Adoption of Online Grocery Shopping: Personal or Household Characteristics?

ABSTRACT

This article analyzes how sociodemographic characteristics impact the adoption of online grocery shopping, and relies on the Motivation-Opportunity-Ability (MOA) model to explore what these sociodemographics actually capture and how they are linked with consumer motivations. The researchers exploit a survey among 468 customers of Belgian supermarket chain Colruyt. Their logistic regression shows that while variables at the personal level do affect adoption of the online channel, consumers’ motivations to adopt in fact lie on the household level. In particular, the effect of age disappears or becomes less strong when it is combined with household characteristics. An examination of respondents’ self-reported motivations confirms that age does not only capture a person’s ability to use the technology but also its usefulness for that person’s household, in that age is correlated with the presence of young children and the working situation in the household.

KEYWORDS:

• Consumer adoption
• online grocery shopping
• sociodemographics

Notes

According to the definition of the European Commission (2015), online groceries include food, drinks, and other necessities of daily use, such as sanitary products, cosmetics, tobacco products, and flowers; in other words, fast-moving consumer goods usually bought in the supermarket.

Goethals and colleagues (2012) alluded to a link between age and the number of children, but did not find any proof. The differences in adoption between the age groups in their sample could not be explained by the number of (young) children in the household.

In most groups, sample size was large enough. When this was not the case, a non-parametric test was performed as a robustness check. Unless mentioned otherwise, these tests yielded similar results.

Although one-way ANOVA has proven to be robust with respect to violations of the assumptions, there is one exception: the case of unequal variances combined with unequal sample sizes. When the researchers encountered such a situation, they again performed non-parametric tests as a robustness check. Again, unless mentioned otherwise, they obtained similar results.

In order to test the significance of Pearson correlation coefficients, the sampling distribution of the variables must be normally distributed. Not all cases passed the normality check. Therefore, non-parametric tests (Spearman’s correlation coefficient and Kendall’s tau) were performed as robustness checks. Unless mentioned otherwise, these tests yielded similar results.

Detailed results can be obtained from the corresponding author.