E-commerce risk perception: development and validation of a measurement scale

Risk perception is considered by researchers an important construct for comprehending online shopping consumption (Bhatnagar, Misra, Rao, 2000; Olivero, Lunt, 2003; ); nonetheless, there is low agreement on how to assess this construct. In scientific papers many measurement scales to assess consumers’ risk perception in online shopping can be found; someone considers risk perception in online shopping like a single construct and uses mono-dimensional scales (i.e. Jarvenpaa, Tractinsky, 1999), someone else measures risk perception in online shopping by using multidimensional scales (i.e Forsythe, Shi, 2003). The goal of this paper is to describe psychometric proprieties of a scale which was used to measure e-Commerce Risk Perception, operationalised as risk perception of events that may occur to an online consumer. A list of eleven negative events was created by reviewing literature and interviewing nine expert pc users and online consumers. These events refer to three areas: “Events caused by consumer incompetence (CI)”, “Events caused by seller mistakes (SM)” and “Events caused by fraudulent behaviors (FB)”. The eleven events were used to evaluate: a) the degree of risk perception ;b) the perception of probability of harm; c) the perception of severity of consequences. A six point Likert-scale was used. We tested the hypothesis that the in each aspect of risk perception (RISK, PROBABIL and CONSEQ) the eleven events are structured in the three factors we described (CI, SM and FB). Concerning method, 501 adult Italians answered the questionnaire (55% Males; age range: 18 - 64 y.o., average = 28.00 y.o., SD=7,36. Psychometric properties were examined by measuring internal consistency, through Cronbach’s alpha, and internal validity, through explorative factor analysis (with maximum likelihood and oblimin rotation methods). Concerning results, Kaiser criterion was used to determinefactors to retain. A two factor model was found for RISK and PROBABIL scales. Concerning CONSEQ scale, a three factor model was found; it does not adequately represent the three expected areas, but matches the issues at risk: delivered goods or personal data. The minimum alpha index is .72 and the maximum is .89.