If, in specifying your model, you fixed the loadings of the first indicator variable for each latent variable H1-H4, then it makes no sense to presume a test of whether the loading is different from zero. That would be my first guess as to what you're seeing.
David is correct. Fixing a parameter simply means that you determine instead of estimating it. Hence, the parameter gets so standard (estimation) error and, hence, no test statistic.
I guess that you forgot to fix the loadings of the first indicator variables for each latent variable, EGOV as well as H1 to H4. In this case, LISREL automatically fixes the latent variance of the independent latent variable to one and the loadings of the first indicator for the dependent latent variables also to one.
a) the fact that there is no test of fixed effects is no problem; it's an implication
b) Karin mentioned that you forget to fix one of the loadings of EGov. Hence, fix it. Problem solved.
If you talk of problems, I surely would worry a bit more about the misfit of the model which is no wonder as the model imposes very hard to belief implications....For instance, the model claims that ALL of the correlations between the indicators of EGov and all other latent (H1-H2) and observed variables are zero once EGov is controlled.
Please not that I mention that with highly constructive intentions.