Based on what you’ve described here, it sounds like you will need to adjust the provided sampling weights. Specifically, the sampling weights allow for representative population estimates for the given survey year. So, if you pool together several single-year ACS files and apply the sampling weights, you’ll calculate population size estimates that are many times larger than the actual population. An approximate adjustment is to simply divide the sampling weight by the number of samples you are pooling together. So, in your case, generate a new weight equal to PERWT/8. The sub-setting of the data that you suggest should not dramatically influence this weighting methodology.