus, the less often we use fresh and unprocessed products, the lower the chance of a high level of neophobia.

The parameters for Model 2 are introduced in Table 4.

Table 4. Numerical results for Model 2 parameters.

Term	Est. Value βi	Std. Dev. βi	−95% CI	+95% CI	Wald Stat.	p-Value	OR	−95% CI	+95% CI
Free	−3.015841	0.568433	−4.129949	−1.901733	28.148709	<0.000001	0.049005	0.016084	0.14931
$X_1^2$	0.494527	0.153014	0.194625	0.794429	10.445241	0.00123	1.639723	1.214856	2.213177
$X_2^2$	0.406695	0.137199	0.137791	0.6756	8.78694	0.003034	1.501847	1.147735	1.965212
$X_3^2$	−0.359535	0.174173	−0.700908	−0.018162	4.261086	0.038995	0.698001	0.496135	0.982002
$X_4^2$	0.423496	0.166683	0.096803	0.75019	6.455278	0.011062	1.527292	1.101643	2.117402

Source: [own study].

Table 4. Numerical results for Model 2 parameters.

Term	Est. Value βi	Std. Dev. βi	−95% CI	+95% CI	Wald Stat.	p-Value	OR	−95% CI	+95% CI
Free	−3.015841	0.568433	−4.129949	−1.901733	28.148709	<0.000001	0.049005	0.016084	0.14931
$X_1^2$	0.494527	0.153014	0.194625	0.794429	10.445241	0.00123	1.639723	1.214856	2.213177
$X_2^2$	0.406695	0.137199	0.137791	0.6756	8.78694	0.003034	1.501847	1.147735	1.965212
$X_3^2$	−0.359535	0.174173	−0.700908	−0.018162	4.261086	0.038995	0.698001	0.496135	0.982002
$X_4^2$	0.423496	0.166683	0.096803	0.75019	6.455278	0.011062	1.527292	1.101643	2.117402

Source: [own study].

The p-values determined in the Wald test for all variables (shown in column 7 in Table 4) are less than 0.05, which means that all explanatory variables are significant for this model.

Next, a covariance matrix was calculated to fully consider the odds ratio in the model. The matrix shows that the variances for individual variables have negative signs (see Table 5).

Table 5. Covariance matrix for Model 2.

	Y2	${\mathit{X}}_1^2$	${\mathit{X}}_2^2$	${\mathit{X}}_3^2$	${\mathit{X}}_4^2$
Y2	0.323116	−0.042463	−0.02799	0.015885	−0.038567
$X_1^2$	−0.042463	0.023413	−0.002729	−0.009561	0.003639
$X_2^2$	−0.02799	−0.002729	0.018824	−0.00333	−0.002009
$X_3^2$	0.015885	−0.009561	−0.00333	0.030336	−0.020661
$X_4^2$	−0.038567	0.003639	−0.002009	−0.020661	0.027783

Source: [own study].

Table 5. Covariance matrix for Model 2.

	Y2	${\mathit{X}}_1^2$	${\mathit{X}}_2^2$	${\mathit{X}}_3^2$	${\mathit{X}}_4^2$
Y2	0.323116	−0.042463	−0.02799	0.015885	−0.038567
$X_1^2$	−0.042463	0.023413	−0.002729	−0.009561	0.003639
$X_2^2$	−0.02799	−0.002729	0.018824	−0.00333	−0.002009
$X_3^2$	0.015885	−0.009561	−0.00333	0.030336	−0.020661
$X_4^2$	−0.038567	0.003639	−0.002009	−0.020661	0.027783

Source: [own study].

Hence, the chance of occurrence of a high level of neophobia depends on the variables mentioned above, as described by the Odds Quotient:

• Variable $X_1^2$—consumption of products enriched with omega-6 and omega-3 fatty acids: OR [95%CI] = 1.64 [1.21; 2.21]. Thus, the less we consume products enriched with omega-6 and omega-3 fatty acids, the greater the chance of a high level of neophobia.
• Variable $X_2^2$—consumption of products containing probiotic bacteria: OR [95%CI] = 1.50 [1.15; 1.97]. Thus, the less we consume products containing probiotic bacteria, the greater the chance of a high level of neophobia.
• Variable $X_3^2$—consumption of products with reduced sugar content OR [95%CI] = 0.69 [0.49; 0.98]. Thus, the less we consume products with reduced sugar content, the lower the chance of a high level of neophobia.
• Variable $X_4^2$—consumption of products with reduced sodium content OR [95%CI] = 1.53 [1.10; 2.12]. Thus, the less we consume products with reduced sodium content, the greater the chance of a high level of neophobia.

The parameters for Model 3 are introduced in Table 6.

Table 6. Numerical results for Model 3 parameters.

Term	Est. Value βi	Std. Dev. βi	−95% CI	+95% CI	Wald Stat.	p-Value	OR	−95% CI	+95% CI
Free	−2.17788	0.703131	−3.555991	−0.799769	9.593908	0.001952	0.113281	0.028553	0.449433
$X_1^3$	0.64867	0.215413	0.226468	1.070872	9.067837	0.002601	1.912995	1.254163	2.917922
$X_2^3$	−0.490427	0.160831	−0.805651	−0.175203	9.298384	0.002294	0.612365	0.446797	0.839286
$X_3^3$	0.554767	0.179564	0.202828	0.906707	9.545121	0.002005	1.741536	1.224861	2.476156

Source: [own study].

Table 6. Numerical results for Model 3 parameters.

Term	Est. Value βi	Std. Dev. βi	−95% CI	+95% CI	Wald Stat.	p-Value	OR	−95% CI	+95% CI
Free	−2.17788	0.703131	−3.555991	−0.799769	9.593908	0.001952	0.113281	0.028553	0.449433
$X_1^3$	0.64867	0.215413	0.226468	1.070872	9.067837	0.002601	1.912995	1.254163	2.917922
$X_2^3$	−0.490427	0.160831	−0.805651	−0.175203	9.298384	0.002294	0.612365	0.446797	0.839286
$X_3^3$	0.554767	0.179564	0.202828	0.906707	9.545121	0.002005	1.741536	1.224861	2.476156

Source: [own study].

The p-value determined in the Wald test for every variable (shown in column 7 in Table 6) is less than 0.05, which means that all explanatory variables are significant for this model.

Next, a covariance matrix was calculated to fully consider the odds ratio in the model. The matrix shows that the variances for individual variables have negative signs (see Table 7).

Table 7. The covariance matrix for Model 3.

	Y3	${\mathit{X}}_1^3$	${\mathit{X}}_2^3$	${\mathit{X}}_3^3$
Y3	0.494393	−0.085593	−0.030864	−0.035558
$X_1^3$	−0.085593	0.046403	−0.012669	−0.013629
$X_2^3$	−0.030864	−0.012669	0.025867	0.000349
$X_3^3$	−0.035558	−0.013629	0.000349	0.032243

Source: [own study].

Table 7. The covariance matrix for Model 3.

	Y3	${\mathit{X}}_1^3$	${\mathit{X}}_2^3$	${\mathit{X}}_3^3$
Y3	0.494393	−0.085593	−0.030864	−0.035558
$X_1^3$	−0.085593	0.046403	−0.012669	−0.013629
$X_2^3$	−0.030864	−0.012669	0.025867	0.000349
$X_3^3$	−0.035558	−0.013629	0.000349	0.032243

Source: [own study].

Hence, the chance of negative attitudes depends on the variables mentioned above, as described by the Odds Quotient:

• Variable $X_1^3$—constantly trying new and varied foods: OR [95%CI] = 1.91 [1.25; 2.92]. Thus, the less we constantly try new and varied foods, the greater the chance that our attitude will be negative.
• Variable $X_2^3$—I eat “almost everything”: OR [95%CI] = 0.61 [0.45; 0.84]. Thus, the less likely we are to eat “almost everything”, the lower the chance that our attitude will be negative.
• Variable $X_3^3$—trying new recipes when preparing dishes: OR [95%CI] = 1.74 [1.22; 2.48]. Thus, the less often we try new recipes, the greater the chance of a negative nutritional attitude.

The parameters for Model 4 are introduced in Table 8.

Table 8. Numerical results for Model 4 parameters.

Term	Est. Value βi	Std. Dev. βi	−95% CI	+95% CI	Wald Stat.	p-Value	OR	−95% CI	+95% CI
Free	−4.810875	0.757113	−6.29479	−3.32696	40.376269	<0.000001	0.008141	0.001846	0.035902
$X_1^4$	0.518452	0.14228	0.239588	0.797315	13.277909	0.000269	1.679425	1.270726	2.219573
$X_2^4$	0.628132	0.149757	0.334613	0.921651	17.592443	0.000027	1.874106	1.3974	2.513436
$X_3^4$	0.423514	0.147358	0.134698	0.71233	8.26016	0.004052	1.527319	1.144191	2.038736

Source: [own study].

Table 8. Numerical results for Model 4 parameters.

Term	Est. Value βi	Std. Dev. βi	−95% CI	+95% CI	Wald Stat.	p-Value	OR	−95% CI	+95% CI
Free	−4.810875	0.757113	−6.29479	−3.32696	40.376269	<0.000001	0.008141	0.001846	0.035902
$X_1^4$	0.518452	0.14228	0.239588	0.797315	13.277909	0.000269	1.679425	1.270726	2.219573
$X_2^4$	0.628132	0.149757	0.334613	0.921651	17.592443	0.000027	1.874106	1.3974	2.513436
$X_3^4$	0.423514	0.147358	0.134698	0.71233	8.26016	0.004052	1.527319	1.144191	2.038736

Source: [own study].

The p-value determined in the Wald test for every variable (shown in column 7 in Table 8) is less than 0.05, which means that all explanatory variables are significant for this model.

Next, a covariance matrix was calculated to consider the odds ratio in the model fully. The matrix shows that the variances for individual variables have negative signs (see Table 9).

Table 9. Covariance matrix for Model 4.

	Y4	${\mathit{X}}_1^4$	${\mathit{X}}_2^4$	${\mathit{X}}_3^4$
Y4	0.573221	−0.049634	−0.061699	−0.059833
$X_1^4$	−0.049634	0.020244	0.001617	−0.004536
$X_2^4$	−0.061699	0.001617	0.022427	−0.000404
$X_3^4$	−0.059833	−0.004536	−0.000404	0.021714

Source: [own study].

Table 9. Covariance matrix for Model 4.

	Y4	${\mathit{X}}_1^4$	${\mathit{X}}_2^4$	${\mathit{X}}_3^4$
Y4	0.573221	−0.049634	−0.061699	−0.059833
$X_1^4$	−0.049634	0.020244	0.001617	−0.004536
$X_2^4$	−0.061699	0.001617	0.022427	−0.000404
$X_3^4$	−0.059833	−0.004536	−0.000404	0.021714

Source: [own study].

Hence, the chance of negative attitudes depends on the variables mentioned above, as described by the Odds Quotient:

• Variable $X_1^4$—consumption of food that lowers cholesterol: OR [95%CI] = 1.68 [1.27;2.22]. Thus, the less we consume products that lower cholesterol, the greater the chance our attitude will be negative.
• Variable $X_2^4$—consumption of products with reduced sugar content: OR [95%CI] = 1.87 [1.39;2.51]. Thus, the less we consume products with reduced sugar content, the greater the chance our attitude will be negative.
• Variable $X_3^4$—consumption of anti-ageing products: OR [95%CI] = 1.53 [1.14;2.04]. Thus, the less we consume anti-ageing products, the greater the chance of a negative nutritional attitude.

The next step is to determine the quality of the built models.

Table 10. Statistical values for particular criteria.

Criterion	Model 1	Model 2	Model 3	Model 4
AIC	194.493235	169.513345	178.974069	116.017673
AICc	194.652175	169.783616	179.131963	116.178747
BIC	203.62351	181.635097	188.123637	125.108987
Pseudo R2	0.122753	0.238491	0.200164	0.481285
R2 (Nagelkerke)	0.208638	0.375357	0.323085	0.649139
R2 (Cox–Snell)	0.156476	0.281514	0.242314	0.486847

Source: [own study].

presents the statistical values of particular criteria for the logit regression model.

Table 10 lists the values of the criteria used to determine the quality of the models. In the case of all models (1–4), it should be clearly stated that the quality of their fit is not high. The Pseudo-R², R² (Nagelkerke), and R² (Cox–Snell) values confirm this.

The next step should be to check the statistical significance of all variables in Models 1–4.

Table 11. Statistics of the likelihood ratio tests for Models 1–4.

Likelihood Ratio Test	Model 1	Model 2	Model 3	Model 4
Chi-square statistic	26.375939	50.583156	43.287852	102.078828
Degrees of freedom	3	4	3	3
p-value	0.000008	<0.000001	<0.000001	<0.000001

Source: [own study].

contains the basic statistics for the likelihood ratio test.

Based on the p-values from the likelihood ratio test (Table 11), it can be concluded that all models (1–4) are statistically significant; hence, all independent variables in the models are statistically significant (

Table 12. The values of statistics for the Hosmer–Lemeshow test.

Hosmer–Lemeshow Test	Model 1	Model 2	Model 3	Model 4
Chi-square statistic	11.136135	7.54777	7.801891	4.944209
Degrees of freedom	6	8	8	8
p-value	0.084259	0.478842	0.453058	0.763522

Source: [own study].

).

When building logistic regression models, it is also possible to determine whether they are sufficient for forecasting.

Table 13. Numerical data for ROC curves—Models 1–4.

ROC (DeLong’s Method)	Model 1	Model 2	Model 3	Model 4
AUC (area under the curve)	0.73643	0.817157	0.794297	0.918746
SE(AUC)	0.040447	0.034232	0.036715	0.020734
−95% CI	0.657156	0.750063	0.722336	0.878107
+95% CI	0.815705	0.88425	0.866257	0.959384
Z statistic	5.081875	6.77261	6.346299	8.941962
p-value	<0.000001	<0.000001	<0.000001	<0.000001
Cut-off line	0.580507	0.469964	0.450886	0.502158

Source: [own study].

presents the empirical data to build and interpret the ROC curves for all models (1–4).

When we consider the area under the ROC curve, indicated by the AUC parameter, we can see that all models are more significant (p < 0.000001) than assumed in the 0.5 regression (see Table 13, rows 1 and 6).

The ROC curves for all models are presented in Figure 1.

Furthermore, the classification parameters are given in

Table 14. Classification parameters.

Classification	Model 1	Model 2	Model 3	Model 4
Cut-off point	0.5	0.5	0.5	0.5
% correct	69.677%	72.549%	73.077%	84.967%
Sensitivity (% correct 1)	67.949%	70.13%	70.513%	84.211%
Specificity (% correct 0)	71.429%	75.00%	75.641%	85.714%

Source: [own study].

.

According to Figure 1 and the values shown in Table 14, we can make the following remarks about the usefulness of Models 1–4 for predicting neophobia and negative attitudes towards functional foods:

• Model 1: The classification determined for a cut-off of 0.5 points reveals 69.67% of the occurrence of positive instances. The need to classify the “high level of neophobia” value is 67.949% (sensitivity), and the “low level of neophobia” value is 71.429% (specificity).
• Model 2: The classification determined for a cut-off of 0.5 points reveals 72.549% of the occurrence of positive instances. The need to classify the “high level of neophobia” value is 70.13% (sensitivity), and the “low level of neophobia” value is 75.00% (specificity).
• Model 3: The classification determined for a cut-off of 0.5 points reveals 73.077% of the occurrence of positive instances. The need to classify the “negative attitudes” value is 70.513% (sensitivity), and the “positive attitudes” value is 75.641% (specificity).
• Model 4: The classification determined for a cut-off of 0.5 points reveals 84.967% of the occurrence of positive instances. The need to classify the “negative attitudes” value is 84.211% (sensitivity), and the “positive attitudes” value is 85.714% (specificity).

Based on the results obtained, it can be concluded that the classification ability of the models is satisfactory. Hence, we can calculate the dependent variable’s predicted value for any given explanatory variable values.

4. Discussion

According to the interpretation of Models 1 and 2, the most significant impact on the occurrence of a high level of food neophobia among consumers aged 60+ is their attention to the health benefits of food. This means that people who ignore this aspect will be 2.33 times more likely to show a high level of neophobia. In turn, people who do not care about a diet rich in vitamins and minerals are 1.78 times more likely to show a high level of neophobia than the rest of the population in this consumer group. This fact can be easily combined with the interpretation of the variables in Model 2. People consuming fewer products enriched with omega-3 and -6 fatty acids will be 1.64 times more likely to display neophobia. Similarly, consumers aged 60+ who do not care about consuming products containing probiotic bacteria and approximately 1.5 less sodium content are more likely to have a high level of neophobia.

However, reducing the daily diet’s inclusion of fresh and unprocessed products and products with reduced sugar content reduces the chance of a high level of neophobia by approximately 0.6 times.

These models are consistent with general trends outlined in the consumer behaviour literature [56,57,58,59,60,62]. The factor that is most important when choosing food products and determining the level of quality in the consumer assessment is the product’s sensory attractiveness [59,60,77]. This influences the positive perception of the product and causes consumers to look for products with specific tastes [78]. Taste is an important determinant of food choice across age groups and cultures [54,55,56,63,79,80,81]. This is not surprising because food’s aroma, taste, texture, and appearance provide pleasure and satisfaction, providing a rich and varied sensory experience [82].

An essential ingredient that consumers pay special attention to when choosing food products is their nutritional value. This term should be understood as the ability of a given product to provide nutrients that are building materials (carbohydrates, proteins, fats, vitamins, and mineral salts) [56,57,58,59,60,83].

On this basis, it should be emphasised that both logistic models for neophobia should work as tools supporting marketing activities related to promoting healthy food. Thanks to these activities, there is a chance to reduce neophobia in the 60+ consumer group.

We will discuss Models 3 and 4 relating to negative attitudes towards functional viability. According to the definition of Functional Food Science in Europe, developed in 1999, functional food has been proven to benefit one or more body functions beyond its nutritional effect. This impact involves improving health and well-being and reducing the risk of disease. Functional foods must resemble conventional foods and demonstrate beneficial effects in amounts expected to be typically consumed in the diet; they are not tablets or capsules but part of a proper diet [50,51,52,53,56,84,85].

With this in mind, it should be stated that people who try new and varied foods less frequently are almost twice as likely to show a negative attitude in the 60+ group of consumers. We will see a comparable increase in the chance (1.74 times) of a negative attitude in those who use new recipes less often when preparing daily meals. In turn, people with restrictions in their food preferences, i.e., those who do not eat everything, have a 0.6 times lower chance of showing a negative attitude towards functional (healthy) food.

The paradox of Model 4’s results is that consumers aged 60+ who consume products with reduced sugar content less frequently have a 1.87 times greater chance of having a negative attitude. Similarly, people who less often reach for cholesterol-lowering or anti-ageing products have a higher chance of having negative attitudes towards functional foods—1.68 times and 1.53 times, respectively. This is presumably because they do not care about consuming a healthy, varied diet. This consumption pattern suggests that senior consumers sometimes cannot connect healthy food with the concept of functional food. For them, eating cholesterol-lowering or low-sugar foods may be an everyday and typical behaviour. Therefore, they may still have negative attitudes towards functional foods. The second possibility is that there is no relationship between healthy food and the concept of functional food in this consumer group.

The main results of these models confirm the following theses:

-

the neophobia as the attitude impacts both consumer food and eating habits, and consumer awareness about the composition (Models 1 and 2);

-

the attitudes towards functional food relates to everyday eating habits and consumer awareness regarding health-promoting food ingredients described by frequency of consumption (Models 3 and 4);

-

the level of neophobia relates to negative attitudes towards functional food (Models 2 and 4).

Furthermore, Models 3 and 4 show the relationship between attitudes towards functional foods and behaviours towards healthy foods.

Classifying the main objects and issues of concern in the models and evaluating their importance and influence is done in accordance with the AA1000SES method [86].

5. Conclusions

This article proposes four analytical and forecasting models to assess the high level of food neophobia and negative attitudes towards functional foods in the 60+ consumer group. For this purpose, logistic regression was used, which is used to determine the relationship between several (more than two) independent descriptive variables influencing the dependent variable.

The models developed based on the results of 510 survey responses are characterised by good quality and fit and can be used to predict the behaviour and attitudes of consumers aged 60+ towards functional foods.

These models, as shown in the Discussion, can support those responsible for marketing strategies for functional food products. According to the results in this article, one way may be to increase awareness of healthy eating and promote a positive attitude towards it. As a result, in the group of consumers aged 60+, this should reduce the chance of high levels of neophobia and negative attitudes towards functional foods.

Thus, the purpose of this study has been achieved. These models can make it easier to manage products and their marketing campaigns. This is closely linked to the UN Agenda 2030 Sustainable Development Goals 2 and 3.

In the near future, as part of continuing research on the attitudes and behaviours of consumers aged 60+, a multidimensional model will be developed that takes into account the respondents’ financial status, gender, and attitude towards healthy eating habits.