Engel’s approach as a tool for estimating consumer behaviour

15 2, XX, 2017 Economics DOI: 10.15240/tul/001/2017-2-002 Introduction Consumer’s preferences and wealth (income, resources) are viewed as exogenous quantities in the neoclassical theory. Methodological individualism, rational behaviour, equilibrium and perfect information of consumer are the further features of this concept. Simple precautions are the necessary condition for model application. Methodological individualism assume, that the principles of the individual subjects behaviour are the most important determining factors of the model functioning. These subjects act according to maximizing of the total utility in the given conditions. The concept of equilibrium is static. The perfect information precondition need not to be necessary fulfi lled because the contemporary concepts of the consumer’s behaviour function in the conditions of risk and uncertainty (Aguirre Sotelo & Block, 2014). Institutional concepts of consumption are derived from the T. Veblen’s theory of the leisure class consumption (Grzega, 2015). The preferences are infl uenced by the position of the consumer subject in the social hierarchy and by activity of fi rms. According to J. Galbraith, fi rms can partially create the desires of the consumers. Not only the preferences, resources, and the individual behaviour, but also social institutions are important factors of consumption in the institutionalism concepts. Consumption is viewed also as a social behaviour. Psychological aspects of the consumer subjects’ behaviour are also important. In the modern concept of consumption, there must be viewed all aspects of it. Consumer behaviour is based on the decision-making of individuals spending their own resources (i.e. time, money and efforts) in order to obtain the items associated with consumption Novotný and Duspiva (2014); Horáková (2015). This form of behaviour involves the reasons why, when, where, how often and what people buy, how often they use the purchased items, how they evaluate them after the purchase, and in which way these factors infl uence their future purchases Šrédl, Soukup and Severová (2013); Wanninayake W.M.C. Bandara (2014). Marketing is so much more than creating a catchy phrase or a jingle people will sing for days. Understanding consumer behavior is a vital aspect of marketing. Consumer behavior is according to Stávková et al. (2006) the study of how people make decisions about what they buy, want, need or act in regards to a product, service, or company. This issue was studied by Krishna and Schwarz (2014) in detail and in relation to psychology aspects. It is critical to understand consumer behavior to know how potential customers will respond to a new product or service (Bačík, Gavurová, & Fedorko, 2015). It also helps companies identify opportunities that are not currently met (Gavurová, Vagasova, & Kovac, 2016). Through research and observation, several models have been developed that help further explain why consumers make decisions, including black box, personal variables, and complex models. The black-box model is based on external stimulus-response, meaning something triggers the consumer to make buying decisions that are infl uenced by many factors, including marketing messages, sampling, product availability, promotions, and price (Michalski, 2008; Andrejovska & Banociova, 2014; Raisova & Durcova, 2014). When infl uenced by the personal-variable model, consumers make decisions based on internal factors. These internal factors may include personal opinions, belief systems, values, traditions, goals, or any other internal motivator (Dúbravská, Mura, Kotulič, & Novotný, 2015). The complex model of consumer behavior considers both internal and external variables. ENGEL’S APPROACH AS A TOOL FOR ESTIMATING CONSUMER BEHAVIOUR Renata Benda-Prokeinová, Kamil Dobeš, Ladislav Mura, Ján Buleca EM_2_2017.indd 15EM_2_2017.indd 15 14.6.2017 9:29:1514.6.2017 9:29:15 16 2017, XX, 2 Ekonomie Human food choice is a complex phenomenon infl uenced by a wide range of factors (Hes et al., 2008; Czarniewski, 2014). The theory of planned behaviour offers a means for trying to understand the roles of some of these factors. Evidence will be presented for the inclusion of measures of self-identity and moral obligation. Problems with the measurement of perceived control and possible means for overcoming them will also be discussed. The choice of foods is an area of concern for many people involved in the production and distribution of foods, and for those concerned with nutrition and health education. Relatively little is known about how and why people choose the foods that constitute their diets or about how their choices can be infl uenced in an effective way. Like any complex human behaviour, food choice will be infl uenced by many factors. There are a number of models in the literature, which seek to delineate the effects of likely infl uences. Piqueras-Fiszman and Jaeger (2016) concentrate on the question why consumers associate certain emotions with food/beverages and meal occasions. This topic is important as people’s food-related behaviour and attitudes are shaped by these associations. However, many such models are simply catalogues of the likely infl uences. Few of them present any indication of the likely mechanisms of action of the multitude of factors identifi ed, nor do they quantify the relative importance of factors or allow any quantitative tests which are predictive of food choice. Although such models can be useful in pointing to the variables to consider in studies in this area, they do not provide a framework for quantitative modelling of food choice behaviour. The factors infl uencing food choice are categorized as those related to the food, to the person making the choice and to the external economic and social environment within which the choice is made (Hildenbrand, 1994; Virglerova, Dobes, & Vojtovic, 2016). Some chemical and physical properties of the food will be perceived by the person in terms of sensory attributes, e.g. fl avour, texture or appearance. However, perceiving these sensory attributes in a particular food does not necessarily mean that a person will or will not choose to consume that food. It is the person’s liking for that attribute in that particular food which will determine whether or not the food is chosen. Other components in the foods will have effects upon the person, e.g. reducing hunger, and the learning of the association between the sensory attributes of a food and its consequences appears to be a major mechanism by which preferences develop. Marketing and economic variables, as well as social, cultural, religious or demographic factors are also likely to be very important. This paper applies a behavioral approach to consumption to connect differences in satiation patterns between innate needs with systematic changes in consumption expenditures when income rises. It is conjectured that, at the level of aggregate expenditure data, these differences translate into different income elasticities of demand for groups of goods and services that are likely to be consumed to serve those needs. This way, it should be able to explain a good deal of the differences in the shapes of Engel curves of the underlying goods and services (Lewbel, 2008; Lades, 2013). We illustrate in the paper the technique of Engel curve computation on the household’s income and expenditure data originated from the Statistical Offi ce of the Slovak Republic. Results of the method will be presented as a set of elasticities, which provide deeper knowledge about consumer’s behavior of the Slovak households. 1. Theoretical Background of Consumer Behavior Factors that infl uence the purchasing behavior of consumers are very signifi cant for businesses, because based on these factors, it is possible to focus the trade policy, which then could lead to better business results (e.g. an increase in sales volume, market share, distribution of a portfolio of clients, etc.). Every human is a consumer, and for the other consumers represents the key factors that infl uence their purchasing behaviour. Theory and practice nowadays lacks greater links between businesses and consumers, based on segmentation by appropriate segmentation criteria. More detailed description of segmentation criteria highlights (Koudelka, 2005). In many areas of law such connections, including research that would facilitate company decisions is inadequate. There are a number of research projects related to lifestyle, dress, smoking, eating (Šoltés & Gavurová, 2014), but research including models and factors EM_2_2017.indd 16EM_2_2017.indd 16 14.6.2017 9:29:1514.6.2017 9:29:15 17 2, XX, 2017 Economics infl uencing purchasing behaviour consumers in many manufacturing industries is missing. The issue of purchasing behavior of consumers was discussed by several prominent authors (Kotler & Armstrong, 2004; Solomon, Marshall, & Stuart 2006; Stávková et al., 2006; and others). The assumption is that the consumer will act rationally and will maximize the overall benefi t that is associated with a combination of goods and services that accompany the product. According to Hes et al. (2008), however, it is limited by the consumer when making decisions, especially by the fi nancial resources. In many fi elds linking businesses and consumers, including research their behavior that would facilitate decision of company management is insuffi cient. One of other approaches to the analysis of consumer behaviour is also monitoring behaviour when there is a change in information cost. This may vary across consumers, which is the case when consumers who search for information are “bounded rational” – making random errors due to limited attention (Manzini & Mariotti, 2014) or “rationally inattentive” – optimally allocate attention (Matějka & McKay, 2015). Following these approaches, it is also important to study how to optimally inform a consumer about the valuations for multiple goods when a consumer can learn about fewer goods than he/she can consume (Saak, 2016). Market orientation is an approach based on marketing conception, introduced by Drucker (1954). Although the principle of market orientation has been known for sixty years, detailed analysis only began at the beginning of the 1990s. In 1990, the Journal of Marketing published the results of two studies sponsored by the Marketing Science Institute. Those studies were made by Kohli and Jaworski (1990) and by Narver and Slater (1990). These two pieces of research had a substantial impact on the formation of a new approach to marketing and subsequent studies on market orientation and took the initial steps towards creating a defi nition of market orientation. The defi nition of Kohli and Jaworski (1990) was totally different to those by Narver and Slater (1990). The defi nition by Kohli and Jaworski (1990) perceives market orientation as an implementation, namely as an instrument for strategic decision-making. Alternatively, the defi nition of Narver and Slater (1990) perceives market orientation from a business learning and corporate culture point of view Novotný and Duspiva (2014). Other approach by Šrédl, Soukup and Severová (2013) and by Peréz et al. (2013) were applied in modern market economy where the supply exceeds demand, and the importance of the “consumer’s behaviour in the market analysis” continuously increases. Consumer’s preferences are viewed as exogenous variables in a neoclassical theory. They are not explained in a framework of the concept but are viewed as given ones. Heikkinen (2015) studied decline and growth in a dynamic equilibrium model with preference heterogeneity. Simulations suggest that decline triggered by voluntary simplicity increases equilibrium welfare under externalities in consumption and leisure. Methodological individualism, rational behavior, equilibrium, and perfect information about the consumers are the next features of this concept. Simple precautions are a necessary condition for the model application. Methodological individualism means that principles of individual subjects behaviour are the most important determining factors of model functioning. These subjects abide by the principle of total utility maximizing in given conditions. The concept of equilibrium is static. Perfect information precautions doesn’t need to be necessarily fulfi lled, because contemporary concepts of consumer’s behavior function in the conditions of risk and uncertainty. Institutional concepts of consumption are derived from Veblen’s theory of leisure class consumption (Herrnstein, 1997). There is also a strong marketing infl uence. Subramanyam and Kumaraswamy (1981) consider the effect of marketing policies and conditions on demand. It is worth to mention a few smart marketing tools infl uencing customer behaviour. Online customer behavior is a new phenomenon and it is a very important case for fi rm´s marketing (Rypakova, Moravcikova, & Krizanova, 2015). Kramolis and Kopeckova (2014) deal with this kind of smart marketing tool (Product placement) which clearly infl uences customers. Indeed, there are not only luxury goods, (Kramolis & Drabkova, 2012) which are promoted by means of product placement. This marketing activity has a simple purpose – to connect brands with appropriate entertainment partners to maximize brand relevance –- which leads to strengthening consumers’ attitudes to promoted products. This smart marketing whisperings can build the preferences wanted. EM_2_2017.indd 17EM_2_2017.indd 17 14.6.2017 9:29:1514.6.2017 9:29:15 18 2017, XX, 2 Ekonomie Moreover, preferences are infl uenced by the position of a consumer subject in the social hierarchy. Not only preferences, resources and individual behavior, but also social institutions represent the important factors of consumption in institutionalists’ concepts. Consumption is also viewed as social behavior. Ultimately psychological aspects of consumer subject’s behavior are also important. The contemporary relationship between brand personality and consumer personality has become a researched issue in recent years and it was studied by Banerjee (2016). The author´s result shows buying decision, personalities of promoted products and corporate brands infl uencing consumers´ preferences. Engel curves describe the change of expenditure (or demand) for different commodities as a function of income. It is common to speak about „cross-section Engel curves“ if one considers expenditures of a population of households (for example, all households in a large countries (Becerra Alonso, Androniceanu, & Georgescu, 2016)) rather than single individuals. Since the work of Engel (1857), cross-section Engel function has been a major research in economies and econometrics. Engel curves have been extensively studied to establish empirical relationships between expenditure and income. Engel’s law states, that the budget shares for the commodity food decrease with an income. Generally, Engel functions can be used to classify commodities into luxuries, necessities and inferior goods. The most known classical study of cross-section Engel curves is that of Prais and Houthakker (1955). Important objective is the determination of budget elasticities for different commodities from the corresponding Engel curves. In demand analysis, the structure of crosssection Engel curves has been exploited to build up demand system. Working (1943) proposes the parametric model m(x) = θ1x + θ2x logx for the cross-section Engel function m(x) of some commodity. Here, x denotes income, the θ1, θ2 are unknown parameters, which are to be estimated from given data. Deaton and Muellbauer (1980) consider this model as an important motivation of their famous „AIDS – almost ideal demand system“. When doing the analysis of food demand, the model of AIDS was used in works of Benda Prokeinova and Hanova (2016), Ulubasoglu et al. (2016) and Hayat et al. (2016). The fi nding of Lewbel (1993), telling that a better parametric model might be written in the form m(x) = θ1x + θ2x log x + θ3x (log x)2 is often interpreted as a justifi cation of the quadratic almost ideal demand system of Blundell et al. (1993), which generalizes the results of Deaton and Muellbauer (1980). These approaches are based on the idea of recoverability of individual behavior from aggregate data. In the present context this corresponds to the assumption that the structure of cross-section Engel curves refl ects the structure of individual ones. The same argument was used by Härdle and Jerison (1991), where a certain variance of estimated cross-section Engel curves over time were used to draw conclusions on individual behavior. Engel curves also play a role in the theoretical economics. Hildenbrand (1994), and Kahneman (1997) provides conditions on the structure of these curves which ensure the law of demand. There is thus strong empirical and theoretical interest to analyze the cross-section Engel function of real populations. A prerequisite of any economic interpretation is a reasonable estimation of these curves from given crosssection data about household expenditures and income. Statistically, cross-section Engel curves are regression function, m(x) equals to the conditional expectation of expenditure for a commodity given the income level x. The classical approaches in regression analysis are based on parametric models, i.e. m(x) = g(x, θ) with a known function g and an unknown parameter θ. The vast majority of studies of Engel curves are based on such parametric models. Apart from the Working model, there are many further parametric approaches (double logarithmic, semilogarithmic, etc.). In any parametric model the functional shape of the estimate is already given by an assumption. The quality of the resulting estimator depends heavily on the correctness of this specifi cation. If a model is correct, good estimators of the parameters and the curves can usually be obtained. However, if the model is poorly specifi ed, the resulting estimator is not even consistent. Since the economic theory doesn’t yield any information about a correct parametric model, there is thus reason to consider the alternative approaches. EM_2_2017.indd 18EM_2_2017.indd 18 14.6.2017 9:29:1514.6.2017 9:29:15 19 2, XX, 2017 Economics 2. Material and Methods According to Friedman’s hypothesis the consumption of individuals or households depends on p income, it does not depend on the current income of economic unit as Keynes insisted. Permanent income can be explained as an average income, which an individual expects to earn during his life using his labor and capital. Šimsová and Reissová (2016) deal with the topic of expected income of selected groups of inhabitants, e.g. university students. According to Gombos, Kiss and Zvara (2016) it depends on household welfare (human capital, education, health included), interest rate, and expected job vacations. The similar approach was described by Engel. Engel curves describe the change of expenditure (or demand) for different commodities as a function of income from the theoretical viewpoint. 2.1 Engel Curves and the Income Elasticity of Demand The general connection between the shape of an Engel curve and the income elasticity of the respective good can easily be illustrated. If an Engel curve for good i is expressed in terms of the expenditures qi spent on i depending on the households’ income y, the slope of the fi tted curve dqi/dy can indirectly be used to derive i′s income elasticity of demand. As a good’s income elasticity is defi ned by the relative change in qi (dqi/qi) divided by the relative change in y (dy/y), it can be estimated by regressing logarithm of qi on logarithm of y Lewbel (2008). Using expenditure shares (wi=qi/ywi=qi/y), instead of the expenditures spent on i, facilitates more directly an inference of the income elasticity of demand from the curve’s slope. While an increasing slope represents more than proportional expenditures, thus luxuries, a decreasing slope points to less than proportional expenditures with rising income, i.e., necessities. Although Engel curves usually show considerable nonlinearities Lewbel (2008), budget share Engel curves facilitate to readily observe the income elasticity of demand for the underlying group of goods and services. 2.2 A Motivational Approach to Consumer Behavior Engel’s approach to consumption never solely focused on food expenditures. In fact, Engel’s original contribution was meant to determine and measure household welfare by Chai and Moneta (2010). He started by categorizing expenditure items according to the underlying “wants” they satisfy. The importance of these wants was subsequently assessed by the empirically derived expenditure patterns. Engel’s work thus essentially focused on a behavioral foundation of the necessity-luxury taxonomy. In a revised version of his original contribution, Engel (1895) made even more explicit, that the motivation of human action, and thus also consumption behavior, is rooted in the satisfaction of universally-shared needs (Chai & Moneta, 2010). To build upon Engel’s legacy, this paper enriches the classifi cation of needs with current scientifi c knowledge on the nature of consumer needs and how they are satisfi ed. However, motivational underpinnings of economic behavior in general and consumer behavior in particular are rarely addressed in economics. Among the existing works, two different explanatory approaches were identifi ed by Witt (2010). While in the utilitarian hedonic approach the explanation refers to the motives of seeking pleasure and avoiding pain, nonhedonistic variants focused on the motivating power that deprived needs and wants have for consumption activities. This paper closely connects to the latter approach. A behavioral-need based– interpretation of the consumption motivation was offered by Witt (2001; 2010) to actually theoretically explain the necessity-luxury distinction that becomes visible through the shape of Engel curves. The theory postulates an intimate relationship between human biological and cultural evolution in the sense that cultural development is based upon as well as constrained by innate behavioral dispositions and cognitive learning abilities, which have emerged during human phylogeny. Hence, the theory focuses on the explanation of long-run economic change from a biological and psychological perspective (Witt, 2008). The theory of the learning consumer (Witt, 2001; 2010) emphasizes the role of human needs as ultimate motives of consumption behavior. The theory distinguishes between genetically determined-innate-needs and both culture and socialization specifi c acquired wants, that result from processes of associative learning (classical and operant conditioning) EM_2_2017.indd 19EM_2_2017.indd 19 14.6.2017 9:29:1614.6.2017 9:29:16 20 2017, XX, 2 Ekonomie and social-cognitive learning. The theory holds the attempt to relieve or reduce deprivation of a limited number of innate needs, consequently an increase in satiation level of these needs, is one major motivation to consume. Deprivation is thus seen to intrinsically motivate consumers to act which creates a rewarding experience. Needs are the contingencies under which deprivation occurs. Although needs can be manifold, in this context only the subset of universally-shared “basic” needs i relevant. Among these are that for sleep, for something to drink, for something to eat, for maintaining body temperature, for physical activity, for status recognition, or for sensory arousal Witt (2011). An important characteristic of these needs is that their satisfaction by an action effects a primary reinforcement in the sense of instrumental or operant conditioning (Staddon & Cerutti, 2003). This, of course, has implications for the allocation of resources. Imagine individuals to be equipped with a fi xed set of innate needs that show different need deprivation states. If individuals allocate their behavior to obtain a reduction of their deprivation states, the likelihood that a particular activity is chosen over another one depends on its relative contribution to reducing deprivation. Individuals are accordingly assumed to shift their behavior to those alternatives which provide the highest average reward. The allocation of income to consumption categories results as proportional to the corresponding need deprivation states. Under reinforcement learning, the frequency distribution over actions converges to a state satisfying the so called “matching law” Herrnstein (1997), which seems to be a good approximation of behavior driven by mainly basic need deprivation states and that is characterized by low levels of cognitive intervention. Among the universally-shared needs, a further distinction relates to the underlying satiation properties they show Witt (2001; 2010). On the one hand, there are basic needs which follow homeostatic features, i.e., deprivation can be reduced relatively easily up to the temporary satiation point once rising income allows for a suffi cient increase in the corresponding consumption expenditures per period of time. The motivation to consume is, then, temporarily reduced or removed. Satisfaction of these needs depends mainly on the intrinsic value of the corresponding goods and services. Examples are the homeostatic needs underlying to eating, drinking, sleeping, and the maintenance of body temperature. On the other hand, there are also basic needs where homeostatic features are absent, and where it is therefore diffi cult, if not impossible, to reduce the average deprivation to zero. Typically, these are needs whose satiation level is defi ned in relative terms, like the need for arousal and for social recognition. Despite interpersonal sources of variance, which can be expected due to individual cognitive and conditioning learning processes, it can be conjectured, that shared innate needs exert some systematic effects on behavior, that are visible at the level of the population means, i.e., at the level of aggregate consumer expenditures. As needs differ with respect to the amount of spending that is necessary to reach satiation, this difference can be expected to become relevant with rising real income Witt (2001; 2011). Being able to spend more, consumers should be able to approach the satiation level of some needs faster than the satiation level of other needs. Their consumption motivation is not equally upheld and their spending should thus not expand equally. Differences in the income elasticity of demand for the products that serve the different needs should express this differential satiation effect. A formal model of the differential satiation effects and its implications for Engel curves has recently been put forward by Lades (2013). Taken together, the behavioral approach to consumption suggests connecting the differences in satiation patterns between innate needs with systematic changes in consumption expenditures when income rises. It is conjectured that, at the level of aggregate expenditure data, these differences translate into different income elasticities of demand for products or groups of goods and services that are likely to be consumed to serve those needs. The behavioral approach to consumption should thus be able to explain a good deal of the differences in the shapes of Engel’s curves and thus in the income elasticities of the underlying goods and services. A double-log specifi cation of the Engel’s function has been chosen in order to estimate the expenditure elasticity of households by using an economic status of households’ head at work for each household by children per EM_2_2017.indd 20EM_2_2017.indd 20 14.6.2017 9:29:1614.6.2017 9:29:16 21 2, XX, 2017 Economics person. The data set were obtained from the Slovak Statistical offi ce and consists of the yearly observations: money incomes of private household by economic status of household head at work per person and month (€); money incomes of private household by children per person and month (€); money expenditures of households by using and economic status of household head at work per person and month (€); money expenditures of households by aim of using and number of children (€). The household Budget Survey of the Slovak Statistical offi ce was used for the period 2004-2014. The use of the household level data offers the potential of richer dataset that may offer an additional insight into the underlying economic relationships. A Double-Log Specifi cation has proven the most appropriate way of estimating the expenditure elasticity of demand; it generates more realistic expenditure elasticities. The general model can be written as follow: ln wj =αj + βj ln y +ηj (1) where: wj is the average annual per capita expenditure share for food group j, αj, and βj are the estimated coeffi cients, y is the average total per capita income calculated as the average annual total per capita expenditure, and ηj is the disturbance term. As pointed out before, the derivation of the Engel’s function assumes constant prices. Equation is estimated for each of the 13 food groups for each household by child and by economic status of household head. We have to include dummy variables. We have supposed there are differences in the elasticity for each food group among several types of households. Households’ Classifi cation Analysis of the income elasticity was determined in 2,418 Slovak households. Different types of households were classifi ed based on the economic status of the household head (household of employees, self-employed household, household of pensioners and others). For better orientation it is essential to characterize the household types as defi ned by the Statistical Offi ce of the Slovak Republic: Economic status of household head was characterized by his/her employment status. Employee: a person, who works for a public or private employer, and who receives compensation in the form of wage or salaries, including pensioners with income from employment and co-operative members. Self-employed person: works in their own business, including pensioners with income from own business. Pensioner: a person without job, who receives retired pension, with no income from job or business; in household could live also other persons without old-age pension. Others: all persons, which were not included therein before (unemployed, maternity or paternity leave, students, etc.). 3. Results and Discussion 3.1 Analysis of the Income Elasticity of Demand Quantity demanded sensitivity to changes in income, ceteris paribus, was measured by the coeffi cient of income elasticity. The income elasticity of demand is the ratio of the percentage change in quantity demanded product X to percentage change in income, and shows how many percent will change the demand for product X if there is 1 percent change in consumer income. For normal goods the income elasticity is positive. As we deepen our classifi cation of goods to essential goods, the 1% change in consumer’s income caused less than 1% change in demand for the goods. Because the share of the goods on a total income decreases, the average propensity to consume is decreasing and true: 0 ≤ Eid ≤ 1. Analysis of the income elasticity was investigated for each type of households by number of dependent children and for households by economic status of the household head. From the Tab. 1 it could be identifi ed that the calculated elasticity were highly signifi cant. All food groups have explained variability over 90%. Almost every elasticity reached the positive value, except these of meat and oils and fats. This means, that all the food groups within the interval 0 ≤ β ≤ 1 are the Slovak households without children needs. 1% change in the consumer’s income results in a change in demand for the good which is less than 1%. EM_2_2017.indd 21EM_2_2017.indd 21 14.6.2017 9:29:1614.6.2017 9:29:16 22 2017, XX, 2 Ekonomie The highest value of the elasticity was found in case of vegetables, including potatoes. If we have increased the revenue by 1%, the demand changed by 0.66%. Soft drinks, coffee, tea, cocoa, and mineral water have relatively similar levels of elasticity at the level of 0.5%. Fish, milk, cheese, and eggs showed the elasticity around 0.2%. The consumption of these products is relatively little affected by the change of income. We have focused on negative elasticity mainly. Very sensitive commodity was the meat. From a nutritional point of view this is an important food, and in fi nancial terms, it represents one of the most expensive food for the Slovak households. This fact was indicates by the elasticity value -0.08. 1% change in income will surely results in change of consumer preferences for meat by -0.08%. Fats & oils have even more negative elasticity value, reaching -0.22% only. In case of households without children the needs for goods such as meat, oil & fats were only inferior, demand for them decreases with an increasing income. Fig. 1 shows the differences in retirement elasticity of households according to number of children. There is a signifi cant difference between households with children and households without them. The number of children per household is not the most important factor for a change in retirement elasticity. The analysis of variance among the groups of households studied shows that the null hypothesis H0 cannot be rejected: All population Household without children Household with 1 child Household with 2 children Household with 3 and more children αβProb R2 EID αβ EID αβ EID αβ EID prob R2 Bread & cereals -1.04 0.55 1.39E-15 0.982 necessary goods -1.38 0.75 necessary goods -1.38 0.81 necessary goods -1.38 0.93 necessary goods 1.23E-17 0.96 Meat 1.28 -0.08 7.67E-16 0.983 inferior goods 0.43 0.37 necessary goods 0.43 0.26 necessary goods 0.43 0.22 necessary goods 1.77E-15 0.93 Fish -2.18 0.21 4.96E-18 0.991 necessary goods -2.48 0.5 necessary goods -2.48 0.4 necessary goods -2.48 0.35 necessary goods 2.24E-16 0.94 Milk, cheese & eggs 0.06 0.29 9.93E-12 0.949 necessary goods -0.77 0.55 necessary goods -0.77 0.54 necessary goods -0.77 0.54 necessary goods 7.07E-14 0.91 Oil & fats 0.13 -0.22 7.33E-14 0.972 inferior goods -0.21 0.23 necessary goods -0.21 0.1 necessary goods -0.21 0.1 necessary goods 2.86E-09 0.8 Fruits -1.95 0.3 5.44E-13 0.964 necessary goods -1.67 0.44 necessary goods -1.67 0.36 necessary goods -1.67 0.29 necessary goods 2.51E-12 0.88 Vegetables incl. potatoes -3.3 0.66 1.17E-10 0.932 necessary goods -3.6 0.92 necessary goods -3.6 0.88 necessary goods -3.6 0.94 necessary goods 1.70E-11 0.87 Sugar, jam, honey, chocolate & confectionery 0.19 0.01 6.01E-13 0.964 necessary goods 0.23 0.2 necessary goods 0.23 0.14 necessary goods 0.23 0.1 necessary goods 1.25E-09 0.81 Coffee, tea & cocoa -3.11 0.5 2.48E-14 0.975 necessary goods -3.22 0.7 necessary goods -3.22 0.67 necessary goods -3.22 0.66 necessary goods 1.52E-15 0.93 Mineral waters, juices & other non-alcoholic drinks -2.18 0.59 6.77E-11 0.936 necessary goods -3.07 0.68 necessary goods -3.07 0.68 necessary goods -3.07 0.58 necessary goods 1.73E-20 0.97 Alcoholic beverages & tobacco -1.86 0.47 4.88E-14 0.973 necessary goods -1.86 0.58 necessary goods -1.86 0.42 necessary goods -1.86 0.29 necessary goods 1.63E-16 0.95 Alcoholic beverages -2.38 0.35 7.13E-17 0.987 necessary goods -2.47 0.58 necessary goods -2.47 0.54 necessary goods -2.47 0.46 necessary goods 9.65E-15 0.92 Source: own calculations Legend: α-intercept, β-regression coeffi cient, EID-income elasticity of demand, prob.-probability value, R2-coeffi cient of determination Tab. 1: Elasticity of households according to number of children per person and month in 2004-2014 (in €) EM_2_2017.indd 22EM_2_2017.indd 22 14.6.2017 9:29:1614.6.2017 9:29:16 23 2, XX, 2017 Economics with regard to elasticity is equal (p-value = 0.14738). Therefore, there is no statistically signifi cant difference among the households according to number of children. Analysis of the relationship between models of consumer demand based on micro and aggregate data were provided by Blundell et al. (1993). According to their fi ndings it is important to establish the presence of nonlinearity in the micro-level Engel’s curves and the need for interactions with householdspecifi c characteristics, since either of these would rule out simple linear aggregation. In our sample, pooled over 10 years, we have found strong evidence of both. Similar to our result were the fi ndings of the foodstuff model described by Chai and Moneta (2010), in which the prices of corn and fruits & vegetables positively affect income; the prices of dairy products affects the income negatively. In the wage income model on the other side the elasticity of corn was the same as in previous models; the elasticity of dairy products was signifi cantly negative, but smaller in absolute values (-0.58 vs. -1.41, and -1.10); and the elasticity of fruits & vegetables was no longer statistically signifi cant. There were minor differences in the estimates of the demand elasticity across the models. Income elasticity was slightly above the levels for cheese & giblets, and notably above the values for bovine (1.43), other cereals (1.37), and mutton (1.50). Income elasticity falls between 0.5 and unity for fi sh (0.56), bread (0.64), eggs (0.66), rice (0.74), yogurt (0.75), milk (0.79), poultry (0.87), fats (0.88), dough product (0.95), and oils (0.98). Our income elasticity for bread constitutes the lower limit of earlier estimates, which range from 0.65 to 0.85, while income elasticities for meat, meat products, dairy products, and eggs were similar to those reported by Hes et al. (2008), Lades (2013), and Witt (2001; 2011). The estimation results indicate that the total per capita expenditure and the household size have signifi cantly negative estimates in both groups of households. The estimate of remittances is positive and signifi cant only in the group of rich households, so when the remittances increase, households above the poverty line increase their share of food consumption Narver (1990). Households with 1 child has the highest value of the elasticity of vegetables including potatoes and bread and cereals. If we increased revenue by 1%, the demand for vegetables increased by 0.92% and the demand for bread and cereals should be increased by 0.75%. The lowest Fig. 1: Box plot of elasticity of households according to number of children Source: own EM_2_2017.indd 23EM_2_2017.indd 23 14.6.2017 9:29:1614.6.2017 9:29:16