European Advances in Consumer Research Volume 2, 1995 Pages 149-156
SOCIO-DEMOGRAPHIC, PSYCHO-SOCIAL AND LIFE-STYLE FACTORS AFFECTING CONSUMPTION OF CONVENIENCE FOOD
K.S. Veenma, TNO Nutrition and Food Research Institute
C. Kistemaker, TNO Nutrition and Food Research Institute
M.R.H. Lowik, TNO Nutrition and Food Research Institute
K.F.A.M. Hulshof, TNO Nutrition and Food Research Institute
A.G.M. Steerneman, University of Groningen
M. Wedel, University of Groningen
[Supported by the Dutch Ministry of Agriculture, Nature Management and Fisheries.]
Convenience food products have gained great popularity in the past decade. However, the nutritional value of these products may be poor. Although it is only the total food pattern that can be qualified as healthy, the chance to realize this is lower as the proportion of nutritionally poor products is greater. We attempted to assess determinants of convenience food usage to gather information for nutrition education programmes, in which messages may be targeted at specific groups. A sample of 1783 female housekeepers, studied within the framework of the 1992 Dutch National Food Consumption Survey, was used. Significance of the effects of psycho-social, socio-demographic and life-style factors was assessed using a covariance structure model. The results indicated that the most important determinants of convenience food usage were nutritional knowledge, socio-economic status (SES), marital status, employment status and stage in the family life cycle. Knowledge of food and nutrition was inversely related to convenience food usage. When aiming at diminishing convenience food usage in the Netherlands by improving nutrition knowledge, nutrition educators might focus their attention on women with a low SES who are in an advanced stage of the family life cycle (older women), are married and do not have a paid job.
In the USA consumption of meals that take less time to prepare and to eat has increased considerably in the 1980s and the early 1990s and is expected to grow even further in the years to come (U.S. Department of Agriculture 1988; Sills-Levy 1989), although consumption of convenience foods is not a new phenomenon (Axelson et al. 1983; Redman 1980). In the Netherlands, this trend has taken place in the past decade as well, albeit some years later than in the USA (van Harten 1985; Kouwenhoven and van der Veer 1990). A comparison of the first (1987/1988) and the second (1992) Dutch National Food Consumption Survey shows that the consumption of easy-to-prepare products has increased by some 30% (Voorlichtingsbureau voor de Voeding 1993).
Several factors may have led to the increased consumption of this kind of food. Some authors have considered psycho-social factors characterizing users of convenience foods (Axelson et al. 1983; Redman 1980; Axelson et al. 1985; Sims 1978; Witte et al. 1991), whereas others studied socio-demographic variables (Axelson 1986; Belante and Foster 1984; Haines et al. 1990; Reilly 1982; Waslien 1988). Because convenience food products have a high energy-to-nutrient ratio, are rich in saturated fat, sodium and additives and lack vegetables (Axelson et al. 1983; Haines et al. 1990), and hence are often considered nutritionally 'poor', knowledge of factors affecting the use of these products is important. They can be a basis for nutrition education programmes, indicating how these meals should be integrated to achieve a healthy diet.
In past research several authors have focused on different aspects of convenience foods: snacks (Cronin et al. 1982), related product groups (Pearson et al. 1986), fast-food restaurants (Gofton and Ness 1991), or the more general variant 'eating away from home' (Axelson 1986; Belante and Foster 1984). In other research convenience (food) is considered more theoretically (Gofton and Ness 1991; Meulenberg 1993). From a theoretical point of view we adopt the definition of Gofton and Ness (1991): "Convenience refers to the ease with which a product may be prepared, served and eaten". Furthermore, the product needs to be bought and consumed in a time-saving form (Meulenberg 1993). Meals consumed in fast-food restaurants and take-out food, as well as prepared and easy-to-prepare food are considered as convenience food in our study. In the analysis only those products which are a component of what is considered to be a hot meal will be taken into account. For modelling consumer behaviour with respect to the consumption of these foods a broader range of psycho-social and socio-demographic characteristics was included and a larger segment of the convenience food products was studied, than in most other studies.
The data used in this study have been collected within the framework of the second (1992) Dutch National Food Consumption Survey (DNFCS), which is part of the Dutch Nutrition Surveillance System (Hulshof and van Staveren 1991).
As a sampling frame we used an existing panel (AGB Scriptpanel), which is a probability sample of non-institutionalized households in the Netherlands. From the AGB Scriptpanel a representative sample was drawn to be used for the 1992 DNFCS (AGB Fresh Foods 1994). Actually within the panel 2,475 households (comprising 6,218 subjects) participated (response rate 72%).
The second DNFCS employed questionnaires related to food consumed at home and away from home, a questionnaire on nutrition knowledge and attitudes, and a questionnaire enquiring some personal characteristics. Detailed information on the foods consumed at home and away from home was collected on two consecutive days by means of a two-day record. Subjects primarily responsible for domestic affairs were requested to complete the questionnaire on nutrition knowledge and attitudes. Most of these subjects agreed to participate (2,209, response rate 89%). Four dimensions of nutrition knowledge were tested in the questionnaire: preparation and preservation of food, definitions of food and nutrition, functional aspects of nutrition, and aspects related to the nutritional composition of some food items (TNO Nutrition and Food Research Institute 1991). In another part of the questionnaire food- and nutrition-related attitudes were measured using statements on (healthy) food. Among other things, respondents evaluated 54 statements by indicating their position on a 5-point Likert scale (1=strongly disagree, 5=strongly agree). Above all, respondents were asked several personal questions (e.g. about leisure activities). Demographic and socio-economic facts about the panel members were also known, both on an individual and a household level.
In this study data obtained by the various questionnaires were combined, which means that only the 2,209 respondents primarily responsible for domestic affairs (who completed the questionnaire on nutrition knowledge and attitudes) were included in the analysis. Respondents following a certain regimen with respect to food (e.g. vegetarian, n=57), or on a diet prescribed by a physician (n=220) or both (n=6) were excluded from the analysis, since their deviating consumption habits might have affected consumption of convenience foods. Furthermore, one female respondent aged 17 who still lived with her parents was excluded from the analysis because she completed the questionnaire on nutrition knowledge and attitudes, but was not the main housekeeper of the household. The remaining sample consisted of 1925 respondents, 142 men and 1783 women aged 20 to 75 (average 43 years). The gender imbalance was due to the fact that in most household women are the main housekeepers.
Significance of psycho-social, socio-demographic and life-style factors was assessed using a covariance structure model, in which because of the nominal and ordinal properties of several variables polychoric and polyserial correlations between the factors were calculated and used as input (J÷reskog and S÷rbom 1988; J÷reskog and S÷rbom 1989). Covariance structure analysis is a technique in which, in its most general form, the researcher posits a causal model among a set of unobservable constructs based on theoretical arguments. These constructs are represented by latent variables, which are empirical measures of the constructs. Each latent variable is measured by a set of observable indicator variables, which may be assumed to be measured with error (Hair et al. 1992). The constructs and the relations among these constructs are based on empirical results (Sills-Levy 1989; Axelson et al. 1983; Redman 1980; Axelson et al. 1985; Sims 1978; Witte et al. 1991; Axelson 1986; Belante and Foster 1984; Haines et al. 1990; Reilly 1982; Waslien 1988; Hulshof and van Staveren 1991; Sapp 1991; Schwartz 1975; White et al. 1988; Hulshof et al. 1992; Eppright et al. 1970; Grotkowski and Sims 1978; Perron and Endres 1985; Sims 1976; Schultz 1981; Courgeau 1989; TNO Nutrition and Food Research Institute 1994) or theoretical considerations (TNO Nutrition and Food Research Institute 1994). Figure 1 shows the model that has been estimated.
A few notes with respect to the specification of the model should be made. The attitude constructs were based on the questionnaire on (healthy) food, and were assessed using factor analysis. Based on the percentage of variance criterion (and to avoid identification problems) four components were used to represent the respondents' attitude towards nutrition (top four constructs in Table 3), together explaining 73% of total variance in all 54 statements. Variables where chosen to represent these four components in the analyses, based on the magnitude of correlation between statements and components. These indicators (three for each construct) are given in Figure 1.
Possible measures of convenience food usage are frequency or quantity of usage. The major disadvantage of both is that they do not measure the relative importance of convenience foods compared to other meals. Also the frequency of usage does not clearly distinguish between respondents when two-day food records are used as a means to register consumption. Using the quantity consumed will be difficult because of aggregation problems concerning different kinds of foods. Therefore, the convenience food usage of the subjects was measured as the average amount of energy obtained from convenience food products for a hot meal and expressed as a percentage of the average total amount of energy consumed during the two-day record. This measure does provide an indication of the relative importance of convenience food compared to other food consumed during the two-day record (Haines et al. 1990).
Nutrition knowledge was measured as the percentage of correct answers on each of the four knowledge dimensions mentioned above.
The value of the indicator socio-economic status (SES) was based on the education and occupation of the main wage earner, and was low (high) if the respondent was in a high (low) class. The value of the degree of urbanization was low if the subject lived in a highly urbanized region. Some respondents skipped meals, so an indicator was calculated on the basis of the use of breakfast, lunch and dinner during the two days covered by the diary, to measure the use of meals (0,1 up to 6). The stage in the family life cycle was composed on the basis of the age of the respondent and (if applicable) the age of the child(ren). Respondents who were on a diet prescribed by a physician were excluded, but a dummy variable was used to assess the effect of a diet followed on a person's own initiative.
Information on the food consumption of each respondent was based on the consumption on two consecutive days. Obviously, in such a short period the use of convenience foods can be missed or over-emphasized within individual records. However, all seven days of the week were randomly distributed over the population in such a manner that every day was equally represented. Associations within the model will therefore possibly be weakened but not biased. The effect of weekend days was assessed with a variable that indicated the number of days in the weekend in the two-day period of the diary (0, 1 or 2).
With respect to the relations among the constructs it is hypothesized that knowledge affects attitudes, following the results found by Grotkowski and Sims (1978). Furthermore, it is hypothesized that nutrition knowledge influences the life-style of the subject, primarily because the life-style is measured (for a large part) in terms of food behaviour (the use of meals) or is the result of this behaviour (body mass index, diet).
Differences in consumption patterns between male and female subjects have been found in several studies (Axelson 1986; Waslien 1988). Moreover, gender differences in life-style (Hulshof and van Staveren 1991), nutrition knowledge (Waslien 1988; Shepherd and Towler 1989), and attitudes (Shepherd and Towler 1989) have been found. On the basis of these results it was decided to estimate the model for female respondents only (n=1783). Estimation of the model for male subjects only was not possible because of the small sample size (n=142) (J÷reskog and S÷rbom 1989; Fornell 1983).
The covariance structure model has been estimated using a two-step procedure (Hair et al. 1992; Anderson and Gerbing 1988). In the first step the measurement model was estimated, i.e. the reliability and the measurement errors of the indicators was assessed. In the second phase the structural model was estimated, assessing the magnitude and significance of the relations among the constructs and the errors in the structural equations. Both steps were performed using maximum likelihood (ML) estimates and the computer programs PRELIS (J÷reskog and S÷rbom 1988) and LISREL7 (J÷reskog and S÷rbom 1989).
Measures for the reliability of the constructs are the construct reliability and the variance extracted. The construct reliability should exceed 0.7, and the variance extracted should measure 0.5 (Hair et al. 1992). Both measures were calculated for the constructs including more than one indicator. Measures for the overall model fit are the (adjusted) goodness-of-fit measure (GFI and AGFI), the chi-square (c2), the normed chi-square (the c2 adjusted for the degrees of freedom), and the root mean squared residual (RMSR). Although these measures should be interpreted with caution (J÷reskog and S÷rbom 1989; Hair et al. 1992; Fornell 1983), a combination of these measures can give an indication of the degree to which the model fits the data. The validity of the measurement and structural model was assessed by re-estimating both models with several sub-samples of the original sample of female subjects and by ascertaining the stability of the solutions. Furthermore, identification of the solutions (an unique solution should be found) was established using different starting values: the parameter estimates of these solutions should be identical.
THE THEORETICALLY SPECIFIED COVARIANCE STRUCTURE MODEL
MEAN PERCENTAGE OF CORRECT ANSWERS (AND STANDARD DEVIATION) ON THE DIMENSIONS OF NUTRITIONAL KNOWLEDGE
Level of convenience food consumption
During the two-day record period the energy derived from convenience food (for a hot meal) as a percentage of total energy intake was about 5% on average. Most of this energy was provided by prepared and easy-to-prepare products (about 70%), ca. 25% by take-out meals, and less then 10% by meals in a fast-food restaurant. In the sample about 60% of the respondents consumed convenience food products on at least one of the two record days. These users of convenience foods obtained about 8.5% (average over two days) of their energy from convenience products.
The average percentage of correct answers on both the definition and functional dimensions of nutrition knowledge was about 70%. For nutrition knowledge about the preparation and preservation of food and for knowledge on food items the average percentage correct answers was lower (Table 1).
Determinants of convenience food usage
The results indicate that the reliability of the indicators of the life-style construct and of the construct used to measure the personal and family situation was low. The construct reliability and the variance extracted measure indicated that the reliability of the knowledge and attitude constructs was at a satisfactory level, but the reliability of the other constructs was somewhat lower than the recommended level (0.7 and 0.5 respectively). Constructs with one indicator are assumed to be measured without error (i.e. a reliability of 1.00) (Table 2 and 3).
In the second phase of the two-step procedure the measurement model was fixed at the estimates obtained in the first step of the analysis, and the structural model was estimated (Figure 2). Convenience food usage was significantly affected by nutrition knowledge, SES, and personal and family situation of the respondent: someone who knew more about food and nutrition used less convenience food, subjects with lower SES used more than those with higher SES, and respondents who were not married, had a paid job, and were in an early stage of the family life cycle used more convenience food than their counterparts. The direct effect of SES on convenience food usage is augmented by an indirect effect, in that SES influences the usage level of convenience food products by affecting nutrition knowledge. Subjects of lower SES had less knowledge about nutritional facts than those of higher SES. Respondents who were in an early stage of the family life cycle, were employed and not married were found in the higher SES. Nutrition knowledge had a significant effect on life-style in that subjects who have more nutrition knowledge skipp main meals less frequently, have a higher BMI, are more likely to be non-smoker and to be on a diet (on their own initiative), than those who do not know much about nutrition. In the lower SES more respondents were smokers, less were on a diet (on their on initiative), had a lower BMI, and skipped meals, more frequently than respondents in higher SES. Female subjects who were not married, had a paid job and were in an early stage of the family life cycle used significantly less meals, had a lower BMI, were less likely to be on a diet and more likely to be smoker. These subjects were predominantly living in more urbanized regions.
Although no absolute threshold values are available, the overall fit measures all indicate that the model fits the data reasonably well (Table 4). In both steps of the analysis different starting values resulted in the same solution, indicating that the models were identified. Validation of the model estimated in the two-step procedure was assessed using random subsamples. The solutions were found to be stable.
Determinants of convenience food usage were assessed for female housekeepers only. Therefore, care should be taken generalizing the results to the Dutch population. Estimating the model including both male and female housekeepers did not result in dramatic changes and women make up the majority of Dutch housekeepers, so that results may be generalized to the population of main housekeepers.
Some difficulties were encountered in measuring the level of convenience food consumption. These problems were related to defining convenience food, measuring convenience food usage, and determining what measure to use in determining the level of this usage. In past research many different interpretations of convenience food have been used (Redman 1980; Reilly 1982; Gofton and Ness 1991; Meulenberg 1993), and different means have been used to indicate the degree of (convenience) food usage (Redman 1980; Haines et al. 1990; Reilly 1982; Cronin et al. 1982; Pearson et al. 1986). Measuring convenience food usage using a two-day diary has its limitations because of the short period during which consumption is measured and problems related to the coding of products consumed during the two-day period (is a pizza coded as pizza or by the ingedients used?). The convenience food measure could be improved by collecting consumption data over a longer period of time and by considering more specific convenience food products.
CONSTRUCTS, LOADINGS AND INDICATORS INCLUDED IN THE CONFIRMATORY FACTOR ANALYSIS
CONSTRUCT RELIABILITY AND VARIANCE EXTRACTED MEANSURE OF THE CONSTRUCTS INCLUDING MORE THAN ONE INDICATOR (STEP ONE)
The absence of significant effects of nutrition knowledge on food- and nutrition-related attitudes, and the low and non-significant effects of these attitudes on convenience food consumption, may be caused by the way the attitudes towards nutrition were measured and implemented in the model. The attitude constructs were measured using general statements with respect to healthy food, whereas more specific statements on convenience food (usage) would probably be better (Axelson et al. 1985). On the other hand, in many other empirical studies the relations between nutrition knowledge and attitudes (Sapp 1991; White et al. 1988), and effects of nutrition attitudes on food behaviour have been found to be non-significant (Sims 1978; Perron and Endres 1985).
Since it is probable that users of convenience food products have a poorer diet than non-users (Redman 1980; Haines et al. 1990), this group may require special attention. The aim of this study was to assess the characteristics of users of convenience foods ('who') as well as to examine in what way this consumption pattern can be amended ('how').
ESTIMATION OF THE STRUCTURAL MODEL (STEP 2)
OVERALL GOODNESS-OF-FIT MEASURES
The results show two segments with a relatively high consumption of these foods. One group consists mainly of married women of low SES who do not have a paid job and are in an advanced stage of family life cycle (older women). The second group consists mainly of women who are not married, have a relatively high SES, have a paid job and are in an early stage of family life cycle (younger women).
Concerning the first group, focusing the attention on those having a poor nutritional knowledge seems a promissing strategy to improve the nutritional quality of day-to-day consumption. Education should not be aimed solely at reduction of convenience food usage, but also include suggestions indicating how these meals can be supplemented to obtain a more complete meal. These suggestions are easier to implement and go along with the trend of increased convenience food consumption.
The model implies, however, that there is also a significant direct effect of SES on convenience food usage. This relationship cannot be influenced by an upgrade of the level of nutritional knowledge of these women.
Examination of the second segment reveals that it is clearly not a lack of these women's nutritional knowledge that is the cause of their increased convenience food usage. Their behaviour is more or less inconsistent with their level of knowledge. Reducing their convenience food consumption level cannot be accomplished by further improving their knowledge. More or less traditional programmes, aimed at a changing of attitudes and stimulating intentions to change consumption behaviour, focused especially on these women, seem more appropriate here.
AGB Fresh Foods. Beschrijvend rapport inzake de opzet en uitvoering van de Voedsel Consumptie Peiling 1992. Dongen, 1994.
Anderson, J.C. and D.W. Gerbing. Structural equation modeling in practice: a review and recommended two-step approach. Psychological Bulletin 103(3):411-423, 1988.
Axelson, M.L., D. Brinberg, and J.H. Durand. Eating at a fast-food restaurant: A social-psychological analysis. Journal of Nutrition Education 15(3):94-98, 1983.
Axelson, M.L., T.L. Federline, and D. Brinberg. A meta-analysis of food- and nutrition-related research. Journal of Nutrition Education 17(2):51-54, 1985.
Axelson, M.L. The impact of culture on food-related behavior. Annual Review of Nutrition 6:345-363, 1986.
Belante, D. and A.C. Foster. Working wives and expenditures on services. Journal of Consumer Research 11:700-707, 1984.
Courgeau, D. Family formation and urbanization. Population 44(1):123-146, 1989.
Cronin, F.J., S.M. Krebs-Smith, B.W. Wyse, and L. Light. Characterizing food usage by demographic variables. Journal of the American Dietetic Association 81:661-673, 1982.
Eppright, E.S., H.M. Fox, B.A. Fryer, G.H. Lamkin, and V.M. Vivian. The North Central Regional study of pre-school children: 2. Nutrition knowledge and attitudes of mothers. Journal of Home Economics 62:327-332, 1970.
Fornell, C. Issues in the application of covariance structure analysis: a comment. Journal of Consumer Research 9:443-448, 1983.
Gofton, L. and M. Ness. Twin trends: health and convenience in food change or who killed the lazzy housewife? Brittish Food Journal 93(7):17-23, 1991.
Grotkowski, M.L. and L.S. Sims. Nutritional knowledge, attitudes, and dietary practices of the elderly. Journal of the American Dietetic Association 72:499-506, 1978.
Haines, P.S., B.M. Popkin, and D.K. Guilkey. Methods of patterning eating behavior of American women. Journal of Nutrition Education 22(3):124-132, 1990.
Hair, J.F.jr., R.E. Anderson, R.L. Tatham, and W.C. Black. Multivariate data analysis. 3rd ed. New York: Macmillan Publishing Company, 1992.
Harten, van C. De opmars van de snelle hap. Elsevier Magazine 13th of June, 16-20, 1985.
Hulshof, K.F.A.M. and W.A. van Staveren. The Dutch National Food Consumption Survey: design, methods and first results. Food Policy 16:257-260, 1991.
Hulshof, K.F.A.M., M. Wedel, M.R.H. L÷wik, F.J. Kok, C. Kistemaker, R.J.J. Hermus, F. ten Hoor, and Th. Ockhuizen. Clustering of dietary variables and other lifestyle factors (Dutch Nutritional Surveillance System). Journal of Epidemiology and Community Health 46:417-424, 1992.
J÷reskog, K.G. and D. S÷rbom. PRELIS: A program for multivariate data screening and data summarization. 2nd ed. Mooresville: Scientific Software Inc., 1988.
J÷reskog, K.G. and D. S÷rbom. LISREL 7: Users reference guide. Mooresville: Scientific Software Inc., 1989.
Kouwenhoven, T. and O. v.d. Veer. Snack, een nieuwe eetcultuur. Voeding 51(1):10-13, 1990.
Meulenberg, M.T.G. De consument van gemak. In: Het eten van gemak. Wageningen, 1993.
Pearson, J.M., M. Capps, and J. Axelson. Convenience food use in households with male food preparers. Journal of the American Dietetic Association 86:339-343, 1986.
Perron, M. and J. Endres. Knowledge, attitudes, and dietary practices of female athletes. Journal of the American Dietetic Association 85:573-576, 1985.
Redman, B.J. The impact of women's time allocation on food expenditure for meals away from home and prepared foods. American Journal of Agricultural Economics 62:234-237, 1980.
Reilly, M.D. Working wives and convenience consumption. Journal of Consumer Research 8:407-418, 1982.
Sapp, S.G. Impact of nutritional knowledge within an expanded rational expectations model of beef consumption. Society for Nutrition Education 23(5):214-222, 1991.
Schultz, T.P. Age of individuals and family composition and income distribution. In: International Population Conference, Manila, solicited papers, vol. 3. Liege: Ordina Edition, 1981, pp. 23-48.
Schwartz, N.E. Nutritional knowledge, attitudes, and practices of high school graduates. Journal of the American Dietetic Association 66:28-31, 1975.
Shepherd, R. and G. Towler. Nutrition knowledge, attitudes and fat intake: application of the theory of reasoned action. Journal of Human Nutrition and Dietetics 5:387-397, 1989.
Sills-Levy, E. U.S. Food trends leading to the year 2000. Food Technology 43(4):128-132, 1989.
Sims, L.S. Demographic and attitudinal correlates of nutrition knowledge. Journal of Nutrition Education 8(3):122-125, 1976.
Sims, L.S. Relation of nutritional knowledge and attitude to nutrient intake. Journal of the American Dietetic Association 73:147-154, 1978.
TNO Nutrition and Food Research Institute. Occupational Toxicology and Nutrition. De ontwikkeling van een vragenlijst naar voedingskennis, by P. Schneijder, G.R.A. Paas, C. de Graaf, A. Stafleu, and M. Wedel. Report no. V91.094. Zeist, 1991.
TNO Nutrition and Food Research Institute. Occupational Toxicology and Nutrition. Socio-demographic, psycho-social, and lifestyle factors affecting consumption of convenience food, by K.S. Veenma. Report no. V94.232. Zeist, 1994.
U.S. Department of Agriculture. Economic Research Service. Take-out food in convenience stores, by C.C. Price. National Food Review, ser.11, no.4, 1988, pp.14-17.
Voorlichtingsbureau voor de Voeding. Zo eet Nederland, 1992. no. 755. The Hague, 1993, pp. 164.
Waslien, C.I. Factors influencing food selection in the American diet. Advances in Food Research 32:239-269, 1988.
White, F.D, A.V.A. Resurreccion, and D.A. Lillard. Consumer characteristics, food related attitudes and consumption of meat and meat products. Nutrition Research 8:1333-1344, 1988.
Witte, D.J., J.D. Skinner, and B.R. Carruth. Relationship of self-concept to nutrient intake and eating patterns in young women. Journal of the American Dietetic Association 91:1068-1073, 1991.