Social class differences in food consumption
The explanatory value of permissiveness and health and cost considerations
CHRISTIANNE L.H HUPKENS, RONALD A. KNIBBE, MARIA J. DROP *
Background: Middle-class people generally have healthier diets than lower-class people. Considerations that underlie
choices of foodstuffs may explain this class difference in eating habits. Qualitative studies on food beliefs show that
lower-class mothers consider health less frequently in their choice of food, while they take the preferences of family
members and expenses more often into account than their middle-class counterparts. In this paper quantitative
survey data are used to explore the explanatory power of these factors. Methods: Data on 849 women living with a
partner and at least one child in three European cities (Maastricht, The Netherlands, Liege, Belgium and Aachen,
Germany) were analysed. Results: Regarding food consumption, most but not all class differences were in line with
expectations. Class differences in food choice considerations corresponded with the findings of other studies.
Middle-class mothers were less permissive and they considered health more often and costs less often than lower-class
mothers. However, regression analyses indicated that these considerations scarcely explain class patterns in food
consumption. Conclusions: This study shows that social class differences in food consumption are hardly explained
by permissiveness, hearth or cost considerations.
Keywords: costs, food consumption, health considerations, permissiveness, social class
M
any studies in Western countries have reported social
inequalities in food consumption. Social epidemiological
and qualitative studies on food consumption patterns
show that diets in the higher social classes are more often
in line with dietary recommendations than those in the
lower classes.1"12 The study reported here explores class
differences in food consumption of women and aims to
explain these class differences.
Most studies on class differences in food choice considerations have a qualitative design. These studies confirm that there is a class difference in the extent to which
mothers take health, costs and the taste preferences of
their family into account when buying and preparing
food. Middle-class mothers mention the 'goodness' or
health value of foods more often as their primary consideration, while lower-class mothers mention the cost of
food and the preferences of their partner and children
more often as important considerations in their choice of
food. 101113 - 15
The present study aims to bridge the gap between the
large-scale population studies showing class differences in
food consumption and the mainly qualitative sociological
studies on class differences in food beliefs. This study uses
survey data on food consumption and food considerations
of mothers in three cities: Maastricht, The Netherlands,
Liege, Walloon part of Belgium and Aachen, Germany.
* C L H . Hupkens', R.A. Knlbbe'. M J Drop1
1 Department of Medical Sociology, Maastricht University, Maastricht,
The Netherlands
Correspondence: Dr. Christianne Hupkens, Statistics Netherlands,
P.O. Box 4481, 6401 CZ Heerlen, The Netherlands
Compared to qualitative studies, the present study allows
an assessment of the extent to which class differences in
food choice considerations are replicated in a relatively
large sample. In addition, whereas studies of food considerations are predominantly British, the present study is
conducted in three countries on the European continent.
Furthermore, this study examines the explanatory value
of food considerations systematically for the consumption
of a wide range of foods and snacks.
This paper explores three topics: class differences in food
consumption, class differences in food considerations and
the extent to which food considerations explain class
differences in consumption. For the analysis of class
differences in food consumption, foods and snacks were
selected which contribute largely to the intake of fat and
fibre. These nutrients are most likely associated with
healthy eating, as health promotion campaigns in the
three countries have recommended that people consume
less fat and more fibre for several years. We hypothesize
that middle-class women consume fewer snacks and foods
that contribute primarily to the intake of fat and more
foods that contain fibre than lower-class women. With
respect to the food considerations, we expected that, in
line with the outcomes of the studies mentioned earlier,
middle-class mothers consider health more often in their
choice of food, while lower-class mothers take costs and
the food preferences of their family into account more
often. We subsequently analysed whether class differences in these considerations explain class differences in
food consumption. Finally, we examined whether the
associations between social class, considerations and food
consumption are similar or different in the three cities.
Class differences in food consumption
DATA AND METHODS
Survey
Data on dietary and eating habits were collected from 849
women in Maastricht (n=304), Liege (n=316) and
Aachen (n=229) who lived with a male partner and at
least one child aged 4-14 years. The response rate was
highest in Maastricht (79%), followed by Liege (78%)
and lowest in Aachen (67%). On average, die women
were approximately 36 years old, the majority had two or
three children, the average age of their children was 8
years and nearly half of the women had a paid or unpaid
job. More detailed information on die composition of the
samples in the three cities has been described elsewhere.' 6
Questionnaire
The questionnaire consisted of two parts: a food frequency
questionnaire and a questionnaire on food practices. The
semi-quantitative food frequency questionnaire was validated on die ability to rank subjects. This questionnaire
was developed by Bausch-Goldbohm et a l . " for a large
cohort study on diet and cancer in The Netherlands. Bodi
the validity and reproducibility were tested and appeared
to be satisfactory.' 8 '' 9 The questionnaire on food practices
was based on the results of a qualitative study of Van
Otterloo and Van Ogtrop. 13 In collaboration with Van
Otterloo, questions were formulated on, for example,
meal patterns, the preparation of meals, food regimes,
health considerations and preferences. More information
on the questionnaire is reported in anorJier publication.
Data
• Social class
In general, social class is based on education, occupation
and/or income. We decided to use die women's educational level as an indicator of social class for two reasons.
First, studies which have compared these tJiree measures
for rJieir power in predicting health behaviours reported
that educational level is usually the most consistent and
reliable measure.
Secondly, in the present study the women's educational
level appeared to be a better predictor of food consumption dian four other proxies for social class, namely the
women's occupational status, rJieir partners' educational
level or occupational status and die family's income
source (a dichotomous variable: either salary or state
benefit). We compared the explanatory value of diese
proxies by regressing die food consumption variables on
die five proxies (stepwise regression). As the women's
educational level was most often included first in die
analyses, we used this variable as an indicator of social
class. Women were classified into diree social classes:
lower educational class (i.e. elementary and lower vocational training for 12-16 years old), middle educational
class and higher educational class (i.e. higher vocational
training for >18 years old and university).
• Food consumption
We computed mean daily intakes of die 120 food items
that were included in die food frequency questionnaire
by multiplying consumption frequency, number of serving
units and weight of a unit. Next, nutrient intake was
calculated using Dutch, Belgian and German food composition tables. 2} ~ 25
We grouped the 120 food items into 30 food groups. For
die present study we selected food groups which contributed considerably to die intake of fat and fibre, namely
meat products, milk products, cheese, dietary oil and fats,
brown bread (including wholemeal bread), grain, fruit
and vegetables and potatoes. In addition, we chose three
categories of snacks, namely chips, savoury snacks (such
as nuts and crisps) and sweets (including cakes and biscuits). We analysed die average number of grams die
women consumed per day for each food group.
• Considerations
Former studies on class differences in food considerations
have generally had a qualitative design. Therefore, die
formulation of die questions on considerations were based
on one of these studies. The mothers were asked how
often they took health into account when diey bought or
prepared food. Answer possibilities ranged from never (1),
dirough sometimes (2), regularly (3) and often (4) to
always (5). Three questions were formulated to measure
the extent to which the mothers considered expense
when they bought bread, meat and vegetables, with
similar answer categories as indicated above. The
reliability of the answers to diese diree questions was
satisfactory (Cronbach's a: Maastricht a = 0.85, Liege a
= 0.70 and Aachen a = 0.81) and, therefore, these
answers were combined to one five-point scale.
Permissiveness concerning food preferences was assessed
by how strictly die modiers controlled dieir children's
consumption widi respect to foods children prefer, but
which their parents generally consider less healthy,
namely white bread, sweet sandwich fillings, chips,
dessert, sweets, sweet yoghurt drinks and soft drinks. For
each of these seven food groups modiers noted whedier
diey restricted dieir children eating these foods or
whedier their children were allowed to eat them whenever diey wanted. The number of foods diat the mothers
restricted their children eating was used as a proxy for
permissiveness: the more foods the modiers restricted, die
less diey were inclined to indulge their children's preferences and, thus, die less permissive they were.
Analysis
We applied analyses of variance to test class differences
in food consumption and in the number of foods mothers
restricted their children eating and we used y} tests to test
class differences in healdi and cost considerations. In
order to examine die correlations between die diree food
choice-related considerations and the food consumption
variables, we calculated correlation coefficients.
For food groups which showed an association with both
educational level and one or more considerations, we
examined whether class differences in consumption could
be explained by die three considerations. Therefore, we
regressed each food consumption variable on educational
level, permissiveness and health and cost considerations
using two regression models. The food variables were first
EUROPEAN JOURNAL OF PUBLIC HEALTH VOL. 10 2000 NO. 2
regressed on two dummy variables for educational level:
one for middle educational level and one for high educational level (regression model 1). Secondly, the three
considerations were added to the regression equation
(regression model 2). Comparison of these models gives
insight into the explanatory power of the considerations.
All analyses were performed for all respondents together
and for each city separately. The results for all respondents
are presented in table 3 and interesting exceptions in die
different cities are described in the text.
RESULTS
Class differences in food consumption
Table 1 shows that, for almost all foods, die expected class
differences in consumption were found. Generally, differences in food consumption between lower-educated and
middle-educated women were more pronounced than
differences between middle-educated and highly-educated women. Regarding the consumption of foodstuffs
which contribute to fat, it appeared diat highly-educated
women consumed less meat and meat products, less milk
and milk products and less edible oil and fats, but more
cheese. With regard to the consumption of fibre-rich
foods, highly-educated women consumed less potatoes,
but more brown bread, grain and slightly more fruit and
vegetables than lower-educated women. Middle-educated and highly-educated women hardly differed in dieir
consumption of chips, savouries and sweets, but they
consumed less dian lower-educated women. The class
patterns in food choice were similar in the three cities,
but not of die same magnitude. Generally, class differences were largest in Maastricht and smallest in Aachen.
Class differences in food choice considerations
Table 2 shows die proportion of mothers who considered
healdi and costs often or very often in their choice of food
and die average number of foods tiiat motxiers of different
educational levels restricted their children eating. The
class differences were as expected: highly-educated modiers
considered health more often and costs less often dian
lower-educated mothers, while die latter restricted fewer
foods, indicating diat diey were more permissive.
Before including diese variables in die regression analyses, die three correlation coefficients between die considerations were inspected. Only die correlation between
healdi considerations and permissiveness was found to be
statistically significant, though not strongly so (r=0.17
and p<0.01). Modiers who considered healdi more often
were somewhat more likely to control their children's
consumption more often.
Explanation of class differences in food consumption
In order to examine whether class differences in food
consumption could be explained by die diree con-
Table 1 Mean daily consumption of food groups by educational level (standard deviation in brackets)
Food groups (g/day)
Meat products
Milk products
Cheese
Oil and fats
Brown bread
Grain
Potatoes
Fruit and vegetables
Chips
Savouries
Sweets
113
307
28
42
107
26
98
279
3
12
20
Educational level
Middle
Low
Total population
(58)
(304)
(24)
(22)
(83)
(30)
(69)
(155)
126
352
(9)
(19)
(25)
7
25
48
86
20
125
264
16
27
(64)
(391)
(28)
(23)
(83)
(32)
(86)
113
289
(161)
(12)
(25)
(33)
276
2
9
(147)
16
(21)
28
41
112
28
92
(54)
(271)
(23)
(21)
(85)
(30)
(58)
(6)
(13)
High
97
279
32
37
124
31
77
300
1
10
16
(53)
(216)
(22)
(20)
(72)
**
**
**
**
**
(27)
(47)
(159)
**
**
*
**
(4)
•*
(17)
(18)
a. Analysis of variance, linearity of class differences. * 0.01<p<0.05, *• p<0.01
Table 2 Proportion of women who consider health and costs often or very often and average number of foods women restrict their children
eating by educational level
Health
Restricted foods
Mean (SD)
n
Costs
Educational level
Low
Middle
High
50
61
69
268
348
221
53
39
28
263
339
211
2.7 (1.8)
3.6 (1.7)
3.9 (1.6)
270
349
222
All
59**"
837
41'
813
3.4 (1.8)*
841
n. number of respondents
a: X2 test, statistically significant class differences: * 0.01<p<0.O5, ** p<0.01
b: Analysis of variance, statistically significant class differences: • 0.01 <p<0 05, • • p<0 01
Class differences m food consumption
siderations, all food consumption variables in table 1 were three considerations and food consumption did not vary
included in the analysis, except in the case where the class
considerably between the cities. Generally, in Maastricht
differences in consumption were small (fruit and
and Aachen educational level and the three convegetables) or when consumption was not related to any
siderations predicted food consumption somewhat better
of the three considerations (sweets). The nine remaining
than in Liege.
consumption variables were first regressed on educational
level; subsequently, the considerations were added. DISCUSSION
Table 3 shows two regression models for each food conMost class patterns in food consumption in the present
sumption group (the dependent variable): the first model
study corresponded with the results of social epiincludes two dummy variables for educational level and
demiological studies and more qualitative studies.1"11
in the second model the three considerations are added.
Middle-class women appeared to consume less snacks and
Table 3 presents the standardized regression coefficients
fewer foods which contribute to the intake of dietary fat
of the independent variables.
and more foods which contribute to the intake of fibre.
Yet, there were two exceptions: they ate more cheese and
Comparison of the standardized regression coefficients of
less potatoes than lower-class women. They probably ate
the dummy variables for educational level between
more cheese instead of meat and more grain (like rice)
models 1 and 2 shows that in most cases these regression
instead of potatoes. In addition, the results on class differcoefficients decreased only slightly when the three conences in permissiveness and health and cost considerations were included in the regression analysis. The
siderations were in line with the results of the studies on
influence of educational level on all food consumption
food beliefs.10'1 lll3>14 Middle-class mothers considered
variables remained significant in model 2. This indicates
health more frequently and costs less often and they
that class differences in food choice were scarcely exapplied more food restrictions, indicating that they were
plained by these considerations.
less permissive than lower-class mothers.
Apart from the class differences in food consumption,
However, the results regarding the explanatory value of
women who took health into consideration consumed
these considerations for class differences in consumption
more milk, cheese, brown bread and grain and less
were not in line with expectations. Class differences in
dietary oil and fats, potatoes and chips compared to
the considerations hardly explained class patterns in food
women who considered health less often. Women who
consumption. This finding corresponded with the conclurestricted their children more often consumed less meat,
sion of a study on determinants of health behaviour
milk, potatoes and savouries and more brown bread. Comamong lower-class mothers in the UK26 that sociodemopared to women who did not consider costs, women who
graphic variables such as housing tenure were dominant
did so more often consumed more oil and fats and more
factors rather than attitudes and beliefs. The main point
potatoes.
of discussion is why class differences in food consumption
The associations of these variables with food consumpwere not explained by class differences in these contion appeared to explain only a small part of the variance
siderations, as suggested by almost all studies on food
(R2 = 3-11%). Regression analyses of the cities separately
beliefs. Three possible reasons are considered here:
showed that the associations between social class, the
Table 3 Multiple regression analyses of educational level (model 1) and of educational level and considerations regarding health, number
of restricted foods and costs (model 2) on nine food groups (standardized regression coefficients)
Independent variables
Model 1
Middle educational
level
High educational
level
Adjusted R2 (xlOO)
Model 2
Middle educational
level
High educational
level
Health
considerations
Number of restricted
foods
Costs considerations
Adjusted Rz (xlOO)
4
0.01<p<005, **p<0.01
Meat
Milk
Cheese
Oil and
fats
-0.10*
-0.10*
0.05
-0.15**
0.16**
-0.22**
3.50
-0.12**
1.00
0.11**
0.70
-0.24"
4.00
-0.07
-0.10*
0.04
-0.19"
-0.12**
0.04
0.14**
-0.11**
0.05
4.50
-0.07*
0.01
2.80
Brown
bread
Potatoes
Chips
Savouries
0.12"
-0.22**
-0.29*
-0.17**
0.21**
3.30
0.14**
1.40
-0.31**
7.50
-0.33"
9.40
-0.13**
2.20
-0.12"
0.09*
0.09*
-0.17**
-0.26**
-0.14"
0.09*
-0.19**
0.12"
0.10*
-0.25**
-0.29**
-0.09*
0.16**
-0.13**
0.11**
0.14**
-0.13"
-0.09*
-0.02
0.02
0.05
-0.04
0.08*
6.00
0.21**
-0.06
9.00
0.05
0.01
3.40
-0.10**
0.10**
10.70
-0.04
0.06
10.50
-0.09*
0.07
3.20
3.40
Grain
EUROPEAN JOURNAL OF PUBLIC HEALTH VOL. 10 2000 NO. 2
• the measurement of food-related considerations,
• the impact of social desirability and
• die extent to which the present sample allows generalization.
Widi regard to die first point, there are no validated
instruments available to measure considerations regarding health, taste and costs with the help of a questionnaire. As studies on class differences in food-related
considerations are mainly sociological studies, we based
the instruments on the findings of a qualitative study on
food regimes.13 We did not apply social psychological
instruments ' such as attitudes, beliefs and perceived
norms as we were interested in behavioural norms.
Analysis of die association between the women's food
consumption and their health and taste considerations
supports the validity of these measures. Women who took
healdi into account more often and women who were less
permissive had healthier food consumption patterns: they
consumed less fat and more fibre.
Secondly, social desirability may have biased the results.
The women may have reported opinions and practices
diey aspire to, but are not (yet) able to accomplish.
Insofar as the results of the women of bodi social classes
were coloured by social desirability, it probably would
hardly affect social class differences in food consumption
and considerations. Van Otterloo reported diat, in an
ethnographic study on food regimes, the interviewers
noted that lower-class women in particular 'felt the urge
to present themselves as competent mothers and
housewives' (p. 113). If this was also die case in die
present study, die social class differences may have been
underestimated.
Thirdly, die selection of die sample may have influenced
the results. In all surveys it appears to be very difficult to
reach the upper classes and the lowest segment of the
lower social classes. In addition, in the present study the
class contrasts were not very large. More than 80% of the
lower-class families in die three cities earned an income
and the working-class women and their partners were
mainly semi-skilled and skilled workers. A Dutch study
on class differences in food regimes13 showed diat mothers
who diemselves or dieir partner were skilled or semiskilled labourers had more in common with highermiddle-class women than motiiers from a socially more
deprived background. Therefore, die conclusions cannot
be generalized to include the most deprived and most
well-off people, but diey are valid for the majority of lower
and higher-middle social class families living in Maastricht, Liege and Aachen.
Anotiier explanation for die discrepancy between the
findings of the present study and die literature on class
differences in food choice is diat die majority of die
studies cited have been conducted in the UK. In this
country die stronger hierarchy of social classes may have
a larger impact on life style dian on die continent. As
income inequalities and the number of households living
in poverty are larger in the UK than in The Netherlands,
Belgium and Germany, it seems that class differences in
terms of material resources are larger in die UK.29 Only
further empirical research can show to what extent die
outcomes of die present study can be replicated in societies differing in class structure and/or cultural variations
between social classes.
The final question is what odier factors may explain die
class patterns in food consumption. The study of Van
Otterloo and Van Ogtrop13 indicated diat die women's
own upbringing differed between the social classes and
diat working-class women in particular took on die eating
habits diey learned at home when diey were young.
Possibly, class differences in die women's own backgrounds explain class differences in food choice. Anodier
explanatory factor was described in a recent Finnish
study.30 This study suggested diat not only healdi considerations but also the status of different foods are explanatory factors. Higher-class people distinguish themselves by choosing modem foods (which are not always
healthy) while lower-class people adhere more to traditional foods (which are not always unhealdiy). These
explanations for class differences in food choice deserve
further study.
We would like to thank all the women who completed the questionnaires for their cooperation and the Dutch, Belgian and German
dietitians and researchers who advised us on the questionnaire and
the design of the study for their constructive comments. We are
particularly grateful to Dr A H. Van Otterloo and Dr R.A. BauschGoldbohm for their valuable contributions to die development of
the questionnaire.
1
PrSttala R. Social class and food in the Nordic countries.
Ernahrungs-Umschau 1995;42:16-20.
2
Ministry of Agriculture, Fisheries and Food. National food
survey 1994. Annual report on household food consumption and
expenditure. London: HMSO, 1995.
3
Mennell S, Murcott A, Van Otterloo AH. The sociology of
food 1 eating, diet and culture. London: Sage Publications, 1992.
4
Prattala R, Berg MA, Puska P. Diminishing or increasing
contrasts? Social class variation in Finnish food consumption
patterns, 1979-1990. Eur J Clin Nutr 1992;46:279-87.
5
Heseker»H, Adolf T, Eberhardt W, et al. Lebensmittel- und
NShrstoffaufnahme Erwachsener in der Bundesrepublik
Deutschland (Food and nutrient intake of adults in the Federal
Republic of Germany). Niederkleen: Wissenschaftlicher Fachverlag
Dr Fleck, 1992.
6
Hulshof KFAM, Lowik MRH, Kok FJ, et al. Diet and other
life-style factors in high and low socio-economic groups (Dutch
Nutrition Surveillance System). Eur J Clin Nutr 1991,45:441-50.
7
De Craene I, De Backer G, Kornitzer M, et al. Determinants
of fat consumption in a general population. Rev Epidemiol Sante
1990,38:539-43.
8
Kromhout D, Doombos G, Hoffmans MDAF. voedselkeuze,
leefwijze en sterfte in relatie tot opleiding (Food choice, lifestyle
and mortality in relation to education). Tijdschrift Sociale
Gezondheidszorg 1988,66:345-8.
9
Lepage Y. Recent dietary trends in Belgiumsocio-economic aspects. In: Diehl JM, Leitzmann C, editors.
Measurement and determinants of food habits and food
preferences. Wageningen: Stichting Nederiands Instituut voor de
Voeding, 1985:109-16.
10
Charles N, Kerr M. Women, food and families.
Manchester: University Press, 1988.
11
Calnan M. Food and health: a comparison of beliefs and
practices in middle-class and working-class households. In:
Cunningham-Burley S, McKeganey NP, editors. Readings in
medical sociology. London: Routledge, 1990:9-36.
Class differences in food consumption
12
Popkin BM, Siega-Riz AM, Haines PS. A comparison of
dietary trends among racial and socioeconomic groups in the
United States N Engl J Med 1996;335:716-2O.
13
Van Otterloo AH, Van Ogtrop J. Het regime van veel,
vet en zoet: praten met moeders over voeding en gezondheid
(The regime of plenty, fat and sweet: talking with mothers on
food and hearth). Amsterdam: VU Uitgeverij, 1989.
14
DeVault M L Feeding the family: the social organization
of caring as gendered work. Chicago: The University of Chicago
Press, 1991.
15
Wilson G. Family food systems, preventive health and
dietary change: a policy to increase the health divide. J Soc Policy
1989; 18:167-85.
16
Hupkens CLH, Knibbe RA, Drop MJ. Social class differences
in women's fat and fibre consumption: a cross-national study.
Appetite 1997;28:131-49.
17
Bausch-Goldbohm RA, Van den Brandt PA, Van 't Veer P,
Sturmans F, Hermus RJJ. Results of the methodological study for
the design of a simplified, serf-administered questionnaire. In:
Riboli E, Saracci R, editors. Diet hormones and cancer,
methodological issues for prospective studies. Lyon: International
Agency for Research on Cancer, 1988:79-89.
18
Goldbohm RA, Van den Brandt PA, Brants HAM, et al.
Validation of a dietary questionnaire used in a large-scale
prospective cohort study on diet and cancer. Eur J Clin Nutr
1994;48:253-65.
19
Goldbohm RA, Van 't Veer P, Van den Brandt PA, et al.
Reproducibility of a food frequency questionnaire and stability of
dietary habits determined from five annually repeated
measurements. Eur J Clin Nutr 1995;49:420-9.
20
Pill R, Peters TJ, Robling MR. Social class and preventive
health behaviour a British example. J Epidemiol Commun Hlth
1995,49:28-32.
21
Winkleby MA, Jatulis DE, Frank E, Fortmann SP.
Socioeconomic status and health: how education, income, and
occupation contribute to risk factors for cardiovascular disease.
Am J Public Hlth 1992,82:816-20.
22
Van Berkel-Van Schaik AB, Tax B. Naar een
standaardoperationalisatie van sociaal-economische status voor
epidemiologisch en sociaal-medisch onderzoek (Towards a
standardised operationalisation of socio-economic status for
epidemiological and social-medical research). Rijswijk: Ministerie
van Welzijn, Volksgezondheid en Curtuur, 1990.
23
Nevo Tabel. Nederlandse voedingsmiddelentabel
(Dutch food composition table). Den Haag: Voorlichtingsbureau
Voor de Voeding, 1993.
24
Nubel. Belgische voedingsmiddelentabel (Belgian food
composition table). Brussels: Nubel, 1992.
25
Heseker B, Heseker H. Nahrstoffe in Lebensmitteln: die
GroBe Energie- und NShrwerttabelle (Nutrients in foodstuffs: the
large energy and nutritional value table). Frankfurt am Main:
Umschau, 1993.
26
Pill R, Peters TJ, Robling MR. Factors associated with
health behaviour among mothers of lower socio-economic status:
a British example. Soc Sci Med 1993;36.1137-44.
27
Stafleu A, De Graaf C. Van Staveren WA, Schroots JJF.
A review of selected studies assessing social-psychological
determinants of fat and cholesterol intake. Food Qual Pref
1991,3:183-200.
28
Van Otterloo AH. Taste, food regimens and fatness: a
study in social stratification. In: De Garine I, Pollack NJ, editors.
Social aspects of obesity. Luxembourg: Gordon and Breach,
1995:111-26.
29
Kooiker S, Christiansen T. Inequalities in health: the
interaction of circumstances and health related behaviour. Sociol
Hlth III 1995; 17:495-524.
30
Roos E, Prattala R, Lahelma E, Kleemola P, Pietinen P.
Modern and healthy? Socioeconomic differences in the quality of
diet Eur J Clin Nutr 1996;50:753-60.
Received 3 }une 1997, accepted 14 October J999