I hate to jump in to this because it isn't my main "lane", but I have supervised theses based on statistical reasoning, though mostly about learning outcomes of pedagogy.
But, I suggest that your question isn't ready for prime time and you need to look at those prior studies to get a better idea about what is going on. They can't all be correct if they can be compared. But perhaps they can't be compared at all.
In particular, I'd ask if the background is the same in the various studies. An outcome in one region of the world might be different than in another. Are the definitions of terms in the studies the same? I have my doubts.
I'll suggest, also, that you first need to examine why those other studies might have come to different results and also examine any assumptions that were made. The next step, before you could form an hypothesis is to look to which set of studies has an underlying structure most similar to yours.
And, of course, the very nature of the question is a bit suspect. Were these before-after studies of the situation (before the laws, then after)? Or are they just guesses/opinions and not true "studies" with a scientific basis. This suggests that you need to get a really firm grasp of the methodologies of the various studies and whether that affected the outcomes in some way.
Something is wrong. First figure out what. Then you have the basis for making an hypothesis that might be tested and finding a methodology for doing so. Otherwise it seems to be nothing more than spinning wheels; going through the motions. That gives you only meaningless "answers".
Sorry to seem harsh, but social science gets a bad name when it is done poorly.
Replying to the changed question. I'd suggest that you make hypotheses about A and B separately, not about Y. If you can properly test those, perhaps quantitatively then you can, perhaps form conclusions about Y from the results. For example "H: Given treatment T, A will decrease by at least 10%" or something like that.
If done carefully you avoid the dilemma, but it still requires figuring out why the previous studies diverge. That might help you quantify the hypothesis also.