Profile

We sought to identify the clients who consume street food in Jardim da Penha Square. Specifically, information regarding gender, the frequency of purchase of street food in the square, age group, place of residence, and household income of the respondents.

In relation to the gender, females were more represented (59 percent) in the samples, as shown in Figure 1. As for the variable age group, respondents between 24 and 29 years are the largest group (19 percent), as depicted in Figure 2. According to the data it is possible to realize that young people and adults up to 35 years of age are those with the highest representation in food consumption, in Jardim da Penha Square.

Source: Research data

Figure 1 Gender 

Source: Research data

Figure 2 Age group 

When asked about the frequency of purchase of street food in the square, a large number of respondents claimed to consume it every week (41.5 percent). Thus, from these results we can say that the majority of respondents have the habit of regularly buying street food in the square, as depicted in Figure 3.

Source: Research data

Figure 3 Frequency of purchase 

To identify the social class of respondents, we asked about the household income, i.e. the combine incomes of all people living at home. The income ranges used in this work were based on the criteria of economic classification of ABEP (Brazilian Marketing Research Association), in which the strata of Brazil Criteria represent the average monthly income for the economic strata. This stratifycation is not based on the value of the current minimum wage. As of January 2015, income strata adopted are described in Figure 4. It is important to note that, in order to facilitate the application of the questionnaire and simplify the reading of the numbers, the digits that represent the cents have been rounded up.

Source: Research data

Figure 4 Household income 

By analyzing Figure 4, it can be noticed that most of the people who participated in this study belong to the economic groups B1 (32 percent) and B2 (32 percent). Thus, we can affirm that consumers of Jardim da Penha Square are mostly people from middle and upper middle classes. This result does not support FAO's report (2011), which states that most people who consume street food have low purchasing power.

Source: Research data

Figure 5 Place of residence 

In relation to the place of residence, 126 people (63 percent) stated that they live in the region of the Jardim da Penha, as shown in Figure 5. This figure clearly indicates that the square is frequented mostly by people from the neighborhood.

Importance of the Attributes in the Purchase Decision

In the second part of the questionnaire, customers were questioned on the importance of attributes in the purchase decision when buying street food, since that in the process leading up to the purchase, as expressed by ^{Blackwell, Miniard and Engel (2005}), consumers evaluate the different possibilities available to make their purchasing decision, in order to best satisfy their demands.

However, because this is an exploratory study, we developed a scale to measure the importance of these attributes, which made necessary the use of exploratory factor analysis (EFA). According to Brown (2006) EFA is generally conducted when the researcher does not have an underlying theory or sufficient empirical evidence which clearly indicates how items of a given instrument must be grouped and evaluated.

Exploratory Factor Analysis (EFA) of the Importance of the Attributes

Considering that the attributes identified in the qualitative stage formed an 18-items scale, it was relevant to verify the distribution of these data. To this end, we used the statistical procedure of exploratory factor analysis (EFA). Factor analysis helps in the evidences relating to the validity of questionnaires, commonly used in research (^{Hair et al., 2005}).

The EFA was held separately in each of the dimensions (categories) in order to assist in the evaluation of the distribution of attributes in the dimensions, thus the distribution and organization of the scale reflected the theory of core attributes and EFA.

The analytical technique employed was the exploratory perspective as it is "[...] useful in searching for structure among a set of variables or as a data reduction method [...]" (^{Hair et al., 2005}, p. 92). Normally the exploratory factor analysis (EFA) is used in initial research with the purpose of exploring the data.

For the development of this analysis we used oblique rotation, identifying the correlation degree of each factor (^{Hair et al., 2005}). Factor rotation was performed aiming factor adjustment in search of simple and more significant solutions, to improve the interpretation and reduce the ambiguity (^{Hair et al., 2005}). The factor analysis was performed using SPSS Software. Next we present the criteria and results of the EFA.

Exploratory Factor Analysis Results

The KMO and Bartlett tests were carried out initially. These tests verify if the factor analysis is appropriate. The test of Kaiser-Meyer-Olklin (KMO) varies between 0 and 1. The closer to 1 the better, with acceptable values between 0.5 and 1 (^{Hair et al., 2005}). Values below 0.5 indicate inadequacy (^{Hair et al., 2005}). For Bartlett's test of sphericity (BTS) the acceptable values are less than 0.05 (p < 0.05). As depicted in Table 1, all values are within the limits indicated.

The number of factors to be extracted was determined by the latent root criterion, in which the eigenvalues greater than 1 should be retained, always trying to reach the recommended minimum of 60 percent of accumulated variance, indicated by ^{Hair et al. (2005}). To perform this procedure, the attributes of the dimension "food" were distributed in three components to a higher cumulative variance. However the variable "healthy food" would form a single construct.

This gave rise to a theoretical reflection on this variable. Despite the studies by ^{Liu and Jang (2009}); ^{Harrington, Ottenbacher and Way (2011}) and Castelo Branco (2000) that highlighted only the variable healthy food, other authors such as Steptoe, Pollard and Wardle (1995) and Yuksel and Yukesel (2002) provide evidence that the variable "healthy food" can be composed of other factors, considering, for example, the caloric content of food and the amount of vitamins, nutrients, proteins and fat. Thus, it appears that this variable is not complete in the way it was measured, so we decided to remove this variable. Two components were then formed and renamed as shown in Table 1.

After the EFA, the remaining dimensions remained with the same structure of the scale originally used, as depicted in Table 2.

The communalities represent the total amount of variance that an original variable shares with all other included in the analysis. ^{Hair et al. (2005}) suggest that the acceptable levels of communality must be greater than or equal to 0.5. Values below that should be evaluated and it is up to the researcher the decision to keep the variable or not (Hair et al., 2005). You can verify that the variable "smell" and "safety" present a communality value below the critical level (0.371), however we decided to keep this variable given the theoretical support.

The values of factor loadings of the components are shown in Table 2. ^{Hair et al. (2005}) explain that factor loading is the correlation between the variable and the factor. It represents the acceptable limit of contribution of the variable to the factor in order to avoid the problem of the indeterminacy of the relationship between variables and factors. For statistical significance of a sample of 200, as in the case of this study, the loads must be equal or greater than 0.4 (^{Hair et al., 2005}).

Reliability was tested with Cronbach's Alpha coefficient, which constitutes one of the most widely adopted procedures to assess the reliability of a scale (Malhotra, 2001). According to ^{Hair et al. (2005}), the acceptable values of Cronbach's alpha should vary between 0.6 and 0.9.

The value of Cronbach's alpha is influenced both by the value of the correlations of the items, and the number of items. Thus, factors with few items tend to have smaller Cronbach's alpha (^{Cortina, 1993}). The constructs "convenience" and "price" have Cronbach's alpha values (see Table 3) below the acceptable values indicated by the literature.

From the development of exploratory factor analysis, it was possible to infer some modifications in relation to the attributes and constructs initially used in this research, so that the original scale suffered adjustments for better reliability and validity to the questionnaire, as noted in Table 4.

Factor analysis led to changes in the scale shown in Table 5. This new structure was considered for the analysis of the importance of attributes presented in the next topic.

Analysis of the importance of the attributes

The guiding hypothesis of this study is based on the theory of ^{Alpert (1971}), which notes that the attributes can have different degrees of importance. To answer the problem of this research, the hypothesis was based on identifying whether there is a difference in degree of importance assigned by the customers to the attributes.

Therefore, we developed hypotheses to test the similarities and differences between the averages, in order to prove or disprove the theory of ^{Alpert (1971}). The first hypothesis is as follows:

Ha: There is no difference in the degree of importance assigned by customers, to attributes.

This test considered a confidence interval of 95 percent, with the significance level (α) of 0.05. According to ^{Levine et al. (2012}), if the p-value falls below the predetermined significance level (α = 0.05) H0 is not confirmed, which shows that at least one of the averages is different from the other. When the p-value is above the previously established significance level (α = 0.05) H0 is not rejected.

To test the hypothesis (Ha), we used Excel tool and conducted the tests for the analysis of variance (ANOVA). The variance measures how much is the mean variation around the arithmetic mean of attributes, in order to check if there are statistical difference between these averages (^{Levine et al., 2012}). The results are presented in Table 1.

The analysis the p-value (4.609E-11) showed that it is less than 0.05. In common numbers that scientific notation is equivalent to approximately 0 (zero). This indicates that H0 is rejected, i.e. at least one of the means is different from the others. This information assumes that at least one of the constructs exerts different influence from the others on the buying decision, then there are different degrees of importance, assigned by the customers to the constructs. In this way, the theory of ^{Alpert (1971}) is verified as to the existence of differences in the degree of importance assigned by customers to the attributes.

Classification of the Attributes

It is also necessary to know which constructs have different means and put them in descending order of importance for, if possible, sort them in determinant, important and salient. Only constructs that presented statistically equal averages were classified in the same way.

In order to identify which construct features distinct mean from the others, we conducted new tests of hypotheses involving two constructs to each test. For this new test of hypotheses we used t test for the difference between two means assuming different variances. This method is used to evaluate the statistical differences between two arithmetic means (^{Levine et al., 2012}). The test was performed with Excel tool. The hypotheses for this test are presented in Table 6, as well as the p-value obtained in each one of these tests and the hypotheses results.

To classify the degree of importance, those constructs with averages no different between themselves were considered within the same classification. Thus, following the hierarchy, for purposes of classification, the construct "convenience" was considered along with "price", but separated from "environment", because "price" and "environment" did not presented equal means (see Table 6), which demonstrates the existence of a gray area, depicted in Figure 6. This shows that there may be other possible forms of classifications. This work followed the theory of ^{Alpert (1971}), thus the classification was limited to a maximum of three alternatives, regarding the importance of the constructs. It was obtained the classification shown in Figure 6.

Source: The authors

Figure 6 Classification of the importance of the constructs, with the gray area highlighted 

Responding to the research problem, it can be observed that the determinant construct is service. This construct is formed by the attributes: fast service, good service, appearance and personal hygiene of employees. The important constructs are: organoleptic characteristics, formed by: presentation, taste, smell, temperature. Food, is formed by variety and hygiene. And environment comprised of: physical structure, familiar atmosphere, open environment and security. The salient constructs are convenience, consisting of: convenience and location, and price, formed by value and ease of payment.

These results will meet those revealed by ^{Mhlanga, Hattingh and Moolman (2014}), in which these authors showed that the level of service is the main factor considered by customers in the context of restaurants. According to the same study, the organoleptic characteristics are essential to provide good food and they were also highlighted as important and come right after service.

Regarding services, ^{Sanches and Salay (2011}) reveal that the personal hygiene of the employees is the most important factor when it comes to choosing a restaurant to have meals away from home.

The study of ^{Namkung and Jang (2007}) showed that the presentation, taste and temperature (organoleptic characteristics) were the most significant for customer satisfaction in restaurants. Similarly ^{Liu and Jang (2009}) revealed that the taste of food is essential to influence purchase decisions. The service offered appears next. In the study of ^{Park (2004}) the taste of food was also considered the most important, followed by hygiene and service.

^{Soriano (2002}) obtained similar results, in which the variety and presentation of the food were the most important influence on customers' decisions to return to a restaurant for another meal. The service was the second most important factor.

Yuksel and Yuksel (2002) revealed that, although important, the environment appears as the last relevant factor influencing the selection and evaluation of restaurants. Similarly, ^{Soriano (2002}) obtained that environment was the least important factor considered in the influence of customers' decisions to return to a restaurant.

^{Harrington, Ottenbacher and Way (2011}) showed that price and distance (location), although important, are not among the first factors in the ranking of importance. The results of ^{Park (2004}) and ^{Sanches and Salay (2011}) also highlight this fact.

The factor that stands out for being classified in different ways is price. In some studies, like ^{Abdullah, Abdurahman and Hamali (2013}), and ^{Ehsan's (2012}), the price was regarded as the most important dimension of customer preference in the food sector to select restaurants. In the studies of Siebeneichle et al. (2007) and ^{Souki (2010}), the price and ease of payment appear as the last factor in the order of importance. Thus, although important in the decision to purchase, there are other factors, such as high performance, that make the consumer less sensitive to price.

The results of this study identified that price is an important variable for the purchasing decision, however it is not the most important. Thus, consistent with the results of ^{Pelissari et al. (2011}), in which the authors claim that, although the price is an influential variable in purchasing decisions, it is not necessarily determinant for customers. However, it is important to highlight that price acts directly as a factor of competitiveness. Similarly, Reis (2004) posits that in face of a market in which products and services are becoming more alike, customers give importance to other factors besides price.

This study confirms the theory of ^{Griffin (1998}), in which he states that consumers are becoming more interested in customer service when buying something, representing the service factor in this study. Thus, the importance of service surpasses price and product factors (^{Griffin, 1998}).