The changes in purchase intention are measured by the difference between PI (the original PI) and INT (the PI after food safety scandal).H1a proposes that the PI would be reduced after food safety scandal.The result reveals that the p values are less than .05 of Pairs 1 to 6(Table 3). and therefore, the H1a is supported.Next, the model is tested with structural equation modelling. After deleting a number of items, the final structural model fit (Figure 2) is acceptable with fit statistics: CMIN = 1016.754, DF = 343, p = .000;CMIN/DF = 2.964; goodness‐of‐fit index (GFI) = 0.897; normed fit index (NFI) = 0.898; CFI = 0.928; and RMSEA = 0.055. In addition, modification indices showed no critical problems of misfit, and the individual relationships between constructs of interest were thus examined in the model. The hypothesis testing results are illustrated in Table 4.With the exception of H1e, the p values of the hypotheses H1b to H1i are below .05, thus the relationships are significant. Therefore, thehypotheses from H1b to H1i were supported with the exception of H1e.