In order to investigate discriminant validity, the method proposed by Fornell and Larcker (1981) is initially applied. According to this method, the extracted variance for each individual construct should be greater than the squared correlation (i.e., shared variance) between constructs. An examination of Table 2 shows that the extracted variance for each of the constructs exceeds the squared correlation. Moreover, to further test discriminant validity, the baseline measurement model is compared with alternative models where covariances between pairs of constructs are constrained to unity (Anderson and Gerbing, 1988). In every case, the restricted model has a significant (p