Content uploaded by Myriam Ertz
Author content
All content in this area was uploaded by Myriam Ertz on Oct 05, 2017
Content may be subject to copyright.
709© Academy of Marketing Science 2017
P. Rossi (ed.), Marketing at the Confluence between Entertainment
and Analytics, Developments in Marketing Science: Proceedings of the
Academy of Marketing Science, DOI 10.1007/978-3-319-47331-4_138
Spillover Effects Between Categories
of Responsible Food Products? An Extended
Abstract
Francine Rodier, Caroline Boivin, Fabien Durif, and Myriam Ertz
Introduction
Many responsible food products are available in supermarkets and other venues.
Through their purchasing behavior, consumers thus have opportunities to express
their concerns toward the society and the environment. The consistency of purchasing
behaviors in different categories may be questioned. The purpose of this research is to
explore potential spillover effects between categories of responsible food products.
Background
The literature on responsible consumption initially viewed this concept as unidimen-
sional (Kinnear et al. 1973, 1974; Webster 1975; Antil 1984; Roberts 1993), thus
suggesting that the different responsible behaviors go hand in hand. A parallel stream
of research has focused on the existence of a positive correlation between different
responsible behaviors. For instance, between recycling and waste prevention
(Thogersen 1999), purchases of organic food products and recycling (Thogersen
2004). Thogersen (2004) suggest that the tendency for a responsible behavior to spill
over on other behaviors depends on the extent to which they share motivational roots.
This paper contributes to this stream of research by testing not only links between two
responsible behaviors but potential multiple dyads within the same product category.
F. Rodier (*) • F. Durif • M. Ertz
ESG UQAM, Montréal, QC, Canada
e-mail: rodier.francine@uqam.ca; durif.fabien@uqam.ca; myriam.ertz@gmail.com
C. Boivin
University of Sherbrooke, Sherbrooke, QC, Canada
e-mail: Caroline.Boivin@USherbrooke.ca
710
Our study examines the effects of four food product categories in a model con-
stituted of feedback loops representing the different expected spillover effects. The
results enable to determine the extent to which purchases in one product category
influence those in other food categories.
Research Methods
Data was gathered in September 2015 via a self-administered e-mail survey from a
randomly selected online panel in Canada (n = 1006). The selection of respondents
was based upon preset quotas relating to age, sex, and place of residence.
The survey instrument consisted of a total of 32 scale items to measure purchase
behavior across the four product categories (i.e., locally cultivated, locally pro-
cessed, organic, and fair trade). Given the fully nonrecursive nature of the model,
these 32 items were considered as reflective of their respective latent dimension of
purchase behavior (Diamantopoulos and Winklhofer 2001). To measure consump-
tion of responsible food products, respondents were asked how often they bought a
series of responsible food products over the last 12 months on a ten-point Likert
scale ranging from 1 “never” to 10 “always.” The items are listed generically and
not by brand so that consumers may focus on the intended product category. The list
of food products contains five locally cultivated products, seven locally processed
products, twelve organic food products, and eight fair-trade products.
Principal component analysis (PCA) and exploratory factor analysis (EFA) (IBM
SPSS) using varimax rotation for the four variables together showed the factors to
be completely distinct. The number of factors was constrained to four. Ten items
exhibited low factor loadings or cross loadings on other factors and were dropped
from further analysis. The item-to-total correlations for all the remaining 22 items
were above the threshold of .5 (Nunnally and Bernstein 1994).
Results and Discussion
A maximum likelihood (ML) confirmatory factor analysis (CFA) using EQS 6.2
was conducted on the four dimensions. Non-normality of the data called for the
robust maximum likelihood (ML-r) estimation method in conjunction with ML
(Finch et al. 1997). The full model showed moderate but acceptable fit
(SB-=
()
c
203
21115 32. , standardized root-mean-square residual [SRMR] = .05, root
mean-square-error of approximation [RMSEA] = 0.067 [CI: .063; .071], non-
normed fit index [NNFI] or Tucker-Lewis index [TLI] = .937, comparative fit index
[CFI] = .945, goodness of fit index [GFI] = .845 (Steiger’s [1989] revised sample
robust GFI* = .924), and adjusted goodness of fit index [AGFI] = .806 (Steiger’s
[1989] revised sample robust AGFI* = .905)). The RMSEA value was found to be
above the .07 threshold for good fit (Steiger 2007). Although the χ2/df value of 5.5
F. Rodier et al.
711
is slightly above Jöreskog’s (1969) threshold of 5 for appropriate fit, this ratio is
interpreted with caution, since the chi-square increases with larger sample sizes
(Marsh et al. 1988), which is the case in this study. In addition to model fit, all stan-
dardized path coefficients were highly significant (p < .001) and ranged between
.698 and .906, while coefficients of determination (R2 values) varied between .5 and
.8 and whose, conjointly with model fit, Cronbach’s alphas (.824–.958) and coeffi-
cients of reliability (.825–.958) provide evidence of content validity, convergent
validity, and the constructs’ high unidimensionality.
Inter-factorial correlations ranged from .18 to .53, with the exception of the cor-
relation between organic product and fair-trade product purchase, which was .74.
Importantly, all the correlations were statistically significant, thus indicating that the
factors are fairly well related to each other. These results support convergent valid-
ity of our measures since these measures assess constructs that theoretically should
be related to each other and which are, in fact, observed to be related to each other.
The exploratory nature of the research investigating spillover effects on respon-
sible food categories presupposes that each of the four constructs may generate
reciprocal causal relationships. This implies that the model is of the nonrecursive
type, but each independent variable is simultaneously also a dependent variable.
The model is constituted of feedback loops in order to estimate spillover effects
between responsible product category purchases. However, dynamic relationships
of this nature cannot be handled by any software program (yet). Consequently, the
core model was divided into four distinctive sub-models (Byrne 2006). For exam-
ple, one sub-model estimates the effect of fair-trade purchase behavior on organic,
locally cultivated, and locally processed product purchase behavior. The same pro-
cedure was repeated for the three other constructs, and the results were then gath-
ered within a unique model framework.
The model fit was similar across all four sub-models ( SB-=
()
c
203
2846 8895. ,
SRMR = .043, RMSEA = 0.057 [CI: .053; .060], NNFI = .955, CFI = .961, GFI = .881
(GFI* = .945), AGFI = .850 (AGFI* = .931)) and satisfied Hu and Bentler’s (1999)
combined rule for good fit under maximum likelihood structural equation modeling
(CFI > .95, SRMR < 0.08, and RMSEA < 0.06). The results indicate that all link-
ages, except two (H1 and H2), have the appropriate signs providing “partial empiri-
cal validation of the theoretically assumed relationships between latent variables”
(Henseler et al. 2009: 304). Six of the twelve hypotheses are supported implying a
sound theoretical rationale of the model specification (Rose et al. 2012). The
explanatory power (coefficient of determination R2) of the constructs is as follows:
fair-trade product purchase (R2 = .63), organic product purchase (R2 = .59), locally
processed product purchase (R2 = .46), and locally cultivated product purchase
(R2 = .29). In other words, the model explained much of the variance for the fair-
trade and organic product categories but also for locally processed product and to a
lesser extent for locally cultivated product purchases.
This study sought to explore the spillover effects between purchases in four
responsible food product categories. An initial lookup on correlations between the
categories revealed strong correlations. Correlations are however typically limited
in that they do not determine causality between variables. Our study alleviates that
Spillover Effects Between Categories of Responsible Food Products? An Extended…
712
drawback by examining the effects of each of the four variables in an original
nonrecursive model constituted solely of feedback loops representing the different
expected spillover effects. The results enable to determine the extent to which pur-
chases in one product category influence those in other food categories.
Our first major result is that while there are significant links between purchases
of some responsible food product categories (namely, between organic and fair-
trade food products, locally processed and locally cultivated food products, locally
processed and fair-trade food products), spillover effects are not present between all
the categories. Following Thogersen’s (2004) argument, this may suggest that dis-
tinct motivational roots may be associated with the purchasing behavior in different
responsible food product categories.
The model explained much of the variance for the organic and fair-trade product
categories, indicating that the purchases in these two food categories reciprocally
generate the strongest spillover effects. A consumer who buys fair-trade products is
thus very likely to also purchase organic products. Conversely, a consumer who
purchases organic products also tends to buy fair-trade products, but this relation-
ship is slightly lower.
Another significant reciprocal causality that was observed relates to locally cul-
tivated and locally processed product purchases. Similarly to the organic-fair-trade
product relationship, the purchase of locally processed food is more amenable to the
purchase of locally cultivated products, whereas the reverse relationship is less pro-
nounced. It may be assumed that the purchases in these two categories share the
motivational root associated with the support of the local economy.
Though to a lesser extent another significant outcome relates to the purchase of
locally processed and fair-trade food products, this result may be partially explained
by the fact that some products could belong to the two categories (for instance,
locally processed coffee with fair-trade beans).
As mentioned earlier, some links between purchases of responsible food product
categories were not significant (namely, between locally cultivated and organic
products, locally processed and organic products, locally cultivated and fair-trade
products). The lack of relationship between purchases of local products (both
locally cultivated and locally processed) and purchases of organic products may be
explained by the presence of different motivational roots in these categories. As
purchases of local products may be motivated by a desire to support the local econ-
omy, purchases of organic products may be strongly motivated by personal health
concerns (Padel and Foster 2005). Moreover, it can be argued that locally cultivated
and organic products may be substitutes for one another (Szmigin et al. 2009).
Finally, the hypothesized feedback loops between locally cultivated product
purchases and fair-trade product purchases are not significant. This result may stem
from the fact that motivational roots to support local producers and to support pro-
ducers from developing countries may be in contradiction (Bray et al. 2010).
References Available Upon Request
F. Rodier et al.
