The Success Factors of Food Events: The Case Study of Umbrian Extra Virgin Olive Oil

Abstract

Gastronomic events are an important part of the “Made in Italy” and a vital tool for rural development. The aim of the study is to identify the complexity of the factors that determine the success of a food event. The authors, after state-of-the-art analysis of the event marketing, formulate the hypothesis of a causal model and form a data collection during the “open oil mills” in the Umbria region. Through in-depth interviews the participants were asked to express themselves about the importance and interest attributed to some aspects directly related to the theme of the event, and finally to indicate the degree of satisfaction and the likelihood of some of their future behavior. Statistical analysis based on structural equation models allowed the authors to highlight which aspects are significant in order to confirm the assumptions made. Results provide guidance for management decisions and organization of food and wine.

Keywords:

• Event marketing
• extra virgin olive oil
• food and wine festival
• SEM
• tourism
• Umbria

Notes

¹ The stakeholders event evaluation (SEE) is also part of this line of research. Developed in the early 1990s, it also considers the event with a system that can affect a multitude of stakeholders.

² The specified model also considers interrelations between exogenous variables. These are associations and relations without a causal interpretation. Neither these connections nor the final results are reported in the path diagram because the main objective of this research is the analysis of the causal links between the variables researched, explained by the assumptions specified in the text.

³ In the analysis we used the covariance-based approach instead of the variance-based approach, like PSL. Considering discussion of some of the main reference in the PLS approach instead the CBSEM (Haenlein & Kaplan, 2004; Henseler, Ringle, & Sinkovics, 2009; Dijkstra, 1983; Fornell & Bookstein, 1982, Jöreskog and Wold, 1982), this method can be considered the other side of the coin, with its own assumption and limitations. On the one hand, PLS has the advantage that do not involve distributional assumption or scale of measurement in the analysis; on the other hand, PLS raises the problem of consistency in finite sample and number of indicators. Starting from this evidence, there are several issues that lead us to consider suitable the CBSEM approach. PLS is typically recommended in situations in which the sample is small: a sample size of 253 observations allow the use of CBSEM. Many researchers recommend a minimum sample size that varies from 100 to 200 observations (Anderson & Gerbing 1988; Boomsma & Hoogland, 2001; Dillon, Kumar, & Mulani, 1987; Marsh, Hau, Balla, & Grayson, 1998; Nasser & Wisenbaker 2003). About the discussion of normality/skewness/kurtosis, it is a frequent fact that the real-word empirical data, gathered from the survey, do not have a univariate normal distribution, and so do not even have the multivariate normal distribution (Gao, Mokhtarian, & Johnston, 2008; Micceri, 1989).

⁴ The Lisrel term is the abbreviation of linear structural relations and is the name of the software developed by the Swedish statistician Karl Jöreskog and his collaborators in the early 1970s (Jöreskog, 1973; Jöreskog and Sörbom, 1996; Jöreskog and Sörbom, 1993).

⁵ Latent variables are “theoretical constructs that are not directly observable, but that have implications for the relationships between the observed variables” (Goldberger, 1972).

⁶ Since 150 is the minimum value of statistical units need for the structural equation analysis (Anderson & Gerbing, 1988), we are satisfied with the sample size obtained. The survey was carried out during the event Festivol within the activities that take place in Open Oil Mills throughout the region.

⁷ The function of the Student t statistic enables us to know if the contribution of the parameter to the model is significant or not. The null hypothesis is that the estimate of the parameter is 0. This hypothesis is rejected by Student t at level α = 0.05, the critical value of which is 1.96.

⁸ In this case, and in the case of H4, the indirect effects will be taken into consideration, which are significant at the 5% threshold.