When calculating PLS consistent algorithm with SmartPLS, some path coefficients could be higher than 1. Why is this? Is there any problem? Does it have a solution?
Are you sure that the coefficients are: (a) standardized; and (b) represent correlations rather than regression coefficients?
Both unstandardized relationships (e.g., covariances) and regression coefficients, whether z-score standardized or not, may exceed an absolute value of 1.
As well, have a look at these two links for some additional explanation:
https://www.statmodel.com/download/Joreskog.pdf
Deegan, J. (1978). On the occurrence of standardized regression coefficients greater than one. Educational and Psychological Measurement, 38 (4), 873-888. doi:10.1177/001316447803800404