Introduction

Maize (Zea mays) tortilla is considered the principal food of Mexican people. This product provides energy because of its high carbohydrate content; in addition, it contributes calcium, potassium, phosphorus, fiber, proteins and some vitamins like thiamine, riboflavin and niacin (^{México Produce, 2012}).

Maize is a commodity with high influence in international markets. In 2011, approximately 945 million tons of this cereal were produced in the world, so it surpassed other grains like wheat, sorghum, barley, oats and rye. The main producers are the US (31.0 %), China (16.0 %), the European Union (6.0 %), Brazil (5.0 %) and México (2.0 %). In 2011, world exports were around 96.7 million tons and the main exporting countries were the US, Argentina, Brazil, China and Ukraine, which exported approximately 88.0 % of the volume sold. Imports of maize in the same year were 93.5 million tons, and the main importing countries were Japan (15.5 mill/t), México (10.5 mill/t) and South Korea (7.5 mill/t) (^{USDA, 2012}).

Maize is the most important crop in México, because around 3.1 million producers cultivate this cereal and it represents 12.5 % of the Agricultural Gross Domestic Product (A GDP). During the 2002 to 2010 period, 21.59 mill/t were produced annually in average, and the principal producing states were Sinaloa (28.1 %), Jalisco (19.6 %), Estado de México, Chiapas and Michoacán; these last three generated approximately 29.8 % of the production (^{SIAP-SAGARPA, 2010}).

México produces and consumes mainly white maize; around 22 million tons are produced of this product, of which 12 million are destined to commercial human consumption (flour industry and boiled maize (nixtamalizado) dough, primarily), close to six million tons from the production are used for the families' auto-consumption, two million are used by the livestock production sector; the rest is distributed between seeds, losses, inventories and exports (^{SE, 2012}).

The tortilla value chain in México is made up of several links, among which there are maize producers, stockpilers, distributers, the industry, and the final consumer. The maize flour industry in México is concentrated in a few companies, with the following standing out: Grupo Industrial MASECA (71.2 %), MINSA (23.54 %), Harimasa (1.4 %), Cargill de México (1.3 %), Molinos Anáhuac (1.1o%) and Productos Manuel José (0.2 %). The dough and tortilla industry is dispersed throughout the country with 85 111 establishments that include mills and tortilla shops or both modalities; 94.0 % of these establishments give jobs to between one and five employees. Of the economic units for dough or tortilla production, 53 % are found in Estado de México, Distrito Federal, Jalisco, Veracruz, Michoacán, Guanajuato and Puebla; 51.0 % of production and employment are generated in these states (^{SE, 2012}).

It has been estimated that the daily consumption of tortilla per person in rural zones is 217.9 g (with an approximate weight of 27 g per tortilla, resulting in 8 tortillas per day), while in urban zones this indicator is 155.4 g (six tortillas per day). Thus, in 2010 the tortilla consumption in the country was calculated to be 6.9 million tons, with a value of 72,481 million pesos (^{CONEVAL, 2010}). Maize consumption is higher in the rural population than in the urban, and it is higher in the masculine population than in the feminine (INE/SEGENLAN/CEDESCA 1991, cited by ^{Bressani, 2008}). Consumption increases with age and is associated to income; there is higher consumption when there is lower income (Berssani R. 2008, cited by ^{Rodríguez G. M. et al. 2008}).

The Metropolitan Zone of Valle de México includes some municipalities in Estado de México: Ecatepec, Atizapán de Zaragoza, Cuautitlán, Nezahualcóyotl, La Paz, etc.; in addition to Distrito Federal, it is the most populated region of the country (20.12 million inhabitants). Of this total population, 55.5 % belongs to the urban municipalities of Estado de México and the remaining 44.5 % to Distrito Federal, but in addition, Estado de México as a whole houses more than 15.17 million people (13.50o% of the national population), of which 87 % live un urban zones (^{INEGI, 2010}). Because of this, and due to the importance of tortilla consumption in the country's population, it was considered necessary to perform a study that will characterize the consumers of this food in aspects such as the type of product that they demand and its characteristics, correlating it with income and consumption. It was decided to carry out the study in Estado de México because it is the state that concentrates the highest population of the country (13.5 %), mostly urban. The working hypothesis is that consumers with medium income and medium consumption are the ones who determine the attributes of the tortilla (type of maize with which it is made, color, consistency, etc.).

Materials and Methods

The study of tortilla consumers in Valle de México included 19 municipalities of Estado de México: Ecatepec de Morelos, Nezahualcóyotl, Naucalpan de Juárez, Toluca, Tlalnepantla, Chimalhuacán, Tultitlán, Cuautitlán Izcalli, Atizapán de Zaragoza, Ixtapaluca, Nicolás Romero, Tecámac, Valle de Chalco Solidaridad, Chalco, Coacalco de Berriozábal, La Paz, Huixquilucan, Texcoco and Metepec. In the study, the multivariate analysis method called CHAID algorithm (Chi-squared Automatic Interaction Detection) and the ordinal regression model were used. The first is an analysis technique of variable segmentation (explained and explicative), which allows establishing relations of dependence between them (^{Escobar, 1992}); the second is a model designed for the analysis of response variables measured in an ordinal scale (^{Rodríguez, 2007}).

According to ^{Binger and Hoffman (1988}), in the theory of demand an individual is considered who maximizes his/her welfare subject to a function of utility U(X); this function is restricted by the budget: M=P’X where: U(X): Welfare function, M: Total income, X: Vector of goods to be assigned, P: Price vector.

The corresponding Lagrangian is:

The first order condition for the i-th good is:

The solution is a demand in the sense of Marshall:

As is noted, in this function the prices and the income or budget are the relevant variables in the demand. However, under the first order conditions between the marginal valuation of the individual, this valuation can change depending on some of its characteristics, such as: demographic profile, age, gender, schooling, number of family members, among others. These characteristics can influence the final amount demanded and they can be captured through ordinal regression, as is the case of this study where, in addition to the price and the income, some of the characteristics mentioned are considered.

Because the dependent variable is an ordinal: low, medium and high consumption, and this is a function of the explicative variables, such as price, income, type of tortilla and place of purchase, it was decided to apply an ordinal regression model which in its estimation takes into account the direction towards which the probabilities of the variables considered move.

To analyze the variables, the SPSS statistical package (Statistical Package for Social Sciences) was used, which includes the algorithm to perform the Chi-square tests, as well as the ordinal regression model (^{Castañeda, 2010}).

To obtain the information of Estado de México consumers, a semi-structured survey was designed with quantitative and categorical variables of economic and social nature. To carry out the calculation of the sample, the population selected was identified (older than 20 years) from the total in Estado de México (4 611 412 inhabitants, according to the National Population and Housing Survey by ^{INEGI, 2010}).

The size of the sample was determined with the following equation:

where: N=Total population of the universe of study, 4 611 412 inhabitants. n=Size of the sample. p=Proportion estimated of positive variability: 0.5. q=100-p (Negative variability). E=Error or estimation precision allowed (6 %). Z=Level of confidence: Z of tables =1.96.

Substituting:

The size of the sample determined to apply the survey to maize tortilla consumers was 267 interviews; however, a total of 269 were carried out. The criterion used to select individuals was their willingness to be interviewed, and it was attempted to interview adults, mostly housewives, who know what the type and characteristics are of the tortilla that the household consumes. The 269 interviews were carried out in the municipalities of higher to lower population, considering the population density. The individuals were interviewed in tortilla shops, parks, traditional markets and shopping centers.

In the survey, the variables were concentrated in two blocks: 1) the first was made up of general data of the individual: name, place of origin, age, sex, schooling and occupation; 2) the second examined the frequency of consumption and purchase of tortilla, place of purchase, type (elaborated with maize flour or boiled maize (nixtamalizado)), price, maize preference for their elaboration (white, yellow and blue), preference for place of purchase, physical attributes (color, resistance, flexibility, etc.), and monthly family income.

The classification of income was performed according to the geographical zone of the country and the total monthly salary that is earned within the interviewee's family. For this purpose, they were asked about the level of total monthly family income, including all of the income from the people who contribute, not only the head of the household. The income categories were: low, of less than $4000.00; medium, of $4001.00 to $10 000.00; and high, of more than $10 000. A maximum income of $4000.00 monthly for the strata of consumers of low income is equivalent to a daily family income of $133.00, which is in the range of the lowest salaries in the region, such as those of domestic workers and day laborers, who receive between $100.00 and $200.00 daily; these workers are not hired with the indicative minimum wage of the zone ($62.33) (^{CONASAMI, 2011}). The level of consumption was classified from the variable "frequency of purchase" in the following manner: low, of less than one kilogram; medium, from one to three kilograms; high, of three kilograms.

The data obtained from the interviews were gathered during the period of November 2011 to May 2012, and they were captured in an Excel spreadsheet; later, with the use of the SPSS 20.0 package, the variables were analyzed through personalized contingency tables that allowed obtaining frequencies and performing hypothesis tests. Due to the categorical nature of the response variable (kilograms purchased, KCom), an ordinal regression model was estimated, performing adjustment tests of the model, of parallel lines, as well as other individual ones for the factors related (income, tortilla price, type of tortilla, preference of white maize consumption, and place of purchase). The calculation of the estimators was carried out with the statistical package mentioned before.

Results and Discussion

The sample of tortilla consumers included mostly women (64.31 %) of 21 to 40 years of age; concerning schooling, 33.46 % had studied up to secondary school, 26.33 % only primary and the rest 26.33 %, including those who studied high school and undergraduate studies. Of the sample evaluated, 56.5 % have a low income; 36.81 %, medium; and 6.69 %, high.

A high proportion of those interviewed (94.80%) stated that they consumed tortilla every day, with an average of eight tortillas daily, in families integrated in average by four members. They purchase the product daily at tortilla shops next to their house with an average price of $11.00 per kilogram.

Of the consumers, 78.44 % prefer that the product is wrapped in a cloth napkin with which they go to the establishment, and they prefer white tortilla (68.03 %) and yellow in a smaller proportion (24.54 %). Likewise, they prefer tortillas that are resistant, flexible and elaborated with boiled maize (nixtamalizado) dough, and that these are of recent elaboration (warm). They also prefer the medium size and appreciate more the attributes of flavor, aroma, texture and durability, and that they are elaborated in a traditional way (boiled maize, nixtamalizado, dough); the consumers care less that tortillas are mixed with other products like cactus (nopal), fiber or vitamins.

The analysis performed with the formation of response trees with relative frequencies shows variables like income and consumption in the upper part, which are related to those located in the initial area of the table, such as preference over type of tortilla, of the maize with which the tortillas are made, place of purchase, type of wrapping, or some physical attributes. An important assumption is that the variables located on both sides of the tables are random and the cross between the boxes shows their correlation.

The analysis of results is sustained also on the χ² tests, which, when obtaining a calculated χ_c ² and contrasting it with a χ_t ² from the tables, will express in their decision rule that if χ_c ²< χ_t ², then the null hypothesis (H₀) is accepted, about the distribution of the responses in the table boxes being uniform; contrary to if χ_c ²< χ_t ², then H₀ is rejected in favor of the alternative hypothesis (H_a) of the distribution in the boxes being different.

The χ² calculated for the income and consumption regarding the type of tortilla that the consumer purchases (elaborated with maize flour or boiled maize, nixtamalizado) are 17.366, 13.258 and 12.663; the probability of finding a χ_c ² less than χ_t ² is low (12.5916), so that the H₀ is rejected in favor of the H_a, which establishes that the distribution between the boxes is different. The correlations of the variables observed indicate that the low, medium and high income with medium consumption have a high correlation with the purchase of tortillas made with maize that is nixtamalizado (Table 1). ^{Soria (2000}) mentions that until the mid-1990s, 58 % of the tortillas sold in México were elaborated from fresh dough made of boiled maize (nixtamalizado). Meaning that tortilla consumers in Estado de México still maintain a preference for those elaborated in a traditional manner (nixtamalizado maize).

The statistical analysis of the χ² from the results indicates that for consumers with medium income and medium consumption with regards to the preference of the type of maize with which the tortillas are made is 52.359 for white maize, and 37.499 for yellow (considering white as the favorite), and the probability of finding an χ_t ² higher than χ_c ² is practically zero. The correlations for medium income with medium consumption are high for preference 5 (most favorite), in the case of tortilla made with white maize, as well as for the level of preference 4 (favorite) for tortilla made with yellow maize; although the correlation in the latter is lower, it is significant for the study (Tables 2 and 3).

The χ_c ² that relates income and consumption with the preference of consumers for purchasing at tortilla shops in neighborhoods next to their homes is high for those who are located in a high income and have medium consumption; because the probability of finding a χ_t ² is low, the null hypothesis is rejected. The high level of income with medium consumption has a high correlation (66.7 %) with the purchase of tortillas in a common tortilla shop (Table 4). According to the studio carried out by ^{Soria in 2000}, 77 % of the tortilla sale in México was done in neighborhood tortilla shops (common tortilla shop). When contrasting the data, it is observed that currently the sale in common tortilla shops has decreased, at least for the low and medium incomes. The different price of tortilla between establishments in the country shows the disadvantage of the small enterprises (tortilla shops) in face of the commercial chains, which is manifested in their lower ability for negotiation, reducing the possibility of facing the pressure from distributors, and whose level of income does not allow them maize storage for more than seven days, so they transfer their high costs to the consumer (^{Poder Legislativo, 2007}).

For the variables that correspond to the physical attributes that consumers prefer, the χ_c ² that relate the attributes of color, resistance and size with the levels of income and medium consumption are high, since it is not very likely to find an χ_t ² higher than χ_c ²; therefore, the null hypothesis is rejected. In the medium income and consumption, high correlations are observed for the preference for tortillas of white color, which are resistant and have in average a medium size (Tables 5, 6 and 7). The first two attributes are strongly related to the preference that consumers have for tortillas made with white maize, as well as the preference for those made in a traditional way (boiled maize, nixtamalizado), which in general tend to be more resistant than those made with maize flour.

In the case of the variables of attributes like flexibility and consistency, the χ_c ² obtained were high for the high income; because the probability of obtaining an χ_t ² higher than χ_c ² is almost zero, the hypothesis that the distribution between boxes is uniform is rejected. For the case of the attribute flexibility, a high correlation is observed for high income and with medium consumption of tortillas that are flexible; this is because tortilla tends to accompany a great variety of foods in which tortilla serves as wrap and spoon. A high correlation was also observed for a high income and médium consumption in relation to the purchase of tortillas that have a soft consistency (Tables 8 and 9).

The second analysis was carried out through an ordinal regression model that allows analyzing categorical variables that contain a scale or order. In this analysis, the kilograms of tortilla purchased (Kcom) were contemplated as the dependent variable, with which the high, medium and low consumptions were determined. The factors that were considered for the model were income (Ing), tortilla Price (PrecioT), type (TTor), preference for tortilla made with white maize (PTMB) and place of purchase (LCom). Table 10 shows the coefficients obtained, as well as the adjustment tests of the model (Table 11) and the test of parallel lines (Table 12).

The estimated model has a good fit, because when comparing the level of significance of 0.006, with α=0.05, the null hypothesis is rejected, which establishes that the model without predictive variables is better than the model with predictive variables.

In the test of parallel lines, the null hypothesis establishes that the coefficients of the slopes are the same for all the categories of the response variable. Therefore, when there is a level of significance of 0.493, the null hypothesis is rejected.

According to the significance tests for each one of the coefficients, it is observed that for α = 0.05 only the variables place of purchase (LCom), type of tortilla (TTor) and price (PrecioT) are related with the tortilla consumption. Specifically, it is said that a decrease in the price (low prices) is less likely to have an impact at high levels of consumption (purchases higher than 3 kilograms). However, it does impact the medium consumption, where most of the population is found, because if the price decreases it is likely that the tortilla consumption will increase.

Concerning the place of purchase, it is observed that it is less likely that purchases in nearby tortilla shops will impact the high consumptions. It is also observed that it is likely that the preference for tortillas made with boiled maize (nixtamalizado) will impact the high consumption; the contrary effect happens with low consumption of tortillas.

It was observed that medium consumption has a higher probability of being affected by the price, place of purchase and type of tortilla. That is, it has a specific level of selection with regards to the characteristics of the tortillas purchased: color, price, place and way of elaboration; most of the population from Estado de México is found in medium consumption.

Conclusions

Tortilla consumption in Estado de México is high due to the high population concentration and to the importance of the product in the diet. All the consumers with different income prefer tortillas made of boiled maize (nixtamalizado). However, only those with medium income prefer them to be white color and made with boiled maize (nixtamalizado).

Only the consumers located at high and medium levels of income prefer to perform their purchases in common tortilla shops, because they are willing to pay for the increased cost.

The continuous increase in the price of tortillas impacts more the consumers who have low income because they disregard their preferences for color, quality, type of tortilla and place of purchase, and they focus on acquiring it at a low price and in the amount that allows them to cover their daily consumption.