Escolar Documentos
Profissional Documentos
Cultura Documentos
Appetite
journal homepage: www.elsevier.com/locate/appet
Research report
Department of Food and Agri-Business Management, University of Ioannina (UoI), George Seferi 2, GR-30100, Agrinio, Greece
MAPP, Department of Marketing and Statistics, Aarhus School of Business (ASB), University of Aarhus, Haslegaardsvej 10, DK-8210 Aarhus V, Denmark
c
Istituto Nazionale di Ricerca per gli Alimenti e la Nutrizione (INRAN), via Ardeatina 546, IT-00178, Roma, Italy
b
A R T I C L E I N F O
A B S T R A C T
Article history:
Received 14 May 2008
Received in revised form 17 September 2008
Accepted 20 September 2008
Recognising the need for a more statistically robust instrument to investigate general food selection
determinants, the research validates and conrms Food Choice Questionnaire (FCQs) factorial design,
develops ad hoc a more robust FCQ version and tests its ability to discriminate between consumer
segments in terms of the importance they assign to the FCQ motivational factors. The original FCQ
appears to represent a comprehensive and reliable research instrument. However, the empirical data do
not support the robustness of its 9-factorial design. On the other hand, segmentation results at the
subpopulation level based on the enhanced FCQ version bring about an optimistic message for the FCQs
ability to predict food selection behaviour. The paper concludes that some of the basic components of the
original FCQ can be used as a basis for a new general food motivation typology. The development of such a
new instrument, with fewer, of higher abstraction FCQ-based dimensions and fewer items per dimension,
is a right step forward; yet such a step should be theory-driven, while a rigorous statistical testing across
and within population would be necessary.
2008 Elsevier Ltd. All rights reserved.
Keywords:
New food motivation typology
Conrmatory factor analysis
Cluster analysis
Introduction
Food product choice is a complex function of preferences for
sensory characteristics, combined with the inuence of nonsensory factors, including food-related expectations and attitudes, health claims, price, ethical concerns and mood (Prescott,
Young, ONeil, Yau, & Stevens, 2002). A variety of social, cultural
and economic factors thus contribute to the development,
maintenance and change of dietary patterns. Intra-individual
determinants, such as physiological and psychological factors,
acquired food preferences, and knowledge can be distinguished
from interpersonal or social factors, such as family and group
inuences (Eertmans, Baeyens, & Van den Bergh, 2001).
Comprehensive models have been developed to sketch out the
way people construct the process of choosing foods in general. For
example, Frust, Connors, Bisogni, Sobal, and Winter Falk (1996)
group the factors involved in food choice into three major
components (life course, inuences and personal systems), the
particular relationship of these components to one another
generating the process or pathway leading to the point of choice.
* Corresponding author.
E-mail address: taspay@gmail.com (A. Pagiaslis).
0195-6663/$ see front matter 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.appet.2008.09.014
1
For a comprehensive review of the concept of human motivation, as well as its
application on the food-related research see Pincus (2004) and Eertmans, Victoir,
Notelaers, Vansant, and Van den Bergh (2006).
200
201
with points as follows: 1 = extremely unimportant, 2 = unimportant, 3 = slightly unimportant, 4 = neither unimportant
nor important, 5 = slightly important, 6 = important and 7 =
extremely important.
It has been decided to use a bipolar 7-point scale instead of the
original unipolar 4-point one for two reasons: rst, to increase the
number of response options within items (i.e. from 4 to 7) and
thereby the opportunities to increase the variability of the
measurement scale, that is to increase the mere quality of the
scale (DeVellis, 1991); second, to take into account a central
neutral point as a means of avoiding a forced preference on behalf
of the subjects (DeVellis, 1991). The food choice decision-making is
a process well known for its complexity and, as such, it naturally
requires a substantial number of response options to allow
respondents more latitude in describing their level of opinion.
The use of 7-point Likert-type scales is not unseen in the food
choice motivation literature. Martins and Pliner (1998) report the
development of a 32-item FCQ-based scale (called Food Motivation
Scale) measured on 7-point Likert-type importance scales.
Table 1
Samples socio-demographic prole (%, N = 997).
1: Central (8%), Crete (8%), South-Central (4.7%), West (4.0%); 2: Temporarily unemployed, pupil/student, pensioner; 3: Entrepreneur or CEO (4.6%), worker/technician (4.4%),
pupil/student (2.8%), temporarily unemployed (1.6%), farmer (0.9%).
202
Moreover, Prescott et al. (2002) report the use of the FCQ measured
on 7-point Likert-type agreement scales. In both of these surveys,
the 7-point scales show satisfactory statistical properties; however, no effort has been made in either study to conrm the
statistical robustness of the FCQs factorial design when measured
on 7-point scales.
The descriptive statistics for the FCQ can be found in Table 2.
The scores on each of the nine FCQ dimensions are computed by
averaging item ratings per dimension.
Data analysis
Conrmatory factor analysis (CFA) is performed (LISREL 8.72) to
conrm and validate the factorial pattern suggested for the FCQ by
Steptoe et al. (1995). Many authors agree in the importance of a
priori theory before implementing CFA (Hurley et al., 1997),
whereas the use of exploratory factor analyses (EFA) is appropriate
for new or ad hoc measures (Conway & Huffcutt, 2003). In this
study, the measurement instrument is established a priori through
the FCQ. Consequently, it is meaningful to directly conrm its
original factorial design trough CFA.
The internal consistency of the various constructs is assessed by
Cronbach alpha coefcients (SPSS 15.0).
The general CFA models t is assessed by: (a) the chi-square
test as a descriptive goodness-of-t index for nested models; and
(b) the Comparative Fit Index (CFI), the Non-Normed Fit Index (TLINNFI) and the root mean square error of approximation (RMSEA).
To evaluate the parsimony of the hypothesized models, the
Parsimonious Normed Fit Index (PNFI) is used to compare nested
and non-nested 1st and 2nd-order models. Finally, the chi-square
difference test, the Akaikes Information Criterion (AIC) and the
Consistent Akaikes Information Criterion (CAIC) are applied to
compare two or more competing models estimated from the same
data.
Hierarchical cluster analysis (HCA) is performed at the nal
stage of the analyses (SPSS 15.0) to determine a number of
consumer segments (clusters) with different level of importance
assigned to the enhanced FCQs motivational factors, after the
enhanced FCQ models robustness being tested at the previous
stage of analyses. The enhanced FCQ motivational factors are thus
used as clustering criteria. Existence and number of the
statistically signicant differences among the emerging clusters
in terms of the importance assigned to the FCQ motivational
factors is assessed through post-hoc one-way ANOVA tests
(p < 0.05) performed at the cluster level. The selection of the
ideal clustering scenario (i.e. ideal number of clusters) is based on
the above statistically signicant differences, as well as on the
convergence between the hierarchical and the k-means solutions
per clustering scenario (Pearson correlations, p < 0.01). Finally,
statistically signicant socio-demographic differences among the
motivational clusters emerged are assessed through chi-square
tests at the cluster level (p < 0.05).
Analysis and results
Scale reliabilities and item statistics
As hypothesized in the original FCQ, the nine motivational
dimensions show moderate to good reliability (Cronbach alphas
from a = 0.61 to a = 0.82; see Table 2), with the exception of the
ethical concern dimension (a = 0.30), whereas the reliability of
the overall FCQ typology is also very high (a = 0.93). Not
surprisingly, the items 20, 32 and 19, which form the lowreliability ethical concern dimension, show the lowest item-tototal correlation (r). Correlations lower than 0.40 appear also in
Table 2
Descriptive statistics and reliabilities of the original FCQ with 7-point scales
(N = 997).
No.
FCQ item
Mean/S.D.
ra
5.77/1.08
5.92/1.06
5.83/1.09
5.44/1.27
5.00/1.24
5.49/1.09
0.62
0.63
0.61
0.40
0.36c
0.49
4.76/1.36
6.12/1.12
4.97/1.36
5.57/1.11
5.11/1.30
5.29/1.19
0.43
0.31c
0.53
0.41
0.53
0.61
5.12/1.42
5.18/1.34
5.19/1.37
5.41/1.20
0.56
0.57
0.61
0.36c
5.47/1.20
0.41
5.58/1.15
5.01/1.31
5.76/1.10
5.85/1.10
0.51
0.28c
0.52
0.50
6.03/1.12
5.76/1.08
5.88/1.17
0.58
0.59
0.66
5.50/1.18
5.42/1.22
5.73/1.06
0.55
0.58
0.49
5.14/1.31
5.10/1.24
5.36/1.32
0.74
0.67
0.60
5.46/1.10
4.93/1.45
4.99/1.41
0.32c
0.44
0.51
4.60/1.63
5.60/1.21
5.27/1.13
0.09c
0.19c
0.24c
r: 0.93
Cronbach a: 0.771
2. Mood
16
Helps me cope with stress
34
Helps me cope with life
26
Helps me relax
24
Keeps me awake/alert
13
Cheers me up
31
Makes me feel good
Cronbach a: 0.736
3. Convenience
1
Is easy to prepare
15
Can be cooked very simply
28
Takes no time to prepare
35
Can be bought in shops close to where
I live or work
11
Is easily available in shops and supermarkets
Cronbach a: 0.742
4. Sensory appeal
14
Smells nice
25
Looks nice
18
Has a pleasant texture
4
Tastes good
Cronbach a: 0.668
5. Natural content
2
Contains no additives
5
Contains natural ingredients
23
Contains no articial ingredients
Cronbach a: 0.779
6. Price
6
Is not expensive
36
Is cheap
12
Is good value for money
Cronbach a: 0.772
7. Weight control
3
Is low in calories
17
Helps me control my weight
7
Is low in fat
Cronbach a: 0.820
8. Familiarity
33
Is what I usually eat
8
Is familiar
21
Is like the food I ate when I was a child
Cronbach a: 0.613
9. Ethical concern
20
Comes from countries I approve of politically
32
Has the country of origin clearly marked
19
Is packaged in an environmentally friendly way
Cronbach a: 0.304
Key: Bold characters correspond to the highest means among FCQ variables and all
r > 0.60.
a
Item-total correlation.
b
7-point bipolar scale with end-points 1 = extremely unimportant and
7 = extremely important.
c
FCQ items with r < 0.40.
1131.45
224
0.064
0.96
0.78
0.96
1283.45
1732.21
4360.38
484
0.090
0.92
0.84
0.92
6164.74
6619.41
4218.11
480
0.088
0.93
0.83
0.92
5977.78
6456.06
6010.74
1403.08
203
5815.78
CFA6 vs. CFA7: 194.96
Chi-sq (5, 0.001): 20.52
B. 2nd-order CFAs
CFA6: 33 items
(excl. it. 19, 20, 32)
2 2nd order factors 8
1st order factors
CFA7: 33 items
(excl. it. 19, 20, 32) 1
2nd order factors
8 1st order factors
C. Adapted CFA
3791.16
467
0.085
0.94
0.84
0.93
5422.73
5977.77
4036.48
517
0.083
0.94
0.81
0.93
5818.72
6485.96
4179.76
542
0.082
0.94
0.80
0.93
6002.25
6734.44
4473.45
558
0.084
0.93
0.82
0.92
6365.89
7003.60
SBS Chi-sq
Df
RMSEA
CFI
PNFI
TLI-NNFI
AIC
CAIC
4080.64
524
0.083
0.94
0.82
0.93
5875.96
6501.87
4435.96
5592.72
CFA4 vs. CFA5: 245.32
Chi-sq (50, 0.001): 86.66
5754.25
CFA3 vs. CFA4: 161.5
Chi-sq (25, 0.001): 52.62
CFA3 vs. CFA5: 1318.29
Chi-sq (75, 0.001): 118.60
5663.96
CFA2 vs. CFA3: 90.2
Chi-sq (18, 0.001): 42.3
CFA2 vs. CFA4: 71.2
Chi-sq (7, 0.001): 24.3
6149.89
N chi-sq
D(N chi-sq)
CFA4: 35 items
(excl. it. 20)
8 factors 2 items
CFA3: 36 items
8 factors 3 items
CFA2: 35 items
(excl. item 20)
9 factors
CFA1:
36 items
9 factors
A. 1st-order CFAs
Goodness-of-t index
2nd-order CFAs
The presence of substantial correlations at the 1st-order factor
levels of CFAs 15 might point out to the existence of higher-order
factors (Fabricar, Wegener, MacCallum, & Strahan, 1999). A
hierarchical sequence of nested partial 2nd-order CFAs is therefore
implemented (CFAs 6 and 7), based on the previously described
1st-order CFA5 model and its estimated correlations.
Table 3
Progress of CFAs conducted on the original FCQ with 7-point scales (N = 997).
1st-order CFAs
Even though the non-normality of the data is only slight, the
selected method of CFA model estimation is Maximum Likelihood
(LISREL 8.72) with its robust correction for non-normality (Robust
Maximum Likelihood RML; Satorra & Bentler, 1994). The t of the
original 36-item CFA measured in 7-point scales hereafter called
CFA1 is marginally accepted (Table 3A). Moreover, most of the
standardized factor loadings result adequate, ranging from 0.47 to
0.87 (cut-off levels from 0.50 to 0.95 are adequate to assess
convergent validity; Kline, 2005). However, the loadings of some
items are quite low (e.g. item 16 = 0.39, item 25 = 0.40 or item
35 = 0.42), while a zero correlation appears for item 20: comes from
countries I approve of politically. The latter result should be
expected for the ethical concern dimension because of its
relatively low reliability a and low item-total correlation r that
affect its internal validity (see Table 2). Furthermore, many of the
estimated correlations at the factor level are higher than 0.85, not
satisfying CFA1s discriminant validity (Kline, 2005).
Based on the above outcome of CFA1, the most natural step
forward is to focus on the problematic ethical concern
motivational dimension. A hierarchical sequence of nested CFAs
is then implemented, as follows: (a) by excluding item 20 due to its
zero correlation with the ethical concern dimension (CFA2); (b)
by separating the ethical dimension into the three items it is
comprised of (i.e. items 20, 32 and 19; CFA3); (c) by combining a
and b above (i.e. excluding item 20 and considering items 32 and
19 as separate ethical concern-related sub-dimensions; CFA4); and
(d) by excluding the three items that from ethical concern all
together (CFA5). Nevertheless, all models goodness-of-t indices
remain marginal (see Table 3A), whereas many of the estimated
correlations at the factor level are still higher than 0.85, not
satisfying discriminant validity (correlation matrices of CFAs 15
available upon request).
CFA5: 33 items
(excl. it. 19, 20 32)
8 factors
204
Fig. 1. Path diagram, adapted CFA8, 24-item FCQ with 7-point scales (N = 997).
2
A two-item ethical concern dimension (by excluding the clearly unrelated
item 20) can be maintained in an adapted 26-item FCQ version, with equally
satisfactory goodness-of-t indices but worse discriminant validity (results
available upon request). However, as postulated by the Lindeman and Vaananen
(2000) work and conrmed by the present results, the ethical concern dimension
as formulated in the original FCQ version remains problematic and its inclusion in a
food choice motivation typology merits careful consideration.
1. H
2. M
3. C
4. SA
5. NC
6. P
7. WC
8. F
1.H
2. M
3. C
4. SA
5. NC
6. P
7. WC
8. F
1.00
0.50
0.50
0.88*
0.95*
0.65
0.46
0.39
1.00
0.54
0.62
0.50
0.53
0.40
0.68
1.00
0.41
0.36
0.52
0.52
0.46
1.00
0.83
0.59
0.23
0.46
1.00
0.62
0.46
0.36
1.00
0.37
0.48
1.00
0.27
1.00
*: Represents correlations between FCQ factors larger than 0.85 that do not satisfy
discriminant validity. Key: H: health; M: mood; C: convenience; SA: sensory appeal;
NC: natural content; P: price; WC: weight control; F: familiarity.
205
Table 5
Motivational cluster proles, statistically signicant differences (%, N = 997).
Key: Bold characters correspond to the higher gures and italicised gures correspond to the lower gures among clusters. 1: 7-point bipolar scale with end-points
1 = extremely unimportant and 7 = extremely important.
*
Duncan and Scheffe post-hoc paired ANOVA tests, p < 0.05.
**
Chi-square tests, p < 0.05.
+
Not statistically signicant differences in cluster pair-wised comparisons, ANOVA tests, p < 0.05.
206
Table 6
Process of the hierarchical cluster analysis, scenarios of 47 clusters (N = 997).
Key: H: health; M: mood; C: convenience; SA: sensory appeal; NC: natural content; P: price; WC: weight control; F: familiarity. 1: Duncan and Scheffe post-hoc ANOVA test
between each pair of clusters in each scenario in terms of importance assigned to the eight motivational factors (p < 0.05). 2: Correlation between the hierarchical cluster
membership and the k-means cluster membership variables (k-means cluster analysis implemented on hierarchical cluster centroids) (p < 0.01).
Discussion
For the participants to the present research, the FCQ appears to
represent a realistic typology of general food selection-related
motives and a comprehensive and easily administrated research
instrument, as concluded by the high mean importance scores
assigned to individual motives, as well as to the overall
instrument-usually higher than 5 in the 17 scale. Compared to
the results from the normative Steptoe et al. (1995) sample, the
Greek consumers seem to attach high importance to more or less
the same motivational dimensions (i.e. convenience, health,
sensory appeal and price); however, the importance assigned
to foods natural content is the highest among all motives. This
result could possibly be the outcome of more than a decade of food
scares since the original Steptoe et al. (1995) research was
conducted, a time period that has shifted consumer attention
closer to food safety-related issues internationally.
Except for the ethical concern dimension, the FCQ also
appears to be a quite reliable research instrument both in parts, as
well as a whole. However, compared to the values reported by
Steptoe et al. (1995) for their normative sample, in the present
sample almost all motivational scales (except for weight control)
show different (lower) reliability indices.3 Overall, the ndings of
the present research conrm the observation that the reliability
values for the UK sample reported by Steptoe et al. (1995) tend to
be higher than those for other western populations, such as the
Canadian, Italian and Belgian samples, as reported by Eertmans
et al. (2006). Nevertheless, this deviation from the original does not
3
Equality of reliabilities of the FCQ scales is tested between the present sample
and the normative Steptoe et al. (1995) sample with the method of Van de Vijver
and Leung (in Eertmans et al., 2006), that is . . .by computing (1 a1)/(1 a2),
where a1 is the reliability obtained for the rst sample and a2 the reliability of the
second sample. When the resulting value exceeded the critical value of the F distribution
with numerator df1 = N1 1 and denominator df2 = N2 1 (N1 being the size of the rst
sample and N2 the size of the second sample), the hypothesis of equal reliabilities was
rejected . . . Equality was tested at the p < 0.01 signicance level. . . (Eertmans et al.,
2006, p. 346). For example, regarding the internal consistency of sensory appeal,
the comparison yielded a value of (1 0.66)/(1 0.72) or 1.21, which exceeded the
critical F-value at the p < 0.01 level (with df1 = 996 and df2 = 357).
207
208
References
Ares, G., & Gambaro, A. (2007). Inuence of gender, age and motives underlying food
choice on perceived healthiness and willingness to try functional foods. Appetite,
49, 148158.
Biloukha, O. O., & Utermohlen, V. (2000). Correlates of food consumption and perceptions of food in an educated urban population in Ukraine. Food Quality and
Preference, 11, 475485.
Byrd-Bredbenner, C., & Abbot, J. (2008). Food choice inuencers of mothers of young
children: implications for nutrition educators. Topics in Clinical Nutrition, 23, 198
215.
Chryssohoidis, G., Krystallis, A., & Perreas, P. (2007). Ethnocentric beliefs and country-oforigin (COO)-effect: impact of country, product and product attributes on Greek
consumers evaluation of food products. European Journal of Marketing, 41, 15181544.
Conway, J. M., & Huffcutt, A. I. (2003). A review and evaluation of exploratory factor
analysis practices in organizational research. Organizational Research Methods, 6,
147168.
DeVellis, R. F. (1991). Scale development. Theory and applications. Applied social research
methods series (Vol. 26). Newbury Park, CA: Sage Publications.
Eertmans, A., Baeyens, F., & Van den Bergh, O. (2001). Food likes and their relative
importance in human eating behaviour: review and preliminary suggestions for
health promotion. Health Education Research, 16, 4430456.
Eertmans, A., Victoir, A., Notelaers, G., Vansant, G., & Van den Bergh, O. (2006). The Food
Choice Questionnaire: factorial invariant over western urban populations? Food
Quality and Preference, 17, 344352.
Fabricar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. (1999). Evaluating the
use of exploratory factor analysis in psychological research. Psychological Methods,
4, 272299.
Frust, T., Connors, M., Bisogni, C. A., Sobal, J., & Winter Falk, L. (1996). Food choice: a
conceptual model of the process. Appetite, 26, 247266.
Glanz, K., Basil, M., Maibach, E., Goldberg, J., & Snyder, D. (1998). Why Americans eat
what they do: taste, nutrition, cost, convenience and weight control concerns as
inuences of food consumption. Journal of the American Dietetic Association, 98,
11181126.
Hurley, A. E., Scandura, T. A., Schriesheim, C. A., Brannick, M. T., Seers, A., Vandenberg,
R. J., et al. (1997). Exploratory and conrmatory factor analysis: guidelines, issues,
and alternatives. Journal of Organizational Behavior, 18, 667683.
Kline, B. R. (2005). Principles and practice of structural equation modelling (second
edition). NY: The Guilford Press.
Krystallis, A., & Chryssohoidis, G. (2006). Does the Country-of-Origin (COO) of Food
Products Inuence Consumer Preferences? An Empirical Examination of Ham and
Cheese Products in the Greek Market. 29 June2 July, Chania, Crete.
Lindeman, M., & Vaananen, M. (2000). Measurement of ethical food choice motives.
Appetite, 34, 5559.
Lockie, S., Lyons, K., Lawrence, G., & Mummery, K. (2002). Eating Green: motivations
behind organic food consumption in Australia. Sociologia Ruralis, 42, 2340.
Martins, Y., & Pliner, P. (1998). The development of the Food Motivation Scale. Appetite,
30, 94.
Pincus, J. (2004). The consequences of unmet needs: the evolving role of motivation in
consumer research. Journal of Consumer Behaviour, 3(4), 375387.
Pollard, T. M., Steptoe, A., & Wardle, J. (1998). Motives underlying healthy eating: using
the Food Choice Questionnaire to explain variation in dietary intake. Journal of
Biosocial Science, 30, 165179.
Prescott, J., Young, O., ONeil, L., Yau, N. J. N., & Stevens, R. (2002). Motives for food
choice: a comparison of consumers from Japan, Taiwan, Malaysia and New
Zealand. Food Quality and Preference, 13, 489495.
Satorra, A., & Bentler, Q. M. (1994). Corrections to test statistics and standard errors in
covariance structure analysis. In A. Von Eye & C. C. Clogg (Eds.), Latent variable
analysis: applications for developmental research (pp. 399419). Thousand Oaks, CA:
Sage Publications.
Steenkamp, J-B.E.M., & Baumgartener, H. (1998). Assessing measurement invariance
in cross-national consumer research. Journal of Consumer Research, 25, 78
90.
Steptoe, A., Polland, T. M., & Wardle, J. (1995). Development of a measure of the motives
underlying the selection of food: the Food Choice Questionnaire. Appetite, 25, 267
284.
Steptoe, A., & Wardle, J. (1999). Motivational factors as mediators of socio-economic
variation in dietary intake patterns. Psychology and Health, 14, 391402.