By the mean scores. If in the same range, based on the litterature, one result validates the other. If they are different, you're in trouble. You have to compare items (or factors if any) to try and find similarities. If you can't, go back to the literature to kill one result.

You sum the satisfied and give them a score of ( 1 ) and you sum the disatisfied and give them a score of ( 2 ) . You do the same with the second scale . You could also do a correlation between the two scales to justify how storng your original scale is .

You need to calculate the TOTAL score for each scale. and you can run Pearson correlation between the 2 scales at the different times (Waves). This will give you the strength of the relashionship and if it is significant. Hope this is helpful, good luck

You can't really compare the two scales directly. The first scale is labelled with relative statements ("very")  while the second has absolute anchors ("completely"), and the first scale is labelled while the second scale is numeric. You can probably come up with a rough and ready comparison by just comparing how the two distributions overlap and assigning scores from one scale as equivalent values in the other, but this won't be much use for comparing over time.

I agree with Conal Smith who says that you can't really compare the two scales directly. However, if for one year at least you have the two measures (look at different sources too) you can build a bridge between them, identifying reasonable cutpoints on the cumulative distribution.

Maybe the following paper which we recently published is of interest to you. In this paper we describe how we derived a reference distribution from the responses to a numerical scale and used this reference distribution to decide at which point verbally labelled response options transit from one state to another, for example from ‘satisfied’ to ‘very satisfied’. Next, these transition points can be used to estimate a population mean for each wave of the time series in which the verbal scale is used on a level that is comparable to that of the mean of the reference distribution. These estimates are appropriate for use in an extended time series based on the responses measured using a verbal and a numerical scale.  

Article Conversion of Verbal Response Scales: Robustness Across Demo...

Thank you to all of you for your help and advice. 

Thanks especially to you, Tineke! After a quick look at your paper, I think it might indeed be helpful for my study.

As a co-author of the paper of Tineke de Jonge,  I have no additional contribution

Conversion of Verbal Response Scales and still trying to determine the robustness across demographic categories is likely to give you dubious results if you have not already hypothesized the ideal situation where you give maximum marks