I guess no one here has ever heard of weighted summation. Oversampling can be done within individual sectors, then weighted appropriately when added to the overall pool of participants to keep from skewing results. Pew discusses it on their web site. It's not difficult, and it reduces the errors for smaller sections of the population. The example Pew gave was a sample of 1000 people. If you go pro-rata, that includes 136 African American participants. The error on that sample is 10.5 points. If you increase the sample size of that sector of the population to 500, you reduce the error on that estimate to 5.5 points. You can do the math to scale the results from 500 down to 136 (reduce as a fraction, not by choosing which 136 you want to include.) It's a common polling method, and it improves accuracy instead of skewing results.
Oversampling is discussed about 2/3 the way down the page:
http://www.pewresearch.org/methodology/u-s-survey-research/sampling/
I'm sure everything is on the up and up. Do you not think for a second there could logically be nefarious reasoning behind the decisions to do it? They are creating a false narrative with the desire of swaying the low info crowd into giving up.